**An In Vitro Assessment of Immunostimulatory Responses to Ten Model Innate Immune Response Modulating Impurities (IIRMIs) and Peptide Drug Product, Teriparatide**

**Claire K. Holley 1,†, Edward Cedrone 1,†, Duncan Donohue 2,†, Barry W. Neun 1, Daniela Verthelyi 3, Eric S. Pang 4,\* and Marina A. Dobrovolskaia 1,\***


**Abstract:** Understanding, predicting, and minimizing the immunogenicity of peptide-based therapeutics are of paramount importance for ensuring the safety and efficacy of these products. The so-called anti-drug antibodies (ADA) may have various clinical consequences, including but not limited to the alteration in the product's distribution, biological activity, and clearance profiles. The immunogenicity of biotherapeutics can be influenced by immunostimulation triggered by the presence of innate immune response modulating impurities (IIRMIs) inadvertently introduced during the manufacturing process. Herein, we evaluate the applicability of several in vitro assays (i.e., complement activation, leukocyte proliferation, and cytokine secretion) for the screening of innate immune responses induced by ten common IIRMIs (*Bacillus subtilis* flagellin, FSL-1, zymosan, ODN2006, poly(I:C) HMW, poly(I:C) LMW, CLO75, MDP, ODN2216, and *Escherichia coli* O111:B4 LPS), and a model biotherapeutic Forteo™ (teriparatide). Our study identifies cytokine secretion from healthy human donor peripheral blood mononuclear cells (PBMC) as a sensitive method for the in vitro monitoring of innate immune responses to individual IIRMIs and teriparatide (TP). We identify signature cytokines, evaluate both broad and narrow multiplex cytokine panels, and discuss how the assay logistics influence the performance of this in vitro assay.

**Keywords:** cytokines; innate immunity; immunogenicity; peptides; teriparatide

#### **1. Introduction**

Repeated administration of therapeutic drug products was shown to trigger unwanted immune responses and the production of antibodies capable of neutralizing both the therapeutic protein and its endogenous counterparts [1–3]. Antibodies to recombinant biotechnology therapeutics come in a variety of isotypes (e.g., IgM vs. IgG vs. IgE), allotypes (e.g., reflecting genetic differences between IgG of biologically unrelated individuals), idiotypes (e.g., reflecting binding to specific epitopes within antibody variable sites), and may ultimately lead to different functional consequences for the host (e.g., binding, PK-altering, neutralizing, hypersensitivity- or anaphylaxis-triggering, and cross-reactive neutralizing). Such anti-drug antibodies (ADA) may lead to severe and, when not timely and properly treated, potentially lethal clinical consequences, loss of treatment efficacy,

**Citation:** Holley, C.K.; Cedrone, E.; Donohue, D.; Neun, B.W.; Verthelyi, D.; Pang, E.S.; Dobrovolskaia, M.A. An In Vitro Assessment of Immunostimulatory Responses to Ten Model Innate Immune Response Modulating Impurities (IIRMIs) and Peptide Drug Product, Teriparatide. *Molecules* **2021**, *26*, 7461. https:// doi.org/10.3390/molecules26247461

Academic Editor: Aleksandra Misicka-Kesik

Received: 2 November 2021 Accepted: 6 December 2021 Published: 9 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and the formation of autoimmunity [4–8]. The frequency of different ADA types and their clinical impact have a reverse relationship, in that binding antibodies occur most frequently and have low clinical impact whereas cross-reacting neutralizing antibodies are rare but have the highest clinical significance.

The immunogenic risk of biotherapeutics and ADA response can be influenced by a multitude of factors. One such factor is the presence of innate immune response modulating impurities (IIRMIs) that might be inadvertently introduced during product manufacturing [4,5]. IIRMIs may have little or no impact on the function of the resulting drug product but may influence the host immune response [4,9–13]. While it is nearly impossible to predict the immunogenicity of a specific biotherapeutic without directly assessing the related immune responses in vivo [3], the presence of IIRMIs contributing to the immunogenicity via priming the immune cells could be identified using in vitro methods detecting innate immunostimulatory responses, including the production of inflammatory cytokines (e.g., IL-1, IFNs, IL-8, TNFα, etc.) and activation of the complement system. Therefore, there is an urgent need in understanding the applicability to, and performance of, in vitro assays in detecting IIRMIs presence in drug products.

Herein, we report the results of an in vitro study analyzing the applicability of several in vitro assays (i.e., complement activation, leukocyte proliferation, and cytokine secretion) to the screening of innate immune responses induced by ten common IIRMIs, including *Bacillus subtilis* flagellin, FSL-1, zymosan, ODN2006, poly(I:C) HMW, poly(I:C) LMW, CLO75, MDP, ODN2216, and *Escherichia coli* O111:B4 LPS, as well as model therapeutic Forteo™ (teriparatide or TP). The selected assays were chosen due to the known roles of the complement system, cytokines, and activated leukocytes in the process of immunogenicity [14]. While these immunostimulatory biomarkers do not directly predict immunogenicity, they serve as important prerequisites to it, which, when monitored in vitro, may allow for the detection of biologically active contaminants contributing to the process of immunogenicity by priming the immune cells [14].

#### **2. Results**

#### *2.1. Initial In Vitro Characterization and Assay Selection*

Forteo™ is a peptide-based therapeutic formulation where the active peptide, teriparatide (TP), is produced using recombinant DNA technology. To characterize the whole product, we first established that TP and its corresponding formulation buffer (FB) had no detectable endotoxin and β-glucans that could activate innate immune responses [15–17] using a commercial turbidity *Limulus* Amebocyte Lysate (LAL) assay and Factor-C depleted LAL (Glucatell) assay respectively (Table 1 and Table S1).

Detection of impurities in cell-based assays requires cells that are sensitive to the presence of IIRMI and can elicit a quantifiable response. Previous studies have shown that very low levels of impurities that trigger pattern recognition receptors (PRRs) can stimulate a local innate immune response at the site of inoculation and suggested that cell-based assays could be used to detect these types of impurities in the products. Since retaining cell viability throughout the assay is critical, we first determined whether TP would alter cell viability and determined the highest concentration of TP that could be used in a PBMC-based study where the cells were in culture for 24 h. As shown in Figure S1, when PBMC were cultured in the presence of TP at concentrations ranging from 0.025 to 25 μg/mL, the viability of the cells was retained, but higher concentrations of the product reduced cell viability to 60% (Figure S1). Based on this data, the highest non-toxic concentration (25 μg/mL) was chosen as the top concentration to be used for subsequent in vitro experiments, including the assessment of TP and/or IIRMI activation of C3a complement, leukocyte proliferation, and cytokine secretion (Table 1).

**Table 1.** Initial Characterization of Teriparatide. Teriparatide (TP) purity and capability of triggering innate immunity activation in vitro, either due to the presence of innate immune response modulating impurities (IIRMIs) in the drug formulation or due to the presence of the drug itself, was assessed through the following assays. Results were below the level of detection, so these assays were not used for future TP immunity experiments. LAL = *Limulus* Amoebocyte Lysate Assay; LLOQ = lower limit of quantification; STE = Sterility Endotoxin assay; ITA = Immuno-Toxicity Assay; CBA = Cell Based Assay; ELISA = Enzyme-Linked Immuno-Sorbent Assay; AO = Acridine Orange; PI = Propidium Iodine.


Next, we determined whether TP, in concentrations that do not interfere with cell viability, can reduce the response to potential impurities. Using an array of purified TLR agonists at concentrations that are close to those shown to elicit a local innate immune response in vivo, we examined whether the presence of TP in the culture would modulate the response to the PRR-agonists. As shown in Figure S2, while PRR-agonist exposure triggered low levels of leukocyte proliferation in a dose-dependent manner, the response was abrogated in TP-treated cultures (Table 1).

In addition to inducing cell proliferation, the activation of innate immune cells could also induce complement activation. Therefore, we next explored whether TP would activate complement. Treatment with TP resulted in an activation of the complement system as evidenced by an increase in detectable C3a split products; this activation was comparable to that detected in Cremophor-EL and Feraheme-treated plasma samples, used as positive controls (Figure S3). Concentrations of IIRMIs capable of inducing detectable complement activation are typically higher than what may potentially be present in drug products as undesirable contaminants. For example, concentrations of zymosan and lipopolysaccharide (LPS) required to produce detectable complement activation are 10 mg/mL or >500 μg/mL, respectively [18,19]. Therefore, this assay was not selected for subsequent experiments (Table 1).

#### *2.2. In Vitro Cytokine Responses to Teriparatide*

PBMCs treated with TP alone noticeably induced PGE-2 and IL-8 production (Figure S4). TP-induced PGE-2 production directly correlated with TP concentrations added to PBMC cultures (Figure 1A). Such correlation for IL-8 induction was only observed in 3 of the 10 tested PBMC cultures (Figure 1B). Cultures from the remaining donors showed increased IL-8 levels at the second-highest concentration (2.5 μg/mL) but not at the highest concentration (25 μg/mL). The reduced levels of IL-8 secreted after incubation with highest concentration of TP (25 μg/mL) suggest a level of PBMC exhaustion resulting from high stimulation over the course of 24 h.

**Figure 1.** A 16-plex Induction of Prostaglandin-E2 and Interleukin-8 by Teriparatide. PBMCs from 10 healthy human donors were treated with 0.025, 0.25, 2.5, and 25 μg/mL teriparatide (TP), compared to a PBS negative control (NC) and LPS/PHA-M/ODN positive control (PC) for 24 h. Supernatants were analyzed for the presence of (**A**) PGE-2 or (**B**) IL-8 by 16-plex multiplex ELISA. Each bar shows mean and standard deviation (N = 2).

#### *2.3. Teriparatide Effects on Cytokine Expression Are Due to the Formulation Buffer (FB)*

To understand whether the induction of PGE-2 and IL-8 observed in TP-treated cultures (Figure S4) was due to the active pharmaceutical ingredient (API) or FB, we conducted a follow-up experiment in which TP was tested side-by-side with FB at equivalent dilutions that resulted in equivalent concentrations of the FB; these dilutions were performed in

PBS. We also performed TP dilutions in the FB and tested them in the same cultures with PBS-diluted FB and TP. The results of this experiment demonstrated that PGE-2 and IL-8 responses to TP were due to the FB (Figure 2A,B and Figure S5).

**Figure 2.** Formulation Buffer is Responsible for Prostaglandin-E2 and Interleukin-8 Cytokine Response to Teriparatide. (**A**,**B**) PBMCs from three healthy human donors were used to test teriparatide (TP) at 0.025, 0.25, 2.5, and 25 μg/mL API, diluted in either PBS or Formulation Buffer (FB), compared to complete FB diluted in PBS to achieve the equivalent API concentrations, compared to a PBS negative control (NC) and LPS/PHA-M/ODN positive control (PC). Each bar shows a mean response and a standard deviation (N = 3); (**C**,**D**) PBMCs from another set of three healthy donors were used to test the components of FB (metacresol, mannitol, glacial acetic acid, and sodium acetate) at concentrations equivalent to 25 μg/mL of API in TP, in comparison to complete FB, TP diluted in PBS, and TP diluted in FB. Each bar shows a mean response and a standard deviation (N = 2).

> Next, we hypothesize that metacresol, a preservative of FB, was the cause of the cytokine response to TP, because an earlier study in THP-1 cells reported that this excipient, at a concentration comparable to that present in our cultures (0.2 mg/mL), induced chemokine MCP-1 (but not TNFα, IL-1, or IL-6) [20]. To verify this hypothesis, metacresol and other components of FB (mannitol, glacial acetic acid, and sodium acetate) at concentrations equivalent to that of API in TP were added to PBMC cultures and the supernatants were analyzed for the presence of cytokines (Figure 2C,D and Figure S6). The result of this experiment demonstrated that, in addition to metacresol, all other individual components of the FB contribute to the cytokine response observed with TP. Contrary to our hypothesis about the potential inflammatory nature of metacresol, the highest cytokine response, specifically IL-8, was observed upon application of mannitol (Figure 2D), a response which

has previously been reported on in vitro PBMCs and in vivo endothelial cells to deleterious effect [21,22].

#### *2.4. In Vitro Cytokine Responses to Individual IIRMIs*

Since IIRMIs activated a broad and often overlapping spectrum of cytokines (Figure 3, Figures S7 and S8), we next performed a global analysis using Euclidian distance and Ward's clustering for the dendrogram and constructed a heatmap of normalized values averaged across all donors and replicates (Figure 4 and Figure S9). This normalization included scaling each cytokine reading across all collected values by dividing each value by that cytokine's standard deviation obtained across all donors, which brought all cytokines onto roughly the same scale. The benefit of using this approach is that one can compare cytokines directly across all 10 donors, while keeping cytokines with very large values from swamping the comparative global analyses. These analyses revealed a pure red band representing the negative control and a bright yellow vertical band representing the positive control. These analyses also identified groupings of cytokines with similar response patterns across all IIRMI treatments (Figure 4). For example, chemokines IL-8 and MIP-1α showed very similar patterns; PGE-2 showed such a high response to the two higher concentrations of zymosan that it overshadowed the positive control; IL-2 and IL-17 were very similar in that they did not appear to be strongly induced by any IIRMI (Figure 4). Alternatively, these analyses also gave us insight on how various IIRMIs, and their concentrations, clustered with respect to the cytokine response patterns that they induced (Figure S9). These analyses demonstrated that the highest concentrations of each IIRMI often clustered together. For example, zymosan and CLO75 clustered together at the bottom of the heatmap. The analyses also highlighted a group of IIRMIs that seemed to have virtually no cytokine response, including the lowest concentrations of poly(I:C) LMW, ODN2006, poly(I:C) HMW, and ODN2216. Finally, these analyses also highlighted a group of IIRMIs, and their concentrations located in the center of the heat map, that predominantly activated IL-8 (Figure S9).

**Figure 3.** Normalized Cytokine Response to Zymosan and/or Teriparatide and Selection of One Signature Cytokine: PBMCs from 10 healthy human donors were treated with (**A**) zymosan alone or (**B**) zymosan in combination with 25 μg/mL TP for 24 h. Supernatants were analyzed for the presence of cytokines by multiplex ELISA. The signature cytokine (red box) is the one for which the IIRMI concentration, when compared to the PBS negative control (NC), results in a *p* < 0.05. The data for which statistical significance was not observed are marked with ns. Statistical significance is shown with an asterisk as follows: \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001; and \*\*\*\* *p* < 0.0001. Similar results for the other nine IIRMIs are available in Figures S7 and S8.

**Figure 4.** Innate Immune Response Modulating Impurity Treatment and Concentration Patterns via Euclidian Distance and Ward's Clustering. PBMCs from 10 healthy human donors were treated with various concentrations of IIRMIs, alone and in combination with 25 μg/mL Teriparatide (TP), compared to a PBS negative control (NC) and LPS/PHA-M/ODN positive control (PC), for 24 h. Supernatants were analyzed for the presence of cytokines by multiplex ELISA. Shown is the mean response of normalized values averaged across all donors, clustered based on cytokine response. Dendrograms were created using complete linkage clustering on the Euclidian distance matrices. Similar results for IIRMI clustering available in Figure S9.

> Further Pearson's correlation analysis allowed for clustering the cytokine responses based on how well cytokine values correlated across all treatment groups (Figure S10). This analysis revealed that the strongest correlations were between IL-6 and TNFα, IL-8 and MIP-1α, as well as IL-1α, IL-1β, and IL-12 (Figure S10A). We also observed that different concentrations of the same IIRMI tended to correlate well, especially at the higher concentration ranges. For example, higher concentrations of ODN2216, poly(I:C) LMW, and poly(I:C) HMW showed a distinct cluster which was anti-correlated with the higher concentrations of zymosan) and to a lesser degree CLO75 (Figure S10B).

#### *2.5. Identification of Signature Cytokines*

A two-sided paired Wilcoxon test was used to compare cytokines induced by individual concentrations of IIRMIs with negative control samples pooled across all donors (Figure 3A). For each IIRMI, a signature cytokine was identified by determining the lowest IIRMI concentration which, when compared to the baseline, resulted in both an elevation of the cytokine and the lowest p-value (i.e., at least *p* < 0.05) (Figure 3A, red box). For each IIRMI concentration, if two cytokines achieved a level of significance, the lower (more significant) *p*-value won. Since many test samples, especially negative controls, resulted in cytokine levels below the assay lower limit of detection, we used a non-parametric test for statistical analysis. This approach ranks cytokine significance values rather than the magnitude of cytokine difference. Therefore, the "winning" cytokine was not always the one that appeared the best with regards to mean differences, but rather the one that both had the fewest overlaps between treated samples and controls and was consistent between individual donors (Table 2; Figure S7). Other cytokines with statistically significant elevation above the baseline, at the same IIRMI concentration as the signature cytokine, were also observed (Table 2; Figure S7).

**Table 2.** Cytokines Induced by Innate Immune Response Modulating Impurities. Individual IIRMIs, their cognate pattern recognition receptors (PRRs), and signature cytokines detected after treatment with IIRMI are summarized. Using a two-sided Wilcoxon test, a signature cytokine was identified for each IIRMI by determining the lowest IIRMI concentration, which, when compared to the baseline, resulted in an elevation of the cytokine, and had the lowest ranking *p*-value (i.e., at least *p* < 0.05). IIRMI = innate immune response modulating impurities; TLR = Toll-Like Receptor; IL = interleukin; IFN = interferon; MCP = monocyte chemoattractant protein; MIP = macrophage inflammatory protein; NOD = nucleotide-binding oligomerization domain; TNF = tumor necrosis factor; PGE = prostaglandin; LPS = lipopolysaccharide; CLO = thiazoloquinolone derivative; MDP = muramyldipeptide; ODN = oligo deoxyribonucleotide; LMW = low molecular weight; HMW = high molecular weight; FSL = Pam2CGDPKHPKSF, a synthetic lipopeptide derived from *Mycoplasma salivarium*.


#### *2.6. Selection of the Cytokine Panel Specific to Teriparatide and Individual IIRMIs*

In order to understand whether the 16-cytokine panel could be narrowed down to three or four cytokines that would be representative of all 10 IIRMIs, we performed additional analysis using the same approach as described above but focused on the top three "winning" cytokines for each IIRMI (Table 3). For this analysis, IIRMIs were grouped based on the intracellular location of their cognate PRRs. Interestingly, all IIRMIs that activate membrane-tethered TLRs consistently induced two cytokines (IL-1α and MIP-1α (Table 3). This finding suggests that any of these two cytokines could be used as a biomarker for the detection of IIRMIs triggering membrane-tethered PRRs. In contrast, no such consistency was observed for IIRMIs that activate endosomal TLRs. Therefore, a combination of cytokines MCP-1 and IL-8 or MCP-1 and IL-6 would be required to suggest the presence of IIRMIs triggering endosomal TLRs (Table 3). One of the following cytokines—IL-6, IL-8, or IP-10—could be used to suggest the presence of IIRMIs triggering cytosolic PRRs (Table 3).

*Molecules* **2021**, *26*, 7461

**Table 3.** Selection of three signature cytokines induced by individual Innate Immune Response Modulating Impurities. A two-sided unpaired Wilcoxon test was used to select the top three cytokines for each IIRMI, which had consistent responses between all donors and the lowest *p*-value. Starting with the lowest concentration for each IIRMI, if three cytokines did not achieve significance of *p* ≤ 0.05, the next highest concentration was evaluated until three cytokines were chosen. If more than three cytokines achieved *p* ≤ 0.05 at the selected concentration, the three with the lowest (most significant) p-values were selected. The top three cytokines selected for any IIRMI are shown as "TRUE" while the remaining less significant cytokines are shown as "FALSE". IIRMIs are grouped based on the intracellular localization of their cognate pattern-recognition receptors (PRRs) and color-coded as follows: BLUE-cellular membrane, RED-endosome, GREEN-cytosol. TRUE values in each group are highlighted in bold and the same color code as that used for corresponding IIRMIs.


For the subsequent experiments, we focused on a seven-cytokine panel which includes a combination of the following signature cytokines (IL-1α, MIP-1α, IP-10, MCP-1, IL-6, and IL-8) representing all tested IIRMIs, and one cytokine (PGE-2) representing the response to TP (Figure 5, Figures S11 and S12). Interestingly, the majority of IIRMIs that activate membrane-tethered TLRs, induced MCP-1 production, with the remaining IIRMI, zymosan, instead inducing MIP-1α expression rather than the expected IL-1α. For the cytosolic PRRs, the overwhelming response was IL-8 expression. For the endosomal PRRs, we again observed that there was no cytokine consistency, with the highest cytokine expression covering MIP-1α, IL-8, and IL-6, with two of the five IIRMIs inducing high levels of MCP-1. For the majority of the IIRMIs, there was little or no IL-1α, IP-10, or PGE-2 expression detected (Figure 5, Figures S11 and S12). However, PGE-2 production had a dose dependent response when PBMCs were treated with increasing concentrations of TP alone (Figure 6). This effect was previously observed (Figure 1A) indicating that PGE-2 is still a hallmark cytokine for tracking TP immunostimulatory activity.

**Figure 5.** Seven-plex Induction of Cytokines in PBMCs. PBMCs collected from 10 healthy human donors were treated with 0.025, 0.25, 2.5, and 25 μg/mL TP (red box) or IIRMIs alone, compared to a PBS negative control (NC) and LPS/PHA-M/ODN positive control (PC), for 24 h. Supernatants were analyzed for the presence of cytokines by multiplex ELISA. Shown is the mean response (N = 2). Shown here are the data generated using PBMC cultures of five representative donors. The data generated using PBMCs of the remaining five donors are presented in Figure S11. Normalized data for each treatment set in all ten donors are also presented in Figure S12.

**Figure 6.** Seven-plex Induction of Prostaglandin-E2 by Teriparatide. PBMCs from 10 healthy human donors were treated with 0.025, 0.25, 2.5, and 25 μg/mL teriparatide (TP), compared to a PBS negative control (NC) and LPS/PHA-M/ODN positive control (PC), for 24 h. Supernatants were analyzed for the presence of PGE-2 by 7-plex multiplex ELISA. Each bar shows mean and standard deviation (N = 2).

#### *2.7. Teriparatide Affects Expression of IIRMI-Induced Cytokines*

The presence of TP in cell cultures affected the induction of cytokines by individual IIRMIs (Table 4; Figure S8). Euclidian distance and Ward's clustering analysis demonstrated that the patterns for chemokines IL-8 and MIP-1α did not change with the addition of TP (right half of the plot) (Figure 4). In contrast, the group of IL-1β, IL-1α, and IL-12, which showed strong responses to the higher concentration of zymosan and CLO75, was strongly inhibited by the addition of TP. The loss of response with TP was also seen at the highest concentration of IIRMI for cytokines IFNλ and IFNα. PGE-2 induced by two higher concentrations of zymosan was also lost with the addition of TP (Figure 4).

**Table 4.** Teriparatide Affects Cytokines Induced by Innate Immune Response Modulating Impurities (IIRMIs). Individual IIRMIs and IIRMI-triggered cytokines in which expression is affected by the presence of 25 μg/mL of teriparatide (TP) are summarized in the table. In the presence of TP, all cytokines shown in the table are inhibited, except for the cytokines highlighted with an asterisk (\*); levels of these cytokines are higher in the presence of TP. Statistical analysis included a two-sided Wilcoxon test.


#### *2.8. Teriparatide Effects on IIRMI-Induced Cytokines Are Due to the Formulation Buffer (FB)*

To understand whether the suppression of IIRMI-induced cytokines by TP was due to the API or FB, we conducted a follow-up experiment in which four concentrations of TP were tested side-by-side with the second highest concentration of IIRMIs alone, as well as IIRMIs in combination with either 25 μg/mL TP or equivalent 25 μg/mL FB. The results of this experiment demonstrated that changes in the expression of IIRMI-induced cytokines by TP were due to the FB (Figure S13).

#### *2.9. Donor's Genetic Background Determines the Magnitude of Cytokine Response to IIRMIs*

We observed that PBMCs from some donors demonstrated more robust (i.e., higher magnitude) responses to TP than cultures from other healthy donors (Figure S4). To understand whether such differences were due to the genetic background of the PBMC donor or variability in the day-to-day handling of donor's blood and PBMCs, we recalled one highest responder (donor G9L1) and two average responders (donors M4W2 and C9M4) for the second time, repeated the TP treatments, and compared the results between two experiments. The results were consistent between the two experiments despite some variability in the individual cytokine levels observed in all donors (Figure S14).

#### *2.10. Influence of Assay Logistics on Cytokine Responses to TP and IIRMIs*

We further examined the influence of blood handling and storage conditions on resultant cytokine responses to IIRMIs or TP. Blood from ten healthy donors was separated into six treatment groups: freshly isolated and freshly treated PBMCs; freshly isolated PBMCs cultured for 24 h prior to the treatment; freshly isolated and cryopreserved PBMCs; PBMCs isolated from blood refrigerated for 24 h or 48 h before PBMC isolation; and whole blood cultures. All groups were then dosed with IIRMIs or TP for 24 h. We then measured PBMC recovery and viability for these treatment groups (Figure 7) as well as the levels of our seven key cytokines, IL-1α, MIP-1α, IP-10, MCP-1, IL-6, and IL-8 (Figures 5, 8, S11, S12 and S15).

**Figure 7.** The Effect of Storage Conditions on PBMC Viability and Cell Recovery. To simulate various handling and storage conditions used in research, PBMCs from three healthy human donors were examined after fresh isolation, cryopreservation, and isolation from refrigerated blood (24 h or 48 h). Cell viability was then assessed using AO/PI. (**A**) Number of PBMCs recovered under the various storage/handling conditions. (**B**) Viability of stored PBMCs compared to their freshly isolated PBMC counterparts. Each bar shows the mean result and standard deviation (N = 3).

**Figure 8.** IL-1α and PGE-2 Responses to Zymosan are Affected by PBMC and Blood Handling Conditions. PBMCs from 10 healthy human donors were exposed to various common laboratory handling conditions (isolated from fresh blood, cultured for 24 h, cryopreserved, isolated from blood refrigerated for 24 h or 48 h, and whole blood cultures) before being treated with IIRMIs for 24 h. Supernatants were analyzed for the presence of cytokines. Shown are the mean cytokine responses to zymosan (red), compared to a PBS negative control (NC, blue). The data for which statistical significance was not observed are marked with ns. Statistical significance is shown with an asterisk as follows: \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001; and \*\*\*\* *p* < 0.0001. Additional results, including zymosan-induced levels of the remaining five cytokines and the cytokine responses for the other nine IIRMIs, are available in Figure 5 (fresh PBMCs) and Figure S15 (all other experimental conditions).

Compared to freshly isolated PBMCs, PBMC viability is reduced to approximately 63% after 48 h of cryopreservation, as compared to the very low viability (~10%) of PBMCs isolated from anti-coagulated blood after refrigeration storage for 24 h or 48 h (Figure 7B). Due to the loss of 90% of usable PBMCs from the stored blood samples, we were only able to treat the remaining cells with a limited selection of IIRMIs for comparison to the other treatment/storage conditions. In addition, this loss of available cells can potentially skew the resultant cytokine production (Figure 7A).

As previously discussed, there was very little general expression of IL-1α, IP-10, or PGE-2 detected even for freshly isolated PBMCs (Figure 5, Figure 8, Figures S11, S12 and S15). Interestingly, the highest levels of IL-1α and PGE-2 were observed after zymosan stimulation in whole blood cultures, indicating that other components of blood may be responsible for increasing the levels of these cytokines.

For the other four cytokines (MCP-1, MIP-1α, IL-8, and IL-6), cultured PBMCs and cryopreserved PBMCs had similar but reduced levels of cytokines compared to freshly isolated PBMCs. Cytokines from refrigerated blood further reduced cytokine levels, even at the highest IIRMI concentrations. This was especially true for IL-6, which were reduced to almost nothing even in the presence of strong LPS or zymosan stimulation (Figures 5, 8, S11, S12 and S15).

#### **3. Discussion**

Based on the results of these initial characterization studies (Table 1), complement activation and leukocyte-proliferation assays were not chosen for subsequent studies as they cannot adequately detect potential differences in IIRMI contamination between different batches of product. These assays, however, could be helpful in studies investigating different formulations of the same API. Examples may include when a product is reformulated, or when a generic or follow-on product elects to have differences in formulation compared to an innovator (reference) product. Therefore, we focused the rest of the study on the cytokine secretion by PBMC after in vitro exposure to TP and formulations containing IIRMIs.

Our study suggested that PGE-2 could be used as a signature cytokine for tracking TP induction of innate immune responses (Figures 1A and 6). We further found that this response is mediated by the FB rather than API. Further investigation found that, unlike our hypothesis about the influence of metacresol, all the FB ingredients contributed to the resultant cytokine response (Figure 2A,C; Figures S5 and S6).

IIRMIs activated a broad and often overlapping spectrum of cytokines (Figures 3, S4, S7 and S8). This finding is consistent with the current literature about PRRs and their cognate ligands [23–25]. Using Euclidian distance and Ward's clustering analyses, we obtained insight on the patterns of IIRMI stimulation and the resultant induced cytokine responses. From these results, we identified groupings of cytokines with similar response patterns across all IIRMI treatments (Figure 4), as well as several cytokines which did not appear to be strongly induced by any IIRMI. These analyses also demonstrated that the highest concentrations of each IIRMI often clustered together (Figure S9).

Further Pearson's correlation analysis allowed for clustering the cytokine responses based on how well cytokine values correlated across all treatment groups and donors (Figure S10A). Strong correlations patterns identified during these analyses were consistent with the currently available literature about the function of these cytokines and the cells that produce them. Specifically, IL-6 and TNFα are produced by monocytes and T-cells, and are responsible for pyrogenicity; IL-8 and MIP-1α are chemokines produced by monocytes and responsible for neutrophils and mixed leukocyte recruitment; IL-1α, IL-1β, and IL-12 are produced by monocytes and DCs and are responsible for the inflammation, fever, and activation of specific subsets of lymphocytes (i.e., IL-1β promotes TH17 differentiation, whereas IL-12 supports TH1 differentiation, NK and T-cell activation to increase IFNγ synthesis and increased cytotoxicity); in addition, IL-1α is a danger signal that indicates damaging effects of IIRMIs that induce its secretion [14]. According to our expectations from the global heatmaps (Figure 4 and Figure S9), Pearson's correlation between the different concentrations of the same IIRMI was consistent with the current knowledge about type and intracellular localization of PRRs stimulated by these IIRMIs (Figure S10B). Specifically, ODN2216, poly(I:C) LMW, and poly(I:C) HMW activate endosomal TLRs (TLR9 and TLR3), whereas zymosan triggers membrane-tethered PRRs (TLR2 and Dectin1) [25,26]. In contrast, CLO75, which showed a lower degree of anti-correlation, is also located in the endosome but is specific to a different PRR (TLR8) [27].

To understand whether our 16-cytokine panel could be narrowed down to three or four cytokines that would be representative of all 10 IIRMIs, we examined the top three "winning" cytokines identified for each IIRMI (Table 3). Due to the overlapping nature of the induced cytokines, we identified two possible panels of three cytokines which would provide at least one positive result for all 10 IIRMIs and potentially could be used by users who do not have access to more than a 3- or 4-plex cytokine detection panel. These panels include the following markers: 1) IL-1α (or MIP-1α), IP-10, and IL-8; or 2) IL-1α (or MIP-1α), MCP-1 and IL-8 (or IL-6).

The results from the subsequent 7-plex panel containing IL-1α, MIP-1α, IP-10, MCP-1, IL-6, IL-8, and PGE-2 (Figure 5, Figures S11 and S12), suggest that our initial 16-cytokine panel can be reduced to a four-cytokine panel, specifically containing MCP-1, MIP-1α, IL-8, and IL-6, which would be representative of all 10 IIRMIs, which can further be expanded

to a five-cytokine TP-specific panel, which includes the TP-signature cytokine, PGE-2, in addition to the four IIRMI-specific cytokines.

TP did not significantly increase the levels of cytokines induced by IIRMIs. TP was, instead, found to decrease the levels of most IIRMI-induced cytokines (Figure 4). Reduced levels of IIRMI-induced cytokine responses in the presence of TP were also the result of the FB rather than the API (Figure S13). Collectively, this finding and the data demonstrating the induction of TP signature cytokine PGE-2 by the FB suggests that the assessment of potential IIRMI contamination of the API could be more informative for comparison of RLD and generic formulations. This data also suggests that a change in the formulation buffer may result in a change in the signature cytokine of the whole product.

The more robust cytokine responses to TP demonstrated by some donors suggests that day-to-day variability in phlebotomy and handling of whole blood and PBMCs may result in quantitative differences (i.e., influence the magnitude of the responses) but would not change the overall qualitative trends and resultant conclusions of the study. However, the genetic background of donors that donate their blood for in vitro experiments does appear to be an important factor in qualitative determination of the PBMC response to individual IIRMIs (Figure S14).

Overall, the PBMC handling and blood storage conditions have a significant effect on the detectable levels of cytokines, with freshly isolated PBMCs being the most preferred condition since it allows for more adequate detection of cytokines as a result of innate immunity activation (Figure 5, Figure 8, Figures S11, S12 and S15).

#### **4. Materials and Methods**

#### *4.1. Materials*

Feraheme (FH) (AMAG Pharmaceuticals, Waltham, MA) and Forteo™ (teriparatide, TP) (Eli Lilly, Indianapolis, IN, USA), were obtained from NIH Pharmacy. All *Limulus* amebocyte lysate (LAL) reagents, LAL grade (endotoxin free) water, Glucatell kits, Glucashield buffer, and *E. coli* lipopolysaccharide (LPS) were from Associates of Cape Cod (East Falmouth, MA, USA). Veronal Buffer was obtained from Boston BioProducts (Ashland, MA, USA). Phosphate Buffered Saline (PBS), RPMI-1640 media, fetal bovine serum (FBS), penicillin and streptomycin solution, L-glutamine, Ficoll-Paque Premium was from GE Life Sciences (Marlborough, MA, USA). Hank's balanced salt solution (HBSS) was from Gibco (Gaithersburg, MD). All IIRMIs—*B. subtilis* flagellin, FSL-1, ODN2006 Class B, poly(I:C) HMW, poly(I:C) LMW, zymosan, CLO75, MDP, ODN2216, and *E. coli* O111:B4 LPS—were from Invivogen (San Diego, CA, USA). Acridine orange (AO)/propidium iodide (PI) staining solution were purchased from Nexcelom Bioscience (Lawrence, MA, USA). The 16-plex and 7-plex cytokine multiplex kits were supplied by Quansys Biosciences (Logan, UT, USA). Cobra venom factor (CVF), Heat Aggregated Gamma Globulins (HAGG), and MicroVue EIA kits were purchased from Quidel Corporation (San Diego, CA, USA). Glacial acetic acid, sodium acetate, mannitol, sodium hydroxide (NaOH), hydrochloric acid (HCl), MTT (3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide), glycine, sodium chloride, dimethyl sulfoxide (DMSO), Phytohemagglutinin (PHA-M), and Cremophor (Cre) were purchased from Sigma-Aldrich (Burlington, MA, USA). Metacresol was from USP (Frederick, MD, USA).

#### *4.2. Innate Immune Response Modulating Impurities*

Ten model innate immune response modulating impurities (IIRMIs) were tested at four concentrations (Table 5) either alone or in combination with teriparatide (TP). Eight IIRMIs (*B. subtilis* flagellin, FSL-1, zymosan, ODN2006, poly(I:C) HMW, poly(I:C) LMW, CLO75, and MDP) were selected based on preliminary studies in HEK-TLR reporter cells [9,10]; two other IIRMIs (ODN2216 and *E. coli* O111:B4 LPS) were selected based on the Nanotechnology Characterization Laboratory (NCL) (https://ncl.cancer.gov/, accessed on 15 October 2020) prior experience using them as immunological assay cascade positive

controls. Taken together, these ten IIRMIs bind Dectin 1, TLRs 2, 3, 4, 5, 6, 8, 9, and NOD2, as summarized in Table 5.

**Table 5.** Innate Immune Response Modulating Impurities used in the present study. IIRMIs and their final concentrations tested in vitro are summarized. LPS = lipopolysaccharide; CLO = thiazoloquinolone derivative; MDP = muramyldipeptide; ODN = oligo deoxyribonucleotide; LMW = low molecular weight; HMW = high molecular weight; FSL = Pam2CGDPKHPKSF, a synthetic lipopeptide derived from *Mycoplasma salivarium*.


#### *4.3. Endotoxin Detection*

Endotoxin levels were evaluated using the kinetic turbidity *Limulus* Amebocyte Lysate (LAL) Assay according to NCL protocol STE-1.2 [28,29]. Briefly, 100 μL of TP (at 250 μg/mL) and the equivalent amount of its formulation buffer (FB) were each mixed with 100 μL of LAL reagent in a glass tube, then measured via spectrophotometer at 660 nm for at least 7200 sec for appropriate development. Using a standard curve prepared with Control Standard Endotoxin of known potency, we calculated the concentration of endotoxin present in the TP and FB solutions.

#### *4.4. β-Glucan Detection*

Levels of β-glucans were evaluated using Glucatell® kit as detailed in NCL protocol STE-4 [15,30]. Briefly, 50 μL of TP (at 250 μg/mL) and the equivalent amount of its formulation buffer (FB) were each mixed with 50 μL of Glucatell reagent in a 96-well plate and incubated at 37 ◦C. The reaction was stopped through the addition of 50 μL of 1N HCl-sodium nitrite solution, 50 μL of ammonium sulfamate solution, and then 50 μL of NEDA solution to each well. Color development was immediately observed and measured at 540–550 nm using a spectrophotometer. Using a β-(1,3)-D-glucan standard curve, we calculated the concentration of β-(1,3)-D-glucan present in the TP and FB solutions.

#### *4.5. Donor Blood*

Blood from healthy human donors was collected in vacutainers containing either Li-heparin or K2-EDTA (BD Biosciences, San Jose, CA, USA) under the NCI-Frederick protocol OH9-C-N046. At the time of blood collection, donors were not on any medications and have never been exposed to the model Forteo™ teriparatide formulation.

#### *4.6. Peripheral Blood Mononuclear Cell (PBMC) Isolation and Culture*

Fresh donor blood anti-coagulated with Li-heparin was mixed with an equal volume of room-temperature PBS. The blood/PBS mixture was then slowly layered on top of Ficoll-Paque solution in a 4:3 ratio. The sample was centrifuged for 30 min at 900× *g*, 18–20 ◦C, without brake. After centrifugation, the upper layer containing plasma and platelets was removed and discarded. The mononuclear cell layer was isolated and washed using an excess (approximately three times volume) of Hank's Balanced Salt Solution (HBSS) and centrifuged for 10–15 min at 400× *g*, 18–20 ◦C. After washing, the supernatant was discarded, and the wash step was repeated once more. The remaining mononuclear cells were then resuspended in complete RPMI-1640 medium, containing 10% FBS (heat inactivated), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. Cell viability was then determined using the acridine orange (AO)/propidium iodide (PI) dualfluorescence viability method, in which an equal volume of staining solution, containing AO (live cells, green) and PI (dead cells, red) was added to cells and analyzed in <60 s using a fluorescent Cellometer instrument. The details of the protocols are publicly available through NCL protocol ITA-10 and were previously described [31,32].

#### *4.7. PBMC Cryopreservation*

Isolated PBMCs were resuspended at a concentration of 5–7.5 × 106 cells/mL in freezing media (10% DMSO in FBS), placed into cryopreservation tubes, and stored in a freezing container containing isopropanol for controlled freezing at −80 ◦C.

#### *4.8. Whole Blood Cell (WBC) Culture*

Fresh donor blood anti-coagulated with Li-heparin was mixed 1:4 with room-temperature PBS (e.g., 10 mL blood added to 30 mL PBS). The blood/PBS mixture was then added directly to 96-well plate for treatment and culture at 37 ◦C. The details of the protocols are publicly available through NCL protocol ITA-10 and were previously described [31,32].

#### *4.9. Teriparatide Cytotoxicity Analysis*

PBMCs in complete 1640-RPMI were incubated with 0–50 μg/mL teriparatide (TP) for 24 h. Cell viability was then determined using the AO/PI staining method [33].

#### *4.10. Leukocyte Proliferation*

PBMCs were cultured at in the presence of controls, 0.025–25 μg/mL TP, four concentrations of IIRMIs (Table 5), or four concentrations of IIRMI + 25 μg/mL TP for 72 h. The proliferation of leukocytes was determined according to NCL protocol ITA-6 [34].

#### *4.11. Complement Activation*

These experiments were conducted according to NCL protocol ITA5.2 [35]. Briefly, K2-EDTA plasma from individual donors was pooled and incubated with controls or 0.025–83.3 μg/mL TP, and veronal buffer for 30 min at 37 ◦C. Following incubation, the samples were analyzed for the presence of complement split product C3a using a commercial multiplex ELISA kit. In this experiment, Cobra venom factor (CVF) and Heat Aggregated Gamma Globulins (HAGG) were used as the assay positive controls (PC). Cremophor (Cre) and Feraheme (FH) were included as additional controls as they are known to cause complement-mediated toxicity in sensitive patients [36–39].

#### *4.12. Cytokine Production*

These experiments followed NCL protocol ITA-10 [31,32]. PBMCs were cultured at in the presence of PBC negative control, LPS/PHA-M/ODN positive control, 0.025–25 μg/mL TP, four concentrations of IIRMIs (Table 5), or four concentrations of IIRMI + 25 μg/mL TP for 24 h in a humidified 37 ◦C, 5% CO2 incubator. After incubation, the plates were centrifuged for 5 min at 700× *g* to pellet the PBMCs. The supernatants were collected for cytokine analysis using custom 16-plex or 7-plex multiplex plates from Quansys Biosciences (Logan, UT, USA). The cytokines present in the multiplex panel included type I interferon (IFNα), type II interferon (IFNγ), type III interferon (IFNλ), interleukins (IL-1α, IL-1β, IL-2, IL-6, IL-8, IL-10, IL-12, IL-17), interferon-gamma inducible protein (IP-10), tumor necrosis factor alpha (TNFα), prostaglandin-E2 (PGE-2), macrophage inflammatory protein (MIP-1α), and monocyte chemoattractant protein (MCP-1). Cytokine levels were each quantified against a standard curve of calibrator controls (provided in the Quansys kit).

#### *4.13. Statistical Analysis*

All experiments were performed with at least two independent samples, tested in duplicate (%CV < 25). Unless otherwise stated, results show the mean and standard deviation generated from these independent samples. For the cytokine multiplex assay, the analysis was performed using custom R scripts. Cytokine concentration values above the detection limit ("ADL") were set to the upper detection limit, and values below the detection limit ("BDL") were set to zero. Statistical analysis of cytokine data was performed using normalized values. The normalization included scaling each cytokine reading across all collected values by dividing each value by that cytokine's standard deviation obtained across all donors. The normalization brought all cytokines onto roughly the same scale. The benefit of using this approach is that one can compare cytokines directly on graphs across 10 donors, and it keeps cytokines with very large values from swamping global analyses. As an initial quality control (QC) step, we looked at the negative control (NC) vs. positive control (PC) values for each cytokine using the Wilcoxon non-parametric test on replicate-averaged cytokine-normalized values. All cytokines showed significantly higher PC than NC values. Additionally, we looked at the correlation between pairs of replicate runs (a vs. b for each treatment) and observed a good correlation for most pairs. Unless otherwise noted, comparisons between cytokine levels were made on normalized values using two-sided Wilcoxon tests.

#### **5. Conclusions**

Cytokine secretion by human PBMCs may be used to assess the innate immune responses to IIRMIs, formulation components, and whole products containing peptide and protein therapeutics. While the whole product needs to be analyzed, the results of our study emphasize that the components of FB are not immunologically inert and can contribute to both the cytokine stimulation by the whole product and inhibition of the IIRMI-mediated cytokines. Statistical analysis helps to identify signature cytokines and select cytokine panel appropriate for the given peptide drug product and any prospective generics and biosimilars. It is expected that signature cytokines maybe different between different products due to differences in formulation components, potential IIRMI contamination, immunological properties of API, and interactions among them, which collectively may lead to both quantitative and qualitative differences. Importantly, the logistics of blood storage and handling may influence the results, and, therefore, should be carefully investigated during assay validation phase.

**Supplementary Materials:** The following are available online, Table S1: Endotoxin and β-glucan Levels in Teriparatide Formulation, Figure S1: PBMC Viability in the Presence of Teriparatide, Figure S2: In vitro Leukocyte Proliferation in the Presence of Teriparatide and/or Innate Immune Response Modulating Impurities, Figure S3: In vitro Complement Activation Induced by Teriparatide, Figure S4: 16-plex Induction of Cytokines in PBMCs, Figure S5: Formulation Buffer is Responsible for the Cytokine Response to Teriparatide, Figure S6: Metacresol and Mannitol are Responsible for the Formulation Buffer Cytokine Response, Figure S7: Normalized 16-plex Cytokine Response to Innate Immune Response Modulating Impurities and Selection of One Signature Cytokine, Figure S8: Normalized 16-plex Cytokine Response in the Combined Presence of Innate Immune Response Modulating Impurities and Teriparatide, Figure S9: Innate Immune Response Modulating Impurity-Induced Cytokine Response Patterns via Euclidian Distance and Ward's Clustering, Figure S10: Cytokine Analysis via Pearson's Correlation, Figure S11: 7-plex Induction of Cytokines in PBMCs, Figure S12: Normalized 7-plex Cytokine Response to Innate Immune Response Modulating Impurities, Figure S13: Formulation Buffer Affects Cytokines Induced by Innate Immune Response Modulating Impurities, Figure S14: Reproducibility of Cytokine Response to Teriparatide in PBMC Cultures, Figure S15: Normalized Cytokine Responses to Innate Immune Response Modulating Impurities are Affected by PBMC and Blood Handling Conditions.

**Author Contributions:** C.K.H., E.C. and B.W.N. conducted experiments. D.D. performed statistical analysis. C.K.H., E.C. and D.D. contributed equally to this study. M.A.D., E.S.P. and D.V. conceived, anddesigned the study. M.A.D. managed the study. All authors analyzed data and wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported in whole by federal funds from the U.S. Food and Drug Administration, Center for Drug Evaluation and Research/Office of Generic Drugs, under an Inter-Agency Agreement with NCL/NCI/NIH (IAA #224-19-3008S).

**Institutional Review Board Statement:** Blood from healthy human donors was collected under the NCI-Frederick protocol OH9-C-N046.

**Informed Consent Statement:** Informed consent was obtained from all human subjects involved in donating blood for this study.

**Data Availability Statement:** Materials used in this study are available from commercial vendors; the data used to produce figures in this manuscript maybe requested by contacting principal investigators of the study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Change in Lipofectamine Carrier as a Tool to Fine-Tune Immunostimulation of Nucleic Acid Nanoparticles**

**Hannah S. Newton 1,†, Yasmine Radwan 2,†, Jie Xu 1, Jeffrey D. Clogston 1, Marina A. Dobrovolskaia 1,\* and Kirill A. Afonin 2,\***


**Abstract:** Nucleic acid nanoparticles (NANPs) require a carrier to allow for their intracellular delivery to immune cells. Cytokine production, specifically type I and III interferons, allows for reliable monitoring of the carrier effect on NANP immunostimulation. Recent studies have shown that changes in the delivery platform (e.g., lipid-based carriers vs. dendrimers) can alter NANPs' immunorecognition and downstream cytokine production in various immune cell populations. Herein, we used flow cytometry and measured cytokine induction to show how compositional variations in commercially available lipofectamine carriers impact the immunostimulatory properties of NANPs with different architectural characteristics.

**Keywords:** nucleic acid nanoparticles; lipofectamine; cytokine; interferons

#### **1. Introduction**

Nucleic acid nanoparticles (NANPs) are therapeutic nucleic acids designed to assemble into various geometric shapes with distinct physicochemical properties and have a host of diagnostic and therapeutic benefits in a wide array of diseases [1–4]. Physicochemical characterization and immunological evaluation of various RNA and DNA NANPs have been performed to fully understand their structure–activity relationship and help bridge gaps that hinder the clinical translation of these novel nanomaterials [3,5–7].

It has been shown that RNA and DNA NANPs require a carrier for their intracellular delivery to immune cells [5–7]. Without a delivery agent, NANPs have repetitively been shown to remain invisible to the immune system and do not stimulate immune responses [5,7,8]. However, upon delivery with, for example, Lipofectamine 2000 (L2K), NANPs are recognized by peripheral blood mononuclear cells (PBMCs), more so by monocytes than lymphocytes [5,7]. Furthermore, NANPs induce interferon (IFN) response in PBMCs, particularly type I (IFNα; IFNβ; IFNω) and III (IFNλ) IFN responses, and NANP composition and structure define the degree of response—for example, RNA NANPs stimulate greater immune response as compared to their DNA counterparts [5–8]. Within the RNA NANP category, the potency of IFN responses is influenced by nanoparticle architectures, shape, and size. For example, 3D RNA cubes are more immunostimulatory than 2D RNA rings, and 1D RNA fibers are the least immunostimulatory NANPs of all [5,8]; likewise, RNA hexagons are more potent than RNA triangles [5].

Moreover, various delivery platforms can tailor NANPs' immunorecognition and subsequent function, including cytokine induction [7–9]. For example, NANPs' delivery with dendrimers influences their uptake and PBMC cytokine induction when compared to L2Kassisted deliveries. NANPs delivered using cationic dendrimers induce pro-inflammatory

**Citation:** Newton, H.S.; Radwan, Y.; Xu, J.; Clogston, J.D.; Dobrovolskaia, M.A.; Afonin, K.A. Change in Lipofectamine Carrier as a Tool to Fine-Tune Immunostimulation of Nucleic Acid Nanoparticles. *Molecules* **2023**, *28*, 4484. https:// doi.org/10.3390/molecules28114484

Academic Editor: Aldo Galeone

Received: 27 April 2023 Revised: 27 May 2023 Accepted: 30 May 2023 Published: 1 June 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

cytokines and danger signals but not type I and III IFNs. In contrast, the same NANPs delivered with L2K induce the IFN response with no/low cytokines and danger signals [7]. To further examine the role of the delivery carrier in the qualitative and quantitative outcomes of NANPs' interactions with the primary human immune cells, we investigate two different commercial lipofectamine carriers.

Lipofectamine is a 3:1 (*w*/*w*) formulation of 2,3-di-oleyloxy-N- [2(spermine-carboxamido) ethyl]-N,N-dimethyl-l-propan-aminium (DOSPA) and dioleoylphosphatidylethanolamine (DOPE) [10]. While L2K and Lipofectamine MessengerMAX (LMM) are both lipofectamines, their composition and chemical structures were optimized to improve the transfection of different types of nucleic acids. L2K is marketed as a more versatile transfection reagent with superior co-transfection performance and the ability to deliver a variety of nucleic acids [11]. LMM, on the other hand, is optimized and recommended for delivery of mRNA without genomic integration [12]. We hypothesized that fine structural variations in lipofectamine might further contribute to controlling the magnitude of NANP-mediated immunostimulation.

Herein, we present results indicating that the type of lipofectamine, L2K vs. LMM, alters NANPs' immunostimulation and cytokine production, thereby providing additional tools to researchers for controlling the magnitude of the IFN response.

#### **2. Results**

#### *2.1. Assembly of NANPs and Formation of Lipoplexes*

A representative panel of NANPs—RNA fibers, RNA rings, RNA cubes, and DNA cubes—were selected to address the effect of NANPs' composition and architectural parameters on their delivery with lipofectamines and immunorecognition. The assembly of NANPs took place in endotoxin-free conditions. The successful assembly of NANPs was confirmed using non-denaturing polyacrylamide gel electrophoresis (native-PAGE) and visualized via atomic force microscopy (AFM), as shown in Figure 1A.

In addition, eight lipoplexes formed between L2K or LMM, and each of the tested NANPs were visualized using transmission electron microscopy (TEM) and compared to free lipofectamines. The change in morphology of the carrier alone compared to the carriers complexed with NANPs suggests that NANPs were successfully complexed in L2K and LMM, as demonstrated in Figure 1B.

#### *2.2. Monocytes Have Greater NANP Uptake Than Lymphocytes Regardless of Lipofectamine Carrier*

To compare the ability of LMM vs. L2K to serve as carriers for NANPs, representative Alexa Fluor 488 (AF488) fluorescent DNA or RNA NANPs (AF488-DNA cubes; AF488- RNA cubes; and AF488-RNA rings) were incubated overnight with PBMCs at a final concentration of 10 nM. The uptake (and/or association with the cellular plasma membrane) of the fluorescent NANPs in both lymphocyte and monocyte populations was determined using flow cytometry. Lymphocyte and monocyte populations were defined via forward and side scatter. The NANP-association with the cells was measured in two ways: (i) the percentage of AF488+ lymphocytes or monocytes, i.e., the proportion of cells that have NANP-associated fluorescence, and (ii) the degree of geometric mean fluorescence intensity (gMFI) in each AF488+ population, i.e., the magnitude of NANP uptake/association by individual cells. Representative gating of the lymphocyte and monocyte populations, along with the AF488+ gating, is shown in Figure 2A.

**Figure 1.** Characterization of NANPs and their lipoplexes. (**A**) 3D models and AFM images of representative NANPs. (**B**) TEM images of NANPs complexed with either L2K (upper panel) or LMM (lower panel).

As previously established by our group, AF488-labeled NANPs have different levels of fluorescence due to the differences in labeling efficiencies of individual oligos. Therefore, the experimental results should not be compared across different NANP types and should be considered qualitatively [5]. Nonetheless, our results were in agreement with previous data from our group, which showed lipofectamine leads to NANP uptake predominately by the monocyte population (Figure 2B,C) [5,7]. Both the percentage of AF488+ monocytes (~60–90%) and the gMFI of AF488+ monocytes (~10 K–40 K arbitrary units (a.u.)) were more significant than the results in the lymphocyte population (~10–50% and 800–1600 a.u., respectively), regardless of lipofectamine type (Figure 2B,C). However, there were a few significant differences when we compared L2K- versus LMM-mediated NANP uptake within a particular NANP type. The only difference between the percentage of AF488+ populations was in lymphocytes, where LMM led to a higher uptake percentage of AF488- RNA rings than L2K (Figure 2B). Furthermore, for the magnitude of NANP uptake, LMM led to lower gMFI for DNA and RNA cubes in the lymphocyte population, while LMM led to lower gMFI for RNA cubes and RNA rings in the monocyte population (Figure 2B,C).

However, while these differences may be statistically significant, biological significance may not follow.

**Figure 2.** Monocytes have greater NANP uptake than lymphocytes regardless of lipofectamine carrier. PBMCs were treated with 10 nM AF488-NANPs for 20 h, fixed, and acquired on the flow cytometer. Cells were gated for lymphocyte and monocyte populations based on side and forward scatter and then gated on the AF488 signal. (**A**) Representative gating strategy for one healthy donor (Q3G6) showing the raw data for the negative controls and AF488-DNA cubes. (**B**) Lymphocytes and (**C**) monocytes were assessed for uptake of AF488-labeled NANPs. The percentage of cells positive for the AF488+ signal (left plots) and the geometric mean fluorescence intensity (gMFI) of AF488+ cells (right plots) were assessed for both populations. Each bar graph represents the mean data ± standard deviation from three healthy donors. Each dot represents the mean for each individual donor (run in duplicate). An asterisk (\*) indicates *p* ≤ 0.05 for paired *t*-test between lipofectamine carriers for a particular NANP type. L2K—LipofectamineTM 2000 reagent; LMM—LipofectamineTM MessengerMAXTM reagent.

#### *2.3. RNA Fibers Delivered with LMM Carrier Decrease IFN Production in PBMCs*

To determine if the lipofectamine-carrier-type-induced changes in NANP uptake affected PBMC biologically, IFN response was determined. Multiplex analysis was used to assess type I (IFNα; IFNβ; IFNω) and type III (IFNλ) interferon production in PBMCs after overnight treatment with 10 nM NANPs delivered using either L2K or LMM. The IFN panel was used in our earlier studies, which identified IFNs as biomarkers of immunostimulation of NANPs delivered using lipofectamine carriers [5,7]. The cytokine levels are presented as a heat map (Figure 3A) and as a bar graph (Figure 3B). It was determined that treatment with RNA cubes led to IFN levels similar to the positive control (ODN2216), and this finding agrees with previous studies (Figure 3) [5]. Furthermore, PBMC treatment with DNA rings, RNA fibers, and RNA rings generally led to lower IFN responses than the positive control.

**Figure 3.** Lipofectamine carrier type alters PBMC IFN production in response to incubation with RNA fibers. PBMCs were treated with 10 nM NANPs for 20 h, and supernatants were collected and analyzed via multiplex for IFN production (IFNα; IFNβ; IFNλ; IFNω). (**A**) Heat map of the different IFN production levels of three healthy donors. Data points for each donor were run in duplicate. (**B**) Bar graphs representing the IFN production levels of three healthy donors. Each bar graph represents the mean data ± standard deviation from three healthy donors. Each dot represents the mean for each individual donor (run in duplicate). An asterisk (\*) indicates *p* ≤ 0.05 or \*\* indicates *p* ≤ 0.01 for paired t-test between lipofectamine carriers for a particular NANP type. NC—negative control (untreated PBMC); PC—positive control (5 μg/mL ODN2216); L2K—LipofectamineTM 2000 reagent; LMM—LipofectamineTM MessengerMAXTM reagent.

When addressing the specific effect of lipofectamine carrier, we determined that lipofectamine carrier type did not affect NANP-induced IFN production except with the RNA fibers. In the case of RNA fibers, delivery of RNA fibers with LMM led to decreased IFN production for all four IFNs tested as compared to L2K (Figure 3B). This difference in IFN production may reflect the design of the LMM carrier, which was optimized to deliver mRNA.

#### **3. Discussion**

The greater degree of NANP uptake in the monocyte population as compared to the lymphocyte population in the presence of a lipofectamine carrier (Figure 2) is consistent with our previous studies [5,7] and data published by other research groups using DNA origami [13]. The uptake of these NANPs in the monocytes was higher than in lymphocytes regardless of tested carriers—L2K, LMM (Figure 2), or dendrimers [7]. Furthermore, while there were differences seen between L2K-mediated and LMM-mediated uptake for a few NANPs in the lymphocyte and monocyte populations, the differences are less than twofold except for AF488 gMFI in monocytes for RNA cubes (Figure 2B,C). Differences less than two-fold are unlikely to lead to a biologically significant change. Moreover, these differences seen in uptake did not correspond to downstream differences in IFN production (Figure 3). We observed decreased PBMC IFN production after treatment with LMMdelivered RNA fibers compared to L2K-delivered RNA fibers. Unfortunately, we did not have AF488-RNA fibers to test RNA fiber uptake in monocyte and lymphocyte populations. Therefore, we do not have data to indicate whether the decrease in IFN production in PBMC from LMM-delivered RNA fibers is due to a lack of NANP uptake or another downstream process.

Interestingly, the current study and one of our earlier studies [7] observed the association of DNA cubes with ~60% of the monocyte population in the absence of a carrier (Figure 2C, left) [7]. This observation was also similar to the study by Du et al. investigating the uptake of DNA origami [13] but in contrast to the initial report by Hong et al., in which the uptake of DNA NANPs by monocytes was detected only in the presence of L2K [5]. We hypothesize that differences in the type of flow cytometer used in these studies may explain the observed discrepancy in the test results. Our current research and reports by Avila et al. and Du et al. utilized digital flow cytometers, which adjust the instrument settings automatically and, thus, are more sensitive at detecting even low fluorescent signal [7,13]. In contrast, the initial study by Hong et al. used a traditional cytometer which involves manual adjustment of instrument settings and often leads to the relocation of objects with weak fluorescence outside of the data collection gates [5]. Furthermore, even though the use of the NovoCyte flow cytometer in our current study revealed ~60% of the monocyte population was positive for DNA cube in the absence of any carrier (Figure 2C, left), this increase was not accompanied by increased AF488 gMFI (Figure 2C, right) nor was it accompanied by detectable IFN production (Figure 3), further suggesting that the association on the individual cell level was relatively low. This could imply that the NovoCyte 3005 (and possibly other digital cytometers with similar properties) is more sensitive than the previously used FACSCalibur [5] in the ability to detect low NANP quantities associated with the cells. A cross-validation between the two instruments would help verify this hypothesis, but it was not feasible because FACSCalibur is no longer available.

#### **4. Materials and Methods**

#### *4.1. Materials*

DNA strands (PCR forward and reverse primers and templates for RNA NANPs and individual oligos for DNA NANPs) and fluorescently labeled oligos (3 - Alexa Fluor 488) were obtained from Integrated DNA Technologies (IDT), Inc. MyTaq Mix, was purchased from Bioline. A DNA Clean & Concentrator kit was obtained from Zymo Research. RQ1 RNase-Free DNase was purchased from Promega (3 u/50 μL). Phosphate-buffered saline (PBS), RPMI-1640 medium, penicillin–streptomycin solution, L-glutamine, ficoll-paque

premium, fetal bovine serum (FBS), and HyPure cell-culture-grade water were all obtained from Cytiva/GE Heathcare Life Sciences (Marlborough, MA, USA). Opti-MEMTM I reduced serum medium and Hank's balanced salt solution (HBSS) were from Gibco (Gaithersburg, MD, USA). Acridine orange (AO)/propidium iodide (PI) staining solution was from Nexcelom Bioscience (Lawrence, MA, USA). Oligodeoxyribonucleotide, a human TLR9 ligand (ODN2216), was from InvivoGen (San Diego, CA, USA). NovoFlow, NovoRinse, and NovoClean were from Agilent Technologies (Santa Clara, CA, USA). LipofectamineTM MessengerMAXTM reagent and LipofectamineTM 2000 reagent were obtained from Invitrogen (Waltham, MA, USA). Paraformaldehyde (PFA) 20% Solution was from Electron Microscopy Science (Hatfield, PA, USA). A custom 4-plex Multiplex (IFNα; IFNβ; IFNλ; IFNω) kit with sample diluent, calibrator 1, calibrator 2, detection solution, streptavidin-HRP, substrate A, substrate B+, and wash buffer was obtained from Quansys BioSciences (Logan, UT, USA).

#### *4.2. NANP Preparation*

All sequences used for NANP preparation are provided in the Supporting Information. DNA templates were amplified via PCR using MyTaq Mix. The DNA Clean & Concentrator kit was used to purify the amplified PCR products, followed by in vitro run-off transcription using T7 RNA Polymerase in 80 mM HEPES-KOH (pH 7.5), 2.5 mM spermidine, 50 mM DTT, 25 mM MgCl2, and 5 mM of each rNTP at 37 ◦C over 3.5 h. Transcription was stopped through adding RQ1 RNase-Free DNase and incubating at 37 ◦C for 30 min. For the purification of RNA strands, denaturing polyacrylamide gel electrophoresis (PAGE, 8%) in the presence of 8 m urea run in 89 mM tris-borate, 2 mM EDTA (TBE, pH 8.2) was run at 13 W for 2 h. UV was used to visualize the RNA bands; the bands were then excised and eluted overnight in 300 mM NaCl, TBE (pH 8.2) at 4 ◦C. To precipitate the RNAs, the elution was mixed with 2.5 volumes of 100% EtOH and stored at −20 ◦C for 3 h. Then, the samples were centrifuged at 10.0× *g* for 30 min at 4 ◦C. The pellet was washed with 90% EtOH for 10 min via centrifugation at 10.0× *g* at 4 ◦C; this step was repeated twice. The pelleted samples were vacuum-dried at 55 ◦C with IR in a CentriVap micro-IR vacuum concentrator (Labconco), then dissolved in HyPure cell-culture-grade water. The concentration of each strand was measured using a NanoDrop 2000 (ThermoFisher) at 260 nm. The fourteen RNA strands were stored at −20 ◦C until use.

All NANPs were assembled in a one-pot thermal anneal through combining each strand in an equimolar ratio with HyPure cell-culture-grade water. The DNA cubes, RNA cubes, AF488-DNA cubes, and AF488-RNA cubes were heated to 95 ◦C for 2 min, then mixed with assembly buffer (89 mM tris-borate (pH 8.2), 2 mM MgCl2, 50 mM KCl) and incubated at 45 ◦C for 30 min, and then stored at 4 ◦C until use. The RNA rings, AF488-RNA fibers, and RNA rings were heated to 95 ◦C for 2 min, snap-cooled on ice for 2 min, mixed with the assembly buffer, and incubated at 30 ◦C for 30 min, then stored at 4 ◦C until use.

#### *4.3. Characterization of NANPs*

Successful assembly of NANPs was confirmed via visualization on 8% native-PAGE (37.5:1 acrylamide:bis-acrylamide). The gel was prepared on a Mini-PROTEAN Tetra Cell system (Bio-Rad), pre-run for 5 min at 150 V with running buffer (89 mM TB (pH 8.2), and 2 mM MgCl2). 2 μL of each sample was mixed with 2 μL loading buffer (Assembly buffer, 30% glycerol, bromophenol blue, xylene cyanol), and loaded per well. The loaded gel was run at 300 V for 30 min in a 4 ◦C cold room. The gel was stained with ethidium bromide (EtBr, 0.5 μg mL−1) for 5 min, then washed twice with double-deionized water (ddiH2O). Then, the gel was imaged using a ChemiDoc MP (Bio-Rad). The Alexa Fluor 488-labeled NANPs' gel was imaged before EtBr staining via the Alexa Fluor 488 setting on the ChemiDoc MP system.

Atomic force microscopy (AFM) imaging of NANPs was performed on a freshly cleaved 1-(3-aminopropyl) silatrane-modified mica surface as previously described [7,14,15]. The AFM imaging was performed in tapping mode on the MultiMode AFM Nanoscope IV system (Bruker Instruments, Billerica, MA, USA).

For TEM imaging, 10 μL of corresponding 1 μM NANP stock and 2 μL of L2K or LMM were repeatedly mixed through pipetting up and down. The complexed samples were incubated at room temperature for 5–30 min. Stock L2K and LMM complexes were used for imaging except for the LMM + RNA fiber, which was diluted 10-fold in water before imaging. Samples were vortexed and 5 μL of each sample was applied to a glow-discharged carbon-coated 200 mesh Cu grid (EMS, Hatfield, PA, USA) for LMM/LMM-NANPs complexes or carbon-coated 400 mesh Cu/Rh grid (Ted Pella, Redding, CA, USA) for L2K/L2K-NANPs complexes and allowed to dry for 1 min at room temperature. Staining with 5 μL of 1% uranyl acetate (EMS, Hatfield, PA for LMM samples or Polysciences, Warrington, PA, USA for L2K samples) was repeated twice followed by final blotting and air-drying the grid. An FEI Tecnai T20 transmission electron microscope operating at 200 kV with a Gatan 2 k × 2 k Eagle camera was used to image the LMM grids and an FEI Talos L120C TEM with Gatan 4 k × 4 k OneView camera was used to image the L2K grids. A bridging experiment analyzing L2K on 200 mesh Cu grids (EMS, Hatfield, PA, USA) and images from an FEI Tecnai T20 transmission electron microscope operating at 200 kV with a Gatan 2 k × 2 k Eagle camera was conducted to verify that differences in instrumentation do not affect the results; the image is included in the Supplementary Materials (Figure S1).

#### *4.4. PBMC Isolation*

Healthy human donor whole blood was collected in li-heparin vacutainers (BD Bio-Sciences) under NCI-Frederick protocol OH9-C-N046. The whole blood was used for PBMC isolation as specified in NCL protocol ITA-10 [16]. In brief, whole blood was diluted with PBS at a 1:1 ratio, layered over ficoll-paque at a ratio of 4:3 (4 mL diluted blood for every 3 mL ficoll-paque), and centrifuged for 30 min at room temperature at 900× *g* with no brake. The mononuclear cell layer containing the PBMCs was then removed, collected, and washed twice with HBSS (centrifuged for 10 min at 400× *g*). The PBMCs were resuspended in complete RPMI-1640 medium (10% heat-inactivated FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine) and counted on a Cellometer using a 1:1 ratio of the cell suspension to AOPI. Once the PBMCs were counted, samples were diluted to 1.25 × <sup>10</sup><sup>6</sup> cells/mL using a complete RPMI-1640 medium.

#### *4.5. Uptake of Alexa Fluor-488 NANPs in PBMCs*

PBMCs were aliquoted into a 96-well round bottom plate with 160 μL cell suspension (1.25 × 106 cells/mL) per well. The AlexaFluor-488 NANPs (AF488-DNA cubes; AF488- RNA cubes; and AF488-RNA rings) and appropriate controls (untreated controls and no-carrier controls) were then prepared in microcentrifuge tubes using Opti-MEMTM I reduced serum medium and lipofectamine reagents. An aliquot of 15 μL of 1 μM stock of appropriate NANPs was combined with 3 μL of lipofectamine reagent (LMM or L2K) or 3 μL of Opti-MEMTM I reduced serum medium for the no-carrier controls and incubated between 5–30 min in the dark at room temperature. The untreated controls consisted of either complete RPMI-1640 media (Complete Media) only or Opti-MEMTM I reduced serum medium (OptiMEM) only. After the incubation, 282 μL of Opti-MEMTM I reduced serum medium was added to each sample (except negative controls) for a total volume of 300 μL and an NANP concentration of 50 nM where applicable. An aliquot of 40 μL prepared sample or control was added to each appropriate well of the prepared 96-well plate with PBMC suspension for a final volume of 200 <sup>μ</sup>L (cells at 1 × 106 cells/mL; NANP at 10 nM final concentration). The 96-well plate was placed in a humidified 37 ◦C/95% CO2 incubator for approximately 20 h.

The PBMC samples were then prepared for acquisition on a NovoCyte 3005 flow cytometer (Agilent Technologies, Inc., Santa Clara, CA, USA). The plate was removed from the incubator and centrifuged for 5 min at 400× *g*. The supernatants from each well were then aspirated and discarded, leaving the cell pellet undisturbed. The samples were

washed twice with 150 μL 1× PBS (centrifuged 400× *g* for 5 min). The cell pellets were then fixed with 2% PFA for 15 min at room temperature and washed twice more with 1× PBS. Each cell pellet was resuspended in 150 μL 1× PBS for acquisition on the flow cytometer. On the NovoExpress software, side-scatter and forward-scatter area and height parameters were selected along with the area and height parameters for the FITC (488) channel. All other parameters remained unselected. Samples were then acquired with the instrumentation and analyzed using GraphPad Prism 9 (Graph Pad Software, Boston, MA, USA) and NovoExpress software version 1.5.6 (Agilent Technologies, Inc., Santa Clara, CA, USA).

#### *4.6. IFN Production of PBMCs after NANPs Treatment*

PBMCs were aliquoted into 96-well round bottom plates with 160 μL cell suspension (1.25 × 106 cells/mL) per well. The NANPs (DNA cubes; RNA cubes; RNA fibers; and RNA rings) and appropriate controls (negative control, positive control (5 μg/mL ODN2216), vehicle controls, no-carrier controls) were then prepared in microcentrifuge tubes. An aliquot of 20 μL of 1 μM stock of appropriate NANPs was combined with 4 μL of lipofectamine reagent (LMM or L2K) or 4 μL of Opti-MEMTM I reduced serum medium for the no-carrier controls and incubated between 5–30 min at room temperature. The vehicle controls consisted of 20 μL Opti-MEMTM I reduced serum medium combined with 4 μL of appropriate lipofectamine reagent. The negative control consisted of Opti-MEMTM I decreased serum medium only. The positive control consisted of 10 μL ODN2216 1 mg/mL stock diluted in 390 μL Opti-MEMTM I reduced serum medium for a 25 μg/mL concentration. After the incubation, 376 μL of Opti-MEMTM I reduced serum medium was added to each sample (except the positive control) for a total volume of 400 μL and a NANPs concentration of 50 nM where applicable. An aliquot of 40 μL prepared sample or control was added to each appropriate well of the prepared 96-well plates with PBMC suspension for a final volume of 200 <sup>μ</sup>L (cells at 1 × <sup>10</sup><sup>6</sup> cell/mL). NANP samples were at a 10 nM final concentration, and the positive control samples were at a final concentration of 5 μg/mL. The well plates were placed in a humidified 37 ◦C/95% CO2 incubator for approximately 20 h. After the incubation, the plates were centrifuged for 10 min at 700× *g*. Supernatant aliquots were then collected in newly labeled 96-well plates and stored at −80 ◦C.

A custom 4-plex Multiplex (IFNα; IFNβ; IFNλ; IFNω) from Quansys BioSciences was then used to analyze the freeze–thawed aliquots according to the manufacturer's manual and NCL Protocol ITA-27 [17]. All reagents needed were included in the kit except for de-ionized water and cell-culture-grade water and prepared when indicated by the manual. In brief, the supernatants were thawed (partially at room temperature and partially at 37 ◦C). The calibration standards were prepared using the sample diluent in a 96-well polypropylene plate. The supernatant samples were diluted 2-fold with the sample diluent. 50 μL aliquots of calibration standards and supernatants were loaded into appropriate wells of the provided multiplex plate and incubated at room temperature for 2 h on a shaker (500 rpm). The multiplex plate was washed 3 times with wash buffer using a plate washer. The detection mix was then added to the multiplex plate and incubated for 1 h at room temperature on a shaker. The multiplex plate was then washed 3 times. A 50 μL aliquot of streptavidin-HRP was added to each well, and the plate was incubated for 20 min on the shaker. The multiplex plate was then washed 6 times, and 50 μL of ChemiLum substrate (Substrate A combined with Substrate B+) was added to each well. The plate was then read using the Quansys ImagePro, and the resulting data were analyzed using Microsoft Excel and GraphPad Prism.

#### **5. Conclusions**

This is the first study to demonstrate that the lipofectamine type of commercial delivery agents can be used as a simple tool to mediate change in the immunorecognition of different NANPs. LMM decreased IFN production in response to RNA fibers, which may be linked

to their linear structure. This sensitivity to lipofectamine carriers could be used to modify PBMC response to NANPs precisely.

**Supplementary Materials:** The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/molecules28114484/s1, Sequences used in this project and Figure S1: Comparison of lipofectamine 2000 (L2K) morphology by TEM performed using different grids.

**Author Contributions:** H.S.N., Y.R., and J.X. conducted experiments and analyzed the data. J.D.C. analyzed the TEM data. M.A.D. and K.A.A. conceived, designed and supervised the study. All authors wrote and reviewed the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** The study was funded in part (M.A.D., J.D.C., J.X and H.S.N.) by federal funds from the National Cancer Institute, National Institutes of Health, under contract 75N91019D00024. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. Research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number R35GM139587 (to K.A.A.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

**Institutional Review Board Statement:** All experiments involving human whole blood were performed according to the IRB-approved NCI-Frederick protocol OH99-C-N046.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data are provided in manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Sample Availability:** Samples of the compounds (NANPs) are available from the authors upon reasonable request.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

### *Review* **Nanomedicine Reformulation of Chloroquine and Hydroxychloroquine**

**David M. Stevens, Rachael M. Crist and Stephan T. Stern \***

Nanotechnology Characterization Laboratory, Cancer Research Technology Program, Leidos Biomedical Research, Inc., Frederick National Laboratory for Cancer Research sponsored by the National Cancer Institute, Frederick, MD 21702, USA; dstevens5654@gmail.com (D.M.S.); cristr@mail.nih.gov (R.M.C.) **\*** Correspondence: sternstephan@mail.nih.gov

**Abstract:** The chloroquine family of antimalarials has a long history of use, spanning many decades. Despite this extensive clinical experience, novel applications, including use in autoimmune disorders, infectious disease, and cancer, have only recently been identified. While short term use of chloroquine or hydroxychloroquine is safe at traditional therapeutic doses in patients without predisposing conditions, administration of higher doses and for longer durations are associated with toxicity, including retinotoxicity. Additional liabilities of these medications include pharmacokinetic profiles that require extended dosing to achieve therapeutic tissue concentrations. To improve chloroquine therapy, researchers have turned toward nanomedicine reformulation of chloroquine and hydroxychloroquine to increase exposure of target tissues relative to off-target tissues, thereby improving the therapeutic index. This review highlights these reformulation efforts to date, identifying issues in experimental designs leading to ambiguity regarding the nanoformulation improvements and lack of thorough pharmacokinetics and safety evaluation. Gaps in our current understanding of these formulations, as well as recommendations for future formulation efforts, are presented.

**Keywords:** chloroquine; hydroxychloroquine; nanomedicine; nanoformulation

**Citation:** Stevens, D.M.; Crist, R.M.; Stern, S.T. Nanomedicine Reformulation of Chloroquine and Hydroxychloroquine. *Molecules* **2021**, *26*, 175. https://doi.org/10.3390/ molecules26010175

Academic Editor: Derek J. McPhee Received: 8 December 2020 Accepted: 29 December 2020 Published: 31 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional clai-ms in published maps and institutio-nal affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

Chloroquine (CQ) and hydroxychloroquine (HCQ) have been used for decades in the prevention and treatment of malaria and in the treatment of some autoimmune diseases such as lupus erythematosus and rheumatoid arthritis due to their immunomodulatory properties [1–3]. Despite being considered old drugs, CQ and HCQ have generated new interest due to their anticancer activity both in preclinical and clinical studies [4,5]. Researchers have shown these drugs act through a variety of antineoplastic mechanisms such as autophagy disruption, tumor vessel normalization, immunomodulation, and inhibition of metastasis, acting both directly on the tumor parenchyma and tumor microenvironment [6,7]. Chloroquines have been shown effective either as monotherapies or as adjunct therapies, sensitizing cancer cells to existing cytostatic agents as well as targeted therapies [7]. For example, HCQ has been shown to synergize with MEK pathway inhibitors for effective treatment of RAS-driven cancers, and CQ has been shown to inhibit melanoma growth through modifying tumor-associated macrophage (TAM) from the M2 immunosuppressive/pro-tumor phenotype to M1 immunostimulatory/antitumor phenotype [8,9].

CQ and HCQ have also recently received worldwide attention due to their potential use in treating coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Previous studies showed in vitro efficacy of these drugs against Middle East respiratory syndrome coronavirus (MERS-CoV) and severe acute respiratory coronavirus (SARS-CoV), and a recent study demonstrated CQ could effectively inhibit viral infection of SARS-CoV-2 in vitro [10–12]. As a result, scientists suggested their assessment in patients, leading to emergency use authorization for HCQ

and the initiation of several clinical trials. However, interest in these drugs sharply declined following a retrospective analysis claiming COVID-19 patients were more likely to die of irregular heart rhythms when taking CQ or HCQ, resulting in revocation of the FDA's emergency use authorization [13,14]. This report was later retracted due to data validity concerns; however, many clinical trials had already been terminated. It should also be noted that recent data have questioned the original in vitro findings supporting inhibition of viral replication by CQ, demonstrating that the CQ-sensitive viral activation mechanism in the Vero cell line utilized was not relevant to human lung cells [15]. For these reasons, the use of these drugs for the prevention or treatment of COVID-19 remains extremely controversial.

CQ and HCQ are both basic amphiphiles that concentrate in the lysosome and inhibit lysosomal function as their primary mechanism of action [16]. While CQ and HCQ also have similar toxicity profiles and are equipotent, chloroquine is much more toxic (2-fold) [16]. Although short-term administration of either drug is generally well-tolerated, except in patients predisposed to arrhythmia, chronic dosing and high-dose regimens can cause severe side effects such as irreversible retinal toxicity [17–19]. CQ and HCQ have similar pharmacokinetic (PK) properties, including high volume of distribution and prolonged plasma half-lives between 40 and 50 days, which requires weeks of dosing to achieve steady-state therapeutic concentrations [20]. Reformulation of CQ and HCQ to improve their PK and safety profile may support the use of these drugs for applications such as cancer and infectious diseases.

Nanoparticle drug delivery is one promising strategy to overcome drug liabilities such as poor PK and toxicity while improving site-specific drug delivery. Nanomedicines can provide a variety of benefits, such as improving the solubility of hydrophobic drugs, protecting drugs from degradation, and altering tissue distribution through passive or active targeting mechanisms [21]. Indeed, various nanomedicines have been developed and clinically approved that enhance the safety and/or efficacy of drugs and legacy formulations [22]. Overall, CQ and HCQ therapy may benefit from reformulation, and this review will discuss the efforts to formulate these drugs through nanomedicine approaches (Figure 1).

**Figure 1.** Nanomedicine formulations of CQ and HCQ. A variety of nanotechnology platforms are being explored in the reformulation efforts of improving the overall safety and efficacy of CQ and HCQ.

#### **2. Liposomes**

Liposomes are spherical vesicles consisting of one or more phospholipid bilayers and are capable of loading drugs within their aqueous core or lipid bilayer. Liposomes are generally very stable with long circulatory half-lives, and changes to their surface chemistry, such as hydrophilic coating (e.g., polyethylene glycol; PEG) or targeting moieties (e.g., antibodies), can result in decreased uptake by the mononuclear phagocytic system (MPS) and site-specific delivery, respectively [23,24]. Liposomal formulations of CQ, with and without erythrocyte-specific antibody targeting fragments, were first developed during the 1980s and provided better suppression of parasitemia compared to unformulated CQ in malaria parasite *P. berghei*-infected animals (Table 1) [25–30]. Despite these early successes, liposomal CQ did not progress toward clinical applications, and only a few liposomal CQ formulations have been published since. For example, Fotoran et al. developed micronsized, multilamellar liposomes for loading CQ through interlayer hydrogen bonding [31]. In comparison to unformulated CQ, this formulation only provided a significant reduction in parasitemia for two of the thirteen-day efficacy study, suggesting only a modest improvement in therapy.

It is worth noting that these studies utilized non-PEGylated liposomes, which are known to be rapidly cleared by resident macrophages in MPS organs such as the liver and spleen [32]. Although this is unfavorable for many applications, since it lowers drug exposure to non-MPS tissues, some researchers have utilized non-PEGylated liposomes as a strategy to increase drug exposure to macrophages and improve treatment of macrophagebased infections. For example, in a *C. neoformans* murine model, liposomal CQ in combination with fluconazole provided better antifungal prophylaxis and treatment compared to free drug controls due to enhanced liposomal drug uptake by macrophages [33,34]. Most modern liposomal formulations contain a PEG surface coating that reduces macrophage clearance and increases circulatory time, which may be desirable for malaria and cancer indications. In one recent example, a CQ formulation using PEGylated liposomes with antibody targeting to the erythrocyte surface protein glycophorin A provided robust CQ delivery to uninfected and Plasmodium-infected red blood cells, resulting in superior efficacy compared to unformulated CQ in *P. falciparum*-infected mice [35]. Overall, these studies support the use of liposomal formulations for delivering CQ to erythrocytes and macrophages for malaria and antifungal applications, but additional PK and toxicology studies would be informative to evaluate their safety profile moving forward.

Liposomes initially found clinical success as drug carriers in cancer treatment with the development of Doxil® (liposomal doxorubicin), which reduced the drug's dose-limiting cardiotoxicity and increased tumor exposure due to the enhanced permeability and retention (EPR) effect [36]. The EPR effect concept was first introduced by Matsumura and Maeda et al. in 1986; this ability of nanoparticle-based formulations to accumulate in tumor tissue is now widely recognized and was recently reviewed by Price et al. [37,38]. In particular, liposomes have become a commonly used formulation to passively target one or multiple drugs to tumors. Due to CQ's anticancer activity, researchers have developed liposomal formulations combining CQ and other chemotherapeutics for enhanced anticancer efficacy. For example, liposomes co-loaded with CQ and paclitaxel (PTX) or doxorubicin (DXR) resulted in tumor growth suppression in A549/T-tumor-bearing mice and MCF-7/ADR-tumor-bearing zebrafish, respectively [39,40]. However, the authors did not compare to unformulated drug controls in the efficacy or drug distribution studies, and therefore, it is unclear if the liposomal formulations provided any benefits to CQ delivery, a major shortcoming of these studies.



**1.**NanomedicineformulationsofCQandHCQtestedin

*Molecules* **2021**, *26*, 175


**Table 1.** *Cont.*

Bis-MPA: 2,2-bis(hydroxymethyl)propionic acid; Chol: cholesterol; DOPC: 2-dioleoyl-sn-glycero-3-phosphocholine; DPGG: 1,2-dipalmitoyl-galloylglycerol; DSPC: 1,2-distearoyl-sn-glycero-3-phosphocholine; DSPE-PEG2000: 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000]; DSPE-PEG2000-Mal: 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[maleimide(polyethylene glycol)-2000]); MPB-PE:maleimido-4-(p-phenylbutyrate)-phosphatidylethanolamine; mPEG: methoxypoly(ethylene glycol)-b-poly(lactic acid); PC: phosphatidylcholine; PCL: polycaprolactone; PEG-PLL: poly(ethylene glycol)-block-poly(L-lysine); PEI: polyethylenimine; PG: phosphatidylglycerol; PLA: polylactic acid; PLGA: poly(lactic-co-glycolic acid); PS: phosphatidylserine.

Several HCQ-loaded liposomes have also been developed for cancer indications. For example, Wang et al. combined HCQ-loaded liposomes with TAT-Beclin 1 peptide to induce autophagy catastrophe in a 4T1 breast cancer model [41]. The combination treatment significantly reduced tumor growth compared to unformulated drug controls and HCQ liposomes alone, suggesting improved liposomal HCQ tumor exposure and supporting the strategy of inducing autophagy catastrophe in tumor cells to treat cancer. A similar strategy was used to combine HCQ-loaded liposomes with Salmonella VNP20009 antitumor peptide in a B16F10 melanoma xenograft model [42]. The liposomes increased HCQ concentrations 4-fold within the tumor compared to the free drug control 24 h following injection, with no difference in liver or spleen concentrations at this same time point. This improvement in tumor drug exposure resulted in 90% survival in comparison to 20% survival for the free drug HCQ + VNP20009 combination control group, and no survival in the HCQ, VNP20009, or HCQ-liposome only groups. These studies strongly support the use of liposome formulations to increase HCQ delivery to the tumor site and improve efficacy when combined with other anticancer drugs.

To further improve the delivery of HCQ to tumors, liposomes have been modified with various targeting ligands to enable tumor-specific drug delivery. For example, liposomes decorated with pH-sensitive RGD peptides for targeting ITGAV-ITGB3/integrin αvβ3 receptors were used for HCQ delivery to melanoma tumors [43]. Both untargeted and targeted versions of the liposomes significantly decreased drug exposure to the heart, spleen, lung, and kidney compared to unformulated drug control, while only the targeted liposome version significantly increased HCQ concentrations within the tumor 24 h post-injection. As a monotherapy, the formulation achieved a median survival of 30 days compared to 25 days from the untargeted liposome treatment group and 15 days from HCQ free drug control. However, when combined with liposomes containing DXR, the median survival improved to >60 days, and tumor growth was significantly inhibited compared to free drug controls or DXR liposomes only. This same ITGAV-ITGB3/integrin αvβ3 receptor-targeted liposome formulation was also used to co-deliver HCQ and PTX for pancreatic cancer therapy [44]. This formulation achieved significantly better tumor growth inhibition and reduction of metastatic tumor nodules in a BxPC-3-luc orthotopic tumor model compared to targeted liposomes containing either HCQ or PTX, untargeted liposomes containing both drugs, and PTX + HCQ free drug control while not affecting body weight. This ITGAV-ITGB3/integrin αvβ3 receptor-targeted liposomal formulation not only significantly changed HCQ distribution toward the tumor, but also provided excellent anticancer efficacy when combined with chemotherapeutics.

In addition to integrin αvβ3 receptors, Yin et al. also targeted neuropilin-1 receptors on melanoma cells for co-delivery of HCQ and PTX [45]. This targeted liposomal formulation significantly inhibited tumor growth and effectively inhibited metastasis in a B16F10 melanoma model compared to an untargeted liposome version and unformulated drug controls. Another liposomal formulation, also targeting the integrin αvβ3 and neuropilin-1 receptors was co-loaded with HCQ and tyrosine kinase inhibitor ZD6474 and evaluated for efficacy in a C6 glioma model [46]. Interestingly, these liposomes did not significantly change drug exposure to the heart, liver, spleen, lung, or kidney compared to untargeted liposome version or free drug controls, but they did achieve a 4.9-fold increase in drug exposure to the brain in C6 intracranial tumor-bearing mice. This improvement in drug delivery across the blood-brain barrier (BBB) resulted in significantly prolonged median survival time with the targeted, co-loaded liposomes (41 days) compared to the untargeted liposome version (35 days) and unformulated, free drug controls (28 days).

Ultrasound (US) is another method that has been investigated as a means to improve nanoparticle delivery across the BBB and is also involved in sonodynamic therapy [70,71]. This strategy was used to improve the delivery of HCQ and sonoactive chlorin e6 to glioma tumors using angiopep-2 peptide-modified liposomes that target low-density lipoprotein receptor-related protein 1 (LRP1) [47]. Combined with ultrasonic pulse, the targeted liposome containing both drugs achieved the greatest median survival time of 52 days

compared to 40 days from the untargeted liposome version and 33 days from chlorin e6 + HCQ unformulated drug controls. Combining autophagy inhibitors with sonodynamic therapy through targeted drug delivery to brain tumors may offer a novel therapeutic strategy for glioma.

Drugs that are not suitable for remote loading into the liposomal aqueous core or are not sufficiently lipophilic to associate with the lipid bilayer can be conjugated to a lipid anchor to facilitate loading within the lipid bilayer [72]. Although researchers have shown HCQ can be successfully incorporated in the liposomal aqueous core with high drug loading, Liu et al. developed a liposome bilayer-loaded cholesterol-modified version of HCQ for the treatment of pulmonary fibrosis [48]. Both cholesterol-modified HCQ liposomes and core-loaded HCQ liposomes inhibited the development of bleomycininduced pulmonary fibrosis in Sprague-Dawley rats; however, the authors did not compare to unformulated HCQ, so the benefits of using a liposomal bilayer-loaded cholesterolmodified HCQ formulation remain unclear.

#### **3. Polymeric Nanoparticles**

Polymer-based nanoparticles have been used to improve the solubility of hydrophobic drugs and facilitate enhanced tumor distribution through the EPR effect. Polymeric micelles, one of several different types of polymer-based nanoparticles, generally consist of amphipathic polymers that co-precipitate with drugs to form a hydrophobic core surrounded by a hydrophilic shell. These formulations have been shown to have low critical micelle concentrations (CMC) and have better stability than traditional surfactant micellar systems due to hydrophobic interactions between the drug and polymer [73]. CQ is a hydrophobic drug with a high logP of 4.72 and is predicted to be suitable for polymeric micelle formulations based on previous analysis of how drug properties influence nanomedicine compatibility [74,75]. Despite this, few examples of CQ-polymeric micelles have been reported. In one study, micelles composed of methoxy PEG-b-poly(L-lactic acid) (mPEG-PLA) were used to co-load CQ with either DXR, PTX, or cis-platin [49]. In all cases, the micellar formulations provided superior efficacy in ovarian cancer models compared to unformulated drug combinations, indicating improved tumor distribution.

In addition to micelles, polymeric nanoparticles can be formed through emulsion techniques. This approach can be used to encapsulate hydrophilic drugs and biologics within the polymer matrix and do not require amphipathic polymers. For example, Yang et al. developed a nanoparticle composed of poly(lactic-co-glycolic acid) (PLGA) for co-delivery of CQ and pDNA expressing the mSurvivin-T34A protein [50]. In this case, CQ was used for pDNA compaction through electrostatic interactions as well as for improving lysosome escape of the pDNA following cell uptake. This formulation provided better tumor growth inhibition compared to pDNA/PLGA nanoparticle without CQ in a CT26 tumor model. However, the authors did not compare to a CQ free drug control or to pDNA/PLGA + CQ administered separately, so it is unclear if the improvement in efficacy is due simply to the addition of CQ or to an improvement in CQ drug delivery.

HCQ has also been formulated with biologics to aid in lysosome escape. For example, Liu et al. developed a PLGA nanoparticle co-loaded with HCQ and ovalbumin (OVA) as a model antigen for a proof-of-concept vaccine delivery formulation [51]. This formulation provided statistically significant tumor growth inhibition in an OVA-sensitive E.G7-OVA xenograft tumor model compared to free OVA or OVA-nanoparticles alone, but the authors did not include controls for unformulated HCQ administered alone or in combination with OVA-nanoparticles. Further studies are required to determine if there is a benefit to formulating CQ or HCQ to facilitate cytosolic delivery of biologics, or if the same effects can be achieved by simply administering the drugs separately.

Similar to liposomes, polymeric nanoparticles can be coated with antibodies to enable tumor-specific drug delivery, but few have been developed for CQ or HCQ. In one example, a cd20-antibody-targeted poly(caprolactone)/PLA nanoparticle was co-loaded with HCQ and chlorambucil and evaluated for efficacy in a Burkitt lymphoma animal model [52]. The targeted nanoparticle provided 90% survival after 120 days compared to 40% survival in animals treated with the antibody alone and 0% survival in animals treated with untargeted, drug-loaded nanoparticles or free drug combination controls. Interestingly, at non-toxic doses, the untargeted version of the nanoparticle provided worse survival (0%) compared to the free drug combination control (33%), indicating an untargeted polymeric nanoparticle may unfavorably change tissue distribution of these drugs.

Although most polymers used in drug delivery are biodegradable, some non-biodegradable polymers such as acrylamide-based polymers have shown success for small molecule and oligonucleotide delivery [76,77]. One major advantage of acrylic polymers is the wide selection of functionalized monomers available to form polymers with different physicochemical properties. For example, poly(N-isopropylacrylamide-co-acrylic acid) (PNIPAM-Aac) is a negatively charged polymer that can undergo electrostatic complexation with positively charged molecules. This approach was used to co-load CQ and DXR within PNIPAM-Aac nanogels to induce autophagy catastrophe within tumor cells [53]. Despite successful drug loading, CQ release in PBS was rapid, with more than 50% in the first two hours and more than 95% over 12 h. This rapid drug release is likely too fast to benefit from any passive tumor targeting of the nanoparticle. Indeed, in an efficacy study in an MCF-7 breast cancer model, the nanoparticles containing both drugs did not achieve a statistically significant decrease in tumor weight compared to CQ-only nanogels. The authors also did not compare to a free drug DXR + CQ control to prove the benefit of nanoparticle delivery.

Overall, due to a lack of appropriate controls, there is limited data to support the utilization of polymeric nanoparticles for improving the delivery of either CQ or HCQ.

#### **4. Dendrimers**

Dendrimers are repetitively branched molecules generally constructed as macromolecular polymers with variable cores and terminal groups to facilitate drug encapsulation and drug delivery [78]. Properties such as size, morphology, and surface chemistry can be controlled through synthetic chemistry steps and designed for specific drug delivery needs. To improve CQ delivery to Plasmodium-infected red blood cells, Marti Coma-Cros et al. designed cationic dendrimers based on Pluronic F127 and 2,2 -bis(glycyloxymethyl)propionic acid as well as a hyperbranched dendrimer derived from 2,2 -bis(hydroxymethyl)propionic acid [54]. Although both dendrimer formulations were capable of loading CQ and demonstrated parasite growth inhibition in vitro, they provided worse survival outcomes (20%) in *P. yoelii*-infected mice compared to CQ control (80%), indicating the formulations significantly reduced the antimalarial efficacy of CQ. One possible explanation for this decrease in efficacy could be due to a reduction in systemic drug exposure. Previously, dendrimers composed of PEG and poly(lysine) with and without galactose terminal groups significantly reduced the maximum concentration (Cmax) and area under the concentration-time curve (AUC) of CQ in comparison to unformulated CQ [55]. A similar CQ-loaded PEGpoly(lysine dendrimer) with a chondroitin sulfate A coating also significantly reduced Cmax compared to free drug (13.85 and 50.23 μg/mL, respectively), but increased AUC from 74.72 to 120.58 μg\*h/mL; however, in this case the differences in PK were likely due to the routes of administration, since the unformulated drug was administered intravenously and the dendrimer formulation was administered intramuscularly [56].

Alternatively, Panagiotaki et al. designed dendrimers composed of poly(ethylenimine) with triphenylphosphate terminal groups to facilitate mitochondrial delivery of DXR and CQ for improved cancer therapy [57]. Dendrimer formulations were developed for each drug and, when administered together, significantly reduced tumor volume in DU145 tumor-bearing mice. However, the efficacy was only slightly better than the DXRonly dendrimer, and the authors did not compare to a CQ-only dendrimer formulation or DXR + CQ free drug control. Therefore, it is unclear whether this formulation provided any benefit to the delivery or anticancer efficacy of CQ.

#### **5. Polyelectrolyte Complexes**

Polyelectrolyte complexes, also sometimes referred to as polyplexes and coacervates, are formed by mixing oppositely charged polyionic species in an aqueous medium, and various ionic polymers have been investigated extensively for their ability to complex with nucleic acids [79]. However, their use for delivering small molecule drugs has been limited, likely due to the necessity of multiple charge sites per drug molecule to allow stable complexation with the polymer.

CQ is positively charged at physiological pH due to its two ionizable amine groups, and because of this, researchers have attempted to load the drug into complexes containing ionic polymers. In one example, Urban et al. developed poly(amidoamine) polymers that formed ~10 nm complexes when mixed with CQ [59]. Drug release from the formulations in PBS was nearly identical to unformulated CQ, indicating formulation instability. Surprisingly, *P. yoelii*-infected mice treated with the polymer/CQ complexes achieved 100% survival 30-days post-infection compared to 0% survival in the unformulated CQ control group. Although the polymers alone were shown to reduce parasitemia in vitro, polymer-only controls were not included in the in vivo efficacy study. Therefore, it is unclear whether the improved survival is due to an improvement in CQ delivery or rather due to additive or synergistic effects of the drug and polymers. Furthermore, the formulations provided no statistically significant improvement in survival compared to CQ alone in *P. yoelii*-infected mice when administered orally [58].

Another CQ-polyelectrolyte complex, composed of chitosan and tripolyphosphate, was shown to reduce parasitemia to a greater extent than unformulated CQ in several efficacy studies in *P. berghei*-infected mice [60–63]. However, the authors did not use vehicle-only controls in any of the studies to rule out the possible antimalarial activity of the polymer complex itself. Overall, these studies support the use of combining ionic polymers with CQ to improve malaria treatment since there is evidence of better survival outcomes and reduced parasitemia, possibly due to additive effects between CQ and the ionic polymers, rather than improved delivery to target cells.

#### **6. Non-Liposomal Lipid-Based Nanoparticles**

In addition to liposomes, there are a variety of other lipid-based nanoparticles including solid lipid nanoparticles (SLN), nanoemulsions, and niosomes. These formulations are generally used for improving the solubility and delivery of hydrophobic drugs and are highly biocompatible and biodegradable due to their physiological lipid compositions.

Unlike other lipid-based carriers, SLN contain a solid lipid core and are often utilized as oral formulations to improve solubility and intestinal absorption of hydrophobic drugs [80,81]. CQ is typically administered orally and has highly variable bioavailability ranging from 52% to 102% as an oral solution and 67–114% as a tablet [82]. It has also been shown that taking CQ with food results in significantly higher Cmax and AUC, and it is recommended to avoid an upset stomach during CQ dosing [83]. Despite having high oral bioavailability, Bhalekar et al. attempted to improve CQ oral delivery and intestinal lymphatic uptake using a SLN formulation for arthritis therapy [64]. The SLN formulation achieved 2-fold increases in Cmax, time of maximum concentration (Tmax), and AUC in comparison to standard CQ suspension, reportedly due to intestinal lymphatic uptake and bypassing first-pass metabolism. Consequently, the SLN formulation achieved greater paw volume reduction compared to the standard CQ suspension in the arthritis mouse model.

In addition to loading drugs, lipid-based carriers have been shown to inhibit malarial parasitemia in erythrocytes [84]. Due to these properties, Baruah et al. developed CQ-loaded, cationic nanoemulsions to improve antimalarial efficacy [65]. The formulation suppressed parasitemia by 99.68% compared to only 76.5% by unformulated CQ in *P. berghei*-infected mice 5 days post-infection. However, the blank lipid emulsion reduced parasitemia by 35.35%, indicating the lipid emulsion alone inhibited malarial infection. Therefore, it is unclear if the efficacy from the CQ nanoemulsion is due to an improvement in drug delivery or simply additive or synergistic effects with the lipid emulsion and drug.

Niosomes are another class of drug delivery vehicle capable of loading both hydrophobic and hydrophilic drugs. Niosomes are similar to liposomes in that they also contain a bilayer and an aqueous core. Unlike liposomes, which typically utilize phospholipids, niosomes are formed from mixtures of non-ionic surfactant molecules and cholesterol. Niosomes have been used for transdermal drug delivery due to their ability to improve drug penetration through the skin and provide local and sustained drug release [85]. This strategy was used to develop a HCQ-loaded niosome formulation dispersed in a Pluronic F-127 gel for the treatment of oral lichen planus [66]. Human patients applied the niosome gel with or without the drug (placebo group) to their lesion every day for four months. Patients receiving the HCQ-containing gel observed an average lesion size reduction of 64.28% compared to only 3.94% reduction in the placebo group. On a pain score from 0 to 10, where 0 is no pain and 10 is the worst pain, patients in the gel and placebo groups reported pain scores of 4 and 3 pre-treatment and 1 and 3 post-treatment, respectively. Although these data support the benefits of this HCQ niosome gel in human patients, the authors did not compare to HCQ gel control, HCQ free drug control, or standard of care (corticosteroids). Therefore, it is unclear whether encapsulation within niosome provided any benefits to the delivery of HCQ.

#### **7. Metal Nanoparticles**

Metallic nanoparticles have been successfully implemented as contrast agents and many are being investigated as therapeutic agents and drug delivery vehicles [86,87]. One of their limitations for drug delivery is the requirement of functional groups on the drug that can undergo chelation with metals. For example, thiol-containing drugs can be conjugated to the surface of gold nanoparticles through Au-thiol bonding. Upon cell entry, thiol-exchange with intracellular glutathione releases the drug. Drugs without thiol groups must be chemically modified as prodrugs in order to conjugate to gold nanoparticles and allow the release of the parent drug. Ruan et al. used this strategy to modify DXR and HCQ as ester prodrugs containing terminal thiol groups to enable coupling to gold nanoparticles and evaluated these nanoparticles for antiglioma efficacy [67]. The nanoparticles containing both drugs resulted in a 56-day median survival in C6 glioma-bearing mice compared to 44 days from nanoparticles containing only DXR; however, the results were not statistically significant. The nanoparticles containing only HCQ resulted in a 38-day median survival compared to 30 days from the free HCQ treatment group, though a better control would have been the modified version of HCQ since this is the molecule that is released from the gold nanoparticle. The authors described in vitro DXR release in PBS at acidic pH, but they did not investigate HCQ release, and drug release in plasma would be a better predictor of nanoparticle stability in vivo since plasma contains both glutathione and esterase enzymes. Therefore, the stability of the HCQ prodrug and its chelation with the nanoparticle surface are unclear.

HCQ has also been used to enhance sonodynamic therapy of metallic nanoparticles through autophagy disruption. For example, Feng et al. designed HCQ-loaded hollow mesoporous titanium dioxide nanoparticles that are coated with a cancer cell membrane to allow homologous targeting to the tumor [68]. HCQ release in PBS from coated nanoparticles was much slower than that of uncoated nanoparticles, but the release became equivalent to the uncoated particles when exposed to US irradiation, suggesting a US responsive drug release mechanism. In MCF-7 tumor-bearing mice, the cancer cell membrane coated nanoparticles extended the systemic half-life of HCQ to 12.3 ± 1.7 h, which was higher than that of uncoated nanoparticles (8.7 ± 1.3 h) and free HCQ (3.4 ± 0.4 h). However, it is unclear if the authors measured the total drug fraction in the blood or the released (pharmacologically active) fraction. The PK of nanomedicines is very complex since total drug concentration in the plasma and blood, as well as tissues, is comprised of encapsulated and unencapsulated drug fractions, and both fractions can contribute to drug efficacy and toxicity [88]. Nevertheless, the cancer cell membrane coated nanoparticles containing HCQ combined with tumor US irradiation significantly reduced tumor growth

compared to empty nanoparticles + US and free HCQ controls, supporting the strategy of combining US with autophagy disruption. However, the degree to which the nanoparticle improved HCQ exposure of the tumor site remains unknown, and treatment of HCQ + nanoparticle + US may have been just as effective.

#### **8. Conclusions and Perspectives**

There is new interest in repurposing CQ and HCQ for novel applications such as cancers, as well as improving therapy for their traditional indications such as infectious and inflammatory diseases. Nanomedicines have been evaluated for their ability to improve the safety and efficacy of chloroquines. There are a variety of nanoparticle types, with each having their own advantages and disadvantages, and it is important to understand the liabilities and physicochemical properties of the drug being formulated in order to select the most appropriate platform. In the case of CQ and HCQ, off-target toxicities can be reduced, and efficacy enhanced using a combination of site-specific drug delivery and controlled release; the balance between delivery and release kinetics being a crucial factor in improving therapeutic index [89]. In order to achieve this, researchers have tested nearly every type of nanomedicine available, with many failing to conclusively demonstrate benefits to CQ or HCQ therapy.

Polymeric nanoparticles, which have been successful in formulating hydrophobic drugs in preclinical and clinical studies, are typically unstable formulations that release their drug immediately after injection, thereby eliminating any potential benefits of nanoparticle distribution and essentially acting as solubilizing formulations. For example, Genexol® PM, a polymeric nanoparticle formulation of PTX that is approved as a cancer therapy in South Korea, has been shown to completely release its drug within 10 min after exposure to plasma [90]. Dendrimers and polyelectrolyte complexes have shown promising preclinical results for gene delivery but have been less successful in formulating small molecule drugs. Dendrimer-drug conjugates of chemotherapeutics are currently undergoing clinical trials, and this may prove to be a more useful strategy since drug release stability is controlled through the linker chemistry [91,92]. Metallic nanoparticles have been approved as contrast and therapeutic agents, but none have proven useful for improving the delivery of small molecule drugs, likely due to insufficiently stable drug-metal interactions. All of these nanoparticle types have been used to reformulate CQ and HCQ, but most have not provided sufficient evidence of improving their efficacy and safety profile. In many cases, appropriate controls were missing, and it was unclear if the efficacy of CQ and HCQ was due to an improvement in drug delivery or if the same results could be achieved using the unformulated drugs. Therefore, additional PK and efficacy studies with appropriate controls are needed to support the use of these nanomedicine formulations for CQ or HCQ delivery. Further, toxicity studies are also rarely performed on these formulations and are necessary for evaluation of improvements to the therapeutic index overall.

On the other hand, liposomal formulations appeared to provide a clear benefit to the delivery of CQ and HCQ in various malaria and tumor models, respectively. With its ionizable amine groups, CQ can be actively loaded into the aqueous liposomal core, and erythrocyte-specific targeting ligands on the surface of the liposomes improve drug uptake within red blood cells, a target for malaria. Since lipids have been shown to inhibit Plasmodium infection, combining CQ with lipid-based carriers may provide not only better drug delivery to uninfected and infected red blood cells, but also synergistic efficacy. Liposomes also make a good choice for improving the delivery of these drugs to tumors. With their ~100 nm size and good stability, liposomes are able to accumulate within the tumor microenvironment via the EPR effect and deliver their therapeutic cargo [93]. With the help of targeting peptides on their surface, liposomes were able to co-deliver HCQ and other chemotherapeutics to significantly improve efficacy and survival outcomes and appear to be a promising strategy for cancer therapy moving forward. However, one disadvantage of liposomes is that they are generally very stable with extremely long drug release half-lives. For example, Doxil has a drug release half-life greater than 100 h,

and there are currently efforts to design less stable liposomes that provide faster drug release rates at the site of interest [94–96].

One notable absence in the above nanotechnology formulation discussion of chloroquines is polymer prodrug systems, a major drug delivery class that has scarcely been evaluated for these drugs and may offer an ideal balance of targeting and stability. Polymer prodrugs can be designed to be biodegradable, provide site-specific targeting, and enable controlled drug release through the polymer-drug linker chemistry [97]. This strategy has proven useful for the delivery of small molecule drugs for cancer and neurological diseases, and there are several candidates in clinical trials [98]. To our knowledge, only a single example of a polymer prodrug of HCQ evaluated in vivo has been published, and it demonstrated substantially better efficacy and lower toxicity compared to unformulated HCQ in a mouse model of colitis [69].

It should be emphasized that despite the promising preclinical data for some of the formulations presented in this review, none of the formulations have made it to the clinical stage. The lack of clinical development is likely due to poor intellectual property protections and uncertain commercial promise for the formulation platforms presented, many of which rely on generic formulation strategies. However, it is expected that the recent commercial success of novel nanotechnology-based delivery platforms and renewed interest in chloroquine drugs for novel indications, such as cancer, will fuel future clinical development of chloroquine nanoformulations [9,99]. Overall, reformulation efforts of CQ and HCQ through nanomedicine approaches have shown some promising improvements in efficacy and safety, but further developments are warranted.

**Author Contributions:** Writing—original draft preparation, D.M.S. and S.T.S.; writing—review and editing, D.M.S., R.M.C. and S.T.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has been funded in whole or in part with Federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. 75N91019D00024. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** No new data were created or analyzed in this study. Data sharing is not applicable to this article.

**Acknowledgments:** The authors thank Joseph Meyer (Leidos Biomedical Research, Inc.) for the figure.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Review* **Dissecting Functional Biological Interactions Using Modular RNA Nanoparticles**

**Kaitlin Klotz 1, Yasmine Radwan <sup>2</sup> and Kausik Chakrabarti 1,\***


**Abstract:** Nucleic acid nanoparticles (NANPs) are an exciting and innovative technology in the context of both basic and biomedical research. Made of DNA, RNA, or their chemical analogs, NANPs are programmed for carrying out specific functions within human cells. NANPs are at the forefront of preventing, detecting, and treating disease. Their nucleic acid composition lends them biocompatibility that provides their cargo with enhanced opportunity for coordinated delivery. Of course, the NANP system of targeting specific cells and tissues is not without its disadvantages. Accumulation of NANPs outside of the target tissue and the potential for off-target effects of NANPmediated cargo delivery present challenges to research and medical professionals and these challenges must be effectively addressed to provide safe treatment to patients. Importantly, development of NANPs with regulated biological activities and immunorecognition becomes a promising route for developing versatile nucleic acid therapeutics. In a basic research context, NANPs can assist investigators in fine-tuning the structure-function relationship of final formulations and in this review, we explore the practical applications of NANPs in laboratory and clinical settings and discuss how we can use established nucleic acid research techniques to design effective NANPs.

**Keywords:** nucleic acid nanoparticle; RNA motif; RNA domain; SHAPE analysis

#### **1. Introduction**

Nucleic Acid Nanoparticles (NANPs) are a subtype of therapeutic nucleic acids (TNAs) exclusively composed of specialized oligonucleotides designed to carry out defined architectures and functions such as delivery of therapeutic agents, biosensing, and immunostimulation [1]. NANPs can be specified by the designer to deliver functional groups capable of modulating their biological activities while adding regulatory control to the intended function of the NANP [2,3]. NANPs can be engineered to associate with specific targets which make them useful in diagnostics as well as the targeted delivery of therapeutic agents to detect and combat disease with fewer off-target impacts than experienced with traditional delivery systems [4]. For that, the sequences amenable to interactions with receptors on cellular surfaces to facilitate NANP uptake through receptor-mediated endocytosis [5]. A 2015 study by Narayan and colleagues established that class A scavenger receptors have enhanced affinity for spherical nucleic acid nanoparticles conjugates exhibiting a high guanine content [6]. High guanine content in the oligonucleotides of the conjugates facilitated adoption of a secondary structure that facilitated uptake of nanoparticles carrying camptothecin by A549 (human lung adenocarcinoma) cells which resulted in the significantly diminished viability of the cancerous cells seven days after treatment with a G-rich spherical conjugates [6].

NANPs are often used as a method of getting nucleic acids past "barriers" that exist in the body. Typically, carrier-free, naked, exogenous nucleic acids introduced without

**Citation:** Klotz, K.; Radwan, Y.; Chakrabarti, K. Dissecting Functional Biological Interactions Using Modular RNA Nanoparticles. *Molecules* **2023**, *28*, 228. https:// doi.org/10.3390/molecules28010228

Academic Editor: Maria Camilla Bergonzi

Received: 8 December 2022 Revised: 22 December 2022 Accepted: 25 December 2022 Published: 27 December 2022

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

any chemical modifications would meet one of two fates within the body: rapid nucleasemediated degradation and/or renal clearance. The engineered larger NANPs or NANPs mixed with the delivery agents assemble in such a way that the cargo they are delivering is protected from degradative nuclease activity which would otherwise stop the NANPs from achieving their therapeutic purposes. Additionally, the architectural parameters (size, shape, composition, functionalization with TNAs, etc.) of NANPs can be engineered in a way that is non-immunogenic and allows the payload to be shuttled to its intended target without triggering any immunological responses [7]. Once the problems of nuclease mediated degradation and immunogenicity are dealt with, NANPs can deliver gene regulatory RNA interference (RNAi) inducers to decrease the expression of overexpressed genes in pathogenic settings [4,8].

#### *1.1. Nucleic Acid Nanoparticle Design and Functionalization*

There is more than one way to design NANPs and the method used to design a particular NANP depends on the material delivered as well as its action upon delivery [9–11]. The physical properties of individual NANPs determine how well the NANP interacts with its intended target and consequently, impacts the efficacy of the NANP's payload delivery. Several computer-assisted approaches allowing for designing RNA and DNA NANPs have been introduced and explored [9,12–17]. Once a particular NANP is selected for further use, the addition of functional moieties to its structure can be achieved through a direct extension of 5 - or 3 -ends of individual strands that enter the NANP's composition. One way in which NANPs demonstrate specificity is through aptamers. Aptamers are singlestranded oligonucleotides that can adopt specific conformations [18]. Due to their specific conformational arrangements, aptamers can interact with cell components, namely receptors and ligands, in a level of specificity like that of antibodies to elicit a response [18,19]. Their ability to modulate pathways within cells makes aptamers an attractive candidate for therapeutics-particularly in an antagonistic indication [18].

In terms of NANP design, aptamers can be added to NANPs to enhance the specificity of NANP binding. This binding of NANPs can be to whole cells (on the cell surfaces), as well as proteins, and in some special cases, viruses [20]. NANPs provide therapeutic agents access to cells without leaving those therapeutic agents subject to degradation and less prone to triggering an immune response. The aptamers, which have undergone several rounds of Selective Evolution of Ligands by Exponential Enrichment (SELEX), are the key to getting the therapeutic agent to exactly where it is needed [19,20]. We can think of aptamers like a ZIP code on a letter such that the contents of the NANP could go to several different places, but when the aptamer is added, the potential delivery locations are narrowed down to a specific place within the cell.

#### *1.2. Selection of Nucleic Acid for NANPs*

In the previous section, it was stated that NANP design is influenced by the cargo that is delivered as well as the intended action. The same dependence on cargo and action must be considered when selecting the nucleic acid for a NANP. Both DNA and RNA NANPs have been developed for therapeutic and research applications, but the two have several distinct qualities that make them unique from one another. In general, DNA is typically regarded as the more stable of the nucleic acids, however, that does not mean that RNA is the "lesser" nucleic acid. In fact, RNA has some properties that make it favorable for use in NANPs.

A clear distinction between DNA and RNA nanoparticles is the base pairing capabilities of each nucleic acid. DNA is confined to Watson–Crick base pairing, meaning that A binds with T and C binds with G. While RNA often does form Watson–Crick base pairs, it also participates in non-canonical base pairing, such as G:U, G:A or C:A type as found in RNA structural folds [21]. These noncanonical base pairs play critical roles in RNA-folding to establish the three-dimensional structures required for diverse functions of RNA. Thus, RNA's ability to form non-canonical base pairs allows it to adopt motifs with

discrete structure and function, setting it apart from DNA [21]. The motifs present in RNA provide it with enhanced thermal stability, which can be variable in relation to base pairs in DNA, creating differentials in thermal stability within the same DNA sequence [21,22]. This stability is further reflected in the free energy of RNA-RNA helices [21]. The free energy for RNA-RNA helices is the lower that of DNA-RNA hybrid helices and DNA-DNA helices [21,23,24].

For a NANP to carry out its intended function, it must be recognized by a receptor on the exterior of a target cell and endocytosed into the cell. Upon endocytosis, the NANP is inside of an endosome that functions to degrade molecules and reuse their components [21,25,26]. The endosome is acidic to aid in the degradation of endocytosed molecules [25–27]. In the case of DNA, this low pH environment leads to depurination-a process in which adenine and guanine (the purines) are protonated and subsequently lost from the sequence, leaving the DNA lacking purines and susceptible to degradation [21,28,29]. On the other hand, RNA is tolerant of lower pH levels than DNA, which will facilitate successful delivery through the endosome [21].

In summary, RNA nanoparticles are more stable thermodynamically and to pH ranges. Additionally, they have greater structural flexibility. When considered together, these features make RNAs preferred over DNAs as NANPs.

#### *1.3. Small Interfering RNA (siRNA)*

Small interfering RNA (or siRNA) is a short, non-coding, regulatory RNA [30]. Like other small RNAs, siRNAs are generally 20–30 nucleotides in length, and they can modulate the expression of genes [30,31]. siRNAs are usually exogenous to the organisms they regulate, and they can be designed by researchers to target a specific gene for downregulation [31]. siRNAs are commonly used in RNA interference (RNAi). RNAi has long existed in nature, however, more recently researchers have begun to harness its power to regulate gene expression [32]. siRNA is used to target specific sequences of RNA or DNA. siRNA originates from double-stranded RNA (dsRNA), which is not common in cells and triggers a cascade to eliminate the dsRNA [31]. The dsRNA is cleaved into small pieces, which become the siRNAs, by a protein called Dicer [31]. siRNAs are loaded into the RNA-Induced Silencing Complex, or RISC [31]. Once in the RISC, the strands of the siRNA are separated and only one strand is retained [31]. When the siRNA-bound RISC recognizes a mRNA sequence complementary to the siRNA strand, the RISC activates a protein called slicer to cleave the mRNA and render it untranslatable [31]. Sometimes, there is not perfect sequence complementarity, but it is still similar enough that the siRNA strand will bind the partially complementary RNA and block it from efficient translation (Figure 1A) [32]. In a therapeutic setting, researchers may introduce synthetically produced siRNAs for targeted knockdown of a gene of interest. In this case, the siRNA is not cleaved from a larger dsRNA inside the cell-it is already the appropriate size for association with the RISC [31].

NANPs are particularly useful in the delivery of siRNA due to their stability [33]. Typically, "naked" RNAs are quickly recognized as foreign and degraded by endogenous nucleases inside the cell or organism to which the RNA is delivered [33]. The stability provided to the siRNA payload delivered by the NANP increases the ability of the siRNA to reach its intended target without being destroyed by the recipient. Additionally, NANPs can be conjugated to aptamers which bind with high specificity to a desired target. The specificity conferred to the siRNA delivery by nanoparticles reduces the possibility of offtarget binding/delivery and increases the concentration of siRNA at the desired target [19].

**Figure 1.** (**A**) Diagram detailing the steps involved in RNA interference (RNAi) through the production of small interfering RNAs (siRNAs). (**B**) A representative diagram of how siRNAs may be delivered to a cell via RNA-functionalized nanoparticles, and how the siRNAs can be used for RNAi upon delivery. Image created with BioRender.com.

#### *1.4. NANP Applications in Medicine*

In addition to the delivery of therapeutics to treat disease, NANPs can be used in disease prevention. RNA or DNA NANPs can be designed to deliver genetic information from pathogens in vaccines to prime the immune system for a natural exposure [34]. This is a new and exciting application of nanoparticles that we will undoubtedly see become more common as nucleic acid vaccines are further developed. NANPs can also be used as adjuvants that allow vaccine materials to be incorporated into the target tissue to stimulate an appropriate immune response [35]. On the flipside, nanoparticles can be altered to be immunologically inert and lack inflammatory activation, which can result in adverse effects to the patient. The future of NANP medicine will rely on careful optimization of these drugs and mitigation of their side effects and off-target effects.

#### **2. Functionalized RNA Nanoparticles**

#### *2.1. RNAi*

RNAi is the phenomenon of dsRNAs knocking down or silencing the expression of endogenous mRNAs initially described by Andrew Fire and Craig Mello nearly 25 years ago [36,37]. RNAi relies on small interfering RNAs (siRNAs) which silence encoded genes (Figure 1A). This occurs through a short double-stranded RNA binding with a target and preventing translation into functional protein [38]. While Fire and Mello characterized RNAi with an exogenous RNAi with dsRNA delivered from outside in C. elegans, the phenomenon is an evolutionarily conserved method of post-transcriptional gene silencing that has served as a means for organisms to protect themselves from exogenous threats as well as regulate gene expression [39]. RNAi has been characterized in plants as a sort of immune protection from plant viruses and insects [39]. A few years ago, Chejanovsky and colleagues used deep sequencing to detect the presence of perfect siRNA matches for three viruses that strongly contribute to colony collapse disorder in the genomes of honey bee colonies that had succumbed to colony collapse disorder [40]. In an experimental setting, RNAi has been used to tackle a wide range of biological problems. RNAi-mediated biological pest control has been used experimentally to protect valuable crops which will provide nutrients to countless people and livestock [41]. Building off of Chejanovsky's work in honey bees, RNAi has been experimentally employed to combat viruses that contribute to colony collapse disorder and threaten global food supplies [42]. NANPs can be used to mediate the delivery of exogenous RNAs for RNAi within cells. NANPs functionalized with the exogenous RNA deliver the RNA inside a cell to silence a target gene and prevent the formation of that gene's product [43]. Once the NANP has been internalized within the cell, the enzyme Dicer acts upon the attached RNAs, allowing them to participate in RNAi-mediated gene silencing (Figure 1B) [43].

Currently, RNAi is a subject of much research and development for clinical applications [44]. Despite a rocky start in the 2000s in which several RNAi therapeutic candidates were pulled from clinical trials due to unintended effects, Patisiran became the first RNAi drug to receive FDA approval in 2018 [37,45]. Patisiran treats hereditary transthyretin amyloidosis-a deadly genetic disorder which is characterized by the deposition of amyloid plaques of the protein transthyretin in key organs such as the heart and kidneys, leading to deterioration of quality of life for patients and eventual death [45–48]. Today, there are several RNAi drugs in clinical trials for treating cancers, inherited genetic disorders such as Sickle Cell Disease and familial hypercholesterolemia, and viruses such as HIV and hepatitis B [37].

Like with all new technologies, RNAi does not come without drawbacks. Naturally, we must consider immunogenicity. RNAi therapeutics work at their best when they are delivered to the appropriate tissue without degradation. When the RNAi drug triggers an immune response, the drug may never reach its intended target, or if it does, it could be in a less effective state [37]. In addition to unintended immunogenic responses, accumulation of RNAi therapeutics in unintended locations and subsequent toxicity are of major concern to RNAi drug developers [37]. While accumulation of RNAi therapeutics in off-target tissue is

a very legitimate problem in and of itself, an additional unintended consequence of RNAi drug treatment is the drug acting as it is supposed to outside of its target tissue and causing interruption to normal tissue function [37].

To continue to develop the applications of RNAi therapeutics, researchers have gotten creative with how they have approached drawbacks. Researchers have incorporated base modifications into the RNA delivery system to get around the problem of immunogenic activation caused by RNAi therapeutics [5,37,49]. In 2007, Robbins and colleagues demonstrated that the addition of a 2 -O-Me modification to siRNA reduced the siRNA's immunogenic potential without diminishing its ability to decrease expression of a target gene [49]. The added benefit of the base modifications is increased RNA stability, which generally increases the amount of time it is able to remain in the body without becoming degraded or neutralized by the immune system [5,37]. When the circulation time is increased, the RNAi therapeutic has an increased probability of arriving at its target and carrying out its intended function.

To mitigate off-target RNAi activity, researchers propose some common-sense measures to enhance safety for RNAi therapeutic recipients. The first proposed measure occurs long before the RNAi therapeutic makes it to the patient-it is extensive quality control to ensure that the drug has as few targets in the human genome as possible [37,50]. In their 2019 review, Setten and colleagues uphold that some off-target binding is inevitable and patients receiving RNAi therapy should be administered the smallest dose capable of achieving the desired effect, while being closely monitored for signs of off-target activity that is threatening to the patient's wellbeing [37].

It becomes more challenging when on-target effects in off-target tissues are considered. The drug is doing what it is supposed to do, just in a suboptimal location, which can imperil the patient's health. Much like mitigating off-target effects in off-target tissue, researchers must design intentional, highly specific therapeutics to ensure that there are as few possible routes for the therapeutic to build up in a non-target location [37]. In addition to designing therapeutics with the awareness of these effects, drug developers have "reverse engineered" compounds to reverse the impacts of the siRNA therapeutics in the event that they excel at performing their intended functions outside of the optimal location [37,51].

#### *2.2. NANP-Induced Immunogenicity*

The use of NANPs in humans as well as other animals carries the potential for immune activation. The immunogenic response must be thoroughly evaluated by the investigators to ensure that it is not activated in an unintended manner (i.e., prior to the delivery of the NANP to the target tissue). Researchers do have some (though not complete) control in how immunostimulation proceeds through the design of their NANPs. A 2017 research article by Guo and colleagues examined the impact of NANP sequence as well as physical properties (size and shape) on their immunostimulatory effects [52]. In this study, the research team demonstrated that increases in the size of RNA squares resulted in enhancement of cytokine secretion, namely TNF-α and IL-6 [52]. Similarly, RNA squares with attached uniform RNA sequences showed the same effect-as the number of attached RNA sequences increased, so did the levels of cytokine secretion [52]. The same study indicated that three-dimensional RNA nanoparticles elicited higher levels of cytokine secretion than their planar counterparts [52]. Figure 2 provides a visual representation of the results that were described in Guo's 2017 article (Figure 2).

**Figure 2.** Structural parameters and other factors affecting immune stimulation by NANPs. Image created with BioRender.com.

#### *2.3. Immuno-Adjuvant*

The use of adjuvants offers a robust method where immunostimulatory compounds are employed to potentiate and modulate the immune response, when used together with vaccines. Among the recent advancements of the immunomodulatory NANPs, is their ability to regulate and modulate the immune responses when encountered. The rapid clinical development of this technology is impeded by several hurdles among which is the unknown immune stimulation by NANPs. However, recent studies have reported that mammalian cells use the same patterns of recognition established to defend themselves against viral and bacterial nucleic acids to process NANPs [2]. Further investigations showed that various interactions of NANPs with different immune cells elicit different immune responses. The regulation of the immune response by NANPs can be controlled by varying their design, specifically by altering numerous architectural parameters including the NANPs' size, shape, functionalization and composition (Figure 2) [2,8,53,54]. The correlation between those variations in NANPs' structures and their effect on the immune response is now being considered carefully during the design process aiding to develop NANPs that would either modulate the immune activation or stay immunoquiescent [54]. Other aspects considered to successfully translate NANP technologies into clinical settings would include the choice of delivery carriers and administration routes for NANP formulations. If NANPs complexed with a carrier are delivered via intravenous administration, they may induce undesired inflammation due to cytokine induction and complement activation. However, if the same system is administered locally, it would serve perfectly as immune-adjuvant as it will induce the same cytokine and interferon response along with complement activation, which would potentiate the vaccine efficacy and enhance immunotherapy efficacy [53].

#### *2.4. NANPs with Regulated Immune Responses*

Human cells have receptors for the recognition of foreign nucleic acids called pattern recognition receptors (PRRs), which can also distinguish and process NANPs. Therefore, understanding the underlying mechanisms of recognition and the NANPs structural parameters that affect their recognition process would allow the tunability of the immunostimulatory effects. This, in turn, will allow engineering the immunoquiescent NANPs intended for drug delivery, or NANPs with regulated immunological properties that could be used in immunotherapies [3]. In the case of immune-adjuvants, PRR agonists help

to stimulate the innate cytokine and interferon production, which endorses the cellular antiviral defenses [3].

Extensive studies have been carried out to assess the immunostimulation of the representative library of NANPs introduced to freshly collected human peripheral blood mononuclear cells (PBMCs) [3,7,8,55–61]. PBMCs were chosen as a highly reliable model for prediction of cytokine storm toxicity in humans [57]. One of the parameters that affects the immunorecognition of NANPs is their chemical composition (e.g., DNAs vs. RNAs vs. DNA/RNA hybrids) (Figure 2-Composition) [55,62]. A study reported that altering the composition of NANPs can modulate the mechanisms and degree of elicited immune responses [63]. It also highlighted that NANPs made of RNA normally demonstrate significantly higher immune activations in comparison to their DNA counterpart, since RNA NANPs can trigger both TLR7 and RIG-I mediated cytokine and interferon response (Figure 2-Composition) [63–65]. Another parameter is NANP size; increases in the size of NANPs may lead to elevated immunostimulation (Figure 2-Size). One more parameter that plays a major role in immunostimulation is the dimensionality of NANPs. Several studies have confirmed that for RNA NANPs, fibrous structures (1D) demonstrate reduced immunostimulation when compared to planar NANPs (2D) and that the globular RNA NANPs (3D) produce the highest levels of immune responses amongst all of them (Figure 2-Globularity) [2,59,60]. It was reported that 2D and 3D RNA NANPs induced interferon production upon activation of TLR7, while 1D NANPs did not [62]. In addition, it was shown that the immunostimulation of NANPs functionalized with therapeutic nucleic acids (TNAs) induce higher production of type I and II IFNs when compared to non-functionalized NANPs and that the extent of activation can be regulated by relative orientation of the TNAs (Figure 2-Functionalization) [8,60].

#### *2.5. The Role of Carriers on NANPs Immunorecognition*

Another hurdle precluding broader clinical application of NANPs is their intracellular delivery [1,66,67]. One of the most essential factors for NANPs delivery and their efficacy is the use of carriers and complexation agents. Extensive studies had previously reported that carrier-free NANPs do not elicit any immune response as they are invisible to the cells, and the use of delivery platforms further tailors the immunorecognition of NANPs [2,53,68]. Hence, for efficient intracellular delivery of NANPs, various delivery agents such as lipidbased carriers [55], exosomes [69], polymeric agents [70], and inorganic materials [71,72] have been investigated. One study employed PBMCs to investigate the use of amineterminated PAMAM dendrimers to deliver NANPs (e.g., RNA and DNA cubes) and compared to commercially available Lipofectamine 2000 (L2K), a well-established lipidbased delivery platform [55]. The results highlighted that the uptake of the NANPs by different human immune blood cells, and their cytokine responses varied based on the delivery system used [55]. NANPs complexed with dendrimers did not induce type I and type III IFNs as opposed to NANPs complexed with L2K. In addition, NANPs complexed with L2K did not induce cytokine production (IL-1α, IL-1 β, IL-6, TNFα), while NANPs complexed with dendrimers induced the production of these stress associated cytokines. The 3D RNA NANPs (RNA cubes) delivered by dendrimers also elicited a more potent profile of cytokine production when compared to their DNA counterparts, which aligned with previous findings highlighting the effect of NANP composition on immunorecognition [53]. Additionally, as was expected, the carrier-free NANPs did not elicit any immune responses [55]. To overcome the barrier of safe and efficient delivery of NANPs while avoiding the carrier-associated toxicity, naturally occurring nanovesicles involved in cellular communication (e.g., exosomes) can be utilized. The exosomes provide a stealth-coating for loaded NANPs, which prevents nuclease degradation of NANPs as well as exposure to PRRs. For example, exosome-mediated delivery of RNA cubes which are known to have high immunostimulatory effects on cells, showed negligible immune activation [69], as compared to other carriers.

As each set of NANPs holds a unique physicochemical and architectural profile, this creates a burden to predict the type of immune response and its magnitude. To overcome this challenge, a computational predictive tool called "artificial immune cell", or AI-cell, was developed to guide the design of NANPs to fit the desired immunological profiles. This unprecedented computational approach is fed by physicochemical and immunological profiles for an array of various NANPs and uses innovative transformer architectures to predict the immunological activity of NANPs based on the entered oligo and their sequence compositions [7]. This freely available web-based implementation is expected to advance the understanding of properties that contribute to immunomodulatory activity of NANPs and draw guidelines for their design principles. The AI-cell shall further promote the therapeutic nucleic acid nanotechnology even further by addressing the public health challenges related to the toxicities of nucleic acid therapies [7].

#### **3. RNA Motifs and Domains, and Their Delivery via Nanoparticles**

#### *3.1. Motifs*

In 1999, P.B. Moore defined an RNA motif as a "discrete sequence or combination of base juxtapositions found in naturally occurring RNAs in unexpectedly high abundance" [73]. Moore's definition of an RNA motif leaves room for RNA sequence motifs as well as structural motifs [74]. RNA motifs occur naturally, can exhibit three-dimensional structure, and can interact with other motifs in RNAs and protein domains to contribute to their overall functionality [73]. Researchers and pharmaceutical developers have taken advantage of naturally occurring RNA motifs and incorporated them into their nanoparticle designs for enhanced stability and increased capacity for payload delivery and tracking [5].

#### *3.2. Domains*

In a 2002 review of protein domains, Ponting and Russell provided three different perspectives from which protein domains could be defined: biochemical, structural, and sequential [75]. Structurally, they defined domains as "spatially distinct units" [75]. In the biochemical context, they were less concerned with structure and specified a domain as a region with a clear-cut function [75]. From a sequential standpoint, Ponting and Russell claim that a domain is characterized by homology to other sequences which achieve similar functions in different environments [75]. While each definition of the domain holds some truth, all three should be considered together to get the full understanding of the domain.

Ponting and Russell did provide a more modern definition of a domain as a structure that could adopt the necessary structural conformation for carrying out its function [75]. While this definition is applied to protein domains, it could be applied in the context of RNA as well. After all, RNA can adopt specific conformations that facilitate biological functions. RNA motifs, described earlier, have the capacity to build upon each other and interact with other motifs in ways that perform biological functions. Those interactions between motifs are key to the establishment of RNA domains. A 2011 review by Reiter, Chan and Mondragón described domains as complex, functional, three-dimensional structures in the RNA that are comprised of (and stabilized by) interacting RNA motifs [76]. Functional RNA motifs are the building blocks of larger functional RNA domains that have unique three-dimensional structures [76].

#### *3.3. Motifs and Domains in Research*

Since the late 1980s, RNA motif research has been rapidly growing. In a 1998 review article, Conn and Draper claim that there are only a few functional RNA motifs, but when these motifs are placed together in combination, the functions that can be carried out by the structured RNA and the specificity with which these functions can be executed is enormous [77]. Nanoparticles, which were in their early days at the time of Conn and Draper's review, take full advantage of combinatorial effects of RNA motifs. RNA motifs themselves are quite frequently incorporated into the designs of nanoparticles for their functionality [5].

RNA tectonics (or tectoRNAs) utilize naturally occurring RNA motifs to form hierarchically folded modular functional RNAs which can be used to construct RNA nanostructures [78,79] and illuminate the functions of already-existing RNA structures [80]. We can think of modular tectoRNAs as jigsaw puzzle pieces that when put together, provide us with the greater unified function of the specific nanostructure much like the pieces of an actual jigsaw puzzle show us an entire picture [79]. However, unlike a jigsaw puzzle, tectoRNAs (or pieces) can be reused in different nanostructures (puzzles) to make and execute entirely new functions [80].

#### *3.4. RNA Functional Augmentation*

Over the last nearly three years, the public has become increasingly aware of the use of RNA in medicine through the SARS-CoV2 vaccine. Two of the three mainstream COVID-19 vaccines contain the mRNA transcript encoding the spike protein, which facilitates viral entry into the cell [81]. The mRNA to be delivered is encased in lipid nanoparticles to prevent rapid degradation of the encoded instructions [81]. Once delivered, the mRNA becomes translated into the viral spike protein and takes on its unique conformation to elicit an immune response against the spike protein [82]. The nanoparticle delivery facilitates non-immunogenic delivery of the mRNA cargo which is selected for its ability to trigger immune activation.

Outside of the COVID-19 context, RNA nanoparticle delivery could be used for functional complementation studies. Previously, conjugates of gold nanoparticles were used as carriers for functional RNA structures in cells and these in-cell structures efficiently contribute to gene expression regulation [83]. X-ray crystallographic determination has provided further evidence that these nanostructures can fold into stable RNA motifs, such as kissing loops and T-junctions, that resemble natural RNA motifs [84]. This opens up the possibility of using RNA structural motifs as nanostructures for genetic complementation studies to restore a normal phenotype to mutants with some defect (Figure 2-Functionalization). In this arena, our team is exploring how crucial discrete structural domains of RNA can be used as nanostructures to compensate functional deficiency in parasitic disease caused by the protozoan parasite *Trypanosoma brucei*, the causative agent of African sleeping sickness. Telomerase RNA is a long noncoding RNA that is an integral functional subunit of a large RNA-protein complex responsible for synthesizing the G-rich strand of telomeres, the physical ends of linear chromosomes. Telomerase is critical for telomere length maintenance, thus preventing chromosome instability in eukaryotic cells [85,86]. Our previous and ongoing works with *T. brucei* telomerase RNA structural domain deletion mutants have demonstrated that certain domains of the telomerase RNA are vital to cell proliferation [87]. Once the functions of the *T. brucei* structural domains have been established, we aim to deliver the missing telomerase RNA structural domains to the domain-deletion mutants in an effort to restore their functions.

#### *3.5. Domain Delivery and Associated Challenges*

Several different types of RNA can be delivered to cells for purposes such as posttranscriptional regulation, enhancement of catalytic activity and augmentation of gene expression [88]. NANPs functionalized with aptamers can deliver cargos (including RNA domains) to highly specific locations within cells (Figure 3) [88]. Additionally, CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) technologies are showing great promise in targeting highly specific sequences for editing (Figure 4) [88]. CRISPR genome editing is inspired by a bacterial immune system to protect bacteria against viral invaders [89]. The CRISPR-associated (Cas9) restriction enzyme is directed by a guide RNA (gRNA) which binds to a protospacer adjacent motif (PAM) on a segment of DNA containing the target for editing [89,90]. Cas9 moves from the PAM to determine if there is base complementarity between the gRNA and the DNA target sequence [89,90]. When complementarity between the gRNA and target DNA is located, the Cas9 endonuclease creates a double strand break (DSB) in the target DNA sequence [89,91].

**Figure 3.** Hypothetical experimental schematic for delivering RNA structural domains to RNA domain-depleted cells for complementation studies with nucleic acid nanoparticles (NANPs). Image created with BioRender.com.

Once the DSB has been made, the host cell's DNA damage repair machinery is activated to prevent cell death [89,92]. Repair processes such as non-homologous end-joining (NHEJ) and homology directed repair (HDR) commonly activate and they cause alterations in the target sequence [89]. A single DSB repaired through NHEJ often incorporates extra base pairs which interrupt the coding sequence and result in the absence of the gene product (Figure 4) [89]. If two different gRNAs are used to make cuts at different places in the target sequence, a segment of the target DNA can be eliminated [89]. If a DNA template is also present in the CRISPR-Cas9 reaction involving two gRNAs, HDR will incorporate the DNA template into the target sequence [89].

Although NANPs are exceptionally promising as tools for modulating gene regulation, they come with considerable delivery and stability issues. NANPs exhibit diminished stability in mammalian serum [63]. In addition to the low stability of NANPs inside of the body, RNAs are rapidly degraded inside of cells by ribonucleases. When an unstable NANP is coupled with RNA domains that could be destroyed upon cellular entry, the prospects of the RNA domains reaching their intended target decrease dramatically. Furthermore, RNA carries a net negative charge, which makes it unlikely to achieve internalization within cells without modifications or incorporation within a carrier that is more amenable to traversing the plasma membrane [2,63]. NANPs carrying exogenous RNA are also quite effective activators of immune responses which can stimulate inflammation harmful to the patient's wellbeing, and to the successful delivery of the NANP payload.

The challenges of NANP stability and rapid RNase-mediated degradation of RNAs delivered by NANPs are not absolute. In fact, over the last decade, significant strides have been made in mitigating NANP serum instability and exogenous RNA degradation. In a 2020 article by Johnson and colleagues, they effectively demonstrated that the composition of the NANP itself carried a significant impact on how stable the NANP was in serum [63]. Replacement of the 2 -OH group in RNA by a 2 -F increased stability for triangular NANPs that contained either an RNA or DNA center [63]. In addition to enhancing the stability of NANPs, the substitution of the 2 hydroxyl group for a fluorine assisted in mitigating RNase-mediated degradation of RNA NANPs [63,93].

**Figure 4.** Basic diagram of the CRISPR-Cas9 genome editing system. Image created with BioRender.com.

#### **4. Determination of RNA Structural Properties in Nanobiotechnology**

*4.1. Importance of RNA Structure Determination*

It is a strong theme throughout biology that structure is crucial in determining function. The structure of RNA is dynamic and typically reflects the RNA's specific function [94]. Whether it is the delivery of RNA nanoparticles in cells or RNAs delivered via gold or other nanostructures, determining thermodynamically stable structures, such as three-way junction (3WJ) motifs or structural RNA domains are important for further investigations. With the advent of high-throughput sequencing based RNA probing and cell-penetrating chemical probes, it is now possible to determine structures of RNAs in vivo. In our research team's work, we have demonstrated the dynamic nature of RNA through different stages in the life cycle of *Trypanosoma brucei* [87]. The needs of cells can change throughout their life cycles and having an understanding of how the changing structure of RNA contributes to meeting those needs is of the utmost importance [95].

#### *4.2. Methods for Structural Prediction*

In some of our recent work, we employed a selective 2 -hydroxyl acylation analyzed by primer extension mutational profiling (SHAPE-MaP) technique to model telomerase RNA secondary structure at different stages of the life cycle in *Trypanosoma brucei* (Figure 5) [87,96]. We chose this technique for its ability to visualize RNA conformations within living cells, as well as its adaptability for immunoprecipitated RNA [87]. The SHAPE protocol works

by applying a SHAPE chemical probe to a sample to facilitate the addition of bulky adducts to RNA bases that are not engaged in a binding arrangements with other bases (from the same RNA or a different RNA), protein, or DNA [96]. For the purpose of comparison, RNA extracts are also treated with a control that does not place bulky adducts on the unbound bases, which is usually the solvent for the SHAPE reagent [96]. Once the adducts are associated with the unbound bases, library preparation is started through the production of cDNA, which will induce mutations in the sequence at the location of SHAPE reagentinflicted adducts. A second strand of DNA is made to stabilize the DNA libraries and prepare them for high-throughput DNA sequencing. The sequences from the SHAPE reagent treated samples is compared with the sequence data from the control treated samples, then aligned to identify where the SHAPE reagent induced mutations are located (Figure 5) [96]. The sequence data is processed by the SHAPE-MaP software program to calculate the flexibility (reactivity) at each base position [96]. Both the sequence data and the flexibility data are used in a structure prediction program (we used RNAstructure) to create minimum free energy models of the RNA secondary structure [87,96,97].

**Figure 5.** The SHAPE-MaP pipeline for RNA structural analysis. Image created with BioRender.com.

More recently, a new application of selective 2 -hydroxyl acylation analyzed by primer extension emerged as juxtaposed merged pairs protocol (SHAPE-JuMP) [98]. SHAPE-JuMP serves the purpose of supporting higher-order RNA structural prediction through crosslinking nearby structures in the same RNA transcript [98]. The crosslinker is a SHAPE reagent called trans-bis-isatoic anhydride (TBIA) which has two functional sites that interact with the 2 hydroxyl groups of bases in RNA structural domains that are near to each

other [98]. Once the RNA structures are crosslinked, RT-C8, a reverse transcriptase capable of "jumping" over the TBIA crosslinker, reverse transcribes the RNA into DNA while leaving out the RNA between the crosslinked bases [98]. The skipped region of RNA appears as a deletion when the sequence is aligned to a reference sequence [98].

#### *4.3. Advantages and Limitations of Structural Modeling*

A crucial advantage of using SHAPE-MaP techniques to predict the secondary structure of RNA is the ability to use it inside living cells in addition to outside of the cells and on deproteinized "naked" RNA [96]. In their protocol paper, SHAPE-MaP developers Smola and Weeks propose using both in-cell and cell-free SHAPE-MaP procedures to identify the locations of likely RNA-protein interactions [96]. Since the SHAPE-MaP software calculates base flexibility (proclivity for interaction with other nucleic acids or proteins) at the individual base level, RNA SHAPE-treated inside of cells can be compared with RNA SHAPE-treated samples after extraction and deproteinization to determine where there are differences in flexibility. The locations that exhibit base flexibility clue investigators into areas that warrant further evaluation to determine (1) if there is in fact some kind of interaction occurring between the RNA of interest and some other molecule(s), and (2) what those other molecules are if there is an interaction occurring [96].

SHAPE-JuMP builds upon the concept of SHAP-MaP. Its creators credit it with being better adapted at handling long range, RNA tertiary interactions that can be more prone to errors in traditional SHAPE-MaP protocols [98]. An additional benefit of the SHAPE-JuMP procedure of RNA modeling is the structural support provided to interacting structures within the transcript [98]. Essentially, TBIA freezes the interacting RNA structures in place for the purpose of reverse transcription and sequencing [98]. The immobilization of the interacting structure makes SHAPE-JuMP a particularly effective method of predicting the structures of large RNAs [98].

While quite powerful, SHAPE-MaP is not perfect. SHAPE-MaP software is unable to differentiate between multiple different isoforms of the same transcript. Different isoforms of RNAs can be predicted by several different RNA processing events which are the results of complex RNA interactions. These interactions have impacts on the RNA's ability to perform its intended function, whether that be translation into a functional protein or a regulatory role. In addition to the inability to differentiate between structural isoforms of an RNA transcript, SHAPE-MaP requires chemical probes that can cross the plasma membrane of the cells being studied [96]. There are several SHAPE probes commercially available, but they do not all have the same inclination to penetrate cellular plasma membranes [96]. SHAPE-MaP analysis relies on effective DNA library preparations and in RNA transcripts that are highly repetitive or structurally inaccessible, these sequencing results and predicted secondary structures are not as reliable as regions that are non-repetitive or structurally accessible [96]. SHAPE-JuMP is quite novel and not all limitations have been fully characterized. The SHAPE-JuMP creators did cite a less than optimal ability to identify tertiary contacts between interacting structures as well as a reliance on amplicon sequencing as major limitations of the protocol [98].

The applications of RNA structural modeling using SHAPE techniques are not limited to RNA in living systems. NANPs themselves take on discrete structures that are key to the effective delivery of their cargo to the appropriate location. The RNA-SHAPE techniques to produce structural models of RNA can be applied after NANP production in a quality control step to ensure that the nanoparticles accurately formed the intended structure. Additionally, SHAPE structural modeling techniques can be employed to evaluate how the structures of NANPs change when the cargo is delivered. Furthermore, we can use RNA-SHAPE to determine if the delivery process makes any changes to the structure of the nucleic acid cargo that may impact its ability to perform its intended function(s). These techniques can assist investigators in evaluating the structural stability and integrity of their delivery systems as well as cargo loads to enhance the efficacy of delivery and incorporation of the functional cargo.

#### **5. Discussion**

Nanotechnology has been a crucial area of research over the last decade. Nanoparticles are a growing area of academic and commercial investment, and the exploration of nucleic acid nanoparticles has opened up a wide array of research and therapeutic avenues. In the clinic, NANPs offer medical providers with potential preventative, diagnostic and therapeutic applications for patient care. Of course, as is the case with all novel technologies, there is room for nucleic acid nanoparticles to improve. Problems surrounding immunogenicity, off target activity, and unintended buildup all persist, and careful experimentation will lead researchers to optimizing specific NANPs for their unique purposes.

Outside of the clinic in a basic research capacity, NANPs are of great academic value. Investigators can use NANPs to deliver nucleic acids to knock down expression of genes to clarify the purpose of the gene. Additionally, in the arena of RNA domains, we propose the use of NANPs to deliver RNA structural domains to cells that have been depleted of the same structural domains. In this planned research, one could examine the impact of RNA structural domain loss followed by complementation of the cells with the missing domain to clarify the purpose of discrete RNA structural domains. Once optimized, this system will allow us to gain a better understanding of how RNA structure influences its function.

As research with nucleic acid nanoparticles progresses, investigators will need to address the challenges and disadvantages that come with working with them. Challenges such as unintended immune neutralization, NANP accumulation, and off-target effects are most common. The most powerful tool researchers have in mitigating these challenges is careful design of the NANPs so that they are non-immunogenic prior to target site delivery, do not build up in inappropriate locations, and do not act on inappropriate tissue. This can be done through robust screening of candidate NANPs to ensure that their specificity is as narrow as possible (i.e., it has only one complementary sequence). When designing NANPs for use in the clinic, specificity is absolutely crucial for proper delivery and appropriate immune activation. Additionally, investigators must consider the stability of the NANP they wish to design. While the NANP should be stable enough to travel to the intended target tissue, once there, it must be able to deliver its cargo and then be broken down and cleared to prevent accumulation of "spent" NANPs. The structure of the NANP is a key element in its activity and special attention must be paid to how the nanostructures will react upon arrival at the target site. Advances in NANP design technology (i.e., software like NanoTiler and SELEX) will undoubtedly assist researchers in creating NANPs that have highly specific sequences to limit target possibilities and the proper three-dimensional structures to act on a specific target, then undergo degradation and clearance [19,20,99].

**Author Contributions:** Conceptualization, K.K., Y.R. and K.C.; visualization, K.K., Y.R. and K.C.; writing—original draft preparation, K.K. and Y.R.; writing—review and editing, K.K., Y.R. and K.C.; supervision, K.C.; all authors critically revised, discussed, and edited the article until it reached its current form. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work received no external funding.

**Institutional Review Board Statement:** Not Applicable.

**Informed Consent Statement:** Not Applicable.

**Data Availability Statement:** Not Applicable.

**Acknowledgments:** This review paper was written as a collaborative effort between the Chakrabarti Lab in the Department of Biological Sciences and the Afonin Lab in the Nanoscale Science Program within the Department of Chemistry at the University of North Carolina at Charlotte.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

### *Review* **The Recognition of and Reactions to Nucleic Acid Nanoparticles by Human Immune Cells**

**Dominika Bila 1, Yasmine Radwan 2, Marina A. Dobrovolskaia 3, Martin Panigaj 1,\* and Kirill A. Afonin 2,\***


**Abstract:** The relatively straightforward methods of designing and assembling various functional nucleic acids into nanoparticles offer advantages for applications in diverse diagnostic and therapeutic approaches. However, due to the novelty of this approach, nucleic acid nanoparticles (NANPs) are not yet used in the clinic. The immune recognition of NANPs is among the areas of preclinical investigation aimed at enabling the translation of these novel materials into clinical settings. NANPs' interactions with the complement system, coagulation systems, and immune cells are essential components of their preclinical safety portfolio. It has been established that NANPs' physicochemical properties—composition, shape, and size—determine their interactions with immune cells (primarily blood plasmacytoid dendritic cells and monocytes), enable recognition by pattern recognition receptors (PRRs) such as Toll-like receptors (TLRs) and RIG-I-like receptors (RLRs), and mediate the subsequent cytokine response. However, unlike traditional therapeutic nucleic acids (e.g., CpG oligonucleotides), NANPs do not trigger a cytokine response unless they are delivered into the cells using a carrier. Recently, it was discovered that the type of carrier provides an additional tool for regulating both the spectrum and the magnitude of the cytokine response to NANPs. Herein, we review the current knowledge of NANPs' interactions with various components of the immune system to emphasize the unique properties of these nanomaterials and highlight opportunities for their use in vaccines and immunotherapy.

**Keywords:** nucleic acid nanoparticles (NANPs); immunorecognition; immunoreaction; Toll-like receptors; cytokine storm syndrome; complement activation-related pseudoallergy

#### **1. Introduction**

#### *Nucleic Acid Nanoparticles*

Nanomedicine is an application of nanotechnology in medical settings for diagnosis, treatment, and prevention. It exploits unique chemical, physical, and biological properties of materials at the nanoscale. One of the perspective branches of nanomedicine is nucleic acid nanotechnology, which uses nucleic acids—DNA, RNA, and their various modifications—to design and formulate nanostructures for therapeutic applications [1].

Due to the programmability and the intrinsic functions of nucleic acids, singlestranded DNA or RNA molecules are rationally designed into modular nucleic acid nanoparticles (NANPs) that are easily customized into supramolecular three-dimensional structures exclusively made of nucleic acids. RNA and DNA form canonical and noncanonical base pairings to assemble into various higher-order structures that serve as a basis for the assembly of different nanostructures including rings, fibers, and polygons [1–6]. Advantageously, the choice of nucleic acid components provides tunability

**Citation:** Bila, D.; Radwan, Y.; Dobrovolskaia, M.A.; Panigaj, M.; Afonin, K.A. The Recognition of and Reactions to Nucleic Acid Nanoparticles by Human Immune Cells. *Molecules* **2021**, *26*, 4231. https://doi.org/10.3390/ molecules26144231

Academic Editor: Ali Nazemi

Received: 1 June 2021 Accepted: 8 July 2021 Published: 12 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

for the physicochemical properties, biological activities, and multifunctionality of NANPs. Many experiments in biotechnology and biomedicine propose applications of NANPs as carriers of bioactive compounds, molecular tools for imaging and biosensing, scaffolds for biochemical reactions, or multifunctional nanoparticles combining the previous functions into one complex [1,7–9]. The rapidly evolving field of nucleic acid nanotechnology had provided multiple synthesis methods for NANPs, established their characterization techniques in vitro and in vivo, and produced proof-of-concept data for using NANPs in various therapeutic applications [10–13].

NANPs can achieve biomedical functions by delivering therapeutic nucleic acids (TNAs) that are designed to perform key functions in gene regulation and expression and protein synthesis to serve in therapeutic applications. The modular functionalization of NANPs with aptamers, antibodies, or small molecules for their targeted delivery allows NANPs to integrate and deliver various TNAs into cells for synergistic therapeutic effects. However, despite these developments, NANPs have yet to advance to clinical translation due to concerns that need to be investigated and resolved including their specific delivery to target cells, their enzymatic degradation, and their ability to induce an immune response upon cellular uptake [4,12,14,15]. While targeting and stability are not immediate lifethreatening issues, the excessive immune recognition of NANPs and overreaction by immune cells can have potentially deleterious effects. Thereby, the immunostimulatory properties of NANPs are being extensively investigated [14,16,17].

Several physicochemical properties of NANPs determine their recognition by the immune cells; the most notable properties are 3D structure, composition (RNA to DNA ratio), molecular size, and the NANP's sequence. In addition, the immune response could be modulated by the type of delivery carriers used [6,14,18–22]. The proper design of NANPs with respect to immunostimulatory properties has the potential to activate innate and adaptive immune responses by activating nucleic acid immune sensors, thus having high potential as vaccine adjuvants and pan-antivirals [2,14,20,21,23–25]. Our emerging knowledge about the individual immunostimulatory abilities of nucleic acids aids in the design of safe NANPs, but it must be stressed that because of the effects of structure, the immunological characteristics of NANPs are not the sum of their individual components. Therefore, each NANP assembly must be experimentally tested and safety validated.

#### **2. Recognition and Reaction of Immune Cells to Nucleic Acids**

Immune cells are equipped with an extensive portfolio of so-called pattern recognition receptors (PRRs) that detect pathogen-associated molecular patterns (PAMPs) and damageassociated molecular patterns (DAMPS). The first line of PRRs include Toll-like receptors (TLRs) located on cell membranes (TLR1, TLR2, TLR4, TLR5, TLR6, TLR10) and in the endosomal compartment (TLR3, TLR7, TLR8, and TLR9) followed by RIG-I-like receptors (RLRs) or DNA sensor cyclic GMP–AMP synthase (cGAS) situated in the cytosol [20,26].

TLR sensing of nucleic acids is specific for RNA or DNA recognition and resides in the endosomal compartment, where TLR3 is specific for double-stranded RNA, including small interfering RNA (siRNA), TLR7 functions as a single-stranded RNA receptor, TLR8 is specific for bacterial and viral RNA immune recognition, and finally, TLR9 responds to bacterial and viral DNA (Figure 1) [20,26]. Recognition of nucleic acids from noncellular origins activates a complex network of signaling cascades that usually culminates in the expression of interferons (IFNs), including other cytokines and various chemokines. The general goal of the response is to alarm adjacent cells and recruit cells of adaptive immunity. The recognition of nucleic acids by TLRs causes signal transduction through Toll/interleukin-1 receptor (TIR)-containing signaling adaptors, TRIF, or MyD88 [27,28]. The downstream acceptor of these signals is NF-κB which, upon activation, translocates into the nucleus and induces the expression of pro-inflammatory genes [26,29]. NF-κB is functioning in both innate and adaptive immune cells. In addition to the mediation of macrophage inflammatory responses, NF-κB promotes the activation and differentiation of T cells and the maturation and differentiation of B cells [30,31]. Finally, the expression of

IFNs modulates further immune defense via paracrine and autocrine signaling through the transcription of IFN-stimulated genes (ISGs). The main effector functions of ISGs are to target pathways and functions required during the pathogens' life cycle as well as to enhance innate immune signaling. In addition, ISGs encode proapoptotic proteins that lead cells to apoptosis under specific conditions [32,33].

**Figure 1.** Toll-like receptors. Cell membrane-bound TLRs include TLR1, TLR2, TLR4, TLR5, and TLR6, while endosomal TLRs include TLR3, TLR7, TLR8, and TLR9. TLR3 recognizes double-stranded RNA (dsRNA). TLR8 recognizes bacterial and viral single-stranded RNA (ssRNA). TLR7 recognizes single-stranded RNA (ssRNA), as well as ring and cube RNA. TLR9 recognizes bacterial and viral double-stranded DNA (dsDNA), along with cube RNA. It is important to note that RNA cube triggers the activation of TLR9 and TLR7 only after its delivery inside the cell using a carrier such as L2K.

> Intracellular surveillance of RNA is carried out by RLRs, mainly the retinoic acidinducible gene-I protein (RIG-I) and melanoma differentiation-associated protein 5 (MDA5) located in the cytosol, although the presence of RIG-I has also been observed in the nucleus. RIG-I and MDA5 are activated by binding short double-stranded RNA (dsRNA) with a 5 -triphosphate and 5 -diphosphate or long dsRNA structures, respectively. Furthermore, for the most efficient activation of RIG-I, the blunt end is required as well as a short double-stranded sequence. Activated RIG-I interacts with the mitochondrial antiviral signaling protein (MAVS) residing on the mitochondrial membrane or peroxisomes. Finally, kinase complexes activated by MAVS induce transcription through IRF3, IRF7, and NFκB. The main cytoplasmic sensors of dsDNA are cyclic GMP-AMP synthase (cGAS) and IFNγ-inducible protein 16 (IFI16), which is also located in the nucleus, where it probably detects naked viral DNA. After binding dsDNA, cGAS synthesizes the second messenger 2 3 -cyclic-GMP-AMP (cGAMP) that subsequently mobilizes the stimulator of IFN genes (STING) on the endoplasmic reticulum that again induces the transcription of antiviral genes through IRF3 and NF-κB [34–36].

#### **3. Recognition and Reaction of Immune Cells to NANPs**

NANPs demonstrate different interactions with various types of immune cells, that, unlike traditional nucleic acid therapeutics, are also determined by the type of carrier or complexation agent used for NANPs' intracellular delivery. Without such agents, plain NANPs are invisible to the immune cells and do not trigger cellular immunological responses. For example, flow cytometric analysis of freshly collected human peripheral blood mononuclear cells (PBMCs) treated with a carefully chosen panel of NANPs with various compositions (RNA, DNA) and connectivity (globular, planar, and fibrous) revealed that after complexation with Lipofectamine 2000 (L2K), most NANPs are associated with the monocyte fraction and less with lymphocytes. Subsequent confocal microscopy showed that in monocytes, L2K-complexed NANPs were located inside the cells. Using a dye labeling the endolysosomal compartment and an inhibitor of endosomal uptake, it was observed that unlike lymphocytes, monocytes transport L2K-complexed NANPs into their interiors via endosomes. Overall, phagocytosis and endosomal acidification are key processes for L2K-complexed NANPs' uptake by monocytes. A further functional study indicated that scavenger receptors (SRs) are the most probable receptors involved with binding and internalization of L2K-complexed NANPs. In addition, the inhibition of SRs also prevented the expression of IFN-α in response to L2K-complexed NANPs [21]. Scavenger receptors are a heterogenous group of cell surface receptors that recognize a broad range of ligands; therefore, we currently do not know the mechanism of how SRs recognize NANPs [37]. Without L2K, NANPs did not show any signs of internalization by immune cells present in PBMCs and did not trigger the activation of PRRs or interferon responses.

Plasmacytoid dendritic cells (pDCs) play a key role in linking the innate immune and adaptive response, and although they constitute less than 1% of the monocyte fraction, pDCs, in comparison with isolated monocytes and myeloid DCs, respond to L2Kcomplexed NANPs with the strongest expression of type I and III IFNs. While in all fractions, RNA cubes appear as the strongest inducer of IFN response, pDCs activated IFNs regardless of the composition (DNA vs. RNA) or 3D structure. The depletion of pDCs from PBMCs leads to a dramatic reduction of IFN production, which means that pDCs are the primary source of immune reaction to NANPs. Interestingly, the distinct expression profile of IFN-α, IFN-β, IFN-ω, and IFN-λ between whole PBMCs and isolated pDCs implies that most likely, there is cellular crosstalk among PBMC subpopulations, which determines the overall response to NANPs [21].

The next important question is, which PRRs are responsible for the recognition and triggering of signaling cascades? The application of a pan oligonucleotide inhibitor of endosomal TLR signaling completely prevented the induction of IFN response upon treatment of PBMCs with any L2K-complexed NANPs used in the study. Similar results were observed in purified pDCs. The model HEK293 cell lines overexpressing either TLR3, TLR7, TLR8, or TLR9 were used to rule out which TLR type recognizes respective NANPs. In this model, the globular NANPs (RNA cubes) were sensed by TLR7, and RNA fibers were sensed by the rest of the examined TLRs [21].

In a follow-up study, we downregulated TLR7 and TLR9 expression in PBMCs by a mix of siRNAs. TLR7 and TLR9 were chosen as TLRs expressed in pDCs that are the primary IFN producers in the PBMC pool. However, the interpretation of observed data is complicated by different levels of downregulation of TLRs among the cells isolated from different healthy donors. Even the extent of silencing between TLR7 and TLR9 in one donor varied. The possible explanation may lay in the inter-individual sequence heterogeneity or regulation of TLRs' expression. The significant reduction in IFN response for the L2Kcomplexed RNA cubes was observed in two out of three donors with silenced TLR7, while no decrease in IFN production was detected upon treatment with L2K-complexed RNA fibers or DNA cubes. The downregulation of TLR9 prevented IFN response only in culture from one donor treated with RNA cubes and from another donor treated with RNA rings [20]. Taken together, TLR7 is responsible for RNA rings' and cubes' immune recognition but not DNA cubes nor RNA fibers (Figure 1).

#### **4. What Makes NANPs Immunostimulatory?**

The recognition of NANPs by the cell defense system depends on several physicochemical characteristics, including composition, 3D structure, sequence, shape, size, and connectivity. One of the first observations that the composition of NANPs (number of RNA vs. DNA strands that enter the composition of a particular NANP) affects their immune recognition came from the earlier study of functionally interdependent shape-switching nanoparticles where we noted that all examined NANPs triggered an IFN-α response, but NANPs assembled from six RNA strands were the most immunostimulatory [38]. A similar trend was observed in a study implementing a new RNA tetra-U helix linking motif in triangles with different DNA vs. RNA composition. In a model of human microglia-like cells, the transfection of RNA triangles induced the highest level of IFN-β production, followed by hybrid DNA/RNA triangles. No expression of IFN-β was stimulated by DNA triangles [19].

Several structure–activity relationship models that link the physicochemical properties of NANPs to their immunostimulation have emerged from a larger analysis of 25 different NANPs [21]. First, globular RNA cubes proved to be the most immunostimulatory NANPs. In comparison to DNA cubes that have almost identical shape and size, RNA cubes induced not only IFN-α and IFN-ω as DNA NANPs did, but also IFN-β and type III IFNs (IFN-λ). In addition, RNA cubes were more immunostimulatory than any other RNA-based NANPs (planar rings or fibers), and planar DNA or RNA structures were more immunostimulatory than chemically corresponding fibrous nanoobjects (Figure 2). In all these examples, NANPs were delivered to the cells using L2K.

**Figure 2.** Influence of physicochemical properties on immune stimulation. The main characteristics of NANPs that affect their immunostimulation are connectivity (how individual NANP strands are assembled), composition (number of RNA strands vs. DNA), and dimensionality (3D shape).

The chemical complexity or diversity of assembled NANPs can be increased by the incorporation of modified bases in individual strands. Especially for RNA bases, the diverse modifications play significant roles in RNA stability and affect the immunostimulatory potential [39]. Various experiments have described that the modification of RNA (herein siRNA) helps to circumvent TLR signaling and renders modified RNA immunoquiescent [40]. Therefore, it is interesting that when used with a carrier (L2K or DOTAP), triangular NANPs that consisted of a DNA strand in their center and 2 fluoropyrimidinemodified RNA strands on their sides induced IFN-β and IL-6 production, unlike all DNA NANPs and NANPs composed of a DNA center and unmodified RNA sides. The results suggest that the presence of 2 fluoro-modification significantly enhances the immunoreactivity of DNA-containing NANPs. The NANPs with RNA in the center and 2 fluoropyrimidine-modified RNA sides stimulated IFN-β and IL-6 production similarly to all RNA NANPs and NANPs composed of an RNA center and DNA sides. This indicates that 2 fluoropyrmidine modification does not affect the immune mediator response. The fully 2 fluoropyrimide-modified RNA triangles stimulated significant IFN-β and IL-6 production similarly to NANPs with either an RNA center and 2 fluoropyrimidine-modified RNA sides or NANPs consisting of a DNA center and 2 fluoropyrimidine-modified RNA sides [13]. Surprisingly, the incorporation of 2 fluoro-modifications into RNA NANPs abrogated the activation of TLR7 in the HEK293 reporter cell line but failed to avoid RIG-I dependent immune responses [14].

The ability to design complementary NANPs (also called anti-NANPs) that are assembled from the reverse complementary strands of evaluated NANPs allows for examining the effects of the sequence of NANPs on the ability to activate an IFN response. NANPs and anti-NANPs had completely different sequences but nearly identical 3D shapes. The RNA rings and DNA cubes were able to stimulate similar levels of IFN to their anti-NANPs analogs and anti-RNA cubes maintained the high response, which indicates that the NANPs' sequences are less important for immunostimulation than their 3D shape and composition (RNA vs. DNA). Except for the RNA rings and RNA fibers that are assembled from pre-formed monomers, all other studied NANPs (cubes, polygons, tetrahedrons, and DNA fibers) create intermolecular bonds (Figure 2). Indeed, free-unpaired nucleotides (ssUs) have enhancing effects on immunogenicity, but it appears only for globular NANPs such as RNA cubes. Interestingly, PBMCs from donors that demonstrated a higher IFN response to a TLR agonist (ODN 2216) reacted stronger to RNA cubes with nine ssUs in their corners than to cubes with a lower number of ssUS (three and six). On the other side, blood cells with lower reactions to the administered TLR agonist induced a similar IFN expression irrespective of the numbers of ssUs.

The size of the nanoparticles is one the main characteristics with potential impact on interactions with cells. Similar to the case of the number of free nucleotides in RNA cubes, the difference was observed only in donor cells with high reactions to ODN 2216, where hexagons activated the stronger response than three-, four-, or five-sided RNA polygons. Adjusting the mass of smaller polygons to be equal to or larger than that of the larger polygons had no effect on IFN production. In cells with low activation by ODN 2216, there was no observed difference between individual NANPs. In the case of DNA polygons, no significant differences were detected between different sizes of NANPs [21].

#### **5. Delivery Method/Carrier: An Unexpected Immunomodulator**

The immunostimulatory potential of NANPs is significantly influenced by the employed delivery method. The NANPs without a delivery agent are not efficiently internalized and thus do not induce IFN production. Even if naked NANPs are delivered to cells via electroporation, no production of IFNs was detected in response to any of the tested NANPs (Figure 3). Moreover, electroporated cells lose the ability to respond to other known inducers of IFN response, such as TLR9 agonist ODN 2216, although the addition of ODN2216 to the non-electroporated cells resulted in high levels of type I and III IFNs. The

results suggest that electroporation negatively affects endosomal TLR signaling, thereby affecting the ability of cells to elicit an immune response [20].

The importance of complexing the NANPs with a carrier for immunorecognition was demonstrated in a study that tested the ability of RNA cubes to induce the type I IFN immune response. The NANPs added to the cell cultures without a delivery carrier were incapable of stimulating an IFN response, while the NANPs complexed with L2K showed the ability to induce the secretion of both type I and type III IFNs. On the other side, ODN 2216, which was used as a positive control, stimulated an IFN response regardless of its complexation with L2K. The application of carrier itself did not cause the induction of the IFN response [21]. L2K does not affect NANPs' structures. Not surprisingly, different carriers demonstrate distinct transfection efficiencies for the same NANP [14].

Although the delivery of NANPs remains a challenge, new carriers are constantly introduced and tested. For instance, the immunostimulatory ability of the lipid-based carrier versus a cationic amphiphilic copolymer was compared. The NANPs delivered via the lipid-based carrier stimulated the production of both IL-6 and IFN-β. In contrast, when the NANPs were delivered using an amphiphilic copolymer, no statistically significant presence of IL-6 or IFN-β was detected. The results suggest that the employment of a cationic amphiphilic copolymer as a delivery carrier can reduce the immunostimulation, therein decreasing off-target effects [41].

Another recent study compared a lipid-based carrier (L2K) and dendrimers (PAMAM) to determine whether the spectrum and the magnitude of the cytokine response to RNA and DNA cubes depend on the type of the utilized carrier. The results showed significant differences in the induction of type I and type III IFNs and pro-inflammatory cytokines between NANPs delivered utilizing a lipid-based carrier and those delivered via dendrimers. The NANPs complexed with L2K stimulated type I and type III IFNs, while the complexation

of NANPs with dendrimers did not induce an IFN response. A remarkable difference was observed for cytokines associated with stress and danger (TNFα, IL-1 β, IL-6). The NANPs delivered via L2K did not stimulate a danger response, whereas those complexed with dendrimer induced the production of the stress- and danger-associated pro-inflammatory cytokines. The examination of chemokines (IL-8, MIP-1α, MIP-1β, MCP-1, MCP-2, and RANTES) showed that dendrimers alone did not stimulate any of the chosen chemokines, while the L2K carrier alone induced the production of all examined chemokines but MCP-2. The induction of MCP-2 was detected only when NANPs were complexed with the lipidbased carrier but not for dendrimer-complexed NANPs. Intriguingly, the induction of IL-8, MIP-1α, MCP-1, and RANTES was comparable between NANPs complexed with the lipidbased carrier and complexed with dendrimers. These results support the hypothesis that the type of carrier used for NANPs' delivery significantly alters their ability to stimulate the immune response, both quantitatively and qualitatively [4].

#### **6. Complement Activation-Related Pseudoallergy (CARPA) and Cytokine Release Syndrome (CRS)**

The systemic administration of pharmacologic or biologic agents can cause a strong and serious response in immune cells. Infusion-related reactions (IRs), a form of anaphylaxis or other hypersensitivity reactions occurring within minutes to hours of infusion, are immune-mediated adverse effects that occur after the administration of various products, including low-molecular-weight drugs, antibodies, and recombinant proteins, therapeutic nucleic acids, and nanotechnology-formulated products. Frequently observed symptoms in patients with IRs comprise flushing or rash, chest and back pain, dyspnea, wheezing, chills, or fever. These manifestations can lead to serious and potentially fatal consequences. Therefore, accurate assessments and early intervention are crucial when these symptoms occur. When IRs are triggered by the complement system, anaphylactoid reactions or CARPA occur. CARPA has the same symptoms and timeline of development as immediate type hypersensitivity (ITH) reactions. However, in contrast to the ITH, which are mediated by the antigen-specific IgE, CARPA is triggered by the complement. Both CARPA and CRS, also known as cytokine storm, are common, and the best understood mechanisms of IRs are associated with nanotechnology-formulated products [42].

The fundamental processes of CARPA include complement system activation, stimulation of blood cells and secretory cells, and the response of effector cells to mediator presence. The complement is activated via an initial trigger. The initial trigger can be radiocontrast agents, therapeutic antibodies, micellar and liposomal formulations, or nanoparticles. After the activation of the complement, anaphylatoxins are released. The anaphylatoxins are primary mediators that bind to target secretory cells (macrophages, mast cells, basophils, other phagocytic cells, and leukocytes), resulting in a release of secondary mediators that include cytokines, proteases, histamine, tryptase, prostaglandins, platelet-activating factor, thromboxane A2, and leukotrienes. The indications of CARPA are like those that occur with common allergies, with some unique exceptions. The most frequent symptoms are asthma, chest pain, chills, confusion, coughing, dermatitis, diaphoresis, dyspnea, edema, erythema, fever, headache, hypertension, hypotension, hypoxemia, nausea, rash, and wheezing [43]. The significant distinguishing feature is that the reaction arises after the first exposure to the drug and then decreases upon repeated exposure. In the case of NANPs, the lipidbased carrier is the most common cause of complement activation, which can subsequently lead to CARPA [44]. The large size and positive or negative surface charge of liposomes were shown to promote complement activation, whereas liposomes of a smaller size and neutral charge had reduced ability for activation [45]. In addition, the susceptibility of liposomes for complement activation was demonstrated to depend on dose and, in the case of PEGylated liposomes, on the presence of anti-PEG antibodies.

The CRS is a systemic inflammatory response caused by the excessive and rapid release of various pro-inflammatory molecules, including but not limited to INF-γ, TNF-α, IL-1, and IL-6. Macrophages, neutrophils, NK cells, and T cells are most often implicated in the pathogenesis of cytokine storm. The activation of primary T cells or immune cells' lysis initiates the production of IFN-γ and TNF-α, which stimulate macrophages, dendritic cells, other immune cells, and endothelial cells to release more pro-inflammatory cytokines (Figure 4). The production of IL-6 is essential for cytokine storm because IL-6 activates T cells and other immune cells, thereby creating a positive feedback loop. The trigger activating CRS can be traditional therapeutic proteins and nucleic acids as well as small molecular drug allergens, whereas nanocarriers can amplify their toxicity. The analysis comparing the ability of adenoviral vectors and lipid-based carriers to induce cytokine production showed that lipid-based carriers exhibit higher immunostimulatory potential than viral vectors. The clinical translation of numerous nanoformulations designed for nucleic acid delivery was terminated in part due to the immune-mediated adverse effects [46].

**Figure 4.** Cytokine storm. Cytokine storm is the result of the rapid release of numerous pro-inflammatory cytokines, including INF-γ, INF-α, IL-1, and IL-6. T cells, macrophages, neutrophils, and NK cells are most often involved in the cytokine storm pathogenesis. The activation of primary T cells or immune cells' lysis stimulates the production of IFN-γ and TNF-α, which activate other immune cells and endothelial cells to release more pro-inflammatory cytokines. The excessive production of IL-6 constantly activates the JAK–STAT3, Akt–mTOR, and MAPK–ERK signaling pathways. Their prolonged activation stimulates immune cells to produce more cytokines, which causes hyperinflammation and multiple organ failure. JAK–STAT3, Janus kinase-signal transducer and activator of transcription 3; MAPK, mitogen-activated protein kinase; mTOR, mammalian target of rapamycin; NF-κB, nuclear factor κB.

#### **7. Conclusions**

It is evident that the programmability, biological compatibility, and modularity of nucleic acids assembled into multifunctional NANPs promotes this class of biologically active molecules into an innovative class of personalized therapeutics. To successfully translate these materials to the clinic, one has to recognize the importance of the indication, route of administration, and complexation of NANPs with delivery carriers. If delivered with a carrier via intravenous administration, the induction of cytokines and/or interferons by NANPs may lead to undesirable inflammation. Moreover, some carriers such as liposomes may also trigger CARPA upon systemic administration. However, the

same type of cytokine or interferon response and complement activation by the carrier upon local administration may contribute to vaccine efficacy and improve the efficacy of immunotherapy. Experimental data from our laboratory provide several ways for controlling NANPs' immunostimulatory properties. Among them are NANPs' physicochemical properties (e.g., size, shape, sequence, connectivity), complexation with a delivery agent (e.g., lipofectamine, dendrimers), and route of administration (e.g., i.c., vs. s.c. or i.d.). Since the relationship between NANPs' physicochemical/bioactive parameters and the immune system has just emerged, it is necessary to improve the current understanding of NANPs' immunostimulatory properties for their successful translation to the clinic. We believe that the recent onset of mRNA vaccines to fight the COVID-19 pandemic will boost the field of therapeutic nucleic acids, including NANPs.

**Funding:** Research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Numbers R01GM120487 and R35GM139587 (to K.A.A.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. M.P. involvement in this publication is the result of the project implementation: Open scientific community for modern interdisciplinary research in medicine (OPENMED), ITMS2014+: 313011V455 supported by the Operational Program Integrated Infrastructure, funded by the ERDF. The study was funded in part by federal funds from the National Cancer Institute, National Institutes of Health, under contract 75N91019D00024 (M.A.D.). The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does the mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. All figures were created with BioRender.com (21 June 2021).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Review* **Innate Immunity Modulating Impurities and the Immunotoxicity of Nanobiotechnology-Based Drug Products**

**Claire K. Holley and Marina A. Dobrovolskaia \***

Nanotechnology Characterization Lab, Cancer Research Technology Program, Frederick National Laboratory for Cancer Research Sponsored by the National Cancer Institute, Frederick, MD 21702, USA; claire.holley@nih.gov **\*** Correspondence: marina@mail.nih.gov

**Abstract:** Innate immunity can be triggered by the presence of microbial antigens and other contaminants inadvertently introduced during the manufacture and purification of bionanopharmaceutical products. Activation of these innate immune responses, including cytokine secretion, complement, and immune cell activation, can result in unexpected and undesirable host immune responses. These innate modulators can also potentially stimulate the activation of adaptive immune responses, including the formation of anti-drug antibodies which can impact drug effectiveness. To prevent induction of these adverse responses, it is important to detect and quantify levels of these innate immunity modulating impurities (IIMIs) that may be present in drug products. However, while it is universally agreed that removal of IIMIs from drug products is crucial for patient safety and to prevent long-term immunogenicity, there is no single assay capable of directly detecting all potential IIMIs or indirectly quantifying downstream biomarkers. Additionally, there is a lack of agreement as to which of the many analytical assays currently employed should be standardized for general IIMI screening. Herein, we review the available literature to highlight cellular and molecular mechanisms underlying IIMI-mediated inflammation and its relevance to the safety and efficacy of pharmaceutical products. We further discuss methodologies used for direct and indirect IIMI identification and quantification.

**Keywords:** immunity; bionanopharmaceuticals; impurities; immunotoxicity; immunogenicity; bioassays; nanomedicine

#### **1. Introduction**

The body's primary "innate" defense against foreign invaders is triggered by an immediate but relatively non-specific localized immune response including both cellular and biochemical components. The cells contain pathogen recognition receptors (PRRs) capable of tightly binding pathogen-associated molecular patterns (PAMPs) common to several classes of infectious agents [1]. PAMP binding by cognate PRRs triggers immune cell activation, chemokine/cytokine secretion, and biochemical mediators, including the complement system (both systemically produced by the liver and cellularly produced by the activated immune cells), ficolins, pentraxins, and the coagulation system. The coordinated function of these components leads to the hallmark signs of acute inflammation: redness due to increased blood flow and tissue permeability, swelling caused by increased leukocyte (neutrophil, basophil, monocyte) recruitment and subsequent fluid retention in affected tissues, heat (local), and fever (systemic) to decrease pathogen replication and activate production of complement proteins for pathogen opsonization, and pain from the previous effects which act as a warning to the host of tissue damage and infection [2,3]. Together, these processes work to destroy invaders as well as prevent and repair any further tissue damage.

Lastly, innate immune effectors promote the secondary "education" of the immune system against similar future attacks. For this, microbial antigens generated via pathogen phagocytosis are displayed on the surface of antigen-presenting cells (APCs), specifically

**Citation:** Holley, C.K.; Dobrovolskaia, M.A. Innate Immunity Modulating Impurities and the Immunotoxicity of Nanobiotechnology-Based Drug Products. *Molecules* **2021**, *26*, 7308. https://doi.org/10.3390/ molecules26237308

Academic Editor: Alejandro Baeza

Received: 2 November 2021 Accepted: 24 November 2021 Published: 1 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

macrophages and dendritic cells (DCs). Through co-stimulation by pro-inflammatory cytokines and APC-antigen presentation, T-cells differentiate into specialized subsets responsible for promoting enhanced B-cell activation (CD4+ helper T-cells), direct pathogen degradation (CD8<sup>+</sup> cytotoxic T-cells), and immune modulation (regulatory T-cells (Tregs)) [3,4]. Upon B-cell activation, gene rearrangement produces large quantities of highly variable and specific antibodies. While this "adaptive" immune response is slow compared to the innate immune response, these antibody-producing plasma cells are maintained long-term, the "memory" of which allows for more rapid recognition and a stronger, more specific immune response upon secondary antigen exposure [1].

Unlike the epigenetic recombination required by the adaptive immune response, trained immunity is a form of non-specific, T-cell independent innate immunity, which relies mainly upon macrophage activation and pro-inflammatory cytokine production for long-term functional reprogramming of innate immune cell responses. Therefore, secondary antigen exposure can lead to temporarily altered cellular responses, either enhanced or reduced, compared to the primary response [5]. Depending on the degree of "training," protection can be conferred against reinfection by a specific microorganism and some additional non-specific protection against other unrelated pathogens [5].

To prevent inadvertent activation of these immune responses, new pharmaceutical compounds must go through several phases of investigation and regulatory review, consisting of discovery/development, preclinical testing, clinical testing, and approval, before being introduced to the market. Drug discovery/development encompasses the isolation (or fabrication) and subsequent characterization of a new compound, whether a molecule, nucleic acid sequence, or peptide/protein, for therapeutic use. This new compound is then subjected to preclinical (laboratory) testing, during which chemical or genetic analysis, pharmacological tools, and animal models are used to determine the safety and effectiveness of this drug towards a specific disease/condition. Due to the need for new drug compounds, half of all drug-related research and development expenditures occur during this stage, even though only one out of every thousand compounds progress to the next stage [6,7]. After successful testing in animal models, a new drug candidate is then deemed ready for clinical testing in humans. The clinical trial phases determine (I) the drug's metabolic and pharmacological actions, side effects, and effective dosage in healthy patients; and then (II) the drug's effectiveness in "diseased" patients as an improvement upon available treatments, if any. Of the compounds entering clinical trials, approximately 90% fail to pass the clinical phase I/II safety and efficacy requirements [7]. Those few compounds that do advance to clinical trial phase III are tested on a larger cohort of diseased patients to find the best balance between drug safety and effectiveness (dosage regimen, duration, etc.). Finally, once a therapeutic candidate has successfully passed these experimental hurdles, it must undergo final approval by a regulatory health agency (e.g., Food and Drug Administration (FDA) in the US) before being registered and sold as an available treatment [6]. Overall, from start to finish, the process of bringing a drug from the bench to the patient's bedside can cost over USD 800 million and take 8–10 years of effort with no guarantee of final approval [6,8].

Due to the financial and societal costs of the extensive process required for drug development, testing, and approval, it is essential that any potential product "failure" not be the result of the inadvertent inclusion of innate immunity modulating impurities (IIMIs, a.k.a innate immune response modulating impurities, IIRMIs [9]), components of a biotherapeutic treatment other than the target product that can potentially trigger the development of an immune response in the recipient [9,10]. Herein, we review the available literature to highlight cellular and molecular mechanisms underlying IIMI-mediated inflammation and its relevance to the safety and efficacy of pharmaceutical products, and to discuss methodologies used for IIMI identification. Challenges with the detection and understanding of the immunotoxic effects of drug products arising from intrinsic immunological properties (e.g., immunosuppression, immunostimulation, immunomodulation, immunogenicity) of activating pharmaceutical ingredients (APIs) or intended formulation

components (e.g., carriers and excipients) are not covered in this review as they have been extensively discussed elsewhere [11–18].

#### **2. Innate Immunity Modulating Impurities**

IIMIs encompass everything from live microbial contamination and pathogen-derived antigens (proteins, sugars, nucleic acids) to compounds introduced during the nanobiotherapeutic manufacturing and purification processes (Figure 1) [19,20]. The first source of IIMIs is adventitiously introduced microbial contaminants including live bacteria, mycoplasma, fungi, viruses, or their by-products. While the most common source of these impurities is contaminated raw materials [10], other sources include non-sterile equipment, improper handling practices, or contaminated facilities, though these sources are less likely in a highly controlled facility that employs appropriate sterilization procedures [10]. The second source of IIMI contamination is from host-cell proteins (HCPs), proteins produced by modified host organisms that are unrelated to the intended recombinant product. The population of HCPs produced during biopharmaceutical manufacture depends on host cell type and strain, location of expressed product (cytoplasm, periplasm, external culture medium), physiochemical properties and modalities expressed by-product (charge, hydrophobicity, structure, post-translational modifications, etc.), and the techniques employed during recovery and purification [21]. Due to the limited subset of physicochemical properties optimized for purification, a sub-population of HCPs with similar attributes to the target product will normally co-purify regardless of the process employed [21]. In addition, the use of chemical additives needed to maintain these modified host cells (e.g., growth medium, transferrin, albumin, insulin), as well as chemical additives and selective pressure agents applied for increased product production and modification (e.g., methotrexate, antibiotics, guanidine HCl) can result in adverse patient reactions and can potentially lead to the formation of antibiotic resistant bacterial strains [10]. Lastly, even processes employed for product filtration and purification can inadvertently introduce impurities that leach into the final product. Common "leachates" include bacterial protein A which is normally used for isolating antibodies, as well as hydroxyapatite, tungsten and stainless-steel fragments, glass and cellulose fibers, surfactants, and silicones which can be introduced by filters or containers used during the manufacture and purification processes [10,22]. Ideally, detection of such "leachates" in a biopharmaceutic product will result in modification and/or augmentation of purification processes, such as the use of high-quality resins, to prevent introducing these impurities [10]. Overall, at each stage biopharmaceutical production, there is the potential to introduce IIMIs which may have little/no impact on the function of the resulting drug product but are potent immune activators that have the potential to trigger an undesirable host immune response [23].

When in the presence of these IIMIs (Figure 2), immune cells (e.g., DCs, macrophages, monocytes, neutrophils, and some epithelial cells) recognize these antigens via a variety of pattern recognition receptors (PRRs) containing leucine-rich repeats (LRR) [24,25], including toll-like receptors (TLRs), nod-like receptors (NLRs), retinoic acid-inducible gene-I (RIG)-I-like receptors (RLRs), and C-type lectin receptors (CLRs). Each of these receptor families binds highly conserved microbial structures containing pathogen-associated molecular patterns (PAMPs), or endogenous structures containing damage-associated molecular patterns (DAMPs) released via cell rupture which are important for augmenting the elimination of pathogens and pathogen-damaged cells [9,26,27].

**Figure 1.** The many levels of possible unintended contamination in drug products. While most often associated with microbial contamination, unintended impurities can actually be introduced into pharmaceutical products from multiple sources, including raw materials and specialized host-cell reagents, and at various stages of production, ranging from fabrication and payload encapsulation in nanocarriers to purification of the final formulation.

**Figure 2.** Impurities in drug products trigger innate cellular responses and produce biomarkers for bioassay detection and Quantification. Currently, only β-glucans and endotoxins can be detected and quantified directly using specialized assays. The remaining population of impurities must instead be detected and quantified indirectly using downstream biomarkers (e.g., proteins, peptides, and nucleic acids) and immune cell activation as hallmarks of contamination.

The most studied and diverse family of PRRs, TLRs are a family of highly varied signaling receptors, each of which binds to a different set of microbial structures to trigger intracellular signaling resulting in cytokine secretion and lymphocyte activation [24,26]. Membrane-tethered TLRs, which often require dimerization for appropriate antigen binding and subsequent intracellular signaling, bind to molecules found on bacterial surfaces, including triacyl lipopeptides/proteins, glycolipids, and peptidoglycans, all of which bind to either the TLR1/2 heterodimer or the Dectin1/TLR2 heterodimer; diacyl lipopeptides, lipoteichoic acid, or zymosan which bind to TLR2/6; lipopolysaccharides (LPS) or endotoxins, which bind to MD2, an extracellular adaptor protein for TLR4; and flagellin, which binds TLR5 [24,25,28]. Several DAMPs can also bind membrane-tethered TLRs, including but not limited to hyaluronic acid and other fatty acids, high-mobility group protein B1 (HMGB1), heat shock proteins, S100 proteins, fibrinogen, and tenascin-C which bind to TLR4 [29,30] and serum amyloid A protein, which binds the TLR2/6 heterodimer [27,31]. On the other hand, intracellular TLRs bind to microbial components released after pathogen endocytosis and phagocytosis, including viral double-stranded (ds) RNA containing poly(I:C) motifs which binds TLR3; unmethylated CpG-rich DNA which binds to TLR9; and Guanosine/Uridine-rich single-stranded (ss) RNA and anti-viral imidazoquinoline compounds that mainly bind to TLR8 but can also bind TLR7 [24,25,27,28] Many intracellular TLRs also recognize DAMPs. For example, TLR7 and TLR9 distinguish between snRNP immunocomplexes vs. immunocomplexes of self-DNA or histones respectively [27].

With the assistance of a variety of signaling adaptor proteins (TIRAP, TRAM) and TRIF/TRAF transcription factors [24,25], all antigen-bound TLRs, except TLR3, activate intracellular signaling through a myeloid differentiation primary response protein (MyD88) dependent NFκB pathway resulting in the secretion of pro-inflammatory cytokines, including type II interferons (IFNs) (e.g., IFNγ), interleukins (ILs) (e.g., IL-1β, IL-6, CXCL8/IL-8, IL-12, and IL-18) and tumor necrosis factor α (TNFα); priming of caspase-1; and the activation of local lymphocytes and vascular endothelium, eventually resulting in antibody production [24,26]. Meanwhile, MyD88-independent activation of IRF3/7 leads to the type I IFNs (IFNα) response critical for antiviral defense [25,30]. However, the continuous stimulation of these PRRs, especially the "bipolar" PRRs involved in DAMP recognition, can lead to inflammatory dysregulation leading to the development of autoimmune and chronic inflammatory diseases [27], as well as blunted responses, also known as tolerance [32]. As such, these pathways are tightly controlled, with some TLRs (TLR2 and TLR4) even having decoy receptors designed to dampen innate responses during severe infection by blocking the interactions between the bacterial ligands and the active TLRs [26].

TLR function also overlaps and integrates with other PRR signaling pathways, including NLRs, RLRs, and CLRs. NLRs, such as NOD1 and NOD2, act as intracellular bacterial sensors by recognizing peptidoglycans (e.g., mDAP and MDP respectively) resulting in inflammasome-mediated NFκB activation leading to the production of IL-1β [26]. The TLR and NLR pathways are clearly integrated for producing IL-1β, as effective NLR activation requires both PAMP activation of the inflammasome and TLR priming, to initiate an inflammatory response [26]. Other NLRs are responsible for triggering the activation and regulation of pro-inflammatory caspase-1 and caspase-5. RLRs, on the other hand, are intracellular viral sensors, binding specifically to dsRNA. Like NLRs, these receptors contain caspase-recruitment domains (CARD) responsible for recruiting adaptor proteins resulting in IRF3 and NFκB activation, leading to the production of type I IFNs (IFNα/β) and pro-inflammatory cytokines (e.g., TNFα, IL-1β, IL-6). Due to these similarities with viral-sensing TLRs (i.e., TLR3, 7, 8, and 9), it is likely that TLRs and RLRs also function together to provide ubiquitous anti-viral protection [26]. Lastly, CLRs are carbohydratebinding receptors located mainly on the surface of DCs [33,34]. Group I CLRs, which bind mannose and fucose, aid in pathogen phagocytosis, degradation, and antigen presentation to T-cells [33]. Group II CLRs, which bind glucan and dectin, appear to be more immunomodulatory; they induce upregulation of IL-10 and the secretion of cytokines

(specifically IL-1β, IL-6, IL-12, and IL-13) required for T-cell polarization into the TH1 or TH17 subsets [33,34]. CLRs also act in collaboration with other TLRs (TLR2, 4, 5, 7, and 9) to amplify preceding TLR-mediated NFκB activation and cytokine induction, in addition to triggering the complement cascade through β-1,3-glucan binding complement receptor-3 (CR3, CD11b/CD18), located in the membrane of many phagocytic cells [24,33,34].

Overall, while the binding domains and adaptor proteins vary, there is a significant overlap between the downstream signaling domains employed by each of these pathways. However, these pathways are far from redundant. While TLR7 and TLR9 are expressed on the endosomes of many cells including DCs, eosinophils, basophils, and B-cells, TLR3 and TLR8 are only expressed by natural killer (NK) cells [24]. In the same way, where TLRs are located mainly on leukocytes (macrophages, DCs, neutrophils, etc.), NLRs and RLRs can be found on all cells except DCs [26]. The complex signaling interplay between these pathways, in response to bacterial and viral antigens, highlights the importance of proinflammatory cytokines and PAMP-PRR detection in providing a tailored front-line defense against a wide variety of invading pathogens [26]. Further, the interplay between these PRR signaling pathways also drives the induction of effective adaptive immune responses, in that IL-1R and caspase-1 play a crucial role in development of both CD4+ and CD8+ T-cells, as well as antibody responses [25]. As such, IIMI-induced immune responses in the presence of biological therapeutics can lead to immunogenicity toward the administered biologic and potentially to other similar endogenous proteins [19], which can result in loss of treatment efficacy as well as severe and potentially lethal clinical consequences including anaphylaxis, serum sickness, and the formation of autoimmunity [19].

#### **3. Impact of IIMIs on the Immunotoxicity of Drug Products**

In the presence of IIMIs, activated phagocytes secrete both stimulatory and inhibitory cytokines to drive and regulate the immune response (Figure 2). These small proteins, which include interferons (IFNs), interleukins (ILs), tissue necrosis factors (TNFs), and chemokines, create a multilevel signaling network that elicits inflammatory responses, angiogenesis, as well as cellular activation, proliferation, and differentiation. IFNs play a central role in innate immunity to viruses and other microbial pathogens [2,29]. ILs function mainly as immune system regulators, responsible for immune cell differentiation and activation [2,29]. Multifunctional TNFs activate vascular endothelium permeability to allow entry of complement proteins and effector cells; increase fluid drainage to lymph nodes to clear pathogens and educate T/B-cells; and stimulate the production of IL-6 responsible for systemic fever, metabolite mobilization, and shock [2,29]. As the largest family of cytokines, chemokines have many diverse functions, ranging from controlling cell migration (e.g., recruitment and activation of local neutrophils and basophils to the site of infection), to such diverse processes as embryogenesis, innate and adaptive immune system development and function, and cancer metastasis [2,3].

Under normal circumstances, cytokine-driven immunostimulation is protective, such as when it is triggered by adjuvants to increase vaccine potency. However, when immune stimulation is unexpected or uncontrolled, especially in the presence of therapeutic compounds, it leads to unintended cellular immune responses and/or antibody production in response to that drug product. Such immunotoxicity encompasses 'any adverse effect on the structure or function of the immune system, or other systems affected by the same biological mediators (e.g., nervous and endocrine systems), as a result of immune system dysfunction' [35]. Immunotoxicity is further classified by the level of response, including (1) non-specific immunostimulation, (2) uncontrolled hypersensitivity (allergy, autoimmunity, and chronic inflammation) leading to tissue damage, and (3) immunosuppression [35].

In the most general terms, immunostimulation is the normal, controlled activation of an immune response ("sensitivity") to an antigen, an important prerequisite for immunogenicity [36,37]. Weak antigen sensitivity responses due to the simple presence of an antigen often fail to elicit sufficient immune activation required to trigger humoral or cellular immunity and subsequent clinical effects [36]; whereas moderate immunostimulatory responses, which might require the assistance of an adjuvant for additional phagocyte activation and cytokine secretion, can result in the eventual downstream production of neutralizing antibodies leading to therapeutic immunogenicity [22,38]. The most common symptoms of immunostimulatory reactions are fever, chills, malaise, hypotension, and localized tissue inflammation (redness, heat, swelling, and pain) around application [2,3,39]. These symptoms are often quickly resolved or can be controlled through the application of immunosuppressive agents such as recombinant chemokines or monoclonal antibodies [36].

Inappropriate or inadequately controlled immunostimulation may lead to hypersensitivity reactions (HSRs) [37]. While no universal classification of HSRs exists, the system proposed by Gell and Coombs, which classifies HSR reactions based on underlying mechanisms, time of symptom occurrence, mediators, and clinical manifestations, is frequently used [40]. Type I HSRs, or classic "acute allergic" reactions such as asthma or food allergies, result from antigen binding to immunoglobulin E (IgE) antibodies on the surface of granulocytes (basophils, mast cells), triggering cellular degranulation and an immediate release of histamine, leukotrienes, and other mediators [40–42]. While also antibody-driven, type II HSRs lead to the production of IgM and IgG antibodies as well as the activation of complement, natural killer (NK) cells, neutrophils, and macrophages [41], all of which result in cellular cytotoxicity and tissue damage. These types are reactions are commonly seen in response to medications such as penicillin, thiazides, or cephalosporins. Type III HSRs are driven by uncontrolled systemic complement activation, resulting in large deposits of IgM immuno-complexes and anaphylatoxins C3a and C5a in tissues which can trigger cell death and compromise organ function [40,43]. Examples of this type of HSR include serum sickness and autoimmune diseases such as rheumatoid arthritis and lupus erythematosus [42]. As both type I and type III HSRs result in the degranulation of basophils and mast cells, true IgE-mediated type I allergy reactions, which are referred to as "anaphylaxis" even though they lack complement involvement, are often difficult to distinguish from IgE-independent complement-activation related pseudoallergy (CARPA) reactions, also known as anaphylactoid or pseudoallergy, which do rely on complement anaphylatoxins C3a and C5a [42,44–46]. Lastly, type IV HSRs such as contact dermatitis or drug sensitivities, are delayed T-cell and macrophage-mediated reactions characterized by increased cytokine release and lymphocyte stimulation [40,42].

Anaphylatoxins and activation of immune cells by PAMPs and DAMPs, also trigger cytokine responses. Since cytokines are pleiotropic and have overlapping functions, they are normally very effective for small-scale localized responses [3]; however, whatever the antigenic trigger, the unregulated overproduction of cytokines due to strong/hyperimmunostimulation (a.k.a cytokine storm or cytokine-response syndrome) can quickly spread unchecked throughout the body via the circulation, resulting in overwhelming systemic inflammation, catastrophic tissue damage, disseminated intravascular coagulation (DIC), and death [2,22,24,38]. Due to their systemic nature, cytokine storms are most often associated with severe, widespread infections, high levels of IIMI contamination (e.g., endotoxins at doses above 5 EU/kg), or massive tissue damage (e.g., shock/trauma) [2,47–49].

Cases of delayed unregulated cytokine secretion coupled with prolonged tissue infiltration by activated macrophages and lymphocytes can also lead to other serious immunological consequences, such as the formation of chronic inflammatory or autoimmune diseases [35,39]. While differentiated by the source of the inflammatory trigger, either endogenous (autoimmune) or exogenous (chronic inflammatory), the general result is the same. Excess TNF production is associated with a number of chronic inflammatory and autoimmune diseases [2,29] while an over-activation of the complement system has been implicated in the pathophysiology of asthma and acute respiratory distress syndrome [43]. Similarly, prolonged exposure to over-activated immune cells, cytokines, and antibody/immune complexes can trigger the formation of granulomas, a common defense mechanism in which harmful components are isolated away from healthy tissue. These chronic HSRs are debilitating as well as life-threatening, since the cells of the immune

system are continuously attacking healthy tissues resulting in chronic pain, injury, and eventually organ failure [35].

Lastly, effective immune responses are normally a delicate and tightly controlled balance between stimulation and suppression. The systemic production of IL-10 is associated with the downregulation of neutrophil and monocyte function, working as an anti-inflammatory response following systemic inflammation [2,29]. While this natural counterbalance is conceptually beneficial in controlling systemic responses to local infections, immunotoxicity can occur when immunosuppression or dysregulation leads to an inappropriately reduced immune response resulting in frequent and serious adverse effects [35]. Since the majority of destructive immune responses are associated with HSRs, as previously discussed, many immunosuppressive therapeutics attempts to dampen overactive pro-inflammatory responses but instead have been reported to exacerbate asthma, eczema, and psoriatic lesions [2,39]. Dampening/deficiency of normal immune functions, such as the inhibition of T-cell function and adaptive immune responses, has also been associated with more frequent opportunistic concomitant infections (e.g., pneumonia, Candida, Kaposi's sarcoma, etc.) [35].

After activation by pro-inflammatory cytokines and PRR binding, local antigenpresenting cells (APCs), such as macrophages and dendritic cells (DCs), endocytose and degrade invading pathogens. APCs present fragments of these degraded pathogens on their membrane-bound major histocompatibility complex (MHC) receptors, which bind to and activate T-cells, initiating their downstream activation of B-cells [3,4,24,33]. The fate of activated T-cells is determined by the levels and types of cytokines induced during the inflammatory response, as well as the type and dose of antigen, type and affinity of MHC binding, route of administration, presence of other adjuvants, and patient genetic predisposition [4]. Major classes of T-cells include CD4+ helper (TH) T-cells activated by MHC class II antigen presentation, CD8<sup>+</sup> cytotoxic T-cells activated by MHC class I presentation, and regulatory T-cells (Tregs) [3,4,33]. In the presence of either IFNγ or a combination of IL-4, IL-6, and PGE-2, naive CD4<sup>+</sup> helper T-cells are further differentiated into specialized subsets of CD4<sup>+</sup> helper T-cells which are responsible for cell-mediated (TH1) or humoral (TH2) responses respectively [3,4]. TH1 T-cells secrete large quantities of IFNγ, in addition to IL-2, IL-3, IL-12, IL-18, GM-CS, and TNFβ, to regulate the inflammatory response and fight intracellular pathogens and viruses [3,4,33]. These cytokines promote macrophage activation and the production of opsonizing and complement-fixing antibodies. However, if not properly regulated, TH1-dependent immune reactions can also lead to antibody-dependent cellular toxicity and delayed HSRs, the most predominant of which can include autoimmune disorders, acute allograft rejection, and chronic inflammatory disorders [2,4,39]. On the other hand, TH2 T-cells secrete large quantities of IL-4, IL-5, IL-13, in addition to IL-3, IL-6, IL-9, IL-10, GM-CSF, and TNF, to induce humoral responses and mucosal immunity, as well as fight helminths and extracellular pathogens [3,33]. These cytokines promote the proliferation of mast cells and eosinophils, favor the differentiation of IgE and IgG-producing B-cells, and facilitate the synthesis of mucosal IgA [3,4]. While TH2 cells predominate in transplantation tolerance, they can also lead to chronic graft vs. host disease, systemic sclerosis, and allergen-reactive atopic disorders [4,39,43].

While it has been observed that cytokines from specific TH cell subsets (e.g. IFNγ from TH1 cells and IL-10 from TH2 cells) usually inhibit the action of the other types of T-cells and their companion phagocytes [3,4], this classic binary model does not account for instances where an immunological response is triggered without any significant shift in TH1/TH2 balance, such as is the case with omega-3 fatty acids, or alternatively where there is TH1/TH2 activation with minimal immunological pathogenesis, such as with melanin, probiotics and zinc [3]. In addition, other sub-classes of T-cells have been identified which were not previously represented by this model, including but not limited to: TH17 cells, which secrete IL-17 to mobilize phagocytes against extracellular fungi and bacteria; and Tregs, which produce FoxP3 to control the activity of the other effector TH cells and maintain immunological tolerance to self-antigens [3,19,23,33]. However, increased levels of

regulatory (TH17, Treg) cytokines such as IL-10 or IL-17 can also be an indication of adverse patient effects such as autoimmune diseases or advantageous concomitant infections [3].

#### **4. Sources of Immunotoxicity in Nanotechnology-Based Products**

The use of nanoscale platforms (e.g., dendrimers, liposomes, nanoparticles, nanotubes, nanogels, etc.) has become a popular technique to reduce drug immunotoxicity while improving therapeutic solubility, biodistribution, and cell-specific delivery compared to the traditionally formulated versions of these drugs. However, it has been noted that some nanocarriers can themselves be immunomodulatory (Figure 1), such as RNA nanoparticles which have been shown to induce pro-inflammatory cytokine secretion and enhance inflammation [11,50]. The raw materials used for nano-platform fabrication can have various immunological effects, either due to previously discussed contamination or due to the chemical properties of the material itself. Some nanomaterials are immunostimulatory, such as lipid-based nanocarriers and carbon nanotubes which have been shown to induce cytokine production and inflammation [50–52], while other nanomaterials are immunosuppressive including PEGylated NPs which lead to TLR9 inhibition and immune cell avoidance [50,51,53]. Similarly, the processes employed during nanocarrier synthesis and purification often use immunotoxic reagents, such as surfactants such as cetyltrimethylammonium bromide (CTAB); peptizing agents such as polystyrene sulfonate (PSS); or complexing agents such as nickel, to improve drug loading or enable molecule crosslinking [54]. While these chemicals are not generally intended to be in the final product, trace elements ("leachates") that remain after washing and filtration can induce cytokine production and inflammation, compounding the other immunomodulatory aspects of the nanocarrier [54].

Once fabricated, the physical properties of the nano-formulation, including size, shape, and surface charge, can also alter immunotoxicity. Nanoparticle interactions with the immune system have been extensively discussed elsewhere [11–13,50,55–58]. Here, we will use some examples to demonstrate structure–activity relationship between nanoparticle physicochemical characteristics and their immunological properties. First, several studies have shown that smaller particles (<500 nm) promote humoral TH2 responses, compared to very large particles (>1 μm) which have been found to stimulate cell-mediated TH1 responses. In addition, very small particles (<100 nm) are associated with increased CD8+ and CD4+ T-cell activation compared to their larger (>500 nm) counterparts, who induce good antibody responses [59]. Thus, small particles may invoke virus-like responses and larger particles induce bacteria-like responses [59]. Second, compared to spherical nanocarriers, oval-shaped liposomes and carbon nanotubes have been shown to activate complement and platelet aggregation with membrane rupture, respectively [50,60]. Finally, cationic carriers are more immunostimulatory than anionic or neutral carriers, triggering cytokine secretion (TNF, IL-12, IFNγ); activation of DCs, T-cells, and neutrophils; and procoagulant leukocyte and platelet activation which can potentially lead to DIC [12,50,61]. Taken together, while a nanocarrier is often designed to reduce the immunotoxicity of a therapeutic payload, the chemical and physical properties of that nanocarrier along with it being a source of undesirable IIMIs contamination may lead to an exaggeration of the immunotoxicity of the final drug product. For example, cationic polyamidoamine (PAMAM) dendrimers in the presence of low amounts of endotoxin have a variety of immunotoxic effects that neither dendrimers nor low levels of endotoxin alone have [11,50]. Therefore, the use of nanomaterial platforms should be considered as yet another source of IIMIs. Translational and regulatory challenges arising from immunomodulatory properties of nanocarriers and their ability to exaggerate immunotoxicity of low levels of IIMIs (e.g., endotoxin) have been extensively discussed elsewhere [12,13,50,56,61]. Immunogenicity of nanoparticles alone and in the context of IIMIs along with nanoparticle contribution to the immunogenicity of protein-based therapeutics have also been reviewed earlier [22].

#### **5. IIMIs Commonly Found in Pharmaceutical Products**

#### *5.1. Microbial Components*

When it comes to assessing biotherapeutic purity, the only current consensus is that it is important for manufacturers to minimize the potential for their formulation to trigger adverse patient reactions and future immunogenicity by removal of microbial or host cellrelated impurities, as summarized by testing standards (Table 1) [29,35,62–68]. Currently only a fraction of the potential IIMIs, specifically lipopolysaccharide (LPS), β−glucan, flagellin, HMGB1, and nucleic acids, are routinely measured during immunotoxicity screening of biotherapeutics [69] to confirm that the levels of these IIMIs fall within the FDA-approved 1–100 ppm range [21]. In addition, due to the breadth and complexity of potential IIMIs, there is currently no single assay that can provide a profile of all IIMIs present within a biotherapeutic [70]. Other than the fact that any assays used to detect IIMIs and evaluate possible immunotoxicity should be tailored to the specific contaminant [62], there is currently very little agreement as to which analytical assays should be standardized for IIMI screening [21]. Therefore, most studies use a series of assays to broadly cover the detection of all possible IIMIs present in biopharmaceuticals [21,71], including single analyte mechanistic assays, basic staining/gel-based assays, immunoassays, and cellularbased assays (Figure 2).

**Table 1.** Examples of guidance documents and international standards for the measurement of impurities in therapeutic products. International standards (IS) and Guidance for Industry (GI) provided through the U.S. FDA, the U.S. Pharmacopeia (USP), and the International Organization for Standardization (ISO) describe the risks of endotoxin and pyrogen contamination in therapeutic products and outline the assays, protocols, and detection limits which have been standardized and approved for universal application in therapeutic safety and purity measurements.


#### *5.2. Whole Microbes*

After biopharmaceutical manufacture and microbial inactivation via low pH adjustment, heat, and solvent/detergent treatments [10], filtration is used for the removal of bulk impurities such as neutralized pathogens (bacteria, viruses), destabilized protein aggregates, or other bulk contaminants [10,22]. Due to the comparatively large size of these impurities, microscopy techniques such as transmission electron microscopy (TEM) [10], have been used to assess the effectiveness of these initial filtration steps. These highresolution microscopy techniques employ lasers or electrons beams and extensive sample preparation to achieve a 0.1–1 mm visualization limit [73,74], which makes them timeand cost-prohibitive. Further, given their inability to provide accurate IIMI quantification, microscopy techniques such as TEM can only provide an indication as to what additional filtration and purification steps may be required; as these filtration techniques may not be sufficient to completely remove all traces of IIMIs, more accurate IIMI detection and quantification must then employ antigen-specific assays [31,73].

#### *5.3. Leachates*

After filtration, a range of chromatographic techniques are used for drug concentration and purification, to remove impurities such as drug by-products, unprocessed raw materials, and other leachates that may have been introduced into the formulation during the manufacturing process [10,22]. For complete sample separation, chromatography exploits the physical characteristics of the target protein/peptide in solution, including size, mass, ionic charge, binding affinity, pH, and electrokinetics, to partition it away from other components that may be present in the solution after fabrication [75]. Some of the chromatography techniques previously used for assessing biotherapeutic purity include ion exchange, size exclusion, capillary electrophoresis (CE), micellar electrokinetic chromatography (MEKC), and reverse-phase high-performance liquid chromatography (HPLC) [75]. Often referred to as "high pressure" liquid chromatography due to how the sample in the mobile phase is pressurized before injection into the absorbent stationary phase column, HPLC has become one of the most popular chromatography methods due to its high-performance detection, separation, and quantification of very small volumes (5–50 μL) of samples including degradation by-products, IIMIs, and unprocessed raw materials. HPLC is often used to separate molecules that are not large enough or charged enough for adequate separation by traditional size-exclusion chromatography or ion exchange-chromatography respectively [75]. While separation efficiency and quantification analysis are highly accurate, this technique requires extensive protocol optimization for the best results [75] in addition to specialized equipment and a trained operator. Additionally, chromatography can typically only separate one IIMI at a time, though multidimensional chromatographic separations paired with fluorescence detection are currently being pursued [71].

Sub-visible particles, which can include anything from small molecules to the components of protein aggregates, can also be identified using mass spectrometry (MS) techniques [21,76]. MS separates charged molecules or fragments by accelerating them through an electric or magnetic field, which separates the molecules based on their mass-to-charge ratio and then identifies them by correlation with known molecule masses and fragmentation patterns. This technique is especially important in identifying the relative concentrations of impurities and degradation products relative to target drug products during pharmaceutical development [77]. As a pivotal technique in the process of molecule structure elucidation [77], high-resolution MS/MS is now also being used to identify and quantify larger, more complex impurities and proteins that can be isolated from the bands of an electrophoresis gel or sampled directly from solution using liquid chromatographytandem mass spectrometry (LC-MS/MS) [20,21]. Due to improvements in high-throughput capabilities combined with improved sample preparation (e.g., chromatography fractionation and 2D gel electrophoresis), LC-MS/MS is now also being used for complete proteomic characterization and identification of complex therapeutic samples [21]. MS analysis is

more precise than immunoassays but requires specialized equipment and analysis software, as well as trained personnel [76].

#### *5.4. Host Cell Proteins*

The most difficult IIMIs to isolate and quantify are host-cell proteins (HCPs) due to the diversity and complexity of the potential protein repertoire, as well as HCP similarities to the target drug product [69]. As there is currently no single assay that can detect and quantify all possible HCP-based IIMIs within a biotherapeutic formulation [70] nor any absolute control limits required by pharmaceutical regulators [21], most quality assurance uses a combination of methodologies to confirm drug product purity. A typical strategy often includes generic IIMI clearance studies such as the *Limulus* amebocyte lysate (LAL) test or mass spectrometry; sensitive silver staining (and immunoblotting) of electrophoretic gels; and quantitative HCP-specific immunoassays such as ELISAs [71], all of which will be discussed below.

#### **6. Immune-Mediated Adverse Effects to Pharmaceutical Products**

The combination of a strong immunostimulatory response [3,35,43] and the activation of specialized subsets of T-cells leads to target-specific destruction of pathogens and cancer cells, either by direct interaction with CD8+ T-cells and natural killer (NK) cells or by CD4<sup>+</sup> T-cell activation and proliferation of B-cells to produce antigen-specific antibodies [19,23,24,78]. This IIMI-driven immunogenicity can lead to the formation of antibodies of different isotypes (e.g., IgM vs. IgG vs. IgE), allotypes (e.g., reflecting genetic differences between IgG of biologically unrelated individuals), and idiotypes (e.g., reflecting binding to specific epitopes within antibody variable sites) [19,23,79–81], resulting in anti-drug antibodies (ADAs) with varying impacts on drug effectiveness. Binding antibodies attach to a non-active portion of the therapeutic and therefore have little/no effect on therapeutic function, whereas cross-reactive neutralizing antibodies bind to therapeutic active sites, thereby neutralizing therapeutic function while also binding similar endogenous proteins and breaking immunological tolerance [19,23,82–84]. The presence of these ADAs can also have different functional consequences to the host including the HSR/anaphylaxis and autoimmune responses previously discussed [19,23,35,79–81]. The relationship between the occurrence of a specific antibody type and the impact on the patient are inversely related; binding antibodies are the most common but have the lowest clinical impact, while crossreacting neutralizing antibodies are rare but have the highest clinical impact [23,79–81,85]. Therefore, it is important to understand, measure, and prevent this response from being induced.

During the fabrication and production of drug compounds, there are many potential sources for the introduction of IIMIs into the final biotherapeutic formulation (Figure 1) [19,20]. In addition to the impurities/contaminants previously discussed, there are also several product-related and host-related factors that may have little/no impact on the function of the resulting drug product but have been shown to impact the immunotoxicity and immunogenicity of biotherapeutics [19,23,78]. Product-related factors include structural properties of the drug (sequence, epitopes, post-translational modifications), exposure to antigenic sites, solubility, formulation stability and storage, downstream processing, presence of impurities/contaminants that might be introduced during processing [19,78]. These factors can be mostly controlled through careful optimization and modification of the fabrication/purification processes. Further compounding the risk of immunogenicity are host-related factors, including host genetic predisposition, endogenous protein genetic variants, concomitant illnesses (e.g., kidney or liver diseases), host immune status (e.g., autoimmunity, prior exposure) as well as the treatment dose, duration, and route of administration [19,23,78].

#### **7. Methods for IIMI Detection**

#### *7.1. Direct Detection Methods*

The first bioassay used to measure the presence of bacterial contamination was the rabbit pyrogen test (RPT) which detected pyrogens, any contaminant that induces a histamine response, fever, chills, and other unwanted inflammatory side effects. The rabbit pyrogen test detects all pyrogens, so it is subject to high variability and low selectivity, in addition to being expensive and requiring extensive use of animals [10,31]. As an improvement, the *Limulus* amebocyte lysate (LAL) test detects the hemolymph coagulation of the American horseshoe crab *Limulus polyphemus* when in the presence of bacterial endotoxin/LPS and is used as a standard for bacterial contamination [86,87]. However, this assay is specific for endotoxin, not general pyrogens [31], and has reduced specificity in the presence of fungal β-glucans because the horseshoe crab lysate used for this assay contains two proteins that trigger activation of the proteolytic cascade: factor C is specific to the presence of endotoxin while factor G is specific to β-glucans [88,89]. Knowing this, a modified version of the LAL assay containing glucan-blocking reagents or recombinant factor C overcomes β-glucan interference during endotoxin detection [90].

While β-(1,3)-d-glucans are not as immunologically potent as bacterial endotoxins, requiring μg/mL concentrations as compared to the endotoxin pg/mL concentrations to elicit an immunomodulatory response, they are a common IIMI present in many pharmaceutical products and solutions [89]. Moreover, while there is currently no compendial standard for β-glucan detection or acceptable levels, a modified version of the LAL assay is growing in popularity [90]. Since LAL factor G is specific to β-glucans, factor C depletion from the LAL lysate enhances the assay's sensitivity solely to β-glucan detection [89]. It is important to note that β-glucans are naturally introduced in a person's diet, so data generated from β-glucan quantification assays need to be from clinically relevant doses of the drug formulation [89].

Challenges with endotoxin and beta-glucan detection in nanoformulations stemming from carrier-, excipient-, or drug-mediated interferences, mechanisms of interferences, and ways for overcoming them have been identified and extensively discussed earlier [11,89,91–95].

#### *7.2. Indirect Detection Methods*

For the development of effective assays, an appropriate biomarker can consist of any compound (e.g., metals, solvents, pathogens, etc.) or useful characteristic, such as a mechanistic by-product, which can be measured or evaluated, either directly or indirectly, and used as an indicator of normal biological, pathogenic, or pharmacologic processes [83,84]. Therefore, any of the product- or host-related impurities previously discussed, as well as raw materials used during the product's manufacture and purification, can technically be considered a potential biomarker [85]. During method development, quantitative assays must be validated using appropriate controls and quantification must employ a standard curve of known analyte concentrations to determine the range of conditions under which appropriate levels of confidence can be attributed to the reproducibility and accuracy of the data [84,96]. Further, the validated assay must then demonstrate both sensitivity and specificity for the biomarker [84], such that the biomarker is correctly identified (i.e., true positive, sensitivity) at clinically relevant (ng/mL to pg/mL) concentrations [96] without also reacting to residual therapeutics or other impurities likely to be present within the therapeutic formulation (i.e., true negative, specificity). Reduced sensitivity can result in mistakenly missing the presence of IIMIs in a formulation (i.e., false negative) resulting in possible dangerous clinical manifestations and immunogenicity, while reduced specificity can result in misidentification of inert compounds as IIMI (i.e., false positive) leading to incorrect quantification and product disposal rather than administration to patients. Overall, when balancing these two parameters, increased sensitivity is often preferred to increased specificity.

#### *7.3. Biological Staining and Gel-Based Methods*

Biological staining is a common technique to detect and visualize the presence of HCPs and other impurities. This technique utilizes Coomassie Blue or silver staining to highlight the presence of protein analyzed by multidimensional (2D or 3D) gel electrophoresis [71] or fixed in histological samples, respectively [21]. While the sensitivity of these staining techniques is quite high, selectivity is not; this technique cannot discriminate between types or sources of proteins so other techniques need to be employed to further identify and quantify the protein contaminants [10]. Newer versions of this method employ fluorescent stains, such as SyproRuby, for 10–100 times increased sensitivity compared to previous stains since these stains are not dependent upon the protein composition [21]. Other stains also have improved specificity by binding to specific cellular elements (i.e., nucleic acids, carbohydrates, chromatin, etc.) though this method is still largely qualitative [21]. Gel electrophoresis and protein staining have progressed to the use of the more quantitative Western blot, a common antibody-dependent detection method [21] that has merit for identifying low (pg/mL) concentrations of protein impurities. Contaminating HCPs and product-related impurities are separated from the target biologic by gel electrophoresis [10], and then transferred to a PVDF or nitrocellulose membrane. Primary antibodies raised against HCPs are incubated with the membrane to allow for the formation of antigen– antibody complexes, which are then detected through secondary enzymatic or fluorescent labeling [21]. While this technique is both sensitive and specific, it requires the use of separate polyclonal antibodies against each impurity for optimal detection, which can be time and cost prohibitive in the long run [10]. In addition, this technique needs to be supplemented with additional immunoassays to help distinguish between process- or product-related impurities and impurities that might comigrate with the product [10].

#### *7.4. Antibody-Based Enzymatic Methods*

Surface plasmon resonance (SPR) uses antigen-ligand binding on a sensor chip to generate a signal due to a change in the refractive index caused by a difference in mass as the analyte binds to the ligand. Most often used to detect the presence of antibodies rather than antigens, this assay is capable of continuous measurements of binding interactions in 'realtime' [84]. For the detection of immunotoxic antigens, SPR assays tend to be less sensitive, less tolerant to therapeutics, and have lower throughput compared to enzyme-linked immunosorbent assays (ELISAs). In fact, SPR is capable of characterizing early immune responses by detecting and isotyping low-affinity antibodies, which other assays might miss, which makes it much more suitable for immunogenicity assays [70,97]. Furthermore, unlike other immunoassays where the reagents are cost-prohibitive, here the detection equipment is expensive and vendor specific [70,97].

Electrochemiluminescence (ECL) also uses antibodies to bind target impurities. However, unlike the commonly used enzyme-labeled secondary antibodies previously discussed, this technique employs a ruthenium-conjugated protein and tripropylamine (TPA) to produce a detectable, quantifiable luminescent signal. Ruthenium labels are stable, non-radioactive, and offer a choice of convenient coupling chemistries [70]. This is a highly sensitive and selective technique; however, this method requires the production and use of specific antibodies for analyte immobilize and detection, indicating that each impurity must be detected separately [98]. In addition, this technique requires the use of specialized, costly equipment containing carbon electrode plates for detection, which are not necessarily standard in most labs [10,70].

Enzyme-based (EIA) or fluorescence-based (FIA) microtiter plate assays were developed to circumvent the need for method-specific instrumentation and resources experienced with ECL and SPR [10]. This assay involves incubating the sample with a couple of biotinylated antigen-specific antibodies which, after binding and forming immunocomplexes, are removed from solution by association with streptavidin-coated paramagnetic beads. Thereafter, the beads are incubated with enzyme-labeled or fluorescence-labeled antibodies for colorimetric development. By substituting the paramagnetic beads for a

solid-substrate surface, the traditional EIA/FIA was transformed into the enzyme-linked immunosorbent assay (ELISA), the most practically useful and commonly employed immunoassay [21]. As previously described, this type of assay employs a series of antibodies to capture specific target antigens. The bound antigen is then complexed with a secondary antibody modified to undergo an enzymatic reaction (colorimetric, fluorescent, or luminescent) for detection via spectrophotometer [70]. However, unlike the EIA/FIA, the use of a solid-substrate surface enables the assay to be set up in various configurations (e.g., sandwich, indirect, bridging, competitive, etc.) for optimal IIMI detection and quantification. ELISAs are relatively sensitive with a detection range of 12–200 ng/mL [10,99]; modern ELISAs have been optimized to improve their sensitivity and allow the detection of analytes at lower (e.g., pg/mL) levels. ELISAs also have high specificity due to their use of analytespecific antibodies and can be performed relatively quickly (completed in one day) [10,21]. However, the dependency on highly specific antibodies also means that each analyte must be known and analyzed individually, which can be cost-prohibitive [70]. Common HCPs detected via ELISA include anaphylatoxins such as complement C3a [100]; inflammatory cytokines such as IL-1β, IL-6, IL-8, and TNFα [2,101]; and other IIMIs including HMGB1 and flagellin.

Due to antibody specificity combined with the progression of fluorophore technology, a large number of biomolecules can now be captured from the same small (μL to mL) sample and then detected simultaneously [3,71]. These "multiplex" assays are usually modified ELISA assays, though the EIA/FIA assay format can similarly be multiplexed, as is often used in flow cytometry [3]. Each analyte is then tagged with either a different fluorescent label or organized in a known array for detection via spectrophotometer. As the basic principles of the assay are unchanged, the sensitivity and specificity are still high, though fluorescence bleed-through increases as the number of analytes and fluorophores with similar excitation/emission spectrums increases. In addition, multiplexed assays are less time consuming and labor intensive, while providing higher throughput analysis, compared to an individual ELISA [3].

#### *7.5. Nucleic Acid Hybridization Methods*

For the detection of nucleic acids in pharmaceutical samples, hybridization techniques such as the dot blot or immunoligand assay (ILA) are often used. The ILA (a.k.a "Threshold Assay") reliably detects very small amounts of DNA and impurities in liquid solution [102]. This assay employs a biotinylated single-stranded binding (SSB) protein and general antissDNA antibody to complex with any host ssDNA available in the sample. Streptavidin filtration then captures any biotinylated complexes on a specialized matrix-embedded silicon chip, after which the DNA is quantified via enzymatic hydrolysis and subsequent light-addressable potentiometric sensor (LAPS) detection [99]. This method has been shown to be 10–100 times more sensitive than traditional colorimetric or ELISA assays, with a detection range of 5–40 ng/mL [99], requires only small amounts of sample, removes steric binding or stability issues inherent in solid-phase systems, and comes in two formats (sandwich or competitive) depending on the size of the analyte being detected [102] though optimal ssDNA fragments tend to be larger than 600 base pairs [99]. However, this method has reduced specificity due to its sequence-independent binding by general ssDNA antibodies. Furthermore, this technique can be expensive as it requires the use of proprietary equipment, software, and consumables (e.g., silicon chips, specialized buffers, etc.) for quantification [10]. On the other hand, the dot blot employs a substrate covered with immobilized "randomly primed" DNA probes from a known microbial source tagged with radio or fluorescent labels. The probes are exposed to the drug sample allowing for binding between host-cell DNA present in the sample and the specific DNA probes. This binding is then detected and quantified to 3–800 pg/mL against a calibration curve by phosphor- or fluorescence-imaging systems [99].

The more popular method of detecting and identifying bacterial and viral nucleic acids is through reverse transcriptase (RT) and quantitative polymerase chain reaction (qPCR) assays [10]. For these assays, trace amounts of DNA or RNA are collected and then amplified through the PCR or RT-PCR method respectively, resulting in many identical copies of the target DNA. The levels of target DNA are then quantified and nucleic acid concentration in the original sample is derived from target copy numbers [99]. Innate immune activation can similarly be assessed by quantifying the levels of pro-inflammatory cytokines, such as IFNγ, IL-1β, IL-6, or other downstream biomarkers by quantifying the levels of target mRNA, amplified as cDNA, which are compared to standard housekeeper genes such as GAPDH or 18S [9,69,103] to determine the fold increase or decrease of the target genes [9,69,103]. Since this process uses specific DNA primers for PCR amplification, the resultant quantification is highly sensitive and specific for the target sequence [99]. However, this also means that species-specific primers must be known. Additionally, as amplification of each nucleic acid fragment requires its own primers, these reactions need to be carried out separately; though, like the previously discussed, multiplexing and proteomics analyses coupled with improvements in high-throughput capabilities have produced arrays of many immobilized primers used to amplify, identify, and quantify many different DNA sequences at the same time [9,21]. This standardization increases the amount of data produced while reducing the required time and labor of these assays [76].

#### *7.6. Cell-Based Methods*

Since the long-term goal of these studies is the prevention of patient immunotoxicity and possible immunogenicity, more recent assays focus on the in vitro and in vivo impact of IIMIs. These cellular assays detect immune cell activation and proliferation or quantify levels of secreted innate immunity biomarkers (e.g., cytokines, prostaglandins, complement), which may contribute to the process of immunogenicity by priming the immune cells.

Cellular proliferation assays examine the activation and proliferation of specific immune cell subsets, usually, macrophages, neutrophils, or lymphocytes, when treated with the biotherapeutic, compared to control cells and the potential adjuvant effect of known IIMIs [31]. For example, T-cells are activated by concanavalin A or phytohemagglutinin, while B-cells proliferate in response to LPS. While it has long been established that immune cell proliferation in vitro is correlated with cell-mediated immunity, these assays have not been extensively standardized and validated [36]. In addition to needing a skilled technician and the appropriate facilities to support these studies, this assay is time prohibitive as culturing these cells takes at least 48–72 h [36].

For a more specific way to determine the type of IIMIs present in a drug formulation, a model of HEK-BLUE cells containing a secreted embryonic alkaline phosphatase (SEAP) reporter inducible by NFκB, transfected with individual TLR receptors, can be used. When bound with their specific agonists alone or in mixtures of IIMIs, the observed NFκB activation for each TLR can be quantified through a colorimetric change. This reporter system has high sensitivity and specificity, similar to what was observed in normal human PBMCs [69,88]. Since therapeutic biologics could mask or interfere with the response of these cell lines, this model necessitates the use of additional inhibition controls. In addition, while this model is effective for detecting TLR-specific IIMIs, it does not yet cover innate immune responses that can be triggered solely through alternate pathways such as CLRs, NLRs, and RLRs. As such, the reporter cells were modified to contain different reporter systems (SEAP, THP-1, and MM6) that would be expressed in the presence of NFκB, TNFα, and mRNA from IL-6 or IL-8 respectively, thereby covering the activation of multiple innate immune responses [69].

Other in vitro models instead directly quantify the levels of cell-secreted immune modulators, such as cytokines and complement proteins (e.g., C3a, C5a), or antibodies [3]. While all of these soluble mediators play an integral role in host defense against microbial invasion, the network of cytokine interactions is responsible for maintaining cellular homeostasis, making them a popular biomarker for gauging the potential immunotoxicity and immunogenicity of new biotherapeutic compounds, especially when compared to normal

(untreated) controls [3,36]. Increased levels of cytokines after application of a new drug product can therefore be associated with a product's immunotoxic effects (either stimulatory or inhibitory), which can lead to adverse patient reactions and reduced therapeutic efficacy due to the formation of ADAs [3]. As such, pharmaceutical immunogenicity is often determined through the use of commercially available multiplexed ELISA assays, chosen based on convenience, affordability, and availability [3], which typically quantify a limited panel of pro-inflammatory cytokines (IL-1, IL-6, or TNFα) [2,87] or subsequent T-effector (TH1/TH2) cytokines, including IL-2, IL-12, IFNγ or IL-4, IL-5, and IL-6 respectively, even though this may bias analysis towards specific immune pathways [2,3]. Despite the sensitivity and specificity of the multiplex ELISAs used for these analyses, the pleiotropic nature of cytokines and their overlapping activation pathways on numerous target cells [36] often make the results difficult to interpret. Hence, there is a lack of consensus as to which cytokines should be measured to accurately characterize the immunological effects of a new drug.

#### *7.7. In Vivo Methods*

A more recent study performed by Haile *et al.* employed an in vivo macaque skin model to better characterize the relationship between type and dose of IIMIs, patterns of innate immune receptors, and pathways triggered by these impurities, and immunogenicity. This model was developed due to the similarity of macaque PBMCs to human PBMCs, and to increase sensitivity compared to traditional murine models which are known to have less sensitive immune cells than those of humans [31]. These studies used mRNA collected after application of known IIMIs, as a basis for comparison to Rasburicase, as a model therapeutic, and measured by qRT-PCR to track the expression of 48 genes involved in the innate immune response, including ILs, TNFs, CD40, GAPDH, etc. [31]. This study demonstrated that, while an increased innate immune response is dependent upon the dose of IIMI administered, the presence of these impurities acted as an adjuvant during coadministration with a protein therapeutic, thereby increasing its immunogenicity. However, it was noted that even trace amounts of IIMIs triggered the transcription of multiple innate immunity genes, emphasizing the need to assess biotherapeutics for a wide variety of possible contaminants and related downstream biomarkers in a more thorough, relevant model, rather than just quantifying levels of specific IIMIs [31]. While the use of animal models is cost-, labor-, and time-prohibitive, these models can provide more applicable data as to the immunotoxicity and immunogenicity of biotherapeutics in humans.

Overall, methods of cell growth and stimulation are more or less optimized and standardized, and cell-based assays (both in vitro and in vivo) provide the most relevant data on IIMI and drug interactions with the immune system [3,76]. However, they are labor intensive, and the evaluation of cell-secreted biomarkers is subjective due to the cross-reactivity of most immunological pathways and the potential confounding influence of other substances that may modulate the activity of the target substance [3,76].

#### **8. Conclusions and Future Directions**

It is well documented that the presence of IIMIs in a biotherapeutic formulation can trigger immunotoxicity and, with repeated exposure, immunogenicity against the therapeutic [31,80–82]. To prevent these adverse patient reactions, the FDA currently requires quantification of five key IIMIs: LPS, HMGB1, β-glucan, flagellin, and nucleic acids [62,69], to demonstrate biotherapeutic safety, quality, and clinical performance. These guidelines aim to mitigate the formation of future ADAs through commonly activated innate immune receptors, specifically TLRs, CLRs, and complement. However, these guidelines do not necessarily account for potential immunotoxic responses to other IIMIs that may be present in the formulation. As such, the FDA panel of IIMIs required for quantification should be expanded to cover a much broader repertoire of impurities, including microbial antigens that can potentially trigger other innate immunity pathways, common manufacturing leachates, and solvents, and toxic additives required for maintaining host cells. The list of

possible leachates, solvents, and host cell additives will be extensive, requiring tailoring to the specific processes employed during manufacturing and purification [62]. As for the microbial IIMIs, most innate immunity receptors and pathways can be covered using ten common IIMIs, some of which are already required and discussed, including flagellin, FSL-1, zymosan, ODN2006, and ODN2216, both high- and low-molecular-weight poly(I:C), MDP, CLO75, and LPS. While this ten IIMI panel necessitates more laboratory testing before new drugs can gain approval, adhering to the ppm levels for these required IIMIs will demonstrate that little/no immunotoxicity will result from trace levels of substances present in the drug formulation, therefore reducing the potential for immunogenicity.

Second, to measure and quantify IIMIs present in biotherapeutic formulations, a variety of available assays have been discussed. As genomic and proteomic technology advances, these assays have become more sensitive and specific, enabling improved detection and quantification of IIMIs. In addition, many of these assays are now being coupled into high-throughput formats which can produce more data with reduced sample and reagent volumes, as well as cost and labor expenditures. However, due to the variety of potential IIMIs, there is currently no single assay that can provide a profile of all IIMIs present within a biotherapeutic [70]. Moreover, there is a lack of agreement as to which analytical assays should be standardized [21] so most studies use a series of assays to broadly cover the detection of all possible IIMIs present in biopharmaceuticals [21,71]. To better standardize results across experiments and laboratories, the use of a single high-throughput platform capable of detecting a wide panel of biomarkers of the same class (small molecules, proteins, or nucleic acids) in parallel, such as multiplexed ELISAs, MS, or genomic arrays, should be employed.

Finally, given that immunostimulation is the overall concern, the use of newer cell-based assays which track levels of biomarkers (e.g., cytokines, transcription factors, mRNA [62]) affected by the presence of IIMIs, rather than the individual IIMIs themselves, can provide a stronger connection between the applied biotherapeutic and its impact on immunotoxicity and immunogenicity [31]. Past cellular studies focusing on a limited selection of cytokines and chemokines, usually, a combination of pro-inflammatory IL-1, IL-8, IL-6, TNFs, and IFNs, have failed to adequately interrogate the entire immune cascade [2,36]. Since immunotoxicity can cover a range of patient responses from immunostimulation and HSR to immunosuppression, measuring a wider assortment of cytokines, including but not limited to IFNs (α, γ, λ); ILs (1α/β, 2, 6, 8, 10, 12, 17); interferon-gamma inducible protein (IP-10); TNFα, prostaglandin-E2 (PGE-2), macrophage inflammatory protein (MIP-1α), and monocyte chemoattractant protein (MCP-1), can provide a more complete picture as to the type and degree of immunotoxic response that can potentially be triggered by a new biotherapeutic formulation.

**Author Contributions:** All authors (C.K.H. and M.A.D.) researched literature and wrote manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported in whole by federal funds from the National Cancer Institute, National Institutes of Health, under contract 75N91019D00024. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### MDPI

St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Molecules* Editorial Office E-mail: molecules@mdpi.com www.mdpi.com/journal/molecules

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-8167-5