**A Nano-Emulsion Platform Functionalized with a Fully Human scFv-Fc Antibody for Atheroma Targeting: Towards a Theranostic Approach to Atherosclerosis**

**Samuel Bonnet 1,2,\* ,† , Geoffrey Prévot 3,† , Stéphane Mornet <sup>2</sup> , Marie-Josée Jacobin-Valat <sup>1</sup> , Yannick Mousli <sup>3</sup> , Audrey Hemadou <sup>1</sup> , Mathieu Duttine <sup>2</sup> , Aurélien Trotier <sup>1</sup> , Stéphane Sanchez <sup>1</sup> , Martine Duonor-Cérutti <sup>4</sup> , Sylvie Crauste-Manciet 3,† and Gisèle Clofent-Sanchez 1,†**

	- **\*** Correspondence: samuel.bonnet@univ-angers.fr
	- † These authors contributed equally.

**Abstract:** Atherosclerosis is at the onset of the cardiovascular diseases that are among the leading causes of death worldwide. Currently, high-risk plaques, also called vulnerable atheromatous plaques, remain often undiagnosed until the occurrence of severe complications, such as stroke or myocardial infarction. Molecular imaging agents that target high-risk atheromatous lesions could greatly improve the diagnosis of atherosclerosis by identifying sites of high disease activity. Moreover, a "theranostic approach" that combines molecular imaging agents (for diagnosis) and therapeutic molecules would be of great value for the local management of atheromatous plaques. The aim of this study was to develop and characterize an innovative theranostic tool for atherosclerosis. We engineered oil-in-water nano-emulsions (NEs) loaded with superparamagnetic iron oxide (SPIO) nanoparticles for magnetic resonance imaging (MRI) purposes. Dynamic MRI showed that NE-SPIO nanoparticles decorated with a polyethylene glycol (PEG) layer reduced their liver uptake and extended their half-life. Next, the NE-SPIO-PEG formulation was functionalized with a fully human scFv-Fc antibody (P3) recognizing galectin 3, an atherosclerosis biomarker. The P3-functionalized formulation targeted atheromatous plaques, as demonstrated in an immunohistochemistry analyses of mouse aorta and human artery sections and in an *Apoe*−/<sup>−</sup> mouse model of atherosclerosis. Moreover, the formulation was loaded with SPIO nanoparticles and/or alpha-tocopherol to be used as a theranostic tool for atherosclerosis imaging (SPIO) and for delivery of drugs that reduce oxidation (here, alpha-tocopherol) in atheromatous plaques. This study paves the way to non-invasive targeted imaging of atherosclerosis and synergistic therapeutic applications.

**Keywords:** atherosclerosis; nano-emulsion; magnetic resonance imaging; stealth; human antibody

#### **1. Introduction**

Atherosclerosis is characterized by the development of lipid-rich plaques, called atheromatous plaques, in the artery wall [1]. Atheromatous plaques can be classified into two types: stable plaques and vulnerable plaques [2]. Stable plaques are usually rich in extracellular matrix and smooth muscle cells that maintain the integrity of these fibrous

**Citation:** Bonnet, S.; Prévot, G.; Mornet, S.; Jacobin-Valat, M.-J.; Mousli, Y.; Hemadou, A.; Duttine, M.; Trotier, A.; Sanchez, S.; Duonor-Cérutti, M.; et al. A Nano-Emulsion Platform Functionalized with a Fully Human scFv-Fc Antibody for Atheroma Targeting: Towards a Theranostic Approach to Atherosclerosis. *Int. J. Mol. Sci.* **2021**, *22*, 5188. https:// doi.org/10.3390/ijms22105188

Academic Editors: Annamaria Sandomenico and Menotti Ruvo

Received: 25 March 2021 Accepted: 9 May 2021 Published: 14 May 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/).

plaques for years. Conversely, vulnerable plaques are rich in macrophages and inflammatory cells that make them prone to rupture, leading to cardiovascular complications [3–5].

Currently, angiography is normally used for imaging peripheral arterial disease [6]. However, it gives information only on vessel lumen reduction (stenosis) but not on the plaque morphology and risk of rupture [7]. Moreover, in two-thirds of ruptured plaques, stenosis is insignificant on angiograms [8,9]. Recently, other imaging modalities have emerged. For instance, intravascular ultrasound and optical coherence tomography [10–13] provide information on plaque morphology but are invasive procedures. Other noninvasive imaging strategies might offer new opportunities for atheroma diagnosis [14,15], particularly magnetic resonance imaging (MRI), which combines excellent soft-tissue contrast, good resolution, and absence of exposure to ionizing radiation [16–19].

MRI with highly specific targeting probes, in which contrast agents are conjugated to antibodies against molecular components of the atheromatous plaque, might allow the contrast agents to be directed specifically to the lesions. Molecular imaging of vulnerable plaques is of utmost interest because plaque composition might contribute to plaque rupture more than artery narrowing. The aim of this study was to develop an improved targeted contrast agent using nano-emulsions (NE) and the first human antibody (HuAb) against galectin-3 (HuAb P3, WO2019068863A1). Indeed, previous work demonstrated that oil-in-water NEs loaded with superparamagnetic iron oxide (SPIO) nanoparticles generated an accurate MRI signal, making them highly suitable as imaging agents [20,21]. The P3 antibody was chosen because galectin-3 is strongly expressed by the TREM2-positive foamy macrophage subset [22], which has been recently identified by single-cell RNA sequencing as the main immune cell subset in atherosclerosis. Importantly, the TREM2 positive subset endowed with specialized functions in lipid metabolism and catabolism was almost exclusively detectable in atherosclerotic aortas and present at different time points of lesion formation [23]. Moreover, using a HuAb decreases the potential immunogenicity in clinical settings. Before conjugation to the P3 HuAb, the NE surface was decorated with polyethylene glycol (PEG), the most common method to reduce clearance from the blood circulation [24]. PEG macromolecules create a protective hydrophilic layer around nanoparticles that can repel binding by opsonin proteins (i.e., opsonization) [25] and increase their half-life in the blood [26]. PEGylation of liposomes [27], micelles [28], and nanoparticles [29] has been widely investigated, but only a few works have focused on NE surface modification with PEG. Hak et al. [30] studied the impact of PEG surface density on NE half-life by blood sampling at different time points after administration in mice. Here, the effect of NE surface modification by PEG layers of different molecular weights (PEG<sup>2000</sup> and/or PEG3400) on liver uptake was studied by dynamic MRI. The HuAb P3 engineered via the single chain fragment variable (scFv)-Fc format was conjugated to the formulation with the lowest liver uptake and a longer half-life according to previous procedures [31].

Finally, theranostic PEGylated NEs loaded with SPIO nanoparticles and alpha-tocopherol were developed. Oral supplementation with alpha-tocopherol failed to show a clear benefit on the reduction of cardiovascular events in clinical trials [32]. Indeed, systemic administration does not allow one to reach the critical active concentration of the specific antioxidant at key sites [33]. In this study, we aimed to tackle this issue by proposing a targeted drug delivery strategy that could overcome the failure of specific tissue concentration of orally administered antioxidants. The antioxidant properties of PEGylated NEs loaded with SPIO nanoparticles and alpha-tocopherol were assessed as a multi-modal tool for atherosclerosis imaging and treatment.

#### **2. Results**

#### *2.1. PEGylated NE Formulations and Characterization*

Formulations were prepared using Miglyol 840, a neutral pharmaceutical oil based on propylene glycol diether of C8 and C10 saturated plant fatty acids. Two surfactants were used to stabilize the oily droplets, Tween 80 and Lipoid E80. The NE formulations were loaded with maghemite-based SPIO nanoparticles to develop T2\*-shortening MR contrast agents with relaxivity values close to the gold standard, as previously described [20] (r ∗ 2 = 42 to 45 mM−<sup>1</sup> ·s <sup>−</sup><sup>1</sup> and r<sup>1</sup> < 0.1 mM−<sup>1</sup> ·s −1 ). To produce stealth NEs with limited liver uptake and, consequently, a longer half-life in the blood circulation, the droplet surface was PEGylated. PEG creates a hydrophilic and biocompatible layer that limits opsonin adsorption [34,35] and non-specific cellular uptake compared with unmodified carriers. Several therapeutic strategies using PEG have been approved by the Food and Drug Administration (FDA) [25]. In this study, two lipid–PEG combinations with different molecular weight and functionalization were used: DSPE–PEG<sup>2000</sup> and DSPE–PEG3400– maleimide. The maleimide linker allows the conjugation of the HuAb P3. The two lipid–PEG combinations were added at the same molar concentration (5 µmol/mL) to the oily phase before the phase inversion step. Some formulations included only lipid–PEG<sup>2000</sup> (#2), only lipid–PEG3400–maleimide (#3), or a mixture of both (#4). Non-PEGylated NEs (#1) were used as controls.

Decoration of the droplet surface with PEG increased the hydrodynamic diameter from 175.8 nm for NE (#1) to 190.9 nm, 197.2 nm, and 191.0 nm, for NE–PEG<sup>2000</sup> (#2), NE–PEG3400–maleimide (#3), and NE–PEG2000/3400–maleimide (#4), respectively. The size of all NE formulations remained in the submicronic range and their polydispersity index (PdI) was <0.2 (monodisperse samples) (Table 1). Thus, the formulation diameter remained smaller than the tiniest blood vessel, preventing their occlusion [36]. Moreover, the nature of the lipids associated with NE plasticity might favor deeper tissue penetration and biological barrier crossing [37].


**Table 1.** NEs' physicochemical data.

NE: nano-emulsion; PEG: polyethylene glycol; SD: standard deviation.

SPIO nanoparticle inclusion in the oily droplets was confirmed by transmission electron microscopy (TEM) analysis (Figure S1). Nanoparticle tracking analysis (NTA) was used to determine the size and number of submicron particles. The size distribution was consistent with the dynamic light scattering (DLS) analysis results and the droplet number was 5.75 <sup>×</sup> <sup>10</sup><sup>13</sup> <sup>±</sup> 3.66 <sup>×</sup> <sup>10</sup><sup>12</sup> droplets per mL. To our knowledge, this is the first time oily droplet number has been determined to control the theoretical antibody:PEGylated NE ratio.

#### *2.2. Stealthy Features of PEGylated NEs*

Among the several promising new drug delivery systems, NEs are an advanced technology used to carry molecules to a specific site, and several NE formulations are already used in clinics [38–40]. Before improving NE targeting thanks to an antibody conjugated to its surface, it was important to characterize the in vivo stealthy feature of each NE formulation. Stealth is a parameter directly related to the half-life in the bloodstream. Here, the stealthy behavior of three PEGylated NE formulations ((#2), (#3), and (#4)) and of non-PEGylated NE ((#1); control) was studied by dynamic MRI after NE injection at the same iron concentration (3 mg/kg bodyweight) in the tail vein of C57BL/6 mice. The iron concentration of 3 mg/kg was chosen based on guidelines and doses used in the literature for ferumoxytol, approved by the FDA for anemia treatment. The dose of 3 mg/kg of ferumoxytol seems to be well tolerated for MRI-based diagnostic imaging without serious adverse events, according to a recent multi-centric study that underlined the positive safety profile [41]. Liver uptake was monitored continuously by MRI for about 10 min before and up to 50 min after injection (Figure 1).

50 min after injection (Figure 1).

**Figure 1.** Estimation of the in vivo liver iron (Fe) uptake in mice by dynamic MRI after injection of the different NE formulations. Non-PEGylated NEs (#1) are rapidly cleared from the blood. NE-PEG3400-maleimide (#3) displays the best stealth properties with very low liver accumulation at 50 min. NEs decorated with PEG2000 (#2) or PEG2000/PEG3400–maleimide (#4) displayed a similar profile, with a relatively rapid clearance. Each NE half-life was estimated from the corresponding graph (dashed lines in Figure 1; Table 2). **Figure 1.** Estimation of the in vivo liver iron (Fe) uptake in mice by dynamic MRI after injection of the different NE formulations. Non-PEGylated NEs (#1) are rapidly cleared from the blood. NE-PEG3400-maleimide (#3) displays the best stealth properties with very low liver accumulation at 50 min. NEs decorated with PEG<sup>2000</sup> (#2) or PEG2000/PEG3400–maleimide (#4) displayed a similar profile, with a relatively rapid clearance. Each NE half-life was estimated from the corresponding graph (dashed lines in Figure 1; Table 2).

ferumoxytol, approved by the FDA for anemia treatment. The dose of 3 mg/kg of ferumoxytol seems to be well tolerated for MRI-based diagnostic imaging without serious adverse events, according to a recent multi-centric study that underlined the positive safety profile [41]. Liver uptake was monitored continuously by MRI for about 10 min before and up to

Depending on the lipid–PEG molecular weight, the formulations displayed different liver uptake patterns. After injection, the non-PEGylated formulation (#1) rapidly accu-**Table 2.** Half-life of the tested NE formulations in the bloodstream.


kidneys remained stable throughout the in vivo experiments, with all tested formulations. On the basis of these in vivo results, the NE–PEG3400–maleimide (#3) formulation was chosen, due to its having the lowest liver uptake, for conjugation with scFv-Fc-2Cys P3 HuAb to obtain NE-P3 (#5). The next experiments focused on the characterization of NE-P3 (#5). *2.3. Bio-Conjugation with P3 HuAb and In Vivo Clearance of NE-P3*  Conjugation of HuAb P3 to the NE–PEG3400–maleimide surface **(**NE-P3, (#5)) induced an increase in NE diameter, assessed by DLS (199.5 ± 2.3), as previously reported [20]. The formulation of NE-P3 (#5) was monodisperse (PdI = 0.167) and its size was still in the Depending on the lipid–PEG molecular weight, the formulations displayed different liver uptake patterns. After injection, the non-PEGylated formulation (#1) rapidly accumulated in the liver and the signal was saturated at 2 min after injection, confirming its less stealthy properties. Such rapid NE clearance is consistent with literature data showing that typically up to 90% of the formulation is taken up by the liver within 5 min [36]. NEs decorated with lipid–PEG<sup>2000</sup> (#2) or the mixture of two PEGs (#4) showed similar clearance profiles, with a rapid liver uptake after injection, although it was slower compared with (#1). NEs functionalized with lipid–PEG3400–maleimide (#3) showed the lowest liver uptake at 50 min post-injection, and therefore the best stealth profile. The MR signal in the kidneys remained stable throughout the in vivo experiments, with all tested formulations.

uptake at 50 min post-injection, and therefore the best stealth profile. The MR signal in the

submicronic range, with a diameter similar to that of commercial NEs used for human parenteral nutrition [36]. To determine whether the addition of antibody affected the formulation's stealthy properties, the in vivo clearance of the NE-P3 (#5) formulation was evaluated by dynamic MRI monitoring of liver uptake for 24 h after injection in the tail On the basis of these in vivo results, the NE–PEG3400–maleimide (#3) formulation was chosen, due to its having the lowest liver uptake, for conjugation with scFv-Fc-2Cys P3 HuAb to obtain NE-P3 (#5). The next experiments focused on the characterization of NE-P3 (#5).

#### *2.3. Bio-Conjugation with P3 HuAb and In Vivo Clearance of NE-P3*

Conjugation of HuAb P3 to the NE–PEG3400–maleimide surface (NE-P3, (#5)) induced an increase in NE diameter, assessed by DLS (199.5 ± 2.3), as previously reported [20]. The formulation of NE-P3 (#5) was monodisperse (PdI = 0.167) and its size was still in the submicronic range, with a diameter similar to that of commercial NEs used for human

parenteral nutrition [36]. To determine whether the addition of antibody affected the formulation's stealthy properties, the in vivo clearance of the NE-P3 (#5) formulation was evaluated by dynamic MRI monitoring of liver uptake for 24 h after injection in the tail vein of one *Apoe*−/<sup>−</sup> mouse (Figure 2). This experiment allowed calculation of the NE-P3 (#5) half-life (103 min) in blood (Table 2). *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 21 vein of one *Apoe*−/− mouse (Figure 2). This experiment allowed calculation of the NE-P3 (#5) half-life (103 min) in blood (Table 2).

**Figure 2.** NE-P3 (#5) half-life in the bloodstream was determined by dynamic MRI. (**a**): MR images at different time points showing the regions of interest (ROI, inside the red dashed lines) that encompass the liver and were used to measure the mean magnitude of the signal. (**b**): Mean normalized magnitude of the signal (red) and the corresponding iron concentration (black) estimated from the transverse relaxivity rଶ <sup>∗</sup> of NE- P3 (#5). Error bars come from the normalized magnitude's standard deviation of the voxels included in each ROI. The conjugation of biological moieties (scFv-Fc-2Cys P3 HuAb) to the PEGylated NE surface decreased the NE's half-life in the blood from 139 min to 103 min (calculated from the graph of the iron concentration, dashed line). **Figure 2.** NE-P3 (#5) half-life in the bloodstream was determined by dynamic MRI. (**a**): MR images at different time points showing the regions of interest (ROI, inside the red dashed lines) that encompass the liver and were used to measure the mean magnitude of the signal. (**b**): Mean normalized magnitude of the signal (red) and the corresponding iron concentration (black) estimated from the transverse relaxivity r ∗ 2 of NE- P3 (#5). Error bars come from the normalized magnitude's standard deviation of the voxels included in each ROI. The conjugation of biological moieties (scFv-Fc-2Cys P3 HuAb) to the PEGylated NE surface decreased the NE's half-life in the blood from 139 min to 103 min (calculated from the graph of the iron concentration, dashed line).

**Table 2.** Half-life of the tested NE formulations in the bloodstream.  **NE NE–PEG2000 NE–PEG3400–Maleimide NE–PEG2000/3400–Maleimide NE–PEG3400–Maleimide-P3**  After calculation of NE-P3's (#5) half-life in the bloodstream, its targeting efficiency was assessed in vitro, in vivo, and ex vivo (see below).

#### Blood half-life (min) ± SD (*n* = 3) *2.4. In Vitro Atheroma Targeting*

\* (*n* = 1) <1 7.2 ± 3.4 14.1 ± 2.5 138.7 ± 11.2 103 \* After calculation of NE-P3's (#5) half-life in the bloodstream, its targeting efficiency was assessed in vitro, in vivo, and ex vivo (see below). *2.4. In Vitro Atheroma Targeting*  First, NE-P3's (#5) ability to specifically target atheromatous plaques was tested by immunohistochemistry (IHC) using aorta tissue sections from hypercholesterolemic First, NE-P3's (#5) ability to specifically target atheromatous plaques was tested by immunohistochemistry (IHC) using aorta tissue sections from hypercholesterolemic *Apoe*−/<sup>−</sup> mice. A strong positive signal was observed only in aorta sections incubated with NE-P3 (#5) (Figure 3(Ab)), but not with NE–PEG3400–maleimide (#3) (without antibody) (Figure 3(Ad)). The uniform intense signal obtained with NE-P3 (#5) compared with the free P3 antibody, used at the same concentration, (compare Figure 3(Aa,Ab)) is certainly due to the ratio of 14 antibody molecules per PEGylated NE droplet used for the bio-conjugation that contributed to increase P3's binding avidity.

> *Apoe*-/- mice. A strong positive signal was observed only in aorta sections incubated with NE-P3 (#5) (Figure 3(Ab)), but not with NE–PEG3400–maleimide (#3) (without antibody) (Figure 3(Ad)). The uniform intense signal obtained with NE-P3 (#5) compared with the free P3 antibody, used at the same concentration, (compare Figure 3(Aa,Ab)) is certainly

**Formulation Number (#1) (#2) (#3) (#4) (#5)** 

galectin-3.

**Figure 3.** IHC analysis of aorta sections from hypercholesterolemic *Apoe<sup>−</sup>/<sup>−</sup>* mice (**A**) and human endarterectomy samples (**B**). NE-P3 (#5) can recognize galectin-3 in mouse aorta sections (**Ab**) and in human samples (**Bb**). No signal was observed with NE–PEG3400–maleimide (#3) (without antibody) (**Ad**,**Bd**) and in the negative control (secondary antibody only: **Ac**,**Bc**). Unconjugated scFv-Fc-2Cys P3 (**Aa**,**Ba**) was used as a positive control. Black arrows highlight specific binding by P3 or NE-P3 (#5). Scale bar: 250 µm (upper panels) and 100 µm (lower panels). Two tissue blocks were used for mouse aorta sections labeling in Panel A (one for a and c, and another for b and d). The same tissue block from the same patient was used in Panel B. The mouse aorta and endarterectomy sections shown are representative images taken from successive stained sections. Three independent experiments were performed with these formulations (a section per formulation). **Figure 3.** IHC analysis of aorta sections from hypercholesterolemic *Apoe*−/<sup>−</sup> mice (**A**) and human endarterectomy samples (**B**). NE-P3 (#5) can recognize galectin-3 in mouse aorta sections (**Ab**) and in human samples (**Bb**). No signal was observed with NE–PEG3400–maleimide (#3) (without antibody) (**Ad**,**Bd**) and in the negative control (secondary antibody only: **Ac**,**Bc**). Unconjugated scFv-Fc-2Cys P3 (**Aa**,**Ba**) was used as a positive control. Black arrows highlight specific binding by P3 or NE-P3 (#5). Scale bar: 250 µm (upper panels) and 100 µm (lower panels). Two tissue blocks were used for mouse aorta sections labeling in Panel A (one for a and c, and another for b and d). The same tissue block from the same patient was used in Panel B. The mouse aorta and endarterectomy sections shown are representative images taken from successive stained sections. Three independent experiments were performed with these formulations (a section per formulation).

*2.5. Preliminary Data on In Vivo Atheroma Targeting by NE-P3*  To determine whether NE-P3 (#5) can also target atheromatous plaques in vivo, this formulation was injected in the tail vein of one *Apoe*−/− mouse and atheroma targeting was monitored by MRI. The T2\* relaxation maps of atheromatous plaques were then computed from a multi-slice (RF)-spoiled gradient echo sequence (multi-echo GRE) obtained at 4.7 T at different time points: before (baseline), 7 h, and 24 h after NE-P3 (#5) injection. Aorta segmentation and T2\* mapping were performed concomitantly by two skilled experimenters. The slice-by-slice diagrams representing the T2\* mean values (in ms) at baseline, 7 h and 24 h post-injection calculated by the two experimenters (Figure 4A) showed a sliceby-slice decrease in the T2\* mean values, with differences between slices particularly at 7 h after injection. As the plaque characteristics (e.g., size, permeability, components) can Next, NE-P3 (#5) and PEG3400-maleimide (#3) were tested on human endarterectomy samples (Figure 3B). As before, a strong signal was observed only in atheromatous plaques incubated with NE-P3 (#5) (Figure 3(Bb)) but not with NE-PEG3400-maleimide (#3) (Figure 3(Bd)). The faint signal observed with NE–PEG3400–maleimide (#3) was mostly due to human immunoglobulins in atheromatous plaques. Moreover, the NE-P3 (#5) signal was stronger than that of the unconjugated P3 antibody (positive control) (Figure 3(Ba)), as observed in the mouse sections. These findings indicate that NE-P3 (#5) efficiently targets atheromatous lesions in mouse aortas (pre-clinical model) and also in human endarterectomy samples (for future translational assays). These results underline NE-P3's (#5) potential as a new molecular contrast agent to target atherosclerosis by binding to galectin-3.

Next, NE-P3 (#5) and PEG3400-maleimide (#3) were tested on human endarterectomy samples (Figure 3B). As before, a strong signal was observed only in atheromatous plaques incubated with NE-P3 (#5) (Figure 3(Bb)) but not with NE-PEG3400-maleimide (#3) (Figure 3(Bd)). The faint signal observed with NE–PEG3400–maleimide (#3) was mostly due to human immunoglobulins in atheromatous plaques. Moreover, the NE-P3 (#5) signal was stronger than that of the unconjugated P3 antibody (positive control) (Figure 3(Ba)), as observed in the mouse sections. These findings indicate that NE-P3 (#5) efficiently targets atheromatous lesions in mouse aortas (pre-clinical model) and also in human endarterectomy samples (for future translational assays). These results underline NE-P3's (#5) potential as a new molecular contrast agent to target atherosclerosis by binding to

#### change along the aorta, it is not surprising to observe differences in T2\* decrease from one slice to another. Figure 4B–D shows the segmented T2\* maps of the abdominal atheroma-*2.5. Preliminary Data on In Vivo Atheroma Targeting by NE-P3*

tous plaque for the same five adjacent slices before, at 7 h, and at 24 h after injection, respectively. For each slice, the T2\* map was overlaid on its corresponding magnitude image (i.e., the last panel, in the lower right corner, for each time point in Figure 4B–D). Comparisons of the data at the three time points (Figure 4B–D) highlighted the decrease in T2\* values at 7 h post-injection and their increase again (but still lower than at baseline) at 24 To determine whether NE-P3 (#5) can also target atheromatous plaques in vivo, this formulation was injected in the tail vein of one *Apoe*−/<sup>−</sup> mouse and atheroma targeting was monitored by MRI. The T2\* relaxation maps of atheromatous plaques were then computed from a multi-slice (RF)-spoiled gradient echo sequence (multi-echo GRE) obtained at 4.7 T at different time points: before (baseline), 7 h, and 24 h after NE-P3 (#5) injection. Aorta segmentation and T2\* mapping were performed concomitantly by two skilled experimenters. The slice-by-slice diagrams representing the T2\* mean values (in ms) at baseline, 7 h and 24 h post-injection calculated by the two experimenters (Figure 4A) showed a slice-by-slice decrease in the T2\* mean values, with differences between slices particularly at 7 h after injection. As the plaque characteristics (e.g., size, permeability, components) can change along the aorta, it is not surprising to observe differences in T2\* decrease from one slice to another. Figure 4B–D shows the segmented T2\* maps of the abdominal atheromatous plaque for the same five adjacent slices before, at 7 h, and at 24 h after injection, respectively. For each slice, the T2\* map was overlaid on its corresponding magnitude image (i.e., the last panel, in the lower right corner, for each time point in Figure 4B–D). Comparisons of the data at the three time points (Figure 4B–D) highlighted the decrease in T2\* values at 7 h post-injection and their increase again (but still lower than at baseline) at 24 h postinjection. These first results showed that NE-P3 (#5) could be used for in vivo targeting

of atheromatous plaques and indicated that the in vivo behavior of NE-P3 (#5) must be accurately monitored over time to determine the optimal imaging window. h post-injection. These first results showed that NE-P3 (#5) could be used for in vivo targeting of atheromatous plaques and indicated that the in vivo behavior of NE-P3 (#5) must be accurately monitored over time to determine the optimal imaging window.

**Figure 4.** In vivo atheroma T2\* variations after injection of NE-P3 (#5). (**A**): Diagrams made by two experimenters representing the T2\* mean values (in ms) for each slice of the abdominal aorta of one *Apoe<sup>−</sup>/<sup>−</sup>* mouse at baseline, 7 h, and 24 h after NE-P3 (#5) injection. Typical segmented T2\* maps of five slices before (**B**), 7 h (**C**), and 24 h (**D**) after NE-P3 (#5) injection, shown with the color scale. Slices 4, 5, 6, 7, and 8 before injection correspond to Slices 3, 4, 5, 6, and 7 after injection. After injection, great care was taken to reposition the mouse in exactly the same way as before injection; however, sections before and after injection could be shifted by one or two slices. The corresponding magnitude image (first echo) of the third slice in each dataset (i.e., Slice 6 at baseline and Slice 5 post-injection) is shown in the rightmost lower panel at each time point. In this image, the atheromatous plaque is clearly visible in white (arrow). Slice-by-slice T2\* maps were obtained using a homemade MATLAB-based tool. **Figure 4.** In vivo atheroma T<sup>2</sup> \* variations after injection of NE-P3 (#5). (**A**): Diagrams made by two experimenters representing the T<sup>2</sup> \* mean values (in ms) for each slice of the abdominal aorta of one *Apoe*−/<sup>−</sup> mouse at baseline, 7 h, and 24 h after NE-P3 (#5) injection. Typical segmented T<sup>2</sup> \* maps of five slices before (**B**), 7 h (**C**), and 24 h (**D**) after NE-P3 (#5) injection, shown with the color scale. Slices 4, 5, 6, 7, and 8 before injection correspond to Slices 3, 4, 5, 6, and 7 after injection. After injection, great care was taken to reposition the mouse in exactly the same way as before injection; however, sections before and after injection could be shifted by one or two slices. The corresponding magnitude image (first echo) of the third slice in each dataset (i.e., Slice 6 at baseline and Slice 5 post-injection) is shown in the rightmost lower panel at each time point. In this image, the atheromatous plaque is clearly visible in white (arrow). Slice-by-slice T<sup>2</sup> \* maps were obtained using a homemade MATLAB-based tool.

#### *2.6. Ex Vivo Analyses of Isolated Aorta Samples*  Ex vivo studies were then carried out after in vivo MRI imaging to confirm the pres-*2.6. Ex Vivo Analyses of Isolated Aorta Samples*

ence of the targeted contrast agent in the aorta using two methods: MRI (to visualize the areas with a decreased T2\* signal) and Electron Spin Resonance (ESR, to quantify the iron present in the aorta). Ex vivo studies were then carried out after in vivo MRI imaging to confirm the presence of the targeted contrast agent in the aorta using two methods: MRI (to visualize the areas with a decreased T2\* signal) and Electron Spin Resonance (ESR, to quantify the iron present in the aorta).

#### 2.6.1. Ex Vivo MRI Ex vivo MRI scans of the aorta embedded in agarose gel after isolation from one 2.6.1. Ex Vivo MRI

*Apoe−/−* mouse following in vivo injection of NE-P3 (#5) in the tail vein highlighted the intimal thickening characteristic of an atherosclerotic plaque. A drop in the MR signal due to the T2\* effect characterized the accumulation of the iron oxide-based targeted contrast agent in the atherosclerotic lesions (black arrows in Figure 5). Ex vivo MRI scans of the aorta embedded in agarose gel after isolation from one *Apoe*−/<sup>−</sup> mouse following in vivo injection of NE-P3 (#5) in the tail vein highlighted the intimal thickening characteristic of an atherosclerotic plaque. A drop in the MR signal due to the T2\* effect characterized the accumulation of the iron oxide-based targeted contrast agent in the atherosclerotic lesions (black arrows in Figure 5).

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 8 of 21

**Figure 5.** MRI magnitude image of the agarose-embedded aorta after isolation from one *Apoe<sup>−</sup>/<sup>−</sup>* mouse following in vivo injection of NE-P3 (#5) in the tail vein. Black arrows highlight the decrease in the MR signal. **Figure 5.** MRI magnitude image of the agarose-embedded aorta after isolation from one *Apoe*−/<sup>−</sup> mouse following in vivo injection of NE-P3 (#5) in the tail vein. Black arrows highlight the decrease in the MR signal. **Figure 5.** MRI magnitude image of the agarose-embedded aorta after isolation from one *Apoe<sup>−</sup>/<sup>−</sup>* mouse following in vivo injection of NE-P3 (#5) in the tail vein. Black arrows highlight the decrease in the MR signal.

#### 2.6.2. Iron Oxide Accumulation Assessment by ESR 2.6.2. Iron Oxide Accumulation Assessment by ESR 2.6.2. Iron Oxide Accumulation Assessment by ESR

Iron quantification by ESR in the same isolated aorta corroborated the detection of iron accumulation in the atheromatous plaque due to NE-P3 (#5) targeting. After injection, the broad and intense ESR signal detected at about 320 mT (g = 2.12) with a peak-to-peak line width of about 80 mT (Figure 6) was due to the presence of Fe2O3 nanoparticles at an estimated concentration of 1.7 ± 0.4 × 10−8 mol/g in the analyzed tissue. Iron quantification by ESR in the same isolated aorta corroborated the detection of iron accumulation in the atheromatous plaque due to NE-P3 (#5) targeting. After injection, the broad and intense ESR signal detected at about 320 mT (g = 2.12) with a peak-to-peak line width of about 80 mT (Figure 6) was due to the presence of Fe2O<sup>3</sup> nanoparticles at an estimated concentration of 1.7 <sup>±</sup> 0.4 <sup>×</sup> <sup>10</sup>−<sup>8</sup> mol/g in the analyzed tissue. Iron quantification by ESR in the same isolated aorta corroborated the detection of iron accumulation in the atheromatous plaque due to NE-P3 (#5) targeting. After injection, the broad and intense ESR signal detected at about 320 mT (g = 2.12) with a peak-to-peak line width of about 80 mT (Figure 6) was due to the presence of Fe2O3 nanoparticles at an estimated concentration of 1.7 ± 0.4 × 10−8 mol/g in the analyzed tissue.

**Figure 6.** Room temperature X-band (9.54 GHz) Electron Spin Resonance (ESR) spectra of (**a**) an aorta isolated from a *Apoe<sup>−</sup>/<sup>−</sup>* mouse untreated (control) and (**b**) from a *Apoe<sup>−</sup>/<sup>−</sup>* mouse after NE-P3 (#5) injection. **Figure 6.** Room temperature X-band (9.54 GHz) Electron Spin Resonance (ESR) spectra of (**a**) an aorta isolated from a *Apoe<sup>−</sup>/<sup>−</sup>* mouse untreated (control) and (**b**) from a *Apoe<sup>−</sup>/<sup>−</sup>* mouse after NE-P3 (#5) injection. **Figure 6.** Room temperature X-band (9.54 GHz) Electron Spin Resonance (ESR) spectra of (**a**) an aorta isolated from a *Apoe*−/<sup>−</sup> mouse untreated (control) and (**b**) from a *Apoe*−/<sup>−</sup> mouse after NE-P3 (#5) injection.

#### *2.7. Antioxidant Alpha-Tocopherol-Containing NEs for a Theranostic Approach: Formulation, Characterization, and Activity 2.7. Antioxidant Alpha-Tocopherol-Containing NEs for a Theranostic Approach: Formulation, Characterization, and Activity 2.7. Antioxidant Alpha-Tocopherol-Containing NEs for a Theranostic Approach: Formulation, Characterization, and Activity*

For this theranostic approach, NE samples decorated with PEG3400–maleimide (4.4 µmol∙mL−1) were combined with alpha-tocopherol and SPIO (12.5 µmol∙mL−1) in the oily phase (NE–PEG3400–maleimide–SPIO–tocopherol, (#10)) (Table 3). To determine the antioxidant or prooxidant contribution of each component of this formulation, four other formulations were assessed as controls (Table 3). All formulations were in the submicronic size range and monodisperse (PdI < 0.3) (Table 4). For this theranostic approach, NE samples decorated with PEG3400–maleimide (4.4 µmol∙mL−1) were combined with alpha-tocopherol and SPIO (12.5 µmol∙mL−1) in the oily phase (NE–PEG3400–maleimide–SPIO–tocopherol, (#10)) (Table 3). To determine the antioxidant or prooxidant contribution of each component of this formulation, four other formulations were assessed as controls (Table 3). All formulations were in the submicronic size range and monodisperse (PdI < 0.3) (Table 4). For this theranostic approach, NE samples decorated with PEG3400–maleimide (4.4 <sup>µ</sup>mol·mL−<sup>1</sup> ) were combined with alpha-tocopherol and SPIO (12.5 <sup>µ</sup>mol·mL−<sup>1</sup> ) in the oily phase (NE–PEG3400–maleimide–SPIO–tocopherol, (#10)) (Table 3). To determine the antioxidant or prooxidant contribution of each component of this formulation, four other formulations were assessed as controls (Table 3). All formulations were in the submicronic size range and monodisperse (PdI < 0.3) (Table 4).

Lipid–PEG3400–


**Table 3.** NE composition for the antioxidant assessment.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 9 of 21

SPIO: superparamagnetic iron oxide nanoparticles; qs: quantum satis. SPIO (µmol/mL) - - 12.5 - 12.5

**Table 4.** NE characterization. SPIO: superparamagnetic iron oxide nanoparticles; qs: quantum satis

maleimide (µmol/mL) - - 4.4 4.4 4.4


First, the antioxidant properties of these NE formulations were assessed by comparing their ability to increase the time to produce the hemolysis of 50% of red blood cells (T50% hemolysis, in minutes) after the initiation of free radical attack (Figure 7). First, the antioxidant properties of these NE formulations were assessed by comparing their ability to increase the time to produce the hemolysis of 50% of red blood cells (T50% hemolysis, in minutes) after the initiation of free radical attack (Figure 7).

**Figure 7.** "Kit Radicaux Libres" test (KRL test) to determine the antioxidant properties of the tested NE formulations (Table 3). The graph shows the change (in percent) of the time in which 50% of red blood cells were lysed (T50% hemolysis in minutes) after the initiation of free radical attack relative to the control sample (*n* = 3). **Figure 7.** "Kit Radicaux Libres" test (KRL test) to determine the antioxidant properties of the tested NE formulations (Table 3). The graph shows the change (in percent) of the time in which 50% of red blood cells were lysed (T50% hemolysis in minutes) after the initiation of free radical attack relative to the control sample (*n* = 3).

The results indicated that NEs without alpha-tocopherol (#6) and NE–PEG3400–maleimide–SPIO (#8)) had a slight prooxidant effect. When used at the concentration of 1000 mg/L, they reduced the T50% hemolysis by 10.95% and 10.66%, respectively, compared with the reference sample. Conversely, alpha-tocopherol-containing formulations ((#7), (#9), and (#10)) at the concentration of 1000 mg/L) significantly and similarly increased the The results indicated that NEs without alpha-tocopherol (#6) and NE–PEG3400– maleimide–SPIO (#8)) had a slight prooxidant effect. When used at the concentration of 1000 mg/L, they reduced the T50% hemolysis by 10.95% and 10.66%, respectively, compared with the reference sample. Conversely, alpha-tocopherol-containing formulations ((#7), (#9), and (#10)) at the concentration of 1000 mg/L) significantly and similarly

increased the T50% hemolysis by up to 104.63%. However, the presence of iron oxide nanoparticles (#10) slightly hindered alpha-tocopherol's antioxidant activity. Furthermore, when the results were expressed as Trolox equivalents (Figure 8), which is a water-soluble analog of Vitamin E commonly used in biological and biochemical applications as an antioxidant reference, the Trolox equivalent of 1 g of alpha-tocopherol-containing NEs ranged from 39.88 mg to 44.62 mg. These values were close to the amount of alpha-tocopherol loaded in the NEs (50 mg), thus proving that the formulation does not impair or hinder alpha-tocopherol's antioxidant activity. slightly hindered alpha-tocopherol's antioxidant activity. Furthermore, when the results were expressed as Trolox equivalents (Figure 8), which is a water-soluble analog of Vitamin E commonly used in biological and biochemical applications as an antioxidant reference, the Trolox equivalent of 1 g of alpha-tocopherol-containing NEs ranged from 39.88 mg to 44.62 mg. These values were close to the amount of alpha-tocopherol loaded in the NEs (50 mg), thus proving that the formulation does not impair or hinder alpha-tocopherol's antioxidant activity.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 10 of 21

**Figure 8.** Antioxidant properties of the tested NE formulations expressed as mg of Trolox equivalent per gram of NE (*n* = 3). **Figure 8.** Antioxidant properties of the tested NE formulations expressed as mg of Trolox equivalent per gram of NE (*n* = 3).

#### **3. Discussion 3. Discussion**

As cardiovascular complications caused by atherosclerosis are the leading cause of death in Western countries, it is crucial to develop tools for the early detection of vulnerable atheromatous plaques. In this context, the combination of nanotechnology and molecular imaging is a promising non-invasive strategy for detecting unstable plaques. Providing multi-territorial imaging of the atherosclerotic disease burden is now seen as necessary for a comprehensive patient assessment. Indeed, the mechanism of "vulnerable" plaque rupture is clearly more complex than initially assumed, and focusing only on the treatment of a single atherosclerotic plaque may not necessarily lead to a survival advantage. Proof of this is a recent systematic review and meta-analysis showing that there is no survival benefit of revascularization among patients with stable ischemic heart disease [42]. Numerous clinical investigations have demonstrated that many plaques rupture without clinical symptoms [43]. Moreover, plaque morphology changes over a few months, highlighting the necessity of longitudinal imaging studies. Our strategy is in keeping with new concepts indicating that risk would be more strongly predicted by detecting the total atheromatous burden of the arterial tree in the whole body [44]. The development of non-invasive molecular MRI imaging modality using contrast agents functionalized with antibodies capable of detecting high-risk plaques would allow longitudinal studies and assess the dynamic nature of atherosclerotic disease for a comprehensive As cardiovascular complications caused by atherosclerosis are the leading cause of death in Western countries, it is crucial to develop tools for the early detection of vulnerable atheromatous plaques. In this context, the combination of nanotechnology and molecular imaging is a promising non-invasive strategy for detecting unstable plaques. Providing multi-territorial imaging of the atherosclerotic disease burden is now seen as necessary for a comprehensive patient assessment. Indeed, the mechanism of "vulnerable" plaque rupture is clearly more complex than initially assumed, and focusing only on the treatment of a single atherosclerotic plaque may not necessarily lead to a survival advantage. Proof of this is a recent systematic review and meta-analysis showing that there is no survival benefit of revascularization among patients with stable ischemic heart disease [42]. Numerous clinical investigations have demonstrated that many plaques rupture without clinical symptoms [43]. Moreover, plaque morphology changes over a few months, highlighting the necessity of longitudinal imaging studies. Our strategy is in keeping with new concepts indicating that risk would be more strongly predicted by detecting the total atheromatous burden of the arterial tree in the whole body [44]. The development of non-invasive molecular MRI imaging modality using contrast agents functionalized with antibodies capable of detecting high-risk plaques would allow longitudinal studies and assess the dynamic nature of atherosclerotic disease for a comprehensive approach to the atheroma burden in the "vulnerable" patient.

approach to the atheroma burden in the "vulnerable" patient. Due to their high magnetic susceptibility and nanometric size, SPIO nanoparticles have been extensively investigated as MRI contrast agents [5]. However, they must be loaded into a nanocarrier to improve their biocompatibility and half-life. In this work, SPIO nanoparticles made hydrophobic by an oleic acid coating were loaded inside the oily core of PEGylated NEs functionalized with the first anti-galectin-3 human antibody for Due to their high magnetic susceptibility and nanometric size, SPIO nanoparticles have been extensively investigated as MRI contrast agents [5]. However, they must be loaded into a nanocarrier to improve their biocompatibility and half-life. In this work, SPIO nanoparticles made hydrophobic by an oleic acid coating were loaded inside the oily core of PEGylated NEs functionalized with the first anti-galectin-3 human antibody for specific atheroma targeting. When classical iron oxide-based contrast agents are used,

specific atheroma targeting. When classical iron oxide-based contrast agents are used, the

the MR signal changes observed during their accumulation in tissues are weighted both by the R ∗ 2 and R<sup>1</sup> effects. In the environment of oily droplets, SPIO nanoparticles do not have any access to water and, therefore, their tissue accumulation mainly affects the local R ∗ 2 relaxation rate (and not the R1). We took advantage of the "pure" R ∗ 2 -increasing effect obtained in oily droplets to estimate the blood half-life of each tested NE formulation.

NEs have been in medical use for more than five decades as a parenteral nutrition system for patients who cannot be fed orally (e.g., Intralipid, approved in Europe in 1962) [36]. The huge potential of NEs as drug delivery systems is currently unexploited despite the advantages compared with other nanocarriers [45,46]. To target atheroma with NEs, their half-life must be increased. This is commonly achieved by decorating the droplet surface with PEG chains (i.e., PEGylation) to delay opsonization. PEG has the advantages of being FDA-approved, soluble in hydrophilic and hydrophobic phases, non-toxic, and non-immunogenic [47]. PEG is also available in different molecular weights. For long-term circulation, PEG chains with a molecular weight of at least 2000 must be used [45,48–50].

To our knowledge, only a few studies have investigated how to prolong the halflife of PEGylated NEs [30,51,52]. Cheng et al. studied how different molecular weights and different concentrations of lipid–PEG affected NE size but did not analyze their pharmacokinetic profiles [52]. Hak et al. studied the influence of lipid–PEG<sup>2000</sup> density on NE pharmacokinetics by fluorescence analysis of animal serum samples [53]. In all these studies, NE pharmacokinetics were assessed by blood sampling, while our approach relies on the in vivo determination of NE uptake by the liver using MRI. In our study, the kinetics of four NE formulations were compared. The mean NE diameters tended to increase with the increase in PEG molecular weight, confirming PEG's brush-like conformation, whereas a mushroom-like conformation would result in a decrease in NE diameter [51–54]. Dynamic MRI is a powerful tool to rapidly follow, in a longitudinal manner, the stealth properties of nanocarriers over time, in contrast to classical methodologies that rely on blood sampling at defined time points.

The dynamic MRI approach showed that the half-life of NE–PEG3400–maleimide (#3) was significantly increased compared with the other formulations. Stealth properties are important for extending the half-life and consequently improving the targeting efficacy. To develop an optimized tool for molecular imaging of atherosclerosis, an atheromaspecific HuAb, P3, was chosen for conjugation to the NE–PEG formulation with the longest half-life (NE–PEG3400–maleimide (#3)). P3 specifically targets galectin-3, a protein that has been highlighted in recent studies as a new atherosclerosis biomarker [55]. P3 was discovered by in vivo phage-display selection in an animal model of atherosclerosis using a human scFv library [56,57]. Its ability to target galectin-3 within atherosclerotic lesions has been demonstrated in vitro and ex vivo (patent WO2019068863A1). P3 variable domains were engineered in the scFv-Fc format with a 2Cys tag for site-specific conjugation to NE–PEG3400–maleimide (#3) using thiol–maleimide "click" chemistry. Site-specific conjugation has many advantages; particularly, it maintains bioreactivity and can be achieved on the Fc fragment. A theoretical ratio of 14 antibodies per droplet was chosen for efficient molecular targeting, without denaturing the system. Indeed, previous work on SPIO nanoparticles showed that too high a ratio causes object flocculation. Compared with other synthetic nanocarriers, such as SPIO and ultrasmall paramagnetic iron oxide (USPIO) nanoparticles, which are rigid objects, oily droplets have a soft consistency and can efficiently cross the vascular fenestrae of the impaired endothelium and penetrate into the targeted lesions, even if their size is >25 nm. Other groups have demonstrated that lipidbased formulations up to 200 nm in diameter can enter a plaque [58–60]. The preliminary results of the assay testing whether in vivo NE–PEG3400–maleimide-P3 (#5) can target the atheromatous plaques located in the abdominal aorta of one *Apoe*−/<sup>−</sup> atherosclerotic mouse are sufficiently encouraging to justify additional in vivo MRI studies to determine the best imaging window. To this end, the contrast agent's accumulation must be accurately monitored at different time points: directly after injection and up to 48 h post-injection.

In this work, IHC experiments showed that the P3 HuAb can recognize galectin-3 in the *Apoe*−/<sup>−</sup> mouse aorta and also in human atheroma samples. It should be noted that the ligands used in the more recent studies for functionalizing nanoparticles, micelles, or liposomes tend to be non-immunogenic molecules that cross-react with different species, including humans [61–63]. This could facilitate the translation from pre-clinical to clinical studies. Additionally, if the developed diagnostic agents reach the clinics, the use of human antibodies will limit the risk of immunogenicity, thus saving the time required for the humanization of murine antibodies.

Finally, the use of such NEs as a potential theranostic tool was demonstrated using alpha-tocopherol as an antioxidant agent that could reduce the proliferation of radical species inside the atheromatous plaque, thus potentially decreasing the risk of rupture. The "Kit Radicaux Libre" test (KRL test) clearly showed that alpha-tocopherol-containing NEs display antioxidant properties (not observed with NEs without alpha-tocopherol), and also that the full potential of the encapsulated antioxidant agent is available to counteract a free radical attack. Indeed, the antioxidant activity expressed in Trolox equivalents was close to the amount of antioxidant agent loaded inside the NEs, demonstrating that all the alpha-tocopherol contained in the formulation was used in the test. However, it is worth noting that adding SPIO nanoparticles may have a slight prooxidant effect. Moreover, a limitation of this study is the lack of a toxicity assessment. If we want to further develop this theranostic approach using NE formulations that include SPIO nanoparticles, this point should be addressed in priority, although, here, NE-P3 was injected at 3 mg/kg, which was lower than the dose of ferumoxytol, an intravenous iron preparation, used in clinics to treat iron deficiencies. Overall, these results encourage us to develop a complex multi-modal theranostic approach for the diagnosis and treatment of atherosclerosis that could be tested longitudinally in small animal models of this disease in pre-clinical studies. The full potential of the theranostic approach would need to be tested in future in vivo studies using the NE–PEG3400–maleimide formulation loaded with SPIO and tocopherol, and functionalized with the scFv-Fc P3 human antibody.

More generally, this work shows the development of a nanomedicine platform that could be advantageously used with stronger reductant molecules chemically derived from tocopherol [33] or other payloads such as prostacycline for its anti-aggregant properties, which are of high value in atherothrombosis [64]. It could also be used with alternative HuAbs. An anti-platelet HuAb under the same format including cysteines for site-specific functionalization has been successfully grafted on the same platform [20]. The SPIO-loaded NE–PEG3400–maleimide formulation could thus be adapted to other pathologies, just by changing the active principle and the targeting HuAb.

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

#### *4.1. Materials*

Purified Miglyol 840 (oil phase) was kindly provided by IOI OLEO GmbH (Hamburg, Germany). Egg lecithin containing 82.3% phosphatidylcholine (Lipoid E80) and N-(carbonylmethoxypolyethylenglycol-2000)-1,2-distearoyl-sn-glycero-3-phosphoethanolamine (DSPE– PEG2000) were provided by Lipoid GmbH (Ludwigshafen, Germany); polysorbate 80 (Tween 80) was purchased from SEPPIC (Paris, France). The hetero-bifunctional linker 1,2-distearoylsn-glycero-3-phosphoethanolamine-N-poly(ethylene glycol)-maleimide (DSPE–PEG3400– maleimide) was purchased from Laysan Bio, Inc. (Arab, AL 35016, USA). SPIO nanoparticles were synthesized and made hydrophobic by following previously described procedures [65–67]. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP; ≥98%), hydrochloric acid, formic acid, the PBS buffer, sodium hydroxide, and IgG were bought from Sigma-Aldrich (St. Louis, MO 63178, USA). The PBS–heparin solution was from Sanofi Aventis (Vitrysur-Seine, France). Glycerol was purchased from Cooperation Pharmaceutique Française (Melun, France). MACS Cell Separation Columns were from Miltenyi Biotec (Bergisch Gladbach, Germany).

#### *4.2. NE Formulations*

NEs were formulated as previously described [20]. Briefly, 12 mg of Lipoid E80 was dispersed in 200 mg of Miglyol 840 (IOI Oleo GmbH, Hamburg, Germany) by heating, followed by addition of 12.5 µmol/mL SPIO nanoparticles. The aqueous phase was a dispersion of 25 mg Tween 80 in 800 mg Milli-Q water. For NE PEGylation, lipid–PEG at different molecular weights (DSPE–PEG<sup>2000</sup> and/or DSPE–PEG3400-maleimide) was added at 5 µmol/mL. A molar ratio of 3:1 was used for the PEG2000/PEG3400–maleimide mixture. The composition of the different NE formulations is described in Table 5. NE-P3 was obtained after bio-conjugation with HuAb P3, as described in Section 4.4.



Alpha-tocopherol-containing NEs were obtained as before, except that the oily phase included 50 mg of alpha-tocopherol and 150 mg of Miglyol 840. The compositions of the different theranostic NE formulations are described in Table 3.

All mixtures were emulsified by phase inversion and homogenized by sonication (Sonic Vibra Cell–VC 250 set at 70% and output 7; Sonics & Materials Inc, Newtown, CT 06470, USA) to obtain oil droplets in the submicron size range. Before the in vitro and in vivo experiments, the pH of the different formulations was adjusted to the physiological value using 0.1 N sodium hydroxide, and 2% glycerol was added to adjust osmolality.

#### *4.3. NE Characterization*

The physical characteristics of the NE formulations were assessed by DLS, Zeta potential measurement, and TEM. The hydrodynamic size was determined using a DLS device (Zetasizer Nano ZS; Malvern Instruments, Malvern, UK) and with the NEs diluted to 1:1000 (*v*/*v*) (mean of 3 independent measurements performed at 25 ◦C). The Zeta potential was measured using a Zetasizer Nano ZS device coupled to a Folded Capillary Cell (DTS1060) from Malvern Instruments. Oily droplets were quantified by NTA using a NanoSight NS300 instrument (Malvern Instruments). A Hitachi H7650 transmission electron microscope linked to an ORIUS SC1000 11MPX (Gatan Inc., Pleasanton, CA, USA) camera run by Digital Micrograph (Gatan Inc.) was used to study the NE samples (1:50 dilution, *v*/*v*) transferred to a carbon-coated copper grid. Iron concentration was quantified by UV spectrometry as previously described [66].

#### *4.4. P3 Antibody Bio-Conjugation to NEs*

To achieve optimal targeting of atheromatous plaques, the HuAb P3 was engineered in the single chain fragment variable (scFv)-Fc format with a 2-cysteine tag (ScFv-Fc-2Cys) for site-specific conjugation to the NE's surface. The production of antibodies with this cysteine tag has been described previously for the TEG4 antibody [31]. Before bio-conjugation to NE–PEG3400–maleimide (#3), the thiol groups on the cysteine tag of ScFv-Fc-2Cys P3 were activated using TCEP (20 mol per mol of P3). Antibody conjugation to the thiol-reactive maleimide of NE–PEG3400–maleimide (#3) to obtain NE-P3 (#5) was performed overnight with a theoretical ratio of 14 antibody molecules per NE droplet. Unconjugated antibodies were removed using a magnetic sorting column (MACS Cell Separation Columns), as previously described [31].

#### *4.5. In Vitro Immunoreactivity Analysis by IHC*

NE-P3 immunoreactivity was assessed by IHC using paraffin-embedded tissue sections of mouse and human atheromatous plaques. Human biopsies were provided by the vascular and general surgery service of the Pellegrin academic hospital, Bordeaux, France. Samples were from patients who underwent endarterectomy after an acute vascular event. All clinical interventions were carried out at Pellegrin hospital and the use of human samples for research was approved by the Bordeaux CPP ethics committee (Comité de Protection des Personnes Sud-Ouest et Outre Mer) and by the French Research Ministry (Authorization number DC-2016-2724). The CPP committee waived the need for the patients' written consent because surgical waste no longer attached to the person is considered res nullius. Nevertheless, the patients were informed by the clinicians; if they did not express their opposition to research, de-identified samples were immediately processed and embedded in paraffin. IHC experiments were performed as previously described [31]. Briefly, after deparaffinization, heat-induced epitope retrieval and non-specific interaction blocking, as described in [56], aorta sections were incubated with NE-P3 ((#5), 53 µg/mL of P3 antibody, 158 mg/L of Fe), unconjugated NE–PEG3400–maleimide ((#3), 158 mg/L of Fe), ScFv-Fc-2Cys P3 HuAb (53 µg/L) (positive control), or the diluent alone (negative control) overnight. This was followed by incubation with the secondary HRP-conjugated goat anti-human antibody (Fcγ-specific; 1:1000 (*v*/*v*) (Jackson ImmunoResearch; West Grove, PA, USA), and antibody binding was revealed using the Dako Liquid DAB+ Substrate Chromogen System (Agilent Technologies Inc, Santa Clara, CA, USA).

#### *4.6. In Vivo Experimental Animal Model*

Six-week-old female C57BL/6 mice (weighing 17 to 18 g) and 8- to 12-month-old JAX *Apoe*−/<sup>−</sup> mice (28 to 32 g) were purchased from Charles River Laboratories (Saint Germain Nuelles, France) and housed under a 12 h light/dark cycle with food and water provided ad libitum. *Apoe*−/<sup>−</sup> mice were fed a high-cholesterol diet (0.15% cholesterol) for 21 weeks to allow the development of atherosclerotic lesions. Experimental animals were cared for in accordance with institutional guidelines, and they were acclimatized for at least 7 days before the initiation of any experiment. All preclinical experiments described in this publication were approved by the Animal Care and Use Committee of Bordeaux, France (N◦50120192-A).

#### *4.7. Magnetic Resonance Imaging*

All MR experiments and relaxometry measurements were performed at 37 ◦C using a 4.7-Tesla Bruker Biospec System (Ettlingen, Germany) equipped with a gradient system with a maximum strength of 660 mT/m and a 110 µs rise time.

#### 4.7.1. In Vivo Assessment of Stealthy Features by Dynamic MRI

Dynamic MRI was performed using a radiofrequency (RF)-spoiled echo gradient sequence (FLASH) with the following parameters: flip angle = 30◦ ; echo time = 3.4 ms; TR ≈ 30 ms; resolution = 0.16 × 0.16; number of slices = 3; FOV = 40 × 30 mm; slice thickness = 1 mm; Nex = 10; dynamic scan time ≈ 60 s.

Mice (12 C57BL/6 females; *n* = 3 per condition) were anesthetized by inhalation of isoflurane (1.5%). Each animal then underwent imaging (50 s for each dynamic image) for 8 to 10 min (depending on the variation of the respiration rate). Imaging was stopped and 1 NE formulation (#1 to #4) (see Table 5) was injected intravenously in the tail vein at 53.7 µmol (Fe) kg−<sup>1</sup> bodyweight. Straight after injection completion, MR image acquisition was restarted and the interval between the injection and the acquisition of the first dynamic image was measured. This interval varied from 1 to 3 min. Finally, the 2 dynamic sessions

were fused, considering the injection delay. The time between 2 MR dynamic images for each dynamic session was estimated as the total scan time (which depended on the respiration rate) divided by the number of dynamic images obtained.

For this study, only the signals of the liver and kidneys (renal cortex and renal pyramids) were quantified. To restrict the motion analysis to these organs, a segmentation procedure was performed before analysis. Briefly, on each slice acquired before injection, an initial region of interest (ROI) was defined and subsequently refined by manual correction. Dynamic motion fields were then estimated for each dynamic MR image using the RealTITracker method described by Zachiu et al. [68,69]. These motion (vector) fields were used to update the ROI position in an elastic manner to each dynamic image to follow the organ's motion throughout the MRI session. The mean MR signals of the liver and one kidney were then extracted from these dynamic ROIs and used to estimate iron accumulation in these organs, as described in the next section. An example of the results of this correction can be found in the Supplementary Materials (Video S1).

#### 4.7.2. Estimation of the NE Half-Life in the Blood

As the NEs used in this study displayed a longitudinal relaxivity (r1) of almost zero [20], the MR signal changes in the liver and kidney over time were estimated to be caused only by R ∗ 2 changes and not by R<sup>1</sup> changes. Therefore, the signal equation of the gradient echo sequence (1) was simplified to Equation (2):

$$\mathbf{S}(\text{TE}, \text{TR}, \alpha)\_t = \mathbf{M}\_0 \sin(\alpha) \mathbf{e}^{-\text{TE} \cdot \mathbf{R}\_2^\*} \times \frac{1 - \mathbf{e}^{-\text{TR} \cdot \mathbf{R}\_1}}{1 - \cos(\alpha) \mathbf{e}^{-\text{TR} \cdot \mathbf{R}\_1}} \tag{1}$$

$$\mathbf{S}\left(\mathbf{t}\right) = \mathbf{M}\_0 \sin(\alpha) \mathbf{e}^{-\text{TE}\cdot(\mathbf{R}\_2^\* + \mathbf{r}\_2^\* \mathbf{C}(\mathbf{t}))} \tag{2}$$

where α is the flip angle, M<sup>0</sup> is the MR signal when the echo time (TE) tends to zero, r ∗ 2 is the transversal relaxivity of the tested NE, and C(t) is the iron concentration originating from the NE sample.

To estimate iron accumulation, the MR signal over time (S(t)) was first normalized to the signal measured at the first dynamic acquisition, before NE injection. Because at t = 0, the iron concentration (C(t)) is equal to 0, Equation (2) can be rewritten as:

$$\mathbf{S}\_0 = \mathbf{M}\_0 \sin(\alpha) \mathbf{e}^{-\text{TE} \cdot \mathbf{R}\_{2(0)}^\*} \tag{3}$$

where R ∗ 2(0) is the original relaxation rate. Therefore, the normalized signal of each new dynamic image (Snorm) is equal to:

$$\mathbf{S}\_{\rm norm}(\mathbf{t}) = \frac{\mathbf{S}(\mathbf{t})}{\mathbf{S}(\mathbf{0})} = \mathbf{e}^{-\rm TE \cdot r\_2^\* \mathbf{C}(\mathbf{t})} \tag{4}$$

and the iron concentration can be estimated over time as:

$$\mathbf{C}(\mathbf{t}) = \frac{-\ln(\mathbf{S}\_{\text{norm}}(\mathbf{t}))}{\text{TE} \cdot \mathbf{r}\_2^\*} \tag{5}$$

In vivo, most injected nanoparticles (and also NEs) are cleared from the bloodstream by different organs (liver, spleen, and bone marrow). In this study, the liver was considered to be the main organ of NE clearance and, therefore, iron accumulation in this organ directly reflected blood clearance of NEs. To estimate the apparent NE half-life in the blood, the iron concentration was assumed to change over time as:

$$\frac{d\mathbf{C}(\mathbf{t})}{d\mathbf{t}} = \boldsymbol{\pi} \left( \mathbf{C}\_{\text{max}} - \mathbf{C}(\mathbf{t}) \right) \tag{6}$$

where Cmax is the maximum iron concentration estimated from the time of NE injection with no surface functionalization and τ is the accumulation rate. Therefore, the clearance rate, τ, was estimated from a Marquardt–Levenberg fit of C(t) using a first-order differential equation (Equation (7)):

$$\mathbf{C}(\mathbf{t}) = \mathbf{C}\_{\text{max}} \left( \mathbf{1} - \mathbf{e}^{-\mathbf{t}\tau} \right) \tag{7}$$

Finally, the apparent NE half-life in the blood was defined as:

$$\mathbf{t}\_{1/2} = \ln(2)/\pi \tag{8}$$

4.7.3. In Vivo Dynamic MRI and T2\* Mapping of Atheromatous Plaques in *Apoe*−/<sup>−</sup> Mice

Tail vein injections of NE-P3 (#5) were performed at 53.7 <sup>µ</sup>mol(Fe)·kg−<sup>1</sup> bodyweight. In one *Apoe*−/<sup>−</sup> female, the NE half-life in the blood was estimated, based on the FLASH sequence obtained in the dynamic study. Atheromatous plaque labeling by NE-P3 (#5) was monitored in another *Apoe*−/<sup>−</sup> female by MRI before and at 7 h and 24 h after injection. Isoflurane concentrations were adjusted over time to maintain the respiration rate between 40 and 60 bpm. For in vivo T2\* mapping of atheromatous plaques, 10 transversal slices passing through the aorta in the thoraco-abdominal region were acquired using a multi-slice RF-spoiled echo gradient sequence: TR/TE1/∆TE = 1300/2.8/3.6 ms; α = 60◦ ; 15 echoes; NEx = 8; BW = 89.2 MHz; voxel size = 0.1 <sup>×</sup> 0.1 <sup>×</sup> 1 mm<sup>3</sup> ; FOV = 2.56 × 1.6 cm. Manual segmentation at the aorta level was performed to include only voxels that corresponded to the plaque using the first echo image. The mean T2\* values were computed for each slice. All calculations were performed using custom scripts written in MATLAB (MathWorks, Natick, MA, USA) and in C++ for the Levenberg–Marquardt algorithm. For each dataset, T2\* maps were estimated after correction of the macroscopic susceptibilities (e.g., arising from air/tissue interfaces, ∆B0z), as proposed by Dahnke et al. [70]. The limit of acceptability for the signal-to-noise ratio in the T2\* calculations was fixed to 4. The optical flow algorithm RealTITracker method described by Zachiu et al. [68,69] was used to avoid any image shifts during the echo times that could hamper the correct T2\* estimate. T2\* values of >30 ms were not considered to be representative of atheromatous plaques at 4.7 T. These few outliers were excluded from the analysis.

#### *4.8. Ex Vivo Analyses of Isolated Aorta Samples*

#### 4.8.1. MRI Analyses

Aorta samples embedded in agarose gel (1%) were analyzed ex vivo by MRI using multiple gradient echoes with positive readouts and number of echoes = 20; delta TE = 3 ms; max TE = 59.48 ms; TE/TR = 2.48/100 ms; acquisition bandwidth = 125 kHz; matrix = 256 × 128 × 128; FOV = 20 × 10 × 10; isotropic resolution = 78 µm; number of repetitions = 16; total acquisition time = 7 h 30 min.

#### 4.8.2. Iron Quantification by ESR

ESR experiments were performed at room temperature with a Bruker ESP300E spectrometer operating at X-band frequency (9.54 GHz). The microwave power was set to 20 mW and the magnetic field modulation frequency and amplitude to 100 kHz and 1 mT, respectively. The spectral resolution was 0.7 mT/pt and the acquisition time was 30 min for each sample. After weighing, each aorta sample was carefully digested in HNO<sup>3</sup> (65%) under a flame. Digestion was repeated 3 times to ensure the sample's complete mineralization. After the last addition of HNO3, 100 mg of NaNO<sup>3</sup> salt was added to obtain a homogeneous solid powder for the ESR analysis. The recorded ESR spectra were normalized to the aorta mass (expressed in g), and their intensity (obtained by double-integration of the first derivative absorption curve) was compared with that of reference samples with a known concentration of Fe2O<sup>3</sup> nanoparticles (ranging from 5 <sup>×</sup> <sup>10</sup>−<sup>10</sup> to 1 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mol/g).

#### *4.9. In Vitro Antioxidant Measurement*

The antioxidant properties of alpha-tocopherol-containing NEs were assessed by Kirial International/Laboratoires Spiral (Couternon, France) using the KRL test according to Caspar-Bauguil et al. [71]. The antioxidant or prooxidant activity was evaluated in whole

blood samples mixed with the different NE formulations and exposed to a controlled free radical attack. The overall resistance of the blood samples against the free radical attack was assessed by calculating the time to produce the hemolysis of 50% of red blood cells (T50% hemolysis, in minutes) after the initiation of the attack. The antioxidant/prooxidant effect of the different NE formulations was then expressed as the percentage of increase/decrease in the T50% hemolysis relative to the control sample (without NE formulation). The results were also expressed in Trolox equivalents (used as a reference).

#### **5. Conclusions**

HuAbs represent a class of ligands that are theoretically safer for clinical translation. The same is true for NE formulations due to their proven biocompatibility and their easy dissemination through biological barriers. The results of this study pave the way to future non-invasive targeted imaging of atherosclerosis (by combining safe magnetic nanoparticles as a contrast agent and a new anti-galectin-3 HuAb as a targeting agent) and synergistic therapeutic applications using active pharmaceutical ingredients (e.g., alpha-tocopherol, an antioxidant, anti-inflammatory, and cardioprotective vitamin) to induce the regression of the vulnerable plaque. Moreover, a more personalized approach to therapies against atherosclerosis could be guided by molecular imaging. Advances in the in vivo targeting efficiency of agents against proteins overexpressed in the plaque, such as galectin-3, could also improve the assessment and monitoring of atherosclerosis.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10.3 390/ijms22105188/s1. Figure S1: Transmission electron microscopy (TEM) picture without negative staining of the nanoemulsions loaded with magnetic particles. Video S1: Example of dynamic correction.

**Author Contributions:** Conceptualization, G.C.-S., S.C.-M., M.D.-C., and S.M.; methodology, S.B., G.P., Y.M., A.H., M.-J.J.-V., M.D., A.T., M.D.-C., and S.S.; software, S.B.; validation, S.B., G.P., and A.H.; formal analysis, S.B.; resources, G.C.-S., S.C.-M., M.D.-C., and S.M.; data curation, S.B.; writing original draft preparation, G.P., S.B., Y.M., M.-J.J.-V., M.D., and M.D.-C.; writing—review and editing, G.C.-S., S.C.-M., S.B., S.M., G.P., and M.-J.J.-V.; visualization, G.C.-S., S.B., G.P., and M.D.; supervision, G.C.-S., S.C.-M., and M.D.-C.; project administration, G.C.-S., S.C.-M., and S.M.; funding acquisition, G.C.-S. and S.C.-M. All authors have contributed substantially to the work reported and have read and agreed to the published version of the manuscript.

**Funding:** This study was achieved within the context of the Laboratory of Excellence TRAIL ANR-10-LABX-57 and LabEx MAbImprove: ANR-10-LABX-53. A public grant from the SVSE5 program, named ATHERANOS, supported this work. Geoffrey Prévot was a recipient of a PhD scholarship from the French Ministry of Education, Research and Technology. Samuel Bonnet was supported by TRAIL ANR-10-LABX-57.

**Institutional Review Board Statement:** Experimental animals were cared for in accordance with the institutional guidelines, and they were acclimatized for at least 7 days before the initiation of any experiment. All preclinical experiments described in this publication were approved by the Animal Care and Use Committee of Bordeaux, France (No. 50120192-A).

**Informed Consent Statement:** Human biopsies were provided by the vascular and general surgery service of the Pellegrin academic hospital, Bordeaux, France. Samples were from patients who underwent endarterectomy after an acute vascular event. All clinical interventions were carried out at Pellegrin hospital and the use of human samples for research was approved by the Bordeaux CPP ethics committee (Comité de Protection des Personnes Sud-Ouest et Outre Mer) and by the French Research Ministry (Authorization Number DC-2016-2724). The CPP committee waived the need for patient written consent because surgical waste no longer attached to the person is considered res nullius. Nevertheless, the patients were informed by the clinicians; if they did not express their opposition to research, de-identified samples were immediately processed and embedded in paraffin.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** We thank Cyril Lorenzato for his help in designing the dynamic MRI experiments.

**Conflicts of Interest:** The authors Marie-Josée Jacobin-Valat, Martine Duonor-Cérutti, Gisèle Clofent-Sanchez, and Audrey Hémadou are listed as inventors on the patent for the human anti-galectin-3 antibody (WO2019068863A1, https://patents.google.com/patent/WO2019068863A1/en, accessed on 18 March 2021). All authors have approved the final article.

#### **References**


## *Article* **Phage Display Screening of Bovine Antibodies to Foot-and-Mouth Disease Virus and Their Application in a Competitive ELISA for Serodiagnosis**

**Sukyo Jeong <sup>1</sup> , Hyun Joo Ahn <sup>1</sup> , Kyung Jin Min <sup>1</sup> , Jae Won Byun <sup>2</sup> , Hyun Mi Pyo <sup>2</sup> , Mi Young Park <sup>2</sup> , Bok Kyung Ku <sup>2</sup> , Jinju Nah <sup>2</sup> , Soyoon Ryoo <sup>2</sup> , Sung Hwan Wee <sup>2</sup> and Sang Jick Kim 1,\***


**Citation:** Jeong, S.; Ahn, H.J.; Min, K.J.; Byun, J.W.; Pyo, H.M.; Park, M.Y.; Ku, B.K.; Nah, J.; Ryoo, S.; Wee, S.H.; et al. Phage Display Screening of Bovine Antibodies to Foot-and-Mouth Disease Virus and Their Application in a Competitive ELISA for Serodiagnosis. *Int. J. Mol. Sci.* **2021**, *22*, 4328. https://doi.org/ 10.3390/ijms22094328

Academic Editors: Menotti Ruvo and Annamaria Sandomenico

Received: 10 March 2021 Accepted: 19 April 2021 Published: 21 April 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/).

**Abstract:** For serodiagnosis of foot-and-mouth disease virus (FMDV), monoclonal antibody (MAb) based competitive ELISA (cELISA) is commonly used since it allows simple and reproducible detection of antibody response to FMDV. However, the use of mouse-origin MAb as a detection reagent is questionable, as antibody responses to FMDV in mice may differ in epitope structure and preference from those in natural hosts such as cattle and pigs. To take advantage of natural host-derived antibodies, a phage-displayed scFv library was constructed from FMDV-immune cattle and subjected to two separate pannings against inactivated FMDV type O and A. Subsequent ELISA screening revealed high-affinity scFv antibodies specific to a serotype (O or A) as well as those with pan-serotype specificity. When BvO17, an scFv antibody specific to FMDV type O, was tested as a detection reagent in cELISA, it successfully detected FMDV type O antibodies for both serum samples from vaccinated cattle and virus-challenged pigs with even higher sensitivity than a mouse MAbbased commercial FMDV type O antibody detection kit. These results demonstrate the feasibility of using natural host-derived antibodies such as bovine scFv instead of mouse MAb in cELISA for serological detection of antibody response to FMDV in the susceptible animals.

**Keywords:** foot-and-mouth disease virus; type O; type A; phage display; antibody; competitive ELISA

#### **1. Introduction**

Foot-and-mouth disease virus (FMDV) causes significant foot-and-mouth disease for cloven-hoofed animals [1]. The FMDV is one of the contagious animal disease resulting in social and economic problem due to its rapid spread in countries across the world [2]. The genome of FMDV is a single-stranded RNA (approximately 8.5 kb) that is surrounded by a protein coat consisting of four capsid proteins, namely, VP1, VP2, VP3, and VP4 [3,4]. FMDV occurs in seven major immunologically distinct serotypes: A, O, C, Asia I, and South African Territories (SAT1, SAT2, and SAT3). Of the serotypes, FMDV type O and A have widely spread around the world. Rapid and accurate diagnosis is paramount to limit their distribution and eradiate the diseases.

One of the effective approaches for FMDV diagnosis is to serologically detect FMDVspecific antibodies, which are generated by host immune response against either the viral non-structural proteins (NSPs) or capsid structural proteins (SPs). Unlike NSP antibodies of which presence are used to differentiate infected from vaccinated animals, SP antibodies are detected in both infected and vaccinated animals. Hence, the SP antibody test has been used for FMD surveillance as well as for post-vaccination monitoring. Virus neutralization

test (VNT) is a gold standard for detecting SP antibodies, which measures the neutralizing activity of SP antibodies that block virus infection in susceptible cell lines [5]. However, the VNT suffers from such drawbacks as unavoidable use of a live virus, poor reproducibility, and difficulty in large-scale testing. As an alternative, competitive ELISA (cELISA)-based methods have been developed using serotype-specific monoclonal antibodies (MAbs) [6,7]. For example, a commercial kit for FMDV type O antibody includes type O-specific MAb as a key reagent and measures the percent of inhibition (PI) of the MAb binding to antigen by a serum tested. When the PI value is greater than 50%, the serum turns out to be FMDV type O antibody positive. Accordingly, kit performance such as sensitivity and specificity is very dependent on the choice of MAb. The concern about using MAb reagent in cELISA is how well the antibody recognizing a single epitope can represent overall antibody response against diverse epitopes on FMDV capsid. Hence the epitope of antibody reagent in cELISA should be immunodominant and overlap with those of the majority of antibodies produced by an anti-FMDV immune response in host animals.

Diverse sets of MAbs against FMDV have been generated by murine hybridoma technology [7–10]. Extensive studies on antigenic features of FMDV, especially type O, have identified major neutralizing epitopes recognized by the mouse MAbs [11–13]. Even though these epitope regions are suggested to be also recognized by bovine antibodies [14,15], the relative preference of each epitope region and fine epitope structure may be different in cattle. In fact, investigations of antibody response in natural host animals such as cattle, pigs, and sheep have revealed that antibody response does not dominantly direct to the G-H loop of VP1, the historically considered immunodominant epitope of FMDV type O in mice [14,16]. The mouse MAb currently used for cELISA targets the linear epitope on the G-H loop [6] and has such an epitope matching issue that improved performance can be expected from cELISA development using antibody reagent derived from a natural FMDV host animal such as cattle. In addition, the bovine antibody gene repertoire has distinct features when compared with that of a mouse. Mouse germline heavy chain variable (VH) gene segments are diverse in sequence and classified into 16 families, while bovine germline VH gene segments have similar sequences and belong to a single family [17]. Instead, the bovine antibody has extraordinary long heavy-chain complementarity-determining region 3 (HCDR3), which is believed to be the way to achieve high antibody diversity even using the limited source of VH frameworks [18].

Bovine anti-FMDV MAbs can be generated using hetero-hybridoma as well as phage display technology, as reported before [15,19]. Phage display is now considered a better option than hybridoma technology because it erases the instability issue of hybridoma cells, gives more flexibility in antibody selection strategy, and allows higher throughput screening. Recent accumulation of sequence information for bovine germline antibody genes enables the construction of a more reliable antibody library when compared with the first bovine Fab library, which was constructed using limited primer sets [20–22]. In this study, we constructed a phage-displayed bovine scFv library from peripheral blood lymphocytes (PBL) of FMDV-vaccinated cattle. Bovine scFv clones with specificity for type O or A as well as pan-serotype could be isolated by panning and screening from the library. Their binding activity and epitope were precisely characterized after the expression and purification of scFv-Fc proteins. Furthermore, the feasibility of their use in cELISA-based serodiagnosis of FMDV was tested using standard FMDV positive sera purchased from Pirbright and serum samples obtained from vaccinated cattle and virus-challenged pigs. Details are reported herein.

#### **2. Results**

#### *2.1. Selection of FMDV-Specific Bovine Antibodies*

A bovine scFv library was constructed from PBLs of FMDV-immunized cattle as described in Materials and Methods, and its diversity was estimated 2 <sup>×</sup> <sup>10</sup><sup>8</sup> cfu (colony forming unit). To select FMDV-specific antibodies from the library, a library phage was prepared and subjected to panning against both type O (O1 Manisa) and type A (A22 Iraq) of inactivated virus antigens.

After three rounds of panning, the successful enrichment of FMDV binding phage was confirmed by monitoring the phage recovery rate (output/input phage ratio). For both pannings against type O and A, a more than 100-fold increase in the rate was observed after the third round when compared with that after the first round. Most of the scFv phage clones selected randomly after the third round of panning showed FMDV-binding activity in phage ELISA (data not shown), and analysis of their scFv sequences revealed six unique HCDR3 sequences (Table 1). For each HCDR3, a representative scFv phage clone was chosen and tested for its binding activity to type O and A antigens. Three clones with comparable binding activity to both type O and A antigens were named BvOA1, BvOA7, and BvOA18. They could be found from both pannings against type O and A. Two clones (BvO17, BvO22) with type O-specific binding activity were derived from the panning against type O, whereas BvA3, the only clone with preferential binding activity to type A antigen, was selected from the panning against type A.

**Table 1.** Analysis of VH and VL gene usage in the selected scFv clones. The nucleotide sequences were analyzed using IMGT/V-QUEST (http://www.imgt.org/IMGT\_vquest/input, last accessed December 2, 2020).


#### *2.2. Characterization of Antibody Specificity*

For precise characterization of the selected clones, scFv-Fc expression construct was made for each clone by simple cut-and-paste cloning of scFv insert into the cassette mammalian expression vector pDR-OriP-Fc1 [23]. After transfection of the construct, transient expression in 293E cells resulted in an scFv-Fc molecule of 55 kDa secreted as dimers of 110 kDa, which was purified from the culture supernatant by affinity chromatography on a protein G column. The purity and integrity of the purified scFv-Fcs were confirmed by SDS-PAGE analysis, which showed more than 95% purity of scFv-Fcs (Figure S1).

The binding affinity of the purified scFv-Fcs for either type O or A antigen was measured by an indirect ELISA (Figure 1). As in Table 2, the apparent dissociation constant (KD) for each binding interaction was determined by binding curve fitting as described previously [24]. As expected, BvOA1, BvOA7, and BvOA18 bound to both O1 Manisa and A22 Iraq antigens in a concentration-dependent manner. Their affinities appeared to be very high for both strains of FMDV, with K<sup>D</sup> values ranging from 18.5 to 93.3 pM. BvO17 and BvO22 also showed very high affinity for O1 Manisa with K<sup>D</sup> values of about 20 pM but negligible binding activity for A22 Iraq. BvA3 showed preferential binding to A22 Iraq, and its affinity was relatively low when compared with those of the other clones.

**Table 2.** Binding affinity of selected scFv clones. Asterisks (\*) are not detected.


O1 Manisa.

**Table 2.** Binding affinity of selected scFv clones. Asterisks (\*) are not detected.

 **Binding Affinity (KD, pM)** 

 **O1 Manisa A22 Iraq**  BvO17 25.9 ± 1.9 \* N.D. BvO22 21.8 ± 1.4 \* N.D. BvOA1 93.3 ± 9.0 91.8േ9.0 BvOA7 27.9 ± 1.2 18.5േ1.3 BvOA18 46.1 ± 3.9 50.5േ5.3 BvA3 \* N.D. 2179.9േ322.0

Next, we further tested the cross-reactivity of scFv-Fcs for the other serotypes (Figure 2). As expected, BvOA1, BvOA7, and BvOA18 were reactive to all the tested serotypes of FMDV serotypes. They may recognize common epitopes of FMDV regardless of serotype and can be used for pan-serotype diagnosis of FMDV. As clearly shown in Figure 2, BvO17 and BvO22 specifically bound to O1 Manisa and were not cross-reactive with the other serotypes of FMDV strains tested. It is likely that they recognize type O-specific epitope and can be used for the detection of type O-specific antibody response. Unlike the type O-

**Figure 1.** Binding activities of selected bovine scFvs toward (**A**) FMDV type O (O1 Manisa) and (**B**) FMDV type A (A22 Iraq) antigens. scFv sequences selected by phage library screening of FMDVimmunized cattle were cloned into a pDR-OriP-Fc1 vector, which allowed scFv-Fc expression in mammalian cells. Serial dilutions of purified scFv-Fc were applied onto FMDV antigen-coated 96 well Maxisorp plate. Bound scFv-Fc proteins were detected using HRP-conjugated anti-human IgG (Fc-specific). Data are shown as mean ± SD of triplicate samples. **Figure 1.** Binding activities of selected bovine scFvs toward (**A**) FMDV type O (O1 Manisa) and (**B**) FMDV type A (A22 Iraq) antigens. scFv sequences selected by phage library screening of FMDV-immunized cattle were cloned into a pDR-OriP-Fc1 vector, which allowed scFv-Fc expression in mammalian cells. Serial dilutions of purified scFv-Fc were applied onto FMDV antigen-coated 96-well Maxisorp plate. Bound scFv-Fc proteins were detected using HRP-conjugated anti-human IgG (Fc-specific). Data are shown as mean ± SD of triplicate samples.

Next, we further tested the cross-reactivity of scFv-Fcs for the other serotypes (Figure 2). As expected, BvOA1, BvOA7, and BvOA18 were reactive to all the tested serotypes of FMDV serotypes. They may recognize common epitopes of FMDV regardless of serotype and can be used for pan-serotype diagnosis of FMDV. As clearly shown in Figure 2, BvO17 and BvO22 specifically bound to O1 Manisa and were not cross-reactive with the other serotypes of FMDV strains tested. It is likely that they recognize type O-specific epitope and can be used for the detection of type O-specific antibody response. Unlike the type O-specific clones, BvA3 was not perfectly specific to type A. Even though BvA3 bound to A22 Iraq preferentially, it seemed to be still reactive to other serotypes of strains, especially O1 Manisa. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 13

**Figure 2.** Cross-reactivities of bovine scFvs to seven serotypes of FMDV. Inactivated FMDV antigens for seven serotypes purchased from Pirbright Institute and BSA (negative control) were coated onto a 96-well Maxisorp plate. A total of 0.1 µg/mL of purified scFv-Fc was applied onto the wells for most clones except BvA3, of which 2 µg/mL was applied. Bound scFv-Fc proteins were detected using HRP-conjugated anti-human IgG (Fc-specific). Data are shown as mean ± SD of triplicate samples. **Figure 2.** Cross-reactivities of bovine scFvs to seven serotypes of FMDV. Inactivated FMDV antigens for seven serotypes purchased from Pirbright Institute and BSA (negative control) were coated onto a 96-well Maxisorp plate. A total of 0.1 µg/mL of purified scFv-Fc was applied onto the wells for most clones except BvA3, of which 2 µg/mL was applied. Bound scFv-Fc proteins were detected using HRP-conjugated anti-human IgG (Fc-specific). Data are shown as mean ± SD of triplicate samples.

#### *2.3. Characterization of the Antibody Binding Epitopes 2.3. Characterization of the Antibody Binding Epitopes*

To investigate whether the selected clones recognize overlapping epitopes or not, cross-competition binding experiments for both O1 Manisa and A22 Iraq antigens were carried out using scFv phage and scFv-Fcs. Figure 3 shows antibody epitope binning data, and the value in each cell represents relative absorbance for the binding of scFv phage shown on the upper side in the presence of competitor scFv-Fc shown on the left side. The To investigate whether the selected clones recognize overlapping epitopes or not, cross-competition binding experiments for both O1 Manisa and A22 Iraq antigens were carried out using scFv phage and scFv-Fcs. Figure 3 shows antibody epitope binning data, and the value in each cell represents relative absorbance for the binding of scFv phage shown on the upper side in the presence of competitor scFv-Fc shown on the left side. The

degree of binding inhibition was also expressed as the darkness of the cell background,

tibodies, BvOA1, BvOA7, and BvOA18, competed with each other for binding to both types of antigens suggesting their recognition of overlapping epitopes common to all serotypes. Type O-specific antibodies, BvO17 and BvO22, also competed with each other for binding to type O antigen, which indicated a shared epitope region on type O antigen recognized by both. Since BvA3 did not compete with the other antibodies, three different

**Figure 3.** Epitope binning of bovine scFvs by cross-competition ELISA. Inactivated FMDV type O (O1 Manisa) (**A**) or type A (A22 Iraq) (**B**) antigens were coated onto a 96-well Maxisorp plate. A

groups could be finally defined in terms of epitope specificity.

ples.

degree of binding inhibition was also expressed as the darkness of the cell background, which allowed easy grouping of antibodies according to their epitopes. Pan-serotype antibodies, BvOA1, BvOA7, and BvOA18, competed with each other for binding to both types of antigens suggesting their recognition of overlapping epitopes common to all serotypes. Type O-specific antibodies, BvO17 and BvO22, also competed with each other for binding to type O antigen, which indicated a shared epitope region on type O antigen recognized by both. Since BvA3 did not compete with the other antibodies, three different groups could be finally defined in terms of epitope specificity. shown on the upper side in the presence of competitor scFv-Fc shown on the left side. The degree of binding inhibition was also expressed as the darkness of the cell background, which allowed easy grouping of antibodies according to their epitopes. Pan-serotype antibodies, BvOA1, BvOA7, and BvOA18, competed with each other for binding to both types of antigens suggesting their recognition of overlapping epitopes common to all serotypes. Type O-specific antibodies, BvO17 and BvO22, also competed with each other for binding to type O antigen, which indicated a shared epitope region on type O antigen recognized by both. Since BvA3 did not compete with the other antibodies, three different groups could be finally defined in terms of epitope specificity.

**Figure 2.** Cross-reactivities of bovine scFvs to seven serotypes of FMDV. Inactivated FMDV antigens for seven serotypes purchased from Pirbright Institute and BSA (negative control) were coated onto a 96-well Maxisorp plate. A total of 0.1 µg/mL of purified scFv-Fc was applied onto the wells for most clones except BvA3, of which 2 µg/mL was applied. Bound scFv-Fc proteins were detected using HRP-conjugated anti-human IgG (Fc-specific). Data are shown as mean ± SD of triplicate sam-

To investigate whether the selected clones recognize overlapping epitopes or not, cross-competition binding experiments for both O1 Manisa and A22 Iraq antigens were carried out using scFv phage and scFv-Fcs. Figure 3 shows antibody epitope binning data, and the value in each cell represents relative absorbance for the binding of scFv phage

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 13

*2.3. Characterization of the Antibody Binding Epitopes* 

**Figure 3.** Epitope binning of bovine scFvs by cross-competition ELISA. Inactivated FMDV type O (O1 Manisa) (**A**) or type A (A22 Iraq) (**B**) antigens were coated onto a 96-well Maxisorp plate. A **Figure 3.** Epitope binning of bovine scFvs by cross-competition ELISA. Inactivated FMDV type O (O1 Manisa) (**A**) or type A (A22 Iraq) (**B**) antigens were coated onto a 96-well Maxisorp plate. A total of 10 µg/mL of purified scFv-Fc was added as a competitor, followed by incubation with scFv phage. Bound scFv phages were detected using HRP-conjugated anti-M13 antibody. The data represents normalized values obtained by dividing the absorbance values by that of scFv phage binding in the absence of competitor scFv-Fc. Strong binding inhibition is also represented by the darker background in the cell.

#### *2.4. Detection of FMDV Antibody in Serum by cELISA Using Bovine Antibodies*

To test whether the selected bovine antibodies can be used for the detection of serum antibodies against SP of FMDV, a solid-phase competitive ELISA (SPCE) was developed using biotinylated scFv-Fcs as described in Materials and Methods. In the SPCE, binding of biotinylated scFv-Fcs to FMDV antigen is detected by premixed to NeutrAvidin-HRP (NA-HRP) and can be inhibited by the presence of anti-SP antibodies in test sera. Because BvA3 lost its binding activity after biotinylation, intact scFv-Fcs was instead used together with anti-human IgG (Fc-specific)-HRP. After a preliminary test using a small size of serum samples from vaccinated cattle and virus-challenged pigs (Figure S2), BvO17 and BvOA7 were selected as representative clones for the respective groups since they showed slightly better response than the other clones in the same group. The final three clones, BvO17, BvA3, and BvOA7, were further tested for the SPCE to detect antibodies against FMDV type O and A. PrioCHECK Kits for type O and A were used for comparison in this study since PrioCHECK family of type-specific FMDV antibody tests are widely used as commercial SPCE kits for primary screening.

First, strong positive control sera for each serotype of FMDV and a negative control serum purchased from Pirbright Institute were tested in SPCE (Figure S3). When BvO17 was used as a reagent, its binding was significantly reduced by type O serum (Figure S3B). Except for type Asia1 serum, it showed very weak or negligible cross-reactivity for nontype O control sera. In all the tested SPCE, type Asia1 serum was reactive regardless of serotype specificity of antibody reagent used. The pattern of binding inhibition by the control sera in BvO17-based SPCE was very similar to that in PrioCHECK type O Kit (Figure S3A,B). BvA3-based SPCE showed significant binding inhibition by type A serum

and weak or negligible cross-reactivity for non-type A control sera except type Asia1 serum (Figure S3E). Unlike BvA3-based SPCE, PrioCHECK type A Kit did not show comparable binding inhibition by type A serum (Figure S3D,E). When excluding the cross-reactive type Asia1 serum, these results indicate BvO17 and BvA3 can be used for serotype-specific detection of FMDV antibodies. In the case of the pan-serotype antibody, BvOA7, its binding inhibition was observed for most of the sera tested. However, the degree of inhibition was dependent on the serotype as well as the FMDV antigen used (Figure S3C,F).

Next, serum samples from vaccinated cattle and virus-challenged pigs were tested in SPCE. The bovine sera were collected from a local slaughterhouse and were supposed to be both positive for type O and A FMDV antibodies due to national vaccination programs in South Korea. The porcine sera expected to be positive for FMDV type O or A antibodies were collected from pigs exposed to infection with Korean isolates, either FMDV (O/Anseong/SKR/2002) or FMDV (A/Yeoncheon/SKR/2017). In BvO17-based SPCE (Figure 4A), the bovine sera were strongly positive, exhibiting PI values near 100%. The values were comparable with those obtained by PrioCHECK type O Kit. The porcine sera derived from FMDV (O/Anseong/SKR/2002)-challenged pigs provided relatively lower PI values than the bovine sera but still had a high average PI value, 94%. When compared with the PrioCHECK type O Kit, which showed a low average PI value (only 59%) with large variance, BvO17-based SPCE seemed to be much better in the sensitive detection of the porcine FMDV antibodies. In BvA3-based SPCE (Figure 4B), the average PI value was 74% for bovine sera, which was comparable with the PrioCHECK type A Kit. However, the variance was huge, unlike the PrioCHECK type A Kit, and even one sample showed a PI value far below 50%. The PI values for the porcine sera derived from FMDV (A/Yeoncheon/SKR/2017)-challenged pigs were relatively low with an average of 60% and showed wide fluctuations. PrioCHECK type A Kit provided even worse data for the porcine sera exhibiting the average PI value less than 50%, also with huge variance. In the case of BvOA7-based SPCE (Figure 4A,B), the PI values were sufficiently high for both type O and A formats, indicating that they can be generally used for detection of the FMDV antibody, regardless of serotype. *Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 7 of 13 exhibiting the average PI value less than 50%, also with huge variance. In the case of BvOA7-based SPCE (Figure 4A,B), the PI values were sufficiently high for both type O and A formats, indicating that they can be generally used for detection of the FMDV antibody, regardless of serotype.

**Figure 4.** Detection of FMDV antibodies in the serum samples from vaccinated and virus-challenged animals using bovine scFv antibody-based SPCE. (**A**) For type O antibody detection in the serum samples from vaccinated cattle (n = 9, ○) and FMDV (O/Anseong/SKR/2002)-challenged pigs (n = 5, ●), BvO17 and BvOA7-based SPCEs were carried out as described in Figure S3, and their performance was compared with a parallel experiment using PrioCHECK type O antibody ELISA Kit. The results were expressed as a percentage of inhibition as described in Materials and Methods. (**B**) For type A antibody detection in the serum samples from vaccinated cattle (n = 9, □) and FMDV (A/Yeoncheon/SKR/2017)-challenged pigs (n = 4, ■), BvA3 and BvOA7-based SPCEs were carried out as described in Figure S3, and their performance was compared with a parallel experiment using PrioCHECK type A antibody ELISA Kit. **Figure 4.** Detection of FMDV antibodies in the serum samples from vaccinated and virus-challenged animals using bovine scFv antibody-based SPCE. (**A**) For type O antibody detection in the serum samples from vaccinated cattle (n = 9, #) and FMDV (O/Anseong/SKR/2002)-challenged pigs (n = 5, •), BvO17 and BvOA7-based SPCEs were carried out as described in Figure S3, and their performance was compared with a parallel experiment using PrioCHECK type O antibody ELISA Kit. The results were expressed as a percentage of inhibition as described in Materials and Methods. (**B**) For type A antibody detection in the serum samples from vaccinated cattle (n = 9, ) and FMDV (A/Yeoncheon/SKR/2017)-challenged pigs (n = 4, ), BvA3 and BvOA7-based SPCEs were carried out as described in Figure S3, and their performance was compared with a parallel experiment using PrioCHECK type A antibody ELISA Kit.

In this study, we successfully isolated bovine anti-FMDV antibodies by phage display using scFv library generated from PBLs of cattle in a local slaughterhouse. The cattle were considered to be adequate for the FMDV-immune antibody library construction because they were supposed to be immune to both FMDV type O and A due to a vaccine policy in South Korea. In fact, they were seropositive for FMDV type O and A when checked with PrioCHECK FMDV antibody detection kits. The bovine scFv library constructed was hence presumed to contain antibodies against FMDV type O and A. Through two separate pannings against type O and A antigens, we could obtain serotype (O or A) specific scFvs that distinguish other serotypes. In addition, scFv clones reactive to both type O and A antigens could be isolated and further characterized to be cross-reactive to

It is well known that the bovine antibody repertoire is derived from a highly limited diversity of germline immunoglobulin variable region genes [17,20,21,25]. For germline VH genes, there are 12 functional VH gene segments, all of which belong to one subgroup, VH1. For germline VL genes, the Vλ1 subfamily dominantly participated in the bovine light chain repertoire that consists of 95% of lambda (λ) and 5% of kappa (κ) light chains expressed. The sequence analysis of our bovine scFvs revealed VH1 and Vλ1 as the only germline VH and VL gene subfamily, respectively (Table 1). For the six selected bovine scFvs, the germline VH1-10 gene segment was most frequently used in our study. It was also one of the two most frequently used germline VH genes in the recent report on analysis of germline gene use of 55 plasmablast-derived MAbs from FMDV-infected cattle [26]. This high frequency of VH1-10 can be simply attributed to its abundance in bovine antibody repertoire rather than FMDV-specific selection. In fact, VH1-10 has been found

**3. Discussion** 

all seven serotypes.

#### **3. Discussion**

In this study, we successfully isolated bovine anti-FMDV antibodies by phage display using scFv library generated from PBLs of cattle in a local slaughterhouse. The cattle were considered to be adequate for the FMDV-immune antibody library construction because they were supposed to be immune to both FMDV type O and A due to a vaccine policy in South Korea. In fact, they were seropositive for FMDV type O and A when checked with PrioCHECK FMDV antibody detection kits. The bovine scFv library constructed was hence presumed to contain antibodies against FMDV type O and A. Through two separate pannings against type O and A antigens, we could obtain serotype (O or A) specific scFvs that distinguish other serotypes. In addition, scFv clones reactive to both type O and A antigens could be isolated and further characterized to be cross-reactive to all seven serotypes.

It is well known that the bovine antibody repertoire is derived from a highly limited diversity of germline immunoglobulin variable region genes [17,20,21,25]. For germline VH genes, there are 12 functional VH gene segments, all of which belong to one subgroup, VH1. For germline VL genes, the Vλ1 subfamily dominantly participated in the bovine light chain repertoire that consists of 95% of lambda (λ) and 5% of kappa (κ) light chains expressed. The sequence analysis of our bovine scFvs revealed VH1 and Vλ1 as the only germline VH and VL gene subfamily, respectively (Table 1). For the six selected bovine scFvs, the germline VH1-10 gene segment was most frequently used in our study. It was also one of the two most frequently used germline VH genes in the recent report on analysis of germline gene use of 55 plasmablast-derived MAbs from FMDV-infected cattle [26]. This high frequency of VH1-10 can be simply attributed to its abundance in bovine antibody repertoire rather than FMDV-specific selection. In fact, VH1-10 has been found in about 70% of sequences in a recent analysis of deep sequencing of bovine VH repertoire [27]

Another feature in the bovine antibody sequence is the long length of HCDR3, which is believed to compensate for the limited diversity of V-D-J recombination. The average length is well over 20 a.a. residues and about 10% of the antibodies have extremely long HCDR3 of between 40 and 70 a.a. [18,28,29]. This ultra-long HCDR3 could be also identified by sequence analysis of the plasmablast-derived anti-FMDV MAbs in the previous study. However, all the selected HCDR3s in our study were normal (from 14 to 29 a.a.) in length. Our scFvs went through repetitive panning rounds, which might enrich binders that have advantages in affinity and expression. Hence, we can assume that they may have such advantages over antibodies with ultra-long HCDR3 and become final survivors after in vitro selection procedure.

Studies on the antigenic profile of FMDV type O using mouse MAbs have revealed five neutralizing epitope sites [11–13]. Site 1 is well known as an immunodominant and linear epitope that includes the G-H loop in VP1, whereas the other sites are all conformational epitopes [30,31]. When the bovine MAbs developed by heterohybridoma technology were tested for epitope specificity [15], none of them recognized site 1. Instead, they all recognized the other conformational epitope sites. Similarly, most bovine plasmablastderived MAbs were found to be specific to conformational epitopes outside of site 1 [26]. Thus, unlike in mice, immunodominant epitopes for FMDV type O in cattle seemed to be conformational rather than linear. BvO17 and BvO22 were selected as type O-specific clones with high affinity in this study. Despite no sequence homology in the HCDR3 region, they showed similar epitope specificity in cross-competition ELISA (Figure 3). When they were tested for western blotting of capsid protein, none could succeed (data not shown). This result indicates that they recognize conformational epitopes but not site 1 at least and may reflect the previous suggestion that antibody response against FMDV type O in cattle differs from that in mice and does not interact dominantly with the linear epitope in the G-H loop of VP1. In this study, we did not further characterize the epitopes recognized by the bovine scFv antibodies. Since they are conformational, it may be difficult to fully characterize the epitope structure. The clues for the structure can be given by several experimental tools such as binding test with recombinant viral capsid proteins, viral epitope library scan, and

selection and identification of the antibody-resistant mutant virus. Such endeavors are very important and will contribute to the understanding of the performance of bovine scFv in cELISA and the further development of improved antibody reagent.

This difference in epitope specificity affected the performance of SPCE for the detection of FMDV type O antibodies. PrioCHECK kit includes HRP-labeled mouse MAb as a key reagent. Like the previous report on solid-phase blocking ELISA for detection of FMDV type O antibody [6], the MAb also targeted the linear epitope in the G-H loop when we checked specificity using synthetic peptide (Data not shown). Compared with the mouse MAb-based PrioCHECK kit, BvO17-based SPCE showed a similar competition pattern for controlling bovine anti-FMDV sera for seven serotypes (Figure S3). However, it was better in the sensitive detection of FMDV type O antibody, especially for serum samples from virus-challenged pigs (Figure 4). This result may indicate that more antibodies are developed against the conformational epitope region recognized by BvO17 than the linear epitope in the G-H loop when cattle and pigs are exposed to FMDV serotype O. It seems to be promising to develop BvO17-based SPCE for sensitive detection of FMDV type O antibody in both bovine and porcine serum samples.

Unlike BvO17, BvA3 did not provide good performance in SPCE. Though BvA3 based SPCE showed better serotype A-specific competition than the PrioCHECK type A kit in a test using control bovine anti-FMDV sera for seven serotypes, its sensitivity turned out to be low in tests using the serum samples from vaccinated cattle and viruschallenged pigs. PrioCHECK type A kit also showed low sensitivity, especially for the porcine samples. More sensitive detection of FMDV type A antibody in serum samples may require further screening of natural host-derived antibodies recognizing immunodominant epitope that is unique to type A. For such purpose, screening of antibody repertoire from host animals immunized with type A only can be considered as an alternative option since it is reasonable to speculate that our approach using both type O- and A-vaccinated cattle provided rare type A-specific antibody clones due to biased enrichment of antibody clones recognizing both serotypes during in vitro selection procedure.

BvOA7-based SPCE could be used successfully to detect both FMDV type O and A antibodies. When it comes to the detection of type A antibodies, it showed even better sensitivity than the BvA3-based SPCE and PrioCHECK type A kit (Figure 4B). Due to panserotype specificity, BvOA7 cannot be used to differentiate serotypes but can be developed as a general detection reagent for FMDV antibodies regardless of serotypes. Other than type-specific assay, there is also a need for a universal serological test as simplified frontline diagnostics as described recently [32].

It looks like some data in Figure S3 seem to be inconsistent with the data in Figure 4. For example, BvOA7-based SPCE and PrioCHECK type A kit did not respond effectively to type A reference serum sample (Figure S3) while both responded very well to vaccinated bovine serum samples (Figure 4B). Since Figure S3 is derived from a single serum sample for each serotype, it has a limitation of data interpretation even though each sample is purchased as a reference control serum for each subtype. When considering sample to sample variation, one should not interpret the data in Figure S3 as a representation for a specific serotype. Each bar in Figure S3 can be just one of the dots in the plot of Figure 4. Hence we relied on Figure 4, representing data collection for diverse test serum samples, to judge the feasibility of using our antibodies in cELISA. There may still remain concerns about a limited number of serum samples tested. As this was a preliminary feasibility study for the bovine scFvs selected, the serum samples were not broadly representative, and the sample size was not big enough. For further validation study, we will expand sample size and diversity by adding positive and negative samples from a broad range of host animals and virus isolates, which can meet the requirement for commercial development.

In conclusion, we herein demonstrate the isolation of anti-FMDV bovine scFv clones specific to a serotype (O or A) as well as those with pan-serotype specificity by phage library screening of antibody repertoire from FMDV-vaccinated cattle. Their performance in SPCE showing better sensitivity than conventional mouse MAbs suggests the feasibility of their application in serodiagnosis of FMDV. The benefit of using the bovine scFvs for FMDV serodiagnosis will be further proved by a validation study using a large number of diverse field serum samples, including negative serum samples. In parallel, further engineering of them as antibody reagents can also be considered for optimization of performance through direct genetic fusion with detection modules such as HRP and fluorescent proteins. Such endeavors will allow commercial development of a novel bovine scFv-based cELISA as a more reliable tool for FMDV surveillance and ultimately contribute to control FMDV spread in the world.

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

#### *4.1. Construction of Bovine scFv Library*

Blood samples of ten Hanwoo (*Bos Taurus coreanae*) cows were collected from a local slaughterhouse. Peripheral blood lymphocytes (PBLs) were isolated using Histopaque-1077 (Sigma-Aldrich, St. Louis, MO, USA) and subjected to total RNA extraction with Trizol (Thermo Fisher Scientific, Waltham, MA, USA) to obtain a high percentage of inhibition (PI) values when tested by PrioCHECK FMDV type O Antibody ELISA Kit (Thermo Fisher Scientific, Waltham, MA, USA). cDNA was synthesized using Superscript IV (Thermo Fisher Scientific, Waltham, MA, USA) and used for PCR amplification of bovine antibody variable regions, VH and VL (Vκ and Vλ). The PCR primer set used for each variable region (Table S1) was designed by considering the sequences of bovine germline antibody gene segments as described previously [33]. The amplified VH and VL sequences were combined by assembly and extension PCR using extension primers (Table S1), and the resulting scFv sequences with (G4S)<sup>3</sup> linker between VH and VL were cloned into phagemid vector, pDR-D1 as described before [23]. Bacteriophages displaying the scFv repertoire were prepared from the transformed *E.coli* ER2738 cells using VCSM13 helper phage (Stratagen, La Jolla, CA, USA) as described previously [23].

#### *4.2. Bio-Panning and Screening*

Immunotubes (Nunc, Maxisorp, Thermo Fisher Scientific, Waltham, MA, USA) were coated with either inactivated FMDV type O (O1 Manisa) or type A (A22 Iraq) antigen (Pirbright Institute, Pirbright, U.K.) for panning. The tubes were washed twice with PBS and blocked with 2% skim milk in 1X PBS supplemented with final 0.05% Tween-20 (PBST). The library phages were added to the antigen-coated tube and incubated at room temperature for 2 h. Then, unbound phages were removed by PBST washing five times. A total of 0.2 M Glycine-HCl (pH 2.7) was used to elute bound phage, and 1 M Tris-HCl (pH 8.0) was added for neutralization. The eluted phages were infected with *E.coli* ER2738 cells, followed by their superinfection with VCSM13 helper phage for amplification. The amplified phages were used for the next round of panning with an increased number of times of PBST washing.

To screen individual scFv phage clones, forty-eight colonies were randomly selected from the output plates after the third round of panning for each panning procedure. After growing them until OD<sup>600</sup> = 0.5, scFv phages were rescued by superinfection with helper phage. The rescued phages were applied onto the FMDV-coated 96-well Maxisorp plate (Nunc, Maxisorp). The phage binding was detected by phage ELISA using horseradish peroxidase (HRP)-conjugated anti-M13 antibody (Sino Biological, Beijing, China). The clones showing specific binding activity were subjected to DNA sequencing, and the nucleotide sequences determined were submitted to IMGT/V-QUEST (http://imgt.org/ IMGT\_vquest/input, last accessed December 2, 2020) for sequence analysis.

#### *4.3. Expression and Purification of scFv-Fc Proteins in 293E Cells*

The genes encoding the selected scFv clones were cloned into the pDR-OriP-Fc vector, which allows genetic fusion of scFv sequences to human gamma-1 Fc and hinges domain sequences for transient expression of scFv-Fc as described previously [23]. The constructed expression plasmid was transfected using polyethyleneimine (PEI, Polysciences, Warring-

ton, PA, USA) into 293E (CRL-10852, ATCC) cells cultivated in suspension with Ex-Cell 293 serum-free medium (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C in 8% CO2. The transfected cells were cultured for seven days at 32 °C in 8% CO<sup>2</sup> while being fed with 15% glucose (Thermo Fisher Scientific, Waltham, MA, USA) and 200× Glutamax (Thermo Fisher Scientific, Waltham, MA, USA) twice. For purification of scFv-Fc, the supernatant was subjected to affinity chromatography based on a protein G-agarose column (Merck Millipore, Darmstadt, Germany).

#### *4.4. ELISA*

96-well Maxisorp plate (Nunc, Maxisorp, Thermo Fisher Scientific, Waltham, MA, USA) were coated with FMDV antigens (O1 Manisa or A22 Iraq) (Pirbright Institute, Pirbright, U.K.) diluted in PBS (pH 7.4) overnight at 4 ◦C, and blocked with 2% skim milk or 1% BSA in PBST. The plates were washed four times with PBST between steps. All incubations were carried out at RT for 1-2 h. The color was developed with tetramethylbenzidine (TMB) substrate reagents (BD Biosciences, San Diego, CA, USA), and the reaction was stopped with 50 µL of 2.5 M H2SO4. The absorbance was measured at 450 nm (A450) using a microtiter reader (Emax, Molecular Devices, Sunnyvale, CA, USA).

The binding activities of the purified scFv-Fc were measured by an indirect ELISA. Serial dilutions of scFv-Fc were applied to FMDV antigen-coated wells and bound scFv-Fc was detected by HRP-conjugated anti-human IgG antibody produced in goat (Jackson ImmunoResearch, West Grove, PA, USA). From the binding data, equilibrium dissociation constant, K<sup>D</sup> was estimated as described previously [24]. By plotting absorbance as y and antibody concentration applied as x, K<sup>D</sup> can be calculated by nonlinear regression using hyperbola model equation, y = Amax ∗ x/(K<sup>D</sup> + x).

For antibody cross-competition binding experiment, 10 µg/mL of purified scFv-Fc was added as a competitor into each well and incubated for 1 h at RT. After washing, an scFv phage diluted with 2% skim milk in PBST was added into each well and incubated for 45 min at RT. After washing, HRP-conjugated anti-M13 monoclonal antibody (1:5000 dilution) was used to detect bound phage.

#### *4.5. Biotinylation of scFv-Fc*

The scFv-Fc proteins were conjugated with Biotin using a Sulfo-NHS-LC-Biotin (sulfosuccinimidyl-6-(biotinamido) hexanoate; Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. A total of 10 mM of Sulfo-NHS-LC-Biotin reagent was added to 1 mg of scFv-Fc proteins and incubated at 4 ◦C for 2 h. The mixture was then washed using amicon ultra centrifugal tube (10 K) several times and stored at 4 ◦C until further use.

#### *4.6. Serum Samples*

As reference sera, bovine control sera strongly positive for seven serotypes (O/UKG, A/A22, C/Oberbayern, SAT1/RHO, SAT2, SAT3/ZIM, and Asia1/Shamir) as well as negative control serum were purchased from Pirbright Institute. Bovine test sera were prepared from blood samples collected from cattle at a local slaughterhouse that were supposed to be immunized by both FMDV type O and A vaccines. Porcine test sera were collected from pigs challenged either by FMDV (O/Anseong/SKR/2002) or by FMDV (A/Yeoncheon/SKR/2017) with support from the Animal and Plant Quarantine Agency (Gimcheon, South Korea).

#### *4.7. Solid-Phase Competitive ELISA (SPCE)*

The biotinylated scFv-Fc together with NeutrAvidin-HRP (NA-HRP, Thermo Fisher Scientific, Waltham, MA, USA) was used to develop bovine scFv-based SPCE. In the case of BvA3, unlabeled scFv-Fc together with HRP-conjugated anti-human IgG (Fc-specific) was used since it lost binding activity after biotinylation. For detection of FMDV type O or A antibodies, a 96-well Maxisorp plate was coated overnight at 4 °C with FMDV type

O (O1 Manisa) or type A (A22 Iraq) antigens, respectively. The wells were blocked with 1% BSA in PBST at RT for 1 h. A total of 1:10 dilutions of serum samples were added to the wells and incubated at RT for 1 h. After washing with PBST, either a premix of biotinylated scFv-Fcs and NA-HRP or a premix of unlabeled scFv-Fc and HRP-conjugated anti-human IgG (Fc-specific) was added and incubated for 45 min at RT. The wells were washed four times with PBST, followed by incubation with 100 µL of TMB substrate for 10 min and subsequent addition of 2.5 M H2SO4. A<sup>450</sup> was measured for each serum sample. Percentage of inhibition (PI) was calculated using the following formula Equation (1):

$$\text{PI}(\%) = \left\{ 1 - \left( \frac{A\_{450} \text{test serum}}{A\_{450} \text{negative reference serum}} \right) \right\} \times 100 \tag{1}$$

For comparison, commercial SPCE (PrioCHECK FMDV type O or A antibody ELISA Kit; Prionics AG, Schlieren-Zurich, Switzerland) was also carried out in parallel according to the manufacturer's instructions. Briefly, 1:10 (type O kit) or 1:5 (type A kit) dilution of serum was incubated in the respective FMDV antigen-coated wells. After washing, dilution of HRP-conjugate provided by the kit was added to the wells and incubated for 1 h at RT. After removing unbound conjugate by repeated washes, color was developed by adding 100 µL of TMB substrate solution and incubating for 15 min. Then the reaction was stopped by adding 100 µL of stop solution. PI value was calculated according to the manufacturer's instructions.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/ijms22094328/s1.

**Author Contributions:** Conceptualization, S.J.K.; methodology, S.J., H.J.A., J.W.B., H.M.P., M.Y.P., J.N., and S.R.; validation, J.W.B., H.M.P., M.Y.P., B.K.K., and S.H.W.; investigation, S.J., H.J.A., and K.J.M.; writing—original draft preparation, S.J. and S.J.K.; writing—review and editing, S.J. and S.J.K.; supervision, S.J.K.; project administration, S.J.K.; funding acquisition, S.J.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by grants from the R&D Convergence Program of the National Research Council of Science & Technology of the Republic of Korea (Grant No. CAP-16-02-KIST) and the KRIBB Research Initiative Program (KGM9942112), Republic of Korea.

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

#### **References**


## *Article* **Monoclonal Antibody Aggregation Associated with Free Radical Induced Oxidation**

**Kai Zheng 1,\*, Diya Ren <sup>2</sup> , Y. John Wang <sup>1</sup> , Wayne Lilyestrom <sup>1</sup> , Thomas Scherer <sup>1</sup> , Justin K. Y. Hong <sup>3</sup> and Junyan A. Ji <sup>1</sup>**


**Abstract:** Oxidation is an important degradation pathway of protein drugs. The susceptibility to oxidation is a common concern for therapeutic proteins as it may impact product efficacy and patient safety. In this work, we used 2,20 -azobis (2-amidinopropane) dihydrochloride (AAPH) as an oxidative stress reagent to evaluate the oxidation of therapeutic antibodies. In addition to the oxidation of methionine (Met) and tryptophan (Trp) residues, we also observed an increase of protein aggregation. Size-exclusion chromatography and multi-angle light scattering showed that the soluble aggregates induced by AAPH consist of dimer, tetramer, and higher-order aggregate species. Sodium dodecyl sulfate polyacrylamide gel electrophoresis indicated that inter-molecular disulfide bonds contributed to the protein aggregation. Furthermore, intrinsic fluorescence spectra suggested that dimerization of tyrosine (Tyr) residues could account for the non-reducible cross-links. An excipient screening study demonstrated that Trp, pyridoxine, or Tyr could effectively reduce protein aggregation due to oxidative stress. This work provides valuable insight into the mechanisms of oxidative-stress induced protein aggregation, as well as strategies to minimize such aggregate formation during the development and storage of therapeutic proteins.

**Keywords:** monoclonal antibody; free radical; protein aggregation; oxidation; excipient

#### **1. Introduction**

Oxidation-induced degradation of therapeutic proteins is commonly observed during pharmaceutical manufacturing, handling, and storage [1–3]. These oxidation reactions occur when there are activated oxygen species including singlet oxygen (1O2), superoxide radical (O2• <sup>−</sup>), hydroxyl radical (•OH), and peroxide (-OO-) in the environment. Reactive oxygen species can be created through light, heat, free radicals, or transition metals [4,5]. These species can oxidize methionine (Met), tryptophan (Trp), histidine (His), tyrosine (Tyr), and cysteine (Cys) residues in proteins [6]. For example, it has been reported that ultraviolet (UV) irradiation induced oxidation of Met and Trp residues in a monoclonal antibody (mAb) [7], thermal or chemical stresses caused Met oxidation in proteins [8–10], and free radicals promoted Tyr oxidation in ATPase [11].

Oxidative modifications can alter protein's secondary and tertiary structures [12], hydrophobicity [13], stability [14], biological activity [14], and plasma circulation halflife [15]. Owing to these potential impacts, oxidation is typically an important quality attribute to be closely monitored during the development of therapeutic proteins. Various stress models including light and chemical stresses are often used to identify potential oxidation hot spots in therapeutic proteins and to characterize the vulnerability of labile

**Citation:** Zheng, K.; Ren, D.; Wang, Y.J.; Lilyestrom, W.; Scherer, T.; Hong, J.K.Y.; Ji, J.A. Monoclonal Antibody Aggregation Associated with Free Radical Induced Oxidation. *Int. J. Mol. Sci.* **2021**, *22*, 3952. https:// doi.org/10.3390/ijms22083952

Academic Editor: Menotti Ruvo

Received: 26 February 2021 Accepted: 2 April 2021 Published: 12 April 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/).

residues to oxidation. *Tert*-butyl hydroperoxide (*t*-BHP), hydrogen peroxide (H2O2), or 2,20 -azobis (2-amidinopropane) dihydrochloride (AAPH) are often used as chemical stress reagents to evaluate protein oxidation [6,8–10,12,16–18]. In general, *t*-BHP and H2O<sup>2</sup> primarily oxidize Met residues via nucleophilic substitution reactions. In contrast, AAPH (a water-soluble radical initiator) oxidizes both Met and Trp residues [8,19] through free radical reactions. The azo compound, AAPH, is thermally unstable and can generate alkyl radicals at elevated temperatures. In the presence of oxygen, alkyl radicals can form peroxyl radicals [20], which can further form alkoxyl radicals [11]. These radical species can effectively oxidize Met, Trp, and other oxidation-labile residues in proteins. One advantage of using an azo-radical initiator, such as AAPH or azobisisobutyronitrile, as an oxidative reagent is that it can produce a controllable and reproducible amount of oxidizing species [19,21,22].

In previous work conducted by Ji et al. [9], researchers used oxidative reagents of *t*-BHP, H2O2, and AAPH to study the mechanism of Met, Trp, and His oxidation in parathyroid hormone. Although Ji et al.'s work provided valuable insights into the oxidation mechanisms and the corresponding stabilization strategy, the model protein, parathyroid hormone, used in the study is a small protein with minimal tertiary structure. On the other hand, oxidation in large proteins, such as mAbs, may not only cause oxidation of Met, Trp, His, or other labile residues, but also induce additional physicochemical degradations. For example, the previous work did not assess aggregation that can be observed during protein oxidation. Therefore, in this work, we evaluated the oxidative degradation of therapeutic mAbs under the AAPH stress, where we observed both mAb oxidation and aggregation. Protein aggregation is often induced by physical stresses, such as agitation [23], freeze/thawing, and freeze/drying processes [24]. The AAPH-induced aggregation observed here is likely formed through covalent bonds as a result of free radical reactions. Protein aggregates can potentially trigger immune responses and are considered as a critical quality attribute for therapeutic proteins [25]. Thus, it is important to have a good understanding of the formation and the nature of protein aggregates associated with protein oxidation. In addition, this study also provides insight to reduce protein aggregation when exposing to oxidative reagents during therapeutic protein production and storage.

#### **2. Results**

#### *2.1. Aggregate Formation under the AAPH Stress*

AAPH was used to assess the oxidative degradation of mAbs during the formulation development. Size exclusion chromatography (SEC) of the oxidized mAb1 revealed a substantial increase of protein aggregates (SEC; Figure 1). The percentages of aggregates, monomer, and fragments were calculated based on SEC peak areas and are summarized in Table 1. Increasing AAPH concentrations results in a quantitative loss of monomer with corresponding increases of aggregates (SEC peaks A-C between 11 and 15 min in Figure 1). The impact on mAb1 fragmentation under the oxidative stress is relatively subtle. SEC coupled with multi-angle light scattering (SEC-MALS) was used to further characterize the aggregates in the mAb1 sample stressed with 5 mM AAPH. SEC-MALS data indicated that the aggregates upon the oxidative stress contained dimer (peak C, 3.0 <sup>×</sup> <sup>10</sup><sup>5</sup> Da) and higher-order aggregates (peaks A/B 5.2 <sup>×</sup> <sup>10</sup>5–1.4 <sup>×</sup> <sup>10</sup><sup>6</sup> Da) (Figure 1).

dashed line.

sion chromatography (SEC) data.

**Figure 1.** Expanded view of mAb1 size exclusion chromatography (SEC) profiles after 2,2′-azobis (2-amidinopropane) dihydrochloride (AAPH) stress. Chromatogram 1: control sample (mAb1 without the AAPH stress); chromatogram 2: mAb1 oxidized by 1 mM AAPH; chromatogram 3: mAb1 oxidized by 3 mM AAPH; 4: mAb1 oxidized by 5 mM AAPH. Molecular weights of peaks (A–D) for mAb1 oxidized by 5 mM AAPH are calculated and overlaid in the figure as the **Figure 1.** Expanded view of mAb1 size exclusion chromatography (SEC) profiles after 2,20 -azobis (2-amidinopropane) dihydrochloride (AAPH) stress. Chromatogram 1: control sample (mAb1 without the AAPH stress); chromatogram 2: mAb1 oxidized by 1 mM AAPH; chromatogram 3: mAb1 oxidized by 3 mM AAPH; 4: mAb1 oxidized by 5 mM AAPH. Molecular weights of peaks (A–D) for mAb1 oxidized by 5 mM AAPH are calculated and overlaid in the figure as the dashed line.

MAb1 was incubated with AAPH for 24 h at 40 °C. Percentages were calculated from size exclu-


to 20 mM sodium acetate buffer at pH 5.5. The sizes and purity of the enriched species were further confirmed by the SEC analysis. Each sample eluted as a single peak in SEC MAb1 was incubated with AAPH for 24 h at 40 ◦C. Percentages were calculated from size exclusion chromatography (SEC) data.

#### (Figure 2). The purity of higher-order aggregates, dimer, and monomer was estimated to be 99.1%, 93.2%, and 98.4%, respectively, based on SEC peak areas. *2.2. Fraction Collection from SEC*

Given that mAb1 stressed by 5 mM AAPH showed the most pronounced changes, we henceforth collected fractions of 5 mM AAPH stressed mAb1 using SEC for further characterization. Higher-order aggregates (peaks A and B in Figure 1), dimer (peak C in Figure 1), and monomer (peak D in Figure 1) peaks were collected and buffer exchanged to 20 mM sodium acetate buffer at pH 5.5. The sizes and purity of the enriched species were further confirmed by the SEC analysis. Each sample eluted as a single peak in SEC (Figure 2). The purity of higher-order aggregates, dimer, and monomer was estimated to be 99.1%, 93.2%, and 98.4%, respectively, based on SEC peak areas.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 4 of 16

**Figure 2.** SEC of the collected fractions of mAb1 oxidized by 5 mM AAPH. Chromatogram 1: before fraction collection; chromatogram 2: re-injected monomer fraction; chromatogram 3: re-injected dimer fraction; chromatogram 4: re-injected higher-order aggregates fraction. **Figure 2.** SEC of the collected fractions of mAb1 oxidized by 5 mM AAPH. Chromatogram 1: before fraction collection; chromatogram 2: re-injected monomer fraction; chromatogram 3: re-injected dimer fraction; chromatogram 4: re-injected higher-order aggregates fraction.

#### *2.3. Aggregate Characterization 2.3. Aggregate Characterization*

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis with Coomassie blue staining was used to characterize properties of the fractions collected from SEC. Both reduced and non-reduced samples were analyzed by SDS-PAGE (Figure 3). Under the non-reduced condition, the un-oxidized control (lane 2 in the SDS-PAGE gel) runs with a molecular weight (MW) of ~150 kDa, as expected. Higher-order aggregates (lane 3 in the gel) showed multiple bands with MWs higher than the 188 kDa standard, indicating that higher-order aggregates consist of covalently linked multimers that do not dissociate under the denaturing condition. The dimer species (lane 4 in the gel) showed a single band with a MW higher than 188 kDa, indicating that the dimer is also covalently linked. The monomer peak (lane 5 in the gel) has a MW similar to the un-oxidized control sample in lane 2. Under the reduced condition, both the un-oxidized control sample and the monomer fraction (lanes 7 and 10 in the gel, respectively) dissociated into light chains (~25 kDa) and heavy chains (~50 kDa). Reduced higher-order aggregates (lane 8 in the gel) dissociated to light chains, heavy chains, and a few faint bands with MWs of 100 kDa or higher. These results indicate that inter-molecular disulfide cross-links contribute to the formation of higher-order aggregates. The aggregates with MWs of 100 kDa or higher suggest that the aggregates also contain non-reducible inter-molecular crosslinks. The mass of 100 kDa suggests that the cross-links may be formed between two heavy chains. The dimer species (lane 9 in the gel) dissociated mainly to light chains and heavy chains upon reduction, indicating that inter-molecular disulfide bonds played a major role in mAb1 dimerization. In addition, a few faint bands with MWs of 100 kDa or higher were observed as well, suggesting other inter-molecular cross-links also contributed to dimerization. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis with Coomassie blue staining was used to characterize properties of the fractions collected from SEC. Both reduced and non-reduced samples were analyzed by SDS-PAGE (Figure 3). Under the non-reduced condition, the un-oxidized control (lane 2 in the SDS-PAGE gel) runs with a molecular weight (MW) of ~150 kDa, as expected. Higher-order aggregates (lane 3 in the gel) showed multiple bands with MWs higher than the 188 kDa standard, indicating that higher-order aggregates consist of covalently linked multimers that do not dissociate under the denaturing condition. The dimer species (lane 4 in the gel) showed a single band with a MW higher than 188 kDa, indicating that the dimer is also covalently linked. The monomer peak (lane 5 in the gel) has a MW similar to the un-oxidized control sample in lane 2. Under the reduced condition, both the un-oxidized control sample and the monomer fraction (lanes 7 and 10 in the gel, respectively) dissociated into light chains (~25 kDa) and heavy chains (~50 kDa). Reduced higher-order aggregates (lane 8 in the gel) dissociated to light chains, heavy chains, and a few faint bands with MWs of 100 kDa or higher. These results indicate that inter-molecular disulfide cross-links contribute to the formation of higher-order aggregates. The aggregates with MWs of 100 kDa or higher suggest that the aggregates also contain non-reducible inter-molecular cross-links. The mass of 100 kDa suggests that the cross-links may be formed between two heavy chains. The dimer species (lane 9 in the gel) dissociated mainly to light chains and heavy chains upon reduction, indicating that inter-molecular disulfide bonds played a major role in mAb1 dimerization. In addition, a few faint bands with MWs of 100 kDa or higher were observed as well, suggesting other inter-molecular cross-links also contributed to dimerization.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 16

**Figure 3.** Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of the collected fractions from SEC. Lane 1: protein standards; lane 2: non-reduced control sample (no AAPH stressed mAb1); lane 3: non-reduced higherorder aggregates; lane 4: non-reduced dimer; lane 5: non-reduced monomer; lane 6: protein standards; lane 7: reduced control sample (no AAPH stressed mAb1); lane 8: reduced higher-order aggregates; lane 9: reduced dimer; lane 10: reduced monomer. **Figure 3.** Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of the collected fractions from SEC. Lane 1: protein standards; lane 2: non-reduced control sample (no AAPH stressed mAb1); lane 3: non-reduced higherorder aggregates; lane 4: non-reduced dimer; lane 5: non-reduced monomer; lane 6: protein standards; lane 7: reduced control sample (no AAPH stressed mAb1); lane 8: reduced higher-order aggregates; lane 9: reduced dimer; lane 10: reduced monomer.

#### *2.4. Mass Spectrometry with Trypsin Digestion 2.4. Mass Spectrometry with Trypsin Digestion*

Because the aggregates contain non-reducible cross-links, we used the tryptic peptide maps to characterize chemical modifications of mAb1 after the AAPH stress (Figure 4). MAb1 contains Trp-33 in the CDR and Met-256 and Met-432 in the Fc region. Oxidation of these residues in higher-order aggregates, dimer, and monomer samples was observed as expected (indicated by arrows in Figure 4). Upon oxidation, the hydrophobicity of oxidized Met and Trp residues decreased and, therefore, eluted early on the reverse phase high-performance liquid chromatography (HPLC). Oxidation of Met-256 and Met-432 resulted in a single peak of MetOx-256 and MetOx-432 on the chromatograms, respectively. On the other hand, oxidation of Trp is more complicated than oxidation of Met because the indole side-chain of Trp could undergo various reactions during oxidation and lead to products with mass increases of +4, +16, +20, and +32 [26]; indeed, multiple peaks representing various oxidation products of Trp-33 were observed (Figure 4). These Trp-33 oxidation products eluted at 116–122 min. Some Trp oxidation products were visible as new peaks in the chromatograms and others co-eluted with other peptides. The mass increase of each Trp oxidation product is labeled in Figure 4. Except Met and Trp oxidation, Because the aggregates contain non-reducible cross-links, we used the tryptic peptide maps to characterize chemical modifications of mAb1 after the AAPH stress (Figure 4). MAb1 contains Trp-33 in the CDR and Met-256 and Met-432 in the Fc region. Oxidation of these residues in higher-order aggregates, dimer, and monomer samples was observed as expected (indicated by arrows in Figure 4). Upon oxidation, the hydrophobicity of oxidized Met and Trp residues decreased and, therefore, eluted early on the reverse phase highperformance liquid chromatography (HPLC). Oxidation of Met-256 and Met-432 resulted in a single peak of MetOx-256 and MetOx-432 on the chromatograms, respectively. On the other hand, oxidation of Trp is more complicated than oxidation of Met because the indole side-chain of Trp could undergo various reactions during oxidation and lead to products with mass increases of +4, +16, +20, and +32 [26]; indeed, multiple peaks representing various oxidation products of Trp-33 were observed (Figure 4). These Trp-33 oxidation products eluted at 116–122 min. Some Trp oxidation products were visible as new peaks in the chromatograms and others co-eluted with other peptides. The mass increase of each Trp oxidation product is labeled in Figure 4. Except Met and Trp oxidation, we did not

observe any other new peaks in the chromatogram of the higher-order aggregate or dimer fraction that indicated non-reducible cross-links, as suggested by the data of SDS-PAGE. A reasonable explanation could be that the tryptic peptide containing cross-linking sites is very large, which has low ionization efficiency and is difficult to be detected by mass spectrometry. Another possible reason is that there are multiple cross-linking sites with low abundance, which can be detected by SDS-PAGE as broad faint bands, but not by mass spectrometry once eluted at different regions on the chromatogram. we did not observe any other new peaks in the chromatogram of the higher-order aggregate or dimer fraction that indicated non-reducible cross-links, as suggested by the data of SDS-PAGE. A reasonable explanation could be that the tryptic peptide containing crosslinking sites is very large, which has low ionization efficiency and is difficult to be detected by mass spectrometry. Another possible reason is that there are multiple cross-linking sites with low abundance, which can be detected by SDS-PAGE as broad faint bands, but not by mass spectrometry once eluted at different regions on the chromatogram.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 6 of 16

**Figure 4.** Expanded view of tryptic maps of collected SEC fractions of mAb1. Chromatogram A: control sample (no AAPH stressed mAb1); chromatogram B: collected monomer fraction from SEC; chromatogram C: collected dimer fraction from SEC; chromatogram D: collected higher-order aggregate fraction from SEC. Oxidized Met and Trp products and their parent peaks are labeled in the figure. **Figure 4.** Expanded view of tryptic maps of collected SEC fractions of mAb1. Chromatogram A: control sample (no AAPH stressed mAb1); chromatogram B: collected monomer fraction from SEC; chromatogram C: collected dimer fraction from SEC; chromatogram D: collected higher-order aggregate fraction from SEC. Oxidized Met and Trp products and their parent peaks are labeled in the figure.

#### *2.5. Intrinsic Fluorescence 2.5. Intrinsic Fluorescence*

Given that mass spectrometry cannot identify non-reducible cross-linking sites, we used intrinsic fluorescence spectroscopy for further characterization. Fluorescence emission spectra of AAPH-oxidized samples under the excitation wavelength of 325 nm are shown in Figure 5A. The broad emission peak between 400 and 430 nm suggests the existence of bityrosine cross-links after AAPH oxidation, as described in a previous report [11]. This result indicated that cross-links between Tyr residues could account for the nonreducible bands in SDS-PAGE (lanes 8 and 9 in Figure 3). In addition, we collected emission spectra of bityrosine standard to further support this hypothesis (Figure 5B). Spectra of the standard and oxidized mAb1 exhibited similar fluorescent peaks between 400 and 430 nm. Furthermore, a higher intensity of fluorescent emission was observed with increased AAPH concentration (Figure 5A), indicating an increasing trend of bityrosine cross-links with higher levels of AAPH. Given that mass spectrometry cannot identify non-reducible cross-linking sites, we used intrinsic fluorescence spectroscopy for further characterization. Fluorescence emission spectra of AAPH-oxidized samples under the excitation wavelength of 325 nm are shown in Figure 5A. The broad emission peak between 400 and 430 nm suggests the existence of bityrosine cross-links after AAPH oxidation, as described in a previous report [11]. This result indicated that cross-links between Tyr residues could account for the nonreducible bands in SDS-PAGE (lanes 8 and 9 in Figure 3). In addition, we collected emission spectra of bityrosine standard to further support this hypothesis (Figure 5B). Spectra of the standard and oxidized mAb1 exhibited similar fluorescent peaks between 400 and 430 nm. Furthermore, a higher intensity of fluorescent emission was observed with increased AAPH concentration (Figure 5A), indicating an increasing trend of bityrosine cross-links with higher levels of AAPH.

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 7 of 16

**Figure 5.** (**A**) Intrinsic fluorescence spectra of the control sample (no AAPH stressed mAb1) and mAb1 after stressing by 1, 3, and 5 mM AAPH. (**B**) Intrinsic fluorescence spectra of bityrosine standard at 5 µM and 10 µM. Inset: chemical structure of bityrosine. **Figure 5.** (**A**) Intrinsic fluorescence spectra of the control sample (no AAPH stressed mAb1) and mAb1 after stressing by 1, 3, and 5 mM AAPH. (**B**) Intrinsic fluorescence spectra of bityrosine standard at 5 µM and 10 µM. Inset: chemical structure of bityrosine.

#### *2.6. Aggregation Inhibition 2.6. Aggregation Inhibition*

Because protein aggregates could induce immune responses and impact patient safety, it is important to protect therapeutic proteins from aggregation, especially when free radicals may be present in the protein solution [27]. In this study, we first screened Met, Trp, Tyr, His, and pyridoxine as aggregation inhibitors for the following reasons: free Met or Trp was added because Met and Trp oxidation was observed in mAb1 under the AAPH stress; free Tyr was evaluated as the fluorescence data suggest the formation of bityrosine cross-links; free His was selected because His oxidation has been reported in other studies [9,28]; and pyridoxine was included as a free radical scavenger [29]. We did not include ascorbic acid and N-acetyl cysteine (NAC), which are typical anti-oxidants, in the screening study as they will not be suitable excipients for therapeutic proteins: ascorbic acid has been shown to cause undesirable reactions with proteins and NAC may cause disulfide exchange [9]. The structures of the screened inhibitors are shown in Figure 6A. After incubation at 40 °C for 72 h, the aggregate levels were evaluated using SEC. The chromatograms shown in Figure 6B indicated that Trp, pyridoxine, and Tyr can effectively Because protein aggregates could induce immune responses and impact patient safety, it is important to protect therapeutic proteins from aggregation, especially when free radicals may be present in the protein solution [27]. In this study, we first screened Met, Trp, Tyr, His, and pyridoxine as aggregation inhibitors for the following reasons: free Met or Trp was added because Met and Trp oxidation was observed in mAb1 under the AAPH stress; free Tyr was evaluated as the fluorescence data suggest the formation of bityrosine cross-links; free His was selected because His oxidation has been reported in other studies [9,28]; and pyridoxine was included as a free radical scavenger [29]. We did not include ascorbic acid and N-acetyl cysteine (NAC), which are typical anti-oxidants, in the screening study as they will not be suitable excipients for therapeutic proteins: ascorbic acid has been shown to cause undesirable reactions with proteins and NAC may cause disulfide exchange [9]. The structures of the screened inhibitors are shown in Figure 6A. After incubation at 40 ◦C for 72 h, the aggregate levels were evaluated using SEC. The chromatograms shown in Figure 6B indicated that Trp, pyridoxine, and Tyr can effectively

protect protein from oxidation-induced aggregation under the AAPH stress. His and Met also decreased protein aggregation, but not to the same extent as Trp, pyridoxine, and Tyr. also decreased protein aggregation, but not to the same extent as Trp, pyridoxine, and Tyr.

(**B**)

**Figure 6.** (**A**) Structure of tested aggregation inhibitors. (**B**) Expanded view of mAb1 SEC profiles for excipient screening (except the negative control, all samples contained 1 mM AAPH). Chromatogram 1: negative control sample (no AAPH stressed mAb1); chromatogram 2: sample contained 2 mM Trp; chromatogram 3: sample contained 2 mM pyridoxine; chromatogram 4: sample contained 2 mM Tyr; chromatogram 5: sample contained 2 mM His; chromatogram 6: sample contained 2 mM Met; chromatogram 7: positive control sample (AAPH stressed mAb1). All samples contained 10 mg/mL mAb1 and were incubated at 40 °C for 72 h prior to SEC analysis. **Figure 6.** (**A**) Structure of tested aggregation inhibitors. (**B**) Expanded view of mAb1 SEC profiles for excipient screening (except the negative control, all samples contained 1 mM AAPH). Chromatogram 1: negative control sample (no AAPH stressed mAb1); chromatogram 2: sample contained 2 mM Trp; chromatogram 3: sample contained 2 mM pyridoxine; chromatogram 4: sample contained 2 mM Tyr; chromatogram 5: sample contained 2 mM His; chromatogram 6: sample contained 2 mM Met; chromatogram 7: positive control sample (AAPH stressed mAb1). All samples contained 10 mg/mL mAb1 and were incubated at 40 ◦C for 72 h prior to SEC analysis.

With this observation, we further expanded the assessment of the aggregation inhibition effect of Trp, pyridoxine, and Tyr on other types of mAbs, including another IgG1 mAb (mAb2), an IgG2 mAb (mAb3), and an IgG4 mAb (mAb4). Under the AAPH stress, all three mAbs showed substantial increases of aggregates (~12–18%) after 72 h of storage

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 9 of 16

at 40 ◦C (Figure 7), while the addition of Trp, pyridoxine, or Tyr can significantly reduce the aggregation, with the order of pyridoxine ≥ Trp > Tyr. at 40 °C (Figure 7), while the addition of Trp, pyridoxine, or Tyr can significantly reduce the aggregation, with the order of pyridoxine ≥ Trp > Tyr.

With this observation, we further expanded the assessment of the aggregation inhibition effect of Trp, pyridoxine, and Tyr on other types of mAbs, including another IgG1 mAb (mAb2), an IgG2 mAb (mAb3), and an IgG4 mAb (mAb4). Under the AAPH stress, all three mAbs showed substantial increases of aggregates (~12–18%) after 72 h of storage

**Figure 7.** Aggregation of various mAbs under 1 mM AAPH stress after 72 h at 40 °C. Grey bar: no AAPH stressed mAb (negative control); light blue bar: AAPH stressed mAb containing pyridoxine; red bar: AAPH stressed mAb containing Trp; yellow bar: AAPH stressed mAb containing Tyr; green bar: AAPH stressed mAb without aggregation inhibitors (positive control). **Figure 7.** Aggregation of various mAbs under 1 mM AAPH stress after 72 h at 40 ◦C. Grey bar: no AAPH stressed mAb (negative control); light blue bar: AAPH stressed mAb containing pyridoxine; red bar: AAPH stressed mAb containing Trp; yellow bar: AAPH stressed mAb containing Tyr; green bar: AAPH stressed mAb without aggregation inhibitors (positive control).

#### **3. Discussion**

**3. Discussion**  Protein oxidation is a very complicated process that could be triggered by chemical [4,6,30–33] or light [34–36] stresses and produces a diverse collection of oxidation products. In this study, we used AAPH as a chemical stress reagent to characterize oxidation of therapeutic mAbs. In addition to expected Met and Trp oxidation, we observed substantial protein aggregation. Although aggregated species of therapeutic proteins could cause immunogenicity risks and impact patient safety, there are few reports having a thorough investigation on oxidative-stress induced protein aggregation. Our work revealed that the newly formed aggregates are primarily covalently linked. Our results also differ from some previous reports. One report described a study that used *t*-BHP to assess Trp oxidation in a recombinant IgG1 mAb [17]. In that study, only a small increase in aggregation was observed by SEC, while our data showed a remarkable increase in aggregation. In another study [11], AAPH was used to oxidize Ca2+-ATPase. There, protein aggregation Protein oxidation is a very complicated process that could be triggered by chemical [4,6,30–33] or light [34–36] stresses and produces a diverse collection of oxidation products. In this study, we used AAPH as a chemical stress reagent to characterize oxidation of therapeutic mAbs. In addition to expected Met and Trp oxidation, we observed substantial protein aggregation. Although aggregated species of therapeutic proteins could cause immunogenicity risks and impact patient safety, there are few reports having a thorough investigation on oxidative-stress induced protein aggregation. Our work revealed that the newly formed aggregates are primarily covalently linked. Our results also differ from some previous reports. One report described a study that used *t*-BHP to assess Trp oxidation in a recombinant IgG1 mAb [17]. In that study, only a small increase in aggregation was observed by SEC, while our data showed a remarkable increase in aggregation. In another study [11], AAPH was used to oxidize Ca2+-ATPase. There, protein aggregation was observed, but the contribution from disulfide cross-links was ruled out, whereas the data in our study suggested that a majority of aggregates in mAbs were covalently linked through inter-molecular disulfide bonds under the AAPH stress. Because an mAb contains a large number of disulfide bridges, inter-molecular disulfide cross-links could begin with the reduction of a disulfide bridge by free radicals and then the formation of a cysteine

thiyl radical intermediate, which is not stable and can react with another disulfide bridge in an adjacent molecule to form an inter-molecular disulfide cross-link [37].

In addition to disulfide cross-links, bityrosine cross-links can also contribute to the covalent aggregates in mAbs. Based on a number of previous reports [11,38,39], bityrosine cross-links may be formed through a phenoxyl radical intermediate, which is the product of the reaction between free radicals and tyrosine side-chains. Two phenoxyl radicals then recombine to form an inter-molecular or intra-molecular cross-link between two benzyl-ring side chains of Tyr residues. In this work, we observed fluorescence emission spectra that are characteristic of bityrosine, which suggests the formation of bityrosine cross-links under the AAPH stress (Figure 5A). As mAbs possess complex structures and numerous Tyr residues, it increases the challenge to identify the specific bityrosine crosslinking site compared with small proteins. For mAb1, in this study, there are eighteen Tyr residues in the heavy chain and ten Tyr residues in the light chain. The observed fluorescent peak in oxidized samples can be contributed from a single bityrosine crosslinking site or multiple cross-linking sites. To further characterize bityrosine cross-links, we generated homology models of heavy chain Fab, heavy chain Fc, and light chain regions (Figure 8A–C). In the heavy chain Fab region, only the side chain of Tyr-57 is fully exposed to solvent (the green color residue in Figure 8A). In heavy chain Fc and light chain regions (Figure 8B,C), all Tyr residues are buried inside. Thus, it is likely that the non-reducible inter-molecular cross-links are formed through heavy chain Fab regions. If that is the case, the G<sup>41</sup> . . . . . . DY57A . . . K<sup>62</sup> peptide that dimerized through Tyr-57 should be observed in tryptic peptide mapping analysis (Figure 4). However, after careful analysis of mass spectroscopy data, such dimer peptide was not identified. This result may be because the dimer peptide is too hydrophobic to elute. In addition, as mentioned previously, the dimer peptide may lack sufficient ionization efficiency. Furthermore, during the AAPH stress, the mAb tertiary structure could be altered, and other Tyr residues may be exposed and cross-linked. In that case, protein aggregates contain heterogeneous bityrosine cross-links and the abundance of tryptic digested peptides that contain a specific cross-link would be too low to be detected by mass spectroscopy. A possible approach to identify cross-linking sites is to synthesize some peptides containing bityrosine cross-links and use them as standard compounds for peptide mapping analysis. Another strategy is to use site-specific mutations to investigate whether specific residues, such as Tyr-57 in the heavy chain, are responsible for the formation of inter-molecular cross-links. In addition, in-gel digestion may be another approach to identify cross-linking sites by combining SDS-PAGE and mass spectroscopy analysis. These alternative approaches are worthy of further investigation and can be explored in a separate study.

In this work, we assessed Trp, pyridoxine, and Tyr as aggregation inhibitors for IgG1, IgG2, and IgG4 types of mAbs, which covers a good range of antibody platforms as potential therapeutic proteins. To our knowledge, this is the first report to assess IgG1, IgG2, and IgG4 mAbs side-by-side for these inhibitors. Our results showed that these inhibitors can effectively reduce protein aggregation induced by free radicals for the tested mAbs. In general, these inhibitors protect proteins by consuming free radicals in the solution. Therefore, they can also provide a certain level of protection to oxidation labile residues in proteins, which was demonstrated in the previous report [9]. Ji et al.'s work showed that free Trp or pyridoxine in the formulation can effectively protect Trp oxidation in parathyroid hormone under the AAPH stress, while Tyr exhibited only slight protection of the Trp residue. Thus, to protect against both aggregation and oxidation of therapeutic proteins, we can consider combining the inhibitors tested in this work, or their homologs [40], with other anti-oxidant excipients in protein formulations. This strategy can be an effective approach, especially for large proteins that can have complex degradation mechanisms under oxidative stresses. Lastly, besides protection effectiveness, researchers also need to assess whether these excipients form adducts to therapeutic proteins and the impact of the oxidative end-products of these excipients. Though these experiments had

not been done in this work because they were outside the scope of the intended study, they should be considered during formulation development when using these excipients. pact of the oxidative end-products of these excipients. Though these experiments had not been done in this work because they were outside the scope of the intended study, they should be considered during formulation development when using these excipients.

effective approach, especially for large proteins that can have complex degradation mechanisms under oxidative stresses. Lastly, besides protection effectiveness, researchers also need to assess whether these excipients form adducts to therapeutic proteins and the im-

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 11 of 16

**Figure 8.** Homology models of mAb1 with the location of Tyr residues. (**A**) The heavy chain Fab region; (**B**) the heavy chain Fc region; and (**C**) the light chain. All Tyr residues are in color; the other residues are grey. **Figure 8.** Homology models of mAb1 with the location of Tyr residues. (**A**) The heavy chain Fab region; (**B**) the heavy chain Fc region; and (**C**) the light chain. All Tyr residues are in color; the other residues are grey.

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

#### *4.1. Reagents 4.1. Reagents*

The reagents used in this study included AAPH (EMD Chemicals, Gibbstown, NJ, USA), tris (2-carboxyethyl) phosphine (TCEP), trifluoroacetic acid (TFA), formic acid (Thermo Scientific, Waltham, MA, USA), DL-dithiothreitol (DTT), iodoacetic acid (IAA), The reagents used in this study included AAPH (EMD Chemicals, Gibbstown, NJ, USA), tris (2-carboxyethyl) phosphine (TCEP), trifluoroacetic acid (TFA), formic acid (Thermo Scientific, Waltham, MA, USA), DL-dithiothreitol (DTT), iodoacetic acid (IAA), L-Met, L-Trp, L-Tyr, pyridoxine (vitamin B6) (Sigma-Aldrich, St. Louis, MO, USA), trypsin (Roche, Indianapolis, IN, USA), water, acetonitrile, potassium phosphate monobasic, potassium phosphate dibasic (J.T. Baker, Phillipsburg, NJ, USA), and potassium chloride (Mallinckrodt Chemicals, Phillipsburg, NJ, USA).

#### *4.2. Protein Production*

MAb1 and mAb2 are humanized mAbs based on a human IgG1 framework containing heavy chain VHIII and light chain VKI subgroup sequences. MAb3 is a humanized mAb based on a human IgG2 framework containing heavy chain VHI<sup>β</sup> and light chain VLKI subgroup sequences. MAb4 is a humanized mAb based on a human IgG4 framework. All mAbs are approximately 150 kDa in this study and were expressed in Chinese Hamster Ovary cells and purified through standard antibody purification procedures, including affinity chromatography, cation-exchange chromatography, anion-exchange chromatography, and final ultrafiltration and diafiltration steps. The purified mAbs thus obtained were further buffer exchanged into 20 mM sodium acetate buffer at pH 5.5 using 10,000 molecular weight cutoff (MWCO) Amicon Ultra-15 centrifugal filter devices (Millipore) and diluted to 20 mg/mL as the starting materials for the following studies in this work.

#### *4.3. AAPH Stress of MAb1*

Two milliliters of mAb1 at 10 mg/mL in 20 mM sodium acetate buffer at pH 5.5 was incubated with 1, 3, or 5 mM AAPH (final concentrations) for 24 h at 40 ◦C in 3 cc glass vials. The incubation was performed in chambers without illumination (in the dark). After incubation, all small molecule materials, including untreated AAPH, degraded AAPH, and other oxidants, were immediately removed from the antibody solution through buffer exchange using Zeba Spin desalting columns (Thermo Scientific), and the antibody was stored in 20 mM sodium acetate buffer at pH 5.5.

#### *4.4. SEC Analysis*

SEC provides quantitative information about the molecular size distribution of the protein. Size variants of the mAbs were separated using a TosoHaas TSK G3000SWXL column (7.8 × 300 mm) eluted isocratically with a mobile phase consisting of potassium phosphate and potassium chloride (pH 6.2). The separation was conducted at 25 ◦C with a flow rate of 0.5 mL/min. The detection wavelength was set at 280 nm.

Fractions of the entire peak of higher-order aggregate, dimer, and monomer species of mAb1 stressed by 5 mM AAPH were collected using the same column and chromatography conditions as mentioned above. The amount of mAb1 injected for fractionation was 0.5 mg. All collected fractions from SEC were concentrated to 10 mg/mL and buffer exchanged to 20 mM sodium acetate buffer at pH 5.5 using 10,000 MWCO Amicon Ultra-15 centrifugal filter devices.

#### *4.5. SEC-MALS Analysis*

An 18-angle Dawn enhanced optical system (EOS) light scattering detector with a 30 mW solid-state laser (λ = 690 nm) from Wyatt Technology was used for all SEC-MALS measurements. The sample temperature was maintained at 25 ◦C by a water-cooled Peltier temperature controller. The instrument was calibrated with 99.9% toluene (chromatography grade). For the SEC-MALS analysis, a detector gain setting of 100× was used for all photodiodes at fixed angles from 38◦ to 148◦ . Because the radius of gyration of mAb1 is <10 nm, 20 µL of a dilute solution (4 mg/mL) of mAb1 was used to normalize the voltage of the photodiodes relative to the 90◦ detector using a photodiode detector gain setting of 100× at the end of each experiment. During the experimental procedure, 10 µL of protein solution at 6 mg/mL was injected for each sample. Astra 5.3.4.20 Software (Wyatt Technology Corporation, Santa Barbara, CA, USA) was used to acquire and process the static light scattering data, with a dn/dc value of 0.185 mL/g applied to calculations with the appropriate extinction coefficient at UV 280 nm.

#### *4.6. SDS-PAGE Analysis*

The enriched monomer, dimer, and higher-order aggregates, along with an unoxidized control, were denatured in the presence or absence of a reducing agent, TCEP, and analyzed by SDS-PAGE. Non-reduced samples were denatured by heating in the SDS-

PAGE sample buffer at 60 ◦C for 10 min, while reduced samples were heated in the same buffer with the presence of TCEP. Each sample (3 µg) was loaded and separated on a 4–12% polyacrylamide gradient gel along with MW standards. The protein components were then visualized by Coomassie R-250 staining solution. The polyacrylamide gel, protein ladder, SDS-PAGE sample buffer, SDS-PAGE running buffer, Coomassie R-250 staining solution, and distaining solution were purchased from Invitrogen.

#### *4.7. Tryptic Peptide Map*

Control, higher-order aggregates, dimer, and monomer species were diluted to 0.5 mg/mL in 360 mM Tris, 2 mM EDTA, 40 mM DTT, 6 M guanidine hydrochloride buffer at pH 8.6. The diluted solution was incubated at 37 ◦C for 1 h to reduce all inter- and intra-chain disulfide bonds in mAb1. After the reduction, free thiol groups were protected through alkylation by adding freshly prepared IAA solution with a final IAA concentration of 90 mM. The alkylation reaction was completed in 15 min at room temperature with protection from light. The excess IAA was quenched by the addition of 1 M DTT solution to the final concentration of 50 mM. Immediately following the reduction and alkylation, samples were buffer exchanged to 25 mM Tris, 2 mM CaCl<sup>2</sup> buffer at pH 8.2 using Zeba Spin desalting columns. Trypsin was then added at a ratio of 1:45 (w:w) to the antibody solution, which was subsequently incubated at 37 ◦C for 5 h. After digestion was complete, 10% TFA was added to a final concentration of 0.3% to quench trypsin activities. Tryptic peptides were injected into an Agilent 1200 HPLC system for liquid chromatography (LC) separation and a Thermo Fisher Scientific LTQ Orbitrap XL mass spectrometer for mass analysis.

Details of the LC separation method were reported previously [41]. In brief, gradient elution was applied to a Jupiter C18 column (2.0 × 250 mm, 5 µm, 300 Å) from Phenomenex. The flow rate was 0.25 mL/min and the column temperature was maintained at 55 ◦C. The injection amount was 35 µg and the detection wavelength was set to 214 nm. The mass spectrometry data of LC eluents were collected by an LTQ Orbitrap XL mass spectrometer and analyzed using Xcalibur 2.0.7 (Thermo Fisher Scientific) [42].

#### *4.8. Intrinsic Fluorescence Spectroscopy*

Intrinsic fluorescence spectra of mAb1 were collected using a Horiba Jobin Yvon Fluoromax-4 spectrofluorometer with a temperature-controlled water bath. All protein solutions contained 0.3 mg/mL mAb1 in 20 mM sodium acetate buffer at pH 5.5. The emission spectrum was collected from 365 to 500 nm with the excitation wavelength of 325 nm at 25 ◦C. A blank spectrum of the formulation buffer was subtracted from each sample spectrum. The excitation and emission slit widths were both set at 5 nm and data were collected with 0.2 nm increments and 0.2 s integration time.

The bityrosine standard was synthesized by Hande Science. The purity of bityrosine standard was >93%, as determined by the HPLC analysis. The emission spectra of 5 and 10 µM bityrosine standard were collected at the same condition as the mAb1 samples.

#### *4.9. Aggregation Inhibitor Screening*

In the initial screening, MAb1 at 10 mg/mL in 20 mM sodium acetate buffer at pH 5.5 with 1 mM AAPH was incubated with 2 mM aggregation inhibitors (Met, Trp, Tyr, His, or pyridoxine) at 40 ◦C for 72 h. In the subsequent screening, mAb2, mAb3, and mAb4 at 10 mg/mL in the same buffer including 1 mM AAPH were incubated with 2 mM aggregation inhibitors (Trp, pyridoxine, or Tyr) at 40 ◦C for 72 h. In addition, mAbs in 20 mM sodium acetate buffer at pH 5.5 without or with 1 mM AAPH were also incubated at 40 ◦C for 72 h as negative and positive controls, respectively, in both screening studies. After incubation, AAPH and other excipients were immediately removed from mAb solutions using Zeba Spin desalting columns. The purified samples were in 20 mM sodium acetate buffer at pH 5.5 and injected into a SEC column for size analysis.

#### **5. Conclusions**

In summary, the study results showed that free radicals can cause substantial mAb oxidation and aggregation. Free radicals can be introduced into therapeutic protein formulations from light exposure, excipient raw materials, and so on. Trp, pyridoxine, Tyr, or its homologs, in combination with other anti-oxidant excipients, can be an effective approach to protect therapeutic proteins against both aggregation and oxidation under oxidative stresses.

**Author Contributions:** Conceptualization, K.Z., Y.J.W. and J.A.J.; methodology, K.Z.; data curation, K.Z., D.R., W.L., T.S. and J.K.Y.H.; writing—original draft preparation, K.Z.; writing—review and editing, K.Z., D.R., Y.J.W., W.L., T.S., J.K.Y.H. and J.A.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors declare no competing financial interest.

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

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank Eileen Y. Ivasauskas for editorial assistance, and Sreedhara Alavattam and Wei Liu for reviewing this manuscript.

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

#### **Abbreviations**

AAPH, 2,20 -azobis (2-amidinopropane) dihydrochloride; Cys, cysteine; EOS, enhanced optical system; HIS, histidine; DTT, DL-dithiothreitol; HPLC, high-performance liquid chromatography; IAA, iodoacetic acid; LC, liquid chromatography; mAb, monoclonal antibody; Met, methionine; MW, molecular weight; MWCO, molecular weight cutoff; NAC, N-acetyl cysteine; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis; SEC, size-exclusion chromatography; SEC-MALS, size-exclusion chromatography and multi-angle light scattering; *t*-BHP, *tert*-butyl hydroperoxide; TCEP, tris(2-carboxyethyl) phosphine; Trp, tryptophan; Tyr, tyrosine; UV, ultraviolet.

#### **References**

