**Incorporation of a Hydrophilic Spacer Reduces Hepatic Uptake of HER2-Targeting A**ffi**body–DM1 Drug Conjugates**

**Haozhong Ding 1,**† **, Mohamed Altai 2,**† **, Sara S. Rinne <sup>3</sup> , Anzhelika Vorobyeva <sup>2</sup> , Vladimir Tolmachev <sup>2</sup> , Torbjörn Gräslund <sup>1</sup> and Anna Orlova 3,4,\***


Received: 17 July 2019; Accepted: 12 August 2019; Published: 14 August 2019

**Abstract:** Affibody molecules are small affinity-engineered scaffold proteins which can be engineered to bind to desired targets. The therapeutic potential of using an affibody molecule targeting HER2, fused to an albumin-binding domain (ABD) and conjugated with the cytotoxic maytansine derivate MC-DM1 (AffiDC), has been validated. Biodistribution studies in mice revealed an elevated hepatic uptake of the AffiDC, but histopathological examination of livers showed no major signs of toxicity. However, previous clinical experience with antibody drug conjugates have revealed a moderateto high-grade hepatotoxicity in treated patients, which merits efforts to also minimize hepatic uptake of the AffiDCs. In this study, the aim was to reduce the hepatic uptake of AffiDCs and optimize their in vivo targeting properties. We have investigated if incorporation of hydrophilic glutamate-based spacers adjacent to MC-DM1 in the AffiDC, (ZHER2:2891)2–ABD–MC-DM1, would counteract the hydrophobic nature of MC-DM1 and, hence, reduce hepatic uptake. Two new AffiDCs including either a triglutamate–spacer–, (ZHER2:2891)2–ABD–E3–MC-DM1, or a hexaglutamate–spacer–, (ZHER2:2891)2–ABD–E6–MC-DM1 next to the site of MC-DM1 conjugation were designed. We radiolabeled the hydrophilized AffiDCs and compared them, both in vitro and in vivo, with the previously investigated (ZHER2:2891)2–ABD–MC-DM1 drug conjugate containing no glutamate spacer. All three AffiDCs demonstrated specific binding to HER2 and comparable in vitro cytotoxicity. A comparative biodistribution study of the three radiolabeled AffiDCs showed that the addition of glutamates reduced drug accumulation in the liver while preserving the tumor uptake. These results confirmed the relation between DM1 hydrophobicity and liver accumulation. We believe that the drug development approach described here may also be useful for other affinity protein-based drug conjugates to further improve their in vivo properties and facilitate their clinical translatability.

**Keywords:** affibody; drug conjugates; hepatic uptake; DM1

#### **1. Introduction**

Drug conjugates (DCs) are an emerging class of potent biopharmaceuticals developed to overcome resistance to conventional targeted therapy and reduce off-target toxicity [1–3]. DCs are composed of a targeting agent, specifically interacting with a particular antigen, attached to a biologically active drug or cytotoxic compound via a linker. Antibody drug conjugates (ADCs) constitute the most studied class of DCs [3]. Two common types of drug molecules utilized in many ADCs are the auristatins/maytansines that inhibit microtubule polymerization and the calicheamicins which target the minor groove of DNA to induce double-stranded cuts, leading to cell death in both cases. Today, five ADCs have received market approval by the US Food and Drug Administration (FDA); gemtuzumab ozogamicin (Mylotarg®), brentuximab vedotin (Adcetris®), ado-trastuzumab emtansine (Kadcyla®), inotuzumab ozogamicin (Besponsa®), polatuzumab vedotin-piiq (Polivy®), and many others are still under development or in clinical trials [4,5].

Despite the current success, ADCs still face many limitations [6]. Many conjugation strategies rely on unspecific drug attachment to abundant lysine or cysteine residues in the monoclonal antibodies (MAbs). Even though many strategies for site-specific attachment have been developed [7], many ADCs still have a variable drug-to-antibody ratio (DAR) and variable sites of drug attachment, thus forming a nonhomogeneous final product [3,8]. The lack of homogeneity may lead to suboptimal stability, pharmacokinetics, and activity [9]. A random distribution of payloads may potentially interfere with critical residues on the antigen binding regions of MAbs. Moreover, the rather large ADCs may suffer from limited localization and penetration into solid tumors, thus restricting their antitumor efficacy.

In recent years, alternatives to MAbs have started to emerge. Engineered scaffold proteins (ESPs) are considered the next-generation non-immunoglobulin-based therapeutics [10]. They are derived from small, robust non-immunoglobulin proteins, which are used as "scaffolds" for supporting a surface with the ability to specifically interact with the desired target antigens with high affinity, such as receptors overexpressed on cancer cells. Affibody molecules (6–7 kDa) are one of the most studied classes of ESPs and they are more than 20-fold smaller than MAbs [11,12]. Affibody molecules are based on a 58 aa cysteine-free three-helix scaffold which is derived from one of the IgG binding domains in protein A expressed by Staphylococcus aureus. Affibody molecules have commonly been created by randomization of 13 surface-localized amino acids on helices 1 and 2, followed by phage display selection of binders to different biological targets. Currently, affibody molecules binding with high affinity to several cancer-associated molecular targets, such as human epidermal growth factor receptor 2 (HER2), epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 3 (HER3), insulin-like growth factor 1 receptor (IGF-1R), platelet-derived growth factor receptor beta (PDGFRβ), and carbonic anhydrase 9 (CAIX), have been developed. The cysteine-free structure of affibody molecules permits site-specific conjugation of payloads by introduction of one or more cysteine amino acids at desired position(s) in the scaffold onto which the drug (or any other prosthetic/functional group) can be site-specifically attached. This results in generation of well-defined and homogenous products. The use of affibody molecules as an alternative to MAbs for targeted drug delivery offers several advantages, including efficient production in simple prokaryotic hosts such as *Escherichia coli* [13], efficient and specific drug attachment [14] as well as a relatively smaller size compared to MAbs, which may lead to more efficient penetration and better distribution in solid tumors [15]. However, an important issue for payload delivery using small proteins like affibody molecules is rapid renal excretion. Short in vivo half-life may decrease potency and worsen patient compliance by requiring more frequent administrations. An albumin-binding domain (ABD) was used to prolong the in vivo residence time of affibody molecules by noncovalent interaction with serum albumin [16,17]. We have recently reported on the feasibility of using an anti-HER2 affibody drug conjugate for treatment of HER2-overexpressing tumors in a preclinical murine model [14]. In that study, a HER2-specific affibody molecule, ZHER2:2891, was site-specifically conjugated to the antimitotic maytansine derivate (MC-DM1) using maleimide–thiol chemistry. Mice bearing HER2-expressing ovarian cancer xenografts SKOV-3, treated with the tripartite AffiDC, (ZHER2:2891)2-ABD-MC-DM1, showed significantly longer survival—twice as long compared to mice in control groups. (ZHER2:2891)2–ABD–MC-DM1 was well-tolerated, and no signs of tissue injury or morphological changes were observed after six cycles of treatment [14]. An interesting finding of that study was the relatively high hepatic uptake of the AffiDC compared to the parental non-MC-DM1-containing HER2-targeting affibody construct. Although no histopathological changes were observed in liver sections of the treated mice, earlier reports

indicate that hepatotoxicity may be a serious adverse event associated with several FDA-approved ADCs. For example, it has been observed in several clinical studies involving ado-trastuzumab emtansine (T-DM1) that treatment was associated with elevation of hepatic transaminases and hepatic toxicity [18–20]. The mechanism underlying this observed hepatotoxicity remains elusive [20]. A recent report by Yan et al. tried to link hepatic expression of the HER2 receptor to the observed T-DM1-induced hepatotoxicity in a murine model [21]. This study demonstrated that HER2-mediated uptake of T-DM1 by hepatocytes followed by release of DM1 in the cytosol induced several changes, including disorganization of microtubules, nuclear fragmentation, and cell growth inhibition. Even though no liver toxicity was observed in the AffiDC study [14], it is possible that prolonged treatment regimens using higher doses could constitute a problem, and minimization of liver uptake is thus desirable.

In the initial AffiDC study [14], an attempt to decrease liver uptake was performed by pretreating mice with a several-fold excess of the non-MC-DM1-conjugated, HER2-targeting affibody molecule, ZHER2:342, to block available HER2 receptors. However, the hepatic uptake of AffiDC was not reduced by this pretreatment strategy. As mentioned above, the uptake of the AffiDC in liver was significantly higher compared to previously reported HER2-targeting affibody constructs lacking MC-DM1 [16,17]. A possible explanation is that the elevated hepatic uptake is mediated, at least in part, by the presence of the relatively lipophilic MC-DM1. It is known that hydrophobic compounds may facilitate greater reticuloendothelial system clearance and, therefore, increased uptake by the liver. Such effect of drug hydrophobicity on tissue distribution was observed earlier for ADCs, especially at high DARs [22].

In this study, we hypothesized that incorporation of a hydrophilic glutamate-based spacer adjacent to MC-DM1 would reduce hepatic uptake by counteracting the hydrophobic nature of the drug. To test this hypothesis, we designed AffiDCs containing either a triglutamate spacer–((ZHER2:2891)2–ABD–E3–MC-DM1) or a hexaglutamate–spacer–((ZHER2:2891)2–ABD–E6–MC-DM1) (Figure 1A).

These two drug conjugates were compared, in vitro, with the previously evaluated AffiDC, (ZHER2:2891)2–ABD–MC-DM1, containing no spacer. The conjugates were also radiolabeled with 99mTc (T1/<sup>2</sup> = 6 h, Eγ = 140 keV), through the N-terminally localized HEHEHE-tag (Scheme 1 in Supplementary Figure S1), and the influence of the glutamate spacer on hepatic uptake and overall biodistribution in a HER2-overexpressing preclinical murine tumor model was investigated.

**Figure 1.** Production and initial biochemical characterization of the conjugates. (**A**) Schematic representation of the proteins. (**B**) Conjugates after final RP-HPLC purification were analyzed on a 4%–12% SDS-PAGE gel under reducing conditions. The numbers to the left are the molecular weight (kDa) of the marker proteins in lane M. (**C**) Analytical size-exclusion chromatography profiles of the conjugates. The numbers above the chromatograms are the molecular weight (kDa) of protein standards. (**D**) RP-HPLC analysis of the conjugates during a 20 min linear gradient from 30% to 60% acetonitrile in water with 0.1% TFA.

#### **2. Results**

#### *2.1. Production and Biochemical Characterization of the A*ffi*body–MC-DM1 Conjugates*

The affibody constructs, schematically represented in Figure 1A, were recombinantly expressed and purified, and MC-DM1 was conjugated to a C-terminal cysteine. A construct lacking MC-DM1 was used ((ZHER2:2891)2–ABD–IAA) as a control, where the C-terminal cysteine was instead alkylated by 2-iodoacetamide (IAA). The purified conjugates were analyzed by SDS-PAGE under reducing conditions, and the gel showed pure proteins with essentially the expected molecular weights (Figure 1B). A weak contaminating band in the lane of (ZHER2:2891)2–ABD–MC-DM1 was visible with a molecular weight of approximately 45 kDa, and could thus constitute a dimer. The conjugates were further analyzed by size-exclusion chromatography under native conditions. The chromatogram from (ZHER2:2891)2–ABD–MC-DM1 showed that the protein was eluted as a double-peak, where the major peak had a retention time corresponding to a dimer and the minor peak had a retention time corresponding to a monomer. The other three conjugates were eluted as a single symmetrical peak with a retention time corresponding to a monomer (Figure 1C). The molecular weights were measured by ESI-TOF (Table 1) and the results showed conjugates matching exactly the molecular weight of monomeric proteins with a drug-to-affibody ratio of 1. The conjugates were further analyzed by passage through a C18 column using a linear gradient of acetonitrile in water in an RP-HPLC setup (Figure 1D). The recorded chromatograms showed that (ZHER2:2891)2–ABD–E6–MC-DM1 was eluted first, followed by (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–MC-DM1, suggesting that incorporation of glutamate residues reduced the hydrophobicity of the conjugates by shielding the MC-DM1 part from interaction with the C18 column. The control (ZHER2:2891)2–ABD–IAA, lacking

MC-DM1, was eluted even earlier than the other three, further suggesting a profound increase in hydrophobicity of the conjugates by addition of MC-DM1.


**Table 1.** Biochemical characterization of the conjugates.

<sup>a</sup> Determined by analytical RP-HPLC; <sup>b</sup> Mass spectrometry was used to determine the molecular weight (Mw) of the conjugates. Deconvolution was performed to determine the monoisotopic molecular weight of the proteins.

#### *2.2. Binding Specificity and A*ffi*nity Determination of A*ffi*body–MC-DM1 Conjugates*

To investigate if MC-DM1 conjugation and glutamic acid insertion would affect the affinity of ZHER2:2891 to HER2, a dilution series of the conjugates were injected into a biosensor over three different surfaces with different levels of immobilized extracellular domain of HER2 (Figure 2). Since each construct contains two affibody molecules, a potential avidity effect could occur if the HER2 receptor molecules are too closely spaced and allow simultaneous interaction with both. The interaction was analyzed assuming a 1:1 interaction, and consistent on- and off-rates were determined from the recorded sensorgrams for the three surfaces, indicating a lack of avidity effect and that a 1:1 interaction occurred. The equilibrium dissociation constant (KD) for the interactions were determined from the on- and off-rates and are displayed in Table 2. The K<sup>D</sup> values were found to be similar for the three MC-DM1 conjugates and the control, and ranged from 17 to 28 nM. The ability of the conjugates to interact with human serum albumin (HSA) and mouse serum albumin (MSA) was investigated by injection of a dilution series over a chip with immobilized HSA or MSA (Figure 3). The kinetic constants were derived from the sensorgrams (Table 3). The affinities (KD) for HSA ranged from 0.57 to 1.2 nM. The affinities for MSA were slightly weaker and ranged from 2.5 to 8.0 nM.

**Figure 2.** Biosensor analysis of the interactions between the conjugates and HER2. Dilution series of the conjugates were sequentially injected over flow cells with immobilized extracellular domain of HER2. All experiments were repeated once and each panel is an overlay of all concentrations, in duplicates, for each conjugate. The numbers to the right of each panel indicate the concentrations of the injected conjugates (nM) corresponding to each sensorgram.

**Table 2.** Affinity constants for conjugates interacting with HER2.

**Figure 3.** Biosensor analysis of the interactions between the conjugates and serum albumin. Serial dilutions of the conjugates were injected over a flow cell with immobilized HSA (**A**) or mouse serum albumin (MSA) (**B**). All experiments were repeated once, and each panel is an overlay of all concentrations in duplicates for each conjugate. The numbers to the right of each panel indicate the concentrations of the injected conjugates (nM) corresponding to each sensorgram.


**Table 3.** Affinity constants for conjugates interacting with serum albumin.

#### *2.3. In Vitro Cytotoxicity Analysis*

The cytotoxicity of the affibody–MC-DM1 conjugates was measured by treating AU565 (high HER2 expression), SKBR3 (high HER2 expression), SKOV3 (high HER2 expression), A549 (moderate HER2 expression), and MCF7 (low HER2 expression) cells, with serial dilutions of the conjugates followed by measurement of cell viability (Figure 4, Table 4). Two controls were also included, the nontoxic control (ZHER2:2891)2–ABD–IAA lacking MC-DM1, and the nontarget control (ZTaq)2–ABD–MC-DM1. The nontarget control was a size matched control where ZHER2:2891 had been replaced with ZTaq, an affibody molecule that specifically binds to DNA polymerase from *Thermus aquaticus*, and was thus not expected to bind to any protein of human origin [14]. (ZTaq)2–ABD–MC-DM1 was previously characterized and was found to be a homogenous protein of the expected molecular weight with

a purity >95% [14]. It was found not to interact with the HER2 receptor and did not induce cell death in cells overexpressing the HER2 receptor [14]. The targeting drug conjugates demonstrated subnanomolar IC<sup>50</sup> values on AU565 and SKBR-3 cell lines. For AU565 cells, the IC<sup>50</sup> values ranged from 0.22 to 0.48 nM, and for SKBR3 cells from 0.14 to 0.38 nM. For SKOV3, the IC<sup>50</sup> values ranged from 47 to 116 nM. The nontoxic control (ZHER2:2891)2–ABD–IAA showed a slight inhibition of cell growth on the AU565 and SKBR3 cell lines at higher concentrations (>10 <sup>−</sup><sup>9</sup> M). For SKOV3 cells, a slight growth-promoting effect was observed at the highest concentration. All conjugates demonstrated a substantially weaker cytotoxic effect on A549 and MCF7 cells. The IC<sup>50</sup> could not be measured at the concentrations used, but from Figure 4, it is evident that they were weaker than 10 <sup>−</sup><sup>6</sup> M in all cases. For all five cell lines, the nontarget control (ZTaq)2–ABD–MC-DM1 required high concentrations to affect cell viability. The IC<sup>50</sup> values could not be determined from the concentration range used, except for SKOV3 cells (IC<sup>50</sup> 350 nM). From Figure 4, it is evident that the IC<sup>50</sup> value is 2 to 3 orders of magnitude weaker for the high expressing cell lines. The nontarget control (ZTaq)2-ABD-MC-DM1 had a cytotoxic potential similar to the ZHER2:2891-containing conjugates on A549 and MCF7 cells.

**Figure 4.** In vitro cytotoxicity of the conjugates. The cytotoxicity was determined by incubating serial dilutions of the conjugates with the cell lines indicated above the panels. The concentration ranges were 0.25–250 nM (AU565), 0.13–250 nM (SKBR3), 0.4 nM–5 µM (SKOV3), 1.2 nM–1 µM (A549), and 1.2 nM–1.35 µM (MCF7). The relative viability of the cells is plotted on the *Y*-axis as a function of the compound concentration on the *X*-axis. The relative viability of cells cultivated in medium was used as reference (100%). Each datapoint corresponds to the average of four independent experiments and the error bars correspond to 1 SD.



<sup>a</sup> Ranges in parenthesis correspond to 95% confidence interval; <sup>b</sup> Not measured.

#### *2.4. Radiolabeling and Stability Test of Radiolabeled Constructs*

For further in vitro characterization and to facilitate in vivo comparison, the conjugates were site-specifically radiolabeled with 99mTc through the N-terminally localized HEHEHE-tag. Data concerning the labeling yield, radiochemical purity, and stability of the conjugates are presented in Table 5. All three AffiDCs were efficiently labeled with 99mTc (radiochemical yield = 58%–61%). The radiochemical purity after purification by size-exclusion chromatography was >99%. Incubation with a 5000-fold molar excess of histidine showed that most of the activity (>97%) was still bound to the AffiDCs even after 24 h (Table 5).


**Table 5.** Labeling yield and radiochemical purity of 99mTc-labeled AffiDCs.

<sup>a</sup> Yield is calculated as % of conjugate-bound radioactivity from total added radioactivity determined by iTLC; <sup>b</sup> Radiochemical purity is calculated as proportion of conjugate-bound radioactivity from total radioactivity after purification.

#### *2.5. In Vitro Specificity and Internalization*

To evaluate the integrity and cell interaction capability of the radiolabeled constructs, a specificity test was conducted. SKOV3 cells were incubated with the conjugates, with or without preincubation with a 500-fold molar excess of nonradiolabeled anti-HER2 affibody molecule ZHER2:342 to block available HER2 receptors. ZHER2:342 binds to the same epitope as ZHER2:2891 [23]. The three constructs could bind to SKOV3 cells in a HER2-dependent manner, since the cell-associated radioactivity was reduced significantly when HER2-receptors were presaturated with ZHER2:342 (Figure 5).

The internalization of the three AffiDCs by SKOV3 cells (high HER2 expression) was performed using a continuous incubation assay (Figure 6). The cell-associated radioactivity showed a continuous growth for the three AffiDCs up to 6 h of incubation, but at slightly different rates. The internalization of the three AffiDCs also increased over time but, again, at different rates. The construct with no glutamate spacer (ZHER2:2891)2–ABD–MC-DM1 demonstrated the highest rate of internalization compared to the glutamate spacer-containing variants (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 at all studied timepoints. The internalized fraction after 6 h incubation accounted for 36.5% ± 1.2%, 26.4% ± 1.3%, and 22.3% ± 2.3% of the total cell-associated radioactivity for (ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1, respectively (Figure 6).

**Figure 5.** In vitro specificity. Specificity of binding of 99mTc-labeled (ZHER2:2891 )2–ABD–MC-DM1 (**A**), (ZHER2:2891 )2–ABD–E3–MC-DM1 (**B**), and (ZHER2:2891 )2–ABD–E6–MC-DM1 (**C**) to HER2-expressing SKOV-3 cells in vitro. Each bar shows the mean of the values measured in 3 dishes and the error bars correspond to SD.

**Figure 6.** In vitro internalization. Internalization of 99mTc-labeled (ZHER2:2891 )2–ABD–MC-DM1 (circle), (ZHER2:2891 )2–ABD–E3–MC-DM1 (triangle), and (ZHER2:2891 )2–ABD–E6–MC-DM1 (square) by HER2-expressing SKOV-3 cells at 37 ◦C. Each datapoint is the average of three individual experiments ± 1 SD.

#### *2.6. Biodistribution and In Vivo Tumor Targeting*

Data concerning in vivo biodistribution and tumor targeting of 99mTc-labeled (ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1 at 4, 24, and 46 h post injection (p.i.) in BALB/c-nu/nu mice bearing HER2-expressing SKOV-3 xenografts are displayed in Figure 7. There was no significant difference in the residence in circulation between the AffiDCs at all studied timepoints. The tumor uptake of the three AffiDCs was comparable at all studied timepoints and showed better retention with time compared to uptake in other organs. By 46 h p.i., the tumor uptake of all three AffiDCs (5.2%–6.5% ID/g) was higher than the uptake in any other organ except the kidneys. The tumor uptake at 46 h p.i. in mice bearing RAMOS lymphoma xenografts (HER2 negative) was 6–10-fold lower compared to that in SKOV-3 xenografts: 0.9% ± 0.1%, 0.6% ± 0.1%, and 0.6% ± 0.2% ID/g for (ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1, respectively (Figure S2).

There was no significant difference in activity concentration in most organs, and it generally followed the kinetics in the blood. However, a striking difference in the uptake in the liver was observed. The activity uptake of (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 was significantly lower compared to (ZHER2:2891)2–ABD–MC-DM1 at 4 h p.i. (8.7% ± 0.2% and 8.6% ± 0.9% vs. 13.4% ± 0.9 % ID/g) and at 24 h p.i. (6.3% ± 1.8% and 5.7% ± 0.3% vs. 9.3% ± 0.7% ID/g). This difference in hepatic uptake disappeared by 46 h p.i. (6.5% ± 1.8% and 5.4% ± 1.3% vs. 5.2% ± 0.9% ID/g). Interestingly, there was no significant difference in radioactivity uptake in the gastrointestinal tract and kidneys, in connection with the reduction in hepatic uptake.

**Figure 7.** *Cont*.

**Figure 7.** In vivo biodistribution. Comparative biodistribution of 99mTc-labeled DM1 conjugates expressed as % ID/g and presented as an average value from 4 animals ± 1 SD at 4 (**A**), 24 (**B**), and 46 (**C**) h post i.v. injection in female BALB/c nude mice bearing SKOV-3 xenografts. a,b Data are presented as % ID per whole sample. Data were assessed by one-way ANOVA with Bonferroni's post hoc multiple comparisons test in order to determine significant differences between groups (*p* < 0.05) at the same timepoint.

#### **3. Discussion**

In this study, the aim has been to investigate if hepatic uptake of AffiDCs could be reduced by incorporation of a hydrophilic glutamate-based spacer adjacent to site of MC-DM1 attachment. Hepatotoxicity is one of the most common reasons for drug development failures and withdrawal of drugs from the market [24,25]. In the field of ADCs, several reports have found a link between treatment and drug-induced liver injuries. For example, it was observed that T-DM1 therapy was associated with serious grade 3 or greater adverse events in some patients, including hepatotoxicity [18–20]. Similarly, several patients treated with the prostate-specific membrane antigen-directed ADC, MLN2704, have experienced elevated dose-dependent levels of hepatic transaminases [26]. Many drug development programs therefore include development of methods aiming to identify potential liver toxicities and their mechanisms [20,21,24,25]. Despite those efforts, hepatotoxicity still remains to be one of the most complex and poorly understood areas of human toxicity. For example, Yan and coworkers tried to understand the molecular basis for hepatotoxicity induced by T-DM1 [21]. This group concluded that HER2-mediated uptake of T-DM1 by hepatocytes is directly linked to DM1-associated liver toxicity.

We have earlier reported on the development of an AffiDC, (ZHER2:2891)2–ABD–MC-DM1, targeting HER2. The AffiDC demonstrated relatively high hepatic uptake in mice post i.v. injection. The accumulation in liver of AffiDC was several-fold higher compared to other ABD-fused affibody molecules [16,17]. As mentioned above, Yan et al. reported earlier that T-DM1 induced liver toxicity through a HER2-mediated uptake of the ADC by hepatocytes. We tested this assumption by preinjecting mice with >100-fold molar excess of parental HER2-targeting affibody molecule to potentially block available HER2 receptors [14]. We found that there was no reduction in hepatic uptake of AffiDC after HER2-blocking, suggesting an unspecific liver uptake of AffiDC [14]. The main difference between AffiDC and other reported ABD-fused affibody molecules [16,17] is the presence of the drug DM1. Such drug-induced hepatic uptake has also been observed for MAbs after addition of the drug molecules [27,28]. Several groups have hypothesized that the increased hepatic uptake of ADCs may result from an increase in overall hydrophobicity of the conjugate after addition of lipophilic linkers or drug molecules [22,27,28]. Based on this, it would therefore be reasonable to suspect that the relatively high hepatic uptake of AffiDC is mainly a drug-mediated effect. We hypothesized that the incorporation of a hydrophilic spacer consisting of glutamic acid residues next to the cysteine used for MC-DM1 conjugation would lead to a decrease in hepatic uptake.

Comparison of (ZHER2:2891)2–ABD–MC-DM1 and the nondrug-conjugated (ZHER2:2891)2–ABD–IAA showed that addition of MC-DM1 increased the retention time during passage through a RP-HPLC column (Figure 1C). This represents evidence of the increased hydrophobicity conferred by MC-DM1. Further comparison of (ZHER2:2891)2–ABD–MC-DM1 with the newly designed polyglutamate spacer-containing variants, (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1, showed that addition of glutamic acid residues decreased the retention time, suggesting a shielding effect on the interaction with the C18 ligand in the column.

The newly designed AffiDCs demonstrated high binding affinity as well as specificity to HER2 receptors (Figure 2). Retaining the capacity to bind HER2 with high affinity is essential for efficient targeting. The setup in the biosensor with immobilized receptor only allows for determination of an apparent affinity since the affinity of the two affibody domains in the AffiDC for HER2 could be different, and we would thus record a mixture of the signal obtained from affibody one and affibody two interacting with HER2. However, since the kinetic constants were similar for the AffiDC/HER2 interaction on three surfaces with different HER2 density, only one of them are engaged with HER2 at any given time, and an avidity in the interaction is between the analyte and the surface is negligible. The setup with immobilized HER2 rather than immobilized AffiDC was chosen since it better mimics the cell experiments where HER2 is part of the plasma membrane and the AffiDC is free in solution. The albumin-binding function was also retained as demonstrated by the biosensor analysis of the interactions between the conjugates and serum albumin (Figure 3). All tested conjugates demonstrated a sub- to single-digit nanomolar affinity (K<sup>D</sup> value) for both HSA and MSA. These K<sup>D</sup> values are similar to results obtained previously for several affibody-based ABD-fused targeting agents [14,16,17,29,30]. The biodistribution experiments confirmed the capacity of ABD to extend AffiDC circulation time. The three AffiDCs demonstrated comparable retention in the blood at all studied timepoints. The blood associated radioactivity was 13% ± 1%, 5% ± 1%, and 2% ± 0.2% ID/g at 4, 24, and 46 h p.i. of the 99mTc-labeled AffiDCs. Affibody molecules, by themselves or as head-to-tail dimers, are generally cleared almost completely from blood within 1 h [31]. For example, in a similarly conducted biodistribution experiment, the blood activity 4 h p.i. of an anti-HER2 monomeric Z and dimeric ZZ affibody molecules (lacking an ABD) was only 1.5% ± 0.2% and 2.5% ± 0.2 % ID/g, respectively.

Being a natural amino acid, inserted glutamates were not expected to affect the degradation of affibody–MC-DM1 conjugates in the lysosomes during the process of cell intoxication. The results from the in vitro toxicity study demonstrated clearly the cytotoxicity potential of the newly designed AffiDCs with IC<sup>50</sup> values similar to the parental (ZHER2:2891)2–ABD–MC-DM1 (Figure 4 and Table 4). This cell killing potential is also comparable to that of the clinically approved trastuzumab emtansine, as was demonstrated earlier [14]. It is evident that HER2 specificity is important for efficient cytotoxic activity of the AffiDCs. The sensitivity of the low-HER2-expressing MCF-7 cells and the moderate-HER2-expressing A549 cells for AffiDCs was almost 3 orders of magnitude lower than the sensitivity of the high-HER2-expressing SKOV3, SKBR3, and AU565 cell lines. Surprisingly, there was a big discrepancy between the sensitivity of the high-HER2-expressing cell lines to our AffiDCs. The measured IC<sup>50</sup> values were in the range of 47 to 116 nM in SKOV-3 cells while it was ca. 300-fold lower in SKBR3 and AU565 (Table 4). As the level of HER2 in the three cell lines is comparable, the difference may be attributed to other factors known to decrease sensitivity to drug conjugates. These may include, among others, differences in the expression level of multidrug resistance transporters, impairment of receptor internalization, and dysfunction of lysosomal degradation mechanisms [32–34].

An unexpected finding of this study was the growth-promoting effect for SKOV-3 cells observed after incubation with the non-DM1-containing (ZHER2:2891)2–ABD–IAA affibody. We may speculate that it might be caused by HER2 dimerization, mediated by the two affibody domains in the construct, followed by an increase in intracellular signaling by the receptor. It is possible that the increased proliferation observed during incubation with a high concentration of (ZHER2:2891)2–ABD–IAA could

enhance the cytotoxic activity of DM1, since the drug is strongly acting/selective towards rapidly dividing cells through prevention of microtubule formation.

The three AffiDCs were site-specifically labeled with 99mTc through the N-terminally localized HEHEHE-tag (Table 5). After histidine challenge for 24 h, most (>97%) of the radioactivity was still associated with the conjugates. Stable labeling of the conjugates is a perquisite for accurate in vivo evaluation. It is important to mention that the spacer in (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 could potentially offer an alternative weak-chelating pocket for 99mTc, due to the electron-donating properties of glutamate sidechains [35]. However, the minimal activity release in the presence of competing histidines revealed that this is not the case for these conjugates.

The three radiolabeled AffiDCs demonstrated HER2-mediated binding to SKOV-3 cells in vitro (Figure 5). This clearly showed that site-specific radiolabeling had no negative influence on the HER2-binding properties. There was an apparent influence of the spacer on the internalization rate of the conjugates where both conjugates containing a polyglutamate spacer demonstrated a slower internalization rate compared to (ZHER2:2891)2–ABD–MC-DM1 at all studied timepoints (Figure 6). Nonetheless, the internalization experiment clearly showed that both (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 are still efficiently internalized and should thus be capable of targeted delivery of the drug DM1 to kill tumor cells similar to the previously evaluated (ZHER2:2891)2–ABD–MC-DM1.

The biodistribution data of the three AffiDCs in BALB/c nu/nu mice were in a good agreement with the data reported earlier for (ZHER2:2891)2–ABD–MC-DM1 [14]. The AffiDCs clearly demonstrated the capacity to bind to tumor xenografts in vivo in a HER2-dependent manner (Figure 7 and Figure S2). The results of the biodistribution experiment confirmed the relation between the hydrophobicity of the DM1-containing AffiDC and liver accumulation. Incorporation of the hydrophilic polyglutamate spacer enabled modulation of liver accumulation. The hydrophilized (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 AffiDCs had nearly 1.5-fold (*p* < 0.05) lower liver accumulation than that of the parental (ZHER2:2891)2–ABD–MC-DM1 (Figure 7). Several overlapping factors may be associated with the selective accumulation of drug conjugates in the liver [36]. These factors include affinity between the construct and the hepatocellular transport proteins residing outside of the cells, the potential to trigger endocytosis, the release from the endosomes or lysosomes inside the hepatic cells, and the rate at which the linker between the targeting agent and the drug is cleaved. Moreover, the affinity between the construct and its catabolites to the hepatocellular efflux transporters might also play a role in hepatic accumulation. It is important to mention that the radiolabel and the drug DM1 are located at different ends of the AffiDCs. This makes it difficult to link any observed differences in hepatic accumulation to the nature of DM1–catabolites formed after lysosomal degradation. The most plausible explanation for the observed difference in hepatic accumulation of radioactivity, stems from the difference in uptake of the three AffiDCs—having different degree of hydrophilicity—by hepatocytes. This is based on earlier findings, where reduction of overall hydrophobicity of targeting agents was found to suppress hepatic uptake [29,37–39]. Decreasing overall hydrophobicity by incorporation of hydrophilic groups or linkers has also resulted in better in vivo targeting properties for bulky ADCs, particularly reduction of hepatic accumulation [22,28]. Since the AffiDCs are approximately 10 times smaller than ADCs, it is expected that the influence of hydrophilization on liver uptake should be more profound for AffiDCs. Surprisingly, the effect on hepatic accumulation was not directly proportional to the number of incorporated glutamate residues and no significant difference in liver accumulation between (ZHER2:2891)2–ABD–E3–MC-DM1 and (ZHER2:2891)2–ABD–E6–MC-DM1 was found at any of the timepoints (Figure 7). Regardless of the underlying reason, a reduction in hepatic uptake could have a positive impact on the maximum tolerated dose of AffiDC.

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

All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) or Merck (Darmstadt, Germany) unless otherwise stated. Restriction enzymes were from New England Biolabs (Ipswitch, MA, USA).

#### *4.1. Construction of Genes Encoding A*ffi*body Constructs*

Genes encoding (ZHER2:2891)2–ABD–Cys and (ZTaq)2–ABD–Cys were constructed previously [14]. Genes encoding (ZHER2:2891)2–ABD–E3–Cys, (ZHER2:2891)2–ABD–E6–Cys flanked by *Nde*I and *Bam*HI restriction sites were synthesized by Thermo Fisher Scientific (Waltham, MA, USA). They were subcloned into the pET-21a(+) plasmid vector (Novagen, Madison, WI, USA) using *Nde*I and *BamH*I restriction enzymes.

#### *4.2. Expression and Purification of A*ffi*body Constructs*

The affibody constructs were expressed at 37 ◦C in shake flask cultures of *Escherichia coli* BL21 Star (DE3) (New England Biolabs). When OD<sup>600</sup> was between 0.6 and 1, protein expression was induced by addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (Appolo Scientific, Stockport, UK). Protein production was carried out for 3 h, after which the cells were harvested by centrifugation and lysed by sonication. The supernatants were clarified by centrifugation and filtration through a 0.45 µm Acrodisc syringe filter (Pall, Port Washington, NY, USA). The recombinantly expressed affibody constructs were purified by affinity chromatography on a HiTrap NHS sepharose column (GE Healthcare, Uppsala, Sweden) with immobilized human serum albumin (HSA) using an ÄKTA system (GE Healthcare), essentially as previously described [14] including elution with 50 mM acetic acid. The fractions containing affibody constructs were pooled and lyophilized.

#### *4.3. Conjugation with MC-DM1*

The lyophilized proteins were dissolved in PBS at pH 6.5 to a final concentration of 0.1 mM and incubated with 5 mM tris(2-carboxyethyl) phosphine (TCEP) for 30 min at room temperature., to reduce the sulfur on the C-terminal cysteine of the constructs, which could potentially have been oxidized during protein production and purification. Freshly prepared MC-DM1 (Levena Biopharma, San Diego, CA, USA), dissolved in DMSO (20 mM), was mixed with the affibody constructs at a molar ratio of 2:1, and the conjugation mixture was incubated overnight at r.t. The conjugation reaction mixture was diluted with HPLC buffer A (0.1% trifluoroacetic acid in H2O) and then loaded on a Zorbax C18 SB column (Agilent, Santa Clara, CA, USA). Bound material was eluted by a 25 min gradient from 20% or 30% to 60% or 80% buffer B (0.1% trifluoroacetic acid in acetonitrile). The fractions containing affibody–MC-DM1 conjugates were pooled followed by lyophilization.

Capping of the C-terminal cysteine to create the nontoxic control (ZHER2:2891)2–ABD–IAA was carried out with 2-iodoacetamide. Lyophilized (ZHER2:2891)2–ABD–Cys was dissolved in alkylation buffer (6M urea, 0.1 M NH4HCO3) after which dithiothreitol was added to a final concentration of 4 mM, followed by incubation for 30 min at 37 ◦C to reduce any potentially oxidized cysteine residues. 2-Iodoacetamide was added to a final concentration of 10 mM followed by incubation for 30 min at r.t. to alkylate the cysteines. The capped proteins were purified by RP-HPLC as described above for the affibody–MC-DM1 conjugates, followed by lyophilization.

The lyophilized proteins were dissolved in sterile PBS buffer and stored at −20 ◦C until use. Purified proteins (5 µg in each sample) were analyzed by SDS-PAGE (Biorad, Hercules, CA, USA) under reducing conditions. The molecular weight of purified affibody–MC-DM1 conjugates was measured by ESI-TOF mass spectrometry (Agilent).

#### *4.4. Binding Specificity and A*ffi*nity Determination*

A Biacore T200 and a Biacore 3000 instrument (GE Healthcare) were used for biosensor analysis. The extracellular domain of HER2 (HER2ECD) (Sino Biological, Beijing, China) was immobilized to 210, 310, and 456 RUs on three different flow cells on a CM5 chip by amine coupling in sodium acetate buffer, pH 4.5. A reference flow cell was created by activation and deactivation. On a second CM-5 chip, HSA (Novozymes, Bagsvaerd, Denmark), MSA (Sigma-Aldrich, St. Louis, MO, USA), and BSA (Merck Millipore) were immobilized in the same way. The final immobilization levels were 869, 584, and 779 RUs, respectively. HBS-EP (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.05% Tween 20, pH 7.4) was used as running buffer and for dilution of the analytes. All experiments were performed at 25 ◦C with a flow rate of 50 µL/min. The chips were regenerated by injection of 15 mM HCl for 30 s. The binding kinetics was analyzed by the Biacore evaluation software using the one-to-one kinetics model.

#### *4.5. In Vitro Cytotoxicity Analysis*

AU565, SKBR-3, SKOV-3, A549, and MCF7 cell lines were obtained from American Type Culture Collection (American Type Culture Collection, ATCC via LGC Promochem, Borås, Sweden) and were grown in McCoy's 5A (SKOV-3, SKBR-3), RPMI-1640 (AU565), or Dulbecco's modified Eagle medium (A549 and MCF7) (Flow, Irvine, UK) supplemented with 10% FBS (Sigma-Aldrich, St. Louis, MO, USA) in a humidified incubator at 37 ◦C in 5% CO<sup>2</sup> atmosphere. Approximately 5000 cells/well (2000 cells/well for SKOV-3) were seeded in a 96-well plate and allowed to attach for 24 h. Subsequently, the medium was replaced with fresh medium containing serial dilutions of affibody–MC-DM1 conjugates or 2-iodoacetamide-capped nontoxic control followed by incubation for 72 h. Cell viability was determined using Cell Counting Kit-8 (CCK-8; Sigma-Aldrich) according to the manufacturer's protocol with measurement of A<sup>450</sup> in each well. The obtained absorbance values were analyzed by GraphPad Prism using a log(inhibitor) vs. response-variable slope (four parameters) model (GraphPad Software, Inc., La Jolla, CA, USA).

#### *4.6. Radiolabeling and Stability Test of Radiolabeled Constructs*

Site-specific radiolabeling of AffiDCs ((ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1) with 99mTc using ( 99mTc(CO)3(H2O)3) + precursor was performed as previously described [14]. In brief, eluted pertechnetate, 99mTcO<sup>4</sup> - , (400–500 µL) from <sup>99</sup>Mo/ 99mTc generator was added to a CRS kit (PSI, Villigen, Switzerland) to generate the (99mTc(CO)3(H2O)3) + (tricarbonyl technetium) precursor. The mixture was vortexed carefully and incubated at 100 ◦C for 20 min. After incubation, 20 µL of the tricarbonyl technetium solution was added to a tube containing 55 µg of the respective AffiDC in 100 µL of PBS and incubated for 60 min at 60 ◦C. To isolate the radiolabeled AffiDCs, the mixture was passed through a NAP-5 size-exclusion column (GE Healthcare) pre-equilibrated and eluted with 2% BSA in PBS. Radiochemical yield and purity of the conjugates were determined using silica-impregnated ITLC strips (150–771 DARK GREEN Tec-Control Chromatography strips (Biodex Medical Systems, Shirley, NY, USA) eluted with PBS and measured using the Cyclone Storage Phosphor System (PerkinElmer, Waltham, MA, USA). To evaluate the stability of the radiolabeled AffiDCs, they were incubated with a 5000-fold molar excess of histidine at 37 ◦C for up to 4 and 24 h, respectively. The percentage of protein-bound radioactivity after histidine challenge was determined using radio-ITLC as mentioned above.

#### *4.7. In vitro Specificity and Internalization*

To confirm the specificity of binding of 99mTc-radiolabeled AffiDCs to HER2-expressing cells in vitro, SKOV-3 cells (5–7.5 × 10<sup>5</sup> ) were incubated with 2 nM of each conjugate at 37 ◦C for 60 min (*n* = 3). For blocking, another set of dishes containing SKOV-3 cells were preincubated with 500-fold molar excess of nonlabeled anti-HER2 affibody molecule ZHER2:342 prior to the addition of radiolabeled AffiDCs. Thereafter, both medium and cells were collected from each dish and measured for radioactivity using an automated γ-spectrometer (1480 Wizard; Wallac Oy, Turku, Finland). Data are presented as mean values from three cell dishes with standard deviation.

The internalization of 99mTc-radiolabeled AffiDCs by HER2-expressing cells was studied using a method described earlier by Altai et al. [14]. For this, four groups of dishes (*n* = 3) containing SKOV-3 cells (5–7.5 × 10<sup>5</sup> cells/dish) were incubated with 2 nM (per dish) of the respective conjugate at 37 ◦C. At determined timepoints (1, 2, 4, and 6 h) after incubation, a group of dishes (*n* = 3) was removed from the incubator. Media was then discarded, and cells were washed with 1 mL of serum-free media. Thereafter, cells were incubated with 0.5 mL urea–glycine buffer pH 2.5 (acid wash) for 5 min on ice. This acid wash was then collected. An additional 0.5 mL of the acid wash was also used to wash the cells, and this fraction was collected immediately. Cells were then incubated with 0.5 mL 1 M NaOH solution for at least 30 min at 37 ◦C to lysate the cells (base wash). Cells were additionally washed with 0.5 mL base wash. Both acid and base washes were measured for radioactivity using automated γ-spectrometer.

#### *4.8. Biodistribution and In Vivo Targeting*

The animal experiments were planned and performed in accordance with national legislation on laboratory animal protection. The animal studies were approved by the local ethics committee for animal research in Uppsala, Sweden (C85/15).

Comparative biodistribution studies of 99mTc-labeled (ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1 were performed in female BALB/c nude mice (Scanbur A/S, Karlslunde, Denmark). Two weeks before the start of the experiment, 36 mice (6–8 weeks old) were injected with 10 × 10<sup>6</sup> SKOV-3 cells/per mouse (HER2+) in the right hind leg. The mice (18.4 ± 1.4 g) were randomized to nine groups, with four mice in each group. Animals were injected intravenously with 6 µg (of each conjugate) per animal in 100 µL PBS containing 2% BSA. The injected radioactivity was calculated to give 30 kBq per mouse by the time of dissection. At predetermined timepoints (4, 24, and 46 h p.i.) mice were euthanized by overdosing of anesthesia (Ketalar (ketamine): 10 mg/mL, Pfizer AB, Sweden; Rompun (xylazine): 1 mg/mL, Bayer AG, Leverkusen, Germany) followed by heart puncture and exsanguination. Organs and tissue samples were collected and weighed, and the radioactivity was measured using an automated γ-spectrometer.

To demonstrate the specific delivery of 99mTc-labeled (ZHER2:2891)2–ABD–MC-DM1, (ZHER2:2891)2–ABD–E3–MC-DM1, and (ZHER2:2891)2–ABD–E6–MC-DM1 to HER2-expressing tumors, an in vivo specificity study was performed. For this, an additional 12 BALB/c nude mice were xenografted with 5 × 10<sup>6</sup> RAMOS (HER2) lymphoma cells in the right hind leg. Each group of four mice (*n* = 4) were i.v. injected with 6 µg (30 kBq) of the respective conjugate in 100 µL PBS containing 2% BSA. Mice were euthanized at 46 h p.i. and treated as mentioned above.

#### **5. Conclusions**

In conclusion, this study demonstrated that insertion of a polyglutamate spacer is an effective strategy to decrease hepatic uptake of affinity protein drug conjugates. The use of the hydrophilic and negatively charged glutamate spacer provided, by far, the lowest level of hepatic uptake for AffiDCs. Accumulation in other organs and tissues was also low, and no influence on the HER2-mediated tumor uptake was observed. We believe that the approach described here represents a means for the development of other targeting affinity protein drug conjugates for treatment of disseminated cancers and to facilitate their clinical translatability.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/11/8/1168/s1, Figure S1: Scheme 1: Structures of the 99mTc(CO)<sup>3</sup> chelated by HEHEHE tag, Figure S2: In vivo specificity.

**Author Contributions:** H.D. and M.A. contributed equally to this study. H.D., M.A., V.T., A.O. and T.G. conceived and designed the experiments. H.D., M.A., S.R., A.V., V.T., A.O. and T.G. performed the experiments and analyzed the data. H.D., M.A. and T.G. wrote the paper. All authors agreed with the accuracy and integrity of all parts of the work.

**Funding:** This research was funded by the Swedish Cancer Society (grants CAN 2018/824 (T.G.), CAN 2017/425 (A.O.) and CAN2015/350 (V.T.)), the Swedish Research Council (2015-02509 (A.O.) and 2015-02353 (V.T.)), an ESCAPE Cancer grant from the Swedish Agency for Innovation VINNOVA (2016-04060 and 2019-00104 A.O.), and the Swedish Society for Medical Research (M.A.).

**Conflicts of Interest:** V.T. and A.O. own shares in Affibody AB. M.A., H.D., A.V., S.R. and T.G. declare no conflict of interest.

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **IL3RA-Targeting Antibody–Drug Conjugate BAY-943 with a Kinesin Spindle Protein Inhibitor Payload Shows E**ffi**cacy in Preclinical Models of Hematologic Malignancies**


Received: 26 October 2020; Accepted: 17 November 2020; Published: 20 November 2020 -

**Simple Summary:** IL3RA (alpha subunit of the interleukin 3 receptor) is a cell membrane protein frequently expressed in acute myeloid leukemia (AML) and Hodgkin lymphoma; therefore, it is a promising therapeutic target for cancer treatment. Here, we introduce BAY-943, a novel IL3RA-targeting antibody–drug conjugate that shows potent and selective efficacy in IL3RA-positive AML and Hodgkin lymphoma cell lines. In IL3RA-positive AML mouse models, BAY-943 improved survival and reduced tumor burden. Impressively, treatment with BAY-943 induced complete tumor remission in 12 out of 13 mice in an IL3RA-positive HL model. BAY-943 showed a favorable safety profile without any signs of toxicity in rats and monkeys. Overall, these preclinical results support the further development of BAY-943 for the treatment of IL3RA-positive hematologic malignancies.

**Abstract:** IL3RA (CD123) is the alpha subunit of the interleukin 3 (IL-3) receptor, which regulates the proliferation, survival, and differentiation of hematopoietic cells. IL3RA is frequently expressed in acute myeloid leukemia (AML) and classical Hodgkin lymphoma (HL), presenting an opportunity to treat AML and HL with an IL3RA-directed antibody–drug conjugate (ADC). Here, we describe BAY-943 (IL3RA-ADC), a novel IL3RA-targeting ADC consisting of a humanized anti-IL3RA antibody conjugated to a potent proprietary kinesin spindle protein inhibitor (KSPi). In vitro, IL3RA-ADC showed potent and selective antiproliferative efficacy in a panel of IL3RA-expressing AML and HL cell lines. In vivo, IL3RA-ADC improved survival and reduced tumor burden in IL3RA-positive human AML cell line-derived (MOLM-13 and MV-4-11) as well as in patient-derived xenograft (PDX) models (AM7577 and AML11655) in mice. Furthermore, IL3RA-ADC induced complete tumor remission in 12 out of 13 mice in an IL3RA-positive HL cell line-derived xenograft model (HDLM-2). IL3RA-ADC was well-tolerated and showed no signs of thrombocytopenia, neutropenia, or liver toxicity in rats, or in cynomolgus monkeys when dosed up to 20 mg/kg. Overall, the preclinical results support the further development of BAY-943 as an innovative approach for the treatment of IL3RA-positive hematologic malignancies.

**Keywords:** acute myeloid leukemia; antibody-drug conjugate; CD123; IL3RA; kinesin spindle protein inhibitor

#### **1. Introduction**

Interleukin 3 receptor subunit alpha (IL3RA; also known as CD123) is the α subunit of the heterodimeric IL-3 receptor. Together with the β subunit, it forms a functional high-affinity receptor for IL-3 [1–3]. IL-3 is a pleiotropic cytokine that is mainly produced by activated T lymphocytes, and it regulates the function and production of hematopoietic and immune cells [4]. IL3RA is expressed at high levels in ≈80% of acute myeloid leukemias (AML) [1,2,5], 59–100% of classical Hodgkin lymphomas (cHL), and the majority of blastic plasmacytoid dendritic cell neoplasms (BPDCN) [6–10]. It is also expressed by close to 100% of myelodysplastic syndrome (MDS) patients, but the expression intensity may vary [11–14]. Importantly, IL3RA overexpression on AML blasts has been associated with an increased number of leukemic blast cells at diagnosis and with a negative prognosis [15]. IL3RA is also expressed in basophils and plasmacytoid dendritic cells [5,16].

Several studies have indicated that IL-3 and its receptor play important roles in the progression of AML [3,17], and indeed, experiments with a monoclonal antibody that blocks the binding of IL-3 to IL3RA have shown increased survival in AML mouse models [18]. Characterization of hematologic malignancies has demonstrated increased IL3RA expression in CD34+CD38<sup>−</sup> AML blasts as compared to expression in normal cells. Furthermore, these IL3RA-overexpressing cells have been shown to be able to initiate and maintain the leukemic process in immuno-deficient mice and thus act as leukemic stem cells [3,19]. Consequently, IL3RA has been shown to be a very useful biomarker for the detection of minimal residual disease, thereby predicting relapse in AML patients [20,21]. Taken together, these results suggest that IL3RA is a very attractive target for an antibody–drug conjugate (ADC) approach for the treatment of AML and other IL3RA-positive hematologic malignancies [10].

Here, we exploited a novel pyrrole subclass payload that potently inhibits the kinesin spindle protein (KSP/KIF11/Eg5) in biochemical and cellular assays to develop an ADC to target IL3RA on cancer cells [22–25]. KSP is a motor protein responsible for an essential event in mitosis, the segregation of duplicated centrosomes during spindle formation in the G2/M phase of the cell cycle, and therefore, it is required for productive cell divisions [26]. High expression of KSP in hematologic indications such as AML blasts and diffuse large B-cell lymphoma (DLBCL) [27] and in solid cancers such as breast, bladder, and pancreatic cancer has been linked to poorer prognosis [28], and thus, KSP presents an attractive target for cancer treatment.

KSP is active in all proliferating cells and therefore, KSP inhibitors (KSPi) representing various structural classes have resulted in neutropenia, mucositis, and stomatitis in clinical trials [28–32]. However, ADCs that combine a cancer cell-targeting antibody and a cytotoxic payload via a linker can deliver a cytotoxic payload specifically to target-expressing cancer cells. This approach could protect healthy tissue from exposure to the cytotoxic compound, thus decreasing overall side effects especially on highly proliferative tissues, thereby expanding the therapeutic window.

To generate the IL3RA-ADC BAY-943, a non-cell-permeable KSPi was conjugated randomly to the lysine residues of a humanized derivative of the anti-IL3RA antibody 7G3 [33], TPP-9476, via a novel protease-cleavable peptide linker [24]. IL3RA-ADC was efficacious in IL3RA-positive AML and HL cell lines in vitro, as well as in IL3RA-expressing AML and HL cell line and patient-derived xenograft (PDX) models in vivo. IL3RA-ADC was well-tolerated in the mouse, rat, and cynomolgus monkey. No signs of neutropenia, mucositis, or stomatitis, the typical side effects of small molecule KSPis, were observed in safety studies performed in rat and cynomolgus monkey. Taken together, these data support the further development of the compound as a novel therapy option for patients with AML or other hematologic malignancies expressing IL3RA.

#### **2. Results**

#### *2.1. Characterization of the IL3RA-Targeting Antibody TPP-9476 and IL3RA-ADC BAY-943*

The binding affinity of the IL3RA-targeting antibody TPP-9476 (IL3RA-Ab) to human and cynomolgus monkey IL3RA was assessed by surface plasmon resonance (SPR) and flow cytometry. IL3RA-Ab showed high affinity to both the human and the cynomolgus monkey IL3RA protein with dissociation constants (KD) of 11 nmol/L and 16 nmol/L, respectively, as determined by SPR. No binding to murine IL3RA was observed. Furthermore, IL3RA-Ab bound specifically to the IL3RA-expressing human hematologic cancer cell lines MOLM-13, MV-4-11, and KG-1 as determined by flow cytometry (Figure 1A).

− − − − − **Figure 1.** Characterization of the interleukin 3 receptor subunit alpha (IL3RA) antibody TPP-9476 and schematic representation of the IL3RA antibody–drug conjugate (ADC) BAY-943. (**A**) The binding of the IL3RA-targeting antibody (IL3RA Ab) TPP-9476 to IL3RA-positive hematologic cell lines as determined by flow cytometry. The obtained EC<sup>50</sup> values were 2.73 <sup>×</sup> <sup>10</sup>−<sup>9</sup> M for MOLM-13, 6.53 <sup>×</sup> <sup>10</sup>−<sup>9</sup> M for MV-4-11, and 4.54 × 10−<sup>9</sup> M for KG-1 cells. (**B**,**C**) The internalization of the IL3RA Ab TPP-9476 and an isotype control antibody into IL3RA-positive MOLM-13 (**B**) and MV-4-11 (**C**) cells as determined by flow cytometry-based imaging. (**D**) The colocalization of the IL3RA Ab TPP-9476 in lysosomes in the IL3RA-positive MOLM-13 and IL3RA-negative HBL-1 cells. (**E**) Schematic representation of the IL3RA-ADC BAY-943. TPP-9476 represents the IL3RA-Ab. The "cell trapper" functionality indicates a non-cell-permeable payload metabolite that enables maximal retention in target cells after cleavage. (**F**) The binding of the IL3RA-ADC BAY-943 to IL3RA-positive MOLM-13 cells as determined by flow cytometry. The obtained EC<sup>50</sup> values were 20.4 <sup>×</sup> <sup>10</sup>−<sup>9</sup> M for ILRA3A-ADC and 18.7 <sup>×</sup> <sup>10</sup>−<sup>9</sup> M for ILRA3A Ab, respectively.

As the prerequisite for ADC activity is to effectively deliver the cytotoxic payload into the cells, we next studied the ability of IL3RA-Ab to internalize upon target binding. The fluorescently labeled IL3RA-Ab showed highly specific, target-dependent internalization in the IL3RA-positive MOLM-13 and MV-4-11 cell lines with >3.5-fold enhancement as compared with the non-specific internalization of the isotype control antibody (Figure 1B,C). In flow cytometry-based imaging, IL3RA-Ab showed lysosomal colocalization in the IL3RA-positive MOLM-13 but not in the IL3RA-negative HBL-1 cell line (Figure 1D). The lysosomal colocalization of IL3RA-Ab indicates that when incorporated into an ADC, it allows the release of the payload metabolite. This can occur via the cleavage of the linker by a lysosomal protease that is active at acidic pH (such as legumain) and/or by proteolytic degradation of the antibody.

Since IL3RA-Ab demonstrated the essential properties of an effective ADC antibody, we conjugated a non-cell permeable KSPi to the lysine residues of the IL3RA-Ab TPP-9476 via a novel legumain-cleavable peptide linker [24] to produce the IL3RA-targeting ADC BAY-943 (IL3RA-ADC; Figure 1E). IL3RA-ADC showed high stability in human plasma (Supplementary Figure S1C) and a comparable binding affinity to the IL3RA-Ab (half-maximal effective concentration, EC<sup>50</sup> 18–21 nmol/L in MOLM-13 cells; Figure 1F), indicating that the attachment of the KSPi payload linker does not impact the binding affinity of the ADC antibody moiety. Furthermore, the active payload metabolite of IL3RA-ADC, BAY-716, showed poor permeability across Caco-2 cells (apparent permeability, Papp A-B = 1.8 nm/s, Papp B-A = 2.7 nm/s) with an efflux ratio (Papp B-A/Papp A-B) of 1.6, indicating that no active efflux takes place in Caco-2 cells. The poor permeability from B-A in Caco-2 cells indicates a long residence time after intracellular release of the active KSPi metabolite BAY-716 in tumor cells. As Caco-2 cells express the efflux transporter P-gP (P-glycoprotein), it also suggests that BAY-716 is a poor substrate for the efflux transporter P-gP.

#### *2.2. IL3RA-ADC Shows Potent and Selective E*ffi*cacy In Vitro*

The in vitro cytotoxicity of the IL3RA-ADC BAY-943 was assessed in a panel of human tumor cell lines with different IL3RA expression levels (Table 1). IL3RA-ADC demonstrated potent antiproliferative activity with half-maximal inhibitory concentration (IC50) values at the subnanomolar to nanomolar range in the IL3RA-positive AML (MV-4-11, MOLM-13) and HL (HDLM-2, L-428) derived cell lines, whereas little activity was observed in the tumor cell lines with low levels of or negative for IL3RA membrane expression (NCI-H292, HT). Moreover, in IL3RA-positive AML and HL cell lines, a 10 to 100-fold higher sensitivity to IL3RA-ADC compared to the non-targeted isotype control ADC was observed (Table 2), demonstrating that the activity of IL3RA-ADC is target-dependent. Furthermore, IL3RA-ADC was found to induce apoptosis specifically in IL3RA-positive cells, as demonstrated by caspase 3/7 activation in MV-4-11 with an EC<sup>50</sup> of 4.33 nmol/L, but not in the IL3RA-negative MDA-MB-231 cells (EC<sup>50</sup> > 300 nmol/L; Supplementary Figure S2), further supporting the selectivity of IL3RA-ADC.

#### *2.3. IL3RA-ADC Improves Survival in the MOLM-13 and MV-4-11 Xenograft Models*

The antitumor efficacy of the IL3RA-ADC BAY-943 was tested in two IL3RA-positive, systemic (intravenous transplantation) cell line-derived xenografts: MOLM-13 human AML and MV-4-11 human biphenotypic leukemia models in mice. Both the MOLM-13 and MV-4-11 cell lines harbor *FLT3-ITD* mutations shown to be associated with an unfavorable prognosis in AML [34]. The median survival time (MST) for the vehicle and isotype control ADC was 22.5 and 46, respectively (Figure 2). By contrast, in the MOLM-13 model, 80–100% of the mice treated with 10 mg/kg IL3RA-ADC survived without signs of leukemia until day 123, when the study was terminated (Figure 2B), while all mice treated with the isotype control ADC were sacrificed due to signs of disease by day 67 after tumor cell inoculation. In the MV-4-11 model, the administration of IL3RA-ADC once weekly (Q7D), every two weeks (Q14D), or every three weeks (Q21D) resulted in a potent and sustained antitumor effect with MSTs of 120.5, 145.5, and 105 at the IL3RA-ADC dose of 2.5 mg/kg (Figure 2C) and 162, 153, and 140 days at the IL3RA-ADC dose of 10 mg/kg, respectively (Figure 2D). The MST for the vehicle and the isotype control ADC was 48 and 148, respectively. No significant differences in efficacy between the tested treatment schedules were observed in either of the models.


**Table 1.** The antiproliferative activity of IL3RA-ADC in a panel of tumor cells.

In vitro cytotoxicity (CellTiter-Glo®, Promega) of the IL3RA-ADC BAY-943 in cancer cell lines with different levels of anti-IL3RA antibody bound per cell (ABC) as determined by quantitative flow cytometry. The mean IC<sup>50</sup> values from up to six individual assays are shown. n.d., not determined. <sup>a</sup> IL3RA expression analyzed by IHC on paraffin-embedded cell pellets; <sup>b</sup> IC<sup>50</sup> determined at 144 h (at 72 h for the other cell lines).



In vitro cytotoxicity (CellTiter-Glo®, Promega) of the IL3RA-ADC BAY-943, isotype control ADC BAY-229, IL3RA antibody TPP-9476, and small molecule KSP inhibitor BAY-331 in the IL3RA-positive AML cell lines MV-4-11, MOLM-13, HDLM-2, THP-1 and in the IL3RA-low expressing NSCLC cell line NCI-H292 after 72 h incubation time. Anti-IL3RA ABC levels as determined by quantitative flow cytometry are indicated in the parentheses after each cell line. Ab, antibody; ADC, antibody–drug conjugate; DAR, drug-to-antibody ratio; n.a., not applicable; NSCLC, non-small-cell lung carcinoma; SMOL, small molecule.

**Figure 2.** Antitumor efficacy of the IL3RA-ADC BAY-943 in the systemic MOLM-13 and MV-4-11 leukemia xenograft models. A-B. Kaplan–Meier survival plots of mice transplanted with the MOLM-13 human acute myeloid leukemia (AML) cells and treated intravenously (i.v.) with the isotype control ADC (10 mg/kg, Q7D) or IL3RA-ADC at 2.5 mg/kg (**A**) or 10 mg/kg (**B**); Q7D, Q14D, or Q21D. C-D. Kaplan–Meier survival plots of mice transplanted with the MV-4-11 human biphenotypic leukemia cells and treated i.v. with the isotype control ADC (10 mg/kg, Q7D) or IL3RA-ADC at 2.5 mg/kg (**C**) or 10 mg/kg (**D**); Q7D, Q14D or Q21D. The vertical dashed gray lines delineate the treatment period, and the arrows indicate time of treatment. Data were analyzed using the Cox proportional hazards model and corrected for family-wise error rate using Sidak's method. Asterisks and hashtags indicate statistical significance in comparison to the vehicle (\*\* *p* < 0.01, \*\*\* *p* < 0.001) or isotype control ADC (### *p* < 0.001).

In the vehicle and the isotype control ADC groups, nearly all mice had symptoms of leukemia, i.e., splenomegaly and paralysis of hind limbs. In addition, the mean body weights decreased in these treatment groups, indicating that the mice suffered from leukemia (Supplementary Figure S3). However, no treatment-related side effects or abnormalities were observed during the study or at gross necropsy in the IL3RA-ADC-treated groups.

The antitumor efficacy of IL3RA-ADC was also tested in the subcutaneous MOLM-13 and MV-4-11 xenograft models in mice. Repetitive dosing with IL3RA-ADC resulted in a significant suppression of tumor growth in both models compared to the isotype control ADC, while the standard-of-care cytarabine showed no activity in these models (Supplementary Figure S4). Furthermore, in the MV-4-11 model, treatment with the unconjugated IL3RA-Ab TPP-9476 at 5 mg/kg, Q7D×2 showed no tumor growth inhibition (Supplementary Figure S4C,D), indicating that the antitumor activity of IL3RA-ADC is conveyed by the targeted delivery of the KSPi payload and not the IL3RA-Ab.

#### *2.4. IL3RA-ADC Suppresses Tumor Burden and Improves Survival in Systemic AM7577 and AML11655 PDX Models*

The efficacy of IL3RA-ADC was further evaluated in the systemic AM7577 and AML11655 patient-derived AML xenograft models in mice. These PDX models showed high IL3RA protein expression (Supplementary Figure S5) and harbor a typical AML genotype with mutations in genes including *NMP1*, *FLT3-ITD*, *IDH1*, *IDH2*, and *DNMT3A* (Supplementary Table S1).

In the AM7577 PDX model, IL3RA-ADC administered at 10 mg/kg intraperitoneally (i.p.) reduced tumor burden compared to the vehicle or isotype control ADC, as indicated by a decreased number of human CD45 (hCD45)/human IL3RA (hIL3RA)-positive cells in blood on day 56 (both *p* < 0.001; Figure 3A). Furthermore, treatment with IL3RA-ADC resulted in improved survival with the MST of 69 days at the dose of 2.5 + 10 mg/kg and 82 days at the dose of 10 mg/kg. The MST for the vehicle and the isotype control ADC was 62 and 64 days, respectively (Figure 3B).

**Figure 3.** Antitumor efficacy of the IL3RA-ADC BAY-943 in the patient-derived AM7577 and AML11655 AML xenograft models. (**A**) Tumor burden on day 56 in mice transplanted with AM7577 cells and treated i.p. with the isotype control ADC (10 mg/kg, Q7D) and IL3RA-ADC (2.5 + 10 mg/kg or 10 mg/kg, Q7D). In the 2.5 + 10 mg/kg IL3RA-ADC treatment group, the first dose was 2.5 mg/kg (day 38) and the two subsequent doses (on days 45 and 59) 10 mg/kg. (**B**) Kaplan–Meier survival plots of mice described in panel A. Treatment days in all groups except for the 2.5 + 10 mg/kg IL3RA-ADC treatment group are indicated with gray arrows. (**C**) Tumor burden on day 54 in mice transplanted with AML11655 cells. Intraperitoneal treatments with the isotype control ADC (10 mg/kg, Q7D) were initiated on day 34 and with IL3RA-ADC (10 mg/kg, Q7D) on day 5 (preventive setting) or 34 (therapeutic setting). (**D**) Kaplan–Meier survival plots of mice described in panel C. Treatment days are indicated with red arrows. The data were analyzed using the Cox proportional-hazards model and corrected for family-wise error rate using Sidak's method. Asterisks and hashtags indicate statistical significance in comparison to vehicle (\* *p* < 0.05, \*\*\* *p* < 0.001) and the isotype control ADC (### *p* < 0.001).

In the AML11655 mouse xenograft model, IL3RA-ADC was administered i.p. using either a preventive (treatment started on day 5) or a therapeutic (treatment started on day 34) setting. IL3RA-ADC administered at 10 mg/kg inhibited the growth of IL3RA-positive AML cells, as indicated by temporarily reduced numbers of hCD45-positive cells in blood compared to vehicle or isotype control ADC in both settings (Figure 3C,D; all *p* < 0.001). In addition, treatment with IL3RA-ADC using either the preventive or therapeutic setting resulted in prolonged MSTs of 107 or 108 days, respectively (Figure 3D). The MST for the vehicle and the isotype control ADC was 79 and 78 days, respectively.

#### *2.5. IL3RA-ADC Demonstrates Antitumor E*ffi*cacy in Subcutaneous HDLM-2 Hodgkin Lymphoma Xenograft Model*

Finally, the antitumor efficacy of the IL3RA-ADC BAY-943 was tested in a subcutaneous HDLM-2 Hodgkin lymphoma xenograft model in mice. This model showed a high IL3RA antigen density with ≈74,300 anti-IL3RA antibodies bound per cell (Table S1), which is in line with the literature [8]. In the ≈

HDLM-2 model, the i.p. injection of IL3RA-ADC at 5 or 10 mg/kg resulted in a strong reduction of tumor growth compared to the vehicle (both *p* < 0.001; Figure 4). This effect was comparable with the clinically studied KSPi ispinesib administered at 10 mg/kg in the same model. In the two IL3RA-ADC treatment groups, total tumor eradication was observed in twelve mice out of thirteen (92%) at the end of the study.

**Figure 4.** Antitumor efficacy of the IL3RA-ADC BAY-943 in the subcutaneous HDLM-2 Hodgkin lymphoma xenograft model. Mice were transplanted with HDLM-2 cells and treatments with IL3RA-ADC (5 or 10 mg/kg, Q7D×2, i.p.) or ispinesib (10 mg/kg, Q7D×3, i.v.) were initiated on day 49. (**A**) Tumor growth curves. ADC treatment days are indicated with red arrows and ispinesib administration is indicated with blue arrows. (**B**) Tumor volume on day 84. Statistical analyses were performed using a linear mixed-effects model with random intercepts and slopes for each subject (n = 6–7). Mean comparisons between the treatment and control groups were performed using the estimated linear mixed-effects model and corrected for family-wise error rate using Sidak's method. Asterisks indicate statistical significance in comparison to vehicle (\*\*\* *p* < 0.001).

#### *2.6. IL3RA-ADC Is Well-Tolerated*

The safety, including possible changes in the hematologic cell populations, of IL3RA-ADC was evaluated in the cynomolgus monkey in two range-finding studies with single or repeated dosing. IL3RA-ADC was well-tolerated without adverse events typically observed with ADCs containing other payload classes, such as thrombocytopenia, neutropenia, or signs of liver toxicity. In addition, mucositis, a dose-limiting toxicity for small molecule KSP inhibitors in clinical studies [35], was not observed. A single dose of IL3RA-ADC up to 20 mg/kg or three repeated doses of IL3RA-ADC up to 10 mg/kg given every three weeks resulted in a transient reduction of IL3RA-positive basophils and plasmacytoid dendritic cells (pDCs), indicating targeting of the IL3RA-ADC to antigen-positive cells (data not shown). Furthermore, the administration of IL3RA-ADC showed no obvious changes in the percentage of the CD34+Lin<sup>−</sup> bone marrow cell population containing the hematopoietic stem cells, as analyzed by flow cytometry (data not shown).

− The effect of the toxophore metabolite BAY-716 was also analyzed after a single dose of 0.25 mg/kg in rats (data not shown). No laboratory or histopathology findings were observed, indicating that this non-cell-permeable toxophore metabolite does not induce toxic effects thereby underlining the good safety profile of IL3RA-ADC.

#### **3. Discussion**

Despite the recent progress in the treatment of AML, clinical outcomes have improved only minimally over the past three decades. Therefore, novel therapeutic agents with a larger therapeutic window and a favorable tolerability profile are urgently needed to improve the therapeutic outcome for AML patients. Increasing evidence indicates that IL3RA is highly expressed in leukemic stem cells but not in normal hematopoietic stem cells, and it associates in AML with treatment response, minimal residual disease detection, and prognosis [3,15,17]. Consequently, several IL3RA-targeting approaches, such as an anti-IL3RA antibody enhanced for antibody-dependent cell-mediated

cytotoxicity, anti-IL3RA-ADCs with highly potent payloads of the pyrrolobenzodiazepine (PBD) or indolinobenzodiazepine pseudodimer (IGN) class, various bispecific T cell recruiting antibodies, or chimeric antigen receptor T cell (CAR-T) therapies are currently under preclinical or clinical development [10,28,36–39].

Here, we explored a novel concept to improve the therapeutic window and safety of KSP inhibition by targeting a non-cell-permeable KSP inhibitor as ADC to AML cells, and thereby, sparing fast-dividing healthy cells from KSP inhibition. This provides a payload with a novel mode of action and would be a new therapeutic option for the treatment of IL3RA-positive malignancies. The investigated IL3RA-targeting ADC (BAY-943, IL3RA-ADC) consists of a humanized anti-IL3RA antibody conjugated with a stable lysine linkage to a potent proprietary KSPi via a protease-cleavable linker, producing a non-cell-permeable payload metabolite.

KSP is an ATP-dependent plus-end directed motor protein, which generates force and moves along microtubules, and it is involved in the separation of the centrosomes, the generation of the bipolar spindle, and thereby plays an important role in mitosis [26]. The inhibition of KSP with small molecules such as monastrol or small interfering RNA (siRNA)-mediated knockdown results in the formation of monopolar spindles (termed a "monoaster"), which lead to aberrant mitotic arrest and apoptosis [30,40]. Thus, KSP presents a convincing target for the development of an anti-mitotic approach for cancer treatment. Accordingly, several allosteric KSP inhibitors such as ispinesib, litronesib, and filanesib (ARRY-520) have been or are in clinical trials [35,41–44]. Filanesib has also been explored in a Phase I clinical trial in AML [43], and clinical studies are ongoing in relapsed refractory multiple myeloma (rrMM). The antitumor activity of filanesib has previously been shown in AML cells in vitro and in xenograft mouse models [27]. The most common side effects of KSP inhibitors with different chemical scaffolds are neutropenia, mucositis, and stomatitis [35]. This has been explained by the inhibition of KSP in highly proliferative cells such as neutrophils and cells lining the mucosa and the stoma, respectively, thus limiting their therapeutic efficacy due to a small therapeutic window.

Antibody–drug conjugates (ADCs) are one solution that has been proposed to mitigate the toxic side effects of anti-mitotic therapies and to broaden their therapeutic window. In fact, currently more than 60 ADCs against multiple targets in solid and hematologic tumors are in clinical trials [45,46], and eight ADCs have meanwhile been approved [47]. The payload classes currently used are confined to microtubule destabilizers (e.g., auristatin, dolastatin, maytansinoid, tubulysine), DNA interacting agents (e.g., calicheamicin, duocarmycin, PBD, IGN), and topoisomerase inhibitors (e.g., camptothecin derivative SN-38, exatecan). Many of these permeable payloads and/or highly potent DNA-interacting payloads have safety issues and therefore result in an insufficient therapeutic window. For example, the clinical trials for the CD33-targeting SGN-CD33A and the IL3RA-targeting SGN-CD123A [38,39] both with a PBD payload were terminated in 2017 and 2018, respectively. Recently, the first IL3RA-targeting therapy was approved for BPDCN [9]. However, the fusion protein tagraxofusp-erzs (formerly called SL-401), which consists of the ligand IL3 fused to a truncated diphtheria toxin, has been reported to cause capillary leak syndrome as a common side effect in more than 55% of patients [10]. This further underlines that efficacious therapies with acceptable safety profiles are still urgently required for targeting IL3RA-positive malignancies.

The IL3RA-ADC BAY-943 demonstrated the capability of delivering a novel cytotoxic payload to IL3RA-positive cells. The IL3RA antibody TPP-9476 to which the KSP inhibitor payload is linked via a legumain-cleavable peptide linker showed high binding affinity and specificity to IL3RA and bound specifically to IL3RA-expressing human AML and HL cell lines. The IL3RA antibody internalized into the lysosomes of IL3RA-positive MOLM-13 and MV-4-11 AML cell lines, and IL3RA-ADC demonstrated high cytotoxic potency in IL3RA-positive MV-4-11 and MOLM-13 AML and HDLM-2 and L-428 HL derived cell lines. Furthermore, in the IL3RA-positive cell lines tested, IL3RA-ADC showed 10 to 1000-fold cytotoxicity compared with the isotype control ADC, indicating high target selectivity. The less prominent selectivity observed in MOLM-13 and THP-1 cells may be explained by a non-specific uptake of the isotype control ADC to AML cells differentiated along the macrophage lineage.

In an in vivo setting, IL3RA-ADC administered at 10 mg/kg increased survival in both the IL3RA-positive MOLM-13 and MV-4-11 cell line-derived and IL3RA-positive AM7577 and AML11655 patient-derived AML xenograft models harboring molecular alterations associated with poor prognosis in AML. The increased survival was accompanied by a reduction in the growth of IL3RA-positive AML cells and tumor size in the systemic and subcutaneous models, respectively. In the IL3RA-positive HDLM-2 subcutaneous Hodgkin lymphoma model, IL3RA-ADC treatment also resulted in significant antitumor efficacy with most animals being tumor-free at the end of the study.

The body weights of the animals decreased over the course of the study, indicating that they suffered from leukemia. However, no IL3RA-ADC treatment-related body weight losses were observed, suggesting good tolerability (Supplementary Figures S3 and S6), particularly in comparison to the small molecule KSPi ispinesib, which induced a transient body weight loss in mice after the second treatment (Supplementary Figure S6C). The treatments with IL3RA-ADC in the HDLM-2 subcutaneous Hodgkin lymphoma model were also well-tolerated.

The good tolerability of the IL3RA-ADC was confirmed by repeated dose safety and immunotoxicity assessments in the cynomolgus monkey with no changes in the portion of the CD34+Lin<sup>−</sup> cell population, and transient decreases in basophils and IL3RA-positive basophils. Importantly, liver toxicity, thrombocytopenia, and neutropenia, which are frequently observed with ADCs in the clinic and in cynomolgus monkey preclinical studies [48,49], were not apparent, which was most likely due to the fact that the IL3RA-ADC toxophore metabolite is non-cell permeable. In addition, neutropenia and mucositis, which were the dose-limiting toxicities for small molecule KSP inhibitors in the clinic, were not observed. Furthermore, the IL3RA-ADC metabolite BAY-716 showed poor permeability across Caco-2 cells, indicating that the metabolite is trapped in tumor cells after its intracellular release. This "cell trapper" functionality enables a long-lasting exposure and at the same time potentially reduces off-target toxicities through the low permeability of KSPi into normal cells.

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

#### *4.1. Cell Lines*

Cell lines were acquired from DSMZ (German Collection of Microorganisms and Cell Cultures GmbH; Braunschweig, Germany) unless otherwise noted and cultured according to the provider's instructions. Human MDA-MB-231 breast cancer, NCI-H292 non-small cell lung cancer, MV-4-11 and THP-1 acute monocytic leukemia, KG-1 acute myelogenous leukemia, Rec-1 mantle cell lymphoma, and Ramos Burkitt's lymphoma cells were obtained from ATCC (American Type Culture Collection; Manassas, VA, USA). Human OVCAR-8 ovarian cancer cells were acquired from the NCI-60 Human Tumor Cell Line Panel (National Cancer Institute, Rockville, MD, USA). The human diffuse B cell lymphoma cell line HBL-1 was obtained from Dr. Georg Lenz (Charité Universitätsklinikum, Berlin, Germany) and cultivated in RPMI 1640 supplemented with 10% fetal calf serum (FCS). The human Hodgkin lymphoma cell line L-428 (source not known) was cultivated in RPMI 1640 supplemented with 10% FCS. Cancer cell lines were obtained between 2002 and 2012, authenticated using short tandem repeat DNA fingerprinting at DSMZ (Table 1), and subjected frequently to mycoplasma testing.

#### *4.2. Compounds*

The anti-IL3RA antibodies TPP-9476 and TPP-8988 (recognizes a different epitope in the extracellular domain of IL3RA than TPP-9476) and the isotype control antibody TPP-754, the IL3RA-ADC BAY-943, the isotype control ADC BAY-229 (with TPP-754), the non-cell-permeable toxophore metabolite BAY-716 (active toxophore metabolite of IL3RA-ADC), and the cell-permeable small molecule KSPi BAY-331 were manufactured at Bayer AG. Ispinesib (SYNT1009) was purchased from Syncom (a contract research organization in organic chemistry, www.syncom.eu; Groningen, the Netherlands), cytarabine (HT0476) from Accord Healthcare GmbH (Neutraubling, Germany), and staurosporine (#S4400) from Sigma-Aldrich (Saint Louis, MO, USA).

The IL3RA-specific hIgG1 antibody (IL3RA-Ab, TPP-9476) was generated by humanization of the murine anti-IL3RA antibody 7G3 [33] as described in Lerchen et al. [22]. During a protein engineering process, which is meant to bring the amino acid sequence as close as possible to the next human germline [50], multiple variants were tested. The final version, TPP-9476, comprises several amino acid exchanges in the light and heavy chain that resulted in enhanced internalization.

The generation and characterization of the IL3RA-Ab is described in the Supplementary Methods. The IL3RA-targeting ADC BAY-943 (IL3RA-ADC) was generated by conjugating the KSPi to the lysine residues of IL3RA-Ab via a protease-cleavable linker [24]. The characterization of IL3RA-ADC is described in the Supplementary Methods. In the in vivo efficacy studies, IL3RA-ADC BAY-943 with a drug-to-antibody ratio (DAR) of 6.3 as determined by mass spectrometry was used. At a DAR of 6.3, no aggregation of the IL3RA-ADC was observed (Supplementary Figure S1)

#### *4.3. Internalization and Lysosomal Colocalization of IL3RA-Ab*

Internalization and colocalization experiments were performed in MOLM-13 and MV-4-11 AML cell lines using flow cytometry-based imaging. The IL3RA-specific antibody TPP-9476, IL3RA-ADC BAY-943, corresponding isotype control antibody TPP-754, and isotype control ADC BAY-229 were lysine-conjugated with a ten-fold molar excess of CypHer 5E mono NHS ester (GE Healthcare, Chicago, IL, USA) at pH 8.3. The reaction mixture was purified by chromatography (PD10 desalting column, GE Healthcare, Chicago, IL, USA), followed by centrifugation (Vivaspin 500, Sartorius Stedim Biotech, Aubagne, France). Alexa 488 (Jackson ImmunoResearch, West Grove, PA, USA) was utilized as a constitutive dye. The fluorescence was measured using the Amnis® FlowSight® or the Guava easyCyteTM flow cytometers (Luminex Corporation, Austin, TX, USA) and analyzed using the IDEAS® software or the guavaSoft 2.6 software (Luminex Corporation, Austin, TX, USA).

For the internalization assay, the tumor cells (5 × 10<sup>4</sup> /well) were incubated with the labeled antibodies (10 µg/mL) at 37 ◦C, 5% CO<sup>2</sup> for 0, 1, 2, and 6 h. The fluorescence was measured using the Amnis® FlowSight® or the Guava easyCyteTM flow cytometers and analyzed using the IDEAS® or the guavaSoft 2.6 software. The kinetics of the internalization were determined based on the analysis of the median fluorescence intensity (MFI) over time.

For the colocalization studies, MOLM-13 and MV-4-11 tumor cells (5 × 10<sup>4</sup> /well) were incubated with the labeled antibodies (20 µg/mL) at 37 ◦C, 5% CO<sup>2</sup> for 0 h, 0.5 h, 2 h, and 6 h. The lysosomal compartment marker CytoPainter LysoGreen (1:2000; Abcam, Cambridge, UK) was added 30 min before the end of the incubation period. After incubation, the cells were washed and resuspended in ice-cold FACS (fluorescence-activated cell sorting) buffer consisting of phosphate-buffered saline (PBS) and 3% FCS. The lysosomal colocalization was analyzed with FACS image analysis using the IDEAS® software.

The assessment of the stability of the IL3RA-ADC BAY-943 in human plasma and the permeability of the KSPi toxophore metabolite BAY-716 in Caco-2 cells are described in the Supplementary Methods.

#### *4.4. In Vitro Cytotoxicity of IL3RA-ADC*

The antiproliferative activity of IL3RA-ADC was determined in a panel of human tumor cell lines using the CellTiter-Glo® assay (Promega Madison, WI, USA). Cells (2000–5000 cells/well) were incubated at 37 ◦C, 5% CO<sup>2</sup> for 24 h and the compounds were added at concentrations of 3 × 10−11–3 × 10−<sup>7</sup> M (or 3 × 10−12–3 × 10−<sup>8</sup> M, depending on the cell line tested) in triplicates. Cell viability was determined at the beginning (day 0) and after 72 h incubation in the presence or absence of ADCs. The IC<sup>50</sup> of the growth inhibition was calculated in comparison to day 0. The IL3RA antigen density was determined with the QIFI (Dako, Glostrup, Denmark) quantitative flow cytometry assay using the murine anti-IL3RA antibody clone 7G3 (Becton Dickinson, Franklin Lakes, NJ, USA).

#### *4.5. In Vivo Studies*

All animal experiments were conducted in accordance with the German Animal Welfare Law and approved by Berlin authorities (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit Berlin, LAGetSi; code number A0378/12). When a body weight loss of >10% was observed, treatment was ceased until recovery. In the systemic models, mice were sacrificed individually when signs of leukemia were observed (>20–30% body weight loss, hind leg paralysis, or general deterioration of health status). The molecular alterations of the tested in vivo models are described in Supplementary Table S1.

For the systemic MOLM-13 and MV-4-11 models, female CB17-SCID (Janvier Labs, Le Genest-Saint-Isle, France) or NOD SCID (Taconic, Køge, Denmark) mice were injected intravenously (i.v.) with 200 µL of 7.5 × 10<sup>6</sup> or 5 × 10<sup>6</sup> cancer cells in 0.9% NaCl, respectively. The mice were treated with i.p. injection of IL3RA-ADC at 2.5 or 10 mg/kg once weekly (Q7D), every two weeks (Q14D), or every three weeks (Q21D). In the MOLM-13 model, treatments were started on day 10, and the study was terminated on day 124. In the MV-4-11 model, treatments were started on day 3, and the study was terminated on day 174 after tumor cell injection.

For the systemic AM7577 PDX model, female NOD/SCID mice (Shanghai Lingchang Bio-Technology Co. Ltd., LC, Shanghai, China) were injected i.v. with 100 µL of 1.4 × 10<sup>6</sup> cancer cells in PBS at CrownBio (Beijing, China). The development of AML was monitored by flow cytometric analysis of the percentage of hCD45 cells in blood. On day 38 after tumor cell injection, when approximately 4% of hCD45-positive cells were present, the mice were randomized, and treatments were started. The mice were treated with i.v. injections of IL3RA-ADC at 2.5 or 10 mg/kg, Q7D or the isotype control ADC at 10 mg/kg, Q7D. In the first IL3RA-ADC treatment group, the initial dose of 2.5 mg/kg, Q7D was increased to 10 mg/kg, Q14D from the second administration onwards (indicated as 2.5 + 10 mg/kg). The study was terminated on day 104 after tumor cell injection.

For the systemic AML11655 PDX model, female CIEA NOG mice® (NOD.Cg-*Prkdcscid Il2rgtm1Sug*/JicTac, Taconic, Køge, Denmark) were injected i.v. with 400 µL of 1 × 10<sup>7</sup> cancer cells in PBS at EPO Berlin-Buch GmbH (Berlin, Germany). The development of AML was monitored by the percentage of hCD45-positive cells in blood as determined by flow cytometry. Treatments were initiated on day 5 after tumor cell injection (preventive setting) or day 34 after tumor cell injection (therapeutic setting) when approximately 46% of hCD45-positive cells were detected in blood. The mice were treated with i.v. injections of IL3RA-ADC at 10 mg/kg, Q7D or the isotype control ADC at 10 mg/kg, Q7D. The study was terminated on day 118 after tumor cell injection.

For the HDLM-2 Hodgkin lymphoma model, female CB17-SCID mice (Janvier Labs, Le Genest-Saint-Isle, France) were injected subcutaneously (s.c.) with 100 µL of 1 × 10<sup>7</sup> cancer cells suspended in 30% Matrigel/70% medium. Tumor volume (0.5 × length × width<sup>2</sup> ) was determined based on twice weekly measurement of tumor area by a caliper (length and width). Treatments with IL3RA-ADC (5 or 10 mg/kg, i.p., Q7D×2) or ispinesib (10 mg/kg, i.v., Q7D×3) were started on day 49 when the tumors had reached a mean size of 100 mm<sup>3</sup> . The study was terminated on day 84 after tumor cell injection.

The subcutaneous MOLM-13 and MV-4-11 models as well as safety studies are described in the Supplementary Methods.

#### *4.6. Statsitical Analyses*

Statistical analyses were performed using R (version 3.3.2 or newer; R Foundation for Statistical Computing, Vienna, Austria) [51]. Flow cytometry and tumor volume data were analyzed using a linear model estimated with generalized least squares that included separate variance parameters for each study group or linear mixed-effects model with random intercepts and slopes for each subject. Mean comparisons between the treatment and control groups were performed using the estimated linear or linear mixed-effects model and corrected for family-wise error rate using Sidak's method. Survival analyses were performed using the Cox proportional-hazards model and corrected for family-wise error rate using Sidak's method. *p* values < 0.05 were considered significant.

#### **5. Conclusions**

The novel IL3RA-ADC with a differentiated mode-of action demonstrates selective binding and internalization to IL3RA-positive cells, which translates into selective and efficacious antitumor activity in IL3RA-positive AML and Hodgkin lymphoma models. By employing a KSP inhibitor, a stable lysine linkage between the payload and the antibody, and a legumain-cleavable linker resulting in a non-cell-permeable payload metabolite, IL3RA-ADC presents a new alternative for the treatment of IL3RA-positive malignancies. Using the KSPi as a payload in an ADC is expected to result in manageable toxicity and a broader therapeutic window compared to that reported for the systemic application of KSPi in clinical trials. Our data support further development of the IL3RA-ADC BAY-943 as an innovative approach for the treatment of patients with IL3RA-positive AML.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/11/3464/s1, Figure S1: Drug-to-antibody ratio (DAR) of the IL3RA-ADC BAY-943, Figure S2: Apoptotic activity of the IL3RA-ADC BAY-943 in IL3RA-positive cells in vitro, Figure S3: Body weights in the systemic MOLM-13 and MV-4-11 leukemia xenograft models, Figure S4: Antitumor efficacy of the IL3RA-ADC BAY-943 in the subcutaneous MOLM-13 human AML and MV-4-11 human biphenotypic leukemia xenograft models, Figure S5: IL3RA expression in the patient-derived AML11655 and AM7577 AML xenografts models and cell line-derived AML and HL models, Figure S6: Time course of relative body weight changes in mouse models, Table S1: Characteristics of the in vivo mouse xenograft models.

**Author Contributions:** Conceptualization, A.S., D.K., H.-G.L. and B.S.-L.; methodology, D.K., A.M.W., B.S.-L. and H.-G.L.; software, M.E., C.M. and S.G.; validation, H.-G.L., R.Z., S.J. and P.B.; investigation, D.K., B.S.-L., H.-G.L., A.M.W., O.v.A., P.B., S.M., L.D., M.E., R.Z., S.J., S.G., C.M. and A.S.; resources, S.M. and H.-G.L.; data curation, A.S. and L.D.; writing—original draft preparation, A.S. and D.K.; writing—review and editing, A.S. and D.K.; visualization, B.S.-L., A.M.W., D.K., O.v.A. and A.S.; supervision, D.K. and D.M.; project administration, A.S.; funding acquisition, D.M. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** Birgit Albrecht, Susanne Bendix, Henryk Bubik, Anna DiBetta, Charlene Döring, Lisa Ehresmann, Claudia Gerressen, Nils Guthof, Beate König, Michael Krzemien, Katja Kauffeldt, Petra Leidenfrost, Stefanie Mai, Bettina Muchow, Christine Nieland, Volker Pickard, Maria Ritter, Holger Spiecker, Rukiye Tamm, Frank Tesche, Ulrike Uhlig, Sebastian Wertz, and Dirk Wolter are acknowledged for excellent technical support. We thank Bertolt Kreft and Lars Linden for fruitful discussions. We thank Xin Tang, Lily Tong, Yuandong Wang, and Kira Böhmer at CrownBio and Antje Siegert, Michael Becker, and Jens Hoffmann at EPO Berlin-Buch GmbH for performing the in vivo studies in AML PDX models. Aurexel Life Sciences Ltd. (www.aurexel.com) is acknowledged for medical writing and editorial support funded by Bayer AG.

**Conflicts of Interest:** All authors are current or former employees of Bayer AG and inventors on Bayer AG patent applications. Anette Sommer, Hans-Georg Lerchen, Stephan Märsch, Michael Erkelenz, and Dominik Mumberg have ownership interest as shares in Bayer AG.

#### **Abbreviations**



#### **References**


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

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

*Article*

### **An Antibody Specific for the Dog Leukocyte Antigen DR (DLA-DR) and Its Novel Methotrexate Conjugate Inhibit the Growth of Canine B Cell Lymphoma**

**Marta Lisowska <sup>1</sup> , Magdalena Milczarek <sup>2</sup> , Jarosław Ciekot <sup>2</sup> , Justyna Kutkowska <sup>2</sup> , Wojciech Hildebrand <sup>3</sup> , Andrzej Rapak 2,\* and Arkadiusz Miazek 4,5,\***


Received: 31 July 2019; Accepted: 24 September 2019; Published: 26 September 2019

**Abstract:** Canine B-cell lymphoma (CBL) is an incurable, spontaneous lymphoid malignancy constituting an accurate animal model for testing novel therapeutic strategies in human medicine. Resources of available species-specific therapeutic monoclonal antibodies (mAbs) targeting CBL are scarce. The aim of the present study was to evaluate the therapeutic potential of mAb B5, specific for the dog leukocyte antigen DR (DLA-DR) and its antibody-drug conjugate with methotrexate (B5-MTX). B5 induced caspase-dependent apoptosis of DLA-DR-expressing canine B cell lymphoma/CLBL1 and CLB70 leukemia lines, but not the GL-1 line not expressing DLA-DR. The cytotoxicity of B5-MTX to sensitive cells was further potentiated by a payload of MTX, but without any substantial off-target effects. The infusion of B5 and B5-MTX in a murine model of disseminated, advanced canine lymphoma, mediated >80% and >90% improvement in survival, respectively, and was well tolerated by the animals. Interestingly, the concentrations of soluble DLA-DR (sDLA-DR) antigens present in the blood serum of tumor-bearing mice were found proportional to the tumor burden. On this basis, sDLA-DR levels were evaluated as a potential biomarker using samples from canine lymphoma patients. In summary, the action of B5 and B5-MTX holds promise for further development as an alternative/complementary option for the diagnosis and treatment of canine lymphoma.

**Keywords:** passive immunotherapy; canine B-cell lymphoma; DLA-DR; HLA-DR; antibody-drug conjugate; ADC; methotrexate

#### **1. Introduction**

Canine B cell lymphoma (CBL) is a spontaneous malignancy bearing numerous molecular, histopathological and clinical similarities to human non-Hodgkin lymphoma (NHL) [1]. For this reason, dogs are considered an important animal model for pre-clinical testing of new therapies for human lymphoma [2,3]. CBL is the most frequent hematological malignancy with various histopathological presentations and accounts for over 60% of all diagnosed lymphoma cases in dogs [4]. Around

16,000 to 80,000 dogs owned in the United States alone suffer from hematological malignancies annually [5,6]. Current clinical management of CBL involves combination chemotherapy, but in contrast to human regimens, it employs lower doses of cytostatics and lacks biologicals. Relapses of the disease are usually observed within 10–14 months post-treatment, with less than 25% of dogs surviving two years [2].

The use of a therapeutic anti-CD20 monoclonal antibody (rituximab) has greatly ameliorated NHL treatment, but direct application of rituximab for CBL treatment is impossible due to the lack of amino acid sequence conservation between human and canine CD20 [7]. Efforts to raise therapeutic monoclonal antibodies to canine CD20 have resulted in the development of several candidate reagents [8–10]. Of those, 1E4 and its caninized derivatives showed a therapeutic effect against the CLBL1 canine lymphoma cell line in vitro and in vivo [11].

Major histocompatibility antigen class II antigen DR (MHC II DR) is an attractive and alternative target to CD20 for passive immunotherapy of NHL and CBL [12]. MHC II DR is highly expressed by B cell neoplasms in humans and dogs with mean cell surface levels exceeding those of CD20 [13,14]. Research on mAbs targeting MHC II DR dates back to 1987, when eradication of murine lymphoma in a syngeneic model stimulated the development of similar human strategies [15]. However, a record of limited success in clinical trials and safety concerns related to immune toxicity raised some doubts about further exploration of these therapeutic mAbs [16]. The renaissance of interest in therapeutic HLA-DR targeting came with the characterization of the humanized murine L243 antibody, named IMMU-114, which is specific for a monomorphic determinant on the HLA-DR alpha chain [12]. In preclinical trials, it demonstrated a better efficacy in killing various hematological malignancies than CD20. Moreover, it displayed a promising efficacy in phase I clinical trial in relapsed or refractory NHL and in chronic lymphocytic leukemia (CLL) [17]. With the advent of antibody-drug conjugate (ADC) technology, IMMU-114 has recently been modified to carry a payload of an active irinotecan metabolite. With this payload, therapeutic effects were observed in preclinical models of IMMU-114-resistant tumors such as acute myeloblastic leukemia and malignant melanoma [18]. A good safety profile of IMMU-114 has been reported both in human and canine patients [14]. However, due to the limited cross-reactivity of IMMU-114 with canine DLA-DR [19], a thorough assessment of the full therapeutic potential of this target in dogs is difficult.

In the search for species-specific mAbs for therapeutic targeting of DLA-DR, we have developed a murine mAb, B5. This antibody binds strongly to a conformational epitope of canine DR alpha chain (DLA-DRα), but shows only minimal cross-reactivity with HLA-DRα. We have previously shown that B5 exerts immune-dependent and direct cytotoxic effects in vitro [19].

Here, we extend these studies and report on the generation and pre-clinical testing of the novel B5 ADC with methotrexate (B5-MTX). Methotrexate (MTX) is an inexpensive and pharmacologically well-characterized antimetabolite drug [20]. Canine lymphoma cell lines were found to be 10 times more sensitive to MTX (IC50 values of 2–3 nM) than the human Raji B cell lymphoma and Jurkat T-ALL cell lines [21]. At high doses, MTX is still used in combination with rituximab to treat Burkitt's lymphoma and primary mediastinal B-cell lymphoma [22]. Conjugation of mAbs with MTX changes their mode of entry to target cells [23]. It has been shown in several studies that the MTX payload increases the rate of tumor cell inhibition due to rapid conjugate uptake and an increase in sensitivity to direct cytotoxic effects of therapeutic mAbs [23]. We hypothesized that conjugating mAb B5 with MTX can exert an additive effect of both components and contribute to a better cytotoxicity profile of this ADC against CBL.

We report as well on a novel enzyme-linked immunosorbent assay (ELISA) for the detection of circulating soluble DLA-DR (sDLA-DR) complexes in the blood serum of dogs. Physiologically, soluble circulating MHC II molecules (sMHC II) loaded with self-peptides contribute to the maintenance of self-tolerance [24]. They can be released from antigen-presenting cells or tumor B cells as well and suppress T cell immune-surveillance by directly competing with membrane-bound MHC II ligands [24]. In this report, we aimed at testing the hypothesis that blood serum levels of sDLA-DR could be

indicative of tumor burden. We present observations supporting sDLA-DR as a potentially useful biomarker for monitoring the outcome of CBL chemotherapy. Overall, our data indicate the potential therapeutic and diagnostic value of anti-DLA-DR-specific antibodies in CBL.

#### **2. Results**

#### *2.1. Characterization of the B5-MTX Conjugate*

Crosslinking of MTX anhydride with mAb B5 resulted in >94% homogenous preparation of the B5-MTX antibody-drug conjugate (Figure 1A). Size-exclusion HPLC analysis of unmodified mAb B5 versus B5-MTX revealed a delay in retention times (tr = 25.05 versus tr = 26.3), which was due to the high average drug to antibody substitution ratio (DAR), estimated at 42.6 (Figure 1A). The visible second peak at retention time tr = 46.317 min corresponded to free MTX dissociated from the conjugate. The resulting B5-MTX conjugate demonstrated approximately 49% loss of target binding activity in comparison to unconjugated B5 and had negligible nonspecific binding activity (Figure 1B).

**Figure 1.** Synthesis of B5-MTX ADC. (**A**) Size-exclusion HPLC of unmodified mAb B5 (top), detected at A280 nm and B5-MTX conjugate (bottom) detected at A372 nm (peak elution absorbance is given in absorbance units—AU) with a molar ratio of MTX to mAb (DAR) of 42.6. The conjugate was >94% pure and monomeric. Retention time (tr) difference of free mAb and B5-MTX resulted from the high substitution rate. (**B**) Flow cytometry assessment of B5 and B5-MTX staining intensity (MFI-mean fluorescence intensity) in DLA-DR expressing CLBL1 and non-expressing GL-1 cell lines. Isotype control mouse IgG2a antibody was used at the concentration of 100ng/mL (grey histogram). Color histograms correspond to signal intensities obtained with the indicated concentrations of antibody or conjugate in [ng/mL].

#### *2.2. Cytotoxicity of B5 and B5-MTX Against Canine Lymphoma*/*Leukemia Lines In Vitro*

To evaluate the cytotoxicity of B5 and B5-MTX, canine B cell lymphoma/leukemia cell lines expressing DLA-DR (CLBL1 and CLB70) and not expressing DLA-DR (GL-1) were exposed to 2 µg/mL of both preparations for 48 h. As previously reported [19], several hallmarks of direct apoptotic cell death, including caspase 3/7 activation, annexin V binding, DNA fragmentation (subG1 DNA content), were observed upon B5 treatment of DLA-DR expressing cell lines (Supplementary Figure S1). In comparison with B5, the B5-MTX conjugate exerted more potency, but also cell-specific cytotoxic effects at the same time, as no significant increase in toxicity against the DLA-DR non-expressing GL-1 cell line was detected. The average percentages of cell death induced by several tested concentrations (0.1–10 µg/mL) of B5 and B5-MTX were used to calculate IC50 and maximum inhibition values of both preparations. B5-MTX showed a higher maximum cytotoxicity (85% to 88% versus 65% to 69%) and lower IC50 values (5–6.25 nM versus 9.53–11.5 nM) against the CLBL1 and CLB70 cell lines than B5 (Table 1).


**Table 1.** In vitro cytotoxicity of mAb B5 and B5-MTX in canine lymphoma/leukemia cell lines.

<sup>1</sup> ADC concentration is given as antibody concentration; n.d.—not determined.

Reportedly, MHC II cross-linking can trigger either caspase-dependent or caspase-independent cell death mechanisms [12]. In order to determine whether B5- and B5-MTX-induced apoptosis was caspase-dependent, cell lines indicated in Figure 2 were pre-incubated with a pan-caspase inhibitor, ZVAD, prior to treatment with individual antibody preparations. On average, a 50% decrease in caspase 3/7 activating cells was detected after ZVAD pretreatment of CLBL1 and CLB70 cells, but without unspecific effects on the GL-1 cell line. Inhibition of apoptosis by ZVAD was minimally lower for the CLB70 cell line treated with B5-MTX than for the similarly treated CLBL1 cell line.

**Figure 2.** Assessment of caspase 3/7 activation after treatment of individual canine cell lines with B5 and B5-MTX in the presence or absence of a pan-caspase inhibitor (ZVAD). The average percentages of caspase 3/7-activating cells with ±SD were calculated from at least two independent experiments. Every sample was assessed in triplicate.

Together, these data indicated that both B5 and B5-MTX displayed a potent caspase-dependent anti-tumor activity against the CLBL1 and CLB70 cell lines, but not against the GL-1 line in vitro. The MTX payload carried by the B5-MTX conjugate enhanced the specific cytotoxicity compared to B5 alone. The ability of a pan-caspase inhibitor, ZVAD, to strongly interfere with B5- and B5-MTX-induced cell killing supported the caspase-dependent mechanism.

#### *2.3. Therapeutic E*ffi*cacy of B5 and B5-MTX in NOD-SCID Mice Xenotransplanted with the CLBL1-Luc Cell Line*

To evaluate the efficacy of B5 and B5-MTX treatment in vivo, we established a disseminated disease model in which 1 × 10<sup>7</sup> cells of the luciferase-expressing CLBL1-derived cell line (CLBL1-Luc) were injected intravenously into immune-deficient NOD-SCID mice. A total of 42 animals were randomly assigned to five groups treated as follows: PBS (*n* = 8), IgG (*n* = 8), MTX (*n* = 8), B5 (*n* = 10), B5-MTX (*n* = 8). Four days after the CLBL1-Luc injection, mice were treated three times a week. On day 15, all mice in PBS, IgG and MTX treatments were sacrificed because of weight loss and signs of morbidity. Five randomly selected mice from the B5 group and five mice from the B5-MTX group, showing no visible symptoms of health deterioration, were sacrificed along with control mice. Blood and organs from sacrificed mice were further analyzed as described below. The remaining animals were treated with B5 and B5-MTX until day 20 and sacrificed once their weight loss exceeded 15% or when they became moribund. On day 15 post-CLBL1-Luc injection, bioluminescence imaging was performed to assess tumor burden. Foci of intense tumor growth in the groups of control mice were located mostly in hind limb bones and in some distant organs. In B5 and B5-MTX treated mice, tumor growth was only localized in the hind limbs of certain mice, whereas other mice had virtually no signs of localized tumor growth (Figure 3A). The intensities of individual bioluminescence measurements for each mouse are plotted in Figure 3B. Groups treated with B5 and B5-MTX presented at least 20 times lower signal intensities than controls.

**Figure 3.** In vivo imaging and quantification of bioluminescence in CLBL1-Luc tumor-bearing mice on day 15 after tumor cell transplantation. (**A**) Control mice were infused with phosphate-buffered saline (Control), isotype-matched mouse IgG immunoglobulin (IgG), free methotrexate (MTX) or were treated with mAb B5 and B5-MTX. Bioluminescence intensity is presented on pseudo-color scales. (†) The IgG group contained seven mice because one mouse was found dead on the day preceding bioluminescence imaging (B) Individual bioluminescence intensities of each mouse were plotted. Statistically significant differences between the groups were marked with an asterisk (\*\*\* *p* < 0.001, \*\* *p* < 0.01).

We sought for more sensitive methods than bioluminescence to quantify tumor cell burden in bone marrows and other tissues of CLBL1-Luc-infused mice. For this purpose, Western blotting and flow cytometry with an anti-pan-DLA-DR antibody, E11 [19], was used. This antibody was chosen because it recognized a different epitope of DLA-DR than B5 and therefore did not interfere with mAbs infused for therapeutic treatment. Tumor burden in tested organs of control, but not B5- and B5-MTX-treated mice, was demonstrated by Western blotting. Specific bands corresponding to DLA-DR were present in all tested tissues except for peripheral blood mononuclear cells, and much weaker bands were found in brains (Figure 4A and Supplementary Figure S2). In bone marrows of control mice (PBS, IgG, MTX) sacrificed on day 15, CLBL1-Luc cell content exceeded 40%, but was less than 10% in the B5- and B5-MTX-treated groups (Figure 4B).

**Figure 4.** Analysis of tumor cell spread and body weight loss in CLBL-Luc tumor-bearing mice. (**A**) Protein lysates were obtained from the following organs/tissues of tumor-bearing mice: peripheral blood mononuclear cells (PBMC), bone marrow (BM), spleen (SPL), liver (LIV), lung (LUN), brain (BRA). Admixture of CLBL1-Luc cells in the above organs/tissues, reflecting tumor burden, was evaluated by Western blotting with anti-DLA-DR antibody. Protein loading was controlled with an anti-beta actin antibody (β-ACT) (**B**) Cell suspensions of bone marrows were prepared from mice treated with PBS (*n* = 8), IgG (*n* = 7), MTX (*n* = 8), B5 (*n* = 5) and B5-MTX (*n* = 5) and assessed for CLBL1-Luc cell content by flow cytometry with an anti-DLA-DR antibody, E11. (**C**) The average body weights of mice from the indicated experimental groups (*n* = 8 mice per group except for B5, *n* = 10 mice per group) were plotted. Statistically significant differences between the groups were marked with an asterisk (\*\*\* *p* < 0.001, \*\* *p* < 0.01).

All control mice succumbed to the tumor by day 15 post CLBL1-Luc cell injection, whereas mice treated with B5 and B5-MTX experienced an >80% (22.0 ± 2.45 days versus 13.5 ± 0.86 days, *p* < 0.001) and >90% (28.0 ± 8.49 days versus 13.5 ± 0.86 days, *p* < 0.05) delay in time to tumor progression, respectively (Table 2) (TTP parameter is defined in the Materials and Methods section). B5 and B5-MTX treatment was well tolerated by the animals because neither evidence of significant weight loss resulting from off-target toxicity (Figure 4C) nor blood parameter changes (Supplementary Table S1) were noted.


**Table 2.** Time to tumor progression (TTP) for CLBL1-Luc bearing mice treated with B5 and B5-MTX.

%PR—percentage of response to treatment, TF—number of tumor-free animals after day 40, *p*—probability, N.A.—not applicable, n.s.—not significant.

#### *2.4. Detection of Soluble, Circulating DLA-DR Complexes with A B5-Based Immunoassay*

Based on a published report [25], we have hypothesized that canine B cell neoplasms can release soluble DLA-DR molecules in quantities proportional to the tumor burden. Monitoring of soluble DLA-DR levels in the blood serum of CLBL1-Luc bearing mice revealed that B5- and B5-MTX-treated groups had statistically significantly lower values than the control groups (except for the difference between the PBS and B5 groups) (Figure 5A). Therefore, we asked if differences in soluble DLA-DR levels would apply to dogs diagnosed with CBL and subjected to chemotherapy as well. To this end, the blood serum of 18 healthy control dogs, 13 dogs diagnosed with B cell lymphoma (CBL group) and 10 dogs subjected to chemotherapy during remission (CBL + CHOP) was assessed for serum sDLA-DR levels. Detailed clinical data of canine patients whose blood was used in the present study is given in Supplementary Table S2. Analysis of variance showed significant differences between the control group and the CBL group (*p* < 0.05) and between the CBL group and the CBL + CHOP group (*p* < 0.01) (Figure 5B). To further determine immunoassay performance, we analyzed sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) using receiver operating characteristic (ROC) analysis; two separate sets of data were analyzed. First, we sought to determine whether elevated serum sDLA-DR levels could be predictive of CBL. The results shown in Figure 6A suggest that this parameter had a strong positive predictive value of 92%, but at the same time it had a relatively low negative predictive value of 56%, and the area under the curve was 0.835. The set of data in Figure 6B was evaluated to see whether the decrease in sDLA-DR level could be used as a biomarker for successful response to chemotherapy. As shown, 100% of PPV and NPV parameters and the AUC value equal to 1 indicated that this test could reliably predict the response to chemotherapy. However, since the sample size was not pre-defined and no paired blood samples from canine patients before and after chemotherapy were available for this analysis, the above results have to be treated with caution and as preliminary.

**Figure 5.** Assessment of soluble DLA-DR levels in blood sera of tumor-bearing NOD-SCID mice (**A**) and healthy dogs or canine lymphoma patients (**B**). Average serum levels (Abs) of DLA-DR in mice in the indicated experimental groups and healthy (control) or diseased dogs upon admission to the veterinary clinic (CBL) or after chemotherapy (CLB + CHOP) were plotted. Statistically significant differences between the groups were marked with asterisk \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

**Figure 6.** Evaluation of the diagnostic potential of soluble DLA-DR levels in canine CBL patients versus healthy control dogs (**A**), and in CBL patients versus post-chemotherapy patients (**B**). Receiver operating characteristics analysis was used to determine sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), area under the curve (AUC) value and cut off (CUT OFF) value for the soluble DLA-DR immunoassay.

#### **3. Discussion**

Targeted delivery of cytotoxic drugs using ADC technology improves their therapeutic window and minimizes chemo-associated side effects [26–28]. Methotrexate (MTX), a first-generation anti-folate chemotherapeutic with a narrow therapeutic window, is clinically approved for the treatment of multiple neoplasms [20,29]. It is also one of the very few chemotherapeutics with a fully known clinical profile in human and canine patients [30,31]. Relatively low potency of MTX as a payload can be ameliorated by increasing the drug to antibody ratio (DAR), while maintaining acceptable biological activity of the mAb. Reported DAR values for MTX-based ADCs depend on available lysyl and arginyl side chains in antibody molecules and range from 10 to 14.6 [23,32]. In the present work, an even higher DAR value was obtained by MTX anhydride crosslinking reaction. Despite the considerable loss of binding activity of the B5-MTX conjugate to DLA-DR, in vitro and in vivo data confirmed an increase in specific cytotoxicity against lymphoma cells expressing the target antigen. This was achieved without any substantial unspecific cytotoxicity towards the DLA-DR negative GL-1 cell line. In vivo data indicated that the B5-MTX conjugate not only showed promising anti-tumor activity in a model of advanced, disseminated lymphoma at a relatively low dose (2.5 mg/kg body weight), but had a good safety profile as well.

Signaling through anti-HLA-DR mAbs in tumor B cells activates multiple, pro-survival and pro-apoptotic pathways, but ultimately leads to cell death [12,33,34]. In the current work, we determined that apoptosis induced by B5 and B5-MTX in canine lymphoma/leukemia cell lines followed the intrinsic, caspase-dependent pathway that could partly be inhibited by ZVAD. Despite decades of clinical use, the precise mechanism of MTX cytotoxicity remains largely unknown. Available data on biological effects of MTX released from ADCs indicate a mechanism of cell sensitization [23]. Our data further support these observations, because many hallmarks of apoptotic cell death typical of B5 treatment (e.g., annexin-V binding levels and sub-G1 DNA levels shown in Supplementary Figure S1) were amplified in the case of B5-MTX.

In our hands, the CLBL1 cell line—a canine model of diffuse large B cell lymphoma (DLBCL)—was equally sensitive to the cytotoxic action of B5 and B5-MTX as the CLB70 cell line, with characteristics of chronic lymphocytic leukemia (CLL). This is in line with the observations reported by Stein et al. in models of human DLBLC and CLL cell lines treated with a humanized anti-HLA-DR antibody, IMMU-114 [12]. In this context, evolutionary conservation of death signaling pathways between human and canine hematological malignances opens interesting possibilities for comparative studies.

Both negative and positive associations of soluble HLA-DR levels in the blood serum were reported for patients with malignant melanoma and non-Hodgkin lymphoma, respectively [25,35]. Various strategies of tumor survival could account for the observed variations in sHLA-DR. We speculate that, on the one hand, melanoma cells could down-modulate HLA-DR expression in order to counteract recognition by tumor-specific effector T cells. On the other hand, a massive release of sHLA-DR by B cell neoplasms might induce tolerogenic T cell responses, which could weaken tumor immune-surveillance [24,36]. In order to assess the levels of soluble, circulating DLA-DR molecules (sDLA-DR) in the blood of tumor-bearing NOD-SCID mice undergoing experimental therapy with B5 and B5-MTX, we devised an immune-enzymatic assay based on the use of two species-specific mAbs recognizing different and non-overlapping epitopes of DLA-DR (B5 and E11) [19]. Our results strongly suggested that there was a direct correlation between tumor burden and the serum sDLA-DR levels. We could extend these observations to groups of unrelated canine patients suffering from CBL at the time of diagnosis and during remission. Although these observations offered a possibility of creating tools to help monitor the course of CBL chemotherapy in dogs, more samples, preferentially paired, from patients undergoing chemotherapy are required to fully validate this immunoassay.

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

#### *4.1. B5-MTX Conjugate Synthesis*

The B5-MTX conjugate was prepared using the method described by Goszczynski et al. [37] Briefly, 1 mg of mAb B5 in bicarbonate buffer pH 8.3 was mixed with MTX anhydride (50-molar excess). The reaction was allowed to proceed for 5 min. Next, the conjugate was separated on a Dionex Ultimate 3000 apparatus equipped (ThermoScientific, Waltham, MA, USA) with a four-component pump, autosampler with a fraction collection and a diode detector using the Superdex 200 10/300 GL resin ((GE Healthcare, Uppsala, Sweden). Isocratic elution was used with 0.1 M sodium bicarbonate at a flow rate of 0.5 mL/min. MTX concentration in conjugate was determined spectrophotometrically using detection at 280 and 372 nm as described by Ciekot, J. et al. [38].

#### *4.2. Cell Lines*

The CLBL1 cell line [39] was kindly provided by Dr. Barbara Rutgen (Veterinary University of Vienna, Vienna, Austria). CLB70 [40] was established by us. GL-1 [41] was kindly provided by Drs Y. Fujino and H.Tsujimoto (University of Tokyo, Tokyo, Japan). All cell lines were cultured in RPMI with 15% FBS. The stable luciferase-expressing CLBL1 cell line was generated using premade lentiviral particles (Amsbio LVP434, Abingdon, UK). Lentivirus particles were admixed with cells (1 × 10<sup>6</sup> /mL) at a ratio of 50 µL virus per 0.5 mL of cells. 24 h after transduction, cell culture medium was supplemented with puromycin sulfate (1.5 µg/mL). After one week of antibiotic selection, cell luminescence was validated after D-luciferin addition using a benchtop luminometer (Turner designs, TD-20/20, San Jose, CA, USA).

#### *4.3. In Vitro Cytotoxicity Assays*

For in vitro cytotoxicity assays, 1.25 × 10<sup>5</sup> cells were seeded on a 24-well plate. The cells were preincubated for 2 h with 20 µM ZVAD. Then, 2 µg/mL of B5 or B5-MTX were added to the cells. Cytotoxicity analysis was performed after 48 h of incubation with a CellEvent™ Caspase-3/7 Green Flow Cytometry Assay Kit (Thermo Fischer Scientific, Waltham, MA, USA), according to manufacturer's instructions. Samples were analyzed with a FACSCalibur flow cytometer (Beckton Dickinson, Franklin Lakes, NJ, USA).

To calculate IC50 values, cell lines were exposed to several concentrations of B5 and B5-MTX ranging from 0.1 to 10 µg/mL. Cell viability was determined after 48 h with propidium iodide using flow cytometry. IC50 calculation was performed with an internet tool: MLA—"Quest Graph™ IC50 Calculator." AAT Bioquest, Inc, 25 July, 2019, https://www.aatbio.com/tools/ic50-calculator. Maximal inhibition was determined by propidium iodide staining after 24-h incubation with 10 µg/mL of B5 or B5-MTX.

#### *4.4. In Vivo Monitoring of Anti-Tumor E*ff*ects of B5 and B5-MTX*

NOD/SCID (NOD.CB17-Prkdcscid/NCrCrl) mice were purchased from Charles River. Mice were housed in IVC cages with a standard sterilized rodent diet and water ad libitum. All experiments using living animals were performed under permission number 117/2017 and 012/2019 from the Local Ethics Committee in Wroclaw (Poland). The anti-tumor activity of the mAb B5 and B5-MTX conjugate was assessed in vivo based on their effect on the growth of CLBL1-Luc cells transplanted intravenously into NOD/SCID mice. Forty-two mice bearing CLBL-1-Luc cells (1 × 10<sup>7</sup> cells/mouse) were randomly divided into five groups (10 mice in the B5 group and eight animals in other groups). The B5-MTX conjugate, mAb B5, control, isotype matched IgG, MTX alone (0.25 mg/kg body weight) and phosphate-buffered saline (PBS) were administered intra-peritoneally on day 4 post CLBL1-Luc transplantation and repeated every two days. On day 15, all mice from the PBS, IgG and MTX groups were sacrificed because of signs of morbidity along with randomly selected five mice from the B5 group and five mice from the B5-MTX group that did not present any visible signs of disease (no weight loss nor behavioral changes). The remaining five mice from the B5 group and three mice from the B5-MTX group were treated until day 20. B5-MTX and mAb B5 were injected intraperitoneally at a dose of 2.5 mg per kg of body weight. The location of CLBL-1-Luc cells was visualized on day 14 after transplantation using an In-Vivo MS FX PRO system (Bruker INC., Billerica, MA, USA). Twenty minutes before imaging, D-luciferin potassium salt (Synchem, Felsberg, Germany) was administered to each mouse intraperitoneally at a dose of 150 mg/kg. Subsequently, animals were anesthetized with a 3% to 5% (v/v) mixture of isoflurane (Forane, Abbott Laboratories, Lake Bluff, IL, USA) in synthetic air (200 mL/min). Anesthesia was maintained by means of individual masks providing a 1.5% to 2% (v/v) mixture of isoflurane and synthetic air. Visualization was carried out using the following settings: for X-Ray t = 60 s., f-stop = 2.50, FOV = 200.0; for luminescence capture t = 4 min, binning 2 × 2, f-stop = 2.50, FOV = 200.0. Images were analyzed using Bruker MI software (Bruker INC., USA). The intensity of the luminescent signal is presented as the net intensity of the region of interest and expressed in arbitrary units [a.u.]. Time to progression (TTP) was defined as a day when either body weight loss exceeding 15% or morbidity or limb paralysis were noticed by the operator in any individual mouse.

For Western blotting analysis, mouse organs (brain, liver, lungs, bone marrow, PBMC, spleen) were suspended in a lysis buffer (20 mM Tris-HCl pH 7.5, 50 mM NaCl, 0.5% NP-40 and protease inhibitor set), and sonicated for 10 s on ice. The suspensions were centrifuged at 10,000 rpm at 4 ◦C for 10 min. Then, non-reducing SDS sample buffer was added to the supernatants and the samples were subjected to a 12% SDS-PAGE gel. The separated proteins were transferred onto a PVDF membrane (Millipore, Burlington, MA, USA) using semi-dry transfer. After transfer, the membrane was blocked with 1% casein in TBS at 4 ◦C, overnight, and subsequently incubated with 1 µg/mL primary antibody: mab E11 and anti-actin (C-4) (Santa Cruz Biotechnology, Santa Cruz, CAUSA) at room temperature for

1 h, followed by secondary horseradish peroxidase-labeled antibody (DAKO, Agilent, Santa Clara, CA, USA). Bound antibodies were visualized using the ECL blotting detection system (Thermo Fischer Scientific, USA).

#### *4.5. Detection of Soluble DLA-DR Levels in the Blood Serum of Mice and Dogs*

Remaining blood serum samples from 41 dogs (18 controls, 13 lymphomas and 10 during CHOP therapy), referred to the "NeoVet" veterinary clinic for periodic blood checks, and blood sera of NOD-SCID mice bearing CLBL1-Luc tumors were used for the detection of soluble DLA-DR levels. In accordance with the provisions of the Act of January 15, 2015, item 266 on the protection of animals used for scientific and educational purposes, the use of blood samples of dogs for clinical veterinary research does not require the consent of local ethics committees.

96 well plates (Nunc) were coated with B5 mAb in PBS (1 µg/mL) overnight at 4 ◦C. On the next day, the plates were blocked with 5% non-fat milk for 1 h at room temperature (RT), then 20 times diluted blood sera of canine tumor-bearing mice or canine patients were incubated in a 0.5% milk solution at RT for 1 h. Next, biotinylated mAb E11 (1 µg/mL) was added to the solution and incubated for 1 h at RT, followed by the incubation with the Streptavidin-HRP conjugate (1:20,000); after final washes, the 3.3′5.5′ -tetramethylbenzidine substrate (Sigma, St. Louis, MO, USA) was added for a 15 to 20-min incubation. The reaction was stopped with 1 N H2SO4. The absorbance was measured at 450 nm on a Wallac Victor plate reader (Perkin Elmer, Waltham, MA, USA). Each sample was prepared in triplicate.

#### *4.6. Statistical Analysis*

Statistical analysis of in vivo bioluminescence imaging was performed using the non-parametric Kruskal-Wallis test, followed by Dunn's multiple comparison tests (\*\* *p* < 0.01, \*\*\* *p* < 0.001). Statistical analysis of all other in vitro assays was performed using one-way ANOVA with Tukey's Multiple Comparison Test. Significance was set at \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001. Receiver operating characteristic analysis was performed by calculating the shape of the ROC curve. This was done by plotting the values of the sensitivity assay on the y-axis, and the values of false-positive rates (1-specificity) on the x-axis. The area under the curve (AUC) was subsequently calculated.

#### **5. Conclusions**

In this report, we tested a novel MTX-based ADC directed against canine lymphoma/leukemia cells. To our knowledge, this is the first pre-clinical study of an ADC designed for veterinary use. The results indicate a significant increase in the specific cytotoxicity of B5-MTX ADC against canine lymphoma/leukemia cell lines in comparison with unmodified mAb in vitro and in vivo. Unlike in humans, DLA-DR antigen, a target of B5 and B5-MTX, is expressed in dogs by both normal B and T cells and by mixed B/T neoplastic cells. Therefore, the use of anti-DLA-DR antibodies in diagnosis or therapy can cover up to 90% of all hematological malignancies in this species. Elevated sensitivity of canine lymphoma/leukemia cell lines to MTX opens up new opportunities for using this antimetabolite as a payload for therapeutic ADCs targeting DLA-DR. Despite the clearly lower potency of MTX in comparison to the second generation cytotoxic payload, such as auristatin, low price, simple conjugation chemistry and lack of intellectual property rights attached to this antimetabolite makes it an interesting option for veterinary use.

In this study, we have shown as well that the observed correlation of soluble DLA-DR levels in the blood serum of canine lymphoma-bearing immune-deficient mice can be further studied in the context of translation into a diagnostic test for monitoring the efficiency of chemotherapy in canine lymphoma.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/11/10/1438/s1, Figure S1: Analysis of B5 and B5-MTX induced apoptosis in canine cell lines, Figure S2: Cytotoxicity plots used for calculating IC50 and maximum inhibition values of B5 and B5-MTX, Figure S3: Distribution of individual weight measurement of mice used for in vivo studies, Figure S4: Original scans of Western blots with densitometry

analysis, Table S1: Haematological analysis of the whole blood of tumor-bearing NOD-SCID mice, Table S2: Clinical description of oncologic canine patients, Table S3: raw absorbance data of anti-sDLA-DR ELISA used for the ROC analysis.

**Author Contributions:** Conceptualization: A.M. and A.R.; methodology: M.L., J.C., M.M.; software: M.L., M.M., J.C.; validation: M.M., M.L. and J.K.; formal analysis: M.L., M.M., J.C., J.K., W.H.; investigation: M.L., M.M., A.R.; resources: M.L., A.R.; data curation: M.L., M.M., W.H.; writing—original draft preparation: A.M.; writing—review and editing: A.M., A.R., M.M., M.L.; visualization: A.M., M.L.; supervision: A.M.; project administration: A.R., M.L., A.M.; funding acquisition: A.R., M.L., A.M.

**Funding:** This research was funded by the National Science Center, Poland, under grant number Preludium/2016/21/N/NZ5/01942 (to M.L.); the National Centre for Research and Development, Poland, under grant number TANGO2/340428/NCBR/2017 (to A.R.); and the Wroclaw Centre of Biotechnology, The Leading National Research Centre (KNOW) program for years 2014 to 2018 (to A.M.).

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

*Communication*

### **The Risks and Benefits of Immune Checkpoint Blockade in Anti-AChR Antibody-Seropositive Non-Small Cell Lung Cancer Patients**

**Koichi Saruwatari 1,†, Ryo Sato 1,†, Shunya Nakane 2,3, Shinya Sakata <sup>1</sup> , Koutaro Takamatsu <sup>2</sup> , Takayuki Jodai <sup>1</sup> , Remi Mito <sup>1</sup> , Yuko Horio <sup>1</sup> , Sho Saeki <sup>1</sup> , Yusuke Tomita 1,\* and Takuro Sakagami <sup>1</sup>**


Received: 30 December 2018; Accepted: 21 January 2019; Published: 24 January 2019

**Abstract:** Background: Anti-programmed cell death 1 (PD-1) monoclonal antibodies (Abs) unleash an immune response to cancer. However, a disruption of the immune checkpoint function by blocking PD-1/PD-ligand 1(PD-L1) signaling may trigger myasthenia gravis (MG) as a life-threatening immune-related adverse event. MG is a neuromuscular disease and is closely associated with being positive for anti-acetylcholine receptor (anti-AChR) Abs, which are high specific and diagnostic Abs for MG. Methods: A 72-year-old man was diagnosed with chemotherapy-refractory lung squamous cell carcinoma and nivolumab was selected as the third-line regimen. We describe the first report of an anti-AChR Ab-seropositive lung cancer patient achieving a durable complete response (CR) to an anti-PD-1 antibody therapy. To further explore this case, we performed multiplex immunofluorescence analysis on a pretreatment tumor. Results: The patient achieved a durable CR without developing MG. However, the levels of anti-AChR Abs were elevated during two years of anti-PD-1 antibody therapy. The tumor of the subclinical MG patient had high PD-L1 expression and an infiltrated–inflamed tumor immune microenvironment. Conclusions: This study suggests that immune checkpoint inhibitors can be safely used and provide the benefits for advanced cancer patients with immunologically 'hot' tumor even if anti-AChR Abs are positive. Although careful monitoring clinical manifestation in consultation with neurologist is needed, immune checkpoint inhibitors should be considered as a treatment option for asymptomatic anti-AChR Ab-seropositive cancer patients.

**Keywords:** anti-PD-1 monoclonal antibodies; anti-acetylcholine receptor (AChR) antibody; B cell; immune checkpoint blockade; immune-related adverse events (irAEs); myasthenia gravis (MG); non-small-cell lung cancer (NSCLC); nivolumab; programmed cell death ligand 1 (PD-L1); T cell

#### **1. Introduction**

Monoclonal antibodies (Abs) acting against programmed cell death 1 (PD-1) such as nivolumab and pembrolizumab are a class of drugs called immune checkpoint inhibitors that inhibit the interaction between PD-1 and programmed cell death ligand 1 (PD-L1) and unleash an immune response to cancer in contrast with chemotherapies that exert direct cytotoxic effects on tumor cells. The development of immune checkpoint blockade therapy has recently led to a paradigm shift in non-small-cell lung cancer (NSCLC) treatment and dramatically changed the treatment landscape of NSCLC patients [1–3].

For patients with advanced NSCLC, the immune checkpoint inhibitors have shown significant and long-lasting clinical responses in addition to a more favorable toxicity profile and improved tolerability than chemotherapy, and is currently a standard of care [2–7]. However, a disruption of the immune checkpoint function caused by blocking PD-1/PD-L1 signaling can lead to imbalances in immune homeostasis and self-tolerance, which results in an unfavorable immune response to normal tissues, which are termed immune-related adverse events (irAEs) [8,9]. The irAEs that emerge with immune checkpoint blockade therapy share clinical features with autoimmune diseases. The irAEs are usually reversible. However, in rare cases, they can be severe and life-threatening [8,10,11]. In addition, as clinical experience with immune checkpoint inhibitors increases, unexpected severe irAEs have emerged in the real-world clinical practice [10–13]. Thus, elucidating mechanisms of irAEs is urgently needed to improve their early diagnosis and develop more precise treatments for irAEs [8,9].

Myasthenia gravis (MG) is an autoimmune neuromuscular disease that is characterized by muscle weakness and fatigue, and is closely associated with a positive result for the anti-acetylcholine receptor (AChR) antibody directed against the AChR at the neuromuscular junction [14]. Anti-PD-1/PD-L1 monoclonal Abs have been known to trigger the onset of MG as one of the life-threatening irAEs [8,9,15]. Anti-AChR Abs is high specific and diagnostic antibody for MG, and the positivity of anti-AChR Abs has been reported to align with the onset of MG as an irAE in cancer patients, which discourages clinicians from using immune checkpoint inhibitors for cancer patients with pre-existing anti-AChR Abs [8,15,16]. Although several studies highlight the severity of MG as an irAE and the risks of the use of immune checkpoint inhibitors for the cancer patients with pre-existing MG or subclinical MG (asymptomatic anti-AChR Ab-seropositive cancer patients), the benefits and safety of immune checkpoint inhibitors in asymptomatic patients with pre-existing anti-AChR Abs, have not been studied [9,14–17].

In this case, we show a case of anti-AChR Ab-seropositive NSCLC patients achieving a durable complete response (CR) to an anti-PD-1 monoclonal antibody therapy (nivolumab) without developing MG. To further explore this case, we performed multiplex immunofluorescence analysis on a pretreatment tumor sample. This study provides new insights into the use of immune checkpoint monoclonal Abs for cancer patients with pre-existing anti-AChR Abs.

#### **2. Results**

#### *2.1. An Anti-AChR Antibody-Seropositive NSCLC Patient Achieving a Durable Complete Response to an Anti-PD-1 Monoclonal Antibody without Developing MG*

A 72-year-old man was diagnosed with lung squamous cell carcinoma and had left upper lobectomy and lymph node resection (pathological T2aN2M0 stage IB, PD-L1 tumor proportion score ≥ 50%). He had a 90 pack-year history of cigarette smoking. He received S-1 monotherapy as postoperative adjuvant chemotherapy for two years. However, he was diagnosed with recurrence of lung squamous carcinoma with right cervical and mediastinal lymph node metastases. He had pulmonary metastases and enlargement of the lymph node metastases after receiving four cycles of carboplatin plus nab-paclitaxel as the first-line chemotherapy regimen and one cycle of docetaxel as the second-line chemotherapy regimen. Thus, an anti-PD-1 monoclonal antibody, nivolumab, was selected as the third-line regimen.

Screening tests for autoimmune diseases including disease-specific autoantibodies were done before administration of nivolumab. He had no history of thymic epithelial tumor and autoimmune disease. He had no symptom associated with autoimmune or neuromuscular diseases. His performance status was 0. Creatine kinase was not elevated. However, he was positive for serum anti-AChR Abs (0.8 nM, normal upper limit, 0.2 nM). The potential risks and benefits of an anti-PD-1 antibody therapy for the anti-AChR Ab-seropositive advanced NSCLC patient were carefully evaluated in consultation with neurologists. Then, after obtaining informed consent, nivolumab was administered 3 mg/kg every two weeks with careful monitoring of clinical symptoms and levels of anti-AChR Abs by neurologists. Following four cycles of nivolumab, he had hypothyroidism as an irAE and hormone replacement therapy was initiated. The common irAEs such as pyrexia, rash, interstitial pneumonia, hepatitis, and colitis were not observed. After 17 cycles of nivolumab, a fluorodeoxyglucose (FDG)-positron emission tomography-computed tomography (PET/CT) scan revealed a remarkable shrinkage of metastatic lesions of lung and lymph nodes and he achieved a CR (see Figure 1A,B). Importantly, the patient achieved a durable CR without developing MG even though the levels of anti-AChR Abs were elevated (0.8–1.80 nM) during two years of anti-PD-1 antibody therapy (Figure 1C). –

**Figure 1.** *Cont.*

**Figure 1.** Key imaging and longitudinal analysis of the levels of anti-AChR Abs in asymptomatic anti-AChR Ab-seropositive patient who had a complete response to an anti-PD-1 antibody therapy. Panel (**A**) and (**B**) show FDG-PET/CT imaging pre-nivolumab and post-nivolumab. Arrows in panel (**B**) indicate supraclavicular lymph node (upper panels) and mediastinal lymph node (lower panels) metastases. Panel (**C**) shows the longitudinal analysis of serum concentrations of anti-AChR Ab (nM) before and after nivolumab. The dashed line indicates a normal upper limit of the concentrations of anti-AChR Abs.

#### *2.2. The Tumor of Subclinical MG Patient with a Durable Complete Response to an Anti-PD-1 Antibody Therapy had an Immunologically 'Hot' Tumor Microenvironment*

To further explore this case, we investigated the immune contexture of pretreatment lung tumor of the anti-AChR Ab-seropositive NSCLC patient who achieved a CR to nivolumab by fluorescent multiplex immunohistochemistry (mIHC). The mIHC analysis has been shown to capture multidimensional data related to tissue architecture, spatial distribution of multiple cell phenotypes, and co-expression of signaling [18,19]. A high density of tumor-infiltrating CD8+ T cells and CD20+ B cells has been shown to correlate with prolonged survival in patients with a wide variety of human cancers including lung cancer [20–22]. Regulatory T cells (Tregs) have immunosuppressive activity and play a critical role in maintaining immune homeostasis and negatively regulating anti-tumor immune responses [23–25]. Thus, a pretreatment tumor sample from the patient was analyzed for tumor-infiltrating CD8+ T cells, CD20+ B cells, and Tregs (FOXP3+ CD3+ T cells) by fluorescent mIHC. Pan-cytokeratin of tumor cells and PD-L1 were simultaneously stained to evaluate the complex relationship among tissue architecture, spatial distribution of immune cells, and expression of PD-L1.

≥ PD-L1 immunohistochemistry using PD-L1 22C3 pharmDx revealed the tumor PD-L1 tumor proportion score ≥ 50% (Figure 2A,B). CD8+ T cells were infiltrated into both tumor stroma and tumor cell nests (Figure 2). CD20+ B cells were mainly localized to the tumor stroma rather than tumor cell nests and infiltrated at the invasive tumor margin (Figure 3). Tregs were infiltrated into both tumor stroma and tumor cell nests (Figure 4), but the number of tumor-infiltrating Tregs was fewer than conventional T cells (FOXP3-negative CD3+ T cells) or CD8+ T cells (Figures 2 and 4). Altogether, these results demonstrate that an anti-AChR Ab-seropositive NSCLC patient who achieved a CR to nivolumab had an infiltrated–inflamed tumor immune micro-environment and immunologically 'hot' tumor [26,27]. The immunologically 'hot' tumor micro-environment might associate with the benefits of immune checkpoint blockade therapy without developing MG.

μ μ **Figure 2.** CD8+ T cells infiltrate pretreatment lung tumor of the anti-AChR Ab-seropositive NSCLC patient who achieved a CR to nivolumab. The surgically resected tumor was analyzed by fluorescent multiplex immuno-histochemistry. Serial formalin-fixed paraffin-embedded (FFPE) sections of the tumor sample were stained with Haematoxylin and Eosin (**A**), PD-L1 IHC 22C3 pharmDx (**B**) and analyzed by fluorescent multiplex immunohistochemistry (**C**,**D**). The panel (**D**) shows the boxed region in the panel (**C**) at high magnification. CD8+ T cells (green) were infiltrated into both tumor stroma and pan-Cytokeratin positive tumor cell nests (dark yellow). The tumor expressed PD-L1 (magenta). Nuclei were stained with DAPI (blue). Scale bars, 50 µm (**A**, **B**, and **D**) and 200 µm (**C**), are shown in each panel.

μ μ **Figure 3.** CD20+ B cells infiltrate pretreatment lung tumor at the invasive tumor margin. Serial FFPE sections were stained with antibodies against CD20 (green) and pan-Cytokeratin, and analyzed by fluorescent multiplex immunohistochemistry. The right panels show the boxed regions in the left panel at high magnification. CD20+ B cells (green) were infiltrated at the invasive tumor margin rather than tumor cell nests. Scale bars, 1000 µm (left panel), or 200 µm (right panels) are shown in each panel. μ μ 

**Figure 4.** FOXP3+ CD3+ T cells (Tregs) sparsely infiltrate pretreatment lung tumor. Serial FFPE sections were stained with antibodies against CD3 (green), FOXP3 (red), and pan-Cytokeratin. The right panel show the boxed region in the left panel at high magnification. Tregs were sparsely infiltrated in this tumor tissue. Scale bars, 200 µm (left panel), and 50 µm (right panel) are shown in each panel.

#### **3. Discussion**

Anti-PD-1 monoclonal Abs block the interaction between PD-1 and its ligand, PD-L1, which unleashes the anti-tumor immune response [1,5,26]. However, the disruption of immune checkpoint signaling can lead to imbalances in immunologic tolerance and result in an unfavorable immune response, which clinically manifest as irAEs [9,11,12,28]. A unique set of inflammatory and autoimmune side effects known as irAEs was quickly recognized in clinical trials in association with the nature of immune checkpoint inhibitors impacting systemic immunity of cancer patients [9,29]. Although the common irAEs are rash, endocrinopathies, interstitial pneumonia, hepatitis, and colitis, rare but serious irAEs have been identified during post-marketing surveillance [8–11,13]. The pathophysiology underlying these irAEs has not been fully understood, which elucidates mechanisms of irAEs. This is urgently needed to improve their early diagnosis and develop more precise treatments for irAEs [8].

As the use of anti-PD-1/PD-L1 monoclonal Abs is extending to various malignancies with unprecedented speed, there is also an unmet need to identify risks and benefits of immune checkpoint blockade therapy in cancer patients with a history of autoimmune disease [8,29]. Most of the evidence regarding irAEs comes from prospective clinical trials, but cancer patients with concurrent autoimmune disease have been excluded from most of the clinical trials because of concerns that these individuals potentially have an elevated risk for developing serious irAEs. Therefore, the safety of anti-PD-1/PD-L1 monoclonal Abs in cancer patients with a history of autoimmune disease is less clear [8,29]. Recent retrospective studies of immune checkpoint blockade in patients with NSCLC and pre-existing autoimmune disease have shown that adverse events were generally manageable and infrequently led to the discontinuation of immunotherapy. The retrospective studies have also shown that anti-PD-1/PD-L1 monoclonal Abs can achieve clinical benefit in those patients. However, the risks and benefits of immune checkpoint inhibitors in asymptomatic patients with pre-existing disease-specific autoantibodies remain unclear [8,15–17,29].

In the current study, we have shown that an anti-AChR Ab-seropositive NSCLC patient achieved a durable CR to an anti-PD-1 monoclonal antibody therapy without developing MG (Figure 1). Makarious et al. showed that the specific MG-related mortality is high (30.4%) in immune checkpoint antibody therapies even though immune checkpoint inhibitor-associated MG is rare [16]. Among the 23 reported cases of irAEs manifesting as MG, 72.7% were de novo, 18.2% were pre-existing MG exacerbations, and only 9.1% (*n* = 2) were exacerbations of subclinical MG (asymptomatic anti-AChR Ab-seropositive cancer patients before administration of immune checkpoint blockade) [16]. One out of the two exacerbations of subclinical MG patients died (the mortality of exacerbations of subclinical MG, 50%). In a study of two-year safety databases based on post-marketing surveys, Suzuki et al. reported that 12 among 9869 cancer patients treated with nivolumab developled MG (0.12%). The nivolumab-induced MG was severe and two MG patients died (MG-related mortality, 17%) [15]. In this study, two cases of exacerbations of subclinical MG have been reported. These studies highlight the importance of recognizing MG as a life-threatening irAE. However, little is known about the potential benefits and the safety of immune checkpoint blockade for subclinical MG [14–16].

Understanding the complex tumor microenvironment offers the opportunity to make better prognostic evaluations and select optimum treatments [26,27,30]. Accumulating evidence suggests that a high density of tumor-infiltrating CD8+ T cells and CD20+ B cells strongly associates with positive clinical outcomes in various cancer types [20–22,31]. However, the immune contexture of anti-AChR Ab-seropositive tumor response to immune checkpoint inhibitors without developing MG remains unknown. Thus, we analyzed pretreatment tissue of the patient. Infiltrated–inflamed tumor immune micro-environments are considered to be immunologically 'hot' tumors and are characterized by high immune infiltrations including CD8+ T cells, B cells, and tumor cells expressing PD-L1 [26,27]. In the current study, the tumor of the subclinical MG patient had high PD-L1 expression and an infiltrated–inflamed tumor immune microenvironment, which suggests similar cases may respond to immune checkpoint blockade therapy without developing MG.

Although anti-PD-1/PD-L1 monoclonal Abs are selectively targeting the PD-1/PD-L1 pathway, the antibodies do not selectively target the PD-1/PD-L1 signaling between tumor antigen-specific T cells and tumor cells. Furthermore, both PD-1 and PD-L1 are expressed not only on effector CD8+ T cells called "killer T cells", but also on a variety of immune subsets including other T cell subsets and B cells [11,13,32–34]. Thus, administered anti-PD-1/PD-L1 monoclonal Abs may bind to the various non-tumor-specific immune subsets and induce the unwanted activation of the immune system, which may disturb the balance established between tolerance and autoimmunity and lead to irAEs such as MG (Figure 5).

A concept of "immune normalization" for the class of drugs called immune checkpoint inhibitors has recently been proposed [1,5]. However, immune checkpoint inhibitors do not always change the immune balance toward a favorable direction for anti-tumor immunity. MG is a B cell–mediated autoimmune disease in which the target auto-antigen is AChR at the neuromuscular junction and also has been known as one of the life-threatening irAEs associated with immune checkpoint blockade for malignancies [14–16,35]. PD-1 expresses on activated B cells as well as activated T cells [33,36,37], which indicates that there is a potential risk of triggering B cell–mediated autoimmune disease such as MG by the blockade of the interaction between PD-1 and PD-L1. The evidence suggests that blocking PD-1/PD-L1 signaling may shift the systemic immune balance from the T cell-mediated immune response (cellular immune response) to the B-cell mediated immune response (humoral immune response) [33,36,37] which enhances pre-existing anti-AChR antibody, and may lead to the onset of MG as an irAE (Figure 5A).

CD4+ T cells include T helper type 1 (Th1), which drives the cellular immune responses, and CD4<sup>+</sup> T helper 2 (Th2), which promotes humoral immune responses. Th2 cells enhance B-cell mediated immunity and promote antibody production [38,39]. PD-1 expresses on Th2 cells as well as Th1 cells and CD8+ T cells. Therefore, the blockade of PD-1/PD-L1 signaling has been shown to promote Th2 cell responses and Th2-type inflammations [13,40], which suggests that immune checkpoint blockade has the potential to modulate the balance between cellular immune response and humoral immune response and may lead to the onset of MG (Figure 5B).

**Figure 5.** *Cont.*

**Figure 5.** Underlying mechanisms of humoral immune response-associated irAEs. Panel (**A**) shows a model demonstrating the immune balance between a T cell-mediated immune response and a B cell-mediated immune response. Immune checkpoint inhibitors can activate both T cells (cellular immune response) and B cells (humoral immune response), and have the potential to modulate the balance between cellular immune response and humoral immune response, since PD-1/PD-L1 express on both T cells and B cells. Panel (**B**) shows a model demonstrating immune balance between the Th1 cell and the Th2 cell. Immune checkpoint inhibitors can activate both Th1 cells (cellular immune response) and Th2 cells (humoral immune response), and have the potential to modulate the balance between cellular immune response and humoral immune response, since PD-1/PD-L1 express on both Th1 cells and Th2 cells.

There is no evidence of the safety of anti-PD-1 Ab therapy for cancer patients who are positive for anti-AChR Abs. [15,16]. Although we demonstrated that an anti-AChR-seropositive lung cancer patient had immunologically 'hot' tumor and achieved a durable CR to an anti-PD-1 monoclonal antibody therapy without developing MG, our study could not uncover enough evidence to explain the reason why the present case did not develop MG. It is conceivable that the patient might have not been susceptible to an increased anti-AChR antibodies by chance. Thus, clinicians should be cautious to use immune checkpoint blockade for cancer patients with subclinical MG.

Because MG as irAE is life-threatening and closely associated with positive for anti-AChR Ab, the pre-existing serum anti-AChR Ab in cancer patients discourages clinicians from using immune checkpoint inhibitors [14–16]. However, the present study indicates that avoiding use of immune checkpoint inhibitors for cancer patients with subclinical MG potentially lead to losing the chance to cure advanced cancers.

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

#### *4.1. Patinet*

The Kumamoto University Institutional Review Board approved the study (IRB number, 2287, Approval Date, 23 January 2018).

#### *4.2. PD-L1 Staining*

PD-L1 expression in the lung cancer specimen was analyzed by immunohistochemical staining using the PD-L1 IHC 22C3 pharmDx antibody (clone 22C3 (Dako North America, Inc., Carpinteria, CA, USA)). The antibody was applied according to DAKO-recommended detection methods. PD-L1 expression in tumor cells was scored as the percentage of stained cells.

#### *4.3. Fluorescent Multiplex Immunohistochemistry*

Fluorescent multiplex immunohistochemistry was performed with OPAL Multiplex Fluorescent Immunohistochemistry Reagents (PerkinElmer, Waltham, MA, USA) following the manufacturer's protocol. As outlined in the Table 1, formalin-fixed paraffin-embedded (FFPE) sections were stained by one of the three sequences of primary antibodies, PD-L1, pan-Cytokeratin and CD8, pan-Cytokeratin and CD20, or pan-Cytokeratin, FOXP3, and CD3, respectively, using the tyramide signal amplification (TSA) system with Opal dye reagents. Each labeling step consisted of the following at room temperature. Sections of formalin-fixed, paraffin-embedded tissue were depleted of paraffin and were then hydrated and processed for antigen retrieval by treatment with 10 mM citrate antigen buffers (pH 6.0) via microwave radiation (except for PD-L1 which was processed by pH 9.0 citrate buffer via autoclave). The sections were incubated with 3% H2O<sup>2</sup> for 5 min to inhibit endogenous peroxidase activity, washed with 0.05% Tween in TBS (TBST), exposed to blocking buffer (5% goat serum, 0.5% bovine serum albumin in PBS) for 20 min at room temperature, and incubated for 60 min at room temperature with primary antibodies. They were then washed with TBST, incubated with anti-mouse or anti-rabbit HRP polymer conjugated secondary antibodies (Nichirei, Tokyo, Japan) for 10 min at room temperature except for PD-L1 (incubated for 30 min at room temperature), and washed again, after which immune complexes were detected with Opal reagents. Nuclei were counterstained with 4′ ,6-diamidino-2-phenylindole dihydrochloride (DAPI) (DOJINDO, Kumamoto, Japan) in water, and whole sections were mounted in ProLong Diamond (Thermo Fisher Scientific, Waltham, MA, USA). Multiplexed slides were observed with a fluorescence microscope (BZ-X700, Keyence, Osaka, Japan). The antibodies used for fluorescent multiplex immunohistochemistry analysis are listed below.


**Table 1.** The list of antibodies used for fluorescent multiplex immunohistochemistry analysis.

#### **5. Conclusions**

In conclusion, to the best of our knowledge, this is the first report of an anti-AChR antibody-seropositive cancer patient achieving a durable CR to immune checkpoint blockade therapy without developing MG. This study suggests that immune checkpoint inhibitors can be safely used and provide benefits for advanced cancer patients with an immunologically 'hot' tumor even if the anti-AChR antibody are positive. Although careful monitoring clinical manifestation in consultation with a neurologist is needed, immune checkpoint blockade therapy should be considered as a treatment option for asymptomatic anti-AChR Ab-seropositive cancer patients. This study not only provides new insights into the use of immune checkpoint monoclonal Abs for cancer patients with pre-existing disease-specific auto-antibodies, but also may improve our understanding of the pathophysiology underlying irAEs and MG.

**Author Contributions:** Conception and design: K.S., R.S., S.N., S.S. (Shinya Sakata), S.S. (Sho Saeki), and Y.T. Acquisition of clinical data and patient care: K.S., R.S., S.N., S.S. (Shinya Sakata), K.T., T.J., R.M., Y.H., S.S. (Sho Saeki), T.S., and Y.T. Acquisition, analysis, and interpretation of biological data: K.S., R.S., S.N., S.S. (Shinya Sakata), K.T., and Y.T. Writing, review, and/or revision of the manuscript: K.S., R.S., S.N., Y.H., T.S., and Y.T. Study supervision: K.S., R.S., S.N., T.S., and Y.T.

**Funding:** This work was supported by the Takeda Science Foundation and JSPS KAKENHI Grant Number JP18K15928.

**Acknowledgments:** We thank the Departments of Thoracic Surgery and Department of Respiratory Medicine, National Hospital Organization Kumamoto Minami Hospital, for their assistance in obtaining the tissue from the patient and thank Misako Takahashi (Department of Respiratory Medicine, Kumamoto University) for technical assistance in fluorescent multiplex immunohistochemistry analysis. We are very grateful to our patient for his participation in this study.

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Structure and Optimization of Checkpoint Inhibitors**

### **Sarah L. Picardo 1,\*, Je**ff**rey Doi <sup>2</sup> and Aaron R. Hansen <sup>1</sup>**


Received: 20 November 2019; Accepted: 16 December 2019; Published: 21 December 2019 -

**Abstract:** With the advent of checkpoint inhibitor treatment for various cancer types, the optimization of drug selection, pharmacokinetics and biomarker assays is an urgent and as yet unresolved dilemma for clinicians, pharmaceutical companies and researchers. Drugs which inhibit cytotoxic T-lymphocyte associated protein-4 (CTLA-4), such as ipilimumab and tremelimumab, programmed cell death protein-1 (PD-1), such as nivolumab and pembrolizumab, and programmed cell death ligand-1 (PD-L1), such as atezolizumab, durvalumab and avelumab, each appear to have varying pharmacokinetics and clinical activity in different cancer types. Each drug differs in terms of dosing, which becomes an issue when drug comparisons are attempted. Here, we examine the various checkpoint inhibitors currently used and in development. We discuss the antibodies and their protein targets, their pharmacokinetics as measured in various tumor types, and their binding affinities to their respective antigens. We also examine the various dosing regimens for these drugs and how they differ. Finally, we examine new developments and methods to optimize delivery and efficacy in the field of checkpoint inhibitors, including non-fucosylation, prodrug formations, bispecific antibodies, and newer small molecule and peptide checkpoint inhibitors.

**Keywords:** checkpoint inhibitors 1; protein structure 2; pharmacokinetics 3; drug optimization 4

#### **1. Introduction**

Checkpoint inhibitors (CPIs) induce an anti-tumor immune response by antagonizing suppressive immune checkpoint regulatory pathways. The recognized function of these immune checkpoints is to modulate or prevent autoimmune responses and or auto-inflammation. The advent of antibodies targeting programmed cell death protein-1 (PD-1), programmed cell death protein ligand-1 (PD-L1) and cytotoxic T-lymphocyte associated protein-4 (CTLA-4) has led to the development of drugs targeting these pathways in the last 10 years. However, their variable pharmacokinetics and response rates has led to efforts to optimize these drugs, as well as to develop new drugs targeting other checkpoint pathways. Here we examine the structure and mechanism of action of these drugs and human pharmacokinetics in terms of their binding affinities, clearance, and the significance of dosing regimens. In addition, we describe efforts to enhance the delivery and formulation of CPIs, while attempting to minimize the immune-related adverse events (irAEs) associated with these treatments.

#### **2. CTLA-4, PD-1 and PD-L1 Proteins and Antibodies**

#### *2.1. Proteins*

#### 2.1.1. CTLA-4

CTLA-4 was first described in 1987 as "a new member of the immunoglobulin superfamily" [1]. It is a 223 amino acid protein which is expressed on activated T cells co-expressing CD28 [2] and has extracellular, transmembrane and intracellular components. Its ligands are CD80 (B7-1) and CD86 (B7-2), found on antigen presenting cells and T-regulatory (T-reg) cells, with binding causing downregulation of activated T cell activity and upregulation of suppressive T-reg function. The importance of CTLA-4 is demonstrated in CTLA-4-knockout mice, who develop early and catastrophic immune hyperactivation causing myocarditis and pancreatitis, and die by 3–4 weeks of age [3].

#### 2.1.2. PD-1 and PD-L1

The PD-1 protein is a 288 amino acid protein which is primarily expressed on T cells, but also on other immune cells, such as B cells, natural killer T cells, and monocytes. It was first identified at a gene level in murine cell lines and was initially thought to be involved in apoptosis, as its expression was induced when thymocyte cell death was induced [4]. Subsequently, it was found to suppress immune responses, and, in particular, it is hypothesized that PD-1 suppresses anti-self-responses [5,6]. This theory is supported by the fact that PD-1 induction is suppressed in the presence of "foreign" antigens such as lipopolysaccharide (LPS) and a stimulatory CpG-containing oligodeoxynucleotide CpG1826 [7]. The protein itself has an intracellular domain, a hydrophobic transmembrane domain and an extracellular immunoglobulin domain which is folded into a β-strand "sandwich" connected by a disulphide bridge. The intracellular domain, or cytoplasmic tail, contains an N-terminal sequence which forms an immunoreceptor tyrosine-based inhibition motif, as well as a C-terminal sequence which forms an immunoreceptor tyrosine-based switch motif. The murine and human forms of PD-1 share a 62% identical sequence, but there are significant differences in the ligand-binding sites, including alterations in size, polarity and charge [8].

The PD-1 protein has two major ligands—PD-L1 and PD-L2. Both ligands contain an N-terminal domain, which binds to PD-1, and a C-terminal domain, the function of which is as yet unknown. Both domains have an immunoglobulin-like fold forming a β-strand sandwich similar to that of PD-1 and are joined by a short linker. Nuclear magnetic resonance characterization suggests that PD-L1 proteins form homodimers, exposing the hydrophobic PD-1 binding sites, although whether this occurs in vivo remains unclear [8–10]. The PD-L2 molecule has a similar structure, with two immunoglobulin domains and a linker region, with most of the residues in the binding interfaces of both ligands conserved [11].

The binding of human PD-1 and PD-L1 proteins forms a 1:1 complex and induces a conformational change in PD-1, with the closure of the CC' loop around PD-L1 and formation of hydrogen bonds, which are hypothesized to stabilize the complex and cause re-arrangements of the PD-1 protein [10,12]. The binding regions contain both hydrophobic and polar sites, with the majority of the interaction occurring in the front strands of both proteins using the large hydrophobic surfaces of the immunoglobulin-V-type domains; the complex between PD-1 and PD-L2 is thought to be similar, although much of this work is only in murine proteins [11].

#### 2.1.3. Significance in Cancer Immunity

CTLA-4 was the first checkpoint molecule targeted in cancer treatment, initially in melanoma with dramatic results, and subsequently in other cancer types. Its significance in anti-tumor immunity was described over 20 years ago in murine models where blockade of CTLA-4 caused tumor rejection both in established tumors and with secondary exposure to tumor cells [13]. PD-1 is mainly expressed on immune cells, in particular T lymphocytes, as well as B lymphocytes, NK cells, dendritic cells and

monocytes, and its expression can be induced by many factors, including interleukins, infectious agents and LPS [14–16]. As described above, its main function is in immune suppression; therefore, in tumors, it can have the detrimental effect of decreasing anti-tumor immunity, particularly because many cancers develop the capability to express the PD-L1 ligand. On presentation of an antigen to a T lymphocyte, a typical T-cell response involves binding the antigen to the specific T-cell receptor, expansion of this T cell clone and, finally, an effector phase of the response. The co-receptors CD28 and CD3 are involved in the induction of this response. Specifically, in the tumor microenvironment, neoantigens from cancer cells are released, captured and processed by antigen-presenting cells. Antigen presentation to T cells must be accompanied by a secondary signal mechanism in order for T cells to be primed and activated. This secondary signal can be via cytokines, such as IL-12 and type 1 interferon, factors released by dying cancer cells or via the gut microbiota [17,18]. Both CTLA-4 and PD-1 suppress CD28-mediated pathways; PD-1 does this by the activation of phosphatidylinositol-3-kinase which in turn inhibits Akt phosphorylation, thereby suppressing T-cell activation, and also inhibits glycolytic pathways, thereby decreasing cellular metabolism [19]. CTLA-4 binds to its B7 ligands with a much higher affinity than CD28, preventing T-cell stimulation.

Tumor cells in many cancer types express PD-L1 and therefore can activate this pathway to escape immune surveillance. The expression of PD-L1 by tumor cells may be an adaptive response to anti-tumor immune response, with PD-L1 expression co-localized with tumor-infiltrating lymphocytes and IFN-δ, an inflammatory cytokine [20]. However, the clinical significance of PD-L1 expression is tumor histology-specific, with some cancers demonstrating improved outcomes with high PD-L1 expression, while, in other tumors, PD-L1 expression does not correlate with better survival [21–26]. The expression of PD-1 and PD-L1 in tumors may also be heterogeneous both intra-tumorally and between primary and metastatic tumor sites [27–30].

#### *2.2. Monoclonal Antibodies*

#### 2.2.1. Anti-CTLA-4

Ipilimumab, which binds to CTLA-4, was the first CPI to be licensed in 2011, and was initially used for the treatment of metastatic melanoma but is now indicated in multiple tumor types. It has a high surface area at its binding site and has a dissociation constant of 5.25 nM, with a large surface area buried at its binding surface with CTLA-4 [31] (Table 1). Tremelimumab is another monoclonal antibody targeting CTLA-4 but has not yet been licensed for any indication, although it has orphan drug status for treatment of mesothelioma. Tremelimumab is an IgG2 antibody; this subtype is thought to have less complement activation and antibody-dependent cell-mediated cytotoxicity [32]. It is currently in ongoing clinical trials, in particular in combination with durvalumab [33].


#### **Table 1.**Checkpoint inhibitors, their pharmacokinetic and dosing profiles and indications.

#### 2.2.2. Anti-PD-1/PD-L1

The first two anti-PD-1 CPIs licensed were nivolumab and pembrolizumab, based on their anti-tumor activity in phase I studies [34–36]. Pembrolizumab is an IgG4 human antibody; these antibodies have a low affinity for C1q and Fc receptors compared to other IgG molecules, making them a good antibody choice for immunotherapy, with the lowest chance of host immunity stimulation [37]. Most IgG4 antibodies are capable of a process called Fab arm exchange, in which half-molecules (a heavy chain and attached light chain) can be exchanged between IgG4 molecules [38]; pembrolizumab has a hinge region containing a S288P mutation, which prevents Fab arm exchange due to a conformational change [39,40]. The structure of nivolumab is very similar; it is an IgG4 antibody which differs from pembrolizumab only in the variable region of epitope binding-pembrolizumab binds to the C'D loop and nivolumab binds to the N-terminal loop on the PD-1 molecule [41].

Atezolizumab was the first anti-PD-L1 antibody licensed in the US. Atezolizumab and the other licensed anti-PD-L1 antibodies avelumab and durvalumab are IgG1 antibodies, which bind to the front beta-sheet of PD-L1. The heavy chain and light chain regions of these antibodies are involved in binding, with varying buried surface areas on each molecule which may affect their binding affinities [42,43]. These three antibodies have been noted to use all three complementarity determining regions from their heavy chains and two from the light chains [43,44].

After ipilimumab was licensed for the treatment of metastatic melanoma in 2011, the anti-PD-1 and anti-PD-L1 CPIs were subsequently approved for the treatment of many other cancer types, in the metastatic, adjuvant and neo-adjuvant settings. Initial approvals were for refractory/advanced melanoma and non-small cell lung cancer (NSCLC) for the anti-PD-1 CPIs, with subsequent licensing for their use in head and neck cancers, renal cell carcinoma, Hodgkin lymphoma and urothelial carcinomas [45]. Interestingly, the anti-PD-1 antibody pembrolizumab was the first oncologic therapy to be approved for use on the basis of a genetic alteration, with FDA approval granted in 2017 for its use in any tumor demonstrating microsatellite instability (MSI) [46]. The anti-PD-L1 antibodies are used in urothelial, kidney, lung and Merkel cell carcinoma, with many further studies ongoing. The presence of high tumor mutational burden (TMB) (the number of somatic tumor mutations per megabase of sequenced DNA) may identify tumors that are more likely to respond to CPI, such as those tumors that are microsatellite-unstable; however, to date, high TMB is not used to select therapy for patients [47]. Interestingly, responses to CPIs can be durable, with subsets of patients achieving long-lasting complete responses in some disease types, although, for many others, immune escape mechanisms develop, allowing tumors to evade the response primed by CPIs [48]. These treatments generally have a high tolerability, although the main toxicities, which are immune-related inflammatory effects, may be serious in a subset of patients.

#### 2.2.3. Binding Affinities and Pharmacokinetics

Nivolumab has a binding affinity to the PD-1 protein of 3.06 nM, while pembrolizumab has an even higher affinity, with a dissociation constant of 27 pM, possibly due to its extensive binding sites to PD-1, which include hydrogen bonds, specifically water-mediated hydrogen bonds, and salt bridges [41,49,50]. Interestingly, pembrolizumab has a much lower affinity for mouse PD-1, which may be explained by specific amino acid substitutions (Asp<sup>85</sup> to Gly85) which, when mutated in human PD-1, abolish pembrolizumab binding. Atezolizumab has a high binding affinity of 0.4 nM, utilizing specific hot-spot residues on the protein binding surface [42,51], while avelumab and durvalumab have dissociation constants of 42.1 pM [43] and 667 pM [52], respectively.

Studies have shown moderate inter-individual variability (IIV) in pharmacokinetics of CPIs. Ipilimumab has stable clearance over dose ranges from 0.3 to 10 mg/kg, with a half-life of 14.7 days and IIV largely influenced by body weight and baseline LDH value, while age, gender, renal and hepatic function do not affect clearance [53]. The steady state trough concentration of ipilimumab is a predictor of response, with higher trough concentrations (in patients receiving higher doses) resulting in improved complete response rates and higher overall survival (OS), but also in increased

rates of irAEs [54,55]. Both the anti-PD-1 antibodies nivolumab and pembrolizumab have linear clearance over dose ranges of 0.1–20 mg/kg and 1–10 mg/kg respectively, with both demonstrating a time-dependent decline in clearance rates, although the decline did not appear to impact clinical outcomes [56–58]. For the anti-PD-L1 antibodies atezolizumab, avelumab and durvalumab, linear clearance is seen again over wide ranges of doses. For atezolizumab, which is usually used at a fixed dose of 1200 mg, clearance was stable at doses between 1–20 mg/kg and rates were affected by body weight and serum albumin [59]. Avelumab has a similar linear clearance, but interestingly, time-dependent clearance changes differed between tumor types, with Merkel cell carcinoma and head and neck squamous cell carcinoma patients having clearance declines of 24–32%, while all other tumor types had minimal decline in clearance over time [60]. Durvalumab had linear clearance at doses higher than 3 mg/kg, with numerous factors influencing clearance including albumin, body weight, cancer type and gender [61]. Interestingly, a factor that influences clearance in all three anti-PD-L1 antibodies is the development of anti-drug antibodies, which develop in 31.7%, 4.16% and 3.1% respectively for atezolizumab, avelumab and durvalumab, but are unlikely to be clinically relevant as they did not affect clearance to a meaningful degree.

The antitumor effect of pembrolizumab is driven by the reactivation of adaptive immune response by inhibiting PD-1 expressed on T-cells. Once the PD-1 on T-cells are fully saturated by pembrolizumab, the shape of the exposure–response relationship within the dose range of 2–10 mg/kg or 200 mg (exposure at 2 mg/kg every three weeks is similar to exposure at 200 mg every three weeks) is flat, as demonstrated in multiple indications [62]. Available pharmacokinetics (PK) results in participants with various indications (melanoma, NSCLC, HNSCC, and MSI-H) supporting a lack of meaningful difference in PK among tumor types. Therefore, the selection of the 200 mg every three weeks dosing for pembrolizumab was supported as an appropriate dose for multiple tumor types.

Similarly, nivolumab, dosed at a fixed dose of either 240 mg every two weeks or 480 mg every four weeks results in a similar time-averaged steady state exposure and safety as 3 mg/kg every two weeks across multiple tumor types in numerous clinical trials, and is approved at a fixed dosing for most indications [63–65]. Peripheral PD-1 receptor occupancy is saturated at doses ≥ 0.3 mg/kg after eight weeks treatment, again supporting minimizing the doses administered, although the degree of intra-tumoral receptor occupancy is not yet known [66]. Some regulatory authorities have suggested weight-based dosing for patients less than 80 kg and fixed dosing above, to avoid unnecessarily high doses for lower-weight patients [67]. Avelumab is currently approved at a weight-based dosing of 10 mg/kg, but simulations suggested that similar risk/benefit profiles would result from fixed dosing at 800 mg, leading to FDA approval of this fixed dose [68]. Issues with cost and drug wastage are also improved with flat dosing [69]; these results are leading to a move towards fixed dosing in many CPI indications and trials, as evidence from the majority of CPIs demonstrates that exposure, efficacy and safety are similar to weight-based dosing.

#### 2.2.4. Immune-Related Adverse Events

A full discussion of the irAEs associated with CPIs is beyond the scope of this review, but, briefly, these side effects are due to off-target activation or dysregulation of the immune system, which can affect any body organ or system. Common organs affected include the bowel, causing colitis, which can be severe, the lungs, causing pneumonitis, the thyroid gland, which can cause both overproduction or underproduction of the thyroid hormone, the adrenal or pituitary glands, the liver and the skin [70]. There appear to be some patterns to the frequency of irAEs with various CPIs, with colitis and hypophysitis more common with the anti-CTLA-4 antibodies and pneumonitis and hypothyroidism more frequently seen with anti-PD-1 therapies [71]. Deaths from irAEs are rare but do occur, with the most common causes being severe colitis and pneumonitis [71]. Rates of grade 3–4 irAEs increase with combination treatment compared with single agent treatment; for example, treatment of metastatic melanoma with ipilimumab and nivolumab resulted in 59% grade 3–4 AEs, compared with 21% for nivolumab alone and 28% for ipilimumab alone [72]. The management of irAEs includes use of steroids

for less severe cases, and immunosuppression for more severe cases, using agents such as infliximab and mycophenolate [73].

#### **3. Optimization of Checkpoint Inhibitors**

While CPIs are part of standard of care in multiple tumor types, efforts to optimize these antibodies to improve their efficacy and safety are currently underway.

#### *3.1. Non-Fucosylated Antibodies*

Non-fucosylated antibodies have been modified so that the glycans in the Fc binding portion of the antibody are not fucose-bound. This modification enhances the antibody-dependent cell-mediated cytotoxicity (ADCC) via the enrichment of Fc-gamma-receptor-expressing effector cells and depletion of T-regulatory cells [74–77]. A non-fucosylated variant of ipilimumab has been constructed, and, in mice, demonstrated increased anti-tumor activity, peripheral T-cell activation and T-reg depletion compared with standard ipilimumab, and also enhanced T-cell-mediated vaccine responses in macaques [76,78]. A modified molecule, similar to atezolizumab but with reduced core fucosylation, demonstrated increased binding to Fc-gamma-receptor-IIIa and enhanced ADCC against PD-L1-expressing tumor cells in a cell-line model [79]. Knockout of the fucosyltransferase gene FUT8 or the pharmacologic inhibition of this gene, which decreased fucosylation, resulted in decreased PD-1 expression and increased T-cell activation in mice, again supporting this as a potential mechanism to enhance the activity of checkpoint inhibitors [80]. Phase I trials of non-fucosylated ipilimumab are enrolling.

#### *3.2. Pro-Drug Formulations*

Prodrug formulations of antibodies utilize a masking peptide that binds to the antigen-binding site of the CPI which reduces systemic activity. When the antibody reaches the tumor site, proteases cleave the masking peptide and the antibody becomes fully functional, allowing tumor-targeted activity and theoretically reducing off-target systemic adverse effects. Prodrug versions of ipilimumab have been developed and demonstrate equivalent anti-tumor and immune activity and reduced lymphohistiocytic inflammation in the gastrointestinal tract and kidneys compared with standard ipilimumab [76,78]. The result is an improved safety profile. ProbodyTM therapeutics are protease-activated antibodies which have shown pre-clinical efficacy targeting PD-L1 with minimal systemic auto-immunity [81,82]; the Probody drug CX-072 is now in phase I/II clinical trial for solid tumors and lymphoma [NCT03013491].

#### *3.3. Bispecific Antibodies*

Another method to optimize CPIs is to fuse them to another antibody which can then simultaneously bind another target molecule. These molecules then have the extracellular domains of two separate antibodies, both of which can bind to their respective ligands and retain their signaling activity. An example of this type of protein is the PD1-Fc-OX40L molecule, which, on testing, retained its high affinity binding for both PD-L1/L2 and OX40, caused T-cell activation and also demonstrated an improved anti-tumor immune response compared with single antibody treatment or the combination of the two separate PD-1 and OX40 antibodies [83]. A bispecific antibody to CTLA-4 and OX40 has also been effective in pre-clinical models, reducing tumor growth and enhancing response to PD-1 targeted therapy, and is now in phase I clinical trials [NCT03782467] [84]. The RANK/RANKL pathway is usually associated with bone homeostasis and is targeted using bone-protective agents, such as denosumab in patients with metastatic bony lesions and with osteoporosis [85]. However, this pathway is also involved in the tumor-associated immune response, with increased RANKL expression seen in tumor-infiltrating T-cells and RANK expression on dendritic cells and immunosuppressive M2 macrophages [86]. While trials are underway combining CPIs with denosumab, bispecific antibodies targeting the PD-1/PD-L1 and RANK/RANKL pathways have been developed, and show significant anti-tumor activity in mouse models, in particular those of colon and lung cancer [87]. This activity was dependent on CD8+ T cells and IFN-G, and could be increased further by combining the bispecific antibody with an anti-CTLA4 antibody.

Bispecific antibodies have already entered early phase clinical trials. A fusion protein consisting of an anti-PD-L1 antibody fused to the extracellular domain of TGF-β receptor II, M7824, showed excellent pre-clinical activity, suppressing metastases, inducing long-term anti-tumor immunity and improving OS in mouse models of breast and colon cancer, both as a single agent and in combination with a therapeutic cancer vaccine [88,89]. It is currently in phase I/II trials in many cancer types including breast, prostate, lung, biliary tract and colorectal, with an early biliary tract cancer trial showing an overall response rate of 27% [PMC6421177, PMC6421170]. Another bispecific antibody, MGD013, which targets PD-L1 and LAG-3, another CPI, has shown pre-clinical activity and is in phase I trials in solid tumors [NCT03219268] [90,91]. Issues that arise with bispecific antibodies include the potential for increased immunogenicity and therefore more adverse events, as well as difficulties with safety assessments in animal models. There are many other bispecific antibodies in pre-clinical development, combining immune checkpoint blockade with other tumor-specific protein binding.

#### **4. New Agents Targeting Immune Checkpoints**

#### *4.1. Small Molecule Checkpoint Inhibitors*

While there has been considerable progress in the development of antibodies targeting the PD-1/PD-L1 pathway, interest has been growing in attempts to block this axis using small molecules. The purported benefits of using small molecules rather than antibodies include potentially better oral bioavailability, fewer immune-related adverse events, improved tumor penetration and a lower production cost. The initial molecules shown to inhibit this pathway were sulfamonomethoxine and sulfamethizole, which could rescue PD-1-mediated inhibition of IFN-g production, a process which was dependent on PD-L2 [92]. Substituting particular rings in the structure of the sulfamethizole compound, such as a phenyl ring instead of a pyridyl ring, improved the efficacy of the compound in restoring IFN-Gexpression. While, ultimately, research into these compounds was not continued, they provided proof of concept for the small molecule inhibition of the PD-1/PD-L1 pathway.

Several other small molecule compounds that inhibit PD-L1 have been patented [93]. These molecules have been shown to bind directly to each dimer of PD-L1 and can dissociate the PD-1/PD-L1 complex, and certain "hot spots" on the PD-L1 molecule, which are targetable by small molecules, have been identified using in vitro studies of these compounds [94,95]. However, one of the major problems with small molecule inhibitors to date has been their large molecular weight, which impairs adequate absorption and distribution in the human body.

The only small molecule currently in human clinical trials is a molecule called Ca-170, which inhibits both the PD-L1 pathway and the V-domain Ig suppressor of the T-cell activation (VISTA) pathway. Pre-clinical work has demonstrated that in mice, this molecule can inhibit tumor growth, enhance peripheral T cell activation and increase activation of tumor-infiltrating CD8+ T-cells [96,97]. Oral bioavailability in mice was 40%, but in monkeys was <10%, again raising the issue of oral administration of these compounds. Ca-170 is in phase 1 clinical trials in patients with advanced solid tumors and lymphoma, and also in phase II trials, with a clinical benefit rate of 59.5% reported, and higher response rates seen at lower doses [98]. Interestingly, a recent study examining the mechanism of binding of Ca-170 has shown that there is no direct binding between the compound and the PD-L1 molecule, suggesting there may be an alternative mechanism of action [99]. To date, the majority of small molecule inhibitors of PD-L1 do not appear to be ready for widespread clinical usage and further pre-clinical work is needed to optimize their formulation and use.

#### *4.2. Peptide Checkpoint Inhibitors*

As described above, the crystal structure of the PD-1 and PD-L1 molecules and the mechanism by which they bind has been clearly defined, and, therefore, interest has grown in designing a peptide

inhibitor that could bind to one of these binding sites. With this data, the first peptide antagonist, (D)PPA-1, was described in 2015, and designed using a mirror-image phage display method, binding to PD-L1 and blocking the PD-1/PD-L1 interaction and decreasing tumor growth in vivo [100]. Replacing the L-amino acids with D-amino acids can improve the stability and oral bioavailability of these drugs. Another more recently developed peptide, PL120131, was designed to interact with the PD-1 molecule, based on the interacting residues on PD-L1 from the amino acid glycine at position 120 to asparagine at position 131 [101]. PL120131 was shown to act as a competitive inhibitor of PD-L1 by associating with the binding groove on PD-1, and to reverse the apoptotic signal induced by soluble PD-L1 in Jurkat cells and primary lymphocytes. Another class of peptides are the macrocyclic peptides, which bind to the PD-1-binding site on the PD-L1 molecule, and can restore T-cell function in vitro [102].

To date, none of the peptide inhibitors of the PD-1/PD-L1 pathway have been used in human trials. The peptide molecule TPP-1 has a high affinity for human PD-L1, and, in a mouse model, could decrease tumor growth by 56% compared with control peptide-treated mice, by re-activating T cells through blocking the PD-1/PD-L1 interaction [103]. A compound called UNP-12 demonstrated a 44% reduction in tumor growth in mice [104,105]. More recently, NP-12, which also inhibits the PD-1/PD-L1 interaction and can inhibit tumor growth and metastases in colon and melanoma mouse models, demonstrated improved efficacy when combined with tumor vaccination or cyclophosphamide [106]. The peptide inhibitors are still in early phases of development but may provide an alternative method through which to inhibit immune checkpoints.

#### **5. Conclusions**

CPIs have changed the landscape of cancer treatment in recent years, with a small proportion of patients with a variety of tumors experiencing deep and durable responses. Understanding the pharmacokinetics of many CPIs has led to a switch from weight-based to fixed dosing, which is likely to continue as more studies of the efficacy and PK of fixed dosing are completed. IrAEs and heterogeneity in responses has led to efforts to optimize existing CPIs and to develop new methods by which to inhibit checkpoint molecules. Understanding the structure of CPIs and their ligands can help in the further enhancement of these therapeutic agents.

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

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

#### **Abbreviations**


### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).
