The engineered anti-GPC3 immunotoxin, HN3-ABD-T20, produces regression in mouse liver cancer xenografts via prolonged serum retention

Abstract

Treatment of hepatocellular carcinomas (HCC) using our glypican-3 (GPC3)-targeting human nanobody (HN3) immunotoxins causes potent tumor regression by blocking protein synthesis and down-regulating the Wnt signaling pathway. However, immunogenicity and a short serum half-life may limit the ability of immunotoxins to transition to the clinic. To address these concerns, we engineered HN3-based immunotoxins to contain various deimmunized Pseudomonas exotoxin (PE) domains. This included HN3-T20, which was modified to remove T-cell epitopes and contains a PE domain II truncation. We compared them to our previously reported B cell deimmunized immunotoxin (HN3-mPE24) and our original HN3-immunotoxin with a wild-type PE domain (HN3-PE38). All of our immunotoxins displayed high affinity to human GPC3, with HN3-T20 having a K_D value of 7.4 nM. HN3-T20 retained 73% enzymatic activity when compared to the wild-type immunotoxin in an ADP-ribosylation assay. Interestingly, a real-time cell growth inhibition assay demonstrated that a single dose of HN3-T20 at 62.5 ng/ml (1.6 nM) was capable of inhibiting nearly all cell proliferation during the 10-day experiment. To enhance HN3-T20’s serum retention, we tested the effect of adding a streptococcal albumin binding domain (ABD) and a llama single-domain antibody fragment specific for mouse and human serum albumin (ALB1). For the detection of immunotoxin in mouse serum, we developed a highly sensitive ELISA and found that HN3-ABD-T20 had a 45-fold higher serum half-life than HN3-T20 (326 min vs 7.3 minutes); consequently, addition of an albumin binding domain resulted in HN3-ABD-T20 mediated tumor regression at 1 mg/kg.

Conclusion:

These data show that albumin binding deimmunized HN3-T20 immunotoxins are high potency therapeutics ready to be evaluated in clinical trials for the treatment of liver cancer.

Lay Summary:

We examine new glypican 3-targeting immunotoxins that have been engineered to contain a series of deimmunized bacterial exotoxin domains and test their potential for the treatment of liver cancer. The HN3-T20 immunotoxin engineered with a T-cell deimmunized domain, retains high tumor cytotoxicity against hepatocellular carcinoma cells and displays low toxicity in mice. HN3-T20 was found to have a short half-life of 7.3 minutes in mice, which may limit its ability to transition to the clinical environment. To address this concern, we incorporated a streptococcal albumin binding domain (ABD) to create the HN3-ABD-T20 immunotoxin. This albumin binding immunotoxin has an extended serum half-life of about 5.5 hours in mice, which represents a 45-fold improvement over HN3-T20. This increase in half-life was associated with a 10-fold reduction in effective dose and an overall decrease in associated side effects, demonstrating that albumin binding HN3-T20 immunotoxins represent the best option to transition into the clinical testing.

Introduction:

Liver cancer is the second-leading cause of cancer-related deaths worldwide (1). Sorafenib and Regorafenib are the only treatments approved by the U.S. Food and Drug Administration (FDA) for the treatment of late-stage hepatocellular carcinoma (HCC). Unfortunately, treatment with these tyrosine kinase inhibitors only provides a 3-month increase in survival (2, 3). When liver cancer is identified in an early stage, tumor resection and liver transplantation offer the best 5-year survival rates, at 47.9% and 59.3%, respectively (4). Sadly, most patients succumb to disease before a suitable donor can be found. New therapies are emerging for the treatment of GPC3-targeting liver cancers in the form of antibody-drug conjugates (ADC) (5) and chimeric antigen receptor T cell (CAR T) therapies (6, 7), but these new technologies are still being developed and may require years of optimization. Therefore, additional therapeutic options are needed for patients currently suffering from liver cancer.

One therapeutic class that can fill this unmet need is glypican-3 (GPC3)-targeting immunotoxins. These chimeric proteins combine the protein synthesis inhibitory activity of the Pseudomonas exotoxin (PE), with the antigen-binding specificity of an antibody. The targeting of GPC3 is intriguing for several reasons. First, GPC3 is a highly specific target that is expressed in 70–80% of HCC cases but has no detectible protein expression on healthy liver cells (8). Second, GPC3 has a high surface expression level and is rapidly internalized. This increases immunotoxin binding and facilitates delivery into HCC cells (9). Third, GPC3 serves as a cofactor in the Wnt and insulin-like growth factor signaling pathways (9, 10). A cysteine-rich domain in the N-lobe of GPC3 has been identified as the binding site for the Wnt protein (11). Blocking this interaction with our human nanobody targeting GPC3 (named HN3) (12), or by mutating key residues on GPC3 (e.g., F41), decreases the level of active β-catenin and reduces the rate of cancer cell proliferation (9, 11). Combining cell signaling pathway inhibition and protein synthesis inhibition has been shown to cause potent regression of HCC (9). Thus, dual targeting of these pathways with GPC3-targeting immunotoxins has a strong therapeutic potential for the treatment of HCC.

Immunotoxins have been used to treat a wide range of cancers in the clinical setting with different levels of success. Patients with relapsed and refractory hairy cell leukemia have showed sustained cancer remission when treated with Lumoxiti, a FDA-approved CD22-targeting immunotoxin (13). Unfortunately, patients with solid tumors like pancreatic, ovarian, and mesothelioma cancers, typically experience partial remission or stable disease following immunotoxin treatment (14, 15). The ability of an immunotoxin to diffuse into solid tumors (16), the formation of neutralizing anti-drug antibodies (13, 17), and their clearance by kidney filtration (18), all contribute to the ineffectiveness of immunotoxin therapy in solid tumors. The short serum half-life and potential to induce neutralizing antibody responses are important issues that must be addressed before HN3-based immunotoxins can be used to treat HCC patients.

In the present study, we constructed a panel of immunotoxins by combining HN3 with deimmunized toxin fragments predicted to be less immunogenic in patients (17, 19). In addition to, HN3-mPE24, a B cell deimmunized immunotoxin previously produced by our lab (20), we constructed three additional deimmunized immunotoxins with mutations to decrease T cell antigenicity. These included HN3-T20 that contained 6 point mutations targeting T cell activation, as well as HN3-T19 (10 point mutations) and HN3-M11(11 point mutations) targeting a combination of both B and T cell antigenicity. We compared our immunotoxins’ binding affinity, enzymatic function and ability to regress in vivo xenografts in mice. We observed that our immunotoxins had binding affinities in the low nanomolar range. Additionally, the deimmunized immunotoxins exhibited ADP-ribosylation activity similar to the wild-type versions, with the exception of HN3-M11, which showed reduced activity. To further improve HN3-T20, we added two different albumin binding domains (ABD)s known to increase the overall circulation time of immunotoxins (21, 22). The additional of a streptococcal ABD resulted in a 45-fold increase in serum half-life in mice (326 minutes vs 7.3 minutes) and was associated with a 10-fold decrease in therapeutic dose requirement to treat Hep3B xenografts in mice. Taken together, these results suggest that HN3-ABD-T20 represents a viable treatment option for HCC patients that have not responded favorably to current treatments.

Methods:

Tissue Staining

The HN3 nanobody was produced as previously described (12). The HN3 antibody was used to stain a complete tissue cross-reactivity cryopreserved tissue microarray (Asterand Bioscience, UK). Tissue staining was conducted and analyzed by Thomas Longerich from the University Hospital in Heidelberg, Germany.

Cell Lines

Human cells lines were cultured in DMEM supplemented with 10% fetal bovine serum (Hyclone, PA), 1% penicillin-streptomycin (Gibco, MD), and 1% GlutaMAX (Gibco, MD) as previously reported (20). Hep3B and HuH-7 cells were obtained from Xin-Wei Wang from National Cancer Institute (NCI) (Bethesda, Maryland). G1 is the A431 cell line that stably expresses GPC3 at a high level (23). Cells were confirmed to be mycoplasma and mouse pathogen-free by the Animal Diagnostic Laboratory Services at NCI-Frederick.

Mouse Xenograft Studies

Mice were treated in accordance with the Institutional Animal Care and Use Committee guidelines at the National Institutes of Health (NIH). Female athymic nu/nu mice (5-week old) were purchased from the NCI Animal Production Facility located in Frederick, MD and were housed in a Maxi-Miser Caging System (Thoren, PA). Tumor injections, volume calculations, and experimental endpoints were conducted as previously reported (20).

ELISA Binding Assay

Nunc Maxisorp ELISA plates (Thermofisher, MA) were coated with 50 μl Human Serum Albumin (HSA) (1 μg/ml, Grifols Therapeutics, NC) or Mouse Serum Albumin (MSA) (1 μg/ml, Sigma-Aldrich, MO). Immunotoxins were serially diluted in phosphate-buffered saline (PBS) and incubated in the wells overnight at 4°C. IP-12 (anti-Pseudomonas exotoxin antibody, 0.5 μg/ml), goat anti-mouse IgG (111–035-146, 1:12,500, Jackson ImmunoResearch, PA), and TMB substrate were used. The IP-12 antibody was provided by Ira Pastan (NCI, Bethesda, Maryland).

Octet Analysis

Bio-layer interferometry was performed using an Octet RED96e with Dip and Read™ Ni-NTA biosensors. GPC3-His lacking heparan sulfate (250 ng/ml) or mouse GPC3-His (250 ng/ml, R&D Systems, MN) and HN3 immunotoxins (100 nM) were prepared in Octet binding buffer [PBS (Gibco, MD), 0.1% bovine serum albumin (Millipore Sigma, MO) and 0.05% Tween (Bio-Rad, CA)]. Data acquisition phases were baseline 1 (3 min), GPC-His loading (10 min), baseline 2 (1 min), immunotoxin association (10 min), and dissociation (30 min). Octet System Data Analysis 8.2 software (Pall Fortebio, CA) was used to calculate immunotoxin affinities.

Pharmacokinetics

Athymic nu/nu mice were injected with 25 μg of immunotoxin by tail vein injection. Blood was collected into heparinized Eppendorf tubes by submandibular bleeding at 5, 30, 120, and 360 minutes (n=3) or 5, 60, and 240 minutes (n=3) for HN3-T20 treated mice. Albumin-binding immunotoxins were collected at 5, 60, 240, and 480 minutes (n=3) or 5, 720, and 1440 minutes (n=4). Plasma was separated at 16,000 × g for 5 min at 4°C. Immunotoxin levels were determined by sandwich ELISA. IP-12, a mouse monoclonal antibody was used as the capture and was paired with a rabbit anti-PE polyclonal antibody (P2318, 1:10,000, Sigma-Aldrich, MO). Detection was performed with goat anti-rabbit IgG (111–035-046, 1:5,000, Jackson ImmunoResearch, PA) and TMB substrate. GraphPad Prism 7.0 was used to determine half-life by one-phase decay modeling.

Statistical Analysis

Statistical significances for averaged IC50s in the WST assays were determined by one-way ANOVA and Tukey’s multiple comparison test. Mantel-Cox was used to determine significance of the Kaplan-Meier curves. A two-way ANOVA and Dunnett’s multiple comparison test were used to determine significance in Hep3B subcutaneous models. Serum analysis, complete blood counts, and necropsy were analyzed by one-way ANOVA. Asterisks are used to indicate significance (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).

Results:

The HN3 nanobody demonstrates high tumor specificity in human tissues

To better understand the binding specificity of HN3 to GPC3, we performed a complete human tissue cross-reactivity tissue microarray (Asterand Bioscience, UK). A cryopreserved tissue microarray was selected to ensure the conformational epitope on GPC3 was retained. Pathological evaluation revealed that none of the 35 normal tissue samples, including vital organs such as the brain, heart, lungs and liver, showed GPC3-specific binding (Figure 1, Supplemental Figure 2). Positive staining was observed in tonsil tissue, but this was limited to the tonsil crypts and most likely due to non-specific staining of detritus. Non-specific staining was also observed in the testis and was attributed to lipofuscin pigmentation of the Leydig cells. The tissue distribution and cell localization we observed in this study appears consistent with previous reports on GPC3 expression (24). These data demonstrate that HN3 binds GPC3-positive tissues with high specificity and displays low off-target tissue cross-reactivity. Additionally, the absence of GPC3 expression in normal adult tissues strongly suggests that GPC3 is a tumor-specific target.

Fig. 1. HN3 staining of a tissue microarray.

A tissue cross-reactivity array from Asterand was stained with purified HN3-hFc(IgG).

Deimmunized immunotoxins are engineered to remove T cell epitopes

To identify an appropriate immunotoxin for clinical development, we constructed a panel of HN3-based immunotoxins using six versions of the Pseudomonas exotoxin (Figure 2). We previously demonstrated that HN3-PE38 containing a wild-type PE domains II-III, and HN3-mPE24 with a furin cleavage linker and a B cell deimmunized domain III, are capable of inhibiting HCC proliferation (9, 20). For this study, we generated HN3-LR containing a wild-type domain III and HN3-T20 which contains 6 point mutations in domain III to reduced T cell antigenicity. We also produced HN3-T19 and HN3-M11 containing 10 and 11 point mutations reducing a combination of B and T cell antigenicity, respectively (Figure 2). All of our new immunotoxins contained the furin-cleavable linker used in the HN3-mPE24 version to replace PE domain II.

Fig. 2. Predicted folding models for HN3 and the Pseudomonas toxin domains.

(A) Combined structure of the HN3-PE38 immunotoxin. Complementarity-determining region (CDR) 1 (Cyan), CDR2 (Lime), and CDR3 (Magenta) of HN3. Pseudomonas exotoxin domain II (green) and domain III (orange). (B) HN3-LR with a furin cleavable flexible linker (yellow). (C) HN3-mPE24 contains 7 point mutations (red) to reduce B cell antigenicity. (D) HN3-T20 has 6 point mutations (blue) to reduce T cell antigenicity. (E) HN3-T19 contains 10 point mutations targeting both B and T cell antigenicity. Residues that were found to play a role in both B and T cell responses are shown in purple. (F) HN3-M11 toxin has 11 point mutations targeting both B and T cell antigenicity. Numbers indicate the following point mutations: #1-R427A, #2-R505A, #3-R456A, #4-R538A, #5-R467A, #6-D463A, #7-R490A, #8-F443A, #9-R494A, #10-L477H, #11-L552E.

Deimmunized immunotoxins retain their ability to inhibit cancer cell proliferation

To determine cytotoxic activity of our immunotoxins, we treated GPC3-positive HCC cell lines with increasing concentrations of immunotoxins. HN3-LR, HN3-mPE24, HN3-T20, and HN3-T19 showed no significant difference in cell inhibitory ability when compared to HN3-PE38. HN3-M11 showed a significant increase in IC50, requiring over 4-fold and 15-fold higher concentrations to inhibit the proliferation of Hep3B (Figure 3A) and HuH-7 (Figure 3B) cell lines, respectively. HN3-M11 required nearly 10-fold higher concentration to inhibit epidermoid carcinoma G1 cells overexpressing GPC3 (Figure 3C). Immunotoxin inhibition was found to be antigen-dependent because the growth of GPC3-negative A431 cell line was unaffected (Supplemental Figure 3). SDS-PAGE analysis showed all proteins were at the predicted weights and had similar purity levels (Figure 3D). An Octet binding kinetics experiment revealed all the immunotoxins bound GPC3 in the low nanomolar range, with HN3-T20 showing a binding affinity of 7.41 nM (Figure 3E). An ADP-ribosylation assay revealed that HN3-LR showed the highest level of EF2 modification with a 19% increase over HN3-PE38 (Figure 3F). We found the deimmunized domains showed 89% (mPE24), 73% (T20), 55% (T19), and 11% (M11) relative activity when compared to HN3-LR (Figure 3F). Taken together, these data suggest our deimmunized immunotoxins have similar in vitro properties as HN3-LR and HN3-PE38, except for HN3-M11, which showed a decreased affinity for GPC3 and a reduced enzymatic activity.

Fig. 3. Cytotoxic activity of deimmunized immunotoxins in vitro.

(A-C) Average IC50 required to inhibit Hep3B (A), HuH-7 (B), and G1 (C) cells with various immunotoxins (n=3). Bars represent mean ± S.D. (D) SDS-PAGE analysis of TSK-purified immunotoxins separated with a 4–20% Tris-Glycine gel. (E) Association/Dissociation properties of HN3 immunotoxins as determined by Octet analysis. K_D values are indicated in the figure key. (F) ADP-ribosylation assay of immunotoxins using Hep3B cell lysates. (G-I) IncuCyte real-time confluency assay of Hep3B cells co-cultured with HN3-mPE24 (G), HN3-T20 (H), or HN3-T19 (I) immunotoxins for 10 days. Percent confluency is indicated ± S.E. (n=3).

To further examine the activity of our deimmunized immunotoxins, we monitored Hep3B cell confluency for 10 days after treating with a single dose of immunotoxin. Treatment with HN3-mPE24, HN3-T20, and HN3-T19 at 62.5 ng/ml inhibited nearly all cell growth during the experiment (Figure 3G–I). The untreated wells reach confluency within the first 4 days of the experiment. HN3-mPE24 and HN3-T20 were able to prevent cells from reaching confluency at a dose of 15.6 ng/ml, while HN3-T19 required 31.25 ng/ml. These data indicate that HN3-mPE24 and HN3-T20 are the most potent deimmunized immunotoxins and suggest that a single treatment of immunotoxin can have a lasting effect on cancer cell proliferation.

Immunotoxins cause regression of Hep3B xenografts

To assess if the deimmunized immunotoxins could inhibit HCC in vivo, we performed a Hep3B xenograft experiment in nude mice. Mice were administered 10 mg/kg of deimmunized immunotoxins or 0.75 mg/kg of HN3-PE38 (9). The mice responded well to HN3-mPE24, HN3-T20, and HN3-T19, showing almost full tumor regression during treatment (Figure 4A). Although no mice remained tumor-free, the mean survival time for the HN3-mPE24, HN3-T20, and HN3-T19-treated mice was increased by 27, 33, and 29 days compared with the PBS control group, respectively (Figure 4B). Interestingly, HN3-T20 treated mice showed the greatest increase in overall survival, despite HN3-mPE24 showing better in vitro activity. All the mice treated with immunotoxins lost weight during the treatment cycle, but HN3-mPE24 caused the greatest amount of weight loss (Figure 4C). Taken together, we believe that HN3-T20 represents a strong candidate for clinical applications due to its high level of biological activity and tolerance in the mice.

Fig. 4. Deimmunized immunotoxins cause regression of Hep3B subcutaneous xenografts.

Female athymic nude mice were injected with 5 × 10⁶ cells in the right dorsal flank. Mice were treated a total of 9 times with HN3-PE38 (0.75 mg/kg), HN3-mPE24 (10 mg/kg), HN3-T20 (10 mg/kg), or HN3-T19 (10 mg/kg) by tail vein injection on days indicated with a black arrow. Experimental groups contained 5 mice, except for HN3-PE38, which contained 4 mice. (A) Experimental groups’ average tumor volume ± S.E. (B) Kaplan-Meier curve of mouse survival. Mice were euthanized when tumors volume exceeded 1500 mm³. (C) Average body weight of mice during experimental treatment. (D-F) Female athymic nude mice were injected with 2 × 10⁶ cells in the right dorsal flank, then treated with 8 injections of HN3-LR or HN3-T20 immunotoxins at 5 mg/kg. (D) Average tumor volume ± S.E., (E) Kaplan-Meier curve of mouse survival, and (F) average body weight (n=4).

To determine if lower concentrations of HN3-T20 would result in tumor regression, we treated Hep3B xenografts at 5 mg/kg. HN3-T20-treated mice showed reduced tumor progression, while the HN3-LR treatment group showed tumor regression (Figure 4D). The HN3-LR group showed a 38 day increase in survival when compared to the PBS-treated group (Figure 4E). This was 18 days better than the HN3-T20 group that showed a 20 day increase in survival (Figure 4E). The HN3-LR treated mice exhibited a steady loss of body weight that was not observed in the HN3-T20 group (Figure 4F). The ability of HN3-LR to cause tumor regression at lower concentrations than HN3-T20 is promising, but the significant loss of body weight observed in mice raises concerns. Taken together, HN3-T20 appears to be a safer alternative than HN3-LR.

Addition of an albumin binding domain increases serum half-life

One concern associated with the use of immunotoxins in the clinical setting is the short serum half-life observed in patients (14). Increasing an immunotoxin’s serum half-life by adding an ABD has been previously reported (21, 22). To see if this would hold true for our HN3-based immunotoxins, we incorporated ABDs from two different sources into HN3-T20. We used a 54 amino acid long domain isolated from the streptococcal protein G to create HN3-ABD-T20 (Figure 5A). This ABD consists of 3 alpha helical domains and shows similar affinities to both mouse and human serum albumin (21, 25). We also constructed HN3-ALB1-T20 containing a 115 amino acid domain isolated from a llama heavy chain antibody (Figure 5B). This nanobody was discovered following immunization of a llama with HSA and is described in the US patent (US20070269422A1) as having a Kd of 0.57 nM to HSA and 6.5 nM to MSA. The addition of an ABD had no impact on the final purity of our immunotoxins (Figure 5C). The addition of ALB1 was well-tolerated, but the addition of the streptococcal ABD resulted in a 2-fold reduction in ADP-ribosylation activity (Figure 5D). A decreased affinity for GPC3 was observed in the albumin binding immunotoxins (Figure 5E), but this reduced affinity had no effect on the ability to inhibit Hep3B cells (Figure 5F). Taken together, these results demonstrate that the ABDs were successfully incorporated into HN3-T20 with only a modest reduction in immunotoxin performance.

Fig. 5. Albumin binding immunotoxins retain high in vitro activity.

(A) Predicted structural model for HN3-ABD-T20. Numbers indicate the following point mutations: #1-R427A, #2-R505A, #8-F443A, #9-R494A, #10-L477H, #11-L552E. (B) Predicted structural model for HN3-ALB1-T20. (C) SDS-PAGE analysis of TSK-purified immunotoxins separated on a 4–20% Tris-Glycine gel. (D) ADP-ribosylation assay of immunotoxins using Hep3B cell lysates. (E) Association/Dissociation properties of HN3 immunotoxins as determined by Octet analysis. K_D values are indicated in the figure key. (F) Averaged IC50 as determined by WST assay for Hep3B cell. Bars represent mean ± S.D. (n=3).

To determine if the ABDs would increase our immunotoxins’ affinity for serum albumin, we performed an ELISA with MSA (Figure 6A) or HSA (Figure 6B). In both cases, the half maximal binding concentration of the albumin binding immunotoxins was found to be in the low nM range, which was several hundred-fold higher affinity than HN3-T20. ABD showed greater affinity for MSA than ALB1, but the opposite was true about HSA. HN3-ALB1-T20 was found to have 4 times higher affinity to HSA than HN3-ABD-T20 (0.17 nM vs 0.78 nM). To test if the increase in albumin binding affinity would correlate with increased serum half-life, we performed a pharmacokinetic study in nude mice. HN3-T20 was found to have a half-life of 7.3 minutes, with HN3-ABD-T20 and HN3-ALB1-T20 having a half-life of 326 minutes and 164 minutes, respectively (Figure 6C). To determine if the increased half-life would correspond to better therapeutic activity, we used our immunotoxins to treat Hep3B cell xenografts. The HN3-ABD-T20-treated mice showed better tumor regression than the HN3-ALB1-T20 group (Figure 6D), despite HN3-ALB1-T20 having twice the maximum tolerated dose. The HN3-ALB1-T20-treated mice showed a delay in tumor formation when compared to the PBS-treated control mice, but treatment did not result in tumor regression. The reduced tumor burden allowed the HN3-ABD-T20 mice to survive an average of 75 days, compared to 63 days (HN3-ALB1-T20) and 46 days (PBS) (Figure 6E). Body weight loss was observed in both treatment groups, with the most significant weight loss occurring in the HN3-ABD-T20 group (Figure 6F). The low average body weight in the HN3-ABD-T20 group was unintentional. Mice in this experiment were grouped to provide the closest average tumor volumes. The low starting body weight resulted in the HN3-ABD-T20 group receiving less immunotoxin per dose. Despite this, the HN3-ABD-T20 showed the highest degree of tumor regression and the greatest increase in overall survival.

Fig. 6. Immunotoxins with albumin binding domains exhibit greater in vivo activity.

(A,B) Binding profile of purified immunotoxins to immobilized mouse(A) and human(B) serum albumin. (C) Level of circulating immunotoxin following tail vein injection in nude mice as determined by sandwich ELISA. (D-F) Female athymic nude mice were injected with 5 × 10⁶ cells in the right dorsal flank. Mice were treated with a total of 9 tail vein injections of PBS, HN3-ABD-T20 (1 mg/kg), or HN3-ALB1-T20 (2 mg/kg) on days indicated by black arrows (n=4). (D) Experimental groups’ average tumor volume ± S.E. (E) Kaplan-Meier curve of mouse survival. Mice were euthanized when tumor volume exceeded 1500 mm³. (F) Average body weight of mice during experimental treatment.

To help identify the best candidate immunotoxin for clinical trials, we compared HN3-LR, HN3-T20, and HN3-ABD-T20 for both anti-tumor activity and toxicity. We treated Hep3B xenografts with 8 doses of immunotoxin. All the mice in the PBS group showed signs of tumor progression and exhibited no change in body weight (Figure 7A,F). Mice treated with HN3-LR showed a high level of tumor regression, but treatment was associated with rapid weight loss that required the fourth dose to be withheld (Figure 7B,F). One mouse in this group was removed from the study due to continued weight loss. HN3-T20 treatment resulted in tumor regression or stable disease (Figure 7C) and was associated with the least amount of body weight loss (Figure 7F). Four of the five mice in the HN3-ABD-T20 treated group showed tumor regression during the experiment (Figure 7D). This was similar to the tumor regression seen in the HN3-T20 group, even though HN3-ABD-T20 was administered at 1/10 the concentration. Although there was weight loss associated with the HN3-ABD-T20 group, it was not as rapid as the HN3-LR group (Figure 7F). IVIS imaging revealed immunotoxin treatment lead to an overall decrease in tumor burden for all groups (Figure 7G). It should be noted, that the HN3-ABD-T20 treated mouse that presented with tumor progression has the largest tumor burden at the start of the experiment.

Fig. 7. Comparing the anti-tumor properties of top immunotoxin candidates.

Female athymic nude mice were injected with 5 × 10⁶ cells in the right dorsal flank. Groups of 5 mice were treated a total of 8 times with PBS, HN3-T20 (10 mg/kg), or HN3-ABD-T20 (1 mg/kg) by tail vein injection. Mice receiving HN3-LR (5 mg/kg) were given 7 doses due to low body weight on day 31, with one mouse requiring euthanasia. Black arrows indicated days when immunotoxin was administered. The grey arrows indicate days where the HN3-LR group did not receive an injection. (A-D) Spider plots of tumor volume for PBS(A), HN3-LR(B), HN3-T20(C), and HN3-ABD-T20(D) treatment groups. (E) Experimental groups’ average tumor volume ± S.E. (F) Average body weight of mice during experimental treatment. (G) IVIS images of mice at the start and end of treatment.

To further assess the toxicities associated with immunotoxin treatment, 3 mice from each group underwent necropsy, hematology, and serum chemistry analysis. Immunotoxin treatment resulted in liver enlargement when compared to the PBS group. The most significant increases in liver weight were observed with the HN3-LR- and the HN3-T20-treated mice, indicating dose concentration plays a factor (Table 1). Other internal organs were unaffected by immunotoxin treatment. Hematology revealed an increase in white blood cell levels consistent with previous reports by our lab (9, 20). The HN3-ABD-T20-treated group had the highest leukocyte levels observed, but whether this will benefit, or hinder tumor clearance remains unclear. Serum chemistry analysis indicated an increase in alanine aminotransferase following immunotoxin treatment. This enzyme is found in high concentrations in the cytoplasm of liver cells, so elevated serum levels may indicate cell lysis and liver damage (26). Taken together, these data suggest that the high potency and low toxicity of HN3-ABD-T20 make it ideal candidate for clinical development.

Table 1:

Blood analysis and necropsy results

	PBS	HN3-LR	HN3-T20	HN3-ABD-T20	Normal Range
Leukocyte (K/μl)	3.80 ± 1.05	13.08 ± 6.36	10.09 ± 1.87	18.43 ± 3.77 **	1.80 − 10.70
Erythrocyte (M/μl)	8.68 ± 0.44	9.09 ± 0.27	8.74 ± 0.37	9.36 ± 0.11	6.36 − 9.42
Albumin (g/dL)	4.27 ± 0.45	3.50 ± 0.60	3.33 ± 0.51	3.40 ± 0.26	2.50 − 4.80
Alkaline Phosphatase (U/L)	62.33 ± 5.86	63.33 ± 19.50	60.00 ± 11.53	30.33 ± 8.96	62.00 − 209.00
Alanine Aminotransferase (U/L)	31.00 ± 1.00	875.67 ± 405.20 **	169.67 ± 14.84	256.00 ± 87.93	28.00 − 132.00
Total Bilirubin (mg/dL)	0.27 ± 0.06	0.17 ± 0.15	0.20 ± 0.00	0.23 ± 0.06	0.10 − 0.90
Blood Urea Nitrogen (mg/dL)	19.00 ± 4.00	18.33 ± 2.31	19.33 ± 3.06	15.33 ± 1.53	18.00 − 29.00
Creatinine (mg/dL)	0.30 ± 0.17	0.23 ± 0.25	0.33 ± 0.12	0.27 ± 0.12	0.20 − 0.80
Globulin (g/dL)	1.60 ± 0.20	0.83 ± 0.46	1.73 ± 0.60	1.50 ± 0.30	0.00 − 0.60
Total Protein (g/dL)	5.87 ± 0.57	4.33 ± 0.31 **	5.07 ± 0.40	4.93 ± 0.32	3.60 − 6.60
Organ weights (g)
Brain	0.42 ± 0.01	0.43 ± 0.01	0.44 ± 0.04	0.43 ± 0.01
Heart	0.10 ± 0.00	0.11 ± 0.01	0.11 ± 0.01	0.12 ± 0.01
Kidney	0.26 ± 0.00	0.29 ± 0.01	0.28 ± 0.01	0.27 ± 0.02
Liver	0.92 ± 0.12	1.26 ± 0.08 *	1.41 ± 0.20 **	1.13 ± 0.05
Lung	0.15 ± 0.03	0.29 ± 0.22	0.18 ± 0.03	0.19 ± 0.03
Spleen	0.10 ± 0.01	0.13 ± 0.03	0.13 ± 0.06	0.11 ± 0.03
Xenograft	0.55 ± 0.34	0.02 ± 0.01 *	0.04 ± 0.04 *	0.02 ± 0.01 *

Athymic nude mice were subcutaneously injected with 5×10⁶ Hep3B cells. Mice were administered PBS, HN3-T20 (10 mg/kg) or HN3-ABD-T20 (1 mg/kg) by tail vein injection every other day for a total of 8 injection. Mice in the HN3-LR (5 mg/kg) group were given a total of 7 injections. Three mice from each group were selected for a comprehensive blood analysis and necropsy. Values represent the means ± S.D. (*p<0.05, **p<0.01)

Discussion:

In this study we describe HN3-ABD-T20, a GPC3-targeting immunotoxin engineered for reduced T cell antigenicity and increased serum retention. The addition of a streptococcal ABD had little effect on in vitro cell proliferation but displayed increased in vivo performance in our Hep3B xenograft model. The tumor regression we observed with HN3-ABD-T20 at 1 mg/kg was comparable to the regression seen with HN3-T20 at 10 mg/kg. The increased potency in our animal model was attributed to a 45-fold increase in serum retention. Our pharmacokinetic study showed HN3-ABD-T20 had a half-life of ~5.5 hours in mice compared to the 7 minutes half-life of HN3-T20. Taken together, HN3-T20 with an ABD appears to be the better suited for clinical application than HN3-T20.

In our mouse model, the addition of the streptococcal ABD caused the greatest increase in serum half-life and correlates with the observed affinity of streptococcal ABD for mouse albumin. It should be noted that ALB1 has a higher affinity for human albumin, but whether this will translate to increased serum retention in humans is unclear. The use of transgenic mice expressing HSA may provide a more accurate understanding of our immunotoxins’ pharmacokinetics in future studies (27). The increased half-life we observed was similar to previous reports that showed an anti-mesothelin immunotoxin had a 12.7 minute half-life in mice, but a 193.8 minute and 101.3 minute half-life with the streptococcal ABD and ALB1 domains, respectively (21). That study used an indirect ELISA with mesothelin-Fc as the capture antigen. However, we have found that GPC3 is a poor capture antigen, most likely due to the denaturing of GPC3 after binding to the ELISA well, resulting in the loss of the conformation epitope required for HN3 binding. In the present study, we developed a new sandwich ELISA targeting PE domain III, thus circumventing the need for GPC3 binding. This new assay allowed us to more accurately detect immunotoxin in mouse serum and could partially explain why our immunotoxins had a 2 hour longer half-life than previous reports (21). Since HN3-ABD-T20 and HN3-T20 showed no difference in Hep3B inhibition in vitro, the increased serum half-life is likely responsible for the increased in vivo activity.

The effectiveness of immunotoxin therapy in patients has been shown to be largely affected by neutralizing antibodies and by off-target toxicities (13–15). Our HN3 nanobody is of human origin, so we expect low immunogenicity issues in patients. Additionally, the low cross-reactivity of HN3 in normal human tissues would suggest low off-target toxicity in patients. Due to the reduced activity we observed in HN3-T19 and HN3-M11, we believe that these would not be well suited for clinical transition. The deimmunization strategy used to generate the T20 domain was designed to reduce CD4+ T cell activation (28). This has the potential to reduce both B cell and CD8+ T cell activation, while the mPE24 strategy was only focused on removing B cell epitopes (19). It was for this reason, and the better performance in our in vivo model, that we felt HN3-T20 was a better alternative than HN3-mPE24. The high activity that we observed with HN3-LR was promising, but there were safety concerns associated with this immunotoxin. The high alanine aminotransferase and sustained body weight loss we observed would suggest that HN3-T20 may be a safer alternative. The level of immunogenicity we can expect from the addition of the bacterial ABD is still unclear. Neutralizing antibodies targeted to streptococcal ABD could negate the increase in serum half-life. While deimmunization of the bacterial domain may be possible, this would most certainly have a negative impact on the affinity for albumin. An easier, more cost-effective method may be to humanize the ALB1 antibody to reduce immunogenicity (29, 30).

The goal of this pre-clinical screen was to identify the best immunotoxin for clinical transition. Using immunocompromised mouse models for the preclinical testing of immunotoxins has been well-established and has provided enough preclinical data to support clinical trials that lead to FDA approval of the first immunotoxin (moxetumomab pasudotox) (31–34). We could have used an orthotopic or patient-derived xenograft model, but these models have a high degree of inherent variability due to inconsistent tumor inoculation and the possibility of graft vs host disease, respectively (35). Additionally, a syngeneic mouse model is possible due to HN3’s cross-reactivity to mouse GPC3 (Supplemental Figure 4). However, there are contrasting reports of GPC3’s role in liver cell proliferation. Our lab has shown that a knockdown or knockout of GPC3 expression leads to reduced human HCC cell proliferation and an inability to form tumors in mice (11, 12). Conversely, the overexpression of GPC3 in human HCC cell lines promotes cell growth and tumor formation (36), and is correlated with poor prognosis in HCC patients (37). In mice the opposite is found, with the overexpression of GPC3 in the liver resulting in lower liver cell proliferation and a lack of correlation with HCC formation (38). The mouse HCC cell line Hepa1–6 has been reported to express GPC3 (39), but we found that this cell line was not sensitive to immunotoxin treatment (Supplemental Figure 5). We detected little to no expression of GPC3 on Hepa1–6 cells when we analyzed these cells by flow cytometry (Supplemental Figure 6). Currently, we are unaware of a mouse HCC tumor model that expresses the high level of GPC3 that is associated with human HCC. Therefore, we believe that the HCC xenograft model is suitable to test the safety and efficacy of our immunotoxins.

HN3-based immunotoxins represent a promising therapeutic option for HCC patients. Immunotoxins have been well tolerated in clinical trials (13, 14, 40, 41). The FDA recently approved a CD22-targeting immunotoxin for use in patients with relapsed or refractory hairy cell leukemia that have received at least two prior systemic therapies, providing proof-of-concept evidence that immunotoxins can be used to treat cancer that have not responded to other therapies. Our HN3-immunotoxins have the advantage of causing both a reduction in Wnt signaling and the inhibition of protein synthesis (9). Single-action agents like PBD-dimer ADCs have been suggested for the treatment of HCC (5), but currently no PBD-dimers ADC have been approved by the FDA. CAR T cells may also be beneficial in HCC, but the immunosuppressive microenvironment of solid tumors has been shown to negatively affect CAR T cell activity (42). CAR T cell therapy may require further optimization before it can successfully treat HCC, such as the disruption of PD-1 on CAR T cells (7). We believe that GPC3-targeting immunotoxins represent the best alternative therapy for HCC patients, because unlike cell-based therapies, our immunotoxin treatment can be done in any hospital around the world, including community hospitals and hospitals in developing countries where liver cancer is abundant.

The work we presented here described the development of HN3-ABD-T20, a GPC3-targeting immunotoxin with an albumin binding domain and reduced T cell antigenicity. This immunotoxin showed high in vivo potency and was capable of producing tumor regression at 1 mg/kg. We believe that albumin binding HN3-T20 immunotoxins represent the best option to transition into the clinical environment because of their long serum retention, high cytotoxic activity, and reduced antigenicity.

Highlights:

• The HN3-ABD-T20 immunotoxin produces regression in mouse liver cancer xenografts via prolonged serum retention.

• Albumin binding immunotoxins are effective at 10-fold lower doses in mice.

• Albumin binding HN3-T20 immunotoxins show promising preclinical profiles and are ready for clinical testing.

Abbreviations:

HCC

hepatocellular carcinoma

GPC3

glypican-3

HN3

human nanobody targeting GPC3

PE

Pseudomonas exotoxin

ABD

albumin binding domain

FDA

U.S. Food and Drug Administration

ADC

antibody-drug conjugate

CAR T

chimeric antigen receptor T cell

CDR

complementarity-determining region

NCI

National Cancer Institute

NIH

National Institutes of Health

HAS

human serum albumin

MSA

mouse serum albumin

PBS

phosphate-buffered saline