Glypican-3 Specific Antibody Drug Conjugates Targeting Hepatocellular Carcinoma (HEP-10-1020)

Abstract

Hepatocellular carcinoma (HCC) is the second most common cause of cancer-related death in the world. Therapeutic outcomes of HCC remain unsatisfactory, and novel treatments are urgently needed. GPC3 is an emerging target for HCC given the findings that: 1) GPC3 is highly expressed in more than 70% of HCC; 2) elevated GPC3 expression is linked with poor HCC prognosis; 3) GPC3-specific therapeutics including immunotoxin, bispecific antibody, and chimeric antigen receptor T-cells (CART) have shown promising results. Here, we postulate that GPC3 is a potential target of antibody-drug conjugates (ADCs) for treating liver cancer. To determine the payload for ADCs against liver cancer, we screened three large drug libraries (>9000 compounds) against HCC cell lines and found that the most potent drugs are DNA damaging agents. Duocarmycin SA and pyrrolobenzodiazepine (PBD) dimer were chosen as the payloads to construct two GPC3-specific ADCs: hYP7-DC and hYP7-PC. Both ADCs showed potency at picomolar concentrations against a panel of GPC3-positive cancer cell lines, but not GPC3 negative cell lines. To improve potency, we investigated the synergetic effect of hYP7-DC with approved drugs. Gemcitabine showed a synergetic effect with hYP7-DC in vitro and in vivo. Furthermore, single treatment of hYP7-PC induced tumor regression in multiple mouse models. Conclusion: We provide one of the first examples of ADCs targeting GPC3, suggesting a novel strategy for liver cancer therapy.

Introduction

Hepatocellular carcinoma (HCC) is the second most common cause of cancer-related death in the world.(1) Current FDA-approved therapeutics including sorafenib and regorafenib provide limited survival benefits.(2) Moreover, failures in multiple phase III studies of small molecules for targeted therapy of HCC reflect the paucity of effective HCC treatments, and emphasize that there is an enormous unmet medical need to apply new strategies for HCC therapy.(3)

Glypican-3 (GPC3) is one of the six members of the mammalian glycosylphosphatidylinositol (GPI)-anchored cell surface glypican protein family. GPC3 consists of a core protein and two heparan sulfate (HS) chains (4). GPC3 has emerged as a target of antibody-based immunotherapies because (a) GPC3 is highly expressed in more than 70% of HCC, but not in normal adult tissues;(5, 6) (b) the relationship between elevated GPC3 expression level and poor HCC prognosis has been described. (7, 8) Interestingly, GPC3 can mediate internalization after antibody binding.(9, 10) Codrituzumab, a humanized monoclonal antibody (mAb) against GPC3, has shown a good safety profile in a phase I study; however, it has not shown satisfactory efficacy in a phase II trial.(11)

Emerging anti-GPC3 chimeric antigen receptor (CAR) T-cell immunotherapy, bispecific mAb and immunotoxin showed high potency in vivo.(12–14) Our previous work showed that immunotoxins consisting of, antibody variable domains (HN3 and YP7) fused to Pseudomonas exotoxin, had potent anti-cancer activity in vitro and in vivo.(10) However, a major drawback for immunotoxin therapy is the severe immunogenicity of the Pseudomonas exotoxin.(15, 16) This can potentially lead to production of neutralizing antibodies within weeks in humans, eventually compromising the efficacy of the immunotoxin.(16) An alternative strategy to avoid the immunogenicity experienced with immunotoxin is to use small molecule cytotoxins to develop novel antibody-drug conjugates (ADCs) for the treatment of liver cancer. In this study, we investigated the possibility of using anti-GPC3 antibodies to deliver small molecule cytotoxins into liver cancer cells to treat HCC patients, and report the preclinical development of GPC3-specific ADCs. We chose YP7 (IgG1) over HN3 (a VH single domain) as YP7 displays a higher affinity for GPC3.(17) The dipeptide linker was chosen because it is susceptible to lysosomal cleavage, liberating the toxic payload. The choice of cytotoxic payload is a crucial design consideration for ADCs. For instance, replacement of DM1 with a more potent duocarmycin payload is currently under clinical development to address the problem of the low eligibility rate and high relapse rate of anti-HER2 ADC therapies.(18, 19) HCC is well-known to have resistance to cancer chemotherapeutics.(20, 21) In a high-throughput screen of three libraries of drugs (>9000 compounds tested in clinical trials) against HCC, DNA-damaging agents were identified as the most potent family. In this work, we used duocarmycin SA and pyrrolobenzodiazepine (PBD) dimer as the payloads to construct GPC3-specific ADCs. The data presented here revealed that GPC3-specific ADCs have activity against a panel of GPC3 positive HCC cell lines in the picomolar range, but not against GPC3 negative cell lines. GPC3-specific ADCs also showed antitumor activity in vivo using multiple preclinical cell and tumor models.

Materials and Methods

hYP7 and synthesis of hYP7-DC and hYP7-PC

The hYP7 mAb against GPC3 has been reported previously.(22) The antibody was linked to duocarmycin SA or PBD dimer (SET0205 and SET0212, Levena Pharma, CA) via a maleimide linker. The payload conjugation was conducted with hYP7, after reduction by three equivalents of TCEP (tris(2-carboxyethyl)phosphine) at 37^oC for 1.5 h. Six equivalents of payload dissolved in dimethylacetamide (DMA) were added, and the reaction (final solution DMA/H₂O at 15%/85%) was incubated at ambient temperature for 1 h. The product was filtered and purified using a Zeba™ spin desalting column (ThermoFisher Scientific, MD), dialyzed overnight in PBS, and collected as hYP7-DC or hYP7-PC (DAR=1.7–2.4). The YP7 antibody is available from Diagnostic BioSystems (Pleasanton, CA).

Animal studies

All mice were housed and treated under the protocol approved by the Institutional Animal Care and Use Committee (IACUC) at the National Institutes of Health (NIH). Peritoneal model: HepG2-Luc and Hep3B-Luc cells (5×10⁶) were injected i.p. into nude mice (4–5 weeks of age, female, Charles River Laboratories, MA) to create the peritoneal xenograft model. Bioluminescence imaging (IVIS Spectrum In Vivo Imaging System, PerkinElmer) was used to monitor tumor status. After one week, mice were treated with either a single dose of control or hYP7-PC (2.5 mg/kg, i.v.). After treatment, mouse tumor status was monitored weekly by bioluminescence imaging.

Xenograft model:

Hep3B or A431-GPC3 cells (10⁷) were suspended in 200 μl of PBS and inoculated in the right flank subcutaneously (s.c.) into 5-week-old female athymic NCr-nu/nu nude mice (NCI Frederick Animal Production Area, Frederick, MD). Tumor dimensions were determined using calipers and the volume was calculated using the following formula: V = 0.5 × (W² × L), where V = tumor volume, W = the smaller perpendicular diameter and L = the larger perpendicular diameter. When the average tumor size reached approximately 150 mm³, the mice were injected intravenously with the indicated doses of ADC. Mice were euthanized when the tumor size reached the maximum size allowed by the protocol for each cell line.

Toxicological analysis:

5-week-old female athymic NCr-nu/nu nude mice were injected with vehicle control (PBS) or ADC (single injection at 5 mg/kg). Three mice from each group were collected for toxicology studies one week after ADC injection. Samples were processed for completed blood counts, serum chemistry and organ weights performed by Pathology/Histotechnology Laboratory in Leidos Biomedical Research (Frederick, MD).

Small-molecule library single agent

Single Agent Library Screening:

A high-throughput screen was conducted in 1536-well white flat bottom plates (Corning) on a Kalypsys robotic system. Cell lines Hep3B, Huh7, and HepG2 were screened against 3 small molecule annotated drug libraries: NPC,(23) MIPE,(24) and NPACT (an NCATS in-house collection of pharmacologically active small molecules)) in dose response (7–11 pt.) measuring cell viability 72 hours after incubation. Briefly, cell lines were plated down at a density of 500 cells/well in 5 μL of media. A 1,536 pintool (Kalypsys) was used to transfer 23 nL of compound in DMSO to the 1,536-well assay plates. After 72 h incubation at 37°C, 2.5 μL of CellTiter-Glo (Promega) was dispensed into each well using a BiorapTR. Plates were incubated at room temperature for 10 min, transferred to a ViewLux (PerkinElmer) and the luminescence was recorded using an exposure time of 2 seconds. Plate data were normalized to in-plate controls (bortezomib as a positive (cytotoxicity) control, DMSO as negative control) and the normalized data was processed using NCATS in-house software. The most potent small molecules from the initial HTS were then retested in triplicate using a 22-point dose response against Hep3B, Huh7, and HepG2 cells lines using identical conditions as above.

Results

Synthesis of GPC3-specific antibody-drug conjugates.

We used hYP7, a humanized GPC3-specific antibody YP7, to construct ADCs in this work.(17, 22) Based on previous experiences in delivering toxin through anti-GPC3 mAbs for liver cancer, we believe that the choice of cytotoxic payload is a crucial design consideration for ADCs against HCC. HCC is known to have resistance to cancer chemotherapeutics.(20, 21) In a screening of three libraries of drugs, cytotoxins, and pre-clinical candidates (>9000 compounds), we found that the most potent candidates are mainly DNA-damaging natural products with IC₅₀ values in the sub-nanomolar range (Fig 1B, C). Duocarmycin SA and PBD dimer were chosen to test our hypothesis that GPC3 is a potential ADC target for liver cancer. We used a dipeptide linker (valine-citrulline or valine-alanine), which can be cleaved by cathepsin B in lysosomes, liberating the toxic payload. The HIC chromatograph of hYP7-DC showed four peaks indicating a heterogeneous mixture of payload labeled hYP7 mAb due to unspecific conjugation between maleimide on the payload and thiol groups on the hYP7 generated from the reduction of interstrand disulfide bonds (Fig. S1). The drug-to-antibody ratios (DAR) for different batches of hYP7-DC and hYP7-PC are between 1.7–2.4, which were calculated based on UV absorbance at 280 and 310 nm/duocarmycin SA or 335 nm/PBD dimer. The structures of hYP7-DC and hYP7-PC are shown in Figure 1C.

Figure 1.

(A) Screening of three libraries of payloads against liver cancer cell lines. (B) Examples of payloads with highest potency against Hep3B cell line. (C) Structure of hYP7-DC and hYP7-PC. (D) Binding assay of hYP7-DC and hYP7-PC against A431-GPC3 and A431. (E) Internalization rate of hYP7-PC and hYP7-DC (both at 100ng/ml, 4 h incubation) against A431-GPC3 cells. (F) Cytotoxicity curve of hYP7-DC. Error bars denote SD. (G) Western blot for cleaved poly ADP-ribose polymerase (cleaved PARP) and cleaved caspase-9 after treatment of ADCs in Hep3B cells for 4 days. (H-I) Toxicity of hYP7-DC (H) and hYP7-PC (I) after cathepsin B digestion against A431 (GPC3 negative) cells.

Binding, trafficking and in vitro cytotoxicity of hYP7-DC and hYP7-PC against GPC3⁺ cells.

The binding selectivity of GPC3-specific ADCs was determined by flow cytometry using a pair of isogenic A431 cells (epidermoid carcinoma) expressing (A431-GPC3) and not expressing (A431) GPC3. hYP7-DC and hYP7-PC demonstrated similarly selective binding to A431-GPC3 starting at 10 ng/ml till 1,000–10,000 ng/ml, whereas no significant binding was observed on A431 cells at the highest concentrations (Fig. 1E). To study whether the binding of hYP7 to GPC3 results in internalization of the antibody, we prepared fluorescently labeled hYP7, hYP7-Alexa647 conjugate, by the same maleimide-thiol conjugation method used for the ADCs. The time-lapse confocal data demonstrated the initial binding on the membrane of the A431-GPC3 cells after incubating hYP7-Alexa647 conjugate with A431-GPC3 cells for 10 min on ice. The hYP7-Alexa647 was observed to move from the cell membrane into the cells in a time-dependent manner. After 11 h, 87±5% of hYP7-Alexa647 molecules were internalized from the cell surface (Fig. 2A). We observed approximately 50% GPC3 internalization after 8 h, which is slower than mesothelin (another glycosylphosphatidylinositol (GPI) anchored protein) mediated internalization.(25) The result is consistent with our previous finding on GPC3: after 1/2 h, approximately 4×10⁴ molecules were internalized by HepG2 and Hep3B cells and over 1×10⁵ GPC3 molecules were internalized after 4 h.(10) In contrast to immunotoxin, which requires trafficking to endoplasmic reticulum to release pseudomonas exotoxin, these ADCs need to travel to lysosome to cleave the dipeptide linker and release the payload. To further examine the trafficking of ADC, we co-stained lysosome with the ADCs in A431-GPC3 and Hep3B cells. After treatment with hYP7-DC or hYP7-PC for 4 h, the results showed the internalization of the GPC3 and ADC complex (Fig. 1E). Using FACS, we quantitatively measured the internalization rate, and showed that approximately 20–30% of the ADCs were localized in lysosomal compartments after 4 h (Fig. 2B, C and S2). In contrast, no fluorescence was observed in A431 (GPC3⁻) cells under identical conditions. The GPC3-specific uptake was also under confocal microscopy in GPC3⁺ liver cancer cells (Hep3B), but not in the GPC3⁻ cells (Sk-Hep-1), which confirmed the cellular uptake after target recognition (Fig. S3).

Figure 2.

Internalization and trafficking of hYP7-Alexa647, hYP7-DC and hYP7-PC. (A) Time-dependent images of Alexa labeled hYP7 (hYP7-Alexa647) in A431-GPC3 cells. (B-C) GPC⁺ (A431-GPC3) and GPC⁻ (A431) cells were incubated with hYP7-DC or hYP7-PC for 4h at 37°C. Cells were fixed, permeabilized, and hYP7-DC or hYP7-PC visualized using a fluorophore labeled secondary antibody (red). Nuclei and lysosomes were stained with anti-Lamp1 (green) and Hoechst 33342 (blue), respectively. Lysosome (green circle) around hYP7-DC (red) or hYP7-PC (red).

The hYP7-DC is highly active against A431-GPC3 cells, with IC₅₀ values of 11 ng/mL (73 pM), but is low against GPC3⁻ cells with IC₅₀ value higher than 1,000 ng/mL (6,700 pM) (Fig. 1F). The unconjugated hYP7 did not show any potency against HCC cell lines at as high as 10⁶ pM. hYP7-DC induced apoptosis in A431-GPC3 cells (GPC3⁺) at a very low dose (50 ng/mL, 300 pM), whereas no apoptotic cells were observed in A431 cells (GPC3⁻) (Fig. S4). To confirm induction of apoptosis upon ADC treatment, we assessed poly (ADP-ribose) polymerase (PARP) cleavage as an apoptosis marker, followed by assessment of cleavage of caspase 9. A dose-dependent cleavage of PARP and caspase 9 was observed after 96 h (Fig. 1G). Cytotoxicity of both ADCs were further assessed in a panel of HCC cell lines with different GPC3 receptor copy numbers. hYP7-DC and hYP7-PC were highly active against GPC3⁺ HCC cell lines with IC₅₀ values in the picomolar range, but low toxicity against GPC3⁻ cells with IC₅₀ value larger than 2,000 pM (Table 1).

Table 1.

IC₅₀ table of GPC3 specific ADCs against GPC⁺ and GPC3⁻ cell lines.

			IC50 (pM)
Cell line	Cell type	GPC3 receptor copy number	hYP7-DC	hYP7-PC	HN3	hYP7
Hep3B	HCC	2.5×105	73	9	>6000	>6000
HepG2	HCC	3.2×105	130	15	>6000	>6000
Huh-7	HCC	9×103	641	96	>6000	>6000
A431-GPC3	epidermoid carcinoma	1.6×106	27	2	>6000	>6000
A431	epidermoid carcinoma	0	>2000	>2000	>6000	>6000
Huh-7-1	HCC	0	>6000	>6000	>6000	>6000
Huh4	HCC	0	>6000	>6000	>6000	>6000
SK-Hep-1	HCC	0	>2000	>2000	>6000	>6000

To further examine the ADC potency in vitro, we tested this hYP7-DC against spheroids of Hep3B cells: the AlgiMatrix 3D system was employed by seeding Hep3B cells in a 96-well AlgiMatrix plate, media was changed every 3–4 days, cells were treated with hYP7-DC for 6 days, live/dead cells were stained by calcein AM (green) and ethidium homodimer-1 (red), respectively (Fig. S4). With a high concentration of hYP7-DC at 1,000 ng/ml (6,700 pM), dead cells labeled by ethidium homodimer-1 were observed throughout the spheroids, which indicates that hYP7-DC is efficacious in this 3D-HCC model.

Gemcitabine and hYP7-DC showed synergistic effect in vitro and in vivo.

Chemotherapy is considered to be more effective when combinations of drugs with differential mechanisms are administered. While this is a common feature of small molecule regimens, we aimed to develop a technique for identifying potential synergistic drugs with ADCs. We developed an approach for combining a therapeutic antibody with small molecules for identification of synergies with FDA approved drugs or those already tested in clinical trials (Fig. 3A). Cells were treated with hYP7-DC (using tip-based transfer of aqueous solution) at six concentrations in combination with each compound from a library of ~2,000 oncology-focused compounds at six concentrations (using DMSO solutions resulting in a 6-by-6 treatment matrix of all possible combinations for each compound pair (Fig. 3A). Cells were treated for 48 hours and then cell viability (ATP content) was measured. Each treatment matrix was scored by 4 metrics for evidence of synergistic inhibition of cell viability.

Figure 3.

Synergistic effect of hYP7-DC and Gemcitabine in vitro and in vivo. (A) Combination screen of hYP7-DC with different small molecule anticancer agents. (B) Summary of hYP7-DC matrix synergy and antagonism. (C) Example of synergistic effect of hYP7-DC with gemcitabine. (D) Tumor volume (left), bodyweight change (middle) and Kaplan-Meier survival curve (right) for each group from the Hep3B xenograft model. Error bars denote SD. *p<0.05, **p<0.01, ***p<0.001.

A list of candidates was generated which showed synergistic effect in the screening involving different mechanism of actions: gemcitabine (ribonucleotide reductase inhibitor), pevonedistat (NEDD8-Activating Enzyme (NAE) Inhibitor), VE-821/2 (ATR Kinase Inhibitor), SN-38 (DNA Topoisomerase I Inhibitor), XL-647 (EGFR Inhibitor), MK-8776 (Chk1 Inhibitor), clofarabine (DNA Polymerase Inhibitor), and ganetespib (Heat Shock Protein 90 Inhibitor) (Table S1). Gemcitabine is an FDA-approved chemotherapeutic used in HCC treatment and showed the highest synergistic effect in combination screening, and reconfirmed as one of the most potent hits (Fig. 3B and 3C). In the Hep3B human liver cancer xenograft model, combination of hYP7-DC and gemcitabine induced tumor regression whereas treatment with a single agent only slowed tumor growth (Fig. 3D). Combining hYP7-DC and gemcitabine also significantly increased the survival more than hYP7-DC or gemcitabine alone (Fig. 3D).

In vivo anti-tumor activity of hYP7-PC in two peritoneal HCC models

To evaluate the therapeutic potential of hYP7-PC, we established two multifocal cancer models using liver cancer cells, which form multiple tumor foci scattered through the peritoneal cavity. To monitor the tumor growth over time quantitatively, HepG2 or Hep3B liver cancer cells were stably transduced with the firefly luciferase (Luc) gene to generate HepG2-Luc or Hep3B-Luc cells. Six million cells were injected intraperitoneally. After one week, the bioluminescence signal was stabilized and started to increase reflecting an increased tumor burden for both models (Fig. 4A). At the time of ADC injection, mice from control (PBS as vehicle control) and ADC groups showed identical luciferase activity. A significant decrease in bioluminescence was observed in ADC treated (2.5 mg/kg, single injection i.v., MTD = 5 mg/ml) mice one week after hYP7-PC treatment whereas a steady increase in tumor burden was found in the control mice (Fig. 4C–F). Negligible deleterious effects at the dose tested were found in both peritoneal models, fluctuations in body weight were well within the normal limits (Fig. 4H, J). Overall survival was increased significantly with the hYP7-PC treatment (Fig. 4G, I).

Figure 4.

Anticancer efficacy of hYP7-PC in the HepG₂-Luc and Hep3B-Luc xenograft models. (A) Treatment scheme. HepG₂-Luc or Hep3B-Luc cells were injected i.p., single-dose treatment of hYP7-PC (2.5 mg/kg) was given 7 days after disease establishment (n=5 per group). (B) Bioluminescence imaging (BLI) of liver burden change before and after treatment with PBS (control) or hYP7-PC (2.5 mg/kg). (C). Spider plot of individual tumor burden from the HepG₂-Luc model. (D) Quantification of tumor burden from the HepG₂-Luc model. (E). Spider plot of individual tumor burden from the Hep3B-Luc model. (F) Quantification of tumor burden from the Hep3B-Luc model. (G-H) Kaplan-Meier survival curve for each group (n=5 per group) and bodyweight change from the HepG₂-Luc model. (I-J) Kaplan-Meier survival curve for each group (n=5 per group) and bodyweight change from the Hep3B-Luc model. Error bars denote SD. *p<0.05, **p<0.01, ***p<0.001.

In vivo anti-tumor activity of hYP7-DC and hYP7-PC in GPC3 positive xenograft models.

The anti-tumor activity of hYP7-DC and hYP7-PC was further evaluated in two subcutaneous GPC3⁺ solid tumor models (A431-GPC3 and Hep3B) (Fig. 5). In the A431-GPC3 model: hYP7-PC demonstrated potent anti-tumor activity with a single-dose of 5 mg/kg (p<0.001). Although the dose for hYP7-PC is twice as high as the dose used in the two i.p. models, we did not find deleterious effect, the fluctuations in body weight was below 10% (Fig. 5B). hYP7-PC caused tumor regression in all seven mice tested whereas hYP7-DC only showed a slow-down of tumor growth. Overall survival after ADC treatment was significantly extended, with five out of seven mice treated by hYP7-PC tumor free at the end of this bioassay, and with no tumor recurrence after maintaining the survivors as long as another 24 weeks. hYP7-PC and hYP7-DC were well tolerated as mice showed no changes in grooming activity and body weight was not significantly reduced by the ADC treatment in this bioassay. hYP7-PC also demonstrated higher anti-tumor efficacy in a Hep3B HCC xenograft model compared to hYP7-DC (Fig. 5D). Reduced tumor growth compared with the control (PBS treated) group and isotype ADC was found in all mice given a single dose of 5 mg/kg of hYP7-PC and hYP7-DC in the Hep3B model. hYP7-DC (single-dose of 5 mg/kg, i.v.) induced a significant extension of survival comparing to the control group. Furthermore, all eight mice treated by hYP7-PC (single-dose of 5 mg/kg, i.v.) survived and no tumor recurrence was observed at the end of the bioassay. Consistent with previous tests, hYP7-PC was well tolerated as mice showed no changes in grooming activity and body weights were unaltered by the treatment. Taken together, these data demonstrate that GPC3-specific ADCs have significant anti-tumor activity in GPC3 positive xenograft models.

Figure 5.

(A-C) Single-dose antitumor activity of hYP7-DC in xenograft model of A431-GPC3 cells. Localized tumor models were developed in nude mice with A431-GPC3 cells, hYP7-DC was injected when tumor volume reached 160 mm³. Tumor volume (A) and bodyweight (C) were measured daily. (C) Kaplan-Meier survival curve for each group (n=7 per group). (D-F) Singe-dose antitumor activity of hYP7-DC in xenograft model of HCC. Localized tumor models were developed in nude mice with Hep3B cells, different doses of hYP7-DC were injected when tumor volume reached 100 mm³. Animals were monitored for 2 weeks. Tumor volume (D) and bodyweight (F) were measured daily. (E) Kaplan-Meier survival curve for each group (n=8 per group). Error bars denote SD. *p<0.05, **p<0.01, ***p<0.001.

Serum chemistry, blood cell counts and organ weights.

Toxicology studies involving serum chemistry and blood cell counts were carried out in mice treated by hYP7-DC (5 mg/kg) and hYP7-PC (5 mg/kg). We observed no significant change to organ weight compared to the control group (PBS). Both ADC-treated groups showed an increase (not statistically significant) in white blood cells (all readings were within normal range: 1.7–10.8 (K μl⁻¹)), but no toxicity related significant changes were found in other parameters including alanine aminotransferase (an indicator of liver function), hemoglobin, total bilirubin, blood urea nitrogen (an indicator of kidney function), total protein, and creatine (Table 2).

Table 2.

Toxicological results.

Parameters	Control	hYP7-DC (5mg/kg)	hYP7-PC (5mg/kg)	Normal range
White blood cells (K μl−1)	2.7 ± 1.0	3.8 ± 1.6	6.0 ± 3.1	1.8–10.7
Red blood cells (M μl−1)	8.5 ± 0.7	8.2 ± 0.7	8.7 ± 0.4	6.36–9.42
Haemoglobin (g dl−1)	13.2 ± 1.3	13.3 ± 0.1	13.8 ± 0.4	11.0–15.1
Albumin (g dl−1)	3.9 ± 0.2	4.0 ± 0.2	4 ± 0.1	2.5–4.8
Alanine aminotransferase (U/I)	57.7 ± 1.2	41.7 ± 1.2*	51.3 ± 4.9	28–132
Total bilirubin (mg dl−1)	0.3 ± 0.1	0.2 ± 0.1	0.3 ± 0.0	0.1–0.9
Blood urea nitrogen (g dl−1)	18.3 ± 3.1	16.3 ± 1.5	19.0 ± 2.6	18–29
Creatine (mg dl−1)	0.3 ± 0.1	0.3 ± 0.1	0.3 ± 0.0	0.2–0.8
Total protein (g dl−1)	5.2 ± 0.4	5.5 ± 0.2	5.3 ± 0.2	5.6–6.6
Organ Weight (g)
Brain	0.44 ± 0.02	0.45 ± 0.05	0.44 ± 0.00
Heart	0.11 ± 0.02	0.12 ± 0.01	0.12 ± 0.01
Kidney	0.32 ± 0.01	0.30 ± 0.02	0.30 ± 0.02
Liver	1.01 ± 0.19	1.01 ± 0.10	1.07 ± 0.12
Lung	0.16 ± 0.01	0.16 ± 0.01	0.15 ± 0.02
Spleen	0.08 ± 0.00	0.10 ± 0.03	0.10 ± 0.02

Values represent mean±s.d (n=3/group).

^*p<0.05 compared with control group.

Discussion

GPC3 is an emerging candidate target for HCC therapy.(5) Vaccines,(26) chimeric antigen receptor (CAR) T-cell therapy,(27) and bispecific mAbs (28) against GPC3 are currently under clinical evaluation. ADCs are one of fastest-growing oncology therapeutics: four ADCs are approved by FDA or the European Medicine Agency (EMA): gemtuzumab ozogamicin (developed by Wyeth/Pfizer), inotuzumab ozogamicin (developed by Wyeth/Pfizer), brentuximab vedotin ADCs (developed by Seattle Genetics), and trastuzumab emtansine (developed by Genentech/Roche).(29) More than 60 are currently in clinical development.(29) However, there are no available ADCs for liver cancer treatment. In this work, we established GPC3 as a target for ADC-based HCC therapy.

First-generation ADCs (e.g., BR96-DOX) typically used clinically approved drugs (e.g. doxorubicin) which have activity in the nanomolar range as their payload.(30) Further improvements such as enhanced payload potencies and improved ADC homogeneity led to the FDA approval of second-generation ADCs.(31) For example, T-DM1 (trastuzumab emtansine, anti-HER2 ADC) employed a potent cytotoxic agent (DM1) as its payload.(32) However, T-DM1 is still not effective enough to kill cancer cells expressing relatively low levels of HER2. Therefore, only approximately 20% of breast cancer patients are eligible for HER2-targeted therapies, and a high relapse rate is observed in a majority of these patients with initial drug response due to intratumorally heterogeneous expression of HER2.(18, 19) Replacement of DM1 with a more potent duocarmycin payload is currently under clinical development to address the problem of the low eligibility rate and high relapse rate of anti-HER2 ADC therapies.(18, 19) HCC is notorious for its resistance to cancer chemotherapeutics.(20, 21) In the search of the suitable payload for HCC, we screened three libraries of drugs which include a significant number of compounds tested in human clinical trials, and a selection of drugs which already showed promising in vivo efficacy. We found that the most potent drugs against HCC cell lines are mainly DNA-damaging natural products with IC₅₀ values in the sub-nanomolar/nanomolar range. Duocarmycin SA and PBD dimer, two most potent compounds in our screening which currently are used in other ADCs in clinical trials, were chosen as payloads for constructing anti-GPC3 ADCs to test the hypothesis that GPC3 is a potential target for ADC against HCC. Although the two chosen payloads (duocarmycin SA and PBD dimer) showed comparable potency against Hep3B, HepG2 and Huh 7 cell lines (Fig. 1B), we found that the ADC using PBD dimer as payload is 5–10 times more potent than the ADC with duocarmycin SA as its payload. We did not observe a significant difference between the two ADCs in terms of binding affinity and internalization rate (Fig. 1). Two possible reasons to explain such potency difference are the difference in cleavage efficiency of the linker by cathepsin B and the transportation ability of the payload to egress from the lysosome (cleavage site) to nucleus (target site).(33)

To improve potency and to establish potential clinical implementation, we developed a platform for screening for synergies between an ADC and small molecules, and carried out an HTS, showing gemcitabine with the strongest synergistic effect in combination with hYP7-DC. The in vivo result confirmed the combinational efficacy of gemcitabine, which is the top hit in our screening and has been used for liver cancer chemotherapy.(34) The in vivo result confirmed the screening data. Compared to either gemcitabine or hYP7-DC as single agent, the combined treatment induced significantly longer survival in Hep3B xenograft model (Fig. 3D).

Xenograft and peritoneal models are widely used in drug development. Hep3B and HepG2 cells are liver cancer models commonly used to evaluate drug efficacy, but it is still rare to observe tumor remissions in those models which is consistent to clinical observations of a low 5-year survival rate for HCC patients.(35) In this work, we consistently observed tumor remissions in HepG2 peritoneal model and Hep3B xenograft model after a single injection of hYP7-PC with the remission rate at 40% and 100% respectively. These data indicate ADC is a strategy that should be explored for HCC therapy. Because cancer stem cells (CSC) can contribute to therapeutic resistance,(36) we analyzed four liver cancer stem cell markers (CD133, EpCam, CD24 and CD44) in Hep3B cells with or without ADC treatment.(37–42) Interestingly, GPC3 and CD24 were reduced upon ADC treatment. We did not observe significant decrease or increase of other biomarkers of CSCs. One explanation is that GPC3 is uniformly expressed in all HCC cells including the subpopulations that express CSC markers. ADC treatment can kill all HCC cells including potentially CSC cells. (Fig. S6) A previous study indicated that GPC3 is also a CSC marker.(43) In summary, our results indicate that GPC3 specific ADCs do not induce enrichment of CSC markers. The payload (PBD dimer) itself did not show preference of killing or enrichment of the subpopulations with CSC markers that we tested (Fig. S6A). Further investigation is needed to uncover the precise role of CSCs in ADC treatment and other therapies targeting GPC3.(40, 44) It has been shown that bystander killing of ADCs can be triggered by the secretion of cathepsin B from cancer cells in the tumor microenvironment.(45) To analyze the bystander effect, we found that after cathepsin B pretreatment, hYP7-DC and hYP7-PC both showed significant increase of toxicity against antigen (GPC3)-negative A431 cells (Fig. 1H–I). The ADCs can potentially exhibit the bystander killing effect because cathepsin overexpression was found in HCC tumor microenvironment.(46) After ADC treatment, dead HCC cells have the potential to release cathepsin B which can facilitate the bystander killing effect for the ADCs using dipeptide linkers.

It is known that the slower internalization of GPC3 is an advantage for antibody-dependent cell-mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) activities.(22) Based on our previous work, hYP7 is found to induce ADCC and CDC. However, we also found that 5mg/kg (single injection) of hYP7 did not show potent anti-tumor activity in mice. Actually, we had to inject 60  mg/kg (9 injections) of hYP7 to achieve modestly significant tumor inhibition in xenograft models.(22) Of note the dosage used in this study is higher than other ADCs tested in vivo, possibly because the slower internalization is induced by GPC3 than other ADC targets such as HER2.(47, 48) Therefore, we elevated the dose to achieve tumor remission. In the long-term incubation assay, we also observed that hYP7-PC required 48–72 hours to start to induce cancer cell death (Fig. S7). Another potential reason is that the conjugation method we used to test our hypothesis generated a portion of unconjugated antibody, which may have lowered the efficacy. This phenomenon was observed in the development of SYD985, which employed the same conjugation strategy.(49, 50) Consistent with previous reports, when we used hydrophobic interaction chromatography (HIC) to remove the free unconjugated antibody for both hYP7-DC and hYP-PC, the potencies of HIC-purified ADCs were 5–10 times greater than that of the unpurified ADCs in vitro (Fig. S5).

In conclusion, we determined the payload for ADC against liver cancer through high throughput screening of drugs and found several potent DNA damaging agents against HCC cells. PBD dimer was identified as one of the most potent small molecules, which was used to construct hYP7-PC. hYP7-PC showed potency in the picomolar range in cancer cell models. Moreover, hYP7-PC caused tumor remission in mouse tumor models. This work unveils a new strategy that could translate into a liver cancer therapy.

Abbreviations:

ADC

antibody-drug conjugate

HCC

hepatocellular carcinoma

GPC3

glypican-3

s.c.

subcutaneously

i.v.

intravenously

i.p.

intraperitoneal injection

FACS

fluorescence activated cell sorting