RN927C, a Site-Specific Trop-2 Antibody–Drug Conjugate (ADC) with Enhanced Stability, Is Highly Efficacious in Preclinical Solid Tumor Models

Antibody–drug conjugates (ADC) were developed to improve the therapeutic indices of cytotoxic anticancer agents. The approach makes use of an immunoconjugate, in which a cytotoxic agent is chemically or enzymatically linked to an antibody that selectively binds to an internalizing tumor-associated antigen (1–3). This strategy allows specific delivery of the cytotoxic agent to the tumor site while minimizing the exposure to normal tissues. Trop-2 (trophoblast cell-surface antigen 2), also referred to as M1S1, GA733-1 (gastric antigen 733-1), EGP-1 (epithelial glycoprotein-1), or TACSTD2 (tumor-associated calcium signal transducer 2), is a type I transmembrane cell-surface glycoprotein originally identified in human placental trophoblasts (4) and subsequently found to be highly expressed in most human carcinomas. Although the biological role of Trop-2 is still under investigation (5), various studies have shown that overexpression of Trop-2 correlates with increased cancer growth (6, 7), tumor aggressiveness, metastasis, and poor prognosis in various human carcinomas (8–14). Studies have also shown that Trop-2 contributes to tumor pathogenesis, at least in part, by activating the ERK1/2 MAPK pathway, which has important implications in cancer cell proliferation, migration, invasion, and survival (15). Regulated proteolysis of Trop-2 is suggested to drive epithelial hyperplasia and stem cell self-renewal via β-catenin signaling (16). Other studies, however, have indicated that the loss of Trop-2 can also contribute to tumorigenesis depending on the cell type and context (17–19).

Immunohistochemical analysis on human tumor samples has demonstrated expression of Trop-2 protein in multiple tumor types, including pancreatic (8), ovarian (9), endometrial (10), non–small cell lung (NSCLC; ref. 11), colon (12), gastric (13), and oral cancers (14). The association of Trop-2 expression with advanced disease and/or clinical outcome in multiple tumor types, and the relatively restricted expression in normal adult tissues, underscores the potential benefit of targeting Trop-2 to fill an unmet need in cancer treatment. An anti-Trop-2 SN38 conjugate, hRS7-CL2A-SN-38 (IMMU-132), consisting of a humanized Trop-2 antibody conjugated to SN38, the active component of irinotecan, has been shown to be efficacious in several epithelial cancer cell line xenograft models (20–22). IMMU-132 (sacituzumab govitecan) is currently being tested in clinical trials for solid tumors and has reported encouraging therapeutic activity in patients with difficult-to-treat cancers (23). The safety profile of IMMU-132 is similar to irinotecan, with neutropenia and diarrhea being the most commonly observed side effects (23).

ADCs synthesized by conjugating to reduced cysteine sulfhydryl groups or lysine side chain amines yield heterogeneous conjugates that have been associated with aggregation, increased toxicity, and inconsistent pharmacokinetics (24–26). One approach that can potentially overcome these limitations is to employ site-specific conjugation of antibodies, and many experimental approaches have been developed, including conjugation through engineered cysteine residues, nonnatural amino acids, antibody carbohydrates, aldehyde tags, and utilizing enzymes such as sortases and phosphopantetheinyl transferases (2, 27–36). A site specifically conjugated anti-CD33 ADC, SGN-CD33A, has shown encouraging activities in targeting drug-resistant acute myeloid leukemia (37). We have previously disclosed the unique qualities of conjugating antibodies through an enzymatic method employing microbial transglutaminase (30). Here, we describe the design and utility of a new Trop-2-ADC, RN927C, composed of a humanized anti-Trop-2 hIgG1 antibody, conjugated with the AcLys-VC-PABC–PF-06380101 linker payload at the C-terminus of the antibody heavy chain via an enzymatic process. This conjugate contains a valine–citrulline cleavable linker and PF-06380101 (38), which is a potent antimitotic agent that inhibits tubulin polymerization. Upon binding to the extracellular portion of Trop-2 on the cell surface, RN927C is internalized and traffics to lysosomes. After being processed by the lysosomal proteases, the payload PF-06380101 is released and induces cell-cycle arrest, resulting in cell death. Indeed, RN927C showed potent in vitro cell killing activity in tumor cell lines with wide range of Trop-2 expression levels. In vivo testing of RN927C in cell line and patient-derived xenograft (PDX) models representing several solid tumor types confirmed robust activity often inducing tumor regression with a single dose. The AcLys-VC-PABC linker in RN927C is shown to be stable in the bloodstream from preclinical mouse pharmacokinetic studies and is expected to provide an enhanced off-target safety profile (30, 39). In addition, our site-specific conjugation technology produces highly homogeneous conjugates (30, 40) and has the potential to minimize unwanted pharmacokinetic behavior and toxicity often associated with highly loaded ADCs (24, 27).

The antibody portion of RN927C is derived from immunization of Balb/c mice using recombinant Trop-2 extracellular domain fused with human Fc. The mouse antibody was then humanized by CDR grafting/affinity maturation and cloned into human IgG1/k constant domains to create the parental antibody. The C-terminal lysine on the heavy chain of the humanized anti-Trop-2 antibody was replaced with the LLQGA tag using standard molecular biology techniques. The DNA encoding this antibody was cloned into an in-house expression plasmid, transiently expressed in HEK293 cells, and then purified with Protein-A MabSelect SuRe columns (GE Healthcare). For AcLys-VC-PABC–PF-06380101 conjugation, the antibody concentration was adjusted to 5 mg/mL in buffer containing 25 mmol/L Tris-HCl, pH 8.0, 150 mmol/L sodium chloride. Linker-payload (40) was added in a 10-fold molar excess over antibody; the conjugation reaction was initiated by addition of 2% (w/v) bacterial transglutaminase (Ajinomoto Activa TI) and incubated with gentle shaking at 37°C for 16 hours. The reaction mixture was adjusted to 0.75 mol/L ammonium sulfate, 25 mmol/L potassium phosphate, pH 7 (buffer A), and the material was applied to a Butyl HP column (Butyl HP, GE Healthcare), washed with 5 column volumes of buffer A, and eluted with a linear gradient over 20 column volumes into 25 mmol/L potassium phosphate, pH 7. Fractions containing the conjugate were pooled, dialyzed into 1× PBS, concentrated using a 10-kDa Amicon Ultra centrifugal filter unit (Millipore), and sterilized with 0.2-μM filter. The final product drug–antibody ratio (DAR) was 2.0 as determined by hydrophobic interaction chromatography (HIC) and mass spectrometry analysis.

A431, Fadu, OVCAR3, BxPC3, Calu-3, NCI-H292, NCI-H1650, HCC-827, MDA-MB-468, Colo205, SKBR3, and SW620 were all purchased from ATCC in 2008. Cell lines were tested for Trop-2 expression using immunofluorescence and FACS. No further authentication was performed. Gemcitabine (Gemzar) was purchased from Eli Lilly and Company. Paclitaxel was purchased from Sigma-Aldrich. Receptor quantitation was performed using Quantum Simply Cellular Kits (Bangs Laboratories, Inc.) according to the manufacturer's instructions.

Human tumor cells from ATCC were cultured in DMEM + 10% FBS (A431, Fadu, OVCAR3, SKBR3, MDA-MB-468, and SW620) or RPMI + 10% FBS (BxPC3, NCI-H1650, NCI-H292, Calu-3, HCC-827, and Colo205). Cells were seeded at 2 × 10³ cells per well (A431, Fadu, MDA-MB-468, SW620, BxPC3, NCI-H1650, NCI-H292, Calu-3, HCC-827, and Colo205) or 3 × 10³ cells per well (OVCAR3 and SKBR3) in growth media on white 96-well plates on day 0. RN927C or negative control ADC were added to each well on day 1 in the dilution range of 0.0006 to 40 μg/mL final concentration (equivalent to 0.004–267 nmol/L). Free PF-06380101 (38) was tested in the final concentration range of 0.0006 to 40 nmol/L. CellTiter-Glo viability assays (Promega) were performed 4 days after treatment. Luminescence signals were detected on the M5 plate reader (Molecular Devices). All the readings were normalized to the percentage of value derived from control untreated wells. IC₅₀s were calculated by logistic nonlinear regression (GraphPad Prism 6.03) and reported as the concentration of antibody (nmol/L) that reduced cell viability by 50%.

HCC-827 cells were cultured until nearly confluent, harvested, and resuspended in 5 mL of cold binding buffer (DMEM + 10% FBS + 10 mmol/L HEPES) at 1.6 × 10⁶ cells/mL. Primary antibody (RN927C or parental Ab) was labeled with DyLight 800 Antibody Labeling Kit per manufacturer's protocol (Thermo Scientific). The labeled antibody was then added to the cells to a final concentration of 2 μg/mL and incubated at 4°C for 1 hour. To initiate internalization, cells were washed and resuspended in 5 mL prewarmed binding buffer supplemented with 2 μg/mL unlabeled parental Ab and incubated in a 37°C water bath. At each duplicate time point, 0.25 mL sample was removed and internalization was stopped by the addition of 4 volumes of quench buffer (cold 1× PBS + 0.2% sodium azide) and placed on ice. Subsequently, cells were treated with 1× trypsin-EDTA (Mediatech) supplemented with 5 mg/mL papain (Sigma-Aldrich) and incubated in a 37°C water bath for 25 minutes to remove uninternalized antibodies on the surface. After the wash, cells were resuspended in 200 μL of fix buffer (1× PBS + 1.5% paraformaldehyde + 1:4,000 Draq5 (BioStatus) and centrifuged onto a 96-well poly-d-lysine coated plate (BD biosciences). For time point zero, a second set of samples was collected without the trypsin/papain step to measure the maximal amount of surface-bound fluorescence signals. The fluorescence signals were acquired by Odyssey CLx Infrared Imaging System (LI-COR Biotechnology) and analyzed using Odyssey software version 3.0.30. For each time point, relative fluorescence signal, F_r(t), was determined from the fluorescence antibody signal divided by the Draq5 DNA signal. Subsequently, the normalized amount of internalized antibody at each time, Y(t), was calculated from F_r(t) divided by F_max, the maximum surface-bound fluorescence signal at time zero, and plotted as a function of time. The resulting graph was fitted to the equation Y(t) = (Bmax * t)/(T½ + t) using GraphPad Prism 6.03, where t is time, Bmax is the maximal amount of internalized antibody, and T½ is the time at which half maximum internalization occurs.

BxPC3 and SW620 cells were seeded on 12-mm coverslips put into 24-well plates in growth media until 70% to 80% confluent. Mouse Ab from which RN927C was derived was used at 2 μg/mL in RPMI + 10% FBS + 20 mmol/L HEPES and incubated at 4°C for 1 hour. The cells were washed twice with 1× PBS and then fixed in 4% formaldehyde solution for 10 minutes at room temperature. Afterwards, the cells were washed twice with PBS and then incubated in 400 μL nonpermeabilizing blocking buffer (1× PBS + 5% donkey serum) for 1 hour. Trop-2 expression was then detected with Cy3-conjugated donkey anti-mouse second Ab (1:300, Jackson ImmunoResearch Labs).

Cells were grown on coverslips as described previously. After removing the growth media, 10 μg/mL of RN927C in RPMI + 10 % FBS + 20 mmol/L HEPES was added to the cells and incubated at 4°C for 1 hour. The antibody solution was then removed, and the cells were washed once with 1× PBS. Growth media (0.5 mL) were then added, and the cells were moved to 37°C incubator for 2 hours. The cells were washed twice with 1× PBS, fixed, and then incubated in 400 μL blocking buffer (1× PBS + 0.3 % TX-100 + 5 % donkey serum) for 1 hour. Lysosomal compartments were labeled with anti-LAMP2 (Abcam) diluted 1:100 in 300 μL blocking buffer for 1 hour at room temperature, followed by second Ab detection (1:300 for Cy3-donkey anti-mouse and 1:200 for Dylight488 goat anti-human, both from Jackson ImmunoResearch Labs). The coverslips were mounted on slides with a drop of VECTASHIELD with DAPI (Vector Laboratories) to stain the nuclei. Images were observed and acquired by a scanning confocal microscope LAS AF from Leica.

Cells were grown on chamber slides (Thermo Scientific) until 50% to 70% confluent and then treated overnight with 0.1, 1, and 10 μg/mL of RN927C. Anti-phospho-histone H3 rabbit antibody (Cell Signaling Technology) was used at 1:1,600 dilution for staining, and detection was done using Alexa 488-F(ab')2 donkey anti-rabbit IgG (Jackson ImmnuoResearch Labs) as second Ab. Images were observed and acquired by a Nikon Eclipse E800 microscope (Nikon Instruments).

All animal studies were conducted in an AAALAC accredited facility under IACUC-approved protocols. Three animals were used at each time point in the pharmacokinetic study and 5 animals per cohort were used in all the efficacy studies. For BxPC3 xenograft model used in both pharmacokinetic and efficacy studies, 2 × 10⁶ cells were implanted subcutaneously into 5- to 6-week-old female CB17/SCID mice (The Jackson Laboratory) until the tumor sizes reached 250 to 300 mm³ before treatment started. For SW620 model, 3 × 10⁶ tumor cells were implanted subcutaneously into 5- to 7-week-old female NCr nu/nu mice (Taconic) until the tumor sizes reached approximately 300 mm³. Pan0123 (pancreatic ductal carcinoma from peritoneal biopsy), Pan0135 (pancreatic adenosquamous carcinoma) and Pan0146 (metastatic pancreatic carcinoma from liver biopsy), and LG0476 (squamous NSCLC) PDX models were acquired from The Jackson Laboratory, CTG-1017 (triple-negative breast cancer) model was acquired from Champion Oncology, and the Ova196756 (ovarian adenocarcinoma metastasized to colon) model was established in-house from surgical tumor specimen propagated first in NSG mice (The Jackson Laboratory). For these models, 1 to 2 mm³ of tumor fragments were implanted subcutaneously into the lateral flanks of female CB17/SCID mice from Taconic. Animals were randomized by tumor sizes once they reached approximately 300 mm³ or otherwise indicated, and RN927C and other reagents were administered through bolus tail vein injection. Tumor volume was calculated with the following formula: tumor volume = (length × width²)/2. Animals were humanely sacrificed before their tumor volumes reached 2,000 mm³. All tumor models except SW620 were tested positive for Trop-2 expression by IHC (data not shown).

Serum samples were analyzed using an ELISA developed on GyroLab immunoassay platform. For the total antibody assay, RN927C or unconjugated Ab were captured using biotinylated polyclonal goat anti-human IgG (H+L) antibody (Southern Biotech) and detected with a polyclonal goat anti-human IgG (H+L) (Bethyl Laboratories) labeled with Alexa Fluor 647. For the ADC assay, RN927C was captured with biotinylated polyclonal goat anti-human IgG (H+L) antibody (Southern Biotech). The detection of the captured RN927C was done with a polyclonal antibody generated in-house that recognizes PF-06380101. The instrument response was used to construct a standard curve and calculate concentration of study samples and QCs. For both antibody and ADC assays, the lower limit of quantitation was 50 ng/mL. Frozen tumor samples were thawed on ice and homogenized (FastPrep-24 tissue homogenizer) in ice-cold lysis buffer (Invitrogen). The resulting homogenate was centrifuged at 12,000 rpm for 20 minutes at 4°C to separate the supernatant containing RN927C. The samples were analyzed for total antibody and ADC using the same analytic procedures as for serum. Pharmacokinetic data analysis was performed by the noncompartmental method using Phoenix software v. 6.3 (Pharsight).

Formalin-fixed, paraffin-embedded tumor sections were processed and stained with anti-phospho-H3 (Ser10) rabbit pAb (#9701, Cell Signaling Technology) according to the manufacturer's instructions, followed by detection with EnVision HRP-labeled polymer anti-rabbit second Ab (DAKO).

The chemical composition of RN927C is shown in Fig. 1A. RN927C is generated by conjugating AcLys-VC-PABC–PF-06380101 (38, 40) to the transglutaminase tag (LLQGA) located at the C-terminus of the antibody heavy chain (see Materials and Methods). The AcLys linker was chosen to improve the stability of the conjugate in circulation (40). The conjugation reaction yielded a product with DAR of 1.9, and after a single HIC purification step, a homogeneous conjugate with DAR of 2.0 was achieved (Fig. 1B). The process is readily scalable, and high degree of reproducibility was observed among multiple gram scale production lots as measured by cytotoxicity on Trop-2–expressing BxPC3 cells (Fig. 1C).

The uptake of ADCs into the target cells and their subsequent release, usually by lysosomal proteases, is important for cytotoxic drug delivery. The internalization kinetics of RN927C was measured and compared with the parental unconjugated Ab in Trop-2 expression NSCLC cell line HCC-827. Both RN927C and the unconjugated parental Ab internalized efficiently, with t_1/2 at 23.4 and 31.6 minutes, respectively (Fig. 2A). The localization of RN927C was further examined by confocal microscopy, and significant amount of RN927C was found to colocalize with the lysosomal marker LAMP-2 after 2-hour incubation (Fig. 2B).

The presumed mechanism through which RN927C elicits cytotoxicity is by intracellular release of the PF-06380101 payload, which results in the disruption of microtubule polymerization, leading to mitotic arrest, apoptosis, and cell death. To confirm this mechanism, Trop-2⁺ BxPC3 and Trop-2⁻ SW620 cells (Fig. 3A) were incubated with various concentrations of RN927C overnight and processed for immunofluorescence (Fig. 3B and C). Cells in mitotic phase were indicated by positive staining of phosphorylated histone H3 (41). RN927C induced a dramatic increase of positive phosphorylated histone H3 staining in BxPC3 cells treated with RN927C at the lowest experimental concentration of 0.1 μg/mL (Fig. 3B), indicating many treated cells were trapped in the mitotic phase. Cells started to detach from the slides after overnight incubation of RN927C at 10 μg/mL, resulting in loss of cells. In contrast, the untreated cells display the typical low percentage of mitotic cells (Fig. 3B). RN927C incubation did not show any increase of phosphorylated histone H3 staining on Trop-2⁻ SW620 cells at 10 μg/mL (Fig. 3C), indicating that the activity of RN927C to induce mitotic arrest in these cells is dependent on Trop-2 expression. RN927C-induced increase in phosphorylated H3 staining is also observed in vivo on a pancreatic PDX model Pan0146, with the peak staining observed 3 to 5 days after RN927C dosing (Supplementary Fig. S1).

Next, we examined the in vitro cytotoxicity of RN927C on a variety of tumor cell lines. Tumor cell lines from multiple tumor types, including skin, pancreas, lung, head and neck, breast, ovary, and colon, were exposed to RN927C and negative control conjugate at concentrations ranging from 0.04 to 267 nmol/L for 4 days. Most cell lines were readily killed by RN927C with IC₅₀ below 1 nmol/L. Negative control conjugate was not active against any of the cell lines, and RN927C did not show any cytotoxicity toward Trop-2⁻ SW620 cells (Table 1). The lack of killing of Trop-2⁻ SW620 cells is not due to intrinsic insensitivity to the PF-06380101, as free payload is active against SW620 cells with IC₅₀ of 0.305 nmol/L, comparable with that of the Trop-2–expressing BxPC3 cells at 0.116 nmol/L or Fadu cells at 0.683 nmol/L (Table 1). These results indicated that RN927C is efficacious in killing cells derived from multiple tumor types, and the killing requires Trop-2 expression.

To evaluate the pharmacokinetic characteristics of RN927C in vivo, single 1.5 mg/kg dose of RN927C was injected intravenously into BxPC3 tumor-bearing animals with tumor sizes approximately 250 mm³. Both blood and tumor samples were collected at the following time points (0, 0.083, 2, 6, 24, 72, 120, 168, 240, and 336 hours). Total antibody (including both conjugated and unconjugated antibodies) and ADC concentration were measured as described in the Materials and Methods. Note that the immunoassay-based measurement cannot distinguish between ADC of DAR1 and DAR2; therefore, certain degree of payload loss can still occur and not be shown by the assay. ADC and total antibody concentration in serum were similar throughout the time course, indicating only a slow release of the payload in circulation (Fig. 4A). The serum AUC_0–336 for total Ab is 2,488 (μg*h/mL) and for ADC is 2,154 (μg*h/mL), suggesting that most of the Ab remaining in circulation still contains payload PF-06380101. There is a slightly faster loss of payload within the tumors, possibly due to increased internalization and processing of RN927C occurring in Trop-2⁺ tumor cells (Fig. 4B). AUC_0–336 percentage of ADC from the total Ab population in serum is 87% (Fig. 4A), showing good serum linker stability of RN927C in vivo.

The in vivo efficacy of RN927C was tested in multiple tumor models. In a panel of pancreatic tumor xenograft models, including cell line BxPC3 and PDX models Pan0123, Pan0135, and Pan0146, one or two doses (every two or three weeks) of RN927C treatment resulted in sustained tumor growth inhibition/regression at doses between 0.75 to 1.5 mg/kg (Fig. 5A–D). A negative control conjugate consisting of a nonbinding antibody conjugated in the same manner as RN927C showed no effect on these tumor models at 3 mg/kg, the highest dose tested in these studies. In addition, in the Pan0146 model, two doses of RN927C at ≥0.75 mg/kg were more efficacious than gemcitabine treatment given at 75 mg/kg twice weekly for 6 doses (Fig. 5D).

RN927C is also highly efficacious in other solid tumor types tested (Fig. 5E–G). Robust antitumor activity was observed in an ovarian PDX model Ova196756 with single-dose treatment of RN927C at doses ≥1.50 mg/kg, resulting in long-term tumor regression and tumor eradication (Fig. 5E). In a lung PDX model, LG0476, 1.5 mg/kg single injection of RN927C is more efficacious in inducing tumor regression than gemcitabine treatment given at 75 mg/kg twice weekly for 8 doses. Four weekly doses of paclitaxel at 20 mg/kg also only achieved partial tumor growth inhibition in the same study (Fig. 5F).

In a triple-negative breast cancer PDX model CTG-1017, single injection of RN927C induces tumor regression of large tumors (∼830 mm³) for more than 60 days. Tumors that regrew were treated again 63 days after the first RN927C injection, and tumor regression was again achieved for long duration (Fig. 5G). The antitumor effect of RN927C is Trop-2 dependent, as an SW620 colon cancer model devoid of Trop-2 expression showed no response to RN927C compared with the control conjugate, indicating target expression is required for efficacy (Fig. 5H). On the other hand, standard-of-care irinotecan resulted in significant tumor growth inhibition in SW620 model after 4 doses treatment at 40 mg/kg (Fig. 5H).

The nonclinical safety profile of RN927C has been characterized in rats (up to 30 mg/kg) and nonhuman primates (up to 6 mg/kg) in repeat-dose studies (two doses; on day 1 and day 15). The key safety signals observed in cynomolgus monkeys were found in tissues that express Trop-2. Reversible findings included increased mitoses and single-cell necrosis in multiple epithelial tissues in the skin, injection sites, upper alimentary canal (oral mucosa and esophagus), and vagina. These were likely related to the activity of the PF-06380101 payload in the Trop-2–expressing tissues. On the other hand, in rats, organ toxicities were more pronounced in the hematopoietic system and lymphoid tissues, presumably due to the ability to test higher doses in rats due to lack of cross-reactivity of RN927C to tissues that express rat Trop-2.

In this article, we described a novel anti-Trop-2 ADC, RN927C, composed of a humanized anti-Trop-2 antibody with a site specifically conjugated novel MTI-payload, PF-06380101 (30, 38, 40), resulting in improved ADC stability and durable antitumor responses after a single dose of therapy. Trop-2 as a tumor antigen has the advantage of high prevalence in many solid tumor types. It also internalizes efficiently with t_1/2 of approximately 30 minutes (Fig. 2), generally considered a prerequisite for efficient ADC delivery into the cells. We and others have found Trop-2 to be highly expressed on multiple epithelial tumor types, including pancreatic, breast, ovarian, NSCLC, prostate, gastric, and oral cancers (5). However, it is also expressed on a number of normal epithelial tissues, such as skin and oral mucosa. For example, human keratinocytes were found to express Trop-2 at approximately 50,000 copies per cell based on receptor quantitation (data not shown). In general, it is determined that there is a 2- to 10-fold increase of Trop-2 expression in tumors (++ to +++) compared with target-positive normal tissues (+ to ++). The design of an anti-Trop-2 ADC, thus, required a careful balance between efficacy and safety. Several factors were taken into consideration: first, different antibody affinities were explored and a carrier Ab of medium affinity to Trop-2 (K_D ∼ 14 nmol/L at 37°C) was chosen to favor the binding of higher Trop-2–expressing tumor tissue over normal tissues. In the cytotoxicity assays, this medium affinity Ab was similarly active to Abs with 10-fold higher affinity on cell lines with high Trop-2 expression (+++) but showed lower killing activity in low Trop-2–expressing cells (+) that have similar target levels as normal tissues (data not shown). Second, use of a transglutaminase mediated site-specific conjugation technology to enable production of nearly homogeneous Trop-2 ADCs carrying two payloads (DAR2; ref. 30). The majority of ADCs that have been used clinically have been manufactured through cysteine disulfide bond or lysine-based conjugations, producing mixtures of ADC species with various DARs with an average around 4. By decreasing the payload-to-antibody ratio, making a homogeneous ADC, and eliminating high DAR species, we aim to reduce the toxicity to lower Trop-2–expressing tissue. The generally high Trop-2 expression on tumor cells and efficient internalization of the ADC should compensate for the reduced drug loading (Fig. 2). Indeed, most of the tumor cell lines we tested are sensitive to RN927C with IC₅₀ in the subnanomolar range (Table 1). RN927C is highly efficacious in vivo, with tumor regression generally achieved with single injection of RN927C at a dose of ≥1.5 mg/kg, with some sensitive models responding to a dose of 0.75 mg/kg (Fig. 5). Efficacy is seen in pancreatic, ovarian, lung, and breast cancer models and is Trop-2 expression dependent. We also found that RN927C is generally more efficacious than standard of care in these models. For instance, in the pancreatic cancer PDX model Pan0146, RN927C treatment at >1 mg/kg every 2 weeks for two doses induces tumor regression, while multiple twice a week gemcitabine dosing only resulted in tumor growth inhibition (Fig. 5D). Favorable comparison is also observed in NSCLC PDX model LG0476, where a single dose of RN927C treatment outperforms multiple doses of gemcitabine or paclitaxel (Fig. 5F).

Many ADCs in the clinic are limited by the off-target toxicity, most notably hematologic toxicity and neurotoxicity. Off-target toxicities can arise from both nontarget-specific uptake and premature release of payloads. Loss of payloads from ADCs in circulation can result in reduced efficacy and a toxicity profile resembling that of the free payload. Several ADC programs utilizing a cleavable linker and MMAE payload reported neutropenia and peripheral neuropathy as dose-limiting toxicities (42). IMMU-132, an ADC targeting Trop-2 with payload SN38 (active metabolite of irinotecan), also concluded their safety profile to be similar to irinotecan (hematologic and gastrointestinal toxicity), with no/little toxicities related to Trop-2–expressing normal tissues (23). In addition, rapid loss of payloads likely results in accumulation of conjugate-free antibodies, as antibody half-lives are usually longer than the intact ADCs. Cardillo and colleagues reported in a mouse pharmacokinetic study that the AUCs for the intact IMMU-132 versus the carrier antibody hRS7 were 1,516 μg*h/mL and 13,112 μg*h/mL, respectively (21). The persistent presence of an unconjugated antibody could have complicated effects to the ADC efficacy. Although some antibodies can elicit antitumor effect and possibly aid to the response, such as in the case of T-DM1 for HER2⁺ breast cancer, the unconjugated antibody is also a potential competitor for target binding and could negatively impact ADC efficacy. In fact, we have observed a strong inhibitory effect from unconjugated Trop-2 antibodies on the in vivo efficacy of RN927C, and this effect is most profound within the first 3 days (Supplementary Fig. S2).

To reduce the off-target toxicities resulted from deconjugation of free payloads or linker cleavage observed with ADCs made with conventional cysteine disulfide bond or lysine-based conjugations, we utilized stable isopeptide linkage and incorporated an AcLys-VC-PABC linker that was shown to be more stable in circulation (30, 39, 40). We have indeed observed improved stability of our Trop-2 ADC in vivo, evidenced by the small difference between ADC and total Ab pharmacokinetic profiles (Fig. 4). Preclinical exploratory safety studies of RN927C conducted in nonhuman primates with doses up to 6 mg/kg for two doses showed mostly on-target epithelial toxicities (rash and mucositis) that were fully recoverable, consistent with Trop-2 expression on skin and oral mucosa. Notably, there were no adverse hematologic findings observed in monkeys, indicating reduction of off-target toxicity, presumably due to the stable nature of our linker-payload. With a more stable compound, we have observed a potential correlation between Trop-2 expression and efficacy. Trop-2⁻ tumor models, such as SW620, are typically insensitive to RN927C treatment, and tumors with high and homogeneous Trop-2 expression respond well with sustained efficacy (Fig. 5). It is likely that a companion diagnostic test for patient selection based on Trop-2 expression level would enhance clinical activity of a Trop-2–targeting ADC. In summary, we have developed a homogeneous site-specific Trop-2 ADC with enhanced stability, and the preclinical efficacy and safety data support clinical testing of RN927C in multiple solid tumor types.

R. Dushin has ownership interest in Pfizer stock and stock options. D.L Shelton has ownership interest (including patents) in Pfizer, Inc. S.-H. Liu has ownership interest (including patents) in Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Strop, T.-T. Tran, M. Dorywalska, R. Dushin, E. Kraynov, L. Aschenbrenner, B. Han, J. Pons, A. Rajpal, D.L. Shelton, S.-H. Liu

Development of methodology: P. Strop, M. Dorywalska, K. Delaria, R. Dushin, O.K. Wong, D. Zhou, C.J. O'Donnell, D.L. Shelton, S.-H. Liu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-T. Tran, M. Dorywalska, O.K. Wong, W.-H. Ho, A. Wu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Strop, M. Dorywalska, R. Dushin, O.K. Wong, E. Kraynov, L. Aschenbrenner, B. Han, A. Rajpal, D.L. Shelton, S.-H. Liu

Writing, review, and/or revision of the manuscript: P. Strop, M. Dorywalska, R. Dushin, E. Kraynov, L. Aschenbrenner, B. Han, C.J. O'Donnell, J. Pons, D.L. Shelton, S.-H. Liu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Strop

Study supervision: L. Aschenbrenner, J. Pons, A. Rajpal, D.L. Shelton

Other (managed all of the nonclinical safety work from design to interpretation): L. Aschenbrenner

The authors would like to thank Jessica Yu and Jeanette Dilley for their hybridoma work, Kevin Lindquist, Christine Bee, and Yasmina Abdiche for Biosensor analysis, Bryant Chau, Colleen Brown, Ishita Barman, and Michael Chin for protein and antibody production, Jyothirmayee Kudaravalli, Ratika Chopra, and Jing-Tyan Ma for antibody characterization and technical assistance, Victor Lui and Santiago Farias for Mass Spectrometry analytic support, Rachel DeVay for confocal microscopy assistance, and Birte Nolting for technical development in scale up. The authors wish to acknowledge the contributions of Andreas Maderna and Matthew Doroski for their discovery of the PF-06380101 payload and Michael Green and Ramalakshmi Chandrasekaran for preparation of key intermediates used in the synthesis of AcLys-VC-PF-06380101.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.