Gastrointestinal cancer is one of the leading causes of cancer-related mortality in men and women worldwide. The adult stem cell marker LGR5 (leucine-rich repeat-containing, G protein–coupled receptor 5) is highly expressed in a significant fraction of gastrointestinal tumors of the colon, liver, pancreas, and stomach, relative to normal tissues. LGR5 is located on the cell surface and undergoes rapid, constitutive internalization independent of ligand. Furthermore, LGR5-high cancer cells have been shown to exhibit the properties of tumor-initiating cells or cancer stem cells (CSC). On the basis of these attributes, we generated two LGR5-targeting antibody–drug conjugates (ADC) by tethering the tubulin-inhibiting cytotoxic drug monomethyl auristatin E to a highly specific anti-LGR5 mAb via a protease cleavable or noncleavable chemical linker and compared them in receptor binding, cell internalization, and cytotoxic efficacy in cancer cells. Here, we show that both ADCs bind LGR5 with high specificity and equivalent nanomolar affinity and rapidly internalize to the lysosomes of LGR5-expressing gastrointestinal cancer cells. The anti-LGR5 ADCs effectively induced cytotoxicity in LGR5-high gastrointestinal cancer cells, but not in LGR5-negative or -knockdown cancer cell lines. Overall, we demonstrate that the cleavable ADC exhibited higher potency in vitro and was able to eradicate tumors and prevent recurrence in a xenograft model of colon cancer. These findings provide preclinical evidence for the potential of LGR5-targeting ADCs as effective new therapeutics for the treatment and eradication of gastrointestinal tumors and CSCs with high LGR5 expression. Mol Cancer Ther; 15(7); 1580–90. ©2016 AACR.
Gastrointestinal cancer is a leading cause of cancer-related deaths worldwide, with an estimated 2.3 million new cases and 1.4 million mortalities occurring in 2012 from colorectal and stomach cancers alone (1). LGR5 (leucine-rich repeat-containing, G protein–coupled receptor 5) expression is highly elevated in gastrointestinal cancers of the colon, liver, pancreas, and stomach (2–5). The LGR5 receptor was initially identified as a marker of intestinal crypt stem cells and has since been established as a bona fide marker of adult stem cells in several major epithelial tissues (6–10). In normal tissues, LGR5 is exclusively expressed at low levels on the cell surface of the adult stem cells (11, 12). Our group and others showed that LGR5 and its related receptors LGR4 and LGR6 bind the R-spondin growth factors (RSPO1-4) and function to potentiate Wnt signaling (13, 14). The RSPO–LGR system has since emerged as a major ligand–receptor system in the control of stem cell survival and tumorigenesis. A series of studies using mouse models with loss of the tumor suppressor APC (adenomatous polyposis coli) gene have established that LGR5-positive stem cells are the cells-of-origin that drive intestinal and gastric tumor development (6, 15). LGR5-high cancer cells have been shown to exhibit the properties of tumor-initiating cells or cancer stem cells (CSC), which retain the capacity for self-renewal (2, 15–20). CSCs can fuel tumor growth, confer resistance to standard chemotherapy and radiation, and are likely to contribute to recurrent disease (21). Correspondingly, high LGR5 expression in gastrointestinal tumors of different tissue types correlates with higher incidence of metastasis, drug resistance, and poor patient survival (16, 22–24). Thus, LGR5 may serve as an effective therapeutic target for the treatment of gastrointestinal cancers and CSC-based therapy.
Antibody–drug conjugates (ADC) are unique therapeutics that utilize highly specific mAbs to deliver extremely potent cytotoxic drugs to the tumor site through antigen targets enriched on the cell surface (25, 26). The innovation of ADC design resides within the chemical linker that tethers the cytotoxic drug to the mAb, preventing nonspecific drug release to normal tissues during systemic circulation by conferring a therapeutic index to the cytotoxic payload (26). In general, cleavable chemical linkers release drug through enzymatic cleavage (e.g., cathepsins) following internalization, whereas noncleavable linkers must undergo lysosomal degradation of the ADC to release active drug. Regardless of the nature of the linker, ADC efficacy is dependent upon its ability to bind tumor-specific antigen with high affinity, robustly internalize within the tumor cell, and traffic to the lysosome for processing and payload release (26). Our previous work showed that LGR5 is rapidly and constitutively internalized through clathrin-mediated endocytosis independent of ligand (27). LGR5 localizes with Rab7/9 (28), which is essential for proper trafficking to the lysosome. As LGR5+ adult stem cells in the mouse have proven to be dispensable (11, 29), its high expression in tumors and rapid internalization make LGR5 a promising candidate for ADC development.
We generated two LGR5-targeting ADCs by conjugating an anti-LGR5 mAb (anti-LGR5 mAb) to the antimitotic, tubulin-inhibiting agent monomethyl auristatin E (MMAE) via a cleavable or noncleavable chemical linker. We compared binding, internalization, and cytotoxicity using gastrointestinal cancer cell lines. Here, we show that the anti-LGR5 ADC with the cleavable linker was more potent at eradicating gastrointestinal cancer cells in vitro and effectively inhibited tumor growth and recurrence in vivo.
Materials and Methods
Plasmids and cloning
Human RSPO3 (Origene; SC128235) and the Fc of mouse IgG2a from mRSPO1-Fc (Calvin Kuo) were subcloned into pIRESpuro3 (Clontech) to generate the hRSPO3-Fc plasmid. Super 8× TOPFlash and pRL-SV40 (Renilla) were purchased from Addgene and Promega, respectively. For the anti-LGR5 mAb (anti-LGR5 mAb) rat–mouse chimera constructs, codon-optimized sequences encoding the light and heavy chain variable regions for the rat 8F2 antibody (14, 30) were synthesized (Epoch Life Science) and subcloned into expression vectors containing the constant region for κ light chain and the mouse IgG2a heavy chain, respectively (Z. An; ref. 31).
Recombinant proteins, chemicals, and commercial antibodies
Wnt3a-conditioned media (Wnt3aCM) were produced as described previously (13). hRSPO1 was purchased from R&D Systems. hRSPO3-Fc CM was produced as described previously (13), and protein was purified using protein A agarose column, verified by SDS-PAGE, and quantified (OD280 nm). Noggin (Peprotech), N-acetylcysteine (Sigma), N2, B27, and mEGF (Life Technologies) were used for organoid cultures. Monomethyl auristatin E (MMAE) was purchased from ALB Technology. All commercial antibodies were used in accordance to the manufacturer's guidelines. Anti-LGR5 (Abcam; ab75732) and anti-β-actin (Cell Signaling Technology; 4970) were used for Western blot analysis. Anti-Lamp-1 (Cell Signaling Technology; 9091), anti-LGR5 (BD Biosciences; 562731), anti-HA-Alexa-594, TO-PRO-3, and secondary antibodies goat anti-rabbit-Alexa-488, goat anti-mouse-Alexa-594, goat anti-mouse-Alexa-488, and goat anti-rat-Alexa-555 (Life Technologies) were used for immunocytochemistry (ICC).
Cell culture, stable cell line generation
HEK293T, AGS, and HCT15 cells were authenticated and purchased from ATCC. LoVo cells were obtained from the laboratory of Dr. Shao-Cong Sun at MD Anderson Cancer Center (Houston, TX). Cell lines were authenticated utilizing short tandem repeat profiling and routinely tested for mycoplasma. HEK293T cells were cultured in DMEM, AGS in F-12K, and LoVo and HCT15 cells were cultured in RPMI medium. All media were supplemented with 10% FBS and penicillin/streptomycin at 37°C with 95% humidity and 5% CO2. To generate stable LGR5 knockdown cells, LoVo cells were infected with lentivirus particles produced by cotransfecting HEK293T cells with a pLKO.1 vector incorporating the LGR5-targeted shRNA and packaging plasmids, psPAX2 and pMD2.G, using Fugene 6 (Roche). Virus-infected cells were selected with puromycin, followed by clonal selection. The shRNA for human LGR5, TRCN0000011589 (shLGR5) was from GE Dharmacon. Stable HEK293T cell lines overexpressing hLGR5 or hLGR4 and vector cells were generated as reported previously (13).
Synthesis of maleimido-functionalized MMAE
MMAE was modified with a short linker to permit conjugation to partially reduced thiols. A solution of MMAE in anhydrous DMSO was treated with N, N-diisopropylethylamine and 3-(maleimido)propionic acid N-hydroxysuccinimide ester (Thermo Fisher Scientific), and the mixture was stirred 2 days at room temperature. The reaction mixture was evaporated and purified by silica gel chromatography using dichloromethane:methanol (100:3) as an eluant (∼67% yield). The expected molecular weight of the purified product was confirmed by mass spectroscopy: ESI-MS (m/z): 869.8(M+H)+; 891.0(M+Na)+.
Anti-LGR5 mAb production and ADC generation
The anti-LGR5 mAb was transiently expressed in 293F suspension cells by c-transfecting the light and heavy chain constructs using polyethylenimine reagent. Medium was harvested approximately 7 days posttransfection. Antibodies were purified using protein A agarose (Life Technologies) affinity chromatography, eluted in PBS (pH 7.4), analyzed for purity and homogeneity [SDS-PAGE and high-performance liquid chromatography (HPLC)], and quantified (OD 280 nm). The cleavable and noncleavable ADCs were both generated using cysteine-based conjugation methods. Anti-LGR5-mc-vc-PAB-MMAE was generated by The Chemistry Research Solution, LLC by conjugating mc-vc-PAB-MMAE to the mAb through accessible disulfide bonds using similar methodology, as has been described previously (32). Anti-LGR5-mp-MMAE was generated inhouse as follows: Purified mAb was diluted with PBS (pH 7.4) to a final concentration of 3 mg/mL and warmed to 37°C. A stock solution of DTT (50 mmol/L) was prepared and a 2.5 molar equivalent was added. After 3 hours, the reaction was cooled to room temperature. The mAb was purified with a Zeba Desalting Spin Column (Thermo Fisher Scientific) and collected in 0.1 mol/L Na2HPO4, pH 8.3. A 5 molar equivalent of maleimido-functionalized MMAE payload (2 mmol/L in DMSO) was added to the mAb, and payload conjugation was performed for 2 hours at 25°C, followed by 4°C overnight. The ADC was Zeba purified and collected in PBS (pH 7.4). ADCs were analyzed by HPLC to identify possible aggregate formation and by UV-VIS spectroscopy to assess drug–antibody ratio (DAR) using previously published methodology (33). An average DAR of 4 was determined for both anti-LGR5 ADCs.
Western blot analysis and SuperTopFlash luciferase assays
SuperTopFlash assays were performed as described previously (13). Cell were treated with serial dilutions of anti-LGR5 mAb, hRSPO1 (30 ng/mL), or Wnt3aCM (1:5 dilution) as indicated and incubated overnight at 37°C. Luciferase measurements were carried out using the Dual-Glo Luciferase Assay Kit (Promega) according to the manufacturer's protocol and EnVision Multilabel Plate Reader (PerkinElmer). Each condition was tested in quadruplicate for three experiments. For Western blot analysis, protein extraction was performed using RIPA buffer (Sigma) supplemented with protease/phosphatase inhibitors. Cell lysates and tumor samples were incubated at 37°C for 1 hour in 2× SDS buffer prior to equal loading on SDS-PAGE. HRP-labeled secondary antibodies were utilized for detection with the standard ECL protocol. Quantification was performed using ImageJ.
Internalization and binding assays
For ICC internalization assays, HEK293T, LoVo, or AGS cells were seeded into 8-well chamber slides (BD Biosciences) and allowed to adhere overnight. The next day, cells were incubated with anti-LGR5 mAb, anti-LGR5 ADCs, or coincubated with hRSPO3-Fc and commercial anti-LGR5 antibody at 37°C for 1 hour. Cells were washed, fixed, and permeabilized. For lysosome colocalization studies, cells were incubated with anti-Lamp-1 for 45 minutes, followed by anti-rabbit-Alexa-488 and anti-mouse-Alexa-594 for 1 hour at room temperature, respectively. For hRSPO3-Fc studies, cells were incubated with anti-mouse-Alexa-488 and anti-rat-Alexa-555 antibodies. Nuclei were counterstained with TO-PRO-3. Images were acquired using confocal microscopy (Leica TCS SP5 microscope) with the LAS AF Lite software (Leica Microsystems, Inc.). For binding assays, stable HEK293T cells overexpressing vector, hLGR5, or hLGR4 were seeded onto poly-D-lysine–coated 96-well plates and allowed to attach overnight. Serial dilutions of anti-LGR5 mAb or anti-LGR5 ADCs were added for 3 hours at 4°C. Plates were washed, fixed, and incubated with anti-mouse-Alexa-594 for 1 hour at room temperature. Fluorescence intensity was quantified using Tecan Infinite M1000 plate reader (Tecan). Data were analyzed with GraphPad Prism software using the logistic nonlinear regression model.
In vitro cytotoxicity
The cytotoxicity of the various antibodies was tested on 293T-hLGR5, 293T-hLGR4, AGS, LoVo, LoVo-shLGR5, and HCT15 cell lines. 293T cells were plated at 500 cells per well, and the cancer cell lines were plated at 1,000 cells per well in 96 half-well plates. Serial dilutions of unconjugated anti-LGR5 antibody, anti-LGR5 ADCs, or free MMAE were added to the plate and allowed to incubate for 4 days at 37°C. Each condition was tested in triplicate in 2 to 4 separate experiments. The resulting cell viability was measured using CellTiter-Glo (Promega) according to the manufacturer's protocol and the EnVision Multilabel Plate Reader (PerkinElmer). Data were analyzed with GraphPad Prism software using the logistic nonlinear regression model.
Evaluation of in vivo efficacy in xenografts
Animal studies were carried out in strict accordance with the recommendations of the Institutional Animal Care and Use Committee of the University of Texas Health Science Center (Houston, TX). Female 6- to 8-week-old nu/nu mice (Charles River Laboratories) were subcutaneously inoculated with 3.5 × 106 LoVo cells in 1:1 mixture with Matrigel (BD Biosciences). After 2 weeks, when tumor size reached approximately 100 mm3, mice were randomized into 3 groups of 6 to 7 mice and intravenously treated with either PBS vehicle, unconjugated anti-LGR5 mAb (8 mg mAb/kg), or anti-LGR5-mc-vc-PAB-MMAE (8 mg ADC/kg) weekly for 3 weeks. Dosing regimen of 8 mg/kg one dose every week for 3 weeks was selected to evaluate ADC efficacy, and any potential toxicity until a complete response was observed. Tumor volumes were measured biweekly and estimated by the formula: tumor volume = length × width2/2. Mice were euthanized when tumor volume reached 1,000 mm3. Tumors and intestines were harvested for Western blot and immunohistochemical analysis, respectively. Intestinal tissue was fixed then paraffin embedded, sectioned, and H&E stained by the Tissue Histopathology Center (University of Texas Health Science Center, Houston, TX). Five mice from the anti-LGR5-mc-vc-PAB-MMAE treatment group were further monitored for tumor recurrence 35 days posttreatment or until the size of first tumor regrowth reached ≥200 mm3.
Ex vivo intestinal organoid culture
Mouse intestinal organoid cultures were established as previously published (34). Briefly, small intestines were harvested from wild-type B6.129P2-Lgr5tm1(cre/ERT2)Cle/J (The Jackson Laboratory) and washed to remove contaminants and villi. Intestinal fragments were incubated in EDTA for 30 minutes, strained, and centrifuged. Crypts were resuspended in Matrigel with DMEM/F12 media containing 10 mmol/L HEPES, GlutaMAX, 1× B27, 1× N2, 1 mmol/L N-acetylcysteine, 50 ng/mL mEGF, 100 ng/mL mNoggin, 20 ng/mL RSPO1, and pen/strep. The media were replenished every 2 to 3 days, and the organoids were mechanically broken down by glass pipette and passaged every 5 to 6 days. Organoids were treated with PBS vehicle, anti-LGR5-mc-vc-PAB-MMAE (1 or 10 μg/mL), or 7 ng/mL MMAE to test toxicity. Organoid viability (∼11–20 organoids/treatment) was quantified using bright field microscopy and image analysis.
Data are expressed as mean ± SEM or SD as indicated in the Results section. Cancer Cell Line Encyclopedia (CCLE) and The Cancer Genome Atlas (TCGA) datasets were partitioned into low, medium, and high expression values based on overall distribution range of each cohort. For TCGA data, normal and tumor samples were compared and analyzed using paired t test. For in vivo studies, differences between groups were analyzed by unpaired t test. Multiple comparisons used one-way ANOVA and Tukey post hoc analysis. P ≤ 0.05 was considered statistically significant.
Increased expression of LGR5 in gastrointestinal tumors
To evaluate the level of LGR5 gene expression in gastrointestinal tumors, we examined whole-transcriptome sequencing datasets of different tumor types from TCGA (35). On the basis of the distribution of log2 RSEM (RNA‐Seq by expectation‐-maximization) or RPKM (reads per kilobase of transcript per million) of the transcripts, LGR5 was shown to be expressed at moderate to high levels in a significant fraction of cases (Fig. 1A). Discretization of the data into low, medium, and high partitioned values demonstrated high level expression of LGR5 in 67%, 28%, 20%, and 22% of cases in the colorectal, liver, pancreatic, and stomach cohorts, respectively (Fig. 1A). Similarly, mining of publicly accessible microarray datasets from the CCLE showed moderate to high LGR5 expression (log2 robust multiarray average normalized) in 49%, 28%, 10%, and 29% of colon, liver, pancreatic, and stomach cancer cell lines, respectively (Fig. 1B; ref. 36). Utilizing individual patient data extracted from the TCGA data matrix, we investigated the difference in LGR5 expression between tumors and matched normal samples from the colon adenocarcinoma (Fig. 1C) and stomach adenocarcinoma (Fig. 1D) cohorts (P < 0.0001). By implementing a fold change cut-off value of 2, we found that 70% of the colon adenocarcinoma and 78% of the stomach adenocarcinoma patient population exhibited increased LGR5 expression in primary tumor compared with normal adjacent tissue (Fig. 1C and D). To assess LGR5 protein expression in gastrointestinal cancer cell lines, we conducted Western blot analysis (Fig. 1E). As we reported previously, HEK293T cells express LGR4 with an insignificant level of LGR5 (13). Therefore, we used established stable cell lines overexpressing recombinant myc-tagged human LGR5 (293T-hLGR5) or vector control (293T-vector) as positive and negative controls for LGR5 expression (Fig. 1E; ref. 13). AGS gastric cancer and LoVo colon cancer cells were found to express high levels of LGR5, whereas no LGR5 was detected in HCT15 colon cancer cells or in Lovo cells with lentivirus-mediated shRNA knockdown of LGR5 (LoVo-shLGR5; Fig. 1E). These Western blot findings were consistent with LGR5 gene expression data extracted from the CCLE for each of the aforementioned cancer cell lines (Fig. 1F). Collectively, these data show that LGR5 expression is significantly elevated in gastrointestinal tumors and cancer cell lines of diverse tissue types.
Characterization of the anti-LGR5 mAb
We produced and purified the 8F2 anti-LGR5 mAb as a rat–mouse chimera (Supplementary Fig. S1A and S1B), which recognizes the N-terminus of LGR5 as shown through epitope mapping (14). Using a fluorescence-based binding assay we showed that anti-LGR5 mAb binds LGR5 expressed on the surface of 293T-hLGR5 cells with high affinity (Kd = 0.92 μg/mL or 6 nmol/L; Fig. 2A). Anti-LGR5 mAb does not bind 293T-vector cells or 293T cells overexpressing related receptor hLGR4 (293T-hLGR4 cells; Supplementary Fig. S2), indicating antigen specificity (Fig. 2A). Next, we tested whether the anti-LGR5 mAb could block binding of RSPO ligand to the LGR5 receptor. 293T-vector, -hLGR4, and -hLGR5 cells lines were incubated with 1.5 μg/mL RSPO3-Fc and an approximate 10-fold molar access of anti-LGR5 mAb and incubated for 1 hour at 37°C. Of note, a commercial rat anti-LGR5 mAb with the identical variable region as our mAb was used as both our mAb and the RSPO3-Fc fusion protein incorporate the mouse IgG2a Fc. Using ICC and confocal analysis we found that both RSPO3-Fc and anti-LGR5 mAb simultaneously bound and cointernalized in 293T-hLGR5 cells (Fig. 2B). As expected, RSPO3-Fc binding was detected in 293T-hLGR4, but not 293T-vector cells, and no binding of anti-LGR5 mAb was detected in either 293T-vector or -hLGR4 cells (Fig. 2B). These data suggest that the anti-LGR5 mAb does not compete with RSPO ligand for binding to the LGR5 receptor and that the mAb epitope is separate from the ligand-binding site of the LGR5 extracellular domain. As RSPO binding to LGR5 potentiates activation of the Wnt/β-catenin signaling pathway, we examined the effect of anti-LGR5 mAb on LGR5 function using the SuperTopFlash assay. This Wnt/β-catenin–responsive luciferase reporter assay measures the level of active or stabilized nuclear β-catenin/TCF (T-cell factor) complexes binding to 8 tandem repeats of the Wnt/TCF transcriptional response element. 293T-vector and -hLGR5 cells were treated with serial dilutions of anti-LGR5 mAb in the presence of vehicle, Wnt3aCM, or Wnt3aCM + RSPO1 (30 ng/mL). As shown in Fig. 2C and D, anti-LGR5 mAb does not inhibit RSPO1-mediated potentiation of the Wnt pathway. Interestingly, at higher concentrations (>0.4 μg/mL or 2.5 nmol/L), anti-LGR5 mAb essentially enhances Wnt/β-catenin signaling in 293T-hLGR5 cells treated with Wnt3aCM and to a greater extent in the presence of exogenous RSPO1 (Fig. 2D). We can speculate that this synergy between the anti-LGR5 mAb and LGR5 ligands is due to changes in receptor conformation, yet the molecular mechanism remains unknown. These studies establish that anti-LGR5 mAb-8F2 is highly specific, binds LGR5 at the cell surface with nanomolar affinity, and can cointernalize with receptor in the presence or absence of ligand without neutralizing RSPO–LGR5 signaling function.
Generation, binding, and internalization of anti-LGR5 ADCs with cleavable and noncleavable chemical linkers
The anti-LGR5 mAb was conjugated to MMAE toxin via the cathepsin B protease–sensitive linker maleimidocaproyl-valine-citrulline-p-aminobenzyloxycarbonyl (mc-vc-PAB), as has been described previously (32, 37), and via a noncleavable maleimidopropionyl linker, resulting in the generation of two separate ADCs, anti-LGR5-mc-vc-PAB-MMAE and anti-LGR5-mp-MMAE, respectively (Fig. 3A). For the latter conjugation method, MMAE was successfully functionalized with a maleimidopropionyl linker and then conjugated to free thiols of the anti-LGR5 mAb under partial reduction conditions. The ADCs were tested for purity and homogeneity using HPLC (not shown), and no aggregates were detected. Coomassie staining of SDS-PAGE run showed a gel shift of the ADCs compared with unconjugated anti-LGR5 mAb, indicating an increased molecular weight due to the addition of MMAE (Supplementary Fig. S1B). Both ADCs were determined to have an equivalent average DAR of 4, making them suitable for direct comparison of binding and functional efficacy.
Using a fluorescence-based binding assay, we found that anti-LGR5-mp-MMAE and anti-LGR5-mc-vc-PAB-MMAE were comparable with unconjugated anti-LGR5 mAb in binding to 293T-hLGR5 cells and lacked interaction with LGR5-negative 293T-vector cells (Figs. 2A and 3B). Furthermore, no measurable binding of ADCs to 293T-hLGR4 cells was detected (Supplementary Fig. S3A). These results indicate that neither conjugation approach affected mAb-binding affinity or specificity. To evaluate the trafficking and internalization of the anti-LGR5 ADCs, AGS (gastric cancer) and LoVo (colon cancer) cells were incubated with anti-LGR5-mp-MMAE or LGR5-mc-vc-PAB-MMAE for 1 hour at 37°C. ICC and confocal microscopy showed that both ADCs were rapidly internalized within the cancer cells (Fig. 3C, top). As ADC efficacy is dependent upon the extent of trafficking to the lysosome for either peptide linker- or total degradation-mediated release of the toxin (26), we investigated intracellular colocalization of ADCs to the acidic lysosomal compartment. Lysosomes were visualized using an antibody specific to the lysosome-associated membrane protein 1 (Lamp-1; Fig. 3C, middle). Both ADCs colocalized with LAMP1 in a distinct and punctate pattern in AGS and LoVo cells (Fig. 3C, bottom). No ADC binding or lysosome colocalization was detected in LGR5-negative LoVo-shLGR5 cells (Fig. 3C, right). Collectively, these results suggest that conjugation of MMAE to anti-LGR5 mAb via a cleavable or noncleavable linker does not affect its binding efficacy and that both anti-LGR5 ADCs are trafficked to the lysosome in a manner consistent with the ADC mechanism of action (26).
Anti-LGR5 ADC with the cleavable linker is more effective at inducing gastrointestinal cancer cell cytotoxicity
To evaluate target-dependent in vitro cytotoxicity, AGS and LoVo cancer cells, which express high levels of LGR5 (Fig. 1E and F), were incubated with increasing concentrations of unconjugated anti-LGR5 mAb, anti-LGR5-mp-MMAE, anti-LGR5-mc-vc-PAB-MMAE, or free MMAE, and cell viability was measured after 4 days (Fig. 4A and B and Supplementary Fig. S3B). Both AGS and LoVo cells demonstrated MMAE sensitivity, with exposure to free drug-mediating cell cytotoxicity at EC50 = 0.08 and 0.46 nmol/L, respectively (Supplementary Fig. S3B). In AGS cells, anti-LGR5-mc-vc-PAB-MMAE exhibited 11-fold greater potency than anti-LGR5-mp-MMAE at cell killing (EC50 = 0.69 μg/mL or 4.5 nmol/L and 7.9 μg/mL or 51 nmol/L, respectively; Fig. 4A). Furthermore, anti-LGR5-mc-vc-PAB-MMAE was more cytotoxic to LoVo cells compared with AGS cells and showed a 20-fold greater potency compared with anti-LGR5-mp-MMAE (EC50 = 0.36 μg/mL or 2.3 nmol/L and 7.3 μg/mL or 47 nmol/L, respectively; Fig. 4B). Unconjugated anti-LGR5 mAb had no significant effect on cell viability in either cell line (Fig. 4A and B). As, anti-LGR5-mc-vc-PAB-MMAE was more effective at killing cancer cells, it was selected as our lead ADC for the remainder of this study. To confirm target antigen specificity, we tested cytotoxicity using LGR5-negative cell lines LoVo-shLGR5 and HCT15 (Figs. 1E and F and 4C and D). Anti-LGR5-mc-vc-PAB-MMAE and anti-LGR5 mAb had no significant effect on either cell line up to 10 μg/mL (Fig. 4C and D). Cytotoxicity was also measured in the receptor overexpression cell lines, 293T-LGR4 and -hLGR5. As shown in Fig. 4E, anti-LGR5-mc-vc-PAB-MMAE killed 293T-hLGR5 cells with high potency (EC50 = 3.6 × 10−3 μg/mL or 2.3 pmol/L) and had no effect on 293T-hLGR4 cells up to 3 μg/mL, further validating target specificity. As expected, anti-LGR5 mAb had no effect on either cell line (Fig. 4E). These data demonstrate that although both cleavable and noncleavable linker anti-LGR5 ADCs can mediate cytotoxicity in LGR5-expressing cancer cell lines in a concentration-dependent manner, the cleavable anti-LGR5-mc-vc-PAB-MMAE exhibits the highest overall potency.
Anti-LGR5 ADC inhibits tumor growth and recurrence in vivo
As LGR5 is normally expressed at low levels throughout the body, with its most notable expression in the stem cells of the intestinal crypts (8, 11), we presumed that if anti-LGR5 ADC treatment were to result in any toxicity in the mouse, it would likely occur within the intestine. Using a fluorescence-based binding assay we showed that anti-LGR5 mAb binds mouse LGR5 expressed on the surface of 293T-mLGR5 cells, however, with much lower affinity and maximum binding compared with human LGR5 (Supplementary Fig. S4A and S4B). Prior to in vivo testing, we established an ex vivo intestinal crypt organoid model and investigated whether anti-LGR5-mc-vc-PAB-MMAE had any effect on organoid viability (Fig. 5A and Supplementary Fig. S4C). Organoids were treated with anti-LGR5-mc-vc-PAB-MMAE (1 or 10 μg/mL) for 1 week and subsequently passaged with no observable cytotoxicity (Fig. 5A and Supplementary Fig. S4). In contrast, exposure to free MMAE resulted in complete cell death (Fig. 5A and Supplementary Fig. S4). This suggests that anti-LGR5-mc-vc-PAB-MMAE should not induce significant target-mediated toxicity in the intestinal system in vivo.
To assess the in vivo efficacy of anti-LGR5-mc-vc-PAB-MMAE, athymic nu/nu mice were implanted with LoVo colon cancer cells and randomized into groups of 6 to 7 mice once tumors reached approximately 100 mm3. Each group was administered intravenously via tail vein with vehicle (PBS), unconjugated anti-LGR5 mAb (8 mg/kg), or anti-LGR5-mc-vc-PAB-MMAE (8 mg/kg) one dose every week for 3 weeks (Fig. 5B). All 7 mice treated with anti-LGR5-mc-vc-PAB-MMAE had a significant reduction in tumor size compared with the other treatment cohorts, with 4 mice exhibiting a complete response (i.e., no detectable tumor; Fig. 5B and Supplementary Fig. S5A). There was a significant reduction in the rate of tumor growth of the anti-LGR5 mAb-treated cohort compared with the vehicle cohort (Fig. 5B). As exposure to anti-LGR5 mAb had no effect on cancer cell viability in vitro (Fig. 4), this suggests an antitumor immune response mechanism, such as antibody-dependent cell-mediated cytotoxicity or complement-dependent cytotoxicity (CDC; refs. 38, 39). No indications of overt toxicity were observed in any group. There were no significant effects in body weight (Supplementary Fig. S5B), and H&E staining of intestinal tissue sections from each treatment group did not show any morphologic changes or signs of cell death (Supplementary Fig. S6). Five mice from the anti-LGR5-mc-vc-PAB-MMAE cohort (2 with residual tumor and 3 with complete regression) were monitored for tumor recurrence subsequent to last treatment dose and until tumor volume reached >200 mm3. Three mice exhibited regrowth, whereas 2 of the mice with complete regression after treatment had no tumor recurrence (Fig. 5C and Supplementary Fig. S7). We then measured the LGR5 expression levels in vehicle, anti-LGR5 mAb, and recurrent tumors from the anti-LGR5-mc-vc-PAB-MMAE group using Western blot analysis (Fig. 5D). On average, the recurrent tumors expressed significantly lower levels of LGR5 compared with vehicle and anti-LGR5 mAb groups. These data provide evidence that anti-LGR5-mc-vc-PAB-MMAE can effectively target and destroy LGR5-expressing tumor and suppress tumor recurrence.
A major dilemma in cancer therapeutics has been the need to target and destroy tumors while minimizing undesired toxicity to normal tissue. First, many cytotoxic drugs fail to have a therapeutic index as they are equally damaging to both cancer and normal cells. Second, several mAb-based therapies that target tumor-specific antigens and limit toxicity to normal tissues do not always destroy cancer cells. ADC technology remedies this conflict by coupling a nonspecific cytotoxic drug to a highly specific mAb (which need not be functionally effective against the tumor on its own) through a chemical linker with high systemic stability. The clinical success of ADCs, brentuximab–vedotin and trastuzumab–emtansine, has validated this approach, and many more ADCs are at various stages of clinical development (25, 26). The selection of target is critical in that it must be highly expressed in tumor compared with normal tissue and be effectively internalized and trafficked to the lysosome for the release of the cytotoxin. We selected LGR5 as a target for ADC development due to its high expression in gastrointestinal cancers of different tissue types relative to its low cell-surface expression in normal tissues, which is restricted to the adult stem cells (2–5, 11). LGR5 is also rapidly and constitutively internalized in a ligand-independent manner, making it an ideal transit for ADC uptake into cancer cells (27).
We generated two LGR5-targeting ADCs, one with the enzyme cleavable valine–citrulline peptide linker (anti-LGR5-mc-vc-PAB-MMAE) and the other with a noncleavable maleimidopropionyl linker (anti-LGR5-mp-MMAE). We found that both ADCs bind LGR5 with equal nanomolar affinity and rapidly internalize to the lysosome of gastrointestinal cancer cells. However, cleavable anti-LGR5-mc-vc-PAB-MMAE exhibited 10- to 20-fold greater potency at killing gastrointestinal cancer cells in vitro. The difference in potency may be a result of enzyme-mediated proteolytic cleavage (e.g., lysosomal cathepsins) being more efficient for drug release than lysosome-mediated degradation of the noncleavable ADC. In fact, cleavable ADCs against poorly internalized targets have shown to be effective in vivo (37), suggesting a role for extracellular proteases in ADC activation. Yet, another mechanism that could explain why the cleavable ADC is more cytotoxic is the bystander effect (38, 40). This phenomenon is attributed to the efflux of hydrophobic cytotoxic drugs, like MMAE, which can diffuse from target cells and subsequently permeate adjacent cells that need not be positive for the target antigen. This bystander effect is not observed with noncleavable ADCs, as its degradation results in charged metabolites (e.g., amino acid-linker-cytotoxin) that are impermeable to the membrane. Although bystander cytotoxicity may reduce ADC specificity, the extent of the effect depends largely on conjugation chemistry and may ultimately be advantageous when treating solid tumors that are composed of heterogeneous cell populations (40).
Using a xenograft model of colon cancer, we demonstrated that treatment with anti-LGR5-mc-vc-PAB-MMAE can effectively eliminate tumors with high LGR5 expression and prevent recurrence. Interestingly, of the tumors that were not completely ablated and eventually grew back, LGR5 expression was significantly lower compared with vehicle and nearly all tumors treated with anti-LGR5 mAb. This suggests that either (i) a subpopulation of LGR5-low/null LoVo cells exists, in which the ADC was ineffective at targeting or (ii) ADC treatment destroyed the majority of LGR5-high LoVo cells, yet the residual LGR5-high cells that were not eliminated essentially gave rise to a LGR5-low/null subpopulation of cells to drive tumor regrowth. However, as 2 mice showed no recurrent tumors subsequent to ADC treatment, it seems more feasible that the latter is true. In fact, it has been reported that both LGR5-positive intestinal crypt stem cells and colon CSCs can interconvert with LGR5-negative cells (19, 41). The conversion of LGR5-positive CSCs was dependent upon exposure to chemotherapy resulting in an LGR5-negative drug-resistant phenotype, although both cell types retained tumor-initiating activity (19). We also observed that anti-LGR5 mAb-treated tumors grew more slowly compared with vehicle and two of them expressed low levels of LGR5. This suggests a potential immune response effect that may have led to an interconversion at some time during the course of treatment. LGR5 mAb-based therapy mediated by CDC has been reported (39). Future studies involving adjustments in ADC dosing conditions and comprehensive testing in LoVo and other cancer models will be required to address these issues.
Several lines of evidence have identified LGR5 as a marker of CSCs, making it an important target for the development of CSC-directed therapeutics. APC deletion in LGR5+ stem cells of the intestinal crypt and stomach leads to rapid transformation and growth of macroscopic adenomas, suggesting LGR5-positive stem cells are tumor-initiating cells (6, 15). LGR5 antibody enrichment of CSCs in colorectal cancer demonstrated that LGR5 enhances clonogenicity (18). In a number of gastrointestinal cancer cell lines, LGR5 was shown to increase drug resistance, and in colon primary tumors, LGR5-high expression correlated with chemoresistance and lower disease-free and overall survival (16, 17, 20, 23, 42, 43). Thus, an LGR5-targeting ADC may not only effectively treat gastrointestinal tumors, but also function to eliminate CSCs. In fact, while this manuscript was in preparation, a report was published further validating the importance and effectiveness of targeting LGR5 for the treatment of colon cancer using different ADCs (44). In summary, our anti-LGR5-mc-vc-PAB-MMAE has demonstrated significant efficacy in gastrointestinal cancer cell lines and in a colon cancer xenograft model, suggesting its potential as a promising new therapeutic to target CSCs, eradicate a LGR5-positive tumors of different tissue types, and prevent recurrence.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: Q. Liu, K.S. Carmon
Development of methodology: Z. An, K.S. Carmon
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Gong, A. Azhdarinia, Q. Liu, K.S. Carmon
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Gong, A. Azhdarinia, S.C. Ghosh, Z. An, Q. Liu, K.S. Carmon
Writing, review, and/or revision of the manuscript: Q. Liu, K.S. Carmon
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Gong, A. Azhdarinia, S.C. Ghosh, W. Xiong, K.S. Carmon
Study supervision: K.S. Carmon
This work is supported by a grant from the Cancer Prevention Research Institute of Texas, RP150640 (to K.S. Carmon).
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.