Research ArticleCancer

A PTK7-targeted antibody-drug conjugate reduces tumor-initiating cells and induces sustained tumor regressions

See allHide authors and affiliations

Science Translational Medicine  11 Jan 2017:
Vol. 9, Issue 372, eaag2611
DOI: 10.1126/scitranslmed.aag2611

Initiating an antitumor attack

Cancer is notorious for relapsing after treatment, making it difficult to eradicate from a patient’s body. Such relapses are driven by tumor-initiating cells, a type of stem cells that give rise to tumors. Damelin et al. determined that a protein called PTK7 is frequently present on tumor-initiating cells and developed an antibody-drug conjugate for targeting it. The authors demonstrated the effectiveness of this therapy in mouse models of several tumor types and confirmed that it reduces tumor-initiating cells and outperforms standard chemotherapy. The antibody-drug conjugate also had some unexpected benefits, reducing tumor angiogenesis and promoting antitumor immunity, all of which may contribute to its effectiveness.

Abstract

Disease relapse after treatment is common in triple-negative breast cancer (TNBC), ovarian cancer (OVCA), and non–small cell lung cancer (NSCLC). Therapies that target tumor-initiating cells (TICs) should improve patient survival by eliminating the cells that can drive tumor recurrence and metastasis. We demonstrate that protein tyrosine kinase 7 (PTK7), a highly conserved but catalytically inactive receptor tyrosine kinase in the Wnt signaling pathway, is enriched on TICs in low-passage TNBC, OVCA, and NSCLC patient–derived xenografts (PDXs). To deliver a potent anticancer drug to PTK7-expressing TICs, we generated a targeted antibody-drug conjugate (ADC) composed of a humanized anti-PTK7 monoclonal antibody, a cleavable valine-citrulline–based linker, and Aur0101, an auristatin microtubule inhibitor. The PTK7-targeted ADC induced sustained tumor regressions and outperformed standard-of-care chemotherapy. Moreover, the ADC specifically reduced the frequency of TICs, as determined by serial transplantation experiments. In addition to reducing the TIC frequency, the PTK7-targeted ADC may have additional antitumor mechanisms of action, including the inhibition of angiogenesis and the stimulation of immune cells. Together, these preclinical data demonstrate the potential for the PTK7-targeted ADC to improve the long-term survival of cancer patients.

INTRODUCTION

The high rate of disease relapse after therapy remains a central problem in the treatment of cancer patients. Although the advent of immuno-oncology agents has greatly increased the number of patients who experience sustained disease remission, the vast majority of patients do not respond to these therapies when administered as single agents. New therapeutics that deliver potent anticancer drugs specifically to tumors and act in synergy with immune checkpoint inhibitors are likely to more broadly improve clinical outcomes.

Several lines of evidence support the hypothesis that a subpopulation of tumor cells, called cancer stem cells (CSCs), drives tumor growth and metastasis as well as disease relapse (1). Tumor-initiating cells (TICs) include CSCs as well as more differentiated but highly proliferative tumor cells that are functionally tumorigenic (termed tumor progenitor cells). The CSC paradigm does not impose any requirement as to the frequency of tumor cells categorized as CSCs or TICs, and the frequency has been found to vary widely among tumors both across and within tumor types, reflecting the heterogeneous nature of the disease (2). The CSC paradigm predicts that therapeutics that successfully eliminate TICs will improve clinical outcome (3). The identification of therapeutic targets on TICs requires the rigorous in vivo characterization of TICs, remaining cognizant of tumor type and subtype. Such studies can rarely be performed with established cancer cell lines given that they have often evolved under extensive in vitro selection and no longer exhibit the cellular heterogeneity observed in cancer patients (4, 5). Recently, we demonstrated the utility of well-annotated libraries of patient-derived xenografts (PDXs) (6, 7) to identify and characterize TICs in triple-negative breast cancer (TNBC) and ovarian cancer (OVCA) (8).

Antibody-drug conjugates (ADCs) are a therapeutic modality in which a drug, such as a cytotoxic anticancer agent, is conjugated to a monoclonal antibody (mAb) via a chemical linker (9, 10). The mAb binds a target that is more highly expressed in tumors than in normal tissues and thus confers tumor specificity to the drug. ADCs that effectively and specifically target TICs should offer a therapeutic advantage by limiting the opportunity for acquired resistance and by sparing normal tissues from damage (8, 11, 12). In addition, ADCs may act in synergy with immuno-oncology agents by multiple mechanisms such as activating dendritic cells and/or inducing immunogenic cell death (ICD) (13, 14).

Protein tyrosine kinase 7 (PTK7) is a highly conserved member of the pseudokinase family of receptor tyrosine kinases. The lack of observable kinase activity across species is likely explained by substitutions at residues that are typically conserved in kinase domains (15). Genetic and biochemical studies have demonstrated a key function for PTK7 in noncanonical Wnt signaling, and PTK7-deficient embryos exhibit severe developmental defects in planar cell polarity (1619). There is also evidence for additional, possibly context-dependent functions of PTK7 in the vascular endothelial growth factor (VEGF), semaphorin/plexin, and canonical Wnt signaling pathways (20). Oncogenic functions of PTK7 have been documented in colon cancer, lung cancer, and esophageal cancer (2125), and PTK7 promotes cell survival and resistance to chemotherapy in acute myeloid leukemia (26). However, because PTK7 has no catalytic activity, it is impractical to develop a small-molecule inhibitor, as typically done with receptor tyrosine kinases.

Here, we establish that PTK7 is enriched on TICs in at least three tumor types, has increased expression in tumors versus normal tissues, and may be leveraged to deliver potent cytotoxins with a PTK7-targeted ADC. The PTK7-targeted ADC (PF-06647020) elicited potent antitumor activity, reduced TIC frequency in low-passage PDXs, and exhibited a favorable safety profile in nonhuman primates. The PTK7-targeted ADC is being tested in clinical trials, and early results were recently reported (27).

RESULTS

Functional enrichment for TICs by PTK7

An unbiased analysis of isolated TIC populations provided the first indication that PTK7 is enriched on TICs. Next-generation whole transcriptome sequencing of functionally tumorigenic cell subpopulations, described previously (8), showed higher PTK7 expression on TICs from TNBC and OVCA PDXs compared to normal tissues (Fig. 1A), with two of three TNBC PDXs showing a modest increase in expression in paired TIC versus nontumorigenic (NTG) cells (Fig. 1A). NTG cells were experimentally defined as cells unable to initiate tumor growth for all the PDXs shown, even if NTG cell RNA was not sequenced in some cases. To evaluate whether increased PTK7 RNA expression translates to the protein level, anti-PTK7 mAbs were generated and screened for utility by flow cytometry, enzyme-linked immunosorbent assay (ELISA), and immunohistochemistry. The amounts of PTK7 protein on the cell surface were interrogated by flow cytometry and found to be heterogeneous among tumor cells (Fig. 1, C and D). Heterogeneous PTK7 expression in the PDXs was also observed by immunohistochemistry using the PTK7-specific mAb 6M60, which exhibited specific and sensitive staining in formalin-fixed, paraffin-embedded (FFPE) samples (figs. S1 to S3).

Fig. 1. Functional enrichment of TICs by PTK7.

(A) Next-generation sequencing data for PTK7 expression in TNBC and OVCA PDXs, including TIC and NTG cells, compared to a panel of normal vital tissues (“Normal”). Each symbol indicates data from one tissue or PDX: black symbols, normal tissues; green symbols, NTG cells; and red symbols, TICs. The lines that connect green and red symbols indicate the PDX-matched NTG cells and TICs. (B) Summary of tumor growth of sorted populations from OVCA and NSCLC PDXs. (C) OVCA PDX OV55 flow cytometry staining for PTK7 expression (black trace; isotype control, shaded gray trace) and tumor growth curves of sorted and implanted PTK7-positive cells (red squares) versus PTK7-negative cells (green circles) implanted in naïve animals. (D) NSCLC PDX LU176 flow cytometry staining for PTK7 expression (black trace; isotype control, shaded gray trace) and tumor growth curves of PTK7-positive cells (red squares) versus PTK7 negative-cells (green circles) implanted in naïve animals (right).

To determine whether PTK7 expression functionally enriches for TICs, tumor cells were isolated on the basis of PTK7 expression by fluorescence-activated cell sorting (FACS) and then implanted into naïve animals. Cells were positively selected for human epithelial-specific antigen (ESA+), negatively selected for all murine lineages (8), and isolated on the basis of PTK7 expression. In six OVCA and non–small cell lung cancer (NSCLC) PDXs, PTK7-positive cells were highly tumorigenic, whereas PTK7-negative cells appeared devoid of tumorigenic capacity (Fig. 1B). The PTK7 expression profiles and the tumor growth curves from sorted cells are shown for OVCA PDX OV55 (Fig. 1C) and NSCLC PDX LU176 (Fig. 1D). Together, these results demonstrate that PTK7 functionally enriches for TICs in multiple tumor types.

PTK7 overexpression in tumors

By querying The Cancer Genome Atlas (TCGA) databases (2830), overexpression of PTK7 was confirmed in primary breast tumors, lung adenocarcinoma, and lung squamous cell carcinoma, as indicated by significantly higher mean mRNA expression in tumors versus their corresponding normal tissues (P < 0.0001; Fig. 2A). Breast tumors classified as TNBC using the PAM50 gene signature (31) exhibited higher expression than non-TNBC (Fig. 2A, “Other BRCA”). PTK7 expression levels in OVCA were comparable to those in breast cancer (Fig. 2A), but TCGA does not include normal ovary, thus the extent of PTK7 overexpression in OVCA could not be determined. Finally, our analysis of TCGA metadata revealed that higher PTK7 expression in NSCLC is associated with significantly shorter survival (Fig. 2B; hazard ratio, 4.01; log-rank P = 2.2 × 10−16), consistent with the documented oncogenic nature of PTK7 in NSCLC (21) and our observation that PTK7 enriches for TICs in NSCLC (Fig. 1).

Fig. 2. Overexpression of PTK7 in tumor versus normal tissue.

(A) PTK7 mRNA expression in primary human tumors from TCGA compared to matched normal tissues. Red lines indicate the sample means. RPKM, reads per kilobase of exon per million reads mapped. (B) Kaplan-Meier plot of NSCLC patient survival for tumors with high (red line) or low (black line) PTK7 mRNA expression. (C) PTK7 mRNA expression (normalized intensity on microarray) in PDXs of TNBC, BRCA other than TNBC (“Non-TNBC”), OVCA, lung adenocarcinoma (LU-Ad), lung squamous cell carcinoma (LU-SCC), and a panel of normal vital tissues (“Norm Tox”). Red lines indicate the sample means. (D) PTK7 protein expression in PDXs and normal tissue lysates was measured on the MSD platform. Red lines indicate sample means.

PDX tumors also exhibited an increased expression of PTK7 relative to normal tissues. Many TNBC and OVCA PDXs exhibited a robust expression of PTK7 mRNA compared to normal vital tissues (excluding reproductive organs; Fig. 2C). Likewise, lung adenocarcinoma and squamous cell carcinoma PDXs exhibited overexpression of PTK7 mRNA relative to normal tissues, including the lung (Fig. 2C). To analyze the protein expression, we performed an electrochemiluminescence-based sandwich ELISA using the Meso Scale Discovery (MSD) platform to quantify the PTK7 protein in a panel of protein lysates from TNBC, OVCA, and NSCLC PDXs and a panel of normal tissues (Fig. 2D). PTK7 expression was observed in 43 of 47 (91%) OVCA PDXs and 100% of TNBC PDXs tested. Overexpression of PTK7 protein relative to normal tissues was observed in a smaller proportion of lung adenocarcinoma PDXs than was overexpression of PTK7 mRNA in lung adenocarcinoma PDXs and primary tumors (TCGA). The ovary was the normal tissue with the highest PTK7 expression (Fig. 2D, “Norm”).

Immunohistochemistry studies also demonstrated increased expression of PTK7 in tumors versus normal tissues (fig. S4). PTK7 expression within each tumor was heterogeneous, and the extent of heterogeneity varied from tumor to tumor. Consistent with the mRNA and ELISA data, PTK7 staining was observed in some normal tissues, including esophagus, urinary bladder, kidney, mammary gland, lung, ovary, uterus, and digestive tract (fig. S4 and table S1). Together, the overexpression of PTK7 in multiple tumor types, its association with poor survival in NSCLC, and its functional enrichment for TICs indicated that PTK7 could be a suitable therapeutic target.

Induction of mitotic arrest and cell death

Anti-PTK7 mAbs were generated in mice with hybridoma technology and characterized for binding specificity and ability to deliver cytotoxic drugs into tumor cells. Clone 6M24 exhibited nanomolar affinity binding to recombinant antigen and demonstrated target-dependent binding and internalization into cells. 6M24 was humanized by grafting the complementarity-determining regions (CDRs) onto a human immunoglobulin 1 (IgG1) framework, resulting in h6M24 mAb. The humanized mAb efficiently mediated the delivery of the saporin toxin into PTK7-expressing cells with a half-maximal effective concentration (EC50) of 2 pM (0.3 ng/ml), comparable to the original hybridoma (Fig. 3A), which suggested that a PTK7-targeted ADC would elicit cytotoxicity in a target-dependent manner. h6M24 exhibited comparable binding to cynomologus monkey and human antigens; no binding to rat or murine antigen was observed (fig. S5A).

Fig. 3. Structure and characterization of the PTK7-targeted ADC.

(A) Saporin-mediated cytotoxicity of murine 6M24 and humanized h6M24 mAbs. (B) Structure of the PTK7-targeted ADC, composed of h6M24, a dipeptide-based linker (blue), and the Aur0101 microtubule inhibitor payload (black). The linker-payload was conjugated to native cysteine residues on h6M24 with an average drug/antibody ratio of 4. (C) In vitro cytotoxicity of PTK7-targeted ADC (red triangles) and control ADC (black squares) against H661 and H446 cancer cell lines, which have endogenous PTK7 expression. Error bars represent SD of the mean. (D) In vitro cytotoxicity of PTK7-targeted ADC against OVCAR3 cancer cell line and endogenous PTK7 expression under 2D versus 3D culture conditions. Error bars represent SD of the mean. (E) Disruption of microtubule structure in H661 cancer cells upon treatment with PTK7-targeted ADC. Scale bars, 10 μm. (F) Mice with TNBC PDX BR22 tumors were administered one dose of PTK7-targeted ADC (3 mg/kg) (red triangles) or control ADC (black circles). Binding of ADC to tumor cells was detected by anti-hIgG1, and mitotic arrest was detected by pHH3. Data points represent the individual animals, and horizontal lines indicate the median value in each group.

The PTK7-targeted ADC was generated by bioconjugation of the auristatin microtubule inhibitor Aur0101 (32) to endogenous cysteine residues on h6M24 via a valine-citrulline (mc-ValCitPABC or “vc”) linker, with an average drug-to-antibody ratio (DAR) of 4 (Fig. 3B). The PTK7-targeted ADC, also known as h6M24-vc0101 or PF-06647020, and the unconjugated mAb exhibited a comparable binding to PTK7, as evidenced by surface plasmon resonance with recombinant antigen and flow cytometry on live PTK7-expressing cells (fig. S5). A control ADC, composed of a nonbinding mAb and the same linker-payload, exhibited comparable DAR and mouse pharmacokinetics as h6M24-vc0101 (fig. S6).

The PTK7-targeted ADC elicited cytotoxicity against cancer cells with endogenous expression of PTK7, yielding EC50 values of 27.5 ± 20.5 ng/ml against H661, 7.6 ± 5.0 ng/ml against H446, and 105 ± 17 ng/ml against OVCAR3 (Fig. 3, C and D). In vitro cytotoxicity was target-dependent and was not observed with the control ADC (Fig. 3C).

When OVCAR3 cells were grown in a three-dimensional (3D) context embedded in Matrigel extracellular matrix and then treated with PTK7-targeted ADC, the EC50 value (38.9 ± 22.1 ng/ml) was significantly lower than in 2D culture (105 ± 17 ng/ml, P = 0.04; Fig. 3D). This effect might be explained at least in part by increased surface PTK7 protein expression in OVCAR3 cells grown in a 3D matrix (Fig. 3D and fig. S7A). These data suggest that PTK7 expression might be enriched in the 3D microenvironment of an intact tumor, but this could not be evaluated in tumor xenografts because we could not define the human protein content (as opposed to mouse protein from stroma) in the lysates.

To confirm the expected mechanism of action of the ADC, H661 cells were treated with ADC for 48 hours and then stained to visualize the microtubules and the DNA. The cells exposed to PTK7-targeted ADC exhibited disrupted microtubules and mitotic arrest, whereas no effect was seen with the control ADC or unconjugated mAb (Fig. 3E and fig. S8). Free Aur0101 elicited the same effect as PTK7-targeted ADC (fig. S8). Together, these results demonstrated that the mechanism of action of the ADC was dependent on the activity of Aur0101 and target binding by h6M24.

To determine whether the PTK7-targeted ADC could induce similar effects in tumors in vivo, animals bearing TNBC PDX BR22 were administered one dose of PTK7-targeted ADC (3 mg/kg) or control ADC, and samples were harvested after 24, 48, and 96 hours. FFPE tumor samples were subjected to immunohistochemical analysis for the ADC and phosphorylated histone H3 (pHH3), an established biomarker of mitosis. Binding of the ADC to tumor cells was highest at the 24-hour time point and then declined gradually (Fig. 3F). The pHH3 biomarker increased from 24 to 48 hours and remained high at 96 hours (Fig. 3F). Staining for ADC and pHH3 was significantly higher for PTK7-targeted ADC relative to control ADC (P = 0.005 for ADC at 24 hours and for pHH3 at 48 hours), which confirmed the specificity of the response. The offset pharmacodynamic profiles of ADC binding and mitotic arrest likely reflect the time required for intracellular release of auristatin from the ADC as well as the cell cycle dependence of auristatin activity.

Sustained tumor regressions and reduced TIC frequency

The in vivo efficacy of the PTK7-targeted ADC was evaluated head-to-head against standard-of-care chemotherapy and the control ADC in TNBC, OVCA, and NSCLC PDXs (table S2). Specifically, naïve nonobese diabetic (NOD) scid mice were implanted with 50,000 PDX tumor cells and, once tumors reached an average volume of 140 to 200 mm3, were randomized into groups of 6 to 10 animals each and treated with chemotherapy at the maximum tolerable dose, multiple dose levels of PTK7-targeted ADC, control ADC, or vehicle.

In TNBC, the PTK7-targeted ADC (3 mg/kg) outperformed docetaxel in all PDX tested (Fig. 4 and Table 1). The PTK7-targeted ADC induced sustained regressions for 200 days in BR22 (Fig. 4A) and 150 days in most animals bearing BR31 xenografts. Even in the BR13 and BR5 PDXs that showed weaker response, the ADC induced tumor regression that lasted ~35 days after the completion of dosing (Fig. 4B and Table 1). The ADC also induced regressions in xenografts of the established TNBC cell line MDA-MB-468 (Table 1). In NSCLC, the PTK7-targeted ADC outperformed cisplatin in LU176 and achieved comparable activity to paclitaxel in LU135 (Fig. 4A and Table 1). In ovarian cancer, the PTK7-targeted ADC induced sustained regressions in OV55 PDX and strongly inhibited tumor growth in OVCAR3 cell line xenografts (Fig. 4A, fig. S7B, and Table 1). The control ADC elicited little or no antitumor activity (Fig. 4, A and B), and the unconjugated mAb did not exhibit any antitumor activity; these results demonstrate, as expected, that the activity of the PTK7-targeted ADC was dependent on the targeted delivery of auristatin to PTK7-expressing cells.

Fig. 4. Tumor regression and TIC frequency reduction.

(A) Tumor growth curves for TNBC PDX BR22 (left), NSCLC PDX LU176 (middle), and OVCA PDX OV55 (right). Animals were dosed by intraperitoneal injection twice a week for four cycles with PTK7-targeted ADC [1 mg/kg (red open circles) or 3 mg/kg (red solid circles)]; control ADC (3 mg/kg) (green circles); or standard-of-care chemotherapy for that tumor type: docetaxel (20 mg/kg) in TNBC or cisplatin (5 mg/kg), once a week for two cycles in NSCLC (blue triangles). Black triangles represent animals in the control group dosed with vehicle. (B) Tumor growth curves of TNBC PDX BR13 with dose regimens and groups as in (A). (C) Reduction of TIC frequency after treatment with PTK7-targeted ADC was measured by implanting the remaining live tumor cells into naïve animals in a limiting dilution analysis (left); %Negative events indicates the percentage of naïve animals that did not grow tumors after implant. Calculations of TIC frequency (right) were based on Poisson distribution; anti-PTK7 ADC yielded significant reduction in TIC frequency (P = 0.0013) but docetaxel did not (P = 0.09).

Table 1. Summary of preclinical efficacy studies with PTK7-targeted ADC.

SOC, standard of care; TTP, time to progression; TGI, tumor growth inhibition; NL, normal-like; Lu-Ad, lung adenocarcinoma; MMMT, malignant mixed Mullerian type; HGS, high-grade serous; ND, no data.

View this table:

An optimized transduction pharmacokinetics/pharmacodynamics (PK/PD) model with tumor growth rates and nonlinear drug effect (33) was used to characterize the PK/PD relationship and predict the efficacious concentration (Ceff) in humans. Using this approach, the predicted human Ceff was 2.9 to 7.0 μg/ml, with the range representing the data from multiple PDX models.

On the basis of the enrichment of PTK7 on TICs, we tested the impact of PTK7-targeted ADC on TICs by directly assessing the tumorigenic potential of cells remaining after treatment. TNBC PDX BR13 was treated with the PTK7-targeted ADC, control ADC, or docetaxel on days 0, 3, and 7. Representative animals were euthanized on day 10, and live human tumor cells were isolated, counted, and reimplanted into naïve animals in limiting dilutions. The frequency of TICs was calculated using Poisson distribution statistics based on the frequency of tumor incidence in these serial transplants as monitored for 40 weeks. On the basis of these experiments, TIC frequency was significantly reduced after treatment with PTK7-targeted ADC relative to control ADC (5.5-fold, P = 0.0013; Fig. 4C). Treatment with docetaxel reduced the TIC frequency only marginally and not significantly (2.1-fold, P = 0.09; Fig. 4C). Together, these results demonstrated the ability of the PTK7-targeted ADC to specifically deplete TICs in vivo, which likely explains the observed delay or lack of tumor progression.

PTK7 expression in the stromal compartment

During immunohistochemical analysis of primary human tumors, PTK7 staining was observed in the stromal compartment in addition to the tumor cells (Fig. 5A). Stromal staining was observed in primary human tumors, PDXs, and cell line xenografts across tumor types and with multiple anti-PTK7 mAbs that recognize both human and murine PTK7 (Fig. 5A and figs. S1 to S4, S9, and S10). We did not observe any relationship between PTK7 expression in the stroma and the tumor cells; that is, there were tumors in which the PTK7 expression was high in the stroma but minimal or undetectable in the tumor cells and vice versa (figs. S1 to S4 and S9).

Fig. 5. PTK7 expression in the stromal compartment.

(A) Immunohistochemistry of PTK7 (brown) in a representative primary human lung squamous cell carcinoma. Scale bars, 100 μm. (B) PTK7-targeted ADC inhibits the sprouting of endothelial cells, which were fixed and stained for vimentin (green) and DNA (blue) after 7 days. Scale bars, 500 μm. (C) Flow cytometry of PTK7 on dendritic cell populations in human blood. Control traces are indicated by green and yellow. The gating scheme is provided in fig. S12. (D) Immunofluorescence of PTK7 and pDC markers CD123 and CD303 in a representative primary human NSCLC tumor. Scale bars, 20 μm. (E) PTK7 serum concentrations in healthy donors (“Normal”) and patients with various types of cancer were measured on the MSD platform. SCLC, small cell lung cancer.

Because of the different patterns of stromal staining observed in various tumor samples, we hypothesized that there were multiple sources of PTK7 in the stromal compartment. In some tumors, the staining pattern appeared to reflect tumor vasculature; subsequently, we observed the PTK7 expression on cultured human umbilical vein endothelial cells (HUVECs; fig. S11), which was consistent with reports linking PTK7 to VEGF signaling (20, 34). These data suggested that the PTK7-targeted ADC might elicit an antiangiogenic effect, thus targeting tumor vasculature in addition to targeting tumor cells, as previously suggested for prostate-specific membrane antigen (PSMA)–targeted ADCs (35). The PTK7-targeted ADC inhibited sprouting angiogenesis of HUVECs in a concentration-dependent manner (Fig. 5B); the effect was specific to the ADC and was not observed with the unconjugated anti-PTK7 mAb or with the control ADC.

Furthermore, evaluation of cells from human blood revealed moderate PTK7 expression on plasmacytoid dendritic cells (pDCs) and lower but consistently observed expression on myeloid dendritic cells (mDCs) (Fig. 5C and fig. S12). PTK7 expression was typically observed on most, if not all, pDCs and was confirmed with multiple anti-PTK7 antibodies. When we evaluated the PTK7 expression on intratumoral pDC by immunohistochemical analysis, we observed costaining of PTK7 and pDC markers in all 10 NSCLC tumors evaluated (Fig. 5D and fig. S13). These data indicate that PTK7 is expressed on pDCs both in circulation and in tumors.

Another potential source of staining in the stromal compartment is shed (cleaved) PTK7, which has been reported previously (36). We developed an assay to quantify the circulating PTK7 and measured its concentrations in serum from healthy individuals and cancer patients. The mean concentration of PTK7 in serum from healthy humans was 12.4 ± 3.3 ng/ml, whereas the mean concentration for cancer patients was slightly higher, ranging up to 24.6 ± 3.8 ng/ml and with a broader distribution of individual values (Fig. 5E). These results suggest that non–cell-associated cleaved PTK7 may also contribute to the stromal staining observed in tumors.

Exploratory toxicology

We characterized the nonclinical safety profile of the PTK7-targeted ADC in cynomolgus monkeys in repeat-dose studies (once every 3 weeks for three cycles), with doses up to 5 mg/kg. The major toxicities, primarily myelosuppression, were attributed to target-independent (off-target) effects because they had been previously observed with other targeted ADCs containing microtubule inhibitors and cleavable linkers (37). There was no indication of target-dependent toxicity in any of the tissues examined, including those with PTK7 expression such as esophagus, kidney, urinary bladder, and lung (fig. S14).

Exposure of the PTK7-targeted ADC was measured with a ligand-binding assay and was found to increase with increasing dose. The mean systemic exposure of conjugated mAb was similar to the mean systemic exposure of total mAb, which implied stability of the ADC. The toxicokinetic parameters are provided in table S3.

Soluble total PTK7 concentration in monkeys increased after dose administration, with greater concentrations observed at higher doses (fig. S15); increases in total target concentration are commonly observed for mAbs and ADCs when there is a circulating target because of the prolonged half-life of the target imparted by its binding to the mAb. Soluble PTK7 concentrations did not affect ADC concentrations in monkeys, as indicated by the lack of difference between free and total ADC concentrations (fig. S15). PK/PD modeling of the monkey data was completed and translated to humans by incorporating human circulating target concentrations (Fig. 5E), binding affinities of the ADC to the human target (fig. S5), and predicted human PK parameters. Sensitivity analysis demonstrated no significant impact over a range of soluble PTK7 concentrations, as observed in cancer patients. These simulations suggested that, as observed in monkeys, the circulating concentrations of PTK7 were not high enough to significantly affect ADC exposure.

The PK parameters in monkeys were then compared to those from the efficacy studies above. The average concentration in monkeys dosed at 5 mg/kg, at which no severe toxicity was observed, was 11.8 μg/ml over the 3-week dosing interval. Because this value is higher than the predicted human Ceff of 2.9 to 7.0 μg/ml, the preclinical data for this compound indicated that it would have a therapeutic window in cancer patients.

DISCUSSION

We have identified PTK7 to be overexpressed on TICs in TNBC, OVCA, and NSCLC and developed a PTK7-targeted ADC that elicited robust antitumor activity in low-passage PDX tumors. Specifically, the ADC induced sustained tumor regressions and substantially outperformed the standard-of-care chemotherapy in most cases. The lack of tumor recurrence observed in many of our studies likely reflects the ability of the ADC to substantially reduce the TIC frequency, as demonstrated in limiting the dilution serial transplants of human tumor cells after treatment. Despite the expression of PTK7 in certain normal tissues, no target-dependent toxicities were observed in monkeys. The collective experience is that normal tissue expression does not consistently predict the toxicity of ADCs, especially in the context of microtubule inhibitors, such as auristatins, that typically require high antigen expression and actively cycling cells to exert an effect (38). Our PK/PD model integrated the preclinical efficacy data with the monkey toxicology data and suggested that the ADC would have a therapeutic window in cancer patients. The broad distributions of PTK7 expression within each tumor type suggest that a PTK7-targeted ADC might be more effective in certain patient populations, but clinical data will be necessary to evaluate this point.

PTK7 was originally named CCK4 based on the discovery of its overexpression in colon carcinomas, where it was implicated in tumorigenesis (39). PTK7 is also overexpressed with oncogenic functions in other cancers (2125). Here, the use of well-annotated PDX tumors in vivo enabled the discovery of PTK7 enrichment on TICs in TNBC, OVCA, and NSCLC. Given that PTK7 is an orphan receptor and not catalytically active, it has not been possible to develop blocking antibodies or small molecules that inhibit its mechanism of action. We have overcome these limitations and achieved robust antitumor activity with a PTK7-targeted ADC. We used cancer cell lines to characterize the potency and mechanism of the ADC, but we do not make inferences about TICs from the cell line data because established cancer cell lines typically lose heterogeneity and TIC-based hierarchy as a result of in vitro selective pressure (4, 5). Despite the expression of PTK7 in stroma, our TIC data pertain only to the tumor cells because the anti-PTK7 mAb used in the ADC does not recognize murine PTK7.

Our preclinical data demonstrate that the incorporation of an antimitotic drug, such as an auristatin, into an ADC can effectively affect TICs. This result is noteworthy in light of reports that have associated TICs with a mesenchymal state (40, 41). Recent studies have indicated that TICs exhibit plasticity between a quiescent mesenchymal state and a proliferative epithelial state (42, 43), which implies that most, if not all, TICs will cycle through periods of susceptibility to antimitotic drugs. There is also growing evidence that a proliferative state of TICs is required to generate high-burden metastases that affect patient survival. Metastatic progression from stem-like cells was associated with their proliferation and was inhibited by a cyclin-dependent kinase inhibitor (44). A proliferative state of TICs may result from genetic changes even when a mesenchymal phenotype is maintained (45). Consistent with these results, the empirical determination of TIC markers in TNBC and OVCA resulted in the discovery of elevated PTK7 in cells that express what is traditionally considered to be an epithelial marker, CD324 (E-cadherin); the CD324 protein robustly enriched for tumor-initiating capacity, whereas CD324 cells were consistently nontumorigenic (8). These previous reports together with the preclinical results in this study provide support for targeting TICs with an ADC containing an antimitotic drug to affect disease progression and improve clinical outcome.

The PTK7-targeted ADC could also affect tumor cells (including TICs) indirectly. The bystander effect of released auristatin could result in the collapse of the microenvironment supporting the tumor cells. The ADC may stimulate antitumor immunity by ICD upon direct and indirect tumor cell killing and by activation of mDCs by the release of auristatin in the tumor microenvironment (13, 46, 47). Although antitumor activity driven by ICD is likely to be underrepresented in the immunocompromised NOD scid mice leveraged for PDX tumor studies, there may be an impact on residual myeloid and/or natural killer cells, thus facilitating recognition of tumor cells as abnormal.

Our characterization of PTK7 expression in tumor stroma suggests additional potential therapeutic mechanisms of the PTK7-targeted ADC. Consistent with reports of PTK7 function in angiogenesis (34), we demonstrated a target-dependent antiangiogenic effect of PTK7-targeted ADC in vitro. We also observed the expression of PTK7 on pDCs in primary tumors and in circulation. Together with reports that pDCs may be immunosuppressive within the tumor microenvironment (48, 49) and that certain microtubule inhibitors, including auristatins, can reverse their immunosuppressive phenotype (47), our data suggest the possibility that the PTK7-targeted ADC may promote antitumor immunity by delivering auristatin directly to pDCs. WNT5a, which promotes noncanonical WNT signaling as does PTK7, was shown to inhibit the activation of pDCs in melanoma (50) and was recently reported to be a ligand for PTK7 (19). Because h6M24 mAb does not bind to murine PTK7, the efficacy data in this study did not reflect either of these potential mechanisms and are driven only by PTK7 expression on TICs. Finally, our finding that concentrations of circulating PTK7 are generally higher in cancer patients than in healthy individuals is consistent with reports of PTK7 shedding from tumor cells (36), but our PK/PD modeling suggested that ADC exposure would not be compromised by shed PTK7 antigen. The PK/PD simulation suggested that shed antigen can retain ADC in the tumor and thus enhance antitumor activity (51).

The PTK7-targeted auristatin conjugate presented here may have multiple mechanisms of action to inhibit tumor growth, with relative contributions in a particular tumor depending on the proportion of tumor cells to stromal cells, the tumor vasculature, and the parameters that regulate PTK7 shedding, as well as genetic or epigenetic factors that modulate sensitivity to auristatin. Several of these mechanisms may help promote an antitumor immune response: tumor cell killing can induce ICD (46), antiangiogenic activity can promote immune cell infiltration into tumors (52), and delivery of auristatins to dendritic cells can promote an activated antitumor phenotype (14). Some of these mechanisms might be common to ADCs, whereas others are particular to the targeting of PTK7 because of its expression on several cell types in tumors. Clinical studies with the PTK7-targeted auristatin conjugate will be needed to elucidate the contributions of these various mechanisms to the survival of cancer patients.

MATERIALS AND METHODS

Study design

The in vivo studies were designed to support the objective of evaluating the antitumor activity of the PTK7-targeted ADC and, in most cases, of comparing the ADC to standard-of-care chemotherapy. The studies were conducted in NOD scid mice to enable the assessment of efficacy against human tumor PDXs. Sample size (n = 6 to 10 per group) is standard for xenograft studies with frequent measurements of tumor volume and was sufficient to detect statistically significant differences between treatments and/or doses. Cohorts were automatically randomized by the ID Business Solutions (IDBS) electronic notebook statistical package Biobook based on tumor volume. The average tumor volume at study inception was 140 to 200 mm3. Tumor measurements were recorded at least once weekly using digital calipers, and tumor volume was estimated using the equation V = (A × B2)/2, where A is the long axis and B is the short axis of an ellipsoid. Researchers were not blinded during these studies. Study groups were followed until either individual mice or entire cohort measurements reached 1200 mm3, at which point sacrifice was indicated in accordance with Institutional Animal Care and Use Committee (IACUC) protocols.

Patient-derived xenografts

After Institutional Review Board approval at the Cooperative Human Tissue Network and National Disease Research Interchange, patients with breast, ovarian, or lung cancer consented to enable live tumor specimen collection. Each tumor specimen was prepared in Matrigel (BD Biosciences) and implanted into 6- to 10-week-old NOD scid mice (obtained from Harlan Laboratories or Charles River Laboratories) near the mammary fat pad. Research animals were housed and handled according to IACUC-approved protocols and procedures in accordance with American Association for Laboratory Animal Science (AALAS) recommendations. Each PDX was authenticated as unique and matching the primary tumor specimen with the Ion AmpliSeq Sample ID Panel (4779790, Life Technologies) (table S2). The resulting freshly resected xenograft tumors were dissociated to a single cell suspension as described previously (53, 54) and then implanted into naïve animals or prepared for flow cytometry.

Survival analysis

Kaplan-Meier analysis was applied to the bioinformatics data using published freeware (55). PTK7 mRNA expression and patient survival data were plotted for 720 lung adenocarcinoma patients using the tool’s autoselect best cutoff. Our analysis did not indicate a significant correlation between high PTK7 expression and overall survival in ovarian cancer patients and could not be completed in TNBC because the data set is not categorized by that subtype.

RNA expression analysis

Functionally, tumorigenic and NTG tumor cell subpopulations were isolated by FACS and either implanted into NOD scid mice to confirm tumorigenicity (or lack thereof) or immediately pelleted and lysed in Qiagen RLT Plus RNA lysis buffer (Qiagen Inc.) to extract RNA for whole transcriptome sequencing or microarray hybridization. PTK7 expression in individual samples is plotted using the metric RPKM (56), and the geometric mean is indicated by the red horizontal bar. For microarray data, the normalized intensity of PTK7 expression in each sample was plotted with the geometric mean indicated by the red bar. Standard industry practices (57) were used to normalize and transform the intensity values to quantify gene expression for each sample. Statistics reflect group comparisons using a two-tailed unpaired t test in GraphPad Prism.

Antibody generation and characterization

Using standard protocols to generate hybridomas (58), mice were immunized with purified extracellular domain of PTK7, and clones were screened for binding to PTK7 protein and PTK7-expressing cells. 6M24 mAb was cloned and then humanized by grafting of murine heavy and light chain CDR sequences onto a human variable heavy chain (VH) and variable kappa light chain (VK) framework that was selected on the basis of sequence homology and structure similarity to the mouse variable gene (59). Biacore analysis of antigen binding was accomplished using standard techniques on a BIAcore3000 (GE Healthcare).

Protein measurements

PTK7 protein concentrations were measured using two anti-PTK7 mAbs (H2.35 and 6M38, Stemcentrx) as capture and detection reagents developed on the MSD platform. PTK7 protein concentrations were determined by interpolating the values from a standard protein concentration curve that was generated using purified recombinant PTK7 protein. This assay was used to determine PTK7 concentrations in human serum samples and tissue lysates.

ADC bioconjugation

The PTK7-targeted ADC is composed of the h6M24 humanized anti-PTK7 IgG1 mAb conjugated to the cytotoxic agent auristatin Aur0101 (PF-06380101) via a valine-citrulline (mcValCitPABC) linker (60). The bioconjugation was accomplished by derivatizations of the side chains of cysteine residues. Partial reduction of these disulfide linkages provides a distribution of free thiols that can be functionalized with the maleimide handle on the linker. Specifically, h6M24 mAb was partially reduced by addition of 2.4 M excess of tris(2-carboxyethyl)phosphine in 100 mM Hepes (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid) buffer (pH 7.0) and 1 mM diethylenetriaminepentaacetic acid (DTPA) for 2 hours at 37°C. The mcValCitPABC-Aur0101 linker-payload in dimethylacetamide (DMA) was added to the reaction mixture [final concentration, 15% (v/v) of DMA in water] and incubated for 60 min at 25°C, followed by incubation with a threefold excess of N-ethylmaleimide to cap the unreacted thiols and addition of sixfold excess l-Cys (l-cysteine) to quench any unreacted linker-payload. The reaction mixture was dialyzed and purified by size exclusion chromatography (SEC) on an AKTA explorer, Superdex 200. The ADC was further characterized by SEC, hydrophobic interaction chromatography, and liquid chromatography electrospray ionization mass spectrometry. The protein concentration was determined via ultraviolet spectrophotometry. The ADC preparation is composed of a heterogeneous mixture of conjugated mAb molecules that contain an average auristatin/antibody ratio of about 4 mol/mol. The bioconjugation process and product are highly reproducible. The control ADC is composed of human mAb 8.8 (which has not exhibited any binding to human, monkey, or rodent antigens) and the same vc0101 linker-payload, with comparable drug loading and distribution and comparable mouse pharmacokinetics as the PTK7-ADC (fig. S6).

In vitro pharmacology

Cell lines were obtained from the American Type Culture Collection and cultured in the recommended media. The effective concentration of ADC that can kill 50% of cells relative to untreated cells (EC50) was calculated by logistic nonlinear regression on the GraphPad Prism software. Three or more biological replicates were performed for cytotoxicity experiments. Two biological replicates were performed for PTK7 expression in 2D versus 3D culture of OVCAR3. The fibrin gel bead angiogenesis assay was performed as described (61), and three biological replicates were conducted. Additional details are provided in the Supplementary Materials and Methods.

Immunohistochemistry

Anti-PTK7 mAbs were screened for robust and specific detection of PTK7 in FFPE samples. Clone 6M60 exhibited sensitive and specific staining for both human and murine PTK7 and yielded comparable results when used in various formats: the original clone, the subclone 6M60.1, and in chimeric form with rabbit heavy chains to avoid background issues when staining mouse (PDX) samples. FFPE fragments of tumors and normal tissues were prepared and stained using standard histological procedures, as detailed in the Supplementary Materials and Methods. Slides were scanned on the Leica/Aperio AT2 whole-slide digital scanner. For digital image analysis in the pharmacodynamics study, a threshold was applied to the sampled area and each cell was identified as positive or negative; data are shown as the percentage of marker-positive cells.

Immunofluorescence

H661 lung cancer cells were seeded onto a four-well chamber slide system with a CC2-coated growth surface (Thermo Scientific) and treated with PTK7-targeted ADC (4 μg/ml) or control ADC for 48 hours. The cells were then fixed in 3% paraformaldehyde, washed with phosphate-buffered saline (PBS), permeabilized with 0.5% Triton X-100 in PBS, and incubated with anti-tubulin antibody (Sigma-Aldrich #T9026, clone DM1A) in 5% normal goat serum and 0.2% Tween 20 in PBS. Cells were washed and incubated with Alexa 488–conjugated secondary antibody (Invitrogen) and 4′,6-diamidino-2-phenylindole (DAPI) to stain the DNA. Cells were visualized on a Zeiss LSM 510 Meta confocal microscope. Three biological replicates of the experiment were performed.

In vivo efficacy studies

Animals were dosed by intraperitoneal injection twice a week for four cycles with PTK7-targeted ADC (1 to 3 mg/kg) or control (nonbinding) ADC or with standard-of-care chemotherapy for that tumor type (see Table 1 footnote). Tumor regression was defined as a reduction in mean tumor volume after dosing. In cases where tumors regressed, TTP was calculated as the number of days between the first dose and the time at which mean tumor volume increased (regrew) after regression. TGI was calculated using the equation %TGI = [1 − (mean tumor volume of treated)/(mean tumor volume of vehicle)]. Two independent efficacy studies were conducted in BR5, BR13, BR22, and OVCAR3, with comparable results.

TIC frequency

TNBC PDX BR13 tumor-bearing mice were treated with the PTK7-targeted ADC or control ADC, and tumors were harvested at day 10 after the first dose on the basis of when tumors were starting to regress and the absence of most, if not all, ADC from circulation (90% of injected ADC cleared circulation after 96 hours; fig. S6). Tumors were harvested, dissociated, and stained as described above. Three tumors per treatment group were pooled, and live human tumor cells (murine Lineage human ESA+) were isolated by FACS, counted, and implanted into naïve animals in limiting dilution (8 to 10 animals per group). Mice bearing tumors that exceeded 200 mm3 within a 40-week observation period were scored as positive. Poisson distribution statistics were generated by the L-Calc software (STEMCELL Technologies).

Statistical analysis

The number of technical and biological replicates, the data plots, and the statistical analysis for each experiment are discussed in the corresponding sections of Materials and Methods, Results, and in the figure legends.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/372/eaag2611/DC1

Materials and Methods

Fig. S1. Immunohistochemistry of PTK7 in OVCA xenografts.

Fig. S2. Immunohistochemistry of PTK7 in NSCLC xenografts.

Fig. S3. Immunohistochemistry of PTK7 in TNBC xenografts.

Fig. S4. Immunohistochemistry of PTK7 in primary human tumors and normal human tissues.

Fig. S5. Characterization of the PTK7-targeted ADC.

Fig. S6. Characterization of the control ADC.

Fig. S7. Characterization of the OVCAR3 ovarian cancer model.

Fig. S8. Control treatments for mechanism of action studies.

Fig. S9. Examples of PTK7 expression in the stroma.

Fig. S10. Comparison of PTK7 immunohistochemistry with two antibodies.

Fig. S11. PTK7 expression on cultured human endothelial cells.

Fig. S12. PTK7 expression on dendritic cells in human whole blood.

Fig. S13. Summary of PTK7 staining on pDCs in primary NSCLC tumors.

Fig. S14. Histopathology from the monkey toxicology study.

Fig. S15. Monkey toxicokinetics.

Table S1. PTK7 expression in normal human tissues.

Table S2. Characterization of PDX tumor models.

Table S3. Toxicokinetics in monkey.

REFERENCES AND NOTES

  1. Acknowledgments: We acknowledge A. Maderna, M. Doroski, A. Porte, H. Risley, and Z. Chen for synthesis of the linker-payload; N. Prasad and E. Muszynska for analytical characterization of the ADC; E. Upeslacis, D. Leahy, M. Cinque, and J. Lucas for the in vivo studies; J. Golas for assistance with immunohistochemistry; A. Opsahl for digital image analysis; S. Roy and S. Fong for TIC analysis; A. He for protein quantitation; L. Saunders and H. Auch for single-nucleotide polymorphism analysis and PDX genomic DNA sequencing; W. Zhong, K. Geles, A. Giannakou, A. Jackson-Fisher, M. Roy, E. Powell, and A. Hooper for assistance and discussions. Funding: All funding for this study was provided by Stemcentrx and Pfizer Worldwide R&D. Author contributions: M.D., A. Bankovich, J.B., J.L., S.W., A.P., J.A., R.D., J.H., L.K., F.B., H.F., H.P.G., P.S., and S.J.D. designed the experiments. A. Bankovich, J.B., J.L., L.C., S.W., A.P., J.A., E.E., M.C., M.A., C.L., H.R., M.M., J.H., S.L., V.P., E.R., and Y.S. performed the experiments. M.D., A. Bankovich, R.D., J.H., L.K., F.B., A. Betts, M.G., H.F., C.J.O., R.S., M.P., P.E., D.L., O.F., H.P.G., P.S., and S.J.D. analyzed the data. M.D., A. Bankovich, and S.J.D. wrote the manuscript. All authors reviewed the manuscript. Competing interests: All authors were either employees and/or shareholders of Stemcentrx Inc., a privately held and financed company, or Pfizer Inc., a publically traded company. Both Pfizer and Stemcentrx have filed patent applications on aspects of this work. Data and materials availability: Antibodies and ADCs are available through Pfizer, subject to materials transfer agreement requirements (contact M.D., marc.damelin{at}pfizer.com). PDXs are available through AbbVie Stemcentrx, subject to materials transfer agreement requirements (contact A.B., alex.bankovich{at}abbvie.com).

Stay Connected to Science Translational Medicine

Navigate This Article