Discoidin domain receptor 2 signaling networks and therapy in lung cancer

Abstract

Discoidin domain receptor 2 (DDR2) is an atypical receptor tyrosine kinase that binds to and is activated by collagen in the extracellular matrix. Recent exon sequencing studies have identified DDR2 to be mutated with a 3-4% incidence in squamous cell cancers (SCCs) of the lung. This article summarizes the current state of knowledge of DDR2 biology and signaling in lung SCC. It also explores the context-dependent role of this receptor as both an oncogene and a tumor suppressor in cancer cells. Promising therapeutic opportunities based on existing and novel targeted small molecule inhibitors against DDR2 may provide new strategies for treating lung SCC patients.

Introduction

Squamous cell cancers (SCCs) of the lung develop from bronchial epithelial cells as a result of squamous metaplasia and are typically found in smokers ¹. The standard of care for this disease is a chemotherapy regimen of four to six cycles of platinum doublets in the first-line setting ². Prognosis is often poor with objective response rates of 30-40% and a median survival of 12 months for patients with stage IIIb/IV disease ³. Clinical trials of targeted therapies in lung SCCs have not shown any patient benefit and in some cases led to greater toxicity. For instance, treatment with the VEGFR inhibitor bevacizumab resulted in an increased risk of bleeding complications in SCC patients in phase II trials and is currently approved only in non-squamous non-small-cell lung carcinoma (NSCLC) ^{4, 5}. Another example is the IGF-1R inhibitor figitumumab which showed increased toxicity when combined with chemotherapy compared to chemotherapy alone, resulting in the closure of phase III clinical trials ⁶. The failure of these trials underscore the need for a comprehensive understanding of the biology of lung SCC, in particular how genetic aberrations manifest at the signaling level to promote tumorigenesis and dictate therapeutic response. The differential responses to targeted therapies between lung adenocarcinomas and SCCs in the clinical setting suggest that the biology and genetic landscape of these two diseases are unique. Indeed, recent lung SCC sequencing studies by Hammerman et al. demonstrate that the genomic aberrations in SCCs are distinct from adenocarcinomas ^{7, 8}.

DDR2 and lung SCCs

In an exon sequencing study, Hammerman et al. identified the Discoidin Domain Receptor 2 (DDR2) gene as a potential oncogenic target in lung SCC ⁸. The authors screened 290 tumors and cell lines and reported a 3.8% incidence of DDR2 point mutations in lung SCC samples. This frequency is similar to the incidence of EML4-ALK in lung adenocarcinomas ⁹. Additional DDR2 mutations at 4.4% frequency have since been identified in an independent cohort of lung SCC patients ¹⁰. It is likely that early targeted sequencing studies failed to identify any aberrations in the DDR2 gene due to small patient cohort size ^{11, 12}. A lower mutation frequency of 1.1% was reported in a large-scale next generation sequencing study in a cohort of 178 SCC patients performed by the TCGA Network ¹³, while no mutations were found in a screen of 166 SCC biopsies from Japanese patients ¹⁴. In the latter case, Sasaki et al proposed that their inability to detect any DDR2 mutations may have been related to ethnic differences in the sample populations¹⁴.

The DDR2 point mutations are not localized to hotspot regions and are distributed throughout the gene, including the extracellular ligand-binding discoidin domain and the cytoplasmic kinase domain. Interestingly, data emerging from lung adenocarcinoma sequencing studies have also identified DDR2 mutations at 2-5% frequency (http://www.cbioportal.org/) ¹⁵. Again, these mutations are spread across the gene but were not found to be significantly enriched over the background mutational rate of the tumors analyzed. The biological role of DDR2 in lung adenocarcinoma remains to be investigated. DDR2 is a receptor tyrosine kinase (RTK) that also functions as an adhesion receptor which is activated by collagen, a major component of the extracellular matrix in the lung ¹⁶. Hammerman et al. showed that a subset of these DDR2 mutants is tumor promoting in cell lines in vitro ⁸. Depletion of mutant DDR2 using RNA interference in lung SCC cells demonstrated oncogene addiction. Importantly, this class of mutations is sensitive to inhibition by the FDA-approved tyrosine kinase inhibitor (TKI) dasatinib in both in vitro assays and in subcutaneous xenograft models in vivo, making it clinically actionable and amenable to rapid advancement into lung SCC trials ^{8, 17}.

There have been two reports of tumor shrinkage in response to dasatinib treatment in lung SCC patients harboring the DDR2 kinase domain S768R mutation ^{8, 18}. In the first case described by Hammerman et al. a combination of erlotinib (an EGFR inhibitor) and dasatinib was administered to a patient whose disease had progressed despite carboplatin and paclitaxel therapy ⁸. Within 2 months of treatment, this patient showed a partial response and tumor shrinkage. She remained on the combination regimen for 14 months prior to discontinuation due to toxicity. In the second case study reported by Pitini et al. the patient had a rare instance of a BCR-ABL positive chronic myelogenous leukemia (CML) and a DDR2 S768R mutation in a lung SCC lesion ¹⁸. Dasatinib therapy resolved both the CML and the lung tumor after 10 weeks and the patient remained clinically well 8 months into the treatment. These case studies provide an indication that a subset of DDR2 mutations are oncogenic in this disease. Ongoing phase II trials assessing the efficacy of dasatinib in lung SCC (NCT01491633 & NCT01514864) are undertaking correlation analysis of DDR2 mutational status with response to therapy in order to validate the clinical relevance of these experimental findings ¹⁹.

DDR2 signaling networks

DDR2 has been implicated in a number of cancer types and has been shown to play a role in driving proliferation and metastasis (see ^{20, 21} for recent excellent reviews on this topic). There is limited information available on the signaling pathways activated by DDR2 upon collagen engagement. Work done by several research groups have shown that DDR2 activates important signaling components including SHC, SRC, JAK, Erk1/2 and PI3K (summarized in Figure 1) ¹⁶. Additionally, Hammerman et al. used phosphorylation of STAT5 as a biological readout for DDR2 activity, although this signaling protein is unlikely to be a bona fide downstream substrate of the DDR2 pathway but rather the result of a survival signaling cascade in the IL-3 dependent Ba/F3 murine cell line ⁸. DDR2 also exhibits crosstalk with other cell surface receptors such as the integrins and RTKs resulting in diversification of downstream signal transduction networks ^22-25. Our laboratory has recently performed a global phosphoproteomic screen of DDR2 signaling activated by collagen I and identified 45 signaling effectors downstream of this receptor ²⁶. In addition to the previously identified signaling nodes Erk1 and PI3K, these effectors also include novel protein substrates such as Lyn, SHP-2, SHIP-2 and PLCL2 (Figure 1). We further show that these signaling events are independent of integrin activation by collagen and are specific to the DDR2 pathway. Similar to previous reports of signal transduction pathway adaptation in the EGFR mutants commonly found in lung adenocarcinoma ^{27, 28}, the identification of DDR2-specific signaling nodes will facilitate future studies on network reprogramming events that occur upon acquisition of DDR2 mutations in cancer cells.

Figure 1.

Depiction of signaling pathways activated downstream of DDR2. Binding of collagen to the extracellular domain of DDR2 triggers the auto-phosphorylation of its cytoplasmic domain. This results in the recruitment of downstream adaptor proteins, kinases and phosphatases including SHC, NCK1, SRC and SHP-2. As a consequence, a series of canonical signaling pathways are initiated including the Erk1/2 and PI3K cascades.

Is DDR2 an oncogene or a tumor suppressor?

There is some controversy regarding the role of DDR2 in cancer. While Hammerman et al. showed that a subset of the DDR2 mutants, including the extracellular discoidin domain variant L63V and kinase domain variant I638F, are oncogenic ⁸, these assays were performed in the absence of its physiological ligand collagen. Furthermore, the activation and phosphorylation status of the receptor in these mutants were not established in this study. Fibrillar collagen inhibits cancer cell growth and one mechanism by which this process occurs is through a DDR2-dependent cell cycle arrest in melanoma and fibrosarcoma cells ^29-31. It is plausible that DDR2 functions in a context-dependent manner and in the presence of its natural ligand collagen, may act as a tumor suppressor rather than an oncogene. In support of this notion, mRNA levels of DDR2 are diminished in lung cancer tumors compared to matched normal lung tissue, suggesting a potential tumor suppressor role ^{11, 14}. This context-dependence is reminiscent of the β1 integrin adhesion receptor that promotes tumor formation in transgenic mouse models of breast cancer but exhibits tumor suppressor-like properties in the TRAMP prostate adenocarcinoma model ^{32, 33}.

Intriguingly, our laboratory has shown that the I638F kinase domain mutant, unlike the L63V and G505 mutants, is incapable of receptor autophosphorylation and activation of its downstream effector SHP-2 when exposed to collagen stimulation ²⁶. These data indicate that I638F is a loss of function mutation which eliminates receptor activation and effector signaling. Consistent with this idea, this kinase domain mutant was able to abrogate the growth suppressive properties of wildtype DDR2 in HEK293 cells grown in 3D collagen gels ²⁶. It is not clear as to the mechanisms by which receptor depletion (by RNA interference) or kinase inhibitor treatment promotes cytotoxicity in lung SCC cell lines bearing the I638F kinase inactive DDR2 mutant ⁸. One potential explanation is that DDR2 may possess additional kinase-independent oncogenic functions that are important for fuelling lung cancer cell growth. Alternatively, in the absence of intrinsic autophosphorylation capacity, DDR2 may be susceptible to transphosphorylation and activation by other tyrosine kinases such as Src or IGF-1R ^{8, 22, 34}. It should be noted that collagen accumulates in the lung during tumor progression which may impact the biological properties of DDR2 ^{35, 36}. In this respect, transgenic mouse models of mutant DDR2-driven lung SCC will undoubtedly shed light on the role of this receptor in the context of the in vivo lung tumor microenvironment ³⁷.

In addition a direct role in regulating tumor cell growth, conflicting literature also exists for the role of DDR2 in supporting metastatic growth. In particular, two groups have examined the function of DDR2 in the host stromal compartment as a regulator of cancer metastasis. In one study, Zhang et al., found that DDR2 expression is enriched in tumor-associated endothelial cells in both colon carcinoma and melanoma ³⁸. Using a transgenic mouse model that is deficient for DDR2 as the host in an experimental system for metastatic melanoma, the authors showed that pulmonary metastasis was significantly reduced in the mutant mice versus control. This was attributed to a decrease in angiogenesis as a result of downregulation of pro-angiogenic genes such as Vegfr2 and upregulation anti-angiogenic components including Ang-1. In contrast, Badiola et al. demonstrated the opposite effect in hepatic metastases of colon carcinoma ³⁹. In their model, downregulation of DDR2 in the host mice generated a pro-metastatic niche in the liver leading to increased VEGF and TGFb expression and metastases. Collectively, these studies emphasize the complex nature of DDR2 interactions in both tumor and stromal cells and highlights the potential challenges associated with targeting this receptor.

DDR2 targeted therapy

There are several candidate small molecule inhibitors that selectively target DDR2. As discussed above, the multi-kinase inhibitors dasatinib, imatinib and nilotinib block DDR2 kinase activity in an ATP-competitive manner with varying levels of potency ¹⁷. Hammerman et al. showed that a panel of lung SCC DDR2 mutants is selectively sensitive to these inhibitors ⁸. A recent report by the same group showed that prolonged exposure of dasatinib to lung cancer cell lines that were dasatinib-sensitive and DDR2-dependent resulted in acquired drug resistance in vitro ⁴⁰. Interestingly, using massively parallel sequencing, the authors identified two distinct mechanisms of dasatinib resistance. One mechanism is the acquisition of the gatekeeper T654I mutation on DDR2 that increases the affinity for ATP and prevents drug binding ⁴¹. The second is loss of NF1 expression through a splice site mutation. NF1 is a negative regulator of Ras and the authors showed that loss of this protein results in the maintenance of Erk1/2 survival signals even when DDR2 is inhibited by dasatinib, conferring drug resistance. This data suggest that in order to overcome potential drug resistance that may arise in lung SCC dasatinib trials, a combination regimen including a Erk1/2 pathway inhibitor may be necessary.

The identification of a gatekeeper mutation also suggests that alternative DDR2 inhibitors may be required to overcome acquired resistance. Additional DDR2 inhibitors that have been isolated include the recently identified alkaloid natural products discoipyrroles A-D as well as the chemotherapeutic Actinomycin D, although the mechanism of action of these compounds is less well characterized and it is not clear if the T654I mutant would be susceptible to inhibition ^42-44. The Gray group has recently developed selective inhibitors against both DDR1 and DDR2 based on a scaffold for type II kinase inhibitors primarily for use as chemical biology tools for functional studies ⁴⁵. These tools will be invaluable in the dissection of mutant DDR2 function in lung SCC progression. Furthermore, these compounds provide the foundation for second generation DDR2 selective inhibitors which have the potential to minimize toxicities seen in multi-target tyrosine kinase inhibitors such as dasatinib. A recent report of a phase II dasatinib trial in lung SCC patient was discontinued due to excess toxicity, including the development of grade 2 pleural effusions ⁴⁶. In addition to DDR2, dasatinib is a broadly-specific TKI that targets BCR-ABL, SRC, PDGFR, c-KIT and DDR1 among others ^{47, 48}. These additional targets may be responsible for the excess adverse effects observed in patients which may be overcome by more selective DDR2 inhibitors. In addition to targeted inhibitors, a chemical proteomic screen has also identified DDR2 as a client protein of the HSP90 chaperone and is susceptible to protein degradation by HSP90 inhibitors ⁴⁹. The potency of this class of inhibitors to specific DDR2 mutants has yet to be established but previous work on EGFR mutants and the EML4-ALK translocation in lung adenocarcinoma would suggest that HSP90 inhibitors may preferentially degrade DDR2 mutants and could have utility as a potential therapeutic in lung SCC ^{50, 51}.

Conclusion

The studies described above highlight the promise of DDR2 as a potential new target in the treatment of lung SCC. Progress has been constrained by our limited knowledge of the biology of the DDR2 receptor and its oncogenic/tumor suppressive properties in lung SCC progression. It is clear that more research is required to delineate the signaling pathways of DDR2 and how these events are modulated in the lung SCC mutants. In light of the failed trials with targeted therapies in SCC, a more comprehensive understanding of the biology of DDR2 and its mutants will be crucial in avoiding similar disappointing outcomes in ongoing and future dasatinib trials.