Abstract P1-07-05: Integrated transcriptional analysis of the triple negative 'proliferation paradox': High proliferation, chemosensitivity, and poor prognosis

Background: In triple-negative breast cancers (TNBC), high proliferation is associated with greater chemosensitivity but, paradoxically, also associated with poor prognosis. We hypothesized that this subset of TNBC has distinct transcriptional features that contribute to poor prognosis.

Approach: To evaluate transcriptional signatures associated with this 'proliferation paradox,' we identified 17 study cohorts of TNBC treated with neoadjuvant chemotherapy (NAC) that reported receptor status, pathologic response, and had expression data from biopsies obtained prior to NAC (n=446). In 6 studies, distant metastasis-free survival (DMFS) data was available for 235 patients with a median follow-up of 31.2 months. We calculated scores for 135 published gene expression signatures for each tumor and evaluated the association with response to chemotherapy and DMFS.

Results: Using recursive partitioning to develop a model of response using a training set (n=340), six of the 135 expression signatures stratify primary tumors into four groups based on signatures of proliferation, BRCA1 mutation, immune, luminal, Ras, and PI3K phenotypes (Table 1.). Response to NAC ranged from 11% to 61% pCR/RCB-I and results were highly concordant when applied to a validation set (n = 106, p = 0.006). The group that was highly proliferative but chemoresistant ('resistant' group) had a distinct transcriptional profile, including lower 'BRCA-ness' and DNA damage expression signatures with higher Ras and stem cell signatures. The 'resistant' group had the poorest DMFS (HR 2.48 [1.52-4.06]; log-rank p=0.002) and this poor survival was validated among chemotherapy-treated TNBCs in a separate dataset, METABRIC. Analyses of only patients with residual disease after NAC demonstrated that the 'resistant' group remained poorest prognosis, with median DMFS of only 31 months from diagnosis.

Conclusions: Using a novel approach to categorize primary TNBC tumors based on six signatures, we can effectively distinguish subgroups with higher versus lower pCR rates. One specific group demonstrated high proliferation but low response to chemotherapy and particularly poor survival. This group demonstrates expression signatures implicating DNA damage repair, stemness, and Ras pathway activity as potential mediators of the phenotype. We identify specific molecular characteristics for investigation in patients within a poor prognosis subgroup of TNBC.

Table 1. Proportion Pathologic Complete Response or RCB-I and Survival Low ProlifHigh Prolif / ResistantHigh Prolif / SensitiveHigh ImmuneSignature StratificationLow GGI + High LuminalHigh GGI + Low BRCA1mut or High RasHigh GGI + High PI3K or Low RasHigh TNBC ImmunepCR/RCB-I rate: Training Set11/105 (10.5%)26/127 (20.5%)42/81 (51.9%)16/27 (59.3%)pCR/RCB-I rate: Validation Set3/23 (13.0%)11/45 (24.4%)13/29 (44.8%)6/9 (66.7%)pCR/RCB-I rate: TOTAL14/128 (10.9%)37/172 (21.5%)55/110 (50.0%)22/36 (61.1%)Overall Survival (n=235)Hazard Ratio (95% CI)1.62 (0.99-2.64)2.48 (1.52-4.06)(ref.)0.47 (0.29-0.77)Signatures GGI (Sotiriou, JNCI 2006); Luminal (Lim, Nat Med 2009); BRCA1 mutation (van't Veer, Nature 2002); Ras (Pratilas, PNAS 2009); PI3K (Gatza, PNAS 2010), TNBC Immune (Lehmann, JCI 2011)

Citation Format: Stover DG, Selfors LM, Winer EP, Partridge AH, Barry WT. Integrated transcriptional analysis of the triple negative 'proliferation paradox': High proliferation, chemosensitivity, and poor prognosis [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P1-07-05.

Abstract P3-07-31: Immune activation signatures identify a subset of ER+ breast cancers with increased pathologic complete response to neoadjuvant chemotherapy

Background: Proliferation is the strongest predictor of response to neoadjuvant chemotherapy in estrogen receptor-positive (ER+) breast cancer. Evidence of immune activation has also been associated with improved response to neoadjuvant chemotherapy in breast cancer. We hypothesized that immune signatures may be associated with response independent of proliferation in ER+ breast cancers.

Approach: We compiled microarray expression data from breast cancer biopsies obtained prior to neoadjuvant chemotherapy on 465 ER-positive/HER2-negative patients by reported pathologic receptor status. We evaluated the association of 118 published gene expression signatures with response to neoadjuvant chemotherapy, based on study-defined pathologic complete response (pCR) versus residual disease (RD).

Results: Overall, 42 of 118 signatures were significantly associated with response to neoadjuvant chemotherapy in ER+ breast cancer (FDR-corrected p<0.05, simple logistic regression). Of those signatures that achieved significance, 52% (22/42) of signatures were proliferation-associated based on correlation to the 11-gene PAM50 proliferation index (Pearson's R2>0.30, p<1e-10). Among signatures that were NOT proliferation-associated, 50% (10/20) were immune-related. Using unsupervised hierarchical clustering of all 118 signatures, these ten immune signatures formed a distinct cluster. Of the 10 signatures, nine were designed to reflect "immune activation" and were highly correlated with each other in ER+ tumors (R2>0.4, p<0.001). The mean of each of these nine signatures was significantly higher in patients with pCR versus RD (FDR-corrected p<0.05, t-test). Patients with higher "immune activation" signatures had increased likelihood of pCR within multiple subgroups of ER+ breast cancer, including luminal B and non-luminal PAM50 subgroups, as well as intermediate- and high-proliferation ER+ breast cancers. For luminal A or low-proliferation breast cancers, "immune activation" signatures were not significantly associated with response, though very few patients achieved pCR in these two subgroups.

Conclusions: Gene expression signatures associated with "immune activation" identify a subset of ER+ breast cancers with higher rates of pCR to neoadjuvant chemotherapy. These "immune activation" signatures appear to be proliferation-independent and may provide additional predictive information to existing gene expression-based approaches for ER+ breast cancer.

Citation Format: Stover DG, Waks AG, Erica ML, Brugge JS, Winer EP, Selfors LM. Immune activation signatures identify a subset of ER+ breast cancers with increased pathologic complete response to neoadjuvant chemotherapy. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P3-07-31.

Abstract P4-08-05: Basement membrane localized tumor cells are protected from HER2-targeted therapy in vivo

Drug resistance compromises the efficacy of HER2-targeted therapy. Results from our laboratory, and previous reports from others, indicate that tumor cell attachment to basement membrane (BM) and other extracellular matrix (ECM) proteins may confer drug resistance. We have discovered a differential drug response between the outer, matrix-attached cells and inner matrix-deprived cells comprising 3D tumor spheroids grown in reconstituted BM (Muranen 2012). The outer matrix-attached cells are resistant to multiple drug therapies due to upregulation of a multi-factor survival program including anti-apoptotic proteins and growth factor receptors. To address whether these observations are relevant in vivo, we utilized a previously described model of human-in-mouse HER2+ ductal carcinoma in situ, which involves the intraductal transplantation of human HER2+ SUM225 tumor cells directly into the ductal network of the mouse mammary gland. The tumors are characterized by organized nests of noninvasive cells confined within a BM surrounded by ECM. We focused on the tumor cell response to short-term lapatinib monotherapy in vivo. A close examination of the tumor architecture revealed that cells closest to the BM, and nearest to the vasculature, display a striking insensitivity to lapatinib whereas the remainder of the tumor undergoes extensive cell death in response to treatment. Further characterization also revealed that cells closest to the BM largely maintain proliferative capacity despite an overall reduction in the total Ki67+ cell population. To further explore the response to treatment, we performed reverse phase protein array (RPPA) analysis on protein lysates prepared from tumor fragments following lapatinib monotherapy. RPPA profile analysis revealed a lapatinib-induced adaptive response program characterized by upregulation of multiple receptor tyrosine kinases, reactivation of AKT and ERK pathway components, and the upregulation of prosurvival BCL2 family proteins. Evaluation of BCL2 in matched SUM225 tumor sections revealed selective upregulation within the BM-localized tumor cells. We predict BCL2 upregulation in the BM-localized tumor cells may represent the critical adaptive response mechanism by which these cell populations would escape lapatinib treatment. BCL2 therefore represents a target for designing combined therapeutic approaches capable of abrogating the protected BM-localized tumor cells. Under 3D culture, lapatinib combined with the BCL2 antagonist ABT737 resulted in greater SUM225 tumor cell synthetic lethality compared to single agent treatment. The combination treatment is currently being explored in vivo. We evaluated the translational relevance of our preclinical observations within the context of HER2+ clinical disease. Patient biopsy samples collected as part of an ongoing clinical trial were assayed for BCL2 before and after short-term HER2-targeted therapy. BCL2 was clearly upregulated post-treatment in a subset of patient samples and largely correlated with a poor response to treatment. Our results suggest that resistant populations may be a source of residual disease post-therapy, therefore identifying and characterizing these cells will be crucial to the prevention of disease recurrence in the clinic.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-08-05.

DL1-1: Adaptive Resistance to Targeted Therapies

While therapies that target specific molecular alterations in human tumors are showing significant efficacy in the clinic, drug resistance remains a major obstacle. Studies from our laboratory and others have provided evidence that PI3K- and HER2− targeted therapies induce a program of molecular responses to drug treatment that significantly contributes to drug resistance. Interestingly, we have found that resistance in vitro and in mouse models is significantly enriched in cells attached to basement membrane. This resistance correlates with specific upregulation of the adaptive response in the matrix-attached cells. Importantly, inhibition of key components of the adaptive response, like BCL-2, abrogates the drug-induced adaptive resistance. We propose a model in which matrix-protection from targeted therapies allows survival of a subpopulation of tumor cells associated with matrix/basement membrane which expand and lead to more stable drug resistance.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr DL1-1.

S5-3: Basement Membrane Localized Tumor Cells Are Protected from HER2−Targeted Therapy In Vivo

Drug resistance compromises the efficacy of HER2−targeted therapy. Results from our laboratory, and previous reports from others, indicate that tumor cell attachment to basement membrane (BM) and other extracellular matrix (ECM) proteins may confer drug resistance. We have discovered a differential drug response between the outer, matrix-attached cells and inner matrix-deprived cells comprising 3-dimensional (3D) tumor spheroids grown in reconstituted basement membrane (T. Muranen and J. Brugge, unpublished data). The outer matrix-attached cells are resistant to multiple drug therapies. To address whether these observations are relevant in vivo, we utilized a previously described model of human-in-mouse HER2−positive ductal carcinoma in situ (DCIS), which involves the intraductal transplantation of human HER2−positive SUM225 tumor cells directly into the ductal network of the mouse mammary gland. The intraductal tumors generated are histologically indistinguishable from human DCIS lesions and recapitulate the architecture of the 3D tumor spheroids. The tumors are characterized by organized nests of noninvasive cells confined within a BM surrounded by ECM. These features permit a direct assessment of differential drug response within the tumor. We focused on the tumor cell response to short-term lapatinib monotherapy in vivo. A close examination of the tumor architecture revealed that cells closest to the basement membrane, and nearest to the vasculature, display a striking insensitivity to lapatinib whereas the remainder of the tumor undergoes extensive cell death in response to treatment. Further characterization also revealed that cells closest to the basement membrane largely maintain proliferative capacity despite an overall significant reduction in the total Ki67-positive cell population. These results provide in vivo evidence that basement membrane-attached tumor cells are protected from lapatinib. We confirmed that these cells maintain HER2 status and also observed an overall reduction in pHER2, pAKT and pERK throughout the tumor suggestive of adaptive response mechanisms, which support the proliferation and survival of these cell populations despite inhibition of the HER2 pathway. To further explore potential mechanisms of the adaptive response, we performed reverse phase protein array (RPPA) analysis on protein lysates prepared from tumor biopsy fragments following lapatinib monotherapy. RPPA profile analysis revealed an adaptive response composed of upregulation of multiple RTKs (HER2, IGFI-R) and altered apoptotic protein levels (Bcl-2, Bim, Bcl-xL) in addition to activation of AKT/S6K and ERK/p38 pathway components. These observations suggest that basement membrane-attached tumor cells may escape from lapatinib response via compensatory activation of these survival mechanisms. Each of these components will serve as targets for designing combined therapeutic approaches capable of targeting the protected basement membrane-attached tumor cells. Our results suggest that resistant populations may be a source of residual disease post-therapy, therefore identifying and characterizing these cells will be crucial to the prevention of disease recurrence in the clinic.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr S5-3.

Vav family proteins couple to diverse cell surface receptors

Vav proteins are guanine nucleotide exchange factors for Rho family GTPases which activate pathways leading to actin cytoskeletal rearrangements and transcriptional alterations. Vav proteins contain several protein binding domains which can link cell surface receptors to downstream signaling proteins. Vav1 is expressed exclusively in hematopoietic cells and tyrosine phosphorylated in response to activation of multiple cell surface receptors. However, it is not known whether the recently identified isoforms Vav2 and Vav3, which are broadly expressed, can couple with similar classes of receptors, nor is it known whether all Vav isoforms possess identical functional activities. We expressed Vav1, Vav2, and Vav3 at equivalent levels to directly compare the responses of the Vav proteins to receptor activation. Although each Vav isoform was tyrosine phosphorylated upon activation of representative receptor tyrosine kinases, integrin, and lymphocyte antigen receptors, we found unique aspects of Vav protein coupling in each receptor pathway. Each Vav protein coprecipitated with activated epidermal growth factor and platelet-derived growth factor (PDGF) receptors, and multiple phosphorylated tyrosine residues on the PDGF receptor were able to mediate Vav2 tyrosine phosphorylation. Integrin-induced tyrosine phosphorylation of Vav proteins was not detected in nonhematopoietic cells unless the protein tyrosine kinase Syk was also expressed, suggesting that integrin activation of Vav proteins may be restricted to cell types that express particular tyrosine kinases. In addition, we found that Vav1, but not Vav2 or Vav3, can efficiently cooperate with T-cell receptor signaling to enhance NFAT-dependent transcription, while Vav1 and Vav3, but not Vav2, can enhance NFkappaB-dependent transcription. Thus, although each Vav isoform can respond to similar cell surface receptors, there are isoform-specific differences in their activation of downstream signaling pathways.

soc-2 encodes a leucine-rich repeat protein implicated in fibroblast growth factor receptor signaling

Activation of fibroblast growth factor (FGF) receptors elicits diverse cellular responses including growth, mitogenesis, migration, and differentiation. The intracellular signaling pathways that mediate these important processes are not well understood. In Caenorhabditis elegans, suppressors of clr-1 identify genes, termed soc genes, that potentially mediate or activate signaling through the EGL-15 FGF receptor. We demonstrate that three soc genes, soc-1, soc-2, and sem-5, suppress the activity of an activated form of the EGL-15 FGF receptor, consistent with the soc genes functioning downstream of EGL-15. We show that soc-2 encodes a protein composed almost entirely of leucine-rich repeats, a domain implicated in protein-protein interactions. We identified a putative human homolog, SHOC-2, which is 54% identical to SOC-2. We find that shoc-2 maps to 10q25, shoc-2 mRNA is expressed in all tissues assayed, and SHOC-2 protein is cytoplasmically localized. Within the leucine-rich repeats of both SOC-2 and SHOC-2 are two YXNX motifs that are potential tyrosine-phosphorylated docking sites for the SEM-5/GRB2 Src homology 2 domain. However, phosphorylation of these residues is not required for SOC-2 function in vivo, and SHOC-2 is not observed to be tyrosine phosphorylated in response to FGF stimulation. We conclude that this genetic system has allowed for the identification of a conserved gene implicated in mediating FGF receptor signaling in C. elegans.

Identification of the cellular source of laminin beta2 in adult and developing vertebrate retinae

The interphotoreceptor matrix (IPM) is a specialized extracellular matrix that surrounds the inner and outer segments of photoreceptors. This matrix contains molecules that may be important in directing photoreceptor differentiation and survival. For example, one molecule that we have previously identified as a component of the IPM, laminin beta2 (formerly known as s-laminin), is implicated in the differentiation of rod photoreceptor cells. Developmentally, laminin beta2 is present before rod birth in a position that is consistent with a role in directing rod differentiation; it is found, in both the rat and skate, in the ventricular space that ultimately becomes the IPM. In this study, we identify the source of laminin beta2 in the adult and developing retina. Both immunohistochemistry in the adult skate retina and in situ hybridizations in the adult rat retina reveal that laminin beta2 is produced by Müller cells. In addition, in the skate but not the rat retina, retinal pigment epithelial cells may be an alternative source of laminin beta2. During development, however, laminin beta2 is present before the birth of Müller glial cells; at this stage of development, laminin beta2 RNA is present within the neuroepithelial layer in a pattern that is consistent with its production by neuroepithelial cells.