Background: The CK12 expression signature consists of genes CCL2, CCL3, CCL4, CCL5, CCL8, CCL18, CCL19, CCL21, CXCL9, CXCL10, CXCL11, CXCL13 and was previously shown to associate with the presence of T and B cell rich tertiary lymphoid structures in melanoma and other cancers, and with better patient survival independent of tumor staging and treatment. I-SPY 2 is a biomarker-rich, phase II neoadjuvant platform trial for high risk early stage breast cancer. Here we leverage the I-SPY 2 biomarker program to test the hypothesis that this signature associates with sensitivity to neoadjuvant immunotherapies and potentially other classes cancer therapeutics in breast cancer.
Methods: Data from 1130 patients across 12 arms of I-SPY2 (control (ctr): 210; veliparib/carboplatin (VC): 71; neratinib (N): 114; MK2206: 93; ganitumab: 106; ganetespib: 93; AMG386: 134; TDM1/pertuzumab(P): 52; H/P: 44; pembrolizumab (pembro): 69; durvalumab/olaparib (durva/olap): 71; pembro/SD101: 72) were available for analysis. Pre-treatment FF (n=987) or FFPE (n=143) biopsies were assayed using Agilent gene expression arrays. Signature scores were calculated as the average expression level across the 12 genes, after z-score normalization. We used logistic modeling to assess association with pCR in each arm in a model adjusting for HR and HER2 (likelihood ratio test, p<0.05). This analysis was also performed within HR/HER2 receptor subsets, numbers permitting. We also assessed differences in levels across HR/HER2 subsets using ANOVA and Tukey post-hoc testing. Our statistics are descriptive rather than inferential and do not adjust for multiplicities of other biomarkers outside this study.
Results:CK12 levels associate with HR/HER2 status (ANOVA p=1.07E-14), with higher levels in TN and HR-HER2+ subsets and lower levels in HR+ groups. Overall, patients with higher levels of CK12 were significantly more likely to achieve pCR in all 3 IO arms: pembro (OR=3.4/1SD), pembro/SD101 (OR=4/1SD), and durva/olaparib (OR=2.5/1SD) (LR p<0.05), in a model adjusting for HR status. The CK12 performed favorably in predicting response to pembro/SD101 compared to several other genomic signatures measuring intratumoral immune response. Higher CK12 also associates with response to the ANG1/2 inhibitor AMG386, an agent known to have immune modulatory activity. Higher CK12 was moderately associated with pCR in the control (OR=2.0/1SD), neratinib (OR=1.7/1SD), veliparib/carboplatin (OR=2.0/1SD), ganitumab (OR= 1.7/1SD) and TDM1/P arms (OR=2.1/1SD). Within the HR+HER2- subset, CK12 associated with pCR in all three IO arms, and in the control, AMG386, ganitumab, and ganetespib arms. Within the smaller TN subset, it associated with response in pembro and pembro/SD101 arms but not in durva/olaparib, and in the neratinib and AMG386 arms. Chemokine12 mostly did not associate with pCR in HER2+ subsets, except for HR+HER2+ patients treated with neratinib, and HR-HER2+ patients in the original control arm (trastuzumab).
Conclusion:The CK12 signature is highly predictive of complete pathologic response to immuno-oncology agents and other therapeutics supporting the role of the crosstalk within the tumor immune microenvironment in predicting response across subtypes. This gene expression signature can be readily obtained from microarrays and warrants further investigation in future arms of ISPY2 as a predictive biomarker.