The correlation between immune subtypes and consensus molecular subtypes in colorectal cancer identifies novel tumour microenvironment profiles, with prognostic and therapeutic implications

BACKGROUND

Solid tumour growth is the consequence of a complex interplay between cancer cells and their microenvironment. Recently, a new global transcriptomic immune classification of solid tumours has identified six immune subtypes (ISs) (C1-C6). Our aim was to specifically characterise ISs in colorectal cancer (CRC) and assess their interplay with the consensus molecular subtypes (CMSs).

METHODS

Clinical and molecular information, including CMSs and ISs, were obtained from The Cancer Genome Atlas (TCGA) (N = 625). Immune cell populations, differential gene expression and gene set enrichment analysis were performed to characterise ISs in the global CRC population by using CMSs.

RESULTS

Only 5 ISs were identified in CRC, predominantly C1 wound healing (77%) and C2 IFN-γ dominant (17%). CMS1 showed the highest proportion of C2 (53%), whereas C1 was particularly dominant in CMS2 (91%). CMS3 had the highest representation of C3 inflammatory (7%) and C4 lymphocyte depleted ISs (4%), whereas all C6 TGF-β dominant cases belonged to CMS4 (2.3%). Prognostic relevance of ISs in CRC substantially differed from that reported for the global TCGA, and ISs had a greater ability to stratify the prognosis of CRC patients than CMS classification. C2 had higher densities of CD8, CD4 activated, follicular helper T cells, regulatory T cells and neutrophils and the highest M1/M2 polarisation. C2 had a heightened activation of pathways related to the immune system, apoptosis and DNA repair, mTOR signalling and oxidative phosphorylation, whereas C1 was more dependent of metabolic pathways.

CONCLUSIONS

The correlation of IS and CMS allows a more precise categorisation of patients with relevant clinical and biological implications, which may be valuable tools to improve tailored therapeutic interventions in CRC patients.

Abstract PD2-09: High-throughput phosphoproteomics (HTPS) in neoadjuvant (NEO) breast cancer (BC) reveals clusters of extreme sensitivity to paclitaxel (T)

Background: NGS has elucidated the mutational landscape in BC. However, the correlation between mutational landscapes and complex phenotypic traits such as drug response is yet unclear. The genomic and transcriptomic aberrations coalesce into a diverse number of phosphorylation-driven patterns of activation of the proteome. These patterns are and associated to clinical outcomes of interest. The specific nature of the phosphopeptides in a profile versus another can be driven by a small number of activated kinases. In the past we relied on mass-spectrometry-based HTPS to build a kinase-based classification of triple-negative breast cancer (TNBC, Nat. Comms, in press), which is more parsimonious than gene-centered classifications and uncovers new actionable targets. We implemented this methodology to find predictors of response to T in early BC.

Methods: Training set: fresh baseline and day+15 tumor biopsies form a trial in NEO HER2-negative BC (N=139) that randomized patients to T (80 mg/m2 weekly x12) plus placebo or nintedanib (150 mg b.i.d) were processed with a hybrid ion trap-orbitrap mass spectrometer after TiO2 phospho-enrichment. Phosphoprofiles were compared pair-wise according to the following factors: day 0 versus day + 15, standard versus experimental arm, pathologic response Symmans/Pusztai (Symmans) 0/1 vs 2/3. Kinases driving each phosphoprofile were solved by an in silico algorithm termed kinase-set enrichment analysis (KSEAS). Validation set: an in-house designed and previously validated mass-spectrometry-to-immunohistochemistry translation algorithm was used to validate the kinases enriched in the profiles from the baseline samples of patients achieving Symmans 0/1 (pCR) or Symmans=3 in the training set. Probes against those kinases were validated in an independent dataset of 160 HER2-negative patients receiving NEO T followed by AC. H-score of each activated kinase was divided in quartiles (Q1 to Q4), and upper-quartiles (Q1) of each kinase were tested in a multivariate logistic regression model to predict pCR adjusted by T, N, age, G, ER/PR and Ki67.

Results: >2.5 millions of spectra were captured, identifying >35000 unique phosphopeptides mapping to >2500 unique proteins per sample. Training set: KSEAS revealed that a high activity of CDK4 and pP70S6K drove the baseline phosphoprofiles of the patients obtaining pCR in the T arm, whereas pSTAT3, pSrc and BARD drove that of the patients achieving Symmans= 3. In the validation set, H-score cut-offs for Q1 were 1.21, 0.69, 0.79, 1.64 and 1.54 for pP70S6K, CDK4, pSTAT3, pSrc and BARD. TNBC patients with Q1 pP70S6K or Q1 CDK4 achieved pCR in 100% of the cases (versus 45% in Q2-4). In the HR+ cases, the pCR rate for Q1 patients was 50% and 61% respectively. In the multivariate model (all patients), having Q1 pP70S6K or CDK4 multiplied by 2.3- and 3.6-fold, respectively, the probability of achieving pCR (P<0.05). Regarding Symmans=3, BARD and pSTAT3 lost significance in the validation but not pSrc (2.5-fold less probability of pCR, P<0.005).

Conclusions: HTPS is a useful tool to find associations with complex traits. TNBC and HR+ patients with high pP70S6K or CDK4 receiving NEO T-based chemotherapy achieve very high rates of pCR.

Citation Format: Quintela-Fandino M, Lluch A, Manso L, Calvo I, Cortes J, Garcia-Saenz JA, Zagorac I, Tapial-Martinez P, Gomez-Lopez G, Fustero C, Muñoz J, Gonzalez-Cortijo L, Mouron S. High-throughput phosphoproteomics (HTPS) in neoadjuvant (NEO) breast cancer (BC) reveals clusters of extreme sensitivity to paclitaxel (T) [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr PD2-09.