IntroductionHumanized immune system (HIS) mouse models, generated by engrafting tumors and hematopoietic cells of human (Hu) origin into immunodeficient host mice, effectively recapitulate key aspects of the crosstalk between human immune cells and tumors. These models represent a valuable tool for the preclinical evaluation of immunotherapies.MethodsIn this study, we provide a comprehensive comparison of two widely used HIS models: the Hu-CD34+ model, which engrafts Hu-hematopoietic cells derived from Hu-CD34+ hematopoietic stem cells (HSCs), and the Hu-PBMC model, which utilizes Hu-peripheral blood mononuclear cells (PBMCs).ResultsWe assess the kinetics, quality and extent of immune cell engraftment, as well as the development of graft-versus-host disease (GVHD). Additionally, we investigate the impact of different immunodeficient host mouse strains on immune cell reconstitution in the Hu-CD34+ model. Both HIS models were engrafted with human tumors derived from either cell lines or patient-derived xenografts (PDX), revealing distinct immune-tumor interactions that influenced antitumor responses. Notably, tumor responses to T-cell-directed therapies, including anti-PD1 antibodies, IL-2-anti-IL-2 antibody complexes, and T-cell engagers, varied across these models.DiscussionOur findings provide novel insights into the properties and limitations of HIS models, offering a critical resource for optimizing next-generation immuno-oncology strategies and guiding the design of future therapeutic interventions.

This study investigates the relationship between genetic ancestry, breast cancer subtypes, and survival outcomes among 951 locally advanced breast cancer cases from Argentina, Brazil, Chile, Mexico, and Uruguay, participating in the Molecular Profile of Breast Cancer Study. Array-based genotyping and ADMIXTURE analysis were used for genetic ancestry evaluation. Breast cancer subtypes were defined by IHC and the gene expression-based PAM50 algorithm. The distribution of genetic ancestry, including European, Indigenous American (IA), African (AFR), and East Asian components, revealed a heterogeneous genetic admixture across countries, with the highest IA ancestry observed in Chile (30.9%) and Mexico (30.8%). Testing the relationship between genetic ancestry and breast cancer subtypes demonstrated that a 10% increase in European ancestry was significantly associated with a 14% decrease in the odds of developing HER2-enriched breast cancer, after adjustment by age, nodal status, and the AFR component (adj. P = 0.021, luminal A as reference). Accordingly, a 10% increase in IA ancestry was associated with a 21% increase in the probability of having HER2-enriched breast cancer (adj. P = 0.022). IA ancestry also significantly increased overall survival after adjustment by age, nodal status, and AFR ancestry, although this result is controversial and may be affected by the size and heterogeneity of the Molecular Profile Breast Cancer Study cohort. Our research confirms previous findings of a high prevalence of HER2-dependent breast tumors among Hispanic/Latina women and strengthens the hypotheses of the existence of either population-specific genetic variant(s) or of other ancestry-correlated factors that impact HER2 expression in breast cancer consistently across different Latin American regions. SIGNIFICANCE:The evidence in this work supports the idea that factors linked to genetic ancestry influence the prevalence of breast cancer subtypes in Latin America, potentially affecting treatment needs in the region.

Rhabdoid tumors (RTs) are aggressive pediatric malignancies with poor prognosis and limited immunotherapy options. Here, we investigate the therapeutic potential of combined PD-L1 (Programmed cell death ligand 1) and TIGIT (T cell immunoreceptor with Ig and ITIM domains) immune checkpoint blockade in RTs using a preclinical murine model that recapitulates key features of human ATRT (Atypical teratoid rhabdoid tumors) subtypes. Transcriptomic analyses of human and murine RTs reveal co-expression of TIGIT and PD-1 (Programmed cell death 1) pathway components and their ligands, particularly in immune-infiltrated subtypes, supporting a rationale for dual blockade. Combination therapy induces complete tumor regression, prolongs survival, and reprograms the tumor immune microenvironment by enriching CD62L⁺ naïve and central memory T cells and promoting selective T-cell clonal expansion. Notably, dual blockade initiates PNAd⁺ (Peripheral node addressin) high endothelial venule (HEV)-like structures, associated with focal lymphocyte clustering and enhanced immune cell recruitment. These findings reveal a mechanistic link between vascular remodeling and immune infiltration and support dual TIGIT and PD-L1 inhibition as a promising immunotherapeutic strategy for RTs.

BACKGROUND:Several guidelines recommend the use of different classifiers to determine the risk of recurrence (ROR) and treatment decisions in patients with HR+HER2- breast cancer. However, data are still lacking for their usefulness in Latin American (LA) patients. Our aim was to evaluate the comparative prognostic and predictive performance of different ROR classifiers in a real-world LA cohort. METHODS:The Molecular Profile of Breast Cancer Study (MPBCS) is an LA case-cohort study with 5-year follow-up. Stages I and II, clinically node-negative HR+HER2- patients (n = 340) who received adjuvant hormone therapy and/or chemotherapy, were analyzed. Time-dependent receiver-operator characteristic-area under the curve, univariate and multivariate Cox proportional hazards regression (CPHR) models were used to compare the prognostic performance of several risk biomarkers. Multivariate CPHR with interaction models tested the predictive ability of selected risk classifiers. RESULTS:Within this cohort, transcriptomic-based classifiers such as the recurrence score (RS), EndoPredict (EP risk and EPClin), and PAM50-risk of recurrence scores (ROR-S and ROR-PC) presented better prognostic performances for node-negative patients (univariate C-index 0.61-0.68, adjusted C-index 0.77-0.80, adjusted hazard ratios [HR] between high and low risk: 4.06-9.97) than the traditional classifiers Ki67 and Nottingham Prognostic Index (univariate C-index 0.53-0.59, adjusted C-index 0.72-0.75, and adjusted HR 1.85-2.54). RS (and to some extent, EndoPredict) also showed predictive capacity for chemotherapy benefit in node-negative patients (interaction P = .0200 and .0510, respectively). CONCLUSION:In summary, we could prove the clinical validity of most transcriptomic-based risk classifiers and their superiority over clinical and immunohistochemical-based methods in the heterogenous, real-world node-negative HR+HER2- MPBCS cohort.

Synchronous bilateral breast cancer (sBBC) occurs after both breasts have been affected by the same germline genetics and environmental exposures. Little evidence exists regarding immune infiltration and response to treatment in sBBCs. Here we show that the impact of the subtype of breast cancer on levels of tumor infiltrating lymphocytes (TILs, n = 277) and on pathologic complete response (pCR) rates ( n = 140) differed according to the concordant or discordant subtype of breast cancer of the contralateral tumor: luminal breast tumors with a discordant contralateral tumor had higher TIL levels and higher pCR rates than those with a concordant contralateral tumor. Tumor sequencing revealed that left and right tumors ( n = 20) were independent regarding somatic mutations, copy number alterations and clonal phylogeny, whereas primary tumor and residual disease were closely related both from the somatic mutation and from the transcriptomic point of view. Our study indicates that tumor-intrinsic characteristics may have a role in the association of tumor immunity and pCR and demonstrates that the characteristics of the contralateral tumor are also associated with immune infiltration and response to treatment.

BackgroundThe success of HER2-positive (HER2+) breast cancer treatment with trastuzumab, an antibody that targets HER2, relies on immune response. We demonstrated that TNFα induces mucin 4 (MUC4) expression, which shields the trastuzumab epitope on the HER2 molecule decreasing its therapeutic effect. Here, we used mouse models and samples from HER2+ breast cancer patients to unravel MUC4 participation in hindering trastuzumab effect by fostering immune evasion.MethodsWe used a dominant negative TNFα inhibitor (DN) selective for soluble TNFα (sTNFα) together with trastuzumab. Preclinical experiments were performed using two models of conditionally MUC4-silenced tumors to characterize the immune cell infiltration. A cohort of 91 patients treated with trastuzumab was used to correlate tumor MUC4 with tumor-infiltrating lymphocytes.ResultsIn mice bearing de novo trastuzumab-resistant HER2+ breast tumors, neutralizing sTNFα with DN induced MUC4 downregulation. Using the conditionally MUC4-silenced tumor models, the antitumor effect of trastuzumab was reinstated and the addition of TNFα-blocking agents did not further decrease tumor burden. DN administration with trastuzumab modifies the immunosuppressive tumor milieu through M1-like phenotype macrophage polarization and NK cells degranulation. Depletion experiments revealed a cross-talk between macrophages and NK cells necessary for trastuzumab antitumor effect. In addition, tumor cells treated with DN are more susceptible to trastuzumab-dependent cellular phagocytosis. Finally, MUC4 expression in HER2+ breast cancer is associated with immune desert tumors.ConclusionsThese findings provide rationale to pursue sTNFα blockade combined with trastuzumab or trastuzumab drug conjugates for MUC4+ and HER2+ breast cancer patients to overcome trastuzumab resistance.

Conventional CD4+ T (Tconv) lymphocytes play important roles in tumor immunity; however, their contribution to tumor elimination remains poorly understood. Here, we describe a subset of tumor-infiltrating Tconv cells characterized by the expression of CD39. In several mouse cancer models, we observed that CD39+ Tconv cells accumulated in tumors but were absent in lymphoid organs. Compared to tumor CD39- counterparts, CD39+ Tconv cells exhibited a cytotoxic and exhausted signature at the transcriptomic level, confirmed by high protein expression of inhibitory receptors and transcription factors related to the exhaustion. Additionally, CD39+ Tconv cells showed increased production of IFNγ, granzyme B, perforin and CD107a expression, but reduced production of TNF. Around 55% of OVA-specific Tconv from B16-OVA tumor-bearing mice, expressed CD39. In vivo CTLA-4 blockade induced the expansion of tumor CD39+ Tconv cells, which maintained their cytotoxic and exhausted features. In breast cancer patients, CD39+ Tconv cells were found in tumors and in metastatic lymph nodes but were less frequent in adjacent non-tumoral mammary tissue and not detected in non-metastatic lymph nodes and blood. Human tumor CD39+ Tconv cells constituted a heterogeneous cell population with features of exhaustion, high expression of inhibitory receptors and CD107a. We found that high CD4 and ENTPD1 (CD39) gene expression in human tumor tissues correlated with a higher overall survival rate in breast cancer patients. Our results identify CD39 as a biomarker of Tconv cells, with characteristics of both exhaustion and cytotoxic potential, and indicate CD39+ Tconv cells as players within the immune response against tumors.

Purposes:Most molecular-based published studies on breast cancer do not adequately represent the unique and diverse genetic admixture of the Latin American population. Searching for similarities and differences in molecular pathways associated with these tumors and evaluating its impact on prognosis may help to select better therapeutic approaches.Patients and Methods:We collected clinical, pathological, and transcriptomic data of a multi-country Latin American cohort of 1,071 stage II-III breast cancer patients of the Molecular Profile of Breast Cancer Study (MPBCS) cohort. The 5-year prognostic ability of intrinsic (transcriptomic-based) PAM50 and immunohistochemical classifications, both at the cancer-specific (OSC) and disease-free survival (DFS) stages, was compared. Pathway analyses (GSEA, GSVA and MetaCore) were performed to explore differences among intrinsic subtypes.Results:PAM50 classification of the MPBCS cohort defined 42·6% of tumors as LumA, 21·3% as LumB, 13·3% as HER2E and 16·6% as Basal. Both OSC and DFS for LumA tumors were significantly better than for other subtypes, while Basal tumors had the worst prognosis. While the prognostic power of traditional subtypes calculated with hormone receptors (HR), HER2 and Ki67 determinations showed an acceptable performance, PAM50-derived risk of recurrence best discriminated low, intermediate and high-risk groups. Transcriptomic pathway analysis showed high proliferation (i.e. cell cycle control and DNA damage repair) associated with LumB, HER2E and Basal tumors, and a strong dependency on the estrogen pathway for LumA. Terms related to both innate and adaptive immune responses were seen predominantly upregulated in Basal tumors, and, to a lesser extent, in HER2E, with respect to LumA and B tumors.Conclusions:This is the first study that assesses molecular features at the transcriptomic level in a multicountry Latin American breast cancer patient cohort. Hormone-related and proliferation pathways that predominate in PAM50 and other breast cancer molecular classifications are also the main tumor-driving mechanisms in this cohort and have prognostic power. The immune-related features seen in the most aggressive subtypes may pave the way for therapeutic approaches not yet disseminated in Latin America.Clinical Trial Registration:ClinicalTrials.gov (Identifier: NCT02326857).

Consistent with better clinical outcome, Epstein-Barr Virus associated gastric cancer (EBVaGC) displays a particular expression pattern of immunological genes and is highly infiltrated by effector immune cells. To decipher potential mechanisms linking tumor infiltrating lymphocytes (TILs) with an expression of immune-related genes in EBVaGC we utilized RNAseq gene expression data from The Cancer Genome Atlas (TCGA) Stomach Adenocarcinoma (STAD) dataset and performed cellular deconvolution analyses by a mixed approach including two complementary approaches (MIXTURE and xCell). In accordance with previous studies, we observed an increase in CD8+ cells and Interferon activity. Further analysis by gene set variation analysis (GSVA) showed an increased exhausted CD8+ signature accompanied by a higher intratumoral Treg signature. Accordingly, an enrichment in both stimulatory and inhibitory immune checkpoint levels on EBVaGC tumors was observed. We next performed differential expression analysis comparing EBVaGC to other STAD subtypes (CIN, GS, HM-SNV, HM-indel) individually. As a result, from this approach, we generated a list of unique EBVaGC-specific genes that was used to construct EBVaGC-specific cell-cell interaction networks. To infer cell-cell interactions underlying the specific tumor immune microenvironment, we implemented the NicheNetR algorithm by including only differentially expressed ligands and targets from the list of EBVaGC genes obtaining a ligand-interaction matrix from which we constructed a EBVaGC-like gene signature. Functional annotation of this ligand-target network showed ligands and targets principally involved in chemotaxis, T-cell apoptosis, and cellular responses to bacterial infection. This signature correlated with better survival not only in EBVaGC but also in other TCGA cohorts such as colon adenocarcinoma and kidney renal clear cell carcinoma. Our findings identified an increase in exhausted CD8+ signature, high intratumoral Treg signature and enrichment in inhibitory immune checkpoint transcripts. We herein propose a new EBVaGC-like gene expression signature that correlates with better survival in multiple cancer types. In-vitro and in-vivo assays will be required for functional validation of these findings as well as clinical trials to confirm the role of our novel EBVaGC-like gene signature in a precision oncology setting. Grant Support: ANID/FONDECYT/POSTDOCTORADO/3201028, CONICYT-FONDAP 15130011, FONDECYT 1191928 & 1180241, IMII P09/016-F, CONICET and Universidad Católica de Córdoba 80020180100029CC, grant 33620180100993CB from the Universidad Nacional de Córdoba. Citation Format: Keila E. Torres, Charlotte N. Hill, Dario Rocha, Elmer Fernández, Ignacio A. Wichmann, Gareth I. Owen, Alejandro H. Corvalan. Characterization of the immune microenvironment of Epstein-Barr virus associated gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2517.

Oral squamous cell carcinoma (OSCC) is one of the most frequent types of oral cancer in developing countries and its burden correlates with exposure to tobacco and excessive alcohol consumption. Toll like receptors (TLRs) are major sensors of inflammatory stimuli, from both microbial and sterile causes and as such, they have been related to tumor progression and metastasis. Here, we evaluated the expression of TLR2, 4 and 9 as well as CD3+, CD8+ and Granzyme B+ cell infiltration by immunohistochemistry in oral samples of 30 patients with OSCC, classified according to their consumption of alcohol. Our findings indicate that there is a significant association between heavy alcohol consumption and tumors with higher expression levels of TLR9. Moreover, patients with TLR9high tumors, as well as those who indicated high consumption of alcohol exhibited a diminished overall survival. TCGA data analysis indicated that TLR9high tumors express a significant increase in some genes related with the oral cavity itself, inflammation and tumor promotion. Our analysis of tumor infiltrating leukocytes demonstrated that the major differences perceived in heavy alcohol consumers was the location of CD8+ T cells infiltrating the tumor, which showed lower numbers intratumorally. Our data suggest the existence of a pathogenic loop that involves alcohol consumption, high TLR9 expression and the immunophenotype, which might have a profound impact on the progression of the disease.

An amendment to this paper has been published and can be accessed via the original article.

The accurate quantification of tumor-infiltrating immune cells turns crucial to uncover their role in tumor immune escape, to determine patient prognosis and to predict response to immune checkpoint blockade. Current state-of-the-art methods that quantify immune cells from tumor biopsies using gene expression data apply computational deconvolution methods that present multicollinearity and estimation errors resulting in the overestimation or underestimation of the diversity of infiltrating immune cells and their quantity. To overcome such limitations, we developed MIXTURE, a new ν-support vector regression-based noise constrained recursive feature selection algorithm based on validated immune cell molecular signatures. MIXTURE provides increased robustness to cell type identification and proportion estimation, outperforms the current methods, and is available to the wider scientific community. We applied MIXTURE to transcriptomic data from tumor biopsies and found relevant novel associations between the components of the immune infiltrate and molecular subtypes, tumor driver biomarkers, tumor mutational burden, microsatellite instability, intratumor heterogeneity, cytolytic score, programmed cell death ligand 1 expression, patients' survival and response to anti-cytotoxic T-lymphocyte-associated antigen 4 and anti-programmed cell death protein 1 immunotherapy.

Studying tissue-independent components of cancer and defining pan-cancer subtypes could be addressed using tissue-specific molecular signatures if classification errors are controlled. Since PAM50 is a well-known, United States Food and Drug Administration (FDA)-approved and commercially available breast cancer signature, we applied it with uncertainty assessment to classify tumor samples from over 33 cancer types, discarded unassigned samples, and studied the emerging tumor-agnostic molecular patterns. The percentage of unassigned samples ranged between 55.5% and 86.9% in non-breast tissues, and gene set analysis suggested that the remaining samples could be grouped into two classes (named C1 and C2) regardless of the tissue. The C2 class was more dedifferentiated, more proliferative, with higher centrosome amplification, and potentially more TP53 and RB1 mutations. We identified 28 gene sets and 95 genes mainly associated with cell-cycle progression, cell-cycle checkpoints, and DNA damage that were consistently exacerbated in the C2 class. In some cancer types, the C1/C2 classification was associated with survival and drug sensitivity, and modulated the prognostic meaning of the immune infiltrate. Our results suggest that PAM50 could be repurposed for a pan-cancer context when paired with uncertainty assessment, resulting in two classes with molecular, biological, and clinical implications.

Abstract Although gene expression-derived PAM50 intrinsic subtypes (LumA, LumB, HER2E and Basal) were reported in Latin American breast cancer, most studies did not adequately represent the unique and diverse genetic admixture of the Latin American population and/or included a small number of individuals. As a result of these limitations, confirmation of the prognostic value of available intrinsic subtype classification signatures in a diverse cohort of Latin American women is of utmost importance. We assessed the general distribution and prognostic performance of PAM50-based intrinsic and immunohistochemistry (IHC)-based surrogate subtype classifications in Latin American women included in the Molecular Profile of Breast Cancer Study (MPBCS), an initiative of the US-Latin America Cancer Research Network (US-LACRN) comprising institutions of Argentina, Brazil, Chile, Mexico and Uruguay. MPBCS focused on stage II-III breast cancer in Latin American women. Eligible enrolled patients (n=1300) were characterized clinically, pathologically and epidemiologically and followed-up for 5 years. IHC subtypes were assessed according to St Gallen's 2013 criteria, using Ki67 to discriminate LumB from LumA tumors. A total of 1071 tumors were characterized by gene-expression microarrays. PAM50 classification defined 45% of tumors as LumA, 19.7% as LumB, 13.8% as HER2E and 17.5% as Basal. Normal-like tumors (6.3%) were excluded from the analysis. The 5-year prognostic ability of PAM50 and IHC classifications, both at the cancer-specific (OS) and progression-free survival (PFS), was tested. The prognosis for LumA tumors was significantly better than for other subtypes, while Basal-like tumors had the worst prognosis. The prognostic power of IHC-based subtypes (C-index 0.698 for OS, 0.635 for PFS) was very similar to that of PAM50 (C-index 0.678 for OS, 0.639 for PFS), indicating that in US-LACRN-MPBCS, contrary to other cohorts, surrogate subtypes are as useful as PAM50 for discriminating recurrence risk. PAM50-derived risks of recurrence (RORs), in particular ROR-S (C-index 0.699 for OS, 0.649 for PFS), clearly discriminated risk into low, intermediate and high-risk groups. Transcriptomic pathway analysis showed high proliferation (i.e. cell cycle control and DNA damage repair) associated with LumB, HER2E and Basal tumors, and a strong dependency on the estrogen pathway for LumA. Overall, a general concordance of the molecular features of US-LACRN-MPBCS breast cancer tumors with those of other cohorts was confirmed. The shift towards non-luminal subtypes could be partly attributable to the recruitment bias towards advanced stages. Further refinement of analyses using molecular ancestry assignation may help to reveal more subtle differences in this heterogeneously admixed population. Citation Format: Andrea S. Llera, Eliana Abdelhay, Osvaldo Podhajcer, Nora Artagaveytia, Adrián Daneri-Navarro, Bettina Müller, Carlos Velázquez Contreras, Darío Rocha, Juan Martín Sendoya, Renata Binato, Elmer Fernández, Elsa Alcoba, Isabel Alonso, Alicia I. Bravo, Natalia Camejo, Dirce Carraro, Mónica Castro, Juan M. Castro-Cervantes, Sandra Cataldi, Alfonso Cayota, Mauricio Cerda, Susanne Crocamo, Raul Delgadillo-Cisterna, Lucía Delgado, Alicia del Toro Arreola, Marisa Dreyer Breitenbach, Jorge Fernández, Wanda Fernández, Ramon A. Franco-Topete, Fancy Gaete, Jorge Gómez, Gonzalo Greif, Marisol Guerrero, Marianne Marianne Henderson, Andres de J Moran-Mendoza, María Aparecida Nagai, Antonio Oceguera-Villanueva, Antonio Quintero-Ramos, Rui Reis, Javier Retamales, Robinson Rodríguez, Cristina Rosales, Efrain Salas-González, Laura Segovia, Araceli Silva-García, Vidya Vedham, Livia Zagame, The US-Latin American Cancer Research Network. Molecular features of breast cancer involved in classification and prognosis of a multi-country Latin American cohort: The US-LACRN-MPBCS breast cancer cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 608.

Stat3 is constitutively activated in several tumor types and plays an essential role in maintaining their malignant phenotype and immunosupression. To take advantage of the promising antitumor activity of Stat3 targeting, it is vital to understand the mechanism by which Stat3 regulates both cell autonomous and non-autonomous processes. Here, we demonstrated that turning off Stat3 constitutive activation in different cancer cell types induces senescence, thus revealing their Stat3 addiction. Taking advantage of the senescence-associated secretory phenotype (SASP) induced by Stat3 silencing (SASP-siStat3), we designed an immunotherapy. The administration of SASP-siStat3 immunotherapy induced a strong inhibition of triple-negative breast cancer and melanoma growth associated with activation of CD4 + T and NK cells. Combining this immunotherapy with anti-PD-1 antibody resulted in survival improvement in mice bearing melanoma. The characterization of the SASP components revealed that type I IFN-related mediators, triggered by the activation of the cyclic GMP-AMP synthase DNA sensing pathway, are important for its immunosurveillance activity. Overall, our findings provided evidence that administration of SASP-siStat3 or low dose of Stat3-blocking agents would benefit patients with Stat3-addicted tumors to unleash an antitumor immune response and to improve the effectiveness of immune checkpoint inhibitors.

In molecular biology, modern technologies generate large databases of certain phenomena allowing to study the same group of individuals from different omics views. This provides data blocks that can be analyzed together through techniques, such as Multiple Factor Analysis (MFA), that allow a multidimensional exploratory approach. When two data blocks provide different information on the same phenomenon, the differences observed could arise from actual biological facts but, they could as well be a result of a confounding factor such as the technology used to obtain the data. In this study, we used MFA to analyze two data blocks (a transcriptomic block and a proteomic block) of gene expression on breast cancer patients, and added a third data block (also transcriptomic but gathered through a different technology) as a positive control. Both transcriptomic data blocks provided highly similar information between them, but different to the information provided by the proteomic block; hence the information provided by the different blocks represent different attributes of a biological phenomenon.

e15198 Background: Immunotherapies have revolutionized cancer treatment, but responses are not universal and patients who initially respond to therapy develop resistance. The accurate quantification of tumor-infiltrating immune cells holds the promise to reveal the role of the immune system in human cancers and its involvement in tumor escape mechanisms and response to therapy. We present MIXTURE, a new algorithm for tumor immune cell-type proportions deconvolution from transcriptomic data that overcomes competitive methods and revealed novel associations of immune cell types with patient survival and immunotherapy response. Methods: We applied MIXTURE to transcriptomic data from BRCA (n = 1095), LUAD (n = 506), SKCM (n = 472), HNSC (n = 499) and COAD (n = 521) cohorts from TCGA and five published datasets of melanoma patients treated with anti-PD-1/anti-CLTA-4. Results: The analysis of TCGA breast cancer biopsies showed high proportions of M2-macrophages associated with poor patient survival (p < 0.001). In contrast, we found that proportions of follicular T helper cells were associated with better outcome (p < 0.001). We observed a differential immune composition in biopsies of lung adenocarcinoma patients with mutations in TP53 (WT; n = 247; Mut; n = 259) and EGFR (WT, n = 438; Mut, n = 68) related with their different response to immunotherapy (p < 0.05). We found correlations between immune cells proportions and current biomarkers of response to immunotherapy such as TMB, intratumoral heterogeneity, MSI and PD-L1 expression in the TCGA cohorts (p < 0.05). The meta-analysis of melanoma patients treated with immunotherapy showed a distinct immune infiltrate in patients who responded to anti-PD-1 with an increase of immune effector cells such as CD8, CD4 memory activated and gamma-delta T cells, and a decrease in immunosuppressive M2-macrophages (p < 0.05; R = 81; NR = 107). According with the latest findings, we observed higher presence of B cells in responders to anti-PD-1 on-treatment (p = 0.033; R = 31; NR = 23) and in baseline of responders to anti-CTLA-4 (p = 0.028; R = 14; NR = 26). Interestingly, patients that previously progressed to anti-CTLA-4 showed a differential immune profile that was associated with response to anti-PD-1 (p < 0.05; Ipi-Prog = 59; Ipi-Naïve = 102). Conclusions: We demonstrated the potential of MIXTURE to understand the tumor immune microenvironment and its relationship with patient survival and response to immunotherapies. MIXTURE is available for the wider scientific community as web application and as packages for R and Python.

RNA sequencing has proved to be an efficient high-throughput technique to robustly characterize the presence and quantity of RNA in tumor biopsies at a given time. Importantly, it can be used to computationally estimate the composition of the tumor immune infiltrate and to infer the immunological phenotypes of those cells. Given the significant impact of anti-cancer immunotherapies and the role of the associated immune tumor microenvironment (ITME) on its prognosis and therapy response, the estimation of the immune cell-type content in the tumor is crucial for designing effective strategies to understand and treat cancer. Current digital estimation of the ITME cell mixture content can be performed using different analytical tools. However, current methods tend to over-estimate the number of cell-types present in the sample, thus under-estimating true proportions, biasing the results. We developed MIXTURE, a noise-constrained recursive feature selection for support vector regression that overcomes such limitations. MIXTURE deconvolutes cell-type proportions of bulk tumor samples for both RNA microarray or RNA-Seq platforms from a leukocyte validated gene signature. We evaluated MIXTURE over simulated and benchmark data sets. It overcomes competitive methods in terms of accuracy on the true number of present cell-types and proportions estimates with increased robustness to estimation bias. It also shows superior robustness to collinearity problems. Finally, we investigated the human immune microenvironment of breast cancer, head and neck squamous cell carcinoma, and melanoma biopsies before and after anti-PD-1 immunotherapy treatment revealing associations to response to therapy which have not seen by previous methods.