GAiN: An integrative tool utilizing generative adversarial neural networks for augmented gene expression analysis.

Big genomic data and artificial intelligence (AI) are ushering in an era of precision medicine, providing opportunities to study previously under-represented subtypes and rare diseases rather than categorize them as variances. However, clinical researchers face challenges in accessing such novel technologies as well as reliable methods to study small datasets or subcohorts with unique phenotypes. To address this need, we developed an integrative approach, GAiN, to capture patterns of gene expression from small datasets on the basis of an ensemble of generative adversarial networks (GANs) while leveraging big population data. Where conventional biostatistical methods fail, GAiN reliably discovers differentially expressed genes (DEGs) and enriched pathways between two cohorts with limited numbers of samples (n = 10) when benchmarked against a gold standard. GAiN is freely available at GitHub. Thus, GAiN may serve as a crucial tool for gene expression analysis in scenarios with limited samples, as in the context of rare diseases, under-represented populations, or limited investigator resources.

Sensitive MRD Detection from Lymphatic Fluid after Surgery in HPV-Associated Oropharyngeal Cancer.

PURPOSE: Our goal was to demonstrate that lymphatic drainage fluid (lymph) has improved sensitivity in quantifying postoperative minimal residual disease (MRD) in locally advanced human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (OPSCC) compared with plasma, and leverage this novel biofluid for patient risk stratification. EXPERIMENTAL DESIGN: We prospectively collected lymph samples from neck drains of 106 patients with HPV (+) OPSCC, along with 67 matched plasma samples, 24 hours after surgery. PCR and next-generation sequencing were used to quantify cancer-associated cell-free HPV (cf-HPV) and tumor-informed variants in lymph and plasma. Next, lymph cf-HPV and variants were compared with TNM stage, extranodal extension (ENE), and composite definitions of high-risk pathology. We then created a machine learning model, informed by lymph MRD and clinicopathologic features, to compare with progression-free survival (PFS). RESULTS: Postoperative lymph was enriched with cf-HPV compared with plasma (P < 0.0001) and correlated with pN2 stage (P = 0.003), ENE (P < 0.0001), and trial-defined pathologic risk criteria (mean AUC = 0.78). In addition, the lymph mutation number and variant allele frequency were higher in pN2 ENE (+) necks than in pN1 ENE (+) (P = 0.03, P = 0.02) or pN0-N1 ENE (-) (P = 0.04, P = 0.03, respectively). The lymph MRD-informed risk model demonstrated inferior PFS in high-risk patients (AUC = 0.96, P < 0.0001). CONCLUSIONS: Variant and cf-HPV quantification, performed in 24-hour postoperative lymph samples, reflects single- and multifeature high-risk pathologic criteria. Incorporating lymphatic MRD and clinicopathologic feature analysis can stratify PFS early after surgery in patients with HPV (+) head and neck cancer. See related commentary by Shannon and Iyer, p. 1223.

Genetic characterization of primary and metastatic high-grade serous ovarian cancer tumors reveals distinct features associated with survival.

High-grade serous ovarian cancer (HGSC) is the most lethal histotype of ovarian cancer and the majority of cases present with metastasis and late-stage disease. Over the last few decades, the overall survival for patients has not significantly improved, and there are limited targeted treatment options. We aimed to better characterize the distinctions between primary and metastatic tumors based on short- or long-term survival. We characterized 39 matched primary and metastatic tumors by whole exome and RNA sequencing. Of these, 23 were short-term (ST) survivors (overall survival (OS) < 3.5 years) and 16 were long-term (LT) survivors (OS > 5 years). We compared somatic mutations, copy number alterations, mutational burden, differential gene expression, immune cell infiltration, and gene fusion predictions between the primary and metastatic tumors and between ST and LT survivor cohorts. There were few differences in RNA expression between paired primary and metastatic tumors, but significant differences between the transcriptomes of LT and ST survivors in both their primary and metastatic tumors. These findings will improve the understanding of the genetic variation in HGSC that exist between patients with different prognoses and better inform treatments by identifying new targets for drug development.

SCCA1/SERPINB3 suppresses antitumor immunity and blunts therapy-induced T cell responses via STAT-dependent chemokine production.

Patients with cancer who have high serum levels of squamous cell carcinoma antigen 1 (SCCA1, now referred to as SERPINB3) commonly experience treatment resistance and have a poor prognosis. Despite being a clinical biomarker, the modulation of SERPINB3 in tumor immunity is poorly understood. We found positive correlations of SERPINB3 with CXCL1, CXCL8 (CXCL8/9), S100A8, and S100A9 (S100A8/A9) myeloid cell infiltration through RNA-Seq analysis of human primary cervical tumors. Induction of SERPINB3 resulted in increased CXCL1/8 and S100A8/A9 expression, which promoted monocyte and myeloid-derived suppressor cell (MDSC) migration in vitro. In mouse models, Serpinb3a tumors showed increased MDSC and tumor-associated macrophage (TAM) infiltration, contributing to T cell inhibition, and this was further augmented upon radiation. Intratumoral knockdown (KD) of Serpinb3a resulted in tumor growth inhibition and reduced CXCL1 and S100A8/A expression and MDSC and M2 macrophage infiltration. These changes led to enhanced cytotoxic T cell function and sensitized tumors to radiotherapy (RT). We further revealed that SERPINB3 promoted STAT-dependent expression of chemokines, whereby inhibition of STAT activation by ruxolitinib or siRNA abrogated CXCL1/8 and S100A8/ A9 expression in SERPINB3 cells. Patients with elevated pretreatment SCCA levels and high phosphorylated STAT3 (p-STAT3) had increased intratumoral CD11b+ myeloid cells compared with patients with low SCCA levels and p-STAT3, who had improved overall survival after RT. These findings provide a preclinical rationale for targeting SERPINB3 in tumors to counteract immunosuppression and improve the response to RT.

Intrinsic tumor resistance to CAR T cells is a dynamic transcriptional state that is exploitable with low-dose radiation.

Chimeric antigen receptor (CAR) T-cell therapy represents a major advancement for hematologic malignancies, with some patients achieving long-term remission. However, the majority of treated patients still die of their disease. A consistent predictor of response is tumor quantity, wherein a higher disease burden before CAR T-cell therapy portends a worse prognosis. Focal radiation to bulky sites of the disease can decrease tumor quantity before CAR T-cell therapy, but whether this strategy improves survival is unknown. We find that substantially reducing systemic tumor quantity using high-dose radiation to areas of bulky disease, which is commonly done clinically, is less impactful on overall survival in mice achieved by CAR T cells than targeting all sites of disease with low-dose total tumor irradiation (TTI) before CAR T-cell therapy. This finding highlights another predictor of response, tumor quality, the intrinsic resistance of an individual patient's tumor cells to CAR T-cell killing. Little is known about whether or how an individual tumor's intrinsic resistance may change under different circumstances. We find a transcriptional "death receptor score" that reflects a tumor's intrinsic sensitivity to CAR T cells can be temporarily increased by low-dose TTI, and the timing of this transcriptional change correlates with improved in vivo leukemia control by an otherwise limited number of CAR T cells. This suggests an actionable method for potentially improving outcomes in patients predicted to respond poorly to this promising therapy and highlights that intrinsic tumor attributes may be equally or more important predictors of CAR T-cell response as tumor burden.

p38MAPKα Stromal Reprogramming Sensitizes Metastatic Breast Cancer to Immunotherapy.

Metastatic breast cancer is an intractable disease that responds poorly to immunotherapy. We show that p38MAPKα inhibition (p38i) limits tumor growth by reprogramming the metastatic tumor microenvironment in a CD4+ T cell-, IFNγ-, and macrophage-dependent manner. To identify targets that further increased p38i efficacy, we utilized a stromal labeling approach and single-cell RNA sequencing. Thus, we combined p38i and an OX40 agonist that synergistically reduced metastatic growth and increased overall survival. Intriguingly, patients with a p38i metastatic stromal signature had better overall survival that was further improved by the presence of an increased mutational load, leading us to ask if our approach would be effective in antigenic breast cancer. The combination of p38i, anti-OX40, and cytotoxic T-cell engagement cured mice of metastatic disease and produced long-term immunologic memory. Our findings demonstrate that a detailed understanding of the stromal compartment can be used to design effective antimetastatic therapies. SIGNIFICANCE: Immunotherapy is rarely effective in breast cancer. We dissected the metastatic tumor stroma, which revealed a novel therapeutic approach that targets the stromal p38MAPK pathway and creates an opportunity to unleash an immunologic response. Our work underscores the importance of understanding the tumor stromal compartment in therapeutic design. This article is highlighted in the In This Issue feature, p. 1275.

Radiation-induced circulating myeloid-derived suppressor cells induce systemic lymphopenia after chemoradiotherapy in patients with glioblastoma.

Severe and prolonged lymphopenia frequently occurs in patients with glioblastoma after standard chemoradiotherapy and has been associated with worse survival, but its underlying biological mechanism is not well understood. To address this, we performed a correlative study in which we collected and analyzed peripheral blood of patients with glioblastoma (n = 20) receiving chemoradiotherapy using genomic and immune monitoring technologies. RNA sequencing analysis of the peripheral blood mononuclear cells (PBMC) showed an elevated concentration of myeloid-derived suppressor cell (MDSC) regulatory genes in patients with lymphopenia when compared with patients without lymphopenia after chemoradiotherapy. Additional analysis including flow cytometry and single-cell RNA sequencing further confirmed increased numbers of circulating MDSC in patients with lymphopenia when compared with patients without lymphopenia after chemoradiotherapy. Preclinical murine models were also established and demonstrated a causal relationship between radiation-induced MDSC and systemic lymphopenia using transfusion and depletion experiments. Pharmacological inhibition of MDSC using an arginase-1 inhibitor (CB1158) or phosphodiesterase-5 inhibitor (tadalafil) during radiation therapy (RT) successfully abrogated radiation-induced lymphopenia and improved survival in the preclinical models. CB1158 and tadalafil are promising drugs in reducing radiation-induced lymphopenia in patients with glioblastoma. These results demonstrate the promise of using these classes of drugs to reduce treatment-related lymphopenia and immunosuppression.

Pan-Cancer Analysis Reveals Recurrent BCAR4 Gene Fusions across Solid Tumors.

Chromosomal rearrangements often result in active regulatory regions juxtaposed upstream of an oncogene to generate an expressed gene fusion. Repeated activation of a common downstream partner-with differing upstream regions across a patient cohort-suggests a conserved oncogenic role. Analysis of 9,638 patients across 32 solid tumor types revealed an annotated long noncoding RNA (lncRNA), Breast Cancer Anti-Estrogen Resistance 4 (BCAR4), was the most prevalent, uncharacterized, downstream gene fusion partner occurring in 11 cancers. Its oncogenic role was confirmed using multiple cell lines with endogenous BCAR4 gene fusions. Furthermore, overexpressing clinically prevalent BCAR4 gene fusions in untransformed cell lines was sufficient to induce an oncogenic phenotype. We show that the minimum common region to all gene fusions harbors an open reading frame that is necessary to drive proliferation. IMPLICATIONS: BCAR4 gene fusions represent an underappreciated class of gene fusions that may have biological and clinical implications across solid tumors.

Loss of H3K27 Trimethylation Promotes Radiotherapy Resistance in Medulloblastoma and Induces an Actionable Vulnerability to BET Inhibition.

Medulloblastoma has been categorized into four subgroups based on genetic, epigenetic, and transcriptional profiling. Radiation is used for treating medulloblastoma regardless of the subgroup. A better understanding of the molecular pathways determining radiotherapy response could help improve medulloblastoma treatment. Here, we investigated the role of the EZH2 (enhancer of zeste 2 polycomb repressive complex 2 subunit)-dependent histone H3K27 trimethylation in radiotherapy response in medulloblastoma. The tumors in 47.2% of patients with group 3 and 4 medulloblastoma displayed H3K27me3 deficiency. Loss of H3K27me3 was associated with a radioresistant phenotype, high relapse rates, and poor overall survival. In H3K27me3-deficient medulloblastoma cells, an epigenetic switch from H3K27me3 to H3K27ac occurred at specific genomic loci, altering the transcriptional profile. The resulting upregulation of EPHA2 stimulated excessive activation of the prosurvival AKT signaling pathway, leading to radiotherapy resistance. Bromodomain and extraterminal motif (BET) inhibition overcame radiation resistance in H3K27me3-deficient medulloblastoma cells by suppressing H3K27ac levels, blunting EPHA2 overexpression, and mitigating excessive AKT signaling. In addition, BET inhibition sensitized medulloblastoma cells to radiation by enhancing the apoptotic response through suppression of Bcl-xL and upregulation of Bim. This work demonstrates a novel mechanism of radiation resistance in medulloblastoma and identifies an epigenetic marker predictive of radiotherapy response. On the basis of these findings, we propose an epigenetically guided treatment approach targeting radiotherapy resistance in patients with medulloblastoma. SIGNIFICANCE: This study demonstrates a novel epigenetic mechanism of radiation resistance in medulloblastoma and identifies a therapeutic approach to improve outcomes in these patients.

DANSR: A Tool for the Detection of Annotated and Novel Small RNAs.

Existing small noncoding RNA analysis tools are optimized for processing short sequencing reads (17-35 nucleotides) to monitor microRNA expression. However, these strategies under-represent many biologically relevant classes of small noncoding RNAs in the 36-200 nucleotides length range (tRNAs, snoRNAs, etc.). To address this, we developed DANSR, a tool for the detection of annotated and novel small RNAs using sequencing reads with variable lengths (ranging from 17-200 nt). While DANSR is broadly applicable to any small RNA dataset, we applied it to a cohort of matched normal, primary, and distant metastatic colorectal cancer specimens to demonstrate its ability to quantify annotated small RNAs, discover novel genes, and calculate differential expression. DANSR is available as an open source tool.

HPV transcript expression affects cervical cancer response to chemoradiation.

Persistent HPV infection is causative for the majority of cervical cancer cases; however, current guidelines do not require HPV testing for newly diagnosed cervical cancer. Using an institutional cohort of 88 patients with cervical cancer treated uniformly with standard-of-care chemoradiation treatment (CRT) with prospectively collected clinical outcome data, we observed that patients with cervical tumors containing HPV genotypes other than HPV 16 have worse survival outcomes after CRT compared with patients with HPV 16+ tumors, consistent with previously published studies. Using RNA sequencing analysis, we quantified viral transcription efficiency and found higher levels of E6 and the alternative transcript E6*I in cervical tumors with HPV genotypes other than HPV 16. These findings were validated using whole transcriptome data from The Cancer Genome Atlas (n = 304). For the first time to our knowledge, transcript expression level of HPV E6*I was identified as a predictive biomarker of CRT outcome in our complete institutional data set (n = 88) and within the HPV 16+ subset (n = 36). In vitro characterization of HPV E6*I and E6 overexpression revealed that both induce CRT resistance through distinct mechanisms dependent upon p53-p21. Our findings suggest that high expression of E6*I and E6 may represent novel biomarkers of CRT efficacy, and these patients may benefit from alternative treatment strategies.

A community challenge to evaluate RNA-seq, fusion detection, and isoform quantification methods for cancer discovery.

The accurate identification and quantitation of RNA isoforms present in the cancer transcriptome is key for analyses ranging from the inference of the impacts of somatic variants to pathway analysis to biomarker development and subtype discovery. The ICGC-TCGA DREAM Somatic Mutation Calling in RNA (SMC-RNA) challenge was a crowd-sourced effort to benchmark methods for RNA isoform quantification and fusion detection from bulk cancer RNA sequencing (RNA-seq) data. It concluded in 2018 with a comparison of 77 fusion detection entries and 65 isoform quantification entries on 51 synthetic tumors and 32 cell lines with spiked-in fusion constructs. We report the entries used to build this benchmark, the leaderboard results, and the experimental features associated with the accurate prediction of RNA species. This challenge required submissions to be in the form of containerized workflows, meaning each of the entries described is easily reusable through CWL and Docker containers at https://github.com/SMC-RNA-challenge. A record of this paper's transparent peer review process is included in the supplemental information.

Standardized Uptake Value for 18F-Fluorodeoxyglucose Is a Marker of Inflammatory State and Immune Infiltrate in Cervical Cancer.

PURPOSE: Chemoradiotherapy for locally advanced cervical cancer fails in over a third of patients. Biomarkers with therapeutic implications are therefore needed. We investigated the relationship between an established prognostic marker, maximum standardized uptake value (SUVmax) on 18F-fluorodeoxyglucose positron emission tomography, and the inflammatory and immune state of cervical cancers. EXPERIMENTAL DESIGN: An SUVmax most prognostic for freedom from progression (FFP) was identified and compared with known prognostic clinical variables in a cohort of 318 patients treated with definitive radiation with prospectively collected clinical data. Gene set enrichment analysis (GSEA) and CIBERSORT of whole-transcriptome data from 68 patients were used to identify biological pathways and immune cell subpopulations associated with high SUVmax. IHC using a tissue microarray (TMA, N = 82) was used to validate the CIBERSORT findings. The impact of macrophages on cervical cancer glucose metabolism was investigated in coculture experiments. RESULTS: SUVmax <11.4 was most prognostic for FFP (P = 0.001). The GSEA showed that high SUVmax is associated with increased gene expression of inflammatory pathways, including JAK/STAT3 signaling. CIBERSORT and CD68 staining of the TMA showed high SUVmax tumors are characterized by a monocyte-predominant immune infiltrate. Coculture of cervical cancer cells with macrophages or macrophage-conditioned media altered glucose uptake, and IL6 and JAK/STAT3 signaling contribute to this effect. CONCLUSIONS: SUVmax is a prognostic marker in cervical cancer that is associated with activation of inflammatory pathways and tumor infiltration of myeloid-derived immune cells, particularly macrophages. Macrophages contribute to changes in cervical cancer glucose metabolism.See related commentary by Williamson et al., p. 4136.

Integrating imaging and RNA-seq improves outcome prediction in cervical cancer.

Approaches using a single type of data have been applied to classify human tumors. Here we integrate imaging features and transcriptomic data using a prospectively collected tumor bank. We demonstrate that increased maximum standardized uptake value on pretreatment 18F-fluorodeoxyglucose-positron emission tomography correlates with epithelial-to-mesenchymal transition (EMT) gene expression. We derived and validated 3 major molecular groups, namely squamous epithelial, squamous mesenchymal, and adenocarcinoma, using prospectively collected institutional (n = 67) and publicly available (n = 304) data sets. Patients with tumors of the squamous mesenchymal subtype showed inferior survival outcomes compared with the other 2 molecular groups. High mesenchymal gene expression in cervical cancer cells positively correlated with the capacity to form spheroids and with resistance to radiation. CaSki organoids were radiation-resistant but sensitive to the glycolysis inhibitor, 2-DG. These experiments provide a strategy for response prediction by integrating large data sets, and highlight the potential for metabolic therapy to influence EMT phenotypes in cervical cancer.

HPV-EM: an accurate HPV detection and genotyping EM algorithm.

Accurate HPV genotyping is crucial in facilitating epidemiology studies, vaccine trials, and HPV-related cancer research. Contemporary HPV genotyping assays only detect < 25% of all known HPV genotypes and are not accurate for low-risk or mixed HPV genotypes. Current genomic HPV genotyping algorithms use a simple read-alignment and filtering strategy that has difficulty handling repeats and homology sequences. Therefore, we have developed an optimized expectation-maximization algorithm, designated HPV-EM, to address the ambiguities caused by repetitive sequencing reads. HPV-EM achieved 97-100% accuracy when benchmarked using cell line data and TCGA cervical cancer data. We also validated HPV-EM using DNA tiling data on an institutional cervical cancer cohort (96.5% accuracy). Using HPV-EM, we demonstrated HPV genotypic differences in recurrence and patient outcomes in cervical and head and neck cancers.