University Cancer Center Hamburg

Many different theories have been advanced concerning the biological roles of the oligosaccharide units of individual classes of glycoconjugates. Analysis of the evidence indicates that while all of these theories are correct, exceptions to each can also be found. The biological roles of oligosaccharides appear to span the spectrum from those that are trivial, to those that are crucial for the development, growth, function or survival of an organism. Some general principles emerge. First, it is difficult to predict a priori the functions a given oligosaccharide on a given glycoconjugate might be mediating, or their relative importance to the organism. Second, the same oligosaccharide sequence may mediate different functions at different locations within the same organism, or at different times in its ontogeny or life cycle. Third, the more specific and crucial biological roles of oligosaccharides are often mediated by unusual oligosaccharide sequences, unusual presentations of common terminal sequences, or by further modifications of the sugars themselves. However, such oligosaccharide sequences are also more likely to be targets for recognition by pathogenic toxins and microorganisms. As such, they are subject to more intra- and inter-species variation because of ongoing host-pathogen interactions during evolution. In the final analysis, the only common features of the varied functions of oligosaccharides are that they either mediate 'specific recognition' events or that they provide 'modulation' of biological processes. In so doing, they generate much of the functional diversity required for the development and differentiation of complex organisms, and for their interactions with other organisms in the environment.

Lists of authors and their affiliations appear in the online version of the paper Breast cancer risk is influenced by rare coding variants in susceptibility genes, such as BRCA1, and many common, mostly non-coding variants. However, much of the genetic contribution to breast cancer risk remains unknown. Here we report the results of a genome-wide association study of breast cancer in 122,977 cases and 105,974 controls of European ancestry and 14,068 cases and 13,104 controls of East Asian ancestry 1 . We identified 65 new loci that are associated with overall breast cancer risk at P < 5 10 -8 . The majority of credible risk single-nucleotide polymorphisms in these loci fall in distal regulatory elements, and by integrating in silico data to predict target genes in breast cells at each locus, we demonstrate a strong overlap between candidate target genes and somatic driver genes in breast tumours. We also find that heritability of breast cancer due to all single-nucleotide polymorphisms in regulatory features was 2-5-fold enriched relative to the genomewide average, with strong enrichment for particular transcription factor binding sites. These results provide further insight into genetic susceptibility to breast cancer and will improve the use of genetic risk scores for individualized screening and prevention.

Microsatellite instability determines whether patients with gastrointestinal cancer respond exceptionally well to immunotherapy. However, in clinical practice, not every patient is tested for MSI, because this requires additional genetic or immunohistochemical tests. Here we show that deep residual learning can predict MSI directly from H&E histology, which is ubiquitously available. This approach has the potential to provide immunotherapy to a much broader subset of patients with gastrointestinal cancer. A deep residual learning framework identifies microsatellite instability in histology slides from patients with cancer and can be used to guide immunotherapy.

UNLABELLED: "Liquid biopsy" focusing on the analysis of circulating tumor cells (CTC) and circulating cell-free tumor DNA (ctDNA) in the blood of patients with cancer has received enormous attention because of its obvious clinical implications for personalized medicine. Analyses of CTCs and ctDNA have paved new diagnostic avenues and are, to date, the cornerstones of liquid biopsy diagnostics. The present review focuses on key areas of clinical applications of CTCs and ctDNA, including detection of cancer, prediction of prognosis in patients with curable disease, monitoring systemic therapies, and stratification of patients based on the detection of therapeutic targets or resistance mechanisms. SIGNIFICANCE: The application of CTCs and ctDNA for the early detection of cancer is of high public interest, but it faces serious challenges regarding specificity and sensitivity of the current assays. Prediction of prognosis in patients with curable disease can already be achieved in several tumor entities, particularly in breast cancer. Monitoring the success or failure of systemic therapies (i.e., chemotherapy, hormonal therapy, or other targeted therapies) by sequential measurements of CTCs or ctDNA is also feasible. Interventional studies on treatment stratification based on the analysis of CTCs and ctDNA are needed to implement liquid biopsy into personalized medicine. Cancer Discov; 6(5); 479-91. ©2016 AACR.

PURPOSE Patients with advanced esophageal cancer have a poor prognosis and limited treatment options after first-line chemotherapy. PATIENTS AND METHODS In this open-label, phase III study, we randomly assigned (1:1) 628 patients with advanced/metastatic squamous cell carcinoma or adenocarcinoma of the esophagus, that progressed after one prior therapy, to pembrolizumab 200 mg every 3 weeks for up to 2 years or chemotherapy (investigator’s choice of paclitaxel, docetaxel, or irinotecan). Primary end points were overall survival (OS) in patients with programmed death ligand-1 (PD-L1) combined positive score (CPS) ≥ 10, in patients with squamous cell carcinoma, and in all patients (one-sided α 0.9%, 0.8%, and 0.8%, respectively). RESULTS At final analysis, conducted 16 months after the last patient was randomly assigned, OS was prolonged with pembrolizumab versus chemotherapy for patients with CPS ≥ 10 (median, 9.3 v 6.7 months; hazard ratio [HR], 0.69 [95% CI, 0.52 to 0.93]; P = .0074). Estimated 12-month OS rate was 43% (95% CI, 33.5% to 52.1%) with pembrolizumab versus 20% (95% CI, 13.5% to 28.3%) with chemotherapy. Median OS was 8.2 months versus 7.1 months (HR, 0.78 [95% CI, 0.63 to 0.96]; P = .0095) in patients with squamous cell carcinoma and 7.1 months versus 7.1 months (HR, 0.89 [95% CI, 0.75 to 1.05]; P = .0560) in all patients. Grade 3-5 treatment-related adverse events occurred in 18.2% of patients with pembrolizumab versus 40.9% in those who underwent chemotherapy. CONCLUSION Pembrolizumab prolonged OS versus chemotherapy as second-line therapy for advanced esophageal cancer in patients with PD-L1 CPS ≥ 10, with fewer treatment-related adverse events.

Importance: Safe and effective therapies for untreated, advanced gastric/gastroesophageal junction (G/GEJ) cancer remain an unmet need. Objective: To evaluate the antitumor activity of pembrolizumab, pembrolizumab plus chemotherapy, or chemotherapy alone in patients with untreated, advanced G/GEJ cancer with programmed cell death ligand 1 (PD-L1) combined positive score (CPS) of 1 or greater. Design, Setting, and Participants: The phase 3 KEYNOTE-062 randomized, controlled, partially blinded interventional trial enrolled 763 patients with untreated, locally advanced/unresectable or metastatic G/GEJ cancer with PD-L1 CPS of 1 or greater from 200 centers in 29 countries between September 18, 2015, and May 26, 2017. Interventions: Patients were randomized 1:1:1 to pembrolizumab 200 mg, pembrolizumab plus chemotherapy (cisplatin 80 mg/m2/d on day 1 plus fluorouracil 800 mg/m2/d on days 1 to 5 or capecitabine 1000 mg/m2 twice daily), or chemotherapy plus placebo, every 3 weeks. Main Outcomes and Measures: Primary end points were overall survival (OS) and progression-free survival (PFS) in patients with PD-L1 CPS of 1 or greater or 10 or greater. Results: A total of 763 patients were randomized to pembrolizumab (n = 256), pembrolizumab plus chemotherapy (n = 257), or chemotherapy (n = 250). The median (range) age of all patients in the study cohort was 62 (20-87) years; 554 of 763 (72.6%) were men. At final analysis, after a median (range) follow-up of 29.4 (22.0-41.3) months, pembrolizumab was noninferior to chemotherapy for OS in patients with CPS of 1 or greater (median, 10.6 vs 11.1 months; hazard ratio [HR], 0.91; 99.2% CI, 0.69-1.18). Pembrolizumab monotherapy was not superior to chemotherapy in patients with CPS of 1 or greater. Pembrolizumab prolonged OS vs chemotherapy in patients with CPS of 10 or greater (median, 17.4 vs 10.8 months; HR, 0.69; 95% CI, 0.49-0.97), but this difference was not statistically tested. Pembrolizumab plus chemotherapy was not superior to chemotherapy for OS in patients with CPS of 1 or greater (12.5 vs 11.1 months; HR, 0.85; 95% CI, 0.70-1.03; P = .05) or CPS of 10 or greater (12.3 vs 10.8 months; HR, 0.85; 95% CI, 0.62-1.17; P = .16) or for PFS in patients with CPS of 1 or greater (6.9 vs 6.4 months; HR, 0.84; 95% CI, 0.70-1.02; P = .04). Grade 3 to 5 treatment-related adverse event rates for pembrolizumab, pembrolizumab plus chemotherapy, and chemotherapy were 17%, 73%, and 69%, respectively. Conclusions and Relevance: This phase 3 randomized clinical trial found that among patients with untreated, advanced G/GEJ cancer, pembrolizumab was noninferior to chemotherapy, with fewer adverse events observed. Pembrolizumab or pembrolizumab plus chemotherapy was not superior to chemotherapy for the OS and PFS end points tested. Trial Registration: ClinicalTrials.gov Identifier: NCT02494583.

BACKGROUND: The detection and molecular characterization of circulating tumor cells (CTCs) are one of the most active areas of translational cancer research, with >400 clinical studies having included CTCs as a biomarker. The aims of research on CTCs include (a) estimation of the risk for metastatic relapse or metastatic progression (prognostic information), (b) stratification and real-time monitoring of therapies, (c) identification of therapeutic targets and resistance mechanisms, and (d) understanding metastasis development in cancer patients. CONTENT: This review focuses on the technologies used for the enrichment and detection of CTCs. We outline and discuss the current technologies that are based on exploiting the physical and biological properties of CTCs. A number of innovative technologies to improve methods for CTC detection have recently been developed, including CTC microchips, filtration devices, quantitative reverse-transcription PCR assays, and automated microscopy systems. Molecular-characterization studies have indicated, however, that CTCs are very heterogeneous, a finding that underscores the need for multiplex approaches to capture all of the relevant CTC subsets. We therefore emphasize the current challenges of increasing the yield and detection of CTCs that have undergone an epithelial-mesenchymal transition. Increasing assay analytical sensitivity may lead, however, to a decrease in analytical specificity (e.g., through the detection of circulating normal epithelial cells). SUMMARY: A considerable number of promising CTC-detection techniques have been developed in recent years. The analytical specificity and clinical utility of these methods must be demonstrated in large prospective multicenter studies to reach the high level of evidence required for their introduction into clinical practice.

Microglia are yolk sac-derived macrophages residing in the parenchyma of brain and spinal cord, where they interact with neurons and other glial. After different conditioning paradigms and bone marrow (BM) or hematopoietic stem cell (HSC) transplantation, graft-derived cells seed the brain and persistently contribute to the parenchymal brain macrophage compartment. Here we establish that graft-derived macrophages acquire, over time, microglia characteristics, including ramified morphology, longevity, radio-resistance and clonal expansion. However, even after prolonged CNS residence, transcriptomes and chromatin accessibility landscapes of engrafted, BM-derived macrophages remain distinct from yolk sac-derived host microglia. Furthermore, engrafted BM-derived cells display discrete responses to peripheral endotoxin challenge, as compared to host microglia. In human HSC transplant recipients, engrafted cells also remain distinct from host microglia, extending our finding to clinical settings. Collectively, our data emphasize the molecular and functional heterogeneity of parenchymal brain macrophages and highlight potential clinical implications for HSC gene therapies aimed to ameliorate lysosomal storage disorders, microgliopathies or general monogenic immuno-deficiencies.

Effective treatment options are limited for patients with acute myeloid leukemia (AML) who cannot tolerate intensive chemotherapy. Adults age ≥18 years with newly diagnosed AML ineligible for intensive chemotherapy were enrolled in this international phase 3 randomized double-blind placebo-controlled trial. Patients (N = 211) were randomized 2:1 to venetoclax (n = 143) or placebo (n = 68) in 28-day cycles, plus low-dose cytarabine (LDAC) on days 1 to 10. Primary end point was overall survival (OS); secondary end points included response rate, transfusion independence, and event-free survival. Median age was 76 years (range, 36-93 years), 38% had secondary AML, and 20% had received prior hypomethylating agent treatment. Planned primary analysis showed a 25% reduction in risk of death with venetoclax plus LDAC vs LDAC alone (hazard ratio [HR], 0.75; 95% confidence interval [CI], 0.52-1.07; P = .11), although not statistically significant; median OS was 7.2 vs 4.1 months, respectively. Unplanned analysis with additional 6-month follow-up demonstrated median OS of 8.4 months for the venetoclax arm (HR, 0.70; 95% CI, 0.50-0.98; P = .04). Complete remission (CR) plus CR with incomplete blood count recovery rates were 48% and 13% for venetoclax plus LDAC and LDAC alone, respectively. Key grade ≥3 adverse events (venetoclax vs LDAC alone) were febrile neutropenia (32% vs 29%), neutropenia (47% vs 16%), and thrombocytopenia (45% vs 37%). Venetoclax plus LDAC demonstrates clinically meaningful improvement in remission rate and OS vs LDAC alone, with a manageable safety profile. Results confirm venetoclax plus LDAC as an important frontline treatment for AML patients unfit for intensive chemotherapy. This trial was registered at www.clinicaltrials.gov as #NCT03069352.

Abstract Physical activity has been associated with lower risks of breast and colorectal cancer in epidemiological studies; however, it is unknown if these associations are causal or confounded. In two-sample Mendelian randomisation analyses, using summary genetic data from the UK Biobank and GWA consortia, we found that a one standard deviation increment in average acceleration was associated with lower risks of breast cancer (odds ratio [OR]: 0.51, 95% confidence interval [CI]: 0.27 to 0.98, P-value = 0.04) and colorectal cancer (OR: 0.66, 95% CI: 0.48 to 0.90, P-value = 0.01). We found similar magnitude inverse associations for estrogen positive (ER +ve ) breast cancer and for colon cancer. Our results support a potentially causal relationship between higher physical activity levels and lower risks of breast cancer and colorectal cancer. Based on these data, the promotion of physical activity is probably an effective strategy in the primary prevention of these commonly diagnosed cancers.

Common variation on 14q24.1, close to RAD51B, has been associated with breast cancer: rs999737 and rs2588809 with the risk of female breast cancer and rs1314913 with the risk of male breast cancer. The aim of this study was to investigate the role of RAD51B variants in breast cancer predisposition, particularly in the context of familial breast cancer in Finland. We sequenced the coding region of RAD51B in 168 Finnish breast cancer patients from the Helsinki region for identification of possible recurrent founder mutations. In addition, we studied the known rs999737, rs2588809, and rs1314913 SNPs and RAD51B haplotypes in 44,791 breast cancer cases and 43,583 controls from 40 studies participating in the Breast Cancer Association Consortium (BCAC) that were genotyped on a custom chip (iCOGS). We identified one putatively pathogenic missense mutation c.541C>T among the Finnish cancer patients and subsequently genotyped the mutation in additional breast cancer cases (n = 5259) and population controls (n = 3586) from Finland and Belarus. No significant association with breast cancer risk was seen in the meta-analysis of the Finnish datasets or in the large BCAC dataset. The association with previously identified risk variants rs999737, rs2588809, and rs1314913 was replicated among all breast cancer cases and also among familial cases in the BCAC dataset. The most significant association was observed for the haplotype carrying the risk-alleles of all the three SNPs both among all cases (odds ratio (OR): 1.15, 95% confidence interval (CI): 1.11-1.19, P = 8.88 x 10-16) and among familial cases (OR: 1.24, 95% CI: 1.16-1.32, P = 6.19 x 10-11), compared to the haplotype with the respective protective alleles. Our results suggest that loss-of-function mutations in RAD51B are rare, but common variation at the RAD51B region is significantly associated with familial breast cancer risk.

Glioblastomas are malignant tumors of the central nervous system hallmarked by subclonal diversity and dynamic adaptation amid developmental hierarchies. The source of dynamic reorganization within the spatial context of these tumors remains elusive. Here, we characterized glioblastomas by spatially resolved transcriptomics, metabolomics, and proteomics. By deciphering regionally shared transcriptional programs across patients, we infer that glioblastoma is organized by spatial segregation of lineage states and adapts to inflammatory and/or metabolic stimuli, reminiscent of the reactive transformation in mature astrocytes. Integration of metabolic imaging and imaging mass cytometry uncovered locoregional tumor-host interdependence, resulting in spatially exclusive adaptive transcriptional programs. Inferring copy-number alterations emphasizes a spatially cohesive organization of subclones associated with reactive transcriptional programs, confirming that environmental stress gives rise to selection pressure. A model of glioblastoma stem cells implanted into human and rodent neocortical tissue mimicking various environments confirmed that transcriptional states originate from dynamic adaptation to various environments.

Four decades ago, angiogenesis was recognized as a therapeutic target for blocking cancer growth. Because of its importance, VEGF has been at the center stage of antiangiogenic therapy. Now, several years after FDA approval of an anti-VEGF antibody as the first antiangiogenic agent, many patients with cancer and ocular neovascularization have benefited from VEGF-targeted therapy; however, this anticancer strategy is challenged by insufficient efficacy, intrinsic refractoriness, and resistance. Here, we examine recent discoveries of new mechanisms underlying angiogenesis, discuss successes and challenges of current antiangiogenic therapy, and highlight emerging antiangiogenic paradigms.

IMPORTANCE: Effective treatment options are limited for patients with advanced, metastatic esophageal cancer progressing after 2 or more lines of systemic therapy. OBJECTIVE: To evaluate the efficacy and safety of pembrolizumab for patients with advanced, metastatic esophageal squamous cell carcinoma (ESCC) or advanced, metastatic adenocarcinoma of the esophagus and gastroesophageal junction that progressed after 2 or more lines of systemic therapy. DESIGN, SETTING, AND PARTICIPANTS: This phase 2, open-label, interventional, single-arm study, KEYNOTE-180, enrolled 121 patients from January 12, 2016, to March 21, 2017, from 57 sites in 10 countries. Patients had advanced, metastatic esophageal cancer that progressed after 2 or more lines of therapy and had evaluable tumor samples for biomarkers. INTERVENTIONS: Pembrolizumab, 200 mg, was administered intravenously every 3 weeks until disease progression, unacceptable toxic effects, or study withdrawal, for up to 2 years. MAIN OUTCOMES AND MEASURES: Primary end point was objective response rate per the Response Evaluation Criteria in Solid Tumors by central imaging review for all patients. RESULTS: As of September 18, 2017, of 121 enrolled patients (100 men and 21 women; median age, 65 years [range, 33-87 years]), 18 (14.9%) had undergone 3 or more prior therapies, 63 (52.1%) had ESCC, and 58 (47.9%) had tumors positive for programmed death ligand-1 (PD-L1), defined as a combined positive score of 10 or higher assessed by immunohistochemistry. Median duration of follow-up was 5.8 months (range, 0.2-18.3 months). Objective response rate was 9.9% (95% CI, 5.2%-16.7%) among all patients (12 of 121), and median duration of response was not reached (range, 1.9-14.4 months). Objective response rate was 14.3% (95% CI, 6.7%-25.4%) among patients with ESCC (9 of 63), 5.2% (95% CI, 1.1%-14.4%) among patients with adenocarcinoma (3 of 58), 13.8% (95% CI, 6.1%-25.4%) among patients with PD-L1-positive tumors (8 of 58), and 6.3% (95% CI, 1.8%-15.5%) among patients with PD-L1-negative tumors (4 of 63). Overall, 15 patients (12.4%) had treatment-related grade 3 to 5 adverse events. Only 5 patients (4.1%) discontinued treatment because of adverse events. There was 1 treatment-related death from pneumonitis. CONCLUSIONS AND RELEVANCE: Where effective treatment options are an unmet need, pembrolizumab provided durable antitumor activity with manageable safety in patients with heavily pretreated esophageal cancer. Phase 3 studies evaluating pembrolizumab vs standard therapy for patients with esophageal cancer progressing after first-line therapy or in combination with chemotherapy as first-line therapy for patients with locally advanced unresectable or metastatic esophageal cancer are ongoing. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT02559687.

The goal of this review is to systematically address a number of issues raised in the American Society of Clinical Oncology-College of American Pathologists (ASCO-CAP) guidelines on testing for the human epidermal growth factor receptor 2 (HER-2) alteration. A group of investigators who are experienced in the conduct and interpretation of HER-2 assay methods reviewed the ASCO-CAP guidelines and address several areas of the HER-2 testing guidelines with a particular emphasis on biologic and methodologic considerations. Although HER-2 status determined by immunohistochemistry (IHC) and the status determined by fluorescent in situ hybridization (FISH) are significantly correlated, we feel that standard considerations of laboratory testing, including test accuracy, reproducibility, and precision, as well as the current data favor FISH over IHC assay methods for determining HER-2 status. These considerations are clearly important in clinical practice because HER2 amplification is directly linked to protein expression levels in breast cancer. However, this protein is not consistently analyzed in formalin-fixed tissues as a result of variability in fixation methods and times and the impact of fixation on HER-2 protein antigenicity. Conversely, gene amplification and FISH are significantly less dependent on tissue fixation methods, making this assay more reproducible between central and peripheral laboratories than IHC. Moreover, review of the existing data demonstrate that FISH is more strongly correlated with responsiveness to either trastuzumab or lapatinib treatment. Until other methods achieve similar test accuracy, reproducibility, and predictive value, we suggest FISH as the primary HER-2 testing modality for women with breast cancer who are candidates for HER-2-targeted therapies.

BACKGROUND: Different technologies, such as quantitative real-time PCR or microarrays, have been developed to measure microRNA (miRNA) expression levels. Quantification of miRNA transcripts implicates data normalization using endogenous and exogenous reference genes for data correction. However, there is no consensus about an optimal normalization strategy. The choice of a reference gene remains problematic and can have a serious impact on the actual available transcript levels and, consequently, on the biological interpretation of data. CONTENT: In this review article we discuss the reliability of the use of small RNAs, commonly reported in the literature as miRNA expression normalizers, and compare different strategies used for data normalization. SUMMARY: A workflow strategy is proposed for normalization of miRNA expression data in an attempt to provide a basis for the establishment of a global standard procedure that will allow comparison across studies.

Abstract Pancreatic ductal adenocarcinoma (PDAC) is expected to be the second most deadly cancer by 2040, owing to the high incidence of metastatic disease and limited responses to treatment 1,2 . Less than half of all patients respond to the primary treatment for PDAC, chemotherapy 3,4 , and genetic alterations alone cannot explain this 5 . Diet is an environmental factor that can influence the response to therapies, but its role in PDAC is unclear. Here, using shotgun metagenomic sequencing and metabolomic screening, we show that the microbiota-derived tryptophan metabolite indole-3-acetic acid (3-IAA) is enriched in patients who respond to treatment. Faecal microbiota transplantation, short-term dietary manipulation of tryptophan and oral 3-IAA administration increase the efficacy of chemotherapy in humanized gnotobiotic mouse models of PDAC. Using a combination of loss- and gain-of-function experiments, we show that the efficacy of 3-IAA and chemotherapy is licensed by neutrophil-derived myeloperoxidase. Myeloperoxidase oxidizes 3-IAA, which in combination with chemotherapy induces a downregulation of the reactive oxygen species (ROS)-degrading enzymes glutathione peroxidase 3 and glutathione peroxidase 7. All of this results in the accumulation of ROS and the downregulation of autophagy in cancer cells, which compromises their metabolic fitness and, ultimately, their proliferation. In humans, we observed a significant correlation between the levels of 3-IAA and the efficacy of therapy in two independent PDAC cohorts. In summary, we identify a microbiota-derived metabolite that has clinical implications in the treatment of PDAC, and provide a motivation for considering nutritional interventions during the treatment of patients with cancer.

The "cancerized field" concept posits that cancer-prone cells in a given tissue share an oncogenic mutation, but only discreet clones within the field initiate tumors. Most benign nevi carry oncogenic BRAF(V600E) mutations but rarely become melanoma. The zebrafish crestin gene is expressed embryonically in neural crest progenitors (NCPs) and specifically reexpressed in melanoma. Live imaging of transgenic zebrafish crestin reporters shows that within a cancerized field (BRAF(V600E)-mutant; p53-deficient), a single melanocyte reactivates the NCP state, revealing a fate change at melanoma initiation in this model. NCP transcription factors, including sox10, regulate crestin expression. Forced sox10 overexpression in melanocytes accelerated melanoma formation, which is consistent with activation of NCP genes and super-enhancers leading to melanoma. Our work highlights NCP state reemergence as a key event in melanoma initiation.

Importance: Cytotoxic CD8+ tumor-infiltrating lymphocytes (TILs) participate in immune control of epithelial ovarian cancer; however, little is known about prognostic patterns of CD8+ TILs by histotype and in relation to other clinical factors. Objective: To define the prognostic role of CD8+ TILs in epithelial ovarian cancer. Design, Setting, and Participants: This was a multicenter observational, prospective survival cohort study of the Ovarian Tumor Tissue Analysis Consortium. More than 5500 patients, including 3196 with high-grade serous ovarian carcinomas (HGSOCs), were followed prospectively for over 24 650 person-years. Exposures: Following immunohistochemical analysis, CD8+ TILs were identified within the epithelial components of tumor islets. Patients were grouped based on the estimated number of CD8+ TILs per high-powered field: negative (none), low (1-2), moderate (3-19), and high (≥20). CD8+ TILs in a subset of patients were also assessed in a quantitative, uncategorized manner, and the functional form of associations with survival was assessed using penalized B-splines. Main Outcomes and Measures: Overall survival time. Results: The final sample included 5577 women; mean age at diagnosis was 58.4 years (median, 58.2 years). Among the 5 major invasive histotypes, HGSOCs showed the most infiltration. CD8+ TILs in HGSOCs were significantly associated with longer overall survival; median survival was 2.8 years for patients with no CD8+ TILs and 3.0 years, 3.8 years, and 5.1 years for patients with low, moderate, or high levels of CD8+ TILs, respectively (P value for trend = 4.2 × 10−16). A survival benefit was also observed among women with endometrioid and mucinous carcinomas, but not for those with the other histotypes. Among HGSOCs, CD8+ TILs were favorable regardless of extent of residual disease following cytoreduction, known standard treatment, and germline BRCA1 pathogenic mutation, but were not prognostic for BRCA2 mutation carriers. Evaluation of uncategorized CD8+ TIL counts showed a near-log-linear functional form. Conclusions and Relevance: This study demonstrates the histotype-specific nature of immune infiltration and provides definitive evidence for a dose-response relationship between CD8+ TILs and HGSOC survival. That the extent of infiltration is prognostic, not merely its presence or absence, suggests that understanding factors that drive infiltration will be the key to unraveling outcome heterogeneity in this cancer.

BACKGROUND & AIMS: Microsatellite instability (MSI) and mismatch-repair deficiency (dMMR) in colorectal tumors are used to select treatment for patients. Deep learning can detect MSI and dMMR in tumor samples on routine histology slides faster and less expensively than molecular assays. However, clinical application of this technology requires high performance and multisite validation, which have not yet been performed. METHODS: We collected H&E-stained slides and findings from molecular analyses for MSI and dMMR from 8836 colorectal tumors (of all stages) included in the MSIDETECT consortium study, from Germany, the Netherlands, the United Kingdom, and the United States. Specimens with dMMR were identified by immunohistochemistry analyses of tissue microarrays for loss of MLH1, MSH2, MSH6, and/or PMS2. Specimens with MSI were identified by genetic analyses. We trained a deep-learning detector to identify samples with MSI from these slides; performance was assessed by cross-validation (N = 6406 specimens) and validated in an external cohort (n = 771 specimens). Prespecified endpoints were area under the receiver operating characteristic (AUROC) curve and area under the precision-recall curve (AUPRC). RESULTS: The deep-learning detector identified specimens with dMMR or MSI with a mean AUROC curve of 0.92 (lower bound, 0.91; upper bound, 0.93) and an AUPRC of 0.63 (range, 0.59-0.65), or 67% specificity and 95% sensitivity, in the cross-validation development cohort. In the validation cohort, the classifier identified samples with dMMR with an AUROC of 0.95 (range, 0.92-0.96) without image preprocessing and an AUROC of 0.96 (range, 0.93-0.98) after color normalization. CONCLUSIONS: We developed a deep-learning system that detects colorectal cancer specimens with dMMR or MSI using H&E-stained slides; it detected tissues with dMMR with an AUROC of 0.96 in a large, international validation cohort. This system might be used for high-throughput, low-cost evaluation of colorectal tissue specimens.