Opening the black box: spatial transcriptomics and the relevance of AI-detected prognostic regions in high grade serous carcinoma.

Image based deep learning models are used to extract new information from standard H&E pathology slides, however, biological interpretation of the features detected by artificial intelligence (AI) remains a challenge. High-grade serous ovarian carcinoma (HGSC) is characterized by aggressive behavior and chemotherapy resistance, but also by striking variability in outcome. Our understanding of this disease is limited, partly due to considerable tumor heterogeneity. We previously trained an AI model to identify HGSC tumor regions that are highly associated with outcome status but are indistinguishable by conventional morphologic methods. Here we applied spatial transcriptomics to further profile the AI-identified tumor regions in 16 patients (8 per outcome group) and identify molecular features related to disease outcome in patients who underwent primary debulking surgery and platinum-based chemotherapy. We examined FFPE tissue from 1) regions identified by the AI model as highly associated with short or extended chemotherapy response, and 2) background tumor regions (not identified by the AI model as highly associated with outcome status) from the same tumors. We show that the transcriptomic profiles of AI-identified regions are more distinct than background regions from the same tumors, are superior in predicting outcome, and differ in several pathways including those associated with chemoresistance in HGSC. Further, we find that poor outcome and good outcome regions are enriched by different tumor subpopulations, suggesting distinctive interaction patterns. In summary, our work presents proof of concept that AI-guided spatial transcriptomic analysis improves recognition of biologic features relevant to patient outcome.

PIK3R1 fusion drives chemoresistance in ovarian cancer by activating ERK1/2 and inducing rod and ring-like structures.

Gene fusions are common in high-grade serous ovarian cancer (HGSC). Such genetic lesions may promote tumorigenesis, but the pathogenic mechanisms are currently poorly understood. Here, we investigated the role of a PIK3R1-CCDC178 fusion identified from a patient with advanced HGSC. We show that the fusion induces HGSC cell migration by regulating ERK1/2 and increases resistance to platinum treatment. Platinum resistance was associated with rod and ring-like cellular structure formation. These structures contained, in addition to the fusion protein, CIN85, a key regulator of PI3K-AKT-mTOR signaling. Our data suggest that the fusion-driven structure formation induces a previously unrecognized cell survival and resistance mechanism, which depends on ERK1/2-activation.

High-Resolution Genotyping of Formalin-Fixed Tissue Accurately Estimates Polygenic Risk Scores in Human Diseases.

Formalin-fixed paraffin-embedded (FFPE) tissues stored in biobanks and pathology archives are a vast but underutilized source for molecular studies on different diseases. Beyond being the "gold standard" for preservation of diagnostic human tissues, FFPE samples retain similar genetic information as matching blood samples, which could make FFPE samples an ideal resource for genomic analysis. However, research on this resource has been hindered by the perception that DNA extracted from FFPE samples is of poor quality. Here, we show that germline disease-predisposing variants and polygenic risk scores (PRS) can be identified from FFPE normal tissue (FFPE-NT) DNA with high accuracy. We optimized the performance of FFPE-NT DNA on a genome-wide array containing 657,675 variants. Via a series of testing and validation phases, we established a protocol for FFPE-NT genotyping with results comparable with blood genotyping. The median call rate of FFPE-NT samples in the validation phase was 99.85% (range 98.26%-99.94%) and median concordance with matching blood samples was 99.79% (range 98.85%-99.9%). We also demonstrated that a rare pathogenic PALB2 genetic variant predisposing to cancer can be correctly identified in FFPE-NT samples. We further imputed the FFPE-NT genotype data and calculated the FFPE-NT genome-wide PRS in 3 diseases and 4 disease risk variables. In all cases, FFPE-NT and matching blood PRS were highly concordant (all Pearson's r > 0.95). The ability to precisely genotype FFPE-NT on a genome-wide array enables translational genomics applications of archived FFPE-NT samples with the possibility to link to corresponding phenotypes and longitudinal health data.

DLL4/Notch3/WNT5B axis mediates bidirectional prometastatic crosstalk between melanoma and lymphatic endothelial cells.

Despite strong indications that interactions between melanoma and lymphatic vessels actively promote melanoma progression, the molecular mechanisms are not yet completely understood. To characterize molecular factors of this crosstalk, we established human primary lymphatic endothelial cell (LEC) cocultures with human melanoma cell lines. Here, we show that coculture with melanoma cells induced transcriptomic changes in LECs and led to multiple changes in their function. WNT5B, a paracrine signaling molecule upregulated in melanoma cells upon LEC interaction, was found to contribute to the functional changes in LECs. Moreover, WNT5B transcription was regulated by Notch3 in melanoma cells following the coculture with LECs, and Notch3 and WNT5B were coexpressed in melanoma patient primary tumor and metastasis samples. Moreover, melanoma cells derived from LEC coculture escaped efficiently from the primary site to the proximal tumor-draining lymph nodes, which was impaired upon WNT5B depletion. This supported the role of WNT5B in promoting the metastatic potential of melanoma cells through its effects on LECs. Finally, DLL4, a Notch ligand expressed in LECs, was identified as an upstream inducer of the Notch3/WNT5B axis in melanoma. This study elucidated WNT5B as a key molecular factor mediating bidirectional crosstalk between melanoma cells and lymphatic endothelium and promoting melanoma metastasis.

Characterization of Expression and Function of the Formins FHOD1, INF2, and DAAM1 in HER2-Positive Breast Cancer.

PURPOSE: Human epidermal growth factor receptor 2 (HER2)-targeted therapies, such as trastuzumab, benefit patients with HER2-positive metastatic breast cancer; however, owing to traditional pathway activation or alternative signaling, resistance persists. Given the crucial role of the formin family in shaping the actin cytoskeleton during cancer progression, these proteins may function downstream of the HER2 signaling pathway. Our aim was to uncover the potential correlations between formins and HER2 expression using a combination of public databases, immunohistochemistry, and functional in vitro assays. METHODS: Using online databases, we identified a negative prognostic correlation between specific formins mRNA expression in HER2-positive cancers. To validate these findings at the protein level, immunohistochemistry was performed on HER2 subtype breast cancer tumors to establish the links between staining patterns and clinical characteristics. We then knocked down individual or combined formins in MDA-MB-453 and SK-BR-3 cells and investigated their effects on wound healing, transwell migration, and proliferation. Furthermore, we investigated the effects of erb-b2 receptor tyrosine kinase 2 (ERBB2)/HER2 small interfering RNA (siRNA)-mediated knockdown on the PI3K/Akt and MEK/ERK1 pathways as well as on selected formins. RESULTS: Our results revealed that correlations between INF2, FHOD1, and DAAM1 mRNA expression and ERBB2 in HER2-subtype breast cancer were associated with worse outcomes. Using immunohistochemistry, we found that high FHOD1 protein expression was linked to higher histological grades and was negatively correlated with estrogen and progesterone receptor positivity. Upon formins knockdown, we observed effects on wound healing and transwell migration, with a minimal impact on proliferation, which was evident through single and combined knockdowns in both cell lines. Notably, siRNA-mediated knockdown of HER2 affected FHOD1 and INF2 expression, along with the phosphorylated Akt/MAPK states. CONCLUSION: Our study highlights the roles of FHOD1 and INF2 as downstream effectors of the HER2/Akt and HER2/MAPK pathways, suggesting that they are potential therapeutic targets in HER2-positive breast cancer.

The impact of implementing current treatment modalities and female sex on gastric cancer outcomes, 2000-2016: a longitudinal nationwide cohort study.

BACKGROUND: The implementation of current treatment modalities and their impact on nationwide gastric cancer outcomes remain poorly understood. Biological differences between females and males could impact survival. We aimed to analyze rates of gastric surgery, chemotherapy, and radiotherapy as well as changes in overall survival among gastric cancer patients diagnosed between 2000-2008 and 2009-2016, respectively, in Finland. MATERIAL AND METHODS: Data on gastric cancer patients were collected from national registries. Cox regression analysis and the Kaplan-Meier method were used to analyze differences in survival. RESULTS: We identified 9223 histologically confirmed gastric cancer patients. The rate of gastric surgery decreased from 44% (n = 2282) to 34% (n = 1368; p < 0.001). The proportion of gastric surgery patients who underwent preoperative oncological treatment increased from 0.5% (n = 12) to 16.2% (n = 222) between the calendar periods (p < 0.001) and stood at 30% in 2016. The median overall survival (OS) improved from 30 months [95% confidence interval (CI) 28-33] to 38 months (95%CI 33-42; p = 0.006) and the period 2009-2016 independently associated with a lower risk of death [hazard ratio (HR) 0.78, 95%CI 0.70-0.87] among patients who underwent gastric surgery. Females exhibited a lower risk of death (HR 0.88, 95%CI 0.81-0.97) among patients who underwent gastric surgery. CONCLUSION: Preoperative oncological treatment was gradually introduced into clinical practice and OS among gastric surgery patients improved. Moreover, female surgical patients exhibited a better survival than male patients.

A platform for efficient establishment and drug-response profiling of high-grade serous ovarian cancer organoids.

The broad research use of organoids from high-grade serous ovarian cancer (HGSC) has been hampered by low culture success rates and limited availability of fresh tumor material. Here, we describe a method for generation and long-term expansion of HGSC organoids with efficacy markedly improved over previous reports (53% vs. 23%-38%). We established organoids from cryopreserved material, demonstrating the feasibility of using viably biobanked tissue for HGSC organoid derivation. Genomic, histologic, and single-cell transcriptomic analyses revealed that organoids recapitulated genetic and phenotypic features of original tumors. Organoid drug responses correlated with clinical treatment outcomes, although in a culture conditions-dependent manner and only in organoids maintained in human plasma-like medium (HPLM). Organoids from consenting patients are available to the research community through a public biobank and organoid genomic data are explorable through an interactive online tool. Taken together, this resource facilitates the application of HGSC organoids in basic and translational ovarian cancer research.

Evolutionary states and trajectories characterized by distinct pathways stratify patients with ovarian high grade serous carcinoma.

Ovarian high-grade serous carcinoma (HGSC) is typically diagnosed at an advanced stage, with multiple genetically heterogeneous clones existing in the tumors long before therapeutic intervention. Herein we integrate clonal composition and topology using whole-genome sequencing data from 510 samples of 148 patients with HGSC in the prospective, longitudinal, multiregion DECIDER study. Our results reveal three evolutionary states, which have distinct features in genomics, pathways, and morphological phenotypes, and significant association with treatment response. Nested pathway analysis suggests two evolutionary trajectories between the states. Experiments with five tumor organoids and three PI3K inhibitors support targeting tumors with enriched PI3K/AKT pathway with alpelisib. Heterogeneity analysis of samples from multiple anatomical sites shows that site-of-origin samples have 70% more unique clones than metastatic tumors or ascites. In conclusion, these analysis and visualization methods enable integrative tumor evolution analysis to identify patient subtypes using data from longitudinal, multiregion cohorts.

Functional Homologous Recombination Assay on FFPE Specimens of Advanced High-Grade Serous Ovarian Cancer Predicts Clinical Outcomes.

PURPOSE: Deficiency in homologous recombination (HR) repair of DNA damage is characteristic of many high-grade serous ovarian cancers (HGSC). It is imperative to identify patients with homologous recombination-deficient (HRD) tumors as they are most likely to benefit from platinum-based chemotherapy and PARP inhibitors (PARPi). Existing methods measure historical, not necessarily current HRD and/or require high tumor cell content, which is not achievable for many patients. We set out to develop a clinically feasible assay for identifying functionally HRD tumors that can predict clinical outcomes. EXPERIMENTAL DESIGN: We quantified RAD51, a key HR protein, in immunostained formalin-fixed, paraffin-embedded (FFPE) tumor samples obtained from chemotherapy-naïve and neoadjuvant chemotherapy (NACT)-treated HGSC patients. We defined cutoffs for functional HRD separately for these sample types, classified the patients accordingly as HRD or HR-proficient, and analyzed correlations with clinical outcomes. From the same specimens, genomics-based HRD estimates (HR gene mutations, genomic signatures, and genomic scars) were also determined, and compared with functional HR (fHR) status. RESULTS: fHR status significantly predicted several clinical outcomes, including progression-free survival (PFS) and overall survival (OS), when determined from chemo-naïve (PFS, P < 0.0001; OS, P < 0.0001) as well as NACT-treated (PFS, P < 0.0001; OS, P = 0.0033) tumor specimens. The fHR test also identified as HRD those PARPi-at-recurrence-treated patients with longer OS (P = 0.0188). CONCLUSIONS: We developed an fHR assay performed on routine FFPE specimens, obtained from either chemo-naïve or NACT-treated HGSC patients, that can significantly predict real-world platinum-based chemotherapy and PARPi response. See related commentary by Garg and Oza, p. 2957.

Effects of Wee1 inhibitor adavosertib on patient-derived high-grade serous ovarian cancer cells are multiple and independent of homologous recombination status.

Objective: A major challenge in the treatment of platinum-resistant high-grade serous ovarian cancer (HGSOC) is lack of effective therapies. Much of ongoing research on drug candidates relies on HGSOC cell lines that are poorly documented. The goal of this study was to screen for effective, state-of-the-art drug candidates using primary HGSOC cells. In addition, our aim was to dissect the inhibitory activities of Wee1 inhibitor adavosertib on primary and conventional HGSOC cell lines. Methods: A comprehensive drug sensitivity and resistance testing (DSRT) on 306 drug compounds was performed on three patient-derived genetically unique HGSOC cell lines and two commonly used ovarian cancer cell lines. The effect of adavosertib on the cell lines was tested in several assays, including cell-cycle analysis, apoptosis induction, proliferation, wound healing, DNA damage, and effect on nuclear integrity. Results: Several compounds exerted cytotoxic activity toward all cell lines, when tested in both adherent and spheroid conditions. In further cytotoxicity tests, adavosertib exerted the most consistent cytotoxic activity. Adavosertib affected cell-cycle control in patient-derived and conventional HGSOC cells, inducing G2/M accumulation and reducing cyclin B1 levels. It induced apoptosis and inhibited proliferation and migration in all cell lines. Furthermore, the DNA damage marker γH2AX and the number of abnormal cell nuclei were clearly increased following adavosertib treatment. Based on the homologous recombination (HR) signature and functional HR assays of the cell lines, the effects of adavosertib were independent of the cells' HR status. Conclusion: Our study indicates that Wee1 inhibitor adavosertib affects several critical functions related to proliferation, cell cycle and division, apoptosis, and invasion. Importantly, the effects are consistent in all tested cell lines, including primary HGSOC cells, and independent of the HR status of the cells. Wee1 inhibition may thus provide treatment opportunities especially for patients, whose cancer has acquired resistance to platinum-based chemotherapy or PARP inhibitors.

A Subset of Secreted Proteins in Ascites Can Predict Platinum-Free Interval in Ovarian Cancer.

The time between the last cycle of chemotherapy and recurrence, the platinum-free interval (PFI), predicts overall survival in high-grade serous ovarian cancer (HGSOC). To identify secreted proteins associated with a shorter PFI, we utilized machine learning to predict the PFI from ascites composition. Ascites from stage III/IV HGSOC patients treated with neoadjuvant chemotherapy (NACT) or primary debulking surgery (PDS) were screened for secreted proteins and Lasso regression models were built to predict the PFI. Through regularization techniques, the number of analytes used in each model was reduced; to minimize overfitting, we utilized an analysis of model robustness. This resulted in models with 26 analytes and a root-mean-square error (RMSE) of 19 days for the NACT cohort and 16 analytes and an RMSE of 7 days for the PDS cohort. High concentrations of MMP-2 and EMMPRIN correlated with a shorter PFI in the NACT patients, whereas high concentrations of uPA Urokinase and MMP-3 correlated with a shorter PFI in PDS patients. Our results suggest that the analysis of ascites may be useful for outcome prediction and identified factors in the tumor microenvironment that may lead to worse outcomes. Our approach to tuning for model stability, rather than only model accuracy, may be applicable to other biomarker discovery tasks.

Single-Cell RNA Sequencing-Based Characterization of Resident Lung Mesenchymal Stromal Cells in Bronchopulmonary Dysplasia.

Late lung development is a period of alveolar and microvascular formation, which is pivotal in ensuring sufficient and effective gas exchange. Defects in late lung development manifest in premature infants as a chronic lung disease named bronchopulmonary dysplasia (BPD). Numerous studies demonstrated the therapeutic properties of exogenous bone marrow and umbilical cord-derived mesenchymal stromal cells (MSCs) in experimental BPD. However, very little is known regarding the regenerative capacity of resident lung MSCs (L-MSCs) during normal development and in BPD. In this study we aimed to characterize the L-MSC population in homeostasis and upon injury. We used single-cell RNA sequencing (scRNA-seq) to profile in situ Ly6a+ L-MSCs in the lungs of normal and O2-exposed neonatal mice (a well-established model to mimic BPD) at 3 developmental timepoints (postnatal days 3, 7, and 14). Hyperoxia exposure increased the number and altered the expression profile of L-MSCs, particularly by increasing the expression of multiple pro-inflammatory, pro-fibrotic, and anti-angiogenic genes. In order to identify potential changes induced in the L-MSCs transcriptome by storage and culture, we profiled 15 000 Ly6a+ L-MSCs after in vitro culture. We observed great differences in expression profiles of in situ and cultured L-MSCs, particularly those derived from healthy lungs. Additionally, we have identified the location of Ly6a+/Col14a1+ L-MSCs in the developing lung and propose Serpinf1 as a novel, culture-stable marker of L-MSCs. Finally, cell communication analysis suggests inflammatory signals from immune and endothelial cells as main drivers of hyperoxia-induced changes in L-MSCs transcriptome.

Ovarian Cancers with Low CIP2A Tumor Expression Constitute an APR-246-Sensitive Disease Subtype.

Identification of ovarian cancer patient subpopulations with increased sensitivity to targeted therapies could offer significant clinical benefit. We report that 22% of the high-grade ovarian cancer tumors at diagnosis express CIP2A oncoprotein at low levels. Furthermore, regardless of their significantly lower likelihood of disease relapse after standard chemotherapy, a portion of relapsed tumors retain their CIP2A-deficient phenotype. Through a screen for therapeutics that would preferentially kill CIP2A-deficient ovarian cancer cells, we identified reactive oxygen species inducer APR-246, tested previously in ovarian cancer clinical trials. Consistent with CIP2A-deficient ovarian cancer subtype in humans, CIP2A is dispensable for development of MISIIR-Tag-driven mouse ovarian cancer tumors. Nevertheless, CIP2A-null ovarian cancer tumor cells from MISIIR-Tag mice displayed APR-246 hypersensitivity both in vitro and in vivo. Mechanistically, the lack of CIP2A expression hypersensitizes the ovarian cancer cells to APR-246 by inhibition of NF-κB activity. Accordingly, combination of APR-246 and NF-κB inhibitor compounds strongly synergized in killing of CIP2A-positive ovarian cancer cells. Collectively, the results warrant consideration of clinical testing of APR-246 for CIP2A-deficient ovarian cancer tumor subtype patients. Results also reveal CIP2A as a candidate APR-246 combination therapy target for ovarian cancer.

Longitudinal single-cell RNA-seq analysis reveals stress-promoted chemoresistance in metastatic ovarian cancer.

Chemotherapy resistance is a critical contributor to cancer mortality and thus an urgent unmet challenge in oncology. To characterize chemotherapy resistance processes in high-grade serous ovarian cancer, we prospectively collected tissue samples before and after chemotherapy and analyzed their transcriptomic profiles at a single-cell resolution. After removing patient-specific signals by a novel analysis approach, PRIMUS, we found a consistent increase in stress-associated cell state during chemotherapy, which was validated by RNA in situ hybridization and bulk RNA sequencing. The stress-associated state exists before chemotherapy, is subclonally enriched during the treatment, and associates with poor progression-free survival. Co-occurrence with an inflammatory cancer-associated fibroblast subtype in tumors implies that chemotherapy is associated with stress response in both cancer cells and stroma, driving a paracrine feed-forward loop. In summary, we have found a resistant state that integrates stromal signaling and subclonal evolution and offers targets to overcome chemotherapy resistance.

QuantISH: RNA in situ hybridization image analysis framework for quantifying cell type-specific target RNA expression and variability.

RNA in situ hybridization (RNA-ISH) is a powerful spatial transcriptomics technology to characterize target RNA abundance and localization in individual cells. This allows analysis of tumor heterogeneity and expression localization, which are not readily obtainable through transcriptomic data analysis. RNA-ISH experiments produce large amounts of data and there is a need for automated analysis methods. Here we present QuantISH, a comprehensive open-source RNA-ISH image analysis pipeline that quantifies marker expressions in individual carcinoma, immune, and stromal cells on chromogenic or fluorescent in situ hybridization images. QuantISH is designed to be modular and can be adapted to various image and sample types and staining protocols. We show that in chromogenic RNA in situ hybridization images of high-grade serous carcinoma (HGSC) QuantISH cancer cell classification has high precision, and signal expression quantification is in line with visual assessment. We further demonstrate the power of QuantISH by showing that CCNE1 average expression and DDIT3 expression variability, as captured by the variability factor developed herein, act as candidate biomarkers in HGSC. Altogether, our results demonstrate that QuantISH can quantify RNA expression levels and their variability in carcinoma cells, and thus paves the way to utilize RNA-ISH technology.

Dubious effects of methadone as an "anticancer" drug on ovarian cancer cell-lines and patient-derived tumor-spheroids.

BACKGROUND: The opioid agonist D,L-methadone exerts analgesic effects via the mu opioid receptor, encoded by OPRM1 and therefore plays a role in chronic pain management. In preclinical tumor-models D,L-methadone shows apoptotic and chemo-sensitizing effects and was therefore hyped as an off-label "anticancer" drug without substantiation from clinical trials. Its effects in ovarian cancer (OC) are completely unexplored. METHODS: We analyzed OPRM1-mRNA expression in six cisplatin-sensitive, two cisplatin-resistant OC cell-lines, 170 OC tissue samples and 12 non-neoplastic control tissues. Pro-angiogenetic, cytotoxic and apoptotic effects of D,L-methadone were evaluated in OC cell-lines and four patient-derived tumor-spheroid models. RESULTS: OPRM1 was transcriptionally expressed in 69% of OC-tissues and in three of eight OC cell-lines. D,L-methadone exposure significantly reduced cell-viability in five OC cell-lines irrespective of OPRM1 expression. D,L-methadone, applied alone or combined with cisplatin, showed no significant effects on apoptosis or VEGF secretion in cell-lines. Notably, in two of the four spheroid models, treatment with D,L-methadone significantly enhanced cell growth (by up to 121%), especially after long-term exposure. This is consistent with the observed attenuation of the inhibitory effects of cisplatin in three spheroid models when adding D,L-methadone. The effect of methadone treatment on VEGF secretion in tumor-spheroids was inconclusive. CONCLUSIONS: Our study demonstrates that certain OC samples express OPRM1, which, however, is not a prerequisite for D,L-methadone function. As such, D,L-methadone may exert also detrimental effects by stimulating the growth of certain OC-cells and abrogating cisplatin's therapeutic effect.

Long-term nationwide trends in the treatment of and outcomes among pancreatic cancer patients.

Whilst treatment modalities for pancreatic cancer patients have evolved in recent years, their impact on outcomes remains relatively unexamined on a national scale. We aimed to analyse changes in overall survival and trends in surgical and oncological treatments in pancreatic cancer patients diagnosed in the periods 2000 through 2008 and 2009 through 2016 in Finland. We collected data for pancreatic cancer patients diagnosed between 2000 and 2016, gathering data from the Finnish national registries on surgeries, oncological treatments and time of death. Follow-up continued through the end of 2018. We compared patients diagnosed between 2000 and 2008 to those diagnosed between 2009 through 2016. Our study comprised 14 712 pancreatic cancer patients. There was no significant change in the national resection rate (8.1% vs 8.0%, p = 0.690). In radical surgery patients, median survival improved from 20 months (95% confidence interval (CI) 18-22) to 28 months (CI 25-31) (p < 0.001), with 1-year survival ranging from 70% to 81%. In the no-surgery group, median survival slightly improved from 3.1 months (CI 3.0-3.3) to 3.3 months (CI 3.1-3.4) (p < 0.001). The proportion of radical surgery patients receiving preoperative oncological treatment increased from 4% to 13% (p < 0.001) and only postoperative treatment from 25% to 47% (p < 0.001). Whilst the resection rate did not increase, the prognosis of pancreatic cancer patients improved, particularly amongst radical surgery patients resulting most likely from the fact that a larger proportion of patients receive more effective oncological treatments.

Artificial intelligence-based image analysis can predict outcome in high-grade serous carcinoma via histology alone.

High-grade extrauterine serous carcinoma (HGSC) is an aggressive tumor with high rates of recurrence, frequent chemotherapy resistance, and overall 5-year survival of less than 50%. Beyond determining and confirming the diagnosis itself, pathologist review of histologic slides provides no prognostic or predictive information, which is in sharp contrast to almost all other carcinoma types. Deep-learning based image analysis has recently been able to predict outcome and/or identify morphology-based representations of underlying molecular alterations in other tumor types, such as colorectal carcinoma, lung carcinoma, breast carcinoma, and melanoma. Using a carefully stratified HGSC patient cohort consisting of women (n = 30) with similar presentations who experienced very different treatment responses (platinum free intervals of either ≤ 6 months or ≥ 18 months), we used whole slide images (WSI, n = 205) to train a convolutional neural network. The neural network was trained, in three steps, to identify morphologic regions (digital biomarkers) that are highly associating with one or the other treatment response group. We tested the classifier using a separate 22 slide test set, and 18/22 slides were correctly classified. We show that a neural network based approach can discriminate extremes in patient response to primary platinum-based chemotherapy with high sensitivity (73%) and specificity (91%). These proof-of-concept results are novel, because for the first time, prospective prognostic information is identified specifically within HGSC tumor morphology.

Network-guided identification of cancer-selective combinatorial therapies in ovarian cancer.

Each patient's cancer consists of multiple cell subpopulations that are inherently heterogeneous and may develop differing phenotypes such as drug sensitivity or resistance. A personalized treatment regimen should therefore target multiple oncoproteins in the cancer cell populations that are driving the treatment resistance or disease progression in a given patient to provide maximal therapeutic effect, while avoiding severe co-inhibition of non-malignant cells that would lead to toxic side effects. To address the intra- and inter-tumoral heterogeneity when designing combinatorial treatment regimens for cancer patients, we have implemented a machine learning-based platform to guide identification of safe and effective combinatorial treatments that selectively inhibit cancer-related dysfunctions or resistance mechanisms in individual patients. In this case study, we show how the platform enables prediction of cancer-selective drug combinations for patients with high-grade serous ovarian cancer using single-cell imaging cytometry drug response assay, combined with genome-wide transcriptomic and genetic profiles. The platform makes use of drug-target interaction networks to prioritize those combinations that warrant further preclinical testing in scarce patient-derived primary cells. During the case study in ovarian cancer patients, we investigated (i) the relative performance of various ensemble learning algorithms for drug response prediction, (ii) the use of matched single-cell RNA-sequencing data to deconvolute cell population-specific transcriptome profiles from bulk RNA-seq data, (iii) and whether multi-patient or patient-specific predictive models lead to better predictive accuracy. The general platform and the comparison results are expected to become useful for future studies that use similar predictive approaches also in other cancer types.

WNT2 activation through proximal germline deletion predisposes to small intestinal neuroendocrine tumors and intestinal adenocarcinomas.

Many hereditary cancer syndromes are associated with an increased risk of small and large intestinal adenocarcinomas. However, conditions bearing a high risk to both adenocarcinomas and neuroendocrine tumors are yet to be described. We studied a family with 16 individuals in four generations affected by a wide spectrum of intestinal tumors, including hyperplastic polyps, adenomas, small intestinal neuroendocrine tumors, and colorectal and small intestinal adenocarcinomas. To assess the genetic susceptibility and understand the novel phenotype, we utilized multiple molecular methods, including whole genome sequencing, RNA sequencing, single cell sequencing, RNA in situ hybridization and organoid culture. We detected a heterozygous deletion at the cystic fibrosis locus (7q31.2) perfectly segregating with the intestinal tumor predisposition in the family. The deletion removes a topologically associating domain border between CFTR and WNT2, aberrantly activating WNT2 in the intestinal epithelium. These consequences suggest that the deletion predisposes to small intestinal neuroendocrine tumors and small and large intestinal adenocarcinomas, and reveals the broad tumorigenic effects of aberrant WNT activation in the human intestine.