Respiratory complex I regulates dendritic cell maturation in explant model of human tumor immune microenvironment.

BACKGROUND: Combining cytotoxic chemotherapy or novel anticancer drugs with T-cell modulators holds great promise in treating advanced cancers. However, the response varies depending on the tumor immune microenvironment (TIME). Therefore, there is a clear need for pharmacologically tractable models of the TIME to dissect its influence on mono- and combination treatment response at the individual level. METHODS: Here we establish a patient-derived explant culture (PDEC) model of breast cancer, which retains the immune contexture of the primary tumor, recapitulating cytokine profiles and CD8+T cell cytotoxic activity. RESULTS: We explored the immunomodulatory action of a synthetic lethal BCL2 inhibitor venetoclax+metformin drug combination ex vivo, discovering metformin cannot overcome the lymphocyte-depleting action of venetoclax. Instead, metformin promotes dendritic cell maturation through inhibition of mitochondrial complex I, increasing their capacity to co-stimulate CD4+T cells and thus facilitating antitumor immunity. CONCLUSIONS: Our results establish PDECs as a feasible model to identify immunomodulatory functions of anticancer drugs in the context of patient-specific TIME.

Characterization and clinical impact of the tumor microenvironment in post-transplant aggressive B-cell lymphomas.

Post-transplant lymphoproliferative disorders (PTLD) are iatrogenic immune deficiency-associated lymphoid/plasmacytic proliferations developing due to immunosuppression in solid organ or hematopoietic stem cell allograft patients. PTLD are characterized by abnormal proliferation of lymphoid cells and have a heterogeneous clinical behavior. We profiled expression of >700 tumor microenvironment (TME)-related genes in 75 post-transplant aggressive B-cell lymphomas (PTABCL). Epstein-Barr virus (EBV)-positive PT-ABCL clustered together and were enriched for type I interferon pathway and antiviral-response genes. Additionally, a cytotoxicity gene signature associated with EBV-positivity and favorable overall survival (OS) (hazard ratio =0.61; P=0.019). In silico immunophenotyping revealed two subgroups with distinct immune cell compositions. The inflamed subgroup with higher proportions of immune cells had better outcome compared to noninflamed subgroup (median OS >200.0 vs. 15.2 months; P=0.006). In multivariable analysis with EBV status, International Prognostic Index, and rituximab-containing treatment, inflamed TME remained as an independent predictor for favorable outcome. We also compared TME between post-transplant and immunocompetent host diffuse large B-cell lymphomas (n=75) and discovered that the proportions of T cells were lower in PT-diffuse large B-cell lymphomas. In conclusion, we provide a comprehensive phenotypic characterization of PT-ABCL, highlighting the importance of immune cell composition of TME in determining the clinical behavior and prognosis of PT-ABCL.

Compressive stress-mediated p38 activation required for ERα + phenotype in breast cancer.

Breast cancer is now globally the most frequent cancer and leading cause of women's death. Two thirds of breast cancers express the luminal estrogen receptor-positive (ERα + ) phenotype that is initially responsive to antihormonal therapies, but drug resistance emerges. A major barrier to the understanding of the ERα-pathway biology and therapeutic discoveries is the restricted repertoire of luminal ERα + breast cancer models. The ERα + phenotype is not stable in cultured cells for reasons not fully understood. We examine 400 patient-derived breast epithelial and breast cancer explant cultures (PDECs) grown in various three-dimensional matrix scaffolds, finding that ERα is primarily regulated by the matrix stiffness. Matrix stiffness upregulates the ERα signaling via stress-mediated p38 activation and H3K27me3-mediated epigenetic regulation. The finding that the matrix stiffness is a central cue to the ERα phenotype reveals a mechanobiological component in breast tissue hormonal signaling and enables the development of novel therapeutic interventions. Subject terms: ER-positive (ER + ), breast cancer, ex vivo model, preclinical model, PDEC, stiffness, p38 SAPK.

Machine Learning of Bone Marrow Histopathology Identifies Genetic and Clinical Determinants in Patients with MDS.

In myelodysplastic syndrome (MDS) and myeloproliferative neoplasm (MPN), bone marrow (BM) histopathology is assessed to identify dysplastic cellular morphology, cellularity, and blast excess. Yet, other morphologic findings may elude the human eye. We used convolutional neural networks to extract morphologic features from 236 MDS, 87 MDS/MPN, and 11 control BM biopsies. These features predicted genetic and cytogenetic aberrations, prognosis, age, and gender in multivariate regression models. Highest prediction accuracy was found for TET2 [area under the receiver operating curve (AUROC) = 0.94] and spliceosome mutations (0.89) and chromosome 7 monosomy (0.89). Mutation prediction probability correlated with variant allele frequency and number of affected genes per pathway, demonstrating the algorithms' ability to identify relevant morphologic patterns. By converting regression models to texture and cellular composition, we reproduced the classical del(5q) MDS morphology consisting of hypolobulated megakaryocytes. In summary, this study highlights the potential of linking deep BM histopathology with genetics and clinical variables. SIGNIFICANCE: Histopathology is elementary in the diagnostics of patients with MDS, but its high-dimensional data are underused. By elucidating the association of morphologic features with clinical variables and molecular genetics, this study highlights the vast potential of convolutional neural networks in understanding MDS pathology and how genetics is reflected in BM morphology. See related commentary by Elemento, p. 195.

Spatial immunoprofiling of the intratumoral and peritumoral tissue of renal cell carcinoma patients.

While the abundance and phenotype of tumor-infiltrating lymphocytes are linked with clinical survival, their spatial coordination and its clinical significance remain unclear. Here, we investigated the immune profile of intratumoral and peritumoral tissue of clear cell renal cell carcinoma patients (n = 64). We trained a cell classifier to detect lymphocytes from hematoxylin and eosin stained tissue slides. Using unsupervised classification, patients were further classified into immune cold, hot and excluded topographies reflecting lymphocyte abundance and localization. The immune topography distribution was further validated with The Cancer Genome Atlas digital image dataset. We showed association between PBRM1 mutation and immune cold topography, STAG1 mutation and immune hot topography and BAP1 mutation and immune excluded topography. With quantitative multiplex immunohistochemistry we analyzed the expression of 23 lymphocyte markers in intratumoral and peritumoral tissue regions. To study spatial interactions, we developed an algorithm quantifying the proportion of adjacent immune cell pairs and their immunophenotypes. Immune excluded tumors were associated with superior overall survival (HR 0.19, p = 0.02) and less extensive metastasis. Intratumoral T cells were characterized with pronounced expression of immunological activation and exhaustion markers such as granzyme B, PD1, and LAG3. Immune cell interaction occurred most frequently in the intratumoral region and correlated with CD45RO expression. Moreover, high proportion of peritumoral CD45RO+ T cells predicted poor overall survival. In summary, intratumoral and peritumoral tissue regions represent distinct immunospatial profiles and are associated with clinicopathologic characteristics.

Oncogenic Ras Disrupts Epithelial Integrity by Activating the Transmembrane Serine Protease Hepsin.

Ras proteins play a causal role in human cancer by activating multiple pathways that promote cancer growth and invasion. However, little is known about how Ras induces the first diagnostic features of invasion in solid tumors, including loss of epithelial integrity and breaching of the basement membrane (BM). In this study, we found that oncogenic Ras strongly promotes the activation of hepsin, a member of the hepsin/TMPRSS type II transmembrane serine protease family. Mechanistically, the Ras-dependent hepsin activation was mediated via Raf-MEK-ERK signaling, which controlled hepsin protein stability through the heat shock transcription factor-1 stress pathway. In Ras-transformed three-dimensional mammary epithelial culture, ablation of hepsin restored desmosomal cell-cell junctions, hemidesmosomes, and BM integrity and epithelial cohesion. In tumor xenografts harboring mutant KRas, silencing of hepsin increased local invasion concomitantly with accumulation of collagen IV. These findings suggest that hepsin is a critical protease for Ras-dependent tumorigenesis, executing cell-cell and cell-matrix pathologies important for early tumor dissemination. SIGNIFICANCE: These findings identify the cell-surface serine protease hepsin as a potential therapeutic target for its role in oncogenic Ras-mediated deregulation of epithelial cell-cell and cell-matrix interactions and cohesion of epithelial structure.

Somatic mutations and T-cell clonality in patients with immunodeficiency.

Common variable immunodeficiency and other late-onset immunodeficiencies often co-manifest with autoimmunity and lymphoproliferation. The pathogenesis of most cases is elusive, as only a minor subset harbors known monogenic germline causes. The involvement of both B and T cells is however implicated. To study whether somatic mutations in CD4+ and CD8+ T cells associate with immunodeficiency, we recruited 17 patients and 21 healthy controls. Eight patients had late-onset common variable immunodeficiency and nine patients other immunodeficiency and/or severe autoimmunity. In total, autoimmunity occurred in 94% and lymphoproliferation in 65%. We performed deep sequencing of 2533 immune-associated genes from CD4+ and CD8+ cells. Deep T-cell receptor beta sequencing was used to characterize CD4+ and CD8+ T-cell receptor repertoires. The prevalence of somatic mutations was 65% in all immunodeficiency patients, 75% in common variable immunodeficiency and 48% in controls. Clonal hematopoiesis-associated variants in both CD4+ and CD8+ cells occurred in 24% of immunodeficiency patients. Results demonstrated mutations in known tumor suppressors, oncogenes, and genes that are critical for immune- and proliferative functions, such as STAT5B (two patients), C5AR1 (two patients), KRAS (one patient), and NOD2 (one patient). Additionally, as a marker of T-cell receptor repertoire perturbation, common variable immunodeficiency patients harbored increased frequencies of clones with identical complementarity determining region 3 sequences despite unique nucleotide sequences when compared to controls. In conclusion, somatic mutations in genes implicated for autoimmunity and lymphoproliferation are common in CD4+ and CD8+ cells of patients with immunodeficiency. They may contribute to immune dysregulation in a subset of immunodeficiency patients.

STAT3 Mutation Is Associated with STAT3 Activation in CD30+ ALK- ALCL.

Peripheral T-cell lymphomas (PTCL) are a heterogeneous, and often aggressive group of non-Hodgkin lymphomas. Recent advances in the molecular and genetic characterization of PTCLs have helped to delineate differences and similarities between the various subtypes, and the JAK/STAT pathway has been found to play an important oncogenic role. Here, we aimed to characterize the JAK/STAT pathway in PTCL subtypes and investigate whether the activation of the pathway correlates with the frequency of STAT gene mutations. Patient samples from AITL (n = 30), ALCL (n = 21) and PTCL-NOS (n = 12) cases were sequenced for STAT3, STAT5B, JAK1, JAK3, and RHOA mutations using amplicon sequencing and stained immunohistochemically for pSTAT3, pMAPK, and pAKT. We discovered STAT3 mutations in 13% of AITL, 13% of ALK+ ALCL, 38% of ALK- ALCL and 17% of PTCL-NOS cases. However, no STAT5B mutations were found and JAK mutations were only present in ALK- ALCL (15%). Concurrent mutations were found in all subgroups except ALK+ ALCL where STAT3 mutations were always seen alone. High pY-STAT3 expression was observed especially in AITL and ALCL samples. When studying JAK-STAT pathway mutations, pY-STAT3 expression was highest in PTCLs harboring either JAK1 or STAT3 mutations and CD30+ phenotype representing primarily ALK- ALCLs. Further investigation is needed to elucidate the molecular mechanisms of JAK-STAT pathway activation in PTCL.

Immune profiles in acute myeloid leukemia bone marrow associate with patient age, T-cell receptor clonality, and survival.

The immunologic microenvironment in various solid tumors is aberrant and correlates with clinical survival. Here, we present a comprehensive analysis of the immune environment of acute myeloid leukemia (AML) bone marrow (BM) at diagnosis. We compared the immunologic landscape of formalin-fixed paraffin-embedded BM trephine samples from AML (n = 69), chronic myeloid leukemia (CML; n = 56), and B-cell acute lymphoblastic leukemia (B-ALL) patients (n = 52) at diagnosis to controls (n = 12) with 30 immunophenotype markers using multiplex immunohistochemistry and computerized image analysis. We identified distinct immunologic profiles specific for leukemia subtypes and controls enabling accurate classification of AML (area under the curve [AUC] = 1.0), CML (AUC = 0.99), B-ALL (AUC = 0.96), and control subjects (AUC = 1.0). Interestingly, 2 major immunologic AML clusters differing in age, T-cell receptor clonality, and survival were discovered. A low proportion of regulatory T cells and pSTAT1+cMAF- monocytes were identified as novel biomarkers of superior event-free survival in intensively treated AML patients. Moreover, we demonstrated that AML BM and peripheral blood samples are dissimilar in terms of immune cell phenotypes. To conclude, our study shows that the immunologic landscape considerably varies by leukemia subtype suggesting disease-specific immunoregulation. Furthermore, the association of the AML immune microenvironment with clinical parameters suggests a rationale for including immunologic parameters to improve disease classification or even patient risk stratification.

The human long non-coding RNA gene RMRP has pleiotropic effects and regulates cell-cycle progression at G2.

RMRP was the first non-coding nuclear RNA gene implicated in a disease. Its mutations cause cartilage-hair hypoplasia (CHH), an autosomal recessive skeletal dysplasia with growth failure, immunodeficiency, and a high risk for malignancies. This study aimed to gain further insight into the role of RNA Component of Mitochondrial RNA Processing Endoribonuclease (RMRP) in cellular physiology and disease pathogenesis. We combined transcriptome analysis with single-cell analysis using fibroblasts from CHH patients and healthy controls. To directly assess cell cycle progression, we followed CHH fibroblasts by pulse-labeling and time-lapse microscopy. Transcriptome analysis identified 35 significantly upregulated and 130 downregulated genes in CHH fibroblasts. The downregulated genes were significantly connected to the cell cycle. Multiple other pathways, involving regulation of apoptosis, bone and cartilage formation, and lymphocyte function, were also affected, as well as PI3K-Akt signaling. Cell-cycle studies indicated that the CHH cells were delayed specifically in the passage from G2 phase to mitosis. Our findings expand the mechanistic understanding of CHH, indicate possible pathways for therapeutic intervention and add to the limited understanding of the functions of RMRP.

Dominant TOM1 mutation associated with combined immunodeficiency and autoimmune disease.

Mutations in several proteins functioning as endolysosomal components cause monogenic autoimmune diseases, of which pathogenesis is linked to increased endoplasmic reticulum stress, inefficient autophagy, and defective recycling of immune receptors. We report here a heterozygous TOM1 p.G307D missense mutation, detected by whole-exome sequencing, in two related patients presenting with early-onset autoimmunity, antibody deficiency, and features of combined immunodeficiency. The index patient suffered from recurrent respiratory tract infections and oligoarthritis since early teens, and later developed persistent low-copy EBV-viremia, as well as an antibody deficiency. Her infant son developed hypogammaglobulinemia, autoimmune enteropathy, interstitial lung disease, profound growth failure, and treatment-resistant psoriasis vulgaris. Consistent with previous knowledge on TOM1 protein function, we detected impaired autophagy and enhanced susceptibility to apoptosis in patient-derived cells. In addition, we noted diminished STAT and ERK1/2 signaling in patient fibroblasts, as well as poor IFN-γ and IL-17 secretion in T cells. The mutant TOM1 failed to interact with TOLLIP, a protein required for IL-1 recycling, PAMP signaling and autophagosome maturation, further strengthening the link between the candidate mutation and patient pathophysiology. In sum, we report here an identification of a novel gene, TOM1, associating with early-onset autoimmunity, antibody deficiency, and features of combined immunodeficiency. Other patient cases from unrelated families are needed to firmly establish a causal relationship between the genotype and the phenotype.

Gain-of-function CEBPE mutation causes noncanonical autoinflammatory inflammasomopathy.

BACKGROUND: CCAAT enhancer-binding protein epsilon (C/EBPε) is a transcription factor involved in late myeloid lineage differentiation and cellular function. The only previously known disorder linked to C/EBPε is autosomal recessive neutrophil-specific granule deficiency leading to severely impaired neutrophil function and early mortality. OBJECTIVE: The aim of this study was to molecularly characterize the effects of C/EBPε transcription factor Arg219His mutation identified in a Finnish family with previously genetically uncharacterized autoinflammatory and immunodeficiency syndrome. METHODS: Genetic analysis, proteomics, genome-wide transcriptional profiling by means of RNA-sequencing, chromatin immunoprecipitation (ChIP) sequencing, and assessment of the inflammasome function of primary macrophages were performed. RESULTS: Studies revealed a novel mechanism of genome-wide gain-of-function that dysregulated transcription of 464 genes. Mechanisms involved dysregulated noncanonical inflammasome activation caused by decreased association with transcriptional repressors, leading to increased chromatin occupancy and considerable changes in transcriptional activity, including increased expression of NLR family, pyrin domain-containing 3 protein (NLRP3) and constitutively expressed caspase-5 in macrophages. CONCLUSION: We describe a novel autoinflammatory disease with defective neutrophil function caused by a homozygous Arg219His mutation in the transcription factor C/EBPε. Mutated C/EBPε acts as a regulator of both the inflammasome and interferome, and the Arg219His mutation causes the first human monogenic neomorphic and noncanonical inflammasomopathy/immunodeficiency. The mechanism, including widely dysregulated transcription, is likely not unique for C/EBPε. Similar multiomics approaches should also be used in studying other transcription factor-associated diseases.

Breast cancer outcome prediction with tumour tissue images and machine learning.

PURPOSE: Recent advances in machine learning have enabled better understanding of large and complex visual data. Here, we aim to investigate patient outcome prediction with a machine learning method using only an image of tumour sample as an input. METHODS: Utilising tissue microarray (TMA) samples obtained from the primary tumour of patients (N = 1299) within a nationwide breast cancer series with long-term-follow-up, we train and validate a machine learning method for patient outcome prediction. The prediction is performed by classifying samples into low or high digital risk score (DRS) groups. The outcome classifier is trained using sample images of 868 patients and evaluated and compared with human expert classification in a test set of 431 patients. RESULTS: In univariate survival analysis, the DRS classification resulted in a hazard ratio of 2.10 (95% CI 1.33-3.32, p = 0.001) for breast cancer-specific survival. The DRS classification remained as an independent predictor of breast cancer-specific survival in a multivariate Cox model with a hazard ratio of 2.04 (95% CI 1.20-3.44, p = 0.007). The accuracy (C-index) of the DRS grouping was 0.60 (95% CI 0.55-0.65), as compared to 0.58 (95% CI 0.53-0.63) for human expert predictions based on the same TMA samples. CONCLUSIONS: Our findings demonstrate the feasibility of learning prognostic signals in tumour tissue images without domain knowledge. Although further validation is needed, our study suggests that machine learning algorithms can extract prognostically relevant information from tumour histology complementing the currently used prognostic factors in breast cancer.

Publisher Correction: Pharmacological reactivation of MYC-dependent apoptosis induces susceptibility to anti-PD-1 immunotherapy.

The original version of this Article contained an error in Fig. 7. In panel b, the survival curves were shifted relative to the y axis. This error has been corrected in both the PDF and HTML versions of the Article.

Pharmacological reactivation of MYC-dependent apoptosis induces susceptibility to anti-PD-1 immunotherapy.

Elevated MYC expression sensitizes tumor cells to apoptosis but the therapeutic potential of this mechanism remains unclear. We find, in a model of MYC-driven breast cancer, that pharmacological activation of AMPK strongly synergizes with BCL-2/BCL-XL inhibitors to activate apoptosis. We demonstrate the translational potential of an AMPK and BCL-2/BCL-XL co-targeting strategy in ex vivo and in vivo models of MYC-high breast cancer. Metformin combined with navitoclax or venetoclax efficiently inhibited tumor growth, conferred survival benefits and induced tumor infiltration by immune cells. However, withdrawal of the drugs allowed tumor re-growth with presentation of PD-1+/CD8+ T cell infiltrates, suggesting immune escape. A two-step treatment regimen, beginning with neoadjuvant metformin+venetoclax to induce apoptosis and followed by adjuvant metformin+venetoclax+anti-PD-1 treatment to overcome immune escape, led to durable antitumor responses even after drug withdrawal. We demonstrate that pharmacological reactivation of MYC-dependent apoptosis is a powerful antitumor strategy involving both tumor cell depletion and immunosurveillance.

Immune cell constitution in bone marrow microenvironment predicts outcome in adult ALL.

As novel immunological treatments are gaining a foothold in the treatment of acute lymphoblastic leukemia (ALL), it is elemental to examine ALL immunobiology in more detail. We used multiplexed immunohistochemistry (mIHC) to study the immune contexture in adult precursor B cell ALL bone marrow (BM). In addition, we developed a multivariate risk prediction model that stratified a poor survival group based on clinical parameters and mIHC data. We analyzed BM biopsy samples of ALL patients (n = 52) and healthy controls (n = 14) using mIHC with 30 different immunophenotype markers and computerized image analysis. In ALL BM, the proportions of M1-like macrophages, granzyme B+CD57+CD8+ T cells, and CD27+ T cells were decreased, whereas the proportions of myeloid-derived suppressor cells and M2-like macrophages were increased. Also, the expression of checkpoint molecules PD1 and CTLA4 was elevated. In the multivariate model, age, platelet count, and the proportion of PD1+TIM3+ double-positive CD4+ T cells differentiated a poor survival group. These results were validated by flow cytometry in a separate cohort (n = 31). In conclusion, the immune cell contexture in ALL BM differs from healthy controls. CD4+PD1+TIM3+ T cells were independent predictors of poor outcome in our multivariate risk model, suggesting that PD1 might serve as an attractive immuno-oncological target in B-ALL.

Immune cell contexture in the bone marrow tumor microenvironment impacts therapy response in CML.

Increasing evidence suggests that the immune system affects prognosis of chronic myeloid leukemia (CML), but the detailed immunological composition of the leukemia bone marrow (BM) microenvironment is unknown. We aimed to characterize the immune landscape of the CML BM and predict the current treatment goal of tyrosine kinase inhibitor (TKI) therapy, molecular remission 4.0 (MR4.0). Using multiplex immunohistochemistry (mIHC) and automated image analysis, we studied BM tissues of CML patients (n = 56) and controls (n = 14) with a total of 30 immunophenotype markers essential in cancer immunology. CML patients' CD4+ and CD8+ T-cells expressed higher levels of putative exhaustion markers PD1, TIM3, and CTLA4 when compared to control. PD1 expression was higher in BM compared to paired peripheral blood (PB) samples, and decreased during TKI therapy. By combining clinical parameters and immune profiles, low CD4+ T-cell proportion, high proportion of PD1+TIM3-CD8+ T cells, and high PB neutrophil count were most predictive of lower MR4.0 likelihood. Low CD4+ T-cell proportion and high PB neutrophil counts predicted MR4.0 also in a validation cohort (n = 52) analyzed with flow cytometry. In summary, the CML BM is characterized by immune suppression and immune biomarkers predicted MR4.0, thus warranting further testing of immunomodulatory drugs in CML treatment.

Aggressive natural killer-cell leukemia mutational landscape and drug profiling highlight JAK-STAT signaling as therapeutic target.

Aggressive natural killer-cell (NK-cell) leukemia (ANKL) is an extremely aggressive malignancy with dismal prognosis and lack of targeted therapies. Here, we elucidate the molecular pathogenesis of ANKL using a combination of genomic and drug sensitivity profiling. We study 14 ANKL patients using whole-exome sequencing (WES) and identify mutations in STAT3 (21%) and RAS-MAPK pathway genes (21%) as well as in DDX3X (29%) and epigenetic modifiers (50%). Additional alterations include JAK-STAT copy gains and tyrosine phosphatase mutations, which we show recurrent also in extranodal NK/T-cell lymphoma, nasal type (NKTCL) through integration of public genomic data. Drug sensitivity profiling further demonstrates the role of the JAK-STAT pathway in the pathogenesis of NK-cell malignancies, identifying NK cells to be highly sensitive to JAK and BCL2 inhibition compared to other hematopoietic cell lineages. Our results provide insight into ANKL genetics and a framework for application of targeted therapies in NK-cell malignancies.

Deep learning based tissue analysis predicts outcome in colorectal cancer.

Image-based machine learning and deep learning in particular has recently shown expert-level accuracy in medical image classification. In this study, we combine convolutional and recurrent architectures to train a deep network to predict colorectal cancer outcome based on images of tumour tissue samples. The novelty of our approach is that we directly predict patient outcome, without any intermediate tissue classification. We evaluate a set of digitized haematoxylin-eosin-stained tumour tissue microarray (TMA) samples from 420 colorectal cancer patients with clinicopathological and outcome data available. The results show that deep learning-based outcome prediction with only small tissue areas as input outperforms (hazard ratio 2.3; CI 95% 1.79-3.03; AUC 0.69) visual histological assessment performed by human experts on both TMA spot (HR 1.67; CI 95% 1.28-2.19; AUC 0.58) and whole-slide level (HR 1.65; CI 95% 1.30-2.15; AUC 0.57) in the stratification into low- and high-risk patients. Our results suggest that state-of-the-art deep learning techniques can extract more prognostic information from the tissue morphology of colorectal cancer than an experienced human observer.