For virtually every patient with colorectal cancer (CRC), hematoxylin–eosin (HE)–stained tissue slides are available. These images contain quantitative information, which is not routinely used to objectively extract prognostic biomarkers. In the present study, we investigated whether deep convolutional neural networks (CNNs) can extract prognosticators directly from these widely available images. We hand-delineated single-tissue regions in 86 CRC tissue slides, yielding more than 100,000 HE image patches, and used these to train a CNN by transfer learning, reaching a nine-class accuracy of >94% in an independent data set of 7,180 images from 25 CRC patients. With this tool, we performed automated tissue decomposition of representative multitissue HE images from 862 HE slides in 500 stage I–IV CRC patients in the The Cancer Genome Atlas (TCGA) cohort, a large international multicenter collection of CRC tissue. Based on the output neuron activations in the CNN, we calculated a “deep stroma score,” which was an independent prognostic factor for overall survival (OS) in a multivariable Cox proportional hazard model (hazard ratio [HR] with 95% confidence interval [CI]: 1.99 [1.27–3.12], p = 0.0028), while in the same cohort, manual quantification of stromal areas and a gene expression signature of cancer-associated fibroblasts (CAFs) were only prognostic in specific tumor stages. We validated these findings in an independent cohort of 409 stage I–IV CRC patients from the “Darmkrebs: Chancen der Verhütung durch Screening” (DACHS) study who were recruited between 2003 and 2007 in multiple institutions in Germany. Again, the score was an independent prognostic factor for OS (HR 1.63 [1.14–2.33], p = 0.008), CRC-specific OS (HR 2.29 [1.5–3.48], p = 0.0004), and relapse-free survival (RFS; HR 1.92 [1.34–2.76], p = 0.0004). A prospective validation is required before this biomarker can be implemented in clinical workflows. In our retrospective study, we show that a CNN can assess the human tumor microenvironment and predict prognosis directly from histopathological images.

PURPOSE: This trial evaluates the addition of the PD-L1 antibody atezolizumab (ATZ) to standard-of-care fluorouracil, leucovorin, oxaliplatin, and docetaxel (FLOT) as a perioperative treatment for patients with resectable esophagogastric adenocarcinoma (EGA). METHODS: DANTE started as multicenter, randomized phase II trial, which was subsequently converted to a phase III trial. Here, we present the results of the phase II proportion, focusing on surgical pathology and safety outcomes on an exploratory basis. Patients with resectable EGA (≥cT2 or cN+) were assigned to either four preoperative and postoperative cycles of FLOT combined with ATZ, followed by eight cycles of ATZ maintenance (arm A) or FLOT alone (arm B). RESULTS: Two hundred ninety-five patients were randomly assigned (A, 146; B, 149) with balanced baseline characteristics between arms. Twenty-three patients (8%) had tumors with microsatellite instability (MSI), and 58% patients had tumors with a PD-L1 combined positive score (CPS) of ≥1. Surgical morbidity (A, 45%; B, 42%) and 60-day mortality (A, 3%; B, 2%) were comparable between arms. Downstaging favored arm A versus arm B (ypT0, 23% v 15% [one-sided P = .044]; ypT0-T2, 61% v 48% [one-sided P = .015]; ypN0, 68% v 54% [one-sided P = .012]). Histopathologic complete regression rates (pathologic complete response or TRG1a) were higher after FLOT plus ATZ (A, 24%; B, 15%; one-sided P = .032), and the difference was more pronounced in the PD-L1 CPS ≥10 (A, 33%; B, 12%) and MSI (A, 63%; B, 27%) subpopulations. Complete margin-free (R0) resection rates were relatively high in both arms (A, 96%; B, 95%). The incidence and severity of adverse events were similar in both groups. CONCLUSION: Within the limitations of the exploratory nature of the data, the addition of ATZ to perioperative FLOT is safe and improved postoperative stage and histopathologic regression.

Like normal colorectal epithelium, colorectal carcinomas (CRCs) are organized hierarchically and include populations of cells with stem-like properties. Leucine-rich-repeat-containing G-protein-coupled receptor 5 (LGR5) is associated with these stem cells in normal colorectal epithelium; however, the precise function of LGR5 in CRC remains largely unknown. Here, we analyzed the functional and molecular consequences of short hairpin RNA-mediated silencing of LGR5 in CRC cell lines SW480 and HT-29. Additionally, we exposed Lgr5-EGFP-IRES-CreERT2 mice to azoxymethane/dextrane sodium sulfate (AOM/DSS), which induces inflammation-driven colon tumors. Tumors were then flow-sorted into fractions of epithelial cells that expressed high or low levels of Lgr5 and were molecularly characterized using gene expression profiling and array comparative genomic hybridization. Silencing of LGR5 in SW480 CRC cells resulted in a depletion of spheres but did not affect adherently growing cells. Spheres expressed higher levels of several stem cell-associated genes than adherent cells, including LGR5. Silencing of LGR5 reduced proliferation, migration and colony formation in vitro and tumorigenicity in vivo. In accordance with these results, NOTCH signaling was downregulated upon LGR5 silencing. In AOM/DSS-induced colon tumors, Lgr5 high cells showed higher levels of several stem cell-associated genes and higher Wnt signaling than Lgr5 low tumor cells and Lgr5 high normal colon cells. Array comparative genomic hybridization revealed no genomic imbalances in either tumor cell fraction. Our data elucidate mechanisms that define the role of LGR5 as a marker for stem-like cells in CRC.

Genetic changes during tumorigenesis are usually acquired sequentially. However, a recent study showed that in 2% to 3% of all cancers a single catastrophic event, termed chromothripsis, can lead to massive genomic rearrangements confined to one or a few chromosomes. To explore whether the degree of genomic instability and chromothripsis influences prognosis in cancer, we retrospectively applied array-comparative genomic hybridization (aCGH) to 20 malignant melanomas that showed, despite comparable conventional clinical and pathologic parameters, a profoundly different clinical course. We compared 10 patients who died of malignant melanoma 3.7 years (median, range 0.9–7.6 years) after diagnosis with 10 patients who survived malignant melanoma and had a median disease-free survival of 14.8 years (range 12.5–16.7 years; P = 0.00001). We observed a striking association between the degree of chromosomal instability, both numerical and structural, and outcome. Malignant melanomas associated with good prognosis showed only few chromosomal imbalances (mean 1.6 alterations per case), predominantly whole chromosome or chromosome arm gains and losses, whereas malignant melanomas with poor prognosis harbored significantly more chromosomal aberrations (13.9 per case; P = 0.008). Array-based CGH showed that these aberrations were mostly focal events, culminating in two cases in a pattern consistent with the phenomenon of chromothripsis, which was confirmed by paired-end sequencing. This is the first description of chromothripsis in primary malignant melanomas. Our study therefore links focal copy number alterations and chromothripsis with poor outcome in patients with malignant melanomas (P = 0.0002) and provides a genetic approach to predict outcome in malignant melanomas. Cancer Res; 73(5); 1454–60. ©2012 AACR.

PURPOSE: High pathologic complete response (pCR) rates and comparably good survival data were seen in a phase II trial combining perioperative fluorouracil, leucovorin, oxaliplatin, and docetaxel (FLOT) chemotherapy with trastuzumab for resectable, esophagogastric adenocarcinoma (EGA). The current trial evaluates the addition of trastuzumab and pertuzumab to FLOT as perioperative treatment for human epidermal growth factor receptor 2–positive resectable EGA. METHODS: In this multicenter, randomized phase II/III trial, patients with human epidermal growth factor receptor 2–positive, resectable EGA (≥ clinical tumor 2 or clinical nodal–positive) were assigned to four pre- and postoperative cycles of either FLOT alone (arm A) or combined with trastuzumab and pertuzumab, followed by nine cycles of trastuzumab/pertuzumab (arm B). The primary end point for the phase II part was the rate of pCR. RESULTS: The trial was closed prematurely, without transition into phase III, after results of the JACOB trial were reported. Eighty-one patients were randomly assigned (A: 41/B: 40) during the phase II part. The pCR rate was significantly improved with the trastuzumab/pertuzumab treatment (A: 12%/B: 35%; P = .02). Similarly, the rate of pathologic lymph node negativity was higher with trastuzumab/pertuzumab (A: 39%/B: 68%), whereas the R0 resection rate (A: 90%/B: 93%) and surgical morbidity (A: 43%/B: 44%) were comparable. Moreover, the inhouse mortality was equal in both arms (overall 2.5%). The median disease-free survival was 26 months in arm A and not yet reached in arm B (hazard ratio, 0.58; P = .14). After a median follow-up of 22 months, the median overall survival was not yet reached (hazard ratio, 0.56; P = .24). Disease-free survival and overall survival rates at 24 months were 54% (95% CI, 38 to 71) and 77% (95% CI, 63 to 90) in arm A and 70% (95% CI, 55 to 85) and 84% (95% CI, 72 to 96) in arm B, respectively. More ≥ grade 3 adverse events were reported with trastuzumab/pertuzumab, especially diarrhea (A: 5%/B: 41%) and leukopenia (A: 13%/B: 23%). CONCLUSION: The addition of trastuzumab/pertuzumab to perioperative FLOT significantly improved pCR and nodal negativity rates at the price of higher rates of diarrhea and leukopenia.