Search

Additional file 1: Supplementary Table 1. Included patients and samples. Supplementary Figure 2. Validation of pipeline and analytical approach. A, mean RPPB detected for indicated species in blood-derived and lymphoblastoid cell line 1000 Genome samples sorted by mean of blood-derived samples. B, proportion of non-human read pairs matching the indicated taxon of read pairs matching any species-level taxon for each external validation sample. C, comparison of Kraken matched read pairs in RNA-seq and WGS data of the same sample. Each dot represents one species-level taxon in one sample with both RNA-seq and WGS data available. The line represents the best-fitted line (log-log linear regression). Pearson correlation coefficients (log-log) are shown with two-sided p-values. D, plot of Pearson correlation coefficient (log-log) distribution of all samples with both RNA-seq and WGS data available. Each dot represents the Pearson correlation coefficient within a single sample. E, comparison of 10% subsample and full dataset. Each dot represents one species-level taxon in one sample for which both the full and the subsampled dataset has been analyzed and indicates the absolute read count identified in both samples. The line represents the best-fitted line (log-log linear regression). Pearson correlation coefficients (log-log) are shown with two-sided p-values. F, Ratio of absolute read counts in the full sample to the 10% subsample for 4 selected taxa. The mean ratio of all samples in which the respective taxon was detected is indicated by the symbol and the error bars indicate the standard error of the mean. The dotted line shows the expected ratio of 10. G, comparison of tumor tissue and matched normal by patient and taxon. Each dot represents one species-level taxon in one patient with both tumor tissue and matched normal analyzed and indicates the RPPB in both samples. The line represents the best-fitted line (log-log linear regression). Pearson correlation coefficients (log-log) are shown with two-sided p-values. Supplementary Figure 3. Alpha diversity. A, counts of 1000 Genome samples by species-level richness. B, counts of tumor tissue samples by species-level richness color-coded by project. C, counts of matched normal samples by species-level richness color-coded by project. D, comparison of richness between projects and sample type. Bars show mean and error bars standard deviation. N in brackets indicates total sample number for each project. Supplementary Figure 4. Coverage distribution for all tumor-linked species-level taxa. Coverage distribution across each species-level taxon identified as tumor-linked. Supplementary Figure 5. Flow chart of taxa filtering strategy. Flow chart of filtering strategy to derive likely tumor-linked species-level taxa. Supplementary Figure 6. Heatmap of filtered taxons. Log2-transformed RPPB of all species-level taxa identified as likely tumor-linked after filtering in all samples. Taxa were hierarchically clustered using Pearson correlation as a distance measure with average-linkage. Samples were hierarchically clustered within each project and type subgroup using Pearson correlation as a distance measure with average-linkage. Supplementary Figure 7. Heatmaps of tumor-linked taxa for all cancers without discernible clusters. A-D, log2-transformed RPPB of all species-level taxa identified as tumor-linked and detected after filtering in all tumor-tissues of the indicated cancer. Results of k-means clustering of samples are shown. Supplementary Figure 8. Host integration. A, integration rate by species for tumor tissue and matched normal sample. B, difference in integration rates between bacterial and viral taxa (p < 0.0001, Wilcoxon rank-sum test, two-tailed). The midline of the boxplot shows the median, the box borders show upper and lower quartile, the whiskers show 5th and 95th percentiles and the dots outliers of species-specific integration rates in tumor tissue or matched normal samples. C, difference in integration rate between tumor tissue and matched normal samples (p = 0.0009, Wilcoxon signed rank test, two-tailed). The midline of the boxplot shows the median, the box borders show upper and lower quartile, the whiskers show 5th and 95th percentiles and the dots outliers of species-specific integration rates. Supplementary Figure 9. “Elbow method” to determine k for k-means clustering. A-J, plot of reducing within group sum of squares for increasing k (number of clusters) in k-means clustering (Supplementary Figures 4 and 5) of log2-transformed RPPB of all species-level taxa identified as tumor-linked and detected after filtering in all tumor-tissues for each indicated cancer.

Additional file 3:. Supplementary note.

Background Host tissue infections by bacteria and viruses can cause cancer. Known viral carcinogenic mechanisms are disruption of the host genome via genomic integration and expression of oncogenic viral proteins. An important bacterial carcinogenic mechanism is chronic inflammation. Massively parallel sequencing now routinely generates datasets large enough to contain detectable traces of bacterial and viral nucleic acids of taxa that colonize the examined tissue or are integrated into the host genome. However, this hidden resource has not been comprehensively studied in large patient cohorts. Methods In the present study, 3025 whole genome sequencing datasets and, where available, corresponding RNA-seq datasets are leveraged to gain insight into novel links between viruses, bacteria, and cancer. Datasets were obtained from multiple International Cancer Genome Consortium studies, with additional controls added from the 1000 genome project. A customized pipeline based on KRAKEN was developed and validated to identify bacterial and viral sequences in the datasets. Raw results were stringently filtered to reduce false positives and remove likely contaminants. Results The resulting map confirms known links and expands current knowledge by identifying novel associations. Moreover, the detection of certain bacteria or viruses is associated with profound differences in patient and tumor phenotypes, such as patient age, tumor stage, survival, and somatic mutations in cancer genes or gene expression profiles. Conclusions Overall, these results provide a detailed, unprecedented map of links between viruses, bacteria, and cancer that can serve as a reference for future studies and further experimental validation.

Purpose/Objectives: SCLC accounts for approximately 15% of newly diagnosed lung cancer cases, is driven by RB1 and TP53 lesions and represents the most aggressive pulmonary carcinoma. Current chemotherapies are initially effective in SCLC patients, however tumors rapidly reoccur and patients die within a few months. Results from KEYNOTE-028 revealed that PD-L1 positive late stage SCLC patients who received a human-ized monoclonal anti-PD-1 antibody, showed 25% response rates. This finding is remarkable for SCLC, since targeted therapies for this tumor entity are rare. However, most SCLC patients harbor a primary resistance or acquire resistance during treatment by an immune suppressive microenvironment. Thus, there is a critical need to combine immune checkpoint inhibitors with other therapies to overcome these resistances. Materials/Methods: We implemented a combined therapy concept including anti-VEGF-A and anti-PD-L1 monoclonal antibodies in an autochthonous SCLC mouse model. As a read out, we used X-ray computed tomography, flow cytometry and end-point immunohistochemistry. Results: We found that combined anti-VEGF-A/anti-PD-L1 treatment synergistically improved overall survival, compared to monotherapies, and enhanced T cell infiltration into the tumor. Moreover, we identified in murine SCLC tumors which acquired resistance to anti-PD-L1, a significant increase in PD-1+/TIM-3+ exhausted CD4 and CD8 T-cells accompanied by significantly increased expression of Galectin-9, the major TIM-3 ligand, on TAMs. In addition, we observed a significantly enriched fraction of TAMs with a similar exhausted phenotype indicated by the expression of LAG-3 and CTLA-4 upon acquired resistance to PD-L1 blockade. Importantly, the exhaustion of T-cells and TAMs was reverted upon combined anti-VEGF-A/anti-PD-L1 therapy. In line with the results obtained from the mouse model, we showed mechanistically in PBMCs of SCLC patients, that VEGF-A triggers co-expression of PD-1 and TIM-3 on T-cells, indicating an immunosuppressive function of VEGF-A in SCLC patients during anti-PD-1/PD-L1 therapy. Furthermore, we elucidated using CRISPR-Cas9 technology, that VEGF-A liberated from the tumor microenvironment is the likely mediator of acquired resistance to PD1/PD-L1 blockade in SCLC. Conclusion: Taken together, in SCLC there is evidence for a substantial benefit of implementing combined anti-angiogenic and anti-immune checkpoint therapy approaches in the clinic in order to overcome acquired resistances delivered by T-cells and macrophages in the tumor microenvironment. Remarks: The authors declare no conflicts of interest. This work was published in Cancer Res. 2018 Aug 1;78(15):4270-4281. doi: 10.1158/0008-5472.CAN-17-2176. Epub 2018 May 18 with the title 'Combined VEGF and PD-L1 Blockade Displays Synergistic Treatment Effects in an Autochthonous Mouse Model of Small Cell Lung Cancer'., Proceedings on Cancer-Immuno-Oncology, Vol 1 No 1 (2018)