Patient sample collection

Twenty six fresh frozen primary tumour and matched normal tissues were obtained from the tissue repository of Tata Memorial hospital (TMH). The Institutional Review Board (IRB) and the Ethics Committee (EC) of Tata Memorial Centre (TMC) - Advanced Centre for Treatment, Research and Education in Cancer (ACTREC) (Mumbai, India) approved the project (#104). Since this was a retrospective analysis, the IRB and the EC waived the need for an informed consent. Patients were randomly selected based on the availability of fresh frozen tissues. The patient characteristics including age, gender, gall stone status and histopathology were recorded.

PCR analysis for Salmonella isolates

The PCR method used for Salmonella detection has been previously described [12]. Nested PCR was carried out in a 25 μl volume containing 10 μl KAPA 2X ready mix master-mix (KapaBiosystems catalog no-KK1024), 10 pmol primer and 100 ng of genomic DNA. Following the first round of PCR (94 °C for 1 min, 55 °C for 1 min 15 s, 72 °C for 3 min – 40 cycles) with ST1 and ST2 primers, 5 μl PCR product was used as template for nested PCR using ST3 and ST4 primers (94 °C for 1 min, 68 °C for 1 min 15 s, 72 °C for 3 min – 40 cycles). We also performed validation of Salmonella sequences using read specific primers. The PCR conditions - 94 °C for 1 min, 59 °C for 30s, 72 °C for 45 s – for 30 cycles. The primer sequences are detailed in Additional file 1: Table S2.

Sequencing and analysis

Exome capture and library preparation was performed using Agilent Sure select in-solution (low-input capture-500 ng) target enrichment technology. Genomic DNA was sheared and size selected (150–200 bp) and ligated to adaptors and run on Illumina Nextseq 500 platform to generate 150 bp paired end reads at a coverage of 100X and above. To detect Salmonella traces, the HPVDetector pipeline was used, as described previously [11]. Briefly, reads were aligned against six known Salmonella species genomes in addition of the HPVDetector nce data set of 143 HPV types, as downloaded from the National Centre for Biotechnology Information (NCBI), using BWA (Burrows wheeler algorithm) aligner (v0.6.2). All reference sequences were annotated and concatenated to compose a multi-fasta sequences using bio-perl modules. The alignment files were parsed using UNIX shell program to detect the types of Salmonella represented by at least one read that aligned to a particular Salmonella type with high confidence.

HPVDetector pipeline identifies Salmonella sequence present in gallbladder cancer samples

We performed PCR based analysis of 26 gallbladder tumour and paired normal samples to detect the presence of Salmonella DNA using pan primers, as described earlier [12]. None of the gallbladder samples were found to be positive for Salmonella (data not shown). As a next step, whole exome data for these 26 samples (generated in house, manuscript in make) were analysed to detect Salmonella traces using HPVDetector pipeline, modified to include additional genome sequence of 6 common Salmonella isolates. The computational approach, in brief, subtracts all reads that align to human genome and aligns remaining reads to HPV and Salmonella reference database from NCBI. While HPV16 was detected in 1 gallbladder sample, Salmonella isolates were found across multiple samples: S. typhi Ty2 (3 samples), S. typhi CT18 (6 samples), S. typhimurium LT2 (10 samples), S. choleraesuis SCB67 (5 samples), S. paratyphi TCC (3 samples), and S. paratyphi SPB7 (7 samples). In total, Salmonella reads were found in 19 of 26 gallbladder tissues (tumor as well as adjacent normal tissues). Interestingly, 10 of 19 samples were co-infected with multiple isolates independent of gallstone status or gender of the patients, as shown in Fig. 1.

Annotation of the Salmonella reads found in gallbladder cancer samples

Variable number of overlapping reads of variant lengths for each isolate were assembled into contigs based on Clustal X2 multiple alignment as shown in Additional file 2: Figure S1. The unique stretch of contigs generated were annotated based on gene annotation database of Salmonella isolates from NCBI (National centre for biotechnology information) database. 114 reads of multiple Salmonella isolates were found in 19 of 26 samples analyzed. 47 of 114 reads of Salmonella ORF (open reading frame) were identified as encoding for bacterium genes known to be involved in metabolism and those related to the toxin-antitoxin system. Rest of the reads aligned to the Salmonella ribosomal genes, understandably due to their relatively higher abundance (Additional file 3: Table S1).

HPVDetector pipeline is specific and highly sensitive to detect true Salmonella traces

To assess specificity of our assay, we re-analyzed whole exome data of all samples by taking their reverse (not complement) to simulate random sequence, but retaining composition of nucleotides and genome complexity, using an in-house perl script, as described earlier [11]. We found no spurious Salmonella reads when the primary tumour whole exome sequence was reversed in any of the 26 samples, suggesting the computational pipeline used was specific to detect Salmonella traces. (Additional file 4: Figure S2A). To test the sensitivity of our assay, raw FASTQ file of a primary tumour sample 16 T that was found positive for Salmonella reads was down-sampled to 1X, 5X, 10X, 15X, 25X, 50X, 75X and 100X coverage using Picard Toolkit’s DownsampleSam function (http://broadinstitute.github.io/picard/), as described earlier [11]. The resulting FASTQ files were analysed for detection of Salmonella reads using the HPVDetector pipeline. Distinct Salmonella reads were detected at as low as 10X whole exome coverage that increased linearly (Additional file 4: Figure S2B).

Sanger validation of Salmonella reads identified in gallbladder cancer samples

We have attempted to validate the presence of Salmonella read sequences identified by HPVDetector in 4 of 16 Salmonella positive samples using Sanger sequencing (Additional file 5: Figure S3).