HeLa Nucleic Acid Contamination in The Cancer Genome Atlas Leads to the Misidentification of Human Papillomavirus 18

Abstract

We searched The Cancer Genome Atlas (TCGA) database for viruses by comparing non-human reads present in transcriptome sequencing (RNA-Seq) and whole-exome sequencing (WXS) data to viral sequence databases. Human papillomavirus 18 (HPV18) is an etiologic agent of cervical cancer, and as expected, we found robust expression of HPV18 genes in cervical cancer samples. In agreement with previous studies, we also found HPV18 transcripts in non-cervical cancer samples, including those from the colon, rectum, and normal kidney. However, in each of these cases, HPV18 gene expression was low, and single-nucleotide variants and positions of genomic alignments matched the integrated portion of HPV18 present in HeLa cells. Chimeric reads that match a known virus-cell junction of HPV18 integrated in HeLa cells were also present in some samples. We hypothesize that HPV18 sequences in these non-cervical samples are due to nucleic acid contamination from HeLa cells. This finding highlights the problems that contamination presents in computational virus detection pipelines. IMPORTANCEViruses associated with cancer can be detected by searching tumor sequence databases. Several studies involving searches of the TCGA database have reported the presence of HPV18, a known cause of cervical cancer, in a small number of additional cancers, including those of the rectum, kidney, and colon. We have determined that the sequences related to HPV18 in non-cervical samples are due to nucleic acid contamination from HeLa cells. To our knowledge, this is the first report of the misidentification of viruses in next-generation sequencing data of tumors due to contamination with a cancer cell line. These results raise awareness of the difficulty of accurately identifying viruses in human sequence databases.

Possible Human Papillomavirus 38 Contamination of Endometrial Cancer RNA Sequencing Samples in The Cancer Genome Atlas Database

“…Computational techniques have been used to quantify viral presence in many cancers, thereby accelerating the identification of virus-disease associations (5)(6)(7). However, the increasing sensitivity of these techniques may also reveal previously unappreciated viral contaminations that could be erroneously associated with the disease (8)(9)(10).…”

Zombies in TCGA

“…Pipas and colleagues confirmed the presence of HPV18 RNA in TCGA data, including some of the individual tumors analyzed in the prior publications, but noticed that the numbers of HPV18 RNA sequence reads in these atypical tumors were much lower than those in cervical cancer specimens and, unlike the case for the cervical cancers, the HPV18 sequence reads were present in transcriptome sequencing (RNA-Seq) data only, not in whole-exome sequencing data (10). These findings raised suspicions, so they examined in more detail the HPV18 sequences reported to be present in these cancers and compared them to the well-characterized HPV18 sequences in HeLa cells.…”

“…But the problem is not restricted to HeLa cells. Pipas and colleagues cite additional examples of apparent cell line or DNA contamination in other next-generation DNA sequencing projects and have detected a low number of hepatitis B virus RNA reads in thyroid and cervical cancer sequences that contain the same SNPs as a contemporaneously sequenced liver cancer sample (10). In the cases studied, it appears that residual HeLa cell DNA contaminated the sequencing machines, but it is possible that in other cases the cells themselves or DNA preparations are contaminated.…”

“…Finally, the contamination is obviously not restricted to viral sequences; rather, the entire HeLa cell genome (or at least transcriptome) is present in the sequencing results. In fact, Cantalupo et al showed enrichment of a rare cellular SNP in the glucose-6-phosphate dehydrogenase gene in the colorectal cancer sequencing results that is characteristic of HeLa cells (10). So their findings may be regarded as the canary in the coal mine, or perhaps the zombie in the cell line, warning that similar cellular genome contamination may also occur in other sequencing projects.…”

“…The latest (but I suspect not the last) reincarnation of HeLa cells is reported in this issue of the Journal of Virology with the discovery by James Pipas and collaborators at the University of Pittsburgh that HeLa cell RNA sequences lurk in the results of large-scale sequencing efforts used to characterize cancer genomes (10).…”