Next-Generation Sequencing of Prostate Tumors Provides Independent Evidence of Xenotropic Murine Leukemia Virus-Related Gammaretrovirus Contamination

LETTER

In 2006, xenotropic murine leukemia virus (MLV)-related gammaretrovirus (XMRV) was isolated from prostate cancer tissue (9). However, subsequent studies have yielded conflicting and controversial results, with widespread detection of XMRV suggested to be the result of contamination with mouse DNA (3). While recent work has focused on PCR-based targeted detection of XMRV (10), the advent of next-generation sequencing (NGS) means that it is now possible to interrogate the entire genomes and transcriptomes of human samples for the unique genomic signatures of thousands of viruses (1, 5, 6).

With this in mind, we analyzed whole-genome (DNA-Seq) and transcriptome (RNA-Seq) data from 9 human prostate tumors (6 primary and 3 metastatic), 3 prostate tumor-derived murine xenografts, and 1 benign tissue sample from a pelvic lymph node. The xenograft tumors carried significant amounts of host mouse tissue, thereby acting as positive controls. We mapped all nonhuman RNA-Seq reads to a custom database of 3,932 viral genomes and 1,387 microbial genomes downloaded from the National Center for Biotechnology Information RefSeq database (July 2011) (7) and filtered for reads mapping specifically to MLVs, the family that includes XMRV. Murine genomes are hosts to hundreds of retroviruses (8); consistent with this, the xenograft tumor transcriptomes yielded thousands of MLV reads. Interestingly, 2 primary tumors (P3 and P5) also demonstrated significant enrichment of 5 MLVs, especially XMRV, in their transcriptomes, compared to the other patient tumors (P = 1.28 × 10⁻⁸ [t test comparing XMRV reads per million]) (Table 1).

Table 1.

Analysis of RNA-Seq and DNA-Seq reads from human prostate tumors mapping to viral genomes^a

Sample	RNA-Seq					DNA-Seq
	No. of nonhuman reads (× million)	No. of reads per million				No. of nonhuman reads (× million)	No. of reads per million
		RefSeq viral genomes	MLVs	XMRVs	Mouse-specific sequences		RefSeq viral genomes	MLVs	XMRVs
P1	26.5	17.86	4.15	0	0	16.2	21.77	0	0
P2	28.6	46.61	3.5	1.11	1.07
P3	31.8	118.54	22.46	11.25	4.99	9.3	15.53	0	0
P4	22.9	8.77	4.71	0.04	0.73
P5	23.5	111.43	21.19	10.97	6.33	36.3	2.56	0	0
P6	16.5	99.01	0.84	0	0.36
M1b	25.8	40.03	4.78	0.87	0.58
M2	29	22.91	4.33	1.47	1.23
M3	15.7	22.58	0.63	0	0.38
M4	15.5	18.04	3.01	0	0
X1	29.9	2,187.91	327.27	169.45	1,251.12	53.3	30.82	10.07	4.11
X2	71.3	1,644.88	421.93	192.94	162.75	59.6	22.32	5.97	2.38
X3	21.3	2,990.69	443.02	220.13	65.92

^a“Nonhuman reads” refers to reads that could not be mapped to the human genome. P, primary prostate tumor; M, lymph node metastasis; X, human-derived xenograft tumor. Note that metastatic samples M1 to M4 were isolated from the same patients as tumor samples P1 to P4.

^bAdjacent benign tissue was sampled.

To search for evidence of viral integration into the tumor genomes of P3 and P5, we mapped the nonhuman DNA-Seq reads to the same viral database. As expected, the xenograft samples showed enrichment of MLV DNA sequences. However, no DNA-Seq reads from patients P3 and P5 mapped to any of the MLVs (Table 1). Note that our DNA-Seq human genome coverage averaged >4×, while XMRV is expected to integrate on every chromosome within infected genomes (4).

Finally, we interrogated RNA-Seq data for 2 mouse-specific segments of the mitochondrial cytB gene (mouse mitochondrial DNA [mtDNA]; NC_005089.1). The xenograft samples were strongly positive, but tumors P3 and P5 also demonstrated enrichment of mouse mtDNA (Table 1). Indeed, the correlation coefficient across all human samples between the numbers of reads per million mapping to XMRV mtDNA and to mouse mtDNA was 0.982 (note that there was no sequence common to the XMRV mtDNA and mouse mtDNA).

To our knowledge, this is first analysis of XMRV in NGS data and provides no evidence for XMRV infection in human prostate tumors. The 2 human samples with enrichment of XMRV reads were also enriched with mouse mtDNA, suggesting a contaminant origin for the viral reads in the samples from those patients. Interestingly, the RNAs from these two samples were extracted at similar time points distinct from those of the other tumor samples and the DNA extractions. The recent emergence of large-scale sequencing projects such as the International Cancer Genome Consortium (2) means that it should soon be possible to mine large data sets in a manner similar to that described here and facilitate the detection of XMRV in cases of prostate cancer.