Detection of Murine Leukemia Virus or Mouse DNA in Commercial RT-PCR Reagents and Human DNAs

The xenotropic murine leukemia virus (MLV)-related viruses (XMRV) have been reported in persons with prostate cancer, chronic fatigue syndrome, and less frequently in blood donors. Polytropic MLVs have also been described in persons with CFS and blood donors. However, many studies have failed to confirm these findings, raising the possibility of contamination as a source of the positive results. One PCR reagent, Platinum Taq polymerase (pol) has been reported to contain mouse DNA that produces false-positive MLV PCR results. We report here the finding of a large number of PCR reagents that have low levels of MLV sequences. We found that recombinant reverse-transcriptase (RT) enzymes from six companies derived from either MLV or avian myeloblastosis virus contained MLV pol DNA sequences but not gag or mouse DNA sequences. Sequence and phylogenetic analysis showed high relatedness to Moloney MLV, suggesting residual contamination with an RT-containing plasmid. In addition, we identified contamination with mouse DNA and a variety of MLV sequences in commercially available human DNAs from leukocytes, brain tissues, and cell lines. These results identify new sources of MLV contamination and highlight the importance of careful pre-screening of commercial specimens and diagnostic reagents to avoid false-positive MLV PCR results.


Introduction
Xenotropic murine leukemia virus (MLV)-related virus (XMRV) is a new gammaretrovirus reported in persons from the U.S. with prostate cancer or chronic fatigue syndrome (CFS), and in blood donors [1]. MLV-like sequences have also been found in specimens from persons with CFS from the US [2]. However, numerous studies using both PCR and serology, both in the U.S. and abroad, were unable to replicate these findings, stimulating much debate and discussion regarding the origin of the positive results seen in the initial studies [3,4,5,6,7,8,9,10,11,12,13,14].
MLVs are endogenous and exogenous retroviruses whose genomes are integrated into mouse chromosomal DNA and can thus be PCR-amplified along with other mouse-specific sequences in specimens contaminated with mouse DNA. Several groups have shown recently that Platinum Taq polymerase from Invitrogen, or reverse transcriptase (RT)-PCR kits containing this enzyme, contain low levels of mouse DNA that produce a positive PCR signal with diagnostic XMRV or MLV primers [1,4,15,16,17]. The contamination source in Platinum Taq from these studies was linked to carry-over mouse DNA from the mouse monoclonal antibody used to keep Taq inactive during hot-start PCR. XMRV, MLV, and murine sequences have also been found recently in Qiagen nucleic acid extraction columns [18]. High levels of infectious MLV and XMRV have also been found in human cell lines [19,20,21]. These results suggest multiple sources of potential contamination of clinical specimens from different cohorts [4,19,21,22,23,24].
We report here the identification of widespread MLV contamination of RT enzymes from six manufacturers as well as mouse DNA contamination of commercially available human cell lines and clinical specimens. Our results highlight the importance of careful pre-screening of diagnostic reagents and commercially available specimens to avoid false-positive PCR results during testing of human clinical specimens.

Contamination of commercial RT enzymes with MLV plasmid DNA
While investigating the prevalence of XMRV and MLV in persons with CFS and prostate cancer we occasionally detected low levels (,10 copies) MLV and XMRV-like protease (pro) sequences in non-template controls (NTC) or negative blood donor plasma by using a quantitative RT-PCR (qRT-PCR) assay that employs the ABI/Ambion AgPath One Step RT-PCR kit (Cat# 4387391, Austin, Tx) [3]. Results for clinical specimens in runs with false-positive negative controls were always considered invalid and were repeated. None of the pro sequences detected in the NTC and negative blood donor samples contained the signature sequence present in the positive control template engineered in our laboratory, indicating that the qPCR standard template was not the source of the pro signals (data not shown). The recombinant RT enzyme used in the pro qRT-PCR testing was included in the ABI/Ambion AgPath One Step RT-PCR kit and according to the manufacturer was derived from an expression plasmid containing the ecotropic Moloney MLV (MoMLV) RT gene. The reagents in this kit are different from those in the Invitrogen One-step RT-PCR kit or Taq enzymes previously found to be contaminated with mouse DNA [1,4,15,17] in that they do not contain mouse monoclonal antibodies to keep the Taq inactive during hot start PCR.
Given the plasmid production history of this enzyme and absence of mouse monoclonal antibodies in the reagents, we suspected that the AgPath RT-PCR kit was contaminated with trace amounts of residual MLV-like plasmid sequences. Furthermore, the enzyme was repeatedly negative for mouse DNA contamination using highly sensitive PCR tests for mitochondrial (mtDNA) and intracisternal A particle (IAP) DNA sequences that can detect attograms of mouse DNA (Table 1) [3]. To evaluate this hypothesis further, we tested multiple replicates of NTC in the pro qRT-PCR test. Low levels ( Table 1, 1-10 copies/reaction) of MLV/XMRV were found in 7 of 16 replicates in a new, previously un-opened AgPath one step RT-PCR kit. All 16 wateronly reactions without the RT enzyme were negative. The proportion of positive NTCs was similar when a larger number of NTC replicates was tested (13/32), but increased significantly when the amount of enzyme was doubled (15/16 positives). Four of five different kit lots tested positive in the pro qPCR assay at a frequency ranging from 1/16 to 7/16 replicates when using 1 ul of enzyme per reaction (Table 1). Representative pro qPCR results are shown in Fig. 1a. These results document frequent contamination of different lots of this enzyme with low levels of MLV sequences.
To confirm the contaminant was from a DNA source, we added the AgPath enzyme into a qPCR pro DNA PCR test using an AmpliTaq Gold enzyme (Cat # N8080244; ABI, Foster City, CA), previously determined to be free of any MLV/XMRV, and amplified using the same conditions used for the pro qRT-PCR test except the first step included a 9 min 95uC step to inactivate the AgPath RT. 7/16 (44%) NTC reactions tested positive ( Table 1). The AgPath enzyme was found to be negative for gag sequences (0/16), but positive for pol sequences (14/16, 88%) using two new qRT-PCR tests to generically detect MLV and XMRV (Table 1).
To investigate the breadth of this contamination in other RT enzymes and kits, we tested both MoMLV-and avian myeloblastosis virus (AMV)-derived recombinant RTs from multiple manufacturers using the pro qRT-PCR test or a nested polymerase (pol) RT-PCR test that generically detects MLV and XMRV sequences about 216-bp in length using the primers XPOLOF, XPOLOR, XPOLIF, and XPOLOR [5,9]. AMV and MLV are highly divergent retroviruses in different genera sharing only 28% pol nucleotide identity and thus the MLV-based primers and probes should not amplify AMV sequences if present. We found that two MoMLV-based RT kits from ABI (TaqMan 1-step Master Mix and TaqMan RNA to Ct kit; cat #s 43019169 and 432938, respectively, Foster City, CA) and one MoMLV-based kit from Agilent (Brilliant II QRT-PCR Master mix; cat # 600884; Santa Clara, CA) contained low copies of XMRV/MLV sequences in 4/16 (25%)-6/16 (38%) replicates, respectively (Table 1). Likewise, a Roche (Transcriptor 1-step RT-PCR kit; cat # 04655877001; Indianapolis, IN) and a Finnzymes kit (RobusT I RT-PCR kit; cat# F580L; Thermo Fisher Scientific, Lafayette, CO) that both include AMV RT were also positive for XMRV/ MLV pro (8/16 (50%)) and pol sequences (4/16 (25%) and 2/16 (13%), respectively ( Table 1). The Finnzymes kit failed to amplify sequences from the standard controls using the pro qRT-PCR primers and probes. Thus, the nested pol PCR assay was used to test the RT in the Finnzymes kit. Fig. 1b shows the detection of pol sequences in 2/16 (12.5%) replicates in the AMV RT from Finnzymes. Increasing amounts of AMV RT from the Finnzymes kit used in the nested pol PCR test also yielded increasing quantities of detectable contaminants (5/16, 31%). We also replaced the AMV RT from Finnzymes with an AMV RT from Promega and repeated the testing and found no positive reactions The reverse transcriptase (RT) enzymes and kits were tested by quantitative protease (pro), polymerase (pol), gag and/or nested pol RT-PCR assays and results shown are number of positive tests out of the total number of replicates. Percentages for replicate testing and copies/reaction (rxn) are in parentheses.
in the water only controls (0/16, data not shown), further supporting the contamination of the Finnzymes RT. In contrast, AMV RTs in two kits from Promega (Access RT-PCR (cat# A1259) and AccessQuick RT-PCR (cat# A1702) systems, Madison, WI) were both negative in all 16 replicates ( Table 1).
None of the enzymes with detectable MLV sequences were positive for mouse mtDNA or IAP sequences in any of the 16 replicates. These results suggest that the source of the MLV sequences in these enzymes is not mouse DNA.
To infer the origin of the contaminants we performed phylogenetic analysis of pol sequences from multiple clones from the PCR-positive RT enzymes and kits, multiple prototypic MLV pol sequences, and those identified with the highest homology by BLAST analysis using the Neighbor joining distance-based method. The pol sequences found in the RT enzymes shared .99% identity with MoMLV (GenBank accession number J02255). Phylogenetic analysis of 168-bp pol sequences showed that the RTs and RT kits have MLV-like sequences genetically related to MoMLV that are distinct from XMRV (Fig. 2). The absence of detectable gag sequences in the contaminated RT reagents is consistent with a provenance from a plasmid containing only pol sequences. As previously reported, the pol sequences did not cluster by virus host cell tropism, reflective of a probable history of recombination occurring in this region [3].  Contamination of commercial human DNAs with mouse DNA We also found that commercially available DNAs are contaminated with mouse DNA. Commercially prepared human cell line and tissue DNAs were purchased for use as negative controls in our PCR assays from Sigma-Aldrich (St. Louis, MO) and Biochain (Hayward, CA). All ten Biochain human specimens, including DNA from leukocytes, the B-and T-cell lines Raji and Jurkat, respectively, two tumor cell lines (K562 and MCF-7) and four DNAs from brain tissues of persons with neurologic disorders (dementia, Alzheimer's, Parkinson's, and multiple sclerosis) were positive for XMRV/MLV pro, and pol sequences, and mouse IAP and mtDNA sequences ( Table 2). In addition, all of the Biochain DNAs contained MLV gag sequences ( Table 2). Although most of the Biochain specimens showed low levels of contamination, the leukocyte DNA had significant amounts of contamination with ,8,000 copies/mg pro, 39,200 copies/mg IAP, and 232,000 copies/mg mouse mtDNA sequences detected (Table 2, Fig. 3). Although the assays showed the same level of sensitivity using plasmid standards, differences in the amounts of mouse mtDNA versus IAP sequences were detected in some DNA samples and can be explained as pipetting errors or due to increased expression of mtDNA in some cells compared to the stability of integrated IAP sequences. Similar discordant levels of mtDNA and IAP sequences have been reported previously in patient samples contaminated with mouse DNA [22].
Analysis of the 168-bp pol sequences confirmed mouse DNA contamination of the samples with the majority (6/7, 86%) showing 98-100% nucleotide identity to an endogenous retrovirus present on mouse chromosome 12 (GenBank accession # CT030142). The pol sequence detected in the MS brain tissue DNA had 99% nucleotide identity to a xenotropic MLV (XMLV15, Genbank accession # HQ154630). In contrast, two human placental DNAs from Sigma-Aldrich both tested negative for XMRV/MLV and mouse IAP and mtDNA sequences ( Table 2). Phylogenetic analysis of 168-bp pol sequences obtained from the Biochain samples inferred genetic relatedness across the diversity of MLV by clustering with different xenotropic and polytropic MLVs, but which were all distinct from XMRV (Fig. 2).
Likewise, phylogenetic analysis of an alignment of 301-bp gag sequences showed that those obtained from the Biochain DNAs were divergent from XMRV and clustered with various xenotropic and polytropic MLVs (Fig. 4). These results are consistent with the genetic diversity of endogenous MLVs in the mouse genome. Interestingly, we also observed the same level of diversity in PMLV sequences identified in CFS patients and blood donors, and in other human specimens and cell lines, congruent with an external mouse cell or DNA contamination source (Fig. 4) [2,17,20,25,26]. Biochain was notified of these results and independently confirmed the presence of mouse mtDNA sequences in the human leukocyte DNA. After implementing measures to prevent carry-over contamination of human DNAs with mouse material also processed at their facility, a new lot of human leukocyte DNA (B210149) tested negative for mouse mtDNA sequences at Biochain, and MLV sequences at CDC, but was still weakly positive for IAP and mtDNA sequences (1-10 copies/rxn) at CDC (Table 2).

Discussion
Our results show the frequent contamination of commercial RT enzymes and RT-PCR kits with residual pol sequences that are highly related to MoMLV. The absence of detectable MLV gag and mouse DNA sequences suggest that the most likely source of the contamination is an expression plasmid vector containing only MoMLV pro and pol sequences. These results are consistent with patented MoMLV RT-expression vectors that consist of only partial pro and 59 RT sequences. The RNase H and integrase regions of the MoMLV pol in this expression vector were selectively removed. The low-level contamination most likely reflects the difficulty in completely removing plasmid DNA with DNase treatment following protein expression. Our data also show that the same MLV sequences are present in AMV RTs, suggesting that the source of contamination of this reagent is not the AMV plasmid but likely carry-over contamination with the MLV RT-containing plasmid. It is possible that such contamina- 1 ug of the human DNAs were tested using quantitative (qPCR) pro and nested pol and gag PCR assays. Copies/mg DNA is provided in parentheses for specimens with evidence of contamination. 2 Mouse DNA contamination was detected using mitochondrial DNA (mtDNA) and intracisternal A particle (IAP) PCR qPCR assays. 3 Second lot of human DNA sample received from Biochain after informing them of the contamination results. 4 NA, information not available. doi:10.1371/journal.pone.0029050.t002 tion can occur during production of both AMV and MLV RTs in the same laboratories. Although we showed that currently used diagnostic PCR can detect this MLV sequence, the phylogenetic analysis was important to demonstrate that the sequences are highly related to MoMLV, an ecotropic virus that cannot infect human cells, and would thus indicate contamination if found in a human specimen. Our data therefore re-iterate the importance of phylogenetic analysis for all specimens testing PCR-positive [19,27]. We also demonstrate that commercially prepared human DNAs from a variety of clinical specimens and cell lines can be contaminated with mouse DNA likely originating from mouse tissue processing in the same facility. Unlike the situation observed for the RT kits, we showed that the level of contamination of these specimens can be high and, thus, easy to detect by assays for mouse or MLV DNA. Our findings should serve to remind commercial providers to institute measures to prevent crosscontamination.

Conclusions
Combined, our results show that MLV and mouse DNA contamination is more widespread in diagnostic reagents and clinical specimens than previously thought. These data reinforce the need for rigorous screening of all diagnostic reagents and specimens by methods capable of detecting trace amounts of MLV and mouse DNA, and to use phylogenetic analysis to assess the biological significance of newly detected sequences. Also, since many of the contaminated RT and PCR reagents are widely used in metagenomics studies, our data call for careful interpretation of any murine retroviral sequences that are detected in these analyses.

RT-PCR kits and human DNAs
The manufacturer and lot numbers of the nine RT-PCR kits tested in our study are listed in Table 1. Nine human DNAs from either disease tissues or cell lines were purchased from Biochain (Hayward, CA), while the male and female placental DNAs were obtained from Sigma-Aldrich (St. Louis, MO) ( Table 2).

PCR testing for MLVs and mouse sequences
The pro qRT-PCR assay was used to screen specimens and reagents for evidence of MLV/XMRV contamination since it generically detects all MLVs and XMRV [3]. This assay uses the Taqman primers Pro-UNV-F1 (59 CCT GAA CCC AGG ATA ACC CT 39) and Pro-UNV-R1 (59 GTG GTC CAG CGA TAC CGC T 39) and probes Pro-UNV-P1C (FAM59 AGA TAC TGG GGC CCA ACA CTC CGT GCT GAC 39BHQ1) and Pro-UNV-PR1 (FAM59 CCT CCA GTA GCC CCT TGG ACC CAG GC 39BHQ1) in a one-step RT-PCR reaction mixture at 45uC for 20 min, 95uC for 10 min, followed by 55 cycles of 95uC for 30 sec, 52uC for 30 sec and 62uC for 30 sec. We also use an engineered standard template for quantification of pro sequences that contains two unique nine-bp nucleotide insertions that code for the amino acids methionine-leucine-valine (MLV)and tryptophan-methionine-threonine (WMT) separated by a 31-bp hamster gammaretrovirus sequence that are not present in wild-type MLV pro sequences [28].
To further evaluate the origin of the MLV sequences detected in the RT reagents and human DNAs we also used three additional PCR tests targeting the gag and pol regions of the MLV and XMRV genomes. For the gag qRT-PCR assay, the primers GAG-UNV-F1, 59AGGTAGGAACCACCTAGTYC and GAG-UNV-R1, 59GTCCTCAGGGTCATAAGGAG and probes GA-G-UNV-P1C, 59FAM AGCGGGTCTCCAAAACGCGGGC 39B-HQ1 and GAG-UNV-PR1, 59FAM CCTTTTACCTTGGCCA-AATTGGTGGG 39BHQ1 were used [5]. For the pol qRT-PCR test, the primers POL-UNV-F1 59CAGAGATGGCTGACTGAG-GC39 and POL-UNV-R1 59AAACAGAGTCC CCGTTTTG-GT39 and probes POL-UNV-P1C 59FAM AGGGAGTTCCTA-GGGACGGCAGGCTTCT 39BHQ1 and POL-UNV-PR1C 59FAM CAAACCCAGGGATCCAGAGGCGACAG 39BHQ1 were used. Both assays used the same conditions as the pro qRT-PCR test and gave sensitivities similar to the pro qRT-PCR test detecting 1-10 copies MLV/XMRV per reaction (data not shown). We also used a nested polymerase (pol) RT-PCR test that generically detects MLV and XMRV sequences about 216-bp in length using the primers XPOLOF and XPOLOR, and XPOLIF and XPOLOR [5,9].
For detection of mouse mtDNA, the primers MCOX2F2 (59  TTC TAC CAG CTG TAA TCC TTA 39)  Dilutions of a plasmid containing murine COX2 sequences that was generated by PCR with the MCOX2F1 (59 ACA TAC AAG CAC AAT AGA TGC 39) and MCOX2R1 of the murine macrophage cell line RAW264.7 (ATCC, Manassas, VI) were used as the assay standard. For detection of IAP sequences we developed a new qPCR assays that uses the generic primers and probes IAP-MH-POLF2: 59 GCCTCAYATGTG ATTCAA-CATTG 39 and IAP-MH-POLR2: 59 TTGRGASGTATAW-GCTGGT CCATT 39 and IAP-MH-POL P2: 59 FAM TTG-AGGCMTGGAGTGCTTGGGGRAAACCCAGA 39BHQ1, respectively. The assay was performed with a hot start at 95uC for 9 min followed by 55 cycles of PCR at 95uC for 30 sec and 62uC for 30 sec. The PCR product amplified from the murine macrophage cell line RAW264.7 DNA with the IAP-MH-POLF2 and IAP-MH-POLR2 primers was cloned and used as the assay standard.
Preparation of RNA standards used in the RT-PCR testing XMRV pol and gag genes were amplified using the primary PCR primers, cloned into pCR2.1 TOPO-vectors (Invitrogen), and transcribed in vitro using the Promega RiboMAX large scale RNA production system (T7). Following several DNase treatments to completely remove plasmid DNA, the RNA was quantitated on a Nanodrop spectrophotometer. Copy number was determined using the formula (ng RNA6Avogadro constant)/(RNA length6(1610 9 )6325) and was verified by comparison of qPCR results using XMRV RNA extracted from 22Rv1 cell culture supernatant quantitated independently [3].

Sequence analysis
PCR products were purified with QiaQuick PCR or gel purification kits (Qiagen, Valencia, CA) and were directly sequenced on both strands by using ABI Prism BigDye terminator kits and an ABI 31306l sequencer (Foster City, CA) or following cloning in the TOPO vector (Invitrogen). Initial sequence identity was determined using BLAST analysis at the National Center for Biotechnology Information web server (http://blast.ncbi.nlm.nih. gov/Blast.cgi) using the blastn search option. Sequences were aligned with those retrieved from the BLAST analysis with the highest nucleotide identity, and other MLV prototypes available at GenBank, using MAFFT [29]. Following manual editing and removal of indels, substitution models and phylogenetic relationships were inferred using the neighbor joining (NJ) method implemented in MEGA v5.03 [30]. Support for the branching order was evaluated using 1,000 nonparametric bootstrap replicates. The 168-bp pol sequences are available from the authors upon request. GenBank does not accept sequences less than 200-bp in length. Accession numbers for the new gag sequences generated in our study are JN629081-JN629087.