Is MMTV associated with human breast cancer? Maybe, but probably not

In an IRB-approved archival study we obtained peripheral blood mononuclear cells (PBMC) from 65 subjects deemed to be at risk for retroviral infections [namely, 18 mothers of HIV-1 infected children and 47 HIV-1, HIV-2, HTLV-1 and/or HTLV-2 infected intravenous drug users (IVDU)], and from 100 volunteer blood donors (VBD). We also obtained 66 samples of formalin fixed paraffin-embedded (FFPE) or snap-frozen (SF) biopsies of breast tissue: 10 SF breast cancer, 9 SF fibrocystic breast, 19 FFPE breast cancer, 19 FFPE fibrocystic breast, and 9 FFPE breast reduction samples. The above samples were collected over several years from multiple clinics and operating rooms at Upstate Medical University and Staten Island Hospital. The breast cancer specimens were originally processed in the Pathology Department at Upstate Medical University. All pre-PCR processing was done by one person, in one room, at Upstate, while all PCR and post-PCR work was done by a different person, in a different room, at Upstate.

We examined 20 breast cancer cell lines (HCC1419, HCC1569, HCC1500, HCC70, MDA-MB-453, HCC1143, HCC1187, BT474, HCC1395, HCC1428, BT-20, T-47D, MDA-MB-468, HCC38, HCC1954, HCC2218, HCC1599, HCC1937, ZR-75-1, MCF7, obtained from ATCC); 15 malignant T-lymphocyte cell lines (HUT78, HUT102, MOT, UMC 11B, CEM, JK, GA, JS, JC, MOLT4, MT2, UMC ATL-20, NIH CTCL11, SLB-1, MLA144), 10 EBV-transformed B lymphocyte lines (HCC38BL, HCC1143BL, HCC1187BL, HCC1395BL, HCC1428BL, HCC1599BL, and HCC1937BL, obtained from ATCC and UMC402-EBV45, UMC406-EBV51, and UMC412-EBV54 (a kind gift of Steve Graziano), and 5 non-lymphoid malignant cell lines [one sarcoma (HOS); three brain tumors (JON, ROS and BAIRD, a kind gift of Gregory Canute), and one malignant histiocytic lymphoma (ALGR)].

In a CHUA approved component of the study, DNA for analyses was extracted from the spleens, livers and tails of 18 wild mice cadavers caught in Upstate New York. DNA of laboratory strains and the wild strains, Mus musculus molassinus, Mus musculus castaneus, Mus spretus, Mus pahari, Mus caroli, WSB/Eij, and Czech II/EI were obtained from Jackson Laboratory, Bar Harbor, ME. DNA was extracted from the livers, spleens and/or tails of laboratory strains of one C3H, one Balb C, one SWR, one NZB, ten C57 Bl6 and ten C57 Bl10 mice. Analyses were also done on DNA in our laboratory from a gerbil, rabbit, rat, squirrel and vole. DNA was extracted from the MMTV infected C3H mammary tumor cell line, Mm5MT (ATCC).

Pre-PCR work and post-PCR analyses were done by different laboratory personnel in different labs with separate ventilation systems, so as to prevent contamination by previously amplified DNA. DNA was extracted from the various human and murine samples listed above, as previously described [28]. Human studies were done antecedent to rodent studies. Human samples were tested for human β-globin DNA, as previously published [29]. Only samples that tested positive for human β-globin at 0.1 μg and 1.0 μg of DNA input were deemed suitable for subsequent retroviral DNA analyses (i.e. they contained amplifiable human DNA).

Similarly, to check for the presence of suitable amplifiable murine DNA, the murine samples were amplified with a rodent tropomyosin primer pair/probe set. The forward primer MTmF1 (5′-ATG GAC GCC ATC AAG AAG AAG ATG-3′), reverse primer MTmR1 (5′-GCA GAC CTG CTG GCT CCG-3′) and probe MTp (5′-CTC GAC AAG GAG AAC GCC TTG GAT CGA GCT GAG CAA GC-3′) were used. These same primers and probe were used to assess for the contamination of the human samples with rodent DNA.

DNA was analyzed for the presence of MMTV env sequences using viral-specific primers that contained “signature” non-human, non-murine, non-viral (NHNMNV) sequences at their 5′ termini. Because only synthetic amplified DNA would contain them, signature primer pairs, comprised of just the (NHNMNV) sequences, can be subsequently used to determine if a positive result originated from the target viral sequence or from “carryover” contamination [30]. The “signature” primers were forward 5′-TAC GAG CTC GCG AAT TCA TGA T-3′ and reverse 5′-ACA GGT ACC TGC AGA TCT AG-3′. The MMTV env primer that contained “signatures” at the 5′- termini were forward BR-7376F 5′-TAC GAG CTC GCG AAT TCA TGA TCC AGA TCG CCT TTA AGA AGG-3′ and reverse BR-7548R 5′-ACA GGT ACC TGC AGA TCT AGT AAT CTG ATC TGA CTG ATC TAC ACT-3′. The probe for amplified MMTV env DNA was BR-7422pr 5′-GCT CCT CCA CGG TGG TTG CCT TGC GCC TTC CCT GAC CAG G-3′. This primer pair/probe set was termed the “clinical” detection set (Fig. 1).

MMTV PCR assays were conducted under varying permutations. Optimal conditions for the MMTV PCR were: 1 cycle at 94 °C for 5 min., 45 cycles at 94 °C for 30 s., 55 °C for 30 s., and 72 °C for 45 s., followed by a final extension of 72 °C for 5 min. The 50 μl PCR mixture contained 1 μg or 0.5 μg of target DNA, 200 mM of primer, 200 mM dNTPs, 2.5 mM MgCl₂, 1 U Taq polymerase. dUTP rather than dTTP was used as a substrate, and all samples were pre-sterilized with uracil-N-glycosylase to prevent carryover contamination [31]. PCR products were electrophoresed into 2% agarose gels containing ethidium bromide and visualized by epi-illumination with UV light. The amplified DNAs were further analyzed, as previously described, by Southern blot hybridization using the above probes [28]. The assay was sensitive to one copy of input MMTV DNA in a Poisson distribution.

Human samples positive for MMTV DNA were reanalyzed with “signature primers” as previously described, to ascertain whether “carryover” contamination could explain the result [30].

The copy number of MMTV env DNA was calculated by comparing the amplification results to those of a serial dilution of Mm5MT cells. The relative copy numbers of MMTV env DNA in a Mm5MT cell was determined by making serial 10 fold dilutions of Mm5MT cells in the human T-cell line HUT 78 DNA. Replicates were analyzed via MMTV env PCR and a Poisson calculation was performed on the results.

Amplicons of MMTV env-pol DNA using the primers described below were ligated into TOPO-TA (Invitrogen), transfected into E. coli on agar plates, and grown o/n at 37 °C (TOP 10 cells; Invitrogen). Positive colonies were identified by blotting and probing with the MMTV specific sequences above. They were then grown o/n in LB broth. DNA was extracted using QI prep Spin Miniprep Kit (Qiagen Sciences, Germantown, MD) and then sent to the Cornell University Life Sciences Core Laboratories Center (Ithaca, NY) for DNA sequencing using overlapping primers: BR-7321F-5′-CCA ATA CAA AAC TGG TCC CTA-3′ and BR-7507R-AAA TCC CAA AGT AAC CCA AGG-3′, BR-7462F-5′-GGG TGA GTT TTT CTC CAA AAG G-3′ and BR-8070R-5′-AAT CAA AGC AGA TAT GCC CAG-3’ [32]. The respective primers used were termed the “sequencing” primers, and dTTP rather than dUTP was used to generate these amplicons (Fig. 1).

Newly obtained MMTV and human mammary tumor virus (HMTV) DNA sequences and those from databases or published papers were aligned with the env/pol regions corresponding to the region amplified by the above PCR conditions (GenBank accession nos: D16249, AY152721, AF228552, KO1788, KO0556, KU184587-KU184607, AY152722, M15122, M22028, M29600, AF243039, AF228551, X64553) [9, 33, 34]. Homologous sequences from the MMTV-related human endogenous HERV-K family of retroviruses served as an outgroup (GenBank accession nos: AF020092, AF012336, AF164612, AF164614, AF164615, AF164610, M14123, AF164613, AF164609, AF164611). Alignments and phylogenetic trees were produced using pileup (Accelrys GCG Seq Web version 3.01 and GCG installed on a UNIX host system). Phylogenetic analysis of the sequences was carried out by using an algorithm for construction of neighbor-joining (NJ) phylogenetic trees as implemented in the phylogeny inference package PHLIP version 3.57C [35]. NJ analysis included 100 resamplings of aligned sequences using the SEQBOOT program. The NJ trees were generated by computing the distance matrixes (Kimura distances, two-parameter model) from bootstrapped sequence data using DND DIST for nucleic acids followed by the Neighbor Program (outgroup rooting and randomized input order options “on”). The majority rule consensus NJ trees were generated using the CONSENSE program. Additionally, the maximum-likelihood method in Puzzle 4.0 package was used [36]. Phylograms were visualized using Tree View [37]. Experimentally derived DNA were subjected to standard nucleotide BLAST searches (http://​www.​nebi.​n/​m.​nih.​gov/​BLAST) to define homologies with those deposited at GenBank, European Molecular Biology Laboratory and the DNA database of Japan.

Differences in the prevalence of MMTV sequences in various populations were analyzed using Fisher’s t-test [38].

Figure 2 and Table 1 show the initial MMTV env DNA PCR results using the “clinical” primer pair/probe system on various human samples. As can be seen in Table 1, a number of positive samples were detected. There were 10 out of 86 (11.6%) positive breast samples and 9 out of 195 (4.6%) positive non-breast samples (p = 0.031). There were 6 out of 37 (16.2%) positive normal breast samples and 4 out of 49 (8.2%) positive breast cancer specimens (p = 0.249). The positive samples were re-amplified with “signature” primer pairs and all were negative indicating the absence of “carry over” contamination. Interestingly, save for one frozen human breast cancer sample (40412) in which the MMTV env copy number was ≥1500 copies per μg of human DNA (Fig. 2), all of the other positive human samples had low copy numbers - on average 10 copies/μg of human DNA (data not shown). Because a retroviral-caused cancer, including murine breast carcinoma caused by MMTV insertional mutagenesis and human T-cell lymphoma/leukemia virus which acts in trans, usually have at least one copy of integrated provirus per cancer cell (i.e. 150,000 copies/μg tumor DNA) this observation made it unlikely that the MMTV detected in most of the human breast cancers could have been the cause of those breast cancers. This observation and the fact that none of the samples in the retroviral risk group nor breast cancer cell lines were positive also raised the possibility of an artifactual cause of positive results.

The MMTV tropomyosin primer pair/probe set gave positive results in all murine, gerbil and rat samples and negative results in the rabbit, squirrel and vole samples. Hence, they could detect DNA from a subset of the Muridine Family of rodents. A possible explanation for a false positive human sample could be contamination with naturally infected (either endogenous or exogenous) MMTV positive rodent DNA. Using this assay, we tested the above MMTV positive human samples for the presence of rodent tropomyosin DNA. A minority of these were positive on any given day that we tested, including aliquots to which we had not added any DNA sample (ie. primer only controls). (A representative experiment is shown in Table 2). The human breast cancer sample (40412) that had the high MMTV copy number was also positive for rodent DNA (Table 2).

Theorizing that a source of contaminating rodent DNA could be rodents living, dying and/or being dissected in our own building, we decided to episodically test for rodent DNA and MMTV DNA over a year’s time and correlate that with three periods of construction in our building. As can be seen in Fig. 3, there was a high correlation between these periods and the detection of both rodent and MMTV DNA.

Table 3 shows the MMTV env DNA prevalence in various wild mice and inbred strains of mice. All laboratory mice and 11 of 18 wild Upstate New York mice were positive. Using the criteria that, if DNA from the tail, liver, and spleen were all positive for MMTV, we would assume an endogenous infection, four of the 11 wild Upstate New York positive mice were deemed endogenously infected and 7 exogenously infected. All of the samples from the laboratory animals were positive suggesting that they were endogenously infected. Of the samples deemed to be endogenously infected, all had high MMTV DNA copy numbers per μg of sample DNA (data not shown).

Only one human sample (HMTV 40412) had enough DNA and MMTV copy number to allow for env sequencing (GenBank accession #AY152722). We obtained MMTV env sequence from one C3H mouse, one C57 Bl 6 mouse, six C57 Bl 10 mice, one SPRET/EiJ mouse, one CASA/RkJ mouse, one MOLF/EiJ mouse, and two wild mice from New York State. Several of these animals had more than one MMTV sequence among the clones analyzed. These new, either human (HMTV) or murine (MMTV), mammary tumor virus sequences were compared phylogenetically to published HMTV and MMTV sequences (Fig. 4). The sequences of the regions chosen as the “clinical” primers and probes in this study were 100% conserved in all published and newly described MMTV and HMTV strains. As seen in Fig. 4, there are at least 5 to 7 major clades of HMTV/MMTV sequences. There were 2 major subclades of MMTV in the C57 Bl 10 mice. Our human strain (HMTV 40412) is identical to an endogenous MMTV sequence found in C57 Bl 10 mice raised in our own facility, suggesting that it could be a contaminant from rodent DNA in our environment. Likewise, several of the published HMTV sequences are very similar or identical to MMTV sequences found in laboratory mice. However, a number of HMTV sequences published by others (eg. HMTV184b, HMTV332, HMTVt6, HMTV612) are quite distinct from any of the MMTV sequences. These HMTV sequences were derived in one lab and presumably collected in metropolitan New York City [34]. The ethnicity of these subjects was not reported.