We identified 6234 integration sites, or tags, in 2199 T-cell tumors. In these tumors, 243 tags were located at or near the Pvt1 locus, distributed over a region of 679,620 bp; additionally 134 tags were located at the Myc locus, distributed over 105,445 bp (Fig. 1). The proviral inserts were in both sense and anti-sense orientations with respect to each transcript encoded by the Myc and Pvt1 loci, respectively. The Mouse Retroviral Tagged Cancer Gene Database 29, which compiles retroviral insertions into the genomic DNA from various non-T cell derived mouse tumors, also lists 37 integrations when searched by the Myc locus, some of which are in fact in the Pvt1 locus. Insertions at the Pvt1 locus were originally reported in myelogenous mouse leukemia 29, and, as mentioned above, in the work defining the Pvt1 locus in T lymphomas induced by MLV in both mice and rats 8910. Remarkably, in a separate screen (data not shown) where we recovered 1798 tags from B lymphomas induced by the MLV strain Akv 1530, only one tag was found at the Myc locus, and none at the Pvt1 locus.

Fig. 1 shows a customized screen print of the UCSC genome web site browser, looking at the Myc and Pvt1 loci. The bars in green represent the retroviral insertions in T lymphomas studied here; below them are the exon-intron structures of Myc and Pvt1, respectively. At the Myc locus, there are two main integration sites clusters flanking the gene upstream and downstream of it. Whereas the Myc transcript is clearly defined, there are several alternative transcripts depicted for Pvt1, a variety that was noted before 234. Notably, there are two reference sequences, AK090048 and Z11981, which do not share any sequences, but are denoted as Pvt1 nevertheless. Furthermore, among the mRNAs from GenBank, there are other fragments of apparently intronic transcripts, and there is AK030859, which represents an extended exon 1. At any rate, there are three main integration site clusters at the Pvt1 locus, as represented by transcript AK090048 – one upstream of the transcript, and two within the locus.

When a genomic region is gene-rich, it is not always straightforward to identify the target gene of insertional mutagenesis. In the past, it has been assumed that the retroviral enhancer can act over a distance of 200 kb in either direction, but without "leapfrogging" a gene promoter. With this assumption, because one of the proximal promoters will always be the retroviral promoter, the orientation of the provirus in regard to the direction of transcription of the gene will be important. Indeed, the two integration clusters into the Myc locus are an example of this prediction: the direction of transcription of the provirus upstream of the Myc gene always points away from the gene (Fig. 2A; with the exception of the three insertions, boxed in red, which presumably are "promoter insertions," i.e., the transcript is driven by the viral promoter rather than the endogenous promoter). In contrast, the cluster downstream of the Myc locus contains proviruses in the same orientation as Myc (Fig. 2B). In both cases, this arrangement allows the retroviral 5' enhancer to interact with the Myc promoter, although other interpretations are viable (see below). Because of the rule that the retroviral enhancer does not "leapfrog" promoters, but synergizes with the two promoters next to it, the two clusters targeting Myc are not expected to directly influence Pvt1 transcription, 50 kb downstream.

The criterion of orientation does not hold in an immediately obvious way if a virus integrates into a transcription unit, as it does at the Pvt1 locus. In this case, especially as many alternative transcripts have been identified, the exact location of the retrovirus – 5'UTR, 3'UTR, intron, or exon is important. Apart from the retroviral enhancer cooperating with the gene promoter in a conventional manner, the retroviral promoter may override the endogenous promoter, or it may initiate a (truncated) transcript, in addition to truncating or destroying one. If the provirus is located with the UTR, it may also affect mRNA stability, although in that case no preference in proviral orientation would be obvious. If the Pvt1 nuclear (primary) transcript encodes miRNAs, we cannot predict the likely consequence of a particular integration – whether the steady-state levels of all or only a few miRNAs change. A low level of Pvt1 transcript does not necessarily mean little miRNA product. For example, NIH-3T3 mouse fibroblasts express very little primary RNA of the mir-17-20 cistron, but as much mature mir-17-3p as T cell tumors with retroviral integrations into the primary transcript 23. This points to the possibility that retroviral insertions do not always have to increase the levels of primary transcripts in order to produce more mature product; instead they might make the processing of miRNA from the primary transcript more efficient.

Fig. 3 shows a higher magnification of the area around exon 1 of Pvt1, where a main cluster of 78 integrations is located. Because a plurality of the Pvt1 integrations clustered around exon 1, we determined the expression levels of that exon (exon 1a) in various tumors by quantitative PCR, using a primer set that covered the 5' end of this exon (see boxed area in transcript AK030859 depicted in Fig. 3; the 5' end of the exon representing AK030859 is shared with exon 1 of the reference sequence AK090048). Of the tumors with integrations shown in Fig. 3, the designations of tumors we selected are shown in bold face type, and are numbered (1) through (28) (only tumors 1 through 24 are shown in Fig. 3; the integration sites of these tumors, and all other tumors studied here, along with the relative transcription orientation of the proviruses, are given in Table 1). As compared to control tumors, which have no integration into the Pvt1 locus, most tumors with the integrations selected in Fig. 3 overexpressed the Pvt1 transcript, up to 40-fold (Fig. 4A). Tumors 10 through 28, with insertions starting right at the 3' boundary of the first exon, mostly overexpress Pvt1 with a few exceptions (tumors 13, 17, and 19). We have noticed that the direction of transcription of the proviruses in tumors 13 and 19 is opposite of all the others in that group (see above for discussion of provirus transcriptional orientation). Tumors 1 through 9, with insertions located 5' to exon 1a, express Pvt1 at levels similar to the control tumors (the control tumors are not listed in Table 1). They possibly overexpress transcripts starting with exon 1b (Fig. 5), although we have not tested this assumption.

Because transcript AK030859 seems to represent a (less frequent) alternative splice product of the putative nuclear transcript, we also performed qPCR analyses with a primer set covering the 3' end of AK030859 (see right boxed area in Fig. 3). In these analyses, tumors with insertions at the Pvt1 locus on average expressed more AK030859 sequences than the control tumors, (Fig. 4B) (the control tumors are not listed in Table 1).

It is possible that the common integration site at the Pvt1 locus is not actually due to selection for tumorigenesis via Pvt1, but to preferred (yet unknown) integration sequences at this locus. In this view, the increased Pvt1 expression would be of no biological consequence, but the insertions actually would increase Myc expression directly. We thus investigated Myc expression in tumors with insertions at the Myc and Pvt1 locus, respectively, and compared them to tumors without insertions at either of these loci; and to normal spleen cells or thymocytes from mock infected (i. e., no virus) mice. Clearly, the normal cell controls expressed less Myc than the tumors (Fig. 4C). But by and large, there was not much difference in Myc expression among the tumors, whether they had an insertion into the Myc locus, the Pvt1 locus, or no such insertion (Fig. 4C). Thus the SL3-3 induced T lymphomas generally have elevated Myc expression, no matter by which insertion that is accomplished, and there is no obvious correlation between location of insert into the Pvt1 locus and Myc expression.

It is surprising that although only 6% of the T lymphomas have insertions directly into the Myc locus, almost all T lymphomas overexpress Myc as compared to normal splenocytes and thymocytes, whether there are insertions into the Myc locus, Pvt1 locus, or into an unknown site. This fact could be taken as an indication that retroviral integrations are capricious and not always the driving force of tumorigenesis. However, we interpret these data to mean that there may be a requirement for MLV induced T lymphomas in BALB/c mice to overexpress Myc, regardless of how this is achieved.

Although at the time of manuscript preparation no miRNAs were listed in the miRNA registry of the The Wellcome Trust Sanger Institute 3132 that map to the Pvt1 locus, the expressed sequence tag pattern indicated the possibility that Pvt1 does encode miRNAs. Indeed, using previously described algorithms that use sequence conservation of putative seed sequences and secondary structural properties of the putative miRNA hairpin structures, Pvt1-based miRNA candidates in human, chimpanzee, canine, mouse and rat have been identified 28, and confirmed experimentally in human and mouse 28. The human miRs have recently been given designations by the Sanger miRBase, and we have adopted the analogous nomenclature for the mouse miRs. Fig. 5 shows the genomic sequences of mouse Pvt1 associated miRNAs and their flanking sequences in mouse; the miRNAs are called mmu-mir-1204, mmu-mir-1205 mmu-mir-1206, mmu-mir-1207-5p, mmu-mir-1207-3p, and mmu-mir-1208. Because in the following, we are only dealing with mouse sequences, we will omit the pre-fix "mmu." The mature miRNA sequences are shown in red. Above the sequences, their relative genomic locations, on chromosome 15, are shown. With mir-1204 closest to the Myc locus, at a distance of approximately 50 kb, and miR-1208 furthest away (305 kb), the pvt-1 primary RNA, if a single transcript, spans at least 255 kb. The exact genomic locations of the Pvt1-encoded miRNA sequences are given in Table 2.

To determine if the retroviral integrations altered expression of these Pvt1 associated miRNAs, we measured the expression of the mature species of the five miRNAs by qRT-PCR using a stem-loop RT primer specific for each miRNA 2433, in tumors with and without Pvt1 insertions (Table 3). For standardization, we compared them to known concentrations of synthesized miRNAs of the relevant sequence. While we could detect a signal for miR-1206 only in one tumor, we did find expression of miR-1204, miR-1205, miR-1207-5p, miR-1207-3p and miR-1208, albeit at quite different levels. On average, mir-1204 was most pronounced as it was expressed nearly four times more in tumors with Pvt1 inserts than in the control tumors (Table 3; Δμ 0.05) – irrespective of the site of retroviral integration within the Pvt1 locus. Because thymocytes and spleen cells represent a mixture of many cells, one cannot directly compare these cells with the tumor cells. Nevertheless, we note that the expression level in the tumors with Pvt1 integrations was not significantly different from thymocytes and non-stimulated spleen cells. It therefore seems as if this miRNA is required for cell survival. The relatively modest overexpression in tumors with insertions into the Pvt1 locus may be a consequence of the retroviral enhancer, but the tumorigenicity of the provirus may be mediated by the persistence of miRNA expression rather than by its overexpression.

Although miR-1205 and miR-1208 gave clear signals, the threshold was only reached after 40 cycles, making the significance of these miRNAs in tumorigenesis less clear. However, in three tumors (#31, #32, #34) with integrations close to its genomic position, miR-1205 is expressed more than in other tumors; and the expression level of miR-1205 in thymus (34 cycles to reach threshold; Table 3) makes it likely that miR-1205 plays a role in normal cell differentiation. In most of the tumors, we did not find miR-1206 expression; although precursor RNA was increased in the mouse myeloma MOPC104E 28, we did not find the mature miRNA expressed (not shown). In fact, there was also no expression in thymocytes and spleen cells, but tumor 16 gave a clear and reproducible signal. Since this tumor does not differ in its integration site or proviral transcriptional orientation from other tumors with insertions in this region, we think that the miR-1206 expression is not mediated by the provirus. Rather it may be the effect of another mutational event, which in myelomas is more frequent. The level of mmu-miR-1207-5p was relatively low in thymus; but the levels of miR-1207-5p and miR-1207-3p in tumors with and without integrations into the Pvt1 locus did not differ much, and thus we cannot correlate expression of these miRs with an oncogenic event. In all the tumors, it is possible that the other allele (with no proviral integration) contributes to the miRNA levels, which may mask differences.

Overall we can conclude that except perhaps for miR-1206, the other Pvt1 encoded miRNAs are expressed in T-lymphocytes. However, we have not yet performed a detailed analysis of the consequences of the various proviral integrations sites. We can assume that the exon 1 overexpressing tumors end their transcripts with the retroviral termination site and poly A tail, which would exclude all the downstream miRNAs. However, the 3' retroviral promoter may also restart a transcript, as has been discussed for integrations into the Notch1 locus 34. An indication for this is the fact that the qPCR primers covering the the 3' end of the intron-less transcript AK030859, also measured increased expression levels in tumors with insertions between the DNA segments of probe sets 1 and 2. At any rate, we feel justified in concluding that except perhaps for miR-1208, all other Pvt1 encoded miRNAs do exist, and that it is likely that murine mir-1204 is oncogenic in T lymphomas when constitutively expressed.

It is well established that tumorigenesis is the result of accumulating several cooperating mutations that drive relentless proliferation and aid in metastases. Co-mutation analyses, where one oncogenic event is fixed by means of a transgene in the mouse to be infected with retrovirus, were very successful in identifying cooperating oncogenes, for example with Myc 14, or with p27Kip1 loss 19. Without fixing any event by a transgene, viral insertional mutagenesis, though perhaps not providing all the mutations necessary for a full-blown tumor, follows the multistep scenario of tumorigenesis. Although in general the superinfection barrier largely prevents multiple proviral integrations within the same cell, re-infection does happen over time. Because it is a rare event, such cells are selected over the others only when these integrations also give a growth advantage. As a consequence, in general, most viral insertions ("co-mutations") in a single tumor are thought to be causative in its formation. With the caveats of potential passenger mutations and potential oligoclonality of tumors, co-mutation analysis may be a powerful way to find cooperating signaling pathways in tumorigenesis. For this analysis, the following two rules can be stated: (i) genes that are co-mutated in a single cancer cell represent different pathways that cooperate during carcinogenesis; and (ii) genes within the same pathway are never co-mutated. These rules assume "linear," non-branched pathways, which is a gross oversimplification. They also assume that it does not help to turn on a pathway (twice) by two integrations rather than one, but, of course, increased signal strength may indeed help tumorigenesis. For example, an obvious exception is Notch1, for which two mutations have been shown to lead to more aggressive growth than just one 35 – a fact that is reflected by our finding of three double mutants (Fig. 6; and unpublished), with mutations in the same two domains that are also co-mutated in patients. Nevertheless, if in a large sample set, one never finds two genes co-mutated, it seems fairly safe to assume that they are in the same pathway.

Fig. 6 shows a matrix for co-mutation analysis, here focusing on Pvt1. The numbers (color blue, underlined) represent the frequency of tagged oncogenes/tumor suppressor genes detected in the T lymphoma screen, in the order of their incidence, horizontally and vertically. The numbers in the boxes at the intersections (color black) indicate the number of tumors the cancer genes were found in the same tumor. In this matrix, Pvt1 is represented by number 3, i.e., it is the third most frequent cancer locus in mouse T lymphomas mutated by MLV. The other high frequency tagged genes (1, 2, 4, 5, 6 and 7) are Evi5, Notch1, Rasgrp1, Ahi1, Gfi1 and Myc, respectively. Because they are found together with Pvt1 in a number of tumors, these genes are all co-mutations, except for Myc – there are no tumors with insertions to both Myc and Pvt1 loci. By the logic above, this apparently places Pvt1 and Myc into the same pathway, although from this analysis it cannot be determined which one of the genes is upstream. Another indication for the two genes sharing a pathway comes from the fact that they have the same co-mutations (Fig. 6); and that they both do not co-mutate with gene 21, which thus ought to be in the same pathway as well.

With 11% of the 2199 T lymphomas studied having insertions within it, the pathway pvt is part of what seems to be one of the most important regulators of T lymphomagenesis in the BALB/c mouse strain, and, by extension, perhaps also in Burkitt's lymphoma and in mouse plasmacytomas, as these are clearly driven by the translocations involving the Myc and Pvt1 loci. If so, it seems peculiar that in our screen with Akv, which induces B cell lymphomas in NMRI mice, only one out of 1798 tags were in the Myc locus, and none in the Pvt1 locus. Similarly, among the resulting 24 tumors analyzed by Lovmand et al. 15, only one tumor, containing an Akv variant, harbored a clonal proviral integration in the c-Myc locus. Because most cells, including B cells, express the Akv receptor, the reason for this may lie in the differentiation stage of the infected cell.

BOSC23 retroviral packaging cells were transfected with plasmids encoding the complete SL3-3 provirus. Viral particles from culture supernatant were injected intraperitoneally into newborn (<3 days) BALB/c mice. The fathers of the injected mice were also mutagenized by ethyl-nitroso-urea as part of another study 36. Mice were monitored everyday for general sickness as well as tumor development. When sickness or tumors of defined size were discovered, mice were euthanized and tumors of the spleen and thymus were removed and frozen at -80°C.

The genomic locations of the proviral integrations were determined using the splinkerette-based PCR method 37. This method recovers genomic DNA directly flanking the 5' LTR of the integrated provirus. Genomic DNA was isolated from tumors using the DNeasy Tissue kit (Qiagen) and digested using restriction enzymes BstYI or NspI. A double-stranded splinkerette adapter molecule 38 containing the appropriate restriction site was ligated to the digested genomic DNA using the Quick Ligation kit (New England Biolabs). These ligation products were then digested with EcoRV to prevent subsequent amplification of internal viral fragments. The resulting mixture was purified using QIAquick PCR purification kits (Qiagen), and subject to three rounds of PCR using nested PCR primers that had homology to the adapter DNA and to the 5' LTR sequence of the SL3-3 virus. After resolving the PCR products by gel electrophoresis, the desired bands were purified using QIAquick Gel Extraction kits (Qiagen) and subject to standard DNA sequencing.

Total RNA was extracted from frozen mouse spleen and thymus tumor samples using the RNeasy Mini Kit (Qiagen). All RNA samples were treated with rDNase (Ambion) prior to reverse transcription. 500 ng RNA from each tumor sample was reverse transcribed with random hexamers using the SuperScript First-Strand Synthesis System III (Invitrogen). qPCR was conducted on the Stratagene MX3000P using Brilliant SYBR Green qPCR Master Mix (Stratagene). SYBR qPCR primers were designed using Beacon Designer 5.0 from Premier Biosoft and ordered from Integrated DNA Technologies. Beta-actin (ACTB) served as an endogenous control gene for all SYBR qPCR runs. qPCR primers were as follows: ACTB: 5'-TTCCAGCCTTCCTTCTTG-3', 5'-GGAGCCAGAGCAGTAATC-3'; Pvt1-exon1: 5'-(GAGCACAT)GGACCCACTG-3' (it contains 8 bp of genomic sequence before the start of AK090048 exon1, genomic part in parenthesis); 5'-GCTGCCAACATCCTTTCC-3'; AK030859 (3'end): 5'-GGCACAAGAGAACCAAGTCC-3', 5'-CGCTTATCCTCCTGCTTCAAC-3'; and Myc-ExJ2-3: 5'-GACACCGCCCACCACCAG-3', 5'-GCCCGACTCCGACCTCTTG-3'.

The qPCR reaction mixture contained 150 nM (final concentration) of each primer and the appropriate dilution of cDNA for each target studied in a final qPCR reaction volume of 25 μl. PCR cycling was as follows: 95°C 10 min; 40 cycles of 95°C 30 sec, Ta (annealing) of 55 to 60°C 60 sec, 72°C 30–45 seconds; followed by a denaturation cycle of 95°C 60 sec, 55°C 30 sec, 95°C 30 sec. Tumor samples containing no integration sites in the region of interest were used as control tumors. Relative expression values (2^-ΔΔCt) were calculated using control tumor 1 as a calibrating sample. All relative expression values were then normalized to set the average of the tumor controls to a value of 1 for each target.

MiRNAs and low molecular weight RNAs were isolated from frozen mouse tumor tissue using the Purelink miRNA Isolation Kit (Invitrogen). The mature species of the miRNAs were measured by RT-qPCR using a stem-loop RT primer specific for each miRNA 33 in the cells listed, in triplicates. Accordingly, 50 ng of each tumor miRNA preparation was reverse transcribed with the SuperScript III First-Strand Synthesis System for RT-PCR using the following stem loop RT primers (50 nM final concentration) 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAGCACT-3'(mmu-mir-1204), 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCTCAAA-3'(mmu-mir-1205), 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACACTTAA-3' (mmu-mir-1206), 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCCCTTC-3'(mmu-mir-1207-5p), 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGAGATG-3'(mmu-mir-1207-3p) and 5'-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCCAGCC-3'(mmu-mir-1208). The reverse transcription reactions were diluted 1:5 and 5 μl of these dilutions were used in the 25 μl qPCR reactions. The annealing step was 50°C for 60s. The qPCR probes and primers were as follows: mmu-mir-1204: 5'-GCGGTGGTGGCCTGCTCT-3', 5'-GTGCAGGGTCCGAGGT-3', 5'-[56-FAM]-CACTGGATACGACAGCACTG-[36-TAMSp]-3'; mmu-mir-1205: 5'-GGCGTCTGCAGGACTGG-3', 5'-GTGCAGGGTCCGAGGT-3', 5'-[56-FAM]-CACTGGATACGACCTCAAAG-[36-TAMSp]-3', mmu-mir-1206: 5'-TTGCGGTATTCACTTGGG-3', 5'-GTGCAGGGTCCGAGGT-3', 5'-[56-FAM]-CACTGGATACGACACTTAAACA-[36-TAMSp]-3'; mmu-mir-1207-5p: 5'-TGCTGGCACGGTGGGTG-3', 5'-GTGCAGGGTCCGAGGT-3', 5'-[56-FAM]-CACTGGATACGACCCCTTCC-[36-TAMSp]-3'; mmu-mir-1207-3p: 5'-TGTCTGTCAGCTGGCCT-3', 5'-GTGCAGGGTCCGAGGT-3', 5'-[56-FAM]-CACTGGATACGACGAGATGA-[36-TAMSp]-3'; mmu-mir-1208: 5'-CCGGTCACTGTTCAGAC-3', 5'-GTGCAGGGTCCGAGGT-3', 5' [56-FAM]-CACTGGATACGACCCAGCCT-[36-TAMSp]-3'.

Synthetic RNA oligos (IDT) were used to generate a calibration curve for each miRNA; for mmu-mir-1204, 5'-UGGUGGCCUGCUCUCAGUGCU-3'; mmu-mir-1205, 5'-UCUGCAGGACUGGCUUUGAG-3'; mmu-mir-1206, 5'-UAUUCACUUGGGUGUUUAAGU-3'; mmu-mir-1207-5p, 5'-UGGCACGGUGGGUGGGAAGGG-3'; mmu-mir-1207-3p, 5'-UCAGCUGGCCUUCAUCUC3'-3'; and mmu-mir-1208, 5'-UCACUGUUCAGACAGGCUGG-3'. Amplification efficiencies of the calibration curves for the 6 mmu-mirs were at least 70%. Concentrations of the mature species were calculated using the calibration curves and then normalized by the average of the control tumors, to calculate relative expression levels.