Antisense Transcription in Gammaretroviruses as a Mechanism of Insertional Activation of Host Genes^▿

Isolation of RNA.

Total RNA from frozen tumor samples and from cell line cultures was isolated using TRIzol Reagent (Invitrogen) according to the recommendations of the manufacturer. RNA was quantified spectrophotometrically and visualized by agarose gel electrophoresis to ensure high quality.

Reverse transcription-PCR (RT-PCR), quantitative real-time PCR (qRT-PCR), and 5′ rapid amplification of cDNA ends (RACE).

For RT-PCR analyses, 1 or 2 μg of total RNA was reverse transcribed into cDNA using a RevertAid H Minus First Strand cDNA synthesis kit (Fermentas). All PCR analyses were carried out in ABI 2720 thermocyclers (Applied Biosystems) using recombinant Taq polymerase (Invitrogen) with the following program: an initial melting step at 95°C for 5 min; 30 to 35 cycles of amplification, each consisting of a melting step at 95°C for 30 s, annealing at 55 to 62°C for 30 s, and elongation at 72°C for 30 s; and a final elongation step at 72°C for 7 min. The annealing temperature for amplicon 1f-3-4 was 62°C, that for Bach2 antisense U3 (here referred to as asU3) screening was 55°C, and that for canonical Nras RT-PCR was 60°C, while the remaining RT-PCRs were done at 58°C. After a single round of Bach2 asU3 RT-PCR, faint bands appeared, which were subsequently amplified using 1 μl PCR product in a seminested PCR using identical conditions except for only 30 cycles. PCR amplification products from asU3 PCRs were cloned into pCR4-TOPO vector (Invitrogen), and the obtained bacterial colonies were screened by PCR. PCR products were sequenced by using an ABI 7300 Biosystems apparatus with BigDye 3.1 (Applied Biosystems). The following primers were utilized: 47, 5′-CTAGACGAGGAAGAAGAGCGAAG-3′; 51, 5′-GAGGTGAAACTGGGCAAGAG-3′; 57, 5′-gCGGTTGAGCATCAGGATAA-3′; 78, 5′-GTTGGGGCCTCTTGCCCAGTTTCAC-3′; 140, 5′-GTTAGCTGGCTAAGCCTTATGA-3′; 164, 5′-TGTGTTGAAAGGATCTGTCAAGCT-3′; 2620, 5′-GAATTCGATATCGATCCCCGGTCATCTGGG-3′; actb1, 5′-ACACAGTGCTGTCTGGTGGT-3′; actb2, 5′-CTGGAAGGTGGACAGTGAGG-3′; 955, 5′-TGGACACTCACTGATCTATCACAA-3′; 369, 5′-CCAGGCTTCTCATCCACAGA-3′; 705, 5′-ACTGTAGCAGTGGCCCAAAG-3′; 849, 5′-TACAAACTGGTGGTGGTTGGAGCA-3′; 970, 5′-GCAGGTCTCACCATCAATCACCACTTG-3′; and 332, 5′-ACTGGTCTCTCATGGCACTGTACT-3′.

qRT-PCR was done using SYBR green fluorescence (Brilliant II SYBR green; Stratagene) with an Mx3000 apparatus (Stratagene) following the manufacturer's recommendations. All reactions were run in triplicate and analyzed using MxPro 4.1 software (Stratagene). Relative quantification was performed for each amplicon by relating to a 10-fold dilution series of a plasmid cloned with the particular amplicon, which was run simultaneously with the tumor samples. Actb signal was used to normalize total Jdp2 signal. Primers for amplicon Actb (actb1 and actb2), total Jdp2 exon 3-4 (47 and 57), total exon 1f-3-4 (164 and 57), and exon asU3-3-4 (140 and 57) are described above. qRT-PCR amplification conditions consisted of an initial melting step of 95°C for 10 min, and then amplification was conducted through 50 repetitions of a 95°C step for 15 s and a combined annealing and extension step at 60°C for 30 s. In the case of amplicon E1f-E4, the annealing/extension step was done at 62°C.

The 5′RACE analyses were performed using a GeneRacer kit (Invitrogen) for the Nras and a Smart RACE cDNA amplification kit (Clontech) for the Jdp2 according to the manufacturers' recommendations and as described previously (25, 38). The 5′ RACE PCR was done with a touchdown program consisting of an initial melting step at 95°C for 5 min and then six cycles of touchdown in which the annealing temperature was decreased 1°C/cycle, from 74°C to 68°C. This was followed by 30 cycles of amplification with an annealing temperature of 67°C and a final elongation step for 10 min at 72°C. Each touchdown and amplification cycle consisted of a 30-s melting step, a 30-s annealing step, and a 30-s elongation step. Gene-specific reverse primers were located in Jdp2 exon 3 (5′-GCATCTGGCTGCAGCGACTTTGT-3′) and Nras exon 3 (5′-GCACTGTACTCCTCTTGTCCAGCT-3′).

Generation of Nras LTR knock-in mice.

A targeting cassette consisting of the loxP-flanked PGK/Tn5 neomycin selection cassette and the Akv1-99 MLV proviral LTR was inserted by homologous recombination in CJ7 embryonic stem (ES) cells (48) at a specific position in intron 1 of the Nras gene. This position has previously been identified as an Akv1-99 proviral integration site involved in B-cell tumorigenesis (tumor 9 in reference 27). ES cells containing the LTR at this position (in either the same or the opposite transcriptional orientation compared to Nras) were generated and injected into blastocysts which, upon surgical implantation into pseudopregnant mice, developed into chimeric animals. Chimeras were then mated with a wild-type background until heterozygous knock-in offspring were identified. Homozygotes could be obtained at the expected frequency and presented a normal phenotype at birth. At the time of termination (30 to 40 days), no gross enlargement of lymphoid organs could be observed. The PGK/Tn5 neomycin selection cassette was removed by crossing the mice with EIIa-Cre transgenic mice (21).

Northern blotting.

NIH 3T3 cells were either not infected or infected with replication-competent SL3-3 or Akv1-99 (EGFP) MLVs and cultured for 2 weeks. Total RNA was isolated as described above, and 10 μg was subjected to Northern blot analyses using capillary transfer under the conditions described in reference 44. We used oligonucleotides specific for either antisense (5′-TTCATAAGGCTTAGCCAGCTAACTGCAG-3′) or sense (5′-AAATTTGAAACTGTTGTTGTTTTAGCTATTTCTGGGG-3′) transcription within the LTR that were end-labeled with [γ-³²P]ATP using T4 polynucleotide kinase (New England BioLabs). Residual nucleotides were removed with Micro Bio-Spin 6 columns (Bio-Rad) before being used for hybridization with buffers and conditions described in reference 44.

Detection of antisense viral RNA in SL3-3-induced T-cell lymphomas in mice.

In an effort to identify novel promoters immediately downstream of a common integration site in the gene encoding the AP-1 repressor c-jun dimerization protein (Jdp2), 5′ RACE analyses were performed with total tumor RNA from a murine thymic T-cell lymphoma induced by SL3-3 MLV (38). Surprisingly, we found several chimeric transcripts initiated at the minus strand of the U3 region of a provirus situated in the second intron of Jdp2 in opposite transcriptional orientation (Fig. 1A, integration cluster D). The majority (14 out of 20; 70%) of these transcription start sites (TSS) were within two narrow regions on the negative strand of the viral 5′ LTR (positions −53 to −60 and −95 to −102 with respect to the 5′ LTR U3/host junction) (Fig. 1B). All transcripts continued across the virus-host junction into Jdp2 intron 2 and were spliced from a consensus splice donor site to the splice acceptor site of exon 3. The splice donor site corresponds to the previously reported splice site of the alternative exon 1f (38). Thus, the identified sequences are genuine transcripts and not amplification of contaminating DNA.

The occurrence of antisense U3 (asU3) transcription could be a special case related to this particular integration event, for instance, as a consequence of aberrant levels of transcription initiation and splicing factors in the tumor sample or mutations within the proviral LTR. We therefore isolated total RNA from a range of tumors with integrations into intron 2 of Jdp2 and performed RT-PCR with viral antisense primer 140 and Jdp2 exon 4 reverse primer 57 (Fig. 1C). PCR products were cloned and amplified in bacteria before being sequenced. We found that chimeric asU3 transcription between viral and Jdp2 exon 3-4 is a common phenomenon in this tumor panel and that in many tumors several distinct integration sites could be found (e.g., tumor 6 in Fig. 1C). Of 25 different tumors, asU3 transcription was observed in 24, with an average of 2.04 independent asU3-Jdp2 mRNA-generating integrations per tumor (M. H. Rasmussen and F. S. Pedersen, unpublished data). Although we isolated transcripts that utilize other splice donor sites in the region, which in all cases splice to the acceptor site of exon 3, splicing from exon 1f was predominant. Also, we never found asU3-Jdp2 transcripts that did not splice in intron 2. In the majority of the tumors, canonical Jdp2 was easily detected (Fig. 1C). We note, however, that in some tumors (tumor 1, 4, 6, 7, and 8), an inverse correlation between canonical and asU3-mediated Jdp2 was seen. The variations in asU3-Jdp2 and canonic Jdp2 were not related to unequal template, as evidenced by similar Actb signal intensities across samples.

In most cases, the integrated viruses were positioned within 1,000 bases from exon 3; hence, it was formally possible that unique cis-acting elements in this local region allowed for asU3 transcription. We therefore isolated total RNA from tumors in which proviral integrations were identified upstream of the RefSeq Jdp2 exon 1 (Fig. 1A, cluster B and cluster C) (37, 38) and screened for asU3-Jdp2 chimeric transcripts by the same RT-PCR strategy as used above but with the Jdp2 exon 2-specific reverse primer 78 situated more than 15 kilobases upstream of exon 3 (Fig. 1D). Among the eight analyzed tumors, a single occurrence of robust asU3 transcription was found. In this case, the integration was situated outside the Jdp2 gene, and asU3 transcription continued into the Jdp2 upstream region and spliced to exon 2 using a cryptic splice donor site. Thus, we conclude that antisense transcription is not unique to a particular tumor sample, nor is it confined to a small genomic region.

Antisense transcription in B-cell lymphomas induced by Akv MLV.

In B-cell lymphomas induced by Akv MLV, the BTB and CNC homology 2 gene Bach2 is a common target of proviral integrations (25). Similar to Jdp2 in T-cell lymphomas induced by SL3-3 MLV, intragenic integrations in Bach2 are mostly inserted in the opposite transcriptional orientation compared to Bach2 (25) (Fig. 1E). To investigate if this locus is also subject to asU3 chimeric transcription, total RNA was isolated from a range of tumors with intragenic Bach2 integrations. By RT-PCR using asU3 primer 140 and the reverse primer 705 in Bach2 exon 5, we could detect in three tumors transcripts generated by the negative strand of Akv MLV, intron 3 sequence, and splicing from a cryptic intron 3 splice donor site (SD′) to the splice acceptor site of Bach2 exon 4 (two cases are depicted in Fig. 1E). Interestingly, in line with what was observed for Jdp2, reduced canonical Bach2 mRNA levels were detected for one tumor with asU3-Bach2 (tumor 4). These results show that antisense transcription is not specific for SL3-3 MLV or T-cell lymphomas.

Antisense U3 transcripts contribute considerably to tumor mRNA levels.

Increased levels of truncated forms of Jdp2, such as those arising from exon 1f-3-4 transcripts, are frequently associated with T-cell lymphomas induced by MLV (38, 47), strongly implying a gain-of-function of truncated Jdp2 mRNA in tumorigenesis. Our data imply that the provirus-induced increment of truncated Jdp2 transcripts in cells harboring a cluster D provirus (in opposite transcriptional orientation) is a sum of two distinct mRNA pools: (i) chimeric transcripts initiating on the negative strand of U3 in the 5′ LTR; and (ii) transcripts initiating from the intronic cellular promoter under more or less influence of proviral cis elements—so-called enhancer activation. In order to estimate the contribution of asU3-1f-3-4 to total Jdp2 1f-3-4 mRNA levels, we employed a semiquantitative RT-PCR strategy (Fig. 2A). Total 1f-3-4 mRNA was measured with forward primer 164 specific for exon 1f-3 splicing together with exon 4 reverse primer 57 (38). Antisense transcription was measured with asU3-specific primer 140 and primer 57. We carefully selected tumors in which we could not detect usage of intron 2 splice events other than exon 1f-3. Furthermore, in order to ensure similar amplification efficiencies, the integration positions were all within 40 bases of the exon 1f splice donor site. The PCR signal from the tumors was compared to the signal from a dilution series of a plasmid containing an asU3-1f-3-4 sequence (Fig. 2B, lanes 1 to 4). Without any identifiable integration in cluster D, no asU3 signal was observed (tumor 1, lanes 6 and 7). In this tumor, most of the signal of total 1f-3-4 mRNA probably stems from the cellular enhancer/promoter. Canonical Jdp2 mRNA was also detected. Conversely, in two tumors expressing large amounts of exon 1f-3-4 (tumors 2 and 3, lanes 8 to 11), the signal from asU3 transcription is not insignificant. For instance, the asU3-1f-3-4 amplification product in tumor 2 (both dilutions) corresponds to the amplification signal generated from between 10,000 and 100,000 plasmid copies, which is comparable to, albeit seemingly a little lower than, the amplification signal for total 1f-3-4. The double band in tumor 3 is due to a second provirus that also utilizes exon 1f, as identified by sequencing of the PCR products. The last tumor (tumor 4, lanes 12 and 13) represents a situation in which the contribution from asU3 initiation appears less predominant. While total Jdp2 (as sampled with amplicon 47+57 across the common exon 3-4) varied little between the tumors, we again observed (for tumors 3 and 4) a tendency toward a negative correlation between asU3-mediated transcription and levels of canonical Jdp2 (compare amplicons 140+57 and 51+57, lanes 10 to 13).

To measure the contribution of asU3-mediated transcription more accurately, we switched to a quantitative real-time PCR strategy using amplicons similar to those in Fig. 2B. In this assay (Fig. 2C), the contribution from asU3-mediated transcription to total 1f-3-4 mRNA was found to be around 50% in some tumors (tumors 2, 3, and 5), while in others (tumors 6 and 7), the contribution was below 5%. We did not observe any correlation across the samples between asU3-mediated transcription and total Jdp2 mRNA levels.

In conclusion, these data are consistent with the notion that asU3-mediated transcription levels in some tumors are comparable to the level mediated by the “enhancer-activated” cellular promoter.

The proviral long terminal repeat alone mediates antisense transcription in vivo.

The results obtained from tumor tissue as described above indicated that asU3 transcription initiated primarily within approximately 100 bp on the negative strand of the proviral 5′ LTR. In order to independently test if the LTR alone is responsible for asU3 transcription, knock-in mice homozygous for a neomycin-LTR cassette in the proto-oncogene Nras were generated (Fig. 3A). The LTR sequence was derived from Akv1-99 and inserted by homologous recombination in either sense or antisense orientation into a position corresponding exactly to that of a previously reported Akv1-99 MLV integration, which was shown to activate Nras by promoter insertion in B-cell lymphomas in mice (27). In this knock-in model, the LTR retains functional activity in adult mice, since Nras expression in different organs correlates with the tissue specificity of the Akv1-99 LTR. For the insertion in sense orientation, Nras expression was found to be high in the spleen relative to the level in wt mice but not in the thymus. Moreover, the presence of chimeric RNA species could be readily detected (B. Ballarín-González, L. B. Lassen, A. Füchtbauer, E.-M. Füchtbauer, and F. S. Pedersen, unpublished data). For studying asU3 transcription, this model has several advantages over transient reporter-based in vitro protocols. First, tumorigenic selection for integrations into Nras seems to favors a promoter activation mechanism since all identified integrations are in the same transcriptional orientation as Nras (data from Retrovirus Tagged Cancer Gene Database, http://rtcgd.abcc.ncifcrf.gov/) (1, 26, 43). Thus, by inverting the LTR orientation at this position, asU3 transcription could be studied at a locus known to be activated by MLV using a mechanism different from that for asU3 transcription. Second, since the LTR will be embedded into chromatin, the model better recapitulates a natural setting than a transient reporter assay. Finally, while we cannot exclude the possibility that asU3 transcription observed in tumors was dependent on secondary mutations in, for instance, genes involved in transcript processing pathways, such mutations are not expected to be selected for in this nontumorigenic model.

Total RNA was extracted from spleens from two mice homozygous for a LTR knock-in inserted in the same transcriptional orientation as Nras (corresponding to “promoter activation”) and two mice homozygous for a LTR knock-in inserted transcriptionally opposite to that of Nras (“enhancer activation”). All four mice expressed total Nras RNA at detectable levels (Fig. 3B, amplicon 849+332). As expected, RT-PCR using asU3 primer 140 and Nras exon 3 reverse primer 970 did not amplify asU3 transcripts in mice with a LTR knock-in in the same transcriptional orientation as Nras (amplicon 140+970, lanes 1 and 2). However, in mice with the LTR in the opposite transcriptional orientation, a robust level of asU3-Nras chimeric transcripts could be seen (lanes 3 and 4). These RT-PCR products were cloned and sequenced. Analogous to what was observed for the Jdp2 and Bach2 loci, a chimeric mRNA from a donor splice site in intron 1 to the acceptor site of exon 2 was detected (Fig. 3A and B, SD+34). However, the most predominant amplification product corresponded to a read-through transcript in which the intronic splice donor is obviated and the last 73 bp of intron 1 are included (Fig. 3A and B, R.T.). When we used an asU3 primer, oligonucleotide 2620, located further into the U3 region, two additional viral cryptic splice donor sites, SD-130 and SD-73, were identified by bulk cloning and sequencing of PCR products (amplicon 2620+970). Notably, the SD-130 transcript was much more frequent among transformants than was the SD-73 transcript. To further delineate the nature of the asU3 transcripts, 5′ RACE was done with spleen tissue, and the identified TSS tags were mapped onto the virus LTR sequence (Fig. 3C). Interestingly, we found a cluster of tags close to the positions also found in T-cell lymphoma tissue (Fig. 1C). In addition, another cluster of initiation sites, approximately at positions −240 to −250 bp, was identified (Fig. 3C). Furthermore, we note that all TSS tags mapping upstream of the −130 splice donor site utilized this splice site, whereas all read-through transcript and transcripts splicing from the intron (R.T. and SD+34 transcripts, respectively) were found to initiate after the SD-73 splice donor. This result is in agreement with the RT-PCR data in which the most commonly spliced transcript was spliced at position −130. However, the fact that the read-through and SD-73 transcripts also were observed in the RT-PCR analysis indicates that the 5′ RACE did not reach saturation.

To rule out the possibility that the results of the RT-PCR in Fig. 3B were influenced by the presence of the PGK/Tn5 neomycin selection marker, we crossed the mice with Cre-expressing mice in order to excise the loxP-flanked PGK/Tn5 neomycin cassette. We then repeated the RT-PCR on spleen total RNA from these mice (Fig. 3D). Again, we found a robust level of asU3-Nras mRNAs species with structures similar to those in Fig. 3B. The results are in agreement with the hypothesis that asU3 transcription is dependent on the LTR alone. Furthermore, the data suggest that asU3 transcription is a general phenomenon of the LTR of SL3-3 and Akv gammaretroviruses.

Since antisense transcripts initiating in the 3′ LTR and continuing into the env gene are known from complex retroviruses, we wanted to investigate this possibility for MLV. For this purpose, NIH 3T3 cells were either not infected or infected with SL3-3- or Akv1-99-based gammaretroviruses. Total RNA was extracted and subjected to Northern blotting using radioactively labeled oligonucleotides specific for sense LTR transcripts or putative antisense LTR transcripts. However, while we could readily detect spliced and unspliced RNA of positive polarity, we did not find evidence of antisense transcription (Fig. 4). This indicates that the level of putative U3-initiated antisense transcripts in chronically infected cultures is several orders of magnitude lower than that of the plus-strand RNA and is below the sensitivity of the Northern blot.