![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Previous Article | Next Article 
Journal of Virology, September 2005, p. 12035-12044, Vol. 79, No. 18
0022-538X/05/$08.00+0 doi:10.1128/JVI.79.18.12035-12044.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.
University of Pennsylvania School of Medicine, Department of Microbiology, 3610 Hamilton Walk, Philadelphia, Pennsylvania,1 Genomic Analysis Laboratory, The Salk Institute, 10010 North Torrey Pines Rd., La Jolla, California 920372
Received 20 April 2005/ Accepted 27 June 2005
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The distribution of the human ERVs (HERVs) has been studied previously in some detail (1-3, 9, 13, 18). A study of HERV distribution by Arian Smit, carried out when the human genome sequence was about 10% completed, indicated that HERVs accumulated mostly outside of genes (27). For the few HERV sequences within genes, most were in the antisense orientation, which would be expected to minimize disruption of cellular mRNA synthesis because the viral poly(A) addition and splicing signals are in antisense orientation. These findings were consistent with the idea that purifying selection eliminated HERVs that disrupted the function of cellular genes (27). However, another explanation for the distribution, noted by Smit, was that the initial targeting of HERV DNA integration might have accounted for the observed distribution. Initial integration of HERV DNA has not been studied; indeed, no replication-competent HERVs have been identified, making characterization of de novo integration targeting impossible. In chickens, however, there is a large class of endogenous retroviruses, members of the ERVK group, that are related to the replication-competent avian sarcoma-leukosis viruses (ASLV). For this virus group, it is possible to compare the placement of de novo sites of integration generated experimentally to the preexisting distribution of ERVK sequences, allowing the forces dictating the accumulation of ERVs to be specified more precisely.
The availability of draft vertebrate genome sequences has allowed retroviral integration targeting to be assessed in detail by cloning and sequencing large numbers of integration sites and then analyzing their distribution relative to other features mapped on the genome sequences. Surprisingly, studies of integration targeting in human cells have shown that different retroviruses favor integration near quite different chromosomal features. Studies of several thousand sites of human immunodeficiency virus (HIV) integration in human cells indicated that transcription units (TUs) are favored for integration, and comparison to transcriptional profiles from target cells indicates that active TUs are particularly favored (19, 26, 31). Murine leukemia virus (MLV), in contrast, showed an integration bias in favor of CpG islands and transcription start sites (31) and a weaker preference for transcription units. ASLV showed a nearly random distribution of integration sites in the human genome, with TUs favored only weakly (19, 22). A variety of factors have been proposed to influence integration targeting in chromosomes (7-10, 17, 23). The finding that integration targeting by the three retroviruses differed suggests that each type of integration complex may bind specific chromosomal proteins, a model consistent with studies of related retroelements in the yeast Saccharomyces cerevisiae (4, 8, 25, 32, 33).
Here we have analyzed de novo integration targeting for ASLV and HIV in the chicken genome. Six hundred fifty-eight sites of ASLV integration were sequenced and compared to 205 sites of HIV integration. Activity of the chicken TUs was monitored by transcriptional profiling. One reason for analyzing integration in chickens was to investigate possible species-specific factors influencing integration targeting. In particular, the previous finding that integration by ASLV favored TUs only weakly in the human genome could potentially have been explained by a lack of a chicken-specific integration targeting factor. Similarly for HIV, analyzing integration in a distantly related genome allowed us to ask whether species-specific factors were important for targeting to active TUs. We found, however, that the integration target preferences in the chicken genome were generally similar to the patterns in humans for both viruses, indicating that any cellular factors important for targeting are apparently conserved between chickens and humans.
To investigate factors that determine the distribution of ERVs in the chicken genome, we compared the distributions of ERVs to the distribution of de novo ASLV integration sites. Chicken ERVs are enriched in gene-sparse chromosomes and in intergenic regions. For the minority of ERVs present within genes, most were in the antisense orientation; consequently, ERV sequences for RNA splicing or polyadenylation within introns do not interfere with host gene function. In contrast, for sites made by de novo infection with ASLV, integration was favored in TUs, and sites within genes showed no orientation bias. This indicates that the present-day distribution of chicken ERVs is likely dominated by purifying selection after integration. Further analysis of the distribution of ERVs in mice and rats and an updated analysis for humans revealed that ERVs in these organisms showed the same biases as in chickens, suggesting that purifying selection dominates the placement of ERVs in many vertebrates.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Preparation of the HIV-1 and ASLV vector particles.
HIV type 1 (HIV-1)/vesicular stomatitis virus G (VSV-G) vector particles (pseudotyped with VSV-G) were generated in 293T cells by the cotransfection of three plasmids: p156RRLsinPPTCMVGFPWPRE (11) (carrying the HIV vector segment), pCMVdeltaR9 (21) (packaging construct), and pMD.G (21) (carrying the VSV-G gene). Forty-eight hours after transfection, supernatants containing the viral particles were harvested and centrifuged for 5 min at 350 x g at 4°C. Supernatants were then filtered through a 0.45-µm filter and concentrated by centrifugation at 23,000 x g for 2 h at 4°C. The viral pellets were resuspended in 
1/100 volume of fresh medium. The concentration of HIV vector particles in stocks was determined by p24 enzyme-linked immunosorbent assay.
ASLV particles were generated by transfecting DF-1 cells with the plasmid pRCASBP(A)GFP (from Steve Hughes, National Cancer Institute, Frederick, Maryland; see http://home.ncifcrf.gov/hivdrp/RCAS/plasmid.html). Forty-eight hours after transfection, supernatants containing the viral particles were harvested, centrifuged, filtered, and concentrated as for the HIV-1 viral particles.
Virus infections.
Prior to infection, HIV-1/VSV-G preparations were digested with DNase I (0.2 U/µl) for 1 h at 37°C. CEF cells (3 x 106) at 40% confluence were infected with HIV-1/VSV-G (500 ng p24) or an aliquot of ASLV-containing supernatant for 48 h at 37°C with 5% CO2. Flow cytometry analysis showed that the percentages of cells expressing green fluorescent protein were 27% and 69% for HIV-1/VSV-G and ASLV, respectively.
Cloning integration sites from chicken cells. Integration sites were cloned by ligation-mediated PCR essentially as described previously (26). Briefly, DNA was extracted using the DNeasy tissue kit (QIAGEN, Valencia, CA) 48 h postinfection and digested with DpnI to eliminate possible plasmid vector carryover, followed by digestions with AvrII, NheI, and SpeI. Linker DNA was ligated to the digested ends, and products were cleaved with SacI to prevent amplification of an internal fragment. This was followed by DNA amplification with two rounds of PCR, cloning, and sequencing. For a summary of restriction enzymes used in analyzing integration sites, see Table S4 in the supplemental material; for sequences of oligonucleotides used in this work, see Table S5 in the supplemental material.
Microarray analysis. RNA was harvested from CEF cells in log-phase growth. ASLV vector infections were carried out as described above except that the RNA was harvested 24 h after infection. Labeling of RNA was performed as described by Affymetrix (Santa Clara, CA). To study the effect of ASLV infection on gene activity, three chips with independent RNA samples were analyzed for uninfected cells and another three chips were analyzed for independently infected cells. Significance Analysis of Microarray software was used to analyze changes in gene activity after ASLV infection (28) via the permuted t test method. For the ASLV integration sites, analysis using mRNA, Unigene, and Ensembl calls returned 249 probe sets. For HIV, a similar analysis returned 117.
Bioinformatic analysis. Integration site locations were identified using the BLAT feature in the Chicken Genome Browser Gateway (http://genome.ucsc.edu/cgi-bin/hgGateway) against the February 2004 freeze of the chicken genome sequence. Integration site sequences were judged to be of acceptable quality if (i) the match to the genome began within 3 bp of the 5'-CA-3' terminus of the viral DNA, (ii) the match proximal to the long terminal repeat (LTR) end showed an identity of at least 98%, and (iii) this match yielded a unique best hit using default parameters in the client-server BLAT (15) ranking (galGal2 draft of chicken genome; hg17 draft of human genome). An integration site was scored as present in a TU if it lay in DNA between the base pairs encoding the 5' and 3' ends of the mRNA as defined by Ensembl or mRNA annotation (http://genome.ucsc.edu/cgi-bin/hgGateway). Analysis of previously published retroviral integration sites into the human genome was updated using the hg17 draft (May 2004) of the human genome. New matched random controls were also generated.
For the expression analysis in Fig. 3, integration site sequences were accepted as within a TU if the sequence was present in the Ensembl, mRNA, or Unigene catalog. For cases where multiple probe sets queried a single TU, all probe sets were accepted in the analysis. Matched random control sites were handled similarly.
|
An artifact was encountered in analyzing ASLV integration sites in the chicken genome. A portion of infecting retroviral DNA becomes circularized by autointegration or other reactions, and such forms could be cloned by the ligation-mediated PCR method used. These scored as integration sites clustering at chromosome 1 position 29522444 to 29532218, which is an ASLV-related ERVK. These events could be shown unambiguously to be artifacts, since apparent integration target sites matched exactly to the RCAS vector used and differed detectably from the genomic ERVK copy. These sites were removed from the analysis.
Nucleotide sequence accession numbers. Integration site sequences in the chicken genome have been deposited in GenBank under accession numbers CZ905239 to CZ905918 (ASLV integration sites) and CZ905919 to CZ906101 (HIV integration sites).
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cloning of integration sites was carried out using ligation-mediated PCR (26, 31). Genomic DNA was digested with restriction enzymes and ligated to DNA linkers. DNA containing junctions between retroviral DNA and target DNA was amplified by PCR using one primer complementary to the linker and a second complementary to the viral DNA end. A second round of PCR was performed with nested primers. PCR products were then cloned and sequenced, yielding 658 unique ASLV and 205 unique HIV integration sites. Integration sites were mapped on the February 2004 draft chicken genome sequence (Fig. 1), and local features at each integration site were quantified. Integration sites in the chicken genome were compared to integration site data sets for HIV, ASLV, and MLV in the human genome (Table 1). All analyses of the human genome were updated to the hg17 draft (May, 2004).
|
|
An issue in this analysis was the possible bias introduced into experimentally determined integration sites by the reliance on restriction enzyme cleavage during cloning. Analysis indicates that integration sites closer to appropriate restriction sites in genomic DNA are more readily cloned by the ligation-mediated PCR protocol used but that the fraction of sites sampled is reasonably representative of the genome as a whole (C. Berry, J. R. Ecker, and F. D. Bushman, unpublished). We nevertheless modified our statistical procedure to account for restriction site bias by generating a matched set of random control sites. For each experimental integration site, 40 random sites were generated in silico that were constrained to be the same distance away from a restriction site in the genome that could have been used for cloning. For example, if an experimental site was 300 bp from the restriction site used for linker ligation, then 40 sites were generated that each were exactly 300 bp from such a restriction site but randomly scattered around the genome. A few random sites landed in sequence gaps and were removed from subsequent analysis. Statistical analysis was then carried out by comparing the pooled experimental sites to the pooled matched random controls.
The importance of this procedure was tested by comparing the matched random control set to an unmatched collection of random chromosomal sites. This revealed that the matching procedure resulted in slight but significant differences in the detection of several features (see Table S1 in the supplemental material).
Relationship of de novo integration sites to TUs in the chicken genome. ASLV and HIV integration sites were mapped onto the chicken genome and the proximity to TUs quantified (Table 1). A complication arises due to the relatively early stage and incomplete analysis available for TUs in the chicken genome. In an effort to obtain a robust overview, we repeated the analysis using both the Ensembl and "mRNA" catalogs, which are the two most populated collections (Table 1). ASLV showed slightly increased integration in each collection of TUs: 1.24- and 1.06-fold, respectively, relative to the matched random control. The results achieved statistical significance for the mRNA data but not for Ensembl (Table 1). The analysis was then extended by pooling data for mRNA and Ensembl calls. This revealed that 62.2% of ASLV sites were within TUs compared to only 55.2% of matched random sites, a highly significant difference (P = 0.0055). For HIV, the bias was more pronounced (1.72- and 1.56-fold) and the difference highly significant (P < 0.0001 for both mRNA and Ensembl calls analyzed individually). Analysis of the pooled data showed that 88.3% of HIV integration sites were called as in TUs. We conclude that ASLV shows a weak bias and HIV shows a strong bias in favor of integration in TUs in chicken cells.
The relationship of integration sites to TU boundaries was then assessed (Fig. 2). Integration in transcription start regions is of particular interest because MLV integration in human cells strongly favored transcription start regions (31). One motive for studying ASLV integration in chickens was to determine whether ASLV adopted such a target preference in its natural host. However, no significant bias was seen for ASLV integration in extragenic regions extending 5 kb from TU 5' and 3' boundaries, though a slight surplus of integration events in these sequences was seen (Fig. 2A). Similarly for HIV, no significant biases were seen in favor of integration in 5' and 3' flanking regions (Fig. 2B).
|
Integration by MLV (Fig. 2E) showed a much stronger bias for upstream regions than either ASLV or HIV, with about fivefold more integration sites present than expected by chance, paralleling the original report (31). MLV also favored integration in TUs and downstream regions, though the extent of the bias was more modest than for the 5' bias (Fig. 2E).
The frequency of integration near CpG islands was also assessed. CpG islands are often associated with promoter regions and are strongly favored for MLV integration in human cells (31) but disfavored for HIV integration (19) (Fig. 2F). In chicken cells, ASLV showed a slight but significant bias in favor of integration near these features, as seen previously in human cells (19). HIV did not show any significant bias for or against integration in CpG islands in chicken cells (19).
In summary, ASLV integration and HIV integration in chicken and human cells were generally similar with respect to TUs, flanking sites, and CpG islands. ASLV integration in flanking sites was slightly favored in both chickens and humans, but the data only achieved significance in humans. At present we attribute the difference in significance to the early stage of gene annotation of the chicken genome and not any major biological difference. ASLV did not adopt an MLV-like integration pattern in its natural host.
Effect of gene activity on de novo integration in the chicken genome. We next analyzed the relationship between gene activity and retroviral integration in CEF cells. Steady-state levels of mRNA were quantified using Affymetrix chicken genome microarrays, which query the activity of about 33,000 transcripts in the chicken genome. In the following, we used the steady-state mRNA levels as an approximation of gene activity, though we note that differential mRNA processing, turnover, etc., could also affect these measurements. Uninfected cells were compared to cells 24 h after infection with ASLV.
Three arrays were analyzed with labeled RNA from uninfected cells and three with RNA from infected cells. Changes in gene activity due to infection were analyzed using the Significance Analysis of Microarray package (28). Infection was found to have very little influence on cellular gene activity: only 20 genes were called as significantly affected even when a relatively high false-discovery rate was accepted (25%). The most significantly affected gene was GFP, which had a 15.7-fold higher expression value after infection. The GFP gene was introduced into cells by infection with the ASLV vector used and so provides a positive control. See Table S2 in the supplemental material for a summary of further genes affected by infection. We conclude that there was little or no effect of infection with the ASLV vector on CEF cell transcription under the conditions studied. Results for all six arrays were thus pooled to improve the statistical resolution in subsequent analysis.
The expression levels of TUs hosting integration events were then compared to that of the remainder of genes queried on the array (Fig. 3A). TUs hosting integration events for both viruses were more active than the average of TUs on the chip (P = 0.0022 for each; Mann-Whitney test). To examine the relationship between transcriptional activity and integration frequency in more detail, the chicken TUs analyzed by the Affymetrix microarray were divided into classes by expression level and the frequency of genes hosting integration events in each class was determined (Fig. 3B and C). One reason for carrying out such an analysis was to examine whether highly transcribed genes were disfavored for integration, as has been suggested in previous studies of ASLV integration in two model genes in quail cells (17, 30). For both ASLV and HIV, comparison to the matched random control indicated that TUs in the lower expression categories were less favored for integration and TUs in the higher expression classes were more favored. There was no reduction in integration frequency in the highest expression class of genes for either virus; in fact, integration was favored. Evidently, under the conditions of this study, even very-high-level transcription favored integration.
Biased integration at the chromosomal level. An unexpected trend was observed in the distribution of integration sites at the chromosomal level. The chicken genome is composed of macrochromosomes, intermediate chromosomes, and microchromsomes (14) (Fig. 1). The larger macrochromosomes are gene sparse, A/T rich, repeat dense, and CpG island sparse and have genes with longer introns. The smaller microchromosomes are opposite in these tendencies, and the intermediate chromosomes are intermediate. Thus the tendency for ASLV and HIV to favor integration in TUs predicts that the microchromosomes would be favored targets. This was seen for HIV, as expected, but not for ASLV (Fig. 4). Paradoxically, ASLV favored integration in the macrochromosomes and disfavored integration in the microchromosomes. Inspection of Fig. 1 indicates that some of the excess integration sites in macrochromosomes appear to have accumulated in the telomere-proximal regions of the longer chromosome arms. The reason for this bias is unknown.
|
Chicken ERVs consist of two main groups, the ERVKs and ERVLs, and a smaller miscellaneous group designated "ERV." An analysis of the positions of each ERV group relative to TUs showed that each has accumulated primarily outside TUs (Table 2). Only 11.2% of ERVKs and 19% of ERVLs were present in TUs, while 37.1% of de novo ASLV integration events were within TUs (P < 0.0001 versus ERVK or ERVL; chi-square test). The distributions of de novo integration events and ERVs are compared in Fig. 1. The genomic distributions of de novo ASLV integration sites could also be compared directly to the distributions of ERVKs and ERVLs. This showed a detectable anticorrelation for ERVKs and a strong anticorrelation for ERVLs (P = 0.0251 and P = 0.0004, respectively; Pearson correlation).
|
Of the integration events in TUs, most of which are within introns, ERVKs and ERVLs showed a significant bias in the orientation of the proviral genome relative to the host TUs. ERVK and ERVL sequences were much more commonly in the antisense orientation (73% and 56.9%, respectively; Table 3; P < 0.0001, binomial test). In the antisense orientation the ERV sequences are expected to have minimal effects on the host cell TUs, because the proviral signals for RNA splicing and poly(A) addition are in reverse orientation and so nonfunctional. For de novo integration, the orientation of ASLV and HIV sequences within TUs did not show a significant bias (50.0% and 48.8%, respectively, P > 0.05). For the subset of ERVs most closely related to ASLVs, the antisense bias was highly significant (78.3% in the antisense orientation; P < 0.0001, binomial test). These findings support the idea that purifying selection against gene disruption dominates the distribution of the chicken ERVs. Further analysis indicates that ERVs and related LTR retrotransposons in many vertebrates show similar biases in their distributions (discussed below) (2, 3, 18, 27).
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Parallels in integration targeting among data sets. A key question in interpreting the data presented here is that of how widely conclusions can be generalized. Do different cell types show major differences in integration targeting when infected by the same retrovirus? How different is integration targeting among related retroviruses? The available data are sparse, but a few generalizations can be tentatively proposed. For HIV integration, comparison among human cell types shows only slight differences (19, 26, 31). Tissue-specific transcription influences targeting detectably, but only modestly (19). For ASLV, similar patterns of integration targeting have been reported for two human cell types (19, 22). Given the similarity among the targeting patterns in two human cell types and CEF cells, we infer that the integration pattern in chicken germ cells, the cell type relevant to ERV formation, would probably be similar. Another question is that of whether related but nonidentical ASLVs would be expected to have similar patterns of integration. Again the data are sparse, but for lentiviruses it has been shown that simian immunodeficiency virus in rhesus monkey cells has a similar targeting preference to that of HIV in human cells (12), suggesting that integration targeting may be similar across the lentiviral group. We thus propose that ASLV-related ERVs likely had the same integration targeting preferences as the ASLV vector studied here, though more data would be helpful to strengthen this point.
Integration targeting in cells of different species. Do species-specific factors influence integration targeting? In human cells, ASLV showed only a weak preference for integration in TUs: statistically significant biases could be detected, but they were quantitatively modest (Fig. 2C). One possible reason is that human cells lacked a species-specific factor for targeting ASLV integration. According to this idea, a chicken-specific factor might have tethered ASLV integration complexes to chromosomal DNA, resulting in highly selective integration, perhaps resembling MLV in human cells. However, no such bias was detected in de novo integration by ASLV in CEF cells: ASLV integration in chickens resembled ASLV integration in humans (compare Fig. 2A and C). While it is not ruled out that ASLV tethering factors might exist in chicken cell types other than the CEFs studied, our findings do not provide evidence for species-specific factors that more precisely target ASLV integration.
Similarly for HIV, we sought to test whether human-specific factors influenced the distribution of HIV integration sites. However, we found that HIV strongly favored integration in active TUs in chicken cells as well as human cells. These findings indicated that potential cellular factors important for HIV integration targeting are conserved between chickens and humans. This may be technically helpful in studies of HIV integration; for any candidate targeting factors identified in human cells, it may be useful to confirm that the chicken counterpart is also functional.
Integration and transcriptional activity in chicken cells. The relationship between integration and transcriptional activity was assessed using newly available Affymetrix microarrays, which contain 38,392 probe sets querying the activity of about 33,000 chicken TUs. For both ASLV and HIV, the TUs hosting integration events showed a significantly higher median expression level than those targeted in the matched random control, indicating favored integration in active genes (19, 26, 31). The trend in chicken cells may differ slightly from that in human cells (19), since the association with gene activity in chicken cells was somewhat stronger for ASLV and weaker for HIV compared to the same trend in human cells. However, given the much earlier state of assembly and analysis of the chicken genome sequence it is uncertain whether these slight differences are meaningful.
Published studies analyzing ASLV integration frequency in two model genes in quail cells have suggested that induction of transcription reduced integration in these model genes (17, 30). In our study, ASLV integration frequency was significantly increased in the most highly expressed class of TUs (P = 0.0022 for the comparison of ASLV integration sites in Fig. 3B, rightmost bin, to the matched random control; chi-square test). It should be noted, however, that the available data are somewhat sparse: our collection of highly expressed TUs targeted for integration consisted of only 53 probe sets (Fig. 3B, rightmost bin, ASLV data), and the data set described in reference 17 contained 55 experimental and control integration sites. Perhaps the divergent results are explained by sparse sampling or by the use of different cell types or analytical methods in the two studies.
Integration site distribution in the chicken chromosomes. The distribution of ASLV integration sites in the chicken chromosomes was unexpected. It would be expected for both ASLV and HIV that integration would be more frequent in the gene-dense microchromosomes and disfavored in the gene-sparse macrochromosomes, since both viruses favor integration in TUs. The expected favoring of microchromosomes was seen for HIV. However, for unclear reasons, ASLV significantly favored integration in the gene-sparse macrochromosomes. The finding that macrochromosomes are enriched in all types of chicken repeated sequences, including new ASLV integrations, points to a possible factor causing new integration events to favor these larger chromosomes. Perhaps HIV showed the opposite bias because it responded more strongly to active transcription.
Comparing de novo integration and ERVs in the chicken genome. The distribution of de novo ASLV integration sites was quite different from that of related ERVs, suggesting models for the forces determining which ERVs persist in the genome. Both total chicken ERVs and the ASLV-related ERV subset have accumulated outside of TUs, and the minority of ERVs within TUs were typically in antisense orientation relative to host cell transcription. For ERVs in the antisense orientation, the viral splicing and polyadenylation signals do not affect mRNA synthesis by the host gene, thereby minimizing the genetic damage of integration. Comparison of the de novo distribution of ASLV integration sites to related ERVs indicates that the present distribution of chicken ERVs relative to TUs was likely not determined by the initial integration targeting, but rather by selective pressures against gene disruption. This trend has been noted previously for human ERVs (27), but the analysis of ASLV presented here provides the first case where the de novo pattern of integration was experimentally determined, allowing the observed biases in ERV distribution to be attributed to forces acting after integration.
Forces dictating the placement of ERV sequences in vertebrate genomes. Tables 2 and 3 present a comparison of ERVs in chickens to those in rats (24), mice (20), and humans (16, 29). Earlier analyses of ERVs in the human genome (18, 27) are updated here to the most recent genome draft (hg17). For some of the annotated ERVs, it is not clear whether they have replicated as retroviruses or LTR retrotransposons (7, 9); our analysis did not attempt to distinguish between these possibilities. For all the vertebrates examined, most ERV families show a highly significant bias in favor of accumulation outside of TUs (Table 2). For the minority of ERVs within TUs, most are present in the antisense orientation relative to host gene transcription (Table 3).
In mice and rats, a small group of ERV sequences, designated LTR55 and MER95, did not show an orientation bias when present within TUs (Table 3). An analysis of these ERV sequences showed them to be quite short, apparently representing single LTRs. Such solo LTRs are known to be formed from integrated retroviruses by intrachromosomal recombination between LTRs (9). An analysis of the LTR55 and MER95 sequences present in the sense orientation in TUs showed that none contained the 5'-AATAAA-3' sequence that directs poly(A) addition. Retroviral LTRs typically also lack splicing signals. Thus the LTR55 and MER95 sequences in sense orientation would not be expected to affect cellular mRNA synthesis, likely resulting in a lack of purifying selection and thereby explaining the lack of orientation bias.
Thus the analysis of the LTR55 and MER95 sequences reinforces the idea that selection acts to remove genes that are impaired for mRNA synthesis due to retroviral DNA integration. Overall, our findings indicate that selection against gene disruption has been the major force dictating the distribution of ERV sequences in chickens and that similar biases in ERV distribution were found in three other vertebrates.
|
ACKNOWLEDGMENTS |
|---|
This work was supported by NIH grants AI52845 and AI34786, the James B. Pendleton Charitable Trust, and Robin and Frederic Withington (to F.D.B.) and the Fritz B. Burns Foundation (to J.R.E.). S.B. was supported by a fellowship from the Alberta Heritage Foundation for Medical Research and the Natural Sciences and Engineering Research Council of Canada.
|
FOOTNOTES |
|---|
Supplemental material for this article may be found at http://jvi.asm.org/.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| J. Bacteriol. | Mol. Cell. Biol. | Microbiol. Mol. Biol. Rev. |
|---|
| Clin. Vaccine Immunol. | ALL ASM JOURNALS |
|---|
