Identification of 14 Novel EBV miRNAs Encoded within the Viral BART Gene
Previously, we reported the cloning and analysis of a set of 11 miRNAs encoded by the pathogenic human herpesvirus KSHV that are expressed in the PEL cell line BC-1, which is latently infected by KSHV [12
]. BC-1 cells are also latently infected by a wild-type strain of EBV, and we also cloned 222 cDNAs representing EBV miRNAs, out of the 557 cDNA clones of cellular and viral miRNAs obtained in total. These 222 EBV miRNAs consisted of 15 distinct sequences () that derived from 13 different predicted primary miRNA stem-loop precursors (Figure S1
). Remarkably, only one of these 15 miRNAs was essentially identical in sequence to one of the five EBV-encoded miRNAs previously reported by Pfeffer et al. [11
]. Specifically, miR-BART1-5p is identical to the previously reported miR-BART1 miRNA except that it is 2 to 3 nt longer at the 3′ end. This difference may be real or, alternately, the previously reported miR-BART1 cDNA, which was only cloned once, may have suffered a 3′ deletion during cloning. In total, these data indicate that wild-type EBV actually encodes at least 17 different viral miRNAs.
Sequence and Genomic Location of EBV miRNAs Cloned from the BC-1 Cell Line
The previous report on EBV miRNAs by Pfeffer et al. [11
] isolated and cloned miRNAs from the human BL cell line BL41/95. BL41/95 cells are infected with the EBV B95–8 laboratory isolate that, when compared to wild-type EBV, suffers from an approximately12-kb deletion that removes a large part of the EBV BART
]. As shown in , and implied by the miRNA names listed in , all of the novel EBV miRNAs identified in this report are derived from a miRNA cluster located within the predicted introns of the BART
gene, as previously also proposed for miR-BART1 and miR-BART2 [11
]. Moreover, miR-BART5 to miR-BART14 are all located within the region deleted in the B95–8 EBV strain, thus explaining their lack of detection by Pfeffer et al. [11
]. In contrast, miR-BART3 and miR-BART4 are still present in B95–8 and are in fact located within the same predicted BART
gene intron as miR-BART1 (A), and hence are presumably coexpressed. However, since miR-BART1 was only cloned once [11
], it seems possible that these two viral miRNAs were simply missed due to their low expression (see below).
Genomic Location of Selected EBV and rLCV miRNAs
EBV miRNAs Are Differentially Expressed in Latently EBV-Infected Cells
In contrast to the single miR-BART1 cDNA obtained, Pfeffer et al. [11
] cloned several copies of three other EBV miRNAs derived from a second, distinct miRNA cluster adjacent to the BHRF1
gene. Specifically, these workers cloned miR-BHRF1–1 twice, miR-BHRF1–2 fifty times and miR-BHRF1–3 twenty-three times. Using PCR analysis, we confirmed that the BHRF1
gene was intact in latently EBV-infected BC-1 cells (unpublished data), and the reason for our inability to clone any of the three previously reported EBV BHRF1 miRNAs was therefore unclear. To address this issue, we analyzed the expression of a selection of EBV miRNAs and mRNAs in the PEL cell line BC-1; in the wild-type EBV-infected BL cell lines Raji, MUTU, Jijoye, and Namalwa; in the EBV strain B95–8-infected BL cell line BL41/95 (used as a source of EBV miRNAs by Pfeffer et al. [11
]); in the LCL IM-9; in the NPC cell line C666–1; and finally in the NPC tumor C15, which is passaged as a xenograft in nude mice [19
]. BCBL-1, a PEL cell line not infected by EBV, was used as a negative control. This analysis included several BL cell lines that have been extensively passed in culture—i.e., Raji, Jijoye, Namalwa, and MUTU III—all of which represent EBV stage III latency. We also analyzed one BL cell line, MUTU I, that is an early passage variant of MUTU III and has been shown to be in stage I latency [20
]. All these EBV-infected cell lines contain type I EBV except for Jijoye, which is infected with a type II EBV [21
As noted above, miRNA analysis using RNA derived from the latently EBV-infected cell line BC-1 resulted in the cloning of the BART miRNAs shown in A but failed to identify EBV miRNAs derived from the distinct BHRF1 cluster. Consistent with this cloning result, we observed readily detectable levels of the mature miR-BART1-3p, miR-BART3-3p, miR-BART5, miR-BART7, miR-BART10, and miR-BART12 miRNAs upon Northern analysis of RNA prepared from BC-1 cells, but failed to detect either miR-BHRF1–1 or miR-BHRF1–2 (A, lane 2). In contrast, Northern analysis of RNA obtained from the BL41/95 cell line (A, lane 4) revealed high-level expression of miR-BHRF1–1 and miR-BHRF1–2 but little or no detectable expression of any of the miRNAs encoded by the EBV BART miRNA cluster. While this was predicted for miR-BART5, BART7, BART10, and BART12, all of which are deleted in EBV strain B95–8 (A), it was also true for miR-BART1-3p and miR-BART3-3p, both of which are still present in the B95–8 strain and both of which were readily detectable in BC-1 cells.
Analysis of EBV miRNA and mRNA Expression in Tumor-Derived Cells
Analysis of RNA obtained from the BL, LCL, and NPC samples, all of which are infected with wild-type EBV strains, indicated that the LCL and the four BL cell lines largely shared the miRNA expression pattern seen in BL41/95, while the two NPC samples revealed an exaggerated form of the miRNA expression pattern seen in BC-1. Specifically, very high-level expression of all EBV miRNAs derived from the BART miRNA cluster was observed in the NPC samples without detectable expression of the miRNAs encoded within the BHRF1 cluster (A, lanes 5 and 6). Conversely, the IM9 LCL and the BL cell lines Raji, MUTU III, and Namalwa had readily detectable expression of miR-BHRF1–1 and miR-BHRF1–2 (A, lanes 3, 7, 9, and 11), yet very low to almost undetectable expression of the viral BART miRNAs, even though the BART gene is intact in all these cell lines. A different miRNA expression pattern was noted in MUTU I, which did not express readily detectable levels of any viral miRNA, and in Jijoye, which expressed high levels of miR-BHRF1–2, no detectable miR-BHRF1–1, and intermediate levels of all the BART miRNAs analyzed (A).
To extend this analysis to additional EBV-infected cell lines, we also performed Northern analyses for miR-BART1-3p, miR-BART3-3p, miR-BART7, miR-BART10, miR-BHRF1–1, and BHRF1–2, using RNA derived from the BL cell lines P3HR-1 and Daudi, the LCLs HMy2.CIR and HCC1739 BL, and the PEL cell line JSC-1. P3HR-1 is a subclone of Jijoye containing an EBV that has lost a segment of the viral genome including the EBNA-2
gene but that should retain all the EBV miRNAs [22
]. While HMy2.CIR is a spontaneous LCL, HCC1739 BL is derived by infection of B cells with the EBV B95–8 laboratory strain, which as noted above is deleted for miR-BART5 through miR-BART14 (). Analysis of the viral miRNA expression pattern showed that JSC-1 is similar to BC-1, the other PEL cell line tested, in that it expressed the viral BART miRNAs but did not express detectable BHRF1 miRNAs (Figure S2
). The LCLs HMy2.CIR and HCC1739 BL were similar to the IM-9 LCL in that they both expressed readily detectable levels of miR-BHRF1–1 and BHRF1–2 but little or no viral BART miRNAs. Daudi was similar to the other BL cell lines examined in , in that it expressed readily detectable levels of the viral BHRF1–1 and BHRF1–2 miRNAs but only low levels of the five BART miRNAs analyzed (Figure S2
). Finally, P3HR-1, a BL cell line that derives from Jiyoye, shared with Jijoye the property of expressing high levels of miR-BHRF1–2 but no detectable miR-BHRF1–1. However, P3HR-1 differed from Jijoye, and was more similar to the other BL cell lines examined, in that the various viral BART miRNAs were not detectable (Figure S2
The discordant expression of the miR-BHRF1–1 and miR-BHRF1–2 miRNAs observed in Jijoye and its subclone P3HR-1 was unexpected, given that these miRNAs are located close to each another and show concordant expression in all other EBV-infected cell lines analyzed (A and S2). We therefore used PCR to clone and sequence a 418-bp region flanking the miR-BHRF1–1
sequence from the EBV genome present in Jijoye cells (Figure S3
A). This analysis revealed only two single nucleotide sequence differences, neither of which was located in the mature miR-BHRF1–1
sequence. However, one sequence change maps to the miR-BHRF1–1
passenger strand and this G to C change is predicted to disrupt a G–C basepair located within the primary miRNA hairpin, thereby generating a 4-nt symmetrical bulge (Figure S3
B). Inspection of a wide range of vertebrate miRNA precursors failed to identify any stems containing two symmetrical 4-nt bulges within the miRNA duplex region (unpublished data), and we therefore hypothesize that this 1-nt mutation is disrupting the appropriate processing of the primary miR-BHRF1–1
transcript in Jijoye and P3HR-1 cells. Whether this mutation is restricted to the EBV isolate present in Jijoye and P3HR-1, or is a general characteristic of EBV type II strains, remains to be established.
Transcriptional Origin of the EBV miRNAs
The 14 EBV miRNAs encoded within the viral BART
gene cluster are all located within the predicted introns of this alternatively spliced gene (A), and one would therefore predict that the BART
miRNAs would be coordinately expressed and that the total level of expression of the various alternatively spliced BART
mRNAs would correlate with the expression level of the BART
miRNAs. The miRNA expression data presented in A and S2 strongly support the hypothesis that the BART
miRNAs are indeed coordinately expressed. To examine whether BART
miRNA expression indeed correlates with BART
mRNA expression, we performed a Northern analysis using the same RNA samples analyzed for miRNA expression in A but this time using a probe specific for the invariant exon 7 found in all spliced BART
mRNAs. As shown in B, we indeed observed high-level expression of BART
mRNA in the NPC cell samples (lanes 6 and 7), an intermediate level of BART
mRNA expression in Jijoye and BC-1 cells (lanes 2 and 11), and low to almost undetectable expression in the other BL cell lines and in the LCL tested. The BART
gene has previously been shown to give rise to several alternatively spliced mRNA variants, with a major species at approximately 4.8 kb and a more minor species at approximately 6.2 kb [23
], and the data presented in B are consistent with this prediction. These data also confirm the previous observation [23
] that BART
mRNAs are expressed at high levels in NPC cells and at much lower levels in most BL cells, with Jijoye an obvious exception. More important, these data largely confirm the hypothesis that the BART
mRNA expression pattern (B) is predictive of the expression pattern of the entire EBV BART
miRNA cluster (A). We do not currently understand why the BART miRNA expression levels seen in BC-1 and Jijoye are comparable, yet Jijoye expresses a significantly higher level of BART mRNA (A and B), although we hypothesize that this may reflect less efficient miRNA processing in Jijoye cells.
Previously, Pfeffer et al. [11
], who first identified the EBV BHRF1
miRNA cluster, proposed that these three miRNAs might be coexpressed with mRNAs encoding the BHRF1
gene product. However, this appears unlikely as BHRF1
is thought to be first expressed early during lytic replication [28
], and the hairpin precursor for miR-BHRF1–1 appears to be located 5′ to the cap site for the promoter that drives lytic BHRF1
mRNA expression. An alternative hypothesis is that the BHRF1
miRNA cluster is actually processed out of the BamHIH intron present in the very long pre-mRNAs that initiate at the viral Cp and Wp promoters and, when processed, are translated to give rise to the viral EBNA proteins. Transcription from Cp and Wp is characteristic of type III EBV latency [29
], while EBV-infected cells that are undergoing stage I or II latency instead use the Qp promoter to express EBNA1 [31
]. Because the Qp promoter differs from the Cp and Wp promoters in being located between the BHRF1
open reading frames, viral pre-mRNAs initiating at Qp could not be processed to yield any of the BHRF1
To test whether expression of the BHRF1
miRNA cluster indeed correlates with the activity of the Cp and/or Wp promoters, we performed RT-PCR using previously described primers [33
] specific for RNAs initiating at Wp, Cp, or Qp. As shown in C, we saw readily detectable levels of transcription from Cp in Namalwa, Jijoye and MUTU III, with weaker activity in Raji and MUTU I. The Wp promoter was active in Jijoye, Namalwa, IM-9, and BL41/95. Finally the Qp promoter, which is characteristic of stage I or stage II latency, was active in MUTU I, C666–1, C15, BC-1 and, less strongly, in Raji. Therefore, these data show that the cells predicted to be undergoing EBV stage III latency (Raji, BL41/95, IM-9, MUTU III, Jijoye, and Namalwa) all utilize the Cp and/or Wp promoters, while the cells predicted to be undergoing stage I or stage II latency (MUTU I, C666–1, BC-1, and C15) all utilize the Qp promoter to express EBNA1. The observation that Raji is weakly positive for Qp function, while MUTU I is weakly positive for Cp function, likely implies that these BL cell lines are actually a mixture of cells in stage I and stage III latency. Nevertheless, the overall conclusion from these data is that expression of the BHRF1
miRNA cluster indeed correlates with usage of the Cp and Wp promoters and is therefore likely to be characteristic of stage III latency.
Expression of Several EBV miRNAs Increases during Lytic Replication
In the case of the miRNAs encoded by the γ herpesvirus KSHV, induction of lytic viral replication fails to significantly enhance the level of expression of ten out of the 11 viral miRNAs, all of which are expressed in latently infected cells [10
]. The exception to this generalization, miR-K10, appears to be expressed at higher levels during lytic replication because, unlike the other ten viral miRNAs, it lies within a viral transcription unit that is activated by lytic replication [10
]. Consideration of the genomic location of the viral EBV miRNAs suggests, in contrast, that expression of many of these viral miRNAs is likely to increase after induction of lytic replication. miR-BHRF1–2 and BHRF1–3 lie within the 3′ untranslated region of the early lytic mRNA encoding BHRF1 [11
] and therefore would be expected to be induced during lytic EBV replication. Moreover, recent data demonstrate that BART mRNA expression is also significantly induced during lytic replication of EBV [34
], so one would predict that BART miRNA expression would increase in parallel. It is less clear whether lytic replication would be likely to result in increased miR-BHRF1–1 expression, as this viral miRNA appears to lie 5′ to the BHRF1 mRNA transcription start site [11
Examination of the level of lytic EBV replication induced by treatment of the various LCL, BL, and PEL cell lines analyzed in A and S2 with TPA and n-butyrate revealed that Daudi and MUTU I were the most responsive. Specifically, using immunofluorescence to detect the EBV Zebra protein, which is activated very early during lytic replication [29
], we observed that TPA/n-butyrate treatment increased the number of Zebra-positive Daudi cells from approximately 1.7% to 19.4%, while the number of Zebra-positive MUTU I cells increased from less than 0.5% to approximately 54% (Figure S3
). Analysis of viral miRNA expression revealed that induction of EBV lytic replication resulted in a clear increase in the expression of the viral miRNAs miR-BART1-3p, miR-BART3-3p, miR-BART7, miR-BART10-3p, and miR-BHRF1–2, but did not significantly enhance expression of miR-BHRF1–1 (). In the case of miR-BHRF1–2, this increase was particularly apparent when the level of expression of the pre-miRNA was analyzed, although an increase in the mature miR-BHRF1–2 level was also detected in MUTU I cells. Together these data therefore argue that the expression level of several EBV miRNAs increases significantly during lytic replication, with the apparent exception of miR-BHRF1–1.
Induction of Lytic Replication Can Increase EBV miRNA Expression
Viral miRNAs Have Been Conserved during Lymphocryptovirus Evolution
While between nine and 18 miRNAs are encoded by each of the herpesviruses analyzed so far, none of these viral miRNAs show any obvious sequence homology [10
]. On the other hand, the herpesviruses that have been analyzed, i.e., EBV, KSHV, cytomegalovirus, and mouse herpesvirus 68, are either from different herpesvirus genera or, in the case of mouse herpesvirus 68, derived from a very different species.
We reasoned that if virally encoded miRNAs are indeed important for aspects of the virus replication cycle in vivo, then viral miRNAs should tend to be conserved during viral evolution. To address the validity of this hypothesis, we sought to identify miRNAs encoded by rhesus lymphocryptovirus (rLCV), a member of the lymphocryptovirus genus of which EBV is the human representative. The primate LCV genus includes members that infect every primate species examined so far, and the sequence divergence between different primate LCVs predicts a phylogenetic tree that parallels that of the primate species themselves [16
]. It has therefore been proposed that primate LCVs have coevolved with their primate host species and that EBV and rLCV evolutionarily diverged up to 23 million years, and at least 13 million years ago [16
]. Sequence analysis of rLCV shows that this primate virus has approximately 65% overall nucleotide homology with EBV, with structural proteins being highly conserved, while genes expressed during EBV latent infection are much less well conserved [15
]. More important, this analysis predicts that the EBV BHRF1
genes are both conserved in rLCV.
We performed cDNA cloning of rLCV miRNAs using RNA derived from the latently infected rhesus B cell line 211–98 [35
]. As shown in , 257 rLCV miRNA clones representing 21 distinct viral miRNA sequences were obtained. These could be further assigned to 15 different primary miRNA stem-loop precursors (Figure S5
). One of these miRNAs, miR-rL1-1, was derived from a region adjacent to the rLCV BHRF1
gene, while the other 20 rLCV miRNAs were derived from the rLCV BART
locus. Indeed, as shown in B, the latter rLCV miRNAs are located at the same genomic position, relative to the viral LF2, LF3,
genes encoded on the opposite DNA strand, as the EBV BART
miRNA cluster. As shown in , expression of rLCV miRNAs could be readily demonstrated in the latently rLCV infected 211–98 cell line as well as in a second, unrelated latently rLCV infected cell line, termed 309–98 [35
]. All the rLCV miRNAs analyzed gave rise to a single band in this Northern analysis except miR-rL1-1.
Sequence and Genomic Location of rLCV miRNAs Cloned from the 211–98 Cell Line
Analysis of rLCV miRNA Expression
Sequence comparison of the EBV and rLCV miRNAs revealed that eight miRNAs, derived from six different stem-loop precursors, have been largely or entirely conserved during the evolutionary divergence of rLCV and EBV, especially in the miRNA “seed” region (position 2 to 8 from the 5′ end; ). Moreover, the genomic 5′ to 3′ order of the EBV BART
miRNAs that are conserved in rLCV is unchanged in this distantly related primate lymphocryptovirus (). The small number of miRNA sequence changes that are observed are generally either at the very ends of these miRNAs, which are known to contribute only minimally to target mRNA recognition [36
], or represent C to U or G to A changes that may imply basepairing to a G or to a U residue, respectively (). Other sequence changes that do imply a real difference in basepairing, e.g., a G to C difference between the EBV miR-BHRF1–1 and the rLCV miR-rL1-1 miRNA, may reflect a real difference in the mRNA target sequence in humans versus rhesus macaques. Of particular interest is the one nucleotide insertion seen upon comparison of EBV miR-BART7 with rLCV miR-rL1-12, while all other viral miRNA pairs show neither insertions nor deletions (). This may imply that the miR-rL1-12/miR-BART7 miRNA pair targets a noncoding region, such as an mRNA 3′ untranslated region, that can readily tolerate a 1-nt deletion or insertion.
Sequence Comparison of miRNAs That Are Evolutionarily Conserved in EBV and rLCV and Expressed in Virus-Infected Cells
It could be argued that the sequence conservation of the EBV and rLCV miRNAs documented in reflects the conservation of longer stretches of viral DNA sequence. Conversely, if this conservation is functionally important, then closely adjacent sequences might show more extensive sequence divergence. To examine this issue, the sequences of the predicted primary miRNA stem-loop precursors for each of these EBV and rLCV miRNAs were compared. Previously, we and others have reported that these RNA stem-loops consist of at least three distinct domains [36
]. The central domain consists of the approximately 22-nt mature miRNA sequence, shown in red in , and its complement, termed the miRNA passenger strand, that forms part of the approximately 22-bp miRNA duplex intermediate, but is generally not incorporated into the RNA-induced silencing complex, or is incorporated less efficiently. The terminal loop is a large (≥10-nt) unstructured loop (although RNA folding programs may predict a smaller loop adjacent to a short, rather unstable stem) whose sequence appears irrelevant as long as it maintains an open structure [37
]. Finally, the base of the stem consists of an approximately 8- to 10-bp helical extension of the miRNA duplex that is critical for efficient nuclear processing of the primary miRNA precursor. Because this sequence does not form part of the miRNA duplex intermediate per se, its sequence is not important in and of itself, although maintenance of a helical structure is required [4
]. The approximately 80-nt primary miRNA stem-loop structure is in turn flanked by largely nonstructured RNA sequences that are not believed to play a sequence-specific role in miRNA processing and expression [39
]. Therefore, we would predict that, even though the mature viral miRNA sequences are well conserved (), the flanking basal stem and, particularly, the terminal loop and adjacent single-stranded RNA regions should show significantly more sequence variation.
Sequence Comparison of the Predicted Primary miRNA Stem-Loop Structures of the Indicated rLCV and EBV miRNAs
In , the RNA sequences of six predicted primary stem-loop precursors of miRNAs that are conserved in EBV and rLCV are compared. As may be observed, there is indeed better conservation of the mature miRNA sequence when compared to the flanking basal stem and, particularly, the terminal loop. Overall, the sequences of the six miRNA duplex intermediates are approximately 89% conserved between EBV and rLCV, the basal stems are approximately 84% conserved, the terminal loops are approximately 62% conserved, while the adjacent unstructured flanking sequences, extending approximately 90 nt each side of the predicted miRNA stem-loops shown in , are approximately 68% conserved. This compares to an overall sequence conservation of 65% between the EBV and rLCV genomes [15
]. The statistical significance of the observed conservation in EBV and rLCV of the pre-miRNA stem extension, the miRNA duplex, and the pre-miRNA terminal loop sequence was analyzed using the paired t-test with the null hypothesis being that these sequences are not more highly conserved than the 5′ and 3′ 100-nt segments flanking each viral pre-miRNA. In fact, both the miRNA duplex regions and the miRNA extended stem regions were found to be significantly more highly conserved from EBV to rLCV than either the terminal loop or the flanking sequences (p
Additional computer analysis of the rLCV genome sequence revealed that this virus contains a sequence that is identical at 18 out of 23 positions to the mature EBV miR-BHRF1–2 miRNA (). Moreover, this sequence is found at the same relative genomic position in rLCV, i.e., immediately 3′ to the BHRF1 open reading frame. This candidate rLCV miRNA, termed miR-rL1-2, also forms part of a predicted RNA hairpin that is closely similar to the RNA hairpin predicted for the primary EBV miR-BHRF1–2 precursor (Figure S6
). To test whether this candidate rLCV miRNA is expressed in latently infected cells, we performed a Northern analysis that confirmed the expression of miR-rL1-2 in rLCV-infected 211–98 and 309–98 cells but not in control, uninfected cells (). It therefore appears that rLCV encodes an additional miRNA, missed during cDNA cloning, that is closely similar to EBV miR-BHRF1–2. This brings the number of distinct miRNAs conserved during the evolutionary divergence of EBV and rLCV to at least seven.