Quantitative Studies of Epstein-Barr Virus-Encoded MicroRNAs Provide Novel Insights into Their Regulation^▿

Cell culture.

The BL lines used in this work included the standard EBV-positive latency I BL lines Dante-BL, Sav-BL, Ezema-BL, Akata-BL, Kem-BL, and MutuI-BL cl.59 (28, 31), the group III BL lines MutuIII-BL and GlorIII-BL, which have drifted in culture to a latency III form of infection, the Wp-restricted BL lines Oku-BL, Sal-BL, and Ava-BL (35), and latency I and Wp-restricted sublines derived from Awia-BL (37). The LCLs carrying natural EBV isolates included IM51.1, IM81.1, IM83.1, and IM93.1, established by spontaneous transformation from EBV-infected donors (10), and EH LCL1 and LCL2, generated by in vitro infection of EBV-naïve B cells with the recombinant 2089 EBV genome (20). In addition, the following reference cell lines served as standards for quantifying EBV transcripts: X50/7 LCL (62), CD+Oku LCL (35), the spontaneously permissive cell line Sal tr-LCL (containing 2% BZLF1-positive cells), and the EBV-positive nasopharyngeal cell line C666-1 (17). EBV-negative DG75 cells were included as a negative control in all experiments. AKBM is a derivative of Akata-BL stably transfected with a green fluorescent protein (GFP) reporter plasmid under the control of the early BMRF1 promoter (51). All cell lines were maintained in exponential growth in RPMI medium-10% selected fetal calf serum (Biosera) supplemented with 2 mM l-glutamine (Gibco) and penicillin-streptomycin (Sigma). Where indicated, RNA transcription was inhibited by the addition of 5 μg/ml actinomycin D (Sigma) to the cell culture medium, and cells were harvested for analysis at the indicated time points.

Reactivation of EBV in AKBM cells and isolation of cells in the lytic cycle.

Viral reactivation in AKBM cells was induced by cross-linking surface immunoglobulin (Ig) with 100 μg/ml goat anti-human IgG antibody (Cappel) as described previously (51). Cells were then transferred into fresh medium, and 2 × 10⁶ cells were harvested at each indicated time point. To monitor the induction of the virus lytic cycle, an aliquot of cells was assayed by flow cytometry, either directly for GFP expression or after being stained with the BZLF1-specific monoclonal antibody (MAb) BZ.1, as described previously (51). To isolate productively infected cells, induced AKBM cells were sorted into GFP-positive and GFP-negative populations using a MoFlo cell sorter (Dako, North America). Where indicated, EBV genome replication was inhibited by the addition of 200 μg/ml acycloguanosine (Sigma) to the culture medium.

Measurement of EBV genome load.

Genomic DNA was extracted using a GenElute mammalian genomic DNA kit (Sigma). EBV genome load was determined by quantitative PCR (QPCR) using primer-probe combinations to amplify EBV DNA polymerase (BALF5) and the cellular beta 2-microglobulin sequences, as described previously (34).

Primary infection of B cells.

Primary B cells were isolated from peripheral blood samples using CD19 Dynabeads (Invitrogen) and subsequently exposed to preparations of recombinant 2089 EBV (20) at a multiplicity of infection (MOI) of 100 using published methods (56). Infected cells were cultured in fresh medium and harvested at the indicated time points.

Quantitative PCR analysis of EBV transcripts.

Total RNA was prepared using a Nucleospin RNA isolation kit (Machery-Nagel), treated with DNase I to remove residual DNA (DNA-free; Ambion), and reverse transcribed using gene-specific cDNA primers (10). Relative levels of Qp-, Cp-, and Wp-initiated transcripts, BARTs (exon 6-7 splice junction), lytic and latent BHRF1 transcripts, and immediate-early BZLF1 transcripts were determined by QPCR using published primers and TaqMan probes (10, 36). In addition, a new assay was designed to detect both latent and lytic BHRF1 transcripts as follows: cDNA primer 5′-TCTTGCTGCTAGCT-3′, forward primer 5′-CCCTCTTAATTACATTTGTGCCAGAT-3′, reverse primer 5′-TCCCGTATACACAGGGCTAACAGT-3′, and probe 5′-FAM (6-carboxyfluorescein)-TAGAGCAAGATGGCCTATTCAACAAGGGAGA-TAMRA (6-carboxytetramethylrhodamine)-3′. Two further assays were also designed to amplify across the exon 2-3 splice junction in conventional BART mRNAs (BART 2-3) or across the exon 1-3 splice junction in the alternative splice variant lacking exon 2 (BART 1-3) using the following primers and probes: common cDNA primer 5′-TCTAAAGTCATACGCCC-3′ (exon 3), BART 2-3 forward primer 5′-TCCACTTTGTGTTACAGGTCCG-3′ (exon 2-3 junction), BART 1-3 forward primer 5′-CTCTTCATGTGAGGTCCGGC-3′ (exon 1-3 junction), common reverse primer 5′-TGTGTCCGGTAAACGCCATA-3′ (exon 3), and BART probe 5′-FAM-CCACGGAGACTCGGACGTAGCCCTT-TAMRA-3′ (exon 3).

Relative quantification of gene expression was performed using an Applied Biosystems 7500 sequence detection system. Each PCR run included duplicate test cDNA samples and serial cDNA dilutions prepared from a suitable reference line (Akata-BL for Qp, X50/7 for Wp, CD+Oku LCL for Cp and latent BHRF1, Sal tr-LCL for BZLF1 and lytic BHRF1, and C666-1 for BARTs), which were used to construct relative standard curves. EBV gene expression data were normalized either to cellular glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or to phosphoglycerate kinase (PGK) expression data quantified using commercially available assays (Applied Biosystems). Normalized data were expressed relative to the level for the appropriate reference line, which was assigned an arbitrary value of 1. Note that in the case of lytic transcripts, the data were further adjusted such that the final values were expressed relative to the level for a culture containing 100% BZLF1-positive cells.

Quantitative PCR analysis of EBV miRNAs.

All EBV miRNA sequences were obtained from the Sanger miRNA database, version 12 (http://microRNA.sanger.ac.uk/) (29). Real-time PCR assays to quantify EBV miRNAs were based on the stem-loop reverse transcriptase (RT) primer method described by Chen et al. (13). Stem-loop RT primers and primer-probe combinations for each target miRNA were either designed using the custom TaqMan small RNA assay service (miR-BART5, miR-BART7, miR-BART22, miR-BART13, and miR-BART2-5p) or obtained commercially (miR-BHRF1-1, miR-BHRF1-2, miR-BHRF1-3, miR-BART3, miR-BART4, miR-BART1-3p, and miR-BART15) (Applied Biosystems). For miRNA analysis, RNA was prepared using a mirVana miRNA isolation kit (Applied Biosystems) according to the manufacturer's instructions. Input RNA (10 ng) was reverse transcribed in a 20-μl reaction volume containing up to 7 RT stem-loop primers using a TaqMan microRNA reverse transcription kit (Applied Biosystems). QPCRs (in 20-μl reaction mixtures) were performed using an Applied Biosystems 7500 sequence detection system. In each PCR, duplicate aliquots of each test cDNA were run alongside serial cDNA dilutions prepared from a suitable reference line (X50/7 for miR-BHRF1 miRNAs, and C666-1 for miR-BART miRNAs), which were used to construct relative calibration curves. To convert from relative to absolute quantitation, target miRNAs were obtained as synthetic oligonucleotides (Eurogentec), pooled, and reverse transcribed as described above. Aliquots of cDNA corresponding to 10⁸ to 10² miRNA copies were then subjected to PCR amplification and the data used to generate calibration curves from which the absolute number of EBV miRNA molecules in reference X50-7 and C666.1 cells could be determined. Expression data were normalized to the level for RNU48 as a loading control (Applied Biosystems).

Sequencing of BHRF1 miRNAs.

BHRF1 sequences spanning all three miRNAs were PCR amplified using the primers 5′-GGACTTGGGGTGATTTTCTATGT-3′ and 5′-CCCACACTCACCTCAGTTATTTC-3′ under the following conditions: 94°C for 5 min; 35 cycles of 94°C for 60 s, 56°C for 60 s, and 72°C for 2 min; and 72°C for 10 min. The 1,600-bp fragment was then gel purified and sequenced using an automated Applied Biosystems 3730 system (Functional Genomics and Proteomics Services, School of Biosciences, University of Birmingham, United Kingdom). The pre-miRNA secondary structure prediction through free energy minimization was calculated using the software program Mfold (http://mfold.rna.albany.edu/) (66).

Validation of miRNA PCR assays.

In the initial experiments, we validated the specificity and sensitivity of a panel of QPCR assays designed to quantify selected EBV miRNAs; these included all three miRNAs derived from the BHRF1 region, five miRNAs from BART cluster 1, three miRNAs from BART cluster 2, and miR-BART2-5p. We also optimized a multiplex cDNA synthesis method incorporating up to 7 miRNA-specific RT primers such that several miRNAs could be assayed from the same cDNA pool but without significantly reducing the sensitivity of detection (data not shown). Using cDNA prepared by this multiplex protocol, we then validated both the specificity and the sensitivity of each PCR assay in a series of control experiments. First, using RNA from six EBV-negative cell lines, we showed that all but two assays always gave negative PCR results; the exceptions were the assays for miR-BART3 and miR-BART4, in which these cell lines occasionally yielded extremely weak signals (data not shown; see Fig. 6B). Second, using cell mixtures containing serial dilutions of an EBV-positive LCL in an EBV-negative cell background, we found that all assays reproducibly detected a single EBV-infected cell (data not shown). Finally, we determined the absolute sensitivity of each assay by using cDNA prepared from serial dilutions of a pool of synthetic oligonucleotides containing all 12 target miRNAs. Data from representative amplification plots for serial dilutions of miR-BHRF1-1 and miR-BART3 are shown in Fig. 2A, while standard curves for all 12 assays are shown in Fig. 2B. All assays demonstrated an excellent correlation between cycle threshold (C_T) values and RNA input over at least a 5-log range and had a lower detection limit of between 100 and 1,000 miRNA copies.

BHRF1 transcription and miR-BHRF1 miRNA expression in EBV-positive B cell lines.

In the next series of experiments, we sought to explore the relationship between BHRF1 transcription and the levels of mature BHRF1 miRNAs in well-characterized EBV-positive B cell lines with different patterns of latent gene expression. These test lines included seven conventional latency I BL lines, five Wp-restricted BL lines, two latency III BL lines, and seven virus-transformed LCLs; note that EH LCL1 and EH LCL2 were generated from a recombinant prototype EBV strain which, like B95-8, lacks many of the BART miRNAs (Fig. 1C). We first verified that these cell lines showed the expected pattern of latent viral promoter usage by quantifying levels of Qp-, Cp-, and Wp-initiated EBNA transcripts (Fig. 3). Thus, latency I BL lines exclusively used Qp with only trace or undetectable levels of Cp/Wp activity. In contrast, the Wp-restricted BL lines were distinguished by high levels of Wp activity, usually in the absence of the other latent promoters. As expected, latency III BLs and LCLs showed variable levels of both Cp and Wp usage, with the exception of X50-7, which lacks Cp sequences (62). Furthermore, all test cell lines were tightly latent since immediate-early BZLF1 transcripts were either undetectable or present at very low levels, equivalent to fewer than 1% BZLF1-positive cells (Fig. 4A).

Having validated our test lines, we next quantified BHRF1 transcripts using two different primer-probe combinations, one designed to detect both lytic and latent BHRF1 transcripts and the other designed to specifically amplify Cp/Wp-initiated latent BHRF1 transcripts. As shown by the data in Fig. 4A, the BHRF1 expression profiles were comparable for both assays and indicated that BHRF1 transcripts were detectable only in cells with Cp and/or Wp activity. Consistent with the very low BZLF1 signals, negligible levels of lytic BHRF1 transcripts were observed in these test lines (data not shown), thus excluding the possibility that a small population of productively infected cells were making a significant contribution to the total levels of BHRF1 transcripts. We then went on to measure the levels of miR-BHRF1 miRNAs in the same panel of cell lines using our newly developed QPCR assays. The data (Fig. 4B) show that BHRF1 miRNA expression varied widely, ranging from 0 to 5 copies per pg RNA in latency I BL lines to more than 300 copies per pg RNA in latency III LCLs. Interestingly, levels of the miR-BHRF1 miRNAs in the LCLs were at least as high as those in the Wp-restricted BL lines, even though the latter has much higher levels of BHRF1 transcripts. In most cases, we noted that all three miR-BHRF1 miRNAs were coordinately expressed within a cell line, with no more than 3-fold variation in absolute values between individual miRNAs. One exception was Sal-BL, which expressed abundant levels of mir-BHRF1-2 and mir-BHRF1-3 but is known to be deleted for miR-BHRF1-1 (35); a second exception was Ava-BL, where levels of miR-BHRF1-2 were unusually low compared to those of the other BHRF1 miRNAs.

BHRF1 miRNA sequence variation.

Since previous studies have revealed that sequence changes can interfere with miRNA processing (21, 49), we considered the possibility that EBV strain polymorphisms might account for some of the observed variation in BHRF1 miRNA levels. To this end, we determined the sequence of the BHRF1 pre-miRNAs in 12 EBV isolates and compared the sequences with the published B95-8 genome; the number and position of any sequence changes are summarized in Fig. 5. Overall, miR-BHRF1-1 and miR-BHRF1-2 sequences were highly conserved. In two cases (Kem-BL and Sav-BL), we found a single change in the mature miR-BHRF1-1 miRNA; the effect of this mutation on pre-miRNA processing or PCR detection is unclear, as the BHRF1 region is not transcribed in these latency I lines (Fig. 4A). One cell line, Ava-BL, contained a single polymorphism in the mature miR-BHRF1-2 sequence which disrupts a G-C base pair in the stem-loop. With Mfold used to calculate the change in minimum free energy, this G-to-A substitution is predicted to significantly destabilize the pre-miRNA hairpin loop structure (Fig. 5) and therefore may account for the low levels of miR-BHRF1-2 detected in this cell line (Fig. 4B). Finally, five cell lines contained six common nucleotide changes in the miR-BHRF1-3 pre-miRNA. Surprisingly, these changes did not decrease the minimum free energy structure of the BHRF1-3 pre-miRNA and are therefore unlikely to affect folding or processing.

BART transcription and miR-BART miRNA expression in EBV-positive B cell lines.

In the next series of experiments, we examined the expression of the BART transcripts and derived miRNAs in the same panel of B cell lines (Fig. 6). Since particular BART splice variants may favor the production of BART miRNAs (21), we quantified BART transcription using two different PCR assays, the first specific for the conventional transcript containing exons 1, 2, and 3 (exon 2-3 assay) and the second specific for the exon 2-deleted variant (exon 1-3 assay). In contrast to BHRF1 transcripts, BART transcripts were detected across all forms of latency, albeit at highly variable levels (Fig. 6A); BART transcripts containing exon 2-3 and exon 1-3 splice junctions appeared to be coordinately expressed. Screening the same samples for the presence of nine representative BART miRNAs, we found levels ranging from fewer than 10 to more than 400 copies per pg RNA (Fig. 6B to D). Overall, it is clear that the B cell lines expressed lower levels of BART miRNAs than the epithelial cell line C666-1 (data included in Fig. 6B to D for comparison). Interestingly, we observed up to a 50-fold variation between different BART miRNAs within a given cell line. BART miRNA levels also varied widely between cell lines, but all cell lines showed the same hierarchy of BART miRNA expression, with miR-BART7 usually the most abundant and miR-BART13 the most scarce. While there was a good correlation between BART transcription and BART miRNAs in the BL lines, levels of BART transcripts were frequently lower in the LCLs despite the fact that BLs and LCLs expressed similar levels of BART miRNAs.

Stability of EBV miRNAs.

The above-mentioned findings highlighted both the dramatic differences in overall levels of viral miRNA expression between different cell lines and the wide variation in levels of individual miRNAs within a cell line. One possible explanation for this variability could be differences in the turnover of individual EBV miRNAs. To investigate this in more detail, we examined the half-life of selected miRNAs in three cell lines, X50/7 LCL, Sal-BL, and Ava-BL. Cells were cultured in the presence of the RNA synthesis inhibitor actinomycin D and then harvested for analysis at various time points after treatment. Residual levels of BHRF1 and BART transcripts and derived miRNAs were then monitored by QPCR (Fig. 7); as a control, we also assayed c-myc RNA since this transcript has been reported to have a very short half-life of around 12 min (58). Figure 7A shows the individual data points from X50-7 cells up to 24 h posttreatment, while Fig. 7B summarizes the data from all three cell lines. In the case of X50/7, c-myc RNA levels decreased rapidly to less than 10% after 4 h of exposure to actinomycin D, while the latent BHRF1 transcripts also decreased, but to a lesser extent. In contrast, the BHRF1 miRNAs were very stable, decreasing by less than 2-fold even after 24 h. We then measured levels of BART transcripts and three miRNAs (miR-BART4, miR-BART7, and miR-BART2-5p) previously shown to have markedly different steady-state levels. We found that both the BART transcripts and the BART miRNAs remained relatively unchanged in X50/7 up to 24 h (Fig. 7A). Similar data for the stability of the transcripts and miRNAs were obtained from Sal-BL and Ava-BL (Fig. 7B). Thus, it appears that EBV miRNAs are very stable, and therefore, the observed variation in miRNA levels cannot be explained by different rates of miRNA turnover.

Expression of EBV miRNAs during lytic replication.

Since it has been reported that BHRF1 and BART transcripts are induced during productive infection (32, 47, 64), we next asked to what extent miR-BHRF1 and miR-BART miRNA levels might also be increased in the EBV lytic cycle. To this end, we exploited the AKBM cell system in which EBV-positive Akata-BL cells are stably transfected with a GFP reporter gene under the control of the EBV lytic cycle BMRF1 promoter; following Ig cross-linking, a significant proportion of cells enter into the lytic cycle and these cells can be identified by GFP expression. In the first instance, we monitored the production of BHRF1 and BART transcripts and derived miRNAs in AKBM cells at various time points up to 24 h postinduction. In the representative experiment whose results are shown in Fig. 8A, BZLF1-positive lytic cells were detected by flow cytometry within 4 h and increased to around 22% of the culture at 24 h. Lytic BHRF1 transcripts appeared with kinetics similar to those of BZLF1 (Fig. 8B) and coincided with marked increases in miR-BHRF1-2 and -1-3 miRNAs but not miR-BHRF1-1 (Fig. 8C); this is consistent with the idea that BHRF1 transcription is initiated from the lytic promoter BHRF1p, rather than the latent promoters Cp/Wp, at these early time points. In contrast, levels of BART-derived miRNAs (here illustrated with data for miR-BART15 and miR-BART2-5p) barely increased during the same 24-h period, despite the fact that BART transcript levels increased more than 45-fold (Fig. 8B and C).

To examine whether the relatively small increases in miR-BART miRNA expression might be due to a delay in processing the BART transcripts, we repeated the experiment but now analyzed gene expression in the induced AKBM cells for up to 120 h postinduction. Lytic BHRF1 and BART transcripts showed levels of induction at 24 h similar to those observed in the previous experiments and thereafter remained relatively constant (Fig. 9B). However, we now observed that latent BHRF1 transcripts became detectable at later time points, coincident with induction of the latency III promoters Wp and Cp (Fig. 9B and data not shown) and leading to expression of all three miR-BHRF1 miRNAs (Fig. 9C). Consistent with our earlier findings, we again noted a rapid and dramatic increase in BART transcription accompanied by only modest 5- to 8-fold increases in BART miRNAs.

The above-mentioned findings suggested that BHRF1 miRNA production during virus replication is biphasic, with early lytic BHRF1 transcripts selectively generating miBHRF1-2 and miR-BHRF1-3 while delayed latent BHRF1 transcripts are processed to give all three miR-BHRF1 miRNAs. To investigate BHRF1 miRNA expression in more detail, we performed two additional experiments. In the first, we confirmed that the lytic and latent BHRF1 transcripts were present in the same cell population. Briefly, AKBM cells were induced into the lytic cycle as before and then sorted into GFP-positive (lytic) and GFP-negative (latent) populations at 48 h postinduction. We found that both latent and lytic BHRF1 transcripts, and all three miR-BHRF1 miRNAs, were greatly enriched in the GFP-positive fraction relative to the level for the GFP-negative population (data not shown).

The appearance of latency III transcripts during the later stages of virus replication has been reported previously (64). We speculated that such transcripts might be expressed from the very high copy numbers of newly replicated EBV genomes, since such genomes are unmethylated and could therefore act as templates for Wp- and Cp-initiated transcription. To test this hypothesis, we induced AKBM cells in the presence of acyclovir (ACV) to block viral DNA replication. By QPCR, we confirmed that the average EBV genome load increased from around 50 genomes per cell to over 15,000 genomes per cell in induced control AKBM cells. The addition of ACV completely blocked this genome amplification (Fig. 10A) and also ablated induction of latency III transcripts and miR-BHRF1-1 miRNA (Fig. 10B and C), supporting the view that these transcripts are dependent on viral DNA synthesis. In contrast, ACV treatment only modestly reduced levels of lytic BHRF1 and BART transcripts (2-fold and 5-fold, respectively) (Fig. 10B) and made little difference to the expression of the two remaining miR-BHRF1 miRNAs or BART miRNAs (Fig. 10C and D).

In the final experiment, we asked whether the failure to significantly increase BART miRNA expression during virus replication was due to changes in BART transcript or BART miRNA stability. Briefly, AKBM cells were induced as before, and then, at 12 h postinduction, the cells were either cultured in normal medium or transferred to medium containing actinomycin D to block further RNA transcription. We found that the BART miRNAs had similarly long half-lives in lytically induced AKBM cultures and in the latently infected cell lines examined previously (data not shown), arguing that the weak induction of the BART miRNAs was not due to increased miRNA turnover during the lytic cycle.

Expression of EBV miRNAs during primary infection of B cells.

In the final set of experiments, we examined the expression of EBV miRNAs in newly infected primary B cells. On this point, the kinetics of EBV gene expression during primary B cell infection are well established. Thus, Wp is activated immediately postinfection (1, 56, 59), leading to the initial expression of EBNA2, EBNA-LP, and also BHRF1 (1, 2, 4, 36). Wp activity then falls after 24 h, as Cp activity becomes dominant, concomitant with the broadening of latent antigen expression to the full spectrum of EBNAs and LMPs. Primary B cells were therefore exposed to EBV preparations in vitro, cultured, and harvested for analysis at the indicated time points. Note that the recombinant 2089 EBV strain used in this work contains the same 12-kb BamHI A deletion as the prototype B95-8 strain (Fig. 1C); therefore, in this experiment it was necessary to detect the BART transcripts using an alternative primer-probe combination which specifically amplified sequences across the exon 6-7 splice junction. The data in Fig. 10A show the results of this BART6-7 assay, alongside those of standard assays used to measure Wp-initiated transcripts, Cp-initiated transcripts, and BHRF1 transcripts. As expected, Wp activity was already induced by 8 h postinfection, peaked at 12 h, and then fell as the level of Cp activity began to increase; latent BHRF1 expression was also robustly induced by 8 h (Fig. 10A). In contrast, the appearance of the BHRF1 miRNAs was significantly delayed relative to that of BHRF1 transcription, with weak signals first detected between 12 and 24 h, which then gradually increased throughout the experiment (Fig. 10B). Kinetic analysis of BART RNAs also showed a similar lag in the appearance of mature BART miRNAs relative to BART transcription (Fig. 10A and B).