Human cytomegalovirus expresses novel microRNAs during productive viral infection
Summary
MicroRNAs (miRNAs) are a large class of ∼22-nucleotide non-coding RNAs that facilitate mRNA cleavage and translation repression through the RNA interference pathway. Until recently, miRNAs have been exclusively found in eukaryotic organisms. A non-immunogenic molecule requiring minimal genomic investment, these RNAs may offer an efficient means for viruses to modulate both their own and the host's gene expression during a productive viral infection. In this study we report that human cytomegalovirus (HCMV) expresses miRNAs during its productive lytic infection of four clinically relevant human cell types: fibroblast, endothelial, epithelial and astrocyte cells. The sequences of the miRNAs, expressed from the UL23 and US24 loci of the viral genome, were conserved among all HCMV strains examined and in chimpanzee cytomegalovirus. Furthermore, their expression was detected from both a laboratory-adapted strain and a clinical isolate of HCMV. The conservation of these miRNAs and their expression in different cell types suggests that they represent an evolutionarily primitive feature in the viral genome, and that virus-encoded miRNAs may be more common than previously believed.
Introduction
MicroRNAs (miRNAs), a class of ∼22-nucleotide (nt) non-coding RNAs found in all classes of metazoans (Lagos-Quintana et al., 2001; Lau et al., 2001; Lee and Ambros, 2001), play an important role in modulating gene expression during numerous biological processes such as development (Bartel, 2004). These RNAs are encoded within the host genome as a part of a predicted RNA hairpin precursor. To generate miRNAs, a long RNA transcript, termed pri-miRNA, is initially processed into a ∼60–110 nt hairpin RNA precursor (pre-miRNA) by the nuclear ribonuclease Drosha (Lee et al., 2003). The resultant pre-miRNA is processed by Dicer to generate a ∼22 bp duplex (Bernstein et al., 2001; Hutvagner et al., 2001). Subsequently, one strand of the duplex is incorporated into the RNA-induced silencing complex (RISC) of the RNA interference (RNAi) pathway and guides the RISC to the target transcript that is complementary to the miRNA sequence (Hammond et al., 2000; Zeng et al., 2003). RISC binding results in either mRNA cleavage or translation inhibition, depending on the degree of complementarity between the miRNA and the target transcript (Bartel, 2004).
Although the functions of many miRNAs have yet to be determined, the importance of miRNA-mediated gene regulation in humans is underscored by the fact that miRNAs have been found in every human tissue type examined thus far (Lagos-Quintana et al., 2001; Bartel, 2004). Interestingly, subsets of these miRNAs can be expressed in very specific tissue types and sometimes only during certain developmental stages. The large number of miRNA genes found in the human and animal genomes, their unique distribution and expression patterns, along with functional studies in Caenorhabditis elegans and Drosophila melanogaster, suggest that miRNAs represent a general mechanism to regulate gene expression in specific cell types and tissues (Bartel, 2004).
Animal viruses contain the smallest genomes among all living organisms and typically replicate in many types of cells and tissues during productive infection. Given their small size, miRNAs may offer an efficient method for viruses to regulate both viral and host gene expression in a temporal and tissue-specific manner in order to successfully infect and replicate in vivo. The first examples of virally encoded miRNAs were recently reported to be expressed in B-cell lines supporting latent infection and reactivation of Epstein–Barr virus (EBV) (Pfeffer et al., 2004), a member of the herpesvirus family (Kieff and Rickinson, 2001), and in HIV persistently infected cells (Omoto et al., 2004). These results potentially expand the scope of organisms that employ miRNAs as gene regulating mechanisms from complex organisms to some of the most primitive infectious agents. However, the importance of virally expressed miRNAs remains to be elucidated.
With a genome size of ∼230–240 kb, human cytomegalovirus (HCMV) has the largest genome in the human herpesvirus family, which also includes the prototype herpes simplex virus 1 (HSV-1), EBV and Kaposi's sarcoma-associated herpesvirus (Mocarski and Courcelle, 2001; Roizman and Knipe, 2001). This virus causes mild or subclinical diseases in immunocompetent adults and leads to severe life-threatening complications in immunocompromised individuals, which include AIDS patients and transplant recipients (Pass, 2001). HCMV is also the leading viral cause of congenital abnormalities and mental retardation in newborns (Britt, 1999). Clinical data indicate that HCMV infection of various tissues and cell types is responsible for a myriad of complications (Britt, 1999; Pass, 2001). Depending on the tissue type and the host's immune state, HCMV engages in acute infections with highly productive growth, persistent infections with low levels of replication or latent infections where no viral progeny is produced (Mocarski and Courcelle, 2001; Pass, 2001). In different cell types, HCMV exhibits various growth rates and different gene expression patterns, suggesting that its replication and gene expression in a particular cell type is tightly regulated and thus, determines the outcome of diseases in specific tissues (Britt, 1999; Pass, 2001). Indeed, recent systemic gene-deletion analysis indicated that different viral determinants regulate HCMV replication rates by supporting or suppressing its growth in a cell type-specific manner (Dunn et al., 2003). Thus, it is conceivable that HCMV may encode miRNAs to regulate its own and the host's gene expression during productive replication in order to successfully infect and replicate in different cell types. In this study, we report the identification of virally encoded miRNAs from HCMV.
Results
Cloning and sequencing of small RNA species from HCMV-infected cells
It has been shown that HCMV infects different types of cells and tissues in vivo, leading to a variety of human diseases (Pass, 2001). To determine whether HCMV expresses miRNAs, two different types of cells, human foreskin fibroblasts (HFFs) and astrocytoma U373MG cells, were infected with HCMV and were used to isolate viral-encoded miRNAs. They are among the most permissive cell types that support HCMV lytic infection and replication in vitro.
Both HFFs and U373MG cells were infected with HCMV TowneBAC, which was generated by inserting a bacterial artificial chromosome (BAC) sequence into the genome of HCMV Towne strain (Marchini et al., 2001). This virus replicates as well as its parental strain, Towne, in HFFs, U373MG and several other cell types (Marchini et al., 2001; Dunn et al., 2003). At 96 h after infection, total RNAs were isolated from infected HFFs and U373MG cells and separated on denaturing polyacrylamide gels. The fractions corresponding to the small RNA population (with a size of 18–24 nt) were isolated from the gels and cloned. Sequencing analysis of 1396 clones indicated that about 1.1% of the cloned small RNAs originated from HCMV TowneBAC (Table 1), and that most of the HCMV sequences were cloned more than once (Table 2). A total of three unique small RNA sequences of ∼22 nt were identified as potential HCMV-encoded miRNAs. The sequences coding for these potential miRNAs are located at two distinct regions in the viral genome: two cloned sequences are located in the non-coding region between open reading frame (ORF) UL22A and UL23 in the viral unique long (UL) sequence, while the other is at the non-coding region between ORF US24 and US25 in the unique short (US) sequence (Fig. 1). Based on their alignment to the HCMV Towne genome (Dunn et al., 2003), these three miRNAs have been designated as miR-UL23-5p, miR-UL23-3p and miR-US24 respectively (Fig. 1). The sequences corresponding to miR-UL23-5p, miR-UL23-3p and miR-US24 were cloned eight, four and three times in the 1396 sequences analysed respectively (Table 2).
| Sequence classification | Number of sequences | Percentage of total (%) |
|---|---|---|
| Viral miRNAs | 15 | 1.07 |
| Cellular miRNAs | 59 | 4.22 |
| mRNA fragments | 588 | 42.12 |
| Non-coding RNA | 656 | 47 |
| Not annotated | 67 | 4.8 |
| Not matched | 11 | 0.79 |
| Total sequences cloned | 1396 | 100 |
| miRNA name | Sequences (5′→3′) | Number of miRNAsequences cloned | Length (nt) | Towne co-ordinates |
|---|---|---|---|---|
| miR-UL23-5p | TAACTAGCCTTCCCGTGAGA[GT] | 4 | 20–22 | 26552–26571, 26552–26573 |
| miR-UL23-3p | TCACCAGAATGCTAGTTTGTAG | 8 | 22 | 26589–26610 |
| miR-US24 | CGGTCCGAGCCACTGAGCGGTT | 3 | 22 | 217996–218017 |
- The co-ordinates are based on the genome sequence of HCMV Towne (Dunn et al., 2003).
Sequences and genome locations of the two pre-miRNAs (pre-miR-UL23 and pre-miR-US24) coding for the identified three HCMV miRNAs (miR-UL23-5p, miR-UL23-3p and miR-US24). The miR-UL23 locus is located at the non-coding region between the coding sequences of open reading frame UL22A and UL23 while miR-US24 is at the non-coding region between the coding sequences of US24 and US25. miR-UL23-5p (in red) and miR-UL23-3p (in blue) are generated from the 5′ and 3′ arms of pre-miR-UL23 while miR-US24 (in red) arises from the 5′ strand of pre-miR-US24. The sequences were derived from our recent results on HCMV TowneBAC sequence (Dunn et al., 2003). UL, unique long region; US, unique short region.
Identification of HCMV-encoded miRNAs
In order for a small RNA to be classified as an miRNA, several criteria must be satisfied (Ambros et al., 2003). In addition to the identification and isolation of the small RNA from a size-fractionated RNA pool, the criteria require the detection of the small RNA as a ∼22 nt RNA and include the prediction of the 60–80 nt pre-miRNA transcript to adopt a fold-back or hairpin structure (Ambros et al., 2003; Bartel, 2004). Accordingly, two sets of experiments were carried out to determine whether the cloned small RNA candidates were indeed HCMV-encoded miRNAs. First, we performed in silico analysis to examine the secondary structure of the hypothesized pre-miRNA transcripts. These analyses suggest that the predicted pre-miRNA sequences can adopt the fold-back or hairpin structure characteristic of miRNA genes (Fig. 1).
In the second set of experiments, Northern analyses were used to determine whether HCMV-infected cells expressed these virally encoded small ∼22 nt RNAs. HFFs were infected with TowneBAC at a multiplicity of infection (moi) of 1. At 96 h after infection, total RNAs were extracted from infected cells, electrophoretically separated on denaturing gels, electrically transferred to a membrane and analysed using a radiolabelled probe specific for the candidate miRNA sequence (i.e. miR-UL23-5p, miR-UL23-3p and miR-US24) (Fig. 2). The expression of human miRNA miR-16 (Lagos-Quintana et al., 2001) was used as the internal control (Fig. 2D). All three HCMV-encoded miRNA candidates identified in the cloning and sequencing process were detected to be expressed as distinct 21–23 nt transcripts by radiolabelled probes specific for the cloned miRNA sequence (Fig. 2, lanes 2, 5 and 8). The ∼60 nt pre-miRNA transcript was also detected for miR-UL23-5p and miR-UL23-3p, but was not found for miR-US24. The inability to detect the pre-miRNA transcript for miR-US24 may be due to its low stability in cells, as observed in previous studies on detection of human pre-miRNAs (Lagos-Quintana et al., 2001; Bartel, 2004). The ∼22 and ∼60 nt RNA species were not found using the probes specific for the complementary strand of the candidate miRNAs (data not shown), indicating that these RNAs are expressed from only one strand of the candidate miRNA genes. Furthermore, detection of these RNA species was limited to cells that were infected by HCMV. Transcripts were not detected from uninfected cells (Fig. 2, lanes 1, 4 and 7). Together, these results suggest that miR-UL23-5p, miR-UL23-3p and miR-US24 represent HCMV-encoded miRNAs.
Expression of HCMV miRNAs. Human foreskin fibroblasts were either mock-infected (lanes 1, 4, 7 and 10) or infected with HCMV TowneBAC (lanes 2, 5, 8 and 11) and NJBAC (lanes 3, 6 and 9). At 96 h after infection, total RNAs were isolated from the cells, separated on 12.5% polyacrylamide gels containing 8 M urea, transferred to membranes and hybridized with 32P-labelled oligonucleotides specific for viral miR-UL23-5p (lanes 1–3), miR-UL23-3p (lanes 4–6), miR-US24 (lanes 7–9) and human miRNA miR-16 (lanes 10 and 11).
Viral miRNA genes and their conservation among all HCMV strains
Previous studies have shown that processing of the ∼60–80 nt pre-miRNA precursor generates a ∼20 base-pairing miRNA duplex bearing 2 nt 3′ overhangs (Bartel, 2004). The strand with the less tightly base-paired 5′ end will be incorporated into RISC while the other strand may be degraded (Schwarz et al., 2003). However, each of the strands can be incorporated if there is little difference in the base pairing at the 5′ termini. We examined the location of miR-UL23-5p, miR-UL23-3p and miR-US24 in their corresponding ∼60–80 nt hairpin pre-miRNA conformation. Consistent with the observations regarding strand selection for RISC complex incorporation (Schwarz et al., 2003), miR-UL23-5p and miR-UL23-3p are derived from the 5′ and 3′ arms of pre-miR-UL23 while miR-US24 is derived from the 5′ strand of pre-miR-US24 (Fig. 1).
Sequencing analysis of both laboratory-passaged viral strains and clinical isolates have previously shown that mutations, such as substantial deletions and rearrangements as well as base substitutions, are commonly found in the HCMV genome (Chee et al., 1990; Mocarski and Courcelle, 2001; Dunn et al., 2003; Murphy et al., 2003; Dolan et al., 2004). A laboratory-adapted strain that has been extensively passaged in vitro, Towne may have accumulated numerous mutations in its genome (Mocarski and Courcelle, 2001; Dunn et al., 2003; Murphy et al., 2003; Dolan et al., 2004). Consequently, it is possible that the viral miRNAs sequenced in this study may have diverged from their original forms. To determine whether this was the case, we analysed the loci corresponding to the genes coding for these three miRNAs among the complete genome sequences of seven HCMV strains. The available genomic data represent sequences from highly passaged laboratory strains such as AD169 to low-passaged viruses, such as the Merlin strain, which is considered to be a true wild-type clinical isolate (Chee et al., 1990; Dunn et al., 2003; Murphy et al., 2003; Dolan et al., 2004). While there is sequence variation in the coding sequence of the ORFs flanking the non-coding regions (Fig. 3A, data not shown), the two loci coding for the ∼60 nt pre-miRNAs for miR-UL23-5p, miR-UL23-3p and miR-US24 miRNAs exhibited 100% sequence conservation among all the HCMV strains (Fig. 3B and C). To determine whether these miRNAs are also expressed from an HCMV strain other than Towne, we examined cells infected with NJBAC, a virus derived from a low-passaged clinical isolate (H. Zhu, unpubl. results). As shown in Fig. 2 (lanes 3, 6 and 9), miR-UL23-5p, miR-UL23-3p and miR-US24 miRNAs were detected. These results suggest that the miR-UL23 and miR-US24 loci are found in all HCMV strains and express viral-encoded miRNAs. Given the fact that they are ∼190 kb apart and conserved within each locus, the miR-UL23 and miR-US24 genes represent two distinct classes of HCMV miRNAs, most likely with different targets and roles in viral replication. Sequences similar to these miRNAs were also found in the closely related chimpanzee cytomegalovirus (CMV) genome (Fig. 3B and C) (Davison et al., 2003). Thus, the miRNAs reported here appear to be conserved among human and primate CMVs.
Alignment of the 63 bp genomic sequence coding for pre-miR-US24 (C) and pre-miR-UL23 (B), and the same length UL22A coding sequence that is adjacent to pre-miR-UL23 (within 200 bp) (A). The sequences were obtained from the complete genome sequences of seven HCMV strains and chimpanzee CMV (CCMV) that are available (Chee et al., 1990; Davison et al., 2003; Dunn et al., 2003; Murphy et al., 2003; Dolan et al., 2004).
Characterization of the expression of HCMV miRNAs
The ability of HCMV to replicate in many types of cells and tissues is responsible for the variety of sequalae associated with HCMV infection. For example, HCMV infects human retinal pigment epithelial (RPE) cells and human microvascular endothelial cells (HMVEC) in vivo, leading to viral-associated retinitis and vascular diseases respectively (Mocarski and Courcelle, 2001; Pass, 2001). Due to the lack of an animal model for study of HCMV pathogenesis, cultured natural host cells such as RPE and HMVECs have been used. Thus, we examined the expression of HCMV-encoded miRNAs in these two types of cells in addition to fibroblasts (HFFs) and astrocytes (U373MG). These four cell types represent the tissue types that HCMV is known, but not limited, to infect in vivo: epithelial, endothelial, connective and neuronal (Mocarski and Courcelle, 2001; Pass, 2001). Cells were infected with TowneBAC at an moi of 1 and total RNAs were extracted from the cells at 96 h after infection. As shown in Fig. 4, miR-UL23-5p, miR-UL23-3p and miR-US24 RNAs were found in these four cell types, suggesting that the viral miRNAs are expressed during HCMV infection in vivo. Consistent with previous reports that HCMV infection in these cells led to productive replication (Mocarski and Courcelle, 2001; Pass, 2001), lytic infection in these cells was evidenced by the observation that supernatants collected from the infected cell culture just before RNA extraction contained infectious viral progeny (data not shown).
Expression of viral miRNAs in human U373MG cells (lanes 1–8), retinal pigment epithelial (RPE) cells (9–14) and microvascular endothelial cells (HMVEC) (lanes 15–20). Cells were either mock-infected (lanes 1, 3, 5, 7, 9, 11, 13, 15, 17 and 19) or infected with HCMV TowneBAC (lanes 2, 4, 6, 8, 10, 12, 14, 16, 18 and 20). At 96 h after infection, total RNAs were isolated from cells, separated on denaturing gels, transferred to membranes and hybridized with 32P-labelled oligonucleotides specific for miR-UL23-5p (lanes 1–2, 9–10 and 15–16), miR-UL23-3p (lanes 3–4, 11–12 and 17–18), miR-US24 (lanes 5–6, 13–14 and 19–20) and human miR-16 (lanes 7 and 8).
During productive viral infection, HCMV genes are expressed in a highly regulated cascade fashion (Mocarski and Courcelle, 2001). Viral immediate-early (α) transcripts are expressed first, followed by viral early (β) genes, and finally with viral late (γ) genes. To determine the expression kinetics of these miRNAs, total RNAs were harvested from the infected HFFs at different time points of infection, and the expression of viral miRNAs was studied using Northern analysis (Fig. 5). The expression of miR-UL23-3p and miR-US24 was detected as early as at 18 h after infection and continued to increase at 48 h after infection (Fig. 5, lanes 1–10). In contrast, these miRNAs were barely detectable at 12 h after infection (Fig. 6, lanes 2 and 7). To further examine the temporal expression of these miRNAs during HCMV lytic replication, HFFs were treated with either cyclohexamide and phosphonoacetic acid (PAA), which block host protein synthesis and viral DNA replication respectively (Mocarski and Courcelle, 2001). In the presence of cyclohexamide, only viral immediate-early transcripts are expressed. Treatment with PAA blocks the expression of viral late genes, resulting in the expression of only the immediate-early and early genes. As shown in Fig. 6, miR-UL23-3p was found in cells treated with cyclohexamide (lane 3) while miR-US24 was barely detected (lane 8). In the presence of PAA, cells expressed both miR-UL23-3p and miR-US24 miRNAs (lanes 4 and 9). These results indicate that miR-UL23-3p is expressed at immediate-early time.
Time-course expression of HCMV miRNAs. Human foreskin fibroblasts were either mock-infected (lanes 1, 6 and 11) or infected with HCMV TowneBAC (lanes 2–5, 7–10 and 12–15). Total RNAs were isolated from cells at 18, 24, 30, 36 and 48 h after infection, separated on denaturing gels, transferred to membranes and hybridized with 32P-labelled oligonucleotides specific for miR-UL23-3p (lanes 1–5), miR-US24 (lanes 5–10) and human miR-16 (lanes 11–15).
Expression of HCMV miRNAs at immediate-early, early and late stages of viral lytic replication. Human foreskin fibroblasts were treated either with cyclohexamide (lanes 3, 8 and 13) or with phosphonoacetic acid (PAA) (lanes 4, 9 and 14) before and during viral infection. Cells were either mock-infected (lanes 1, 6 and 11) or infected with HCMV TowneBAC (lanes 2–5, 7–10 and 12–15). Total RNAs were isolated from cells at 12 (lanes 1–2, 6–7 and 11–12), 24 (lanes 3, 8 and 13), 48 (lanes 4, 9 and 14) and 72 h (lanes 5, 10 and 15) after infection, separated on denaturing gels, transferred to membranes and hybridized with 32P-labelled oligonucleotides specific for miR-UL23-3p (lanes 1–5), miR-US24 (lanes 5–10) and human miR-16 (lanes 11–15).
Discussion
In this study, we provide direct evidence that three miRNAs are expressed from two distinct loci in the HCMV genome. The HCMV-encoded miRNAs identified in this study may represent viral miRNAs that are expressed during de novo productive infection by a human virus. While it is possible that these RNAs may be aberrant transcripts, several lines of evidence suggest that they are authentic miRNAs encoded by HCMV. First, all three miRNAs (miR-UL23-5p, miR-UL23-3p and miR-US24) were cloned from a pool of small RNAs isolated and fractionated from cells infected by HCMV. The cloned sequences were found to align to two regions within the HCMV genome and their predicted pre-miRNA transcripts were shown to be capable of adopting hairpin structures by in silico analysis (Fig. 1 and Table 1). Second, all these miRNAs were detected as ∼22 nt transcripts in the examined cell types permissive to HCMV infection, and their expression was found exclusively in HCMV-infected cells (2, 4). Third, the sequences coding for these miRNAs are identical among the HCMV strains for which complete genome sequences have been determined (Fig. 3). Additionally, the miRNAs were found to be expressed from viruses derived from both a laboratory-adapted strain, Towne, and a low-passaged clinical isolate, NJ101 (Fig. 2). Thus, the observations made in this report suggest that these RNA species represent HCMV-encoded miRNAs.
The cloning procedure reported here is only an initial step towards the identification of HCMV miRNAs. It is possible that HCMV may encode many other miRNAs in addition to the three reported in this study, which were isolated from specific cell types during lytic viral infection. In our study, three different viral miRNA sequences were found by analysis of more than 1390 clones that were derived from RNA samples isolated from HCMV-infected fibroblasts and U373MG cells. Further cloning studies, including using RNAs collected at different stages of viral infection and from different cell types, may identify additional viral miRNAs that are expressed only during very narrow time windows in the virus’ life cycle and only in a cell type-specific fashion.
Genomic analyses of the viral miRNAs in our study suggest that they share some general properties with human and animal miRNAs, but possess some unique characteristics. Our data indicate that the distribution of the viral miRNAs within the HCMV genome mirrors that of animal miRNAs in that multiple miRNAs arise from definitive loci within the genome, but that the loci themselves are separate and distinct from each other. Because the distance between miR-UL23 and miR-US24 is approximately 190 kb, it is unlikely that the two groups of miRNAs arise from the same pri-miRNA transcript or that they share a common promoter. Thus, the miRNAs may represent two distinct classes of HCMV miRNAs with different targets and roles in HCMV productive infection. The different classes of HCMV miRNAs are also highly conserved among the HCMV strains and chimpanzee CMV, suggesting that these viral miRNAs may also be expressed during productive infection of animal CMVs other than HCMV.
An immediate question from our results is what are the functions of the miRNAs expressed during productive viral infection. Although previous miRNAs have been implicated in the latent life cycle of other herpesviruses (Pfeffer et al., 2004), the miRNAs identified here are perhaps better examples of miRNAs involved in activities important for viral dissemination and replication within the host population. The capacity to enter latency is an effective strategy of the herpesvirus to establish its presence within the host population. Meanwhile, the ability to actively infect and replicate is essential for the virus to produce infectious viral progeny. Therefore, from the perspective of facilitating a productive viral infection, a very likely function would be the downregulation of the expression of innate host antiviral factors. In terms of viral targets, these miRNAs may facilitate productive viral infection by modulating the expression of viral determinants important for replication, such as those temperance factors that suppress HCMV growth and replication (Dunn et al., 2003). Indeed, an EBV-associated miRNA is proposed to support viral latent infection by shutting down the expression of the viral polymerase by targeting its mRNA (Pfeffer et al., 2004). Thus, it is of significant interest to identify and validate the targets of these miRNAs. Several host targets (Table 3) have been predicted using a computational approach recently developed for miRNA target analysis (John et al., 2004). Further studies to validate these targets and identify other novel targets should reveal whether they may represent key universal components of cellular antiviral response which the virus needs to modulate in order to mount a successful infection.
Human and animal miRNAs function to modulate gene expression by base pairing with their target mRNAs, inducing the RISC complex of the RNAi pathway, and leading to mRNA cleavage and translation repression (Bartel, 2004). However, the unique nature of viral miRNAs facilitating a productive viral infection introduces the prospect that these miRNAs do not actually participate in the RNAi pathway to downregulate protein expression. It is possible that these miRNAs may function as decoys that overwhelm the Dicer complex and RISC, which are also involved in processing small interfering RNAs (siRNAs) in the RNAi pathway (Zeng et al., 2003; Bartel, 2004). We raise this issue because RNAi is believed to be an ancient antiviral mechanism. Upon induction by natural or synthetic siRNAs, the RNAi pathway is effective in blocking gene expression and replication of human viruses including herpesviruses (Scherer and Rossi, 2003). Indeed, anti-RNAi functions have been found in viruses from plants and more recently, from animals (Carrington et al., 2001; Li et al., 2002; 2004). It will be interesting to determine whether overexpression of these miRNAs diminishes the induced RNAi effect in HCMV-infected cells. While synthetic siRNAs are being considered as promising antiviral therapeutics (Scherer and Rossi, 2003), developing novel molecules to shut down the expression of these counter-RNAi miRNAs may represent a novel approach for blocking viral infections. Further studies on identification of the viral miRNA targets and how they interact with RNAi components will elucidate the function of these miRNAs in supporting viral infection and the mechanism of how they achieve their functions.
Our identification of viral miRNAs expressed during productive infection substantiates the initial discoveries in EBV and HIV persistently infected cells (Omoto et al., 2004; Pfeffer et al., 2004) that bona fide miRNAs exist in the primitive virus world. Furthermore, the conservation of these miRNAs among human and chimpanzee CMVs suggests that these viral miRNAs may be an ancient component within the HCMV genome. This notion is consistent with our observations that the degree of sequence conservation of these miRNAs is even greater than those of ‘core’ HCMV proteins, which are highly conserved among all herpesviruses, and have no sequence homologues in the human genome. Thus, like human and animal miRNAs, which are a component of the RNAi pathway, viral miRNAs may also represent a primordial feature in the viral genome that has co-evolved with the host. HCMV's ability to infect and replicate in many tissues in vivo requires a multitude of genes to address the antiviral responses employed by the host. Given the limited coding capacity of a viral genome, even one as large as HCMV, it would be unlikely that the virus contends with every tissue-specific response to viral infection with a protein-based molecule. Thus, miRNAs offer an efficient method to regulate gene expression with minimal investment of viral genomic real estate. Our findings serve to further cement the inclusion of viruses within the ever growing establishment of organisms that encode miRNAs.
Experimental procedures
Virus and cells
Human cytomegalovirus Towne strain (ATCC, Manassas, VA) and human cells were propagated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (Dunn et al., 2001). HFFs and HMVEC were obtained from Clonetics (San Diego, CA), while human RPE cells immortalized with human telomerase reverse transcriptase (hTERT) were purchased from Clontech (Palo Alto, CA) (Dunn et al., 2003). Human astrocytoma cell line U373MG was obtained from ATCC. HCMV TowneBAC was previously constructed by inserting a BAC sequence into the genome of HCMV laboratory-adapted strain Towne (Marchini et al., 2001). HCMV NJBAC, a gift from Hua Zhu (UMDNJ-Medical School, Newark, NJ), was constructed by inserting a BAC sequence into the genome of HCMV clinical strain NJ101. To avoid extensive passaging of the low-passaged clinical NJBAC strain, BAC DNAs were purified from the bacteria carrying the NJBAC sequence and then transfected into HFFs, using the protocols as described previously (Dunn et al., 2003). The NJBAC virus stocks were grown from these transfected cells and were used in the infection experiments.
Cloning and sequencing analysis of the sequences corresponding to HCMV miRNAs
Cloning of miRNA sequences was performed using a modified version of the protocol as described previously (Lau et al., 2001). Briefly, HFFs and human astrocytoma cell line U373MG were infected with TowneBAC at an moi of 1. At 96 h after infection, cells were harvested and total RNAs extracted using the Ambion RNA extraction kit (Ambion, Austin, TX), following the manufacturer's recommendations. The 96 h time point was chosen to allow the virus to undergo at least one complete lytic replication cycle, which takes about 72 h (Mocarski and Courcelle, 2001), and to initiate a second round of infection. The rationale of this design was to obtain RNAs from infected cells representing each phase of the viral lytic replication cycle.
Total RNA was then extracted with the Ambion mirVana miRNA isolation kit (Cat♯ 1560) using the total RNA recovery protocol (Ambion). The direct cloning of the miRNA sequences was carried out using a modified version of the protocols described previously (Lagos-Quintana et al., 2001; Lau et al., 2001). Total RNA (250 µg) was run on 15% denaturing polyacrylamide gels and the small RNAs between 18 and 24 nt in size were collected. 32P-end-labelled RNA markers of 18 and 24 nt were run with the total RNA sample to delineate the gel region containing the RNAs of interest. The gel was exposed on film to visualize the location of the radiolabelled markers. The regions between the markers were excised and the RNA was eluted from the gel and precipitated. The 3′ end donor oligonucleotide oligo3′d (5′-CTGTAG GCACCATCAA-inverted dT-3′) was then ligated to the RNA pool and run on the gel (Lagos-Quintana et al., 2001; Lau et al., 2001). The ligated RNA was again extracted from the gel and the 5′ end donor oligonucleotide oligo5′d (5′-ATCGTaggcacctgaaaa-3′, lower case letters represent ribonucleotides) ligated to the 5′ end of the RNA pool (Lagos-Quintana et al., 2001; Lau et al., 2001). The RNA pool containing the 3′ and 5′ adaptor oligonucleotide sequences was reverse transcribed and the products were digested with BanI. The digested fragments were concatamerized with T4 DNA ligase and the products run on 2% agarose gels. Concatamerized fragments ranging from 300 to 800 nt were extracted from the agarose gel. The ends of the concatamers were filled in with Taq polymerase and cloned into the TA-TOPO vector (Invitrogen, Carlsbad, CA). Plasmid DNAs were isolated from these clones and sequenced at the sequencing facility at UC-Berkeley.
Detection of human and viral miRNAs using Northern analysis
Cells were infected with HCMV at an moi of 1, and harvested at different time points after infection. To detect the expression of immediate-early (α), early (β) and late (γ) transcripts, cells were also treated with cyclohexamide (50 µg ml−1) and PAA (100 µg ml−1), respectively, before and during viral infection. Total RNAs were isolated from the cells using the Ambion RNA extraction kit, following the manufacturer's recommendations, separated on 12.5% polyacrylamide gels that contained 8 M urea, electrically transferred to nitrocellulose membranes, and hybridized to radiolabelled chemically synthesized oligonucleotides that were complementary to specific human and HCMV sequences. The hybridized membranes were analysed using a Molecular Dynamics Storm840 phosphorimager.
The oligonucleotide probes used to detect miRNA miR-UL23-5p, miR-UL23-3p and miR-US24 are oligoUL23-5P (5′-TCT CACGGGAAGGCTAGTTA-3′), oligoUL23-3P (5′-TACAAACTAG CATTCTGGTGA-3′) and oligoUS24 (5′-CGGTCCGAGCCACT GAGCGGTT-3′) respectively. Oligonucleotide probes oligoUL23-5P-s (5′-TAACTAGCCTTCCCGTGAGA-3′), oligoUL23-3P-s (5′-TCACCAGAATGCTAGTTTGTA-3′) and oligoUS24-s (5′-AACCG CTCAGTGGCTCGGACCG-3′) were specific for the sequence complementary to that of the cloned miRNA candidates and were used as negative controls for the detection of miR-UL23-5p, miR-UL23-3p and miR-US24 respectively. Oligonucleotide probe oligoR-16 (5′-CGCCAATATTTACGTGCTGCTA-3′) was used to detect the expression of human miRNA miR-16 (Lagos-Quintana et al., 2001). All probes were chemically synthesized and 3′ end 32P-labelled using Starfire system (IDT, Coralville, IA).
Time-course viral infection experiments
To determine the expression profile of HCMV miRNAs during the three phases of the lytic viral replication cycle, infected HFFs were treated with cyclohexamide and PAA before and during viral infection. Immediate-early (α) viral gene expression is defined by the set of viral genes expressed before de novo protein synthesis, and are the first set of genes to be transcribed after viral entry. Treatment with cyclohexamide blocks protein translation but allows transcription to occur and, therefore, does not affect immediate-early gene expression (Mocarski and Courcelle, 2001). To determine whether the viral miRNAs are expressed during this immediate-early phase, HFFs were treated with cyclohexamide at a concentration of 50 µg ml−1. Twenty-four hours after the addition of the viral innoculum, the infected cells were harvested and total RNAs were isolated.
To determine whether HCMV miRNAs were expressed during the early phase of the viral life cycle, PAA was added to the cell culture to block viral DNA synthesis. In the absence of viral DNA synthesis, the expression of viral late (γ) genes is inhibited, and only the expression of immediate-early (α) and early (β) genes is not affected (Mocarski and Courcelle, 2001). Human fibroblast cells were treated with PAA at a concentration of 100 µg ml−1. The infected cells were harvested at 48 h after infection and total RNAs were isolated for Northern analysis of viral mRNA expression.
Late phase (γ) gene expression is defined by all the viral genes that are not classified as immediate-early or early, and are expressed after viral DNA replication (Mocarski and Courcelle, 2001). To determine whether the HCMV miRNAs were also expressed during the late phase of viral infection, human fibroblasts were infected and total RNA was harvested 72 h after infection.
Bioinformatic analysis of HCMV miRNAs
In silico analysis using MFOLD 3.1 (Zuker, 2003) was performed to determine whether the pre-miRNA transcripts that were located in the UL23 and US24 non-coding regions of HCMV Towne adopted low-energy stem-loop secondary structures (GenBank Accession No. AY315197) (Dunn et al., 2003). Human 3′ untranslated region (UTR) sequences were extracted from both NCBI and Ensembl databases using the BioPerl application programming interface (Birney et al., 2004). The miRanda target-prediction algorithm was used to predict potential target sequences with the following requirements: miRanda-specific score-threshold of 125, ΔG ≤ −20.0 kcal mole−1 and perfect seed homology (nucleotides 2–8 on the 5′ end of the miRNA) (John et al., 2004).
Acknowledgements
We thank Hua Zhu for HCMV clinical strain NJBAC, and Yong Bai and R. Tyler Hillman for technical assistance. W.D. is a recipient of a Predoctoral Fellowship of State of California Universitywide AIDS Research Program. F.L. is a Scholar of the Leukemia and Lymphoma Society, and an Established Investigator of the American Heart Association. This study was supported by a Chancellor's Initiative Grant (UC-Berkeley) and NIH.
