MicroRNAs (miRNAs) are a family of ∼22-nucleotide (nt) noncoding RNAs that are capable of binding to specific target mRNAs and inhibiting their expression (reviewed in reference
1). They are typically derived from one arm of RNA stem-loops found within noncoding regions of capped and polyadenylated transcripts (
4,
26). Successive cleavage of these hairpin structures by the RNase III enzymes Drosha in the nucleus (
25) and Dicer in the cytoplasm (
7,
20) generates a miRNA duplex structure of ∼20 bp with 2-nt 3′ overhangs. One arm of this duplex is then loaded into the RNA-induced silencing complex (RISC), where it is used as a guide to target complementary transcripts for inhibition (
19,
28). In mammalian cells, miRNAs usually guide the RISC to imperfectly complementary target sites, resulting in the translational arrest of bound mRNAs and a modest but detectable mRNA destabilization (
12,
31,
43).
Due to their small size and nonimmunogenic nature, miRNAs appear ideally suited for use as regulatory molecules by viruses, and indeed, a number of human DNA viruses, including many herpesviruses, have now been reported to encode miRNAs (
39). Herpesviruses can be divided into three subfamilies, the alpha-, beta-, and gammaherpesviruses, based on replication characteristics, genomic organization, and preferred latency sites. Members of all three subfamilies have been found to encode miRNAs, ranging from a low of 3 in the alphaherpesvirus herpes simplex virus 2 (HSV-2) (
37,
38) to a high of 25 in Epstein-Barr virus (EBV) (
5,
17,
33,
46). The fact that all herpesviruses examined to date express miRNAs suggests that miRNAs play important roles in the herpesvirus life cycle, and several studies have in fact demonstrated the downregulation of cellular and/or viral mRNA targets by herpesvirus miRNAs (reviewed in reference
16).
HSV-1 and varicella-zoster virus (VZV) are pathogenic human viruses both of which belong to the alphaherpesvirus subfamily. HSV-1, the prototypic alphaherpesvirus, typically initiates productive replication in the mucosal epithelia of the face and establishes latency in neurons of multiple cranial nerve ganglia, including the trigeminal ganglia (TG) (
35,
41). VZV replicates in the mucosal epithelia of the respiratory tract and establishes latency not only in the TG but also in the dorsal root and autonomic ganglia (
15). During latency, transcription of the HSV-1 genome is largely restricted to a single RNA: the latency-associated transcript (LAT) (
2,
36). Although the LAT is capped and polyadenylated, it does not appear to encode a protein. The LAT is unusual in that the spliced ∼6.3-kb transcript is highly unstable, while the single ∼2-kb intron accumulates to significant levels within latently infected cells, although its function remains unknown (
13,
21).
Previously, we used deep sequencing of RNA harvested from murine TG latently infected with HSV-1 to demonstrate that HSV-1 expresses at least five miRNAs, four of which (miR-H2, miR-H3, miR-H4, and miR-H5) are derived from the unstable exonic regions of the LAT (
40). A fifth miRNA, miR-H6, was found to lie in the opposite transcriptional orientation, just upstream of the LAT and antisense to another HSV-1 miRNA, miR-H1, that is expressed exclusively during productive replication (
11). Two of the HSV-1 LAT-derived miRNAs, miR-H2 and miR-H6, have been found to downregulate the expression of the viral proteins ICP0 and ICP4, respectively (
40). ICP0 and IPC4 are viral immediate-early proteins that function as potent activators of productive HSV-1 replication (
3,
18,
34), and it has been hypothesized that their repression by miR-H2 and miR-H6 might facilitate the establishment and/or maintenance of the latent state in HSV-1-infected neurons (
40).
In order to identify viral miRNAs that are expressed by HSV-1 or VZV in latently infected human neurons in vivo, we used Solexa/Illumina technology for deep sequencing of cDNA libraries prepared from postmortem human TG samples naturally infected with latent HSV-1 and/or VZV. Based on these data, we demonstrate the in vivo expression of the five previously reported HSV-1 miRNAs, miR-H2 to miR-H6, and we identify two novel HSV-1 LAT-derived miRNAs, miR-H7 and miR-H8. Surprisingly, despite the successful recovery of large numbers of HSV-1 miRNAs from the same samples, no latently expressed VZV miRNAs were identified.
DISCUSSION
Although HSV-1 and VZV both readily establish productive infections in culture, these viruses are able to establish latency only in primary neurons, an environment that has so far proven impossible to fully recapitulate in culture. While HSV-1, but not VZV, can also establish latency in vivo in the sensory ganglia of laboratory animals, such as mice and rabbits, these model systems do not faithfully mimic the latent infections seen in humans, the only natural host for HSV-1. We therefore felt that it was important to assess the miRNA profiles of these viruses in naturally infected human neurons by using deep sequencing to analyze RNA samples derived from human TG latently infected with HSV-1 and/or VZV.
Using latently infected human TG obtained postmortem, we were able to identify two new HSV-1 miRNAs (Fig.
1A and Table
3). The existence of one of these novel HSV-1 miRNAs, miR-H7, was verified in both latently and productively infected cells by qRT-PCR (Fig.
2), while the other miRNA, miR-H8, was detected only in productively infected cells, even though it was initially sequenced from latently infected TG (Table
3; Fig.
2). Overall, there was only a modest level of correlation in miRNA expression levels as determined by deep sequencing (Table
3) or qRT-PCR (Fig.
2A). This may reflect differences in linker ligation efficiency during cDNA synthesis, differential PCR amplification of the cDNA library during sample preparation and sequencing, and/or different annealing efficiencies of the primers used for the qRT-PCR analysis.
The genomic locations of miR-H7 and miR-H8 place both antisense to the first intron of
ICP0 (Fig.
1B). This location suggests that they are unlikely to downregulate ICP0 expression as reported for miR-H2, which lies antisense to an exonic region of
ICP0 (Fig.
1B), since RNA interference is thought to operate exclusively in the cytoplasm (
42). However, there is a report of alternative splicing of
ICP0 intron 1, which generates as many as four different introns of variable size, although the significance and function, if any, of these alternatively spliced
ICP0 mRNAs are unknown (
6). It is possible that miR-H7 and miR-H8 either act directly on these splice variants or, conversely, act on currently unknown cellular mRNA targets, as previously reported for several other herpesvirus miRNAs (
16).
Analysis of the mature HSV-1 miRNA sequences recovered revealed a significant level of sequence heterogeneity at both the 5′ ends and (especially) the 3′ ends of some of the mature HSV-1 miRNAs (Table
4). The 5′ region of miRNAs, especially nt 2 to 8, has been shown to be particularly important for target mRNA binding, and full sequence complementarity to this so-called “seed” region is usually, but not invariably, essential for mRNA translational inhibition (
1). Therefore, even small sequence differences at the 5′ end of the miRNA, as seen with miR-H2-3p, miR-H6-3p, and miR-H7-5p (Table
4), have the potential to affect the identity of the mRNAs targeted by these viral miRNAs. A comparison to cellular miRNA sequences obtained from the same RNA samples in the same sequencing run showed very little sequence variation at the 5′ end, with 96.0% of the 3,634,951 individual reads of let-7a and 96.3% of the 1,489,388 reads of let-7b recovered from these two HSV-1-positive small-RNA samples showing the same, predicted 5′ end (data not shown). Therefore, it is unlikely that the sequence variation observed at the 5′ ends of miR-H2-3p and miR-H6-3p, but not at the 5′ end of miR-H3-3p or miR-H4-5p (Table
4), is due to degradation during RNA isolation or some other sequencing artifact. It is also noteworthy that the minor 5′ sequence variants observed with miR-H6-3p and miR-H7-5p are actually 1 nt longer than the consensus sequence (Table
4), which is clearly inconsistent with exonucleolytic RNA degradation. Of note, because miR-H2-3p lies antisense to its only known target,
ICP0 (Fig.
1B), a 1-nt change at the 5′ end would not affect the level of miRNA complementarity to this transcript. In contrast, a 1-nt addition at the 5′ end of miR-H6, which has been reported to target ICP4 mRNAs via the seed region (
40), would be predicted to disrupt this interaction and hence may attenuate ICP4 downregulation.
The lack of detectable VZV-derived miRNAs in any of the three VZV-infected TG samples was unexpected, given that all members of the herpesviruses superfamily examined thus far, including primate, murine, and avian herpesviruses, have been found to encode multiple viral miRNAs (
39). It is possible that, despite analyzing more than 1.6 × 10
7 cDNA sequence reads, this method was still not sensitive enough to detect VZV miRNAs, even though ∼3,000 HSV-1 miRNAs were recovered from the same samples (Table
3). As is commonly the case in coinfected TG, the HSV-1 genomic DNA load was modestly higher than that of VZV (Table
2) (
10), but this minor difference seems unlikely to explain the lack of VZV miRNAs reported here. Interestingly, a previously reported computational analysis of the genomes of HSV-1, HSV-2, and VZV, as well as of those of several other herpesviruses, predicted that HSV-1 and HSV-2 would encode multiple viral miRNAs but that VZV was unlikely to produce viral miRNAs in infected cells (
32).
It should also be noted that the mechanism underlying VZV latency is thought to differ significantly from that observed for HSV-1. During HSV-1 latency, the only transcript expressed at significant levels is the LAT, which is contained within the repeat regions of the viral genome (Fig.
1B). Although the VZV genome shares significant homology with HSV-1, VZV lacks sequences complementary to much of the HSV-1 long-internal-repeat and long-terminal-repeat regions (Fig.
1B) (
29) and is therefore not predicted to encode a LAT (
10). In addition, VZV differs from HSV-1 in that it expresses at least five viral protein-coding mRNAs during latency, albeit at low levels (
8,
9,
22). Since most of the HSV-1 miRNAs are LAT derived, it is perhaps not surprising that VZV does not appear to express any latency-specific viral miRNAs. The apparent lack of VZV latency-associated miRNAs, however, does not preclude the possibility of expression of VZV miRNAs during productive replication, as is indeed observed with HSV-1 miR-H1 (
11) (Fig.
2).