Numerous Conserved and Divergent MicroRNAs Expressed by Herpes Simplex Viruses 1 and 2

Human embryonic kidney HEK-293 cells (ATCC, CRL-1573) and African green monkey Vero cells (ATCC, CCL-81) were maintained at 37°C in Dulbecco's modified Eagle's medium (Mediatech, Inc.) supplemented with 10% fetal bovine serum or 5% newborn bovine serum, respectively (Sigma-Aldrich). HSV-1 strain KOS (henceforth referred to as HSV-1) and a thymidine kinase (TK)-negative mutant of HSV-2 strain 186syn⁺,186ΔKpn (30) (henceforth referred to as HSV-2) were propagated and titrated on Vero cells as previously described (14).

For productive infections, HEK-293 cells were infected with HSV-1 or HSV-2 at a multiplicity of infection (MOI) of 5 or 1, respectively, for 18 h. For latent infection, 7-week-old male CD-1 mice (Harlan) were corneally inoculated with 2 × 10⁶ PFU HSV-1 in a volume of 5 μl as previously described (34) or infected intranasally with 1 × 10⁶ PFU of HSV-2 as previously described (30). Trigeminal ganglia were harvested 30 days after infection and immediately frozen as previously described (33).

RNA enriched for small RNA species was extracted from infected cells or trigeminal ganglia using the mirVana kit (Ambion) according to the manufacturer's recommended protocol. For each sample, 10 μg of small-enriched RNA was size fractioned by polyacrylamide gel electrophoresis (PAGE) on a 15% denaturing gel (19:1 acrylamide-Bis solution; Bio-Rad) containing 8 M urea in 0.089 M Tris-0.089 M borate-0.002 M EDTA buffer. RNAs that were 18 to 30 nucleotides were identified by comparison to an SRA ladder (Illumina) and excised from the gel. The gel slices were shredded, and RNA was eluted from the gel for 4 h with 300 μl of 0.3 M NaCl. Gel debris was removed using Spin X cellulose acetate filters (Illumina), and the RNA was precipitated with 1 μl glycogen and 750 μl of 100% ethanol. After being washed with 75% ethanol, the RNA pellet was resuspended in H₂O. Sequencing libraries were created by successive adapter ligation (SRA 5′ and SRA 3′ adapter; Illumina) to the fractioned RNA, followed by cDNA synthesis and amplification by PCR according to the manufacturer's manual (small RNA sample prep kit; Illumina). The libraries were sequenced on the Genome analyzer II (Illumina) at the Biopolymers Facility, Department of Genetics, Harvard Medical School.

The sequences from each library, which consist of 36 nucleotide reads, were first processed with the short oligonucleotide alignment program, SOAP (43), to identify and remove the 5′ flanking sequencing adapter sequence, allowing no mismatches in the adapter. The sequences of the resulting 16 to 26 nucleotide reads from each library were then aligned with SOAP to reference sequences of the HSV-1 strain 17+ genome (NC_001806) or HSV-2 strain HG52 genome (NC_001798), allowing two nucleotide mismatches between sequence reads and the reference genome. Genomic loci represented with three or more sequence reads of ≥19 nucleotides between HSV-1 and HSV-2 in productive or latent infection were then tested by the RNA secondary structure predictions programs mfold and RNAfold (27, 69) for the lowest-free-energy hairpin structures (1). For the hairpin structure analysis, 150 nucleotides of genomic sequence surrounding the locus of the mapped reads was tested.

Vero cells were mock infected or infected with HSV-1 or HSV-2 at an MOI of 10 or 1, respectively. Infected cells were harvested at 18 h postinfection (hpi), and small enriched RNA was prepared as described above. RNA samples, together with a ³²P-labeled Decade marker (Ambion), were separated by 15% denaturing PAGE as described above. Gels were soaked briefly in SYBR gold nucleic acid gel stain solution (Molecular Probes). The RNA was then transferred to a BrightStar-Plus positively charged membrane (Ambion) by electroblotting, cross-linked to the membrane using UV light, and prehybridized in ULTRAhyb-Oligo solution (Ambion) for 30 min at 62°C. DNA oligonucleotides (Integrated DNA Technologies) with sequences complementary to miRNAs (see supplemental Table 1 at https://coen.med.harvard.edu ) were end labeled with [γ-³²P]ATP (MP Biomedicals) and purified using the QIAquick nucleotide removal kit (Qiagen). Overnight hybridizations were performed in ULTRAhyb-Oligo solution (Ambion) at 32°C. Hybridized blots were washed three times briefly and then three times for 15 min in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.5% SDS at 32°C and exposed to a phosphorimager screen. Signals were detected using the Bio-Rad personal molecular imager system. Analyzed blots were stripped using boiling 0.1% SDS solution as previously described (15), exposed to a phosphorimager screen to confirm complete removal of the probe, and successively rehybridized with different probes.

To more completely define the various miRNA species expressed by HSV-1 and HSV-2 during productive and latent infections, we applied massively parallel sequencing of small-RNA libraries using the Illumina Genome Analyzer II. HSV-1 infections were performed with wild-type strain KOS, while HSV-2 infections were performed with 186ΔKpn, a thymidine kinase-negative mutant virus that can establish latent infection in trigeminal ganglia without lethality to mice (30). RNA species of 15 to 30 nt were isolated either from productively infected HEK-293 cells or from latently infected trigeminal ganglia. The RNA was ligated to sequencing adapters, reverse transcribed, and PCR amplified to generate libraries corresponding to small RNA sequences, and sequenced. We acquired between 5 × 10⁵ and 6 × 10⁶ sequence tags from the different samples (see supplemental Table 2 at https://coen.med.harvard.edu ). The number of reads for two known cellular miRNAs correlated with the number of all reads obtained in each sample, suggesting that the four sequencing analyses were of similar quality (see supplemental Table 2 at https://coen.med.harvard.edu ). Each sequence was 36 nt in length, including adapter sequences, with 13 to 26 nt corresponding to RNA. We limit our analysis here to reads of 19 nt or longer that correspond to sequences within HSV genomes. We were able to assign 6,682 and 856 reads to HSV-1 and 1,6883 and 2,177 reads to HSV-2 from productive and latent infections, respectively (see supplemental Table 2 at https://coen.med.harvard.edu ). HSV-specific reads were further analyzed manually. A sequence was considered an miRNA candidate when all of three criteria were satisfied. (i, Multiple reads) A particular genomic locus was represented with 3 or more sequencing reads among the four samples. (ii, Hairpin) The sequence was within a hairpin formed by the 150 nucleotides centered on the candidate miRNA and that hairpin was predicted to have lowest free energy by both mfold and RNAfold (1, 27, 69). (iii, Hairpin-loop proximal) When two possible miRNAs were present in the same lowest-energy hairpin, only the one closest to the loop of the hairpin was considered a candidate. Sixty-seven and 363 genomic loci from HSV-1 and HSV-2, respectively (see supplemental Tables 3 and 4 at https://coen.med.harvard.edu ), fulfilled criterion 1 and were subsequently tested by mfold and RNAfold. This resulted in selection of 16 and 17 hairpin precursors of miRNAs encoded by HSV-1 and HSV-2, respectively (Tables 1 and 2). All recovered sequence reads of the HSV-1- and HSV-2-encoded miRNAs are listed in supplemental Tables 5 and 6, respectively, at https://coen.med.harvard.edu, and the genomic loci are presented in Fig. 1. When there were different sequences obtained for given miRNAs, the sequences shown in Tables 1 and 2 are the most abundant reads found. We will first summarize the HSV-1 miRNAs, starting with those that have been previously identified, and then discuss the HSV-2 miRNAs, the homologies between HSV-1 and HSV-2 miRNAs, and particular miRNAs of interest.

At the time we initiated this study, six HSV-1 miRNAs had been experimentally identified, miR-H1 to miR-H6 (15, 64). As an initial check on our sequencing analysis, we first asked if it identified these 6 miRNAs. Indeed, all 6 were identified, including both mature miRNAs and their corresponding star strands (Tables 1 and 3). Previously, we had not detected star strands for miR-H1, miR-H3, and miR-H5 (15, 64). All of these miRNAs (but not necessarily both strands) were found in both productively and latently infected samples; however, the relative proportions of sequencing reads in the two samples differed considerably. We caution that the relative abundance of sequence reads does not necessarily reflect the relative abundance of one miRNA versus another, as the technique appears to capture certain miRNAs more efficiently than others. However, when the number of reads for one miRNA is greater than the number of reads for a second miRNA in one sample, but lower in another sample, that is likely to reflect differential expression in the two samples. In the HSV-1 productive infection sample, reads for miR-H1 and H6 were the most abundant, whereas in the latent samples they were greatly outnumbered by reads for miR-H2, -H3, and -H4 (Table 1). The analysis also provided the exact sequence of HSV-1 miR-H1, which had previously been detected only by Northern blot hybridization (15). A number of reads corresponding to the original predicted miR-H1 sequence were recovered (see supplemental Table 5 at https://coen.med.harvard.edu ). Additionally, reads with mismatches and/or 3′-terminal nontemplated A or U were frequently detected. However, the most abundant miRNA sequence read, which starts 2 nucleotides upstream from the miR-H1 sequence predicted by Cui et al. (15), was selected as HSV-1 miRNA-H1. We will discuss miR-H1 and miR-H6 in more detail in a subsequent section.

Next we asked whether HSV-1 miRNAs whose existence had previously been predicted, but not experimentally verified, were present. Five such miRNAs were found that fulfilled our three criteria, corresponding to predictions by Pfeffer et al. (49), called T-1, T-4, and T-7 by Cui et al. (15), and two predictions by Cui et al. (15), called #6 and #6-RC (Table 1). One of these (T-1) was recently detected in latently infected human ganglia (65) and designated as HSV-1 miR-H8. We designate the other four as HSV-1 miR-H11 (T-4), H12 (#6-RC), H13 (#6) and H14 (T-7). All four of these miRNAs could be detected in productive infection, including both the mature miRNA and the star strand for miR-H14 (Table 1). Among these four, only miR-H8 was also detected in latently infected mouse ganglia (Tables 1 and 3). MiR-H8 is encoded within the second LAT exon, antisense to the first intron of ICP0, consistent with its detection in latently infected ganglia, whereas miR-H14 is encoded within the third exon of ICP0 mRNA, antisense to miR-H2. Interestingly, miR-H11, -H12, and -H13 are encoded within viral origins of replication (oriL and oriS) and could be derived from previously reported transcripts that span the origins (7, 28, 68). We note that Cui et al. predicted that miR-H13 (#6) would be conserved in HSV-2 (15). We found only one read corresponding to this predicted miRNA from our HSV-1 samples, but found numerous sequencing reads in our sample from HEK-293 cells productively infected with HSV-2, thus fulfilling our multiple-reads criterion; this sequence was also hairpin associated and proximal to the hairpin loop (Tables 1 and 2). We will discuss the origin-encoded miRNAs (miR-H11 to -H13) in more detail below.

Three loci predicted by Cui et al., called #5, #10, and #11, were each recovered as a single sequence read in productively infected samples (see supplemental Table 3 at https://coen.med.harvard.edu ); however, these reads fulfilled neither the multiple-reads criterion nor the hairpin criterion and were not designated as HSV-1 miRNAs. The failure of these loci to fulfill the hairpin criterion here, while fulfilling a similar criterion previously (15), might be due to differences in the sequences tested between the two studies. Regardless, we cannot exclude the possibility that the recovered reads are bona fide miRNAs as predicted by Cui et al.

We also identified 5 additional HSV-1 miRNAs, HSV-1 miR-H7, which has also recently been detected in latently infected human ganglia (65), and miR-H15 to -H18 (Tables 1 and 3). HSV-1 miR-H7, like miR-H8, is encoded within LAT. It is antisense to intron 1 of ICP0 transcripts. The proportion of HSV-1 miR-H7 reads from latently infected mouse ganglia was higher than that from productively infected HEK-293 cells, which would be consistent with miR-H7 being derived from LAT. HSV-1 miR-H15, which was detected in productively infected cells, is encoded downstream of HSV-1 miR-H6 on the same strand. HSV-1 miR-H16, which was also detected in productively infected cells, is complementary to UL32 and presumably arises from the 5′ UTR of the transcript encoding UL33. Among the HSV-1 miRNAs that we detected, only HSV-1 miR-H11 and -H16 are encoded within unique sequences rather than repeat sequences.

We detected two additional HSV-1 RNAs in both productively and latently infected samples that fulfilled our three criteria. One is encoded within ICP4 coding sequences, and thus could arise from ICP4 mRNA, which has been detected in productively and in latently infected ganglia at low abundance (10, 32, 33, 48). The other is encoded on the opposite strand from ICP4, between ICP4 and oriS, and could arise from anti-4 transcripts, which we have previously detected in latently infected ganglia, and may arise from LAT readthrough transcripts or others (10). We note that these two RNAs are very G-C rich. Nevertheless, there are numerous precedents for G-C-rich miRNAs in the miRNA database (e.g., human miR-762 or miR-718; http://www.mirbase.org ). Thus, although our identification of these species as miRNAs is somewhat tentative, we have designated them as HSV-1 miR-H17 (from ICP4) and HSV-1 miR-H18 (between ICP4 and oriS).

Similar to results from the HSV-1 analysis, we detected all HSV-2 miRNAs that had been identified (57, 58) when we initiated our study (Tables 2 and 3): HSV-2 miR-H2 (originally termed miR-III), HSV-2 miR-H3 (originally termed miR-I), and both strands of HSV-2 miR-H4 (originally termed miR-II-5p and -3p). One of these miRNAs (HSV-2 miR-H4, miR-II) had been predicted computationally prior to its identification (49). These miRNAs were readily detected in both the productive infection and latent infection samples (Table 2). Tang et al. (58) found that HSV-2 miR-H3 (miR-I) was more abundantly expressed in productively infected cells than HSV-2 miR-H2 and miR-H4 (miR-III and miR-II), which is consistent with the relative abundance of sequence reads found in our study. We have extended their results by showing that HSV-2 miR-H2 and miR-H4 (miR-III and miR-II) can also be detected in latently infected mouse ganglia. Recently, Umbach et al. (66) have detected two of these miRNAs (HSV-2 miR-H3/miR-I and HSV-2 miR-H4/miR-II) in human sacral ganglia infected with HSV-2.

We also found sequence reads representing miRNAs that had previously been predicted by Cui et al. (15), Pfeffer et al. (49), or both, but not experimentally verified (Tables 2 and 3). The first of these, which we designate as HSV-2 miR-H5, was detected as the -5p strand in both productive and latent infection, was called #4 by Cui et al. (15), and is homologous to HSV-1 miR-H5. Only the -3p version of HSV-1 miR-H5 was originally detected (64). However, in our current study, we also detected HSV-1 miR-H5-5p in the latently infected sample (Tables 1 and 2). Tang et al. (58) reported that they could not detect either pre-miRNA or mature miRNA species in HSV-2-infected Vero cells using a probe complementary to the -3p strand of an miRNA predicted to be homologous to HSV-1 miR-H5. However, we were able to detect a species corresponding to pre-miRNA in HSV-2-infected Vero and HEK-293 cells using probes complementary to either strand (Fig. 2). In their study, Tang et al. used a different HSV-2 strain, a different procedure for RNA extraction, and also a slightly different probe, which might explain the discrepancy with our results. Although the HSV-1- and HSV-2-specific probes used in the Northern blot analysis were highly similar in sequence, there was no cross-reaction (i.e., the HSV-1 probe did not detect a signal from HSV-2 RNA and vice versa), indicating a highly specific assay (Fig. 2).

A second HSV-2 miRNA predicted by Cui et al. was termed #6-RC by those authors (15). This miRNA, as discussed above, arises from a viral origin of replication (oriS), and is homologous to HSV-1 miR-H12. We also detected two HSV-2 miRNAs that had been predicted by Pfeffer et al. (49), which, like HSV-1 miR-H12, are homologous to HSV-1 origin-encoded miRNAs. One of these (prediction termed T-4 by Cui et al. [15]) is encoded within a viral origin of replication (oriL) and is homologous to HSV-1 miR-H11. A second (prediction termed #6 by Cui et al. [15]), which is homologous to HSV-1 miR-H13, is also encoded within a viral origin of replication, in this case, oriS. The three HSV-2 origin-encoded miRNAs (miR-H11 to -H13) were detected only in the productive infection sample and will be discussed in more detail below.

A third miRNA predicted by Pfeffer et al. (49) is homologous to HSV-1 miR-H7, antisense to ICP0 intron sequences, and could arise from LATs. However, unlike HSV-1 miR-H7, which was readily detected in latent infection (Tables 1 and 3), HSV-2 miR-H7 was detected only in productive infection (Tables 2 and 3). However, the failure to detect this miRNA in the latently infected sample may be due to poor sensitivity. Recently, Umbach et al. detected one read of this miRNA in sacral ganglia samples from one of three latently infected human subjects (66).

There were also two HSV-2 miRNAs predicted by Pfeffer et al. that we have now experimentally verified and for which we have not detected HSV-1 orthologs. Both of these were detected only in the productive infection sample. The first of these, which we designate as HSV-2 miR-H19, arises from the 5′ UTR of the transcript encoding ICP34.5 and is complementary to miR-H4. The second of these, which we designate as HSV-2 miR-H20, is encoded within an ICP0 intron and is complementary to miR-H7. Of note, we detected one and four reads in HSV-1 productively infected samples that would correspond to homologs of HSV-2 miR-H19 and -H20, respectively. However, these loci did not meet the hairpin criterion and were not regarded as miRNAs.

Of the remaining HSV-2 miRNAs that we detected (Tables 2 and 3), one, which we designate as HSV-2 miR-H6, is a positional homolog of HSV-1 miR-H6, lying upstream of and on the strand opposite to LAT coding sequences. We will discuss this abundantly expressed miRNA in more detail below.

The other newly identified HSV-2 miRNAs, HSV-2 miR-H21 to -H25, are not homologs of HSV-1 miRNAs. HSV-2 miR-H21 to -H23 are encoded within unique long (U_L) sequences and were detected only in the productive infection sample. HSV-2 miR-H21 is encoded within UL3, HSV-2 miR-H22 is encoded antisense to coding sequences of UL36 and might arise from run-through transcripts from the upstream region (UL30 and UL33 to -35 genes), and HSV-2 miR-H23 is encoded within UL42. HSV-2 miR-H9, -H24, and -H25 are encoded within repeated sequences. HSV-2 miR-H9 and -H24 could arise from LAT sequences, and, indeed, we detected HSV-2 miR-H24 in both productive and latent infection samples. Interestingly, Umbach et al. have recently detected the miR-H9 star strand in three sequence reads from a sacral ganglion sample from one of three subjects infected with HSV-2 (66). HSV-2 miR-H25, which was detected in the productive infection sample, is probably encoded within the same oriS-spanning transcript that encodes miR-H13. The sequence of miR-H25 is very G+C rich, as compared to H13, which may account for the different numbers of recovered reads between H25 and H13. Of note, we did not detect any reads corresponding to an HSV-2 miRNA recently reported by Umbach et al., which they termed HSV-2 miR-H10 (66). This species, which is 85% G+C, is encoded downstream of UL15 and was found in sequence reads from sacral ganglia from two of three subjects infected with HSV-2. Our failure to detect this RNA could be due to several possible factors. One interesting possibility would be differences in HSV-2 miRNA expression between human sacral ganglia and mouse trigeminal ganglia.

Among the miRNAs that we detected, there are nine positional and/or sequence homologs shared between HSV-1 and HSV-2 (Table 4). Of these nine miRNAs, we found both -5p and -3p strands from both viruses for two. It is interesting to compare the sequences of these orthologs, particularly conservation of the seed region, a major determinant of miRNA function. We first asked what fraction of the seven bases in the seed region were identical between orthologs; a high fraction of identity strongly suggests conserved targets for the orthologous miRNAs. However, even when sequences are not identical between orthologs, when the differences in sequence are G:A or C:U, it is possible that these miRNAs still recognize the same mRNA targets (i.e., by G-U and A-U pairing and C-G and U-G pairing). Therefore, we also assessed “seed consistency” (41, 42), the fraction of bases in the seed region of orthologous miRNAs that can retain their ability to base pair with a common target sequence (Table 4). We note that for some miRNAs, the sequence reads differ at their ends, and for the conservation analysis, we used the most abundantly found reads between productive and latent infection (Table 4).

Tang et al. (58) described HSV-1 and HSV-2 miR-H2 (HSV-2 miR-III) as having minimal sequence similarity. Our analysis shows that they share 72% identity overall and completely identical seed regions. This raises the possibility that the miR-H2 orthologs not only can repress expression of ICP0, to whose mRNA they are completely complementary, but that they may repress the expression of other conserved targets as well.

As noted by Tang et al. (57, 58), comparison of the miR-H3 and miR-H4 orthologs of HSV-1 and HSV-2 reveals incomplete sequence identity in the seed region (Table 4). This can be interpreted (57, 58) as implying that these miRNAs are unlikely to have conserved targets other than ICP34.5, whose mRNA is complementary to these miRNAs. However, the miR-H4-3p orthologs exhibit 6/7 seed identity and 7/7 seed consistency, and the miR-H3-3p orthologs exhibit 5/7 seed identity and 6/7 seed consistency. Thus, these orthologs might have conserved targets other than ICP34.5.

Similarly, HSV-2 miR-H5 shares 74% overall sequence identity with HSV-1 miR-H5, including six of seven bases in the seed region and 7/7 seed consistency (Table 4), which might suggest conserved targets. Of the remaining orthologs, all but one (miR-H6; see below) show considerable seed identity (≥6/7 identity), and all show maximal seed consistency (7/7), again suggesting conserved targets for these orthologs.

The miR-H5 orthologs, which are expressed in latent infection, are not completely complementary to any known viral transcript. Their well-conserved seed regions are very G rich. Perhaps they interact with mRNAs with homopolymeric runs of C, of which there are several HSV examples (12, 46), or, given the unusual structures that runs of G's can form, perhaps they function using interactions other than Watson-Crick base pairing. The miR-H7 orthologs are also expressed in latency (66; this report). These miRNAs are highly homologous (86% identical with 7/7 identity in the seed region). Interestingly, they are antisense to ICP0 first intron sequences and, as pointed out by Umbach et al. (66), are therefore unlikely to downregulate ICP0 expression. We have previously described ICP0 transcripts that contain the first intron in latently infected ganglia (11), which may be related to similar transcripts found in productively infected cells (9). However, mutations that drastically reduce LAT expression, and would therefore be expected to drastically reduce miR-H7 expression, did not result in an increase in the levels of these intron-containing transcripts (11), as would be expected if miR-H7 downregulates their expression at the transcript level. Thus, especially given the high identity of the seed region of the two orthologs, we favor miR-H7 having other conserved targets, either viral, cellular, or both.

The miRH1/6 locus lies about 300 bp upstream of the LAT transcription start site. In HSV-1, miR-H1 and miR-H6 are derived from hairpins that are encoded on opposite strands (Fig. 1 and 3). As noted above, the most abundant sequence reads in the HSV-1 productive infection sample were for HSV-1 miR-H1-5p, consistent with its ready detection by Northern blot hybridization (15). Interestingly, we found 11 reads of HSV-1 miR-H1-5p in the latent infection sample. We had not previously detected this miRNA in latently infected material using 454 sequencing or stem-loop reverse transcription-PCR (RT-PCR). However, the former method generates fewer sequence reads than the Illumina sequencing technology used here, and the RT-PCR assay for this miRNA has a high background (64; M. F. Kramer, I. Jurak, and D. M. Coen, unpublished data). Thus, it is possible that HSV-1 miR-H1 is an authentic latent transcript, expressed consistently in latently infected ganglia. It is also possible that this miRNA is expressed less consistently at low levels, like a number of other productive cycle transcripts (10, 11, 16, 32, 33, 48, 56). Resolving this question will require further investigation. Interestingly, we did not detect a positional homolog of HSV-1 miR-H1 in either the HSV-2 productive or latent infection sample (but see below).

The second most abundant sequence reads in the HSV-1 productive infection sample were for HSV-1 miR-H6-3p. HSV-1 miR-H6-5p had substantially fewer reads. In contrast, the most abundant sequence reads in the HSV-2 productive infection sample were for HSV-2 miR-H6-5p, with very few reads for HSV-2 miR-H6-3p. Consistent with the abundant sequence reads in the productive infection samples, we could readily detect the pre-miRNA and mature forms of HSV-1 and HSV-2 miR-H6 in RNA from productively infected Vero and 293 cells using Northern blot hybridization, probing for the strands that were more abundant in the sequence reads (Fig. 4). We also detected HSV-1 miR-H6-3p in the latent infection sample, consistent with its previous detection in latently infected mouse and human trigeminal ganglia (64, 65). We also found one read of HSV-1 miR-H6-5p in the latent sample. For HSV-2, we detected miR-H6-5p in 28 reads from the latent sample and one read of miR-H6-3p.

It is interesting to compare the sequences of the different miRNAs from the miR-H1/miR-H6 locus (Fig. 3). HSV-1 miR-H6-5p and HSV-2 miR-H6-5p are 74% identical, with 6/7 seed identity and 7/7 seed consistency (Table 4), suggesting functional conservation. In contrast, although HSV-1 miR-H6-3p and HSV-2 miR-H6-3p share considerable sequence identity, most of that is in their 3′ ends, with very little seed identity (Table 4). This raises questions regarding whether miR-H6-3p exhibits functional conservation, especially as HSV-1 miR-H6-3p exhibits seed complementarity with HSV-1 ICP4 mRNA and can translationally repress HSV-1 ICP4 expression (64). However, despite the lack of seed identity between the miR-H6-3p orthologs, HSV-2 miR-H6-3p exhibits considerable complementarity, including complete seed complementarity with sequences in the 3′ UTR and coding region of HSV-2 ICP4 mRNA (Fig. 5). Moreover, the two miR-H6-3p orthologs exhibit high seed consistency (7/7; Table 4), consistent with the possibility of conserved function.

Perhaps most intriguingly, two miRNAs that are highly abundant during productive infection, HSV-1 miR-H1-5p and HSV-2 miR-H6-5p, exhibit complete seed identity, despite <60% overall sequence identity (Fig. 3D). This suggests that these miRNAs function to downregulate similar targets during infection. It will be interesting to determine the targets of the various miRNAs expressed from this locus, especially as these miRNAs are both highly abundant in productive infection and can be detected in latently infected ganglia.

Unexpectedly, we detected abundant sequence reads for ∼22-nt RNAs that arise from the same hairpin that is predicted for miR-H6. These miRNAs are encoded immediately adjacent to the two strands of miR-H6, distal to the loop of the hairpin (Fig. 3B). However, despite showing similar abundance in the sequence reads to miR-H6, and despite our being readily able to detect miR-H6 by Northern blot hybridization, we were unable to detect these RNAs by the same method (Fig. 4). Similarly, we observed small RNAs with the same properties arising from the HSV-2 miR-H4 (miR-II) locus (Fig. 3C), but not the HSV-1 miR-H4 locus. These RNAs did not fulfill our hairpin-loop proximal criterion for miRNAs; thus, we are reluctant to designate them as miRNAs. Rather, we designate these RNAs as miRNA-offset RNAs (moRs), a term that has been applied to small RNAs detected in the simple chordate Ciona intestinalis. There have recently been reported moRs in Drosophila melanogaster, mouse embryonic stem cells, and humans, albeit at very low expression levels (2, 36, 52, 54). Moreover, Umbach et al. detected moRs derived from pri-miRNAs encoded by Kaposi's sarcoma-associated herpesvirus (KSHV) in a latently infected cell line (62). The biogenesis of this new class of small RNA molecules is poorly understood; however, it has been proposed that moRs might arise from Drosha processing of pri-miRNAs rather than successive Dicer processing of one long pre-miRNA (54). Our Northern blot analysis of moR-H6 is consistent with the Drosha biogenesis model, as we could not detect pre-miRNA using the moR antisense probe (Fig. 4). Further investigation will be required to understand the expression and actual abundance of these RNAs.

As described above, both HSV-1 and HSV-2 express miRNAs that are encoded within origins of replication: miR-H11, encoded within oriL, and miR-H12 and miR-H13, encoded within oriS. The sequences of these miRNAs and their locations relative to origin sequences are shown in Fig. 6. Notably, the HSV-1 oriL hairpin, being a 144-bp perfect palindrome, does not have any internal mismatch bulges that are typically found in pri-miRNAs or pre-miRNAs (1), and the HSV-2 hairpin has only 1 single nucleotide mismatch (Tables 1 and 2). MiR-H11 lies entirely within oriL (Fig. 6). Thus, there are two copies of miR-H11 coding sequences; whether both are expressed is not clear. In contrast, miR-H12 and miR-H13 are encoded from opposite strands and extend outside the oriS palindromes, so that the two miRNAs have 5′ ends in common but divergent 3′ ends. Adding further complexity, HSV-2 contains 2 copies of oriS in each short repeat (R_S) (Fig. 6A). These two oriS palindromes are surrounded by highly similar, but nevertheless distinctive sequences, and since miR-H13 extends outside of oriS, it is possible to distinguish, not just the orientation, but also the exact oriS copy from which this miRNA arises (Fig. 6A). Notably, miR-H11 to -H13 share considerable sequence identity throughout, both within each virus species and between species (Fig. 6B). In particular, the seed regions show 7/7 consistency and are identical for three of the six miRNAs (both miR-H11 and HSV-2 miR-H13), differing by only one base for two (HSV-2 miR-H12 and HSV-1 miR-H13) and only two bases for the remaining one (HSV-1 miR-H12) (Table 4 and Fig. 6B). Thus, these miRNAs constitute a family and may have common targets.

In HSV-1, the DNA sequences encoding HSV-1 miR-H11 include sites for binding by the origin-binding protein, UL9, and also glucocorticoid-response element half-sites (25, 26). In mice following corneal inoculation, a double-point mutant that would affect both coding sequences for miR-H11 exhibited modestly decreased acute replication and reactivation from latency accompanied by reduced pathology and mortality (4). Although there is considerable reason to think that these phenotypes in mice are due to effects of the mutations on DNA replication (3, 4, 26), we raise the possibility that one or more of the phenotypes may be due to loss of the function of miR-H11.

As described previously (64), HSV-1 encodes two miRNAs encoded on complementary strands: miR-H1 and miR-H6. Our analysis has identified several other such pairs. These are HSV-1 miR-H2 and miR-H14, HSV-2 miR-H4 and miR-H19, and HSV-2 miR-H7 and miR-H20 (Fig. 1). All complementary pairs are transcribed from palindrome-like sequences, and each miRNA pair shares substantial sequence identity, which suggests that they might target conserved mRNAs, as has been recently proposed for some D. melanogaster antisense miRNA pairs (61). An additional intriguing possibility is that each miRNA in the pair might have a role in repression of the transcripts encoding the complementary miRNA or might even block the functions of the mature complementary miRNA.

The miRNAs that we have detected are encoded disproportionately within the repeat sequences of the HSV genome (Fig. 1). We do not think it is an accident that this is also the region that is transcriptionally most active during latency. Indeed, only a handful of HSV-1 miRNAs were not detected in the latent infection sample; these were the two unique long region (U_L)-derived miRNAs, the two oriS-derived miRNAs miR-H14 (which is encoded within ICP0), and miR-H15 (which is encoded on the same strand as and downstream of miR-H6). Perhaps more telling, all of the HSV-1 miRNAs that had more abundant reads in the latent infection sample than in the productive infection sample and all but one of the HSV-2 miRNAs that were detected in the latent infection sample are encoded on the same strand as LATs and are downstream of the LAT promoter. It will be particularly interesting to identify promoters that drive expression of miRNAs such as miR-H6 that are encoded from the opposite strand from LATs.

In this study, we have confirmed and extended previous identifications of HSV miRNAs (15, 57, 58, 64, 65), so that 16 HSV-1 miRNAs and 17 HSV-2 miRNAs are now known. Do we now know the entire complement of HSV miRNAs? There are several grounds to suggest that we do not. First, only one HSV-1 and one HSV-2 strain are completely sequenced and available. Small sequence differences between the strains with which we worked and these sequenced strains might have excluded certain hairpins and thus certain loci from further analysis. In fact, we have identified numerous loci with a high number of sequence reads that did not pass the hairpin criterion. Second, different strains might express strain-specific miRNAs, which might account for differences in pathogenicity. Thus, it would be particularly interesting to compare clinical isolates with laboratory strains. Third, in our analysis we did not detect a single sequence read for HSV-2 miR-H10 (UL15) detected by Umbach et al. (66) from latently infected human ganglia. We cannot exclude the possibility that this miRNAs is strain specific or that it is only expressed in the human host. Along these lines, our productive infection samples were all harvested at a rather late time point. We cannot exclude the possibility of miRNAs that are more abundant earlier in infection. In other words, experimental settings might account for important differences in analyses. Finally, the development of new technologies and bioinformatic tools might increase the sensitivity of detection, leading to the discovery of additional HSV miRNAs.

A key question is what are the roles of the HSV miRNAs that have been identified? For some previously identified miRNAs (i.e., miR-H2 to -H4), identification of viral targets has been facilitated by the fact that these miRNAs are transcribed from the opposite strand of known viral mRNAs, which makes them obvious targets (57, 58, 64, 65). However, for most miRNAs there is no obvious target. The challenge will be to identify them. Moreover, the functions of these miRNAs might not be limited just to RISC-dependent posttranscriptional regulation, but might also include transcriptional regulation (31) and direct effects on the stability of mRNAs that contain hairpins from which miRNAs, in principle, could be derived (24, 60). Regardless, the ability to compare miRNAs from HSV-1 and HSV-2 should greatly aid efforts to identify HSV miRNA functions.