High-resolution profiling and analysis of viral and host small RNAs during human cytomegalovirus infection

Abstract

Human cytomegalovirus (HCMV) contributes its own set of microRNAs (miRNAs) during lytic infection of cells, likely fine-tuning conditions important for viral replication. To enhance our understanding of this component of the HCMV-host transcriptome, we have conducted deep-sequencing analysis of small RNAs (smRNA-seq) from infected human fibroblast cells. We found that HCMV-encoded miRNAs accumulate to ∼20% of the total smRNA population at late stages of infection, and our analysis led to improvements in viral miRNA annotations and identification of two novel HCMV miRNAs, miR-US22 and miR-US33as. Both of these miRNAs were capable of functionally repressing synthetic targets in transient transfection experiments. Additionally, through cross-linking and immunoprecipitation (CLIP) of Argonaute (Ago)-bound RNAs from infected cells, followed by high-throughput sequencing, we have obtained direct evidence for incorporation of all HCMV miRNAs into the endogenous host silencing machinery. Surprisingly, three HCMV miRNA precursors exhibited differential incorporation of their mature miRNA arms between Ago2 and Ago1 complexes. Host miRNA abundances were also affected by HCMV infection, with significant upregulation observed for an miRNA cluster containing miR-96, miR-182, and miR-183. In addition to miRNAs, we also identified novel forms of virus-derived smRNAs, revealing greater complexity within the smRNA population during HCMV infection.

Fig 1

Genomic analysis of smRNAs during HCMV infection of HFFs. (A) Distribution of smRNA-seq reads across categories of annotated smRNAs (pre-miRNAs, snoRNAs, tRNAs, and rRNAs) in HCMV-infected (Towne strain) or mock-treated HFFs at 24 and 72 hpi. Reads that did not match current annotations were classified as “other.” For human miRNA fractions, values listed in parentheses indicate the total number of unique pre-miRNAs detected. (B) Expression levels of HCMV mature miRNAs during infection, measured by smRNA-seq. Total read counts for each viral miRNA, including reads that exactly matched the mature miRNA sequence and those that were uniquely mapped to its genomic coordinates, were normalized by the total number of miRNA reads (human and viral) in the data set.

Fig 2

Identification of novel HCMV miRNA precursors miR-US22 and miR-US33as. (A and B) Uniquely mapped smRNA-seq reads align to predicted RNA stem-loop structures characteristic of a pre-miR hairpin. RNA stem-loop structures of pre-miR-US22 (A) and pre-miR-US33as (B) were generated using mFold software (http://mfold.rna.albany.edu). Highlighted arms indicate where the majority of smRNA-seq reads aligned at 72 hpi (percentages are listed in Table 3). Genomic locations of the pre-miRs are depicted below the stem-loop structures (direction of transcription is indicated by arrowheads). Only nonredundant sequences represented by at least five smRNA-seq reads from the 72-hpi data set are depicted. The predicted RNA structures of the precursors are illustrated by parentheses indicating bases that are paired and dots that correspond to unpaired bases. (C and D) Northern blot analysis of miR-US22-5p (C) and miR-US33as-5p (D) in infected HFFs. Mock-treated cells were harvested at 72 h for this analysis. The U6 snRNA serves as a loading control. (E) Processing of HCMV miR-US22 in 293T cells and repression of a synthetic target. An miR-US22 expression construct or empty vector was transfected into 293T cells, and Northern blot analysis was performed to compare miRNA levels in these samples to those of HFFs at 72 hpi. A luciferase assay was used to demonstrate repression of a perfectly complementary target sequence by miR-US22-5p (P value of <0.01). 293T cells were transfected with miR-US22 expression vector and a construct containing a luciferase reporter with either no targeting site (control), a target consisting of the perfectly complementary target of miR-US22-5p (full target), or a canonical seed match target corresponding to only positions 2 to 7 of miR-US22-5p (seed target). Luciferase measurements shown are averages of three independent experiments. Within each experiment two independent reporter plasmid preparations were used to assess each type of target site. Error bars represent standard deviation, and significance values were calculated using an unpaired, two-tailed Student's t test. (F) Analysis of miR-US33as-5p in transfected 293T cells, performed as described above for panel E (P value of <0.01 for repressed full and seed versions of synthetic targets).

Fig 3

Functional association of HCMV miRNAs with endogenous Argonaute proteins. (A) Experimental strategy for generation of Ago CLIP-seq libraries. Mock-infected and infected cells were UV cross-linked at 24 and 72 hpi. Immunoprecipitated (IP) RNA associated with Ago complexes was further processed and subjected to Illumina sequencing. (B) Differential incorporation of particular HCMV miRNAs between Ago2 and Ago1 complexes, as measured by the ratio of 5p over 3p mature miRNA arm incorporation between Ago2 and Ago1 complexes. Statistical significance was assessed by chi-square analysis (P value threshold of 6 × 10⁻⁵). Read count heights at each position represent the total number of overlapping CLIP-seq reads, normalized by the total number of miRNA reads in the data set. (C) Nondifferential efficiencies of miRNA association with Ago2 and Ago1 demonstrated by other HCMV miRNAs. Three representative examples are shown. (D) Human miRNAs that exhibit differential 5p-3p miRNA arm incorporation into Ago2 and Ago1 complexes. CLIP-seq read counts were used from infected data sets (24 and 72 hpi) for this assessment. Similar differential incorporation levels of these miRNAs could be observed in Ago2-Ago1 comparisons for the mock data sets.

Fig 4

Effects of HCMV infection on human miRNA levels in HFFs. (A) Scatter plot comparisons of infected and mock-infected human miRNA expression values at 24 and 72 hpi. Highlighted in red and green are significantly upregulated and downregulated miRNAs during infection, respectively (P value of < 0.05). Listed above the scatter plots are the names of the significantly altered miRNAs (any that changed between the two mock infection time points are not listed) (B) Expression of miR-182/-96/-183 during infection. The region of chromosome 7 encoding these miRNAs is depicted below the distribution of smRNA-seq reads that aligned to the 182, 96, and 183 miRNA precursors (transcription is from right to left). Read count heights at each position represent log₂-transformed values of the total number of overlapping smRNA-seq reads, normalized by the total number of miRNA reads in the data set. (C) Northern blot analysis of mature miR-182 in both HCMV Towne- and AD169-infected cells. HFFs were either mock-treated or infected with the Towne strain at an MOI of 3 or AD169 at an MOI of 5. The U6 snRNA serves as a loading control.

Fig 5

Analysis of regions of the HCMV genome enriched for production of non-miRNA smRNAs. (A and B) Distribution of uniquely aligned non-miRNA smRNA-seq reads at 24 and 72 hpi, with peaks representing stacked reads that aligned to the Watson (above) and Crick (below) strands of the HCMV genome. Reads overlapping annotated and predicted miRNA precursor regions were excluded, and the viral genome was interrogated in 1-kb segments for smRNA-seq alignments. (C) smRNAs that are generated in sense orientation to the area encoding lncRNA2.7. (D) Representation of smRNA-seq alignments across a transcriptionally active region that falls within the UL61-UL68 region. Expression data from RNA-seq experiments is shown to illustrate transcription of the regions. Coverage by cDNA clones from Zhang et al. (42) is shown, providing independent support of RNA-seq-detected expression of the RNAs.