Dissecting HSV-1-induced host shut-off at RNA level

Herpes simplex virus 1 (HSV-1) installs a profound host shut-off during lytic infection. The virion host shut-off (vhs) protein plays a key role in this process by efficiently cleaving both host and viral mRNAs in a translation-initiation-dependent manner. Furthermore, the onset of viral DNA replication is accompanied by a rapid decline in transcriptional activity of the host genome. Both mechanisms have tremendous impact on the RNA expression profile of the infected cells. To dissect their relative contributions and elucidate gene-specific host transcriptional responses throughout the first 8h of lytic HSV-1 infection, we here employed RNA-seq of total, newly transcribed (4sU-labelled) and chromatin-associated RNA in wild-type (WT) and Δvhs infection of primary human fibroblasts. Following virus entry, vhs activity rapidly plateaued at an elimination rate of around 30% of cellular mRNAs per hour until 8h p.i. In parallel, host transcriptional activity dropped down to 10-20%. While the combined effects of both phenomena dominated infection-induced changes in total RNA, extensive gene-specific transcriptional regulation was observable in chromatin-associated RNA. This was surprisingly concordant between WT HSV-1 and its Δvhs mutant and at least in parts mediated by the embryonic transcription factor DUX4. Furthermore, both WT and Δvhs infection induced strong transcriptional up-regulation of a small subset of genes. Most of these were either poorly or not at all expressed prior to infection but already primed by H3K4me3 histone marks at their promoters. Most interestingly, analysis of chromatin-associated RNA revealed vhs-nuclease-activity-dependent transcriptional down-regulation of at least 150 cellular genes, in particular of many genes encoding integrin adhesome and extracellular matrix components. This was accompanied by a vhs-dependent reduction in protein levels by 8h p.i. for many of these genes. In summary, our study provides a comprehensive picture of the molecular mechanisms that govern cellular RNA metabolism during the first 8h of lytic HSV-1 infection. Author Summary The HSV-1 virion host shut-off (vhs) protein efficiently cleaves both host and viral mRNAs in a translation-dependent manner. In this study, we model and quantify changes in vhs activity as well as the virus-induced global loss of host transcriptional activity during productive HSV-1 infection. In general, HSV-1-induced alterations in total RNA levels were found to be predominantly shaped by these two global processes rather than gene-specific regulation. In contrast, chromatin-associated RNA depicted gene-specific transcriptional changes. This revealed highly concordant transcriptional changes in WT and Δvhs infection, confirmed DUX4 as a key transcriptional regulator in HSV-1 infection and depicted vhs-dependent, transcriptional down-regulation of the integrin adhesome and extracellular matrix. The latter explained some of the gene-specific effects previously attributed to vhs-mediated mRNA degradation and resulted in a concordant loss in protein levels by 8h p.i. for many of the respective genes.


Introduction
to additional, vhs-cleavage-activity-dependent effects on their transcription. Strikingly, 126 vhs-dependent down-regulation of transcriptional activity resulted in reduced protein 127 levels of many of the respective genes already at 8h p.i. in WT but not in vhs infection 128 as confirmed by quantitative whole-proteome mass spectrometry.  [19]. We infected HFF with vhs at a high MOI of 10 and performed 4sU-seq in hourly 136 intervals and total RNA-seq every two hours during the first 8h of infection (2 biological 137 replicates; Fig 1A). It is important to note that although the WT and vhs time-course 138 experiments were performed independently, we carefully standardized the 139 experimental conditions, e.g. by infecting the same batch of cells following the same 140 number of splits after thawing as well as using the same batch of fetal bovine serum 141 (FBS), to achieve a maximum level of reproducibility. Consistent with our previous 142 8 findings [19,21] and with the modest attenuation of vhs in HFF, HSV-1-induced DoTT 143 and subsequent poly(A) read-through transcription in vhs infection was similar but 144 slightly reduced compared to WT infection (Fig A,B in S2 File,S3 Dataset). Since read-145 in transcription into downstream genes due to HSV-1-induced DoTT from upstream 146 genes can be mistaken for "induction" of these downstream genes [19], we excluded 147 genes with read-in transcription from all following analyses (see methods for details).

157
The only exceptions were the first two 4sU-seq time points (0-1 and 1-2h p.i.), when 158 essentially no (n≤2) cellular genes were differentially expressed in both WT and Δvhs 159 infection (multiple testing adjusted p≤0.001, |log2 fold-change| ≥1). This was expected 160 as fold-changes were only very small (median |log2 fold-change| ≤0.1) and dominated 161 by experimental noise. The highest correlations between 4sU-seq fold-changes in WT 162 and Δvhs infection compared to mock were observed at 4-5h and 5-6h p.i. (Spearman 163 rank correlation ≈ 0.8, Fig 1C, Fig C in S2 File). Correlations decreased towards the 164 end of the time-course in particular for genes down-regulated in WT (Fig 1D), 165 consistent with the well described effects of vhs on cellular RNA levels late in infection 166 [23]. Notably, the later stages of Δvhs infection (from 6-7h p.i.) were better correlated 167 9 to slightly earlier stages (4-5h, 5-6h p.i.) of WT infection (Fig C in S2 File), indicating 168 slightly slower progression of Δvhs infection. 169 In contrast to 4sU-RNA, fold-changes in total RNA obtained from WT and Δvhs 170 infection were only poorly correlated ( ≤ 0.11,Fig 1E,F,Fig D in S2 File). Consistent 171 with the cleavage activity of vhs, this was particularly prominent for genes down-172 regulated in WT infection. As 4sU-RNA was purified from total RNA, the poor 173 correlation for total RNA fold-changes cannot be explained by poor reproducibility 174 between the two independent experiments. We conclude that this instead reflects the 175 expected strong impact of vhs cleavage activity on the cellular mRNAs. In principle, 176 vhs cleavage activity should more strongly affect total mRNA levels of long-lived 177 mRNAs than of short-lived mRNAs, as the former have much weaker de novo 178 transcription relative to total RNA levels and are thus much more slowly replaced. On 179 the contrary, HSV-1-induced global loss in transcriptional activity should more strongly 180 affect total RNA levels of unstable, short-lived mRNAs. To test this hypothesis, we 181 correlated the observed changes in total RNA upon WT and vhs infection with the 182 RNA half-life of the respective transcripts. RNA half-lives were obtained based on 183 newly transcribed RNA to total RNA ratios from uninfected HFF as previously 184 described [24]. This revealed the expected striking differences between WT and Δvhs 185 infection. In WT infection, total RNA fold-changes and mRNA half-lives were negatively 186 correlated ( = −0.38 at 8h p.i. , Fig 2A), i.e. total RNA levels of stable cellular mRNAs 187 tended to decrease more strongly than of unstable mRNAs. This was already 188 observable at 2h p.i. ( = −0.31) consistent with mRNA cleavage and degradation by 189 tegument-delivered vhs protein. The negative correlation to RNA half-lives was also 190 confirmed for total RNA fold-changes from the study of Fig E in S2 File), while at 12h p.i., a weaker, but still highly significant, negative 192 correlation was observed ( = -0.15).

193
In Δvhs infection, total RNA fold-changes and RNA half-lives were positively correlated 194 from 4h p.i. onwards ( = 0.55 at 8h p.i., Fig 2B). Thus, total RNA levels of short-lived 195 cellular RNAs were more strongly reduced than of long-lived ones. This effect is 196 consistent with the well described gradual decline in global transcriptional activity 197 starting around 3-4h p.i. [15,19]. Accordingly, total RNA fold-changes in vhs infection  To quantify vhs activity throughout infection, we developed a mathematical model to 212 estimate both the extent of loss in transcriptional activity as well as vhs endonuclease 213 activity during HSV-1 infection (S1 Text) based on our total RNA-seq time-course data 214 for 0, 2, 4, 6 and 8h p.i. in WT and Δvhs infection. Our results indicate that by 8h p.i., 215 transcriptional activity dropped down to 10-20% of the level in uninfected cells during 216 Δvhs infection (Fig 2C). Assuming an at least similar drop in transcriptional activity in 217 WT infection, our model suggests that at the height of vhs activity, ~30% of RNAs are 218 lost per hour due to vhs-mediated degradation ( Fig 2D). This rate reached 26% as 219 early as 2h p.i. and remained fairly constant until 8h p.i. Our data exclude a significant 220 drop in vhs activity before 8h p.i. as the drop in transcriptional activity would otherwise 221 result in positive correlations between total RNA fold-changes and mRNA half-lives in 222 WT infection (S1 Text). Similarly, if the loss of transcriptional activity in WT infection 223 were dramatically higher than in Δvhs infection, vhs-mediated degradation would have 224 to increase even faster and to higher levels to achieve the observed negative 225 correlations.

226
Although statistically significant correlations were also observed between 4sU-RNA 227 fold-changes and RNA half-lives, these were relatively small ( and loss of transcriptional activity, but substantially less strongly than for total RNA. In 231 summary, these results indicate that the poor correlation in total RNA fold-changes 232 between WT and Δvhs infection is a direct consequence of global effects of vhs on 233 RNA stability throughout the first 8h of lytic infection. Accordingly, our model implies 234 that the wide range of total RNA fold-changes observed between genes in HSV-1 235 infection can be largely explained by differences in RNA half-lives between genes and 236 does not require extensive gene-specific differences in vhs-mediated mRNA cleavage. WT-and vhs-infected cells were obtained and sequenced in the same experiment.

245
Only the data from mock and WT-infected cells have previously been published [20].

246
The efficient separation of the cytoplasmic from the nuclear RNA fraction was 247 confirmed by the enrichment of well-described nuclear lincRNAs (MALAT1, NEAT1,    Fig 3A). These are further analyzed 280 below.

281
Notably, 4sU-RNA fold-changes were better correlated to fold-changes in chromatin-282 associated RNA than to nucleoplasmic or cytoplasmic RNA, while total RNA fold-    Strikingly, however, we also observed a strong functional enrichment for several GO 307 terms (adj. p≤0.001, S5 Dataset), in particular "extracellular matrix (ECM) organization" 308 (>32-fold enriched, adj. < 10 −25 ). This included fibronectin (FN1), integrin beta 1 309 (ITGB1), a subunit of integrin complexes binding fibronectin, and several genes 310 encoding for collagen alpha chains. Enrichment was also observed for "focal 311 adhesion", i.e. the integrin-containing, multi-protein complexes that anchor the cell to 312 the ECM and connect it to the actin cytoskeleton [27,28].   PCR, they showed that this reduction was vhs-dependent based on two sets of genes 397 that exhibited either high (COL6A, MMP3, MMP1) or low reduction (GAPDH, ACTB, 398 RPLP0) in total RNA levels in WT infection. As Actinomycin D treatment confirmed 399 similar stability of corresponding mRNAs, they concluded that these differences were 400 not due to differences in transcription rates or mRNA stability between these genes but 401 rather due to differences in the susceptibility of the respective transcripts to vhs 402 cleavage activity. We noted that two (COL6A1, MMP1) of the PCR-confirmed highly 403 vhs-sensitive genes belonged to our set of genes transcriptionally downregulated in a 404 vhs-dependent manner. The third gene (MMP3), though originally not included in our 405 analysis due to its proximity to nearby genes, is also involved in ECM organization.  Concordantly, down-regulated proteins were significantly (adj. p≤0.001, S10 Dataset) 530 enriched for a number of GO terms, including "nucleotide-sugar biosynthetic process" 531 (>77-fold enriched), "canonical glycolysis" (>15-fold), "viral budding" (>9-fold) and 532 "activation of MAPK activity" (>4-fold). Interestingly, meta-adhesome (but not HFF   confirmed that the obtained RNA-seq data could indeed be directly compared. to be up-regulated by DUX4 and 39% of down-regulated genes were down-regulated 636 by DUX4, Fig 4E). Although there was some overlap between DUX4 and IFN-induced 637 genes amongst the HSV-1-induced genes, it was not significantly higher than expected 638 at random. Interestingly, the DUX4 up-regulated gene TRIM43 was recently identified 639 as a herpesvirus-specific antiviral factor independent of the type I interferon response

640
[48], suggesting that DUX4-mediated regulation in HSV-1 infection may represent an 641 alternative pathway which augments the host intrinsic immune response.

642
A key finding of our study is the vhs-dependent, transcriptional down-regulation of  The full description of H3K4me3 ChIPmentation is included in S11 Text.

754
Preparation of samples for proteomic analysis 755 HFF were infected with WT HSV-1 or its vhs-inactivated mutant for 8h at an MOI of 10.

756
Infections were conducted in triplicate, with 4 uninfected controls (10 samples in total).  The mathematical model of WT and Δvhs infection is described in S1 Text.