RNA-interactome capture identifies SRSF3 as a key protein for herpesviral gene expression

Abstract

The primary mRNA sequence determines its secondary structure and the repertoire of interacting RNA-binding proteins (RBPs). The resulting mRNA ribonucleoprotein complex (mRNP) then influences all stages of the life of an mRNA. Here, we determined the mRNP composition of individual Kaposi sarcoma herpesviral (KSHV) mRNAs. Like all herpesviruses, KSHV switches between a latent and lytic stage of the viral life cycle. During reactivation from latency, the viral RNA regulator ORF57 ensures the translation of viral mRNAs by increasing their mRNA stability and nuclear export. We optimized an LNA/DNA mixmer RNA capture protocol for both transfection and viral infection settings. In combination with eCLIP, we confirmed that ORF57 directly binds to an AU-rich RNA motif, which may enable ORF57 to discriminate viral from cellular RNAs based on the nucleotide bias of KSHV lytic RNAs. In addition, we captured the RBPome of two ORF57-dependent viral transcripts and identified the host RNA processing factor SRSF3 as a key regulator of viral replication.

Significance Statement.

Each mRNA assembles a unique ribonucleoprotein complex (mRNP). The interacting RNA-binding proteins (RBPs) determine the fate of mRNAs. The mRNAs of complex DNA viruses picture a mixture of cellular and viral RBPs. To understand how cellular constraints of gene expression and viral replication compete with each other, information on the individual viral mRNP composition is needed. Here, we optimized an RNA capture method to determine the RBPome of herpesviral transcripts. Our data along with eCLIP demonstrated a direct RNA interaction of the main viral RNA regulator. In addition, the splicing factor SRSF3 serves as a key factor for viral gene expression and infectivity. The method presented here may uncover the RBPome of other individual viral and cellular genes.

Introduction

Viruses have evolved different strategies to ensure efficient viral gene expression. For example, many nuclear viruses use alternative splicing to expand their coding capacity (1, 2). The process of splicing merges exons into a continuous reading frame while also recruiting nuclear export factors to the nascent transcripts (3, 4). However, the majority of genes in herpesviruses are intronless (5). Intronless transcripts present herpesviruses with the challenge of how to funnel them efficiently into RNA export pathways (3, 4). Thus, herpesviruses have evolved a viral RNA regulator that ensures the efficient expression of their intronless genes, with ICP27 from herpes simplex virus-1 being the paradigmatic example (6, 7). The Kaposi sarcoma herpesviral (KSHV) homolog is ORF57, a multifunctional protein that recruits the cellular export machinery, namely Aly/REF or UAP56-interacting factor. Upon recruitment, ORF57 stabilizes the viral transcripts in the nucleus and/or promotes their export to the cytosol, thereby ultimately enabling the translation of a large set of viral mRNAs (8–11).

In general, herpesviruses oscillate between a latent state (no replication) and the lytic cycle (genome replication and viral progeny formation). KSHV reactivates rarely, and its default state is latency (12). Reactivation leads to a burst in lytic gene expression, viral particle formation, and cytotoxicity. To switch between these states, the expression of the transcriptional regulator RTA and ORF57 is required (13–15). In fact, many lytic genes are dependent on ORF57 and characterized by a high adenosine/thymidine (AT) content compared with cellular genes (8, 16). AT richness of mRNAs is considered to be unfavorable for gene expression, with HIV as a paradigmatic example (17). Computational analysis by the algorithm RBP map (rbpmap.technion.ac.il) predicted that the AT-rich nucleotide composition of KSHV creates motifs for the hnRNP family of proteins. In sum, the uncommon nucleotide composition and the bound protein repertoire may prevent these transcripts from being efficiently expressed.

RNA-binding proteins (RBPs) play an important role in determining the fate of transcripts. Together with the mRNAs, they form large ribonucleoprotein (mRNP) complexes influencing RNA expression on many levels, such as stability, localization, modification, and translation. Novel RNA-interactome studies and subsequent proteomic analyses resulted in a more complete picture of mRNPs, including RBPs that lack a canonical RNA-binding domain (18).

Here, we set out to identify RBPs that bind to specific ORF57-dependent viral transcripts of KSHV and to determine how they influence the fate of their bound transcripts. By adapting a powerful and stringent RNA capture method (19, 20), we identified the RBPome of the viral ORF47 and ORF6 transcripts in transfections and reactivated virus-infected cells. We identified the cellular serine/arginine-rich (SR) protein SRSF3 as a key RBP that regulates ORF57-dependent viral gene expression and influences viral infectivity. Overall, this work will facilitate RNA-interactome studies of other specific viral mRNAs, which will shed light on how the mRNP composition orchestrates gene expression.

Results

LNA mixmer capture enables the analysis of the RBPome of transfected lytic KHSV transcripts

To identify RBPs that are responsible for the instability and ORF57 dependency of viral transcripts, we applied an LNA/DNA mixmer capture in combination with ultraviolet (UV) crosslinking, as established by the Hentze laboratory (20) (Fig. S1A). This method uses antisense oligonucleotides containing locked nucleic acids. These high-affinity RNA analogs allow more stringent washing steps. Thus, only direct RNA binders are retained, and indirect protein–protein interactions are removed.

Prior to our study, this method had only been used for highly abundant transcripts such as ribosomal RNAs. Here, we adapted it to capture the protein binding repertoire of a single, specific viral transcript. We sought to enrich the viral ORF47 transcript from transiently transfected HEK 293T cells in the presence of ORF57. ORF47 RNA is a strictly lytic, ORF57-dependent transcript of short length (504 nucleotides [nts]) and a high AT content of 49.6%. In the absence of ORF57, the expression of ORF47 is very low (8). This is true for all ORF57-dependent transcripts (8). It is important to note that UV crosslinking, in general, is very inefficient, as <5% of a given RBP can be crosslinked to RNA (21), and thus large quantities of the transcript of interest are required. For these reasons, an ORF57 minus control could not be included in the capture experiments.

We validated the enrichment of our mRNA of interest over random transcripts (GAPDH RNA) and compared it with samples from a control capture with a scrambled LNA/DNA mixmer by RT-qPCR. The no-crosslink samples represent the most reliable estimate for the capture efficiency, as in the crosslinked samples, covalently bound peptides cannot be removed by proteinase K and subsequently interfere with the reverse transcription reaction (Fig. 1E). ORF47 transcripts are enriched over GAPDH RNA when compared with input RNA (Fig. 1E), while contamination with genomic DNA was negligible (Table S1). However, we were unable to remove all plasmid contamination due to the high DNA amount used in transient transfections in 293T cells (Table S1). We also sequenced the captured transcripts by Illumina RNA sequencing to confirm a broad distribution of reads across the ORF47 mRNA (Fig. S1B). The RNA-seq data show full coverage across the ORF47 mRNA. Given the short length of the ORF47 transcript (504 nts) and the read length, we assume capture of the intact transcripts and not simply RNA fragment recovery. Other obtained reads were non-specific or mapped to a mitochondrial rRNA cluster.

Fig. 1.

A) Volcano plot comparing the fold change difference and P-value of the proteins identified in the crosslinked (crlk) and not crosslinked (no crlk) samples for the ORF47 capture experiments (cotransfected with ORF57) (n = 3). Two follow-up hits are indicated in red, the three most upregulated hits are marked in blue. The dotted lines represent the border for significance. B) Gene ontology analysis using GOrilla shows the percentage of the significantly enriched captured proteins, from the ORF47 wild type capture, involved in certain biological processes. Note that each protein can be mapped to multiple processes. C) Volcano plot comparing the fold change difference and P-value of the proteins identified in the crosslinked (crlk) and not crosslinked (no crlk) samples for the synthetic ORF47 transcript (without ORF57 cotransfection) (n = 3). D) Venn diagram showing the overlap between significantly enriched proteins identified in the ORF47 wild type and synthetic capture experiments. E) RNA quality controls include RT-qPCR to control the capture efficiency of ORF47 and synthetic ORF47 transcripts compared with GAPDH RNA and ORF47 RNA in control samples with scrambled LNA/DNA mixmers displayed relative to the corresponding input RNA levels.

The enrichment of full-length ORF47 RNA prompted us to decipher the RBPome by mass spectrometry. The volcano plot in Fig. 1A shows several highly enriched proteins binding to the ORF47 mRNA compared with the “no crosslink” sample. Proteins were considered to be significantly enriched when the Student's t test difference compared with the “no crosslink” control was above ≥2 and the threshold P-value ≤0.05 (corresponding to a −log(P-value) above 1.3). We performed gene set enrichment analysis on these proteins and recovered mainly mRNA processes (Fig. 1B). Among the candidates we identified a large number of nuclear RBPs, many of them with a preference for AU-rich sequences (Table 1), which is in agreement with the sequence bias of ORF57-dependent viral RNAs such as ORF47 (8). Additionally, we used a sequence-optimized, ORF57-independent, ORF47 version (synORF47) as a comparator to test whether a higher GC content may change the interacting RBPs and thus result in a distinct mRNP composition (Fig. 1C) (8). In fact, optimization results in an 11.5% increase in GC content for the synORF47. We compared the significantly enriched proteins over the no crosslink control from the ORF47 wild type (Fig. 1A) and the ORF47 synthetic capture (Fig. 1C). The sequence optimization yields overlapping as well as unique RBPs (Fig. 1D; Fig. S1E lists the proteins from the Venn diagram). We sought to relate this finding to the ORF57 independence of the synthetic ORF47 RNA (8) and recognized a decrease in the fraction of bound hnRNP proteins and a relatively constant number of SR proteins, as we previously showed by in silico RBP motif prediction (Fig. S1D (8)). A protein–protein association network (STRING) additionally revealed striking connectivity of the ORF47-enriched proteins, indicating involvement in similar or closely related processes (Fig. S1C). In sum, these findings demonstrate that our optimized/adapted LNA mixture capture protocol, in combination with high-resolution LC-MS/MS, is suitable for identifying specific RBPs that bind to the viral ORF47 transcript.

Table 1.

Selection of captured RBPs with known AU-binding preference and described function in RNA stability or decay.

Candidate	AU-rich binding	Evidence RNA for degradation/stability	References
RBM15		Nuclear accumulation	(22)
TIA1	3′-UTR AU-rich region binding, Poly (A)	RNA decay/stability	(23, 24)
TRA2A	Preferentially GAA	Nuclear sequestration of HIV-1 Gag RNA, RNA stability	(25, 26)
PUF60	Poly (U)	HBV RNA degradation	(27)
KHDRBS1/Sam68	Poly (U), Poly (A)	HIV Rev RNA export	(28, 29)
hnRNPD/AUF1	(3′-UTR) AU-rich region binding	RNA decay	(30)
SRSF3		RNA decay, export	(31, 32)
ILF2	AU-rich	RNA decay, export	(33–35)

ORF57 directly binds to a preferentially AU-rich sequence motif in the viral transcripts

The ORF47 capture experiment (Figs. 1A and 3A, left panel) confirms the direct binding of ORF57 to viral RNAs without additional co- or auxiliary factors. To verify this finding and to identify the specific binding sites/motifs of ORF57 on the viral transcripts, we employed eCLIP (36) with several modifications as described in Dvir et al. (37). For these experiments, we used reactivated iSLK rKSHV.219 (38) cells, which were crosslinked and harvested 48 h post induction. iSLK rKSHV.219 cells harbor a recombinant, stable KSHV genome that can be an additional, Dox-inducible copy of the master transcriptional regulator RTA (ORF50; compare with Fig. 6A). For the two biological replicates of ORF57 eCLIP (CLIP1 and 2), between 16 and 21% of processed reads align with the KSHV genome. This contrasts with the 6–10% of reads aligned to the KSHV genome from the size-matched input (SMI) controls. Not surprisingly, most KSHV reads (∼70–78%), for both ORF57 CLIP and SMI samples, align with the viral PAN RNA, which is the most abundant transcript during lytic replication (39) (Fig. 2A and D). As expected, the number of detected binding sites for ORF57, in both the KSHV and the human genome, is much higher in the CLIP samples compared with the SMI controls (Fig. 2B). From the ORF57 peaks, we could recover an AU-rich sequence motif that is present in 46% of the target reads aligning to the viral genome. A highly similar sequence motif, present in 19% of the target reads, could also be identified in the human transcriptome (Fig. 2C) but not in any of the SMI samples. Overall, the alignment of the eCLIP data to the KSHV genome confirms binding sites for ORF57 in many viral genes (Fig. 2D) with the most prominent peak in the PAN RNA, while also of note is the high number of binding sites in the ORF57-dependent K15 and ORF59 RNA (far right, reverse strand).

Fig. 3.

A) Dot plots showing the log₂ value of intensity for ORF57, RBM15, and SRFS3 in the individual capture experiments obtained by the mass spectrometry analysis. Each colored dot represents an individual experiment. The crosslinked (crlk) samples were compared with the no crosslink (no crlk) samples and, for RBM15 and SRSF3, also with the results from the capture experiments with the synthetic/sequence-optimized ORF47 transcript (syn crlk and syn no crlk). B) Immunoblot using lysates from transiently transfected cells as indicated. Cells were treated with siRNAs targeting SRSF3 or control siRNAs (si neg ctrl). An anti-HA antibody was utilized for the detection. C) Immunoblot as in (B). Knockdown of SRSF3 was performed using lentivirally encoded shRNAs or scrambled shRNA (sh scr). ORF47 expression was assayed in the presence or absence of ORF57. D) Immunoblot as in (B) performed with expression vectors for ORF6 and K15 expression in the presence or absence of ORF57. Here, a second SRSF3 siRNA (II) was used to exclude unspecific effects. In all cases, transfected HEK293T cells were used.

Fig. 6.

A) Depiction of the KSHV infectivity assay after SRSF3 shRNA knockdown (created with BioRender.com). The recombinant rKSHV genome expresses GFP in the latent state and RFP upon reactivation. Induction was performed using dox and sodium butyrate (see Materials and methods). B) Immunoblot from iSLK rKSHV cell lysates, collected after harvesting the viral supernatants (72 h post induction), confirming the shRNA-mediated knockdown and overexpression of SRSF3. C) Flow cytometry results of infectivity assays. The data represent three biological replicates. Either showing the GFP-positive cells (left) in percent for different dilutions of the KSHV supernatants, with the samples of the scrambled (scr) shRNA set to 100%, or showing the infectious units per milliliter (right) for the different dilutions. Significant changes as determined by paired t tests are indicated. The MFI and IFU in percent can be found in Fig. S3B. D) qPCR data to analyze the viral genome copy numbers in the collected supernatants treated with benzonase to remove all cellular and viral nucleic acids. The viral ORF25 gene was used for the absolute quantification of genome copy numbers. The data represent three biological replicates; significant changes as determined by standard t test are indicated. E) qPCR data to compare the viral DNA replication using whole cell lysates from reactivated cells (n = 3).

Fig. 2.

A) Bar graphs showing the percentage of total processed eCLIP reads that align to the KSHV genome (left) and the percentage of KSHV reads that align to PAN RNA (right). The two biological replicates of ORF57 CLIP (CLIP1 and 2, yellow) are shown in comparison to the SMI samples (SMI1 and 2, blue). B) Bar graph showing the number of binding sites mapping to the human genome (light green) and the KSHV genome (purple) for CLIP and SMI samples. C) AU-rich sequence motif discovered in ORF57-binding sites (top, targ = target reads, bg = background) mapping to the KSHV genome (HHV8 GK18 NC_009333.1) and the human genome (HG38). Motif identified for the SMI samples (bottom). D) ORF57 eCLIP data aligned to the KSHV genome showing the ORF57 CLIP peaks (yellow) and peaks for the SMI samples (blue).

SRSF3 is a relevant factor for viral RNA stability

We next examined the identities of other captured RBPs by surveying the literature for involvement in RNA stability or degradation or binding to AU-rich sequences (Table 1) and testing them for ORF47 mRNA and protein expression in the absence of ORF57. For example, we detected RBM15 (Fig. 3A, middle panel), which is involved in ORF57-mediated gene expression (22, 40) and further confirmed the capture results. However, in our follow-up analyses, siRNA knockdown experiments showed that RBM15 is not a key factor for ORF57 dependency (Fig. S2A). Most of the tested candidates had no observable effect on ORF47 expression; only PUF60 knockdown showed a reduction of ORF47 expression in the presence and absence of ORF57 (Fig. S2A–G). These results illustrate the general redundancy of RBPs and the cumulative effect of mRNP complexes. Interestingly, the splicing factor SRSF3 was highly enriched in the capture assays (Figs. 1A and 3A, right panel). SRSF3 binds to both the wild type and the sequence-optimized version of ORF47 (Fig. 3A, right panel, and Fig. S1E). Thus, we tested the effect of SRSF3 knockdown on viral gene expression (Fig. 3B). The transfection of the ORF47-encoding plasmid alone only leads to a very weak expression of the ORF47 protein (Fig. 3B, lane 2). This phenotype is due to RNA instability and nuclear retention, as shown previously (8). Coexpression of ORF57 completely rescues ORF47 expression (Fig. 3B, lane 5). In the case of SRSF3 knockdown, we observed a partial rescue of ORF47 reporter gene expression in the absence of ORF57 (Fig. 3B, lane 3). To rule out secondary effects due to siRNA application, we lentivirally expressed an anti-SRSF3 shRNA, which binds to the SRSF3 3′-UTR (Fig. 3C), and again confirmed the rescue effect of the SRSF3 knockdown (Fig. 3C, lane 2). In contrast, the rescue effect is not mirrored on the ORF47 RNA level. As shown in Fig. S2J, the knockdown of SRSF3 did not lead to an increase in ORF47 RNA. Thus, the depletion of SRSF3 allowed for a partial rescue of ORF47 expression, but not via increased RNA stability. However, ORF57 coexpression further increased ORF47 expression in all conditions (Fig. 3C, lanes 4–6). Finally, we applied the same approach to other ORF57-dependent viral genes such as K15 and ORF6. This choice was based on the characteristics of these lytic genes. K15 undergoes extensive splicing, which is unique amongst herpesviruses, and ORF6 harbors a long 3′-UTR and exceeds ORF47 RNA by a factor of 6 in length. In addition, both genes have a comparable AT content with ORF47. Surprisingly, no rescue was detectable upon SRSF3 knockdown in the absence of ORF57 for ORF6 and K15, indicating a specific effect on the ORF47 transcripts (Fig. 3D). For all tested transcripts, SRSF3 knockdown decreases the expression of the target gene in the presence of ORF57 (Fig. 3B–D). In sum, the follow-up of our captured candidate RBPs revealed that SRSF3 is an important cofactor for ORF57 but at the same time further potentiates RNA instability for the ORF47 RNA.

SRSF3 levels are critical for the expression of viral lytic transcript

Next, we overexpressed SRSF3 by cotransfection experiments along with ORF57-dependent lytic genes, leading to a small, but detectable, increase in SRSF3 levels compared with the endogenous SRSF3 levels (Fig. 4A–C). Nonetheless, increasing the SRSF3 levels, even to a small extent, affected all ORF57-dependent transcripts tested. Both protein and RNA analysis detected reporter gene expression even in the absence of ORF57 (Fig. 4A–F). This effect is strongest for ORF47 and K15 (Fig. 4A and B, lane 2). The Northern blots in Fig. 4D–F also show a slight increase in the RNA signal for all tested genes in the presence of SRSF3 (compare lanes 2 and 1). The effect of ORF57 and SRSF3 seems to be partially additive, as the signal is further enhanced upon cotransfection, which is best seen for K15 (Fig. 4B and E). We also tested the cytomegalovirus AT-rich UL11 gene, which depends on the CMV homolog of ORF57, pUL69 (8). Also, here UL11 expression could be increased by SRSF3 overexpression (Fig. S2H) and is not enhanced by UAP56 overexpression. We also tested two different GFP reporter constructs, which we used in our previous work (8), to rule out a general effect of higher SRSF3 levels. In the case of unmodified eGFP, neither ORF57 nor SRSF3 cotransfection affected the MFI of the reporter (Fig. S2I, right bar graph). We also tested an HIVGFP reporter (41), where the GFP sequence was adapted to the A-rich nucleotide bias of HIV. No enhancing effect of SRSF3 was detectable compared with ORF57 (Fig. S2I, left bar graph). This indicates that our SRSF3 phenotype is specific for particular sequences enriched in some herpesvirus RNAs. In summary, the phenotype is complex since the knockdown rescues only certain transcripts and not all lytic RNAs. On the other hand, SRSF3 overexpression enables KSHV gene expression in the absence of ORF57 for all tested genes.

Fig. 4.

A–C) Immunoblots using lysates from transiently transfected HEK293T cells showing the effect of SRSF3 overexpression on ORF47, K15, and ORF6 expression. Bar graphs show the quantification of SRSF3 expression levels for each blot, normalized to actin. D–F) Northern blots showing the effect of SRSF3 overexpression on ORF47, K15, and ORF6 RNA expression in cotransfected HEK293T cells.

RNA-interactome capture in virus-infected cells identifies RBPs specifically recognizing the lytic ORF6 transcript

Next, we aimed to confirm our findings in the more relevant context of viral reactivation in KSHV-infected cells.

First, we performed proteomic analyses of the reactivated iSLK rKSHV.219 cells at 24 and 48 h after induction to check for protein expression of individual viral proteins (Fig. S3A, upper panel). At later time points, such as 72 h post induction, virus-producing cells deteriorate and were considered unsuitable for proteomic analysis. We first checked for the endogenous levels of SRSF3, but they remained unchanged after reactivation (Fig. S3D). Although multiple viral proteins could be readily detected, the ORF47 protein was not among them at these time points. We turned to ORF6 as a target due to its abundance throughout lytic reactivation and the identical AT content (ORF6 49.5%, ORF47 49.6%; Fig. S3A, upper panel), while being much larger (3,399 nt compared with 504 nt for ORF47). Even though ORF59 was more abundant, we excluded it from the analysis (Fig. S3A, lower panel) because of its lower AT content (44.3%). We confirmed the ORF57-dependent expression of ORF6 using transient cotransfections (Fig. 5A). RNA quality controls confirmed that the ORF6 transcript could be efficiently enriched compared with GAPDH or other viral RNAs like K15 or ORF59 (Fig. 5B).

Fig. 5.

A) Immunoblot using cotransfections of ORF6 with ORF57 to confirm ORF57 dependency in HEK293T cells. Workflow for the capture method in the context of viral reactivation in iSLK rKSHV219 cells. B) RT-qPCR to control capture efficiency of the ORF6 transcript compared with GAPDH, viral K15, and ORF59 RNAs displayed relative to the corresponding input RNA levels as in Fig. 1E. C) Volcano plot comparing the fold change difference and P-value of the proteins identified in the crosslinked (crlk) and not crosslinked (no crlk) samples for the ORF6 capture experiments (n = 4). Known hits are colored in red, ORF6 in green, and all hnRNPs in blue. D) Venn diagram showing the overlap of identified proteins between the ORF47 captures in the cotransfection context (Fig. 1A) and the ORF6 capture experiments.

Mass spectrometry analyses of the ORF6 RNA capture identified a considerable number of proteins that were enriched for the ORF6 transcript compared with the “no crosslink” sample (Fig. 5C). Among the highly enriched candidates, we found many RBPs belonging to the hnRNP protein family. As for the ORF47 capture, we found ORF57, RBM15, and SRSF3. We also identified the viral protein ORF6 itself, which is the major DNA-binding protein of KSHV and initiates replication (42). Its affinity for nucleic acids might be the reason, but the relevance of ORF6 RNA-binding has to be further analyzed. The Venn diagram (Fig. 5D) displays overlapping and unique interactors with 57 common interactors, representing 28 and 65% of all enriched candidates, respectively (Fig. 5D). This argues that the similarities of both transcripts regarding AT richness and ORF57-dependency result in a similar interactome.

SRSF3 influences viral infectivity

SRSF3 binds to viral RNAs in both transfections and viral reactivation assays. Since it is also required for the expression of some viral genes, we reasoned that knockdown or overexpression of SRSF3 would impact viral infectivity (Fig. 6A). We first confirmed both knockdown and overexpression of SRSF3 by immunoblot (Fig. 6B) and unaltered ORF57 levels in the KSHV-harboring iSLK cells (Fig. S3C). The flow cytometry results of the infectivity assay clearly showed a >2-fold reduction in the percentage of GFP-positive cells infected with the supernatants of SRSF3 knockdown cells for all dilutions (Fig. 6C, left panel). Conversely, when cells were transduced with both the SRSF3 shRNA and the SRSF3 overexpression vector, a strong rescue phenotype was observed, and infectivity reached the same level as for the supernatants of untreated cells. The same results could also be observed by looking at infectious units per milliliter (Fig. 6C, right panel). We also compared the mean fluorescence intensity (MFI) of the samples, which again showed an identical result, with a strong reduction in MFI for the knockdown samples and a clear rescue of MFI for the cotransduction (Fig. S3B).

To identify which step of the viral life cycle is affected by SRSF3 knockdown, we measured genome copy numbers in the viral supernatants (Fig. 6D), showing a 60% reduction in viral genomes, suggesting a lower number of released viral particles. Conversely, viral DNA replication is not affected (Fig. 6E), indicating a possible effect on late gene expression and/or assembly instead of a general effect of the SRSF3 knockdown. Taken together, our results from the virus reactivation demonstrate the importance of SRSF3 for viral infectivity.

Discussion

Here, we applied and adapted the LNA/DNA mixmer capture method to identify the RNA interactome of specific herpesviral transcripts. The recently published SHIFTR method was only applied to highly abundant endogenous RNAs such as U1snRNA and housekeeping genes (43). In our approach, we focus on the identification of proteins that might influence the stability or promote the degradation of viral lytic transcripts, thereby making them dependent on the viral transactivator ORF57. Moreover, we sought to decipher the mRNP composition of KSHV lytic genes.

To our knowledge, we adapted the LNA capture protocol to transient transfections and viral lytic reactivation for the first time. Nevertheless, transient transfections display more variability in efficiency and peptide intensity compared with captures performed after lytic reactivation of virus-containing cells, in addition to a certain degree of background plasmid contamination as a bycatch, which cannot be completely removed (Table S1). However, the crosslinking conditions are optimized for RNA binding, and crosslinking to DNA is very weak, since it would require higher energy (44, 45). In addition, our results showed a clear enrichment for RBPs and hardly any DNA-binding proteins.

The identification of direct-interacting RBPs on individual viral RNAs deciphers the mRNP-code, an intrinsic feature of each RNA, in which sequence and structure dictate the interacting RBPs and thus the fate of the respective RNA (46). First, we compared the RNA interactome of wild type vs. sequence (codon) optimized ORF47 (Figs. 1D and S1E). We show for the first time how adaptation to human codon usage, mainly by increasing the GC content of mRNAs, changes the composition of bound RBPs. These changes will have profound effects on mRNA fate and are uncoupled from translation. Previously, we showed that sequence (codon) optimization increases nuclear RNA stability and the export of KSHV ORF57-dependent genes (8). We observed a decrease in bound hnRNP protein both in silico and in vitro. In sum, codon optimization manipulates the mRNP of the respective RNAs. These changes cannot be neglected if translational effects of codon substitutions are analyzed, at least for RNAs synthesized in the nucleus (33).

The direct binding of ORF57 without any cofactors to viral RNAs has been debated for a long time due to the lack of a typical additional RNA-binding domain present in other herpesviral homologs (47). With the advent of CLIP, crosslinking of ORF57 to some viral clusters was observed (48) in addition to the defined ORF57-binding sites, e.g. in the viral PAN RNA (48). In agreement with recent findings (49), our results now clearly show a direct binding of ORF57 to viral RNA using very stringent conditions. We speculate that ORF57's N-terminal disordered region folds upon contact with a viral RNA sequence into an RRM-like domain and clamps the RNA with the aid of the basic residues in NLS1/2 and the AT hook (50). This raises the question of the nature of the ORF57-binding motif in viral RNAs. The eCLIP experiments mapped ORF57-binding sites/motifs across the KSHV genome, corresponding well with ORF57-dependent genes. Some of these more prominent peaks also appear as lower background in the SMI samples, which is a reported problem with the SMI control for highly abundant RBPs (51), like the abundant ORF57 after reactivation (Fig. S3A). Strikingly, from half of the ORF57-binding sites, an AU-rich motif could be derived. We hypothesize that the AT richness of ORF57-dependent genes constitutes direct binding motifs for ORF57 and that this distinction favors viral over cellular genes and represents the core of ORF57's target specificity.

An unexpected finding in this study was the role of SRSF3 in restricting ORF47 expression. It was shown previously in Epstein–Barr virus (EBV) that SRSF3 targets viral intronless RNAs for degradation by the cellular RNA exosome. EB2, the EBV homolog of ORF57, counteracts this degradation by protecting the viral transcripts (31). However, their data contradict findings from cellular RNA processing where SRSF3 plays an export-promoting function (32). Interestingly, SRSF3 behaves similarly for the KSHV ORF47 transcripts, which encode one of the viral glycoproteins. Its expression can be partially rescued by the knockdown of SRSF3, which is in agreement with the findings of (31). However, SRSF3 depletion did not increase ORF47 RNA levels (Fig. S2J). We can only speculate that nuclear export, translation, or subcellular localization may play a role. Surprisingly, this effect could not be detected for the other KSHV RNAs tested, even with similar AT content. Moreover, for all genes analyzed, including ORF47, an expression-promoting effect of SRSF3 overexpression can be observed. This is remarkable in the sense that genes such as K15 are expressed only at very low levels in the absence of ORF57. This phenotype suggests that the amount of available SRSF3 might be limiting. We speculate that SRSF3 has a dual role in KSHV lytic gene expression. It may provide a second fail-safe mechanism to avoid the accidental expression of certain KSHV proteins in the latent phase. In the absence of ORF57, SRSF3 may alter the fate of ORF47 RNA and block expression. During overexpression, this effect is masked by the positive effects of SRSF3 on nuclear RNA export and translation. Similarly, ORF57 can override the negative effect of SRSF3 on certain transcripts by recruiting Aly and additional export/stabilization factors.

Our data obtained using the rKSHV-harboring iSLK cells demonstrate a strong decrease of infectivity upon SRSF3 knockdown. This decrease of viral genomes in the supernatant argues for a role of SRSF3 in late gene expression. In conclusion, we demonstrated that SRSF3 is an important determinant for the fate of viral transcripts and functions both in synergy with and in opposition to ORF57 to influence viral infectivity. However, the exact mechanism of action and whether only certain subsets of lytic RNAs are bound and regulated by SRSF3 require further studies.

Materials and methods

Cells

HEK293T cells were grown in Dulbecco-modified Eagle medium supplemented with 10% fetal calf serum and 1% penicillin–streptomycin. iSLK rKSHV.219 cells were cultured in DMEM supplemented with 10% fetal calf serum, 1% penicillin–streptomycin, 250 mg/mL G418, and 40 µg/mL Puromycin.

Plasmids

The pCDNA3.1(+)-based ORF47, ORF57, and K15 plasmids were described previously in (8). The ORF6 ORF was PCR-amplified using BAC36 genomic DNA as the template and then inserted into pCDNA3.1(+) using HindIII and XbaI restriction sites. An HA-tag was introduced at the C-terminus.

The SRSF3 cDNA was ordered as a gBlocks gene fragment at IDT and then introduced into the pCDNA3.1(+) using ApaI and EcoRI restriction sites. For the lentiviral SRSF3-RRLPPTSF-based vector, the SRSF3 sequence was inserted using the SalI restriction site. The previously described deoptimized green fluorescent protein (HIVGFP) was kindly provided by R. Wagner (University Regensburg).

LNA capture (transfection and viral capture)

Cell preparation

For the capture experiments in HEK293T cells, 12.5 × 10⁶ cells were seeded per 15 cm dish 1 day before transfection. For each sample (crosslink, no crosslink control, and scrambled crosslink control sample), eight dishes were used. Transfections by calcium phosphate used 12.5 µg ORF47 plus 6.25 µg ORF57 plasmids and eGFP as a transfection control (1.25 µg). The medium was changed after 6 h or the next morning. Cells were crosslinked and harvested 48 h post transfection. For the capture experiments in iSLK rKSHV cells, 3 × 10⁶ cells were seeded per 15 cm dish. For each sample, 10 dishes were used. The next day, cells were induced using 1 mM sodium butyrate and 1 µg/mL doxycycline. Cells were crosslinked and harvested 48 h post induction. The capture method was adapted from (20) and (19).

Cell harvest

Cells were washed with PBS twice, removed, placed on ice, and crosslinked with 150 mJ/cm² of UV light at 254 nm. After irradiation, 10 mL of cold PBS was added immediately, and the dish was placed in the fridge while crosslinking continued. Subsequently, cells were harvested with a cell scraper, centrifuged, and lysed in 7–10 mL of 1× lysis buffer (20 mM Tris-HCl [pH 7.5], 500 mM LiCl, 0.5% LiDS [wt/vol, stock 10%], 1 mM EDTA, and 5 mM DTT [added directly before use]) with protease inhibitors (complete EDTA-free, Roche) and RNAsin Plus, Promega. The samples were incubated for 15 min on ice, and 100 U/10 mL TurboDNAse was added. The lysates were vortexed a few times during the incubation. Then, lysates were homogenized through a narrow gauge needle (0.4 mm diameter, 5 mL syringe) and frozen in liquid nitrogen.

Preparation of beads

The antisense LNA/DNA mixmers were designed as described previously (20, 52) and ordered from Biomers: ORF47LNAg + aag + tgg + ataga + gtg + ga3AmMO; (SynORF47LNA)a + tga + tga + tgt + tcac + gtt + gt3AmMO; ScrambleLNAatac + ttaac + cgta + tat + tcc + ta3AmMO; ORF6LNAc + tcc + tct + ctg + attc + ttc + tga + ggca + gc3AmMO.

The coupling of the mixmers to the carboxylated magnetic beads (PerkinElmer MPVAC11) was performed as described (20).

RNP capture

On the day of the capture, lysates were thawed at 37 °C and additional TurboDNase was added. The falcon tubes with the samples were prewarmed for 15 min at 60 °C, quickly cooled down on ice, and the lysates were cleared for 5 min at 3,810 × g at 4 °C. Five millimolars extra DTT were added. One hundred microliters input were saved for the RNA quality controls. The LNA/DNA mixmer coupled beads were equilibrated by three washes in lysis buffer and heated for 3 min at 95 °C. Then the beads were added to the lysates and incubated for 1 h at 41 °C. After hybridization, beads were washed in 10 mL lysis buffer for 5 min at 41 °C and subsequently twice in 10 mL of buffers with decreasing concentrations of LiDS and LiCl (buffer 1 mix 20 mM Tris-HCl (pH 7.5), 500 mM LiCl, 0.1% LiDS (wt/vol), 1 mM EDTA, and 5 mM DTT; buffer 2 mix 20 mM Tris-HCl (pH 7.5), 500 mM LiCl, 1 mM EDTA, and 5 mM DTT + 0.02% NP40; buffer 3 mix 20 mM Tris-HCl (pH 7.5), 200 mM LiCl, 1 mM EDTA, and 5 mM DTT + 0.02% NP40). All buffers were prewarmed. A pre-elution was carried out with 200 µL nuclease-free water for 5 min at 40 °C and 800 rpm in an Eppendorf ThermoMixer C. Finally, the samples were heat eluted in 150 µL nuclease-free water for 3 min at 90 °C and 800 rpm. For each experiment, three capture rounds were carried out, with recycling of beads after the first round. The beads were solved in 400 µL nuclease-free water and incubated for 5 min at 95 °C and 800 rpm. Then, they were washed three times in water and lysis buffer before they were added again to the lysates. The lysates from the capture rounds were pooled, frozen in liquid nitrogen, and stored at −80 °C.

Preparation for mass spectrometry and RNA quality controls

Samples were thawed and placed on the magnet again to remove any remaining beads. Twenty microliters of each sample were kept for quality controls. The rest of the sample was prepared for proteomic analysis by Benzonase (Millipore >250 U/µL) digestion for 1 h at 37 °C.

RNA quality controls

Twenty microliters of input and 20 µL eluate were proteinase K digested (Ambion AM2546 20 mg/mL; 1 µL proteinase K, 5 µL proteinase K buffer [50 mM Tris-HCl pH 7.5, 750 mM NaCl, 1% (wt/vol) sodium dodecyl sulfate (SDS), 50 mM EDTA, 2–5 mM DTT (add directly before use), 25 mM CaCl₂ autoclave]) at 65 °C for 1 h. Afterward, the proteinase K was inactivated at 95 °C for 10 min. Samples were then Turbo DNase digested in 50 µL total volume for 1 h at 37 °C. RNA was isolated with the RNeasy kit (Qiagen). And 6 µL was used for an RT reaction with QuantiTect Reverse Transcription Kit (Qiagen). qPCR was performed using the QuantiFast SYBR Green kit (Qiagen) and transcript-specific primers, whose sequences can be provided upon request.

For the next-generation sequencing (NGS), between 7.4 and 11.6 ng cDNA were used as an input in the library preparation, using a KAPA HTP Kit, following the manufacturer's protocol. The resulting libraries were checked on a bioanalyzer and loaded at 12 pM on an Illumina MiSeq 600v3 Kit, giving each sample 3% flow cell coverage. The reads were trimmed with fastp 0.23.2 and mapped against the ORF47 sequence, using Geneious Prime 2019 and the bowtie2 plugin. As a control step, the reads were mapped against the human Genome hg38, using bowtie2 2.3.5.1 on the command line, and the statistics were examined.

Mass spectrometry

Samples from RNP capture were prepared as described above. For analyses of viral protein expression upon viral reactivation, 4 × 10⁵ iSLK rKSHV cells were seeded per 6 cm dish. The lytic cycle was induced the next day as described above with 1 mM sodium butyrate and 1 µg/mL doxycycline. Cells were harvested 24 and 48 h post induction, washed twice with PBS, and transferred into protein low-bind tubes; 1 × 10⁶ cells were lysed for 10 min on ice in 200 µL 50 mM HEPES pH 8.5 containing 1% dodecylmaltoside, 10 mM NaF, 0.5 mM Na₂VO₃, 1 mM EDTA, 1 mM EGTA, 100 mM NaCl supplemented with protease inhibitors (cOmplete, EDTA-free). Protein digestion and peptide purification for LC–MS/MS were performed using a slightly adapted SP3 protocol (53). Samples were vacuum-dried and resuspended in 0.1% folic acid before applying to C18 Evotips (EV-2001; Evosep), according to the manufacturer's protocol. Evotips were then loaded on an Evosep One HPLC (Evosep) connected to a TimsTOFPro mass spectrometer (Bruker). The Evosep One HPLC was operated with the standard 60 sample-per-day method (21 min gradient at a flow rate of 1.0 µL/min; buffer A: 0.1% formic acid, buffer B: 0.1% formic acid in acetonitrile). For the TimsTOFPro mass spectrometer, the standard MSMS Bruker method “DDA PASEF method for short gradients with 0.5 s cycletime” was employed. MS settings in detail were: scan begin 100 m/z; scan end 1,700 m/z; ion polarity: positive; scan mode: PASEF. Tims settings were: mode custom; number of PASEF ramps: 4; charge minimum: 0; charge maximum: 5; 1/K0 start 0.75 Vs/cm²; 1/K0 end 1.4 Vs/cm²; ramp time: 100.0 ms; MS average: 1. Data analysis and peptide quantification were performed using PEAKS Studio XPro (version 10.6). Further data analysis, data visualization, and statistics were performed using the Perseus software (54), an online tool for Venn diagrams from VIB/UGent Bioinformatics & Evolutionary Genomics (http://bioinformatics.psb.ugent.be/webtools/Venn/), GraphPad Prism 10, GOrilla (55, 56), and STRING protein–protein interaction networks (57).

Capture experiments were performed in at least three independent biological replicates (n ≥ 3). Ratios of protein abundances to compare the “crosslinked” and “no crosslinked” samples in a label-free quantification manner were calculated, missing values were imputed by random low values from a normal distribution, and significances were tested using Student's t test. Calculations, imputation, and t testing were performed using the Perseus software, version 2.0.3.1 (54).

siRNA knockdown

For the knockdown experiments, 4 × 10⁵ HEK293T cells were seeded per 6 wells. The next day cells were transfected with 30 pmol Silencer Select siRNAs targeting the gene of interest and control siRNA using Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher). On the following day, the cells were split 1:4 to ensure an appropriate cell density for the plasmid transfection. On day 4, transfections were performed using Viafect transfection reagent (Promega) with 1 µg of the reporter plasmid (e.g. ORF47-HA) plus 400 ng of ORF57, 40 ng of enhanced GFP (eGFP) as a transfection control, and 600 ng of empty plasmid DNA. The cells were harvested 72 h after siRNA transfection.

Silencer select siRNAs: SRSF3 (SRSF3 I) s12733, SRSF3 (SRSF3 II) s12731, HNRNPD (s6723), RBM15 (s34936), PUF60 (s22447), KHDRBS1 (s223144), ILF2 (s7400), TIA1 (s14131), TRA2A (s26664).

shRNA knockdown

The SRSF3 shRNA targets the SRSF3 3′-UTR. Top and bottom oligonucleotides (shSRSF3 top GATCCGGATCATACGAGGCATGTAATTCAAGAGATTACATGCCTCGTATGATCTTTTTTACGCGT

G; shSRSF3 bottom AATTCACGCGTAAAAAAGATCATACGAGGCATGTAATCTCTTGAATTACATGCCTCGTATGATCCG) were annealed by mixing 32.5 pmol of each oligo with 10× annealing buffer (PNK buffer A, ThermoFisher) in a total volume of 20 μL. Then, the mixture was incubated for 3 min at 98 °C and cooled to room temperature. The annealed oligonucleotides were phosphorylated using the T4 polynucleotide kinase. Finally, the annealed oligonucleotides were inserted into the retroviral vector pSIREN (Clontech) using the BamHI/EcoRI restriction sites.

Retroviral vector production

HEK 293T cells were transfected by the calcium phosphate with 5 μg pSIREN (retroviral SRSF3 shRNA vector), 10 μg gag-pol, and 0.5 μg vsvg (packaging and envelope plasmids). Supernatants were collected 36 and 48 h after transfection and passed through a 0.22-μm filter (Millipore). For the shRNA knockdown, 4 × 10⁵ 293T cells were seeded in a 6-well plate. The cells were transduced with 200 μL of the collected supernatants in the presence of 4 mg/mL protamine sulfate.

RNA preparation and northern blot

RNA preparations for northern blot analysis were performed, as described previously (58). GAPDH, K15, and ORF57 probes were generated as described in (8). The ORF6 probe was prepared by digestion of the pCDNA3.1(+) based ORF6 plasmid with BsrGI and NheI, and the 526 bp fragment was radioactively labeled using ³²P dCTP.

Western blot

For protein detection, the cells were lysed with 1× SDS protein lysis buffer (62.5 mM Tris-HCl [pH 6.8], 2% [wt/vol] SDS, 10% glycerol, 50 mM dithiothreitol, 0.01% [wt/vol] bromophenol blue) and boiled for 5 min. Only for the detection of K15 protein, cell pellets were incubated on ice for 10 min with RIPA-100 buffer (20 mM Tris [pH 7.5], 1 mM EDTA, 100 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) and then centrifuged at maximum speed in an Eppendorf 5425 centrifuge for 20 min to clear the lysates.

Lysates were analyzed by immunoblotting. HA-tagged proteins were detected with rat monoclonal anti-HA 3F10 in a 1:2,000 dilution (Roche). K15 protein was detected with a rat antibody directed against the cytoplasmic tail of K15 in a 1:250 dilution (provided by E. Kremmer, Helmholtz Center, Munich). Actin was detected with a mouse monoclonal anti-actin antibody in a 1:1,000 dilution (Sigma-Aldrich) or hFAB Rhodamine anti-actin primary antibody (Bio-Rad). Tubulin was detected with hFAB Rhodamine Anti-Tubulin Primary Antibody in a 1:1,000 dilution (Bio-Rad). SRSF3 was detected with mouse monoclonal antibody 7B4A12 in a 1:1,000 dilution (ThermoFisher). TRA2A was detected with rabbit polyclonal A303-779A antibody in a 1:2,000 dilution, PUF60 with rabbit polyclonal A303-819A antibody in a 1:2,000 dilution, KHDRBS1/Sam68 with rabbit polyclonal antibody A302-110A in a 1:2,000 dilution, RBM15 with rabbit polyclonal antibody A300-821A in a 1:10,000, TIA with rabbit polyclonal TIA-1 antibody A303-990A in a 1:1,000 dilution and ILF2 (NF45) with rabbit polyclonal A303-147A antibody in a 1:1,000 dilution (all Bethyl). hnRNPD was detected with rabbit polyclonal antibody PA5-95948 in a 1:2,000 dilution (Invitrogen). Detection was carried out by using a standard enhanced chemiluminescence reaction.

Infectivity assay

Production of rKSHV supernatants: 2 × 10⁵ iSLK rKSHV cells per 6 wells were seeded. The next day, the cells were transduced with 200 µL lentiviral supernatant for the SRSF3 knockdown and 1,000 µL for SRSF3 overexpression. Twenty-four hours post transduction, the lytic cycle was induced by the addition of 1 mM sodium butyrate and 1 µg/mL doxycycline. Seventy-two hours post induction, the supernatants were harvested by filtering through a 0.45-µm filter and then stored at 4 °C overnight. The iSLK rKSHV cells were harvested for western blot controls in 100 µL 1× SDS loading dye. For the actual infectivity assay, 8 × 10⁴ HEK293T cells were seeded per 24 wells. Before the HEK293T cells were incubated with the stored iSLK rKSHV supernatants in different dilutions, cells were shaken for 20 min at room temperature and centrifuged for 20 min at 450 × g. After 48 h of incubation, the cells were harvested for flow cytometry analysis in 250 µL fluorescence-activated cell sorting (FACS) buffer (PBS + 4% FCS) and fixated with 250 µL 4% PFA. GFP-positive cells were measured by fluorescence-activated cell sorting analysis using the CytoFLEX Flow Cytometer from Beckman Coulter.

qPCR analysis of viral genome copy number

For the analysis of viral genomes in the supernatant, 48 h post induction of the KSHV lytic phase, 1.5 mL of viral supernatant was collected and passed through a 0.45-micron filter. Each sample was incubated with 5 µL of Benzonase at 37 °C for 1 h. Four hundred microliters of viral supernatant were used for DNA extraction using the Qiagen Blood and Tissue Kit. For the analysis of genome replication, 48 h post induction, the iSLK.219 cells were counted and normalized. The cell pellets were suspended in 200 µL of PBS and were used for DNA extraction using the Qiagen Blood & Tissue Kit. The genome copy number was calculated for a million cells.

A TaqMan probe-based qPCR was used for the absolute quantification of the viral ORF25 gene. A dual-labeled probe was designed against ORF25 (5′[6FAM]CCACCCAGTCAGCCCAGGCACTAAAC[BHQ1]3′)(59). A 68-bp amplicon of the ORF25 was first isolated by PCR using the following primers (forward primer: 5′CCACCCTCGAATGCACAAC3′ and reverse primer: 5′GTCGGGATCGGGAAAAGCT3′) (59) and cloned in the pCR-Blunt vector. The plasmid copy number was determined using the formula: number of copies/µL = (measured concentration of plasmid [ng]×Avogadro's constant)/(plasmid length [bp]×10⁹×∼ 660) (“Weight” of the plasmid = Plasmid length [3,580 bp]×Weight of individual base pairs in g/mol = 2.34 × 10⁶ g/mol). A serial dilution was performed accordingly, and a standard curve was generated with a 20-µL qPCR reaction using Luna Universal Probe qPCR Master Mix Kit on a qTOWER³ Real-Time Thermocycler. With a similar PCR reaction, C_t values were generated for the viral supernatant and cell samples, and the genome copy number was calculated using the equation derived from the standard curve.

ORF57 eCLIP

For the eCLIP experiment, 3 × 10⁶ iSLK rKSHV cells were seeded per 15 cm dish. For each sample, five dishes were seeded. The next day, the lytic cycle was induced as described above. After 48 h post induction, the cells were washed twice with cold PBS and then placed on ice for crosslinking with 150 mJ/cm² of UV light at 254 nm. After irradiation, 10 mL of cold PBS was added immediately. Subsequently, cells were harvested with a cell scraper, centrifuged, and the pellets were snap frozen in liquid nitrogen and stored at −80 °C. The pellets were lysed in 0.9 mL cold lysis buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1% NP40, 0.1% SDS, 0.5% sodium deoxycholate) with protease inhibitors (complete EDTA-free, Roche) and 15 µL RNAsin Plus, Promega. Then, the lysates were homogenized through a narrow gauge needle before sonication with the Diagenode Bioruptor Sonication System with a low setting, at 4 °C for 5 min 30 s on/30 s off. The lysates were placed in a low retention tube, and 15 µL of Turbo DNAse was added. RNase I was diluted 1:25 in prechilled 1× PBS, and 10 µL of diluted RNase was added to the lysates, mixed, and immediately incubated at 1,200 rpm, at 37 °C for 3 min. After RNase I treatment, the lysate was centrifuged at maximum speed at 4 °C for 10 min, and the supernatant was transferred to a new tube. For preclearing, 20 µL of prewashed protein G beads were added per sample and incubated at 17 rpm for 30 min at 4 °C. Beads were then separated magnetically, the supernatant transferred into a new tube, and snap frozen in liquid nitrogen.

The immunoprecipitation of ORF57-RNA complexes was performed in two independent biological replicates along with two SMI controls, as described previously (36, 37). Concentrated lysates were incubated with 10 μg of anti-ORF57 antibody (Anti-KSHV ORF57 [rabbit] antibody 600-401-A94 Rockland) for 4 h, followed by incubation with 50 μL of protein G magnetic beads (Invitrogen) for 1.5 h. After washing, end repair, and on-bead 3′ adaptor ligation, the protein–RNA complexes were eluted and then resolved by denaturing gel electrophoresis and transferred onto a nitrocellulose membrane. Library preparation was carried out, as described in (37). Libraries were sequenced on the Illumina NextSeq 500 platform.

eCLIP analysis

Demultiplexed paired-end fastq files underwent two rounds of trimming using cutadapt (60) prior to alignment against a combined human (HG38) and KSHV (NC_009333.1) genome file, following the protocol outlined here https://pureclip.readthedocs.io/en/latest/GettingStarted/preprocessing.html. The resulting BAM file was indexed using SAMtools v1.15 (61) and parsed to retain only primary alignments against the main chromosomes and KSHV genome [samtools view -f 2]. Alignments arising from PCR duplicates were removed using UMI_tools (62) prior to further filtering with SAMtools to retain only the read 2 alignments [samtools view -f 130]. Biological replicates were merged for both CLIP and SMI datasets and putative binding sites/regions identified using PureClip v1.3.1 (63). Here, HMM training was performed using HG38 chromosomes 1–3 and the KSHV genome, while output binding sites were clustered into regions if within 10 nucleotides [-iv ‘chr1;chr2;chr3;KSHV-GK18-NC_009333’ -nt 10 -dm 10]. Binding regions/sites were analyzed separately and in combination for HG38 and KSHV using findgenomemotifs.pl from HOMER v4.10 (64) [-size given -len 5,6,7,8 −rna]. Genome coverage bedgraphs were generated using BEDtools v2.27.1 (65) and visualized using R v4.1.1 with the package Gviz (66).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This study was funded by the German Research Council grant (BO 2512/8-1). Ruchria Bhowmik is supported by the German Exchange program (DAAD) Doctoral Programmes in Germany, 2022/23 (57588370).

Author Contributions

Carolin Vogt (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing—original draft, Writing—review & editing), Marco van Ham (Data curation, Formal analysis, Methodology, Visualization), Ruchira Bhowmik (Formal analysis, Methodology, Visualization, Writing—review & editing), Amir Argoetti (Data curation, Formal analysis, Methodology), Daniel P. Depledge (Data curation, Formal analysis, Methodology, Software, Validation, Visualization, Writing—review & editing), Lars Steinbrück (Data curation, Formal analysis, Methodology, Validation), Jasper Götting (Data curation, Formal analysis, Validation), Carolina Henkel (Methodology, Validation), Andrea Cuadra Granados (Methodology, Validation), Yael Mandel-Gutfreund (Funding acquisition, Investigation, Methodology, Resources, Software, Writing—review & editing), Lothar Jänsch (Data curation, Formal analysis, Project administration, Supervision, Validation), and Jens Bohne (Conceptualization, Funding acquisition, Investigation, Project administration, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing)

Data Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [1] partner repository with the dataset identifier PXD056253. The eCLIP raw data was uploaded to GEO: GSE253314 RNA-interactome capture identifies key proteins for herpesviral gene expression (2025 January 15), GSM8018690 Kaposi's sarcoma-derived cell line, eCLIP1 (2025 January 15), GSM8018691 Kaposi's sarcoma-derived cell line, eCLIP2 (2025 January 15), GSM8018692 Kaposi's sarcoma-derived cell line, size-matched input 1 (2025 January 15), and GSM8018693 Kaposi's sarcoma-derived cell line, size-matched input 2 (2025 January 15).