The anti-viral protein Shiftless blocks p-body formation during KSHV infection

TL;DR

This study shows that the protein Shiftless disrupts RNA processing bodies during KSHV infection, which helps block viral replication.

Contribution

The study identifies a novel region in Shiftless responsible for disrupting P-bodies and links this activity to anti-viral defense.

Findings

Shiftless restricts P-body formation even under oxidative stress.

A specific region of Shiftless is necessary for P-body disruption and anti-viral activity.

Disruption of P-bodies by Shiftless increases anti-viral cytokine expression.

Abstract

Processing bodies (P-bodies/PBs) are non-membranous foci involved in coordinating RNA fate by regulating translation and mRNA decay. In this study, we characterize the anti-viral factor Shiftless (SHFL) as a potent disruptor of PB dynamics. We show SHFL expression restricts PB accumulation even in the context of oxidative stress, suggesting that SHFL expression impedes PB formation. Mutational approaches revealed that SHFL RNA-binding activity is not required to restrict PB formation. However, we have identified a new region of SHFL, a bridge between two distant SHFL domains, as necessary for SHFL-mediated PB disruption. Furthermore, we show that SHFL’s ability to disrupt PB formation also impacts its anti-viral activity during infection by the gammaherpesvirus Kaposi’s sarcoma-associated herpesvirus (KSHV). While WT SHFL efficiently restricts KSHV lytic reactivation, SHFL mutants defective in PB disruption no longer restrict KSHV reactivation. SHFL-mediated PB disruption also leads to increased expression of several anti-viral cytokines, further emphasizing the connection among SHFL, PB dynamics and the SHFL-dependent anti-viral state. Taken together, our observations suggest a role of SHFL in inhibiting PB formation to restrict KSHV lytic replication, reinforcing the importance of crosstalk between RNA fate and the innate immune response to viral infection.

Figures

SHFL blocks de novo synthesis of PB. (a) HEK293T cells were transfected with either a Flag-tagged SHFL (or mock). Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). (b) HEK293T cells were transfected with either a Flag-tagged SHFL (or mock). Briefly, 24 h later, cells were treated for 30 min with 0.25 mM NaAs, then subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). Here and in all figures, the scale bar represents 10 µm. The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, , P<0.01, *, P<0.001.

SHFL RNA-binding activity is not required for PB disruption. (a) HEK293T cells were transfected with either a full-length Flag-tagged SHFL or SHFL RNA-binding domain mutants (ΔRBD or Δ3R as indicated). Briefly, 24 h later, cells were harvested and subjected to immunoblot and stained with the indicated antibodies. (b) and (c) HEK293T cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, , P<0.01, *, P<0.001.

Several domains of SHFL contribute to PB restriction. (a) Schematic representation of SHFL incremental deletion mutants. (b) and (c) Cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, , P<0.01, *, P<0.001.

Domains 151–200 and 241–290 form an interface important necessary for PB restriction. (a) HEK293T cells were transfected with the indicated Flag-tagged SHFL constructs, and 24 h later, cells were harvested and subjected to immunoblot and stained with the indicated antibodies. (b) and (c) Cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. (d) AlphaFold prediction of SHFL folding, highlighting closeness of domains 151–200 and 241–290, and amino acids G259 and W191 possible bond. G259A-W191A (ΔGW) mutant of SHFL was constructed and transfected into HEK293T cells, confirming that these mutations did not affect the stability of SHFL. Immunofluorescence of stained PB, however, revealed no loss of PB upon expression of this mutant. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 or n=5 independent biological replicates as indicated. n.s.=not significant, , P<0.01, *, P<0.001.

PB loss upon SHFL expression results in ARE-mRNA overexpression and KSHV lytic restriction. (a) Cells were transfected with SHFL WT or ΔGW (or mock vector) and treated with purified TNF-α (to induce ARE-mRNA expression) at a final concentration of 10 ng ml−1. Total RNA was then harvested and subjected to RT-qPCR to measure mRNA levels of the indicated ARE-containing mRNA. Statistical difference between control and experimental conditions was determined using one-way ANOVA followed by Dunnett’s multiple comparisons. ns=P>0.05, * = P<0.05, = P<0.01. (b) and (c) KSHV-positive iSLK.219 cells were either treated with siRNAs targeting SHFL (or control non-target siRNAs – siSCRB) for 48 h, then transfected with the indicated Flag-tagged SHFL. Cells were then reactivated with doxycycline and sodium butyrate for 48 h, then checked for reactivation efficiency by monitoring the expression of GFP (encoded on the KSHV genome and therefore a marker for infected cells) and RFP (under the control of the PAN promoter and therefore a marker for lytic reactivation). Quantification of RFP-positive cells is to the right. Values represent at least five independent views of the infected cells. Statistical difference between control and experimental conditions was determined using one-way ANOVA followed by Dunnett’s multiple comparisons. ns=P>0.05, * = P<0.05, = P<0.01. (d) iSLK.219 was left latent or reactivated and transfected with the indicated SHFL constructs. Briefly, 48 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6).

Full text

Introduction

The battle between virus and host often culminates around the critical distribution of gene expression resources. A growing body of work highlights how viruses cleverly modulate RNA stability and localization to reshape their host cell gene expression environment. For viruses such as the gammaherpesvirus Kaposi’s sarcoma-associated herpesvirus (KSHV), during lytic replication, it is imperative to rapidly co-opt the host gene expression machinery and repurpose it towards viral gene expression and replication. To achieve this, KSHV induces a widespread RNA decay event that relies on a virus-encoded endoribonuclease, the shutoff alkaline exonuclease SOX. This extensive RNA decay event is accompanied by numerous downstream effects, including a global re-localization of RNA-binding proteins (RBPs) from their bound cytoplasmic mRNA targets to the nucleus [1], a subsequent widespread shutdown of RNA pol-II-based transcription [23], and even hyperadenylation of nascent host transcripts leading to a blockade of nuclear export [4]. All of these events culminate in a large-scale remodelling of the gene expression environment and reallocating of post-transcriptional and post-translational regulatory resources towards viral replication while simultaneously dampening the host immune response [56].

We have previously shown that while KSHV SOX targets the majority of the host transcriptome, some key host transcripts actively evade virus-induced mRNA decay [710]. To date, no consensus appears to unify these escaping mRNAs with either a pro- or anti-viral function. Some, like the host IL-6, even have well-documented pro-viral roles [1115]. However, recently, we identified a novel host mRNA escapee encoding the protein Shiftless (SHFL, C19ORF66, IRAV) – which stringently resists SOX cleavage and encodes a potent anti-viral protein that restricts KSHV lytic replication [1617]. SHFL is an interferon-stimulated gene (ISG) that is emerging as a critical piece of the innate immune response to viral infection, capable of suppressing the replication of multiple DNA, RNA and retroviruses [18]. Recent reports over the past 5 years revealed that SHFL can negatively modulate viral RNA stability, viral gene translation and even viral protein stability [17,1923]. In the context of KSHV infection, we showed that SHFL broadly restricts KSHV lytic gene expression, including that of the master latent-to-lytic switch protein, KSHV ORF50 (RTA), resulting in a global downregulation of the KSHV lytic gene cascade [16]. Surprisingly, we also found that SHFL expression appears to restrict the accumulation of processing bodies (P-bodies/PBs) both outside of and in the context of KSHV infection. The capacity of SHFL to modulate cytoplasmic RNA granules represents one of the first examples of an ISG capable of influencing PB dynamics during viral infection.

In mammalian cells, formation of RNA granules allows flexibility for survival during adverse biological conditions [24]. PBs are non-membranous cytoplasmic ribonucleoprotein (RNP) foci long believed to be solely sites of RNA decay. However, accumulating studies have uncovered roles for PBs as sites of RNA stasis, acting as a ‘storage facility’ for constitutive turnover and regulated release of key transcripts into the translation pool [2530]. While PBs are constitutively present in most cell types, they are also dynamic, changing in size and number in response to shifts in cellular environmental conditions and gene expression [3135]. This plasticity of PBs as biomolecular condensates is attributed to their ability to undergo liquid–liquid phase separation [24]. PB formation occurs via the sequential accumulation of multivalent RNP complexes, often enriched with RBP constituents that contain intrinsically disordered regions that serve as essential scaffolds for PB maturation. Excitingly, recent lines of research have begun to uncover cellular factors that drive the regulation of PB assembly, fission and turnover via autophagy [36]. While numerous viral proteins have been shown to modulate PB dynamics, to date, there are very few examples of cellular proteins that actively regulate the formation and/or drive the disassembly of PB in human cells [3738].

Several studies have shown that the RNA composition of PBs often includes pro-inflammatory cytokine transcripts such as IL-6, TNF and IL-8 (CXCL8) [3339]. The finely regulated timing of the translation and turnover of these transcripts – especially when the cell is not under threat by pathogens – is critical to maintaining cellular homeostasis [3940]. This is exemplified by the fact that many of these same mRNAs bear RNA elements located in their 3′UTR, known as AU-rich elements (AREs), that can actively signal their sequestration into PBs [4143]. Thus, the loss of PBs has been hypothesized to be an anti-viral strategy, priming a robust defence for the cell [44]. However, to date, there are scarce examples of how the cell might induce PB loss to facilitate this process [16373845]. Complicating this model further is an increasing number of reports that viral infection by several DNA and RNA viruses induces PB loss via viral gene products that directly interfere with the scaffolding of PB-associated RNPs [41,44, 4648].

Given its significant role in the innate immune response, we hypothesized that SHFL’s impact on PB dynamics could serve as a host-driven defense mechanism during KSHV infection. Here, we show that SHFL directly blocks PB formation, even in the context of chemically induced oxidative stress. We also found that SHFL RNA-binding activity is not required for its induction of PB loss. Instead, we define the interaction of two amino acids, W191 and G259, within the middle 151 to 200 aa region and 190 to 241 aa region of SHFL as a new motif within SHFL critical for inhibiting PB formation. Reinforcing the connection between SHFL-mediated PB disruption and its anti-viral functions, we also show that SHFL-induced PB disruption induces the expression of inflammatory cytokines, which could further bolster SHFL’s anti-viral capacity. Lastly, we show that SHFL’s ability to disrupt PB is linked to its anti-KSHV activity, as mutants lacking the ability to disrupt PBs are unable to restrict KSHV lytic replication. Collectively, our results demonstrate that SHFL acts as a regulator of PB dynamics during herpesvirus infection, restricting their formation to promote the expression of pro-inflammatory cytokines and contribute to a global anti-viral state.

Methods

Cells and transfections

HEK293T cells (ATCC) were grown in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen) supplemented with 10% FBS. The KSHV-infected renal carcinoma human cell line iSLK.BAC16 (iSLK.WT) (a kind gift from Dr B. Glaunsinger) bearing doxycycline-inducible RTA (Replication and Transcription Activator) was grown in DMEM supplemented with 10% FBS. KSHV lytic reactivation was induced by the addition of 1 µg ml^−1^ doxycycline (BD Biosciences) and 1 mM sodium butyrate for 48 h as reported above. For DNA transfections, cells were plated and transfected after 24 h when 70% confluent using PolyJet (SignaGen). For small interfering RNA (siRNA) transfections, cells were reverse transfected in six-well plates by INTERFERin (Polyplus Transfection) with 10 µM siRNAs. SHFL-targeting siRNA was obtained from IDT as Dicer-substrate siRNA and targets the 3′UTR of endogenous SHFL (AGCUCAAACCGAGAUAUGAAUGACC). NaAs treatment was performed on cells grown on coverslips, transfected as above and treated with NaAs (0.25 mM) for 30 min and incubated at 37 °C. Following treatment, cells were harvested and fixed via 4% paraformaldehyde. Where indicated, cells were treated with purified TNF-α at a final concentration of 10 ng ml^−1^.

Plasmids

The C19ORF66 coding region was obtained as a gBlock from IDT and cloned into a pcDNA4 Nter-3xFLAG vector (FLAG-SHFL). A FLAG-mock (pcDNA4 Nter-3xFLAG) was used as a control where indicated. The SHFL mutant library was cloned by truncating 50 aa starting from the N-terminus. From the full-length SHFL, the FLAG-SHFLΔRBD mutant was constructed to lack amino acids 102–150 (FLAG-SHFLΔ102–150). The FLAG-SHFL Δ3R was ordered as a G-block from IDT with the three point mutations R131A, R133A and R136A and cloned into a pcDNA4 Nter-3xFLAG vector. The mutants FLAG-SHFLΔNES and FLAG-SHFLΔ151–200+NES were also ordered from IDT as G-blocks with the point mutations L261A, L264A, L267A and L269A and cloned into pcDNA4 Nter-3xFLAG vectors. FLAG-SHFLΔ151–200, FLAG-SHFLΔ151–200, FLAG-SHFLΔ241–291 and FLAG-SHFLΔW191A, G259A (ΔGW) were ordered as G-blocks from IDT and cloned into a pcDNA4 Nter-3xFLAG vector. G-blocks were cloned into the vector backbones via in-fusion cloning (Clonetech-takara); all final constructs were verified via Sanger Sequencing.

RT-PCR

Total RNA was harvested using TRIzol according to the manufacturer’s protocol. cDNAs were synthesized from 1 µg of total RNA using AMV reverse transcriptase (Promega) and used directly for quantitative PCR (qPCR) analysis with the SYBR green qPCR kit (Bio-Rad). Signals obtained by qPCR were normalized to 18S unless otherwise noted. qPCR primers used in the study are as follows: COX2 (Fwd: CCCTTGGGTGTCAAAGGTAA; Rv: GCCCTCGCTTATGATCTGTC); CXCL8 (Fwd: AAATCTGGCAACCCTAGTCTG; Rv: GTGAGGTAAGATGGTGGCTAAT); ATF3 (Fwd: CAAGAACGAGAAGCAGCATTTG; Rv: TTCTGGAGTCCTCCCATTCT); JUNB (Fwd: TCTACCACGACGACTCATACA; Rv: GGCTCGGTTTCAGGAGTTT).

Immunoblotting

Cell lysates were prepared in lysis buffer (NaCl, 150 mM; Tris, 50 mM; NP-40, 0.5%; DTT, 1 mM; and protease inhibitor tablets) and quantified by Bradford assay. Equivalent amounts of each sample were resolved by SDS-PAGE and immunoblotted with each respective antibody in TBST (Tris-buffered saline, 0.1% Tween 20). Antibodies used are mouse anti-Flag (Invitrogen, 1:1,000); rabbit anti-DDX6 (Invitrogen, 1:100) or mouse anti-DDX6 (Sigma, 1:100); and rabbit anti-vinculin (Invitrogen, 1:1,000). Primary antibody incubations were followed by HRP-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies (1:5,000; Southern Biotechnology).

Immunofluorescence

HEK293T or iSLK.219 cells were grown on coverslips and fixed in 4% formaldehyde for 20 min at room temperature. Cells were then permeabilized in 1% Triton X-100 and 0.1% sodium citrate in PBS for 10 min, saturated in BSA for 30 min and incubated with the designated antibodies. After 1 h, coverslips were washed in PBS and incubated with Alexa Fluor 680, 594 or 488 secondary antibodies at 1:1,500 (Invitrogen). Coverslips were washed again in PBS and mounted in DAPI-containing Vectashield mounting medium (Vector Labs) to stain cell nuclei before visualization by confocal microscopy on a Nikon A1 resonant scanning confocal microscope (A1R-SIMe). The microscopy data were gathered in the Light Microscopy Facility and Nikon Center of Excellence at the Institute for Applied Life Sciences, UMass Amherst, with support from the Massachusetts Life Sciences Center.

RNA granule quantification

PBs were quantified using an unbiased image analysis pipeline generated in the freeware CellProfiler (cellprofiler.org). Briefly, detection of nuclei in the DAPI channel image was performed by applying a binary threshold and executing primary object detection between 40 and 200 pixels. From each identified nuclear object, the ‘Propagation’ function analyses the 488 nm channel image (identifying FLAG-SHFL) to identify transfected cell borders. The pipeline then subtracts out the nuclei from the total area to define the cytoplasmic spaces. Using this cytoplasmic mask, the 594 nm channel image, with the fluorescing PB puncta (stained with DDX6), could be overlayed to only identify cytoplasmic foci (i.e. excluding any nuclear signals). The ‘Enhance Speckles’ function reduced background staining in the cytoplasmic channel. The ‘global thresholding with robust background adjustments’ function specifically defines and denotes certain puncta based upon strict size and intensity ranges to reduce the amount of background artefacts quantified. All sizes, intensity ranges and thresholds throughout the pipeline remained unchanged between experiments. Those puncta of specific size and intensity were quantified and used for data analysis.

Statistical analysis

All results are expressed as means±sem of experiments independently repeated at least three times. Statistical analyses for PB counts were performed using GraphPad Prism 10. Each cell PB count is displayed in dot plots, and each independent biological replicate is visualized on a single graph. Given the non-parametric nature of cell counts, statistical differences between experimental conditions were tested using the Wilcoxon rank test. Data were then summarized into bar graphs for clarity in the main figures. For qPCRs, an unpaired Student’s t-test was used to evaluate the statistical difference between samples. Significance was evaluated with P values as follows: *P<0.05; **P<0.01; ***P<0.001.

Results

SHLF blocks PB formation

We recently showed that during KSHV infection, upon expression of the anti-viral factor SHFL, PBs are lost [16] both within and outside the context of viral infection. Given the stringent reduction in PB numbers in cells overexpressing SHFL, we next sought to understand if SHFL blocks de novo formation of PB or triggers their disassembly. To test this, we treated cells with sodium arsenite (NaAs), a known oxidative stressor for cells that induces a marked PB increase (Figs 1b and S1, available in the online Supplementary Material) [49]. NaAs treatment was performed over a period of 30 min before fixing the cells and staining with the PB marker DDX6. Upon treatment with 0.25 mM NaAs, we observed an increase in PB numbers as expected (Fig. 1b). This treatment was then performed in cells overexpressing SHFL to address whether SHFL blocks de novo formation of PB, where we would expect to see no increase in PB numbers, or if SHFL triggers the disassembly of PB, where we would expect to still see a large increase in PB after 30 min following NaAs treatment. In line with our former hypothesis, we observed that when SHFL is overexpressed, NaAs-treated cells no longer display higher PB counts, suggesting that SHFL actively prevents PB formation.

*SHFL blocks de novo synthesis of PB. (a) HEK293T cells were transfected with either a Flag-tagged SHFL (or mock). Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). (b) HEK293T cells were transfected with either a Flag-tagged SHFL (or mock). Briefly, 24 h later, cells were treated for 30 min with 0.25 mM NaAs, then subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). Here and in all figures, the scale bar represents 10 µm. The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, **, P<0.01, **, P<0.001.

SHFL-mediated PB loss does not rely on SHFL RNA-binding activity

PBs are condensates composed of RNA and proteins. SHFL has previously been shown to bind to mRNA as a key feature of its anti-viral capacity, so we hypothesized that SHFL could impede PB formation by binding and sequestering RNA molecules away from PB, preventing their assembly/nucleation. To test this hypothesis, we constructed mutants to abrogate SHFL RNA-binding activity: one mutant that lacks amino acids 102–150 (ΔRBD), hypothesized to contain the SHFL RNA-binding domain and previously reported to be unable to bind to mRNA [50]. We also generated a second SHFL mutant, introducing point mutations at arginines 131, 133 and 136 (Δ3R), which have been implicated as the critical amino acids responsible for SHFL RNA binding in vitro. We first verified that these mutants expressed properly and then proceeded to quantify PB content in cells expressing each of them (Figs 2 and S1). We observed that, like WT SHFL, these mutants also drastically reduced PB count, suggesting that SHFL RNA-binding activity is not required for PB disruption.

*SHFL RNA-binding activity is not required for PB disruption. (a) HEK293T cells were transfected with either a full-length Flag-tagged SHFL or SHFL RNA-binding domain mutants (ΔRBD or Δ3R as indicated). Briefly, 24 h later, cells were harvested and subjected to immunoblot and stained with the indicated antibodies. (b) and (c) HEK293T cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, **, P<0.01, **, P<0.001.

SHFL bridge domain is responsible for PB loss

We next sought to identify the domain in SHFL that may trigger PB loss and thus constructed a library of SHFL mutants with incremental 50 aa deletions from the N-terminus. As shown in Figs 3 and S1, two domain deletion mutants appear to result in the restoration of PB numbers, suggesting that individually, these domains contribute to the disruption of PB. We therefore focused on these two domains and constructed deletion mutants for each domain (Δ151–200 and Δ241–291) individually or in combination. However, deleting both domains at once did not result in an additive effect on PB restoration (Fig. 4). Thus, we hypothesized that these domains could broadly affect SHFL subcellular localization. However, both single deletion mutant and double deletion mutant have cytoplasmic localization like WT SHFL (Fig. 4b). Given that these two domains are not contiguous, it was puzzling to understand how and why they could both impact SHFL-mediated PB restriction. To better understand this, we generated a structural model of WT SHFL in AlphaFold3 [51] (Figs 4d and S2), and we saw that these domains are predicted to be in proximity, creating a common interface that we hypothesized could come together as a potential binding platform. Between these domains are two amino acids (G259 and W191) that are predicted to form contact via a three angstrom H-bond (Figs 4d and S2). We therefore mutated these amino acids into alanine residues to abrogate the predicted H-bond formation (mutant referred to as ΔGW). Using this mutant, we demonstrate that SHFL’s ability to restrict PB formation in fact stems from this bridge formed by these two residues, further suggesting that the proximity of these domains in SHFL’s tertiary structure is critical for its function(s) and may serve as a binding platform for a specific interactor to prevent PB formation.

*Several domains of SHFL contribute to PB restriction. (a) Schematic representation of SHFL incremental deletion mutants. (b) and (c) Cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 independent biological replicates. n.s.=not significant, **, P<0.01, **, P<0.001.

*Domains 151–200 and 241–290 form an interface important necessary for PB restriction. (a) HEK293T cells were transfected with the indicated Flag-tagged SHFL constructs, and 24 h later, cells were harvested and subjected to immunoblot and stained with the indicated antibodies. (b) and (c) Cells were transfected with the indicated SHFL constructs. Briefly, 24 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6). The number of PB puncta per cell was quantified using the CellProfiler pipeline as described in the methods and normalized to the mock control within each replicate. (d) AlphaFold prediction of SHFL folding, highlighting closeness of domains 151–200 and 241–290, and amino acids G259 and W191 possible bond. G259A-W191A (ΔGW) mutant of SHFL was constructed and transfected into HEK293T cells, confirming that these mutations did not affect the stability of SHFL. Immunofluorescence of stained PB, however, revealed no loss of PB upon expression of this mutant. Statistics were determined using the non-parametric Wilcoxon rank test between control and experimental groups; error bars represent sem; n=3 or n=5 independent biological replicates as indicated. n.s.=not significant, **, P<0.01, **, P<0.001.

SHFL-mediated PB loss is required to restrict KSHV lytic replication

SHFL acts as a potent anti-viral factor during KSHV infection, and we hypothesized that its role in PB repression may further contribute to establishing an anti-viral state in KSHV-infected cells. PBs are often implicated as sites of translational arrest and RNA storage [3452]. The loss of PB upon SHFL expression could thus alter the availability of certain transcripts. AREs containing transcripts often encode for pro-inflammatory cytokines and can thus be tied to the inflammatory response and have been found to be modulated by PB abundance. We wanted to test if SHFL expression and subsequent PB loss would also result in alteration of ARE-mRNA levels. To test this, we next analysed the abundance of select ARE-mRNAs and found that several of these genes had increased transcript levels upon SHFL overexpression in uninfected cells (Fig. 5a). Intriguingly, as opposed to WT SHFL, SHFL ΔGW mutant did not result in the same increase in expression for these ARE-mRNAs, reinforcing the idea that their increase is tightly linked to PB loss. Given this global anti-viral potential due to SHFL-mediated PB restriction, we next wondered if SHFL’s capacity to restrict KSHV lytic replication could also be globally attributed to its ability to inhibit PB formation. To this end, we first wanted to confirm the contribution of endogenous SHFL expression to PB numbers during KSHV infection. Upon knockdown of SHFL in iSLK.219 cells, as previously observed [17], there is a marked increase in KSHV lytic replication as indicated by increased RFP fluorescence (Figs 5c and S1). This corresponded with a marked increase in PB numbers observed in both KSHV latent and lytic cells. Next, we measured the global effect of SHFL on KSHV lytic reactivation using a complementation approach in the KSHV.219 model (?Fig. 5b). We first verified that SHFL could be efficiently knocked down in these cells and that SHFL WT and ΔGW had similar levels of expression. We noted that SHFL expression (WT or mutant) is increased upon lytic reactivation. As expected, after knocking down endogenous SHFL levels, overexpression of WT SHFL led to a reduction in RFP-positive cells upon KSHV lytic reactivation, confirming SHFL’s anti-viral role during infection. However, SHFL ΔGW is unable to restrict PB formation and also does not restrict the lytic cycle (Figs 5d??? and S3). Our results therefore suggest that SHFL anti-KSHV activity at least partially results from its ability to modulate PB formation and thus to globally reshape RNA flux in KSHV-infected cells.

PB loss upon SHFL expression results in ARE-mRNA overexpression and KSHV lytic restriction. (a) Cells were transfected with SHFL WT or ΔGW (or mock vector) and treated with purified TNF-α (to induce ARE-mRNA expression) at a final concentration of 10 ng ml−1. Total RNA was then harvested and subjected to RT-qPCR to measure mRNA levels of the indicated ARE-containing mRNA. Statistical difference between control and experimental conditions was determined using one-way ANOVA followed by Dunnett’s multiple comparisons. ns=P>0.05, * = P<0.05, ** = P<0.01. (b) and (c) KSHV-positive iSLK.219 cells were either treated with siRNAs targeting SHFL (or control non-target siRNAs – siSCRB) for 48 h, then transfected with the indicated Flag-tagged SHFL. Cells were then reactivated with doxycycline and sodium butyrate for 48 h, then checked for reactivation efficiency by monitoring the expression of GFP (encoded on the KSHV genome and therefore a marker for infected cells) and RFP (under the control of the PAN promoter and therefore a marker for lytic reactivation). Quantification of RFP-positive cells is to the right. Values represent at least five independent views of the infected cells. Statistical difference between control and experimental conditions was determined using one-way ANOVA followed by Dunnett’s multiple comparisons. ns=P>0.05, * = P<0.05, ** = P<0.01. (d) iSLK.219 was left latent or reactivated and transfected with the indicated SHFL constructs. Briefly, 48 h later, cells were subjected to immunofluorescence assay and stained for the indicated PB marker (DDX6).

Discussion

The ability of the cell to regulate the fate of RNA in response to viral infection is crucial for its survival and adaptation. As such, both cell and virus have evolved a myriad of RBPs that clash for control of post-transcriptional surveillance of RNA trafficking, translation and stability. In this context, RNA granules have emerged as discrete sites of intense RNA regulation and thus as particularly important hijacking hubs during viral infection [53]. The condensation of RNA into PBs is especially impactful, as many of the RNA species that constitute this granule type have been tied to critical cellular processes such as the pro-inflammatory response. Indeed, several previous studies have shown an enrichment of ARE-bearing cytokine mRNA within PBs of several cell types. These observations imply that the cell may use PB formation as a means of preventing the premature translation of these transcripts or to carefully time their turnover in an ever-changing cellular environment. This also suggests that proper disassembly of PBs may correlate with a coordinated inflammatory cascade that would ultimately promote cell survival and signalling to the greater cell microenvironment. It is thus perhaps unsurprising that several viruses have evolved to block the formation of, or degrade, PBs. In this study, we focus on the gammaherpesvirus KSHV, an oncogenic virus that has been shown to restrict PBs early on following lytic reactivation from latency [41485455]. It should be noted that this restriction was also observed in KSHV latent cells, but this phenotype appears to vary between latent cell types. During lytic replication, this restriction is mediated by the early lytic gene ORF57, which blocks de novo PB assembly by interfering with interactions between critical PB scaffolding factors such as GW182 and Ago2 [48]. This suggests that KSHV and other viruses may actively disassemble PB to facilitate replication at the expense of triggering the host cell cytokine response. However, as is well understood for KSHV, this could contribute to establishing a pro-tumorigenic environment. Other viruses appear to also benefit from PB disruption, as was shown recently by Kleer et al., whereby dispersed PBs during coronavirus infection lead to an upregulation of cytokine mRNA stability and translation [44]. The authors propose that this may be a mechanism that supports a pro-viral dysregulation of the cytokine response and therein ultimately promotes coronavirus pathogenesis. Collectively, these reports highlight the need for a better understanding of how the host cell differentially regulates these RNP condensates to control gene expression dynamics during viral infection.

Here, we focus on the host protein SHFL, which we previously identified as a transcript that actively evades herpesvirus-induced RNA decay [1618]. This broad escape from viral endonuclease cleavage mirrors SHFL’s own broad anti-viral capacity, as SHFL can restrict the replication of multiple DNA, RNA and retroviruses through diverse mechanisms [18]. Upon further investigation, we found that following this escape from cleavage, the SHFL protein goes on to impede KSHV lytic reactivation from latency and broadly suppresses viral gene expression. This potent anti-viral capacity appears to stem from restricting the expression of several critical KSHV lytic genes, including the master regulator of the latent-to-lytic switch, KSHV ORF50, and the master regulator of KSHV RNA fate, KSHV ORF57 [16]. Intriguingly, we also found that SHFL expression restricts PB numbers both outside of and within the context of KSHV infection [12]. Little is known about cellular factors that have the capacity to disassemble PB [3738]. As such, here, we set out to investigate the molecular underpinnings of the influence of SHFL over PBs, hypothesizing that it may be tied to SHFL anti-viral response to KSHV lytic replication. To our surprise, we found that SHFL restricts PBs even under sodium arsenite stress, suggesting SHFL impedes PB assembly/nucleation rather than inducing disassembly. Given that SHFL is a known RBP and its RNA-dependent interaction with ORF57, we thought that the ability of SHFL to bind to RNA was required for this capacity to impede PB assembly, perhaps by blocking the incorporation of specific RNA species into these phase-separated condensates. To test this, we generated mutants of SHFL that have been previously shown to preclude SHFL RNA interactions. However, we found that the ability of SHFL to bind to RNA is not required for its ability to restrict PB formation. A recent report identified another host protein, Sbp1, as able to trigger PB disassembly through direct protein–protein interaction with a decapping enzyme [37]. This suggested that SHFL’s role in blocking PB formation can perhaps stem from its protein interaction network. It is important to note that while SHFL restricts de novo formation of PBs in the presence of NaAs, this does not rule out a possible SHFL-mediated increase in the autophagic flux of PBs or select proteasomal degradation of RBP constituents, as was observed in other viral contexts [56]. Supporting this, we have previously mapped the SHFL interaction network and found several E3 ligases [37]. It would be interesting to test if these interactions are needed for SHFL-induced PB restriction both within and outside the context of KSHV infection.

To better understand which domain of SHFL was required for restriction of PB formation, we next generated an SHFL deletion mutant library and tested its capacity to restrict PB numbers. From this screen, we identified two domains that individually contribute to the disruption of PB, including aa151–200 and the C-terminal end of SHFL aa241–291. Previous work by Suzuki et al. and others has demonstrated that there is a functional Nuclear Export Signal within the SHFL C-terminus [20]. Given the nearly exclusive cytoplasmic localization of SHFL across multiple cell types, we thought that perhaps the loss of these domains may obstruct proper SHFL subcellular localization. However, both the single domain and double domain deletion mutants still demonstrate cytoplasmic localization like WT SHFL. Furthermore, SHFL appears to be diffusely expressed in the cytoplasm, as opposed to a possible localization at the PB sites. This could suggest that SHFL contribution to blocking PB assembly does not occur from within the granule but rather by preventing the aggregation of the core granule components and sequestering key scaffolding proteins.

The need for two genetically separated domains to promote SHFL effect on PB was puzzling. Thus, we turned to predictive modelling of SHFL structure via application of AlphaFold3 to gain further structural insights in the absence of a current SHFL crystal structure. To our surprise, we found that these two regions appear to come into close proximity in the 3D structure and there is an amino acid pair (G259 and W191) that bridges these distinct SHFL domains via a predicted H-bond. Interestingly, this same bridge formed between these distinct domains of SHFL was also observed in the recent crystal structure of SHFL resolved by Li and colleagues (PDB 9KBO). In this crystal of SHFL, the same H-bond between W191 and G259 that was predicted by AlphaFold is retained, and the distance between these residues is even smaller (two angstroms), supporting our hypothesis that these residues must come together to stabilize an interface on SHFL. A single point mutagenesis of these residues resulted in an SHFL mutant unable to restrict PB formation. This result, combined with the deletion of individual domains surrounding these GW residues, points to a possible mechanism by which the proximity and stabilization of these two domains are necessary for PB restriction. Given our previous observation that SHFL protein interaction rather than its RNA-binding abilities is required for PB disruption, it is likely that these amino acids stabilize a protein-binding platform. By mutating these residues, SHFL may no longer be able to engage in specific protein–protein interactions, which, in turn, allow the PB to form as normal. Given that we have previously established the SHFL interaction network by mass spectrometry, it would be interesting to compare the interactome of WT SHFL to that of SHFL ΔGW.

Lastly, we wanted to determine the contribution of SHFL-mediated suppression of PB on KSHV lytic replication. PBs have been known to house pro-inflammatory mRNA transcripts that contain AREs, including IL-6 and CXCL8 [3339]. The release of these transcripts from PB has been tied to a direct enhancement of mRNA stability and translation. Concordantly, we found that SHFL expression does in fact lead to increased mRNA levels of multiple pro-inflammatory cytokines both outside the context of viral infection and during KSHV latency and lytic replication. This effect on these specific cytokine mRNA levels is likely post-transcriptional, as we did not detect any transcriptional increase by 4SU labelling of nascent RNA for these cytokines. This increase in global cytokine production would contribute to a global anti-viral state instated by SHFL expression. Thus, we next wanted to confirm that the influence of SHFL over PB contributes to its capacity to restrict KSHV lytic replication following reactivation from latency. In line with our previous findings, we found that SHFL ΔGW led to a marked defect in restricting KSHV lytic replication when compared to WT SHFL, mirroring the ability of SHFL to disrupt PB in infected cells. Of note, when checking the effect of SHFL WT or ΔGW on the late protein K8.1 (Fig. S3), we noticed an intriguing difference: SHFL expression leads to a stringent reduction of K8.1 overall expression, but particularly the higher molecular weight isoform of K8.1 detected. On the other end, SHFL ΔGW does not affect this isoform and instead seems to result in the reduction of the lower molecular weight isoform. K8.1 is known to be expressed as multiple isoforms resulting from a complex regulation of splicing variants during infection [5759]. It could thus suggest that these isoforms are differentially targeted into RNA granules and therefore differently affected by PB disruption. It would be interesting to investigate if this selectivity extends to other KSHV transcripts and whether this contributes to regulating the balance of expression of viral isoforms throughout the KSHV life cycle. All together, these results suggest that the ability of SHFL to restrict PB directly contributes to its ability, at least in part, to restrict KSHV lytic replication.

One critical factor of our investigation of SHFL and KSHV is the timing of SHFL expression. In our pilot study of SHFL, we found that the expression of SHFL climbs over the course of lytic replication [17]. We also noted here that transient expression of SHFL in lytic cells results in a marked increase in SHFL expression over levels observed in latently infected cells. Of course, here, this increased expression could be simply due to the addition of sodium butyrate to trigger reactivation, which could also influence expression of the plasmid. But it seems that high expression of SHFL protein is needed to modulate KSHV infection. This makes sense from the host perspective, as SHFL mRNA has evolved to escape KSHV-mediated RNA decay and therein contribute to restricting herpesviral replication. Therefore, our observation that SHFL expression climbs over the course of KSHV lytic replication may represent a gradual suppression of PB by the host to combat KSHV infection. However, alternatively, the expression of SHFL may be considered a mechanism promoted by KSHV to promote PB disassembly in synchrony with virus-encoded factors such as ORF57 and KapB. Thus, it appears that control of the timing of PB disassembly and how host factors such as SHFL or viral factors such as ORF57 and KapB may come together to dictate the ultimate outcomes of the inflammatory response to KSHV infection. For an oncogenic virus like KSHV, knowing more about the long-term consequences of this viral–host battle for the control of PB will be critical to better decipher pathways involved in both intracellular and paracrine signalling during tumourigenesis.

In summary, we identify SHFL as among the few cellular genes capable of restricting PB formation and that this influence over RNA granule dynamics contributes, at least in part, to its capacity to restrict KSHV infection. It remains to be understood what host factors SHFL needs to interact with to facilitate its influence over RNA flux in the cytoplasm. Our results highlight a tight interplay between host factors like SHFL and viral factors like ORF57 for controlling PB during KSHV infection. This relationship will undoubtedly shed light on how SHFL establishes a global anti-viral state. The work presented here highlights the efforts by the host cell to not only combat viral infection but also actively modulate the fate of specific mRNA through the shield of innate immune signalling. Furthermore, our findings underline a need to better understand how the host cell regulates RNA granule dynamics to mediate a versatile control over broader cellular responses to a diverse range of environmental challenges and cellular transformation.

Supplementary material

10.1099/jgv.0.002229Uncited Supplementary Material 1.