Systematic identification of NF90 target RNAs by iCLIP analysis

RNA-binding proteins (RBPs) interact with and determine the fate of many cellular RNAs directing numerous essential roles in cellular physiology. Nuclear Factor 90 (NF90) is an RBP encoded by the interleukin enhancer-binding factor 3 (ILF3) gene that has been found to influence RNA metabolism at several levels, including pre-RNA splicing, mRNA turnover, and translation. To systematically identify the RNAs that interact with NF90, we carried out iCLIP (individual-nucleotide resolution UV crosslinking and immunoprecipitation) analysis in the human embryonic fibroblast cell line HEK-293. Interestingly, many of the identified RNAs encoded proteins involved in the response to viral infection and RNA metabolism. We validated a subset of targets and investigated the impact of NF90 on their expression levels. Two of the top targets, IRF3 and IRF9 mRNAs, encode the proteins IRF3 and IRF9, crucial regulators of the interferon pathway involved in the SARS-CoV-2 immune response. Our results support a role for NF90 in modulating key genes implicated in the immune response and offer insight into the immunological response to the SARS-CoV-2 infection.

RNA-binding proteins (RBPs) robustly influence gene expression programs by modulating a range of posttranscriptional steps, including pre-mRNA splicing and processing, as well as mRNA export to the cytoplasm, mRNA turnover, and translation 1 . Consequently, RBPs are major factors in the response of cells to both intracellular and extracellular stimuli. Nuclear Factor 90 (NF90) is a multifunctional DNA-and RNA-binding protein found mainly in the nucleus and encoded by the interleukin enhancer-binding factor 3 (ILF3) gene. The role of NF90 as a DNA-binding transcription factor was first identified in 1994 by Crothesy and colleagues while studying the human transcription factor complex NFAT (nuclear factor of activated T cells), which is involved in the transcriptional regulation of the IL2 and IL13 genes 2,3 . Deeper characterization of NF90 as transcription factor revealed that it induces the proliferation of K562 erythroleukemia cells and regulates the expression of immediate early genes 4,5 , genes which are rapidly activated in response to several stimuli 6,7 .
As an RBP, NF90 possesses two double-stranded RNA-binding motifs (dsRBM), a zinc finger domain (ZNF), and a nucleic acid-binding arginine/glycine-rich (RGG) motif 8 , through which it binds single-and doublestranded RNA from viruses and mammalian cells 4 . In turn, NF90 modulates RNA metabolism by influencing mRNA splicing, export to the cytoplasm, turnover, and translation 5,[9][10][11] . NF90 regulates mRNA stability and translation efficiency by binding to the 3'-untranslated region (UTR) of target mRNAs, many of them encoding proteins involved in cell proliferation and in the immune response. NF90 was shown to bind a subset of RNAs containing AU-rich sequences in the 3'UTR, including IL2 mRNA during T cell activation and MKP-1 mRNA in response to oxidative damage 9,12 . NF90 stabilizes the mRNAs encoding the cyclin-dependent kinase inhibitor CDKN1A (p21) and the myogenic transcription factor MYOD in developing skeletal muscle 13 , while it promotes cell proliferation by binding the 3'UTR of CCNE1 mRNA, encoding cyclin E1, and stabilizing it, an effect that was further enhanced by the long noncoding (lnc) RNA LINC00470 14,15 . NF90 was found to suppress the translation of several mRNAs encoding proinflammatory factors; NF90 levels declined in senescent fibroblasts, leading to derepression and increased translation of several inflammatory cytokines 16 . Our studies found that NF90 regulates the stability of several cytokines following exposure to Plasmodium falciparum antigens 17 , repressed the translation of the cytokine BAFF in THP-1 cells, and cooperated with microRNA miR-15a in binding BAFF mRNA to inhibit translation 18,19 .

Results
Identification of NF90 target mRNAs. The collective of NF90 target RNAs in human HEK-293 cells was examined by overexpressing MYC-tagged NF90 (Fig. 1A) to isolate NF90 ribonucleoprotein (RNP) complexes. After crosslinking using ultraviolet light, a MYC-specific antibody was used to immunoprecipitate NF90 RNPs; the complexes were end-labeled using 32 P-ATP, size-separated by using SDS-containing polyacrylamide gels, transferred to a nitrocellulose membrane, and recovered from the membrane by digesting the protein with proteinase K. After reverse transcription of the RNA into cDNA, circularization, and linearization, the protected RNA was identified by high-throughput RNA-sequencing (RNA-seq) analysis (Fig. 1B).
The analysis revealed 2936 NF90 target transcripts; a partial list of NF90 target is presented in Fig. 1C, and a complete list in Supplementary Table S1. iCLIP signals were preferentially found within CDS (coding DNA sequence) regions and untranslated regions (UTRs) (50% in total) of protein-coding mRNAs (85%) (Fig. 1D,E). Analysis of the distribution of the signals throughout RNAs expressed from the different chromosomes (Chr.) revealed that most of the signals are present at level of the Chr. 1 and 19, suggesting that there is no apparent correlation between chromosome length and number of signals (R 2 = 0.02036, p-value = 0.516). Coverage profile of query regions at feature boundaries indicated that the signals distributed preferentially towards the end of the 5'UTR and the beginning of the 3'UTR (Supplemental Fig. S1A,B). Transcripts interacting specifically with NF90 were used to computationally identify the presence of conserved RNA signature motifs. Among the candidate motifs identified from our dataset, five motifs showed the highest frequency of hits (Supplemental Fig. S1C). In line with previous results 27 these motifs were highly enriched in A and U residues.
We next performed Reactome pathway (Reactome PA), Kyoto Encyclopedia of Genes and Genome (KEGG) pathway analyses and Gene Ontology (GO) to identify functional information from NF90 targets. Reactome PA revealed that the proteins encoded by NF90 target mRNAs were mainly implicated in biological processes related to RNA degradation, SARS-CoV-2 infection, and WNT signaling pathways. Similarly, enriched KEGG pathways included eukaryotic translation, nonsense-mediated RNA decay independent of exon junction complexes, and regulation of viral translation (Fig. 1F, Supplemental Fig. S1D). These data indicate that NF90 target RNAs are involved in important cellular processes including metabolism of mammalian and viral RNA. GO analysis of the iCLIP targets identified an enrichment in genes implicated in cell-cell interactions and in particular in neuronal interactions, as well as many ribosomal RNA-related functions (Cellular Component, Supplemental Fig. S2A). Regarding biological processes, we found several target RNAs implicated in RNA and viral metabolism, and several genes implicated in neural and epithelial tube formation pathways (Supplemental Fig. S2B).
Validation of endogenous NF90 target transcripts. We validated a subset of NF90 target RNAs identified by iCLIP analysis by performing RNP IP (RIP) analysis using native (non-denaturing) conditions that preserved the binding of endogenous RNAs to NF90, and measuring their enrichment in NF90 RIP samples by reverse transcription (RT) followed by quantitative (q)PCR analysis 28 . NF90 RNP complexes were isolated from HEK-293 lysates using a specific antibody directed at NF90 (Fig. 2A); in parallel, control IP reactions were carried out using pre-immune IgG. Western blot analysis was used to detect NF90 in the anti-NF90 IP samples (Fig. 2B). Purified RNAs were then measured by RT-qPCR analysis employing specific primer pairs as previously described 28 . The vast majority of transcripts tested by RIP analysis confirmed the iCLIP data; some transcripts, such as ZFR, LARP4, ADAM10 and ATP5B mRNAs, showed particularly high enrichment levels in Figure 1. Identification of NF90 target mRNAs in HEK-293 cells by iCLIP analysis. (A) Western blot analysis of MYC-NF90 in HEK-293 cells overexpressing a wild-type, full-length, MYC-tagged human NF90. The levels of ACTB (β-Actin) were assessed to monitor even loading of samples. (B) Schematic representation of the iCLIP methodology. Cells transfected with MYC-NF90 were initially crosslinked with 254 nm UV before being lysed. The MYC-NF90 RNP complexes were immunoprecipitated and visualized after 32 P-end-labeling of the RNA, followed by size-separation through SDS-containing polyacrylamide gels and transfer to nitrocellulose membranes. After the MYC-NF90 complexes were cut from the membrane and digested with Proteinase K, the RNA was recovered from the membrane and reverse transcribed into cDNA. After circularization, followed by linearization and PCR amplification, the RNA fragments were identified by RNA-sequencing. (C) Most abundant NF90 RNA targets identified by iCLIP analysis; p-value of the association, chromosome (Chr) from where the RNA is transcribed and type of RNA are indicated. (D) Percentage of all NF90 iCLIP reads mapping to RNA subdivided by transcript type/region. (E) Percentage of RNA type identified by our analysis. (F) KEGG pathway enrichment analysis of the NF90-target mRNAs identified by iCLIP. www.nature.com/scientificreports/ NF90 IP as compared to IgG IP samples, as measured by RT-qPCR analysis (Fig. 2C). A small percentage of the targets identified with the iCLIP, such as PIN1 and EPS15L1 mRNAs, were not validated by RIP analysis; in general, these may represent target transcripts in which NF90 binds introns or mRNA isoforms that were not amplified by the primers selected. Binding of NF90 to ACTB mRNA, which encodes a housekeeping protein, was measured in order to normalize data input, since ACTB mRNA is not a target of NF90 and thus the ACTB mRNA detected represented the background RNA binding in a nonspecific manner to beads, plastic, and other components of the IP reactions. A small number of lncRNAs was also analyzed by RIP analysis (Supplemental Fig. S3A). In particular, we observed a strong interaction of NF90 with the lncRNA XIST, a key lncRNA involved in the regulation of X chromosome inactivation 29 . Interestingly, in cells expressing low levels of NF90, all the lncRNAs analyzed were downregulated (Supplemental Fig. S3B). Due to the relevance of XIST we decided to show a representative genome browser snapshots of the iCLIP peaks (Supplemental Fig. S3C). We further confirmed the interactions of NF90 with target RNAs using biochemical methods. Five NF90 target mRNAs were biotinylated using in vitro transcription methods, and their interactions with NF90 were monitored by biotin-pull down analysis. Specifically, we analyzed the binding of NF90 with partial biotinylated RNAs corresponding to the Zinc Finger RNA Binding Protein (ZFR) and proteasome 20S subunit alpha 3 (PSMA3) mRNAs, which showed strong interaction with NF90 by RIP analysis, the Tyrosine 3-Monooxygenase/ Tryptophan 5-Monooxygenase Activation Protein Eta (YWHAH) mRNA, which showed weaker interaction, and two transcripts encoding Interferon Regulatory Factors (IRFs), which are associated with SARS-CoV-2 immune response, IRF3 and IRF9 mRNAs ( Fig. 3A-E). Biotinylated transcripts spanning the 3'UTR, CDS and 5'UTR of these mRNAs were synthesized in vitro and subsequently incubated with HEK-293 cell lysates. The biotinylated RNA-protein complexes were pulled down using streptavidin-coated beads and the presence of NF90 in association with the biotinylated RNA was analyzed using western blot analysis. As shown (Fig. 3), binding was validated for all the biotinylated RNAs tested (ZFR, PMSA3, YWHAH, IRF3 and IRF9 mRNAs). ZFR, PMSA3, YWHAH mRNAs showed stronger binding at the 3'UTR RNAs, while for IRF3 and IRF9 mRNAs, preferential binding was seen with the 5'UTR RNAs. Biotinylated RNA fragments from coding regions of PSMA3 and IRF9 mRNAs were also found to bind NF90. A biotinylated fragment synthesized to contain a region of BAFF 3'UTR mRNA, which we previously demonstrated not to bind with NF90 30 , was included as a negative control (Neg) (Fig. 3A-E).

NF90 regulates mRNAs and proteins involved in SARS-CoV-2 immune response.
To investigate the influence of NF90 on the fate of target RNAs, we first assessed the steady-state levels of mature target mRNAs after NF90 silencing. NF90 was silenced using small interfering (si)RNA specifically targeting ILF3 mRNA (which encodes NF90) and 48 h later, total RNA was collected, and the levels of target mRNAs were The levels of target mRNAs in the ribonucleoprotein (RNP) complexes present in the IgG IP and the NF90 IP samples were analysed by RT-qPCR and the levels of enrichment of the mRNAs in NF90 IP relative to IgG IP were calculated after normalization to ACTB mRNA levels in each IP group. Data are the means and standard deviation (+ SD) from at least three independent experiments. *, P < 0.05; **, P < 0.01, ***, P < 0.005.  www.nature.com/scientificreports/ measured by RT-qPCR analysis. NF90 downregulation did not dramatically influence the steady-state levels of most mRNAs analyzed ( Fig. 4A), despite robust silencing (Fig. 4B). The most significant effects were observed for IRF3 and IRF9 mRNAs. These mRNAs were strongly downregulated in NF90-silenced samples, while RGS14, TWIST1 and CLEC16A mRNAs were modestly elevated. The impact of NF90 on total RNA levels was tested globally by RNA sequencing (RNA-seq) analysis of HEK-293 cells in which NF90 was left unchanged or was silenced, as shown in Fig. 4 (Supplemental Fig. S4A). Comparison of total RNA levels in cells expressing low levels of NF90 (NF90 siRNA) to cells expressing normal levels of NF90 (Ctrl siRNA) identified 854 genes differentially expressed, 265 significantly upregulated and 589 significantly downregulated (Supplemental Fig. S4B; Supplementary Table S2 for complete list). For the RNAs most differentially abundant, relative (fold) changes were calculated by comparing RNA-seq results from NF90 siRNA cells with Ctrl siRNA (Supplemental Fig. S4C,D). In line with our iCLIP results, among the most differentially expressed mRNAs were those encoding eukaryotic and viral translation factors, as assessed by KEGG pathway analysis (Supplemental Fig. S4E). However, comparison of the mRNA targets identified by iCLIP and those found by RNA-seq analysis revealed only a small overlap of 3.7% (134 genes) between the two RNA sets, supporting a substantial role for NF90 in regulating translational events as an RNA-binding protein in HEK-293 cells (Supplemental Fig. S4F). Interestingly, among the 134 genes overlapping between the two analysis we observed a trend towards more downregulated genes, 99 (73%) out of 134. This trend suggests that NF90 positively regulates the levels of many target RNAs.

Scientific Reports
To study if the impact of NF90 on target mRNAs was associated with changes in the levels of encoded proteins, we performed Western blot analysis of several target mRNAs and quantified them using ImageJ (Fig. 4B, left and right panels). Interestingly, a greater magnitude of effect was seen at the protein level ( Fig. 4B) than at the mRNA level (Fig. 4A). Furthermore, most proteins showed reduced abundance after NF90 silencing, with the exception of YWHAH, which appeared slightly elevated. Overall, these results support a model whereby in basal conditions, NF90 binds to many mRNAs in HEK-293 cells and does not substantially affect their abundance, and instead appears to lower the steady-state abundance of proteins encoded by the target mRNAs. showed in Fig. 4 demonstrated that the silencing of NF90 did not significantly modify the steady-state levels of the endogenous target mRNAs analyzed (Fig. 4A). To directly test if NF90 controls the stability of the selected mRNAs we measured the half-lives of a subset of NF90 target mRNAs by treating cells with actinomycin D (an inhibitor of RNA polymerase II, used here to block de novo transcription) and measuring the rate of mRNA decay. In cells transfected with Ctrl or NF90 siRNAs actinomycin D was added 24 h later, total RNA was collected at the times shown, and mRNA levels were measured to evaluate the rates of decline (Fig. 5). Although the rates of decline were not totally equal, and silencing NF90 seemed to render some mRNAs slightly more stable (Fig. 5), the differences were quite minor at steady-state conditions. The loss of unstable MYC mRNA was included in the analysis as a positive control for the effect of actinomycin D 31 , while a stable transcript, GAPDH mRNA, was used as a negative control. We then turned our attention to the possibility that NF90 might influence the translation of some target mRNAs. We specifically assessed whether the association of the mRNA with polysomes might be influenced by NF90 levels (Fig. 6). Cells expressing control or silenced NF90 levels (prepared as described in Fig. 4) were collected and the cytoplasmic components were fractionated by centrifugation through linear sucrose gradients (10-50%), whereupon fractions were collected, and RNA was isolated from each fraction. Fractions 1 and 2 contained free RNA, not associated with ribosomal components; fractions 3-5 contained ribosomal subunits (40S, 60S) and monosomes (80S); and the remainder contained low-molecular-weight (LMW, fractions 6-8) and high molecular-weight (HMW, fractions 9-12) polysomes. As shown, silencing NF90 did not change the global distribution profiles (Fig. 6A). In this analysis, a shift in the direction of larger polysomes is interpreted to reflect more active translation for a given mRNA, while a shift to smaller polysomes is associated with reduced translation.

Figure 5.
Influence of NF90 on the stability of target mRNAs. HEK-293 cells that were processed as described in Fig. 4 to express either normal (Ctrl siRNA) or reduced (NF90 siRNA) levels were used to assess the relative rates of clearance of NF90 target mRNAs by RT-qPCR analysis after treatment with actinomycin D for the indicated times. GAPDH mRNA, a stable transcript, and MYC mRNA, an unstable transcript, were included as controls. The levels of mRNAs were first normalized to 18S mRNA levels and then plotted using GraphPad Prism on semi-logarithmic scales relative to 100%, the abundance of each mRNA before adding actinomycin D. Data are the means and standard deviation (+ SD) from at least three independent experiments. *, P < 0.05; **, P < 0.01, ***, P < 0.005. www.nature.com/scientificreports/ We then performed RT-qPCR analysis of the same subset of mRNAs in Fig. 5. We included analysis of the distribution of UBC mRNA, encoding the housekeeping protein Ubiquitin C, as its association with polysomes does not change in an NF90-dependent manner. The relative distribution of a specific target mRNA in each fraction was plotted as the relative distribution (%) of that mRNA in each fraction calculated over the total levels of that mRNA in the entire gradient. For the majority of the mRNAs tested (Fig. 6B), cells with lower NF90 levels showed reduced presence in polysomal fractions. For example, for IRF3 and IRF9 mRNAs, the peaks of abundance (fractions 8 and 9) were reduced after NF90 silencing. Similar shifts towards lighter fractions (less translation) were observed for other mRNAs after silencing NF90, including RBM6, PSMA3 and ZFR mRNAs. These data suggest that NF90 promotes the translation of several NF90 target mRNAs in unstimulated HEK-293 cells.

NF90 is downregulated in PBMC of patients affected by SARS-CoV-2 infection.
Considering that some of the identified NF90 targets encode proteins with direct roles in fighting the SARS-CoV-2 infection, www.nature.com/scientificreports/ we further explored the binding of NF90 to IRF3 and IRF9 mRNAs in immune cells and then studied the impact of NF90 on their expression levels in peripheral blood mononuclear cells (PBMCs) purified from a small heterogenous group of patients infected with the SARS-CoV-2 virus. We first assessed the binding of NF90 to IRF3 and IRF9 mRNAs by RIP analysis using Jurkat cells (an immortalized T lymphocyte line), which are known to express high levels of both IRF3 and IRF9 32 . NF90 RIP analysis in Jurkat cells was performed as described in Fig. 2; the levels of NF90 protein present in the IP samples were examined by western blot analysis (Fig. 7A, upper), and purified RNAs were analyzed by RT-qPCR analysis employing specific primer pairs as described above and in 28 . As shown, IRF3 and IRF9 mRNAs were highly enriched in NF90 IP as compared to IgG IP samples (Fig. 7A). To extend this analysis to human primary cells, we then performed RIP analysis using purified PBMCs from healthy human donors (Materials and Methods). Similar to Jurkat and HEK-293 cells, NF90 also associated with target IRF3 and IRF9 target mRNAs in human PBMCs (Fig. 7B).
Finally, we analyzed the levels of NF90, IRF3, and IRF9 in a small subset of patients infected with the SARS-CoV-2 virus as compared with healthy controls. Western blot analysis followed by quantification by densitometry www.nature.com/scientificreports/ using ImageJ revealed drastically reduced levels of NF90 in PBMCs purified from SARS-CoV-2-infected patients (PZ) (Fig. 7C), and also strongly reduced levels of the two type-I interferon response proteins IRF3 and IRF9 levels in the same patients (Fig. 7C). These data indicate that future studies are warranted to analyze the relevance of NF90 in the immune response against SARS-CoV-2 infection and, hopefully to identify new therapeutic strategies in combatting this virus.

Discussion
In this study, we report a comprehensive analysis of RNA targets of NF90 in HEK-293 cells using iCLIP. We found that many NF90 target mRNAs encode proteins with roles in eukaryotic and viral RNA metabolism (Fig. 1), including IRF3 and IRF9, which are directly implicated in the response to SARS-CoV-2 infection. In unstimulated HEK-293 cells, NF90 influenced only slightly the stability of some target mRNAs, but appeared to have a clearer effect on translation, as silencing NF90 decreased the size of polysomes associated with many NF90 target mRNAs and lowered the levels of the encoded proteins. These findings expand our previous knowledge of the functions of NF90 on target RNAs 13,17-19 .
One of the few NF90 targets that appeared to be degraded in an NF90-dependent manner was YWHAH mRNA, which showed increased stability after silencing NF90 (Fig. 5). YWHAH belongs to the 14-3-3 protein family and has been implicated in a variety of neurological disorders including Parkinson and Creutzfeldt-Jakob disease 33 . A deeper analysis to understand the involvement of NF90 and YWHAH in neurodegeneration is warranted. The small but significant change in the stability of IRF9 mRNA also deserves further attention, particularly to elucidate the molecular mechanisms mediating this influence. Even more importantly, future studies should determine whether the influence of NF90 on the stability of target mRNAs changes in response to stimuli that affect the immune and nervous systems.
For other target mRNAs examined, particularly IRF3, IRF9, RBM6, and ZFR mRNAs (Fig. 6), NF90 modified their association with polysomes, an indirect measure of their translation efficiency. Here too, the precise mechanistic effect needs to be analyzed in future studies. As with other RBPs, changes in the translation of NF90 target mRNAs could be mediated through either cooperation or competition with RBPs and/or microRNAs that can influence their translation. An example of cooperation of NF90 with miR-15a leading to the repression of BAFF mRNA translation was recently reported 30 . Identifying the specific factors that modulate translation of NF90 target mRNAs found in this study will need to be undertaken individually for each mRNA and its cellular context, as the collection of mRNA-binding factors (proteins and RNAs) associating with a given mRNA varies depending on the mRNA sequence, the cell type, and the growth conditions. IRF3 and IRF9 deserve particular attention, as these transcription factors are directly involved in the regulation of the type I interferon (IFN) response against viral infection 34,35 . Interestingly, it was recently reported that SARS-CoV-2 infected patients produce aberrant levels of type I IFNs and alterations in the function of IRF proteins 25 . Indeed, several SARS-CoV-2 proteins, including proteins of the viral non-structural proteins (nsp) and the nucleocapsid (N) protein, have been reported to inhibit type I IFN activation and the corresponding responses through regulation of IRF proteins 36 . For example, Wang and colleagues found that the SARS-CoV-2 nsp12 protein attenuates type I IFN induction by inhibiting IRF3 nuclear translocation 37 . Given the potential relevance to the COVID-19 pandemic, we sought to confirm the association of NF90 with IRF3 and IRF9 mRNAs in PBMCs and the steady-state levels of IRF3, IRF9 and NF90 in PBMCs purified from SARS-CoV-2-positive patients and healthy controls. Although in PBMCs and immortalized lymphocytes NF90 associated with IRF3 and IRF9 mRNAs, there was a striking loss of NF90 in PBMCs from SARS-CoV-2-positive individuals associated with a decline in IRF3 and IRF9 levels (Fig. 7). Interestingly, this regulation appears to vary across tissues; for example, RNA samples extracted from nasal swabs revealed that persons with SARS-CoV-2 had significantly elevated IRF3 and IFN relative to uninfected patients 38,39 . The role of NF90 in establishing a type I IFN program upon dsRNA stimulation was described by Watson and colleagues in cervical cancer cells 40 ; the authors found that NF90 modulated gene expression programs after dsRNA stimulation, and observed that IRF1 was less induced when NF90 levels were low 40 , in line with what we observed for IRF3 and IRF9. Taken together, our findings underscore the relevance of future studies to systematically analyze the impact of NF90 on IFN and IRF proteins, which are deeply implicated in the devastating COVID-19 pandemic.
The role of NF90 in the regulation of genes involved in the immune response was previously demonstrated in other different contexts. For example, NF90 was shown to bind a subset of RNAs containing AU-rich elements in the 3'UTR, including IL2 mRNA during T cell activation 9 . We also reported the role of NF90 in modulating several cytokines in response to Plasmodium falciparum antigens 17 and the role of NF90 as a repressor of the translation of the cytokine BAFF in THP-1 cells 18 . Additionally, we confirmed that NF90 binds A-and U-rich sequences present in target mRNAs as previously reported 27 (Fig. S1). Other RBPs, including AUF1 and HuR also preferentially bind AU-rich target RNAs 41 ; thus, it is likely that other RBPs compete or cooperate with NF90 to define the post-transcriptional outcome of specific transcripts.
Besides binding and regulating mammalian mRNAs, NF90 and its isoforms also bind viral genomes (RNA and DNA). Depending on the virus, NF90 can act as either a negative or a positive regulator of the expression of viral genes 42 . For example, NF90 had antiviral activity by antagonizing the inhibitory role of the nonstructural protein 1 (NS1) on PKR phosphorylation during influenza A infection 43 .
In closing, using comprehensive, high-throughput approaches, we have identified numerous NF90 target mRNAs, and found that the encoded proteins are predicted to influence mammalian and viral RNA metabolism. We confirmed earlier observations that NF90 can influence mRNA turnover and translation, and characterized in molecular detail how NF90 affects the stability and translation of IRF3 and IRF9 mRNAs, two key factors in the IFN response. Our findings set the stage for a deeper investigation of the impact of NF90 on the immune response during SARS-CoV-2 infection, and a search for therapeutic modulators of NF90 function. RNA isolation, reverse transcription (RT)-quantitative (q)PCR analysis. Total RNA was isolated using the TRIzol isolation reagent (Thermo Fisher) following the manufacturer's protocol and RNA integrity was verified by agarose gel electrophoresis. First-strand cDNA synthesis was performed using Maxima reverse transcriptase (Thermo Fisher) and random hexamers. The generated cDNA was diluted tenfold and used as a template for quantitative (q)PCR analysis using SYBR Green mix (Kapa Biosystems). Relative levels of RNA were calculated using the 2 −ΔΔCt method and the levels of β-actin (ACTB) mRNA were used for normalization. Genespecific primer pairs are listed in Table 1. RNA was isolated from PBMCs using the M-PER Mammalian Protein Extraction Reagent (Invitrogen Life Technologies).
RNA sequencing analysis. RNA sequencing (RNA-seq) and RNA-seq analysis were performed by Quick Biology. Briefly, RNA integrity was first assessed by using the Agilent Bioanalyzer 2100. Only samples with clean rRNA profiles were used for subsequent steps. Libraries for RNA-seq analysis were prepared according to KAPA Stranded RNA-Seq Kit with RiboErase (KAPA Biosystems) system. Paired-end sequencing was performed using Illumina HighSeq 4000 instrument (Illumina Inc.).

RIP (ribonucleoprotein immunoprecipitation) analysis.
RIP analysis was performed as previously described 28 . Briefly, PEB buffer (10 mM Hepes, 100 mM KCl, 5 mM MgCl 2 , 25 mM EDTA, 0.5% IGEPAL, 2 mM DTT, 50 U/ml RNase out and protease inhibitors) was used to prepare lysates of HEK-293 cells and M-PER Mammalian Protein Extraction Reagent (Thermo Scientific) was used to prepare lysates of PBMCs. After preclearing (10 min on ice followed by 15 min centrifugation), 500 μg of lysates were incubated (4 h, 4 °C) with a suspension of protein-A Sepharose beads (GE Healthcare) precoated overnight with 5 μg of a NF90 specific antibody (BD Transduction Laboratories) or normal mouse IgG (Santa Cruz Biotechnology) used as a control. Subsequently, the beads were washed with NT2 buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 0.05% IGEPAL) and then treated with RNase-free DNase I (15 min, 30 °C). RNAs that associated with this NF90 were then isolated using TRIzol and their levels of enrichment were assessed by RT-qPCR analysis using the primers listed in Table 1. Target mRNA levels were first normalized to ACTB mRNA levels, and then enrichments were calculated as the relative abundance of a target mRNA in NF90 IP as compared to IgG control IP. Analysis of polysome gradients. To identify mRNAs engaged in translation, cells were incubated with 0.1 mg/ml cycloheximide for 10 min and lysates were prepared from HEK-293 cells (Ctrl or NF90 siRNA). Cytoplasmic extracts were then loaded on top of a linear sucrose gradient [10-50% (w/v)] and size-separated by ultracentrifugation as previously reported 44 . Fractions were collected using a density gradient fractionation system and monitored by optical density measurement (A 254 ) (Brandel). The RNA in each fraction was isolated using TRIzol reagent and reverse-transcribed; the relative levels of a given mRNA in a specific gradient fraction were calculated using qPCR analysis with the primers listed (Table 1) and represented as a percentage of the total mRNA on the gradient.

Synthesis of biotinylated fragments and pull-down assay.
To synthesize biotinylated RNA, genespecific forward PCR primers containing the T7 RNA polymerase promoter sequence were used. After purification of the PCR products, the following biotin-RNAs (bi-RNA) were synthetized using MaxiScript rT7 kit (Ambion): bi-ZFR, bi-YHAW, bi-PSMA3, bi-IRF3, and bi-IRF9. A fragment of BAFF 3'UTR RNA that was previously demonstrated not to bind with NF90 was used as negative control 30 . Bi-RNA pulldown assay was performed as previously described 45   For library preparation and visualization, the RNA was dephosphorylated, an adapter was ligated at the 3' end and it was radioactively labeled at the 5' end with γ-32 P-ATP. The protein-RNA complexes were separated by SDS-PAGE and then isolated from a nitrocellulose membrane according to the expected size; protein digestion was performed using proteinase K and RNA was isolated. RNA fragments were reverse-transcribed using  Biot-YWHAH-2  FW: AGC GAC ATG GGG GAC CGG GAGCA   RV: TCA GTT GCC TTC TCC TGC TT   Biot-YWHAH-3  FW: AAC TGA AGA TCC TTC AGG TC   RV: CAT TTT TGC ATG CAC AAG   www.nature.com/scientificreports/ primers that introduce two cleavable adapter regions and barcode sequences. Circularization of the cDNA was carried out followed by linearization using the restriction enzyme BamHI. Linearization generated appropriate templates for PCR amplification optimized to avoid the generation of secondary products. Libraries were finally sequenced using high-throughput RNA-seq analysis.
iCLIP pipeline. Raw reads were obtained using bcl2fastq from the NextSeq500 run. Adapter and low-quality bases (< Q30) were trimmed using cutadapt 47 and sickle 48 . More than 75 million reads were retained. The overlapping forward and reverse reads were merged using Pandaseq to generate longer and complete sequences 49 . The randomized 5-bp molecular barcodes were extracted using UMI-tools 50 and the reads were demultiplexed using reaper 51 . A total of 2,518,715 reads were obtained and demultiplexed reads were then mapped to hg19 using STAR aligner 52 . In more details, the demultiplexed reads were processed following the eCLIP ENCODE pipeline described in https:// www. encod eproj ect. org/ pipel ines/ ENCPL 357ADL/: (i) reads were mapped to human specific version of RepBase (accessed 18 set 2018) with the STAR alignment tool to remove repetitive elements; (ii) unmapped reads from the first STAR run were than sorted and aligned to the human genome (hg19 with Gencode V19 annotations); (iii) mapped reads were then deduplicated (to remove PCR duplicates) with the command umi_tools dedup. Finally, (iv) output BAM files were then merged into an unique indexed and sorted BAM file. Identification of Crosslinked sites was performed with iCount, a Python module (Python software 4.10.2), and associated command-line interface (CLI), which provides all the commands needed to process protein-RNA iCLIP interaction data and to identify and quantify sites of protein-RNA interactions on RNA 53 . Binding signals were also confirmed with the Clipper tool 54 . Mean coverage of iCLIP data was calculated from different genomic regions using only signal with significant p-value (< 0.05). Motif analysis was performed using HOMER (Hypergeometric Optimization of Motif EnRichment) a suite of tools for Motif Discovery (HOMER software v4.11.) 55 . Pathway enrichment analysis was conducted with the R library (R software 4.1.2) cluster Profiler [https:// bioco nduct or. org/ packa ges/ relea se/ bioc/ html/ clust erPro filer. html]. Given a vector of genes, the function enrichGO returns the enrichment GO categories after FDR control.

PBMC purification. Peripheral blood samples were collected from adult donors positive for COVID-19
infection and adult healthy donors. All of the human samples used in this study were previously anonymized as required by the Italian Data Protection Code (Legislative Decree 196/2003) and the general authorizations issued by the Data Protection Authority. Ethics Committee approval was deemed unnecessary because, under Italian law, it is only required in the case of prospective clinical trials (Art. 6 and Art. 3 of Legislative Decree 211/2003). However, all of the patients gave their written informed consent to the medical procedures/interventions carried out for routine treatment purposes. Peripheral blood mononuclear cells (PBMCs) were purified from a small heterogenous group of patients infected with the SARS-CoV-2 virus and healthy human donors using Histopaque 1077. About 15 ml of blood was collected from each donor using BD Vacutainer CPT tubes containing 0.1 M sodium citrate and PBMCs were isolated following the manufacturer's instructions. The isolated PBMCs were then used for Western Blot and RIP analysis.