NF90 interacts with proteins involved in translational repression and RNA processing

In order to understand the role of NF90 in the cytoplasm, we determined its cytoplasmic interactome by performing mass spectrometry on a HEK293T cell line stably overexpressing NF90-FLAG-HA (Additional file 1: Fig. S1A), after tandem affinity purification of the cytoplasmic protein fraction (Additional file 1: Fig. S1B). Excluding the proteins for which one peptide or more was found in the mock HEK293T sample, 318 proteins were detected associated with NF90 in the cytoplasm (Additional file 8: Table S4). Of these, 209 were detected with 3 or more peptides and FC>2 compared to the mock sample (Additional file 8: Table S4). As expected, the most abundant protein identified was NF90 (ILF3). ILF2 (NF45), a well-known NF90 protein partner, was found among the most abundant interactants. Interestingly, several proteins associated with RISC were identified among the interactants. Notably, the 5′-to-3′ helicase MOV10 was highly represented. Gene ontology of the significantly enriched NF90 interactants (3 or more peptides, FC>2) identified a number of pathways, such as translation, translational initiation, mRNA processing, and regulation of mRNA stability (Fig. 1A). These findings are consistent with previous observations suggesting the implication of NF90 in the regulation of mRNA stability and mRNA translation for specific target mRNAs [6, 16].

Fig. 1

NF90 interacts with proteins involved in translational repression and RNA processing. A Representative gene ontology pathways (biological pathways) of NF90-associated proteins (n = 209; 3 or more peptides, FC>2) identified by mass spectrometry. B Cytoplasmic extracts from HEK-293T cells were used for immunoprecipitation using an anti-IgG control or anti-NF90 antibody followed by RNAse A/T treatment or mock treatment, as indicated. Immunoprecipitates and an aliquot of extract (In) were analyzed by western blot, using the indicated antibodies. C FLAG immunoprecipitates of cytoplasmic extracts of NF90-FH overexpressing HEK293T cells were separated by glycerol gradient sedimentation. NF90-containing fractions were analyzed by western blot using the antibodies indicated

Since NF90 interactants are well-known RNA-binding proteins (Additional file 1: Fig. S1C), we wondered if their interaction with NF90 was RNA-dependent. In order to validate the results of mass spectrometry using endogenous proteins and to determine if the binding of NF90 is RNA dependent, we performed co-IP in HEK293T cell line, with and without RNAse A/T treatment. The binding of NF90 to NF45 was found to be RNA-independent (Fig. 1B), as previously reported [27, 28]. The results confirmed the binding of endogenous NF90 to PACT, MOV10, and Dicer and furthermore showed that these interactions are RNA-dependent (Fig. 1B). Since Ago2, the main component of RISC, was also found among NF90 interactants, albeit below the threshold for interactants, we tested its interaction with NF90. Figure 1B shows that Ago2 was associated with NF90, as described previously [19]. Although previously reported to be RNA-independent [19], under the conditions used in this study, NF90 binding to Ago2 was RNA-dependent, in agreement with a recent study [16]. These findings identify an RNA-mediated interaction between NF90 and RISC-silencing complex in the cytoplasm.

To determine whether RISC subunits can be found in the same complex with NF90, we performed glycerol gradient sedimentation of immunopurified NF90-Flag-HA complexes (Additional file 1: Fig. S1D and E). NF90 was detected in the top 10 fractions (data not shown) where it co-sedimented with MOV10, AGO2, PACT, DDX6, and NF45 (Fig. 1C). In agreement with these results, NF90 has been shown to co-sediment with Ago2, NF45 and other helicases, such as DDX47, DDX36, and DDX30, on a sucrose gradient [19]. These findings suggest that cytoplasmic NF90/NF45 co-sediments with subunits of RISC.

NF90 and MOV10 can bind the same target mRNAs

Among the most abundant interactants of NF90 was MOV10 helicase that is required for optimal RISC activity. To further investigate the interaction between NF90 and MOV10 and a possible role in RISC function, we took advantage of an enhanced UV crosslinking followed by immunoprecipitation of NF90 (eCLIP) [29] and an individual-nucleotide resolution UV crosslinking and IP of MOV10 (iCLIP) [26] datasets. Analysis of MOV10 iCLIP identified 1103 mRNAs significantly bound by MOV10. Approximately half of the bound mRNAs (542 mRNAs) contained a MOV10 peak in the 3’UTR (Fig. 2A), consistent with previous findings suggesting that MOV10 mainly binds 3’UTRs of mRNAs [20]. The remaining MOV10-bound mRNAs contained peaks in introns (279 mRNAs), exons (255 mRNAs) and 5′ UTRs (27 mRNAs) (Fig. 2A). On the other hand, analyses of NF90 eCLIP suggest that the majority of mRNAs significantly associated with NF90 were bound in their introns (3217 out of 3942 NF90-bound mRNAs). However, some mRNAs were also bound in their in exons, 5′ UTRs and 3′ UTRs (301, 70, and 354, respectively) (Fig. 2B).

Fig. 2

NF90 and MOV10 can bind common target mRNAs. A MOV10 iCLIP data were analyzed to show the distribution of MOV10 binding along MOV10-associated mRNAs. B NF90 eCLIP data were analyzed to show the distribution of NF90 binding along NF90-associated mRNAs. C Venn diagram showing the intersection between target RNAs bound by MOV10 or NF90. P-value was obtained using Fisher’s exact test. D Venn diagram showing the intersection between target RNAs bound in the 3′ UTR by MOV10 or NF90. P-value was obtained using Fisher’s exact test

Since the interaction between NF90 and MOV10 was found to be RNA-dependent (Fig. 1B), we wondered if these two proteins could bind the same mRNAs. Intersection of the mRNAs bearing at least one peak of NF90 and MOV10 identified 456 mRNAs that could be potentially bound by both proteins (Fig. 2C). Interestingly, this corresponds to around 41% of all mRNAs bound by MOV10 and it is significantly enriched (P = 1.7e−48, Fisher’s exact test). Next, since the unwinding of 3′ UTRs by MOV10 is implicated in RISC-mediated silencing, we wondered whether NF90 and MOV10 were associated with the 3′ UTRs of common mRNAs. We identified 52 mRNAs that bear at least one peak of both NF90 and MOV10 in their 3′ UTRs (Fig. 2D), which is significantly enriched (P = 3.21e−21, Fisher’s exact test). These findings suggest that NF90 and MOV10 can bind the 3′ UTR of a common set of target mRNAs.

NF90 and MOV10 influence each other’s binding to target mRNAs

MOV10 has been shown to facilitate RISC-mediated silencing [23] while, on the contrary, NF90 was found to increase specific target mRNAs stability [9]. We therefore wondered if NF90 could interfere with the binding of MOV10 to the target mRNAs and vice versa. In order to understand the function of NF90 and MOV10 binding to the same target mRNAs, we performed RIP in HEK293T cell line after RNAi against NF90 and NF45 (NF90/NF45), MOV10 or a non-targeting control (Scr), followed by quantitative PCR (qPCR) of target mRNAs selected on the basis of MOV10 iCLIP and NF90 eCLIP analyses (Fig. 2D and Additional file 2: Fig. S2).

RIP of NF90, MOV10, or IgG control was performed after downregulation of MOV10 (Additional file 3: Fig. S3A). Loss of MOV10 did not significantly affect the abundance of the target mRNAs (Additional file 3: Fig. S3B). RIP results revealed that MOV10 association with the selected target mRNAs was significantly decreased after downregulation of MOV10, as expected (Fig. 3). On the other hand, NF90 binding to the same target mRNAs was significantly increased after downregulation of MOV10, while its binding to the negative control, H2BC1, did not significantly change (Fig. 3). Similar results were obtained using a second small-interfering RNA (siRNA) targeting MOV10 (Additional file 4: Fig. S4A and B).

Fig. 3

MOV10 modulates NF90 association with common target mRNAs. RIP analysis of HEK293T cells transfected with MOV10-targeting siRNA or a non-targeting control (Scr), as indicated. RIPs were performed using anti-NF90, anti-MOV10 or a control IgG antibody, as indicated. An aliquot of input and immunoprecipitates were analyzed by RT-qPCR using specific primers, as indicated. Data represent mean ± SEM obtained from 4 independent experiments (*P < 0.05, **P < 0.01, ***P < 0.001, NS indicates not significant, independent Student’s t test) (Additional file 9)

In order to further investigate this mechanism, we performed RIP of NF90, MOV10, or IgG control after downregulation of NF90 and its protein partner NF45 (Additional file 5: Fig. S5A). Interestingly, the loss of NF90/NF45 significantly decreased the total level of the selected target mRNAs (Additional file 5: Fig. S5B), consistent with its role in increasing mRNA stability [9]. As expected, NF90 association with the target mRNAs was significantly decreased after NF90/NF45 downregulation (Fig. 4). On the other hand, RIP analysis revealed that NF90/NF45 downregulation led to a significant increase in MOV10 association with the selected target mRNAs while its binding to the negative control, H2BC1, did not significantly change (Fig. 4). Similar results were obtained using additional siRNAs targeting NF90 and NF45 (Additional file 6: Fig. S6A and B). These results suggest that the binding of NF90 and MOV10 at common target mRNAs is mutually influenced by the presence of the other factor.

Fig. 4

NF90 modulates MOV10 association with common target mRNAs. RIP analysis of HEK293T cells transfected with NF90/NF45-targeting siRNAs or a non-targeting control (Scr), as indicated. RIPs were performed using anti-NF90, anti-MOV10 or a control IgG antibody. An aliquot of input and immunoprecipitates were analyzed using RT-qPCR using specific primers, as indicated. ND indicates “Not Detected.” Data represent mean ± SEM obtained from 4 independent experiments (*P < 0.05, **P < 0.01, ***P < 0.001, NS indicates not significant, independent Student’s t test) (Additional file 9).

Downregulation of NF90/NF45 complex increases Ago2 binding to target mRNAs

MOV10 is known to promote Ago2 association to mRNAs, thereby enhancing RISC-mediated silencing. Since NF90 and MOV10 influence each other’s binding to common target mRNAs, we wondered if alteration of the association NF90/NF45 complex could have similar effects on Ago2 binding to target mRNAs. To this end, we performed RIP of Ago2 or IgG control after downregulation of NF90/NF45. Depletion of NF90/NF45 had no effect on Ago2 expression in extracts or immunoprecipitates (Additional file 7: Fig. S7A and B). Notably, loss of NF90/NF45 significantly increased Ago2 binding to the target mRNAs tested, while the negative control mRNA, H2BC1, which was poorly associated with Ago2, was not significantly increased relative to the IgG control (Fig. 5). Interestingly, the increase in Ago2 binding to target mRNAs is consistent with the observed reduction in the abundance of the same mRNAs following loss of NF90/NF45 (Additional file 5: Fig. S5B).

Fig. 5

Downregulation of NF90/NF45 increases Ago2 binding to selected target mRNAs. RIP analysis of HEK293T cells transfected with NF90/NF45-targeting siRNAs or a non-targeting control (Scr), as indicated. RIPs were performed using anti-Ago2 or a control IgG antibody. Immunoprecipitates were analyzed using RT-qPCR. ND indicates ‘Not Detected’. Data represent mean ± SEM obtained from 3 independent experiments (*P < 0.05, **P < 0.01, ***P < 0.001, NS indicates not significant, independent Student’s t test) (Additional file 9)

NF90/NF45 impedes binding of Ago2 to VEGF mRNA during hypoxia

NF90 is known to stabilize VEGF mRNA during cancer-induced hypoxia [6]. We wondered whether the stabilization of VEGF by NF90 during hypoxia might be mediated by the ability of NF90 to influence Ago2 binding to VEGF mRNA. To this end, we performed RIP of Ago2 or IgG control after downregulation of NF90/NF45 and treatment with the hypoxia-inducing drug, CoCl2. HEK293T treated with 500 μM of CoCl2 show increased HIF1α expression (Fig. 6A), confirming induction of hypoxia, and this concentration was used for further experiments. Depletion of NF90/NF45 had no effect on Ago2 expression in extracts or immunoprecipitates (Fig. 6B, C). However, loss of NF90 significantly decreased the mRNA level of VEGF while increasing the level of H2BC1 mRNA (Fig. 6D). RIP analysis showed that the binding of Ago2 to VEGF mRNA was significantly increased upon loss of NF90/NF45, while the binding to the negative control mRNA, H2BC1, did not significantly change (Fig. 6E).

Fig. 6

NF90/NF45 impedes binding of Ago2 to VEGF mRNA during hypoxia. A Extracts of HEK293T cells treated with different concentrations of CoCl2, as indicated, were analyzed by Western blot using antibodies to HIF1a and TBP. B Extracts of HEK293T cells transfected with siRNAs targeting NF90/NF45 or a non-targeting control (Scr) and treated with CoCl2 were analyzed by Western blot using the indicated antibodies. C Immunoprecipitates obtained using anti-Ago2 or control IgG antibodies from extracts described in (B) were analyzed by Western blot using the indicated antibodies. D Total RNA obtained from HEK293T transfected with siRNAs targeting NF90 and NF45 or a non-targeting control (Scr) and treated with CoCl2 were analyzed by RT-qPCR. Data represent fold mock (IgG) relative to the control samples (siScr), which was attributed a value of 1 (red line), obtained from 3 independent experiments (*P < 0.05, **P < 0.01, independent Student’s t test) (Additional file 9). E RIP analysis of HEK293T cells transfected with NF90/NF45-targeting siRNAs or a non-targeting control (Scr), as indicated, and treated with CoCl2. RIPs were performed using anti-Ago2 or a control IgG antibody. Immunoprecipitates were analyzed using RT-qPCR. ND indicates ‘Not Detected’. Data represent mean ± SEM obtained from 3 independent experiments (*P < 0.05, **P < 0.01, ***P < 0.001, NS indicates Not Significant, independent Student’s t test) (Additional file 9)