In this study, we show that specific arginines within the RGG box are required for normal polyribosome association and RNA binding. Arginines 533 and 538 are required for normal polyribosome association (Fig. ), whereas all four arginine residues play a role in RNA binding (Figs and ). Arginines 533 and 538 are required for sc1 binding but not AATYK association (Figs and ), suggesting for the first time that different arginines of the RGG box are involved in binding different target RNAs. Additionally, we show that methylation of arginine residues 533 and 538 or 543 and 545 leads to loss of sc1 RNA binding (Fig. ). We also provide the first evidence that FMRP associates with PRMT1 (Fig. ) and that PRMT1 methylates FMRP in cells (Fig. ).
The dependence on different arginines for binding to sc1 and AATYK was unexpected. sc1 has been experimentally demonstrated to form a G-quadruplex, and AATYK encodes a guanine-rich configuration of nucleotides that is predicted by the RNABob algorithm to form a G-quadruplex (17
); thus, we hypothesized that sc1 and AATYK would associate similarly with the RGG box. Our observation that they did not was surprising and suggests the alternative hypothesis that sc1 and AATYK form different secondary structures that interact distinctly with the RGG box. We would predict that AATYK is not bound through a G-quadruplex structure, but rather through a G-rich region analogous to the APP mRNA. Similar to the identification of the FMRP-binding site in the APP mRNA (49
), it would be interesting to identify the exact sequence in the AATYK mRNA that associates with FMRP and then perform biophysical methods to query its secondary structure (19
). This approach was used to demonstrate that another FMRP-co-immunoprecipitated mRNA, semaphorin 3A, forms a G-quadruplex that is required for FMRP binding (21
Interestingly, methylation of arginines 533 and 538 encoded by 543,545m or 543 and 545 encoded by 533,545m had no effect on AATYK association (Fig. ). Since arginines 533 and 538 played no role in AATYK binding, this result was not surprising; however, since 543 and 545 did participate in RNA capture (Fig. ), it was unexpected that their methylation had no effect, especially since we showed earlier that methylation of WT FMRP inhibits AATYK binding (23
). One explanation for this result is that arginines 533, 538, 543 and 545 are not as important for AATYK binding as they are for sc1 binding [note that the absence of all four arginines only yields a 33% decrease in capture (Fig. ) compared with an 85% decrease in capture for sc1 (Fig. )]. Other residues in the RGG box clearly are important for the association with AATYK because its removal leads to a nearly 80% loss of capture (Fig. ). Thus, PRMT1 may methylate other arginines in the RGG box that collectively inhibit binding of AATYK by the WT protein.
This study supports the previous observation that the RGG box is required for normal polyribosome association (25
). A different study showing the opposite result where the RGG box was not required for polyribosome association may be due to the phenotype being subtle, especially when compared with the KH domain deletions and the I304N mutation (25
); our results also required multiple trials and examining smaller volume fractions within each polyribosome experiment to determine a statistically significant difference between FMRP and ΔRGG in the early and late fractions. The latter, conflicting study also used a neuronal cell type that expresses endogenous FMRP (40
). Dimerization of endogenous FMRP with transgene may obscure the reduction in transgene polyribosome association.
The polyribosome experiments in Figure demonstrate that arginines 533 and 538 are important for normal FMRP association, yet the RNA capture experiments demonstrated that arginines 533 and 538 are important only for sc1 association and not for AATYK binding. Further, arginines 543 and 545, which play no discernible role in polyribosome association, interacted with both RNAs. A possible explanation for this disparity is that the RNA capture is a more sensitive assay, i.e. polyribosome association does not allow elucidation of somewhat compromised RNA binding like the 44% of 543,545m captured by sc1 compared with the severely compromised binding of 533,538m (24% captured). An alternative explanation is that additional factors, like the methylation state of FMRP, that occur in cells but not in the capture experiments may contribute to polyribosome association. For example, if methylation inhibits association with RNAs and arginines 543 and 545 are constitutively methylated, then in cells, they are functionally unavailable for RNA binding; therefore, their substitution would have no effect on polyribosome association.
We have also provided evidence that FMRP is methylated by PRMT1 in cells. Because we were unable to achieve complete loss of FMRP methylation using siRNAs against PRMT1, it is possible that other PRMTs methylate FMRP, as has been suggested (24
). At this point in time, we do not know whether methylation occurs to inhibit further RNA binding in the cell or whether it occurs on selective arginines to control the identity of the RNAs bound. Regarding the first possibility, the resulting loss of RNA binding by the RGG box may lead to the KH domains becoming more important in RNA binding, altering the potential targets for FMRP RNA binding. It has also been shown that protein methylation occurs in a distributive manner, so proteins can be partially methylated and released from the enzyme to re-associate later and become fully methylated (53
). With this in mind, a second possible scenario is that partial methylation of FMRP by PRMT1 alters the identity or quantity of RNA associated with FMRP. Characterization of the methylation state of the individual arginines of FMRP in vivo
will give insight into this question. Cellular FMRP is not completely methylated, as it can be further methylated in vitro
(Fig. ) (23
). Thus, only a portion of the FMRP present in a cell is methylated at any one time. It will be informative to determine whether FMRP exists in two pools, completely methylated and unmethylated, or whether it exists as a collection of proteins with distinct methylated arginines. If the latter were true, methylation would allow the cell to express FMRPs with varying RNA-binding abilities.