A simple procedure for bacterial expression and purification of the fragile X protein family

The fragile X protein family consists of three RNA-binding proteins involved in translational regulation. Fragile X mental retardation protein (FMRP) is well-studied, as its loss leads to fragile X syndrome, a neurodevelopmental disorder which is the most prevalent form of inherited mental retardation and the primary monogenetic cause of autism. Fragile X related proteins 1 and 2 (FXR1P and FXR2P) are autosomal paralogs of FMRP that are involved in promoting muscle development and neural development, respectively. There is great interest in studying this family of proteins, yet researchers have faced much difficulty in expressing and purifying the full-length versions of these proteins in sufficient quantities. We have developed a simple, rapid, and inexpensive procedure that allows for the recombinant expression and purification of full-length human FMRP, FXR1P, and FXR2P from Escherichia coli in high yields, free of protein and nucleic acid contamination. In order to assess the proteins’ function after purification, we confirmed their binding to pseudoknot and G-quadruplex forming RNAs as well as their ability to regulate translation in vitro.

. Domains and diprolyl/polyproline stretches of the fragile X protein family. (A) FMRP isoform 1, (B) FXR1P isoform 2, and (C) FXR2P with relevant protein domains and motifs labeled. Diprolyl/polyproline locations are displayed, and shortened FXR2P constructs and FXR2P/FMRP mutants are listed. The fragile X proteins are well-conserved (72-77% identity) through the sequence RQIG of each protein (located after KH2 domain), but the sequences diverge after this point (31-61% identity). Sequence identities were determined in MUSCLE 63 .
cytoplasmic FMRP interacting proteins 1 and 2 (CYFIP1/2) [27][28][29] . Interestingly, despite the high homology of the FXPs in the N-terminal region, the interactions between FMRP and CYFIP1 or NUF1P appear unique, whereas CYFIP2 can also interact with FXR1P and FXR2P 28,29 . Such interactions appear to be important for modulating the translation regulation activity of the FXP family. For example, recruitment of the CYFIP1-FMRP complex to mRNAs by transactive response DNA-protein 43 kDa (TDP-43) was found to repress translation initiation 30 . While some of these assays are performed in vivo, or with protein purified from eukaryotic systems, others have utilized protein purified from E. coli for such assays 2,26,27,30 .
Additionally, much emphasis has been placed on proposing mechanisms of translational regulation for the FXP family, with a particular focus on determining their mRNA targets. Many studies have attempted to identify and validate the mRNA targets of FMRP, while several papers have identified targets of FXR1P and FXR2P 12,[31][32][33][34][35][36][37] . Although there appears to be overlap in the mRNA targets of the FXP family, there is evidence that each protein has unique mRNA targets 11,12,31,34,36,38 . In order to validate, analyze, or compare the mRNA targets of the FXP family, researchers often test the direct binding of each protein to its mRNA targets in vitro. These studies allow researchers to identify binding sites within a target mRNA or test binding to in vitro selected RNAs, leading to the identification of sequence motifs or structural features the proteins may recognize in vivo [39][40][41] . Such studies have identified G-quadruplexes and kissing complexes as RNA features recognized by FMRP 11,[39][40][41] . Thus, it is important to purify these proteins in sufficient quantities, with sufficient purity for in vitro assays.
However, researchers have faced difficulty in purifying full-length FXPs due to their poor expression, the production of truncated proteins (TPs), their tendency to aggregate and precipitate, and their instability in solution when not bound to RNA 36,[42][43][44] . In order to overcome such obstacles, researchers have implemented strategies such as plasmids with tRNAs for rare codons to improve expression, extensive and stringent washes to remove contaminant proteins and TPs, purification from inclusion bodies, and purification under denaturing conditions which requires protein refolding and often lengthy dialysis steps 36,43,45 . Many researchers have also purified specific regions or domains instead of purifying the full-length proteins 11,39,42,45,46 .
Others have purified the FXPs from mammalian cells or the SF9/baculovirus system, but these systems are more expensive and time consuming than purifying from E. coli, and yields can still be low 26,28,32,36,41,45 . One group noted that when purifying from HEK293 cells, it was challenging to obtain high yields of full-length human FMRP, or to obtain the protein at concentrations above ~ 1 μM, noting that this could be due to low expression or a tendency of the protein to precipitate at higher concentrations 47 . Purification from E. coli produces proteins lacking post-translational modifications unique to mammalian systems, which have been proposed to have a role in the FMRP's function 48,49 . Thus, our method would not be suitable for researchers who desire to characterize the function of the FXPs with the post-translational modifications that occur in mammalian systems. However, our purification from E. coli would be advantageous for those who do not require post-translational modifications for their in vitro assays, or for those who wish to analyze the effects of specific, individual posttranslational modifications.
Our purification protocol improves upon previous methods by allowing the FXP family to be purified using a single, simple protocol. This protocol is fast and inexpensive as the proteins are all recombinantly expressed and purified in E. coli, and the materials required are available in most biochemistry laboratories. Briefly, mutations were implemented to disrupt ribosomal stalling proline-rich motifs within the protein sequences. These mutations in tandem with a maltose-binding protein (MBP) tag dramatically boosted the expression of the proteins. The mutations also reduced the production of TPs. An ammonium sulfate precipitation step removed the majority of protein contaminants, while the use of a heparin column removed remaining protein contaminants, TPs, and nucleic acid contamination. The final protein samples were pure and obtained in high yields of 1-9 mg from 2 L of culture. Finally, the purified proteins bound to G-quadruplex and kissing complex RNAs and inhibited translation in vitro, demonstrating that they are functional.

Results
expression of recombinant fragile X proteins. We initially attempted to purify FXR2P from E. coli using an N-terminal 6X His-tag, and in doing so, encountered extremely poor expression of the full-length protein. In fact, the primary protein obtained was E. coli bifunctional polymyxin resistance protein ArnA, which has a similar molecular weight (74 kDa) to His 6 -FXR2P (75 kDa). Furthermore, ArnA forms a hexamer with surface-exposed patches of histidine residues that bind to nickel beads 50 . To improve recombinant expression, codon optimized sequences were purchased for human FXR2P and FMRP (isoform 1); FXR1P (isoform 2) expression was sufficient, so a codon optimized sequence was not used. All three genes were cloned into standard protein expression vectors and transformed into E. coli Rosetta 2(DE3)pLysS for expression tests.
After codon optimization, the expression of FXR2P was still low, so several fusion tags were tested to boost protein expression. Only an N-terminal SUMO or MBP tag seemed to boost the expression of FXR2P, which is supported by the observation that MBP and thioredoxin (TrX) tags are the best N-terminal tags for promoting protein solubility 51 . There is a decreasing likelihood of soluble expression of mammalian proteins as their molecular weight increases, while the presence of low complexity regions within a protein correlates with reduced soluble expression in E. coli, which could both explain FXR2P's poor expression without the MBP tag 51 . Although the His 6 -MBP tag is large (~ 44 kDa), in some cases it promotes the proper folding of the attached protein into the biologically active conformation 52 . Furthermore, it did not contribute to the RNA-binding or translation regulation capabilities of the FXPs. Moreover, it may be possible to cleave the MBP-tag after purification (data not shown). Thus, we selected the MBP tag for our studies.
The MBP tag boosted expression of FXR2P, but we observed the production of many TPs, which we hypothesized were a result of diprolyl and polyproline stretches within FXR2P (Figs. 1C and 2C). As the ring structure of proline makes it a poor peptide bond donor and acceptor, two or more consecutive prolines can cause www.nature.com/scientificreports/ ribosomes to stall during translation 53,54 . Due to nascent chain-mediated stalling of ribosomes, ribosomal rescue mechanisms may release the ribosomes and unfinished proteins from the mRNA chain, leading to the TPs we observe 55,56 .
To enhance expression of the full-length protein and reduce the production of TPs we attempted two approaches (1) co-expression with elongation factor P (EF-P) which alleviates ribosomal stalling in short prolinerich motifs by stimulating peptide bond formation and (2) mutations to disrupt the diprolyl and polyproline motifs 57 . Although co-expression with EF-P enhanced FXR2P expression, we did not see similar results with FMRP ( Supplementary Fig. 2). Furthermore, expression with EF-P did not appear to alleviate the production of TPs. In contrast, mutations dramatically boosted expression of full-length FXR2P and FMRP while reducing the production of TPs ( Fig. 2A,C).
In creating mutations, we aimed to preserve the original protein sequence as much as possible ( Supplementary  Fig. 3). We therefore created and compared the expression of two shortened constructs of FXR2P and seven mutants, which enabled us to determine that the polyproline stretch from residues 492-494 has the greatest contribution towards ribosomal stalling and the production of TPs ( Supplementary Fig. 4). However, we found FMRP P451N and FMRP P451S mutants exhibited a sevenfold increase in expression compared to wild type FMRP, and a reduction in truncated proteins (TP). FXR1P was not mutated, and a non-codon optimized sequence was used, which may explain the lower expression compared to FMRP/FXR2P mutants. FXR2P P474S and P493N was selected as it had the highest expression (130-fold greater than wild type FXR2P), less TPs than FXR2P P493N (100-fold greater), and fewer mutations than FXR2P P400 (120-fold greater). www.nature.com/scientificreports/ we could further enhance expression and reduce TPs by also disrupting the nearby diprolyl motif from residues 473-474. Interestingly, the diprolyl and polyproline stretches at residues 529-530 and 538-540, which are situated closer to one another, do not seem to cause much ribosomal stalling or TPs (see FXR2P P500 mutant, Supplementary Fig. 4). This ultimately led us to select a FXR2P double mutant (FXR2P P474S and P493N) for future purification attempts (Fig. 2C). Based on the results for FXR2P, we decided to mutate the polyproline motif in FMRP that was also located within its disordered region, residues 450-452. Mass spectrometry results of the major TP of FMRP indicated that truncation was occurring after P451, providing further evidence that the TPs are a result of ribosomal stalling at proline-rich motifs (data not shown). Of the two mutants, we selected FMRP P451S for purification attempts ( Fig. 2A). FXR1P did not need to be mutated as it contains only one polyproline motif, however we did observe a truncation for FXR1P (Figs. 1B and 2B). This TP does not appear to be due to ribosomal stalling as mass spectrometry results suggested the truncation occurs within the KH1 domain (data not shown). This TP is lacking the KH2 domain and C-terminal disordered region and appeared to be more stable than full-length FXR1P.
Purification of the fragile X proteins. After selecting a FXR2P and FMRP mutant for purification, we set out to identify a single expression and purification scheme for all three proteins. We grew our cells at 37 °C until the OD 600 was ~ 0.4, then allowed the cells to grow an additional 14-16 h at ~ 14 °C after induction with Isopropyl β-d-1-thiogalactopyranoside (IPTG). Initially we attempted amylose resin for protein affinity chromatography, however, we encountered poor binding to the amylose resin unless we implemented a purification step prior to batch binding, and we were unable to remove all the TPs. We therefore switched to an ammonium sulfate precipitation followed by a heparin column. After harvesting and lysing the cells and clarifying the lysate, we were pleased to discover that all three proteins could be precipitated at relatively low percentages of ammonium sulfate, while the majority of E. coli proteins remained in the supernatant (Supplementary Figs. 5-7). After allowing the FXP to precipitate, the sample was centrifuged, the supernatant removed, and the pellet resuspended. The resulting solution was dialyzed overnight to remove ammonium sulfate as salt must be removed prior to the heparin column. After dialysis, soluble protein was further purified through fast protein liquid chromatography (FPLC) with a heparin column. Using an increasing salt gradient from 0-1 M NaCl (or KCl) we were able to remove nucleic acid contamination, residual protein contaminants, and TPs from the FXPs ( Supplementary  Fig. 8). Interestingly, we observed that the C-terminally truncated His 6 -MBP-FXR1P that is predicted to be missing the KH2 domain and disordered region (~ 75 kDa) did not appear to bind to the heparin column and was present predominantly in the flow-through ( Supplementary Fig. 6). Additionally, the FXR2P TPs began eluting prior to the full-length protein, while the full-length was predominantly eluted at high salt concentrations (Supplementary Fig. 7). In fact, in subsequent purifications, a step gradient was implemented to remove the TPs prior to eluting full-length FXR2P. Thus, it appears the FXPs interact with the heparin column through their C-terminal disordered regions. Through the use of a single ammonium sulfate precipitation step and a heparin column, we were able to purify all three FXPs (Fig. 3).
concentration and storage of the fragile X proteins. After purification we concentrated FMRP and FXR1P by centrifugation. However, we observed a loss in yield, perhaps due to the protein sticking to the concentrator and due to protein precipitation. In our hands, even with a His 6 -MBP tag which promotes solubility, the proteins will precipitate if concentrated too much after removal of nucleic acid contamination 51,52 . FMRP has www.nature.com/scientificreports/ been concentrated to ~ 16 μM but is not stable for long at this concentration and drops to ~ 3-6 μM over time. FXR1P and FXR2P precipitate more readily. We were able to concentrate FXR1P to 8 μM but precipitation began occurring at ~ 3 μM. FXR1P and FXR2P fractions from the heparin column at ~ 5-7 μM seem stable whereas fractions at ~ 9-13 μM had visible precipitation as they eluted. Once purified from nucleic acids, FMRP seems stable at 16 μM and FXR1P/FXR2P at ~ 6 μM initially, but the proteins appear to precipitate slowly over time if not stored at − 80 °C.
To maximize yield, we suggest researchers avoid concentrating, concentrate minimally, or concentrate right before use or storage at − 80 °C. When we centrifuged our concentrated protein samples after storage at 4 °C or − 15 °C to remove precipitated protein prior to use, we noticed a decrease in concentration of the samples over time. This is likely due to precipitation. Storage at − 80 °C appears prevent this, although thawing may induce precipitation. The tendencies of the FXPs to form aggregates, precipitate during concentration, or not concentrate past a certain concentration have been noted by other researchers [42][43][44]47 . Concentrating appears to induce aggregation and precipitation, which may have a function in vivo, namely in the proteins' presence in ribonucleoprotein granules 44 . In rat brain, FXR1P predominantly forms oligomers or insoluble aggregates, while monomers are nearly undetectable 45 . We therefore recommend the MBP tagged FXPs be stored at − 80 °C for long-term storage. Prior to use in assays, we suggest researchers centrifuge stored samples to remove insoluble aggregates then remeasure protein concentration.
Confirmation of fragile X protein identities. After purifying the FXPs, the final samples were analyzed by mass spectrometry. The correct identity of each protein was confirmed with 94% sequence coverage obtained for FXR1P, and 96% for FMRP and FXR2P.
Analysis of the RnA-binding activity of the fragile X proteins. The functionality of the purified proteins was verified by testing the FXP family's binding to a G-quadruplex forming RNA, a well-known target of FMRP, which is bound by the RGG motif, and a target of FXR1P 11,26,35,40,41,58 . We chose to test the proteins' binding to poly-G 17 U as our lab has identified poly-G 17 U as a G-quadruplex forming RNA 46 . For a negative control we used CR1, an RNA with no predicted G-quadruplex forming capability 46 .
As predicted based on our previous observations for an N-terminally truncated FMRP and the FMRP RGG motif, all proteins showed high affinity binding to poly-G 17 U, and no binding to CR1 in the concentration range tested (Fig. 4A-C) 46 . Our fluorescence anisotropy results reinforce the hypothesis that the FXPs have different affinities for mRNA targets: FXR2P showed the greatest affinity for poly-G 17 U, and FXR1P the least (FXR2P K D = 3.1 ± 0.4 nM, FMRP K D = 5.6 ± 0.6 nM, and FXR1P K D = 11.7 ± 1 nM). To ensure that the His 6 -MBP tag did not contribute to the observed RNA-binding, we tested its binding to poly-G 17 U and CR1 and observed no binding (Fig. 4D). Furthermore, it appears that the mutation implemented to boost FMRP expression does not impair the protein's ability to bind poly-G 17 U RNA, as we obtained higher affinity binding than previously determined for an N-terminally truncated FMRP construct (R218-P632 of FMRP) or a glutathione S-transferase tagged FMRP RGG fusion protein (G531-P632 of FMRP): K D of 14 ± 2 nM and 8.6 ± 1.2 nM, respectively 46 .
After assessing the binding of the FXPs to a G-quadruplex forming RNA, we tested for binding to a looploop pseudoknot, or "kissing complex" RNA, ΔKC2, a shortened version of an in vitro selected target of FMRP called KC2 39 . The KH2 domain of FMRP was found to be necessary and sufficient for FMRP binding to KC2 39 . Furthermore, the KH2 domains of FMRP, FXR1P, and FXR2P bind KC2 RNA with equal affinity 11 . As binding of FMRP to KC2 is dependent on the integrity of the KH2 domain, we felt it valuable to assess the FXP family's ability to bind ΔKC2 11,39 .
Due to the size of our ΔKC2 (72 nucleotides), we tested for binding by electrophoretic mobility shift analysis (EMSA) using CR1 and poly-G 17 U as negative and positive controls, respectively. As predicted, all proteins showed binding to poly-G 17 U and ΔKC2, indicated by the reduction in free RNA upon the addition of protein, but not to CR1 (Fig. 5A). It is worth mentioning that there are two free RNA species visible in the control lane for ΔKC2. As this is a native gel, it is likely that the faster migrating species is more compact and folded in the correct confirmation. Darnell et al. also observed two free RNA species for KC2 RNA, and only the faster migrating species was observed to shift with added protein 39 . in vitro translation regulation by the fragile X proteins. We further analyzed the functionality of the FXPs by testing their ability to regulate translation in an in vitro translation system (IVTS) comprised of rabbit reticulocyte lysate treated to reduce endogenous mRNAs. We chose Renilla luciferase mRNA as the reporter for protein synthesis as other researchers have used it previously with the FXPs, it has 3 G-rich sequences that may form G-quadruplex structures, and we have previously observed inhibition of this mRNA by an N-terminally truncated Drosophila FMRP in Drosophila embryo extract 35,59 . We monitored the translation of a 5′ capped Renilla luciferase mRNA with a 3′ poly (A) tail through bioluminescence and observed that all the FXPs inhibited translation, but to different extents (Fig. 5B). Similar to our anisotropy results for poly-G 17 U, we observed the greatest translation inhibition by FXR2P, and the least by FXR1P (FXR2P 2.06% ± 0.54 percent luciferase activity, FMRP 11.7% ± 2.1, and FXR1P 20.2% ± 4.3). As expected, the His-MBP tag did not inhibit the translation of Renilla luciferase mRNA (103.8% ± 7.4% activity). Our results suggest that our purified FXPs maintain their ability to regulate translation.

Discussion
Our RNA-binding and in vitro translation studies suggest the FXPs are functional. Thus, mutations implemented to boost expression of FMRP and FXR2P do not appear to impact their RNA-binding specificity or their ability to repress translation. We were intrigued by the results of our binding studies to poly-G 17 11 . The results we observe may be due to the fact that sc1 was selected using FMRP, while poly-G 17 U may form a generic G-quadruplex structure recognized by all three proteins 40 . Additionally, our results may have been impacted by assessing the binding of the full-length proteins. Future assays testing the RNA-binding specificity of the full-length proteins should yield insightful results as it has been proposed that the multiple RNA-binding domains of the FXPs function cooperatively 31,33 .
Our purification protocol opens the door for compelling research on the FXP family, particularly FXR1P and FXR2P, which have not been studied as extensively as FMRP. As an example, we observed high affinity binding of FXR1P isoform 2 to poly-G 17 U, despite previous results suggesting only the muscle-specific isoforms exhibit high affinity binding for G-quadruplex forming RNAs 26 . Additionally, the ability of FXR2P to bind to G-quadruplexes is not well-documented, yet our results suggest this could be a worthwhile avenue for further research. Finally, the trend we observed for the extent of translation inhibition of Renilla luciferase mRNA, which has potential G-quadruplex structures, matches the trend for the FXPs' binding affinity for the G-quadruplex In summary, we have identified a rapid, simple, and inexpensive purification protocol for the human FXP family, while many of our techniques can be broadly applied. We found the MBP tag very efficient at enhancing the expression and solubility of our proteins, which may be useful to researchers working with large eukaryotic proteins, or proteins with disordered regions 51,52 . By disrupting ribosomal stalling proline-rich motifs within FMRP and FXR2P we drastically boosted recombinant expression while reducing the production of TPs. This technique, or co-expression with EF-P, may assist in the recombinant expression of eukaryotic proteins, 10% of which possess polyproline motifs 60 . An ammonium sulfate precipitation followed by a heparin column allowed the FXPs to be obtained in high yields, free of E. coli protein and nucleic acid contamination. Additionally, this procedure removed the vast majority, and in some cases all, C-terminally TPs. We found the heparin column to be a quick and effective method for removing nucleic acid contamination from nucleic acid binding proteins. All three proteins demonstrated RNA-binding activity through their binding to G-quadruplex and kissing complex forming RNAs and were successful in repressing the translation of Renilla luciferase mRNA. We hope this procedure will mitigate obstacles faced in studying the important roles of the FXP family in translational regulation, and in doing so, promote diverse research questions. Moreover, the techniques described will aid researchers in recombinantly expressing and purifying proteins with poor expression, proline-rich regions, disordered regions, or nucleic acid binding properties. + F M R P + F X R 1 P + F X R 2 P CR1 + F M R P + F X R 1 P + F X R 2 P poly-G 17 U + F M R P + F X R 1 P + F X R 2 P

Creation of EF-P and FMRP/FXR2P co-expression vectors.
A codon optimized sequence for Elongation Factor P (NCBI Reference Sequence: P0A6N4.2) was purchased as His 6 -EF-P in pUC57 (Gene Universal). Seamless cloning was used to insert His 6 -EF-P into a pDSG310 vector (a gift from Ingmar Riedel-Kruse: Addgene plasmid #115,611; https ://n2t.net/addge ne:11561 1; RRID: Addgene_115611). The pDSG310 vector was selected as it has an arabinose regulated promoter (pBAD), p15A origin of replication, and kanamycin resistance, which are all distinct from those of the pMCSG9 vector used for FMRP and FXR2P 62 . Using primers that enabled seamless cloning, PCR was used to prepare the His 6 -EF-P DNA for insertion into pDSG310, and the backbone of pDSG310 was likewise prepared. The insert and pDSG310 backbone were digested with BbsI and ligated, and the ligated plasmid was transformed into DH5α E. coli cells which produced colonies containing viable EF-P containing plasmids. The resulting plasmids were purified, and the sequences verified by Sanger sequencing (GeneWiz). To make cells co-expressing EF-P and FMRP or FXR2P, chemically competent Rosetta 2(DE3)pLysS E. coli cells (Novagen) were transformed with 1:1 EF-P plasmid: FMRP/FXR2P plasmid (200 ng each). A control cell stock containing only EF-P in Rosetta 2(DE3)pLysS E. coli cells was also produced. These cultures were cooled for 10 min at ~ 0 °C to slow cell growth prior to the addition of IPTG. After expression, samples were prepared for analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The OD 600 of a ¼th dilution of each culture was obtained and used to determine the OD 600 of the stock solution. For each culture, 500 μL of sample was centrifuged at 16,100 RCF to pellet the cells. After removing the supernatant, the pellet was resuspended in 25 μL of resuspension buffer (10 mM Tris pH 7.5, 20 mM NaCl) per 0.5 OD 600 unit. After resuspending, 60 μL of the cell pellet resuspension was combined with 15 μL of 5X SDS-PAGE loading dye. The samples were boiled at ~ 95 °C for 10 min, then spun down for a few seconds. The samples were mixed, and 10 μL of each sample was loaded onto a 10% SDS polyacrylamide gel that was run for 10 min at 100 V, followed by 180 V until the dye front ran off the gel (~ 55 min). The protein bands were visualized by staining the gels with Coomassie Brilliant Blue.

Primers to insert EF-P into
Comparison of full-length FMRP and FXR2P expression with and without mutations was performed by analyzing band intensities for each full-length FXP in ImageJ. The background intensity was subtracted from each sample by subtracting the band intensity in the corresponding uninduced samples from the band intensity in the induced samples. For FMRP P451S only, there appears to be leaky expression as a band corresponding to FMRP appears present in the uninduced sample. Therefore, for FMRP P451S only, the uninduced sample band intensity for FMRP P451N was subtracted instead of the FMRP P451S uninduced intensity. Fold change of expression was then calculated by normalizing the band intensity of each full-length FXP mutant to the band intensity for the respective unmutated full-length FXP. The same analysis was performed to analyze full-length FXR2P expression with and without EF-P co-expression. After subtracting the background occurring in the uninduced sample, the fold change of expression was calculated by normalizing the band intensity of full-length FXR2P when co-expressed with EF-P to the band intensity of full-length FXR2P without EF-P expression. After induction the cells were split into six 500 mL centrifuge flasks and pelleted by centrifuging at 4,420 RCF, 4 °C, 15 min (Beckman J2-HC Centrifuge). Pellets were transferred to a 50 mL polypropylene Falcon tube and weighed; the typical weight was 6.4-7.4 g from 2 L of culture. The pellet was then resuspended in lysis buffer with no salt (50 mM Tris pH 7.5, 1 mM ethylenediaminetetraacetic acid (EDTA), 1 mM dithiothreitol (DTT), and 1 mM phenylmethylsulfonyl fluoride (PMSF)) to a total volume of ~ 50 mL. The resuspended cells were sonicated (Branson Digital Sonifier) on ice with ten 8 s pulses at an amplitude of 60% interspersed with 1 min pauses. The crude lysate was then clarified by centrifuging at 50,271 RCF, 30 min, 4 °C (Beckman Coulter Optima LE-80 K Ultracentrifuge). The clarified lysate was placed in a beaker on a stir plate at 4 °C and concentrated ammonium sulfate (5 mM HEPES pH 7.5) at 4 °C (concentration is temperature dependent, but ~ 3.8 M at 0 °C) was added to a final concentration of 20% for FXR1P and FXR2P and 25% for FMRP.
After ammonium sulfate addition, the clarified lysate was allowed to sit at 4˚C for at least an hour. Subsequently, the solution was centrifuged at 5,087 RCF, 15 min, 4˚C (Sigma 4K15C) and the pellet was resuspended in 60-75 mL of lysis buffer. The resuspension was dialyzed in 2L of dialysis buffer (50 mM Tris pH 7.5, 1 mM EDTA, and 1 mM DTT) for at least 16 h. After dialyzing, the protein solution was centrifuged at 5,087 RCF, 30 min, 4˚C in a swinging bucket centrifuge (Sigma 4K15C) to remove insoluble protein.
Fifty milliliters of the supernatant was loaded into a 50 mL superloop (Amersham Biosciences) and bound to a 5 mL heparin column (HiTrap Heparin HP 1X5 mL, GE Healthcare) by fast protein liquid chromatography (ÄKTApurifier, Amersham Pharmacia Biotech) that had been pre-equilibrated with at least 5 column volumes (CV) of 0 M salt buffer (25 mM Tris pH 7.5, 1 mM EDTA, 1 mM DTT, 25% glycerol). The typical mass of protein loaded onto the column ranged from 17-130 mg. After collecting the flow-through the column was washed with 2 CV of 0 M salt buffer. The proteins were subsequently eluted with a linear salt gradient that started with 0 M salt buffer (25 mM Tris pH 7.5, 1 mM EDTA, 1 mM DTT, 25% glycerol) and increased the salt concentration over 30 CV, ending with 1 M salt buffer (25 mM Tris pH 7.5, 1 mM EDTA, 1 mM DTT, 25% glycerol, 1 M NaCl). Each of the FXPs eluted within a unique range of salt concentrations. Pure FMRP fractions eluted from ~ 500-600 mM NaCl, with the peak max at ~ 560 mM, FXR1P at ~ 560-700 mM NaCl, peak max at ~ 640 mM, and FXR2P over a large range, however the most full-length with the least TPs eluted from ~ 700-830 mM NaCl with the peak max at ~ 730 mM . It is important to note that KCl can be used instead of NaCl. This appears to elute the proteins at lower salt concentrations: ~ 400-500 mM for FMRP and ~ 500 mM for FXR1P and FXR2P. Additionally, a step gradient has also been successful for separating the FXPs from contaminant proteins and TPs.
After analyzing the elution fractions by SDS-PAGE, the desired fractions were either pooled and concentrated by centrifugation at 2,493 RCF, 4 °C through a 15 mL 50 kDa MW cutoff concentrator (Amicon Ultra -15 Centrifugal Filters, Millipore Sigma) or individual elution fractions from the heparin column were stored. In either case, the final sample(s) was centrifuged at 16,100 RCF for 10 min at 4˚C to remove any precipitated protein immediately prior to concentration measurements and storage. The concentration of the supernatant was determined from the absorbance at 280 nM (A280 values) (Thermo Scientific NanoDrop 2000/2000c spectrophotometer). The pure FXPs were then stored at 4 °C, -15 °C, or -80 °C. The best temperature for storing the proteins appears to be -80 °C as precipitation appears to occur over time at 4 °C and -15 °C. For samples stored at any temperature, precipitation may occur over time or upon thawing. We therefore recommend that samples are centrifuged to remove precipitated protein and the protein concentration remeasured prior to use in assays.
The A260/280 ratio of stored samples is typically ~ 0.53-0.62, indicating the nucleic acid contamination has been removed (https ://www.biote k.com/resou rces/docs/Power Wave2 00_Nucle ic_Acid_Purit y_Asses sment .pdf). Protein yield ranged from ~ 1.30-8.59 mg; lower yields around 1-2 mg occurred when the fractions from the heparin column were pooled and concentrated. A loss in concentration occurs during concentration steps as discussed previously. The yield reflects the mass of protein at the end of the purification after removing precipitated protein. Storage buffers for use as blanks for concentration readings and for use in assays were created by mixing 0 M and 1 M salt buffers to create buffers with salt concentrations matching that of the final stored protein samples. The concentration of salt in the protein sample was determined from the elution plots from the FPLC machine.
Mass spectrometry analysis of the fragile X proteins. The identities of the FXPs were confirmed through mass spectrometry. The band corresponding to each protein was excised from an SDS-PAGE gel and trypsin digested. The samples were subsequently analyzed by liquid chromatography-mass spectrometry (1.5-h Reverse phase C18 gradient).

G-quadruplex RNA-binding of the fragile × proteins.
To confirm that the FXPs were functional, we tested their binding to G-quadruplex forming RNA through fluorescence anisotropy. To ensure the RNA-binding we observed was not due to the tag we used, we also assessed the binding of His 6 -MBP.
The purified FXPs were centrifuged at 16,100 RCF, 10 min, 4 °C with a benchtop centrifuge to remove any precipitated protein prior to each experiment. The supernatants containing soluble protein were obtained and the concentration of protein determined using A280 readings ( www.nature.com/scientificreports/ spectrophotometer). Poly-G 17 U and CR1 labeled with a 3′ fluorescein (Dharmacon) were diluted to 5X concentrations (~ 25 nM) and these solutions were kept in the dark during the experiment. The RNAs in these 5X solutions were renatured in renaturation buffer (50 mM Tris pH 7.5, 75 mM KCl, 2 mM MgCl 2 ) by heating at 68 °C for 5 min, then slow cooled from 68 °C to ~ 28 °C for ~ 1 h in a water bath. Water, binding buffer, protein storage buffer, protein, and the 5X RNA solution were added in the order listed and mixed together for a final reaction volume of 200 μL. The final reactions contained 20 mM Tris pH 7.5, 75 mM KCl, 5 mM MgCl 2 , 1 μM BSA, 1 mM DTT, 100 ng/μL tRNA (to prevent non-specific binding), and ~ 5 nM RNA. The protein concentrations tested were 0, 1, 2, 5, 10, 20, 50, 100, and 200 nM (for FMRP, only one trial for 1 and 2 nM points). It is important to note that for each protein concentration tested the total volume of protein + protein storage buffer remained constant. In each trial the binding buffer was adjusted to account for the Tris pH 7.5 and DTT that were contributed from the protein storage buffer. For trials with His 6 -MBP we assumed the storage buffer contributions were negligible since the protein was diluted in FXP storage buffer prior to use, and the FXP storage buffer was used as the protein storage buffer in the anisotropy reactions. Reactions were thoroughly mixed and incubated in the dark at room temperature for 1 h. After incubation, each reaction was added into a 96-well non-binding plate (Greiner Bio-One) for fluorescence anisotropy using a multimode microplate reader (SPARK TECAN). Samples were excited at 485 nm and emission was measured at 535 nm. To determine binding affinities, the anisotropy data from each binding assay were normalized to initial values without protein, plotted, and fit to a quadratic equation as previously described 46 . Three independent trials were performed (except for FMRP 1 and 2 nM points) to determine standard deviations.
RNA sequences. 18   RnA-binding of fragile X proteins by electrophoretic mobility shift assay. To confirm the FXPs were functional, we tested their binding to ΔKC2 RNA through an electrophoretic mobility shift assay. Based on the results from fluorescence anisotropy, we used poly-G 17 U and CR1 RNAs as positive and negative controls respectively.
The purified FXPs were centrifuged at 16,100 RCF, 10 min, 4 °C with a benchtop centrifuge to remove any precipitated protein prior to each experiment. The supernatants containing soluble protein were obtained and the concentration of protein determined using A280 readings (Thermo Scientific NanoDrop 2000/2000c spectrophotometer). Fluorescein-labeled Poly-G 17 U, CR1, and ΔKC2 were diluted to 10X concentrations (1 μM) and these solutions were kept in the dark during the experiment. The 10X RNA solutions were renatured in renaturation buffer (50 mM Tris pH 7.5, 100 mM KCl, 5 mM MgCl 2 ) by heating at 68 °C for 5 min, then slow cooled from 68 °C to ~ 29 °C for ~ 1 h in a water bath. Water, 10X binding buffer, protein storage buffer, fragile X protein, and the 10X RNA solution were added in the order listed and mixed together for a final reaction volume of 26 μL. The final reactions contained 50 mM Tris pH 7.5, 145 mM KCl, 5 mM MgCl 2 , 1 μM BSA, 10 mM DTT, 50 ng/ μL tRNA (to prevent non-specific binding), ~ 100 nM fluorescein-labeled RNA, and for reactions containing protein, 250 nM of protein. For each protein concentration tested the total volume of protein + protein storage buffer remained constant. In each reaction the binding buffer was adjusted to account for the Tris pH 7.5, KCl, and DTT that were contributed from the protein storage buffer. The reactions were thoroughly mixed and incubated in the dark at room temperature for 1 h. After incubation, 3 μL of loading dye (xylene cyanol in 50% glycerol) was added to each reaction. A 0.8% agarose gel (SeaKem GTG agarose) was prepared in 1X TBE buffer (100 mM Tris pH 8.3, 100 mM borate, 2 mM Na 2 EDTA). After loading 13 μL of each sample, the gel was run at 4 °C for 2 h at 66 V in 1X TBE buffer. The gel was then scanned using a laser scanner (Typhoon FLA 9500, GE Healthcare) and the gel was analyzed in ImageJ.

Scientific RepoRtS
| (2020) 10:15858 | https://doi.org/10.1038/s41598-020-72984-7 www.nature.com/scientificreports/ Analysis of in vitro translation regulation by the fragile X proteins. The purified FXPs and His 6 -MBP were centrifuged at 16,100 RCF, 10 min, 4 °C with a benchtop centrifuge to remove any precipitated protein prior to each experiment. The supernatants containing soluble protein were obtained and the concentration of protein determined using A280 readings (Thermo Scientific NanoDrop 2000/2000c spectrophotometer). Water, 5′ capped Renilla luciferase mRNA with a 25-nucleotide 3′ poly(A) tail, protein storage buffer, and the corresponding FXP or His 6 -MBP were combined in the order listed, mixed, and allowed to incubate for 10 min at room temperature. Subsequently, we added 2X rabbit reticulocyte lysate that was treated with micrococcal nuclease to reduce endogenous mRNAs and reduce background translation. The reactions were then allowed to incubate at 30 °C for 1.5 h. The final 50 μL reactions contained 10 nM Renilla luciferase mRNA and 500 nM FXP or His 6 -MBP. After incubation, 45 μL of each reaction was combined with 5 μL of 30 μM colenterazine to achieve a final concentration of 3 μM colenterazine. Each reaction was then added into a 384-well plate (Greiner Bio-One) and the luminescence determined using a multimode microplate reader (SPARK TECAN). For each in vitro translation reaction with protein added, the raw luminescence values were compared to the raw luminescence values of the reaction with protein storage buffer only added in order to account for any effect on translation resulting from salts or other reagents in the protein storage buffers.