The Drosophila hnRNP F/H homolog Glorund recruits dFMRP to inhibit nanos translation elongation

Abstract Translational control of maternal mRNAs generates spatial and temporal patterns of protein expression necessary to begin animal development. Translational repression of unlocalized nanos (nos) mRNA in late-stage Drosophila oocytes by the hnRNP F/H homolog, Glorund (Glo), is important for embryonic body patterning. While previous work has suggested that repression occurs at both the translation initiation and elongation phases, the molecular mechanism by which Glo regulates nos translation remains elusive. Here, we have identified the Drosophila fragile X mental retardation protein, dFMRP, as a Glo interaction partner with links to the translational machinery. Using an oocyte-based in vitro translation system, we confirmed that Glo regulates both initiation and elongation of a nos translational reporter and showed that dFMRP specifically represses translation elongation and promotes ribosome stalling. Furthermore, we combined mutational analysis and in vivo and in vitro binding assays to show that Glo's qRRM2 domain specifically and directly interacts with dFMRP. Our findings suggest that Glo regulates nos translation elongation by recruiting dFMRP and that Glo's RNA-binding domains can also function as protein-protein interaction interfaces critical for its regulatory functions. Additionally, they reveal a mechanism for targeting dFMRP to specific transcripts.


INTRODUCTION
Translational control, coupled with other types of posttranscriptional regulation including mRNA localization and degradation, is a key mechanism for establishing and maintaining protein asymmetries. Translational control plays a particularly important role in early embryonic development in organisms like Drosophila (1,2), Caenorhabditis elegans (3) and Xenopus (1), where transcription of the zygotic genome is delayed for many cell divisions. The embryo therefore relies on spatial and temporal control over the translation of a stockpile of maternally synthesized mRNAs to execute the earliest cell fate determination and axial patterning programs.
The Drosophila anterior-posterior body axis is patterned by opposing protein gradients of Bicoid (Bcd) and Nanos (Nos) in the early embryo. Both gradients originate from localized translation of maternal bcd and nos mRNAs at the anterior and posterior poles of the embryo, respectively (4). Unlike bcd, which is tightly sequestered at the anterior pole (5), only a small proportion (4%) of nos resides at the posterior and is translated (6). Therefore, whereas RNA localization is sufficient to sequester Bcd protein synthesis at the anterior (7), an additional level of control is required to restrict production of Nos to the posterior. This is accomplished by maintaining nos in a translationally repressed state that is alleviated only by its posterior localization (6,8,9). Failure to repress unlocalized nos mRNA and the consequent accumulation of Nos throughout the embryo result in loss of anterior structures (10,11).
As a maternally supplied mRNA, nos is synthesized in the ovarian nurse cells (12) and then transferred to the oocyte during the latter part of oogenesis, where localization ensues. nos is translated initially in the nurse cells but becomes repressed when it enters the oocyte. Upon localization to the posterior of the oocyte, translation is reactivated (9). Translational repression of the unlocalized nos is mediated primarily by a 90nucleotide cis-acting translational control element (TCE) in the nos 3 untranslated region (3 UTR) (13)(14)(15). The nos TCE consists of a base stem, TCEI, connected to two stem loops, termed TCEII and TCEIII, which have temporally distinct activities (9,16). TCEI and TCEIII together confer translational regulation on nos during late oogenesis through their interaction with the hnRNP F/H protein Glorund (Glo) (16)(17)(18). TCEII confers repression on unlocalized nos during early embryogenesis through its interaction with a different RNA-binding protein, Smaug (Smg) (15,19).
Translational repression of unlocalized nos during late stages of oogenesis appears to be bipartite. Previous in vitro translation experiments showed that TCE-mediated repression of a reporter RNA in embryo extract was abolished if the reporter RNA was unable to bind eukaryotic translation initiation factor 4E (eIF4E) (20). This cap-dependence of repression suggests that the TCE acts to block initiation in the embryo and is consistent with the ability of Smg to recruit the Drosophila 4Ebinding protein (4E-BP), Cup, and disrupt the formation of the translation initiation complex eIF4F (21). By contrast, TCE-mediated repression of the same reporter in ovary extract was only partially abrogated (20), indicating that repression in the ovary is only cap-dependent in part. Furthermore, sucrose density gradient sedimentation showed that approximately 50% of nos is polysomeassociated whereas only 4% of nos is posteriorly localized and translationally active (22), consistent with repression at the elongation step. Compared to initiation control, an elongation-based mechanism could afford a more rapid on/off switch to shut down translation as nos enters the oocyte and then reactivate translation as it becomes localized.
Glo is the only known repressor of nos in the ovary, suggesting that it might mediate both the initiation and elongation-based repression mechanisms. Like its mammalian counterparts, Glo is a multi-functional RNAbinding protein. In addition to translationally repressing nos, Glo has been shown to repress the translation of nuclear-encoded mitochondrial respiratory chain complex mRNAs (23). Glo also impacts ovarian nurse cell chromatin organization and dorsal-ventral patterning, possibly acting as a regulator of alternative splicing of ovarian tumor mRNA through its interaction with Hrp48 and Hfp (17,24). Most recently, Glo has been implicated in triglyceride homeostasis in adults by regulating mRNAs encoding lipid transport proteins (25). Finally, Glo is required for viability to adulthood (17), although the target RNAs and regulatory mechanism used in this capacity are not known.
How a single protein, Glo, regulates multiple mRNA metabolic pathways and how it distinguishes regulatory needs among its RNA targets remain largely elusive. One possibility is suggested by Glo's dual modes of RNA recognition. Glo contains 3 highly conserved quasi-RNA recognition motifs (qRRMs), each of which harbors 2 RNA-binding interfaces with distinct sequence specificities, allowing different combinations of RNA-binding modes to select subsets of target RNAs (18). Apart from the qRRMs, Glo contains no known protein domains. Thus, the qRRMs may also act as protein-protein interaction interfaces to recruit different effector proteins that diversify Glo's regulatory activities. To gain further insight into Glo as a multi-faceted translational repressor and into qRRM function beyond RNA binding, we isolated Glo-interacting proteins from late-stage oocytes and investigated whether the interaction is mediated by one or more of Glo's qRRMs.
Here, we identify the Drosophila fragile X mental retardation protein, dFMRP, as a Glo-interacting protein required for translational repression of nos mRNA during late stages of oogenesis. We show that dFMRP negatively regulates nos translation in late-stage oocytes and that dFMRP associates with polysome-bound nos, consistent with its proposed role as a translation elongation repressor.
Using an in vitro translation run-off assay that distinguishes between effects on translation initiation and elongation, we show that Glo represses a reporter containing the nos TCE at both the initiation and elongation phases and that dFMRP specifically represses translation elongation. Puromycin release and [ 35 S]-methionine incorporation assays demonstrated that dFMRP stalls ribosomes on nos and decreases elongation rate, respectively. Analysis of engineered Glo variants lacking individual qRRMs revealed that qRRM2 is specifically required for dFMRP binding to Glo and consequently to nos in vivo. Furthermore, an in vitro binding assay using purified qRRMs and dFMRP showed that qRRM2 interacts directly with dFMRP. Taken together, these results support a model in which Glo represses translation elongation of nos by directly recruiting dFMRP through qRRM2.

Construction of transgenes and transgenic lines
The gfp-glo transgene was previously described (18). The transgene was maintained in flies homozygous for the glo 162x null allele (17), so that GFP-Glo constituted the only source of Glo protein. This transgene fully rescues glo 162x phenotypes (24). qRRM and CTX deletions were generated in the glo coding sequence by overlap extension PCR mutagenesis. All transgenes were inserted into the attP40 landing site by phiC31-mediated integration. glo 162x mutant germline clones expressing gfp-glo or gfp-gloΔqRRM transgenes were generated using the dominant femalesterile method (26). The gfp-dfmr1 transgenic line was previously described (27).

Collection of ovary and oocyte samples
To obtain late-stage-oocyte-enriched ovaries (late-ovaries), female flies were fed for 4 days at 25 • C in bottles of corn meal agar food supplemented with yeast paste and then starved in bottles with only wet Kimwipes (Kimtech) for 18 h at 25 • C. Ovaries were hand-dissected on ice in PBS and used immediately to prepare extract. For late-stage oocyte isolation, hand-dissected ovaries were teased apart, and late-stage oocytes were separated out according to morphology, pooled in PBS, and used immediately. The Oregon-R strain was used for preparation of wild-type tissues. dfmr1 RNAi tissues were produced by mat-tub-Gal4 > UAS-dfmr1RNAi (TRiP GL0075) females.

Immunoprecipitation and mass spectrometry (IP-MS)
Late-ovaries expressing GFP-tagged proteins were homogenized in IP buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 0.5% IGEPAL-CA630, 1x cOmplete EDTA-Free Protease Inhibitor Cocktail (Roche), 1 mM PMSF) supplemented with 20 l of RNase A/T1 Mix (Thermo Scientific) and 20 U of RNase ONE (Promega) per 1 ml of IP buffer and cleared by centrifugation at 13,200 rpm for 30 min at 4 • C. The supernatant was incubated with GFP-Trap A beads (Chromotek) at 4 • C for 1.5 h, and the beads were then washed with IP buffer without IGEPAL-CA630.
Nucleic Acids Research, 2022, Vol. 50, No. 12 7069 Immunoprecipitates were eluted in 0.1% (v/v) TFA. For crosslinked IP-MS, dissected ovaries were incubated in PBS containing 1 mM DSP (Thermo Scientific) or in PBS without DSP (control) for 30 min at room temperature, quenched by addition of Tris-HCl pH 7.5 to a final concentration of 20 mM, and incubated for 15 min. Extract was prepared and immunoprecipitation with GFP-Trap A beads was performed as described above, except that the beads were washed with IP buffer containing 1 M NaCl and no IGEPAL-CA630. Immunoprecipitates were eluted by boiling the beads in SDS-PAGE sample buffer (2% SDS, 62.5 mM Tris base, 10% glycerol, 100 mM DTT) which also cleaves the disulfide linker in DSP to reverse-crosslink. For both native and crosslinked IP-MS, LC-MS/MS was performed by the Proteomics and Mass Spectrometry Core at Princeton University. The resulting MS datasets were analyzed using Scaffold 4 (Proteome Software). Fold enrichment was defined as follows.
Native IP-MS: Fold enrichment = # total spectral counts i n G F P-Gl o I P # total spectral counts i n MC P-G F P I P

Crossliked IP-MS:
Fold enrichment = # total spectra count i n cr ossl i n ked (+ DSP) I P sample # total spectra count i n uncr ossl i nked (− DSP) contr ol

Immunoprecipitation of endogenous Glo
10 g of anti-Glo antibody (5B7) (17), or anti-GFP antibody (JL-8; Clontech) was crosslinked to 50 l of Dynabeads Protein G (Thermo Fisher) by dimethyl pimelimidate. Wild-type Drosophila late-ovary lysate was prepared, incubated with beads, and washed as described above (native IP-MS), except that the IP buffer contained 200 mM NaCl.

RNA co-immunoprecipitation
Late-ovary extracts were prepared as follows. 50  For RNA co-IP following polysome fractionation, RNP (fractions 1-5), monosomal (fractions 6-11), and polysomal (fractions 12-24) fractions were separately pooled. EDTA pH 8.0 was added to a final concentration of 50 mM. Diluted extract or pooled fractions were incubated overnight at 4 • C with 75 l of Dynabeads Protein G (Invitrogen) coupled to 5 g of mouse anti-GFP monoclonal antibody (3E6; Invitrogen). After incubation, beads were washed 4 times in ice-cold wash buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM MgCl 2 and 0.05% IGEPAL CA-630) and 2 times in wash buffer supplemented with 2 M urea. Washed beads were treated first with 100 l of wash buffer containing 8 l of RQ1 RNase-free DNase (Promega) for 15 min at 37 • C and then with 150 g/ml proteinase K (NEB) in the presence of 1% SDS for 30 min at 55 • C. RNA was extracted with phenol:chloroform, precipitated with isopropanol, and re-dissolved in 20 l of nuclease-free water.
For RNA extraction, fractions were treated first with 10 l of RQ1 RNase-free DNase for 30 min at 37 • C and then with 150 g/ml proteinase K in the presence of 1% SDS and 50 mM EDTA for 30 min at 55 • C, followed by phenol:chloroform extraction and isopropanol precipitation, and re-dissolved in 20 l of nuclease-free water.

Puromycin release and polysome analysis
100 pairs of late-ovaries were dissected in PBS (-Puro) or PBS containing 3 mM puromycin (+Puro; Sigma-Aldrich) on ice. Dissected ovaries were transferred to 500 l of lysis buffer (50 mM Tris-HCl pH 7.5, 250 mM NaCl, 5 mM Mg(OAc) 2 , 0.2% Triton X-100, 2 mg/ml heparin, 1 mM DTT, 200 U/ml SUPERase·In and 1× cOmplete EDTA-Free Protease Inhibitor Cocktail; -Puro) or lysis buffer containing 3 mM puromycin (+Puro) and incubated at 37 • C for 15 min. Ovaries were then homogenized and incubated at 4 • C for 15 min with constant mixing, incubated at 37 • C for 10 min, and supplemented with cycloheximide (1 mg/ml final concentration) and Mg(OAc) 2 (50 mM final concentration). The extract was cleared by centrifugation at 13,200 rpm for 15 min at 4 • C, and the supernatant was layered onto a 12 ml, 10-50% sucrose gradient in 50 mM Tris-HCl pH 7.5, 250 mM NaCl, and 50 mM Mg(OAc) 2 . Gradients were centrifuged in a Beckman SW41Ti rotor at 36,000 rpm for 1.5 h at 4 • C and were fractionated and RNA extracted as described in Polysome fractionation, except that 2 l of 1 nM Renilla luciferase RNA was added to each fraction prior to RNA extraction as internal control for subsequent RT-qPCR analysis.

RT-PCR
For analysis of RNA from immunoprecipitates and from sucrose gradient fractions, respectively, 1/2 and 1/10 of the total amount of RNA purified was reverse transcribed using SuperScript III Reverse Transcriptase (Invitrogen) with the following condition. Annealing: 1 l of 50 M oligo d(T) 20 (Invitrogen), 1 l of 10 mM dNTP Mix (Invitrogen), 2 l (gradient fraction) or 10 l (RNA co-IP) of extracted RNA was combined in a total volume of 13 l, incubated at 65 • C for 5 min, and chilled on ice for 2 min. First-strand cDNA synthesis: 4 l of 5× First-Strand Buffer (Invitrogen), 1 l of 0.1 M DTT, 1 l of SUPERase·In, and 1 l of SuperScript III RT (200 units/l; Invitrogen) was added to a total volume of 20 l, then incubated at 50 • C for 50 min and treated with 1 l of RNase H (NEB) at 37 • C for 15 min. 2 l of the RT products were PCR-amplified using standard Taq protocol with the following parameters. Annealing temperature: 54 • C for his3.3b and 57 • C for nos and act5c; extension time: 90 s for his3.3b, 60 s for nos, and 15 s for act5c; cycle number: 25 for his3.3b and act5c and 28 for nos. PCR products were resolved on a 1% (his3.3b and nos) or a 1.5% (act5c) agarose gel. For analysis of nos distribution in sucrose gradient, ethidium bromide-stained gels were imaged on an iBright FL1000 Imaging System (Thermo Fisher), and band intensity was quantified using the built-in analysis function. Band intensity of each fraction was normalized to the total intensity of all fractions, therefore the distribution result is independent of absolute transcript levels of nos. The following primers were used: his3.3b-F:

RT-qPCR
For analysis of oocyte samples, RNA was extracted by homogenizing late-stage oocytes in TRIzol (Invitrogen) followed by chloroform extraction and isopropanol precipitation. 1 g of total RNA was used for reverse transcription as described above. For both oocyte and gradient samples, 2 l of cDNA was combined with 10 l of 2× SYBR Green PCR Master Mix (Applied Biosystems) and 300 nM forward and reverse primers (final concentration) in a total volume of 20 l. qPCR was performed on a StepOnePlus Real-Time PCR System (Applied Biosystems). Two (gradient samples) or three (oocyte samples) biological replicates were analyzed with three technical replicates each, all using auto thresholding and the following parameters: Holding, 95 • C for 10 min; Cycling, 95 • C for 15 s, 60 • C for 1 min, repeated for 39 cycles; Melt curve, 95 • C for 15 s, 60 • C for 1 min, + 0.3 • C step & hold, 95 • C for 15 s. rpl32 and rluc were used as internal control for oocyte and gradient samples, respectively, unless otherwise stated. Relative levels of other target transcripts were calculated as follows (Ct of un-detected sample was set as 40). The following primers were used: rpl32-F:

Generation of reporter and competitor RNAs for in vitro translation assay
The luciferase reporter RNAs were constructed in a derivative of a BSK-A vector containing the firefly luciferase coding sequence followed by 73 adenosine residues (kindly provided by F. Gebauer and M. W. Hentze). For firefly luciferase (Fluc) reporters, nos 5 UTR sequence was sub-cloned from the pA25-nos5 UTR plasmid (9) and inserted between the SacI and SmaI sites. 3xTCE, 3xTCEIIIA, 3xSREand tub 3 UTR sequences were subcloned from respective plasmids (9) and inserted between the BglII and BamHI sites. For the Renilla luciferase (Rluc) reporter, the SmaI-BamHI fragment containing the firefly luciferase coding sequence in the BSK-A derivative plasmid was replaced by the Renilla luciferase coding sequence. HAtagged Fluc reporters were generated as follows. The nos 5 UTR sequence was sub-cloned from the pA25-nos5 UTR plasmid (9) and inserted between the SacI and NcoI sites of a FLAG-3xHA vector (kindly provided by T. Aoki and P. Schedl). Firefly luciferase coding sequence with 3xTCE or 3xTCEIIIA sequences followed by 73 adenosine residues was sub-cloned from the respective Fluc reporter plasmids (described above) and inserted between the SmaI and HindIII sites. Capped reporter RNA was transcribed from 1 g of HindIII-linearized plasmid DNA using Nucleic Acids Research, 2022, Vol. 50, No. 12 7071 the mMessage mMachine T3 Transcription kit (Ambion). TCEII and TCEIII competitor RNAs were generated from BamHI-linearized plasmids containing individual TCE stem-loop II and TCE stem-loop III sequences (17), respectively, using the MEGAscript T7 Transcription kit (Ambion). In all cases, in vitro transcription reactions were treated with 1 l of TURBO DNase (Ambion) for 15 min at 37 • C, and RNA was extracted with phenol:chloroform and precipitated first with isopropanol and then with ethanol. Prior to use, the competitor RNA was de-natured at 65 • C for 5 min and slowly cooled to room temperature (17).

In vitro translation and translation run-off assay
Late-ovary translation extract was prepared as previously described (20) except that 1× cOmplete EDTA-Free Protease Inhibitor Cocktail was used in place of 0.5 mM PMSF. Extract was made, aliquoted, flash frozen in liquid N 2 , and stored at -80 • C for no more than 2 weeks before use.
The in vitro translation reaction (20 l) contained 50% late-ovary translation extract, 1× translation mix (25 mM HEPES-NaOH pH 7.5, 3 mM Mg(OAc) 2 , 2.5 mM DTT, 25 M Complete Amino Acid Mixtures (Promega), 1.2 mM ATP, and 0.3 mM GTP), energy regeneration system (15 mM creatine phosphate and 0.1 mg/ml creatine phosphokinase (Sigma-Aldrich)), 0.1 nM Renilla luciferase reporter RNA (internal control), and 1 nM firefly luciferase reporter RNA. Reactions were incubated at 28 • C for 2 h. Additional external control reactions (25 l) were assembled as follows. The same firefly-Renilla luciferase reporter pair (0.1 nM each, final concentration) was incubated in 17.5 l of Nuclease-Treated Rabbit Reticulocyte Lysate (Promega), 40 M Complete Amino Acid Mixtures, and 1 l of Human Placenta RNase Inhibitor (NEB). The reaction was allowed to proceed for 90 min at 28 • C. Luciferase activity was assayed using the Dual-Luciferase Reporter Assay System (Promega), and luminescence was measured by a Glomax 20/20 Luminometer (Promega) using the built-in DLR protocol. The relative reporter activity and fold repression were defined as follows. Data were averaged over three biological replicates using different preparations of extract.
For the translation run-off assay, the reaction mixture was assembled without the energy regeneration system, pre-incubated at 16 • C for 5 min. Either 0.1 l of 10 mg/ml 4E1RCat (in DMSO; Sigma-Aldrich) (the final concentration of 4E1RCat is 50 g/ml which effectively blocks 80S ribosome formation as determined by the 80S formation assay below) or 0.1 l of DMSO was added, followed by 1 l of TCEII or TCEIII competitor RNA (2 g/l) or nuclease-free water. Finally, the energy regeneration system was added, and the reaction was incubated at 28 • C for 2 h. Fold repression was calculated as described above except that the value was normalized to the value of 3xTCEIIIA without 4E1RCat or competitor RNA treatment. The data were averaged over three biological replicates using different preparations of extract.

80S formation assay
10 l of 10 nM Rluc reporter was incubated in a total volume of 100 l of in vitro translation reaction mixture containing 50% Drosophila ovary translation extract (as described above) in the presence of 1 mg/ml cycloheximide. 0.5 l of 10 mg/ml 4E1RCat (+4E1RCat) or DMSO (-4E1RCat) was added, and the mixture was incubated at 25 • C for 30 min followed by polysome fractionation. 2 l of 1 nM HA-Fluc reporter RNA was added to each fraction as an internal control for subsequent RT-qPCR analysis, and RNA from each fraction was prepared as described in Puromycin release and polysome analysis. PerkinElmer NEG709A001MC), and 10 l of nucleasefree water were pre-incubated at 25 • C for 10 min. The concentration of unlabelled methionine was optimized to sustain efficient elongation while retaining sufficient [ 35 S]-Met incorporation. 10% of initial volume was removed at 0 and 5 min for anti-HA IP and scintillation counting. At 10 min, 2.5 l of 10 mg/ml 4E1RCat, 15 l of 100 mM creatine phosphate, and 2 l of 5 mg/ml creatine phosphokinase were added to the reaction mixture to block de novo initiation and start translation run-off. For experiments involving addition of dFMRP or BSA, 10 l of 1 mg/ml purified NT-dFMRP (see Protein expression and purification below) or 1 mg/ml BSA (Sigma-Aldrich) was also added to the reaction mixture at 10 min. Additional aliquots equivalent to 10% of initial volume were removed at 10, 15, 20, 30, 45, 60, 90 and 120 min for anti-HA IP and scintillation counting. For anti-HA IP, aliquots were incubated with 5 l of Pierce Anti-HA Magnetic Beads (Thermo Fisher) in 250 l of RIPA buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% IGEPAL CA-630, 1% sodium deoxycholate (Sigma-Aldrich) and 0.1% SDS) containing 1 mM methionine at room temperature for 1 h with constant mixing. The beads were washed three times in 500 l of TBST containing 0.1% SDS, re-suspended twice with 500 l of UniverSol-ES Liquid Scintillation Cocktail (MP Biomedicals), and radioactivity was quantified by liquid scintillation counting in a TRI-CARB 4810TR Liquid Scintillation Counter (PerkinElmer) using the built-in S-35 counting protocol. Values for each time-point were averaged over two (+dFMRP and +BSA) or three (all others) biological replicates using different preparations of extract and were normalized to the values for the TCEIIIA reporter at 120 min.
For purification of individual qRRM domains, E. coli cultures were grown in LB medium to an OD 600 of 0.6-0.7 and induced overnight with 0.5 mM IPTG at 16 • C. The cells were pelleted, resuspended in buffer containing 50 mM Tris-HCl pH 8.0, 500 mM NaCl and 2 mM PMSF, and lysed using a cell disrupter. Cell lysate was cleared by centrifugation and the qRRM proteins were purified by Ni-NTA agarose (Qiagen) chromatography. Eluates were further purified by sizeexclusion chromatography (Superdex 200 Increase 10/300, GE Healthcare), concentrated, flash-frozen, and stored at -80 • C in 20 mM HEPES-NaOH pH 8.0 and 200 mM NaCl (qRRM1 and qRRM3) or 20 mM HEPES-NaOH pH 8.0 and 250 mM NaCl (qRRM2).
For purification of NT-dFMRP, E. coli culture was grown in LB medium to an OD 600 of 0.4-0.5 and induced overnight with 0.5 mM IPTG at 16 • C. The cells were pelleted, re-suspended in buffer containing 50 mM Tris-HCl pH 8.0, 500 mM NaCl, 20 mM imidazole, 0.1% CHAPS, 5 mM ␤-mercaptoethanol, and 2 mM PMSF. The cells were lysed using a cell disruptor, and the lysate was cleared by centrifugation and applied to Ni-NTA agarose chromatography. The eluted protein was further purified by size-exclusion chromatography (Superdex 200 Increase 10/300), concentrated, flash-frozen, and stored at -80 • C in 20 mM HEPES-NaOH pH 8.0, 250 mM NaCl and 2 mM DTT.

His 6 -tag pull-down assay
Purified individual qRRMs (50 g each) were immobilized on Ni-NTA agarose beads separately, washed in 20 mM Tris-HCl pH 7.5, 500 mM NaCl and 0.5% IGEPAL CA-630, and incubated with late-ovary extract (prepared as described in the IP-MS section) expressing GFP-dFMRP for 1.5 h at 4 • C. The beads were washed three times with IP buffer (above) and eluted in boiling SDS-PAGE sample buffer. The IP products were resolved on a 15% SDS-PAGE gel, transferred onto nitrocellulose membrane, and detected by Ponceau S staining (qRRMs) or anti-GFP immunoblotting (GFP-dFMRP).

Analytical size-exclusion chromatography binding assay
Proteins were diluted to 2.5 M ( NT-dFMRP) or 20 M (individual qRRM domains) in 20 mM HEPES-NaOH pH 8.0, 150 mM NaCl, and 5 mM TCEP in a volume of 60 l, incubated at 4 • C for 48 h, and then loaded onto a Superdex 200 Increase 3.2/300 column. The fractions were collected, and the proteins were concentrated by TCA-precipitation, resolved on a 15% SDS-PAGE gel, and visualized by staining with Coomassie Brilliant Blue R-250.

Statistics
All statistical analyses were performed in GraphPad Prism 9.0.2. A two-tailed Welch t-test was used for comparisons of two means, and post hoc Holm-Sidak test was used for multiple comparison correction when necessary. Brown-Forsythe and Welch one-way ANOVA with post hoc Dunnett T3 test was used for one-dimensional comparisons of more than two means. Two-way full-model ANOVA with post hoc Tukey test was used for twodimensional comparisons of means within and between genotype groups. Linear regressions of [ 35 S]-methionine incorporation data were performed using the Simple Linear Regression function which also tested the significance of the difference between slopes. Data were plotted as mean or mean ± SEM from two or three biological replicates as indicated in the figure legends. Individual data points are shown unless for visibility reasons.

Identification of dFMRP as an RNA-independent Glointeracting protein
To understand how Glo -which contains only annotated RNA-binding domains -represses nos translation, we isolated Glo interacting proteins from ovaries enriched for late-stage oocytes (late-ovaries) (20) by immunoprecipitation coupled to mass spectrometry (IP-MS). Anti-GFP immunoprecipitation was performed using late-ovaries expressing functional GFP-tagged Glo (GFP-Glo) in the absence of endogenous, native Glo. Late-ovaries expressing GFP-tagged bacteriophage MS2 coat protein (MCP-GFP) were used as a control ( Figure 1A). Ovary extracts were treated with RNase prior to immunoprecipitation to eliminate RNA-dependent protein interactions. Mass spectrometry (MS) analysis identified 94 proteins, including Glo itself, that were at least 2-fold enriched in GFP-Glo immunoprecipitates relative to MCP-GFP immunoprecipitates (Supplementary  Table S1). To facilitate recovery of potentially weak and transient Glo-interactors, we also performed crosslinkingbased IP-MS under stringent conditions ( Figure 1B) and detected 170 proteins that were at least 2-fold enriched in the GFP-Glo immunoprecipitates from the crosslinked sample compared to those from the un-crosslinked control (Supplementary Table S1). A single experiment was performed for each condition. 23 interactors were common to both the native and the crosslinked IP-MS experiments and were therefore considered high-confidence Glo-interacting proteins ( Figure 1C; Supplementary  Table S1). Among the high-confidence interactors are eIF4E and ribosomal subunits, consistent with Glo's implicated functions in regulating translation initiation and elongation of nos, respectively. The list also contains known translation regulators, including dFMRP (29,30), Caprin (31) and Lingerer (32,33), consistent with the function of Glo as a translational repressor more generally. Our high-confidence interactors (11 out of 23) overlap those identified in a recently reported Glo native IP-MS experiment using non-staged ovaries (34) (Figure 1D; Supplementary Table S1), further supporting the validity of the interactors identified here.
We focused on dFMRP because studies of dFMRP and its mammalian homologs have identified multiple modes of action in translational control, including inhibition of translation initiation (35,36) and ribosome stalling (28,37). Interaction of Glo and dFMRP in late-ovaries was further tested by reciprocal in vivo co-immunoprecipitation (co-IP) in the presence of RNase. Endogenous dFMRP and Glo were detected in anti-GFP immunoprecipitates from late-ovaries expressing GFP-dFMRP or GFP-Glo, respectively, but were not detected in control immunoprecipitates ( Figure 1E). Endogenous dFMRP also co-immunoprecipitated with endogenous Glo (Supplementary Figure S1), confirming that dFMRP is an RNA-independent Glo-interacting protein during late oogenesis.

dFMRP negatively regulates Nos protein levels
We next sought to determine whether dFMRP, like Glo, participates in nos regulation. To test whether dFMRP interacts with nos mRNA in vivo, we performed RNA co-IP experiments using wild-type, GFP-Glo, or GFP-dFMRP late-ovaries. nos could be detected by RT-PCR in both GFP-Glo and GFP-dFMRP immunoprecipitates, whereas a control maternal transcript, his3.3b, was not detected (Figure 2A). To probe a requirement for dFMRP in translational repression of nos, we knocked down dfmr1 expression in the female germline. Expression of either of two independent UAS-dfmr1 RNAi transgenes with a maternal-tubulin-Gal4 driver resulted in almost complete knock-down (KD) of dFMRP protein expression (un- detectable by western blotting, Figure 2B). For both RNAi lines, Nos protein levels were increased by ∼4.5-fold in latestage ovaries ( Figure 2B) as compared to wild-type. This upregulation was not a result of a decrease in the amount of Glo protein ( Figure 2B) or an increase in nos mRNA levels ( Figure 2C). These results suggest that dFMRP interacts with nos to negatively regulate its translation during late oogenesis. Since the dfmr1 RNAi lines behaved similarly, we used the TRiP GL0075 line in all subsequent experiments requiring dFMRP KD.

dFMRP reduces ribosome load on nos
To determine how dFMRP regulates nos translation, we first performed sucrose density gradient sedimentation to monitor ribosome load on nos in wild-type and in dFMRP KD late-ovaries. We hypothesized that if dFMRP only affects translation initiation of nos mRNA, dFMRP KD would increase polysome association of nos but leave its distribution within the polysomal fractions unchanged. In contrast, if dFMRP plays a role in regulating translation elongation, dFMRP KD would affect the number of ribosomes on nos mRNA, hence the relative level of nos in light (fewer ribosomes) versus heavy (more ribosomes) polysome fractions. Consistent with previous studies (22,38), the overall polysome signal was low ( Figure 3A), presumably due to generally low translational activity during the later stages of oogenesis. While the overall polysome profile did not show an obvious change upon dFMRP KD, there was a redistribution of nos mRNA toward the heavier fractions ( Figure 3B). Specifically, the shift of nos mRNA from lighter to heavier polysome fractions suggests that dFMRP regulates translation elongation of nos mRNA. In addition, we observed a moderate increase in the percentage of polysome-associated nos, from ∼50% in wild-type, which is consistent with our previous findings (20,22), to ∼65% in dFMRP KD late-ovaries ( Figure 3C), indicating either that dFMRP also represses nos translation initiation or that initiation-based repression plays a secondary role and, by itself, does not completely eliminate ribosome loading onto nos mRNA. Together with the results presented above (Figure 2), these data provide further evidence that dFMRP is a translational repressor of nos.

dFMRP represses translation elongation of nos through the TCE
To determine more definitively if dFMRP participates in one or both modes of TCE/Glo-mediated nos repression, we optimized a previously established in vitro translation system (9,20) based on late-ovary extract that confers TCEmediated repression on a reporter RNA and performed a translation run-off assay. First, we compared the ability of extracts from wild-type and dFMRP KD late-ovaries to support TCE-mediated repression of a firefly luciferase (Fluc) reporter RNA. Extracts were programmed with capped and polyadenylated Fluc fused to either 3 tandem copies of the wild-type nos TCE (Fluc-3xTCE) or the unregulated α-tubulin (tub) 3 UTR (Fluc-tub3 UTR), along with Renilla luciferase (Rluc) RNA that served as an internal control for experimental variability ( Figure 4A, B). In the wild-type extract, Fluc-3xTCE RNA was repressed 9-fold compared to the unregulated Fluc-tub3 UTR RNA ( Figure 4C). By contrast, only 4.5-fold repression of Fluc-3xTCE RNA was observed for the dFMRP KD extract ( Figure 4C). As we showed previously, the TCEIIIA mutation, which disrupts Glo binding and nos repression in vivo (9,17), completely abolished repression of the reporter RNA in the wildtype extract (Fluc-3xTCEIIIA), whereas repression was not affected by mutation of the Smg-binding site in TCEII (Fluc-3xSRE -) ( Figure 4B, C). Similarly, the TCEIIIA mutation eliminated repression of the reporter in the dFMRP KD extract, whereas the SREmutation had no effect ( Figure 4C). These results indicate that the in vitro translation system recapitulatea nos regulation in vivo and that dFMRP accounts in part for Glo/TCE-mediated nos translational repression.
Next, we performed a translation run-off assay to biochemically dissect nos regulation at the translation initiation and elongation phases ( Figure 4D, also see Materials and Methods). Reporter RNA was first incubated in late-ovary extract without the ATP/GTP regeneration system to allow the assembly of the repression complex and the initial loading of ribosomes (39). Subsequently, the translation initiation inhibitor 4E1RCat, which competitively disrupts the interaction of eIF4G with eIF4E and hence the formation of the eIF4F complex (40), was added to block de novo initiation on the reporter RNA. 4E1RCat reduced the absolute expression level of luciferase reporters in a dosage-dependent manner (Supplementary Figure S2A) but did not affect TCE-mediated translational repression (Supplementary Figure S2B). Sucrose gradient analysis confirmed that 4E1RCat effectively prevented 80S ribosome formation on the reporter RNA in late-ovary extract (Supplementary Figure S2C, D). Competitor TCEIII RNA was then added to titrate the TCE-bound repression complex from the reporter RNA. Finally, the ATP/GTP regeneration system was provided to resume Renilla luciferase coding sequence did not contain specific 3 UTR sequence and was not subject to translational regulation. Both reporters contained a 5 7-methylated guanosine (m 7 G) cap and 3 poly (A) 73 tail. (B) Schematic representation of the nos TCE. Glo binds to both TCEI and TCEIII, whereas Smg binds to TCEII. The TCEIIIA and SREmutations disrupt Glo and Smg binding, respectively. (C) In vitro translation assay using reporters shown in (A). For each indicated reporter, fold repression is the inverse of the normalized relative luciferase activity, which was calculated as the ratio of firefly to Renilla luciferase activity in late-ovary extract normalized first to activity in rabbit reticulocyte lysate and then to the value for the tub3 UTR reporter. (D) Experimental setup of the translation run-off assay. Reporter RNAs were pre-incubated in late-ovary extract without the energy regeneration system (t 0 ). The translation initiation inhibitor 4E1RCat (50 g/ml) was then added to block de novo initiation (t 1 ), followed by the addition of competitor RNAs (t 2 ; control TCEII or specific TCEIII). Finally, the energy regeneration system was added back to restore translation (t 3 ), and luciferase activity was assayed (t 4 ). (E) Results of the translation run-off assay in wild-type (WT) and dFMRP KD ovary extracts. Normalized relative luciferase activity and fold repression were calculated as in (C), except that the luciferase activity was normalized to the value for TCEIIIA (without 4E1RCat or competitor RNA  TCE+TCEIII . In (C), (E) and (F), values are mean ± SEM from three biological replicates from different ovary extract preparations. Statistical significance was analyzed using two-way full-model ANOVA with post hoc Tukey test (in C), Brown-Forsythe and Welch one-way ANOVA with post hoc Dunnett T3 test (in E, significance symbol for wildtype/dFMRP KD), and two-tailed Welch t-test with post hoc Holm-Sidak test (in F); *P < 0.05, **P < 0.01 and ***P < 0.001, not all tested P-values are shown.
translation. In this way, we were able to directly test if the nos TCE imposes blocks at both translation initiation and elongation, manifested by partial de-repression by TCEIII in the presence of 4E1RCat, and to de-couple the two repression modes to assess the impact of dFMRP on each independently.
As expected, addition of TCEIII competitor RNA to the wild-type extract almost fully relieved repression of the Fluc-3xTCE reporter (1.1-fold repression for Fluc-3xTCE + TCEIII versus 9.0-fold for Fluc-3xTCE), whereas a control competitor, TCEII RNA, had no effect as expected (20) (Figure 4E). In the presence of 4E1RCat, addition of the TCEIII competitor, but not TCEII, partially de-repressed Fluc-3xTCE RNA (2.6-fold repression for Fluc-3xTCE + TCEIII + 4E1RCat versus 9.0-fold for Fluc-3xTCE, 3.6-fold up-regulated; Figure 4E). Since de novo initiation was inhibited, the 3.6-fold increase in reporter activity reflects the elimination of a TCE-mediated elongation block. Maximal de-repression was observed only when initiation was allowed to occur (-4E1RCat), however, indicating that full de-repression resulted from elimination of both an initiation and an elongation block ( Figure 4E). Therefore, the remaining repression when initiation was prevented (2.6-fold for + TCEIII, +4E1RCat versus 1.1-fold for +TCEIII, -4E1RCat, 2.3-fold repressed) corresponds to a TCE-mediated initiation block ( Figure  4E). In summary, TCE-based regulation in wild-type extract comprises 2.3-fold repression of initiation and 3.6fold repression of elongation ( Figure 4F).

dFMRP stalls ribosomes and decreases elongation rate on nos
The substantial polysome association of nos in late-ovaries and the specific inhibition of nos translation elongation by dFMRP led us to hypothesize that dFMRP stalls elongating ribosomes on nos. To test if dFMRP promotes ribosome stalling on nos, we first measured puromycin release of ribosomes on endogenous nos in wild-type and dFMRP KD late-ovaries. Puromycin releases actively elongating, but not stalled ribosomes and caused an overall loss of polysomes in both wild-type ( Figure 5A) and dFMRP KD ( Figure 5B) late-ovaries. The polysome profile of nos, however, responded differently to puromycin treatment in the presence or absence of dFMRP. In wild-type late-ovary extract, puromycin treatment did not cause significant loss of polysomal nos ( Figure 5C, +Puro, Fractions 10 to 24), as compared to the mock-treated extract ( Figure 5C, -Puro). In contrast, significant redistribution of nos from heavier to lighter polysomal fractions was observed in puromycin-treated dFMRP KD extract ( Figure 5D, +/-Puro). The differential puromycin release of ribosomes from nos in wild-type and dFMRP KD extract was not a global effect of dFMRP KD, as ribosomes on a control maternal transcript, act5c, were released to a similar extent in both wild-type (Supplementary Figure  S3A-C) and dFMRP KD (Supplementary Figure S3D-F) late-ovary extracts. Thus, these results indicate that dFMRP stalls elongating ribosomes on nos.
Next, we sought to directly assess if the elongation rate on nos is affected by dFMRP in a TCE-dependent manner. To do this, we modified the reporter RNAs to encode an N-terminal triple HA epitope tag (HA-Fluc-3xTCE and HA-Fluc-3xTCEIIIA), allowing immunoprecipitation of nascent polypeptide chains with anti-HA antibody. We then performed an in vitro translation run-off assay with each reporter RNA in the presence the initiation inhibitor 4E1RCat and [ 35 S]-Met. Following anti-HA immunoprecipitation, incorporation of [ 35 S]-Met into nascent polypeptide chains was monitored by liquid scintillation counting. Because of the low total Met concentration necessary for labeling with [ 35 S]-Met (see Materials and Methods), the absolute elongation rate could not be easily estimated. However, we were able to compare qualitatively the elongation rates on the HA-Fluc-3xTCE and HA-Fluc-3xTCEIIIA RNAs and to determine if the difference in elongation rates depends on dFMRP. Translation run-off with HA-Fluc-3xTCE RNA resulted in ∼50% less [ 35 S]-Met incorporation than with HA-Fluc-3xTCEIIIA in the wild-type late-ovary extract ( Figure 5E) but not in the dFMRP KD extract ( Figure 5F). This is consistent with the luciferase assay results (Figure 4). We fitted the relative [ 35 S]-Met incorporation during first 20 min of the run-off, when neither [ 35 S]-Met nor elongating ribosomes were rate-limiting, to a linear model. After regression, the slope for HA-Fluc-3xTCEIIIA RNA in the wild-type extract was signifcantly steeper than that for HA-Fluc-3xTCE, indicating a faster elongation rate for the mutant reporter RNA ( Figure 5G). However, the elongation rates for the two reporters were indistinguishable in the dFMRP KD extract ( Figure 5H), consistent with dFMRP impeding elongation. Moreover, adding purified dFMRP, but not BSA, to the dFMRP KD extract slowed [ 35 S]-Met incorporation during translation run-off and led to an earlier plateau for HA-Fluc-3xTCE, but not for HA-Fluc-3xTCEIIIA RNA ( Figure 5I, S4). Taken together, results from polysome profiling and [ 35 S]-Met incorporation analyses demonstrate that dFMRP stalls elongating ribosomes on nos via the TCE.

qRRM2 is necessary and sufficient for direct recruitment of dFMRP by Glo
Given that Glo lacks known functional domains other than the qRRMs, we hypothesized that Glo interacts with dFMRP through one or more of its qRRMs. To test this, a series of Glo deletion variants were made by removing individual qRRMs. Two of Glo's qRRMs, qRRM1 and qRRM2, have additional alpha helical extensions (CTX) C-terminal to the core RRM fold (18). Since similar CTX structures have been shown to serve as proteinprotein interaction platforms in other RRM-containing proteins (41), we also generated variants lacking the CTX residues either individually or in combination ( Figure 6A). The deletions were engineered in a functional GFP-Glo transgene, and the resulting transgenes were each integrated at the same chromosomal location.
We then performed anti-GFP IPs from late-ovary extracts to determine whether any of the deletions affected the ability of Glo to interact with dFMRP. Western blot analysis showed that dFMRP co-immunoprecipitated with all Glo variants except Glo qRRM2 ( Figure 6B), indicating that Glo specifically requires qRRM2 to bind to dFMRP. Notably, RNA co-IP analysis showed that all of the deletion variants including qRRM2 bound nos (Supplementary Figure S5), confirming that protein structure is maintained and that, consistent with previous biochemical analysis (18), two qRRMs are sufficient for TCE recognition.
Next, we asked if qRRM2 is sufficient for dFMRP interaction. We purified recombinant hexahistidine-tagged qRRMs (18) (Figure 6A) from E. coli using Ni-NTA agarose chromatography followed by size-exclusion chromatography. Following purification, qRRMs were immobilized on Ni-NTA agarose and incubated with lateovary extract expressing GFP-dFMRP. qRRM2, but not qRRM1 or qRRM3, was able to pull down GFP-dFMRP ( Figure 6C), indicating that qRRM2 alone is sufficient for Glo-dFMRP interaction.
We then sought to determine if dFMRP binds directly to qRRM2. We expressed and purified a hexahistidine-tagged N-terminally truncated dFMRP protein ( NT-dFMRP), which lacks the first 219 amino acid residues containing the nuclear localization signal but is folded and retains full translational repression capability as determined by both cryo-EM and in vitro translation (28). Purified NT-dFMRP was incubated with each of the purified qRRM proteins and the mixture was subjected to size-exclusion chromatography. When chromatographed alone, each Glo qRRM behaved as a monomer. However, when mixed with dFMRP, a proportion of the Glo qRRM2 protein eluted in higher molecular weight fractions, together with NT-dFMRP ( Figure 6D). Neither Glo qRRM1 nor qRRM3 proteins were shifted to higher molecular weight fractions (Supplementary Figure S6). Together, our results show that interaction of Glo with dFMRP is mediated directly and specifically by qRRM2.
Finally, we tested if Glo is required for recruitment of dFMRP to nos by RNA co-IP. dFMRP was immunoprecipitated from glo null mutant late-ovaries expressing either GFP-Glo or individual GFP-Glo qRRM variants, using anti-FMRP antibody. nos mRNA was detected by RT-PCR in all dFMRP immunoprecipitates except when the only Glo protein expressed was GFP-Glo qRRM2 ( Figure 6E). The failure of nos to co-immunoprecipitate with dFMRP when Glo lacks qRRM2 indicates that qRRM2, and thus the interaction between Glo and dFMRP, is required for dFMRP binding to nos.

DISCUSSION
The Drosophila hnRNP F/H homolog, Glo, represses the translation of unlocalized nos during late stages of oogenesis to ensure proper anterior-posterior patterning of the embryo. Previous work suggested that Glo imposes both an initiation and a post-initiation block on nos translation (20,22), but the molecular mechanism and the nature of the post-initiation block remained unknown. Moreover, how a single protein composed primarily of RNA-binding domains can confer repression by two different mechanisms is unclear. Here, we have begun to fill these gaps by identifying dFMRP as a Glo-interacting protein during late stages of oogenesis and biochemically dissecting its role in nos regulation. Our results lead to a new model wherein Glo recruits dFMRP specifically through qRRM2 to repress elongating ribosomes on nos ( Figure 7) and suggest that Glo's distinct activities are dictated by different effector proteins recruited through different qRRMs.
Since the discovery of FMRP, numerous highthroughput analyses along with transcript-specific studies have uncovered a multiplicity of mechanisms by which it is thought to regulate translation in both neuronal and non-neuronal contexts. These include repression of cap-dependent translation initiation (35,36), modulation of microRNA activity (42,43), and stalling of elongating ribosomes (28,37). Among its large set of targets, FMRP is capable of binding to a wide variety of sequence motifs (44), further complicating the understanding of its target selectivity. In this work, we show that an FMRPinteracting protein recognizing an mRNA signature specific to an individual target mRNA can provide the required target selectivity.
A recent ribosome profiling study found that dFMRP activates translation of a collection of stored mRNAs, particularly those encoding large autism-related proteins, in Drosophila oocytes (45). Our work, however, suggests that dFMRP, recruited by specificity factors, can also function as translational repressor of individual targets in oocytes by blocking translation elongation. Indeed, dFMRP has been reported to interact with the 60S ribosomal protein L5, which may prevent tRNA binding during elongation (28). The Glo/TCE-dependent elongationrepressing activity of dFMRP together with the increased ribosome load following dFMRP KD leads to two nonexclusive hypotheses: 1) dFMRP can repress nos translation initiation independently of Glo/TCE and/or 2) in the absence of the dFMRP-mediated elongation block, the initiation block, regardless of its dFMRP dependency, is insufficient for complete shutdown of translation. Although we cannot yet distinguish among these possibilities, the latter is consistent with the relative magnitude of the initiation versus elongation-based repression (2.4-fold versus 3.6-fold) measured by our translation run-off assay ( Figure 4F).
Whereas Glo-RNA interactions have been characterized in some detail (18), much less is known about Glo-protein interactions and how they contribute to Glo's functional Figure 6. Glo qRRM2 is necessary and sufficient for direct recruitment of dFMRP. (A) Schematic representation of Glo protein variants analyzed. Green: GFP tag; yellow: qRRM1; blue: qRRM2; orange: qRRM3; black: His 6 tag. qRRM1: deletion of residues 47-136; qRRM2: deletion of residues 145-232; qRRM3: deletion of residues 478-559; CTX1: deletion of residues 129-136; CTX2: deletion of residues 226-232; CTX1 + 2: deletion of residues 129-136 and 226-232. His 6 -qRRM1: Glo residues 45-141; His 6 -qRRM2: Glo residues 142-234; His 6 -qRRM3: Glo residues 475-562. (B) Anti-dFMRP and anti-GFP immunoblots of input extracts (top two panels) and anti-GFP immunoprecipitates (bottom two panels) from late-ovaries expressing GFP or GFP-Glo variants. dFMRP co-immunprecipitated with all Glo variants except Glo qRRM2. (C) Pull-down assay with His 6 -qRRMs. Top panel, anti-GFP immunoblot to detect GFP-dFMRP (dFMRP) in input extract and after pull-down with qRRM1, qRRM2, or qRRM3. Bottom panel, Ponceau S staining of His 6 -qRRMs in input extract and after pull-down. Only qRRM2 was able to pull down GFP-dFMRP. (D) Analytical size-exclusion chromatography binding assays. For chromatograph (top), red profile: 2.5 M His 6 -NT-dFMRP ( NT-dFMRP) alone (the same chromatograph is also used in Supplementary Figure S6A, B); blue profile: 20 M His 6 -qRRM2 (qRRM2) alone; green profile: 2.5 M NT-dFMRP + 20 M qRRM2 ( NT-dFMRP + qRRM2). Coomassie stained SDS-PAGE gels below show corresponding fractions from the chromatograph. Mixing NT-dFMRP and qRRM2 resulted in elution of qRRM2 in higher molecular weight fractions. Because Coomassie binding is roughly proportional to the size of the protein (∼53 kDa for NT-dFMRP and ∼10 kDa for qRRM2), the qRRM2 bands in the higher molecular weight fractions represent a significant amount of qRRM2/ NT-dFMRP complex. The same NT-dFMRP gel image is also used in Supplementary Figure S6A, B. (E) Ethidium bromide-stained gel analysis of RT-PCR for nos RNA in anti-dFMRP immunoprecipitates from glo null mutant ovaries expressing GFP-Glo and GFP-Glo qRRM variants. nos RNA co-immunoprecipitated with dFMRP when GFP-Glo, GFP-Glo qRRM1, or GFP-Glo qRRM3, but not GFP-Glo qRRM2, was expressed. diversity. The lack of apparent functional domains apart from the qRRMs raises the possibility that these RNAbinding domains also mediate protein-protein interactions. Notably, the RRM and its variants have been implicated in RRM-RRM inter-domain packing (46,47) and RRMprotein interactions (48,49). Among these protein-binding RRM family members, the U2AF homology motif (UHM) has emerged as a specialized protein-interacting platform rather than an RNA-interacting platform (50). Interestingly, the mammalian hnRNP F qRRM2 was recently shown to physically interact with FOXP3, which modulates the activity of hnRNP F in regulating alternative splicing (51). Our demonstration that the interaction between Glo and dFMRP also involves qRRM2 suggests that qRRM2 might be an important protein-interaction platform as well as an RNA-binding platform for the hnRNP F/H family more generally. This finding also indicates that despite their structural similarity, the 3 Glo qRRMs differ in their protein binding specificity. Followup studies of other high-confidence Glo-interacting proteins identified by our IP-MS analysis may provide further evidence for differential protein-protein interaction capacity among Glo's qRRMs.
The incorporation of protein-binding and RNA-binding platforms into a single qRRM may enable specific protein interactors to influence RNA target selectivity. Depending on which part of qRRM is used for interaction, a qRRMinteracting protein could affect the accessibility of RNAbinding residues to RNA targets. While the FOXP3-hnRNP F qRRM2 interaction prevents hnRNP F from binding to BCL-X pre-mRNA (51), the interaction of SF3b155 to p14, a human RRM-containing U2 and U11/12 snRNP component, exposes secondary RNAbinding residues on the latter (52). Previously, we showed that each Glo qRRM harbors 2 distinct RNA-binding interfaces that allow Glo to recognize two different sequence motifs: a single-stranded G-tract and a doublestranded UA-rich motif (18). The integration of both binding modes into a single qRRM and the combination of 3 qRRMs are thought to diversify the RNA target repertoire of Glo. Both the G-tract and UA-rich motif binding modes are required for Glo-TCE interaction, and hence nos regulation, as well as for regulation of as yet unknown targets required for viability. By contrast, only the G-tract binding mode is required for Glo's function as a putative splicing regulator in dorsal-ventral patterning and ovarian nurse cell chromatin structure remodeling. It remains to be determined if and how binding of dFMRP affects the RNA-binding affinity and specificity of qRRM2. Regardless, the choice of protein interactor could influence the combinatorial use of the 6 RNA-binding interfaces of Glo. For example, binding of Glo to dFMRP or other nos regulatory factors may leave both RNA-binding surfaces exposed, favoring nos TCE recognition and translational repression, whereas binding of Glo to Hrp48 and Hfp may leave only the G-tract binding residues exposed, favoring a different subset of targets for alternative splicing.
Using nos regulation as a model, we have shown that Glo represses translation elongation by interacting with a polysome-associated protein, dFMRP, and that Glo's RNA-binding qRRM domains also mediate proteinprotein interaction. Since Glo is a multi-functional posttranscriptional regulator, the same principles could also be applied to its other activities where integration of multiple regulatory modes into a single node is required. Identification of such targets and additional regulatory factors involved will forward our understanding of how qRRMs of Glo contribute to its functional diversity and, more generally, how RNA-binding proteins confer posttranscriptional gene regulation.

MATERIALS AND CORRESPONDENCE
Reagent requests and other correspondence should be addressed to Elizabeth R. Gavis.

DATA AVAILABILITY
The datasets generated or analyzed during this study are included in this published article (and its supplementary information files). Data are available via ProteomeXchange with identifier PXD033174 and PXD033175.