Introduction

The localization of mRNAs to distinct cytoplasmic domains regulates many biological processes in eukaryotic cells such as embryonic patterning, cell polarity and neuronal function(Besse and Ephrussi, 2008; Holt and Bullock, 2009; Sutton and Schuman, 2006). This process is an effective way to restrict expression of proteins in cytoplasmic sub-compartments, thus spatially controlling gene expression(Martin and Ephrussi, 2009; St Johnston, 2005).

The transport of most localized mRNAs depends on cytoskeletal motors(Martin and Ephrussi, 2009; Tekotte and Davis, 2002). The localizing mRNAs are recognized by trans-acting protein factors that link them to microtubule-based motors. This recognition depends on cis-acting elements located mainly in the 3’ untranslated regions (3’UTRs) of mRNAs(St Johnston, 1995). The molecular mechanisms of how trans-acting factors recognize the RNA signals and link the mRNAs to the motor machinery are poorly understood.

Many mRNAs move towards the minus ends of microtubules by the dynein motor and its accessory complex, dynactin(Herbert et al., 2017; Trovisco et al., 2016; Vazquez-Pianzola et al., 2017). In Drosophila melanogaster the dynein-dynactin complex is involved in the localization of several essential patterning factors during early development. Dynein-dynactin-dependent transport has been described in several Drosophila cell types including oocytes, syncytial blastoderm embryos, neuroblasts and sensory neurons(Bullock and Ish-Horowicz, 2001; Ho et al., 2022; Hughes et al., 2004). mRNAs are linked to the dynein-dynactin machinery in these cells by the evolutionarily conserved coiled-coiled adaptor Bicaudal D (BicD) and the RNA-binding protein Egalitarian (Egl)(Bullock and Ish-Horowicz, 2001; Claussen and Suter, 2005; Dienstbier et al., 2009; Hughes et al., 2004; Mach and Lehmann, 1997; Navarro et al., 2004; Oh et al., 2000). Egl directly binds to mRNA localization elements in different mRNAs as well as the C-terminal domain (CTD) of BicD(Dienstbier et al., 2009). In addition, Egl also directly binds to the Dynein light chain (Dlc), facilitating cargo transport(Hoogenraad et al., 2001; Navarro et al., 2004) Transport complexes consisting of Egl, BicD, dynein, dynactin and mRNAs reconstituted in vitro are sufficient for long distance mRNA transport on microtubules(McClintock et al., 2018; Sladewski et al., 2018), revealing this is the minimal machinery for trafficking.

Minimal mRNA signals required for Egl-mediated minus-end-directed transport have been characterized in several mRNAs, including bicoid (bcd), fushi tarazu, gurken, hairy, fs(1)K10 (K10), wingless, and the I Factor retrotransposon(Bullock et al., 2003; Cohen et al., 2005, 2005; dos Santos et al., 2008; Macdonald and Kerr, 1998; Serano and Cohen, 1995; Snee et al., 2005; Van De Bor et al., 2005). These localization signals are predicted to form stem-loop structures of ∼40–65 nucleotides (nt)(Bullock et al., 2003; Cohen et al., 2005, 2005; dos Santos et al., 2008; Macdonald and Kerr, 1998; Serano and Cohen, 1995; Snee et al., 2005; Van De Bor et al., 2005). As these cis-acting elements show little primary sequence similarity to each other, it is unclear how they are recognized by the same protein.

K10 is one of the most comprehensively studied mRNA cargoes for Egl. This mRNA is synthesized in the nurse cells and transported to the oocyte by Egl-BicD-dynein-dynactin where it is needed to establish dorso-ventral polarity of the future embryo(Cheung et al., 1992; Serano and Cohen, 1995). The 44-nucleotide transport and localization signal (TLS) of K10 is necessary and sufficient for the localization of transcripts during early oogenesis (stage 2), late oogenesis (until stage 10) and in blastoderm embryos(Bullock and Ish-Horowicz, 2001; Serano and Cohen, 1995). Comprehensive mutagenesis of the K10 TLS combined with injection-based localization assays in blastoderm embryos revealed that both primary and secondary structural elements are important for recognition by the transport machinery(Bullock et al., 2010).

The molecular details of how Egl recognizes the K10 TLS or other RNA-localization elements are currently unknown. Egl is a 1004-amino-acid protein predicted to include a large uncharacterized N-terminal region, a 3’-5’ exonuclease (EXO) domain from the DEDDy superfamily and a C-terminal part containing a short consensus Dlc-binding region(Dienstbier et al., 2009; Mach and Lehmann, 1997; Navarro et al., 2004). Biochemical assays revealed that a large N-terminal portion of the protein encompassing the EXO domain is able to bind RNA(Dienstbier et al., 2009).

Here, we present the X-ray crystal structure of the minimal RNA-binding region of Egl in complex with the K10 TLS. The structure reveals that RNA recognition is mediated by the EXO domain, a small positively charged domain and a helical linker separating these elements. Egl recognises the K10 TLS by a combination of shape and sequence-specific interactions suggesting a mechanism for mRNA target discrimination. Editing of the egl locus in Drosophila using CRISPR/Cas9 shows that the residues involved in RNA recognition are essential for in vivo function, thus validating the structure.

Results

The central region of Egl is sufficient for RNA binding in vitro

Based on sequence analysis, three main conserved regions can be identified in Egl (Figure 1A; Figure S1A). The N-terminal region (Egl^N, residues 1-500) includes an uncharacterized domain that bears the minimal binding region for BicD (BicD-BR, residues 1-79; (Dienstbier et al., 2009)), which links the RNP complex to dynein-dynactin(Dienstbier et al., 2009; Liu et al., 2013; Mach and Lehmann, 1997; Navarro et al., 2004). The central region of Egl (Egl^MID, residues 500-850) is predicted to include a 3’-to-5’ exonuclease domain (EXO) that is essential for RNA-binding in vitro (Dienstbier et al., 2009), followed by a small domain predicted to contain helical features. The C-terminal portion of Egl (Egl^C; residues 850-1004) is uncharacterized apart from and a short Dynein light chain-binding motif (Dlc-BM)(Hoogenraad et al., 2001; Navarro et al., 2004). Previous analysis of RNA binding in vitro showed that a protein encompassing Egl^N and Egl^Mid (Egl^1-819) is sufficient for specific recognition of mRNA signals in vitro(Dienstbier et al., 2009).

Figure 1.

To further define the boundaries of the Egl RNA-binding region, we produced a series of truncations (Figure 1A), and measured binding affinities to the K10 TLS by Fluorescence Anisotropy (FA). After confirming that the presence of a GST tag does not affect binding of Egl^512-819 to RNA significantly in control experiments (Figure S1B), each Egl variant was purified as a GST fusion to improve solubility and enhance FA sensitivity. All truncated versions of Egl bind to K10 TLS in the sub-micromolar range, similar to Egl^1-819 (Figure 1B-D), revealing that Egl^512-819 is sufficient for full RNA binding. Including regions of the C-terminus (Egl^512-1004) did not significantly affect RNA binding (Figure 1B-D). Thus, we conclude that Egl^512-819 is sufficient to bind to K10 TLS in vitro.

The Egl^512-819 -K10 TLS complex is stable and monomeric in solution

To understand the molecular mechanisms of Egl recognition of K10 TLS, we first reconstituted a minimal complex in vitro using Egl^512-819. We added two G-C-pairs at the base of the K10 TLS stem that have been shown to stabilize the structure in solution and showed no loss of in vivo localization (Bullock et al., 2010). The RNA-protein complex is stable as shown by size-exclusion chromatography (SEC) and elutes as a well-defined peak with a larger elution volume compared with that of apo Egl^512-819 (Figure 1E). Both Egl^512-819 alone and in complex with K10 TLS eluted with an elution volume consistent with that of a monomer. This result was also confirmed by SEC coupled to right-angle light scattering (SEC-RALS) that showed that Egl^512-819 alone or in complex with K10 TLS is monomeric in solution and binds to K10 TLS with a 1:1 protein-RNA stoichiometry (Figure S1E-F).

Overview of the structure of Egl^512-819 in complex with K10 TLS

We obtained crystals of the SEC-purified complex of Egl^512-819-K10 TLS. The structure was determined by molecular replacement (MR) at 3.1-Å resolution (Table 1) using a model generated by I-TASSER based on a homologous template, RNAse D(Barbosa et al., 2014). The crystal lattice includes two complexes of Egl^512-819-K10 TLS in the asymmetric unit (ASU). The two complexes superimpose with root mean square deviation (RMSD) of 0.634 Å over 1971 atoms. The crystal contact interface between them is stabilized through hydrogen bonds and salt bridges (Figure S2A-C).

Table 1

In the final Egl^512-819 model, the C-terminal region of the construct is not visible (aa 769-819) and several loop regions (aa 512-519, 552-553, 652-655, 720-723) are disordered and could not be modeled (Figure 2A-B). In the K10 RNA model, the loop from A20 to C27 (loop^R) is not well defined and was also not modeled. The C-terminal part of the protein contains a region that we termed the Egl helical domain (EHD). To further study the EHD that could not be modeled unambiguously, we generated a construct encoding residues 743 to 819 of Egl (Linker (L)-EHD) fused to the Maltose Binding Protein (MBP) as a crystallization driver (Figure 2A). We crystallized and determined the crystal structure of MBP-L-EHD at 2.5-Å resolution by MR using MBP as a search model (Table 1). The obtained electron density is well defined and allowed us to model residues 743 to 806 with confidence (Figure 2C; Figure S2D). The final Egl^512–819–K10 TLS model was obtained by superimposing the two crystal structures using the shared rigid helical linker (L) as the alignment reference (Figure 2E).

Figure 2.

The overall structure of Egl^512-819 consists of an N-terminal EXO domain and the small, positively charged EHD connected by a helical linker (Figure 2E; Figure S2D). K10 TLS forms a typical A-form dsRNA duplex with an overall conformation similar to ideal A-form dsRNA (RMSD 3.1 Å over 40 nucleotides) (Figure S3; Table 1). In the complex, RNA base pairing is mostly by Watson-Crick interactions with the exception of a wobble base pair (between G6-U43) and two bulges at C35 and A41 named Bulge1 and 2, respectively (Figure 3B). In the structure, Bulge2 lies on the opposite face of the helix to Bulge1, thus pointing to solvent and not interacting directly with the protein. Therefore, when in complex with Egl^512-819, Bulge2 of K10 TLS is in a similar structural position compared with in silico predictions (mFold;(Zuker, 2003)). However, it differs from the conformation of unbound K10 TLS in solution, where Bulge2 is spatially arranged on the same side of the RNA stem as Bulge1 and involves nucleotide A39 (Figure S3;(Bullock et al., 2010)). Conversely, in the complex, Bulge2 consists of A41. Conformational flexibility could explain the less kinked conformation of the stem loop in the complex as compared with the previous NMR structure of K10 TLS(Bullock et al., 2010). The folding of K10 TLS in the complex is similar to a partial crystal structure we obtained of unbound K10 TLS that shows A-form conformation (Figure S3; Table 1).

Figure 3.

The interaction surface of Egl^512-819 with K10 TLS covers 905.7 Ų. Similarly, the K10 TLS interaction surface involved in binding Egl^512-819 is 1175.2 Ų. In the structure of the Egl^512-819-K10 TLS complex, the first 3’-most five nucleotides and the last 5’ complementary five at nt 44-48 associate tightly with the EXO domain (Figure 2E and 3; Figure S2D). The region that spans from nt C35 to A37, located around Bulge1, forms contacts with both linker and EHD. Interactions in this region are charged, and occur mostly with the RNA phosphate backbone (Figure 2D and 3).

The EXO domain adopts a typical α/β fold, consisting of a six-stranded β-sheet flanked by α-helices. A structure-based search using the DALI server(Holm and Laakso, 2016) identified several EXO domain-containing proteins including E. coli RNase D, T. brucei RRP6 and B. mori Exonuclease 3’-5’ domain-containing protein 1 (Exd1) (Figure S4A). Among these, RNase D shares the highest structural similarity with the EXO domain of Egl (RMSD 2.255 Å over 670 atoms) albeit with low sequence identity (21%) (Figure S4A). Based on structural comparisons, and as predicted by previous sequence analysis ((Navarro et al., 2004)Moser et al., 1997; Figure S1), the EXO domain of Egl belongs to the DEDDy superfamily. The conserved signature of DEDDy proteins are the catalytic tetrad and a tyrosine in so-called motif III. The EXO domain of Egl contains the four highly conserved acidic residues Asp561, Glu563, Asp621, and Asp708, and also the tyrosine (Tyr704) in motif III (Figure S1). The close arrangement of these residues in the Egl structure is consistent with the EXO domain of Egl retaining 3’-5’ exonuclease activity. However, mutation of all five putative catalytic residues does not affect RNA binding and in vivo function(Dienstbier et al., 2009; Navarro et al., 2004).

The EHD domain is composed of two positively charged α-helices facing each other in antiparallel arrangement and connected by unstructured loops (helix-loop-helix). Both helices contact the linker helix of Egl while helix α14 contacts the middle and upper regions of the RNA stem around bulge 1 (Figure 2E and 3). No significant structural homology of the EHD to known domains/motifs was retrieved upon databases searches but this could be due to its small size.

A 35 Å long helical linker connects the EXO and EHD domain, which is also highly positively charged. A comparison between Egl^512-819-K10 and the L-EHD structures shows that the linker is in a fixed position relative to the EXO and EHD domains and superposes with an RMSD of 1.031 Å over 117 atoms (Figure 2). Indeed, the linker position is stabilized by several hydrophobic and salt bridges interactions with EHD and EXO domain residues (Figure 2 and S4B).

Protein-RNA interactions within the Egl^512-819-K10 TLS complex

The association between Egl^512-819 and the K10 TLS RNA is mainly driven by electrostatic interactions. Positively charged residues are concentrated at the RNA interface (Figure 2D; Figure S4C). These positively charged residues are conserved in evolution (Figure S4D), supporting an important functional role.

The EXO domain clasps the lower part of K10 TLS mainly by unspecific sugar- and phosphate-backbone interactions but also with a few direct base recognitions (Figure 3). The Watson-Crick base pairs G1-C48 and C3-G46 as well as A44 are involved in base-directed recognition. The side chains of Asn566 and Gln689 form hydrogen bonds with G1 and the aromatic side chain of Tyr685 is π-stacked against the 3’-terminal C48 (Figure 3G; Figure S2D). This interaction occurs with the first of the two RNA stabilizing G-C pairs we added to the sequence and involves a 3’OH base-specific recognition. However, the natural base pair predicted to form the base of the RNA stem is C-G, which would also be engaged in a similar and specific recognition in the absence of the artificial G-C doublet.

In addition, the conserved residues Asp617, Arg619 and Asn620 recognise the third base pair C3-G46 through hydrogen bonding interactions (Figure 3E). A direct interaction is also observed between Lys659 and A44 (Figure 3F). The nonspecific contacts between the Egl^512-819 EXO domain and the RNA backbone can be divided into three groups: those forming hydrogen bonds with O2’ of the phosphate moiety (Gln744, Leu569 and Leu567), those forming hydrogen bonds or salt bridges with phosphate groups (Glu563, Lys678, Ser662, Leu663, His616, Gly564 and Glu563), and hydrophobic interactions with the sugar backbone (Gln640 and Gln744) (Figure 3B). Of the five putative catalytic residues only Glu563 interacts directly with the 3’-OH of the RNA backbone.

In the mid region of the RNA stem, four positively charged residues (Lys751, Lys754, Lys755 and Lys758) in the linker of Egl^512-819 (Figure 3D) and three (Arg782, Arg785 and Arg788) in the EHD (Figure 3C) are in close contact with the negatively charged RNA backbone. Although there are no base-specific contacts in this region, the phosphate-backbone interaction in correspondence with Bulge1 at C35 is involved in electrostatic interactions with conserved positively charged amino acids and might play an important role in Egl substrate selectivity through a shape preference. There are no contacts between Egl^512-819 and the upper region of K10 TLS that include the upper stem and loop.

Conserved residues of Egl^512-819 are involved in RNA binding

The Egl EXO domain, linker and EHD domain all contact the K10 TLS. To investigate the individual contribution of each region to RNA binding, we first generated a series of truncated versions of Egl that included the EXO domain (EXO, Egl^512-751), EXO and linker (EXO-L, Egl^512-770), and linker and EHD (L-EHD, Egl^743-819) (Figure 4A) and measured binding affinities to 5’-6-carboxy-fluorescein (6-FAM)-labelled K10 TLS by FA. The linker and EHD could not be purified separately and were not used for affinity measurements.

Figure 4.

The affinity of the Egl EXO domain for K10 TLS is at least 50 times lower than that of Egl^512-819 as no binding up to 5 μM RNA could be detected. In contrast, L-EHD bound with a K_d of 262 ± 9 nM, only two-fold weaker than Egl^512-819 (Figure 4C-E). To examine the role of the linker region, we extended the EXO construct to include the linker. This EXO-L construct binds in the high nanomolar range with a K_d of 807 ± 104 nM, suggesting that the linker contributes to RNA binding and that the EHD acts as effector of RNA binding (Figure 4C-E). Overall, these results indicate that Egl EXO domain, linker and EHD domain all contribute to binding, consistent with the structure.

To identify key residues involved in RNA-binding, we selected 16 conserved residues on each region for alanine scanning mutagenesis (Figure 4B). All mutants exhibit circular dichroism (CD) spectra similar to that of Egl^512-819 wild type (wt), indicating that mutants are properly folded (Figure S5A). Most individual Ala mutations reduce the affinity to K10 TLS only up to two-fold (Figure S5B-C). Next, to disrupt salt bridges we individually replaced all of the positively charged residues by Glu. These reverse-charge mutants had either a similar or slightly stronger effect than the respective Ala mutants with an up to a 3-fold reduction in affinity (Figure 4F-H). Next, we tried to combine single mutations to more strongly impair RNA binding. Three combinations (D617A/R619A/N620A, R753E/K754E/K755E and K659E/K754E/R788E) were selected for measurements. Asp617, Arg619 and Asn620 form base-directed hydrogen bonds with the base pair C3-G46 (Figure 3A). The D617A/R619A/N620A triple mutant shows two-fold reduction in binding with a K_d of 273 ± 12 nM (Figure S5B-C). These residues are all located on the EXO domain and their mutation reduces binding affinity to a similar degree as the EXO-domain deletion. This result confirms the contribution of the EXO domain to RNA binding and reveals residues that functionally contribute to RNA binding. The next triple mutant we tested disrupts a positively charged patch (Arg753, Lys754 and Lys755) in the linker region (L-Triple). This mutant shows a ten-fold reduction in binding affinity (1461 ± 302 nM) (Figure 4F-H). Finally, we combined mutations affecting each domain (EXO, K659E; linker, K745E; EHD, R788E). We did not detect RNA binding for this triple mutant (Figure S5B-C).

Determinants of RNA shape and sequence for Egl binding

To test the specificity of Egl^512-819 binding towards RNA targets, we first measured direct binding affinity to different FAM-labeled RNA oligomers by FA. However, with this method, we could not detect large differences in affinities between Egl^512-819 and various target and non-target RNAs (Figure S6A-C). As an alternative approach, we set up a competition assay in which the preformed Egl^512-819-FAM K10 complex was incubated with increasing amounts of various unlabeled RNAs.

Previous work suggested that RNA secondary structure is important for specific recognition by Egl(Bullock et al., 2003; Bullock and Ish-Horowicz, 2001; Claussen and Suter, 2005; Hughes et al., 2004; Mach and Lehmann, 1997; Navarro et al., 2004; Oh et al., 2000). To test this, we measured the half-maximal inhibitory concentration (IC₅₀) against single-stranded RNA (ssRNA), double-stranded RNA (dsRNA) and a non-target stem loop RNA derived from a bacteriophage (MS2) (Figure 5A-C and S6D). ssRNA was a weak competitor of K10 TLS for Egl^512-819 binding with an IC₅₀ of 112 ± 19 μM, which is 2000-fold weaker than control competitor K10 TLS. In contrast, dsRNA and MS2 stem loop are two-fold weaker competitors than K10 TLS (Figure 5A-C). Thus, a double helix is required for efficient competition with K10 TLS. We also tested both ssDNA and dsDNA. As expected, dsDNA was a much weaker competition compared to dsRNA. ssDNA competition could not be detected even at a 40,000-fold excess over K10 TLS (Figure 5A-C).

Figure 5.

As controls for the assay, we included another Egl target, bcd stem-loop 5 (bcd_VbL;(Lazzaretti et al., 2016)) and an Egl-independent dsRNA element from the human ARF1 transcript(Kim et al., 2005) (Figure 5D-F and S6D). bcd_VbL competed very efficiently with K10 RNA, whereas ARF1 RNA was a 1.9-fold weaker competitor.

In the structure we observed five base-directed interactions involving the G1-C48 pair, the C3-G46 pair, and A44. We focused on the G1-C48 pair because its substitution to an AU pair removes one hydrogen bond to Egl. This leads to a small but consistent (1.2-fold) reduction in competition. The structure also revealed an interaction of Egl to a bulge at C35. We inserted the bulge on the opposite strand in the same position. This K10-b1as RNA also showed a mild reduction in competition. When we combined the two substitutions, we found an additive effect on competition with a 1.6-fold reduction. This RNA species competes similarly to the antisense K10 RNA, an RNA that is unable to localize in an Egl-dependent manner in vivo(Dienstbier et al., 2009) (Figure 5D-F). Deletion of both bulges (K10-Δb1b2) also leads to defects in in vivo localization and to a 1.2-fold reduction in competition(Bullock et al., 2010).

The addition of native 10 nt flanking sequence of the RNA (K10-flank) and the addition of the two stabilizing base pairs, as used for structure determination (K10-2GC) at the base of the stem did not alter competitive ability as compared to K10 TLS (Figure 5D-F), this aligns with previous in vivo findings showing that such G–C substitutions do not disrupt K10 mRNA localization. In summary, specific interactions involving both sequence-specific and shape recognition contribute to preference of Egl for different RNA targets. Some of the mutations with a small effect on in vitro competition have a large effect on Egl-dependent in vivo localization. We cannot exclude the possibility that other cellular components contribute to in vivo target selection by Egl.

RNA binding is required for Egl function in vivo

We next evaluated the in vivo significance of RNA-protein interactions defined by our structure. We first used CRISPR/Cas9-mediated homology directed repair ((Port et al., 2014); Figure S7A) to create an allele of the endogenous Drosophila egl gene coding for a protein with the R753E, K754E and K755E mutations in the linker (egl^L-triple). As described above, the wild-type residues contact in K10 TLS and the presence of all three mutations lowers the affinity of Egl^512-819 for the K10 TLS in vitro by approximately one order of magnitude (Figure 4F-H).

Null mutations in egl have been described previously, and are most commonly associated with premature termination of translation. Flies homozygous for these alleles survive to adulthood but females are infertile. These females lack eggs due to a failure to differentiate an oocyte(Carpenter, 1994; Mach and Lehmann, 1997; Theurkauf et al., 1993). Homozygous mutant egl^L-triple females were also infertile and had a complete absence of egg production (Figure 6A). Immunostaining of ovarioles for a marker of the oocyte (Orb; (Lantz and Schedl, 1994)) and the synaptonemal complex (Corolla;(Page and Hawley, 2001) revealed the egl^L-triplemutation, like previously existing egl null mutations, prevented differentiation of the oocyte within the germarium (Figure 6B). An indistinguishable phenotype was observed in ovarioles of egl^L-triple/egl^null females (Figure 6B). The finding that the phenotype associated with the egl^L-triple allele was not complemented by an egl^null chromosome generated in an independent study reveals that the phenotype is not due to an off-target mutation. Immunostaining of egg chambers with an antibody that specifically recognises Egl(Mach and Lehmann, 1997) showed that the egl^L-triplemutation does not destabilise the Egl protein (Figure S7). Collectively, these observations reveal that the egl^L-triple mutation strongly disrupts the function of the Egl protein in vivo.

Figure 6.

To further examine the physiological relevance of our in vitro results, we used the same CRISPR methodology to introduce the K659E mutation into Egl (Figure 6C). As described above, K659 resides in EXO domain and contacts A44 of the K10 TLS in the crystal structure. K659E lowers the affinity of Egl^512-819 for the K10 TLS in vitro, although to a lesser degree than the triple mutation R753E/K754E/K755E (∼3-fold vs ∼10-fold (Figure 4)). Females homozygous for the K659E mutation were fertile and exhibited no overt defects in oogenesis (Figure 6C and D). egl^K659E/egl^nullfemales were also fertile. However, ∼30% of early egg chambers from these animals did not restrict the oocyte marker to a single cell, whereas ∼20% did not enrich the synaptonemal complex marker in a single cell (Figure 6C and D). These observations indicate a failure to specify or maintain an oocyte in a subset of cases. There was no overt decrease in Egl protein levels in the mutant egg chambers (Figure S7). Collectively, these data show that K659E partially disrupts the function of the Egl protein in vivo.

In summary, our phenotypic analysis reveals that the RNA binding residues in Egl^512-819 are also important for the function of the full length Egl protein in vivo. The correlation between the extent to which the mutations we tested inhibit RNA binding in vitro and the degree of disruption of oocyte differentiation in vivo further supports the physiological relevance of our in vitro studies.

Discussion

Here, we solved the X-ray crystal structure of the RNA-binding region of Egl in complex with the K10 transport and localization signal (TLS). The structure revealed how Egl recognizes the K10 TLS by a combination of shape and sequence-specific interactions. Our structure-based in vivo mutagenesis in Drosophila showed that key residues involved in K10 TLS recognition are also important for in vivo function.

The EXO domain of Egl contains the conserved signature of DEDDy proteins consisting of five highly conserved residues (Figure S1 and S4) and shares the highest structural similarity with RNase D. RNase D is an active 3′ to 5′ DEDDy-group exonuclease with a function in the maturation of RNAs(Li et al., 1998). Our structural analysis of Egl reveals that the DEDDy residues are appropriately positioned with respect to each other in a putative catalytic centre. Remarkably, mutations of the active site residues do not affect binding to mRNA localization signals and the in vivo function during oogenesis(Dienstbier et al., 2009; Navarro et al., 2004). The evolutionary conservation of the DEDDy motif indicates that Egl may have exonuclease activity that contributes to other aspects of its function. However, Egl’s major function in mRNA localization appears to have evolved in parallel to the exonucleolytic activity.

The origin of new RNA-binding proteins from enzymes involved in RNA metabolism seems to have occurred repeatedly in evolution(Lazzaretti et al., 2016). For example, the RNA-binding protein Exuperantia (Exu) binds RNA by a modified exonuclease-like domain and a SAM domain. The exonuclease active-site is lost in Drosophila Exu but is present in some mollusc and vertebrate Exu homologs(Lazzaretti et al., 2016). Another example is Exd1, a mammalian protein related to Egl that recruits small RNAs in the piRNA pathway through its inactive exonuclease-like domain(Yang et al., 2016). In Egl, RNA binds around the catalytic site and a positively charged surface formed by the helical linker and the EHD domain. In Exu, RNA-binding has been mapped by MS cross-linking and also localizes to a patch around the active site and the SAM domain(Lazzaretti et al., 2016). In both Egl and Exu, novel RNA-binding activity may have evolved by the addition of positively charged surfaces to stabilize the interactions between the RNA substrate and the exonuclease.

Previous biochemical and single-molecule imaging studies indicate that two Egl molecules form a complex with one mRNA containing a single stem-loop structure(McClintock et al., 2018; Sladewski et al., 2018). In our SEC-RALS experiment and crystal structure, the minimal RNA-binding region of Egl binds to K10 TLS with a 1:1 stoichiometry. This difference could be due to the shorter RNA or truncated Egl used in our study. It is unclear how two molecules of full-length Egl can bind to an mRNA with one stem loop. In our structure, we identified weak interactions between two Egl molecules in the asymmetric unit. These or other parts of Egl lacking from our structure may contribute to the formation of a 2:1 Egl-mRNA complex. In addition, the two Egl molecules may bind to different surfaces of the RNA stem loop. In our structure, both Egl molecules bind to both K10 TLS stem loops. The second interaction site has a much smaller interaction surface with the RNA. This weak interaction or interactions to neighbouring ssRNA might recruit a second molecule of Egl to the mRNA. Our structural work paves the way for understanding higher order RNA-protein interactions in the transported RNP.

Our crystal structure and in vitro affinity measurements provided insight into how Egl recognizes a specific localization element. Although Egl is also able to bind to ssRNA, dsRNA and other stem-loop RNA structures, the affinity of binding was higher for K10 TLS. This suggests that Egl can discriminate RNA secondary structure and shows some sequence specificity. Several lines of evidence indicate that in vivo Egl binds stem-loop structures with bulges in localizing RNAs. First, the K10 TLS is required for Egl binding and localization of the full-length mRNA (Cohen papers, Bullock 2001; 2010). Second, several mRNAs that depend on Egl for minus-end-directed transport during oogenesis and embryogenesis have similar predicted structures containing bulged regions(Bullock et al., 2010, 2003; Cohen et al., 2005; dos Santos et al., 2008; Macdonald and Kerr, 1998; Serano and Cohen, 1995; Snee et al., 2005; Van De Bor et al., 2005).

The Egl-K10 TLS structure and our biochemical and in vivo experiments revealed the structural details and relevance of Egl-stem-loop interactions. We identified specific interactions between Egl and K10-TLS-Bulge1 that create a unique electrostatic signature for protein recognition. The importance of bulges in stem-loop elements has also been described for other RNA-protein interactions(Hermann and Patel, 2000). In the genomic RNA of HIV-1 there are bulged adenine residues that provide recognition sites for viral proteins(Ennifar et al., 1999). These interactions depend on the secondary structure or shape of the RNAs, rather than their primary sequence. Mutation of the residues in Egl that mediate interactions to Bulge1 render Egl unable to exert its function in Drosophila oogenesis. In addition, there are also residues of Egl that mediate base-directed interactions and backbone interactions to the stem of the RNA stem-loop. Mutations of these residues also reduced binding and in vivo function. These results indicate that the recognition of stem-loop structures is due to a combination of base- and shape-specific interactions.

Methods

Protein expression and purification

To produce the Egl protein for crystallization and FA, Egl from D. melanogaster was cloned into a pET-MCN vector(Diebold et al., 2011) with an N-terminal glutathione S-transferase (GST) tag followed by a TEV protease cleavage site. Mutants and truncations were generated by site-directed mutagenesis.

GST-Egl constructs were expressed E. coli BL21 (DE3) Gold (Life Technologies) cells in TB medium at 37 °C until an OD600 of 0.6-0.8 and induced at 20 °C with 0.5 mM IPTG overnight. Cells were harvested (5000 x g, 20 min, 4 °C), resuspended in buffer A (20 mM Tris-HCl, pH 7.5, 1 M NaCl, 10% glycerol, 1 mM DTT), and lysed by homogenization. The lysate was cleared by centrifugation (18000 x g, 1 hour, 4 °C), and the supernatant was incubated with Glutathione Agarose 4B resin (Macherey-Nagel). Bound protein was washed with buffer A and buffer B (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol, 1 mM DTT), eluted with buffer C (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol, 1 mM DTT, 50 mM reduced glutathione). The protein was applied to Heparin HP Column (GE Healthcare), eluted with a gradient from 300 mM to 1 M NaCl and further purified by SEC on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) in buffer B.

The Egl^512-819 construct was expressed and purified as described above, except that the GST-tag was removed by TEV protease overnight at 4 °C. The Egl^512-819-K10 TLS complex was assembled by incubating Egl^512-819 with a 1.2-M excess of K10 TLS on ice overnight and further purified by SEC on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) in buffer B.

The MBP- (GSAM)- L-EHD (Egl^743-819) construct was expressed E. coli BL21 (DE3) Gold (Life Technologies) cells in autoinduction ZY medium at 37 °C until OD₆₀₀ = 2.0, and induced at 20 °C overnight. Cells were harvested and lysed as described above. The supernatant was incubated with Amylose Resin (NEB). Bound protein was washed with buffer A and buffer B (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol, 1 mM DTT), eluted with buffer D (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol, 1 mM DTT, 50 mM Maltose) and further purified by SEC on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) in buffer E (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, 10 mM Maltose).

Crystallization, data collection and analysis

Crystallization conditions were initially screened at 4°C and at 20°C using a Mosquito nanodrop dispenser (TTP Labtech) in 96-well sitting-drop plates with 100 nl protein plus 100 nl reservoir. Crystals of Egl^512-819-K10 TLS complex grew in 50 mM sodium citrate, pH 4.8, 30% (w/v) MPD, 10 mM CaCl₂, 2.5 mM Spermine at 20 °C. For X-ray data collection, a single crystal was harvested without cryoprotectant and flash cooled in liquid nitrogen. Diffraction data were collected at Beamline PXII of the Swiss Light Source. The diffraction data were indexed, integrated, and scaled using the XDS program package(Kabsch, 2010). A weak molecular replacement solution [TFZ=8.4] was obtained by using a trimmed poly-Ala model (RNAse D, PDB ID: 4NLB) generated by I-TASSER(Yang and Zhang, 2015). The model building of Egl^512-819 was performed with MR-Rosetta(Terwilliger et al., 2012) and manual building in COOT(Emsley et al., 2010), and the model of K10 TLS was built using Brickworx(Chojnowski et al., 2015) and manual run in COOT(Emsley et al., 2010). Refinement was carried out in Refmac5 using TLS and jelly body(Vagin et al., 2004).

Crystals of MBP- (GSAM)- L-EHD grew in 100 mM Tris-HCl, pH 7.5, 18% (w/v) PEG 6000, 50 mM MgCl₂ at 4 °C. For X-ray data collection, a single crystal was transferred to a cryoprotectant solution with mother liquor supplemented with addition of 20% glycerol, and flash cooled in liquid nitrogen. Diffraction data were collected and processed as described above. The structure was solved by molecular replacement with PHASER(McCoy et al., 2007) using the structure of maltose binding protein (PDB ID: 1ANF)(Quiocho et al., 1997) as the search model. All subsequent model building was performed in COOT(Emsley et al., 2010) and restrained refinement in Phenix(Terwilliger et al., 2012).

Protein structure figures were generated using PyMOL (Schrödinger, LLC.). The quality of the final models was validated with the wwPDB Validation Server (https://validate-rcsb.wwpdb.org/). Interaction surfaces were calculated with PISA(Krissinel and Henrick, 2007).

Fluorescence anisotropy

For direct binding assay, measurements were carried out using an Infinite F200 plate reader (Tecan) at 20°C. 5’-6-FAM-labeled RNA was added to each reaction at a fixed concentration of 10 nM with different concentrations of purified Egl truncations or mutants in 50 µl buffer F (20 mM HEPES-NaOH pH 7.5, 100 mM NaCl, 10% Glycerol). Each titration point was measured three times, with an integration time of 40 µs, with 485 nm and 535 nm as the excitation and emission wavelength, respectively. The data were analyzed using the Hill equation with the Prism 6 software (GraphPad).

For competition binding assay, 150 nM Egl^512-819 was preincubated with 10 nM 5’-FAM-labeled K10 TLS RNA for 10 min at 20°C and further incubated with different concentrations of unlabeled RNAs for 10 min at 20°C in 50 µl buffer F before measurement. The change of anisotropy was plotted and fitted to sigmoidal curve to obtain the IC₅₀ with Prism 6 (GraphPad).

The sequences of the RNA and DNA oligonucleotides (Integrated DNA Technologies) are shown as follows (5’ to 3’): K10 (CUUGAUUGUAUUUUUAAAUUAAUUCUUAAAAACUACAAAUUAAG); K10-Δbulges (CUUGAUUGUAUUUUUAAAUUAAUUCUUAAAAAUACAAUUAAG); K10-antisense (GAAUUAAACAUCAAAAAUUCUUAAUUAAAUUUUUAUGUUAGUUC); K10-anti-C35 (CUUGAUUGUACUUUUUAAAUUAAUUCUUAAAAAUACAAAUUAAG); K10-AU (AUUGAUUGUAUUUUUAAAUUAAUUCUUAAAAACUACAAAUUAAU); K10-anti-C35-AU (AUUGAUUGUACUUUUUAAAUUAAUUCUUAAAAAUACAAAUUAAU); K10-2GC (GGCUUGAUUGUAUUUUUAAAUUAAUUCUUAAAAACUACAAAUUAAGCC); K10-flank (CUUGAUUGUAUUUUUAAAUUAAUUCUUAAAAACUACAAAUUAAGAUCACUCUGU); ARF1 sense (dsRNA sense) (UGAGUGCCAGAAGCUGCCUC); ARF1 antisense (GAGGCAGUUUCUGGUACUCA); dsRNA antisense (GAGGCAGCUUCUGGCACUCA); ssRNA (polyU20) (UUUUUUUUUUUUUUUUUUUU); ssRNA (polyAU20) (AUAUAUAUAUAUAUAUAUAU); bcd_Vb (long) (GGGCCCAAAAUGAAAAAUGUUUCUCUUGGGCGUAAUCUCAUACAAUGAUUACCC UUAAAGAUCGAACAUUUAAACAAUAAUAUUUGGG); bcd_Vb (short) (AAAUGUUUCUCUUGGGCGUAAUCUCAUACAAUGAUUACCCUUAAAGAUCGAACA UUU); SOLE (CGAUAUCGAGCAUCAAGAGUGAAUAUCG); ssDNA (dsDNA sense) (AGGCAGTTTCTGGTACTCAG); dsDNA antisense (CTGAGTACCAGAAACTGCCT); MS2 (ACAUGAGGAUUACCCAUGU).

Right angle light scattering (RALS)

Samples were loaded onto a Superdex 75 5/150 column (GE Healthcare) connected to a TDA302 detector array (Viscotek), in buffer B at room temperature. Chromatography runs were performed with 2 mg/ml of purified Egl alone, or pre-incubated with a 1.1-M excess of K10 TLS on ice for 10 min. The sample volume was 20 µl. Data were analyzed with OmniSEC 4.5 software.

Analytical size-exclusion chromatography

Egl^512-819 (32 uM, 500 ul) alone, or pre-incubated with a 1.2-M excess of K10 TLS were loaded on a Superdex 75 10/300 GL column (GE Healthcare) in buffer B, and UV absorbance was monitored at 280 and 260 nm.

Circular dichroism

CD spectroscopy was performed on a JASCO-810 CD system. Wavelength scans were averages of five scans between 200 nm and 250 nm collected at 20°C in a quartz cuvette with a 1-mm path length with 0.2 mg/ml protein in buffer B.

Generation of Drosophila egl alleles by CRISPR/Cas9-mediated homology-directed repair

We used our previously reported CRISPR/Cas9 methodology(Port et al., 2014) to produce egl alleles encoding proteins with mutations of specific RNA-binding residues. Briefly, single-stranded oligonucleotide donors coding for the desired mutations (Ultramers, IDT; 500 ng/μl) were injected into the posterior of embryos that had a nos-cas9 transgene(Port et al., 2014) (CFD2) and a transgene (integrated at attP40) that expresses a single guide (sg)RNA targeting the egl locus adjacent to the region to be repaired by the respective donor (embryos were generated by crossing nos-cas9 females with sgRNA males). Spacer sequences for the two sgRNAs (sgRNA^L-Triple and sgRNA^K659E) had previously been cloned via complementary oligonucleotides into the transformation vector pCFD3(Port et al., 2014), which expresses sgRNA from the U6:3 promotor. CRISPR RGEN tools (http://www.rgenome.net/) revealed the selected sgRNAs have no other predicted targets in the Drosophila genome. Following injection of the donor into the transgenic embryos, flies heterozygous for the desired mutation were identified by Sanger sequencing of PCR products, as described previously(Port and Bullock, 2016). Sequences of spacer and donor oligos, as well as oligos used for sequencing the targeted region, are provided in Table S1. Balanced mutant stocks were established from the offspring of positive animals, followed by generation of homozygous females for phenotypic analysis.

Immunostaining of ovarioles

Ovarioles were dissected in PBS with 0.5% Tween (PBT0.5) and fixed in 4% paraformaldehyde/PBT0.5 for 20 minutes. Following washing in PBT0.5 (3 x 5 min and 4 x 15 min), samples were blocked in 20% Western Blocking Reagent (Roche)/PBT0.5 (Blocking Buffer) for 1 h. The following primary antibodies were then used overnight at 4°C in Blocking Buffer: (1) a combination of mouse anti-Orb (mixture of clones 4H8 and 6H4(Lantz et al., 1994) (Developmental Studies Hybridoma Bank; each diluted 1:200)) and rabbit anti-Corolla(Collins et al., 2014) (gift of Scott Hawley, Stowers Institute (diluted 1:2000)), or (2) rabbit anti-Egl (Mach and Lehmann, 1997)(gift of Ruth Lehmann, Skirball Institute; diluted 1:3000)). Following 3 x 5 min and 4 x 15 min PBT0.5 washes, secondary antibodies in Blocking Buffer (Alexa488-conjugated donkey-anti-rabbit (Life Technologies (A21206)); Alexa555-conjugated donkey-anti-mouse (Life Technologies (A31570) (each diluted 1:500)) were added for 2 h at room temperature. Washing steps were repeated and, after 2 brief washes in PBS, ovarioles were mounted in Vectashield containing DAPI (VectorLabs). Ovarioles were imaged with a Zeiss 780 confocal microscope using a Zeiss 40 x 1.3NA oil objective (except for Figure S7A (Zeiss 20 x 0.8NA objective)).

Supplementary figures

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Figure S6

Figure S7

Data availability

The coordinates and structure factors have been deposited in the Macromolecular Structure Database of European Bioinformatic Institute (EBI) with ID code 9UJU, 9UJY and 9UUG for Egl512-819-K10 TLS complex, MBP-(GSAM)-L-EHD and K10-TR.

Acknowledgements

We wish to thank the MPI-Tuebingen Crystallization Facility. We also thank the staff at the Swiss Light Source synchrotron for assistance during data collection, Claire Basquin for assistance with MALLS measurements and Stefanie Wachter for initial stages of the project. We thank G. Jékely and A. Cook for discussion and critical reading of the manuscript. This project has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013), ERC grant agreement n° 310957 and the Deutsche Forschungsgemeinschaft (FOR2333 to F.B.). L.J. and S.L.B. are supported by the MRC (file reference number MC_U105178790). Z.H is supported by A*STAR, National Medical Research Council (NMRC; reference number OFYIRG23jul-0062).

Additional information

Author Contributions

Biochemical, biophysical and crystallization work was done by Z.H. and J.M.; Drosophila work was done by L.J.; F.B. supervised the project. All authors wrote the paper.

Funding

MOH | National Medical Research Council (NMRC) (OFYIRG23jul-0062)

• Zebin Hong