Structure of Drosophila Oskar reveals a novel RNA binding protein

Significance

A profound question in animal biology concerns the origin of germ cells. In the model organism Drosophila melanogaster, oskar directs the assembly of germ plasm that forms and specifies germ cells, but the precise function of its gene product, the Oskar protein, remains poorly understood. Using X-ray crystallography, we determined the 3D structures of an N-terminal fragment (Osk-N) and a C-terminal domain of Oskar (Osk-C). Surprisingly, we discovered that Osk-C has RNA binding function. It binds the 3′UTR of its own mRNA, as well as that of nanos, another gene important for germ-line development. This discovery should direct future mechanistic studies of Oskar for a better understanding of its function in germ cell development.

Abstract

Oskar (Osk) protein plays critical roles during Drosophila germ cell development, yet its functions in germ-line formation and body patterning remain poorly understood. This situation contrasts sharply with the vast knowledge about the function and mechanism of osk mRNA localization. Osk is predicted to have an N-terminal LOTUS domain (Osk-N), which has been suggested to bind RNA, and a C-terminal hydrolase-like domain (Osk-C) of unknown function. Here, we report the crystal structures of Osk-N and Osk-C. Osk-N shows a homodimer of winged-helix–fold modules, but without detectable RNA-binding activity. Osk-C has a lipase-fold structure but lacks critical catalytic residues at the putative active site. Surprisingly, we found that Osk-C binds the 3′UTRs of osk and nanos mRNA in vitro. Mutational studies identified a region of Osk-C important for mRNA binding. These results suggest possible functions of Osk in the regulation of stability, regulation of translation, and localization of relevant mRNAs through direct interaction with their 3′UTRs, and provide structural insights into a novel protein–RNA interaction motif involving a hydrolase-related domain.

Spatial and temporal control of mRNA localization and translation play critical roles in oocyte development, cell polarity, and neuronal function (1, 2). The assembly of germ plasm in Drosophila is illustrative of such strict regulations. During oogenesis, the germ plasm is assembled at the posterior pole of the oocyte. Germ plasm is composed of RNAs and proteins, which are specifically required for the formation of germ cells in the early embryo and specification of germ cell fate (3). Germ plasm is also the source of the Nanos protein gradient required for embryonic polarity. Aberrant mRNA localization and untimely protein synthesis during oogenesis can lead to germ cell formation defects and mis-patterning, causing embryonic lethality and sterility (4).

A critical component of germ plasm is oskar. Oskar RNA is synthesized during oogenesis in the nurse cells, the sister cells of the oocyte, and transported toward the oocyte’s posterior pole by a process involving the exon junction complex, the oskar 3′UTR, and microtubule-based movement (5, 6). Oskar protein translation is repressed during transport by the RNA-binding protein Bruno and this repression is released by the binding of activators, such as Orb, once the RNA reaches the posterior pole (7–11). Oskar organizes germ plasm by recruiting other proteins, such as Vasa, Tudor, and Aubergine (12–15). An important function of germ plasm is the localization of 50–200 germ plasm-associated RNAs (16, 17). Among these, nanos (nos), germ cell-less (gcl), and polar granule component (pgc) have important roles in abdominal patterning and germ cell specification, germ cell formation, and germ cell transcriptional repression, respectively (18–22). All these functions fail in the absence of Osk and can be initiated at an ectopic location upon mislocalization of Osk (4, 23).

Germ plasm-associated RNAs are translationally repressed outside of the germ plasm and are translated in the germ plasm or germ cells. Osk has been implicated not only in the localization but also in the translational activation of nos mRNA, possibly by releasing the inhibitor Smaug (24). It is not known how Osk disrupts the interaction between Smaug and nos 3′UTR and whether Osk is also responsible for translational regulation of other germ plasm-enriched RNAs.

Although the function and the mechanism of osk mRNA localization have been studied in great detail, the role of Osk protein in germ-line formation and body patterning remains poorly understood. Translation of osk mRNA generates two protein isoforms, a long form and a short Osk protein spanning residues 1–606 and 139–606, respectively. The two isoforms have different localization patterns and the short Osk alone can rescue the defects of osk mutants, whereas the long Osk cannot (25). Here we focus our study on the short Osk protein, which is composed of an N-terminal LOTUS/OST-HTH domain (Osk-N) and a C-terminal hydrolase-like domain (Osk-C) (Fig. 1A). We have solved the crystal structure of each domain, and our combined structural and biochemical studies offer mechanistic insights into the molecular function of Oskar.

Fig. 1.

Structure of Osk-N. (A) A schematic drawing of domain organization of Osk. The shaded regions indicate Osk-N (cyan) and Osk-C (yellow). Starting positions of long and short forms of Osk are indicated with red arrows. Residue numbers at domain boundaries are labeled. (B) Overall structure of Osk-N homodimer. Two Osk-N molecules, Mol A and Mol B, are shown as a ribbon model colored in cyan and green, respectively. (C) Dimerization interactions involving intermolecular hydrogen bonding (dashed lines) between main-chain groups of two β1 stands. β1 stands are shown in a stick model (carbon, cyan and green; nitrogen, blue; oxygen, red; sulfur, yellow). (D) Intermolecular interactions between the two α3 helixes. Residues involved in interactions are shown in a stick model.

Results

The LOTUS Domain of Osk Is a Winged-Helix–Fold Dimerization Module.

An ∼80-residue N-terminal segment of short Osk, termed the LOTUS domain, was predicted to form an RNA-binding domain resembling the structure of a winged-helix DNA-binding domain. We crystallized two Osk fragments encompassing the delineated region, one wild-type fragment spanning residues 150–240 and a L199M mutant (animo acids 150–224), which is used for the SeMet-based single wavelength anomalous dispersion (SAD) method of structure determination. The two structures have been determined to 3.0 Å and 2.5 Å, respectively (Table S1). Despite being crystallized in different space groups, there are six protein monomers per asymmetric unit for both the wild-type and the mutant Osk fragments that can be separated into three homodimers with the same protein–protein interaction interface (Fig. S1). Furthermore, the ordered protein regions end within residues 220–223, and there are no significant differences, with an rmsd of 1.5 Å, between the structure models. Thus, the two structures will be collectively referred to as Osk-N henceforth, unless explicitly specified.

Table S1.

Statistics of crystallographic analysis

Datasets	Osk-N 150–224 L199M	Osk-N 150–240	Osk-C 393–606
Data collection
Wave length (Å)	0.9790	0.9641	0.9788
Space group	P21	P32	P3121
Cell dimensions
a, b, c (Å)	62.96, 38.31, 102.13	66.90, 66.95, 112.83	69.00, 69.00, 100.69
α, β, γ (°)	90.00, 95.54,90.00	90.00, 90.00, 120.00	90.00, 90.00, 120.00
Resolution range (Å)*	50–2.50 (2.59–2.50)	50–3.00 (3.11–3.00)	50–2.10 (2.18–2.10)
Rmerge	0.075 (0.428)	0.055 (0.456)	0.115 (0.423)
I/σ	26.3 (4.9)	31.2 (4.1)	11.8 (3.0)
Completeness (%)	99.8 (100)	99.8 (99.7)	99.8 (99.9)
Redundancy	7.3 (7.5)	5.8 (5.8)	3.8 (3.9)
Refinement
Resolution (Å)	50–2.50	50–3.00	50–2.10
No. reflections	17,054	11,302	16,668
Rwork/Rfree (%)	21.4 /27.0	18.9 /24.2	16.1 /20.3
No. atoms
Protein	3,387	3,447	1,726
Ligand/ion	24	12	35
Water	41	7	241
B-factor (Å2)	89.1	112.3	27.8
Main chain	88.0	110.6	24.5
Side chain	90.3	114.0	28.7
Ligand/ion	77.2	125.0	34.2
Water	60.1	72.4	35.6
Rmsd
Bond length (Å)	0.005	0.005	0.005
Bond angles (°)	0.900	0.913	0.934
Ramachandran plots
Favored	408 (96.9%)	399 (93.2%)	205 (99.0%)
Allowed	13 (3.1%)	29 (6.8%)	2 (1.0%)

^*Values in parentheses are that of the highest-resolution shell.

Fig. S1.

Packing of Osk-N in an asymmetric unit of the crystal lattice. There are six Osk-N monomers per asymmetric unit for both L199M (amino acids 150–224) (A) and the wild-type Osk-N (amino acids 150–240), despite being crystallized in different space groups (B). The six monomers in both cases can be separated into three identical homodimers, indicated as dimers A-B, C-D, and E-F, colored in cyan-green, magenta-yellow, and pink-gray, respectively. The A-B dimer is highlighted with a red box.

The Osk-N monomer adopts a winged-helix–fold, a typical DNA or RNA binding motif, with two antiparallel β-strands (β1 and β2) and three α-helixes (α1–α3) forming the “wing” and “helix,” respectively (Fig. 1B). Osk-N is predicted to be a RNA-binding motif, termed the LOTUS domain (26, 27). The structure shows that although Osk-N assumes a fold capable of RNA binding, it homodimerizes through the wing and helix α3, masking the potential nucleic acid binding surface (Fig. S2). The dimerization interface buries a total surface area of 881 Ų, indicating a moderately strong intermolecular interaction. The dimeric interface has an obvious hydrophobic patch involving Leu200 (located on α3) and Ala207 (located on β1) from both monomers. Principal dimerization interactions occur between the two β1-strands through intermolecular hydrogen bonding of main-chain groups (Fig. 1C). Ala207 is situated next to the axis of the pseudo twofold symmetry, and they interact with each other through a pair of intermolecular hydrogen bonds between the main-chain carbonyl and amide groups. Ser210 is located in a loop connecting β1 and β2, which appears to be in a location to influence proper positioning of β1. In addition, Arg215 located on β2 interacts with Asp197 located on α3 of the partner monomer (Fig. 1D).

Fig. S2.

Structural alignment of the Osk-N homodimer with the structure of a winged-helix domain transcription factor MecI complexed with DNA (PDB ID code 1SAX). The Osk-N monomer aligns well with the N-terminal domain of MecI (shown in gray), with an rmsd of 2.6 Å. MecI interacts with dsDNA through helix α3, which is longer than that of Osk-N. Homodimeriziton of Osk-N through helix α3 masks its potential nucleic acid binding surface, which is indicated by the pink shade.

To confirm homodimerization of Osk-N in solution, we analyzed the molecular mass of Osk-N (150–224) by multiangle light scattering (MALS). The measured molecular mass of Osk-N (150–224) is 13.2 kDa (Fig. S3A), which is ∼54% larger than the calculated molecular weight of 8.55 kDa, suggesting that dimeric and monomeric forms of Osk-N are in equilibrium in solution. To ensure that the dimeric Osk-N results from specific association, we hoped to identify Osk-N mutants that will disrupt the dimerization interaction. Based on the structural information, we designed a S210P mutant of Osk-N, with the expectation that introduction of a proline will alter the main-chain conformation of β1 that leads to the disruption of the dimer. MALS measurements show that S210P has a molecular mass of 8.77 kDa, a value close to the calculated mass of monomeric Osk-N (Fig. S3A). This result indicates that the observed Osk-N dimer is a result of specific association of two monomers.

Fig. S3.

Osk-N forms a dimer and does not bind RNA. (A) MALS measurement of the molecular mass of Osk-N (amino acids 150–224) and its S210P mutant are shown in red and green curves, respectively. (B) EMSA results of the binding of Osk-N and its S210P mutant to the 3′UTR of osk.

Overall Structure of Osk-C.

A 2.1 Å structure of the C-terminal domain of Oskar (Osk-C, amino acids 393–606) was solved, and the structure reveals a globular domain composed of seven α-helices (α1–α7), two β-strands (β1 and β2), and several long loops. β1 is located between α2 and α3, and β2 is inserted between α4 and α5. The two β-strands form a parallel sheet. Three long loops stand out in the structure: they are the N-terminal loop (LN), the one connecting α1 and α2 (L1), and another connecting α6 and α7 (L2) (Fig. 2A). These loops are longer than 20 residues and some are embedded with short secondary structural elements. Structural alignment shows that Osk-C has a lipase-like fold, having an rmsd of 2.7 Å when superimposed with a GDSL-like lipase (PDB ID code 3P94). Overall, secondary structural elements (α1–α7 and β1–β2) aligned well between the two structures, but significant differences occur in the loop segments (Fig. 2B). The N-terminal ends of Osk-C and the lipase protrude out of the core domain at different directions, and the short turn in the N-terminal loop is replaced by a distorted helix in the lipase structure (Fig. 2B). In addition, a pair of short strands—βa preceding α1 and βb between α5 and α6—join β1 and β2 to form an extended β-sheet, and an extra short helix (αa) is found C-terminal to α1 in the lipase structure. Furthermore, the lipase L1 loop and the loop connecting β2 and α5 are longer than the corresponding ones in Osk-C, and these loops take significantly different conformations in the two structures. In contrast, the L2 loop in Osk-C is much longer and extends over a significant portion of the protein surface (Fig. 2B).

Fig. 2.

Structure of Osk-C. (A) Overall Structure of Osk-C shown as a ribbon model (α-helixes and loops in yellow and β-sheets in purple). The three long loops are labeled as LN (N-terminal loop), L1 (loop between α1 and α2), and L2 (loop between α6 and α7). (B) Structural alignment of Osk-C with a GDSL-like lipase (PDB ID code 3P94). Osk-C is colored the same as in A and the lipase is shown in gray. Main differences between the two structures occur in loop regions, notably those of LN, L1, L2, and the one connecting β2 and α5. The area shaded in pink indicates the location of the lipase active site. An enlarged view of the active site is displayed in the pink-shaded inset, in which the catalytic triad of the lipase are highlighted in a stick representation and the corresponding Osk-C residues are shown in green.

The catalytic active site of the lipase is located on the loop corresponding to loop L2 in Osk-C. The lipase loop is much shorter, and the catalytic triad residues Ser200, Asp202, and His205 are spatially adjacent. The Osk-C residues corresponding to these positions are Phe571, Leu586, and Phe587 (Fig. 2B). Clearly, these hydrocarbon side-chains of the Osk-C residues are incapable to carry out any catalytic activities, as they lack polar functional groups suitable for carrying out nucleophillic attack and electron transfers. Despite the protein-fold similarity, no lipase or similar enzymatic activities are anticipated for Oskar.

RNA Binding Properties of Oskar.

Previous studies have shown that the mRNA localization signals of oskar and nanos are situated in the 3′UTR region of their mRNAs (28, 29). The 3′UTR region often harbors the binding sites of RNA localization and translational control factors. Gene products of cappuccino, orb, staufen, bruno, and oskar itself, have been implicated as posttranscriptional regulators of oskar mRNA (29). For nanos mRNA, Smaug and Osk are major regulators for repression and activation of translation (24, 30). However, direct biochemical evidence of Osk as an RNA-binding protein is lacking.

Two incidental observations suggest that Osk-C may be an RNA binding module: (i) during protein purification we noticed that Osk-C always associates with nucleic acids; and (ii) the protein surface is highly basic, a number of sulfate ions from the crystallization buffer are bound in the structure (Fig. S4). The sulfate ions may mimic the negatively charged phosphate backbone of RNA. Prompted by these observations, we set out to biochemically test the binding of Osk-C and Osk-N fragments to RNA. We first generated the 3′UTR of osk mRNA by in vitro transcription, and tested whether Osk-N and Osk-C are able to bind it in an EMSA. The results show that Osk-C binds osk 3′UTR avidly, but Osk-N shows no detectable binding under our assay conditions (Fig. 3A). Next, we tested whether Osk-C specifically binds RNA or indiscriminately associates with nucleic acids, such as double-stranded DNA (dsDNA), using a competition assay. We generated a fluorescein-labeled osk 3′UTR by incorporation of fluorescein-12 dUTP into the probe during in vitro transcription (Fig. 3B). Addition of fourfold excess of unlabeled osk 3′UTR completely abolished the binding of labeled RNA, whereas dsDNA could not compete for binding even at eightfold excess. Therefore, we conclude that Osk-C is an RNA binding domain that can interact with its own 3′UTR. We further tested the binding of Osk-C to the 3′UTRs of nos and pgc (Fig. 3C). Osk-C bound to the translational control element (TCE) of nos 3′UTR, but not to the 3′UTR of pgc under the same assay condition and a protein:RNA molar ratio up to 30:1. Osk-C appeared to bind to the 3′UTR of nos stronger than to that of oskar (Fig. 3C). By increasing the protein to RNA ratio we could detect a shift of the 3′UTR of pgc, but the implication is unclear. It should be emphasized that our observation does not exclude the possibility that Osk-C may bind full-length pgc vigorously.

Fig. S4.

Electrostatic potential on the surface of Osk-C. The surface of Osk-C is mostly basic (shown in blue) and binds a number of sulfate ions (shown in a stick model) from the crystallization buffer.

Fig. 3.

EMSA analysis of Osk-RNA interaction. (A) EMSA results show that Osk-C binds osk 3′UTR but Osk-N shows no detectable binding. (B) Fluorescein-labeled osk 3′UTR, shows as green bands, shifted by increasing concentration of Osk-C and the binding is competed by fourfold-excess of unlabeled osk 3′UTR but not dsDNA, even at eightfold excess. (C) Binding of Osk-C to the 3′UTRs of osk, TCE of nanos, and pgc.

RNA Binding Surface of Osk-C.

The structure of Osk-C reveals several positively charged patches on the surface that may be RNA binding sites (Fig. 4A). We chose the residues forming positively charged patches and involved in binding sulfate ions for mutational studies. Eleven single or combinational substitution mutants—R448E, R470E, R490E, R567E, R576E, R593E, R593A, R595E, R436E/R442E, R445E/H446D, and R538E/K541E/R544E—were generated. We were able to express and purify all but the R593E mutant, and they were subjected to EMSA analysis using oskar and nanos 3′UTRs as probes (Fig. 4B and Fig. S5). Compared with wild-type Osk-C, two mutants, R436E/R442E and R576E, showed reduced affinities toward oskar 3′UTR. R436E/R442E almost completely lost the ability to bind oskar 3′UTR. Similar levels of reduction of binding abilities of these two mutants toward nos 3′UTR were observed (Fig. 4C). This analysis identifies the regions occupied by the three residues, Arg436, Arg442, and Arg576 as part of the RNA binding surface. Arg436 is located on α1, and Arg442 and Arg576 are situated in the long loops between α1 and α2 and α5 and α6, respectively. These three residues lie on the same side of the Osk-C surface on a positively charged narrow ridge, which appears adequate for accommodating RNA (Fig. 4D).

Fig. 4.

RNA binding surface of Osk-C. (A) Positively charged residues on Osk-C surface selected for mutational analyses are shown in a stick model superimposed with the surface representation of Osk-C colored according to electrostatic potential distribution. (B and C) Two Osk-C mutants, R436E/R442E and R576E, displayed reduced binding affinity toward the 3′UTRs of osk (B) and nanos (C), compared with that of wild-type Osk-C. (D) Arg436, Arg442, and Arg576 form a narrow positively charged ridge, which may be involved in RNA binding.

Fig. S5.

EMSA results of the binding of eight Osk-C mutants to the 3′UTRs of osk and TCE of nanos.

Discussion

Osk is critically important for germ cell development in Drosophila. Most studies to date have focused on the function and regulation of osk mRNA, and little is known about the biochemical function of Osk protein. It has been reported that Osk interacts with Vasa, Staufen, and Tudor proteins (12), but we were not able to detect robust interactions between Osk and these proteins in our hands by coelution in a gel-filtration column using purified proteins (Fig. S6). Speculations regarding the biochemical properties of Osk from bioinformatic analyses predicted Osk-C to have a hydrolase-like fold and Osk-N to form a LOTUS winged-helix domain with RNA binding properties. Our structural analysis largely confirms the protein-fold predictions, and offers additional and novel insights into the structure and function of Osk. First, the winged-helix LOTUS domain forms a dimer and displays no detectable RNA binding activity. We also do not detect RNA binding activity of monomeric Osk-N using the S210P mutant (Fig. S3B). However, it should be cautioned that we could not absolutely rule out possible RNA binding activities of Osk-N because we only tested a specific set of RNAs. The role of the LOTUS domain and the function of the predicted Oskar protein dimer need yet to be determined in vivo. Besides stop codon mutations that lead to protein truncations, specific point mutations mapping to this region and affecting Oskar’s posterior patterning function have not been identified.

Fig. S6.

Test of interaction between Osk and Vasa. (A) Bacterially expressed full-length Vasa and Osk-N (amino acids 150–240) were mixed and analyzed using a Superdex 200 10/300 gel-filtration column. (Upper) SDS/PAGE showing Vasa and Osk-N are present in two separate elution peaks. (Lower) UV (280 nm) absorption profile of the gel-filtration run. The buffer for gel-filtration contains 20 mM Tris⋅HCl 8.0, 150 mM NaCl, and 0.1% β-ME. (B) Gel-filtration analysis of the interaction between Osk-C and Vasa. All conditions are the same as in A, except that 300 mM NaCl was used in this case. This is because Osk-C has a tendency to precipitate at low salt concentrations.

It comes as a surprise that Osk-C binds RNA. To our knowledge, this is the first time that a lipase-like domain has been reported to be a nucleic acid binding domain. Interestingly, a number of metabolic enzymes have recently been shown to “moonlight” as RNA binding proteins (31). We further find that the 3′UTRs of osk and nanos are potential binding partners of Osk-C. This discovery provides new insights into the biochemical functions of Oskar in embryonic patterning and germ cell development. Mutational studies also identified surface regions of Osk-C that are important for RNA binding. Interestingly, the biochemically identified residues that are key for RNA binding are in proximity to several genetically identified osk alleles defective in posterior pattering (Fig. S7 and Table S2) (32–35). This correlation provides a possible link between the biochemical and developmental functions of Oskar. To gain further insights into the RNA binding property of Osk, we attempted to determine the RNA sequence specificity for Osk-C binding. However, we were unable to discern a pattern by simple deletion analysis. It is possible that there may not be strict RNA sequence specificity for Osk binding, as other factors, such as RNA secondary structures, may regulate RNA binding. Nevertheless, a clearer picture of RNA binding, including a structure of Osk in complex with RNA, should greatly benefit mechanistic understandings of the biological functions of Osk.

Fig. S7.

Location of osk mutations within the protein. Residues corresponding to the genetically identified missense mutations defective in posterior pattering are shown as a stick model with the carbon bonds colored magenta. Three arginine residues identified in our EMSA assay to have an effect in RNA binding are shown in a stick model with the carbon bonds colored green. A ribbon representation of Osk-C is superimposed. Please note that several genetically identified mutations (Table S2) occur at places in proximity to the biochemically identified arginine residues. Dashed lines indicate hydrogen bonds.

Table S2.

Genetically identified osk mutant alleles described in previous studies (34, 35)

Mutant allele	Amino acid change	osk RNA localization in embryo	Abdomen defects	Interactions with R436, R442, R576
osk54	180 = novel stop codon	Not localized	Strong	Not being included in the structure
osk84	254 Arg → Stop	Not localized	Strong	Not being included in the structure
Osk88	N561Y	Posterior	Strong	—
Osk150	R593Q	Posterior	Strong	The positon of R593 maybe important because it lies on the binding ridge formed by R436, R442 and R576. However, we cannot get soluble R593E mutant expressed in Escherichia coli
Osk166	R593W	Posterior	Strong
Osk255	S452L	Posterior	Weak and variable	Side-chain hydroxyl group of S452 interacts with main chain carbonyl group of R442, distance 2.7 Å
Osk301	P424S	18 °C Posterior	Weak to none (ts)	—
		29 °C Posterior	Strong
Osk366	324 Gln → Stop	Not localized	Strong	Not being included in the structure
oskCE1	A435T	18 °C Posterior	Weak to none (ts)	—
		29 °C Posterior	Strong
oskCE2	P584L	18 °C Posterior	Weak to none (ts)	—
		29 °C Posterior	Strong
oskCE3	T361I	Posterior	Strong	Main chain carboxyl group of C441 interact with ω-NG of R436, distance 3.1 Å; Side chain mercapto group of C441 interact with main chain amide group of R436, distance 3.4 Å
	C441Y
oskCE4	R595H	18 °C Posterior	Weak to none (ts)	—
		29 °C Posterior	Strong
Osk123	E438K	Posterior	Strong	Side chain hydroxyl group form hydrogen bonds with ε-NG of R593, distance 3.1 Å
oskGB10	S457F	Posterior	Strong	—
Osk4G2	V566E	Posterior	Strong	—

Possible structural relationship between some of these alleles and the biochemically identified arginine residues are noted.

Experimental Procedures

Protein Expression and Purification.

The short isoform of Oskar, generated from translation initiated from an internal start codon, spans residues 139–606. cDNA fragments encoding Osk-N (residues 150–240) and a L199M mutant (residues 150–224) were cloned into a pET28a-SUMO vector between the BamHI and SalI sites and expressed in the BL21-Codon Plus (DE3)-RIL strain of Escherichia coli. Once the cell density reached OD₆₀₀ ∼0.8, protein expression was induced with 0.3 mM isfopropyl-β-d-thiogalactopyranoside (IPTG) at 16 °C for 18 h. Cells were harvested by centrifugation at 4,670 × g and lysed by passing through an Emulsiflex C3 cell disruptor in the lysis buffer (20 mM Tris⋅HCl pH 8.0, 500 mM NaCl, 10 mM imidazole). The homogenate was centrifuged at 26,400 × g and supernatant was loaded onto a Ni-NTA (Novagen) column pre-equilibrated with the lysis Buffer. After extensive washing with the washing buffer (20 mM Tris⋅HCl pH 8.0, 500 mM NaCl, 20 mM imidazole), the his-SUMO tagged protein was eluted with the elution buffer (20 mM Tris⋅HCl pH 8.0, 500 mM NaCl, 500 mM imidazole). The his-SUMO tag was cleaved with SUMO protease during dialysis against buffer A (20 mM MES pH 6.0, 100 mM NaCl, 0.1% β-ME). The tag-free protein was loaded onto a heparin column pre-equilibrated with buffer A and eluted with a 100–1,000 mM NaCl gradient. The desired fractions were pooled, concentrated and further purified on a Superdex 75 10/60 XK column (GE Healthcare) in a buffer containing 20 mM Tris⋅HCl pH 8.0, 500 mM NaCl, and 0.1% β-ME.

Osk-C and its mutants were expressed using a modified pET28a-SUMO vector with the coding sequences inserted between the BamHI and SalI sites. Protein expression and purification follow the same procedures as that for Osk-N, except that buffer A used in the Heparin column contains 20 mM Tris⋅HCl pH 8.0, 150 mM NaCl, 0.1% β-ME, 0.5 mM EDTA, and the final step of size-exclusion column chromatography was carried out using a Superdex 75 10/300 GL column (GE Healthcare) in a buffer containing 20 mM Tris⋅HCl pH 8.0, 300 mM NaCl, 0.1% β-ME, and 0.5 mM EDTA.

Crystallization and Structure Determination.

Osk-N concentrated to ∼5 mg/mL was crystallized using a reservoir buffer containing 100 mM sodium acetate pH 5.6, 20% (vol/vol) PEG 400 and 200 mM magnesium acetate by hanging-drop vapor diffusion at 16 °C. The L199M mutant was concentrated to 6.5 mg/mL and the crystallization buffer contains 100 mM sodium acetate pH 5.6, 17–22% (vol/vol) PEG 3350 and 150 mM NaCl. Osk-C was concentrated to ∼15 mg/mL and crystals for X-ray diffraction were obtained from a condition containing 100 mM sodium acetate pH 4.6, 800 mM ammonium sulfate and 200 mM lithium sulfate, also by hanging-drop vapor diffusion at 16 °C.

X-ray diffraction data were collected at Beamline BL17U of Shanghai Synchrotron Radiation Facility using an ADSC Q315r detector. All diffraction data were processed using HKL2000 (36). The 2.5 Å SAD data for the SeMet L199M mutant were collected at wavelength 0.9790 Å. The crystal belongs to the P21 space group and there are six molecules in one asymmetric unit. The 2.1 Å SAD data of SeMet Osk-C were collected at wavelength 0.9788 Å. The crystal belongs to the P3121 space group and there is one molecule per asymmetry unit. In both cases, six Se sites were found using SHELXD (37) and phasing was carried out with PHENIX (38). A 3.0 Å dataset of SeMet wild-type Osk-N (150–240) was collected at wavelength 0.9641 Å. The crystal belongs to the P32 space group and there are six molecules in one asymmetric unit. The structure was solved by molecular replacement with PHASER (39) using the L199M structure as the search model. Initial model building, model fitting, and refinement of all of the three structures were done with PHENIX and COOT (40). Detailed statistics for data collection and refinement are shown in Table S1.

MALS.

Osk-N was diluted to 2 mg/mL in a buffer containing 20 mM Tris⋅HCl pH 8.0 and 150 mM NaCl. After being centrifuged at 15,871 × g for 15 min, a 50-μL sample was loaded onto a WTC-030S5 size-exclusion column (Wyatt Technologies) connected to a MALS instrument (DAWN HELEOS II; Wyatt Technologies). Molecular mass of particles in a single elution peak was calculated based on light scattering data using the ASTRA software package.

In Vitro Transcription of RNA.

cDNAs encoding osk 3′UTR, nanos 3′UTR, and pgc 3′UTR were cloned into a pBluescript II SK(+) vector between KpnI and HindIII sites for in vitro transcription. The recombinant plasmid DNAs were linearized with HindIII downstream of the insert to be transcribed as the template. The RNAs were transcribed with T7 polymerase by MEGAscript Kit (AM1334, Ambion) following the procedures recommended by the manufacturer. For in vitro transcription of fluorescein-labeled osk 3′UTR, fluorescein-12-UTP (Roche) was added in the reaction together with the unlabeled one. The transcribed RNAs were purified with MEGAclear Purification Kit (Ambion). Quality and quantity of RNAs were examined both by UV spectroscopy and 5% denaturing polyacrylamide gel.

EMSA.

In each EMSA reaction (total volume, 20 μL), an 80 ng heat-denatured RNA probe of osk 3′UTR, nanos 3′UTR, or pgc 3′UTR was incubated individually with various quantities of Osk or Osk mutants in a binding buffer containing 20 mM Tris⋅HCl, pH 7.9, 1 mM DTT, 10 mM KCl, 15 mM NaCl, 40 μg/mL calf BSA, and 5% glycerol for 30 min on ice. After incubation, protein-bound and free RNAs were separated by electrophoresis on nondenaturing 6% polyacrylamide gels (80:1) running in a 0.5×TBE buffer at 10 V/cm⁻¹ at 4 °C. For unlabeled RNA probes, gels were stained by SYBR Gold (Life Technologies) and scanned by FluorChem M System (ProteinSimple). For fluorescein-labeled RNAs, gels were directly scanned at 607 nm by FluorChem M System (ProteinSimple). The RNA probes were quantified by UV spectrophotometer Nanodrop 2000 (Thermo Scientific) at wavelength 260 nm.

Accession Numbers.

Atomic coordinates and diffraction data of the structures have been deposited with PDB under accession codes 5CD8, 5CD7 and 5CD9 for Osk-N, its L199M mutant, and Osk-C, respectively. Structural figures were prepared using Pymol (www.pymol.org).

Note Added in Proof.

While this manuscript was being reviewed, a report by Jeske et al. describing similar studies of Osk was published online (41). Crystal structures of Osk fragments equivalent to Osk-N and Osk-C denoted in our work were reported, and the structures are nearly identical (rmsd 0.34 Å and 0.69 Å for Osk-C and Osk-N, respectively), supporting the dimeric form of Osk-N from both of our studies. Using UV cross-linking and immunoprecipitation, Jeske et al. (41) showed that full-length short Oskar binds to nos, pgc, and gcl mRNAs in vivo, and the RNA binding activity was attributed to Osk-C based on in vitro pulldown results. In our work, we showed that purified Osk-C directly binds the 3′UTR regions of osk and nos in vitro. We detected no meaningful binding of Osk-C to the 3′UTR of pgc in our assay, but we cannot rule out that Osk-C may bind full-length pgc mRNA. Another difference between the two studies is that Osk-N was shown to interact with Vasa by Jeske et al. using GST-pulldown, isothermal titration calorimetry, and yeast two-hybrid assays. We were not able to detect such interaction by gel-filtration column chromatography using purified proteins.