Makorin 1 controls embryonic patterning by alleviating Bruno1-mediated repression of oskar translation

Makorins are evolutionary conserved proteins that contain C3H-type zinc finger modules and a RING E3 ubiquitin ligase domain. In Drosophila, maternal Makorin 1 (Mkrn1) has been linked to embryonic patterning but the mechanism remained unsolved. Here, we show that Mkrn1 is essential for axis specification and pole plasm assembly by translational activation of oskar (osk). We demonstrate that Mkrn1 interacts with poly(A) binding protein (pAbp) and binds specifically to osk 3’ UTR in a region adjacent to A-rich sequences. Using Drosophila S2R+ cultured cells we show that this binding site overlaps with a Bruno1 (Bru1) responsive element (BREs) that regulates osk translation. We observe increased association of the translational repressor Bru1 with osk mRNA upon depletion of Mkrn1, indicating that both proteins compete for osk binding. Consistently, reducing Bru1 dosage partially rescues viability and Osk protein level in ovaries from Mkrn1 females. We conclude that Mkrn1 controls embryonic patterning and germ cell formation by specifically activating osk translation, most likely by competing with Bru1 to bind to osk 3’ UTR.

Makorins are evolutionary conserved proteins that contain C 3 H-type zinc finger modules and a RING E3 ubiquitin ligase domain. In Drosophila, maternal Makorin 1 (Mkrn1) has been linked to embryonic patterning but the mechanism remained unsolved. Here, we show that Mkrn1 is essential for axis specification and pole plasm assembly by translational activation of oskar (osk). We demonstrate that Mkrn1 interacts with poly(A) binding protein (pAbp) and binds specifically to osk 3' UTR in a region adjacent to A-rich sequences. Using Drosophila S2R+ cultured cells we show that this binding site overlaps with a Bruno1 (Bru1) responsive element (BREs) that regulates osk translation. We observe increased association of the translational repressor Bru1 with osk mRNA upon depletion of Mkrn1, indicating that both proteins compete for osk binding. Consistently, reducing Bru1 dosage partially rescues viability and Osk protein level in ovaries from Mkrn1 females. We conclude that Mkrn1 controls embryonic patterning and germ cell formation by specifically activating osk translation, most likely by competing with Bru1 to bind to osk 3' UTR.

Author summary
To ensure accurate development of the Drosophila embryo, proteins and mRNAs are positioned at specific sites within the embryo. Many of these factors are produced and localized during the development of the egg in the mother. One protein essential for this process that has been heavily studied is Oskar (Osk), which is positioned at the posterior pole. During the localization of osk mRNA, its translation is repressed by the RNA-binding protein Bruno1 (Bru1), ensuring that Osk protein is not present outside of the posterior where it is harmful. At the posterior pole, osk mRNA is activated a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
In the Drosophila embryo, the maternally deposited pole plasm is a site of specialized translation of mRNAs required for germ cell specification and posterior patterning [1]. Numerous mRNAs accumulate in the pole plasm during oogenesis and early embryogenesis through several different localization mechanisms [2,3]. Among these mRNAs is oskar (osk), which localizes during oogenesis to the posterior along a polarized microtubule network [4] and via a trapping mechanism [5]. Several lines of evidence indicate that osk is the primary determinant that specifies germ cells and posterior patterning. Ectopic expression of osk at the anterior can induce a second set of pole cells and a bicaudal embryonic segmentation pattern with mirrorimage posterior segments [6,7]. Mutations such as Bicaudal-D (Bic-D), ik2, and others that produce a duplicated anterior focus of osk mRNA also produce bicaudal embryos [8][9][10]. Conversely, embryos from females carrying hypomorphic loss-of-function mutations of osk lack posterior segmentation and pole cells [11]. Mutations in a number of other genes can produce a similar phenotype, and these are collectively known as posterior-group genes [12]. Some of these genes (for example cappuccino, chickadee, spire and staufen (stau)) are required for posterior localization of osk. A failure to deploy osk produces the posterior-group phenotype in these mutants [13,14]. Other posterior-group genes (for example vasa (vas), tudor, nanos (nos), and aubergine (aub)), produce mRNAs and/or proteins that also accumulate in pole plasm and operate downstream of osk [15][16][17].
osk translation is under elaborate temporal and spatial regulation, ensuring that Osk protein becomes abundant only in the posterior pole plasm and not before stage 9 of oogenesis [18,19]. A key repressor of osk translation prior to that stage and outside the pole plasm is Bruno1 (Bru1), which interacts with binding sites called Bru1 response elements (BREs) in the osk 3' UTR. Mutations affecting the BREs result in premature and excessive Osk expression [18][19][20]. Two mechanisms have been proposed for Bru1-mediated repression of osk translation. In the first, Bru1 recruits Cup, which inhibits assembly of an active cap-binding complex by competitively inhibiting eIF4G for binding to eIF4E [21]. The second mechanism involves oligomerization of osk mRNA into large ribonucleoprotein particles (RNPs) that are inaccessible to the translational machinery [22][23][24]. These mechanisms may be connected, in that physically concentrating osk mRNA molecules in RNPs can enable regulation in trans through inter-molecular interactions. For instance, Bru1 bound to the 3' end of one osk mRNA molecule could recruit Cup to eIF4E bound to the 5' cap structure of another osk mRNA molecule, thus repressing its translation [25,26]. Importantly, osk translation and osk mRNA localization are tightly coupled. In nonsense osk alleles or when 3' UTR elements required for translational activation are mutated, osk mRNA localizes transiently to the pole plasm but its accumulation is not maintained [9,27,28].
While several proteins have been implicated in activating osk translation in the pole plasm [14,29,30], a comprehensive picture of how this is achieved has not yet emerged. For instance, it has been proposed that activation of osk translation involves inhibition of Bru1 [24]. Related to this, a BRE-containing region in the distal part of the osk 3' UTR (BRE-C) functions in repression as well as in activation [25]. Nevertheless, the mechanism underlying the dual function of this element has not yet been solved.
Large-scale in situ hybridization screens have identified many other mRNAs that localize to the pole plasm [3,31], and some of the corresponding genes could potentially be involved in osk regulation. To search for new posterior-group genes, we previously expressed shRNAs targeting 51 different mRNAs that accumulate in the pole plasm to determine if doing so produced defects in posterior patterning or pole cell formation. We observed that a substantial proportion of embryos produced by Makorin 1 (Mkrn1) knockdown females showed a posterior-group phenotype [32]. Makorin proteins are conserved in plants, fungi, and animals, and contain a RING-domain as well as one or more C 3 H-type zinc fingers (ZnF) [33]. The role of Makorin proteins is somewhat enigmatic despite their widespread evolutionary conservation. Mammalian MKRN1 has been identified as an E3 ubiquitin ligase that promotes degradation of target proteins [34][35][36], but proteomic analysis does not support an association with proteasome components [37,38]. Furthermore MKRN1's shorter isoform stimulates translation in rat forebrain neurons but the mechanism is currently unknown [39].
Here, we analyzed the function of Mkrn1 during oogenesis and early embryogenesis. We generated several alleles that alter different domains of the Mkrn1 coding sequence. Using these mutants, we found that Mkrn1 is required for accumulation of Osk protein at the posterior pole of the oocyte. We also show that Mkrn1 is not required for osk mRNA localization, but essential for its translation. Furthermore, we present evidence that Mkrn1 directly binds to the osk 3' UTR via its N-terminal zinc finger domain. Using Drosophila S2R+ cells we further found that binding of Mkrn1 to osk partially overlaps with the BRE-C domain, adjacent to an A-rich region that recruits pAbp to the osk 3' UTR [40]. Moreover, the association between Mkrn1 and osk mRNA is stabilized by the physical association with pAbp. Strikingly, depletion of Mkrn1 results in an increased level of Bru1 binding to osk 3' UTR and reduction of bru1 gene dosage partially rescues Mkrn1 mutant phenotypes. Based on these data we propose that Mkrn1 competes with Bru1 for osk mRNA binding, thus positively regulating osk translation and explaining the specific role of BRE-C in translational activation.

The Drosophila genome includes four Makorin-related genes
In many organisms, up to four distinct genes encoding members of the Makorin family exist, but only one such gene, Mkrn1, has been annotated in Drosophila. To investigate whether flies are unusual in this regard, we searched for sequences similar to human MKRN1. This analysis uncovered four Drosophila genes, Mkrn1, CG5334, CG5347, and CG12477, with substantial similarity to MKRN1 (S1A Fig). All four predicted polypeptides from these genes contain a region of approximately 130 amino acids that is highly conserved and contains a RINGdomain as well as C 3 H-type zinc fingers (ZnF). The proteins are otherwise more divergent from one another, with the exception that all but CG12477 contain a ZnF domain near the amino-terminus.
To analyze the differences in their functionalities, we first determined the expression profile of all four Makorin genes during development. Mkrn1 mRNA is expressed at detectable levels at all developmental stages ( Fig 1A) and clearly peaks in early (0-2.5 h) embryos and ovaries. In contrast, expression of the other three Makorin genes is undetectable during early development but peaks in pupae and adult males (S1B Fig). From these results, we conclude that Mkrn1 is the gene of the family most predominantly expressed in ovaries and early embryos and suggest that the three other genes could be specifically expressed in testes.

Mkrn1 mutants reveal essential roles in oogenesis and embryogenesis
To elucidate the role of Mkrn1, we used CRISPR/Cas9 to produce three different mutant alleles: Mkrn1 N , a complete deletion of the coding sequence, Mkrn1 S , a frameshift mutation that is predicted to produce a C-terminally truncated protein of 124 amino acids, including  only ZnF1 among conserved domains, and Mkrn1 W , a small in-frame deletion that disrupts only the ZnF1 domain ( Fig 1B). In the strong Mkrn1 mutants (Mkrn1 S and Mkrn1 N ), most egg chambers cease development at or before stage 10 ( Fig 1C-1J). The nuclei of Mkrn1 S oocytes that progress as far as stage 9 or later remain at the posterior, failing to migrate to the anterodorsal corner (Fig 1M). The very few eggs laid by Mkrn1 S females have no dorsal appendages and do not develop ( Fig 1Q). Mkrn1 N egg chambers do not progress as far as stage 9 and showed variable defects in early oogenesis including failure of oocyte differentiation ( Fig 1N) and inappropriate follicle cell migration (Fig 1J and 1N). On the other hand, ovaries of females homozygous for Mkrn1 W have a similar morphology to wild-type (Fig 1C and 1D). Mkrn1 W mutant ovaries completed oogenesis and produced fertilizable eggs in similar numbers as wild-type controls (Fig 1G, 1H, 1K and 1L, Table 1). To examine the role of Mkrn1 in embryonic patterning, we compared cuticle preparations of wild-type and Mkrn1 W embryos (Fig 1O  and 1P). We found that most Mkrn1 W embryos lack posterior segments, a phenotype similar to osk mutants and to what we previously observed at lower frequency from females expressing shRNA targeting Mkrn1 [32]. To investigate this more closely, we stained wild-type and Mkrn1 W embryos for Fushi tarazu (Ftz) and Vas proteins. In wild-type blastoderm-stage embryos, Ftz is expressed in seven stripes along the anterior-posterior axis, while in posteriorgroup embryos the number of Ftz stripes is usually reduced to four (Fig 1R and 1T, [41]). At blastoderm stage in wild-type embryos, Vas-positive pole cells are clustered at the posterior pole [16] (Fig 1R). Consistent with a posterior-group phenotype, we observed four Ftz stripes in 65% of Mkrn1 W embryos (30/46) (Fig 1S and 1U). Most of the remaining embryos had even fewer Ftz stripes with an additional broad domain of Ftz expression. In addition, 95% of Mkrn1 W embryos (63/66) showed no pole cells by Vas staining (Fig 1S). The posterior patterning defects result in lethality for most (97%) Mkrn1 W embryos. However, the small number of Mkrn1 W embryos that hatch into viable larvae (40/1222, 3.3%, Table 1) can complete development to adulthood.

Mkrn1 accumulates in the pole plasm during oogenesis
To examine the distribution of Mkrn1 in the germline, we expressed transgenic Venus-and FLAG-tagged Mkrn1 using a nos>GAL4 driver. Since Mkrn1 females could be rescued to fertility by germline specific expression of these transgenes (Table 1), we concluded that these tagged transgenes are functional, and thus inferred that their localization should reflect the endogenous one. When expressed in ovaries, Venus-or FLAG-tagged Mkrn1 (subsequently called Mkrn1) becomes detectable in a uniform distribution in germline cells from early oogenesis. We observed a mild accumulation of Mkrn1 in cytoplasmic particles resembling  To obtain insights into the link between Mkrn1 and pole cell determination, we collected embryos from females trans-heterozygous for a Mkrn1 allele and for either a vas or osk allele. Next, we compared the number of pole cells with single heterozygous controls. When heterozygous, Mkrn1 W or Mkrn1 S had little effect on pole cell number. However, either allele reduced the number of pole cells produced by vas 1 heterozygotes, and further by osk 54 heterozygotes (S4 Fig). These data support a genetic interaction between Mkrn1 and genes involved in embryonic patterning and pole cell specification.

Mkrn1 ensures correct deployment of specific mRNAs and proteins involved in embryonic patterning
To address whether Mkrn1 mutations may affect the distribution of proteins involved in embryonic patterning, we performed immunostaining experiments. We found a striking reduction in posterior accumulation of Osk in oocytes from all Mkrn1 alleles (Fig 3A-3D). For Stau, we observed weaker and more diffuse posterior localization in Mkrn1 W , as compared to wild-type (Fig 3E and 3F) and no localized protein in Mkrn1 S and Mkrn1 N (Fig 3G and 3H). On the other hand, Grk localized normally to the antero-dorsal corner of Mkrn1 W oocytes ( Fig  3I and 3J). In Mkrn1 S , Grk was observed at reduced levels associated at the posterior with the mislocalized oocyte nucleus ( Fig 3K) and diffusely distributed at a very low level in Mkrn1 N (Fig 3L). Posterior localization of Aub and Vas was lost in oocytes of all Mkrn1 mutant alleles (Fig 3M-3T). Finally, Orb localization was unaffected in Mkrn1 W (Fig 3U and 3V), but was concentrated at the posterior in Mkrn1 S (Fig 3W). Many Mkrn1 N egg chambers included a single Orb-positive cell (Fig 3X), indicating that, in these cases, oocyte differentiation had taken place. Importantly, normal accumulation of all proteins could be restored by  To investigate whether the primary effect of Mkrn1 W on Osk deployment occurred at the level of RNA localization or translation, we examined more closely the effects of Mkrn1 W on osk expression. Therefore, we performed in situ hybridization for osk mRNA and immunostaining for Osk protein on the same samples ( Fig 4A). In stage 7 egg chambers, osk mRNA accumulation is robust in both wild-type and Mkrn1 W oocytes. Faint expression of Osk protein is occasionally visible in wild-type stage 7 oocytes but never in similarly staged Mkrn1 W oocytes. This difference becomes more pronounced in stages 9 and 10A. While 95% (41/43) of stage 9-10A Mkrn1 W oocytes show posterior accumulation of osk mRNA, only 7% (3/43) of Mkrn1 W oocytes show even a faint posterior signal for Osk protein. All (20/20) similarly-staged wildtype oocytes show strong posterior accumulation of both osk mRNA and Osk protein ( Fig 4A). Subsequently, posterior accumulation of osk mRNA is lost in Mkrn1 W oocytes. In 88% of stage 10B Mkrn1 W oocytes (30/34), osk mRNA remains in a tight focus, but is no longer anchored at Accumulation of Orb is similar in wild-type and Mkrn1 W oocytes, but Orb is more concentrated in the posterior of Mkrn1 S oocytes. In early-stage Mkrn1 N egg chambers there is usually a single Orb-positive cell, indicating that some steps toward oocyte differentiation are able to take place. Scale bars, (U-W) 20 μm, (X) 25 μm.
https://doi.org/10.1371/journal.pgen.1008581.g003 In wild-type oocytes posterior accumulation of osk mRNA and Osk protein is robust and stable from stage 9 onward. In Mkrn1 W oocytes accumulation of osk mRNA resembles the wildtype pattern through stage 10A but is not maintained, while Osk protein is rarely detectable at the oocyte posterior at any stage. Scale bars, 25 μm. (B) Western blot analysis from ovary lysates of various genotypes stained for Osk, pAbp and β-tubulin. Osk protein levels are greatly reduced in all Mkrn1 mutant alleles. 1 day-old young females have not yet completed oogenesis and were used as a control for Mkrn1 S and Mkrn1 N ovaries which also lack late-stage egg chambers, where Osk is most abundant. (C) RT-qPCR experiments measuring ovarian osk mRNA levels (normalized to RpL15 mRNA) in the same genotypes as (B). mRNA levels of ovaries from adult females were compared to Mkrn1 W ovaries. For Mkrn1 S and Mkrn1 N ovaries, mRNA levels of 1 day-old young ovaries was used as normalization. Error bars depict Stdev, n = 2.
https://doi.org/10.1371/journal.pgen.1008581.g004 the posterior pole ( Fig 4A). osk translation remains repressed with only 6% of stage 10B Mkrn1 W oocytes (2/34) having detectable posterior Osk. In later Mkrn1 W oocytes, neither osk mRNA nor Osk protein is detectable at the posterior. We conclude from these experiments that Mkrn1 is not directly required for the localization of osk mRNA at the posterior pole but is necessary for its translation. Consistent with previous observations in other conditions that abrogate Osk anchoring or translation [9,27,28,42], posterior accumulation of osk mRNA is not maintained in Mkrn1 W oocytes because of their failure to accumulate localized Osk protein. To confirm these results, we compared Osk protein levels by western blot analysis and osk mRNA levels by quantitative PCR (qPCR). Consistent with our immunostaining experiments, we observed a pronounced reduction in Osk protein (Fig 4B), but only minor positive or negative effects on osk mRNA depending on the particular Mkrn1 allele (Fig 4C).

Mkrn1 associates with factors involved in osk mRNA regulation
To gain further insights into the molecular pathways underlying Mkrn1 function we sought to identify potential cofactors. For this purpose, we expressed Myc-tagged Mkrn1 in Drosophila S2R+ cultured cells and carried out immunoprecipitation (IP) experiments followed by mass spectrometry analysis. We also repeated this experiment using a version of Mkrn1 carrying a point mutation in the RING domain (Mkrn1 RING ), as we noticed that this construct was expressed at a higher level compared to the wildtype one, which appears to be unstable after transfection in cells (S7A and S7B Fig). Similar stability characteristics have been reported for mammalian MKRN1 [34]. Numerous RNA-binding proteins were enriched after IP (S7C and S7D Fig). Among those, several have been already linked to osk mRNA localization and translation [21,40,[43][44][45]. To validate these interactions, we performed co-IP experiments in S2R+ cells between Mkrn1 RING and various interaction partners in the presence or absence of RNase T1. Using this approach, we could confirm the interaction of Mkrn1 RING with poly(A) binding protein (pAbp), IGF-II mRNA-binding protein (Imp), eukaryotic initiation factor 4G (eIF4G), Squid (Sqd) and maternal expression at 31B (Me31B) (S8A- S8E Fig). All of these interactions persisted upon RNase treatment, suggesting they are direct. We can however not exclude the possibility that the poly(A) tail, which is not affected by treatment with RNase T1, mediates these interactions. Interestingly, several of the identified proteins have already been shown to interact with each other [43,44,[46][47][48]. We also confirmed that interactions between FLAGtagged Mkrn1 and pAbp as well as eIF4G also occur in ovaries ( Fig 5A and S8F Fig).
Furthermore, we found that the stability of Mkrn1 itself was enhanced upon co-transfection of pAbp in S2R+ cells ( Fig 5B). Mammalian MKRN1 contains a PCI/PINT associated module 2 (PAM2) motif, that is present in several pAbp binding proteins and is required in MKRN1 for binding to PABP [39,49]. We identified a similar motif in Drosophila Mkrn1 (Fig 5C), but with one variation compared to human (V instead of E at position 9) that likely explains why this PAM2 motif was not recognized previously. To address the functionality of this motif we repeated the co-IPs, and found that when this domain was mutated (Mkrn1 PAM2 , Fig 5C), the interaction between Mkrn1 and pAbp was compromised in S2R+ cells (Fig 5D). Based on these data, we conclude that Mkrn1 exists in one or several complexes that contain factors involved in the regulation of osk mRNA translation, and stably interacts with pAbp via its PAM2 motif. If the interaction of Mkrn1 with pAbp is indeed required for the function of Mkrn1, Mkrn1 PAM2 should not be able to rescue the Mkrn1 N phenotype. Indeed, we found that overexpression of Mkrn1 PAM2 in Mkrn1 N ovaries could not rescue Osk protein levels and posterior localization compared to wild-type Mkrn1 (Fig 5E and S9B Fig). Moreover, we observed a strong modifier effect between Mkrn1 W and pAbp mutations; Mkrn1 W homozygotes and pAbp heterozygotes complete oogenesis, but many egg chambers from pAbp/+; Mkrn1 W females show severe defects, including supernumerary germ cells, missing or duplicated oocytes, or more extreme dysmorphologies (Fig 5F). These females can produce eggs, but only for the first 2-3 days of life (Fig 5G), and these eggs did not support embryonic development. We thus conclude that the interaction of Mkrn1 with pAbp is essential for its function during oogenesis.

Mkrn1 associates specifically with osk mRNA in vivo
The effects we observed on osk translation in late Mkrn1 egg chambers prompted us to test whether Mkrn1 can interact with osk mRNA. First, to assess whether Mkrn1 can bind to RNA in general, we immunoprecipitated FLAG-Mkrn1 after transfection of S2R+ cells and UV crosslinking. Mkrn1-bound RNA was subsequently labeled, and the protein-RNA complexes were visualized by autoradiography (S10A To confirm that Mkrn1 binds osk mRNA in vivo, we overexpressed either FLAG-tagged wild-type Mkrn1 or Mkrn1 ΔZnF1 in ovaries and performed RNA IP (RIP) experiments. The enrichment of different mRNAs was analyzed by qPCR using primers that bind to the 3' UTRs of the respective transcripts. Interestingly, we observed that osk mRNA was substantially enriched in Mkrn1 IPs, but much less so when using Mkrn1 ΔZnF1 (Fig 6A and S10B Fig). On the other hand, bicoid (bcd) and grk mRNAs were not detected above background levels in either RIP experiment. This provides evidence that Mkrn1 binds specifically to osk mRNA in ovaries and that its ZnF1 domain is important for this interaction.
To determine precisely where Mkrn1 binds to osk mRNA, we performed individual nucleotide resolution cross-linking and immunoprecipitation (iCLIP), an unbiased approach to identify RNA-binding sites of a protein of interest [50]. This condition ensured that all Mkrn1 protein-RNA complexes were potentially recoverable in the immunoprecipitations. To this end, we overexpressed FLAG-tagged Mkrn1 in S2R+ cells and performed iCLIP experiments (S10C Fig). Since osk is poorly expressed in these cells, we co-transfected a genomic construct of osk under the control of an actin promoter. We found specific binding sites in a handful of genes, including osk, ranking at third position in term of read coverage (S10D Fig), in which Mkrn1 binding sites were located in the distal part of the 3' UTR ( Fig 6B). These sites fall just upstream of an A-rich sequence that associates with pAbp [40]. Moreover, the binding site of Mkrn1 partially overlaps with the BRE-C site, which is bound by Bru1 and is required for both repression and activation of osk translation [25].
To validate the identified Mkrn1 binding sites, we performed RIP experiments in S2R + cells using different 3' UTRs fused to the firefly luciferase coding sequence. We found that Mkrn1 binds strongly to osk 3' UTR but not significantly above background levels to grk 3' UTR (S11A Fig). In addition, deletion of the Mkrn1-bound site identified with iCLIP in cells (osk ΔMkrn1 , deletion of nucleotides 955-978 of osk 3' UTR) greatly reduced the interaction of Mkrn1 to osk 3' UTR ( Fig 6C and S11B Fig).
As the Mkrn1 binding site in osk 3' UTR lies just upstream of the A-rich region (AR), we wondered whether the AR would also have an impact on Mkrn1 binding. To test this possibility, we deleted the AR (osk ΔAR , deletion of nucleotides 987-1019 of osk 3' UTR) and examined Mkrn1 binding in S2R+ cells. We observed a decrease of Mkrn1 binding similar to that observed when deleting the Mkrn1 binding sites (Fig 6D and S11C Fig). Thus, we conclude that the AR enhances Mkrn1 binding to osk 3' UTR. As Mkrn1 forms a stable complex with pAbp, our results further suggest that pAbp binding to the AR stabilizes Mkrn1 and therefore enhances its interaction with osk. Accordingly, reducing pAbp levels by RNAi, but not the level of Imp, another Mkrn1 interactor, dramatically decreased Mkrn1 association with osk mRNA in Drosophila cultured cells (Fig 6E, S11D and S11E Fig). Mutation of the PAM2 domain that enables the interaction with pAbp also resulted in reduced binding to osk 3' UTR ( Fig 6F and S11F Fig). Consistent with these results, mutating both the ZnF1 domain and the PAM2 motif led to almost complete loss of Mkrn1 binding to osk 3' UTR in S2R+ cells. Collectively, our results indicate that Mkrn1 binds specifically to the 3' end of osk 3' UTR via its ZnF1 domain and this association is further stabilized through the interaction with pAbp.

Mkrn1 competes with Bru1 for binding to osk 3' UTR
Our observation that Mkrn1 binds to the osk BRE-C region prompted us to test whether Mkrn1 and Bru1 may compete for binding to osk 3' UTR. To this end, we first examined whether we can recapitulate Bru1 binding to osk mRNA in S2R+ cells. As Bru1 is normally not expressed in this cell type, cells were co-transfected with GFP-tagged Bru1 along with the luciferase-osk-3'UTR reporter. RIP experiments confirmed previous findings that Bru1 strongly associates with osk 3' UTR (S12A Fig, [18]). We next repeated this experiment upon knockdown of Mkrn1 mRNA. Strikingly, while the protein level of Bru1 was not affected (S12B Fig), its binding to osk 3' UTR was significantly increased in Mkrn1-depleted cells ( Fig  7A and S12B-S12D Fig). As pAbp is required to stabilize the interaction of Mkrn1 to osk 3' UTR, we wondered whether it is necessary for modulating Bru1 binding as well. Indeed, knockdown of pAbp in S2R+ cells resulted in an increased association of Bru1 to osk 3' UTR ( Fig 7B and S12E Fig).
Next, to assess if these observations made in cultured cells are relevant to Mkrn1 function in vivo, we repeated RIP experiments using ovarian extracts. Bru1 binding was assessed using an antibody directed against endogenous Bru1 and its association with osk mRNA was subsequently analyzed by qPCR. Similar to S2R+ cells, the interaction of Bru1 with osk was significantly increased in Mkrn1 W mutant ovaries (Fig 7 and S12F Fig). Thus, we conclude that Mkrn1 restricts Bru1 binding to osk 3' UTR and this effect is enhanced by the interaction of Mkrn1 with pAbp.
To further address the relationship between Mkrn1 and Bru1, we examined whether the Mkrn1 W mutation affects Bru1 accumulation during oogenesis. In wild-type ovaries, Bru1 is expressed in all germline cells and accumulates to a modest degree in the oocyte during early oogenesis (Fig 7D and 7E, [20]). However, in Mkrn1 W ovaries, oocyte accumulation of Bru1 during early stages was much more pronounced (Fig 7F and 7G). As osk mRNA accumulates in early oocytes, this result is consistent with Bru1 having an increased binding affinity for osk in the absence of Mkrn1, even though Bru1 can also accumulate in the absence of osk RNA [51].
If Mkrn1 activates osk translation by displacing Bru1, we would predict that lowering bru1 genetic dosage should suppress the Mkrn1 phenotype. To test this hypothesis, we used a strong bru1 allele (bru1 QB , [20,52]). We found that removing one copy of bru1 was sufficient to substantially rescue Osk protein level and posterior localization in Mkrn1 W female oocytes as analyzed by immunostainings (Fig 7H-7M). Semi-quantitative analysis of Osk-immunostained ovaries confirmed a clear suppression of the Mkrn1 W phenotype by one copy of bru1 QB ( Fig  7N) that was further confirmed by immunoblotting (Fig 7O). We also observed a higher survival rate of embryos produced from Mkrn1 W females that were heterozygous for bru1 QB , as compared to controls (620/2058, 30.1% vs 40/1222, 3.3%). Taken together, these experiments demonstrate that Mkrn1 activates osk translation, most likely by displacing Bru1 binding at the osk 3' UTR (Fig 8).

Discussion
Our data indicates that Mkrn1 is essential for oogenesis, embryonic patterning, and germ cell specification. An essential role for Mkrn1 in oogenesis has also been recently reported [53]. By taking advantage of a new allele that specifically disrupts Mkrn1 binding to RNA, we demonstrate that Mkrn1 exerts its function in embryogenesis and germ cell specification, primarily via regulating osk translation by antagonizing Bru1 binding.
Control of osk translation has been studied in depth, revealing a complex spatio-temporal interplay between repressing and activating factors [19]. Relief of translational repression and activation of osk translation is likely to involve multiple redundant mechanisms. For example, Bru1 can be phosphorylated on several residues, and phosphomimetic mutations in these residues inhibit Cup binding in pulldown assays. However, these do not seem to affect translational repression activity in vivo [24]. Stau, Aub, Orb and pAbp have also been implicated in activating osk translation [14,29,40,54]. However, it is unlikely that Mkrn1 controls osk translation by recruiting Stau, as Stau still colocalizes with osk mRNA in Mkrn1 W oocytes (S12G-S12I Fig). Instead, we propose that Mkrn1 exerts its positive activity by competing with Bru1 binding to osk 3' UTR (Fig 8). This is evidenced by the overlap of their binding sites, the increased association of Bru1 to osk mRNA upon Mkrn1 knockdown and by our observation that reducing bru1 dosage is sufficient to partially alleviate osk translational repression.
Two distinct Bru1 binding regions (AB and C) are present in the osk 3' UTR and are required for translational repression. However, the C region has an additional function in translational activation. Indeed, it was hypothesized that an activator binds the C region to relieve translational repression [25]. This activator was proposed to either be Bru1 itself, or a different protein that can bind the BRE-C, which is what we observed for Mkrn1 (Fig 6A). Our results suggest that the interaction of pAbp with the nearby AR region, and the consequent Other factors may also be involved. For instance, Bicoid Stability Factor (BSF) binds the C region in vitro at the 3' type II Bru1-binding site [55], at a similar site to where Mkrn1 binds osk. Deletion of this site impacts embryonic patterning, yet depletion of BSF has no effect on Osk protein expression up to stage 10, indicating that initial activation of osk translation is effective even in the absence of BSF [55]. In this case, only late stage oocytes display reduced Osk accumulation. Therefore, it is possible that a concerted action of Mkrn1 and BSF exists at the osk 3' UTR site to activate translation and sustain it at later stages.
The binding of Mkrn1 to mRNA seems to be extremely specific. We found that the binding to osk is dependent on a downstream A-rich sequence and on interaction with pAbp. A few other targets we identified also display enrichment for downstream AA nucleotides (S13 Fig). and human MKRN1 has recently been shown to associate preferentially to such sequences [49]. Relevant to this, Bru1 binds to grk 3' UTR in addition to osk [56,57], and several proteins that associate with Mkrn1 also associate with grk mRNA [46,47]. However, we found no evidence that Mkrn1 binds specifically to grk, which lacks poly(A) stretches in the proximity of its Bru1 binding sites, and consistently, we did not observe a regulatory role of Mkrn1 on Grk translation.
In addition to pAbp, it is noteworthy that Mkrn1 associates with other proteins previously implicated in osk localization and translational activation. Its interaction with eIF4G would be consistent with a role in alleviating Cup-mediated repression, as it could recruit eIF4G to the cap-binding complex at the expense of Cup. However, we did not observe an interaction between Mkrn1 and eIF4E (S7C and S7D Fig, S1 and S2 Tables). The association between Mkrn1 and Imp is also intriguing as the osk 3' UTR contains 13 copies of a five-nucleotide motif that interacts with Imp [28]. This region is essential for osk translation but Osk accumulation is unaffected in Imp mutants, suggesting the involvement of another factor that binds these motifs [26,28,46]. In contrast to pAbp, we did not observe alteration of Mkrn1 binding when Imp was depleted, indicating that Imp is not required to stabilize Mkrn1 on osk mRNA.
The molecular links we uncovered between Mkrn1 and RNA-dependent processes in Drosophila are consistent with recent high-throughput analysis of mammalian MKRN1 interacting proteins [37,49]. RNA binding proteins, including PABPC1, PABPC4, and eIF4G1, were highly enriched among the interactors. Moreover, human MKRN1 was also recently shown to bind to RNA, dependent on the PAM2 motif and the interaction with PABPC1 [49]. In addition, the short isoform of rat MKRN1 was shown to activate translation but the underlying mechanism remained unknown [39]. Since in vertebrates MKRN genes are highly expressed in gonads and early embryos as well, it is possible that similar molecular mechanisms are employed to regulate gene expression at these stages [33]. Consistent with this, MKRN2 was recently found to be essential for male fertility in mice [58]. Thus, our study provides a mechanism that explains the role of Mkrn1 in translation and constitutes a solid framework for future investigations deciphering the roles of vertebrate MKRNs in post-transcriptional control of gene expression during gametogenesis and early development.

Generation of Mkrn1 mutants using CRISPR/Cas9
The guide RNAs used were cloned into expression vector pDFD3-dU63gRNA (Addgene) according to manufacturer's instructions. Different guide RNAs were used either alone (gRNA1 starting at nucleotide 64 of Mkrn1 CDS and gRNA2 starting at nucleotide 363 of Mkrn1 CDS) or in combination (gRNA3 starting at 387 nt of Mkrn1 gene and gRNA4 starting at position 2239 nt). vas-Cas9 Drosophila embryos were injected with the purified plasmids containing the gRNA (500 ng/μl in H 2 O) and allowed to develop to adulthood. Each male was crossed with double balancer females. Genomic PCR from single flies was prepared and tested for CRISPR/Cas9 induced mutations using the T7 endo I (BioLabs) assay or by PCR using primers that bind in proximity to the guide RNA targeting site. A list of gRNAs as well as primers is appended (S3 Table).

Immunostaining and confocal imaging
Ovaries were dissected and fixed in 4% paraformaldehyde in PBS for 20 min at room temperature (RT). After 4-5 washes in PBST (PBS containing 0.3% Triton-X100) ovaries were permeabilized with 1% freshly prepared Triton-X100 in PBS for 1 h. The ovaries were blocked in 2% BSA/PBST overnight. Dispersed egg chambers were then incubated with the primary antibodies diluted in 2% BSA/PBST at RT for 4 h or overnight at 4˚C. The washed egg chambers were incubated with conjugated secondary antibodies at 1:500 at RT for 4 h or overnight at 4˚C. DAPI (1 ng/ml) was added in the last wash to counter-stain the nuclei for 30 min. After 2-3 washes with PBST the mounting medium containing 1% DABCO was added and the samples were equilibrated for 30 min or overnight. The stained samples were mounted on glass slides and sealed with nail varnish for microscopy imaging. Rabbit polyclonal antibodies against Vas, Osk, Aub, Grk were generated in the Lasko lab; rabbit α-Stau is from the St Johnston lab; rabbit α-Bru1 is from the Ephrussi lab and rabbit α-pAbp is from the Sonenberg lab; mouse monoclonal antibodies against Orb, Sqd, and Lamin were purchased from the Developmental Studies Hybridoma Bank; mouse α-GFP and rabbit α-Flag were purchased from Abcam and Sigma. Alexa Fluor 488 or 555 conjugated secondary antibodies were purchased from Molecular Probes, and pre-absorbed with fixed and blocked wild type ovaries to reduce background. Stained egg chambers were examined using a confocal microscope (Leica). Images were taken under 40 x oil lens by laser scanning and processed with ImageJ.

In situ hybridization of embryos and ovaries and RNA-protein double labeling
cDNAs were used as templates for PCR to generate an amplified gene fragment with promoter sequences on each end. PCR products were purified via agarose gel extraction and used for in vitro transcription to generate digoxigenin-labeled RNA antisense probes with MAXIscript kit (Ambion). The length of each probe was about 1000 nt. In situ hybridization experiments were performed as described [59], using biotinylated α-DIG antibody and streptavidin-HRP followed by tyramide conjugation for development of FISH signal. For RNA-protein double labeling, ovaries or embryos were incubated in primary antibody against the protein of interest along with biotinylated α-DIG antibody at 4˚C overnight. The tissue was washed, then detection reagent (fluorochrome-conjugated secondary antibody) along with streptavidin-HRP was added and incubated at 4˚C overnight. Images were taken with confocal microscope (Leica).

Embryo cuticle preparation and staining
Flies were transferred into egg-laying cages with apple juice agar plates and incubated at 25˚C in the dark. Embryos were collected when 50-100 eggs had been laid and allowed to age for 24 h at 25˚C. Embryos were collected in a sieve, dechorionated with 50% bleach for 2.5 min, washed with water, then transferred into PBST buffer (PBS + 0.1% Tween 20). For cuticle preparations, PBST buffer was removed, then 40-50 μl of Hoyer's solution was added and embryos were kept at 4˚C overnight. Embryos in Hoyer's solution were mounted on a glass slide, covered with a cover slip and incubated at 60-65˚C overnight. Dark-field images were taken with Leica DM6000B microscope.

Staging
Staging experiment was performed as described [60] using D. melanogaster w 1118 flies.

Generation of transgenic flies
The constructs were made using Gateway technology (Invitrogen). Full-length wild-type Mkrn1 cDNA was cloned into the pENTR entry vector using the pENTR/D-TOPO cloning kit (ThermoFisher). After verifying the sequence by PCR, the insert was subcloned into the expression vectors containing UASp promoter with different tags (pPVW and pPFW with Venus and FLAG tags at N-terminal respectively), and (pPWV and pPWF with Venus and FLAG tags at C-terminal respectively) by LR in vitro recombination. For Mkrn1 ΔZnF1 pPFMW vector was used. Constructs were verified by sequencing and then injected into yw embryos. Progeny harboring the transgenes were crossed with double balancer flies to establish a variety of lines, and the insertion sites were mapped to either the second or third chromosome. Mkrn1 expression was then driven by crossing the transgenic lines with a nos>Gal4 line (MTD). Expression of tagged Mkrn1 was verified by western blot analysis and immunostaining using anti-GFP (Abcam) or anti-FLAG (Sigma).

Rescue experiment
Flies carrying Venus-or FLAG-Mkrn1 on the second chromosome were crossed with the nos>Gal4 driver lines in three different Mkrn1 mutant backgrounds (Mkrn1 W , Mkrn1 S , and Mkrn1 N ). Progeny were collected and separated into two groups: (1) Venus-or FLAG-Mkrn1/ nos>Gal4; Mkrn1 W (S or N) / Mkrn1 W (S or N) and (2) nos>Gal4/CyO; Mkrn1 W (S or N) / Mkrn1 W (S or N) . To perform hatching tests the same number of flies from each group was fed with yeast butter on apple juice plates for 1 d, embryos were collected and incubated at 25˚C for 48 h to allow completion of hatching. Hatched and unhatched embryos were counted for each group. The data from several tests in the same group were pooled and the hatching percentage was calculated.

Cell culture, RNAi, transfection
Drosophila S2R+ cells were grown in Schneider's medium (Gibco) supplemented with 10% FBS (Sigma) and 1% Penicillin-Streptomycin (Sigma). For RNAi experiments, PCR templates for the dsRNA were prepared using T7 Megascript Kit (NEB). S2R+ cells were seeded at the density of 10 6 cells/ml in serum-free medium and 15 μg/ml of dsRNA was added to the cells. After 6 h of cell starvation, serum supplemented medium was added to the cells. dsRNA treatment was repeated after 48 h and cells were collected 24 h after the last treatment. A list of primers used to create dsRNA templates by PCR is appended (S4 Table). Effectene (Qiagen) was used to transfect vector constructs in all overexpression experiments following the manufacturer's protocol.

Immunoprecipitations (IPs)
For IP experiments in S2R+ cultured cells, protocol was followed as described [60] with minor changes: 2 mg of the protein lysates was incubated for 2 h with 10 μl of either Myc-Trap or GFP-Trap beads (Chromotek). To determine the dependence of interactions on RNA, 50 U of RNaseT1 (ThermoFisher) were added to the respective IP. To ensure the activity of RNase T1, lysates were incubated 10 min at RT prior to the incubation of lysate with antibody.
For IP experiments in ovaries, 150 μl of wet ovaries from 3-5 day old flies expressing Venus-Mkrn1 was homogenized on ice in 2 ml of cold IP buffer (1 X PBS, 0.4% Triton X-100, 1 mM MgCl 2 , 5% glycerol), containing protease inhibitors and PMSF. The extracts were diluted to 1.5 mg protein/ml. Each extract (0.66 ml) was mixed with 24 μg of anti-pAbp Fab antibody (Smibert lab, [61]), 17 μg of α-eIF4G rabbit antibody, or 15 μl of rabbit anti-α−-Tubulin antibody (Abcam). When present, 100 μg RNase A (Qiagen) was added to the samples. Samples were incubated with rotation at 4˚C overnight, then mixed with 30 μl of protein A agarose beads (wet volume, Invitrogen) and incubated with rotation at RT for 1.5 h. The beads were washed three times with IP buffer. Bound material on the beads was eluted by boiling for 2 min in 40 μl of SDS loading buffer. 20 μl of the eluted sample, together with input samples, was used for western blot.

RNA-Immunoprecipitation (RIP)
For RIP, S2R+ cells or ovaries were harvested and lysed in RIP buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 0.5% NP-40, 1 mM EDTA) supplemented with proteinase inhibitors (1.5 μg/ ml Leupeptin, 1.5 μg/ml Pepstatin, 1.5 μg/ml Aprotinin and 1.5 mM PMSF) and RNase inhibitors (20 U/μl). S2R+ cells were lysed for 20 min at 4˚C, subtracted to 2 cycles of sonication on a bioruptor (Diagenode) with 30 sec "ON"/"OFF" at low setting and the remaining cell debris was removed by centrifugation at 21,000 g for 10 min at 4˚C. To remove lipids and cell debris, ovary lysates were centrifuged 4 times. Protein concentrations were determined using Bradford reagent (BioRad). 2 mg of protein lysate were incubated for 3 h with 2 μg of α-FLAG M2 antibody (Sigma-Aldrich) pre-coupled to 20 μl of protein G Dynabeads (Thermo Fisher Scientific) head-over-tail at 4˚C. For RIP experiments analysing binding of Bru1 in ovaries, either 1 μl of rabbit α-Bru1 (gift from A. Ephrussi) or 2 μg of rabbit IgG (Millipore) were incubated with ovarian lysate over night at 4˚C. 20 μl of protein G Dynabeads were added for 2 h after the incubation. For every RIP experiment, beads were washed 4 x for 10 min in RIP buffer at 4˚C.
For immunoprecipitation of GFP-tagged Imp and Bru1 15 μl of GFP-Trap (Chromotek) were used. Lysates were prepared similar as above using RIPA buffer (140 mM NaCl, 50 mM Tris pH 7.5, 1 mM EDTA pH 8, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate) supplemented with proteinase and RNase inhibitors. IP was performed for 2 h at 4˚C and subsequently washed 4 x for 10 min with RIPA buffer.
RNA was eluted in TRIzol Reagent (ThermoFisher), 10 min at RT and subjected to RNA isolation and RT-qPCR. To obtain the depicted fold enrichment, individual transcripts were normalized to either 18S or RpL15 (S5 Table). At least three biological replicates were performed for each experiment. If not stated differently statistical analysis was performed using one sample t-test.
To analyze IPs, 30% of beads were eluted in 1x SDS buffer (50 mM Tris pH 6.8, 2% SDS, 10% glycerol, 100 mM DTT, 0.05% Bromphenol Blue) at 95˚C for 10 min. Eluted IP proteins were removed from the beads and analyzed by western blot together with input samples.

RNA isolation and measurement of RNA levels
Cells or tissues were harvested in TRIzol Reagent (ThermoFisher) and RNA was isolated according to the manufacturer's instructions. DNA was removed with DNaseI treatment (NEB) and cDNA was prepared with M-MLV Reverse Transcriptase (Promega). The transcript levels were quantified using Power SYBR Green PCR Master Mix (ThermoFisher) using the indicated primer (S5 Table).

LC-MS/MS
To identify binding partners of Mkrn1, either Myc-GFP as control or Myc-Mkrn1 were ectopically expressed in S2R+ cells. Upon lysis, Myc-GFP or Myc-Mkrn1 were immunoprecipitated as described above (see IP methods) with small adjustments: The IP buffer was additionally supplemented with 10 mM N-ethylmaleimide, 1 mM sodium orthovanadate, 5 mM β-glycerophosphate and 5 mM sodium fluoride. After IP, samples were eluted in 2x LDS buffer (Life Technologies) supplemented with 1 mM dithiothreitol for 10 min at 70˚C and incubated with 5.5 mM 2-chloracetamide for 30 min at room temperature in the dark. All samples were prepared in parallel.
Conventional interactome analysis of the IP samples was performed as described before [62] with the following changes: The enriched proteins were separated by SDS-PAGE with a 4-12% Bis-Tris protein gel (NuPAGE, Thermo Scientific) and stained with Colloidal Blue Staining Kit (Life Technologies). Subsequently, proteins were in-gel digested using trypsin and digested peptides were then extracted from the gel. Concentration, clearance and acidification of peptides, mass spectrometry analysis, and peptide identification were performed as described before [62]. For peptide identification in MaxQuant (version 1.5.28), the DROME database from UniProtKB (release May 2016) was used. For label-free quantification (LFQ) at least 2 LFQ ratio counts (without fast LFQ) were activated.
The data table of LFQ values resulting from MaxQuant was filtered for potential contaminants, reverse binders and protein groups only identified by site. Furthermore, protein groups with less than two peptides and less than one unique peptide were also removed from further analysis. After log-transforming all remaining LFQ values, missing values were imputed by beta distributed random numbers between 0.1% and 1.5% of the lowest measured values. As a final filtering step, only protein groups having measured values for at least two replicates of at least one experimental condition were kept for further analysis. All filter and imputing steps were done with an in-house R script.
Differential protein abundance analysis was performed on log-transformed LFQ values between two conditions at the time using the R package limma (version 3.34.9, [63]). For each such comparison, only protein groups found in at least two replicates of at least one condition were kept and used. To visualize the interactome, the R package ggplot2 [64] was used. All protein groups with an FDR � 0.05 and a log 2 fold change of � 2 were considered significantly changed.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD011802.
For high-throughput sequencing of iCLIP experiments, libraries of 6 technical replicates for FLAG-Mkrn1 and 1 replicate for FLAG-GFP (Control) were prepared as described [69]. Barcodes used for the libraries are listed in S6 Table. Multiplexed iCLIP libraries were sequenced as 75-nt single-end reads on an Illumina MiSeq sequencing system.
Sequencing qualities were checked for all reads using FastQC (version 0.11.5) (https://www. bioinformatics.babraham.ac.uk/projects/fastqc/). Afterwards, reads were filtered based on sequencing qualities (Phred score) of the barcode region. Reads with more than one position with a Phred score < 20 in the experimental barcode (positions 4 to 7 of the reads) or any position with a Phred score < 17 in the random barcode (positions 1 to 3 and 8 to 9) were excluded from subsequent analysis. Remaining reads were de-multiplexed based on the experimental barcode (positions 4 to 7) using Flexbar (version 3.0.0, [70]) without allowing any mismatch.
All following steps of the analysis were performed on the individual samples after de-multiplexing. Remaining adapter sequences were removed from the read ends using Flexbar (version 3.0.0) with a maximal error rate of 0.1 and a minimal overlap of 1 nt between the beginning of the adapter and the end of the read. Following adapter trimming, the first 9 nt of each read containing the experimental and random barcodes were trimmed off and added to the read name in the fastq files in order to keep this information for downstream analysis. Reads shorter than 15 nt were removed from further analysis.
Trimmed and filtered reads were mapped to the Drosophila melanogaster genome (Ensembl genome assembly version BDGP6) and its annotation (Ensembl release 90, [71]) using STAR (version 2.5.2b, [72]). When running STAR, up to two mismatches were allowed, soft-clipping was prohibited at the 5' ends of reads and only uniquely mapping reads were kept for further analysis. For further analysis, only unspliced reads were kept and analyzed.
Following mapping, duplicate reads were marked using the dedup function of bamUtil (version 1.0.13), which defines duplicates as reads whose 5' ends map to the same position in the genome (https://github.com/statgen/bamUtil). Subsequently, marked duplicates with identical random barcodes were removed since they are considered technical duplicates, while biological duplicates showing unequal random barcodes were kept.
Resulting bam files were sorted and indexed using SAMtools (version 1.3.1, [73]). Afterwards, bedgraph files were created based on bam files, using bamToBed of the BEDTools suite (version 2.25.0; [74]), considering only the position upstream of the 5' mapping position of the read, since this nucleotide is considered as the crosslinked nucleotide. Using bedGraphToBig-Wig of the UCSC tool suite [74], all bedgraph files were converted into BigWig files [75].
In order to estimate binding site strength and to facilitate comparisons between binding sites (S10D Fig) we corrected for transcript abundance by representing the crosslink events within a binding site as a 'signal-over-background' ratio (SOB). The respective background was calculated as the sum of crosslink events outside of binding sites (plus 5 nt to either side) by the merged length of all exons. 3' UTR lengths were restricted to 10 nt past the last Mkrn1 binding site or 500 nt if no binding site was present. SOB calculations were performed separately for each replicate and then averaged. No SOB value was assigned for ribosomal genes and genes with a background of < 10 crosslink events, resulting in SOB values for 184 binding sites in 46 targets.
The sequence content around the predicted Makorin 1 binding sites was estimated by counting homopolymeric 2-mers (S13 Fig). All 262 binding sites that fall in 3' UTRs were centered at their midpoint and a symmetric window was extended 100 nucleotides up-and downstream. For each position the mean occurrence of each 2-mer was counted.

Cloning
To overexpress tagged proteins, the respective coding sequences were amplified and cloned into Gateway plasmids using AscI and NotI restriction sites for plasmids pPFMW and pAHW. The coding sequences of Imp, bru1 and pAbp were cloned using KpnI and XbaI into plasmid pAWG. A list of primers used to introduce the coding sequences as well as to introduce mutations in the Mkrn1 coding sequence is appended (S7 Table). To analyze the binding of Mkrn1 to osk, either the 3' UTR was cloned downstream of Firefly coding sequence or the complete osk gene was cloned in the same vector backbone of pAc5.1B-EGFP (gift from Elisa Izaurralde, Addgene plasmid #21181). For both, restriction sites of KpnI and SalI were used. To ensure the proper usage of the endogenous poly(A) signal the osk 3' UTR and the osk gene included 220 and 248 nucleotides of the downstream sequence, respectively. The primers used for cloning and to introduce mutations are listed in S8 Table. Supporting information S1 Fig. Four Makorin- datasets with Mkrn1-FLAG in S2R+ cells. The x axis displays maximum binding strength per gene (SOB) and the y axis shows the gene identity. The genes are sorted by SOB with osk mRNA appearing at the third place. Note that ribosomal genes have been excluded for clarity. (TIF) S11 Fig. Mkrn1 binding to osk 3' UTR is dependent on pAbp. (A) Mkrn1, but not Mkrn1 ΔZnF1 binds to the osk 3' UTR in S2R+ cells. FLAG-RIP of GFP, Mkrn1 and Mkrn1 ΔZnF1 was performed co-expressing either luciferase-grk-3'UTR or luciferase-osk-3'UTR reporter. Top: qPCR analysis of the RIP experiments. Fold enrichment is illustrated relative to GFP RIP. Error bars depict SEM, n = 3. Bottom: Immunoblots of a representative RIP experiment. (B-F) Western blot analysis of a representative FLAG-RIP experiment in S2R+ cells summarized in Fig 6C-6F. (B) Either FLAG-tagged GFP or Mkrn1 RING was overexpressed. RIP experiments were performed in the presence of a luciferase-osk-3'UTR reporter containing osk 3' UTR wild-type sequence or deletion of the Mkrn1 binding site (osk ΔMkrn1 ) (C) FLAG-RIP of GFP or Mkrn1 RING were performed either using the luciferase-osk-3'UTR reporter. Binding was compared between wild-type osk 3' UTR and a mutation in the AR region (osk ΔAR ).