The target specificity of the RNA binding protein Pumilio is determined by distinct co-factors

Puf family proteins are translational regulators essential to a wide range of biological processes, including cell fate specification, stem cell self-renewal, and neural function. Yet, despite being associated with hundreds of RNAs, the underlying mechanisms of Puf target specification remain to be fully elucidated. In Drosophila, Pumilio – a sole Puf family protein – is known to collaborate with cofactors Nanos (Nos) and Brain Tumor (Brat); however, their roles in target specification are not clearly defined. Here, we identify Bag-of-marbles (Bam) as a new Pum cofactor in repression of Mothers against dpp (mad) mRNAs, for which Nos is known to be dispensable. Notably, our data show that Nos (but not Bam) was required for Pum association with hunchback (hb) mRNAs, a well-known target of Pum and Nos. In contrast, Bam (but not Nos) was required for Pum association with mad mRNAs. These findings show for the first time that Pum target specificity is determined not independently but in collaboration with cofactors.


Introduction
The post-transcriptional regulation of gene expression can provide spatially localized protein synthesis. This process is necessary in whole embryos to give rise to the intricate expression patterns that support proper development, and, within the different parts of a single neuron, to ensure the subcellular protein localizations required for proper neural function. Post-transcriptional gene regulation is generally carried out by repressor proteins that recognize specific sequences in the 3 -UTRs of target mRNAs [1]. Genetic studies of pattern formation in Drosophila embryos identified pumilio (pum) and nanos (nos), both of which limit the expression of hunchback (hb) in the anterior segments of the embryo [2]. These studies also identified specific sequences-Nanos Response Elements (NREs)-in the 3 -UTR of hb mRNAs that are required for translation inhibition [3,4]. The repressive complex on the NREs, comprising Pum, Nos, and Brain Tumor (Brat), recruits d4EHP, Cup, or components of deadenylase, which induce either the decay of hb mRNAs or the inhibition of their translation [4][5][6][7][8][9][10][11][12][13][14][15][16][17].
In addition to hb mRNAs, Pum, Brat, and Nos cooperate to regulate diverse mRNAs in diverse cell types. For example, Pum, Brat, and Nos regulate paralytic mRNAs to control the excitability of motor neurons and dendrite morphogenesis in peripheral neurons [35,36]. In other cases, however, neither Brat nor Nos are required for Pum-mediated translational repression. Pum and Nos (but not Brat) are required to inhibit cycB mRNAs and mei-P26 mRNAs in pole cells and germline stem cells (GSCs) of the Drosophila ovary, respectively [10,14,37]. Pum and Brat (but not Nos) are required to inhibit Mothers against dpp (mad), shnurri, and medea mRNAs in the cystoblasts of the Drosophila ovary [38,39]. The mechanisms underlying these variations in the combinatorial requirements for the differential targeting and repression of different mRNAs remain poorly understood. In particular, the mechanisms underlying mad mRNA repression by Pum and Brat are incompletely understood. Here, we use a cell-based assay with the mad 3 -UTR to explore the mechanisms underlying mad mRNA repression. We hypothesized that the translational repressors Bag-of-marbles (Bam) and Bgcn may collaborate with Pum, as Bam physically interacts with Pum and Brat [40]. Here, we show that Bam and Bgcn do in fact collaborate with Pum and Brat in translational repression via the mad 3 UTR. Notably, we find that Pum binding to mad mRNAs in cells requires Bam and Bgcn, but not Nos, while Pum binding to hb mRNAs-the best-known targets of Pum and Nos-requires Nos. These results indicate that Pum uses distinct cofactors when binding different target mRNAs.

Bam is required for repression through the mad 3 UTR
Previously, it was shown that Bam and its obligate partner Bgcn physically interact with Pum and Brat [40,41]; this raises the possibility that Bam and Bgcn collaborate with Brat and Pum, which regulate mad at the post-transcriptional level through the mad 3 UTR in the absence of Nos. We examined this possibility using a reporter bearing the mad 3 UTR in Schneider's 2 cells (S2 cells). Co-expression of all four factors (Bam, Bgcn, Brat, and Pum) did not affect expression of a control reporter lacking the mad 3 UTR, but greatly repressed luciferase (luc) expression in a dose-dependent manner from a luc reporter bearing the mad 3 UTR ( Figure 1A). The level of luc-mad 3 UTR mR-NAs decreased as expression of all four factors increased ( Figure 1B). These results suggest that luc repression by the four factors is in part due to the destabilization of luc-mad 3 UTR mRNAs. Increasing the doses of expression vectors in the cells increased the expression of their corresponding genes ( Figure 1C).
To determine whether all factors are required for repression, we omitted each expression vector individually. Interestingly, omitting any vector did not affect repression by the other three factors ( Figure 2A). This could be ascribed to the presence of endogenous protein for repression. Each factor is indeed expressed in S2 cells (Supplementary Figure  S1), making it necessary to remove the endogenous proteins. To identify siRNAs that can effectively knockdown target proteins, S2 cells were co-transfected with siRNAs and vectors expressing tagged proteins. For each gene, we tested two siRNAs and identified one that effectively reduced the target protein levels relative to control (scrambled) siRNAs ( Figure 2B). We used these siRNAs to knockdown the corresponding endogenous protein in subsequent experiments.
Co-transfecting any three factors along with an siRNA for the fourth factor abolished repression ( Figure 2A). These results indicate that each of Bam, Bgcn, Brat, and Pum is required for repression. In contrast, S2 cells treated with siRNAs specific for translational regulators Nos, Mei-P26, or Ago1 did not abrogate repression by four factors (Bam, Bgcn, Brat, and Pum) ( Figure 2A). Taken together, these data show that Bam, Bgcn, Brat, and Pum, but not Nos, Mei-P26, or Ago1, are required for repression through the mad 3 UTR in S2 cells.

Bam association with mad mRNA
Next, we examined whether the four factors (Bam, Bgcn, Brat, and Pum) were associated with mad mRNA in S2 cells. Lysates from S2 cells individually expressing tagged-Bam, -Bgcn, -Brat, or -Pum were immunoprecipitated using tag antibodies. The immunoprecipitates were subjected to reverse transcription (RT) followed by polymerase chain reaction (PCR) amplification to determine association with mad mRNAs. Hb and string-of-pearls (sop) (also known as ribosomal protein S2 (rps2)) mRNAs were used as a positive and a negative control, respectively. The results revealed that Bam, Bgcn, Brat, and Pum were all associated with mad mRNA (Figure 3), but not with the negative control sop mRNAs, which does not have PRE elements. Pum and Brat were associated with positive control hb mRNA, consistent with known interactions. In contrast, Bam and Bgcn were not associated with hb mRNA (Figure 3). The translational regulators Mei-P26 and Ago1 were not associated with either mad mRNA or hb mRNA ( Figure 3). Nos was associated with hb mRNA as expected, but not with mad mRNA, which is consistent with Nos being required for hb repression but not mad repression [2,5,38] (Figure 3). Mock-immunoprecipitates with control antibodies (IgG) contained neither mad mRNA, hb mRNA, nor sop mRNA ( Figure 3). Taken together, these results show that Bam and Bgcn are specifically associated with mad mRNA, but not hb mRNA, while Nos is specific to hb but not mad mRNA. Brat and Pum interact with both mad and hb mRNAs.

Bam and Bgcn, but not Brat or Nos, are required for Pum binding to mad mRNA
We were curious whether the association of Bam with mad mRNAs requires Pum. In lysates from control (scrambled) siRNA treated cells, Bam immunoprecipitates contained mad mRNA, but not hb mRNA ( Figure 4A and Supplementary Figure S2). However, Bam immunoprecipitates of lysates from cells treated with Pum or Bgcn siRNA did not contain mad mRNA ( Figure 4A and Supplementary Figure S2). In contrast, Bam immunoprecipitates of lysates from cells treated with Brat, Ago1, or Nos siRNA did contain mad mRNA ( Figure 4A and Supplementary Figure S2). These demonstrate that the association of Bam with mad mRNA requires Pum and Bgcn, but not Brat, Ago1, or Nos. Similar experiments showed that Bgcn requires Bam and Pum, but not Brat, Ago, or Nos ( Figure 4B and Supplementary Figure S2) in association with mad mRNAs. Meanwhile, Brat requires Bam, Pum, and Bgcn but not Ago1 or Nos when in association with mad mRNAs, and does not require other factors when binding hb mRNAs ( Figure 4C and Supplementary Figure S2); the latter is consistent with a recent report that Brat can bind hb mRNAs by itself. Pum association with hb mRNAs does not require Bam, Bgcn, Brat, or Ago1 but does require Nos, while its association with mad mRNAs requires Bam and Bgcn, but not Brat or Nos ( Figure 4D and Supplementary Figure S2). Taken together, the key conclusion is that Pum requires Bam and Nos for binding mad and hb mRNAs, respectively.

Pum binding sites in the mad 3 UTR mediates repression and binding
To identify the cis-element that mediates repression, mad 3 UTR was divided into three parts (1-300, 300-600, 600-900) and examined to see which region was responsible for the repression. The 1-300 region, but not the other regions, of mad 3 UTR-mediated repression of a reporter by Bam, Bgcn, Brat, and Pum ( Figure 5A). We further narrowed the region to 121-220 which mediated the repression ( Figure 5A). The 121-220 region contains two UGUA sequences, a core sequence found at many Pum-binding sites [26]. We mutated these two sites from UGUA to ACUA and examined whether the mutations affected the repression. Mutating both sequences abrogated the Bam, Bgcn, Brat, and Pum mediated repressions ( Figure 5A), indicating that Bam, Bgcn, Pum, and Brat repressed translation through the UGUA sequence in the mad 3 UTR. Endogenous depletion of each of Bam, Bgcn, Brat, and Pum, but not of Nos, Meip-26, and Ago1 abrogated the repression of the reporter with a 121-220 region of the mad 3 UTR ( Figure 5B). The RNA level of luc-mad, but not luc-mad mt (mutated), was reduced when Bam, Bgcn, Brat, and Pum were co-expressed ( Figure 5C).

Discussion
Recent evidence from in vitro assays indicates that the binding of Pum to target mRNAs requires Nos, suggesting that Nos plays an active role in Pum target binding [29,42]. However, other mRNAs associated with Pum were not dependent on Nos [29,42], suggesting that other factors might also play a role in the targeting of Pum to particular mRNAs. Here, we identify Bam and Bgcn as novel Pum cofactors in mad repression. Our data reveal that Bam, but not Nos, plays an active role in targeting Pum to mad mRNAs.
We found that nucleotides 121-220, but not other regions, of the mad 3 UTR are sufficient for repression. There are two putative Pum binding sites in this segment: 5 -UGUACGGC-3 and 5 -UGUAGGCC-3 (hereafter, referred to as madPREs). Except for four invariant nucleotides (i.e., UGUA) these imperfect PREs are distinct from the standard eight nucleotide Pum consensus binding sequence (i.e., 5 -UGUAYAUA-3 ) [26]. The first two nucleotides (UG) are critical for Pum binding to hb mRNAs [23]. We show that mutating these nucleotides from UG to AC in madPREs prevents repression and abolishes Bam, Bgcn, Pum, and Brat binding, demonstrating that these two madPREs are essential cis-elements for mad repression and for the assembly of protein complexes on mad mRNAs.
Nos was recently shown to determine a subset of Pum targets, including hb mRNAs [41,42]. These Nos-Pum target mRNAs contain a Nos recognition sequence immediately upstream of a Pum binding sequence (5 -AAU UGUANAUA-3 (AAU is the Nos binding sequence, the PRE is in bold)) [29]. Nos binds the three nucleotides to enhance Pum binding to hb mRNAs [29]. The sequence immediately upstream of the Pum binding sequence in mad mRNAs (madPREs, UGA UGUACGGC, GUC UGUAGGCC) does not contain the Nos recognition sequence, accounting for why Nos is not associated with mad mRNAs and why Nos is dispensable for Pum binding to and repression of mad mRNAs.
Brat is associated with both hb and mad mRNAs and is required for their repression [6,33,38]. Unlike Bam or Nos, Brat depletion does not abrogate Pum binding to mad and hb mRNA, suggesting Brat is not involved in the binding of Pum to its target mRNAs. In addition, we found that Brat has differential association with mad and hb mRNAs. Our in vivo binding assay showed that Brat associates with hb mRNA in the absence of other collaborators. This is consistent with recent data suggesting that Brat binds hb mRNA alone [33]. In contrast, Brat was not associated with mad mRNA in cells treated with siRNAs against Bam, Bgcn, or Pum. Given that Brat physically interacts with Bam and Pum, these results suggest that Brat is recruited to mad RNA indirectly via these interactions. The reason for this discrepancy between its binding to hb and mad mRNAs remains unclear. Although the mechanisms of Brat recruitment seem to differ, Brat binds to and is required for the repression of both mad and hb mRNAs. What could the role of Brat in repression be? Brat was recently shown to repress mRNAs independently of Pum and Nos [33]. Thus Brat is likely to increase the activity of repressive complexes formed on cis-elements, rather than being specifically involved in Pum targeting.
In conclusion, we propose that Pum targeting shifts depending on cofactors that are expressed differentially in cells. Given that Pum is associated with hundreds of mRNAs in Drosophila and thousands in Caenorhabditis elegans and mammals [25,26,43], while expression of Nos and Bam is more limited, we expect that there are as-yet-unidentified co-factors that specify Pum targets in different cells during embryonic development, during stem cell differentiation, and in the different subcellular regions of neurons.

Chemicals and antibodies
All chemicals were obtained from Sigma-Aldrich. Primary antibodies were obtained as follows; anti-FLAG (M2; Sigma) (mouse), anti-Myc (rabbit) (Cell Signaling), and anti-HA (rat) (Roche Applied Science). All secondary antibodies were obtained from Roche Applied Science.

S2 cells
The Drosophila S2 cells were maintained in a Shields and Sang M3 insect medium (Sigma-Aldrich) with 10% insect medium supplement (Sigma-Aldrich) and 1% antibiotics (HyClone) in a humidified atmosphere at 25 • C.

RNA immunoprecipitation assay
S2 cell extracts were generated from three six-well plates transiently transfected with the appropriate set of expression plasmids using the DDAB method [44]. The cleared extracts (6 mg) were mixed with 40 μl of anti-FLAG M2-conjugated agarose beads (Sigma) for precipitating Flag-tagged proteins incubated for overnight at 4 • C. In case of HA or Myc tagged proteins, the extracts were mixed with 10 μl of anti-HA antibody (mouse) or 10 μl of anti-Myc antibody (mouse) for 16 h at 4 • C followed by incubation with the 20 μl anti-IgA/G-conjugated agarose beads (Sigma) at 4 • C overnight. For knockdown with siRNA, siRNAs were added in a concentration of 100 pmol siRNA/well and 1 μg expression plasmids/well of a six-well plate. Protocols after immunoprecipitation were described [45]. Different cycles (25)(26)(27)(28)(29)(30) were then performed, and the resulting products were run on 1% agarose gels and visualized using EtBr. Primer sequence used for RT-PCR as follows: mad (forward, GACCTCGACGCCCTG; reverse, CACTTGAAATAGACTTGA), hb (forward, GTTCCCCATCACCATCAC; reverse, TGAAAGGTGGCTACAGTT), sop (forward, ACTCCTTACCAGGC;reverse, ACATTTAAAAGTTTATGAC).

Statistical analysis
Luciferase assays and qRT-PCR were performed in triplicate at least three times. Data calculation and statistical analysis were performed using GraphPad Prism 6.0 software. P-values ware determined by one-way ANOVA in GraphPad Prism 6.0, with post-hoc analysis using Bonferroni's multiple comparison test.