Differential Translation of Dazap1 Transcripts during Spermatogenesis

Deleted in AZoospermia Associated Protein 1 (DAZAP1) is a ubiquitous hnRNP protein that has been implicated in RNA transcription, splicing, and translation. It is highly expressed in testes, predominantly in late stage spermatocytes and post-meiotic spermatids. Dazap1 deficiency in mice results in growth retardation and spermatogenic arrest. The gene produces two major transcripts of 2.4 and 1.8 kb, designated Dazap1-L and Dazap1-S, respectively. Results of our previous RNA in situ hybridization and immunostaining suggested translational regulation of the Dazap1 transcripts during spermatogenesis. The main objectives of the study were to determine the origin of the two Dazap1 transcripts and to investigate whether they were similarly translated. Our Northern and 3′ RACE analyses showed that the two transcripts were generated through alternative polyadenylation. In mouse testes, the levels of both transcripts were low at postnatal day 12 (P12), increased significantly at P18, and reached maximum at P27. Sucrose gradient analyses showed that at P12 both transcripts were actively translated. Afterward, an increasing portion of Dazap1-S became associated with the translationally inactive mRNPs, and the translational repression was accompanied by an increase in the length of its poly(A) tail. A much smaller portion of Dazap1-L was also sequestered to mRNPs as testes matured, but there was no changes in its poly(A) tail length. Using RNA pull-down followed by mass spectrometry, we identified DAZL, a germ-cell specific translation regulator, as one of the proteins that bound to the 3′UTR region specific for Dazap1-L. We further showed that DAZL preferentially bound to Dazap1-L in testis lysates and stimulated the translation of a reporter gene carrying Dazap1-L 3′UTR. In summary, our study shows that the translation of the two Dazap1 transcripts is differentially regulated. It also provides a new example of translational repression associated with poly(A) tail elongation during spermatogenesis.


Introduction
Most eukaryotic mRNAs are translated soon after they are processed and exported to the cytoplasm. However, some mRNAs are stored in the cytoplasm as inert messenger ribonucleoprotein particles (mRNPs) for a substantial period of time before they are translated. Such translation repression or mRNA masking is most frequently observed during gametogenesis and early embryogenesis [1]. A classic example of translational regulation during spermatogenesis is the Prm1 gene (encoding protamine 1) which is transcribed in round spermatids and translated several days later in elongated spermatids [2]. Storage of Prm1 mRNA for later translation is a necessary step for male germ cell development because protamine is needed at a stage when the nuclear transcription has been shut down. Premature translation of Prm1 mRNA in round spermatids results in precocious condensation of nuclear DNA and dominant male sterility [3]. Translational repression of Prm1 mRNA is mediated through its 39UTR, and several proteins that bind to Prm1 39UTR have been identified although how they regulate the translation remains unclear [4,5,6,7]. Additional patterns of gene repression are also observed in spermatogenic cells [8]. Some mRNAs appear to be partially repressed at all time, while others are almost totally inactivated in free mRNPs and generate little proteins. Recent profiling of testicular mRNAs in polyribosomes (polysomes) and mRNP fractions of prepuberal and adult mice identified over seven hundreds of transcripts that are differentially translated during testis development [9].
Deleted in AZoospermia (DAZ) proteins are a family of RNAbinding proteins that regulate mRNA translation in germ cells [10]. They include DAZ, a spermatogenic factor encoded by the AZFc region of the Y chromosome that is frequently deleted in azoospermic men [11], and the autosomal encoded DAZL and BOULE [12,13,14]. DAZ proteins are expressed predominantly, if not exclusively, in germ cells and their deficiencies impair reproduction [15,16]. A role of DAZ proteins in mRNA translation is supported by their presence on polysomes [17,18] and their effects on translation in cultured cells [18], intact oocytes [19,20] and zebrafish embryos [21]. A recent study using tethered function assay in Xenopus oocytes suggests that DAZL facilitates translation initiation by binding to the 39UTR of its target mRNAs to recruit poly(A) binding proteins (PABPs) and stimulate the formation of 80 S ribosomes [19]. DAZL recognizes GUU triplet in different sequence context [18,22,23], and several putative natural substrates of DAZL have been identified though some of them need further verification in vivo [20,24,25,26].
DAZ Associated Protein 1 (DAZAP1) is a DAZ interacting protein that is expressed ubiquitously but most abundantly in the testis [27,28]. It is a component of the hnRNP particles [29] and has been implicated in several cellular processes including RNA transcription [30,31], splicing [32,33,34], and translation [35]. Very recently, DAZAP1 was suggested as a potential disease gene for amyotrophic lateral sclerosis [36]. Most mice deficient in DAZAP1 die perinatally, and the few survivors manifest growth retardation and spermatogenic arrest [37]. In mouse testes, DAZAP1 first appears abundantly in the nuclei of mid-pachytene spermatocytes, whereas our previous RNA in situ hybridization detected high levels of Dazap1 transcripts in spermatogonia and early spermatocytes [38]. We thus set out to investigate possible delayed translation of Dazap1 transcripts during spermatogenesis. Our results showed that two Dazap1 transcripts were generated through alternative polyadenylation and their differential translation during spermatogenesis involved poly(A) tail length alteration and DAZL binding.

Results
Two Dazap1 transcripts are generated through alternative polyadenylation Mouse testes express two major Dazap1 transcripts of 2.4 and 1.8 kb, designated Dazap1-L and Dazap1-S, respectively [28]. We failed to detect alternative splicing and promoter usage by RT-PCR and 59 Rapid Amplification of cDNA Ends (RACE), respectively, suggesting that the two Dazap1 transcripts were generated through alternative polyadenylation (data not shown). The 39UTR of Dazap1 contains three canonical AATAAA polyadenylation signals (PASs) (Fig. 1A), but only the second and the third ones are evolutionally conserved throughout vertebrates (data not shown). To identify the polyadenylation sites of Dazap1 transcripts, we performed 39 RACE of adult mouse testis RNA and detected two major fragments around 0.8 and 0.3 kb (Fig. 1B). Sequencing of the fragments revealed that polyadenylation of the 0.3 kb fragment initiated from three CA sites downstream of the second PAS, whereas that of the 0.8 kb product started downstream of the third PAS (Fig. 1A). The minor 0.5 kb fragment represented a PCR artifact amplified from the 0.8 kb fragment since it was also produced by PCR amplification of a clone containing the 0.8 kb fragment. Thus the two Dazap1 transcripts were produced through alternative polyadenylation utilizing the second and the third PASs. This was further confirmed by Northern blot (Fig. 1C, D). Probe-A which covers a large portion of the Dazap1 cDNA detected both transcripts, whereas probe-B which contains a segment downstream of the second PAS hybridized only to Dazap1-L.
The Dazap1 transcripts are differentially expressed during spermatogenesis We next studied the levels of the two Dazap1 transcripts in prepuberal and adult mouse testes. Spermatogenesis is initiated soon after birth, and day 8 mouse testes contain only type A and type B spermatogonia [39]. Spermatocytes start to appear afterward. In postnatal day 12 (P12) testes, germ cells make up 60% of the total cell population and the most advanced germ cell is zygotene spermatocytes. At P18, pachytene spermatocytes dominate the germ cell population, and haploid spermatids start to appear after day 20. P27 testes are enriched in late round spermatids [40], and P34 and adult testes contain spermatids as the major cell types. Northern analyses of mouse testes at different ages showed that the levels of Dazap1 transcripts were low at P12, increased significantly at P18, and reached the highest levels at P27 ( Fig. 2A, B), consistent with Dazap1 expression mainly in late pachytene spermatocytes and spermatids. The ratio of the two transcripts also changed as spermatogenesis progresses (Fig. 2B). The two transcripts were present in similar levels at P12 and P18, but Dazap1-S became the dominant form at P27 and after.
The Dazap1 transcripts are differentially translated in the testis To investigate whether both Dazap1 transcripts are equally translated during spermatogenesis, we fractionated mouse testis lysates on 10-50% sucrose gradients and monitored the distribution of Dazap1 transcripts by 39RACE as well as Northern blot when Dazap1 transcripts were sufficiently abundant for detection (Fig. 3). Both methods showed that in adult testes Dazap1-L had a bimodal distribution, with a large fraction associated with the polysomes, whereas a vast majority of Dazap1-S was present near the top of the gradient in the mRNP fraction (Figs. 3A, G). As a control, we also determined the distribution of the Prm1 transcript and found it in both mRNP and polysomal fractions as expected [9]. Upon EDTA treatment, the polysomes dissociated and all Dazap1 and Prm1 transcripts were found near the top of the gradient (Fig. 3B). The polysomal profiles of Dazap1 mRNAs in P27, P24, and P18 testes were similar to that of the adult testes, with a large fraction of Dazap1-S near the top of the gradients and most Dazap1-L associated with polysomes (Figs. 3C, D, E; G). On the other hand, the Prm1 transcripts in P27 and P24 testes were present only in the mRNP fractions because the most advanced spermatogenic cells in these testes are round spermatids. At P18 and P12, Prm1 transcripts were undetectable. There was a major change in the distribution of Dazap1 transcripts in P12 testes, with both transcripts mainly associated with polysomes being actively translated (Figs. 3F, G). Since the major difference in germ cell constitution between P12 and P18 testes is the appearance of pachytene spermatocytes in the latter, our results suggest that after the meiotic germ cells reach the pachytene stage, a large fraction of Dazap1-S and a much smaller fraction of Dazap1-L start to be sequestered to mRNPs and become translationally inactive.
It has been reported that mouse Prm1, Prm2, Tnp1, and Tnp2 transcripts being actively translated on polysomes have shorter Poly(A) tails than those that are translationally repressed [41]. To investigate whether translational regulation of Dazap1 transcripts also involved changes in poly(A) tail length, we used a recently developed extension Poly(A) Test (ePAT) method [42] to monitor poly(A) tail lengths of Dazap1 transcripts during mouse testis development and in sucrose-gradient fractions. Our initial test of ePAT on Prm1 transcripts confirmed previous Northern hybridization results, i.e., translationally active polysomal Prm1 transcripts had much shorter poly(A) tails than the major Prm1 transcript present in the free mRNP fraction ( Fig. 4A and Fig. S1). It is noticed that a small fraction of Prm1 transcripts in the mRNPs contained short poly(A) tails. Based on the size of the TVN-PAT product that had an invariant poly(A) tail of 12 As as well as the sequences of five or more randomly selected ePAT clones (Table  S1), the lengths of the poly(A) tails in the short and long Prm1 transcripts were determined to be approximately 25 and 120 bases, respectively, compared with previous estimations of 30 and 150 bases [41]. We also found that when Prm1 transcripts were first detected in P27 testes, the vast majority of them had long poly(A) tails (designated as LATs). As the testis matured to contain later stages of germ cells with active Prm1 translation, the fraction of Prm1 transcripts with short poly(A) tails (designated as SATs) increased progressively, consistent with translation activation of Prm1 in elongated spermatids.
We next used a forward primer common to both Dazap1 transcripts in the e-PAT reactions ( Fig. 4B and Fig. S1). Dazap1-S also produced two major ePAT products with poly(A) tails approximately 30 and 170 bases long, respectively (Table S1). Before P27 most Dazap1-S had SATs. However, the level of Dazap1-S with LATs increased progressively as the mice grew older, and reached a level comparable to that of Dazap1-S with SATs at P27. Similar to the Prm1 transcripts, Dazap1-S with both SATs and LATs were present in the translationally inactive mRNP fraction whereas only Dazap1-S with SATs was present in the polysomal fraction. In contrast, the poly(A) tail length of Dazap1-L remained unchanged during testis development. The tail was determined to be approximately 30 bases long using a different Dazap1-L specific forward primer that produced shorter ePAT fragments ( Fig. 4C and Table S1). Dazap1-L in both mRNP and polysomal fractions had the same SATs. Additional ePAT reactions showed that the germ-cell specific Sycp3 transcript also had constant SATs of approximately 30 bases ( Fig. 4D and Table  S1). On the other hand, the house-keeping Gapdh transcript produced a minor band representing LATs in the mRNP fraction, suggesting possible translation repression on a small fraction associated with poly(A) tail elongation (Fig. 4E). Our results therefore showed a correlation between poly(A) tail length and translational status for a subset of transcripts in the testis, with LATs of 120,170 bases associated with translation repression.
Several RNA-binding proteins bind to Dazap1 39UTR 39UTRs are known to contain cis-acting elements for the binding of translational regulators. We therefore searched for testicular proteins that bound to Dazap1 39UTR. We separated Dazap1 39UTR into three regions (Fig. 5A). The short (S) region (nt1378-1561) consists of the short 39 UTR present in Dazap1-S; the conserved (C) (nt1568-1793) and the non-conserved (NC) (nt1788-2072) regions include the 59 and 39 segments, respectively, of the distal portion of Dazap1 39UTR that is present only in Dazap1-L. RNA fragments corresponding to these regions were incubated with testis lysates prepared from mice at various ages, and the RNA-protein complexes were UV cross-linked and analyzed by SDS-PAGE (Fig. 5B). Probe S and probe C, but not probe NC, detected multiple proteins, and some of them showed increased binding during testis development.
We further separated the C region into the CU region (nt1568-1646) that contains a noticeable CU-rich sequence and the AU region (nt1644-1793) that contains several AU-rich elements (AREs), and dissected the CU region again into C_40 (not CUrich, nt1568-1610) and CU_40 (nt1607-1646) subregions (Figs. 1A, 5A). UV cross-linking assays showed that most proteins detected by probe C bound to the CU_40 region (Fig. 5C).
We next performed RNA affinity pull-down assays followed by LC/MS/MS spectrometry to identify proteins that bound specifically to Dazap1 39UTR. Using this approach we found several RNA-bind proteins, including MSY1 (Mouse Y box protein 1) and MSY2 (Mouse Y box protein 2) which bound to the S region, and KSRP (KH-type Splicing Regulatory Protein), PTBP2 (Polypyrimidine Tract-Binding Protein 2), and DAZL (Deleted in AZoospermia-Like) which bound to the C region (Fig. 5D). Additional pull-down experiments with smaller probes in the C-region showed that KSRP bound to the AU region whereas PTBP2 and DAZL bound to the CU_40 region.
DAZL preferentially binds to Dazap1-L DAZL has been reported to regulate the translation of several mRNAs in germ cells [10], and DAZL consensus binding sequences are present in different regions of Dazap1 39UTR. We therefore performed RNA gel-shift assays using His-tagged recombinant DAZL to confirm the binding of DAZL to Dazap1 39UTR (Fig. 6). We first mapped the DAZL binding site to the C region that is present only in Dazap1-L, and used smaller probes to narrow down the binding site to the CU_40 region which contains three partially overlapping copies of the DAZL consensus sequence (Fig. 6A). The binding between labeled CU_40 and DAZL could be competed out by increasing levels of cold CU_40 but not cold C_40, confirming the specificity of the binding (Fig. 6B).
We next used RNA immunoprecipitation to investigate whether DAZL preferentially binds to Dazap1-L. An anti-DAZL antibody, but not the preimmune serum, specifically precipitated DAZL from adult testis lysates (Fig. 6C, top panel). 39RACE analyses of the RNAs associated with the immunoprecipitates showed the presence of Dazap1-L but not Dazap1-S (Fig. 6C, bottom panel). The anti-DAZL antibody also brought down Mvh/Ddx4, Sycp3, and Tex19.1 transcripts of which the translation have been previously shown to be regulated by DAZL [20,21,25,26]. Prm1 and Gapdh transcripts which are not known to bind DAZL served as negative controls.

DAZL enhances the translation of a reporter with Dazap1-L 39UTR
We used luciferase reporter assays to indirectly study the effect of DAZL on the translation of Dazap1 transcripts. We transfected 3T3 cells with a DAZL expression vector or the empty expression vector, together with the empty luciferase (Luc) reporter vector or a vector with Luc linked to the 39UTRs of various genes. Western and RT-PCR analyses showed no significant differences in the levels of DAZL (Fig. S2) and Luc mRNA (Fig. 7) in the cells transfected with the DAZL expression vector and different Luc reporters. However, cells transfected with the reporter linked to the 39UTR of Dazap1-L exhibited significantly higher luciferase activities than those transfected with the empty vector or the reporter linked to the 39UTR of Dazap1-S (Fig. 7). Deletion of the entire CU-rich sequence from the Dazap1-L 39UTR (Dazap1-L-dCU) abolished the translation enhancement. The degree of DAZL-induced translation enhancement of Dazap1-L was comparable to those of the reporters linked to the 39UTRs of Mvh, Sycp3, and Tex19.1, which were much lower than what had previously been observed in other systems. DAZL was reported to increase approximately four-folds the translation of Mvh and Sycp3 luciferase reporters in Xenopus oocytes [25,26] and about 10 folds that of a Tex19.1 reporter in mouse oocytes [20]. Differences in the expression levels of DAZL and other trans-acting factors in the various systems may account for the disparity in the observed translational enhancement.

Discussion
Alternative polyadenylation changes the content of the 39UTR and offers a means to regulate gene expression at the posttranscriptional stage because most regulatory elements controlling the localization, stability and translation of a transcript are located Radioactive bands on the hybridization members were excised and counted. Signals at P12 were used as the reference (100%). The results represent the averages of three independent experiments. Statistical significance of the differences was determined using one-way ANOVA. * denotes p,0.05, ** denotes p,0.01, *** denotes p,0.001. The relative levels (%) between 1.8 and 2.4 kb transcripts are given inside the bars. doi:10.1371/journal.pone.0060873.g002 in the 39UTR [43]. Here we show that two Dazap1 transcripts were generated through alternative polyadenylation, and that they were differentially expressed and translated during spermatogenesis. In P12 testes that contain spermatogonia and spermatocytes up to the zygotene stage, the transcripts were present at low levels and were both actively translated. However, with increased Dazap1 expression at P18, there was a major shift in the polysomal profile of Dazap1-S, with an increasing portion of the transcript sequestered to translationally inactive mRNPs, whereas a large portion of Dazap1-L remained on polysomes. In P27 testes that contain mainly postmeiotic spermatids, the level of Dazap1-S increased significantly, consistent with previous EST analyses showing that round spermatids tend to have shorter truncated 39UTRs [44]. Since most of the increased Dazap1-S was present in mRNPs, no increase in protein synthesis was expected and our previous immunostaining of mouse testicular sections also failed to detect significant increase in DAZAP1 signals in spermatids [38]. The situation is similar to that of the Rnf4 transcripts encoding SNURF/RNF4 transcription modulator [45]. Rnf4 has two transcripts of 3.0 and 1.6 kb that differ in their 39UTRs. The 1.6 kb transcript is expressed abundantly in round spermatids but most of them are not associated with polysomes. It is unclear why germ cells spend their energy and resources to make mRNAs that are not used for protein production.
The Northern analyses on Dazap1 expression in prepuberal testes, presented here, suggest that the expression of Dazap1 expression in spermatogonia and early spermatocytes is much lower than that in pachytene spermatocytes. This contradicts the results of our previous RNA in situ hybridization on adult mouse testicular sections which showed highest levels of Dazap1 mRNA in spermatogonia and early preleptotene spermatocytes [38]. We have since repeated the in situ experiments and were able to reproduce the same hybridization pattern with the same 59 probe, but not with a probe from a different region of the Dazap1 cDNA. We also purified different stage germ cells from adult mouse testes and found the highest levels of Dazap1 transcripts in spermatocytes (data not shown). In addition, previous microarray analyses on purified spermatogenic cells also showed almost three times as much Dazap1 transcripts in pachytene spermatocytes as in type A and type B spermatogonia [46]. Therefore our previous RNA in situ hybridization result likely represents an experimental artifact and there is no delayed translation of Dazap1 transcripts as we previously suggested.
We investigated why Dazap1-S was translated on polysomes in early germ cells and sequestered to mRNPS in later germ cells, and found that translational repression of Dazap1-S was associated with an increase in its poly(A) tail length. Translational regulation mediated through poly(A) tail is well documented for some maternal mRNAs during oocyte maturation and early embryogenesis [47]. In the oocytes, masked mRNAs usually have short poly(A) tails fewer than 20 bases, which elongate to about 80 to 150 bases before the mRNAs are recruited to the polysomes for translation. It is thought that longer poly(A) tails facilitate the binding of poly(A) binding protein (PABP) and subsequent recruitment of translational initiation factors. Some translating mRNAs with long poly(A) tails, on the other hand, undergo deadenylation during oocytes maturation, resulting in their translational repression. As first reported by Kleene [41], male germ cells apparently have a reverse pattern of poly(A) tail mediated translational regulation, with the translational activation of several post-meiotically expressed genes accompanied by poly(A) shortening instead of lengthening. The shortened poly(A) tails still contain about 30 As, sufficient for one PABP molecule which requires approximately 27 As for binding [48,49]. Here we confirmed poly(A) shortening of the Prm1 transcript during translational activation using a recently developed ePAT method which offers higher sensitivity and resolution than the previously used Northern blot [41]. The Prm1 short poly(A) tails gave a band that was quite discrete, suggesting that they may not be as heterogeneous as previously suggested. In addition, we found that a small fraction of the mRNP-bound Prm1 transcripts contained SATs. Previous Northern analyses on sucrose gradient mRNAs enriched for transcripts with SATs did detect the presence of trace amount of Prm1, Prm2, Tnp1, and Tnp2 transcripts with SATs in the mRNP fractions [50]. Our ePAT results are likely biased for transcripts with SATs since it is well known that in PCR reactions short fragments are usually amplified with higher efficiency. It remains unclear whether poly(A) shortening is required for translational activation or it is merely a consequence of translation [49,51]. Our observations that a small fraction of the mRNPbound Prm1 transcript contained SATs whereas very little polysomal Prm1 transcript contained LATs would argue for the occurrence of poly(A) tail shortening at mRNPs instead of on polysomes as a consequence of translation. The poly(A) tail elongation of Dazap1-S accompanying its translational repression represents the other side of the coin. Similar to the Prm1 transcript, the mRNP-associated Dazap1-S had both LATs and SATs whereas the polysomal Dazap1-S contained only SATs. However for Dazap1-L which remained largely translationally active during spermatogenesis, there was no obvious changes in its poly(A) tail length. The transcript contained only SATs regardless whether it was sequestered to mRNPs or was actively translating on polysomes. Our results therefore showed that translational repression of Dazap1-S, but not Dazap1-L, was accompanied by poly(A) tail elongation. Whether the poly(A) tail plays an active role in modulating the translational activity of Dazap1-S remains to be investigated. Previous studies on testicular expression of Ldh-x and Rpl32 failed to find an association between poly(A) tail length and translational activity [52,53]. However, the Northern blots in those studies were not of high resolution/quality and the results need to be confirmed with the new ePAT method.
We have identified several proteins, including KSRP, PTBP2, MSY2, MSY1, and DAZL, that bound to Dazap1 39UTR. The only protein that we found to have any effects on Dazap1 expression when ectopically expressed in cultured cells was DAZL. DAZL is known to stimulate mRNA translation, and requires multiple DAZL binding sites for efficient translation activation [19]. DAZL consensus sequence U 2-10 [G/C]U 2-10 is present in different regions of Dazap1 39UTR. The S region contains a UUUCUUUU sequence, the CU_40 region contains three partially overlapping UUCUU sequences, and the AU region contains a UUUUCUU and a UUUUGUUUU sequence. Our RNA gel-shift assays showed that DAZL bound much stronger to the CU_40 region than the S or the AU region. Expression of luciferase reporters in 3T3 cells also indicated that the CU_40 region was responsible for the DAZL induced translation enhancement of the Luc gene linked to Dazap1-L 39UTR. We also found that an anti-DAZL antibody preferentially precipitated Dazap1-L from mouse testis lysates. There is a good correlation between the expression of DAZL and DAZAP1 during spermatogenesis. Previous immunostaining of adult mouse testicular sections detected dramatic increases in the expression of both DAZAP1 and DAZL in pachytene spermatocytes [38,54]. Our current analyses of Dazap1 expression during spermatogenesis showed that the level of Dazap1-L also increased significantly in pachytene spermatocytes. Therefore, increased expression of both Dazap1-L and DAZL in pachytene spermatocytes likely accounts for the robust translation of DAZAP1 in the cells.

Materials and Methods
Animals C57BL/6 mice used in the study were housed in a specific pathogen-free animal facility. Adult mice were 2-4 months old. The experiments were approved by Academia Sinica's Institutional Animal Care & Utilization Committee.

Northern hybridization
Northern hybridization was performed according to the standard protocol [56] using Sure Blot nylon membranes (Intergen, Purchase, NY) and a Stratalinker UV cross-linker 1800 (Stratagene, La Jolla, CA). Blots were hybridized with radioactive probes at 68uC in the ExpressHyb solution (Clontech, Palo Alto, CA) and analyzed with the Typhoon 9410 Imager (GE Healthcare, Piscataway, NJ). For quantification, the hybridization membranes were exposed to X-ray films and radioactive bands were located, excised, and counted.

Sucrose gradient analysis
Post-mitochondrial fractions of mouse testis lysates were fractionated on 10-50% sucrose gradients as previously described [57]. Briefly, mouse testes were homogenized with a motor-driven glass-Teflon homogenizer (Wheaton, Millville, NJ) on ice in 0.5 ml lysis buffer (100 mM NaCl, 3 mM MgCl 2 , 20 mM Hepes, pH 7.4 and 0.5% Triton X-100). The nuclei and mitochondria were pelleted by centrifugation for 5 min at 12,000 g. The supernatants were loaded onto linear gradients of 10%-50% sucrose (w/w) in the lysis buffer without Triton X-100 and centrifuged for 2 h at 36,000 rpm in a SW-41 rotor (Beckman, Palo Alto, CA). The gradients were pumped through a flow cell (Teledyne Isco, Lincoln, NE) to record the absorbance at 254 nm and fractionated into 0.5-ml fractions. Each fraction was extracted twice with phenol:chloroform (1:1) and the RNA was ethanol precipitated. Polysome profiles of the Dazap1 transcripts were analyzed by 39RACE-PCR or Northern hybridization. Control gradients, containing 16 mM EDTA in the postmitochondrial supernatants, were centrifuged and analyzed in parallel with the polysome gradients.

UV cross-linking assays
Testis extracts were prepared from mice at indicated age according to Xu and Hecht [58]. 32 P-labelled single stranded RNA probes were in vitro transcribed using Riboprobe systems (Promega) and purified on 4.5% denaturing polyacrylamide gels containing 8 M urea. UV cross-linking of RNA-protein complexes was performed according to Xu and Hecht [58] with minor modifications. RNA-protein binding was carried out at room temperature for 20 min in 20ml reaction mixtures containing 20 mM Hepes [pH 7.6], 3 mM MgCl 2 , 40 mM KCl, 2 mM DTT, 5% [w/v] glycerol, 40 mg testis extracts, and 5610 4 cpm of 32 P-labelled RNA probes. Heparin was then added to a final concentration of 5 mg/ml to reduce nonspecific RNA-protein interactions. After10 min, the reaction mixtures were exposed to UV light (Stratalinker 1800) for 11 min on ice, followed by RNase A (0.5 mg/ml) digestion for 15 min at 37uC. The RNA-protein complexes were resolved by 10% SDS-PAGE. The gels were dried and visualized with the Typhoon 9410 Imager.

RNA affinity pulldown assays
RNA affinity pulldown assay was done according to a published protocol [58] with minor modifications. Biotinylated RNA probes were in vitro transcribed using the Riboprobe system containing biotin-11-UTP (PerkinElmer, Boston, MA). Testes extracts containing 1 mg total proteins were mixed with 20 mg biotinylated RNA in 400 ml binding buffer (BB) containing 20 mM Tris-HCl [pH 7.4], 2.5 mM MgCl 2 , 120 mM KCl, 1 mM DTT, 5% [w/v] glycerol, and16 protease inhibitor cocktail for 30 min at room temperature and treated with 5 mg/ml heparin for 15 min. Streptavidin-agarose beads were then added to the mixture and incubated for 60 min. After five washes with BB, bound proteins were released by boiling in 46 SDS sample buffer, resolved by 10% SDS-PAGE and visualized by SYPRO Ruby staining. The protein bands of interest were excised and analyzed by LC/MS/ MS with a Thermo Scientific LTQ XL mass spectrometer (Thermo Scientific, Waltham, MA).

RNA immunoprecipitation assays
RNA immunoprecipitation assay was performed according to Xu and Hecht [58]. Two mg testes extracts were precleaned using 50 ml protein G beads (GE Healthcare) and 100 mg/ml yeast tRNA in 1 ml TKM buffer (20 mM Tris [pH 7.4], 120 mM KCl, 1.5 mM MgCl 2 , 0.1% NP-40) with 500 U/ml SUPERase-In (Ambion, Austin,TX) for 4 hours at 4uC. After spinning down the beads, the supernatants were incubated overnight at 4uC with 10 mg anti-DAZL-C antibody which was generated against an oligopeptide containing the last 21 amino acid residues of mouse DAZL and affinity purified using the services of Bethyl Laboratories, Inc (Montgomery, TX). The mixtures were then incubated  with 20 ml protein G beads for 3 hours at 4uC. The beads were spundown and washed five times by the TKM buffer. Immunoprecipitated RNAs were extracted by Trizol (Invitrogen) and analyzed by RT-PCR. Sequences of the PCR primers used in RNA immunoprecipitation assays are given in Table S2.

RNA gel shift assays
Gel shift assays were carried out according to Venables et al [22] with some modifications. The 10ml binding reactions contained 1 mg recombinant DAZL, 20 mM Tris [pH 7.4], 2.5 mM MgCl 2 , 60 mM KCl, 1 mM DTT, 5% [w/v] glycerol, and 10 4 cpm of 32 P-labelled RNA probes. After incubation at room temperature for 5 min, the reactions were treated with 5 mg/ml heparin and 0.5 mg/ml yeast tRNA for another 15 min. The RNA-protein complexes were separated on 1.5% agarose gels in 0.56TBE buffer and visualized with the Typhoon9410 Imager.

Plasmid construction
The entire 39UTRs of Dazap1, Mvh, Sycp3 and Tex19.1 were PCR amplified from mouse genomic DNA and cloned into pCRII-TOPO. pcDNA-Luc was generated by inserting the Luc gene into pcDNA3.1 (Invitrogen). The 39UTRs of the various genes in pCRII-TOPO were then cloned into pcDNA-Luc to generate the reporter vectors. Deletion of the CU-rich sequence in Dazap1 39UTR was generated by a PCR-mediated deletion approach [59]. For in vitro transcription, different regions of Dazap1 39UTR were generated by PCR amplification and cloned into pBluescript (Stratagene). Dazl coding region was RT-PCR amplified from mouse testis RNA and cloned into pCRII-TOPO for sequencing. It was subsequently cloned into p3xFLAG-CMV-14 (Sigma, Saint Louis, MI) to generate pDAZL-Flag, and into pET-28a (Novagen, Madison, WI) to generate pET-DAZL for the production of His-tagged DAZL protein in E. coli [22,60].

Analysis of the translation of luciferase reporters
3T3 cells were maintained in DMEM (Hyclone, Fremont, CA) supplemented with 10% fetal bovine serum and 16 penicillin/ streptomycin (GIBCO, Eggenstein, Germany). Transfection was performed using the TurboFect reagent (Fermentas, Burlington, Ontario, Canada). The pDAZL-Flag or the empty p36 FLAG-CMV-14 were co-transfected into 3T3 Cells with different firefly luciferase reporter constructs and pRL-TK (Promega, San Luis Obispo, CA) which encodes the Renilla luciferase. The cells were harvested after 24 hours, and luciferase activities in the cell lysates were measured using the dual-luciferase reporter assay system (Promega) and normalized to Renilla luciferase activities. Luciferase RNA levels were determined by semi-quantitative RT-PCR and normalized to the levels of Renilla luciferase transcripts. Luciferase activities and RNA levels in cells transfected with pDAZL-Flag were normalized to those in cells transfected with p36 FLAG-CMV-14 to reflect the effects of DAZL, and the results of cells transfected with different Luc reporter genes were compared using those of cells transfected with the empty Luc reporter as a reference (100%). DAZL expression levels were determined by Western blot and normalized to the levels of tubulin.

Poly(A) tail length determination
A newly developed extension poly(A) test (ePAT) method was used to measure the poly(A) tail of mRNA [42]. This method makes use of the ability of the Klenow polymerase to extend the 39 ends of RNA molecules with dNTPs from an annealed DNA template in standard reverse transcription buffers. The PATanchor primer 59-cgctacgtaacggcatgacagtgt 22 contains a unique universal DNA segment plus 22 Ts at the 39 end and anneals to the end of poly(A) tails to serve as the DNA template for the Klenow polymerase. After the initial extension, the PAT-anchor primer also serves as the primer for the standard reverse transcription. The incubation was carried out at 55uC to eliminate unwanted priming from internal poly(A) tracts. The protocol started with heating 1 mg RNA together with 1 ml of 100 mM PAT-anchor primer in a volume of 8 ml at 80uC for 5 min followed by cooling down to 37uC. A master mixture containing 4 ml DEPC-treated water, 4 ml 56 Superscript III buffer (Invitrogen), 1 ml 100 mM DTT, 1 ml 10 mM dNTPs, 1 ml RNaseOUT (Invitrogen), and 1 ml (5 U) Klenow polymerase (New England Biolabs, Ipswich, MA) was added and the solution was incubated at 37uC for 1 hour. The reaction mixture was then heated at 80uC for 10 min to inactivate the polymerase and cooled down to 55uC. One ml (200 U) Superscript III (Invitrogen) was added and incubation was kept at 55uC for 1 hour. The reverse transcriptase was inactivated at 80uC for 10 min at the end of the reaction. PCR amplification of the 39 ends of specific genes was performed using a universal reverse primer 59-cgctacgtaacggcatgacagtg complementary to the 59 end of the PAT-anchor primer and gene-specific forward primers, the sequences of which are provided in Table S3. A separate TVN-PAT reaction was also performed to generate a fragment with a defined poly(A) tail of 12 As. It served as a size standard for other PAT products that were amplified from the same gene specific forward primer. The TVN-PAT primer has an identical sequence as the PAT anchor primer except the addition of two 39 variable bases V and N (V is A, G or C and N is any base) which fix the primer to the polyadenylation site during reverse transcription. The e-PAT products were analyzed on 2% agarose gels. Fragments of interest were purified from the gels, cloned into pCRII-TOPO and sequenced to determine the lengths of the poly(A) tails.