Dicer1 Ablation in the Mouse Epididymis Causes Dedifferentiation of the Epithelium and Imbalance in Sex Steroid Signaling

Background The postnatal development of the epididymis is a complex process that results in a highly differentiated epithelium, divided into several segments. Recent studies indicate a role for RNA interference (RNAi) in the development of the epididymis, however, the actual requirement for RNAi has remained elusive. Here, we present the first evidence of a direct need for RNAi in the differentiation of the epididymal epithelium. Methodology/Principal Findings By utilizing the Cre-LoxP system we have generated a conditional knock-out of Dicer1 in the two most proximal segments of the mouse epididymis. Recombination of Dicer1, catalyzed by Defb41iCre/wt, took place before puberty, starting from 12 days postpartum. Shortly thereafter, downregulation of the expression of two genes specific for the most proximal epididymis (lipocalin 8 and cystatin 8) was observed. Following this, segment development continued until week 5 at which age the epithelium started to regress back to an undifferentiated state. The dedifferentiated epithelium also showed an increase in estrogen receptor 1 expression while the expression of androgen receptor and its target genes; glutathione peroxidase 5, lipocalin 5 and cysteine-rich secretory protein 1 was downregulated, indicating imbalanced sex steroid signaling. Conclusions/Significance At the time of the final epididymal development, Dicer1 acts as a regulator of signaling pathways essential for maintaining epithelial cell differentiation.


Introduction
After development in the testis, the spermatozoa travel through the epididymis where they mature, gaining motility and the ability to fertilize the oocyte. Despite consisting of a single long duct, the epididymis is a complex organ, highly convoluted and divided into several anatomical and functional segments, i.e. initial segment (IS), caput (CAP), corpus (COR) and cauda (CAU). Each segment synthesizes and secretes a specific set of proteins, thus creating the unique luminal environment needed for the sperm maturation process [1][2][3]. The most proximal segments (IS and CAP) have been proven essential for sperm maturation as the disruption of their development or function often leads to male infertility [4][5][6][7]. The epididymis develops from the mesonephric tubules and the proximal Wolffian duct (WD) [8]. During the embryonic stage, and before epithelial differentiation, mesenchymal androgen receptor (AR), along with inhibin beta A (Inhba), facilitates the elongation and convolution/coiling of the tubule [9][10][11]. Development of the epididymis continues after birth with differentiation of the epithelial cells into principal, basal and narrow/clear cells [2,2,12,12,13]. At the onset of spermatogenesis, the epididymal epithelium develops segment-specific gene expression [14]. From studies with genetically modified mice, it has become evident that leucine-rich repeat-containing G protein-coupled receptor 4 (LGR4, also known as GPR48) is needed for the postnatal epididymal coiling and the differentiation of IS [15,16]. In addition, the proto-oncogene Ros1 (ROS1, also known as c-ros) is necessary for the formation of IS [5]. Recent studies with epididymal AR knock-out mice have revealed that androgen signaling is required for the formation of IS and differentiation of principal and basal cells [6,7,10]. However, there are still many unresolved issues regarding the regulatory pathways responsible for differentiation of the epididymal segments and their specific gene expression.
Many aspects of development are regulated by RNA interference (RNAi). Small non-coding RNAs bind to complimentary mRNA sequences and cause translational silencing and mRNA cleavage [17]. One class of small non-coding RNAs is the microRNAs (miRNAs) which are initially produced as longer precursors that need to be processed by the RNaseIII enzyme Dicer1 to become fully functional [18]. The ,22 nt-long mature miRNAs then control protein expression and modulate diverse cellular events such as differentiation, proliferation, apoptosis and cell metabolism [19,20]. Consequently, Dicer1 deficient mice die already on embryonic day 7.5 owing to complete loss of pluripotent stem cells [21]. To study the effect of RNAi in the development and function of specific organs, Dicer1 conditional knock-out (Dicer1 cKO) mice can be used. The need for Dicer1 in the differentiation of several cell types, including sensory epithelial cells, T cells and pancreatic b-cells, has been shown in Dicer1 cKO mice by crossing Dicer1 fl/fl mice with those expressing Cre in the cells of interest [12,[22][23][24]. Furthermore, a requirement for Dicer1 in the patterning of the liver and colon was also confirmed with Dicer1 cKO mice. Hepatocyte specific Dicer1 ablation compromises region-specific protein expression while Dicer1 depletion in the developing colon leads to a disorganized epithelium [25,26].
Previous studies on human and rat epididymides have shown that miRNAs are differentially expressed at juvenile and adult stages [27,28], indicating a role for miRNAs in the postnatal development of the epididymis. To study the function of RNAi in the developing epididymis we generated epididymis-specific Dicer1 cKO mice by crossing Dicer1 fl/fl mice with Defb41 iCre mice. This resulted in the elimination of Dicer1 expression from the prepubertal epididymis. Our results demonstrate the importance of Dicer1 in sex steroid signaling and in maintenance of the differentiated state of the epididymal epithelium.

Results
Generation of Dicer1 fl/fl ; Defb41 iCre/wt Mice Quantitative RT-PCR of the mouse epididymis showed a continuous expression of Dicer1 from birth into adulthood ( Figure 1C). As the full knock-out of Dicer1 is embryonically lethal [21], we generated a Dicer1 cKO mouse line by crossing Dicer1 fl/fl mice [29] with a mouse line expressing iCre under the Defensin beta 41 (Defb41) promoter. The Dicer1 fl allele consists of two loxP sites flanking exon 24, which contains a major part of the second RNaseIII domain ( Figure 1A). The heterozygous Defb41 iCre mouse line did not show any phenotypic defects or fertility problems and expressed iCre in the epithelium of the most proximal part of the epididymis, IS and CAP. Recombination of Dicer1 was observed in 12 day-old Dicer1 fl/fl ; Defb41 iCre/wt mouse IS and CAP by genomic PCR ( Figure 1B) and qRT-PCR studies revealed a significant reduction in Dicer1 expression levels at the age of 2 months ( Figure 1D).
Morphology of the Epididymis and Fertility of Dicer1 fl/fl ; Defb41 iCre/wt Males Macroscopic evaluation of 2 month-old Dicer1 fl/fl ; Defb41 iCre/wt mice epididymides revealed an underdeveloped IS and, in addition, the mice frequently presented with enlarged efferent ducts (Figure 2A, B). The IS of control mice can be clearly visualized owing to the endogenous b-galactosidase activity in the segment. The much smaller IS of Dicer1 fl/fl ; Defb41 iCre/wt mice could not be distinguished from CAP with X-gal staining ( Figure 2A). Furthermore, the intense vasculature typical of WT IS was missing from the Dicer1 cKO IS. Histological evaluation showed a division of the epididymis into different segments ( Figure 2B) but the epithelial cell layer of both IS and CAP was disorganized ( Figure 3H). Dicer1 fl/fl mice have a similar phenotype to WT mice epididymides and were used as controls throughout the study. Dicer1 cKO epididymides were significantly smaller than those of control mice (30.461.5 mg, control: 35.460.7 mg, P#0.01). No significant difference in the weight of 6 month-old Dicer1 cKO and control mice epididymides was observed. However, the epithelial cell layer of the 6 month-old Dicer1 fl/fl ; Defb41 iCre/wt mouse was further disturbed, with neoplastic changes in the efferent ducts causing their progressive obstruction ( Figure  S1). Even though sperm were detected in the CAU, 2-to 3-monthold Dicer1 fl/fl ; Defb41 iCre/wt male mice failed to produce offspring when mated with WT females ( Table 1). The number of sperm was reduced in Dicer1 fl/fl ; Defb41 iCre/wt mouse epididymides as histological staining showed some tubular cross sections with no sperm. At 6 months of age the number of tubular cross sections without sperm was further increased due to the obstruction of the Dicer1 fl/fl ; Defb41 iCre/wt mouse efferent ducts. The testis of the 6 month-old Dicer1 fl/fl ; Defb41 iCre/wt mouse also displayed disruption of the seminiferous epithelium owing to fluid back-pressure (data not shown). Further morphological analyses revealed that the number of sperm with angulated tails was not significantly increased in 2 month-old Dicer1 fl/fl ; Defb41 iCre/wt mouse CAU (22.963.4% of all sperm, control: 16.361.2%).

Effects of Dicer1 cKO on Epididymal Cell Differentiation
At the age of one month, the IS of both control and Dicer1 fl/fl ; Defb41 iCre/wt mice was distinguishable from the CAP by its high, columnar-shaped epithelial cells ( Figure 3C, D). However, the height of the epithelium of Dicer1 cKO IS was reduced below that of control mouse IS. In the 45 day-old mouse, the epithelial cells of the Dicer1 cKO IS had regressed to an undifferentiated state with a disorganized epithelial cell pattern ( Figure 3E, F), resembling those of the epididymis of 14 day-old mice ( Figure 3A, B). Furthermore, the height of the epithelial cells of IS was significantly reduced (average cell height in control IS: 45 mm, in Dicer1 cKO IS: 28 mm, P#0.0001) in 2 month-old Dicer1 fl/fl ; Defb41 iCre/wt mice. However, the phenotype of the Dicer1 cKO IS varied between individuals from a severely disorganized epithelial cell layer to a thin epithelial cell layer resembling that of COR ( Figure S2) in some individuals.
To study further the effect of Dicer1 ablation on epithelial cell differentiation, the presence of different epithelial cell types of the epididymal epithelium was analyzed by immunohistochemistry. Phalloidin conjugated to TRITC was used to stain the F-actin of principal cells, and antibodies against vacuolar H + -ATPase (V-ATPase) and Keratin 5 (Krt5) were used to stain clear/narrow cells and basal cells, respectively. The results indicated that all cell types were present in the Dicer1 cKO epididymis (Figure 4). The epithelial cell types were found in similar numbers and locations in the Dicer1 cKO epithelium as in control animals. However, a-actin staining revealed an increase in muscle cell layer thickness ( Figure 4G, H). The Dicer1 fl/fl ; Defb41 iCre/wt mice displayed three layers of muscle cells surrounding the epididymal duct, whereas control mice typically had one.
Genes specifically expressed in the mature proximal epididymis, IS and proximal CAP; lipocalin 8 (Lcn8), brain expressed myelocytomatosis oncogene (Bmyc) and cystatin 8 (Cst8), showed a marked downregulation in their expression in IS and CAP of 2 month-old Dicer1 fl/fl ; Defb41 iCre/wt mice. The only exceptions were Ros1 and G protein-coupled receptor 64 (Gpr64, also known as HE6) that showed no significant difference in expression levels compared with the control ( Figure 4I). The expression of Cst8, Lcn8 and Ros1 is initiated at postnatal days 20, 21 and 16, respectively [5,30,31]. The qRT-PCR results from different time points during epididymal development showed significantly reduced expression of both Cst8 and Lcn8 in Dicer1 fl/fl ; Defb41 iCre/wt mice, from 21 days postpartum onward ( Figure S3A, B). However, the expression levels of Ros1 did not differ significantly from that of control mice ( Figure S3C).

Cell Proliferation and Apoptosis
Immunohistochemical studies showed a marked increase in cell proliferation throughout the proximal epididymis of 2 month-old Dicer1 fl/fl ; Defb41 iCre/wt mice ( Figure 5B). The number of Ki-67 positive cells was on average 2 times higher in Dicer1 cKO IS (15.263.0 cells/mm, control: 6.460.6 cells/mm, P#0.05) and almost 6 times higher in Dicer1 cKO CAP (11.161.7 cells/mm, control: 1.960.5 cells/mm, P#0.001) compared with those of control mice epididymides ( Figure 5C). Although the epithelium was highly proliferative, it did not lead to a marked increase in the size of the Dicer1 fl/fl ; Defb41 iCre/wt mouse epididymides. On the contrary, at the age of 2 months the Dicer1 cKO epididymis was significantly smaller than that of the control mouse. Furthermore, the number of apoptotic cells was increased in Dicer1 cKO IS and CAP (Dicer1 cKO IS: 0.8960.39 cells/mm, control IS: 0.0760.07 cells/mm; Dicer1 cKO CAP: 1.1260.34 cells/mm, control CAP: 0.0760.07 cells/mm, P#0.05) ( Figure 5D-F). The difference seen in IS was not statistically significant owing to high variation between the individual mice.

Expression of Sex Steroid and Fibroblast Growth Factor Receptors in the Dicer1 cKO IS and CAP
To study whether the observed epididymal phenotype is caused by changes in sex steroid signaling, the expression of Ar, estrogen receptor 1 (Esr1, also known as ERa) and estrogen receptor 2 (Esr2, also known as ERb) was analysed by qRT-PCR and immunohistochemistry in 2 month-old Dicer1 cKO and control epididymides. The qRT-PCR results showed a significantly weaker expression of Ar in Dicer1 cKO IS and CAP ( Figure 6E), and downregulation of AR target gene expression, cysteine-rich secretory protein 1 (Crisp1), glutathione peroxidase 5 (Gpx5) and lipocalin 5 (Lcn5), was also observed ( Figure 6F). Esr2 showed a similar reduction in mRNA expression levels as Ar, while the expression of Esr1 did not significantly differ between Dicer1 cKO and control IS and CAP ( Figure 6E). However, immunohistological staining displayed an altered ESR1 expression pattern in Dicer1 cKO IS. ESR1 was foremost expressed in the narrow cells of control mouse IS, whereas the Dicer1 cKO IS showed equal expression of the receptor in almost all epithelial cells ( Figure 6A, B). Furthermore, immunohistochemical analyses clearly showed weaker AR staining in a number of the Dicer1 cKO epididymides analysed ( Figure 6D). Nevertheless, the AR findings were variable and some Dicer1 cKO epididymides had an AR staining equal to that of the controls. Furthermore, the relative expression values of fibroblast growth factor receptors 1-4 (Fgfr1-4) were not statistically different when comparing qRT-PCR results from Dicer1 fl/fl ; Defb41 iCre/wt mice with control mouse IS and CAP (data not shown).

Discussion
At birth, the epididymis consists of a long convoluted duct lacking both cell type and segment specific markers. In the mouse epididymis, differentiation of the epithelial cell types takes place between day 14 and day 21 [32] after which segment-specific gene expression can be observed [14]. Nonetheless, very little is known about the regulatory pathways responsible for the epididymal segmentation and the differentiation of epithelial cell types. In this study, we demonstrate that post-transcriptional regulation via RNAi signaling is an important regulator in the development of the proximal epididymal segments.
In our Dicer1 fl/fl ; Defb41 iCre/wt mouse model the ablation of Dicer1 begins 12 days pp but, at the time of the final cell type differentiation during puberty, the epididymal epithelium of Dicer1 fl/fl ; Defb41 iCre/wt mice still resembles that of control mice.
Studies revealed a significant reduction in the expression of IS specific genes at 21 days pp, and about 2 weeks later, the epithelium began to regress. At 45 days of age, the Dicer1 KO IS morphologically resembled that of an undifferentiated prepubertal epididymis. Despite the morphological changes, Dicer1 ablation did not affect cell type differentiation as all major epithelial cell types were detected in the epididymides of 2 monthold Dicer1 fl/fl ; Defb41 iCre/wt mice. However, the function of the principal cells was compromised as they displayed a significant reduction in segment specific gene expression.
When studying the different epithelial cell types, we detected a thicker layer of smooth muscle cells surrounding the duct of the Dicer1 fl/fl ; Defb41 iCre/wt mouse epididymis. As Dicer1 ablation is not likely to occur in the stromal tissue, the observed increase in smooth muscle cell number may indicate altered epithelialmesenchymal signaling. It has previously been shown that cross talk between different cell types is essential for normal epididymal functions, and for example the communication between clear cells and other epithelial cell types is required to maintain luminal pH [33]. Interestingly, there are species-specific differences in thickness and distribution of the epididymal smooth muscle cell layer [34][35][36]. Further investigations regarding smooth muscle cell differentiation could thus benefit from comparative studies of different species.
Previous studies have shown that ROS1 is one of the master regulators of IS development, as the lack of both ROS1 and its negative regulator, protein tyrosine phosphatase SHP-1, causes a defect in IS differentiation [5,37]. This affects also the regulation of sperm volume as the majority of sperm from the Ros1 KO CAU present with angulated tails [38]. However, Dicer1 cKO IS still has ,70% of the Ros1 expression of the control IS and the dedifferentiated IS does not cause a significant increase in sperm tail angulation. Furthermore, males with one intact Ros1 allele have normal epididymal development [5]. These data indicate that the epididymal phenotype, observed in the present study, is independent of ROS1 signaling. Other proteins that have a regulative role in IS are the Fibroblast growth factors. They control gene expression by binding to FGFRs in the epithelium of IS [39,40]. Studies in other male reproductive organs also indicate a role for FGFRs in regulating stromal-epithelial induction of cell proliferation and tissue homeostasis [41]. To assess the effect of FGF signaling on the observed epididymal phenotype, we studied the expression of Fgfrs in the epididymis. However, no changes were observed and it is therefore unlikely that FGF signaling contributes to the dedifferentiation of the epithelium or to the increased smooth muscle cell layer thickness of the Dicer1 fl/fl ; Defb41 iCre/wt mice.
Sex steroids are important regulators of epididymal development and function. In the epididymis, epithelial expression of Ar starts around E14.5-15.5 [10,42] and is required for IS development [7] and for the differentiation of principal and basal cells in the proximal epididymis [6,10]. A further role for androgen signaling is observed after orchidectomy, where androgen depletion leads to extensive apoptosis throughout the epididymal  epithelium of both prepubertal and postpubertal animals [43]. In male reproductive organs, estrogens are produced in the testis, spermatozoa and epididymis [44,45]. Esr1 is highly expressed throughout the efferent ducts, where it regulates fluid reabsorption of non-ciliated epithelial cells [46,47]. Less is known about the function of ESR1 in the epididymis, although high expression of Esr1 is detected in the narrow cells of IS and throughout the epithelium of CAP [46]. Recent studies have indicated a role for ESR1 in luminal pH maintenance [48,49] as well as in smooth muscle contractility [50]. Esr2 is transcribed in all cells of the epithelium, however, its role in the epididymis is still unknown since the full knock-out of Esr2 does not display any difference in morphology or function of the epididymis compared with that of WT mice [51]. The Dicer1 fl/fl ; Defb41 iCre/wt mouse model presents with a significant downregulation in Ar, known AR target genes and Esr2 mRNA expression in the proximal epididymis. However, on the basis of the current study, we cannot distinguish whether the down-regulation of the target genes is due to a direct lack of AR signaling or an effect of the observed epithelial cell dedifferentiation. ESR1, on the other hand, displayed an altered expression pattern in the Dicer1 KO IS, with expression not only in narrow cells, as in IS of control mice, but throughout the epithelium. As ESR1 is known to promote cell proliferation [52], high ESR1 levels in the Dicer1 cKO IS could explain the observed increase in cell proliferation in that epididymal segment.
Studies on excessive estrogen signaling in males suggest that an imbalance in the testosterone-estrogen ratio could be of more importance than the actual increase in estrogen signaling [53]. Administration of a single high dose of diethylstilbestrol (DES), a potent synthetic estrogen, to neonatal rats, caused a reduction in the epithelial height of the epididymis. However, the increased estrogen levels gave rise to an additional downregulation of AR in the entire epididymis and thus an imbalance in sex steroid signaling. When the estrogen-testosterone ratio was adjusted by co-administration of testosterone, no phenotypic abnormalities were observed [54]. The reduction in Ar expression observed in Dicer1 fl/fl ; Defb41 iCre/wt mice could therefore not only have significant consequences for epididymal development but also be directly caused by the increased ESR1 levels. However, a recent study on a proximal epididymis-specific AR cKO, also showed ESR1 expression in all cell types of the proximal epididymal epithelium [7]. This would indicate a role for AR in the repression of Esr1 expression. AR cKO epididymides also showed similar features to those of the Dicer1 fl/fl ; Defb41 iCre/wt mice with an altered epididymal stroma showing a disrupted smooth muscle cell layer or a thickened mesenchyme [7,55]. In light of these studies, AR and ESR1 expression in the epididymis seem to be tightly linked and although our current results show an imbalance in sex steroid receptor ratio we cannot with certainty say if the imbalance was initiated by a lowered AR expression or an increase in ESR1 expression.
Adams et al. were the first to show that Esr1 is a direct target of miR-206 [56]. Since then, around 14 evolutionarily conserved miRNAs have been found to target directly the 39-UTR of mammalian Esr1 [57]. There is also evidence of a miRNA-induced downregulation of Esr1 after testosterone treatment in the female mouse liver [58]. In light of these previous studies the upregulation of ESR1 in Dicer1 KO IS might be directly ascribable to the ablation of mature miRNAs. Interestingly, previous studies have shown induction of miRNA expression by AR in vivo [59]. Androgen Responsive Elements (AREs) are also found in the promoter region of several miRNAs [60,61]. Although AR would not induce expression of all epididymal miRNAs, the observed phenotype of Dicer1 fl/fl ; Defb41 iCre/wt mice could be partially AR dependent. To understand better the role of the RNAi pathway during epididymal development, miRNA expression data from different developmental time points is needed. These results could be further compared to already available array data from the human and rat epididymis [27,28] to demonstrate evolutionarily conserved RNAi regulation.
In conclusion, the present study shows the importance of the RNAi pathway in the postnatal development of the proximal epididymis. Ablation of Dicer1 in the epididymal epithelium causes a regression in cell differentiation. The epithelium maintains cell type identity but lacks a segment-specific gene expression pattern. Furthermore, the epididymal phenotype of Dicer1 fl/fl ; Defb41 iCre/wt mice resembles that of mice with an imbalanced ratio of sex steroid receptors, AR cKO mice and mice with excess estrogen signaling. (I) qRT-PCR for proximal epididymis-specific gene expression. Expression of lipocalin 8 (Lcn8), brain expressed myelocytomatosis oncogene (Bmyc), cystatin 8 (Cst8), Ros1 proto-oncogene (Ros1) and G protein-coupled receptor 64 (Gpr64) in the initial segment and caput of 2 month-old control and Dicer1 conditional knock-out (cKO) mouse epididymides. Values relative to L19 expression. Statistical significance was calculated from the expression levels of 3 control and 3 Dicer1fl/fl;Defb41iCre/wt mouse (Ros1:8 control and 9 Dicer1fl/fl;Defb41iCre/ wt mouse) samples using the unpaired t-test. Statistical significance of changes is indicated as follows: ns, not significant; *, P#0.05; **, P#0.01. Scale bars 100 mm. doi:10.1371/journal.pone.0038457.g004 This suggests that Dicer1-dependent pathways are important regulators of balanced estrogen-androgen action in the mouse epididymis.

Ethics Statement
All mice were handled in accordance with the institutional animal care policies of the University of Turku (Turku, Finland), and every effort was made to minimize suffering of the animals. Mice were specific pathogen-free, fed with complete pelleted chow and tap water ad libitum in a room with controlled light (12 hours light, 12 hours darkness) and temperature (2161uC). Animal experiments were approved by the Finnish Animal Ethics Committee (license number: 20072-Sipilä and 2010-05926-Poutanen), and the institutional policies on animal experimentation fully met the requirements as defined in the NIH Guide on animal experimentation.

Experimental Animals
To inactivate Dicer1 conditionally from the epithelial cells of the proximal epididymis, Dicer1 fl/fl mice [29] were crossed with a heterozygous Defb41 iCre knock-in (KI) mouse line. The KI was generated by inserting the iCre cDNA allelle into the translation initiation site of the Defb41 gene. A detailed characterization of the Defb41 iCre KI mouse line will be described elsewhere. Defb41 iCre/wt mice were genotyped with the primers Defb41 Fw:  TCCATTGCCTTTTCTTGTCC  and  Defb41 Re: TTGTCTTACCAGGTTTCCTCCT, Tm 56uC, for the wildtype (WT) allele and iCre Fw: TCTCCAACCTGCTGACTGTG and iCre Re: AGGGACACAGCATTGGAGTC, Tm 59uC, for the iCre allele. Dicer fl/fl mice were genotyped as previously described [29]. For detection of the floxed allele, IS together with CAP, COR together with CAU and testes from 12 day-old mice were collected. After DNA extraction, the recombined allele was detected by genomic PCR as previously described [29]. Dicer1 fl/fl ; Defb41 iCre/wt mice and the controls, Dicer1 fl/fl mice, were obtained from the same litters. The genetic background of these mice was mixed C57Bl/6N and SV129.
Sperm Morphology and Fertility of Dicer fl/fl ; Defb41 iCre/wt Mice To study male fertility, 2-3 month-old male Dicer1 fl/fl ; Defb41 iCre/wt mice and control mice were mated with FVB/N female mice. The female mice were superovulated by intraperitoneal injections of 5 IU pregnant mare serum gonadotropin (PMSG, NHPP, Dr. Parlow) and 5 IU human chorionic gonadotrophin (hCG, Pregnyl, Schering-Plough) 26 and 2 hours before mating, respectively. Females were checked for copulatory plugs the following morning. The mated females were followed for 3-4 weeks to determine the number of litters and offspring produced by each male. For sperm analyses, CAU of two-monthold control and Dicer1 fl/fl ; Defb41 iCre/wt mice were dissected and incubated in 300 ml of HTF medium (William A. Cook Australia Pty. Ltd, Brisbane, Australia) at 37uC, 5% CO2 for 30 min, to allow the sperm to swim out. The sperm were spread on microscope slides and stained using the Papanicolaou technique (Haematoxylin, OG-6 and EA-50). The morphology of 100 sperm/sample from 4 control and 5 Dicer1 fl/fl ; Defb41 iCre/wt mice samples was analyzed.

Histology, Immunohistochemistry and b-galactosidase Staining
Epididymides from 19 day -6 month-old control and Dicer1 fl/fl ; Defb41 iCre/wt mice were fixed overnight in 4% paraformaldehyde (PFA), embedded in paraffin and prepared for histological analyses by standard procedures. Haematoxylin and eosin (HE) staining was performed by standard procedures. Epithelial cell height was measured by using the Leica IM500 imaging software. To detect the different epithelial cell types of the epididymis, the slides were incubated with the following antibodies overnight at 4uC: rabbit monoclonal anti-Keratin 5 1:100 (RM-2106, Thermo Scientific), rabbit polyclonal anti-V-ATPase 1:100 (a gift from S. Breton, Program in Membrane Biology, MGH Simches Research Center, Boston, MA) and mouse monoclonal anti-a-actin 1:500 (sc-32251, Santa Cruz Biotechnology). All the primary antibodies were diluted in PBS supplemented with 1% bovine serum albumin. The antibody-antigen complexes were visualized by incubation for 30 min at room temperature with 1:200 Alexa Fluor 488-and 594-conjugated goat anti-mouse and goat anti-rabbit antibodies (Invitrogen). For detection of F-actin, the epididymides of 2 month-old mice were fixed for 15 min in 4% paraformaldehyde and rapidly frozen. The frozen sections (8 mm) were immunostained with Phalloidin-TRITC 1:400 (P1951, Sigma-Aldrich). All slides used for the above mentioned stainings were counterstained with 49,6-diamino-2-phenylindole dihydrochloride (DAPI, Sigma) and mounted in Mowiol 4-88 (Sigma). For sex steroid receptor detection the following antibodies were used: rabbit polyclonal anti-AR (N-20) 1:1000 (sc-816, Santa Cruz Biotechnology), mouse monoclonal anti-estrogen receptor a 1:200 (M7040, Dako) and mouse monoclonal anti-estrogen receptor b1 1:100 (M7292, Dako). The antibody-antigen complexes were visualized by using anti-rabbit and anti-mouse HRP labeled polymer (EnVision, Dako) combined with DAB+ chromogen system (Dako).

RNA Isolation and qRT-PCR
For analyzes of gene expression, 1 day -42 day-old WT mice (mixed background, C57Bl/6N and SV129) and 21 day -2 month-old control and Dicer1 fl/fl ; Defb41 iCre/wt mice were used. The epididymides, efferent ducts and testes were dissected out and weighed. The epididymis of control and Dicer1 fl/fl ; Defb41 iCre/wt mice was cut into three segments; IS and CAP, COR and CAU. All tissues were frozen in liquid nitrogen and stored at 280uC. Total RNA was isolated from the tissues using Tri Reagent according to the manufacturer's instructions (Molecular Research Center, Inc.). For RT-PCR, 1 mg of total RNA was treated with deoxyribonuclease I (DNaseI, Amplification Grade, Invitrogen) and reverse-transcribed by using the DyNAmo cDNA synthesis kit (Thermo Scientific). The cDNA was diluted 1:50-1:100 for quantitative PCR. Quantitative PCR was performed using the DyNAmo Flash SYBR Green qPCR Kit (Thermo Scientific). All samples were run in triplicate reactions and standards in duplicate reactions. L19 and Ppia were used as endogenous controls to equalize for the amounts of RNA in the epididymis and testis, respectively. Primer sequences and qRT-PCR conditions for analyzing the expression of Dicer1, Ros1, Bmyc, Lcn8, Cst8, Gpr64, Ar, Esr1, Esr2, Gpx5, Lcn5, Crisp1 and Fgfr 1-4 are described in Table S1.

Cell Proliferation and Apoptosis
To detect cell proliferation, epididymis sections were stained with rat monoclonal anti-Ki67 1:500 (M7249, Dako). The antibody-antigen complexes were visualized by using 1:200 rabbit polyclonal anti-rat (Dako) combined with anti-rabbit HRP-labeled polymer and DAB+ chromogen system. Terminal Uridine Deoxynucleotidyl Nick End Labeling (TUNEL) was used to detect apoptotic cells. Labeling was performed by using 0.8 U/ml TdT (Terminal transferase, recombinant, Roche) and Biotin-16-dUTP (Roche), 1 h at 37uC. The reaction was visualized with Biotin coupled with ExtrAvidine-Peroxidase, 1:200 (Sigma-Aldrich) combined with DAB+ chromogen system. Proliferating and apoptotic cells were counted from ten tubular cross sections of 5 control and 5-8 Dicer1 fl/fl ; Defb41 iCre/wt mouse IS and CAP, respectively. The number of proliferative and apoptotic cells was then divided by the circumference (mm) of the tubular cross sections.

Statistical Analyses
For statistical analyses of organ weights, epithelial cell heights, animal fertility, sperm morphology, cell proliferation/apoptosis and qRT-PCR, the GraphPad Prism 5 software (GraphPad Software, Inc., La Jolla, CA) was used. Unpaired t-test was used to determine statistical significances, assigning P#0.05 as the limit of statistical significance.