Sertoli cell-specific ablation of miR-17-92 cluster significantly alters whole testis transcriptome without apparent phenotypic effects

MicroRNAs are frequently organized into polycistronic clusters whose transcription is controlled by a single promoter. The miR-17-92 cluster is expressed in most embryonic and postnatal organs. It is a potent oncogene associated to several types of cancer and it is involved in several important developmental processes. In the testis, expression of the miR-17-92 cluster in the germ cells is necessary to maintain normal spermatogenesis. This cluster is also expressed in Sertoli cells (the somatic cells of the seminiferous tubules), which require miRNAs for correct cell development and survival. To study the possible role of miR-17-92 in Sertoli cell development and function and, in order to overcome the postnatal lethality of miR-17-92-/ mice, we conditionally deleted it in embryonic Sertoli cells shortly after the sex determination stage using an Amh-Cre allele. Mutant mice developed apparently normal testes and were fertile, but their testis transcriptomes contained hundreds of moderately deregulated genes, indicating that testis homeostasis is tightly controlled in mammals and that miR-17-92 expression in Sertoli cells contribute to maintain normal gene expression levels, but is unnecessary for testis development and function. Our results show that significant deregulation of hundreds of genes might have no functional consequences.


Introduction
Sertoli cells (SCs) are the epithelial supporting cells within the seminiferous tubules of mammalian adult testes. Their main function is to sustain spermatogenesis by providing structural support, nursing and regulating the function of germ cells (GCs) via signaling molecules. They also produce the seminiferous fluid and form the blood-testis barrier (BTB), an inter-SC a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 specialized junctional complex formed by cell adhesion molecules that defines a basal and an adluminal compartment into the seminiferous epithelium and serves as an immunological barrier that creates a unique micro-environment for GC development. In addition, SCs control spermatogonial self-renewal and survival, as well as phagocytose apoptotic spermatocytes and cell debris derived from spermiogenesis [1]. SCs play an essential role in testis determination and differentiation. The expression of the testis-determining gene SRY in embryonic SC precursors leads to SC differentiation and commits them to surround GCs and to form testis cords, the anlagen of the seminiferous tubules. During embryonic development, SC products are necessary for preventing GCs from meiosis entry, as well as for Leydig and peritubular myoid cell differentiation, and regression of the Müllerian ducts, the anlagens of female secondary sexual organs [2,3].
In the testis, several studies have shown that the members of the miR-17-92 cluster are expressed in GCs and conditional inactivation of the cluster in either embryonic GCs or the whole adult testis resulted in spermatogenic defects [12,13]. The SC-specific deletion of the RNase III enzyme Dicer in mice revealed that miRNAs also play an important role in the development and survival of SCs [14,15]. Several studies have provided evidence that the members of the miR-17-92 cluster are expressed in SCs. All the members of the cluster were cloned from purified P6 SCs [14], in situ hybridization with LNA (Locked Nucleic Acid) on adult testes showed miR-17 and miR-20a expression in SCs [12], and ulterior analysis of the small RNA transcriptome of SCs purified from mice at postnatal day 6 revealed high levels of expression for miR-19a and miR-19b, intermediated levels for miR-17 and miR-20a and low levels for miR-18a and miR-92a [16].
Overall, available data show that miRNAs are required for normal development of SCs and that miR-17-92 expression in GCs plays a role in spermatogenesis maintenance, but nothing is currently known on the role of this miRNAs cluster in postnatal and adult SCs. In order to overcome the postnatal lethality of miR-17-92 -/mice and to uncover novel functions for miR-17-92 during SC development, we conditionally deleted the miR-17-92 cluster in embryonic SCs shortly after the sex determination stage using an Amh-Cre allele.

Mice
To induce SC-specific deletion of miR-17-92, we mated mice with a floxed allele of miR-17-92 [5] to mice harbouring the SC-expressed transgene Amh-Cre [17], obtained from the Jackson Laboratory (Bar Harbor, ME, USA; Stock No: 007915 and 008458, respectively). They were bred in our animal house facilities and the resulting double heterozygous offspring carrying the Cre allele was backcrossed to homozygous miR-17-92 flox mice to obtain Amh-Cre;miR17-92 flox/flox mice. To evaluate Cre-activity, Amh-Cre and Amh-Cre;miR17-92 flox/flox mice were bred with the Cre-reporter mouse line R26R-LacZ [18] (Jackson Laboratory, Bar Harbor, ME, USA; Stock No: 003309). Primers and PCR conditions for miR-17-92 flox and for R26R-EYFP and Cre have been previously described [5,19]. Mice with genotype miR-17-92 flox/flox lacking the Cre allele were used as controls. All animal experiments in this study were approved by the University of Granada Ethics Committee for Animal Experimentation (exp. No: 2011-341), and were performed in accordance with the relevant guidelines and regulations dictated by this Committee.

Analysis of apoptosis
Apoptotic cells were labelled using the Roche TUNEL kit (Fluorescent In Situ Cell Death Detection Kit) according to the manufacturer's instructions.

Analysis of BTB permeability
An in vivo test was performed to study the permeability of the BTB in the testes of control and mutant mice. We used a biotin-labelled tracer compound (EZ-Link Sulfo-NHS-LC-Biotin tracer, Thermo Scientific) as described [20].

Quantitative RT-PCR
Total RNA from E15.5 testes was isolated using the RNeasy Micro kit (Qiagen) including DNase I treatment. After checking RNA integrity, synthesis of cDNA was performed using Superscript II reverse transcriptase (Invitrogen). PCR was performed using the NZY qPCR Green Master Mix (Nzytech). Efficiency of the primers was tested by a standard curve assay, and only primers with amplification efficiency close to 100% were used. Each sample was measured in triplicates in two independent experiments. Ct values of samples were normalized to the corresponding Ct values of Gapdh, and relative expression levels were calculated by the ΔΔCt method [21]. Primer pairs used were 5'-GGTGGGGATTGTGACCAG-3' and 5'-CGAGCAAACACGAAAATGAA-3' for miR-17-92hg; and 5'-GATGACATCAAGAAGGTGGTG-3' and 5'-TCATACCAGGAAATGAGCTTG-3' for Gapdh.

Transcriptome preparation
Total RNA was isolated from three Amh-Cre;miR17-92 flox/flox and three control (miR17-92 flox/flox ) testes at postnatal day 60 (P60) using the Qiagen RNeasy Midi kit following the manufacturer's instructions. Subsequently, the six samples were sent to Macrogen Inc. (Seoul, South Korea). After passing a quality check, libraries were prepared using TrueSeq RNA Sample Prep Kit V2 and they were paired-end sequenced separately in an Illumina HiSeq 4000 platform, generating between 120 and 135 million reads per sample.

Bioinformatics
The RNA-seq reads were mapped to the UCSC mm10 mouse genome using the STAR RNAseq aligner (version 2.5) [22], and subsequently they were counted with the featureCounts function from the R subread package [23]. Only genes with 5 or more RPKM (reads per kilobase per million) in at least two of the samples were considered to be expressed and were used for further analysis. Analysis of differential gene expression was performed with edgeR [24].
Gene Ontology analysis was performed for differentially expressed genes (FDR < 0.01) using the Database for Annotation, Visualization and Integrated Discovery (DAVID) [25]. All the expressed genes were used as background.
To generate a circos plot, raw data from the previously transcriptomes reported by Soumillon et al. [26] where downloaded from the Gene Expression Omnibus (Acc. GSE43717) and processed with the same pipeline as our own data. We performed pairwise comparisons of the transcriptomes of the different cell types with the edgeR package [24,27] from bioconductor and used the glmQLFTest() function to Conduct gene wise statistical tests. We then constructed lists of genes for each cell type. Genes more expressed in a cell type than in any of the others were included in the list of this particular cell type. After the genes were distributed among the different cell types we calculated the intersection of each gene set with the differentially expressed genes in our SC-miR-17-92 KO testis. We then constructed tables as input data for the biocircos.js library [28] including the distribution of all differentially expressed genes in our SC-miR-17-92 KO testis among the different cell types, their relative expression levels in log 2 FC and the predicted target sites for the four miR-17-92 seed families obtained from Tar-getScan (version 7.1, http://www.targetscan.org). The tables where exported from R in json format and included in javascript scripts that used the biocircos classes to draw the circos plot.
The transcriptome datasets generated during the current study are available in the Arrayexpress repository E-MTAB-5914. All bioinformatic data generated and analysed during this study are included in this published article (and its Supporting Information files).
Next, we studied the expression of several somatic and GC markers by immunofluorescence. SOX9 and WT1, two transcription factors expressed in adult SCs, are necessary for testis function and maintenance [19,29]. Both control and SC-miR-17-92 mutant testes exhibited the same expression pattern for the two proteins in SCs (Fig 3A). PCNA is expressed in mitotic spermatogonia as well as in zygotene and early pachytene spermatocytes [30] and the meiotic recombination protein, DMC1, is expressed in early spermatocytes [31] (leptotene-pachytene). No difference between mutant and control testes was observed in the expression patterns of these two proteins at both P60 ( Fig 3B) and P365 stages (S1 Fig). Also, Laminin, a component of the basement membrane [32], and α-smooth muscle actin (ACTA2), a marker of peritubular myoid cells and arterial muscle fibers, showed similar expression in the two study groups (Fig 3C). Likewise, the expression of P450scc, a steroidogenic cytochrome expressed in Leydig cells, was also similar in both SC-miR-17-92 KO and control testes (Fig 3D).
In addition, Claudin 11 (Cldn11), a major component of the tight junctions forming the BTB, did not show a different pattern of expression between the control and mutant condition miR17-92 in Sertoli cells is not required for fertility (Fig 4Aa and 4Ad). Consistently, injection of a biotin tracer showed that the BTB in mutant males was impermeable, as observed in control testes (Fig 4Ab and 4Ae). Finally, TUNEL assay showed no increase in the frequency of apoptotic cells in SC-miR-17-92 KO testes compared to control ones (n = 3; control: 14.65 ± 10.07 apoptotic cells/mm 2 ; mutant: 13.96 ± 8.87  miR17-92 in Sertoli cells is not required for fertility apoptotic cells/mm 2 ; p = 0.23, two tail t-test ; Fig 4Ba-4Bc). Altogether, these results show that normal testis function is preserved after SC-specific ablation of miR-17-92.

The testicular transcriptome is altered in Sertoli cell-miR17-92 KO male mice
We performed RNA-seq experiments on the adult testes of three SC-miR-17-92 KO and three control males at postnatal day 60. The differences between the expression profiles of these samples were examined using a hierarchical clustering analysis (Fig 5A). Replicate samples from the same condition clustered together, indicating that the whole testis transcriptome is significantly and consistently altered when miR-17-92 is deleted in SCs. To confirm this, we checked the presence of differentially expressed (DE) genes when comparing the two experimental conditions, and found 809 genes deregulated at P adjusted (FDR) < 0.01 (418 upregulated and 391 downregulated; see Table A in S1 File). Smear plot of the differential expression test showed that the vast majority of the DE genes were deregulated by less than 2 times, indicating that, despite a reproducible deregulation of many genes, the magnitude of the change was generally moderate (Fig 5B, see Table A in S1 File). Gene ontology (GO) analysis of DE genes revealed a significant enrichment (P adj. < 0.05) in terms related to normal testicular functions including spermatogenesis, meiotic cell cycle, chromosome segregation, cell-cell adhesion, cell-projection, and DNA-repair, among others ( Fig 5C, see Table B in S1 File). Interestingly, although we deleted miR-17-92 specifically in SCs, several GO terms referred to processes normally occurring in GCs (e.g. spermatid development and meiotic cell cycle), suggesting that spermatogenesis was also affected in the mutant mice. This observation is consistent with the well known notion that alteration of SC functionality is frequently associated to a failure in GC maturation [19,33,34]. Thus, we used the recently published transcriptomes of the five major testis cell populations (Sertoli, spermatogonia, spermatocyte, spermatid and spermatozoa) involved in spermatogenesis [26] to develop an algorithm to assign every differentially expressed genes in the SC-miR-17-92 KO testes to particular cell types. First, we searched for differentially expressed genes performing pairwise comparisons of the transcriptomes of the cell types from the Soumillon's et al. [26] study. Then we assigned the differentially expressed genes of our SC-miR-17-92 KO to the specific cell type in which they showed the highest expression level in the Soumillon's et al. [26] transcriptomes. We tested the genes assigned to particular cell types by looking for known specific gene markers of each cell type (see Tables C-G in S1 File). Consistently, major cell specific markers were assigned to the correct cell type: Sertoli (Wt1, Cldn11, Cdh2, Sox8, Gata1, Gata4), spermatogonia (Gdnf, Cd9, Nanos2, Nanos3, Thy1, Bcl6, Gfra1, Epcam) spermatocyte (Sycp1, Sycp2, Sycp3, Rad51, Spo11, Syce1, Tex12, Eme1), spermatids (Prm1, Rsbn1, Rsb1nl, Tnp1, Tnp2, Theg, Nme8, Txnd8), indicating that our approach clearly reflects cellspecific expression patterns.
The circos plot in Fig 6A and Tables H-L in S1 File show the distribution of the DE genes in our SC-miR-17-92 KO testes among the different cell types. It is noteworthy that both overand under-expressed genes are not randomly distributed among the different cell types but most of the over-expressed genes belong to spermatids while most of the under-expressed genes belong to all other cell types, including SCs. In this cell type, we found 55 deregulated genes, 34 being downregulated and 21 upregulated (see Table H in S1 File). GO analysis of these genes revealed an overrepresentation in categories related to cell-cell adhesion and endoplasmatic reticulum (Fig 6B, see Table M in S1 File). Likewise, GO analysis of the deregulated genes in the other cell types showed enrichment in terms related to 1) cell junctions and cell miR17-92 in Sertoli cells is not required for fertility cycle for spermatids, 2) meiotic cell cycle, covalent modification of chromatin and DNA repair for spermatocytes, and 3) cilium morphogenesis and spermatid development for spermatids (Fig 6C-6E, see Tables N-Q in S1 File), all of them referring to processes normally occurring in these cell types during spermatogenesis and spermiogenesis.
The individual miRNAs included in the miR-17-92 cluster can be grouped into 4 families based on their seed sequences, which are the most important regions for target selection.

Fig 5. Transcriptome analysis of SC-miR-17-92 KO testes. A)
Hierarchical clustering analysis showed that the replicate testis samples of both SC-miR-17-92 KO and controls grouped together. B) Smear plot of the differential expression test showed that the vast majority of the deregulated genes changed their expression levels by less than one log2 fold change (log2FC). C) GO analysis of SC-miR-17-92 KO deregulated genes revealed a significant enrichment (P adj. < 0.05, blue dashed line) in terms associated to normal testicular functions. Therefore, we searched for target genes that were deregulated in our mutant testis transcriptome, and found a number of target genes for each seed family (see Tables R-U in S1 File). However, they were not enriched in upregulated genes and were not preferentially located in the SC-specific list of deregulated genes (Fig 6A). Genes assigned to the different cell types appears to belong to GO terms normally related to the cell type to which they have been assigned according to the algorithm described. P adj. < 0.05, blue dashed line.
https://doi.org/10.1371/journal.pone.0197685.g006 miR17-92 in Sertoli cells is not required for fertility Discussion SC-specific ablation of Dicer resulted in abnormal testis development and spermatogenic failure, evidencing the importance of miRNAs in the differentiation and function of this cell type [14,15]. Several authors have reported SC-specific expression of a number of miRNAs [14,16] but little is known about the role of these translational regulators in SCs. Prompted by several studies showing the expression of members of the mir-17-92 cluster in SCs [12,14,16,35], and knowing their relevant roles in other organs and disease conditions (see the Introduction section), we decided to induce the conditional ablation of this miRNA cluster in embryonic SCs shortly after testis differentiation. Although the SC-miR-17-92 KO adult testes showed apparently normal development and function, RNA-seq analyses evidenced consistent deregulation of many testicular genes in comparison to control testes, indicating that expression of the mir-17-92 cluster in SCs is necessary to maintain the gene expression levels of the whole testis within normal values. Thus, the question arises as to why such an alteration of the transcriptome homeostasis does not result in an observable phenotype. In this regard, it should be noted that, despite the high number of deregulated genes in SC-miR17-92 KO testes, the magnitude of the changes in their expression levels was generally modest (less than two times in most cases). Such a scenario, implying many deregulated genes showing just moderate changes in their expression levels, has been previously described for miR-17-92, and it is consistent with the proposed notion that its members act as fine-tuners of large gene networks rather than as major genetic switches of specific pathways [36]. The phenotypic consequences of transcriptome alterations induced by the absence of miR-17-92 may be variable and difficult to predict. As an example, despite the fact that the number of deregulated genes in the embryonic tail buds of mice harboring a homozygous deletion for the miR-17 seed family (~500) was lower than the observed in the same tissue of homozygous mutants for the miR-19 seed family (~700), skeletal malformations were observed in the former mice (defects in axial patterning regulation), but not in the latter ones. Hence, it appears that the severity of the phenotype derived from the absence the miR-17-92 cluster depends on the spatial and temporal cellular context. Most importantly, our result evidence that testis homeostasis must be strictly controlled as even when hundreds of genes were deregulated in all testicular cell types of miR-17-92 KO testes, these showed no evident pathologic phenotype, conserving normal structure and function. Probably many other regulatory factors contribute to maintain testis homeostasis, as the simple ablation of miR-17-92 is not enough to disturb it significantly. However, it is likely that mutant testes become more sensitive to stressful situations. Consistent with this idea, it was recently reported that hepatocyte-specific miR-17-92-deficient mice were fertile and apparently healthy, but their capacity of liver regeneration after tissue injury had resulted significantly reduced [37]. Hence, although no obvious testicular phenotype is observed, the transcriptome alteration we detected in our miR-17-92 mutants might cause a suboptimal testis functional status.
Genes predicted to be miRNA targets become preferentially downregulated in cultured cells shortly after miRNA transfection [38]. Accordingly, transcriptome analysis of the embryonic tail buds of miR-17-92-deleted embryos at E9.5 (before the appearance of any developmental abnormalities) showed preferential upregulation of the predicted targets [36]. These observations contrast with our results, as we did not find a preferential upregulation of the predicted miR-17-92 target genes in the transcriptomes of mutant testes. In our mutant mice, miR-17-92 was depleted in SCs at the embryonic stage E14, but the testis transcriptomes were analysed nine weeks later, at the adult stage P60, as we wanted to see the effects in postnatal and adult testes. Hence, our transcriptomes do not reflect the primary effects of such a miRNA depletion, but probably many others induced subsequently by a chain reaction of deregulated miR17-92 in Sertoli cells is not required for fertility genes that proceeded with time. The fact that we did not observe a severe or even moderate phenotype in the mutant testes, made it difficult to select a more convenient time point to analyze the transcriptome. Another limitation of our study is that we could only analyse the transcriptome of the whole testis, where spermatocytes and especially spermatids contribute very significantly to the total cell count and testis volume [26]. Hence, the search of miRNA targets using our transcriptome data is probably futile or even misleading.
GO analysis of deregulated genes in SC-miR-17-92 KO testes showed a significant enrichment in biological processes associated with some GC-specific functions (e.g. spermatogenesis, meiosis, and spermatid development). As this alteration may not be a direct consequence of the misregulation of miR17-92 target genes and their associated gene networks in SCs, we searched for SC-specific deregulated genes in SC-miR-17-92 KO testes. GO analysis revealed that this SC-specific gene set was enriched in terms associated with cell adhesion, actin binding, and cytoskeleton. This is consistent with the fact that differentiating GCs progressively migrate across the seminiferous epithelium during spermatogenesis until they reach the tubular lumen. Throughout this migration, GCs remain attached to SCs via cell adhesion molecules, which prevent the sloughing of immature GCs from the seminiferous epithelium, a phenomenon that may lead to infertility [20,39]. Moreover, specialized cell adhering structures established between adjacent SCs form the BTB, that isolates post-meiotic GCs from the systemic circulation [40]. In addition, SCs have a complex, well organized and functionally active cytoskeleton that requires abundant filament and microtubule proteins to maintain their structural integrity and hold neighbor germ cells [41]. For instance, Cldn11, which encodes Claudin 11, a principal component of the tight junctions forming the BTB, is one of the genes shown to be deregulated in our mutant mice. Mutation of this gene results in spermatogenic failure and male sterility [42]. Other affected genes in SC-miR-17-92 KO testes were Itgb1, that encodes β1-integrin, an essential adhesion receptor necessary for homing the spermatogonial stem cells into the GC niche [43], Tjp1 (or ZO-1) that encodes the tight junction protein 1, and Jup that encodes the junction plakoglobin. The two latter are essential adhesion proteins in SC-SC and SC-GC junctions [40] (see Tables M and N in S1 File). In the light of all these evidences, it seems plausible that SC-specific ablation of the miR-17-92 cluster leads to the deregulation of the cell adhesion molecules present in the cell contacts by which the SC-GC cross-talk is established, thus resulting in an alteration of their transcriptomes.
In the mouse genome, two paralogues of the miR-17-92 cluster exist, i.e. miR-106b-25 and miR-106a-363 and functional redundancy between the two first clusters has been revealed at the embryonic stage. On the other hand, expression of the third one has not been detected in most of the tissues analyzed to date including adult testes [5]. Individual members of the miR-106b-25 cluster are known to be expressed in SCs [16], so they might compensate for the absence of the miR17-92 cluster in the SCs of our mutant mice. To check this possibility, the testicular phenotype of miR-17-92; miR-106b-25 double mutants must be analyzed, a work currently under way in our laboratory.
In summary, we generated conditional mutants in which SC-specific depletion of miR-17-92 was induced during the embryonic development of the testis. Despite the fact that no phenotypic alteration was apparent in mutant adult testes, RNA-seq analysis revealed a consistent and reproducible alteration of their transcriptome, hundreds of genes being moderately deregulated. This indicates that testis homeostasis is tightly controlled in mammals and that the expression in SCs of the miRNAs included in the miR-17-92 cluster contribute to maintain testis transcriptome within normal values, but are not essential to warrant normal testis development and function. While it is likely correct to consider that transcriptome profiles identify gene expression changes that do not necessarily link to the phenotype, this is usually not the case, as most studies report phenotypic defects in mutant animals. However, the reason is probably related to the rationale of most studies, in which profiling assays are performed once, and maybe because, a phenotype is observed, selecting the critical time point for "worst defect" or "putative defect of interest" in order to identify the candidate genes. Hence, our results evidence that significant deregulation of hundreds of genes, a situation frequently observed in many studies, not necessarily will have functionally relevant consequences.