The experimental procedures used in this study were approved by the Institutional Animal Use and Care Committee at Hokkaido University and by the Animal Care and Use Committee of the National Research Institute for Child Health and Development. C57Bl/6Ncl mice (CLEA Japan Inc., Tokyo, Japan) and knockout (KO) mice (ICR and mixed background of C57Bl/6 and DBA2, Sankyo Labo Service Corporation) were maintained on 14 hr light/10 hr dark cycles at 25°C and given food and water ad libitum. In the analysis of Start-KO mice, we used male mice at 64-83 dpp as 2.5-month-old adult mice.

Total RNAs were extracted with ISOGEN and ISOGEN II (Nippon Gene, Tokyo, Japan) for tissues and cultured cells, respectively, according to the manufacturer’s instructions. After treatment with TurboDNase (Thermo Fisher Scientific, Rockford, IL, USA), the RNAs were reverse-transcribed into cDNAs with the oligo(dT) primer or random hexamer using Superscript III (Invitrogen, Carlsbad, CA), according to the manufacturer’s instructions. A random hexamer was used for RT-PCR to check tissue distribution, and oligo(dT) was used for other experiments. PCR was performed using ExTaq polymerase (Takara, Otsu, Japan) or KOD Fx Neo (Toyobo, Tokyo, Japan). For strand-specific RT, “Start full length” reverse and “Start full length” forward primers were used to synthesize Start sense and antisense cDNAs, respectively, and a reaction temperature of 55°C was used to avoid non-specific binding of primers. Primer sequences are shown in Table 1 . We repeated experiments with three or more biological replicates to confirm the reproducibility and showed a representative result.

For 5’RACE, cDNA was generated using a gene-specific primer 1 (GSP1) and total RNAs from mouse male germ cells. After purification of cDNA with a QIAquick PCR Purification Kit (Qiagen, Hilden, Germany), oligodeoxycytidine was added by terminal deoxynucleotidyl transferase (Takara). The first-round PCR was performed with an abridged anchor primer (AAP) and GSP2. For the second nested amplification, GSP3 and an abridged universal amplification primer (AUAP) were used.

For 3’RACE, cDNA was prepared by reverse transcription with oligo(dT) connected to an adaptor sequence (AP). The first-round PCR was conducted by using primers, AP and GSP4. The second nested amplification was carried out with the same adaptor primer and GSP5.

All of the PCR products were subcloned into the EcoRV site of a pBluescript II KS(+) vector (Stratagene La Jolla, CA, USA) by the TA-cloning method, and 10 subclones for each sample were sequenced. All of the primer sequences are listed in Table 1 .

cDNAs were synthesized as described above using the oligo(dT) primer. PCR was performed by using SYBR Green PCR Master Mix for protein-coding genes (Thermo Fisher Scientific) and KOD SYBR qPCR Mix for Start (Toyobo). In both cases, the 7300 real-time PCR system (Applied Biosystems, Foster City, CA, USA) was used as previously described (49). The relative expression levels were normalized to endogenous Gapdh or Aip mRNA and calculated by the comparative Ct method (50). Primer sequences are shown in Table 1 and Supplemental Table 1 .

Native germ cells were isolated from two adult mouse testes, and their purity was checked by marker gene expression as previously described (46, 49). Sertoli cells were obtained from 11-day-old testes for primary culture as previously described (37, 51). The Sertoli/Leydig cell fraction was obtained from 6-month-old testes by using a discontinuous Percoll gradient (20%, 37%, and 53%) as previously described (44).

Subcellular fractions were obtained by using NP-40 lysis buffer (10 mM Tris HCl [pH 7.5], 10 mM NaCl, 3 mM MgCl₂, 0.5% NP 40) containing 1× proteinase inhibitor cocktail (Roche Molecular Biochemicals, Mannheim, Germany) as previously described (44, 52). The nuclear fraction was collected as a pellet after centrifugation, and the supernatant was used as the cytosolic fraction.

The RNA probes to detect Start were prepared as follows. The full length of Start was amplified by RT-PCR with mouse testis cDNA and a primer pair of Start full length ( Table 1 ) using KOD Fx Neo and was subcloned into the EcoRV site of a pBluescript II KS+ vector. After the sequence had been confirmed, the full length was cut out by EcoRI and HindIII digestion, blunted by T4 DNA polymerase (Takara), and inserted into the blunted PstI site of a pGEM-3Zf(+) vector. The resulting plasmid was linearized with ApaI and NotI digestion for synthesizing sense and antisense RNA probes, respectively. Digoxigenin (DIG)-labeled RNA probes were synthesized using a DIG RNA labeling kit (Roche Molecular Biochemicals) with SP6 RNA polymerase for a sense probe and T7 RNA polymerase for an antisense probe.

In situ hybridization with the Tyramide Signal Amplification Plus system (PerkinElmer, Waltham, MA, USA) was performed according to the procedure reported previously (44, 53). Briefly, adult mouse testes were fixed for 3 h at 4°C with 4% paraformaldehyde in phosphate buffered saline (PBS). After washing with PBS, each testis was dissected into two pieces with a razor blade under a dissecting microscope. The two pieces of testis were dehydrated, embedded in paraffin, and cut into 7-μm-thick sections. The sections were hybridized with 2 ng/μl of the DIG-labeled RNA probe overnight at 45°C. After washing, the sections were incubated with an anti-DIG-horseradish peroxidase antibody (1:500 dilution; Cat# 1207733, Roche Molecular Biochemicals) for 30 min at room temperature. The reaction with tyramide-Cy3 (1:50 dilution in 1X Plus Amplification Diluent (PerkinElmer), followed by 1:100 dilution with distilled water) was performed at room temperature for 20 min. To detect nuclei, the sections were incubated with 10 μg/ml Hoechst 33258 at room temperature for 10 min. The samples were mounted with a Fluoro-KEEPER Antifade Reagent (Nacalai Tesque, Kyoto, Japan) and observed under an LSM 5 LIVE confocal microscope (Carl Zeiss, Oberkochen, Germany).

sgRNA sequences were designed by CRISPRdirect (https://crispr.dbcls.jp/) (54). sgRNAs were synthesized and purified by using a CUGA7 gRNA Synthesis Kit (Nippon Gene) according to the manufacturer’s instructions. The selected sequences were Start sgRNA-UP (AGTACTTCCCTAGGTAAATGAGG) and Start sgRNA-DOWN (GTAACATCCTGTCATATTGCGGG), in which proto-spacer adjacent motif (PAM) sequences are underlined. Oligo DNA sequences used for sgRNA synthesis are listed in Table 1 .

For injection, oocytes were collected from F1 hybrid (C57BL/6 x DBA/2) BDF1 female mice that were superovulated by standard procedures (55) and fertilized in vitro with sperms from male mice of the same genetic background. CAS9 protein (100 ng/µl; Nippon Gene) and sgRNAs (250 ng/µl each) were microinjected into the cytoplasm of fertilized oocytes. The oocytes were cultured in KSOM medium overnight, and embryos at the two-cell stage were transferred to pseudo-pregnant ICR female mice on the next day.

Tail tips were collected from mice 3-4 weeks after birth and treated with Proteinase K solution (50 mM Tris-HCl, pH 8.0, 100 mM EDTA, 0.5% SDS, and 100 μg/ml proteinase K) at 50°C for 24 hr. After extraction with phenol/chloroform/isoamyl alcohol, genomic DNA was precipitated with isopropanol. Purified DNAs were examined by PCR using a pair of primers listed in Table 1 .

Wild-type and Start-KO mice were euthanized, and their body weights were measured. Then testes were isolated and their weights were measured. The testes were fixed with Bouin’s solution at room temperature for 2 days, dehydrated, embedded in paraffin, and sectioned at 7 µm in thickness. The sections were stained with hematoxylin and eosin and observed under a light microscope.

One cauda epididymis was collected from wild-type and Start-KO adult mice, soaked in 0.7 ml of HTF medium (Nippon Medical and Chemical Instruments, Osaka, Japan), and torn into pieces by a sterile 26G needle. After incubation at 37°C with 5% CO₂ in air for 2 hr, whole medium was transferred into a new tube and mixed with 0.7 ml of 10% neutral buffered formalin (Fujifilm Wako Pure Chemical Corporation, Osaka, Japan) for fixation. The number of sperms was counted manually by using a hemocytometer.

RNA-seq analysis was performed as previously described (36, 56). In brief, total RNAs were isolated from a wild-type testis and Start-KO testis (line #2) of mice at the age of 2.5 months using ISOGEN according to the manufacturer’s protocol. The RNAs were treated with TurboDNase to completely remove genome DNA, and the quality of the RNA samples was evaluated by electrophoresis and BioAnalyzer (Agilent Technologies, Palo Alto, CA, USA) with RNA6000 nano tip. cDNA synthesis was performed with 500 ng RNA from each sample by using a TruSeq Stranded mRNA Sample preparation kit (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol. The resulting cDNA library was validated using BioAnalyzer with DNA1000 Chip and quantified using NEBNext Library Quant Kit for Illumina (NEB, Ipswich, MA, USA). A total of 101-cycles single-end sequencing was performed using HiSeq1500 (Illumina) in the rapid mode. After conversion to fastq format, reads were mapped to the mouse reference genome, mm10, with Hisat2 software (v2.1.0) (57). The expression level of each gene was calculated as gene-specific Transcript per Million mapped reads (TPM) using Stringtie (v1.3.3b) (57) with the optional –conservative setting enabled. Total reads, mapping rates, and accession numbers are summarized in Table 2 .

A blood sample from each mouse was collected into a microtube and kept on ice for 1-2 hr until it became clotted. The sample was then centrifuged at 1,000g for 10 min at 4°C. The supernatant was collected as serum and quickly frozen by liquified nitrogen. To measure intratesticular testosterone, tunica albuginea of each adult testis was removed, and the tissue was homogenized in 1 ml of ice-cold PBS. The crude sample was centrifuged at 2,500 rpm for 5 min at 4°C, and the supernatant was collected into a new tube. This centrifugation step was performed twice. The sample was mixed with an equal volume of chloroform, vortexed for 1 min, and centrifuged at 15,000 rpm for 5 min at room temperature. After the centrifugation, the organic phase was collected into a new tube whose cap was removed, and the tube was centrifuged until chloroform was evaporated completely. The precipitate was dissolved in 50 μl methanol and diluted with 350 μl enzyme-linked immunosorbent assay (ELISA) buffer (Cayman Chemical, Ann Arbor, MI, USA). The serum from adult mice was diluted 10 or 20 times with milli-Q water for ELISA analysis, and that from 8-days postpartum (dpp) mice was used after 3.3 times dilution. The intratesticular sample was diluted 100 times with ELISA buffer. Intratesticular and serum testosterone concentration were measured by using a Cayman Testosterone EIA kit according to the manufacturer’s instructions.

Paraffin sections of testes were prepared as above at 7 µm in thickness, and de-paraffinized and hydrophilized with xylene, ethanol, and Milli-Q water. The sections were boiled for 45 min with 10 mM sodium citrate-HCl (pH 6.0) and treated with 3% H₂O₂-containing PBS. After the H₂O₂ treatment, the blocking was performed by soaking the sections with 1% bovine serum albumin in 10 mM Tris-HCl (pH 7.5), 0.88% NaCl, and 0.1% Tween20 for 1 hr at room temperature. An anti-HSD3B antibody (no dilution; Cat# sc-515120, Santa Cruz Biotechnology, Dallas, TX) was reacted with the section for 1 hr at room temperature. After washing with PBS, the sections were subjected to the 2nd antibody reaction for 1 hr at room temperature with a HRP-Labelled Polymer Anti-Mouse (1:4 dilution, Dako). After washing with PBS, the sections were stained with ImmPACT Vector Red (Vector Laboratories, Burlingame, CA) according to the manufacturer’s instructions.

Serum samples were collected as described above. Serum LH concentration was measured by using a Luteinizing Hormone ELISA kit (CUSABIO, Wuhan, China) according to the manufacturer’s instructions.

The full length of Start was amplified and inserted into a pBluescript II KS+ vector as described above. The Start sequence was obtained by EcoRI and HindIII digestion from this vector, and the fragment was blunted by T4 DNA polymerase and inserted into the EcoRV site of a pCAG-Hyg vector (Fujifilm Wako Pure Chemicals). Constructs with insertion of the Start sequence in both directions were obtained.

For each of mouse Leydig cell lines, TM3 and MA-10, the hygromycin concentration for selection was preliminarily determined. TM3 cells were cultured in Dulbecco Modified Eagle Medium (DMEM)/F12 containing 2.5% fetal bovine serum, 5% horse serum, 100 U/ml penicillin, 100 µg/ml streptomycin, and 292 µg/ml L-glutamine (Invitrogen, Carlsbad, CA, USA). After coating a 3.5-cm dish with 0.1% gelatin before use, 2.0 × 10⁵ cells were spread in the dish one day before transfection. The next day, 6.0 µl PEI Max (Polyscience, Warrington, PA, USA) was mixed with 200 µl Opti-MEM (Thermo Fisher Scientific) and 2 μg DNA in that order. The mixture was added to the dish after 20-min incubation. Selection was done with the medium containing 250 µg/ml of hygromycin for 8 days.

MA-10 cells were cultured in DMEM/F12 containing 15% horse serum, 100 U/ml penicillin, 100 µg/ml streptomycin, and 292 µg/ml L-glutamine on a dish coated with 0.1% gelatin. One-tenth of confluent cells in a 6-cm dish was spread in a 3.5-cm dish one day before transfection. The next day, 24.0 µl PEI Max was mixed with 200 µl Opti-MEM and 8 μg DNA in that order. The mixture was added to the dish after 20-min incubation. Selection was done with the medium containing 100 µg/ml of hygromycin for 8 days.

Results are presented as the mean value ± standard deviation (S.D.) of at least three independent experiments. The data were statistically analyzed by Student’s t-test, Welch’s t-test or one-way analysis of variance (ANOVA) followed by the Tukey HSD test. A P value less than 0.05 was considered statistically significant. All statistical calculations were done by R software (Ver. 3.5.0; https://cran.ism.ac.jp/bin/windows/).

Previously, we revealed the presence of two testis-specific lncRNAs at the mouse Prss/Tessp locus and we suggested significant roles of these noncoding regions. In this study, we searched for other lncRNAs and found by RT-PCR that a 1418-bp sequence was transcribed in an intergenic region between Prss43/Tessp-3 and Prss45/TESPL genes in the testis ( Figure 2A ). We then determined the 5’- and 3’-ends of this transcript by RACE analyses. In 3’RACE, two bands were specifically amplified by the second PCR (marked as “a” and “b” in Figure 2B ), and 10 subclones from each were sequenced. It was found that all subclones contained the same sequence for each band. Due to stronger band intensity, a cytosine from band “a” was determined to be a major transcriptional termination site ( Figure 2C ). In 5’RACE, two bands were detected by the second PCR (marked as “c” and “d” in Figure 2D ), and 10 subclones were checked for each band. For band “c”, three different sequences were detected, but the subclones from band “d” contained a single sequence. Because band “d” was more intense than band “c”, a guanine from band “d” was determined to be a major transcriptional start site ( Figure 2E ). Note that the transcriptional start site was within the primer sequence used for the initial RT-PCR amplifying the 1418-bp product ( Figure 2A ). Consequently, this transcript was defined to be a single-exon structure of a 1822-nucleotide sequence with a poly(A) tail at its 3’ end. Based on the function reported here, we named this transcript Start.

RT-PCR was performed using 9 tissues to examine the tissue specificity of Start, and a signal was only observed in the testis ( Figure 3A ). During the first wave of spermatogenesis, qRT-PCR showed that Start expression was significantly elevated from 7 dpp to 14 dpp and remained at high levels thereafter ( Figure 3B ). We then separated testicular cells into three fractions, germ cells, Leydig/Sertoli cells and Sertoli cells, and performed RT-PCR. As shown in Figure 3C , Start was expressed in both germ cells and Leydig cells but not in Sertoli cells. To investigate the subcellular localization, germ cells were separated into nuclear and cytosolic fractions. RT-PCR showed more dominant localization in the cytosol, but the signal was also found in the nucleus ( Figure 3D ). These results indicated that Start is a germ cell and Leydig cell-specific lncRNA and is activated during sperm development.

To further determine the localization of Start, we performed in situ hybridization by the highly sensitive tyramide signal amplification method (53). Antisense and sense probes for Start were hybridized with adult testis sections and the signal was detected as red fluorescence. Consistent with results shown in Figure 3C , a strong signal was detected in Leydig cells, mostly in the cytosolic region, by the antisense probe, while few signals were seen in the section hybridized with the sense probe ( Figures 4A, B ). In the seminiferous tubules, many more red dots were observed with the antisense probe ( Figures 4C–E ) than with the sense probe ( Figure 4F ). At seminiferous epithelial stages II-IV and VI-VII ( Figures 4C, D ), the signals were found in all types of spermatogenetic cells including spermatogonia (SG), primary spermatocytes at the early and mid-pachytene stages (ePS and mPS), and round and elongating spermatids (RS and ES, respectively). Consistent with the results of RT-PCR shown in Figure 3D , more signals were present in the cytosol of these cell types. At stage X, most of the spermatogenetic cells were stained in a cytosol-predominant manner as in earlier stages ( Figure 4E ), but late pachytene spermatocytes (lPS) appeared to contain more signals in the nuclei than did other types of germ cells. Similar results were obtained with a testis from another mouse. Collectively, the results showed that Start was dominantly present in the cytosol of Leydig cells and all types of spermatogenetic cells, though transcripts were also localized in the nucleus. Noticeably, Leydig cells carried the strongest signals among all cell types, suggesting that Start plays an important role in Leydig cells.

To reveal the physiological role of Start, we generated Start-knockout (KO) mice by the CRISPR/Cas9 system using two guide RNAs designed at the 5’- and 3’-flanking sequences of Start ( Figure 5A ). Injection of the guide RNAs and CAS9 protein into fertilized eggs resulted in the generation of four founder mice that had a deletion of Start, and two KO lines, lines #1 and #2, were successfully established by mating with wild-type mice. These KO mice contained 2222-bp (#1) and 2208-bp (#2) deletions, which meant that #1 contained 5 and 9 nucleotides longer deletions than #2 at the 5’ and 3’ regions, respectively ( Figures 5A, B ). RT-PCR confirmed no expression of Start in these two KO lines ( Figure 5C ). The two lines were used in this study, and all experiments were done with three or more mice except for RNA-seq.

Both #1 and #2 lines of Start-KO mice were fertile, and we observed morphology of the adult testis by collecting testes from 2.5-month-old wild-type and Start-KO mice. The testis weight normalized to body weight was not significantly different between wild-type and KO mice ( Figure 5D ), and observations of testis sections stained with hematoxylin and eosin did not show any morphological difference ( Figure 5E ). In addition, the number of sperms collected from the cauda epididymides was not changed in KO adult mice ( Figure 5F ). To see whether sex differentiation was affected, we investigated the sex ratio among 16 littermates and we found no significant difference (17 males and 11 females in wild-type vs 22 males and 13 females in Start-KO, no significant difference by the chi-square test). Therefore, the adult Start-KO testis appeared to be morphologically normal.

Subsequently, we attempted to detect differentially expressed genes. Total RNA was purified from whole testes of a 2.5-month-old wild-type and Start-KO mice from the same litter and used for RNA-seq analysis. The resultant reads, mapping rate to the mouse cDNA library, and fastq accessions in the NCBI SRA database are summarized in Table 2 . Interestingly, the Star gene encoding a protein to trigger steroidogenesis was found to be upregulated by 2.3 fold in the Start-KO testis ( Figure 6A ). Furthermore, expression levels of other steroidogenic genes including Cyp11a1, Cyp17a1, Hsd3b1, and Hsd17b3 showed 1.3-1.6-fold increases in the Start-KO testis ( Figure 6A ). qRT-PCR for these genes using four sets of wild-type and KO mice confirmed that Star and Hsd3b1 genes were significantly upregulated in the Start-KO testis ( Figure 6B ). We then measured testosterone concentration in serum and testis. Surprisingly, ELISA analysis showed that the testosterone level was significantly reduced in both serum and testis of KO mice ( Figures 6C, D ). This is inconsistent with the increase of steroidogenic gene expression, but the results suggested that Start participates in regulation of the steroidogenesis pathway in Leydig cells.

To understand why the testosterone level was low despite high expression of steroidogenic genes in adult Start-KO testes, we performed several experiments. First, we considered the possibility that conversion into 5α-dihydrotestosterone might be augmented in KO mice. However, qRT-PCR indicated that expression of the Srd5a1 gene, encoding 5α-reductase, was not changed in Start-KO testes ( Supplemental Figure 1 ). Second, we investigated the effect of Start deficiency on Androgen receptor (Ar) and found by qRT-PCR that Ar expression was not different in wild-type and Start-KO testes ( Supplemental Figure 2A ). This suggests that the low level of testosterone might affect the androgen-AR autocrine pathway in adult Start-KO testes. Third, we investigated the effect of Start deficiency on the HPG axis. In the pituitary, an ELISA assay and qRT-PCR showed that the LH level and Lhb mRNA expression were not significantly different in wild-type and Start-KO mice ( Figures 7A, B ) but that Ar expression was dramatically elevated in KO mice ( Figure 7C ). In the testis, the expression of Lhcgr, a receptor of LH, which is supposed to be specific to Leydig cells (58), was significantly increased in Start-KO mice ( Figure 7D ). Presumably, the upregulation of Star and Hsd3b1 in adult Start-KO testes resulted from a secondary or indirect effect of the Start deficiency on the AR autocrine pathway or on the HPG axis (see “Discussion”).

To gain more insight into effects of Start deficiency on steroidogenesis, we investigated Start-KO mice at 8 dpp. We first conducted morphological observation by preparing cross-sections of testes of wild-type and Start-KO mice from the same litter and staining them with hematoxylin and eosin. As shown in Figure 8A , no significant differences were observed within the seminiferous tubules, whereas more interstitial regions containing few cells were observed in the Start-KO testis than in the wild-type testis (see regions encircled in red). Indeed, the number of interstitial cells in randomly selected areas (0.25 mm²) in the sections was significantly smaller in Start-KO testes at 8 dpp than in wild-type testes ( Figure 8B ). In contrast, no significant change was observed in interstitial cells between KO and wild-type mice at 2.5 months of age ( Figure 8B ). Since interstitial cells mainly consist of Leydig cells but also contain small populations of other types of cells (59, 60), we investigated by immunohistochemistry of HSD3B, a marker of Leydig cells (61, 62), whether the altered number of interstitial cells reflected the Leydig cell population. The number of HSD3B-positive cells was significantly decreased in Start-KO testes at 8 dpp, while it was not altered at 2.5 months of age ( Figure 8C ). This result was supported by two additional data. First, qRT-PCR indicated a significant decrease of Insl3 expression, another marker gene for Leydig cells (63, 64), in Start-KO testes at 8 dpp but not in adults ( Supplemental Figure 3A ). Second, the expression of Hsd3b6, a marker gene for adult Leydig cells (65), showed no difference between wild-type and Start-KO testes in adults ( Supplemental Figure 3B ). These results indicated that the number of Leydig cells was decreased in Start-KO mice at 8 dpp.

Next, we investigated the expression of steroidogenic genes in wild-type and Start-KO testes at 8 dpp. Our qRT-PCR indicated that expression levels of Star and Hsd3b1 genes were significantly low in Start-KO testes ( Figure 9A ), and serum testosterone level was also found to be lower in Start-KO mice than in wild-type mice at 8 dpp ( Figure 9B ). This is in contrast to the results for Start-KO mice at 2.5 months of age. To clarify the difference from adults, we examined the expression of hormone receptors in the testis at 8 dpp. While qRT-PCR showed that Ar mRNA expression was not different in wild-type and Start-KO mice as in adults ( Supplemental Figure 2B ), expression of the Lhcgr gene, which was increased in Start-KO adult testes ( Figure 7D ), was not changed by Start deficiency at 8 dpp ( Figure 9C ). Collectively, the results showed that steroidogenic genes were downregulated in the testis and that the testosterone level was decreased in Start-KO mice at 8 dpp.

To determine how Start regulates steroidogenic genes in Leydig cells, we utilized two mouse Leydig cell lines, TM3 and MA-10 (66, 67), to perform Start overexpression. The antisense sequence of Start was overexpressed as a negative control. Strand-specific RT-PCR confirmed successful overexpression of Start and its antisense transcript in both cell lines ( Figures 10A, C ). In TM3 cells, Star gene expression was significantly increased by Start overexpression ( Figure 10B ), and in MA-10 cells, not only Star but also Cyp17a1, Hsd3b1, and Hsd17b3 genes were significantly upregulated by Start ( Figure 10D ). We failed to detect testosterone in the culture media of MA-10 cells overexpressing Start, probably due to low capability of testosterone production of this cell line as previously reported (67, 68). These results supported the notion that Start acts as an activator of steroidogenic genes in Leydig cells, as shown by results for 8-day-old mice ( Figure 9A ). Based on these results, we concluded that Start activates a subset of genes involved in steroidogenesis in Leydig cells.

While we found a role of Start in Leydig cells, it is presumed that Start also has some function in germ cells, given its localization in meiotic cells. In general, lncRNAs regulate the expression of their target genes at various levels depending on their subcellular localization (69, 70). Start was found to be localized in the cytosol of most germ cells, which raises the possibility that Start is related to the regulation of RNA stability or translation of meiotic genes, as was previously reported (71, 72). At the late pachytene stage, Start was also localized in the nucleus, suggesting a role in transcriptional gene regulation as was observed for other lncRNAs (73–75). Since expression of various meiotic genes is changed during spermatogenesis (76–78), Start may be involved in the regulation of such genes in germ cells.

It is interesting that Start, Tesra, and lncRNA-HSVIII, three lncRNAs at the Prss/Tessp locus ( Figure 1 ), are all expressed in germ cells and Leydig cells. At the late pachytene stage, when some of the Prss/Tessp genes are activated, they are localized in the nucleus. This indicates the possibility that Prss/Tessp genes are regulated by these lncRNAs. Indeed, Tesra contributes to transcriptional activation of the Prss42/Tessp-2 gene by collaborating with an enhancer adjacent to lncRNA-HSVIII (45). However, the contribution of lncRNA-HSVIII to expression of the Prss42/Tessp-2 gene is not clear, and the function of Start in germ cells remains to be elucidated. Therefore, analysis considering the functional relationship of the three lncRNAs at the Prss/Tessp locus is necessary to clarify the function of Start in germ cells.

Our data indicated a decrease in the number of interstitial Leydig cells in Start-KO mice at 8 dpp. This was based on the significantly small number of HSD3B-positive cells shown by immunohistochemistry ( Figure 8C ) and a significantly low level of Insl3 expression shown by qRT-PCR ( Supplemental Figure 3A ). Both changes were observed in Start-KO testes at 8 dpp but not in adults. No difference in Hsd3b6 gene expression between wild-type and Start-KO testes supports the notion that Start does not affect adult Leydig cells ( Supplemental Figure 3B ). Therefore, we concluded that the number of fetal Leydig cells was decreased by deletion of Start at 8 dpp. This suggests the involvement of Start in the development of fetal Leydig cells. According to previous studies, fetal and adult Leydig cells are differentiated by distinct regulatory mechanisms (79, 80). However, the mechanism of the differentiation and proliferation of Leydig cells is not fully understood and is even controversial (80). Therefore, Start may play some role in the differentiation or proliferation of fetal Leydig cells or their progenitor cells, but the detailed mechanism needs further research. Elucidation of the molecular mechanism will provide an insight into the function of Start in the development of fetal Leydig cells.

In the present study, we obtained data supporting the involvement of Start in the regulation of steroidogenesis in Leydig cells. First, many Start transcripts were localized in Leydig cells, where steroid hormones are synthesized ( Figure 4A ). Second, expression of Star and Hsd3b1 genes was changed in Start-KO mice at 8 dpp and in adults ( Figures 6B , 9A ). Third, serum testosterone level was lower in Start-KO mice than in wild-type mice both in adults and at 8 dpp ( Figures 6C , 9B ). Fourth, intratesticular testosterone level was lower in adult Start-KO mice than in wild-type mice ( Figure 6D ). Fifth, overexpression of Start in two Leydig cell lines significantly increased the expression of steroidogenic genes ( Figure 10 ). Although the decreases in the gene expression and testosterone level might be partially due to the reduction in the number of Leydig cells at 8 dpp, these results strongly suggest that Start is a regulator of steroidogenesis in the mouse testis by controlling at least Star and Hsd3b1 gene expression in Leydig cells.

Then, how does Start regulate the expression of steroidogenic genes? In Leydig cells, Start was mostly localized in the cytosol, suggesting its function in post-transcriptional regulation such as mRNA stabilization and translation. In the case of regulation of mRNA stability, lncRNAs are known to form double-strand RNAs with their target mRNAs and induce molecular events leading to RNA degradation or stabilization (81, 82). In translational regulation, lncRNAs generally interact with miRNAs to block their function, resulting in enhancement of the translation of miRNA-targeted mRNAs (75, 83, 84). Given that Start is involved in the regulation of multiple target genes and that one of the targets, Star, has been reported to be regulated by an miRNA (85), we prefer the possibility of the mechanism involving the miRNA pathway. This may also be one explanation for the reason why Cyp17a1 and Hsd17b3 genes were upregulated by Start overexpression in MA-10 cells but were not changed in Start-KO testes. There might be miRNAs that can regulate Cyp17a1 and Hsd17b3 genes and interact with Start in MA-10 cells, but they might not be physiologically significant in Start-KO testes. Elucidation of the detailed molecular mechanism by which Start controls the expression of steroidogenic genes is currently in progress.

It is notable that Start-KO mice were fertile despite low levels of testosterone in serum and testis. In general, testosterone production plays a key role in the generation of mature sperm, and many studies have shown that low testosterone levels were associated with male infertility in mammals (86, 87). Meiotic arrest observed in Sertoli cell-specific AR-KO mice supports the notion that testosterone is essential for spermatogenesis (88, 89). However, in some cases, normal progression of spermatogenesis was reported even when testosterone levels were greatly reduced (90, 91), and testosterone levels were not necessarily correlated with fertility (92). These results suggest that a high testosterone level is important for complete spermatogenesis but is not an absolute requirement. In Start-KO mice, the reduced level of testosterone might still be sufficient for spermatogenesis.

Interestingly, Star and Hsd3b1 gene expression was upregulated in adult Start-KO testes despite the low level of testosterone ( Figure 6 ). We do not have a clear explanation for this counterintuitive result, but we have several suggestions based on our data. One possibility is that testosterone is more actively metabolized to other molecules. Although we cannot rule out this possibility, enhanced conversion to 5α-dihydrotestosterone was unlikely considering the results showing no significant change in the expression of Srd5a1 gene in Start-KO testes ( Supplemental Figure 1 ). Another possibility is that enhancement of Star and Hsd3b1 expression is caused by autocrine signaling of the androgen-AR trail in Leydig cells. We did not detect any difference in Ar expression between wild-type and Start-KO adult testes ( Supplemental Figure 2 ), but the low level of testosterone might cause a defect in the AR-autocrine pathway. It has been reported that the expression levels of some steroidogenic genes were significantly decreased but that the expression levels of other steroidogenic genes were increased in Leydig cell-specific AR KO testes (93). This indicates that the deficiency of autocrine signaling via AR in Leydig cells can cause a disturbance of steroidogenic gene expression. In adult Start-KO males, the low level of testosterone might cause a deficiency in this autocrine system and lead to abnormal elevation of Star and Hsd3b1 expression.

There is also the possibility that the low level of testosterone affects the HPG axis in Start-KO adults. In the pituitary, a low level of testosterone often causes an increase in LH secretion (94–96), but in Start-KO mice, neither LH secretion nor Lhb mRNA expression was significantly changed ( Figures 7A, B ). This was probably associated with a dramatic increase in Ar expression in the pituitary of Start-KO mice ( Figure 7C ), as shown by studies in other tissues (97–99). Notably, Lhcgr expression was found to be significantly increased in Start-KO adult testes ( Figure 7D ), which is consistent with the upregulation of this receptor gene in AR hypomorphic mutants (100). This was only observed in adults and suggests that the LH signaling pathway in Leydig cells might be enhanced under the condition of normal LH secretion, by which LH-responsive genes such as Star and Hsd3b1 might be upregulated in Start-KO adult testes. Given that mouse models overexpressing human chorionic gonadotropin or a constitutively active LH receptor in the testis showed higher levels of testosterone (101–103), the enhancement of the LH pathway might intrinsically cause an increase in testosterone production. Nonetheless, Start deficiency caused a significant decrease in testosterone level, suggesting that Start is a critical regulator of steroidogenesis in the testis.

At 8 dpp, although an effect of pituitary hormones might remain by minipuberty (104, 105), we verified more direct effects of Start deficiency than in adults and observed downregulation of Star and Hsd3b1 genes. Taken together with the increased expression of steroidogenic genes in Start-overexpressing TM3 and MA-10 cells ( Figure 10 ), the function of Start is likely to be enhancement of the expression of steroidogenic genes.

Most previous studies focused on the functions of testis lncRNAs in germ cells (33–36, 41–43, 45, 106) but not on their roles in Leydig cells (40, 107). The results of the present study shed light on the in vivo function of Start, which plays roles in maintenance of the steroid synthesis pathway in the testis. Thus, this report is the first report on a testis-specific lncRNA that finetunes the steroid synthesis pathway in Leydig cells.