Loss of Cx43 in Murine Sertoli Cells Leads to Altered Prepubertal Sertoli Cell Maturation and Impairment of the Mitosis-Meiosis Switch

Male factor infertility is a problem in today’s society but many underlying causes are still unknown. The generation of a conditional Sertoli cell (SC)-specific connexin 43 (Cx43) knockout mouse line (SCCx43KO) has provided a translational model. Expression of the gap junction protein Cx43 between adjacent SCs as well as between SCs and germ cells (GCs) is known to be essential for the initiation and maintenance of spermatogenesis in different species and men. Adult SCCx43KO males show altered spermatogenesis and are infertile. Thus, the present study aims to identify molecular mechanisms leading to testicular alterations in prepubertal SCCx43KO mice. Transcriptome analysis of 8-, 10- and 12-day-old mice was performed by next-generation sequencing (NGS). Additionally, candidate genes were examined by qRT-PCR and immunohistochemistry. NGS revealed many significantly differentially expressed genes in the SCCx43KO mice. For example, GC-specific genes were mostly downregulated and found to be involved in meiosis and spermatogonial differentiation (e.g., Dmrtb1, Sohlh1). In contrast, SC-specific genes implicated in SC maturation and proliferation were mostly upregulated (e.g., Amh, Fshr). In conclusion, Cx43 in SCs appears to be required for normal progression of the first wave of spermatogenesis, especially for the mitosis-meiosis switch, and also for the regulation of prepubertal SC maturation.


Introduction
A steady decrease in birth rate is one of the most severe social problems in industrialized countries nowadays [1]. In this context, a significant decline in male reproductive function has been reported in After paraffin-embedding and sectioning, the slides were deparaffinized and rehydrated. To block the endogenous peroxidase, all sections were incubated in 196 mL 80% alcohol comprising 4 mL hydrogen peroxide at room temperature for 30 min. Following this, heat-induced antigen retrieval was performed using sodium citrate buffer (pH 6.0) and a heat plate at 96 to 99 • C for 20 min (SOX9, AMH) or slides were either microwaved at 800 watts for three minutes and 12 min at 600 watts (SOHLH1) or at 800 watts for 15 min (Cx43) or for 10 min (LIN28A, SALL4). β-galactosidase IHC was conducted without antigen retrieval. Next, the slides were incubated with the primary antibody (Table 1) and 5% bovine serum albumin in a humidified chamber at 4 • C overnight (LIN28A, SALL4) or the slides were either blocked with 3% bovine serum albumin for 20 min (SOX9, AMH, Cx43), 30 min (SOHLH1), or with normal goat serum (1:5) at room temperature for 20 min (β-galactosidase). Then, these slides were incubated with the appropriate primary antibody in a humidified chamber at 4 • C overnight. Thereafter, all slides were exposed to a compatible horseradish peroxidase (HRP)-conjugated secondary antibody (EnVision, DAKO, Hamburg, Germany, ready to use) for 30 min (AMH, β-galactosidase, SOX9, LIN28A, SALL4) or to a biotinylated goat anti-rabbit secondary antibody (Vector Laboratories, Burlingame, CA, USA) either for 30 min (Cx43) or for 60 min (SOHLH1), followed by 30-min incubation with avidin-biotin peroxidase complex (VECTASTAIN ® Elite ® ABC Kit, Vector Laboratories, Burlingame, CA, USA) (Cx43, SOHLH1). Immunoreactivity was visualized by 3, 3 -diaminobenzidine (DAB+) substrate-chromogen. Finally, the sections were counterstained with hematoxylin, followed by dehydration and covering with Eukitt (O. Kindler GmbH, Freiburg, Germany).

RNA Extraction
The RNeasy Lipid Tissue Mini Kit with QIAzol Lysis Reagent (Quiagen, MD, USA) was used for the extraction according to the manufacturer's protocol. The obtained RNA was quality controlled on an Agilent 2100 Bioanalyzer system with an RNA 6000 Nano Kit (Agilent Technologies, Santa Clara, CA, USA). Only those samples with an RNA Integrity Number (RIN) value ≥ 8 were chosen for further analysis.

mRNA Sequencing
The prepared RNA samples were quantified on a Qubit Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) and adjusted to 1000 ng. Library preparation was performed using the TruSeq Cells 2020, 9, 676 5 of 37 Stranded mRNA Sample Preparation Kit (Illumina, San Diego, CA, USA). In total, 18 samples were run on a High Output Kit v2 (Illumina, San Diego, CA, USA) and a further six samples on a Mid Output Kit v2 on the NextSeq 500 system for 2 × 75 bp in paired-end mode. The mRNA-seq data are deposited at Sequence Read Archive (NCBI, https://www.ncbi.nlm.nih.gov/sra, project ID: PRJNA603458). The submission IDs of fastq-files, samples and genotypes used for next-generation sequencing of mRNA are shown in Table S1.

cDNA Synthesis and Quantitative Real-Time Reverse Transcription-PCR
To validate NGS data by qRT-PCR, five candidate genes were chosen according to the following criteria: GC-specific and significantly downregulated (p(FDR) < 0.05) in KO mice (Sohlh1, Sohlh2), SC-specific and significantly upregulated (p(FDR) < 0.05) in murine KO animals (Amh, Fshr) as well as Gja1 that codes for Cx43. β-Actin was selected as a housekeeping gene because no significant regulation at any age and group was observed by NGS.
Initially, obtained RNA samples with a RIN value ≥ 8 were applied to cDNA synthesis by reverse transcription using the Biometra T-Professional Thermocycler (Biometra GmbH, Göttingen, Germany) according to the manufacturer's instructions. In order to do so, a reaction medium comprising 1x Taq DNA-Polymerase-PCR buffer (20 mM Tris HCl (pH 8.4), 50 mM KCl, 18067017, Invitrogen™, Darmstadt, Germany), 5 mM MgCl 2 (18067017, Invitrogen™, Darmstadt, Germany), 1 mM dNTP (NU-0010-10, Eurogentec, Cologne, Germany), 2.5 µM random hexamers (N8080127, Applied Biosystems™, Darmstadt, Germany), 20 U RNAse Inhibitor (N8080119, Applied Biosystems™, Darmstadt, Germany), and 50 U moloney murine leukemia virus reverse transcriptase (M-MLV-RT) (28025013, Invitrogen™, Darmstadt, Germany) was used. Then, 200 ng of total RNA were applied to achieve a final cDNA concentration of 5 ng/µL after reverse transcription reaction. Next, the reaction volume was replenished with nuclease-free water (Fresenius Kabi AG, Bad Homburg, Germany) to a final volume of 20 µL. Subsequently, the reverse transcription reaction was performed for ten minutes at 25 • C, followed by one-hour incubation at 42 • C. Denaturation was carried out at 99 • C for five minutes, and following this the samples were flash-cooled on ice. Negative controls were conducted by renouncing reverse transcriptase and RNase inhibitor to assess DNA contamination, by omitting RNA, reverse transcriptase and RNAse inhibitor to detect possible RNA contamination as well as by excluding RNA and replacing it with Ampuwa ® water (Fresenius Kabi AG, Bad Homburg, Germany) as control for contamination of the reagents.
For qRT-PCR, the BioRad Real Time System CFX96 1000 Touch (BioRad, Munich, Germany) and a reaction mix, which contained 10 µL of Mesa Green PCR MasterMix Plus for SYBR ® Assay (Eurogentec, Cologne, Germany), and 0.2 µM concentrations of each primer (Eurofins MWG Operon, Ebersberg, Germany) for the genes of interest was used. The primers were either previously described or constructed by means of the NCBI primer designing tool [36], respectively, Primer3: WWW primer tool (http://biotools.umassmed.edu/bioapps/primer3_www.cgi) [37]; the respective sequences and accession numbers are shown in Table 2. All reactions were run in duplicates with a 20 µL volume including 1.0 µL cDNA (with a concentration of 5 ng/µL). After adding the corresponding amount of primers, the volume was filled up with nuclease-free water (Fresenius Kabi AG, Bad Homburg, Germany). For each primer pair, a negative control was implemented comprising nuclease-free water instead of cDNA. Additionally, negative controls containing RNA but no RNAse inhibitor and no reverse transcriptase as well as a negative control without RNA, RNAse inhibitor and no reverse transcriptase and one negative control comprising RNAse inhibitor and reverse transcriptase but no RNA were included as controls for the contamination of the cDNA. Cycling conditions were as follows: denaturation at 95 • C for ten minutes and then 43 cycles of following steps: denaturation at 95 • C for 15 s, annealing at 60 • C for 30 s and elongation at 72 • C for 30 s. Transcripts were visualized using Mesa Green (Eurogentec, Cologne, Germany). A subsequent melting curve was generated to verify the PCR fragments and was initiated at 55 • C to a final temperature of 95 • C with an increase of 0.5 • C every ten seconds. The relative abundance of total RNA was determined relative to the housekeeping Cells 2020, 9, 676 6 of 37 gene (β-Actin), which did not differ between KO and WT littermates of all ages (p > 0.05). The ∆Ct (delta cycle threshold) method was used to calculate the relative abundance of total RNA in the testis samples. The mean value of the Ct-merit of a specific gene was subtracted from the mean value of the housekeeping gene (β-Actin), and calculated by the formula 2 −∆Ct .
The subsequent counting of mapped reads per gene was done using the RSEM software (V 1.3.0) [41]. RSEM values were further normalized by either GC or SC counts and prepared for further analysis using the VOOM transformation in the R-package limma to obtain log2 counts [42]. The final expression matrix was explored by means of principal component analysis (PCA) and hierarchical clustering to identify a potential grouping of samples. The identification of differentially expressed genes between WT and KO animals and between the three time points (days 8, 10 and 12) was again performed using the functionality of the limma package [43]. Raw p-values were adjusted by the method of Benjamini and Hochberg to control a false discovery rate (FDR) of 5% [44]. Enrichment of Gene Ontology (GO) terms [45] was studied using the gene-set enrichment analysis (GSEA) with the Kolmogorov-Smirnov (KS) test [46].
Cell counts are presented as mean ± standard deviation for each experimental group (n = 4 per age group and genotype). Next, the effect of group (KO, WT) and time (days 8, 10, 12) as well as the interaction between group and time were analysed using 2-way analysis of variance (Table 3). Subsequent comparisons between KO and WT cell counts were performed by t-tests ( Figure 3). The individual time effect for the KO and WT group, respectively, was studied by 1-way analysis of variance. The level of significance was set to alpha = 0.05. PANTHER pathway analysis was conducted using all significantly differentially expressed genes (p(FDR) < 0.05) for days 8, 10 and 12 p.n. and PANTHER [47,48].
Delta-Ct values (2 −∆Ct ) from qRT-PCR were compared between KO and WT samples by t-tests, and the resulting p-values were adjusted according the Bonferroni-Holm method.

Further Analysis of Differentially Expressed Genes
To identify potentially interesting candidate genes (Table S2), a candidate gene list for the terms 'infertility', 'idiopathic infertility', 'disturbed meiosis' and 'spermatogenic arrest' was prepared using NCBI Gene database, for the terms 'Sertoli cell' and 'Sertoli cell-specific' generated with the UniProt Knowledgebase, and supplemented with further genes ascertained in previous studies [9,49,50] as well as with genes belonging to the GO terms which were found to be significant at all three time points (see below). Finally, only significantly expressed genes at all three time points (p(FDR) < 0.05) were selected as potential candidate genes and were further investigated using PANTHER [47,48,51].

Confirmation of Cx43 Gene Loss on Protein Level
In addition to PCR genotyping, β-galactosidase ( Figure 1) and Cx43 ( Figure 2) IHC were conducted to confirm the successful deletion of Cx43 protein in SCs of SCCx43KO mice.
3.1.1. β-galactosidase Immunohistochemistry A nuclear staining was solely detected in SCs in the seminiferous tubules of SCCx43KO mice ( Figure 1A-C) and thus confirming a successful Cx43 gene deletion as described previously [26]. All other cells showed no immunostaining. In the seminiferous tubules of WT littermates, SCs were immunonegative ( Figure 1D-F). None of the negative controls showed any immunoreaction ( Figure S1A).

Further Analysis of Differentially Expressed Genes
To identify potentially interesting candidate genes (Table S2), a candidate gene list for the terms 'infertility', 'idiopathic infertility', 'disturbed meiosis' and 'spermatogenic arrest' was prepared using NCBI Gene database, for the terms 'Sertoli cell' and 'Sertoli cell-specific' generated with the UniProt Knowledgebase, and supplemented with further genes ascertained in previous studies [9,49,50] as well as with genes belonging to the GO terms which were found to be significant at all three time points (see below). Finally, only significantly expressed genes at all three time points (p(FDR) < 0.05) were selected as potential candidate genes and were further investigated using PANTHER [47,48,51].

Confirmation of Cx43 Gene Loss on Protein Level
In addition to PCR genotyping, β-galactosidase ( Figure 1) and Cx43 ( Figure 2) IHC were conducted to confirm the successful deletion of Cx43 protein in SCs of SCCx43KO mice.
3.1.1. β-galactosidase Immunohistochemistry A nuclear staining was solely detected in SCs in the seminiferous tubules of SCCx43KO mice ( Figure 1A-C) and thus confirming a successful Cx43 gene deletion as described previously [26]. All other cells showed no immunostaining. In the seminiferous tubules of WT littermates, SCs were immunonegative ( Figure 1D-F). None of the negative controls showed any immunoreaction ( Figure  S1A).

Cx43 Immunohistochemistry
In SCCx43KO mice, no immunoreaction was observed in any of the seminiferous tubules, regardless of the age group (Figure 2A,C,E,G). Hence, confirmation of successful Cx43 gene loss in SCs was given again. According to Bravo-Moreno and colleagues [20] as well as Gerber and fellow workers [52], immunostaining for Cx43 was detected in seminiferous tubules of WT mice at all age groups, but with an age-dependent expression pattern ( Figure 2B,D,F,H). In 8-and 10-day-old WT animals, the immunosignal appeared in the cytoplasm of SCs ( Figure 2B,D). At the age of 12 days, the immunoreaction for Cx43 weakened in the cytoplasm of SCs and a shift towards the basal compartment of the seminiferous epithelium was observed ( Figure 2F). Finally, in adult WT mice, as expected, Cx43 expression was detected at the location of the blood-testis barrier between adjoining SCs as well as between SCs and GCs ( Figure 2H). No immunostaining was detected in the negative controls ( Figure S1B).

Prepubertal SCCx43KO Mice Show Evident Differences in the Composition of Intratubular Cells
Morphological, histological and immunohistochemical examination ( Figure 3A-F) revealed evident differences in the composition of intratubular cells between 8-, 10-and 12-day-old WT and KO littermates. GC numbers per seminiferous tubule amounted on average 2 ± 0.78 GCs in 8-day-old SCCx43KO mice, 3.96 ± 0.54 GCs in 10-day-old SCCx43KO animals and 4.21 ± 0.43 GCs in 12-day-old SCCx43KO mice, whereas WT littermates had, on average, 9.49 ± 3.90 (8 days p.n.), 16.44 ± 1.80 (10 days p.n.) and 21.37 ± 3.50 GCs (12 days p.n.) per seminiferous tubule ( Figure 3G).   For GCs, the main effects as well as the interaction between them were significant ( Table 3). The significant interaction means that the group effect is of different size at the individual time points or, vice versa, the time effect is different for KO and WT. The results from the t-tests show that the group effect was strongest at Day 12. The individual time effect was significant for the KO as well as for the WT group (each p < 0.01).
For SCs, the group and time effects, but not the interaction between them, were significant (Table 3). Time specific comparisons did not provide further evidence for a significant difference between KO and WT samples. Thus, there is also an overall group effect. The individual time effect was significant for the KO (p < 0.01) as well as for the WT (p = 0.02) group.
In comparison to WT mice, the GC number was reduced by approximately three quarters in KO mice. Moreover, an increasing (but not significant) tendency of the SC number was observed in prepubertal SCCx43KO mice. Based on these findings, NGS data were normalized regarding GC numbers or SC numbers (see Chapter 3.3.2.). Table 3. Results from 2-way analysis of variance for testing the influence of group, time and their interaction on either germ or Sertoli cell counts.

Cells Factor p-Value
Germ cells

Preprocessing and Data Exploration
Mapping and counting resulted in expression levels for 41,097 genes in total. Exploration by PCA ( Figure 4) showed a clear separation of WT and KO samples on the second principal component as well as a separation of time points on the first principal component.

Preprocessing and Data Exploration
Mapping and counting resulted in expression levels for 41,097 genes in total. Exploration by PCA ( Figure 4) showed a clear separation of WT and KO samples on the second principal component as well as a separation of time points on the first principal component. Hierarchical clustering ( Figure 5) also showed that WT and KO samples were sometimes more similar than samples between time points. Thus, the time effect appears to be larger than the group effect. Hierarchical clustering ( Figure 5) also showed that WT and KO samples were sometimes more similar than samples between time points. Thus, the time effect appears to be larger than the group effect. Hierarchical clustering ( Figure 5) also showed that WT and KO samples were sometimes more similar than samples between time points. Thus, the time effect appears to be larger than the group effect.

Differential Expression Analysis
The numbers of differentially expressed genes were studied in different models (Table 4). Comparisons were either performed in all samples or in subgroups divided by group (WT/KO) or time (day8, 10, 12). Analyses were either performed without cell count specific normalization or with two variants of normalization (Tables S3-S14). Note: Fc-values were calculated relating values from WT mice to values from KO mice. Hence, a positive fc-value means a downregulation of corresponding gene transcripts in the KO mice and a negative fc-value means that transcripts are upregulated in the KO animals (Tables S2-S14 and S21).

Gene Ontology Analysis by GSEA
In total, 8904 GO terms were related to the genes of the transcriptome expression matrix. GO term enrichment was either studied for the comparison of KO versus WT based on all samples or based on time (Day 8, 10,12). Due to the resampling strategy of GSEA and the use of the nonparametric KS test, the resulting p-values had only a precision of three digits. With this limitation, 158 GO terms yielded a p-value of 0 for the comparison of KO versus WT based on all samples (Table S15). The 'GO:0007141 male meiosis I' was found among others. The time specific comparison of KO versus WT samples revealed 180 GO terms at day8, 47 GO terms at Day 10 and 246 GO terms at Day 12 with a p-value of 0 (Tables S16-S18). Ten GO terms with a p-value of 0 were found at all three time points. Table 5 lists these ten GO terms including their corresponding genes.

PANTHER Pathway Analysis of Significantly differentially Expressed Genes
In order to gain a general overview of involved pathways, PANTHER pathway analysis was performed with all significantly differentially expressed genes in 8-, 10-and 12-day-old SCCx43KO mice (p(FDR) < 0.05). In total, 133 pathways were identified (Table S19). Among these, the 'gonadotropin-releasing hormone receptor pathway' showed the most gene hits, followed by 'Wnt signaling pathway', 'CCKR signaling' and 'inflammation mediated by chemokine and cytokine signaling pathway'.

Further Characterization of Differentially Expressed Genes
Preparation of the candidate gene list (Table S2) revealed 254 potential candidate genes that were significantly differentially expressed (p(FDR) < 0.05) in the SCCx43KO animals in all three age groups ( Figure 6). Each column represents a sample of SCCx43KO mice or their WT littermates at different ages (n = 4 per genotype and age group, t8: 8-day-old, t10: 10-day-old, t12: 12-day-old). Genes are depicted in rows. Red indicates downregulation and blue indicates upregulation. KO: knockout, SCCx43KO: Sertoli cell-specific Cx43 knockout, WT: wild type.
In order to identify the most significantly differentially expressed candidate genes in 10-day-old animals, the candidate gene list was sorted by p(FDR)-values and a literature review was conducted for the first 15 genes (Table 6). Noteworthy, all of these 15 genes were significantly downregulated in the 8-, 10-and 12-day-old SCCx43KO mice. Each column represents a sample of SCCx43KO mice or their WT littermates at different ages (n = 4 per genotype and age group, t8: 8-day-old, t10: 10-day-old, t12: 12-day-old). Genes are depicted in rows. Red indicates downregulation and blue indicates upregulation. KO: knockout, SCCx43KO: Sertoli cell-specific Cx43 knockout, WT: wild type.
In order to identify the most significantly differentially expressed candidate genes in 10-day-old animals, the candidate gene list was sorted by p(FDR)-values and a literature review was conducted for the first 15 genes (Table 6). Noteworthy, all of these 15 genes were significantly downregulated in the 8-, 10-and 12-day-old SCCx43KO mice. Spg, decreases in Spc at leptotene/zygotene stage [55]; round and elongated Spd, localized at the blood-testis-barrier and near SC-GC interface [56]; and in human SCs and Leydig cells [57] Does not play a decisive role in murine gametogenesis [58] Possible role in GC movement along the seminiferous epithelium [ Moreover, functional classification of the candidate genes was performed for the molecular function, cellular component, protein class (Figure 7), biological process and pathways to assign the biological information of the transcriptome analyses using PANTHER [47,48,51].
Regarding the biological process, especially genes involved in 'cellular process', 'biological regulation' and 'metabolic process' were found to be highly represented. For this survey, all GO categories and associated genes are summarized in Table 7. Finally, the PANTHER pathway analysis yielded 46 pathways in relation to the candidate genes (Table S20). 'Gonadotropin-releasing hormone receptor pathway', 'p53 pathway' and 'CCKR signaling map' were the three categories with the most gene hits but the majority of PANTHER pathways were associated with four or fewer of the candidate genes ( Figure S2).

Confirmation of NGS Candidate Genes by qRT-PCR
In order to validate results obtained by NGS, five candidate genes that were found to be differentially expressed in the prepubertal SCCx43KO animals and related to spermatogenesis were selected for qRT-PCR analysis (Figure 8). 12-day-old age group. Furthermore, NGS revealed a significant downregulation of Gja1, Sohlh1 and Sohlh2 in SCCx43KO mice in each age group. For Gja1, Sohlh1 and Sohlh2, the relative abundance of total RNA was found to be lower in SCCx43KO animals in comparison with WT mice by qRT-PCR. Differences in Gja1 and Sohlh2 gene expression were significant in all three age groups, whereas Sohlh1 expression was significantly regulated in the 8-day-old age group. In general, qRT-PCR data were found to correlate with results received by NGS.  Among the chosen genes, SC-specific Amh and Fshr were found to be significantly upregulated by NGS analysis showing a decreasing p(FDR)-value with increasing age. Relative gene expression levels observed by qRT-PCR were higher in the SCCx43KO animals compared with their WT littermates for Amh and Fshr but differences in Amh gene expression were only significant in the 12-day-old age group. Furthermore, NGS revealed a significant downregulation of Gja1, Sohlh1 and Sohlh2 in SCCx43KO mice in each age group. For Gja1, Sohlh1 and Sohlh2, the relative abundance of total RNA was found to be lower in SCCx43KO animals in comparison with WT mice by qRT-PCR. Differences in Gja1 and Sohlh2 gene expression were significant in all three age groups, whereas Sohlh1 expression was significantly regulated in the 8-day-old age group. In general, qRT-PCR data were found to correlate with results received by NGS.

AMH
Amh gene, coding for the anti-Müllerian-Hormone, a factor synthesized by prepubertal SCs, was selected for validation because it was among the significantly and differentially upregulated genes in SCCx43KO mice at all three age groups and it is known as marker of immature SCs (reviewed in [83,84], see also [35]).
Immunohistochemical staining for AMH revealed a strong cytoplasmic signal in SCs of seminiferous tubules of 8-day-old SCCx43KO mice ( Figure 9A). In WT littermates (8 days p.n.), immunostaining was observed at the same localization but the immunoreaction was not that strong ( Figure 9B). SC cytoplasm in all seminiferous tubules of 10-day-old SCCx43KO mice also showed a distinct immunosignal ( Figure 9C), whereas in the 10-day-old WT littermates ( Figure 9D), the immunostaining weakened and single seminiferous tubules were found to be almost immunonegative for AMH (asterisks). In addition to SCs, these seminiferous tubules comprised first primary spermatocytes while immunopositive tubules so far only contained spermatogonia. In 12-day-old SCCx43KO animals, the cytoplasm of SCs was still found to be clearly immunopositive ( Figure 9E) but in seminiferous tubules of WT mice, a faint or almost no (asterisk) immunosignal was observed ( Figure 9F). The manufacturer recommends using mouse ovary tissue as a control tissue for AMH IHC.
Thus, mouse ovary tissue was added as a positive control to ensure a correct immunohistochemical staining against AMH ( Figure 9G). As negative controls, both mouse ovary ( Figure 9H) as well as testicular samples ( Figure S1C) were used. None of the negative controls showed any immunostaining ( Figure S1C).

LIN28A and SALL4
In order to examine the protein levels of two genes that are markers for undifferentiated spermatogonia (reviewed in [85]), LIN28A and SALL4 IHC were performed ( Figure 10). Almost all spermatogonia were found to be immunolabeled for LIN28A in the prepubertal KO mice with the exception of a few single GCs. In adult SCCx43KO mice, seminiferous tubules showed SCO tubules or an arrested spermatogenesis at the spermatogonial level. The remaining spermatogonia were immunopositive for LIN28A. Immunoreactivity for LIN28A was detectable in GCs in prepubertal and adult WT mice. Moreover, immunopositive cells for SALL4 were observed in both genotypes and at all ages. Except for the difference regarding GC numbers, no alterations in the staining patterns were detected between SCCx43KO and WT animals.

LIN28A and SALL4
In order to examine the protein levels of two genes that are markers for undifferentiated spermatogonia (reviewed in [85]), LIN28A and SALL4 IHC were performed ( Figure 10). Almost all spermatogonia were found to be immunolabeled for LIN28A in the prepubertal KO mice with the exception of a few single GCs. In adult SCCx43KO mice, seminiferous tubules showed SCO tubules or an arrested spermatogenesis at the spermatogonial level. The remaining spermatogonia were immunopositive for LIN28A. Immunoreactivity for LIN28A was detectable in GCs in prepubertal and adult WT mice. Moreover, immunopositive cells for SALL4 were observed in both genotypes and at all ages. Except for the difference regarding GC numbers, no alterations in the staining patterns were detected between SCCx43KO and WT animals.

SOHLH1
IHC was conducted to examine the protein expression level of a further gene, which is GC-specific [50] and which was significantly differentially expressed (downregulated) at the mRNA level in KO mice in comparison with WT littermates.
In seminiferous tubules of KO mice at all age groups, SOHLH1 protein was rarely detectable in some spermatogonia but the vast majority of this GC population did not show any immunoreaction ( Figure 11A,C,E), whereas in seminiferous tubules of WT littermates at all age groups, a strong immunostaining was observable in spermatogonia ( Figure 11B,D,F). In addition, adult SCCx43KO mice ( Figure 11G) and adult WT animals ( Figure 11H) were immunohistochemically examined. In adult seminiferous tubules of SCCx43KO mice, no immunosignal in GCs was detected. Immunopositive spermatogonia were observed in seminiferous tubules of adult WT mice according to previous findings [50]. SCs, Leydig cells and peritubular myoid cells appeared as constantly immunonegative in both genotypes. Negative controls showed no immunoreaction ( Figure S1D).

Discussion
Spermatogenesis is a highly complex process by which some spermatogonial stem cells undergo phases of mitotic proliferation and differentiation, meiotic division as well as transformation to give rise to mature sperm [86,87]. SCs assume the role of 'nurse cells', for example providing nutrients required by GCs and coordinating intraepithelial movement; in brief, SCs support spermatogenesis [88]. Thus, to enable successful spermatogenesis, a functional intercellular communication between SCs and GCs is necessary. In mammalians, intercellular communication either takes place through tubular structures (tunneling nanotubes (TNTs)) joining adjacent cells, through secreted molecules, which can be soluble or enclosed in vesicles, as well as through intercellular channels that connect the cytoplasm of neighboring cells [89]. Recently, Cx43 has been shown to be involved in all these types of intercellular communication (reviewed in [89]).
Hence, the current study further investigates the intriguing functions of Cx43 in SCs for spermatogenesis and SC maturation. As it has recently been reported, the expression pattern of Cx43 is altered in different forms of human testicular disorders as well as in preinvasive GCNIS and TGCTs developing therefrom [6][7][8][27][28][29]90]. Establishing a SCCx43KO mouse model provided a good opportunity for studying underlying molecular mechanisms leading to spermatogenic impairment [25,26]. A previous microarray-analysis has already shown that the gene expression of many genes is affected in 8-day-old mice by the deletion of Cx43 [9]. In order to examine the effects of Cx43 loss over a greater period of time and more precisely, supplementary to 8-day-old animals, 10and 12-day-old mice were used.
In addition to PCR genotyping and morphological evaluation by H&E staining, IHC for β-galactosidase and Cx43 were conducted to determine and confirm murine geno-/phenotypes. Observed β-galactosidase immunostaining of SC nuclei in KO mice of the different age groups ( Figure 1A-C) again confirmed successful deletion of Cx43 gene because transcription of the LacZ-reporter gene is only possible when the previous gene sequence (Cx43 coding region) is absent [91,92]. These results are in accordance with observations of Cx43 IHC (Figure 2). Thereby, Cx43 expression was not detectable in any KO mice of each age group, again confirming the successful lack of Cx43 in SCs.
Furthermore, to clearly distinguish between GCs and SCs in the present study, SOX9 IHC was used to mark SC nuclei and then to determine intratubular cell numbers revealing reduced GC numbers in KO mice pursuant to recent findings [9,93]. However, in contrast to Giese et al. [9], increased SC numbers per seminiferous tubule have already been detected in 8-day-old KO mice and become more evident in 10-and 12-day-old KO animals (Figure 3), implying an enhanced SC proliferation already at this stage. Thus, these present findings appear to be a reasonable explanation for finally increased SC numbers in adult KO mice [26] as well as for the observed adult SC proliferation [25].
In order to analyze and compare the testicular transcriptome of prepubertal KO mice and their WT littermates, NGS was performed, because it has been shown to be a more precise method than others [94]. As suspected, transcriptome analysis revealed much more significantly and differentially expressed genes in 8-day-old KO mice in comparison with the previous microarray study (after GC-specific normalization: 4253 versus 658 genes) [9]. However, former candidate genes, such as Stra8 and Dazl, were also found to be significantly regulated, confirming previous findings [9]. The greatest changes in differentially expressed genes (34,325 genes after GC-specific normalization) were detected by comparison of the 10-day-old KO mice with their WT littermates. These findings are according to the observed morphologic differences in the composition of intratubular cell populations between KO and WT mice. As described by Bellve in 1977 [32], spermatocytes were first found in 10-day-old WT mice, while in their KO littermates, only spermatogonia were present. By comparing the 10-day-old WT group with the 8-day-old WT mice, 32,042 genes were found to be significantly regulated (after GC-specific normalization), whereas the comparison of 10-day-old KO mice with the 8-day-old KO group yielded 532 significantly differentially expressed genes indicating big differences in gene expression in aging WT animals in relation to the KO mice and an impairment or delay on the KO side. These findings appear to be consistent with the occurring morphologic differences in WT mice (10 days p.n. versus 8 days p.n.) and the minor changes in KO littermates. Moreover, with advancing age, the numbers of significantly differentially expressed genes are lower in WT mice as in KO mice (12 days p.n. versus 8 days p.n.). Rising SC numbers in KO mice may explain these results, but further research is required to investigate the gene expression in an age-dependent manner.
Based on the generated NGS data, different hypotheses are conceivable to explain the development of arrested spermatogenesis at the spermatogonial level and infertility in SCCx43KO mice: (1) retinoic acid (RA) signaling might be altered; (2) maturation of SCs might be affected by the Cx43 deletion, leading to a sustained 'immature' cell fate which prevents SCs from completing functions required for successful spermatogenesis; (3) spermatogonial stem cell self-renewal might be disturbed; (4) differentiation of spermatogonia might be altered; (5) initiation of meiotic prophase I might be impaired; and (6) cell-to-cell communication between SCs and GCs might be interrupted.
In order to gain an overview of gene sets affected by the Cx43 deletion, GO enrichment analysis was performed, highlighting primarily processes related to meiosis. Thus, this seemed to be consistent with the observed phenotypes of KO animals that show an arrested spermatogenesis at the level of spermatogonia, and additionally, supported the valid experimental design.
Meiosis describes the production of haploid cells from diploid progenitor cells, and formation of preleptotene spermatocytes is considered as the initiation of this process [54]. In this case, the transition of undifferentiated A-spermatogonia into A1 differentiating spermatogonia could be regarded as commitment to meiosis [54]. For the progression into differentiating A1-spermatogonia, at least RA is required and subsequently STRA8 protein has been shown to be essential for inducing meiosis [54,62,63,95].
Hence, one hypothesis to explain the observed testicular phenotype in SCCx43KO mice might be that the deletion of Cx43 in SCs alters RA signaling, whereby the initiation of meiosis is disturbed and consequently results in arrested spermatogenesis at the spermatogonial level. Genes involved in RA signaling were found to be significantly differentially expressed in the prepubertal SCCx43KO mice (e.g., Stra6, Stra8, Rarg, Cyp26b1 Crabp1, Crabp2 Aldh1a1 and Aldh1a3).
In the seminiferous epithelium, both SCs and spermatogonia have RA receptors (RARs) and enzymes that are required for the oxidation of retinol to RA [54]. Noteworthy, in the testis, the only putative receptor that is able to take up the circulating form of retinol (ROL-RBP4-TTR) appears to be STRA6, which is expressed primarily in SCs (reviewed in [54]). Thus, SCs are assumed as source of RA and then undifferentiated GCs represent the target cells [54]. However, it has been suggested that preleptotene spermatocytes may be another source of retinoic acid at the time of meiotic initiation [96]. Meanwhile, SC-derived RA has been shown to be pivotal for the initial differentiation of spermatogonia and the onset of meiosis in juvenile mice but not necessary after the first wave of spermatogonial differentiation [96,97]. RA promotes the spermatogenic process, on the one hand by promoting the expression of spermatogonial stem cells to differentiating spermatogonia through activation of KIT expression, and on the other hand by promoting the expression of STRA8 [95,98], thus inducing meiosis [63].
Stra8/STRA8 expression has previously been examined in prepubertal and adult SCCx43KO mice in comparison with their WT littermates and found to be altered or absent [9,93]. NGS data of the present study revealed a significant downregulation of Stra8 in 8-, 10-, and 12-day-old SCCx43KO mice compared to their WT littermates. Additionally, Stra6, Aldh1a1, Aldh1a3 and Cyp26b1 were significantly upregulated, while Rarg, Crabp1, Crabp2 were downregulated. These findings might indicate alterations in the RA signaling pathway caused by the absence of Cx43. In particular, it is thinkable that altered expression of cellular retinol-binding proteins leads to problems regarding storage, cell uptake and transportation in the prepubertal SCCx43KO mice. Furthermore, significant upregulation of Cyp26b1 may imply an induced degradation of RA and in the following missing stimulation of RA-target genes, for example SOHLH1 and STRA8, which in turn may lead to altered spermatogonial differentiation and impaired meiosis in the mutant mice. Upregulation of Stra6, Aldh1a1 and Aldh1a3 may then be the response of reduced RA levels.
In addition, Rarg (encoding the RARγ receptor) was found to be significantly downregulated in the 8-, 10-and 12-day-old SCCx43KO mice. Recently, it has been suggested that this receptor is pivotal for spermatogonial differentiation after the first wave of spermatogenesis; as in mutant mice, the transition of A al into A 1 spermatogonia has been shown to be impaired [99]. During the first wave of spermatogenesis, meiotic cells appear but adult mice lacking RARγ show an arrested spermatogenesis at the spermatogonial level or SC-only tubules [99] comparable to the adult phenotype of the SCCx43KO mice [25,26]. Therefore, alterations of Rarg gene expression might also contribute to the disturbed spermatogenesis in SCCx43KO mice.
So far, various studies of different tissues, including testicular tissue, demonstrated that RA also stimulates the expression of Cx43 [17,[100][101][102][103][104]. The other way round, it has been shown that defective RA signaling leads to altered gap-junction-based cell coupling as well as to alterations in tight junction formation and increased blood-testis barrier permeability [105][106][107]. However, both the exact mechanism and the link to Cx43 still remain unclear and further research needs to be performed to address this issue.
Moreover, PANTHER pathway analysis identified 133 pathways for all significantly differentially expressed genes and 46 pathways for the compiled candidate gene list that are affected by Cx43 loss. Interestingly, the pathways with the most gene hits were identical. Hence, this underlines the importance of Cx43 for these pathways and direct cell-cell communication. Furthermore, it strengthens the value of the established candidate gene list for further evaluation and for identifying possible candidate genes for future examination of corresponding human deficiencies.
For example, the 'gonadotropin-releasing hormone receptor pathway' showed the most gene hits (e.g., Amh, Gata4, Inha, and Inhbb). Gonadotropin-releasing hormone (GnRH) receptor signaling plays a key role in the reproduction of vertebrates [108]. Pulsatile GnRH release from the hypothalamus activates GnRH receptors in gonadotrophs of the pituitary gland leading to the secretion of e.g., follicle-stimulating hormone (FSH) [108,109]. FSH binds receptors on SCs, stimulating various cellular mechanisms, e.g., protein synthesis by SCs and SC growth [109]. In addition to others, inhibin is produced and secreted by SCs, which negatively influences hormone release in the hypothalamus and pituitary gland [84].
Another hypothesis to explain the testicular phenotype of SCCx43KO mice might be that the loss of Cx43 alters SC proliferation/maturation in the mutant mice, resulting in arrested spermatogenesis at the spermatogonial level due to asynchronous differentiation of both SCs and GCs. This assumption is supported by the observation of SCs possessing immature characteristics in mice and men with impaired spermatogenesis [35,[110][111][112][113][114][115].
Inha (coding for inhibin A and inhibin B), Inhbb (coding for activin B) and Amh are all known to be crucial for functional regulation of reproduction and development [116]. Especially during the onset of the first wave of spermatogenesis, the modulation of synthesis and action of inhibins and activins appear to be important [117]. In contrast to activin A, which has been shown to be highly concentrated in testes during SC proliferation in newborn mice and whose concentration decreases in aging mice to a low level in adult animals, concomitantly with decreasing SC proliferation, less is known about activin B [116,117]. NGS data revealed a significant upregulation of Inhbb in SCCx43KO mice in all three age groups in the present study. Considering the increase in SC numbers in SCCx43KO compared to their WT littermates, these results might indicate that activin B is also involved in SC proliferation.
Interestingly, Inha was also found to be significantly upregulated in SCCx43KO mice at each investigated age. Physiologically, Inha mRNA levels progressively decrease to 30% of the 0 days p.n. value within six, ten and 12 days p.n. [117]. Total testicular levels of inhibin have been shown to increase from 0 days p.n., peaking at four days p.n., showing a plateau to Day 16 p.n., and subsequently decreasing, thereby reflecting a period of SC proliferation followed by achieving a stable, non-mitotic SC population [117]. Taking these findings [117] into account, the observed Inha upregulation in SCCx43KO mice compared to their WT littermates might indicate a prolonged and enhanced period of SC proliferation in the KO animals.
Additionally, Amh was found to be significantly upregulated in SCCx43KO mice compared to their WT littermates in all three age groups by NGS. Using qRT-PCR, this upregulation has been confirmed ( Figure 8). However, Amh gene expression was not yet significantly altered in 8-and 10-day-old KO mice (performing qRT-PCR) underlining the precise measuring of transcript levels by NGS. Consistent with these data, an altered AMH protein expression pattern was observed in KO mice (Figure 9) according to Weider et al. [35]. AMH is the first protein secreted by fetal SCs and its expression is maintained until puberty [118,119]. In particular, androgens are known to regulate AMH expression by SCs [119,120], and also maturation of GCs negatively affects AMH production, whereas AMH secretion is stimulated by FSH in the absence of the inhibiting effect of androgens [121]. Prolonged AMH expression in the KO mice suggests that SCs are in a sustained 'immature' state of proliferation and therefore unable to fulfil nurse-like and scaffolding roles required for successful spermatogenesis. This suggestion is supported by the upregulation of further genes, such as the above-mentioned genes (Inha and Inhbb) as well as Thra, Krt18 (markers for immature SCs [113,122]), Fshr, Fgf9 and Kitl (also known as SCF).
FGF9 is a meiotic inhibiting substance normally produced by Sertoli cells to prevent gonocytes from entering meiosis [123]. SCs also produce KIT ligand (KITL), which has been reported to stimulate KIT expression of differentiating spermatogonia thereby promoting spermatogonial differentiation and gene expression of genes that are specific for early meiotic phases [123]. In the SCCx43KO mice, Fgf9 (ten and twelve days p.n.) and Kitl (eight, ten and twelve days p.n.) were found to be significantly upregulated while Kit (also known as c-kit) was significantly downregulated in 8-, 10-and 12-day-old animals. Hence, imbalances of SC signaling might be traced back to altered SC maturation in consequence of Cx43 absence and, together with the detected alterations of Kit expression, these might be involved in the development of the observed testicular phenotype in SCCx43KO animals.
Interestingly, in this case, Gata4 was significantly upregulated in the prepubertal SCCx43KO mice, too. SCs express GATA4 from the onset of their differentiation through to adulthood (reviewed in [124]). Among others, Dmrt1, Inha, Inhbb and Amh have been identified as putative target genes for GATA4 in the testis (reviewed in [124]). Thus, upregulated Gata4 gene expression in the SCCx43KO animals may contribute to the differential gene expression of Amh, Inha and Inhbb in the transgenic mice.
Furthermore, Dmrt1 gene knockout leads to disorganized seminiferous tubules, pre-meiotic GC death and abundance of immature SCs [125]. In 7-day-old Dmrt1 KO mice, Gata1 mRNA levels have been found reduced, Gata4 and Krt18 mRNA levels increased [126]. In this context, it has been suggested that DMRT1 plays an important role for SC maturation and its absence prevents cessation of SC proliferation as consequence of failed terminal differentiation [127]. As Dmrt1 KO mice have a similar testicular phenotype and significantly, downregulated Dmrt1 transcripts were observed in the prepubertal SCCx43KO mice, it seems likely that loss of Cx43 may alter SC proliferation via a same/similar regulatory mechanism resulting in impaired spermatogenesis.
Moreover, the candidate gene list drew attention to Rb1 (also known as Rb) that encodes the retinoblastoma-associated protein, a key regulator of cell cycle entry as well as cellular division, and a tumor suppressor [128,129]. It was generally assumed that adult SCs are terminally differentiated. Recent studies, however, supposed that SCs are rather in a state of continuous cell cycle repression and that the non-proliferative state of adult SCs is maintained by RB1 [129][130][131]. Rotgers and colleagues have shown that repression of the transcription factor E2F3 by RB1 is decisive for the maintenance of cell cycle quiescence in adult SCs [122]. Loss of Rb1 leads to impaired cell cycle control in the form of cell cycle re-entry mediated by E2F3 and hence to increased SC proliferation and loss of SC function [122,131]. Furthermore, follistatin (Fst) expression was found to be increased in the absence of Rb1, which seems to be due to a potentially E2F3-mediated mechanism [122].
With regard to the SCCx43KO mice, NGS data revealed a significant downregulation of Rb1 in all age groups, no changes in E2F3 gene expression and a significant upregulation of Fst in 10-and 12-day-old KO mice, suggesting a direct relation between Cx43 and Rb1 expression and that the cell cycle progression of SCs might be traced back to this regulator mechanism. This again supports the assumption of a sustained 'immature' state of SC differentiation/maturation which might cause the spermatogenic arrest at the spermatogonial level or SC-only tubules in SCCx43KO mice.
Interestingly, alterations of both Cx43 and RB1 seem to be associated with cancer development [132][133][134][135]. However, it remains the question as to how precisely Cx43 affects Rb1. This appears to be an interesting approach for further research, and also with regard to the pathogenesis of TGCTs as a deregulation of RB pathway has already been shown in humans [136][137][138].
Apart from the impact on SCs, SC-specific Cx43 deletion leads to great changes in GC-specific genes and to an arrested spermatogenesis at the level of spermatogonia in prepubertal mice. As expected, based on the histological findings in 10-day-old KO mice, most differences in gene expression were observed compared to their WT littermates. Predominantly, genes associated with meiosis (e.g., Sycp1 and Spo11) and spermatid markers such as Piwil1 were found to be significantly downregulated in the KO mice. In order to ensure that these alterations in gene expression were not just caused by the absence of corresponding GC populations, NGS data were normalized by means of the determined cell numbers. Even after this normalization, genes related to meiosis or haploid GC differentiation were significantly altered in KO animals. Interestingly, some genes that are specific for undifferentiated spermatogonia (e.g., Lin28 and Sall4, see review [85]) also showed downregulated transcript levels in the prepubertal SCCx43KO mice, suggesting alterations already at this early GC stage.
Rode and colleagues [93] reported that during puberty, the GC component of SCCx43KO mice comprised for the most part undifferentiated spermatogonia and that in adult KO animals, spermatogenesis is arrested at the level of undifferentiated spermatogonia, observed alterations in gene expression may indicate that loss of Cx43 in SCs already influences a subpopulation of undifferentiated spermatogonia. However, apart from the different GC numbers, no obvious differences could be detected for LIN28A and SALL4 expression on protein level between the KO mice and their WT littermates. Additionally, testicular samples of adult KO and WT mice were immunostained for LIN28A and SALL4 ( Figure 10). According to the previous study [93], remaining GCs in seminiferous tubules of adult SCCx43KO mice were identified as undifferentiated spermatogonia. Thus, prepubertal SCCx43KO mice seem to have significantly reduced transcript levels of Lin28 and Sall4 but still a sufficient amount of these for translation into detectable protein volumes.
Notably, further genes implicated in spermatogonial stem cell self-renewal and maintenance were found to be significantly altered in the SCCx43KO mice in comparison with their WT littermates. These include, for example, Zbtb16 (also known as Plzf ), Taf4b and Atm, which were found to be significantly downregulated, as well as Gdnf, which was significantly upregulated in 8-, 10-and 12-day-old SCCx43KO mice. Therefore, another hypothesis might be that Cx43 deletion causes alterations of intrinsic and extrinsic factors involved in the regulation of spermatogonial stem cell self-renewal and maintenance, resulting in the observed testicular phenotype of SCCx43KO mice.
Regarding this hypothesis, over-expression of the SC-derived factor GDNF has been reported to inhibit germ cell differentiation, resulting in an accumulation of undifferentiated spermatogonia [139], hence supporting our hypothesis, as SCCx43KO mice exhibit a spermatogenic arrest at the level of undifferentiated spermatogonia [93].
Moreover, in the hematopoietic stem cell (HSC) niche, Cx43 has already been identified as a main regulator of cell-cell communication between HSCs and stromal cells as well as for the connection with the extracellular medium via hemichannels thereby regulating HSC self-renewal [140][141][142]. In this context, it is worth mentioning that embryonic mice with a general Cx43 KO show reduced fetal liver HSCs and progenitor cells (reviewed in [141]). Additionally, the Cx43 lack in stromal cells leads to alterations in HSCs and progenitor cells in the fetal liver, preventing successful hematopoiesis (reviewed in [141]). In addition, Cx43 seems to play key roles in the neural stem cell and the skin stem cell niche (reviewed in [141]. Thus, it appears likely that Cx43 also plays a relevant role in the spermatogonial stem cell niche. Future studies are required to address this interesting issue. Furthermore, with the aid of the created candidate gene list, promising candidate genes such as Sohlh1 and Sohlh2 were highlighted. NGS data of selected genes of interest could be verified using qRT-PCR and/or IHC (Figures 8 and 11). The GC-specific transcription factors Sohlh1 and Sohlh2 play a pivotal role in spermatogonial differentiation [50,143]. As SCCx43KO mice, homozygous Sohlh1 and Sohlh2 KO mice are infertile and show similar phenotypes, in particular, an arrested spermatogenesis at the level of spermatogonia [50,73,74]. Thus, disturbed Sohlh1 and Sohlh2 gene expression appears to be an additional explanation for the testicular phenotype of the SCCx43KO mice.
In a recent study, however, fewer numbers of spermatocytes were also detected in the seminiferous tubules of Sohlh2 KO mice, and SOHLH2 was found to be crucial for synaptonemal complex formation by regulating Sycp1 expression during spermatogonial differentiation and therefore important for progression of meiosis [76].
Moreover, it has been shown that SOHLH proteins act as heterodimers and homodimers and directly regulate the transcription of Gfra1, Sox3, Sohlh1, Sohlh2 and Kit, genes which are pivotal for spermatogonial development and differentiation [143]. In addition, SOHLH1 and SOHLH2 suppress genes implicated in spermatogonial stem cell maintenance/self-renewal, such as Ret, Nanos2 and Pou5f1, affect genes implicated in RNA metabolism, for example Lin28, and induce the expression of genes associated with meiosis, such as Dmrtc2 (also known as Dmrt7) and Piwil1 [143]. A further notable target gene is Stra8, which is directly downregulated by both SOHLH1 and SOHLH2 [144]. Consequently, SOHLH proteins are believed to coordinate spermatogonial differentiation and RA-induced meiosis in cooperation with other transcription factors [144]. In this context, Dmrt1 and Dmrtb1 (also known as Dmrt6) should be mentioned because they have been assumed to play an essential role in regulating meiosis and Stra8 expression [33,[145][146][147]. Additionally, Fignl1 has been proposed to play an important role in controlling meiosis in mice [148].
Remarkably, in the SCCx43KO mice, Sohlh2 and Dmrtb1 were among the most significantly altered genes and above-mentioned target genes of the SOHLH proteins (except for Ret, Nanos2 and Pou5f1) as well as Fignl1 were also significantly differentially expressed in the prepubertal SCCx43KO mice. Hence, a lack of Cx43 in SCs seems to have great impact on genes, which are essential for the necessary mitosis-meiosis switch.
As murine Dmrtb1 was already investigated in 8-day-old SCCx43KO mice [9], and a successful translation of previous findings [9,33] by means of examining human testicular biopsy specimens of corresponding human pathologies [34] was possible, the current study focused on investigating Sohlh1 and Sohlh2. NGS data were confirmed by qRT-PCR. In addition, SOHLH1 protein was immunolocalized in spermatogonia of WT animals, but was rarely detectable in SCCx43KO littermates, which is again consistent with NGS data. Therefore, alterations of genes required for spermatogonial differentiation, especially of Sohlh1 and Sohlh2 expression, might provoke the spermatogenic arrest in the SCCx43KO mice.
Furthermore, SOHLH1 has been shown to be exclusively expressed in pre-meiotic cells in the testis of rhesus monkeys [149]. Its translocation from the cytoplasm into the nucleus at the onset of puberty has been associated with the initiation of spermatogenesis [149]. This also implies a role of SOHLH1 for regulating spermatogonial differentiation in primates [149].
Taken together, SOHLH1 and SOHLH2 appear to be interesting candidate genes for investigating corresponding human deficiencies. At the genetic level, mutations in SOHLH1 gene have already been associated with non-obstructive azoospermia [150,151] as well as genetic variants in SOHLH1 and SOHLH2 with non-obstructive azoospermia risk in Chinese men [152]. In females, too, crucial roles of SOHLH2 have been discovered, for example in the etiology of human premature ovarian failure and in the context of ovarian cancer [153][154][155]. Nevertheless, to the best of our knowledge, no reliable data about the protein localization in human testis have been published so far. Thus, further research in this field is necessary and appears to be promising.
With regard to Cx43, the data of this study indicate that Cx43 in SCs might be involved in regulating the above-mentioned genes in GCs at a subordinate level. Concerning this matter, intercellular communication through TNTs seems to be interesting. TNTs allow the exchange of various materials, such as miRNA, proteins, vesicles derived from the endoplasmic reticulum and Golgi complex, endocytic vesicles and lysosomes, between connected cells [89,156,157]. Their biological relevance has been shown in diverse processes, including embryogenesis, stem cell differentiation and cancer progression, and Cx43 has been associated with TNTs [156,[158][159][160][161][162][163][164].
As in SCs of SCCx43KO mice Cx43 is absent, it might be possible that Cx43-deficient SCs are unable to communicate with other cells through TNTs. In addition, alpha-tubulin has been identified as microtubule component of TNTs in human prostate cancer cells and Ezr as a constituent protein of these intercellular bridges in the rat testis [165,166]. Previous studies have shown that Cx43 directly interacts with alpha-and beta-tubulin [167,168]. For example, Tuba3a, Tuba3b, Tuba4a and Tubb4b have been found to be significantly downregulated in 8-, 10-and 12-day-old and Tuba1c, Tubb1, Tubal3, Tubb4a, Tuba8 as well as Ezr in 10-day-old SCCx43KO mice, which might also suggest that cell-to-cell communication via TNTs might be disturbed in the SCCx43KO mice.
For various reasons (reviewed in [89]), investigating TNTs might be challenging but an auspicious approach.
Nevertheless, the first interesting findings about TNTs implicated in spermatogenesis have been reported in the Drosophila testis as well as in connection with the ectoplasmic specializations (ES) within the seminiferous epithelium of rat testes [169,170].
Additionally, Cx43, Cx45 and Cx32 have been shown to be involved in a new mechanism for rapid and efficient cell-to-cell communication by mediating the interaction of extracellular vesicles with target cells and/or the transfer of miRNAs and small molecules [89]. Apart from Gja1 (encoding Cx43), significant alterations in Gjc1 (coding for Cx45) and Gjb1 (encoding Cx32) gene expression have also been detected in the SCCx43KO mice (Table S21), which might suggest that this form of cell-to-cell communication could be disturbed and might contribute to the development of the observed SCCx43KO phenotype. In order to address this question, future studies are necessary.
Moreover, the 'classical' functions of Cx43 as gap junction protein should not be forgotten. It seems that the reported selective and unidirectional transfer between SCs and spermatogonia [171][172][173] is altered in the SCCx43KO mice and might explain the arrested spermatogenesis in the mutant mice. Based on these permeability studies, gap junctions between adjoining SCs appear to differ in Cx composition from those between SCs and GCs [171][172][173][174]. This suggestion is supported by the impaired spermatogenesis found in SCCx43KO mice, whereas GC-specific Cx43 KO results in functional spermatogenesis and fertile mice [25,26,175]. Therefore, it has been assumed that spermatogonia form a heterotypic connexon [174,175].
Furthermore, the role of Cx43 protein itself seems to be promising. A loss-of-function study indicates that another property of Cx43, other than its ability to form gap junctions, appears to be responsible for its indispensability in germ line development [174,176]. Moreover, recent studies have reported additional functions of Cxs, such as regulating gene transcription and interacting with cell growth and cell death modulators, as well as having chemical roles [177][178][179]. Thus, further research on the different functions of Cx43 is required.
All things considered, this study provided different approaches regarding the underlying mechanisms resulting in the testicular phenotype of the SCCx43KO mice. These hypotheses need to be further investigated in future studies.

Conclusions
In conclusion, this study demonstrates that loss of Cx43 in SCs leads to significant genetic as well as morphologic changes already in prepubertal SCCx43KO mice in comparison with their WT littermates. More precisely, Cx43 in SCs seems to be essential for the normal progression of the first wave of spermatogenesis, especially for the mitosis-meiosis switch. Additionally, its expression appears to be required for successful regulation of prepubertal SC maturation. These findings seem to be in accordance with findings in corresponding human pathologies, thus the transgenic SCCx43KO mouse line proves to be a useful translational model for investigating and understanding the possible causes leading to impaired spermatogenesis and human male factor infertility.