Aging, inflammation and DNA damage in the somatic testicular niche with idiopathic germ cell aplasia

Single-cell transcriptome profiling revealed seven somatic cell clusters in iGCA

In iNOA men with negative sperm retrieval at microdissection TEsticular Sperm Extraction (microTESE), the presence of seminiferous tubules lacking any germ cell configured the histological diagnosis of GCA (Fig. 1a). Conversely, the histological analysis of a case of obstructive azoospermia (OA), associated with cystic fibrosis transmembrane conductance regulator (CFTR) gene mutation, showed a condition of normal spermatogenesis, with sperm cells at different stages of maturation in the seminiferous tubules (Fig. 1a).

To investigate the molecular pathogenesis of iGCA and provide a comprehensive description of the testicular somatic cell populations, we performed single-cell RNA-sequencing (scRNA-seq) on freshly isolated tissue specimens from microTESE. A total of 3880 cells, obtained after filtering out poor-quality cells from 3 iGCA donors (Supplementary information, Fig. S1a–1d, Dataset S1), were integrated to reduce batch effects, and the Seurat standard graph-based clustering approach was used for cell partitioning and cluster identification. Overall, 8 cell clusters were identified as shown in the Uniform Manifold Approximation and Projection (UMAP)¹⁴ plot (Fig. 1b). Using previously determined cell type marker genes^15,16 (Supplementary information, Dataset S2), the main somatic cell populations were identified, Leydig (LEY), myoid (MYD), Sertoli (SRT), and endothelial (END) cells (Fig. 1c; Supplementary information, Fig. S1e–1j, Dataset S1). Furthermore, immune cells were also identified, such as macrophages (MCR) and T-cells (TCL) (Fig. 1c; Supplementary information, Fig. S1e–1j). Based on the expression of the marker genes RGS5 and TPM2 we assigned the stromal (STRO) cluster to pericytes^17,18 or vascular smooth muscle cells according to the expression of FABP4^19,20(Fig. 1c; Supplementary information, Fig. S1j), which represent progenitors stem cell of LEY cells²¹. The relative abundance of somatic cell populations (Supplementary information, Fig. S1k), along with the lack of germline markers expression (Supplementary information, Fig. S1l), confirm the absence of germ cells at any stage throughout spermatogenesis in men with iNOA. Of all, one cluster of cells remained undetermined (UND) (Fig. 1b), since it lacked clear maker genes and had a low number of RNA features (Fig. 1d; Supplementary information, Fig. S1m).

To depict the unique characteristics of each somatic cell population in iGCA testis, we applied functional enrichment analysis to each cluster (Fig. 1d; Supplementary information, Dataset S3–10). Pathways specific for LEY and MYD cells included “collagen biosynthesis” and “extracellular matrix (ECM) organization”. SRT cells were enriched in the “pregnenolone biosynthesis” pathway, the first substrate along the steroid hormones synthesis pathway²². In MCR and TCL groups we found highly enriched pathways related to activation of the immune system that include “cytokine-mediated signaling”, and “PD-1 signaling” pathways. Finally, in END and STRO cells, “vasculogenesis” and “extracellular matrix organization” pathways were enriched, respectively.

Single-cell transcriptomics of somatic cell clusters in normal spermatogenesis

In parallel, scRNA-seq of testicular tissue of one man with OA associated with CFTR gene mutation was used as a positive control for cell cluster identification (Supplementary information, Fig. S2a, Dataset S1). By applying the same cell partitioning protocol used for iGCA samples, we also analyzed published data from three autoptic testicular tissues with normal spermatogenesis found in a published study¹⁵ which reported data for 6490 cells, 1676 of which were somatic (Supplementary information, Fig. S2b, Dataset S11). After cluster identification, OA cell types were assigned based on the expression of somatic and germ cell marker genes, according to the consolidated descriptions of testicular cell type¹⁵. Cells identity was also characterized by projecting the PCA structure of the reference dataset from Guo and colleagues¹⁵ onto both OA and iGCA datasets. The unsupervised graph-based clustering approach and the projection of reference data led to a similar outcome in terms of cluster identity, with the majority of cells isolated from OA testis being ascribed to the spermatogenic pathway (Fig. 1e; Supplementary information, Fig. S2a), while only somatic cells were found in iGCA testis (Fig. 1f).

Cell-specific upregulated pathways in somatic cell populations associated with germ cell aplasia

To unravel differences in terms of somatic testicular environment, we compared gene expression in each somatic population from iGCA testes and from testes with normal spermatogenesis, respectively. Data from OA testis have not been used as a reference for normal tissues since the number of somatic cells in this sample was low and not all the niche populations were fully represented; moreover, in OA patients bearing CFTR mutations, there is overexpression of genes in the inflammatory response pathway²³ that might affect the enrichment analyses. To enable comparison and exclude batch effects, the datasets from our 3 iGCA testes were integrated with those from the five testicular samples with normal spermatogenesis retrieved from two different datasets, Guo¹⁵ and Sohni²⁴. Cell partitioning was performed as described above, and cluster identity was assigned based on the expression of somatic cell marker genes (Supplementary information, Fig. S3, Dataset S12), with 7 somatic cell types clearly identified (Fig. 1g). We observed that the relative abundance of the somatic cell populations varied between iGCA and samples with normal spermatogenesis, and the TCL cluster was only present in the iGCA testes (Fig. 1h).

Functional enrichment analysis was used to identify those gene ontologies and pathways upregulated in testicular somatic cells from iGCA (Supplementary information, Fig. S4, Dataset S13–19). As for MYD, LEY, and STRO cells, the top upregulated pathways were related to ECM deposition and organization. Further upregulated pathways common to MYD and LEY cells were also “Cholesterol biosynthesis” and “Scavenging by Class A Receptors (SR-A)”, which included the overexpression of SCARA5 in LEY cells as previously reported²⁵. “Antigen presentation” and “Pre-NOTCH transcription and translation” pathways were upregulated in MYD cells, while “Nuclear-transcribed mRNAs catabolic process” and “Cellular response to corticosteroid stimulus” were specifically upregulated only in LEY cells (Fig. 2a–c). Upregulated pathways in SRT cells were related to energy metabolism, including the glyceraldehyde-3-phosphate metabolic process (Fig. 2d).

The analysis of resident immune cells revealed a drift toward a pro-inflammatory phenotype. In MCR we found upregulated pathways related to MHC class II antigen presentation, interferon-gamma, and NF-kB signaling (Fig. 2e; Supplementary information, Fig. S4). CD8+/CD69+ T-cells were identified in iGCA testes but not in the datasets of control testes from Guo¹⁵ and Sohni²⁴ (Fig. 2f). In addition, projecting the iGCA transcriptome onto a dataset of T cells isolated from human tissues²⁶ confirmed the phenotype of the T-cells from iGCA testis as tissue-resident CD8 + cytotoxic T cells (Fig. 2g). Furthermore, CD8 + CD69 + TCL expressed granzyme K and M, the pro-inflammatory chemokines CCL4/MIP1β and CCL5/RANTES (Fig. 2h) and cytokine IL-32 (Supplementary information, Dataset S12), supporting the conclusion that these somatic tissue-resident cytotoxic T cells might be generated by a previous inflammatory or autoimmune insult and contribute to chronic local inflammation. All TCL in the iGCA were cytotoxic T cells; conversely, a recent dataset on human testes with normal spermatogenesis showed the presence of a few T-cells, of which only 23% of expressed CD8²⁷.

To investigate the differences in cellular communication between iGCA and healthy donor, we performed an unbiased ligand–receptor interaction analysis between these testicular cell subsets by CellphoneDB²⁸. Overall, we found that the total number of interactions was significantly increased in iGCA donors, especially among MCR, LEY, and MYD cells (Fig. 2i, Supplementary information, Dataset S20). The increased interaction was prompt by the overexpression of the cell surface receptor CD74 in iGCA MCR (Fig. 2j, k).

Cell-specific downregulated pathways in somatic cell populations associated with iGCA

Fifty-six downmodulated pathways were common to all somatic cell populations in iGCA, with the aforementioned exception of the TCL cluster (Fig. 3a; Supplementary information, Fig. S5, Dataset S21–28). Most of those pathways were related to protein metabolism and ribosomal biogenesis. In all somatic iGCA populations, we also found downregulation of the selenocysteine synthesis and selenoamino acid metabolism pathways. In line with these findings, we identified the downregulation of INMT, an enzyme involved in selenium metabolism and detoxification²⁹ in LEY cells, and the glutathione peroxidases Gpx1 and Gpx4 (Fig. 3b) in SRT cells. Since selenium is also involved in redox detoxification of the extracellular milieu, we wondered whether also redoxin gene family was differentially expressed. The transcripts for several selenoproteins were downregulated in all iGCA somatic cells (Fig. 3b; Dataset S29), including the SELENOP (i.e., the most important and abundant selenoprotein in plasma with extracellular antioxidant functions) and SELENOH (which has redox capabilities linked to genome damage and DNA stress and a nucleolar localization)³⁰.

Other downregulated pathways of particular interest were the mitochondrial respiratory chain complex and the oxygen-dependent proline hydroxylation of hypoxia-inducible factor (Fig. 3c; Supplementary information, Dataset S29).

Yet not exclusively, the common pathways were mainly guided by significant downregulation of ribosomal protein genes expression across all cell types. Indeed, a differential expression of ribosomal genes has recently been reported in the somatic cells of fetal and adult human testis and associated with a differentiated metabolic activity²⁴. To exclude the presence of potential artifacts due to different library preparations, we quantified the ribosomal genes expression global levels in samples from our cohort and from the Guo¹⁵ and Sohni²⁴ datasets; overall, ribosomal protein genes were overexpressed in somatic cells compared to germ cells and the read coverage onto ribosomal protein genes was comparable in our OA sample and the ones from Guo and Sohni, thus pointing to a real biological difference rather than batch effects (Supplementary information, Fig. S6a).

The same up and downregulated pathways identified by comparing somatic cells from iGCA vs. CTL testis were identified through the comparison of iGCA vs. the testis with obstructive azoospermia (Supplementary Fig. S6b–e).

Transcriptional and phenotypic immaturity state of LEY cells in iGCA

Paracrine cell-cell signaling is involved in the control of germ cell development and spermatogenesis³¹. Based on our previous work, showing that iNOA men present with higher levels of anti-mullerian hormone (AMH)³² and altered biochemical ECM composition associated with impaired metabolism of the retinoic acid⁹, we hypothesized that the somatic testicular niche of iGCA might be representative of a testis stuck at an immature stage. Here, we provided several observations to support our hypothesis. First, we examined the expression of the Delta Like Non-Canonical Notch Ligand 1 (DLK1) gene, which is considered a marker of the immaturity of LEY cells. DLK1 is expressed in the human testis during fetal, neonatal, and pre-pubertal stages^24,33, is an inhibitory modulator of Notch signaling³⁴, whose block determines the final maturation of fetal LEY cells³⁵. Here, we found overexpression of DLK1 and NOTCH2 in LEY and MYD cells in iGCA vs. adult testis (Fig. 4a, Supplementary information, Dataset S29 and validation at the protein level in Supplementary Fig. S7). Along with the immature transcriptional phenotype of LEY cells, we considered the IGF/IGFR pathway whose signaling on adult SRT and LEY cells³⁶ is directly responsible for the testicular size and sperm production^37,38 and for steroidogenic activity³⁹, respectively. We found that IGF1 was expressed in LEY and MYD cells, and IGF2 down-modulated in LEY, MYD, and SRT cells in iGCA (Fig. 4b; Supplementary information Fig. S8), while no modulation of the relative receptors IGF1R and IGFR2 was observed.

Overall, DLK1 overexpression and decreased IGF2 expression underpin two main pathways that might advocate for LEY cell immaturity. INSL3 is a marker for mature LEY cell⁴⁰ and we here report lower expression of INSL3 at both gene and protein levels in iGCA LEY cells vs. adult testis (Fig. 4c). Transcripts for HSD17B3, an enzyme that converts androstenedione to testosterone and is associated with virilization⁴¹, were present with lower levels of expression only in a fraction of LEY cells (Fig. 4d; Supplementary information, Dataset S29), further supporting the conclusion of a delayed LEY cells maturation in iGCA. Cairns group^15,42 recently reported differentiation trajectories for LEY cells based on the analyses of single-cell RNA seq from neonatal, prepubertal, and adult human testes (Supplementary information, Fig. S9, Dataset S30). We extracted from Cairn’s datasets the transcriptional signatures regarding the early, intermediate, and late stage of LEY cell differentiation (A, B, and C, respectively) and compared them with the transcriptional profiles of iGCA LEY cells. We report that iGCA LEY has a greater expression of the Stage A and B signatures, typical of the pre-pubertal maturation stage, while the controls have a greater expression of the Stage C signature, which is typical of the post-pubertal maturation stage (Fig. 4e). The signature for LEY immaturity included the overexpression of EGR1 (Supplementary information, Fig. S9f), which has been reported to mediate inflammation and fibrosis^43,44, and of JUNB, recently proposed as a marker gene of SRT cell immaturity¹⁹. Furthermore, available single-cell datasets from multiple studies^15,24,42,45, showed that DLK1 expression in LEY and MYD follows a decreasing trend from the neonatal to the prepubertal and adult stages. Consistent with these findings, DLK1 expression in iGCA was similar to what was observed in the neonatal testis (Fig. 4f). Finally, INSL3 expression was found to increase from the neonatal to adult age, but in iGCA LEY the expression was similar to that of LEY cells from the neonatal testis (Fig. 4f).

Deregulated expression of paternally and maternally imprinted genes

Starting from the observation that both DLK1 and IGF2 are paternally imprinted genes, we investigated the genome-wide transcriptional landscape of all the other imprinted genes (http://www.geneimprint.org/site/what-is-imprinting). By comparing the integrated datasets from the Guo¹⁵ and Sohni²⁴ with ours, we found that imprinted genes are dramatically either over or under-expressed in LEY and MYD populations associated with iGCA, (Fig. 4g; Supplementary information, Fig. S10a, b, Dataset S31, 32). Even more important, several imprinted genes are highly expressed in the vast majority of LEY and MYD cells from iGCA testis while completely absent in control cells. The same holds true for genes absent in iGCA cells and expressed in CTL cells. Unexpectedly, a common signature of biallelic expression in iGCA somatic cells was found for some imprinted genes, such as IGF2 (Fig. 4h), with some genes imprinted in all three iGCA and some others specific for each iNOA men (Supplementary Dataset S33) indicating genome-wide deregulation of imprinting molecular mechanisms.

Inflammatory signature at the tissue level

As for the upregulated pathways related to ECM, collagen formation, and collagen organization, we found that COL1A1 and COL1A2 were overexpressed in LEY and MYD cells in iGCA (Fig. 5a). The levels of collagen I and collagen IV proteins, the main types of collagen forming the basal membranes, were quantified by immune-histochemical analyses, using testicular specimens from men with iNOA men depicting iGCA and an age-matched internal cohort of men with normal spermatogenesis. As a further control, we also included testicular specimens from iNOA men with severe hypospermatogenesis, where germ cell-devoid seminiferous tubules intermingled with tubules positive for sperm cells at microTESE.

Thin layers of collagen I were detected around the seminiferous tubules in men with normal germline maturation while tubules from iNOA patients with either complete GCA or hypospermatogenesis showed thick accumulation of collagen I (Fig. 5b). Conversely, collagen type IV was present at the basal membrane of seminiferous tubules with normal germline maturation and absent in iNOA patients with either complete GCA or hypospermatogenesis (Fig. 5c). Of interest, collagen type IV deposition was normal at the basal membrane of the vessels, regardless of the presence of seminiferous tubules with or without sperm production (Fig. 5c).

The median value of basal membrane thickness of 4, 6, and 20 µm was observed in the testis with normal spermatogenesis, iNOA with hypospermatogenesis, or iGCA, respectively (Fig. 5d). We found that the increased thickness of the basal membrane in the tubules was associated with a steady decrease in sperm retrieval (Fig. 5e). In terms of collagen organization, we observed alignment of collagen fibrils in the basal membrane of tubules with normal spermatogenesis, but a scattered distribution in the tubules with iGCA (Fig. 5f), thus resulting in a lower degree of collagen fibrils organization (Fig. 5g). These findings highlighted a pathologic modulation of collagen I and IV both at transcriptional, protein, and organizational levels in MYD and LEY as compared with END cells environment. The “leopard pattern” hyalinization of the seminiferous tubules is indicative of a potential spreading of the disease.

The deregulated collagen deposition at the basal lamina as a feature of tissue aging added to inflammatory features and the detection of both T-cell and activated macrophages prompted us to evaluate a possibly active innate inflammatory response in iGCA testis. We analyzed the cellular localization of HMGB1, the prototypical nuclear danger-associated molecular pattern (DAMP) protein⁴⁶, which at first relocates to the cytoplasm, and then is secreted in the extracellular milieu as a danger signal to neighboring cells upon inflammatory insults. HMGB1 also belongs to the large group of senescent activated secretory phenotype proteins (SASP)⁴⁷, whose secretion correlates with senescence. Somatic cells in three non-neoplastic testes (median age 34, interquartile ranges (IQR) 29–36), showed nuclear HMGB1 localization with a similar degree of staining. Conversely, HMGB1 was found in the cytoplasm of SRT, LEY, and MYD cells in all seven iGCA tested (Fig. 6a). These results indicate that the somatic fraction of iGCA testis cells is actively responding to an inflammatory environment.

DNA damage and structural alteration of the nuclear envelope in somatic cells

The higher tumor frequency reported for men with azoospermia prompted us to proceed from chronic tissue inflammation to senescence and DNA damage as possible causes of genomic instability in iGCA.

To support the senescent phenotype inferred by SASP detection, we looked for the presence of known indicators of cellular senescence-like cell cycle-related proteins. Indeed, although p16/CDKN2A (Cyclin-Dependent Kinase Inhibitor 2A) is not differentially expressed in iGCA somatic cells, the staining revealed a strong accumulation of p16 in iGCA LEY cells as opposed to control testis (Supplementary Fig. S11a).

SASP is also linked to chronic DNA damage response (DDR) that is detected by the nuclear presence of phosphorylated histone H2AX (γH2AX), which is considered an early marker of DNA damage⁴⁸. By immunofluorescence, DNA damage was observed in the cell line treated with H₂O₂ (Supplementary Fig. S12) and meiotic chromosomes in spermatocytes (leptotene/pachytene) of control testis showed the expected dotted nuclear staining for γH2AX⁴⁹ (Fig. 6b). SRT, MYD, and LEY nuclei were negative in the control testis but stained positive in iGCA tissues (Fig. 6b, up to 35% of positive cells in each population, n = 3), thus pointing to an active DDR. In addition, by DAPI staining, we noted very distorted nuclei in the LEY and SRT cell population, reminiscent of nuclei in patients with nuclear envelopathies, a group of rare genetic disorders characterized by a variety of clinical symptoms connected with premature aging and caused by mutations in genes encoding for components of the nuclear lamina⁵⁰. Indeed, the membrane of LEY nuclei from iGCA, but not SRT nuclei, were significantly more indented than the CTL cell counterparts (P < 0.001) as assessed by the nuclear circularity parameter, whereby the value 1 corresponds to a smooth nuclear surface and 0 to high levels of nuclear indentation (Fig. 6c). Likewise, both LEY and SRT nuclei in iGCA showed a significantly lower degree of roundness, a parameter that indicates a shift toward an oval shape as opposed to the roundish shape detected in CTLs (Fig. 6c).

Consistent with the pathogenesis of envelopathies, we found a mutation in the lamin A/C gene (LMNA, ex.10, H566H, C > T, c.1698; rs4641, the allelic frequency of approx. 0.2, depending on the population considered⁵¹) in our three iGCA patients which were in homozygosity (TT) for iGCA#2 and in heterozygosity (C/T) for iGCA #1 and #3; the same mutation was not observed in the 5 CTL datasets (Fig. 6d), by manually inspecting the scRNA-seq reads accumulated on the LMNA locus. So far, this base transition has been considered a synonym polymorphism due to the absence of pathologic phenotypes besides very mild associations with metabolic syndromes^52,53. No mutations in other envelope genes were detected, either.

Of note, the rs4641 LMNA polymorphism hits the −1 position of the splicing donor (SD) site at the 3′ end of exon 10 (Supplementary Fig. S13) which rules the delicate balance between the A and C alternatively spliced forms.

Since the wildtype LMNA gene bears the infrequent SD sequence “AC” in ex10, which has a higher affinity for the U1 spliceosome complex compared to the consensus “AG” SD (half-life = 1.34 vs. 1 a.u., respectively⁵⁴), we hypothesized that the T substitution in the mutant alleles would (i) lead to a shorter U1 spliceosome complex half-life (1.12 a.u.), (ii) decrease the splicing events probability at that SD site, and (iii) both decrease the production of the full-length LMNA transcript and favor the truncated form LMNC. Indeed, the kallisto⁵⁵ isoform analysis applied to the high-coverage RNA-seq on the LMNA/C locus (3000–6000 reads) identified five major LMNA/C transcript isoforms (Fig. 6d) in CTLs and iGCA cases. LMNA and LMNC transcripts are reduced by almost twofold in iGCA as compared to CTLs (Var1 and Var2; p = 0.018 and 0.01, respectively) while a significant fivefold enrichment of Variant 5 was detected (p = 0.02). The overexpression of Variant 5 functionally supports the hypothesis of the weak splice donor site: indeed, the transcript terminates with the exon10 sequence plus a short sequence from intron 10 (18 bp plus stop codon and 3′UTR in intron 10), which codes for the same 6aa present in Variant 2/LMN C. These data suggest that exon10-to-exon11 splicing events have lower odds in iGCA with rs4641 LMNA polymorphism than in CTLs with the wildtype gene splice sequence. In addition, the 5′end is transcribed from an alternative ex1, which excludes the first 100 aa in the protein that is then replaced by an alternative unrelated 20 aa sequence (Fig. S13).

Overall, our results in iNOA patients phenotypically support a p16/CDKN2A+ senescent phenotype in LEY cells that are cell cycle blocked and with high levels of DNA damage to deal with. In addition, the genetic mutation in the LMNA gene in iGCA cells might drive an unbalanced representation of Lamin protein isoforms at the nuclear envelope determining misshapen nuclei reminiscent of those from aged individuals and progeric children⁵⁶.

Features of inflammation in iGCA at a systemic level

A further step was to investigate if the above reported transcriptional findings are reflected at the systemic level. Indeed, the upregulation of pregnenolone biosynthesis transcriptional pathway identified in iGCA SRT cells (Fig. 1d) was paralleled by an increased peripheral level of estradiol in 41 iNOA men with iGCA (median age 37 years; IQR 33–39) vs. 41 normozoospermic men (median age 37 years; IQR 34–41) with proven fertility (Fig. 6e).

We also found decreased levels of circulating total testosterone (tT) in the 41 iNOA men with GCA vs. the 41 normozoospermic individuals with proven fertility (Fig. 6f). These data further support the concept that LEY immaturity contributes to a general and systemic phenotype. Circulating tT and luteinizing hormone (LH) undergo an age-dependent gradual shift in males, with a decrease of tT along with an LH increase⁵⁷. In a cohort of 102 fertile men (median (IQR) 37 (33–40) years) we observed that LH/tT ratio was independent of sperm parameters, whereas the same ratio was increased in an age-matched cohort of iNOA men with GCA (n = 43; 37 (33–39) years) (Fig. 6g). Therefore, low levels of tT potentially associated with an immature status of LEY cells were associated with compensatory positive feedback in terms of LH synthesis.

The deficiency of testosterone, known to inhibit the production of inflammatory cytokines⁵⁸, associated with the cytoplasmic localization of HMGB1, the pro-inflammatory profile of MCR, and the presence of tissue-resident CD8+ TCL in iGCA patients, prompted us to investigate whether the inflammatory processes are only localized in the testis environment or are more systemic. Prompted by the overexpression of cytokines and chemokines in TCL, we quantified blood circulating CCL4 and CCL5, which belong to the SASP proteins secreted in inflammation and senescence. In iNOA men with iGCA vs. age-matched controls, levels of CCL4 and CCL5 chemokines are significantly higher (Fig. 6h, i).

iGCA testis show a transcriptional aged phenotype

As a whole, the upregulated proteins of the innate immunity and cytokine signals, a pro-inflammatory environment, p16 staining, low levels of circulating testosterone, along with the downregulated pathways of RNA processing and amino acids metabolism, reported for iNOA men, are consistent with altered biological pathways orchestrating the process of ageing⁵⁹. Comparative analysis of Differentially Expressed Genes vs. the aging-associated genes annotated in the GenAge database (https://genomics.senescence.info/legal.html) clearly revealed an overlap with all ribosomal genes, UBA52 and NACA reported being associated with aging, for all seven somatic iGCA clusters (Fig. 6j).