The human sperm proteome—Toward a panel for male fertility testing

2.1 Study cohorts

Permission for sample collection was granted by the Ethics Committee of the Medical Faculty at the Martin Luther University Halle/Saale (approval number 218/14.04.10/2; date of decision: April 19, 2010; date of approved amendment: March 14, 2013). All participants were recruited from the fertility outpatient care of the University Hospital Halle (Saale) and gave written informed consent. Ejaculates were provided by 76 men (Table S1). Spermiogram parameters were re-assessed according to the updated version of the World Health Organization (WHO) laboratory manual for the examination and processing of human semen (6th edition, 2021),²⁶ with no changes in the diagnoses of the study participants. For clarity reasons, we use descriptive terms for abnormal spermiograms (O, asthenozoospermia, teratozoospermia, and mixed forms) as proposed in previous WHO laboratory manuals (5th edition and earlier). Nonetheless, the authors acknowledge the fact that the cut-off values used for defining these terms do not necessarily determine the fertility status of an individual male. Thirty-eight infertile men (mean and median age = 32 years, each) presented with O (21), OA (7), or OAT (10), and were unintentionally childless, that is, their partners did not conceive despite regular unprotected intercourse within at least 12 months. In the 38 infertile patients, median sperm count per ejaculate was 36.5 mio (mean = 41.4 mio). The control cohort included 38 men (mean and median age = 30 years, each) presenting with normozoospermia, who already had fathered offspring without medical assistance. Median sperm count of the corresponding ejaculates was 155.0 mio (mean = 205.7 mio) (Table S1).

2.2 Sample preparation for MS

Upon swim-up, spermatozoa were separated from seminal plasma by centrifugation for 5 min at 1000 g. After discarding the plasma, cells were resuspended twice in phosphate-buffered saline and re-pelleted by centrifugation at 2000 g for 5 min, each time succeeded by discard of the supernatant. Washed spermatozoa were shock-frosted in liquid nitrogen and stored at –80°C. Following thawing on ice, samples were boiled in 1× lithium dodecyl sulfate buffer and separated by polyacrylamide gel electrophoresis on a 4%–12% gradient NOVEX gel (Thermo Scientific) for 10 min at 180 V. The gel was fixated and stained with Coomassie blue for 30 min and subsequently washed with water overnight. In-gel digestion was performed as previously described.³¹ In short, each sample was cut from the gel, minced, and transferred to a reaction tube. The gel pieces were destained with 50% EtOH/25 mm ammonium bicarbonate (ABC) buffer (pH 8.0). After dehydration of the gel pieces with pure acetonitrile (ACN), samples were dried for 5 min in a concentrator (Eppendorf) and afterward incubated with reduction buffer (10 mm dithiothreitol in 50 mm ABC) for 30 min at 56°C. The reduction buffer was removed, substituted with alkylation buffer (50 mm iodoacetamide in 50 mm ABC), and then subjected to 30 min incubation at room temperature in the dark. Gel pieces were completely dehydrated with pure ACN and covered in trypsin solution (1 µg trypsin in 50 mm ABC per sample). Proteins were digested overnight at 37°C. Tryptic peptides were extracted twice by incubation with extraction buffer (3% trifluoroacetic acid and 30% ACN) for 15 min and afterward with pure ACN. After concentration of the elution fraction to about 10%–20% in a concentrator (Eppendorf), the peptides were passed through a StageTip.³² StageTips were prepared using two layers of C18 material (Empore), which was activated with methanol, washed with solution B (80% ACN, 0.1% formic acid), and equilibrated once with solution A (50 mm ABC, 0.1% formic acid). Extracted peptides were loaded on the StageTips and washed with solution A. Peptides were eluted with 30 µL solution B, ACN was removed by use of a concentrator (Eppendorf), and samples were diluted with 6 µL solution A.

2.3 MS measurement

The samples were injected via an autosampler into an uHPLC (EASY-nLC 1000, Thermo Scientific). Peptides were loaded on a 25 cm capillary (75 µm inner diameter; New Objective) packed in-house with Reprosil C18-AQ 1.9 µm resin (Dr. Maisch) for reverse-phase chromatography. The EASY-nLC 1000 HPLC system was directly mounted to a Q Exactive Plus mass spectrometer (Thermo Scientific). Peptides were eluted from the column with a 208 min optimized gradient from 2% to 40% ACN with 0.1% formic acid at a flow rate of 225 nL/min with a column oven set-up operating at 40°C (Sonation). The heated capillary temperature was set to 250°C. Spray voltage ranged from 2.2 to 2.4 kV. The mass spectrometer was operated in data-dependent acquisition mode with one MS full scan and up to 10 triggered MS/MS scans using higher energy collisional dissociation. MS full scans were obtained in the orbitrap at 70,000 resolution with a maximal injection time of 20 ms, while MS/MS scan resolution was set to 17,500 and maximal injection for 120 ms. Unassigned and charge state 1 were excluded from MS/MS selection and peptide match was preferred. Detailed information can be obtained from the measurement files available on the public ProteomeXchange repository (PXD037531).

2.4 MS data analysis

Raw data analysis was performed with MaxQuant v1.5.2.8³³ with standard settings except label-free quantification (LFQ) and match between runs were activated. Time windows for matching and aligning were 0.7 and 20 min, respectively. MaxQuant analysis including detailed settings has been deposited at ProteomeXchange under the above accession number. After MaxQuant, potential contaminants, reverse database hits and proteins only identified by site were removed. Further, we filtered for a minimum of two identified peptides (minimum 1 unique). For quantitative proteome analyses, lacking values were imputed assuming a beta distribution within 0.2 and 2.5 percentiles of measured LFQ intensities per sample. Downstream comparisons of protein abundances focused on spermatozoa of infertile men diagnosed with reduced sperm count and fertile men with normal semen parameters. Differentially abundant sperm proteins were grouped according to infertility diagnoses, thus generating O list (O vs. normozoospermia), OA list (OA vs. normozoospermia), and OAT list (OAT vs. normozoospermia). For being contained in a list, a protein needed to be detected in at least two samples of a diagnosis, and abundances had to vary by a factor of three at minimum (false discovery rate [FDR] corrected p-value ≤ 0.05, Welch two sample t-test). This corresponds to a minimum difference of log2-transformed values of 1.585. We additionally highlight sperm proteins showing at least eightfold differential abundances, corresponding to a minimum difference of log2-transformed values of 3.000.

2.5 Further analyses

We ran ShinyGO v0.76³⁴ on ENSEMBL stable IDs of the genes encoding MS-determined sperm proteins for deriving the lengths of coding sequences and transcripts. We also adopted from the ShinyGO output the grouping of genes according to higher level biological process gene ontologies (GOs). Moreover, we matched ENSEMBL stable gene IDs from present MS analysis with the genes to the sperm proteins collected by Castillo et al.⁸ For the latter compilation, UniProt/SwissProt IDs had been converted into ENSEMBL IDs using BioMart (Ensembl Genes 106, Human genes, GRCh38.p13).

Overlaps between above O-associated lists of genes/proteins were rendered in a Venn diagram (https://bioinformatics.psb.ugent.be/webtools/Venn/). We tested for the overrepresentation of biological process GOs in O, OA, and OAT lists using ClueGO v2.5.9³⁵ within Cytoscape v3.9.1.³⁶ In addition, we matched the list for the most comprehensive infertility diagnosis, that is, the OAT list, with entries in NCBI's Online Mendelian Inheritance in Men (OMIM) database (state February 1, 2022) referencing male fertility impairment. Corresponding OMIM entries were related to O and non-obstructive azoospermia using all possible forms of spermatogenetic failure (SPGF) as search items: SPGF1, SPGF2, …, SPGF65, SPGFX1, SPGFX2, SPGFX3, SPGFY1, and SPGFY2. We further collected OMIM entries on asthenozoospermia, teratozoospermia, and male infertility. We screened the WWW for publications using the combination of symbol and infertil* as initial search items (e.g., APAF1 infertil*). We additionally queried The Human Protein Atlas v22 (http://proteinatlas.org/; accessed on January 10, 2023) for the expressional landscape of the genes behind differentially abundant sperm proteins.

Subsequently, we assessed if the candidate markers presented herein and additional ones emerging from external sources might be connected at a higher level. For this purpose, we reconstructed three protein–protein interaction (PPI) networks of growing complexity. Using standard settings in STRING v11.5,³⁷ the first network was confined to the biomarkers as inferred from present empirical data. The second network additionally considered candidate markers as emerging from a previous computational approach. Corresponding genes/proteins were calculated to have elevated fertility probability (≥0.35) given enhanced sequence conservation, heightened testicular transcript abundance, and increased connectivity in a body-wide PPI network. Furthermore, knockouts of their murine orthologs associated with male fertility impairment.¹⁹ We extracted the aforementioned proteins from Dataset M at the PreFer Genes website (https://prefer-genes.uni-mainz.de/). A third network was reconstructed after further expansion of the sample by all proteins with previous mention in the context of male infertility in OMIM database (see above). The networks were compared by their overall coherence and average node degree, that is, the mean number of PPIs per protein or node. Each network was tested for PPI enrichment.

Raw p-values from chi-square tests, GO enrichment tests, and PPI enrichment tests were transformed into FDRs.³⁸ Further consideration of test results required an FDR of 0.001 or lower.

3.1 Fine-scale reconstruction of the human sperm proteome

Present MS analysis identified 5778 protein groups (q ≤ 0.05, each) across sperm samples provided by 76 men and purified by swim-up (Table S1). These contained 33,191 gene products to which 6981 ENSEMBL stable gene IDs were available. Of the latter, 4545 had counterparts in the compilation of sperm proteins by Castillo et al.⁸ Thus, 65% of the sperm proteins for which we found evidence in human spermatozoa can be regarded as reproduced (Table S2). The 6939 genes mapped by ShinyGO significantly differed from the rest in the human genome by having shorter coding sequences (FDR = 6E-15) and transcripts (FDR = 1.7E-20; t-test, each) (Figure 1). A closer look for the 50 biological process GOs occurring with highest frequencies corroborated the importance of sperm motility for fertilization. Processes such as cell proliferation and reproduction, adhesion, and responsiveness to stimuli were also represented, as were involvements in developmental processes and the immune system. The latter included genes coding for spermatogenesis-associated protein 20 (SPATA20), zona pellucida-binding protein 1 (ZPBP), and sperm-associated antigen 4 (SPAG4) (Table S3). Additional higher ranking terms were of more general nature such as “catabolic process” (Figure 2). The GOs of the remaining 50 smaller protein groups partially confirmed roles in male gametogenesis (meiotic cell cycle process, spermatid nucleus differentiation) and fertilization (acrosome reaction), while others referred to additional entanglements (Table S3).

3.2 Exactness of MS analysis

For being considered, a protein had to be determined in at least one unique and one razor peptide across the 76 samples analyzed. However, much higher values were achieved in the validated and novel candidate markers detailed below (Tables 1–4 and S2). In case of the validated marker candidates, we had 4–246 (mean = 51, median = 32) determinations of unique and razor peptides. These were mostly unique peptides, as illustrated by corresponding mean and median numbers of 47 and 29, respectively. In the novel candidate markers, the span of unique and razor peptides ranged from 7 to 177 (mean = 33, median = 21). Again, these were mostly specific to a particular protein group, as indicated by mean (32) and median (21) values of unique peptides. We consider these counts as a confirmation of the reliability of the detections.

3.3 Quantitative MS analysis

The pipeline applied unveiled numerous protein groups with at least threefold differential, mostly lowered, abundances (FDR ≤ 0.05, each; t-test) in spermatozoa of infertile men diagnosed with O, OA, and OAT, relative to normozoospermic spermatozoa of normal-fertile men (Table S4). Thus, 133 of 153 (87%) differentially abundant protein groups exhibited downregulation in O spermatozoa (O list). With 154, the total number of protein groups varying in abundance was almost the same in the OA versus normozoospermia comparison (OA list), whereby most of these showed declines in abundance under infertility once more (N = 110; 71%). The OAT versus normozoospermia comparison (OAT list) revealed the highest number of differentially abundant protein groups (N = 368), with lowered levels predominating in the infertility cohort again (N = 315; 86%) (Table S4). Fifty-five protein groups were shared among the three O-associated lists. In addition, there was an increasing number of unique protein groups with growing complexity of the diagnosis, and thus from the O via the OA to the OAT list (Figure 3). Also, the number of overrepresented GOs (FDR ≤ 0.001, each) was much higher in the OAT list (N = 72) than in O (N = 26) and OA (N = 20) lists. Nevertheless, cilium-dependent cell motility yielded the largest fraction of GOs in all three infertility diagnoses. The three lists also had in common that 75% of the overrepresentations pertained to the structure and function of cilia or the sperm flagellum. Additional overrepresentations referred to male gamete generation and fertilization (O, OA, and OAT lists), cAMP signaling (OA and OAT lists), and sodium–potassium exchange (OAT list) (Figure 4).

Twelve percent of the altogether 198 genes with OMIM entries relating to male fertility impairment had counterparts in the most comprehensive of our three lists, the OAT list (Table S5). Broken down to individual aberrations from normozoospermia, we observed the following pattern: the gene list compiled from OMIM entries on teratozoospermia matched to 10% (3 of 29) with genes or proteins in our OAT list. In OMIM lists for asthenozoospermia and teratozoospermia, 17% (12 of 70) and 35% of the genes (23 of 66) had matches in our OAT list. Thus, the proportion of matching genes in all three OMIM lists considerably exceeded the random expectation of <1%, as given by the total number of genes in single OMIM lists (N = 23–70) divided by 20,471 protein-coding human genes according to ENSEMBL Genes 106: GRCh38.p13.

3.4 Candidate markers of male fertility

Thirty-one sperm proteins in our OAT list had entries in the OMIM database suggestive of an involvement in male fertility impairment (Table 1). Their GOs related to sperm flagellum assembly, structure and beating, as well as male gametogenesis. We refer to these proteins as to validated candidate markers in the following. Out of these, apoptosis trigger APAF1 was the only one showing higher abundance in OAT than normozoospermic spermatozoa (Table 1). The remaining proteins with previous mention relating to male fertility impairment exhibited reduced levels under OAT. Seven of these displayed decreased abundances across all three O-associated diagnoses: actin-like protein 9 (ACTL9), calicin (CCIN), CFAP47, CFAP65, CFAP251 (WDR66), dynein axonemal heavy chain 1 (DNAH1), and sperm maturation protein 1 (SPEM1) (Table 1). In-depth search for previous evidence (Table 2) revealed that most validated candidate markers were previously associated with malformed and dysfunctional sperm flagella and impaired sperm motility. Corresponding terms included multiple morphological abnormalities of the flagella (MMAF) and asthenozoospermia (N = 20). This was followed by associations with morphological defects such as acephalic spermatozoa syndrome (ASS), acrosomal defects, and lacking mitochondria (N = 10). The least connections were found to reduced sperm count (N = 3). Notably, 18 of the validated candidate markers of male fertility were covered by an 81 gene panel and hence are already in use in male fertility diagnostics (Table 2).

Lack of OMIM entries relating to male fertility impairment let us refer to a group of 18 sperm proteins with extremely deviating abundances under infertility as to novel candidate markers. These exhibited at least eightfold reduced abundances in the OAT–normozoospermia comparison (Table 3). According to the data available, the novel candidate markers engage in spermiogenesis, including ciliogenesis and acrosome biogenesis. For all novel candidate markers, we found previous indications of downregulated expression under reduced male fertility. For example, no SPATA48 transcript was determined in testicular samples of six azoospermic men, while corresponding transcripts were highly abundant in healthy subjects (Table 4). Furthermore, seven proteins had significantly reduced abundances in O and OA spermatozoa, and, hence, might elucidate the deeper causes of male fertility impairment whenever reduced sperm count is suspected to play a role.

There were three protein groups, each containing two proteins, with decreased abundances in spermatozoa under male infertility (Table 5). Half of these proteins had OMIM entries relating to male fertility. Until clarification whether one or both group members occur at lowered abundances in spermatozoa of infertile men, the corresponding proteins lend themselves less readily as candidate markers.

Downstream, we checked for potential somatic expression of the genes to the proteins in Tables 1, 3, and 5. According to version 22 of The Human Protein Atlas, transcripts of most of the genes with validated male fertility relevance (Table 1) were specific to (category I: seven genes) or much more frequent in male germline (category II: six genes), or transcript abundance was elevated in male germline and somatic cell types, which will not have contaminated our samples (category III: 15 genes). The latter was exemplified in respiratory ciliated and endometrial ciliated cells (Figures 5 and S1). Genes of only few validated candidate markers showed the expected expression in male germline and additionally displayed elevated transcript abundances in somatic cell types that could have been contained in the samples analyzed (category IV: three genes). Prevalence of categories I–III was reproduced in genes encoding novel candidate markers (Table 3 and Figure S2). Most of the coding genes displayed exclusive (category I: four genes) or skewed expression in male germline (category II: 12 genes), and one gene additionally exhibited raised transcript amounts in somatic cells, which unlikely were contained in the samples (category III: one gene). The transcriptional profile of only one of the novel candidate markers allowed for the possibility that somatic cell types could have been included in the sample in addition to the targeted spermatozoa (category IV). The genes underlying protein groups even spread without exception across categories I–III (Table 5 and Figure S3). Thus, for only few of the proteins contamination with somatic cells might have contributed to the finding of differential abundance. But also, these few ones had OMIM entries indicating male fertility relevance (Table 1: APAF1, CATIP, CFAP69) or were already found to be downregulated in spermatozoa under male infertility (Table 3: SYPL1). We therefore assume that contaminations of our sperm samples with somatic cells should not have compromised present results.

3.5 The male infertility network

For all protein samples considered, STRING gathered more PPIs than to be expected for the proteins mapped (FDR < 0.001, each). In addition, all network parameters rose with growing number of proteins included (Tables 6 and S6). Thus, STRING built a network of 50 nodes, thereof 66% contained in the largest connected component (LCC), and 75 edges when run on the proteins in Tables 1, 3, and 5. Higher coherence was reached upon addition of 41 candidate markers for which we had previously calculated elevated fertility relevance probability.¹⁹ Especially, the number of disconnected nodes decreased, despite an almost doubled number of proteins mapped (N = 91). With 208, this network had a disproportionally larger number of edges. Correspondingly, the number of PPIs per node was higher in the expanded network (node degree = 4.57) than in the one restricted to the proteins highlighted in the present study (node degree = 3.00). Addition of all other proteins with OMIM entries suggestive of male fertility relevance more than doubled the number of nodes to 225. The total number of PPIs increased to 891, and the average node degree to 7.92 (Table 6). The corresponding LCC was comprised of 192 nodes (Figure 6), thereunder half of the novel candidate markers and all of the validated candidate markers. Also included in the LCC were 37 out of the abovementioned 41 proteins with increased fertility relevance probability,¹⁹ and most of the proteins with OMIM entries pointing to male fertility relevance. We refer to this LCC as to the male infertility network.

4.1 The human sperm proteome and its coding genes

According to present MS analysis, about 6981 genes participate in the human sperm proteome (Tables S1 and S2). This represents a considerable increase compared with previous estimates from MS studies, which had determined 3444²¹ and 4718²² sperm proteins. The current number also exceeds the 6871 sperm proteins as gathered in a recent survey.⁸ Merging the compilation from the latter study with present MS results, about 9300 genes might participate in the human sperm proteome. This would be ca. 45% of the approximately 20,500 coding human genes according to genome annotation ENSEMBL Gene 106: GRCh38.p13. The complexity of the sperm proteome will be higher, considering diversification by alternative splicing and post-translational modification.^{29, 98, 99}

Overall shorter cDNAs, as observed in this study, could reflect adaptation to better swimming performance. In line with such possibility, there is previous evidence that selection for increased motility might drive changes in sperm morphology.^{100, 101} This might involve remodeling of protamines which are well-known for evolving under sexual selection.¹⁰² In the present reconstruction of the human sperm proteome, importance of sperm motility was reflected in implications of numerous coding genes in cilium assembly and function (Figure 2). GOs referring to cellular responsiveness and adhesion were also conclusive considering the challenges spermatozoa have to cope with.^{103, 104} This might also be true for annotations referencing the immune system. In fact, male fertility and fertilization require evasion from the own immune response and the one of the partner.^{99, 105} Still, corresponding GO groups contained genes encoding proteins with clear fertility relevance, such as SPATA20,^{11, 106} ZPBP,¹⁰⁷ and SPAG4¹⁰⁸ (Table S3). Thus, reference to the immune system should not be overestimated in our opinion. Not least, present GO analyses (Figure 2) confirmed that spermatozoa contain proteins that are known primarily for their roles in male gametogenesis.¹⁰⁹

4.2 Molecular causes of male fertility impairment

Results of present overrepresentation tests revealed that the fewer spermatozoa produced under O, OA, and OAT carry signatures of disturbed gametogenesis. Functional annotations overrepresented in differentially abundant proteins were further consistent with impairment of sperm motility and fertilization competence in infertile men (Figure 4). This includes proteins involved in cAMP signaling and sodium–potassium exchange. The first is crucial for all steps of sperm function in mammals,¹¹⁰ and importance for motility and fertilization has been assigned to the latter.¹¹¹Significance of proper sperm motility for reproduction¹¹² was additionally reflected in frequent references to the structure and function of the cilium or flagellum (Figure 4). The pattern was even more strongly reflected in the data on the individual proteins and coding genes. Thus, for the validated and novel candidate markers, there was evidence of involvements in malformations such as ASS and MMAF and, related to the latter, impaired motility or asthenozoospermia (Tables 1–4). This suggests that reduced sperm count could be less decisive in O and syndromes including it. The ultimate reason might rather be that the fewer spermatozoa produced are dysfunctional to an increased proportion. This seems particularly noteworthy given that the samples analyzed had undergone a swim-up, meaning that the spermatozoa represented a selection of spermatozoa with increased fertilization competence. Even these would be functionally compromised according to our results, to an increased proportion at least. This could also explain why lowered spermiogram parameters can reduce success rates in ARTs.^{113, 114}

4.3 Toward a marker panel of male fertility assessment

Quantitative MS analysis of the entire sperm proteome might complement diagnostics of the male fertility status in the future. If it comes to this point, special attention may be paid to the abundances of particular sperm proteins. Estimating the fertility status from differential abundances of sperm proteins on scales as found here (N = 153–368) should then be more feasible than based on numbers in the range of 1000 or higher as previously reported.^{21, 22} This was achieved by applying a more conservative threshold of differential abundance in the present study (at least 3×) than was the case in both referenced investigations (at least 1.5×).^{21, 22} Still, the definition of a further reduced number of male fertility markers remains to be a worthwhile goal for standardized fertility testing.^{19, 28, 115} The proteins summarized in Tables 1 and 3, and to a lesser extent in Table 5, may contribute to this. It will be interesting to validate in future independent studies the corresponding proteins using alternative methods such as western blot.

Diagnostic potential should especially have the 31 sperm proteins with previous mention in the OMIM database hinting to roles in the etiology of male infertility (Table 1). Increased importance of these validated candidate markers for reproduction is underscored by transcript data. Indeed, we found expression of most of the coding genes to be tailored to male germline (Figure S1). Numerous associations with male fertility impairment lend additional support for their marker potential, as did their already-mentioned involvements in sperm flagellum structure and functioning, sperm head formation, and spermiogenesis (Tables 1 and 2). Against this background, it also seems plausible for most validated candidate markers that a decrease in abundance should be detrimental for male fertility maintenance. There were only two exceptions from the pattern. Exceptional in terms of function was 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase zeta-1 (PLCZ1). The protein proposedly belongs to the sperm factors transferred to the zygote and as such is likely to initiate the development of the zygote into the blastocyst.^116-118 Consistently, a PLCZ1 variant in an infertile man was found to associate with oocyte activation deficiency.⁵⁴ Thus, infertility at PLCZ1 shortage could hint to disturbed development following fertilization of the ovum. Still, PLCZ1 seems to be also important for spermiogenesis considering spermatogenic defect in male Plcz1 knockout mice.¹¹⁹ The second exception from the above pattern refers to the direction of abundance change in reduced-relative to normal-fertile men. In fact, apoptotic protease-activating factor 1 (APAF1) was the only validated candidate marker showing increased abundance under infertility. Although exceptional in the present study, the present finding fits previous evidence of raised testicular abundance of APAF1 transcript in men with impaired versus normal spermiogenesis.⁶⁵ It appears to be plausible for an apoptosis trigger¹²⁰ that oversupply negatively affects sperm number and viability.²⁰

Seven of the proteins, to which we here refer to as validated candidate markers, displayed deviant abundances under O, OA, and OAT (Table 1): ACTL9, CCIN, CFAP47, CFAP65, CFAP251 (WDR66), DNAH1, and SPEM1. Functional data would be in line with a predictive value of corresponding protein abundances whenever a reduction in sperm count is observed. Thus, SPEM1 should have importance for proper sperm formation including cytoplasm removal as suggested by deformed spermatozoa and infertility in male knockout mice.⁴⁹ Furthermore, ACTL9 plays a significant role in the fusion of proacrosomal vesicles and the anchoring of the acrosome to the perinuclear theca during spermiogenesis. Correspondingly, there are variants that associate with fertilization disorders and male infertility.³⁹ CCIN also localizes to the acrosomal area, thereby showing affinity to F-actin.¹²¹ As a basic cytoskeleton protein of the mammalian sperm head, absence or dysregulation of CCIN is known to cause sperm malformation.¹²² Relevance for proper flagellum formation and functioning is further inherent to three cilia and flagella-associated proteins (CFAP47, CFAP65, CFAP251/WDR66) and DNAH1.^{42, 43, 45, 46, 91, 123, 124} Nevertheless, all sperm proteins in Table 1 could be implicated in the development of male fertility disorders, and hence might be further evaluated for their potential as male fertility markers. In support of such utilization, the coding genes of 18 out of the 31 validated candidate markers are already included in an 81 gene panel (ID192.04) for male fertility testing (e.g., www.zhma.de).

Diagnostic potential might also have 18 proteins displaying at least eightfold and thus extreme deregulation of abundance (Table 3). We refer to these proteins as novel candidate markers because OMIM search did not uncover respective previous entries. Nevertheless, downregulation of testicular and spermatozoic expression under reduced male fertility was documented for the coding genes to all novel candidate markers (Table 4). Predictive power of corresponding protein levels would also be consistent with the functional and transcriptional data available, as will be detailed below in the ones showing reduced levels in all three diagnoses including reduced sperm count. In case of SPATA31E1, SPATA31D1, and SPATA48, relevance for male fertility is reflected by transcript enrichment in the male germline (Figure S2). Transcriptional data additionally suggest that EF-hand domain family member B (EFHB, alias CFAP21), chromosome 2 open reading frame 16 (C2orf16, alias CB016), and family with sequence similarity 161 member A (FAM161A) participate in male fertility maintenance (Figure S2). Indeed, EFHB is a regulator of store-operated Ca²⁺ entry (SOCE) impairment of which causes male infertility in the murine model.^{125, 126} Furthermore, FAM161A is a component of the cilia–basal body complex¹²⁷ modulating mammalian sperm movement.^{127, 128} Another protein with strongly deregulated abundance in association with O, cylicin 1 (CYLC1), is involved in the formation of the sperm head.¹²⁹ As in several other novel candidate markers, CYLC1 transcription is highly enriched in male germ line, suggestive of special importance for male fertility (Figure S2).