MGAT1 and Complex N-Glycans Regulate ERK Signaling During Spermatogenesis

Mechanisms that regulate spermatogenesis in mice are important to define as they often apply to fertility in man. We previously showed that conditional deletion of the mouse Mgat1 gene (Mgat1 cKO) in spermatogonia causes a germ-cell autonomous defect leading to infertility. MGAT1 is the N-acetylglucosaminyltransferase (GlcNAcT-I) that initiates the synthesis of complex N-glycans. Mechanistic bases of MGAT1 loss were investigated in germ cells from 22- and 23-day males, before any changes in germ cell morphology were apparent. Gene expression changes induced by deletion of Mgat1 were determined using the Affymetrix gene chip Mouse Mogene 2.0 ST array, and relationships were investigated by bioinformatics including Gene Ontology (GO), Ingenuity Pathway Analysis (IPA), and Gene Set Enrichment Analysis (GSEA). The loss of complex N-glycans promoted the premature up-regulation of genes normally expressed later in spermatogenesis and spermiogenesis, and IPA and GSEA implicated ERK signaling. EGFR and PDGFRA transcripts and ERK1/2 signaling were reduced in 22-day Mgat1 cKO germ cells. Basigin, a germ cell target of MGAT1, activated ERK1/2 in CHO cells, but not in a Lec1 CHO mutant that lacks MGAT1 and complex N-glycans. Thus, MGAT1 is required to regulate ERK1/2 signaling during spermatogenesis, potentially via different mechanisms.


Results
Early testicular changes associated with deletion of Mgat1 in spermatogonia. Our previous study characterized Mgat1[F/F]:Stra8-iCre males at 7 weeks 3 . To identify the earliest stage at which defective spermatogenesis is observed, testes from control and Mgat1 cKO males from 15 to 28 dpp were compared by histology (Fig. 1A). At 15 dpp, no apparent differences in seminiferous tubule size or the population of germ cells present in 50 tubules were observed (n = 3 mice/group). At 22 and 23 dpp, round spermatids were present in both control and mutant tubules, and there were still no apparent histological differences (Fig. 1A). At 24 and 25 dpp, fusion of cells adjacent to the lumen was observed in a few tubules (Supplementary Table S1; Fig. 1A). Spermatids were identified based on nuclear size, morphology, location in the tubule or detection of acrosomes by periodic Schiff stain (PAS) at 22-25 dpp (Fig. 1A,B), or the acrosomal protein sp56 at 28 dpp ( Supplementary Fig. S1). At 28 dpp, mature spermatozoa were present in control but not Mgat1 cKO mutant testis sections (Fig. 1A). The number of tubules with elongated spermatids was significantly reduced in 28 dpp mutant testes, and MNC were present (Supplementary Table S1). Mgat1 cKO and control testis sections were analyzed at 24-26 dpp to detect Sertoli cells (SOX9), spermatogonia (PCNA), spermatocytes (SYCP3), and spermatids (PAS) ( Fig. 1B; Supplementary  Fig. S2). The number of Sertoli cells, spermatogonia, spermatocytes and Stage VI and beyond tubules were not significantly reduced in Mgat1 cKO versus control tubules (Fig. 1B histograms).

Gene expression in Mgat1 cKO testicular germ cells.
To gain insights into molecular mechanisms that occur early, before morphological changes are apparent, and thus give rise to the defective spermatogenesis of Mgat1 cKO males, microarray analyses were performed on RNA from germ cells from 22 and 23 dpp males (Supplementary Table S2). To investigate the relative purity of germ cells, qRT-PCR for cell-type specific genes was performed on cDNA prepared from 22 dpp germ cells and compared to 22 dpp testis. Transcripts of Rhox5 (Sertoli cells) and Cyp11a1 (Leydig cells) were greatly reduced in germ cell compared to testis preparations  Supplementary Fig. S3A). By contrast, the expression of germ cell-specific genes Sycp3 (spermatocytes) and Dbil5 (round spermatids) were similar in testis and germ cell preparations. Acrv transcripts (late spermatids) were poorly represented in germ cell preparations. Importantly, the expression of Mgat1 was greatly reduced in Mgat1 cKO germ cells at 22 and 23 dpp ( Supplementary Fig. S3B). PCR genotyping of germ cell genomic DNA 11 showed that Mgat1 cKO germ cells contained only deleted Mgat1 alleles (520 bp), whereas control germ cells contained only floxed Mgat1 alleles (560 bp; Supplementary Fig. S3C). Therefore, Mgat1 transcripts in germ cell preparations were contributed by a small fraction of non-germ cells.
The nature of the N-glycans on basigin (~37 kDa), a germ cell glycoprotein target of MGAT1 3 , confirmed efficient deletion of Mgat1 at 22 and 23 dpp. Thus, control germ cell basigin with complex N-glycans was resistant to digestion by endoglycosidase H (Endo H). By contrast, Mgat1 cKO germ cell basigin was sensitive to Endo H digestion, showing that it carried oligomannosyl N-glycans ( Fig. 2A and Supplementary Fig. S3D). Incomplete Endo H digestion probably accounts for traces of undigested basigin in some Mgat1 cKO lysates. Importantly, comparison with ~25 kDa non-glycosylated basigin observed at 22 dpp, showed that levels of basigin were similar in control and Mgat1 cKO germ cells ( Fig. 2A).   Table S2). For the 23 dpp experiment, RNA (RIN > 9) aliquots from six control and six Mgat1 cKO germ cell preparations were separately pooled, and three aliquots from the control and Mgat1 cKO pool, respectively, were converted to cDNA for analysis. Affymetrix Expression Console was used to process the .CEL files of the array. .CHP files were generated using the RMA sketch workflow after signal summarization and data normalization. Gene level analysis was further conducted with Affymetrix Transcriptome Analysis Console v2.0 software (TAC). Hierarchical clustering differentiated control and Mgat1 cKO samples (Fig. 2B). Volcano plots representing gene distributions are shown in Supplementary Fig. S4. Genes that were differentially expressed based on ANOVA p < 0.05 and FDR p < 0.05 were identified. There were 1,643 DEGs in 22 dpp Mgat1 cKO versus control germ cells, with an absolute fold-change (linear) of <−2 to >+2.0. Of those, 1,400 genes were up-regulated and 243 down-regulated at 22 dpp. At 23 dpp, 784 genes were differentially regulated at <−2 to >+2.0, 771 genes up-regulated, and 13 genes down-regulated in Mgat1 cKO germ cells. Many of the DEGs in Mgat1 cKO germ cells (Supplementary Table S3) are involved in spermatogenesis and later stages of spermiogenesis 12 . Validation of microarray experiments by qRT-PCR was performed on cDNA prepared from germ cell RNA obtained from the individual mice used in microarray experiments (Fig. 2C,D). PCR primers are given in Supplementary Table S4. Microarray data are deposited in NCBI's Gene Expression Omnibus (GEO) and are accessible through GEO serial accession number GSE99035. Top upstream transcriptional regulators. IPA upstream functional analysis was used to predict the top upstream transcriptional regulators in DEGs of Mgat1 cKO germ cells at 22 and 23 dpp based on their gene targets. An overlap p value was calculated on the basis of significant overlap between genes in the test dataset and target genes regulated by the same regulator in the IPA knowledge base. The activation Z score algorithm was used to make predictions. IPA predicted the top transcriptional regulator activated in the 22 and 23 dpp datasets to be TAF7L (Z score 4.96; overlap p value 8.89E-31 for 22 dpp; Z score 4.24, overlap p value 6.16E-24 for 23 dpp; Fig. 3C). FIGLA is predicted to be inhibited in both the 22 dpp (Z score −3.22; overlap p value 1.56E-12) and 23 dpp datasets (Z score −3.28, p 1.17E-11) (Fig. 3C). Target genes of TAF7L were up-regulated in our datasets, and predict activation of this regulator. For FIGLA, the target genes from the input datasets were up-regulated, but the regulator is predicted to inhibit transcription. Comparison analysis for upstream regulators at 22 and 23 dpp showed similar activation Z scores for each of the top regulators at both developmental stages (Fig. 3C, right panel).

Differentially expressed genes (DEGs) in control versus
We tested whether the expression of TAF7L itself was up-regulated by IHC, qRT-PCR and western blot analyses ( Supplementary Fig. S6). TAF7L is expressed in the nucleus or cytoplasm of germ cells, depending on their stage of differentiation 13 . We observed TAF7L in the cytoplasm of spermatocytes at 23 dpp. There was no apparent difference in the intensity of the signal between control and Mgat1 cKO sections ( Supplementary Fig. S6A). By qRT-PCR, TAF7L transcripts in 23 dpp Mgat1 cKO germ cells were slightly reduced ( Supplementary Fig. S6B). However, TAF7L protein levels were similar by western blot analyses of germ cell lysates ( Supplementary  Fig. S6C). Nevertheless, microarray data showed up-regulation of numerous TAF7L target genes (Fig. 3C), and qRT-PCR data validated several of these genes including Spert, Spem1, Prm2, Spata3, Tnp1, and Tnp2 (Figs 2C, 4B). Thus, TAF7L activation leading to up-regulation of target genes appears to be due to factors other than increased expression of TAF7L.
Correlation with disease genes. DEGs of Mgat1 cKO germ cells were compared by IPA with genes implicated in diseases. Disease categories impacted in 22 and 23 dpp Mgat1 cKO germ cells include Teratozoospermia and Asthenozoospermia diseases of men. Up-regulated DEGs from the 22 dpp dataset are shown in Fig. 3D. The blue dotted lines indicate the predicted suppression of disease in Mgat1 cKO germ cells, as up-regulation of spermatogenic and spermiogenic genes is predicted to reduce development of diseases such as Teratozoospermia and Asthenozoospermia. Comparison of heat maps for disease conditions at 22 and 23 dpp shows that sperm disorder, Asthenozoopermia, Teratozoospermia, and Oligoszoospermia, as defined in men with fertility disorders, were reduced, whereas sperm capacitation and quantity of male germ cells were increased (Fig. 3D, right panel). Genes related to fertility were increased relatively more at 22 than 23 dpp, whereas genes related to cell movement of sperm were relatively increased at 23 compared to 22 dpp (Fig. 3D).
Interaction networks of DEGs. IPA was used to map biological relationships of the Mgat1 cKO germ cell DEGs into networks based on published literature. At 22 dpp, a 26-gene network with a score of 44 was the most significant. The functions associated with network 1 are cellular development, reproductive system development and function, and cell morphology. The central molecule in this network is ERK1/2, but it is not indicated as activated or inhibited, and neither ERK1 nor ERK2 are amongst the DEGs in our datasets. PKA, several AKAPs and CAM kinase 4 genes were predicted to be up-regulated. Spermiogenesis-specific genes that include transition protein 2 (Tnp2) and protamine 2 (Prm2) were also predicted to be up-regulated (Fig. 4A). Some extracellular matrix genes such as laminin and collagen, indirectly associated with network 1 and not predicted to change, were removed for clarity. The genes in network 1 overlapped with canonical pathways that show a direct or indirect relation with ERK1/2 (Fig. 4A). Five of the up-regulated genes (Akap1, Akap3, Akap4, Tnp1, Camk4) and 1 down-regulated gene (Figf) were validated by qRT-PCR (Fig. 4B). Camk4 is a protein kinase that phosphorylates Signaling through ERK1/2 is reduced in Mgat1 cKO germ cells. GSEA analysis showed that one of the most significant oncogenic signatures (C6) is PDGF ERK (FDR < 0.001), which was enriched in control versus Mgat1 cKO germ cells at 22 dpp (Fig. 6A). The heat map shows core enrichment genes that affect the pathway that were up-regulated in control germ cells, including PDGFRA, which regulates ERK1/2 signaling (Fig. 6B). Microarray data and qRT-PCR identified PDGFRA as well as EGFR as down-regulated genes in Mgat1 cKO germ cells at 22 and 23 dpp (Fig. 2D). Consistent with this, western blot analysis showed that at 22 dpp, pERK1/ERK1 and pERK2/ERK2 levels were markedly reduced in Mgat1 cKO germ cells compared to controls. However, pAKT/ AKT levels were not significantly altered by the loss of Mgat1 (Fig. 6C,D; Supplemental Fig. S7).

Basigin signaling is reduced in Mgat1-null Lec1 CHO cells. Basigin is a germ cell glycoprotein with
complex N-glycans, and the efficient deletion of Mgat1 in spermatogonia generates basigin with oligomannosyl N-glycans 3 ( Fig. 2A). Deletion of basigin causes the complete loss of Griffonia simplicifolia II (GSA) lectin binding to germ cells 10 , suggesting that basigin is a major carrier of germ cell complex N-glycans. Thus, the phenotype of Mgat1 cKO males may arise, in part, from functional defects in signaling by basigin carrying oligomannosyl, rather than complex, N-glycans. To investigate this hypothesis, we examined ERK signaling induced by mouse basigin in Chinese hamster ovary (CHO) cells that express complex N-glycans, compared to Lec1 Mgat1-null CHO mutant cells. The Lec1 CHO mutant 5 is a cell-based model for Mgat1 cKO germ cells that lacks MGAT1 activity and expresses only oligomannosyl N-glycans on glycoproteins 6 . We also investigated the expression of connexin 43 (Cx43), a component of the blood-testis-barrier (BTB). The BTB is altered in basigin-null testes 10 .
Cx43 immunoreactivity was observed along the basal compartment of seminiferous tubules between spermatogonia and spermatocytes of 28 dpp control and Mgat1 cKO testis sections, in a pattern typical of BTB proteins (Fig. 7). Immunoreactivity was scored blindly as high (score 3), medium (score 2), or low (score 1). There was a significant reduction in score 2 and score 3 staining in Mgat1 cKO sections, suggesting that the BTB may be compromised in Mgat1 cKO testis. There are numerous ligands for basigin (including basigin itself and the soluble extracellular domain of basigin), that induce ERK1/2 phosphorylation 15 . To determine if the nature of the N-glycans on basigin could affect basigin signaling, CHO and Lec1 cells were co-transfected with a mouse basigin cDNA, and a plasmid encoding neomycin resistance. G418-resistant transfectants were sorted for high expression of basigin on the cell surface (Fig. 8A,B). Sorted populations of CHO and Lec1 cells expressing equivalent amounts of cell surface basigin, were serum-starved for 24 hr, and medium was replaced with serum-free medium with and without potential ligands, including cyclophilin A (CypA 100, 250 or 500 ng/ml), or 10% fetal calf serum, or 10 μg/ml anti-basigin Ab. After incubation for 15 min at 37 °C, cells were washed and lysates prepared for western blot analysis. Surprisingly, the inclusion of CypA, FCS or anti-basigin Ab in serum-free medium did not consistently stimulate signaling in CHO or Lec1 basigin transfectants (Supplementary Fig. S8 and not shown). By contrast, activation of ERK1/2 was observed in CHO cells expressing basigin (CHO + Bsg) compared to control CHO cells (CHO) after incubation in serum-free medium alone (Fig. 8C,D; supplemental Fig. S9). The "ligand" in this case may be homomeric interactions between basigin on adjacent cells 15 . Lec1 cells exhibited somewhat higher pERK1/2 levels compared to the CHO cells from which they were derived 5 . However, in contrast to CHO + Bsg cells, signaling was not further stimulated by the presence of basigin in Lec1 + Bsg cells (Fig. 8E,D). A significant difference between CHO and Lec1 cells expressing basigin was apparent when the ratios of CHO + Bsg/CHO and Lec1 + Bsg/Lec1 were compared for ERK1/2 (Fig. 8D). Thus, the combined data are consistent with the hypothesis that basigin carrying oligomannosyl N-glycans signals less well than basigin carrying complex N-glycans.

Discussion
In this paper we investigated mechanisms that lead to defective spermatogenesis in Mgat1 cKO males following Mgat1 deletion in spermatogonia at 3 dpp. Morphological changes began to appear at 24-25 dpp in a small proportion of mutant tubules. About 14% of tubules showed spermatid MNC by 28 dpp. Sertoli cell, spermatogonia and spermatocyte numbers per round tubule were unaffected in Mgat1 cKO testes at the same stage. To obtain insights into early events that might be the basis of defective spermatogenesis in Mgat1 cKO males, we interrogated gene expression changes in morphologically normal Mgat1 cKO germ cells from 22 and 23 dpp males. Surprisingly, we discovered that many genes involved in later stages of spermatogenesis and spermiogenesis were up-regulated in 22 and 23 dpp Mgat1 cKO germ cells. The majority of up-regulated genes (>85%) encode transcripts that promote spermatogenesis. Therefore, the loss of Mgat1 in spermatogonia at 3 dpp leads to mutant germ cells at 22 and 23 dpp in which genes normally turned on much later in spermatogenesis, are prematurely up-regulated. To gain insights into the repertoire of genes connected with functions of MGAT1 in germ cells, we applied molecular network analyses using IPA, GO and GSEA. IPA identified sperm motility as the top canonical pathway with a positive activation Z score, and cAMP signaling as the top canonical pathway with a negative activation Z score. Top biological functions in both 22 and 23 dpp Mgat1 cKO germ cells were cellular function and maintenance, and reproductive system. Top upstream regulators were TAF7L and FIGLA. Taf7l null males have a post-meiotic block in spermiogenesis and are sterile 16 . This block occurs beyond the stage at which spermatogenesis is disrupted in Mgat1 cKO males. FIGLA functions in oocytes to suppress male germ cell genes involved in spermatogenesis and spermiogenesis 17 . YBX2/MSY2 is a germ cell specific RNA and DNA binding protein of the cytoplasm, most highly expressed in round spermatids during spermatogenesis. Deletion of Ybx2/Msy2 leads to incomplete nuclear condensation in spermatids, and a block in spermatogenesis 18,19 . The poly(A) polymerase PAPOLB/TRAP polyadenylates a subset of transcription factors and other mRNAs in the cytoplasm. It is expressed in round spermatids and is required for spermatogenesis 20 . Top disease processes Teratozoospermia and Asthenozoospermia were identified as being inhibited. This suggests that the spermatogenesis-specific genes up-regulated in Mgat1 cKO germ cells normally guard against the development of both these fertility diseases in men. Consistent with this interpretation, GSEA analysis revealed gene enrichment in Mgat1 cKO germ cells overlapping most significantly with gamete generation, Matzuk spermatid differentiation and Matzuk spermatozoa formation. In summary, the complement of genes up-regulated in 22 and 23 dpp Mgat1 cKO germ cells suggest that germ cells that lack Mgat1 are attempting to differentiate prematurely by up-regulating genes involved in spermatogenesis and spermiogenesis. This means that MGAT1 and complex N-glycans on germ cell glycoproteins function during normal spermatogenesis to control differentiation by retarding the expression of genes early in spermatogenesis that are required for later stages of spermatogenesis and spermiogenesis.
The mechanism(s) by which MGAT1 and complex N-glycans on glycoproteins regulate gene expression must necessarily be indirect. It is well established in cell-based experiments that loss or reduced branching of complex N-glycans leads to reduced cell surface residence time of glycoprotein receptors due to the weakening of their interactions with a galectin lattice [21][22][23] . On this basis it would be predicted that growth factor signaling should be reduced in Mgat1 cKO germ cells. In fact, our microarray and qRT-PCR data showed that transcripts of both the EGF and PDGF receptors (Egfr and Pdgfra) were markedly reduced in Mgat1 cKO germ cells, and the top IPA network identified ERK1/2 as regulating numerous genes in Mgat1 cKO germ cells. Thus, on several counts, we predicted that growth factor signaling should be reduced in Mgat1 cKO germ cells. Western blot analysis showed that pERK1/2 levels were indeed markedly reduced in Mgat1 cKO germ cells. The MAP kinase pathway that leads to ERK1/2 activation is active in spermatogonia and primary spermatocytes and diminished in pachytene spermatocytes [24][25][26] . Interestingly, pAKT levels were not reduced, consistent with the fact that the block in spermatogenesis in Mgat1 cKO males is distinct from that in males lacking AKT which is essential for survival and proliferation of pre-and post-meiotic cells 27 . When signaling via basigin was investigated in CHO versus Lec1 cells that lack Mgat1, pERK1/2 levels were enhanced by the presence of basigin in CHO but not Lec1 cells. Basigin is a substrate of MGAT1 3 , and is a major carrier of complex N-glycans of germ cell glycoproteins 10 . Basigin has potential ligands in testis that might lead to activation of ERK1/2 in germ cells 15 including basigin itself and soluble basigin extracellular domain 28 . Importantly, the spermatogenesis defect in basigin-null males is similar to that in Mgat1 cKO males 10 . The combined data suggest that defective signaling via ERK1/2 due to the loss of complex N-glycans on basigin leads, along with other factors that reduce ERK1/2 signaling in Mgat1 cKO germ cells, to the block in spermatogenesis in germ cells lacking Mgat1. Testing this hypothesis forms the basis of future experiments in males conditionally lacking basigin in germ cells, and related mouse models. Data are mean ± SEM from 9-13 gels of lysates run in 6 independent experiments. *p < 0.05 in (D) left panel is based on a one-tailed, unpaired Student's t test with Welch's correction; in (E, left panel) significance is based on the non-parametric, two-tailed Wilcoxon matched-pairs signed rank test; in (E) right panel it is based on the unpaired, two-tailed Student's t test with Welch's correction. **p < 0.01, ***p < 0.005 and ****p < 0.001.

Endoglycosidase H digestion.
Lysate containing 20-60 μg testis protein was treated with 5 mU Endo H from S. plicatus (#11088726001, Roche Diagnostics, Manheim, Germany) or water in 20 μl manufacturer's buffer at 37 °C for 2 hr. Reactions were stopped by adding SDS gel loading buffer and heating at 95 °C, 10 min. Microarray analysis. Testis RNA (150 ng, RIN ≥ 9) was provided to the Genomics Core Facility of the Albert Einstein College of Medicine for conversion to cDNA, labeling and hybridization to Affymetrix GeneChip ™ Mouse Gene 2.0 ST Array (Affymetrix, Santa Clara, CA, USA). After gene level normalization, signal summarization and background subtraction, raw intensity data (.CEL files) were transformed to .CHP files using Affymetrix Expression Console software. Genes up-or down-regulated with a fold-change >2.0 or <−2.0 were determined using the Affymetrix Transcriptome Analysis Console. Differences between medians (log 2 ) were determined and transformed to linear fold-change. Statistical significance was assessed using ANOVA and FDR p values.
Quantitative PCR. RNA (500 ng) was reverse-transcribed into cDNA using the verso cDNA synthesis kit (#AB1453/B, Thermo Fisher) with an oligo-dT primer according to the manufacturer's protocol. Real time PCR was performed using Absolute Blue QPCR mix (#AB4162, Thermo Fisher) on a master cycler (ViiA 7, Thermo Fisher). PCR conditions were 95 °C 30 sec, followed by 40 cycles at 95 °C 15 sec, 60 °C 15 sec and 72 °C 20 sec. Gene expression relative to actin (Actb) was calculated as log 2 dCT , Mgat1 cKO values were subtracted from control (log 2 ddCT ) and converted to fold-change.