The Golgi Glycoprotein MGAT4D Protects Testicular Germ Cells From Mild Heat Stress

MGAT4D was identified as an inhibitor of MGAT1 in the Golgi of transfected cells. MGAT1 is the GlcNAc-transferase that initiates complex and hybrid N-glycan synthesis in mammals. Conditional deletion of Mgat1 in spermatogonia leads to a block in spermatogenesis and infertility in mice. Mgat4d is highly expressed in spermatocytes and spermatids. Here we show that, unexpectedly, when the Mgat4d gene was inactivated globally or conditionally in spermatogonia, or mis-expressed in spermatogonia, spermatocytes or spermatids, neither spermatogenesis nor fertility were affected. MGAT1 activity was also not inhibited in germ cells. On the other hand, when males were subjected to mild heat stress of the testis (43°C for 25 min), a novel function for MGAT4D was revealed. Germ cells with inactivated Mgat4d were markedly more sensitive to the effects of mild heat stress, and transgenic mice expressing Mgat4d in spermatogonia, spermatocytes or spermatids were partially protected from heat stress. Microarray data from germ cells of heat-stressed mutant versus control males showed that germ cells lacking Mgat4d generally mounted a similar heat shock response to control germ cells, but could not maintain that response. Several pathways and genes activated by heat stress in wild type were induced to a lesser extent in Mgat4d[-/-] heat-stressed germ cells, including the NFκB response, pro-inflammatory pathways such as TNF and TGFβ signaling, and genes that promote proliferation such as Hif1α and Myc. Thus, MGAT4D protects male germ cells from the deleterious consequences of heat stress.

such as Hif1a and Myc. Thus, MGAT4D protects male germ cells from the deleterious consequences of heat stress.

Introduction
MGAT4D has been designated family member D of the MGAT4 gene family by the Human Genome Nomenclature Committee based on sequences similarity to other members, including MGAT4A and MGAT4B. The latter are N-acetylglucosaminyltransferases (GlcNAcTs) that add a b1,4GlcNAc to complex N-glycans. However, when MGAT4D is transfected into cultured cells, it does not appear to have GlcNAcT activity. Rather, it inhibits MGAT1 activity, the GlcNAcT responsible for initiating complex N-glycan synthesis (Huang & Stanley, 2010). Because of this inhibitory activity, the protein was termed GnT1IP for GlcNAcT1 Inhibitory Protein. The Mgat4d gene is most highly expressed in mouse testis with very low expression in other mouse tissues . RNA-seq analysis showed it to be expressed in spermatocytes and spermatids, but not in spermatogonia, sperm or Sertoli cells (Huang et al., 2015). MGAT4D is the most abundant protein in purified Golgi from rat testis germ cells (Au et al., 2015). Characterization of the interactions of MGAT4D in the Golgi using a fluorescence resonance energy transfer (FRET) assay showed that it interacts with MGAT1 but not MGAT2, MGAT3, MGAT4B or MGAT5 (Huang et al., 2015). Since knockout of Mgat1 in spermatogonia disrupts spermatogenesis and results in infertility (Batista, Lu, Williams, & Stanley, 2012;Biswas, Batista, Sundaram, & Stanley, 2018), deletion or overexpression of Mgat4d in germ cells were both expected to have effects on spermatogenesis. In this paper, we show that unexpectedly, deletion of Mgat4d globally, or specifically in spermatogonia, or mis-expression of Mgat4d in spermatogonia, spermatocytes or spermatids, do not appear to alter spermatogenesis in young or aged mice, and do not affect fertility. However, mild heat stress of the testis in aged mice revealed that germ cells lacking Mgat4d exhibited more damage and apoptosis following heat stress. By contrast, a Mgat4d transgene expressed in spermatogonia, spermatocytes or spermatids under respective germ cell specific promoters, led to comparative resistance to mild heat stress. Gene expression microarray and bioinformatics analyses showed that germ cells lacking Mgat4d responded to heat stress by initially upregulating heat shock and related genes. However, in contrast to controls, germ cells lacking Mgat4d did not sustain this response, nor upregulate anti-inflammatory and anti-apoptotic protective genes to the same degree as wild type germ cells. The combined data identify a new function for MGAT4D, a glycoprotein that resides in the Golgi, as a protector against male germ cell heat stress.

Effects of global and conditional deletion of Mgat4d on spermatogenesis and fertility.
Embryonic stem cells (ES Cells) carrying the construct Mgat4d tm1a(KOMP)Wtsi designed to conditionally delete exon 4 of the Mgat4d gene ( Figure 1A) were obtained from the Knockout Mouse Project (KOMP) repository. Following injection into C57BL/6J blastocysts, chimeras were crossed to C57BL/6J to obtain mice carrying the conditional Mgat4d tm1a(KOMP)Wtsi allele. Male progeny were crossed with FVB Stra8-iCre (Sadate-Ngatchou, Payne, Dearth, & Braun, 2008) or Flp1-Cre transgenic females (129S4/SvJaeSor-Gt(ROSA)26Sor tm1(FLP1)Dym /J) (Farley, Soriano, Steffen, & Dymecki, 2000). Stra8 is expressed in spermatogonia from 3 days post-partum (dpp) and the Flp1-Cre was expressed from the ROSA26 locus. Male mice with global (Mgat4d[-/-]) or conditional (Mgat4d[F/F]:Stra8-iCre) inactivation of the Mgat4d gene were generated, and males expressing LacZ from the Mgat4d promoter were also obtained ( Figure 1A). Both strains were crossed to FVB mice and maintained on a FVB background because Mgat1 deletion was performed on the FVB background (Batista et al., 2012). Genotyping PCR identified Mgat4d [+], , Mgat4d[F] alleles and Stra8-iCre ( Figure 1B). Primer sequences, locations and expected product sizes are given in Figure 1-table supplement 1. Polyclonal rabbit antibodies (pAb) prepared against a C-terminal peptide of MGAT4D identified the long form (MGAT4D-L) and the short form (MGAT4D-S) which lacks 44 amino acids at the N-terminus of MGAT4D-L, and mice with inactivated Mgat4d had no signal, as expected ( Figure 1C). Detection of LacZ expression by beta-galactosidase activity showed that the Mgat4d promoter is active mostly in spermatocytes and spermatids in testis tubules ( Figure 1D), consistent with results of RNA-seq analysis (Huang et al., 2015). Immunohistochemistry for MGAT4D on testis sections from Mgat4d[+/-] or wild type males shows staining in the Golgi of spermatocytes and round spermatids, but not in spermatogonia or spermatozoa ( Figure 1E), as observed in rat testis (Au et al., 2015). Testis sections from Mgat4d[-/-] males showed no staining, as expected ( Figure 1E).
Mgat4d [-/-] males and females were fertile and transmitted the inactivated gene according to the expected Mendelian distribution (Table 1). Male mice with conditional deletion of Mgat4d in spermatogonia also showed no defects in fertility on a FVB background, or after backcrossing 10 generations to C57BL/6J mice (Table 1). Based on histological analyses, testicular weight and analysis of sperm parameters (sperm count, viability, morphology, motility and acrosome reaction), no obvious defects in spermatogenesis were observed in Mgat4d [-/-] males. In addition, aging (up to 596 dpp for FVB and 482 dpp for C57BL/6J) did not reveal apparent histological differences in spermatogenesis between mutant and control males (data not shown).
As discussed in the Introduction, MGAT4D was initially described as an inhibitor of MGAT1 activity and termed GnT1IP (Huang & Stanley, 2010). By deleting such an inhibitor, we expected MGAT1 activity might increase, and the level of complex N-glycans on glycoproteins might also increase. We determined MGAT1 GlcNAc transferase activity in germ cell extracts. Germ cells were purified from 28 dpp C57BL/6J wild type (n=4) and Mgat4d [-/-] (Huang et al., 2015).
We also investigated testis lysates and testis sections for an increase in complex N-glycan levels. Basigin (CD147, BSG) carries complex and some oligomannosyl N-glycans in testis (Batista et al., 2012;Biswas et al., 2018). Thus, loss of an MGAT1 inhibitor might increase the level of complex N-glycans on BSG. However, no significant increase in the level of endoglycosidase H (Endo H) resistant BSG (carries complex N-glycans) was observed between Males lacking Mgat4d are more sensitive to mild heat stress of the testis Given the apparent lack of significant consequences for spermatogenesis of removing Mgat4d, we investigated whether stressing testicular germ cells would reveal any effects of Mgat4d loss.
Spermatogenesis is sensitive to an increase in temperature (Durairajanayagam, Agarwal, & Ong, 2015;Moreno, Lagos-Cabré, Bunay, Urzúa, & Bustamante Marin, 2012) and we reasoned that disturbing tissue homeostasis using mild heat stress might reveal roles for MGAT4D in testis. The remaining cohort of aged Mgat4d[+/-] and Mgat4d[-/-] FVB mice of between 592 and 596 dpp were anesthetized and subjected to mild heat stress by immersing the lower half of the body in water at 43°C for 25 min. Mock treatment involved the same procedure with a water temperature of 33ºC. After recovery for 24 hr, testes were harvested. One testis was used for histological analysis and the other for RNA and protein extraction. While testis sections from males treated at 33ºC appeared normal, 43ºC treatment caused the appearance of enlarged (³10 µm) multinucleated cells, large vacuoles (³10 µm), small vacuoles and pyknotic cells in testis tubules ( Figure 2A). Spermatozoa in the epididymis also included pyknotic cells following heat stress ( Figure 2B). Compared to controls, Mgat4d[-/-] testis sections exhibited an increased number of tubules (~3.5-fold) with enlarged cells, and a decrease in undamaged tubules (~2-fold). ( Figure   2C). No significant difference was found in testis weights of heat-treated versus control mice Heat stress is reported to increase apoptosis in differentiating germ cells (Durairajanayagam et al., 2015;Y. Li et al., 2009;Y. S. Li et al., 2013) and so testis sections from heat-and mock-treated aged FVB males were subjected to the "Apoptag" assay and staining was quantified using FIJI software (https://fiji.sc/). As expected, apoptosis increased in sections from control heat-treated males sacrificed 24 hr after heat treatment ( Figure 3A, 3B). However, testes from Mgat4d[-/-] mice showed ~2-fold more apoptotic germ cells than Mgat4d[+/-] controls ( Figure   3B-3D). Thus, based on histology and levels of apoptosis, the effects of heat stress were more severe for aged Mgat4d[-/-] testes than for heterozygous testes.

Mgat4d transgenic mice are resistant to the effects of heat stress.
Mice with targeted deletion of Mgat1 in testicular germ cells exhibit defective spermatogenesis and are infertile (Batista et al., 2012). Thus, it was expected that inhibiting MGAT1 activity by increasing the level of MGAT4D in germ cells, would induce defects in mouse spermatogenesis.
To investigate, C57BL/6J transgenic males expressing a Mgat4d-L-Myc cDNA in specific germ cell types were generated. This transgene has previously been shown to inhibit MGAT1 in transfected cells (Huang et al., 2015;Huang & Stanley, 2010). The Stra8 (Stimulated By Retinoic Acid 8) promoter was used to express the transgene in spermatogonia (Batista et al., 2012;Biswas et al., 2018;Sadate-Ngatchou et al., 2008), the Ldhc (Lactate Dehydrogenase C) promoter was used to express in spermatocytes (S. Li, Zhou, Doglio, & Goldberg, 1998;Tang, Kung, & Goldberg, 2008), and the Prm1 (Protamine 1) promoter was used to express in respectively. They were genotyped by PCR of genomic DNA using primers described in Figure   1-table supplement 1, and transgene expression was shown to be 3-6-fold greater than endogenous Mgat4d-L levels using quantitative RT-PCR (qRT-PCR) on cDNA from testis ( Figure   4B). qRT-PCR using primers specific for the Myc sequence gave a similar level of expression based on Ct values (not shown). Myc transcripts could not be quantitated relative to the control that has no transgene. By contrast, attempts to determine MGAT4D-L-Myc protein levels in testis extracts by western blot analysis using anti-Myc monoclonal antibodies (mAb) from several species were not successful, although MGAT4D-L-Myc overexpressed in CHO cells is detected by anti-Myc mAb (Huang et al., 2015). We generated C-and N-terminal peptide-purified rabbit pAbs that detect MGAT4D-L-Myc or Myc-MGAT4D-L, respectively, in transfected CHO cells Histological analysis of testis sections showed no obvious changes in spermatogenesis or testicular structure in adult transgenic mice ( Figure 4C). In addition, the fertility of transgenic males was normal, although Stra8-Mgat4d-L-Myc mice showed low transgene transmission from transgenic males (Table 1). Males from the three transgenic mouse strains and non-transgenic littermates or wild type C57BL/6J controls were subjected to mild heat stress. No significant difference was observed in testis weights of mock-versus heat-treated mice (Figure 2-table   supplement 1). Importantly, however, each transgenic strain showed an ~3-fold reduction in the number of tubules with enlarged germ cells, and ~2-fold fewer had tubules with large vacuoles ( Figure 5A, 5B). The number of undamaged tubules was also increased but small vacuoles and pyknotic cells were present in heat-treated transgenic mice ( Figure 5C, 5D). The "Apoptag" assay revealed an ~2-fold reduction in apoptotic germ cells in all three transgenic strains ( Figure 6). We also investigated previously reported gene expression changes due to heat stress. In wild type Thus, on the basis of several criteria, the presence of a Mgat4d-L-Myc transgene in germ cells gave significant protection from heat stress.

Molecular basis of the increased sensitivity of Mgat4d[-/-] germ cells to heat stress
The histological and apoptotic changes induced by heat stress reported above were observed in an aged cohort of 1.6 year FVB mice. We subsequently tested 7 month C57BL/6J Mgat4d [-/-] mice and did not observe increased sensitivity to heat stress (data not shown). However, protection from heat stress was observed in adult C57BL/6J transgenic mice as shown here.
Thus, to determine whether Mgat4d[-/-] mice on a C57BL/6J background exhibited a more sensitive response to heat stress than controls, and to also gain insights into molecular mechanisms that underlie this phenotype, microarray analyses were performed on cDNA from purified germ cells of control and Mgat4d[-/-] C57BL/6J mice of 2 months.

Mgat4d[+/+] and
Mgat4d [-/-] males were treated at 33ºC or 43ºC for 25 min and sacrificed after 8 hr, a time when no visible histological changes to germ cells were observed (data not shown). Testes were enzymatically dissociated and germ cells were isolated and counted. RNA preparations with a RIN value >7.9 were used to make cDNA for microarray analysis. Purity of germ cells was assessed by relative expression of germ cell-specific and non-germ cell genes to the same genes expressed in testis RNA as previously described (Biswas et al., 2018). The Mouse Clariom TM D GeneChip™ Mouse Transcriptome Array 1.0 from Affymetrix was used. Custom scripts using the R/Bioconductor tools affymetrix and limma were used to process the raw (.CEL) files and to compare Mgat4d[-/-] versus Mgat4d[+/+] microarray data from 33ºC-and 43ºC-treated mice. The samples displayed a moderate clustering by genotype, as seen in PCA plots ( Figure 8A). Importantly, significant differences between genotypes were much less pronounced at 33ºC than at 43ºC, as witnessed by the tighter correlation in the heat maps, and a lower number of differentially expressed genes (DEGs) between genotypes shown in volcano plots ( Figure 8B, 8C). However, a clear difference was evident between wild type and Mgat4d[-/-] arrays from germ cells of mice treated at 43ºC. Given the importance of the temperature as a confounding variable, it was included in modelling differential gene expression between genotypes. DEGs in mutant versus wild type germ cells at 33ºC and at 43ºC were determined, and the interaction between

To find the most significantly represented pathways differentially altered in Mgat4d[-/-] versus
Mgat4d[+/+] germ cells following heat stress, we examined the relationship between DEGs at a ±1.5 fold change with adjusted FDR<0.05 and p<0.05 using IPA. Interestingly, the top canonical pathways were mostly down-regulated or with "no activity pattern available" in 43°C-treated Mgat4d[-/-] versus control germ cells, and were related to recovery from stress conditions ( Figure   9C). Ranked by -log(p-value), the top down-regulated pathway was Acute Phase Response Signaling (p=5.2; Z-score -2.45) followed by LXR/RXR Activation (p=5.04; Z-score 0.707), the only pathway with a positive z-score and -log(p-value) higher than 2. NRF2-mediated Oxidative Stress Response (p=4.98; Z-score -2.111) and Glutathione-mediated Detoxification (p=3.37; Zscore -2) were also top pathways.

Top upstream transcriptional regulators.
IPA was used to predict the top upstream transcriptional regulators in the DEGs based on their gene targets. The algorithm calculates a p-value on the basis of significant overlap between genes in our 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. This analysis identified 323 upstream regulators with a p-value of overlap <0.05 and a Z-score greater than or equal to +/-2.

Most represented networks, toxicological functions, diseases and biological functions.
DEGs in germ cells from heat-treated mice were compared by IPA with genes belonging to specific biological networks or implicated in diseases. The most highly ranked network was "DNA Replication, Recombination, and Repair, Nucleic Acid Metabolism, Small Molecule Biochemistry" with 28 focus molecules (Figure 9-table supplement 3). The top diseases and biological functions were related to "Organismal Survival" -19 biological functions were predicted to be increased with an activation Z-score between 6.131 and 2.01, mostly related to inflammation, injury and disease ( Figure 9-table supplement 4), but a higher number of diseases or functions were predicted to be decreased (71). The top category was "Lipid Metabolism, Small Molecule Biochemistry,Vitamin and Mineral Metabolism" and the most represented of these were related to cellular function.
Of the eight categories of gene sets, the Hallmark collection summarizes well-defined biological processes and states from v4.0 MSigDB collections C1 through C6 (Liberzon et al., 2015). Hallmark EnrichR (E. Y. Chen et al., 2013). Heat maps highlight some of the informative EnrichR gene sets and also show illustrative gene expression differences identified ( Figure 10A). The overall results suggest that Mgat4d[-/-] germ cells have a problem responding to heat shock stress, e.g. coping with hyperthermic stress through clearance of damaged proteins (Casp8; Figure 10B). A number of pathways and genes were induced to a lesser extent in Mgat4d[-/-] heat-stressed mice, including Hif1a, the NFk kB response, pro-inflammatory pathways such as TNF and TGFb signaling, and genes that promote proliferation such as Myc ( Figure 10B) .

Discussion
In this paper we reveal a novel function for MGAT4D as a protector of male germ cells against heat stress. This conclusion is supported by several different pieces of evidence. First, we showed that an old cohort of Mgat4d[-/-] males were more sensitive to mild testicular heat stress than heterozygote controls, as evidenced by increased germ cell defects and apoptosis at 24 hr after heat stress. Second, we found that mice expressing a Mgat4d-L-Myc transgene in either spermatogonia (Stra8 promoter), spermatocytes (Ldhc promoter) or spermatids (Prm1 promoter) were less sensitive to testicular heat stress than wild type controls, based on reduced germ cell defects and reduced apoptosis. Characterization of individual gene expression changes for genes known to exhibit increased or decreased expression following heat stress, showed that males expressing the Stra8-Mgat4d-L-Myc transgene were comparatively resistant to heat stress and, at 43ºC, behaved similarly to non-transgenic germ cells treated at 33ºC, whereas non-transgenic males treated at 43C showed the marked gene expression changes predicted.
To investigate gene expression differences in more depth, we performed microarray analyses on Mgat4d wild type and Mgat4d[-/-] germ cells prepared only 8 hr after males were treated at 43ºC for 25 min, when no histological changes were apparent. Comparisons of DEGs and bioinformatics analyses using IPA, GSEA and EnrichR revealed that Mgat4d[-/-] heat-treated germ cells responded initially to heat stress, but did not sustain that response like wild type, heat- A key question for the future is to determine how MGAT4D protects against heat shock in male germ cells. Interestingly, Mgat4d transcripts are markedly reduced by the 43ºC treatment and yet if MGAT4D is not present, germ cells are more sensitive to heat treatment, and if a Mgat4d transgene is present, germ cells are comparatively protected. Thus, the presence of MGAT4D, which may perdure in wild type germ cells after Mgat4d transcripts are reduced after heat stress, appears to facilitate the sustained heat stress response observed in wild type germ cells. How this is accomplished by a type II transmembrane Golgi glycoprotein may be related to the effects of Golgi glycosyltransferases on Golgi fragmentation. Some Golgi glycosyltransferases of the medial and trans Golgi compartments have been shown to facilitate Golgi fragmentation after heat shock (Petrosyan & Cheng, 2013, 2014. For example, the mucin O-glycan GlcNAcT CGNT3 promotes Golgi fragmentation following heat shock by interacting with myosin IIA via its cytoplasmic tail (Petrosyan & Cheng, 2014). MGAT4D is the most abundant protein in rat Golgi of male germ cells (Au et al., 2015) and its loss after heat shock may protect the Golgi from fragmentation and protect Golgi glycosyltransferases and other Golgi residents, including molecules that protect from Inflammation and autophagy and that promote proliferation and survival, from degradation by the proteasome (Petrosyan & Cheng, 2014). No: 001800 respectively) and used for breeding. All mice carrying a transgene were kept as heterozygotes by crossing +/Tg with homozygote wild-type (+/+) mice. Mice were sacrificed by carbon dioxide asphyxiation followed by cervical dislocation. Testes were dissected free of surrounding tissue and weighed. Animal experiments were performed following the relevant guidelines approved by the Albert Einstein Institutional Animal Care and Use Committee.

Immunohistochemistry
Testes were fixed in Bouin's fixative (#100503-962, Electron Microscopic Sciences, Radnor, PA) for 48 hr at room temperature (RT) then processed and paraffin-embedded by the Einstein Histology and Comparative Pathology Facility. Serial sections (5-6 μm) were collected on positively-charged slides.
Immunohistochemistry was performed following the "IHC staining protocol for paraffin-embedded sections" from Abcam (http://www.abcam.com/protocols/). Briefly, testis sections were deparaffinized using Histo-Clear reagent Cat no. HS-200 (National Diagnostics, Atlanta, GA). We performed a heatinduced epitope retrieval with citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) at 100°C for 20 min followed by 20 min period at room temperature in the same buffer. The tissue was permeabilized with 0.1% Triton X-100 in Tris-buffered saline (TBS) for 10 min and blocked for 1 hr at room temperature with 10% normal serum (from the same species as secondary antibody) and 1% BSA in TBS. The primary antibody was diluted in TBS with 1% BSA and incubated overnight at 4°C (unless otherwise indicated). Endogenous peroxidase was quenched by incubating slides in 1.5% H 2O2 in TBS for 10 min and rinsed before incubation with the Biotinylated secondary antibody diluted in TBS containing 1% BSA, for 1 hr at room temperature. The samples were washed and Vectastain® ABC-HRP reagent (cat no. PK-6100, Vector laboratories, Inc. Burlingame, CA) was added and incubated at room temperature for 30 min. After rinsing, peroxidase substrate 3,3′diaminobenzidine (DAB) (Vector laboratories, Cat# SK-4100) was used to detect the antibody, following the manufacturer protocol. The tissue was counter-stained with Mayer's Hematoxylin solution (cat no. MHS16-500ML, Sigma-Aldrich).
The specimens were dehydrated with histo-clear and mounted using Permount® reagent (cat no. SP15-100, Fisher Scientific, Fair Lawn, NJ). Testis section images were produced using 3DHistec Panoramic 250 Flash II slide scanner obtained with the Shared instrumentation Grant SIG# 1S10OD019961-01 to the Analytical Imaging Facility (AIF) of the Albert Einstein College of Medicine.

Western-blot analysis
Testis tissue lysates were prepared using Millipore,Temecula,CA) and following the protocol "Preparation of lysate from tissues" from Abcam with modifications. Briefly, the testis tissue was homogenized in 1X RIPA, 01% SDS, 1X protease inhibitor cocktail (cat no. 05892791001, Roche Diagnostics GmbH, Mannheim, Germany) at a ratio of 0.5 ml buffer for 0.05 g of tissue. The lysate was incubated with constant agitation (orbital shaker) at 4°C for 2 hr and then centrifugated for 20 min at 12000 rpm at 4°C. The supernatant was transferred to a fresh tube and supplemented with 100% glycerol to a final concentration of 20% glycerol. Protein yield was measured using Bradford based colorimetric assay, (cat no. 500-0006, Bio-Rad Protein assay, Bio-Rad, Hercules, CA). Isolated germ cell were lysed in buffer containing 1% IGEPAL, 1%TX-100, 0.5% Deoxycholate and 1X protease inhibitor cocktail in water. Briefly, 100 µl of lysis buffer was used to homogenize 10 7 cells.
The lysate was incubated for 30 min on ice, then centrifugated 5 minutes at 5000 g. The supernatant was transferred to a fresh tube and supplemented with 100% glycerol to a final concentration of 20% glycerol.
Protein levels were measured using the Bradford-based colorimetric assay. All samples were stored at -80°C.

Apoptosis assay
Apoptosis induced DNA damage was measured using the ApopTag® Peroxidase In Situ Apoptosis Detection Kit (cat no. S7100, EMD Millipore, Temecula, CA) following the manufacturer's protocol for paraffin-embedded tissue. Testis sections were deparaffinized using Histo-Clear reagent (cat no. HS-200, National Diagnostics, Atlanta, GA). Stained slides were scanned using a Perkin Elmer P250 high capacity slide scanner and images were analyzed using FIJI software to count foci (Schindelin et al., 2012).

Germ cells isolation
Male germ cells were purified from testis following a modified protocol (Abou-Haila & Tulsiani, 2001;Chang, Lee-Chang, Panneerdoss, MacLean, & Rao, 2011;Romrell, Bellvé, & Fawcett, 1976). Mice were sacrificed by CO 2 asphyxiation followed by cervical dislocation and both testes were collected in 2 ml DMEM: F12 medium (cat no. 11330-032, Gibco, Grand Island, NY) on ice. The tunica albuginea was removed and tubules were transferred to 10 ml enzyme solution I (0.5 mg/ml collagenase Type I (cat no.C0130-1G, Sigma), 200 µg/ml DNase I (cat no. DN25-100 mg, Sigma) in F12 medium), briefly vortexed and incubated 30 min at 33°C in a shaking water bath (100 oscillations/min). Every 10 min an additional manual shaking was done to help tissue dissociation. The dispersed seminiferous tubules were allowed to sediment and the supernatant was discarded. Tubules were washed with 10 ml fresh F12 medium and resuspended in fresh F12 medium. The mixture was layered on 40 ml of 5% Percoll (cat no. 17-0891-02, GE Healthcare Bio-sciences AB, Uppsala, Sweden) in HBSS (cat no. 55-022-PB, Mediatech, Inc. Manassas, VA) and allowed to settle for 20 min at room temperature. The top 45 ml containing Leydig cells was discarded and the remaining 5 ml were transferred to a new tube containing 10 ml of enzyme solution II (200 µg/ml DNase I, 1 mg/ml trypsin (cat no.T4799-5G, Sigma-Aldrich, St Louis, MO) in F12 medium). The mixture was incubated for 40 min at 33°C in a shaking water bath (100 oscillations/min) and every 10 min, manual shaking. After tissue dissociation, 3 ml charcoal-stripped FBS were added and cells were resuspended using a 10 ml pipette to dissociate clumps. The suspension was filtered sequentially through a 70 µm (cat no. 352350, Falcon Corning Incorporated, Corning, NY) then 40 µm (cat no 352340) nylon cell strainer and centrifugated at 500 g for 10 min at 4°C. The cell pellet was resuspended in 1 ml PBS (calcium and magnesium free) and counted. Cells were stored as a dry pellet at -80°C and used for protein or RNA extraction.

RNA isolation and RT-PCR
Testes or isolated germ cells were homogenized in TRIZOL reagent (cat no. 15596018, Invitrogen) following the manufacturer's protocol for tissue or cell pellet, respectively. The isolated total RNA was dissolved in RNase-free water, an aliquot (2 µl) was used to measure nucleic acid concentration and the remainder was immediately stored at -80°C. Total RNA (3 µg) was used to synthesize cDNA (75 µl final volume) with the Verso cDNA Synthesis Kit (cat no. AB-1453/A, Appliedbiosystems, Thermo scientific Baltics UAB, Vilnius, Lithuania) following the manufacturer's protocol. cDNA was tested for genomic DNA contamination using end-point PCR with Actb primers flanking an exon and intron sequence (Figure 9table supplement 1 for primer sequence).

Quantitative PCR (qRT-PCR)
cDNA obtained as described above was used to perform real time PCR. PowerUp™ SYBR™ Green Master Mix (cat no. A25742, Applied Biosystems, Thermo Scientific Baltics UAB, Vilnius, Lithuania) was mixed with each sample to a primer final concentration of 150 nm, following the manufacturer's protocol and run on a master cycler (ViiA 7, Thermo Fisher). PCR conditions were 95°C for 30 sec, followed by 40 cycles at 95°C for 15 sec, 60°C for 15 sec and 72°C for 20 sec. Unless otherwise stated, gene expression relative to Actb and Rps2 was calculated by the log2 ddCT method (Vandesompele et al., 2002).

Histological analysis
Hematoxylin and eosin (H&E) counter stained testis sections were analyzed by light microscopy (Zeiss Axiovert 200M, Göttingen, GERMANY) or scanned using a Perkin Elmer P250 high capacity slide scanner and processed using the proprietary software CaseViewer (3D Histech P250 high capacity slide scanner, Perkin Elmer, Waltham, MA).

Mild heat stress treatment
This protocol was adapted from (Y. S. Li et al., 2013;Rockett et al., 2001;Yin et al., 2013). Briefly, an adult male mouse was anaesthetized in an isoflurane chamber with a constant oxygen flow of 2 L/min and 3 % isoflurane for 1 min followed by 2.5 % isoflurane for 3 min. The mouse was quickly removed from the chamber and its nose was introduced into a nose cone with the same anaesthesia parameters for another 1 min. Testes were secured in the scrotum by manual massage and one third of the body (hind legs, tail and scrotum) was immersed in a 43ºC or 33°C (control) water bath, supported by a plastic tube for 25 min. During the experiment, the isoflurane flow was reduced every 10 min by 0.5 % to reach 1.5 % at the end of the treatment (2.5 % for 5 min after introduction into the water bath, then 2 % for 10 min and followed by 1.5 % for another 10 min). After the heat treatment, mice were dried on paper towel, allowed to recover in a chamber with oxygen flow at 2 L/min and 0% isoflurane for 5 to 10 min, then returned to a cage to recover from the effects of anaesthesia on a heating pad. Testes and epididymis were harvested 8 hr or 24 hr after treatment.

Statistical analysis
The bar graphs in all figures represent the mean ± SEM. Unpaired, two-tailed Student's t test or one way ANOVA was used to calculate p-value using Graph Pad Prism 7.0 (Graph Pad Software Inc., La Jolla, CA). Statistical significance was indicated by *p<0.05, **p<0.01, ***p<0.001 or ****p<0.0001. David Goldman The funders had no role in study design, data collection or interpretation, or the decision to submit the work for publication.

Author contributions
Ayodele Akintayo performed all experiments on transgenic mice, all microarray and validation experiments, data curation, bioinformatics analyses, and co-wrote the manuscript; Meng Liang