Sphingomyelin Synthase 1 Is Essential for Male Fertility in Mice

Sphingolipids and the derived gangliosides have critical functions in spermatogenesis, thus mutations in genes involved in sphingolipid biogenesis are often associated with male infertility. We have generated a transgenic mouse line carrying an insertion in the sphingomyelin synthase gene Sms1, the enzyme which generates sphingomyelin species in the Golgi apparatus. We describe the spermatogenesis defect of Sms1-/- mice, which is characterized by sloughing of spermatocytes and spermatids, causing progressive infertility of male homozygotes. Lipid profiling revealed a reduction in several long chain unsaturated phosphatidylcholins, lysophosphatidylcholins and sphingolipids in the testes of mutants. Multi-Spectral Optoacoustic Tomography indicated blood-testis barrier dysfunction. A supplementary diet of the essential omega-3 docosahexaenoic acid and eicosapentaenoic acid diminished germ cell sloughing from the seminiferous epithelium and restored spermatogenesis and fertility in 50% of previously infertile mutants. Our findings indicate that SMS1 has a wider than anticipated role in testis polyunsaturated fatty acid homeostasis and for male fertility.


Introduction
Sphingolipids are common components of mammalian cell membranes [1]. Sphingomyelin synthase 1 (SMS1/SGMS1) and the isoenzyme SMS2 both catalyze the de novo synthesis of sphingomyelin (SM) and diacylglycerol (DAG) from ceramide (Cer) and phosphatidylcholine (PC), while the paralog sphingomyelin synthase-related (SMSr) has no SM synthase activity [2][3][4]. Several tissue-specific isoforms of SMS1, produced from different promoters as well as through alternative splicing, give rise to two different proteins, of which only one has enzymatic function [5,6]. SMS1 is located in the cis-Golgi apparatus and responsible for the bulk production of SM [2,3,7,8]. However, due to vesicle-mediated trafficking between the Golgi apparatus and the plasma membrane, the highest concentration of SM and other complex sphingolipids (SL) is found in the outer leaflet of the plasma membrane [9]. Here, the diverse SL species have various functions in structure, adhesion and signaling. At the plasma membrane, these functions are mediated by SMS2, which unlike SMS1 plays a major role in controlling Cer levels at the plasma membrane [1,10,11]. However, sphingolipids have similar intracellular functions in all cellular compartments, including the nucleus [12] and an imbalance in lipid composition can lead to disease.
In our study, we analysed an insertional mutation in the anabolic enzyme SMS1 and its consequences for male fertility. Indeed, infertility was observed in 66% of Sms1 homozygous mutant adult males. Infertility was caused by sloughing of meiotic and spermatid stages and their subsequent accumulation in the epididymides. Lipid profiling of testes revealed a decrease of LC-PUFAs in Sms1 -/animals. Male infertility was ameliorated by supplementary diet of DHA and eicosapentaenic acid (EPA), indicating an essential role for SMS1 in the supply with n-3 unsaturated FA substrates.

Sms1 -/animals
All animal experiments were carried out in accordance with the European Communities Council Directive (20/0/63/EU). Animal experiments, which included complementation of food by PUFA and subsequent imaging using fluorescent dyes, were performed according to an allowance by the District Government of Bavaria (number TVA 189-12) to TF from 06/18/2013 after previous reviewing and ethical commission of the Bavarian Government on 04/10/2013 according to §15 paragraph 1 and §8 paragraph 1 of the German Animal Protection law. All surgery was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize suffering. Proper animal caretaking was provided by the Helmholtz Center animal facility (AVM) according to standard rules. Animals were group housed in ventilated standard cages on a 12-hour light-dark cycle and had ad libitum access to food and water. The mutant Sms1 allele (MGI: Sms1 Gt(E201D11)Wrst , in the following referred to as Sms1 -/-) was generated by gene trap technology as described previously [22]. Males were backcrossed six times to C57Bl/ 6J mice (Charles River Laboratories) and SMS1 +/+ and Sms1 mutant littermates were used in this study. Genotyping was done by triplet-PCR amplification using the following primers: Tmem23_fw (5'-gccggaaattaaaaagaacaatga-3'), Tmem23_rv (5'-gcaatgcagggtgcttcttcc-3') and Spli_rv2 (5'-gccaaacctacaggtggggtcttt-3' . A band of 577bp represents the mutant fragment and 999bp represents the wild-type fragment. All PCR fragments were sequence-verified initially. PCR cycling settings were 95°C, 5'-(94°C, 1'-60°C, 1'-72°C, 1') x 35-72°C, 7' for all reactions.
Organ preparation for Protein-and RNA-based analysis Animals were sacrificed using CO 2 . Tail clips were taken for regenotyping. Organs were dissected, kept on dry ice and subsequently stored at -80°C until further investigations.

DNA isolation
Genomic DNA was isolated from tail tips by incubation with 500μl of Lysis-Buffer (10mM KCl, 20mM Tris, 0.1% Non.N40, 0.1% Tween) and 20μg of Proteinase-K per tail tip at 60°C for at least 3h. For Proteinase-K inactivation samples were heated to 95°C for 10' prior to centrifugation and separation of the supernatant.

Assessment of protein concentrations
Concentration of RIPA-soluble protein was measured by the bicinchoninic acid (BCA) Protein Assay Reagent (Thermo scientific) according to the manufacturer's instructions. Briefly, 25μl of BSA standards, ranging from 25μg/μl to 2mg/μl, and sample dilutions were applied to a 96 well flat bottom plate and incubated for 30' at 37°C with 200μl of BCA working reagent (CuSO 4 /bicinchoninic acid (1:50) per well. Readout of the colorimetric assay was performed at 562nm.

Western Blot
40ng of RIPA-soluble protein per sample was mixed with 5μl NuPAGE1 LDS Sample Buffer (4x; Invitrogen) and 0.2μl β-Mercaptoethanol and was heated to 95°C for 3' . For SDS-PAGE protein samples and 10μl of SeeBlue1 Plus2 Pre-Stained Standard (Invitrogen) were applied to a NuPAGE1 Novex Bis-Tris 4-12% Gel (Invitrogen) for 1h at 200V, RT. Proteins were transferred to PVDF membranes (Immobilon-P, Immobilon1 Transfer Membranes, Millipore) for 1.5h at 30V, RT. The membrane was blocked with 5% skim milk (Sigma, in TBS-T (0.05%) for 1h, RT and subsequently incubated with primary antibodies at 4°C, overnight. After washing with 0.05% TBS-T, the secondary antibody was applied to the membrane in a 1:10000 dilution for 1h, RT. After 3 x 10' in 0.05% TBS-T, the blot was developed using Hyperfilm ECL and the ECL detection reagents (Amersham).
Quantification of the protein signal, was performed using ImageJ and statistical analysis of signal intensity was done using the student's T-test. Statistical significance was considered at Ã p 0.05, ÃÃ p 0.01 and ÃÃÃ p 0.001.

RNA isolation
1ml Trizol was added per sample. Samples were homogenized and incubated for 5' at RT prior to centrifugation at 15000g for 10' at 4°C. The supernatant was mixed with 200μl of chloroform, incubated for 8' at RT and centrifugated at 15000g for 15' at 4°C. The supernatant was mixed with 530μl of isopropanol and incubated at -20°C, overnight. Remaining RNA-isolation steps were carried out using the RNeasy Kit (Qiagen) according to the manufacturer's instructions, including DNAse treatment. RNA was stored at -80°C until further processing.

Histological analysis and organ pathology
Histological analysis and organ pathology was performed within the German Mouse Clinic as described [23]. A total of 46 mice of the Sms1 mouse line at different ages were analysed for morphological changes.
Tails were stored at -70°C for genetic reconfirmation. Body weight, body length and weight of internal organs such as heart, spleen and liver were determined. All organs were fixed in 4% buffered formalin and embedded in paraffin, 4μm sections were stained with hematoxylin and eosin (H&E) for histological examination. Periodic acid shiff (PAS) staining was applied for visualization of glycolipids. Images of tissue sections were acquired with an automated slidescanner (NanoZoomer-NDP-Hamamatsu, Germany). All slides were independently reviewed and interpreted by 2 pathologists, experienced in mouse pathology. Statistical analysis was done using the Student's T-test, statistical significance was considered at p value (<0,05).
Tissues were fixed in 2.5% electron microscopy grade glutaraldehyde in 0.1 M sodium cacodylate buffer pH 7.4 (Science Services, Munich, Germany), postfixed in 2% aqueous osmium tetraoxide (Dalton, 1955), dehydrated in gradual ethanol (30-100%) and propylene oxide, embedded in Epon (Merck, Darmstadt, Germany) and cured for 24 hours at 60°C. Semithin sections were cut and stained with toluidine blue. Ultrathin sections of 50 nm were collected onto 200 mesh copper grids, stained with uranyl acetate and lead citrate before examination by transmission electron microscopy (Zeiss Libra 120 Plus, Carl Zeiss NTS GmbH, Oberkochen, Germany). Pictures were acquired using a Slow Scan CCD-camera and iTEM software (Olympus Soft Imaging Solutions, Münster, Germany).

In-vivo Optoacoustic Imaging of BTB Permeability
Animals were placed under isoflurane anesthesia and were imaged using a multispectral optoacoustic tomography (MSOT) system described previously [24]. In brief, mice were shaved using depilation cream around the testes region for imaging, and then placed in a prone position in the MSOT imaging system. Ultrasound coupling gel and water were used to provide acoustically-matched coupling between the tissue and acoustic detector array. The animal was positioned in the detector array to provide tomographic slices in the region of the testes. Tissue was illuminated over a range of wavelengths (700, 730, 760, 800, and 860 nm) and the resulting optoacoustic responses were used to reconstruct cross-sectional images through the testes using a back-projection algorithm [25]. During imaging, 2mg/kg bodyweight Indocyanine Green (ICG) was injected via the tail vein and serial multispectral images were acquired following injection. Finally, multispectral unmixing was done using linear regression [26,27] to derive the ICG absorbance signal. Region of interest (ROI) analysis of the unmixed signal provided the quantitative time course of ICG distribution and clearance within the testes with a time resolution of~5 seconds. Quantification of the ICG absorbance signal was determined by taking the mean pixel intensity value of the ICG signal within the region of the testes.

Lipid profiling
a. Testes preparation. Testes were dissected and homogenized in H 2 O on ice using a Pot-terS (Braun, Melsungen, Germany) at 1500 revolutions per minute and 50 strokes. Homogenates were adjusted to Protein amount of 10mg/ml with H 2 O, snapfrozen in liquid nitrogen and stored at -80°C for further experiments.
b. Protein amount determination. Protein determination was performed according to Smith et al. [28]. Briefly, we used a standard curve of bovine serum albumin with a concentration range of 0.1-1.2 μg/μl. 20μl of the standard solutions was pipetted onto a 96-well plate (Nunc, Langenselbold, Germany) and further 1-2μl of each sample was pipetted in triplicates onto the 96-well plate. 200μl reagent buffer, consisting of CuSO 4 /bicinchoninic acid (1:39; v/v) was added to each well using a multichannel pipette (Eppendorf, Germany). The plate was incubated for 15' at 37°C followed by incubation at RT for 15' while shaking (IKA, Staufen, Germany) at 300 revolutions per minute. Absorbance was measured using a MultiscanEX (Thermo Fisher Scientific, Schwerte, Germany) at a wavelength of 550nm.
c. Mass spectrometry. All measurements were performed using a 4000 quadrupole linearion trap (QTrap) mass spectrometer equipped with a Turbo Spray ion source (AB SCIEX, Darmstadt, Germany): The QTrap was connected to a Agilent Cap HPLC comprising of 1200 Capillary Pump, 1200 Micro Wellplate Sampler and 1200 Thermostat Module (Agilent, Böblingen, Germany). Data analysis was carried out using Analyst 1.5 software (AB SCIEX, Darmstadt, Germany). For determination of phosphatidylcholine, phospatidylcholine-plasmalogen, lyso-phosphatidylcholine, sphingomyelin and ceramide levels 20μl sample was injected into sample loop with the following running solvent gradient (0.0-2.4min, 30μl; 2.4-3.0min, 200μl; 3.0min, 30μl). Mass spectrometry settings and composition of running solvents used is specified below.
g. Measurement of ceramide levels. Ceramide measurement was performed according to Gu et al. [30]. In brief, 50μl of each samples was mixed with 500μl CH 3 OH/H 2 O/12M HCl (95/ 5/0.5; v/v/v) and sonificated for 1' at 4°C. Mixture was shaken for 10' at RT at maximum speed (Multireax, Heidolph Instruments, Schwabach, Germany). After adding 300μl H 2 O and 500μl CHCl 3 samples were shaken for another 10' at RT. Samples were centrifuges at 5000rpm for 10' and the lower phase was transferred to a new tube. 700μl CHCl 3 was added to the remaining two phases and mixture was shaken for 10' at RT at maximum speed. After another centrifugation step as described above, the two lipid containing phases were combined and dried in a vacuum concentrator. Finally, lipids were dissolved in 300μl CHCl 3 /CH 3 OH (95/5; v/v) and transferred into a 96-deep well plate (Nunc, Langenselbold, Germany). Prior to injection the samples were diluted with 600μl 5mM ammonium acetate dissolved in CH 3 OH and shaken at 300rpm (IKA, Staufen, Germany) for 10' . The running solvent for ceramide analyses was composed of CH 3 h. Statistical analysis. All experiments were performed using at least five different animals per genotype and analysed in triplicates. Statistical significance was set at Ã p 0.05, ÃÃ p 0.01 and ÃÃÃ p 0.001; n.s. = not significant and statistical analysis was determined by two-tailed Students t-test.

TUNEL staining and quantification of apoptotic events
TUNEL staining was performed with the ApopTag1 Fluorescein or Peroxidase In Situ Apoptosis Detection Kit (Millipore, California) according to the manufacturer's instructions. Briefly, formalin fixed, paraffin embedded testes sections were deparaffinated, hydrated and treated with Proteinase K (20μg/ml). For fluorescent or peroxidase-mediated labeling, single/ double strand brakes were prolonged by addition of dNTPs (labeled and unlabeled) via terminal deoxynucleotidyl transferase (TdT) for 1h at 37°C. The reaction was terminated and fluorescein-conjugated anti-digoxygenin antibody or anti-digoxigenin conjugate was applied for 30' , RT, respectively. Peroxidase-mediated staining required additional quenching of endogenous peroxidase after Proteinase K treatment and incubation with peroxidase substrate after applying of the conjugate. Sections were counterstained using DAPI (0.5μg/ml) or Haematoxylin and analysed by fluorescent or light microscopy, respectively.
TUNEL-positive cells in an average of 400 tubuli per animal (wild-type n = 3, Sms1 -/n = 4) were counted for quantification of apoptotic events.

PUFA (DHA/EPA) diet
For the dietary assessment 20 week old male Sms1 mice were used for a fertility-test. For this, Sms1 wildtype (n = 2) and Sms1 +/-(n = 4) males, serving as controls, and Sms1 -/-(n = 6) males were each provided with one eleven week old C57Bl/6J female. After 3 weeks these females were separated from the males and kept for additional 3 weeks. The litters obtained were recorded. Subsequently to separation of the first batch of females, Sms1 males of all genotypes were provided with a second female each. In parallel, a mix of n-3 fatty acid ethylesters (Omega-3 90 EE) including 38% DHA and 46% EPA (kindly provided by KD Pharma (Bexbach)) was manually fed to the male Sms1 mutants in an oral dose of 9mg (total n-3 fatty acids)/day/animal for 13 weeks.

Results
The Sms1 gene trap mutation The Sms1 gene comprises 14 exons, of which the upstream exons 1-6a are non-coding. The Sms1 mutation described here is an insertion of a conditional gene trap vector in intron 3, which locates upstream of the first open reading frame in exon 7 (Fig 1A; MGI: Sgms1 Gt(E201D11)Wrst ). Sms1 +/animals were interbred to homozygosity which was determined by triplet-PCR (Fig 1B). Sms1 -/animals were obtained at Mendelian ratios, viable with no obvious phenotypes and had life expectancies of at least 24 months. When animals were further examined for subtle morphological alterations, we determined a slightly higher body weight, and a reduction in liver weights of male Sms1 -/-, while body, heart and spleen weights were not significantly affected (S2 Table). SMS1 protein expression in testes samples was analysed using a polyclonal antibody (Santa Cruz). SMS1 protein was reduced to 5.5 ± 1.2% (p = 0.0005) in Sms1 -/animals, as compared to protein levels in the wild-type (Fig 1F and 1G).

Alternative splicing and alternative promoters in Sms1
Sms1 is widely expressed in a number of tissues and organs [5]. Various transcripts have been described which are derived from either alternative splicing (Sms1-001) or alternative promoter activity (Sms1-002, Sms1-004 and Sms1-014; Fig 1C, [31]). Two alternative promoters in mouse Sms1 are predicted downstream of the insertion based on exclusive use of the upstream non-coding exons 6 or 6a (Sms1-004 and Sms1-014; Fig 1C). In order to determine and quantify, which Sms1 transcripts were expressed and which variants were disrupted by the intronic gene-trap insertion, we performed Northern blots using a probe covering 801bp of the first coding exon 7, which was at least partially present in all four protein coding transcripts of the gene (Fig 1C). The alternative transcripts Sms1-004 (2.5kb) and Sms1-001 (3.7kb) were the major residual transcripts in testes, showing a strong mean reduction to 3.1 ± 0.7% (p = 0.0001) as compared to wild-type levels, reflecting the detected amount of SMS1 protein (Fig 1D and 1E). The transcript Sms1-002 (4.0kb), which initiates upstream of the insertion site, was expressed in marginal amounts in wildtype but was not detected in testes of Sms1 -/males. The short Sms1-014 transcript was neither detected in wild-type nor Sms1 -/testes samples (Fig 1C and 1D). Lipid profiling of testis tissue Next, to determine SMS1 enzyme activity in vivo, we established lipid profiles from testes of Sms1 +/+ vs. Sms1 -/littermates. Substrates of SMS1, Cer and PCs such as diacylphosphatidylcholines (PCaa) and phosphatidylcholine-plasmogenes (PCae) as well as SM and lyso-phosphatidylcholines (lysoPC) of different chain length and saturation states were analysed. In comparison to wild-type levels, SMs were found reduced to 78.9 ± 2.0% (p<0.0001) in Sms1 -/animals compared to wild-types (Fig 2A). Overall Cer levels appeared almost unaffected in testis samples (Fig 2A), which was not surprising considering the role of SMS2 and the multiple effects of Cer as cytostatic, inflammatory and pro-apoptotic factor and the function of Cer as a crucial branching point for several other lipid pathways [32]. A detailed analyses of Cer species however revealed shifts in their relative amounts in testis. Namely, there was a significant increase in the C18:1/18:1 species relative to most other Cer species, with the exception of C18:1/26:1 (S3 Table).
Analysis of PCaa revealed a reduction to 91.9 ± 2.4% (p = 0.0089) in Sms1 -/-. Detailed analysis pointed out, that this unexpected reduction was due to low levels of unsaturated PCaa (91.3 ± 2.4%; p = 0.0056), while the amount of saturated PCaa was comparable to wild-type levels (Fig 2C). Neither PCae levels of total saturated, nor of total unsaturated species were altered in Sms1 -/samples, compared to wild-type levels. (Fig 2C).
A detailed analysis of the most abundant PCaa and PCae species of Sms1 -/against wildtype levels revealed a significant reduction in mono-(MUFA) and PUFA containing species (Table 1A).
Among the FAs with a highly significant reduction, we found the most striking reduction for the PCaa species C38:5 and C38:6, indicative of a loss of arachidonic acid (AA; C20:4), EPA (C20:5), docosapentaenoic acid (DPA; C22:5) or DHA (C22:6) all of which belong either to n-3 or n-6 FAs (Table 1A). While total PCae were unaltered, again PCae C38:6 was among the strongly reduced lipids (Table 1B).The total level of lysoPCs was significantly reduced to 73.4 ± 0.2% (p = 0.001) in testes of Sms1 -/males in comparison to the level in wild-type samples (Fig 2B). This reduction was present in all major lysoPC species (Table 1C). These results suggests that SMS1 has an unexpected role in testis PC, including n-3/6 PUFA, homeostasis.

Sms1 -/males show display altered luminal contents of testes and epididymides
In epididymides of Sms1 -/males, mature spermatozoa were rarely present and instead, round spermatids, lacking acrosome structures and tails, were accumulated (Fig 3A and 3B). Sms1 expression indicates two expression domains in seminiferous tubules: 1. in Sertoli cells forming the blood-testis-barrier (BTB) and 2. in elongating spermatid cells (Fig 3C and 3D). In epididymis, Sms1 expression was detected in principal cells, which are forming the blood-epididymisbarrier (BEB; Fig 3E and 3F). In Sms1 -/seminiferous tubules of different stages, large vacuoles were frequently encountered (Fig 3D; see also Fig 4B and 4F). The Sms1 expression in testis and epididymis was also reflected by the β-gal reporter (see S2A and S2B Fig). In contrast, SMS2 protein expression was found expressed along with SMS1 in elongating spermatids and spermatozoa, where both co-localized to the developing acrosome [33]. This expression domain was strongly reduced in Sms1 -/males (Fig 3G and 3H). Compared to the wild-type, round and elongated spermatids were strongly reduced in number and only occasional spermatozoa were observed in testis tubules of Sms1 -/males (Fig 3G and 3H).
Furthermore, apoptotic cells were noted in seminiferous tubules of Sms1 -/testes (Fig 3I and  3J) and wild type (not shown). Furthermore, often there was accumulation of apoptotic material in the lumen of the tubules (Fig 3J). Quantification of TUNEL-positive cells per tubule revealed a 1.9-fold higher rate of apoptosis in Sms1 -/testes (1.06 ± 0.06 TUNEL-pos. cells/ tubule) compared to wild-types (0.56 ± 0.04 TUNEL-pos. cells/tuble; Fig 3K). These observations were made using animals of 20 weeks of age and in comparison to Sms1 +/+ males. Further investigation of the number of apoptotic cells in TUNEL-positive tubules of stage XII and I-XI revealed, however, that the number of dying cells/tubule was similar in mutant and wild-type testes. In average there were 3.43 apoptotic cells per stage XII wild-type tubule (+-0.40 SE) versus 3.71 (+-0.47) cells/tubule in the mutant. Cell death-positive stage I-XI tubuli (in the absence Table 1. Lipid profile of Sms1 mutant testes PCaa, PCae and lysoPC species, compared to Sms1 wild-type levels. Lipid levels with the highest intensities are shown for diacylphosphatidylcholines (PCaa, A) plasmalogens (PCaes, B) and lyos-phosphatidylcholines (lysoPCs, C). Additionally, predicted major fatty acids are listed. Data is presented as mean percentage ± SEM, with wild-type levels set to 100% (Sms1 wild-type, n = 15; Sms1 mutant, n = 17); *p 0.05, **p 0.01, ***p 0.001. of haploid cells staging could only be crudely done) displayed in average 3.68 (+-0.29) and 4.0 (+-0.28) apoptotic cells per wild-type and mutant tubule, respectively. Our data show that there is an overall increase of cell death in the mutant, which likely relates to the disruption of the BTB and the spermiogenic epithelium. Similar rates of apoptosis in stage XII tubules furthermore indicate that spindle defects or chromosomal missegregation are not increased in the mutant.

Sloughing involves spermiogenic cells
From our observations, we hypothesized sloughing of mainly elongating spermatid stages into testis lumen of the mutants and their accumulation in the epididymides. In order to determine, the exact developmental stage at which the germ cells of Sms1 -/males lose their contact to Sertoli cells, we analysed the expression of characteristic histone modifications. Histone modifications during meiosis include trimethylation of histone H3 at lysine residue 4 (H3K4me3) prior to meiotic foci formation, a modification that is associated with active chromatin [34][35][36]. We determined the expression of H3K4me3 and found this histone modification to be expressed at normal intensity from the spermatogonia stage to the end of meiosis II in both, wild-type ( Fig  4A) and Sms1 -/testes ( Fig 4B). In both genotypes, H3K4me3 was weakly expressed in Sertoli cells but detectable in pachytene spermatocytes (Fig 4A and 4B). Histone H3 lysine 9 becomes first trimethylated (H3K9me3) in spermatogonia, a modification that is associated with inactive chromatin. While H3K9me3 is not found in the XY-body of pachytene spermatocytes, it is detectable in the heterochromatin clusters of spermatogonia until the round spermatid stages [37,38]. This modification was seen in wild-type as well as in testes of Sms1 -/animals (Fig 4C and 4D). Wild-type epididymal spermatozoa have already condensed their chromatin and thus, wild-type epididymal spermatozoa were negative for both H3K9 and H3K4 trimethylation (Fig 4G and 4I). In the epididymides of Sms1 -/animals, trimethylation of both H3K9 and H3K4 was determined in spermatocytes and spermatids, indicating sloughing of histone-retaining spermatogenic cells up to spermatid stages (Fig 4H  and 4J).

The blood-testis barrier appears compromised in Sms1 mutants
We hypothesized that the specific reduction of PUFAs in Sms1 -/testes may cause sloughing of spermatids due to a disturbance of the integrity of the BTB, as proposed by Stoffel et al. [43]. The BTB, built by special tight-, gap-and adherens-junction proteins between Sertoli cells, separates the adluminal stages of spermatogenesis from the blood supply and the cellular immune system, thereby creating and controlling a specific microenvironment required for sperm maturation (for a recent review see: [44]. In order to test the hypothesis of a compromised BTB, we quantified the expression of representative junction markers in testical and epididymal protein samples by western blotting (S1G Fig). Interestingly β-catenin (283.3 ± 26.0%, p = 0.0054) and connexin 43 (112.0 ± 1.7%, p = 0.0084) as subunits of adherens-or gap-junctions, respectively, were significantly elevated in testes of Sms1 -/males (S1G and S1H Fig), while their protein levels in the epididymides were unchanged. Occludin levels in the mutants showed a tendency to be elevated in testes (149.1 ± 18.7%, p = 0.0776) but were unaltered in the epididymides compared to wild-type levels (Fig 1G and 1H). Immunostaining, using antiβ-catenin and anti-connexin 43 antibodies did not reveal any dislocation of junctional proteins (S1A, S1B, S1D, and S1E Fig), while occludin was more abundantly expressed in Sms1 -/-(S1C and S1F Fig). Further studies of BTB integrity were done by transmission electron microscopy (TEM). It appeared that tight junctions between Sertoli cells were established equally in wildtype and Sms1 -/testes (Fig 5E and 5F). In order to elucidate the possibility of a functionally compromised BTB in vivo, Multi-Spectral Optoacoustic Tomography (MSOT) imaging of Indocyanine Green (ICG) distribution in the testes was performed. Following ICG injection, wild-type testes show little ICG peak signal within testes, while, a qualitatively higher signal was detected in Sms1 -/testes (Fig 5A and 5B). Representative time courses of ICG distribution in the Sms1 -/and wild-type mice are shown in Fig 5C, indicating higher peak ICG levels in the Sms1 -/animals (19.21e -5 ± 2.88e -5 optical absorbance units (A.U.) when compared to the wild-type (8.43e -5 ± 1.25e -5 A.U., p = 0.040, Fig 5D). The clearance of ICG from the testes was also delayed in the Sms1 -/mice when compared to wild-type controls. However, this trend was not significant. The higher ICG signal in the Sms1 -/animals indicated increased permeability of the BTB, since such loss of integrity would result in ICG leaking from vessels to testicular tissue.
Sms1 -/animals show subfertility or infertility, which is age-related In agreement with a previous report [45] we noted that Sms1 -/males had reduced reproductive capabilities. To systematically analyze the fertility of Sms1 males at two different ages, twelve male animals covering all genotypes were analysed (SMS1 +/+ n = 2, Sms1 +/n = 4, Sms1 -/n = 6). We first determined the fertility status at the age of eight weeks. Sms1 males were mated to C57Bl/6J females (age: 11 weeks) for a three week period. All wild-type and Sms1 +/animals were able to produce offspring, four of the six Sms1 -/animals were fertile (66%), while two failed to produce offspring with eight to eleven weeks of age (33%). At the age of 20 weeks we observed the described testis phenotype in six out of six mutant males (compare Fig 3A and 3B), which also impacted the fertility state of Sms1 males. While all wild-type and Sms1 +/males tested produced offspring, only two out of the six male Sms1 -/appeared fertile (33%), whereas four failed to produce any offspring (66%). These results pointed to an early age-related decrease of total fertility in Sms1 -/males compared to Sms1 +/and wild-type animals, starting already at the age of 8 weeks. We also analysed the litter size. Infertility was assumed in this experiment when a male failed to produce offspring during the mating time of 3 weeks (Fig 6A). Fertile Sms1 -/males produced 5.50 (IQR: 3.50) pups per litter, while Sms1 +/and wild-type males produced 8.50 (IQR: 3.25) and 7.50 (IQR: 0.50) pups per litter, respectively (Fig 6B). At the age of 20-23 weeks the fertile Sms1 -/males were considered subfertile as compared to wild-type and Sms1 +/males.
Ameliorated spermatogenesis phenotype in Sms1 -/males treated with a PUFA diet The observed reduction in AA, EPA, DPA or DHA (Tab1A, B) and the profound role of PUFAs in spermatogenesis suggested an involvement for the observed male sub-/infertility phenotype ( Fig 6B). As AA treatment was expected to be little suitable as a potential therapeutic intervention we supplemented males with a diet of 38% DHA and 46% EPA, in order to investigate whether the subfertility/infertility could be reversed.
The dietary supplementation experiment was initiated following fertility analysis. Once, the fertility status of the males was determined, Sms1 males of all genotypes were remated to eleven week old C57Bl/6J females and were manually fed an oral dose of the n-3 FA containing oil on a daily basis for 13 weeks as described in Materials and Methods (Fig 6A). During the testing period Sms1 +/+ and subfertile Sms1 -/males produced an average of 2.50 ± 0.50 litters, each. Sms1 +/males produced 3.00 ± 0.00 litters. Of the confirmed infertile Sms1 -/males 50% produced offspring with feeding, while the other 50% failed to generate offspring. Sms1 -/animals with rescued fertility produced one litter each during the testing period of 13 weeks. We noted, that for Sms1 +/+ and Sms1 +/males the number of offspring was reduced with the feeding of the n-3 FA diet to an average of 5.33 (IQR: 0.33) pups per litter for Sms1 +/+ and to 6.33 (IQR: 2.50) pups per litter for Sms1 +/males (Fig 6B). This was likely due to the stress that was associated with the daily manual feeding procedure. Previously subfertile males also produced less offspring after feeding (4.84 (IQR: 1.17) pups per litter). Previously infertile males, produced  (Fig 6B). We determined the average latencies between litters to check for prolonged reproduction times. The results for Sms +/+ (24.00 (IQR: 1.0) days) and Sms1 +/males (23.00 (IQR: 2.5) days) were similar (Fig 6C). Subfertile Sms1 -/and infertile Sms1 -/males required more time for reproduction (subfertile Sms1 -/-: 31.50 (IQR: 9.5) days, infertile Sms1 -/-: 84.00 (IQR: 48.0) days; Fig 6C). Natural occurrence of unproductive matings in the strain C57BL6/6J is 12.6% (http://phenome.jax.org/db/qp?rtn=views/ measplot&brieflook=14934&ustrainids=7&brieflooktool=&projhint=Jax3), which insufficient to explain the amount of infertility observed in these matings.
All offspring was heterozygous, confirming homozygosity of the males, which was also confirmed by re-genotyping via triplex-PCR. Matings were terminated and males sacrificed after one additional week of PUFA enriched diet. As shown in Fig 6D, spermatogenesis had progressed almost normally in subfertile Sms1 -/males after 14 weeks of PUFA diet. Only few round spermatids were visible in mutant epididymides while the majority of cells had matured into spermatozoa (Fig 6D). In contrast infertile Sms1 -/males, which could not be rescued still showed impaired spermatogenesis and accumulation of immature germ cells in the epididymal cauda ( Fig 6D). In previously infertile Sms1 -/animals, which produced offspring after PUFA diet, the histology of seminiferous tubules ranged from being 'impaired' up to 'fully restored' . Their epididymal tubules were similar to those of subfertile Sms1 -/males with a few round, but a majority of elongated germ cells (Fig 6D).

Discussion
In sperm cell membranes, phospholipids represent the most prominent lipid fraction with a high content of PUFAs, e.g. VLC-PUFA-SM, which have a role in sperm capacitation, but are also found in Sertoli cells [46][47][48]. SM is generated by two different SMSs-SMS1 and SMS2. Redundant and non-redundant functions of both SMSs in different organs were illustrated by knockout animals of Sms1 [17,49,50], and Sms2 [51,52]. While no spermatogenesis or fertility defects and normal reproduction were reported from Sms2 knockout animals [51,53,54] impaired reproduction has been noted from male Sms1 knockout animals [45]. Unlike Dong et al., [45] where complete infertility of male Sms1 knockout animals was seen, we observed residual age-dependent fertility in a number of mice, which might be explained by different genetic background or potential differences in diet and age of the animals analysed. In the present study, we show, that disruption of Sms1 expression leads to distinct changes in the testicular lipid profile, especially concerning unsaturated FA containing species and sloughing of pachytene to late spermatid stages. The testicular SM levels of Sms1 -/males were 20% reduced as compared to wild-type. This reduction is in the expected range [17,45] with the remaining 80% likely resulting from SMS2 activity and dietary uptake. One explanation for the different impact of SMS1 and SMS2 on male fertility is the subcellular localization and functional differences of both enzymes. SMS1 activity is restricted to the Golgi apparatus, while SMS2 mainly locates to the plasma membrane where it is important for Cer homeostasis [10,55] SMS1 has a role in the maintenance of intracellular SM homeostasis, normally generating the bulk SM pool, which also supplies the plasma membrane [7]. The imbalance in the lipid metabolism, caused by disruption of Sms1 therefore affects intracellular, but also cell surface lipid composition, altering protein delivery, enzymatic activity and lipid based cell signaling.

The Sms1 mutation altered the lipid profile of mutant testis
Although Cer and PC are the direct substrates of SMS1, their overall levels did not rise in Sms1 -/testes, rather Sms1 -/maintained steady levels of otherwise potentially cytotoxic Cer. This is in agreement with Cer homeostasis to be maintained mainly by enzymes other than SMS1, which was also seen for Sms1 knockout mice, which had only marginal effects on Cer levels, while a Sms2 depletion led to Cer elevation in plasma [17,52,54]. Parallel to unaltered total Cer, unsaturated PCaa and PCae species were decreased. In the PCaa and less pronounced in the PCae profile, we found a decrease of specific PUFA (AA, EPA, DPA and DHA) containing lipids and an overall reduction in MUFA/SFA and PUFA/SFA ratios, which likely reduced the ability of Sertoli cells to support accumulation of LC-PUFA and VLC-PUFA in spermatozoa [56]. The production of these PUFA-SMs appears to be primarily derived from the activity of SMS1. The observed reduction in several unsaturated testicular PCaes, one of the major phospholipids of sperm membranes, together with an alteration in SM levels, leads to changes in membrane properties and thus affects the special composition needed for interaction and communication between Sertoli cells and spermatocytes necessary for the differentiation into mature sperm.
BTB functionality was compromised in Sms1 -/males As described, immature spermatogenic cell types ranging from pachytene to late spermatid stages were sloughed and accumulated in the epididymal lumen of Sms1 -/males. Several lines of evidence support a contribution of BTB malfunctioning to the phenotype. Usually, the BTB is crossed by spermatocytes between preleptotene and leptotene stage, while meiotic prophase stage cells are held back, for as long as they are primed for the passage and provided with their special set of proteins and lipids. The BTB is made up of several distinct junction types, which set up a dynamic network between Sertoli/Sertoli and Sertoli/germ cells, regulating the passage of germ cells from the basal to the adluminal compartment [44,57]. Indications for a BTB deficiency in Sms1 -/males arose from altered junction protein levels, such as elevated β-catenin and connexin-43, which represent subunits of adherens-and gap-junctions, respectively [44,56]. Connexins are continuously synthesized and degraded representing a highly flexible system for intercellular contact and communication, known to influence expression levels of other junctional proteins [44,58,59]. An increase in junction proteins may result from disrupted Trans-Golgi-network (TGN)-mediated protein trafficking and protein secretion, which was shown to be dependent on SMS activity [60]. Although the subcellular localization of junctional proteins was found mostly unaltered in Sms1 -/males, their elevated expression indicates an impaired Sertoli cell/germ cell interaction. The protein levels of occludin, one of the major component of tight-junctions, tended to be elevated as well in Sms1 -/males. Our analysis of the structure of tight-junctions between Sertoli cells by TEM did not reveal obvious morphological differences in the Sms1 -/in comparison to wild-type testes, however, the BTB is not established through tight-junctions only. Adherens-junctions, gap-junctions, channel proteins and transporters also play important roles in intercellular contacts and communication [61]. After injection of ICG into tail veins of living animals followed by life imaging using MSOT we determined a significantly higher fluorescent signal from the testes lumen of Sms1 -/animals, clearly indicating BTB dysfunction and leakiness.
PUFA diet ameliorated the sloughing phenotype of infertile Sms1 -/- In breeding experiments the majority of Sms1 -/males were either infertile or subfertile. We noted that the grade of infertility was age-related, as indicated by a progressive loss in number of offspring and a diminished ability to reproduce.
A crucial role for PUFAs for male fertility has been proposed by a number of studies in different organisms [62][63][64][65]. Epidemiological studies suggested a negative impact of a diet rich in saturated fatty acids ("Western diet") on the sperm quality in young men [66]. Evidence for the importance of PUFAs for the integrity of the BTB has been recently suggested, after analyzing Δ 6 -desaturase (Fads2) mutant mice [43,67]. The reduction in PUFA levels, phenotypic similarities of Sms1 and Fads2 mutants in respect to spermatogenesis, including a significant decrease in PUFA-PCaa and possible leakiness of the BTB prompted us to supplement the diet of Sms1 -/with n-3 PUFAs (DHA/EPA). PUFA supplementation reconstituted the fertility status in 50% of these animals. By taking into account that pregnancies last approximately three weeks and that the first litter was obtained 36 days after mating, we calculated in retrospect that treatment for a minimum of approximately two weeks or more with a PUFA supplementation was needed to restore fertility in Sms1 -/males. Given that a full cycle of spermatogenesis takes 34 days in mice and is made up of four subcycles (8.6 days each), our data support the hypothesis that PUFA supplementation affects pachytene spermatocytes and spermatid stages [68]. Rabionet et al. [69], suggested that SM (C16) and glycosphingolipid production in mouse testis starts around prophase and that after passing the BTB, with the onset of CerS3 expression, the cells switch to the production of VLC-PUFAs, which would be in good agreement with our findings, although we can not exclude other mechanisms to be involved. More recently, Rabionet and coworkers demonstrated through a germ cell-specific knockout of CerS3 that the enzyme is crucial for the formation of intercellular bridges (ICB) between syncytial spermatids. ICB stability is based on ultra long polyunsaturated SMs. Similarly to Sms1 -/animals, the germ cell-specific CerS3 knockout resulted in elevated apoptosis of spermatogenic cells [13]. The sloughing phenotype of Sms1 -/was significantly ameliorated after dietary supplementation. The histological analysis revealed high numbers of mature spermatozoa in epididymides and only few round spermatids of PUFA-fed Sms1 -/males, which produced offspring. This suggests further that a longer treatment duration might help to restore male fertility further.
In summary, the exact cause for sloughing of germ cells in the Sms1 -/males is not entirely clear, but our findings indicate the following mechanisms to be involved. 1) A dysfunctional BTB, which is initiated in Sms1 -/males, as was demonstrated by the presence of tight junctions by TEM and marker analysis, but also appeared leaky as was shown by MSOT. 2) A hampered switch to the production of LC-PUFAs, which appears to gain importance after the passage of the BTB, as it was shown by the reduction in AA, EPA, DPA and DHA containing PCaas and PCaes 3) Abnormal regulation of intercellular contacts and communication after the passage of the BTB, indicated by the increase of adluminal junctional proteins. dual function of Sms1: 1. Sms1 is involved in the BTB (Sert) and 2. Sms1 is involved in maintaining the integrity of syncytial elongating spermatid clusters (elSp). B) β-gal staining in epididymis indicates expression of Sms1 at the site of the BEB (PC; principal cells). β-gal staining was also detectable in epididymal spermatozoa (Sz). (TIFF) S1 Table. Primary and secondary antibodies used for Western blotting (WB) and Immunohistochemistry (IHC). Primary and secondary antibodies are listed according to their immunogen. Dilutions of the antibodies are given for Western blotting (WB) and immunohistochemistry (IHC). (XLSX) S2 Table. Differences between Sms1 WT and mutant animals in respect to body and organ weights. Animals of the Sms1 mouse line were weighed and dissected. Body weights, as well as weights of the organs are listed according to genotype and sex. Data is shown as mean ± SD. (DOCX) S3 Table. Lipid profile testicular Cer species of Sms1 -/-, compared to Sms1 WT levels. Levels of Cer species, sorted by intensities, are shown. Data is presented as mean percentage ± SEM, with Sms1 WT levels set to 100% (Sms1 WT , n = 15; Sms1 -/-, n = 17); Ã p 0.05, ÃÃ p 0.01, ÃÃÃ p 0.001. (XLSX)