Merlin Isoforms 1 and 2 Both Act as Tumour Suppressors and Are Required for Optimal Sperm Maturation

The tumour suppressor Merlin, encoded by the gene NF2, is frequently mutated in the autosomal dominant disorder neurofibromatosis type II, characterised primarily by the development of schwannoma and other glial cell tumours. However, NF2 is expressed in virtually all analysed human and rodent organs, and its deletion in mice causes early embryonic lethality. Additionally, NF2 encodes for two major isoforms of Merlin of unknown functionality. Specifically, the tumour suppressor potential of isoform 2 remains controversial. In this study, we used Nf2 isoform-specific knockout mouse models to analyse the function of each isoform during development and organ homeostasis. We found that both isoforms carry full tumour suppressor functionality and can completely compensate the loss of the other isoform during development and in most adult organs. Surprisingly, we discovered that spermatogenesis is strictly dependent on the presence of both isoforms. While the testis primarily expresses isoform 1, we noticed an enrichment of isoform 2 in spermatogonial stem cells. Deletion of either isoform was found to cause decreased sperm quality as observed by maturation defects and head/midpiece abnormalities. These defects led to impaired sperm functionality as assessed by decreased sperm capacitation. Thus, we describe spermatogenesis as a new Nf2-dependent process. Additionally, we provide for the first time in vivo evidence for equal tumour suppressor potentials of Merlin isoform 1 and isoform 2.


Introduction
The neurofibromatosis type II (NF2) gene encodes the tumour suppressor protein Merlin, which belongs to the ezrin-radixin-moesin (ERM) family of actin-binding proteins. Functionally, Merlin is unique within this family, being the only tumour suppressor. Loss of Merlin due to germline NF2 gene mutations leads to neurofibromatosis type 2 (NF2) disease, a nervous system cancer syndrome [1] where patients primarily develop schwannomas and meningiomas. Furthermore, somatic NF2 mutations have also been reported in a large proportion of sporadic peripheral and central nervous system tumours [2].
Studies on the expression pattern of NF2/Nf2 in humans, rats and mouse embryos have shown that it is expressed in virtually all analysed organs [3][4][5][6][7]. Although the hallmark of NF2 disease is glial cell tumours, there is increasing evidence that the ubiquitously expressed Merlin protein functions as a tumour suppressor in multiple cell types, including, for example, breast carcinoma and melanoma [6,8]. In addition, in vivo models involving the deletion of Merlin have facilitated not only the understanding of Merlin's function as a tumour suppressor in multiple cell types but also the elucidation of a critical role for Merlin in development [9,10].
Heterozygous Nf2 knockout mouse models develop a variety of cancers, including hepatomas and osteosarcomas [10]. In addition, liver-specific deletion of Merlin results in liver overgrowth and tumourigenesis [11] [12] [13]. Interestingly, Merlin loss in the skin shows no effect on progenitor proliferation but instead leads to loss of progenitor cell polarity via an alteration of adhesion junctions [14]. Moreover, in bone tissue, conditional deletion of both Nf2 alleles has revealed a role for Merlin in the regulation of the haematopoietic stem cell microenvironment [15]. Finally, we have recently shown that Merlin plays a neuron-intrinsic role in the regulation of dendrite formation and axon calibre control in the central and peripheral nervous systems [16,17]. Collectively, these studies demonstrate not only Merlin's broad function but also its cell type-specific mechanism of action.
The human gene NF2 and its close murine homologue Nf2 encode for 17 exons that are subject to alternative splicing [18]. By far the most abundant isoforms are isoform 1 and isoform 2 (sometimes denoted as isoform 3), which comprise the full-length isoforms [5]. Exonskipping of exon 16 leads to the production of isoform 1 mRNA that encodes 16 amino acids (LTLQSAKSRVAFFEEL) specific to Merlin isoform 1. Retention of exon 16 leads to translation of an altered C-terminus with 11 unique amino acids (PQAQGRRPICI) in Merlin isoform 2 [5].
The specific function of isoform 1 and 2 is controversial. Of the two major Merlin isoforms, only isoform 1 was originally thought to have tumour suppressor potential [19][20][21]. However, more recent studies suggest that both isoforms have relatively equal growth suppressive potentials and act similarly in vitro [22][23][24][25][26].
We recently reported the first specific function of Nf2 isoform 2 in vivo, demonstrating that axon calibre regulation depends on isoform 2 in neurons. We found that mice deficient specifically in Nf2 isoform 2 develop polyneuropathies similar to patients with NF2 disease [17]. Our discovery of an isoform-specific function in neurons raises the possibility of functional specialisation of the two major Nf2 isoforms. To the best of our knowledge, no pathogenic mutation that specifically strikes one isoform of NF2 has been described to date-tumourigenic mutations always inactivate both isoforms [27,28]. Therefore, NF2 syndromes do not offer an insight into the specific functions of Merlin isoform 1 and isoform 2. Since positive evolutionary selection pressure must be exerted on the maintenance of two different C-termini [18], it is likely that they perform distinct functions in some cells or at some point during development.
To investigate this notion further, the present study examined the in vivo significance of the two major Merlin isoforms by employing mouse models deficient in either Nf2 isoform 1 or isoform 2. Specifically, we aimed to clarify the tumour suppressor potential of both isoforms and analyse the effect of isoform inversion in organs physiologically expressing primarily one isoform. In this study we show that deletion of either Nf2 isoform does not cause tumour formation in mice but unexpectantly decreases sperm quality, demonstrating a novel role for Merlin in spermatogenesis.

Materials and Methods
Animal housing and genotyping Animals were kept according to local governmental and institutional animal care regulations (Zweckverband Veterinär-und Lebensmittelüberwachungsamt Jena-Saale-Holzland). Animal experiments were approved and licensed by the same authorities (see above; license number: "J-SHK-2684-04-08-02/14"). Mice were maintained under a 12-h light/12-h dark cycle at 22°C and provided with standard laboratory chow and tap water ad libitum. Mice were culled by cervical dislocation or CO 2 asphyxiation at the stated age. To prevent suffering, ageing mice were sacrificed upon signs of injury (e.g. scratch wound) or illness (e.g. tumour development). The Nf2 iso1 ko and iso2 ko animals-described in [17]-were backcrossed at least six times into a C57Bl/6 background from an FVB/NJ-C57Bl/6 background.
Tail clip lysates were used to genotype the mouse lines. Genotyping of iso1 ko and iso2 ko animals was conducted using the iso1 ko and iso2 ko primer pairs (Table 1), respectively, yielding a wildtype (product size of iso1 ko: 267 bp, iso2 ko: 378 bp) and a knockout (product size of iso1 ko: 391 bp, iso2 ko: 235 bp) PCR product.
Animals of different ages were analysed. Young mice are denoted according to age in days as P8, P18 or P35 (post-natal day). Adult animals are denoted as young (2-3 months old) or old (16-18 months old) if age is not specified.

Organ sample preparation and storage
Organ samples were obtained from animals killed by cervical dislocation. For histological preparations, liver and skeletal muscle tissue was fixed overnight at 4°C in phosphate-buffered saline (PBS)-buffered 4% paraformaldehyde (PFA), pH 7.0. Testis tissue was fixed in Bouin's fluid (0.9% picric acid, 10% formaldehyde, 5% glacial acetic acid) overnight and then washed multiple times in 70% ethanol. Following fixation, tissue was embedded in paraffin.
Testis used for in situ hybridisation (ISH) was frozen in isopentane on dry ice and then stored at -80°C. Cryosections of 20 μm thickness were cut for the ISH procedure on a cryostat (Leica CM 3050S) at -20°C, thaw mounted onto silane-treated slides (Superfrost Plus; Thermo Scientific), air dried and stored at -80°C.
For mRNA extraction and protein lysate preparation, organs were snap frozen in liquid nitrogen and stored at -80°C.

Histological stainings
Paraffin blocks were sliced to generate 5-μm thick sections which were dewaxed and rehydrated using a graded alcohol series and then stained with H&E (haematoxylin & eosin), PAS (periodic acid-Schiff) or Masson' trichrome stains according to standard protocols [29]. Stained tissues were embedded in xylene based medium under a coverslip and analysed by light microscopy using an Olympus AX70 microscope. Picture quality was optimized by Photoshop (Adobe).

In situ hybridisation (ISH)
A cDNA fragment corresponding to nt 497-997 of mouse Nf2 cDNA (transcript ID NM 001252250.1) recognising exons 1-4 as a total Nf2 transcript probe was generated by PCR using the ISH-mmNf2-ex2 primer pair (Table 2) and subcloned into the pCR2-TOPO vector (Invitrogen). Radiolabelled riboprobes were generated using [ 35 S]UTP (Hartmann Analytic GmbH) as substrate for the in vitro transcription reaction and T7 as well as SP6 polymerases for preparing antisense and sense riboprobes, respectively, according to standard protocols. The detailed ISH procedure is described in the supplementary methods. Experiments conducted with the respective sense probes did not produce any hybridisation signals.

RT-PCR and qPCR
mRNA was isolated from liquid nitrogen snap-frozen organs with Qiagen's RNeasy kit (#75142) according to the instructions using DNase (#79254, Qiagen) on-column digestion, Trizol (Invitrogen) pre-lysis and mRNA extraction using the Precellys homogenisation system (Peqlab) with zirconium beads. For small sample sizes and cells, Macherey-Nagel's NucleoSpin RNA XS kit (#740902) was used. cDNA was produced using the Omniscript kit from Qiagen (#205111). RT-PCR was performed using MyTaq 2× Mastermix (#BIO-25043, Bioline) and 0.5 μM primer concentration (Table 3). qPCR (quantitative real-time PCR) was performed using Invitrogen's Fast SYBR Green Master Mix (#4385616) on a Roche Lightcycler 480. In order to calculate relative mRNA levels, one sample was serially diluted to produce a 4x, 2x, 1x, 0.5x and 0.25x standard. Amplification efficiency was calculated for each primer pair using these standards and the second derivative maximum of the fluorescence intensity curve as a cycle call. mRNA levels of all other samples were then calculated according to the standard curve and therefore describe relative abundance to the sample used as a standard (in figures this is referred to as "normalized to X", where x is the standard sample).

Fertility and sperm analysis
Assessment of fertility was carried out by mating one male with one female at the age of 2 months and documenting the litter frequency and size for 6 months.
Sperm motility was assessed by manually classifying sperm into three categories: progressive motility (sperm moving in one direction or around in large circles), non-progressive motility (sperm tail is moving but sperm does not cover distance while moving, either being stuck in one place or moving in small circles) or immobility (no movement of sperm or sperm tail).
Sperm count and morphology were assessed from immobilised sperm preparations in 0.1% PFA. Iso1 ko and iso2 ko sperm without cytoplasmic droplets (CD) and head-midpiece-agglutinations (AG) were counted as normal.
Capacitation assay was performed by incubating minced cauda epididymis for 20 min at 37°C in non-capacitating Whitten's medium to let sperm swim out (swim-up sperm). Swimup sperm were counted using an improved Neubauer counting chamber. Sperm were then filtered through a 70-μm nylon gauze and incubated for 1 h in capacitating Whitten's medium (non-capacitating medium supplemented with 15 mM NaHCO 3 and 5 mg/ml BSA (bovine serum albumin); both Sigma-Aldrich). Non-capacitated and capacitated sperm were washed with PBS, centrifuged at 2000 g for 1 min and the pellet snap frozen in liquid nitrogen.

Electron microscopy
Swim-up sperm was promptly fixed in Karnovsky fixative (2% PFA, 2.5% glutaraldehyde (Roth) in PBS, pH 7.3) for 12 h. Sperm were then embedded in 3% agar (Roth). Samples were processed on a tissue processor EMTP (Leica). Samples were washed with PBS, fixed again with 1% osmic acid (Roth) and 1% potassium hexacyanoferrate (Roth) in PBS for 2.5 h at 4°C. Samples were then dehydrated using increasing concentrations of acetone and embedded in epon (glycid ether 100; Serva) in gelatine capsules. Epon was allowed to polymerise for 48 h at 60°C. Ultra-thin sections were cut (Reichert Ultracut S (Leica) and diamond blade 35°(Diatome)) with further contrasting onto copper grids. Sections were analysed with a JEM 1400 (JEOL) with an accelerating voltage of 80 kV.

Fluorescence-activated cell sorting (FACS)
FACS of spermatogonial stem cells was performed according to the method described by Takubo et al. [31]. Mice were sacrificed and testes dissected and placed in warm PBS. The tunica albuginea was removed and the testicular cords pre-digested for 20 min with 0.1% collagenase at 37°C. Another round of collagenase digestion for 30 min was followed by digestion with 1× Trypsin-EDTA solution (PAA) and subsequent filtering through a 40-μm nylon mesh. The single cell solution was then incubated with an antibody against EpCam (clone G8.8 1:100; #14-5791, eBioscience) and MACS (magnet-activated cell sorting) sorted using the autoMACS system from Miltenyi Biotec GmbH. The positive cells were prepared for FACS by incubation with anti-Integrin α6-PE (clone eBioGoH3 1:100; #12-0495, eBioscience) and anit-c-Kit-APC (clone 2B8 1:100; #17-1171, eBioscience). The sorting was performed on a BD Biosciences FACS Aria 2. Viable cells were gated according to DAPI (Sigma-Aldrich) staining and forward scatter. Cell fractions were sorted according to c-Kit and Integrin α6 staining intensity.

Oil Red O staining
Oil Red O staining was performed on cryosections according to standard procedures [29]. Briefly, samples were incubated in Oil Red O solution for 20 min, washed with water, counterstained with haematoxylin and embedded under coverslips in aqueous mounting medium.

Luciferase assays
Luciferase assays were performed using the Promega Dual Glo Luciferase assay kit according to the manufacturer's instructions. Cells were transfected with a firefly luciferase reporter plasmid containing the CTGF promoter [32] and the pRL-TK (Promega) vector containing the renilla luciferase under the control of the thymidine kinase promoter using Lipofectamine 2000 (Invitrogen). For overexpression of merlin, pcDNA3-based Nf2 full-length sequences for isoform 1 and 2 was used. The empty backbone vector pcDNA3 was used for the control transfection. Merlin c-terminal fragments consisted of amino acids 300-595 (iso1 c-term) and 300-590 (iso2 c-term).

Hanging wire test
The "hanging wire" test for muscle strength was performed as follows: Mice were put onto a metal grid, which was then turned upside down. Precautions were taken so that the mice could not climb up the side of the grid. The time until mice fell off the grid was measured or the experiment was terminated after 60 s.

Statistical analysis
Quantitative data were analysed using the software R (open source software available at http:// www.r-project.org/). Pairwise Student's t-tests were performed for analysis of statistical significance of differences between two groups. One-way ANOVA (analysis of variance) in combination with post-hoc t-tests were used in order to compare multiple groups. P-values were corrected according to Holm [33]. Statistical significance was accepted for an α-error of less than 5% (p<0.05). P-values are denoted in figures. Statistical signifcance of tumour incidence and testicular atrophy was tested by Fisher's exact test for which we accepted an α < 0.1. Error bars either denote standard deviation (SD) or standard error of the mean (SEM).

Expression of Nf2 isoforms in vivo
Earlier studies provided only limited data regarding NF2/Nf2 isoform expression in humans and rodents [4][5][6][7]. In order to investigate the specific in vivo function of the two major Nf2 isoforms and their role during development and organ homeostasis, we first established the specific isoform expression pattern in a wide range of murine organs. This analysis was performed on organs from young adult wildtype C57Bl/6 female and male animals. Expression levels were similar in the two genders. Quantification of Nf2 isoform mRNA levels by RT-PCR ( Fig 1A) revealed Nf2 expression in all analysed organs. We used the mmNf2ex14-17 primer pair, which spans the alternatively spliced exon 16 and thus amplifies both an isoform 1-and an isoform 2-specific PCR product. In addition, we quantified levels of total Nf2 expression, as well as isoform 1 and isoform 2 expression by qPCR (Fig 1B and 1C). Prostate, lung, heart, skeletal muscle and white adipose tissue showed a strikingly high expression level of Nf2 compared to other organs ( Fig 1B). We observed a relatively equal split in Nf2 isoform 1 and isoform 2 expression in most organs; however, some organs showed a predominant expression of either isoform 1 or 2. Specifically, the testis and spleen displayed an isoform 1 preference, while the liver, heart and skeletal muscle preferentially expressed isoform 2 (Fig 1A and 1C).

Establishment of Nf2 iso1 ko and iso2 ko animal models
In order to investigate the in vivo significance of the two isoforms, we utilised isoform-specific knockout mice generated by M. Niwa-Kawakita in the Giovannini laboratory [17]. In these mice, alternative splicing of Nf2 exon 16 is altered in such a way that the cells must specifically express one isoform, thus generating a "knockout" (ko) for the other isoform (Fig 2A, S1 Fig).
The knockout of isoform 2 is achieved by deletion of exon 16, therefore only isoform 1 can be generated. This mouse line is denoted as iso2 ko. The knockout of isoform 1 is achieved by fusion of exon 15 and exon 16, i.e. deletion of intron 15, which abolishes alternative splicing and thus forces expression of isoform 2. This mouse line is denoted as iso1 ko (Fig 2A).
Tumour suppressor function of Nf2 isoform 1 and 2 Gross observation of the iso1 ko and iso2 ko animals did not reveal any obvious phenotype. The animals developed normally without any alterations in growth and body weight gain ( Fig  2B). We did not observe increased lethality for the analysed time period (up to 26 months of age) in either iso1 ko or iso2 ko animals of mixed C57Bl/6-FVB and pure C57Bl/6 background. More importantly, we did not detect increased tumour formation in young or old knockout animals as judged from histological stainings of all major organs (Table 4). Therefore, both Nf2 isoforms must be considered as genuine tumour suppressors, considering heterozygous Nf2 deletion causes the development of multiple tumours within one year [10]. In vitro, both isoforms were equally able to inhibit known Merlin regulated pathways such as the Hippo pathway (S2 Fig) in cell culture assays (also [22][23][24][25][26]). Why then do organs like the liver or the testis predominantely express isoform 2 or isoform 1, respectively? In order to answer this question, we analysed organs expressing primarily one of the two isoforms, i.e. the testis for isoform 1 and liver and skeletal muscle for isoform 2 in the respective isoform knockout animals.
No muscle or liver phenotype observed in iso2 ko animals Testicular atrophy in aged Nf2 iso2 ko mice First, despite the testis predominantly expressing isoform 1 and only a low level of isoform 2 ( Fig 1A and 1C), we surprisingly found a high incidence of intermediate testicular atrophy in aged iso2 ko animals (Fig 2C and 2D). Additonally, the affected animals also displayed a virtual absence of sperm in the epididymis (Fig 2D). We therefore investigated the role of both isoforms in the testis and male reproductive system in more detail.

Expression of Nf2 isoform 1 and 2 in the testis and epididymis
Considering the low expression level of Nf2 isoform 2 in the testis, we assumed two possibilities: either all cell types express only a small amount of isoform 2 in favour of isoform 1 or a small subset of cell types in the testis express isoform 2. We therefore aimed to identify the cells expressing isoform 1 and 2. We performed in situ hybridisation (ISH) analysis to visualize total Nf2 mRNA distribution pattern in the adult testis, which revealed strong expression of Nf2 in the seminiferous tubules as indicated by the circle-like staining pattern (Fig 3A). Thus, Nf2 is expressed in Sertoli cells or germ cells and not in the interstitium, i.e. it is not expressed by Leydig cells. Additionally, we found that Nf2 is expressed in every tubule of the epididymis ( Fig  3A). Immuno histo chemistry (IHC) staining of Merlin using an antibody that recognizes both isoforms confirmed expression in all tubules of the epididymis and revealed an apical localisation of the protein (Fig 3B). The staining pattern in the testis did not mirror the ISH data, however, this suggested an accumulation of Merlin protein in late spermatid stages and most prominently in residual bodies. Analysis of testicular protein levels at different postnatal stages also demonstrated an upregulation of Merlin around P30, at a time when the first elongated spermatids appear (Fig 3C). RT-PCR using primers spanning exon 14 to exon 17 of Nf2. Isoform 1 (iso1) mRNA lacks exon 16, making it 45 bp smaller than isoform 2 (iso2). Cyclophilin D levels were used as an mRNA loading control. Arrows denote organs primarily expressing one of the two isoforms (testis, heart, muscle, spleen, liver). (B) Determination of mRNA expression levels in male organs using qPCR with primers specific for exon 2 (total Nf2 mRNA); normalized to lung levels. (C) Determination of mRNA expression levels using qPCR with specific primers for isoform 1 mRNA or isoform 2 mRNA; normalized to lung levels. Relative isoform levels were consistent between the two methods (epi., epididymis; sem. ves., seminal vesicles; ant., anterior; wat, white adipose tissue). doi:10.1371/journal.pone.0129151.g001

Nf2-Merlin in Spermatogenesis
The exon specific for isoform 2 is only 45 bases long and all exons of isoform 1 are present in isoform 2 mRNA severely limiting ISH probe sensitivity. Therefore, we could not use isoform-specific probes. Instead, we aimed to use the sensitive qPCR technique with specific primers for each isoform. The ISH revealed high Nf2 expression in the germ cell-Sertoli cell compartment, and since one of the smallest subpopulations of the seminiferous tubules is  Nf2-Merlin in Spermatogenesis comprised of spermatogonial stem cells (SSCs), we made use of established SSC enrichment protocols using MACS/FACS [31]. Spermatogonia were MACS pre-sorted for EpCam + cells. Subsequent FACS analysis sorted the cells according to Integrin α6 + and c-Kit expression. Cells were sorted into three groups: non-SSCs (Integrin α6 + , c-Kit + ), differentiated SSCs (Integrin α6 + , c-Kit + ) and undifferentiated SSCs (Integrin α6 + , c-Kit + ) (Fig 4A). We used the undifferentiated SSC marker Zbtb16 to control the purity of the fractions (Fig 4B). Analysis of isoform mRNA expression in the purified cells revealed an interesting pattern of isoform expression in the testis. Spermatogonia (Integrin α6 + ) displayed an almost 20-fold increase in isoform 2 expression compared to non-SSCs (Integrin α6 + ) and a six-fold increase compared to whole testis lysates (Fig 4D). In contrast, isoform 1 expression was highest in the total lysate, thus indicating that this isoform is expressed primarily in a non-SSC compartment, either the Sertoli cells or differentiated germ cells like spermatocytes or spermatids (Fig 4C).

Sperm analysis of Nf2 iso1 ko and iso2 ko animals
The relative distribution of motile and immotile sperm was similar in the two isoform knockouts ( Fig 5A). Surprisingly, the iso1 ko mice produced about twice as much sperm as the wildtype, whereas iso2 ko animals exhibited normal sperm levels ( Fig 5C). Iso1 ko and iso2 ko mice displayed reduced sperm quality compared to wildtype animals (Fig 5D), as characterised by sperm coiling (SC) and increased occurrence of cytoplasmic droplets (CD) in vas deferens sperm ( Fig 5B). Interestingly, sperm preparations from iso1 ko animals featured an increased percentage of abnormal sperm compared to iso2 ko animals ( Fig 5D).
In order to further refine the phenotype of the knockout sperm, we analysed sperm by transmission electron microscopy (EM). While some sperm appeared normal, we detected a wide variety of sperm head defects and abnormal midpiece structures in the atypical sperm of both iso1 ko and iso2 ko animals compared to wildtype littermates (Fig 5E). The flagella seemed to be formed normally and we observed the typical "9+2" formation of microtubules in flagellum cross-sections [34].
Capacitation reaction of sperm deficient in Nf2 isoform 1 or 2 Finally, we aimed to investigate whether the defects found in the knockout sperm of both isoform animals would have any consequences on sperm function. After ejaculation, sperm undergo capacitation, enabling the acrosome reaction and hypermotility, which are mandatory for successful fertilisation [35]. We tested the sperm in an in vitro capacitation assay. Stimulation of wildtype sperm with BSA and NaHCO 3 leads to activation of the capacitation reaction, visible by an increase in tyrosine phosphorylated proteins in Western blot analysis. We found that sperm from both isoform knockout mice displayed the typical tyrosine phosphorylation pattern seen in wildtypes, but with a reduction in intensity best seen in the prominent bands Nf2-Merlin in Spermatogenesis above 70 kDa, which were reduced to approximately 50% of that of wildtype littermates ( Fig  5F). Thus, the knockout of either Merlin isoform has a detrimental effect on sperm function.

Fertility impairment varied with genetic background
Decreased sperm functionality should also lead to an impairment of fertility. We investigated fertility of the knockout lines by recording the littersize over a timespan of 6 months for different combinations of breeding pairs. Despite lower sperm quality in both iso1 ko and iso2 ko Nf2-Merlin in Spermatogenesis animals, we observed a reduction in litter size only in iso2 ko breedings, which could be attributed to the males (Fig 6A). However, fertility was only impaired in animals on a C57Bl/6 background, animals on a mixed C57Bl/6-FVB/NJ background did not show any signs of decreased fertility (Fig 6B).
In sum, deletion of Nf2 isoform 1 or 2 led to sperm defects and in the case of iso2 ko animals to impaired fertility. Therefore, we next investigated whether this phenotype orignated in the testis or the epididymis.  ko sperm showed lower sperm quality as measured by cytoplasmic droplets and sperm coiling. (E) Electron microscopy of sperm. Iso1 ko and iso2 ko sperm preparations showed normal flagellum morphology but frequently aberrant sperm head morphology as seen by, for example, midpiece deformation (arrows). (F) In vitro sperm capacitation assay analysed by Western blot phospho tyrosine staining. Both iso1 ko and iso2 ko sperm showed decreased phospho tyrosine levels compared to their littermates (quantification correlated the prominent band around 72 kDa against the constitutively phosphorylated protein band at 110 kDa, arrowheads).

Morphology of testis and epididymis
We first analysed testis weight as an indicator of functional spermatogenesis. Additionally, since the isoform knockout is total and applies to all cells, the hypothalamus-pituitary-testis (HPT) axis might be affected by the knockout and, via disturbed hormone levels, influence spermatogenesis. As a read-out for intact HPT homeostasis, we used seminal vesicle weight which correlates tightly with active testosterone levels in the mouse [36]. The testes' and seminal vesicles' weight of both iso1 ko and iso2 ko animals was unchanged in comparison to wildtypes at 3 months of age (Fig 6C), indicating functional spermatogenesis and normal testosterone levels.
Since testis weight was unchanged sperm production in the testis was judged to be grossly normal. We therefore investigated the testicular histology of 3 months old iso1ko and iso2ko animals (Fig 7). However, we found no pathological alterations nor other hints for the cause of the sperm defects in young adults, indicating that the development of atrophy in iso2 ko testes is strictly age dependent and likely independent of the sperm defect.
Additonally, we analysed the status of the two most prominent Merlin-regulated pathways, the MAPK and Hippo pathways. However, no alterations of Erk phosphorylation, as a measurement of MAPK activity, or Yap phosphorylation, as an indicator of Hippo signalling, were observed (Fig 6D and 6E). Interestingly, while iso2 ko testes showed similar Merlin isoform 1 levels as the wildtype (Fig 6D), deletion of isoform 1 caused a decrease in total Merlin levels ( Fig 6E).

Nf2-Merlin in Spermatogenesis
Increased lysosome abundance in the caput epididymis We found no evidence for altered spermatogenesis in the testis despite a clear defect in sperm maturation and an increased sperm count of iso1 ko animals. However, both Merlin isoforms were expressed in the epididymis (Fig 1), which plays a vital role in the maturation of sperm. Additionally, sperm maturation is regulated by the microenvironment of the epididymal fluid [37,38]. Microvilli of epithelial cells, the stererocilia, harbor various proton pumps and ion channels that allow for modification of fluid parameters [39]. In general, microvilli structure and maintenance requires the proteins Ezrin, Paxillin and filamentous Actin [40,41]. Because Merlin belongs to the Ezrin, Radixin, Moesin (ERM) familiy of actin-linker proteins and is known to bind Paxillin we considered that loss of Merlin isoforms might alter the microvilli structure and function. Therefore, we investigated the epididymis as a potential origin of the sperm defects. Although histological analysis of the epididymis did not reveal morphological changes (Fig 8), electron microscopy uncovered increased lysosome formation and a lower Nf2-Merlin in Spermatogenesis abundance of small secreted vesicles in the caput epididymis (Fig 9). On the other hand, microvilli structure and abundance was judged to be normal.
In summary, we observed testicular atrophy in old iso2 ko animals and sperm defects in young animals of both knockout lines. While we found testicular appeareance to be normal in 3 months old animals, we discovered evidence for alterations of vesicle and lysosome abundance in the epididymal epithelium. It is possible that this has an effect on sperm maturation during the epididymal transit.

Discussion
Various mouse models with conditional or total deletion of Nf2 have provided ample evidence for an indispensable role of Merlin in tissue homeostasis regulation and development [9-11, 13-15, 42]. Although Nf2 is widely expressed during mammalian embryogenesis [3], its role during development is still poorly understood. In addition, the in vivo role of alternative splicing of the Nf2 gene into two major Merlin isoforms is currently unknown. In this study, we examined the functional uniqueness of the Nf2 isoforms in vivo by employing mouse models genetically altered to express only one isoform. We provide for the first time in vivo evidence Nf2-Merlin in Spermatogenesis for equal tumour suppressor function of isoform 1 and isoform 2. Neither isoform knockout affected animal development or caused increased tumour formation in the mice in our hands, both on a mixed FVB-Bl6 as well as a pure C57Bl/6 background. Additionally, we add spermatogenesis to the list of cellular processes critically dependent on the fine-tuned regulation of Nf2/Merlin expression. Surprisingly, deletion of either isoform impaired sperm quality and caused decreased sperm functionality.
Equal tumour suppressor activities of Nf2 isoform 1 and isoform 2 We demonstrate in this work that both Merlin isoform 1 and isoform 2 are fully functional tumour suppressors. Our finding that several organs specifically expressed predominantely one isoform in adult mice (Fig 1) supports the hypothesis of equal tumour suppressor potentials of both isoforms, since these organs have to rely on only one isoform to control growth signalling. It is important to note that these organs do not simply use other tumour suppressors for growth regulation. The liver mainly expresses Nf2 isoform 2 without any liver tumours observed, even in old iso2 ko mice. However, deletion of Nf2 in the liver reproducibly leads to a massive increase in liver mass and eventual tumours within one year [11,13]. Additionally, we did not observe increased tumour formation in any organ of iso1 ko or iso2 ko animals up to 26 months of age, regardless of its initial Nf2 isoform expression pattern, although heterozygous inactivation of Nf2 in mice causes the development of multiple tumours within one year [10].
Surprisingly, our results so far indicate that both isoforms can functionally compensate the loss of the other almost entirely, even in organs originally expressing primarily a single isoform such as the liver and skeletal muscle, the exception being the testes. It is possible, however, that injury models of liver or muscle would retrieve a regeneration response phenotype in the isoform knockout mouse models similar to deletion of Nf2 in adult liver [11]. Additionally, since even NF2 disease tumours are thought to arise only after multiple mutation hits [43], an alternate model would be the simultaneous deletion of other tumour suppressors to lower the tumour barrier and thus elicit differential tumour formation behaviour in iso1ko and iso2ko mice. A potential candidate for this approach would be the Cdkn2a gene (encoding P16 Ink4a and P19 ARF ) that has recently been shown to genetically interact with Nf2 mutations in the development of mesothelioma [44].

Inactivation of either Nf2 isoform impairs spermatogenesis
Unexpectantly, we found that deletion of either isoform of the mouse Nf2 gene disturbed the formation of functional sperm (Fig 5). While the deletion of isoform 1 caused a more severe decrease in functional sperm, deletion of isoform 2 led to testicular atrophy in aged animals. It is possible that germ cell loss is iso2 ko mice is caused by aberrant spermatogonial stem cell maintenance as Nf2 isoform 2 was enriched in the testicular stem cell population. Interestingly, we observed impaired fertility only for iso2 ko animals and only on a C57Bl/6 background, not on a mixed C57Bl/6-FVB/NJ background. Noteworthy though, it has previously been reported that inactivation of protein 4.1G, which belongs to the protein 4.1 superfamily together with Merlin, displays impaired male fertility only on a mixed 129SV-C57Bl/6 background but not on a pure C57Bl/6 background [45]. The authors of this study suggest "the presence of major modifier genes that influence 4.1G function" on different backgrounds. Furthermore, Merlin has already been described to play a role in Drosophila melanogaster spermatogenesis in regulating mitochondria morphogenesis of maturing sperm [46,47]. Additionally, Giovannini et al. reported that accidental deletion of Nf2 in the testis causes infertility in mice [42]. Collectively, the existent data strongly argue for a vital role of Merlin during spermatogenesis.
The Nf2 isoforms are expressed at different stages of spermatogenesis, namely isoform 2 in spermatogonia and isoform 1 in differentiated germ cells and possibly Sertoli cells (Fig 4). Our data additionally suggests that Merlin protein levels peak during the final phases of spermatid elongation but that Merlin is then concentrated in residual bodies, as sperm in the epididymis displayed only weak Merlin immuno-reactivity. It is therefore likely that the singular action of each isoform plays a vital role in spermatogenesis regulation. Importantly, comparison of seminal vesicles' weight as an indicator of testosterone levels excluded disturbance of hormonal regulation due to the knockout of the isoforms in all cells as a cause of spermatogenesis impairment. Rossi et al. reported in 2004 that Nf2 is three-fold enriched in mouse spermatocytes compared to spermatogonia [48]. This supports our finding that isoform 1, the predominant testicular Nf2 isoform, is expressed in later germ cell stages. It is unclear, however, whether Sertoli cells also express Nf2, since our FACS and ISH data do not distinguish between Sertoli cells and later germ cell stages. In conclusion, these results clearly demonstrate that fine-tuned expression and regulation of Nf2/Merlin are important for spermatogenesis in the testis. We could not, however, satisfactorly conclude the molecular cause of the sperm defect. It would therefore be interesting to determine whether conditional deletion of Nf2 in Sertoli cells or germ cells can recapitulate the sperm phenotype of the isoform deletion mouse models, potentially providing better tools to study the molecular role of Merlin in the testis.
While visible sperm head aberrations ( Fig 5E) are usually due to sperm differentiation defects in the testis, sperm concentration changes and increased sperm coiling can be caused by alterations of the epididymis, specifically its secretory behaviour controlling osmolarity and volume of the epididymal fluids [37,38]. Regulation of fluid parameters depends on pumps and channels present in microvilli [39]. Microvilli are structured by Actin filaments which are stabilized by Paxillin and the ERM proteins [40,41], all of which have been shown to interact with Merlin [49][50][51]. Furthermore, the microvilli actin structure is anchored to the terminal web, which among others, is composed of spectrin proteins. β2-Spectrin has been shown to preferentially interact with Merlin isoform 2 [52], hence alterations of Nf2 isoform expression may impact epididymal microvilli, fluid absorption and sperm transport. Our electron microscopy data do not support alterations of the microvilli structure of epididymal epithelial cells, however we noticed an increase in lysosomes and a decrease in small secretory vesicles. While we have no data on epididymal fluid parameters, one can speculate that loss of Nf2 isoforms impairs epithelial secretion and absorption processes, which in turn could promote sperm coiling and might affect sperm concentration parameters as seen for the iso1 ko animals (Fig 5). On the other hand, the above mentioned proteins are also part of the apical ectoplasmic specialisation [53,54]. Disturbance thereof could likewise cause sperm differentiation defects and could potentially explain the observed testicular atrophy in iso2 ko animals at older age ( Fig  2D). Furthermore, increased sperm production does not necessarily lead to a proportional increase in testis weight [55]. Increased sperm concentration observed in iso1 ko mice might therefore also be explained by pathological testicular spermatogenesis. Finally, altered epididymal fluid transport could be the cause for the observed testicular atrophy in aged iso2 ko mice as occlusion of e.g. the efferent ducts is known to lead to long-term tubular atrophy [56].
It is unclear whether the complete inactivation of one isoform is comparable to the heterozygous deletion of NF2, as observed in NF2 patients. Complete loss of one isoform might be more obstructive to spermatogenesis than a gene dosage decrease of both isoforms. Moreover, we noticed a change in total Merlin levels both after knockout of isoform 2 in the liver and isoform 1 in the testis, further shrouding the effect of single isoform action. Thus, we cannot presently conclude that spermatogenesis is likely to be affected by inheritance of NF2 mutations in humans.

Conclusion
The functional role of Merlin on the molecular level appears to be complex. Various studies reported that Merlin plays a role in the formation of adherens junctions, the distribution of cell-surface receptors, and regulation of a diverse array of signalling cascades such as the MAPK-, Rac-PAK, mTORC1 and Hippo-YAP pathways (reviewed by [28]). In addition, Nf2 encoding for two different Merlin isoforms increases the potential interactions even further. We found no evidence for isoform-specific regulation of Hippo and MAPK signalling in iso1 ko or iso2 ko tissues, with the exception of the liver. However, we attribute the increase in Erk phosphorylation to the general decrease in Merlin abundance in the iso2 ko liver.
While the signalling pathways responsible for Merlin's complex tumour suppressor function remain controversial, our understanding of Merlin as a general tumour suppressor and regulator of organogenesis is becoming more apparent. In summary, we present new in vivo evidence for equal tumour suppressor qualities of Merlin isoform 1 and isoform 2. Additionally, isoform substitution does not lead to major developmental disorders. One exception is spermatogenesis, which is sensitive to Nf2 gene disruption independent of the targeted isoform, thus revealing a novel regulatory role of Merlin in the male reproductive system.
Supporting Information S1 Fig. Analysis of Nf2 mRNA levels in iso1 ko and iso2 ko organs. (A) qPCR analysis of Nf2 expression in testes demonstrating the specificity of the isoform primers used. Deletion of isoform 1 clearly increased isoform 2 mRNA levels in the testis (n = 2 for each genotype). (B) qPCR analysis of iso2 ko liver Nf2 mRNA showed successful deletion of isoform 2 mRNA, but in contrast to the clear induction of isoform 1 protein levels, mRNA levels remained similar to the wildtype (n = 3 for each genotype). Note that isoform 1 and 2 do not add up to total Nf2 levels due to the use of a dT 18 primer, lessening the amount of cDNA containing the 5' mRNA region. In all cases the total level of Nf2 mRNA was reduced compared to the wildtype. However, Merlin protein levels did not correspond clearly to the mRNA levels ( Fig 6D and 6E The median liver fat content of male iso ko mice was slightly increased (Mann-Whitney U-test, #: significance level α<0.10, n(wt) = 8, n(iso2 ko) = 9). (C) Western blot analysis of iso2 ko muscle showed similar Hippo (phospho Yap) and MAPK (phospho Erk) activity in knockout and wildtype. (D) Loss of isoform 2 in liver decreased Erk phosphorylation in Western blot analysis. Western blot analysis of muscle tissue (E) confirmed similar total merlin levels in both knockouts, whereas the level of total Merlin was upregulated in iso2 ko liver (F). (TIF)