DNMT3L Is a Regulator of X Chromosome Compaction and Post-Meiotic Gene Transcription

Previous studies on the epigenetic regulator DNA methyltransferase 3-Like (DNMT3L), have demonstrated it is an essential regulator of paternal imprinting and early male meiosis. Dnmt3L is also a paternal effect gene, i.e., wild type offspring of heterozygous mutant sires display abnormal phenotypes suggesting the inheritance of aberrant epigenetic marks on the paternal chromosomes. In order to reveal the mechanisms underlying these paternal effects, we have assessed X chromosome meiotic compaction, XY chromosome aneuploidy rates and global transcription in meiotic and haploid germ cells from male mice heterozygous for Dnmt3L. XY bodies from Dnmt3L heterozygous males were significantly longer than those from wild types, and were associated with a three-fold increase in XY bearing sperm. Loss of a Dnmt3L allele resulted in deregulated expression of a large number of both X-linked and autosomal genes within meiotic cells, but more prominently in haploid germ cells. Data demonstrate that similar to embryonic stem cells, DNMT3L is involved in an auto-regulatory loop in germ cells wherein the loss of a Dnmt3L allele resulted in increased transcription from the remaining wild type allele. In contrast, however, within round spermatids, this auto-regulatory loop incorporated the alternative non-coding alternative transcripts. Consistent with the mRNA data, we have localized DNMT3L within spermatids and sperm and shown that the loss of a Dnmt3L allele results in a decreased DNMT3L content within sperm. These data demonstrate previously unrecognised roles for DNMT3L in late meiosis and in the transcriptional regulation of meiotic and post-meiotic germ cells. These data provide a potential mechanism for some cases of human Klinefelter's and Turner's syndromes.


Introduction
Infertility affects 1 in 20 men in the Western world. Increasingly data is showing that in addition to genetic changes, aberrant epigenetic regulation during spermatogenesis can lead to human infertility [1][2][3]. Previous studies on the epigenetic regulator DNMT3L, have shown that deletion of Dnmt3L in mice resulted in hypomethylation of paternally imprinted regions and transposable elements in gonocytes, as well as developmental arrest and death of early meiotic cells [4][5][6]. Consequently Dnmt3L knockout males were sterile.
Dnmt3L mRNA and protein were originally reported only in the primordial germ cells, where DNMT3L is needed to establish epigenetic marks required in early primary spermatocytes [4,7]. Subsequently, it was shown that Dnmt3L mRNA was present in spermatocytes and spermatids [5,8]. Furthermore, it was shown that a second promoter in intron 9 initiates three alternative transcripts (Forms 1, 2 and 3) in late pachytene spermatocytes and round spermatids [9]. While the function of the alternative isoforms is unknown, they are thought to be non-coding and to suppress translation of the full length Dnmt3L isoform [9]. The localization of DNMT3L protein within the adult testis or sperm has not been reported.
Dnmt3L was recently reported as the first example of a mammalian paternal effect gene [10]. Animals with mutations in such genes are of particular interest as they exhibit a male heterozygous phenotype, that can affect the phenotype of their wild type progeny. Specifically, when Dnmt3L heterozygous males were mated to wild type females, there was an increased incidence of XO monosomy in wild type E10.5 embryos and adults compared to wild type matings [10]. These data suggested that the Dnmt3L heterozygous males may have a meiotic defect related to X and Y chromosome segregation and encouraged us to study the effects of the loss of a Dnmt3L allele in germ cells in more detail.
We hypothesised that decreased availability of DNMT3L in Dnmt3L heterozygous males may lead to the inappropriate epigenetic marking of the DNA, and a partial failure of sex chromosomes to convert from an euchromatic state to a highly condensed heterochromatic state leading to a meiotic instability. We hypothesised that if true, abnormal XY compaction would be reflected by aberrant patterns of gene expression in meiotic and haploid germ cells. To investigate these hypotheses we measured the lengths of the XY bodies in meiotic cells from heterozygous and wild type males, we defined the incidence of XY sperm aneuploidies, and monitored transcriptional activity in spermatocytes and round spermatids.
Dnmt3L heterozygous males had XY bodies which were significantly longer than those from their wild type brothers, and exhibited a significantly increased incidence of XY non-disjunction. We demonstrated that a loss of a Dnmt3L allele was associated with disregulation of a number of X-linked and autosomal genes in spermatocytes and spermatids. Consistent with these observations, we localized DNMT3L protein within spermatids and mature sperm. Furthermore, our data strongly suggested a complex Dnmt3L self-regulatory mechanism in germ cells similar to that previously reported in embryos and ES cells [11], but additionally involving the alternative transcripts.

Dnmt3L heterozygous males displayed XY body abnormalities
During male meiosis, the X and Y chromosomes form an XY body and become transcriptionally silent in a process known as meiotic sex chromosome inactivation (MSCI) [12,13]. Unlike the autosomes, in males sex chromosomes are heterologous and only synapse at their pseudoautosomal regions. To test the hypothesis that DNMT3L has a role in X chromosome meiotic condensation, we measured the lengths of the XY bodies in Dnmt3L (2/+) and Dnmt3L (+/+) males. The X and Y chromosomes typically join at their pseudoautosomal regions to form a highly condensed and inactive XY body during the pachytene stage of meiosis (Fig. 1A). Many studies have conducted meiotic spreads in order to look for the localization of proteins on the XY body during pachytene [14,15], however, to our knowledge there have not been any studies on the physical conformation of the XY body. As such this study necessitated defining the normal spectrum of XY lengths within pachytene spermatocytes from wild type mice.
To assess XY body length in a quantitative and unbiased manner we measured the lengths of XY bodies in meiotic spreads from wild type and Dnmt3L heterozygous males. All measurements were conducted without knowledge of genotype. For each group (C57Bl/6 wild type, Dnmt3L (+/+) and Dnmt3L (2/+) ) the percentage of XY bodies of a particular length was determined (Fig. 1C). Phosphorylated Histone 2AX (cH2AX) was used as an XY body marker. Synaptonemal complex protein 3 (SCP3) was used to label chromosome pairs. In order to assess the possibility that Dnmt3L wild type males may also have a meiotic defect we included a C57Bl/6 control mouse group in addition to wild type males from the Dnmt3L colony.
Consistent with previously published data showing that the degree of XY body compaction increases throughout pachytene [16], meiotic spreads from wild type males displayed a range of lengths spanning 6 to 30 mm (Fig. 1C). Compared to meiotic spreads from wild type animals, the XY bodies from heterozygous males appeared abnormal and longer more often than those from wild type males ( Fig. 1A and B). A random-effects general linear model found that the mean length for the Dnmt3L heterozygotes XY bodies was 2.4 mm longer than the mean length for the Dnmt3L wild type XY bodies (95% CI: 0.6 to 4.2 mm, P = 0.01). It is evident that the Dnmt3L wild type males had a larger percentage of XY bodies with shorter lengths when compared to the Dnmt3L heterozygotes, (Fig. 1C). No significant difference was observed between the lengths and distribution of the Dnmt3L wild type and the C57Bl/6 wild type XY bodies (95% CI: 22.6 to +1.1, P = 0.42) (data not shown). These data suggest that the loss of a Dnmt3L allele resulted in a partial failure of XY body chromatin compaction during meiosis. Interestingly, we observed no abnormalities in autosome pairing in germ cells from Dnmt3L heterozygous males (data not shown).
Dnmt3L heterozygous males had an increased risk of producing XY sperm To address the hypothesis that a partial failure of XY body compaction would lead to an increase in XY bearing sperm in Dnmt3L heterozygous males, we used chromosome paints to label decondensed sperm for X and Y chromosomes (Fig. 2). The visual assessments and counting were all conducted without knowledge of genotype. Sperm with no labelling were not included in the analysis in order to avoid the possibility of false negative sperm as a consequence of incomplete sperm head decondensation. The number of sperm lacking a sex chromosome should however, equal the number carrying both an X and a Y chromosome.
From a total of 2,063 sperm (from 9 mice) we found 7 sperm with a XY chromosome content in wild type males, compared to 22 sperm (from 8 mice) from a total of 2,155 sperm in Dnmt3L heterozygous males (Supporting Information File S1). The total number of X and/or Y bearing sperm from wild type and Dnmt3L heterozygous males (analysed separately) was the same, suggesting that the FISH probes worked equally well and that the loss of a Dnmt3L allele did not result in a sex chromosome bias at the level of the sperm (data not shown). A negative binomial regression analysis showed the heterozygous males had three times the risk of sperm having a XY content, compared to sperm from wild type males (P = 0.024). As Dnmt3L wild type and heterozygous males had equivalent rates of daily sperm production (Supplementary Information File S1), this meant that Dnmt3L heterozygous males produced three times more XY bearing sperm. In humans XY sperm leads to the most common chromosome aneuploidy, Klinefelter's syndrome which occurs with a frequency of greater than 1 in 1000 male live births [17,18].
An analysis using a probe against chromosome 16 (the equivalent to human chromosome 21), did not show a difference in the frequency of aneuploid sperm from Dnmt3L heterozygous males compared to wild type males (data not shown). This result fits in with our observation of no observable increase in autosomal pairing defects during meiosis.
A number of genes were differentially expressed in Dnmt3L heterozygous males As an extension of the incomplete XY chromatin condensation, we hypothesised that Dnmt3L heterozygosity would lead to the inappropriate transcription of genes in meiotic and post-meiotic germ cells. To test this, we conducted microarray analyses on mRNA from purified primary spermatocyte and spermatid populations from wild type and heterozygous males. Cells were purified from 10 animals per replicate, and three biological replicates (30 animals) were assessed. About 60% of the Illumina probes (27,994 probes) were found to be expressed in primary spermatocytes or spermatids. Of these, fifty three were differentially expressed in spermatocytes between Dnmt3L wild type and heterozygous males at a 5% false discovery rate (FDR) (Supplementary Information File S1). Twenty nine probes (27 unique genes) were up-regulated and twenty four probes (23 genes) were down-regulated in Dnmt3L heterozygous spermatocytes compared to spermatocytes from wild types. Within spermatids, one hundred and twenty three probes were differentially expressed between Dnmt3L wild type and heterozygous animals (FDR ,0.05) (Supplementary Information File S1). Of these, seventy five probes (73 genes) were up-regulated and forty eight probes (46 genes) were down-regulated in Dnmt3L heterozygous spermatids compared to the wild type. The raw data from our microarray analyses have been submitted to GenBank GEO.
When we restricted the analysis of the microarray data to the 855 X chromosome probes, twelve genes were differentially expressed between spermatocytes from Dnmt3L wild type and heterozygous males, at FDR ,0.20 (Supplementary Information File S1) and 57 genes were differentially expressed between Dnmt3L wild type and heterozygous spermatids at FDR ,0.20 (Supporting Information File S1). An MA-plot showed a trend whereby X-linked genes which were relatively highly expressed in spermatocytes tended to be over-expressed in Dnmt3L heterozygous males (data not shown). In contrast, all the differentially expressed X chromosome genes in spermatids were downregulated in Dnmt3L heterozygous males.
To independently assess the microarray data, we conducted QPCR on RNA from purified spermatocytes and spermatids. We examined 6 X-linked genes and 7 autosomal genes that were differentially expressed in spermatocytes ( Fig. 3A and 3B), and 7 autosomal genes that were differentially expressed in spermatids   (Fig. 3C). Genes were selected based on their relative change in expression as found in the microarray data and their previously recognized roles in spermatogenesis. As for the microarrays, cells were purified from 10 males per biological replicate and a minimum of 3 biological replicates were assessed per experiment. QPCR verified all of the changes in gene expression detected by microarray (Fig. 3).
Collectively, the microarray and QPCR data showed that Dnmt3L heterozygosity resulted in significant changes in spermatocyte and spermatid autosomal gene transcription. Altered spermatocyte X chromosome gene expression was also evident, however, these changes were relatively few in number and often low in magnitude. This data was supported by RNA polymerase II staining of the XY body (as a marker of transcriptional activity) (Supplementary Information File S1), which showed that the XY bodies from Dnmt3L heterozygous males contained on average 10% more RNA polymerase II labelling than XY bodies from wild type males. This data supports a role for DNMT3L in the regulation of autosomal gene transcription in spermatocytes and round spermatids.

Deregulated Dnmt3L expression in germ cells from Dnmt3L heterozygous males
In addition to measuring thousands of other genes, the microarrays assessed Dnmt3L expression. Consistent with the data of Hata et al and La Salle et al [5,8], our data clearly showed that Dnmt3L was expressed in wild type spermatocytes and spermatids (Fig. 4A). It also showed that spermatids have a relatively higher Dnmt3L mRNA content than spermatocytes (Fig. 4A).
Contrary to our expectations, germ cells (both pachytene spermatocytes and round spermatids) from heterozygous males did not have a decreased level of Dnmt3L mRNA compared to those from wild type animals, as determined using a probe complementary to exon 13 (Fig. 4A). Heterozygous males carry both a Dnmt3L knockout and a wild type allele (Fig. 4B). It is of note, that male germ cells develop within a syncitium wherein mRNA and proteins produced in one cell can be shared across cytoplasmic bridges, meaning that although individual spermatids may have a different genotype i.e. heterozygous or wild type for Dnmt3L, they are equivalent at a protein and mRNA level [19]. The analysis done herein also showed that the expression of Dnmt3L was not significantly different between pachytene spermatocytes and round spermatids from Dnmt3L heterozygous and wild type males (Fig. 4A). In light of the observed meiosis and transcription phenotype, this led us to investigate the relative expression of each of the four Dnmt3L transcripts in spermatocytes and spermatids from Dnmt3L heterozygous and wild type.
Exon 13 is contained within the Dnmt3L full protein coding mRNA and all of the alternative transcripts. Forms 1, 2 and 3 transcripts start with an alternative exon 9, designated exon 9b ( Fig. 4C) [9]. Form 1 contains exon 9b and a second novel exon, exon 9cL. Form 2 contains exon 9b and a shorter version of exon 9c, exon 9cS. Form 3 contains exon 9b only. All three forms contain the same exons 10, 11, 12 and 13 as in the Dnmt3L full form mRNA, and were therefore detected on the microarray.
In order to assess the expression levels of individual Dnmt3L isoforms in spermatocytes and spermatids, we designed QPCR assays specific for the three alternative transcripts as well as assays to detect the 59, middle and 39 regions of the Dnmt3L full form transcript (Fig. 4C, Supplementary Information File S1). The primers used for Form 1, Form 2 and Form 3 were as used by Shovlin et al (Supplementary Information File S1) [9]. Based on previously published data, only mRNA transcribed from the Dnmt3L full form transcript results in functional protein [9]. While the mRNA for Forms 1, 2 and 3 isoforms share significant homology with the protein coding splice variant at the level of mRNA sequence, they do not contain an open reading frame of any length and as such can not produce protein. Relative expression was measured in a minimum of three biological replicates and each replicate contained purified germ cells from 10 males.
QPCR on spermatocytes from Dnmt3L heterozygous and wild type males showed that the loss of a Dnmt3L allele resulted in a significant up-regulation of the 59 (P = 0.0223) and mid (P = 0.0417) regions of the full Dnmt3L transcripts in the heterozygous males (Fig. 5A). While the up-regulation in mRNA containing the 59 sequence could be from either the wild type or knockout allele, the up-regulation in mRNA containing the Dnmt3L-mid sequence suggested an up-regulation from the wild type allele i.e. translation from the knockout allele would be blocked by the stop codons in the LacZ cassette (Fig. 4B). The expression of Form 1, 2, 3 and the Dnmt3L-3 products was comparable between spermatocytes from Dnmt3L heterozygous and wild type males (Fig. 5A).
Similarly, QPCR on spermatids from Dnmt3L heterozygous and wild type males showed a significant up-regulation of the 59 (P = 0.0027), mid (P = 0.0094) and 39 (P = 0.0391) regions of the full Dnmt3L transcript in the heterozygous males (Fig. 5B). In addition, there was a significant increase in the expression of Form 1 (P = 0.008) and 2 (P = 0.0129) transcripts in Dnmt3L heterozygous spermatids (Fig. 5B).
Collectively, these data showed that in addition to a compensatory up-regulation of the Dnmt3L full length transcript within spermatocytes and spermatids from Dnmt3L heterozygous males, there was an up-regulation in the use of the 39 promoter and the transcription of the alternative non-coding Dnmt3L forms, Forms 1 and 2, within spermatids. The up-regulation of the Dnmt3L isoforms suggests that the Dnmt3L knockout reported by Webster et al did not result in full ablation of the gene (Fig. 4B).

DNMT3L is found in the adult testis and sperm
Our microarray and QPCR data showed that splice variants of Dnmt3L were present in spermatocytes and spermatids, consistent with previously published data [8]. To assess the possibility that the Dnmt3L full length transcript was translated, we determined the localization of DNMT3L in testis sections using immunohistochemistry with antibodies directed against both the N-terminal and C-terminal of the full length isoform. Using both antibodies we detected DNMT3L in the nuclei of round spermatids, the heads of the elongating spermatids and in peritubular cells (Fig. 6A). This is the first report of DNMT3L in the adult testis.
The presence of DNMT3L in sperm from Dnmt3L wild type and heterozygous males was conclusively shown by Western blotting (Fig. 6B). Further, and consistent with the Dnmt3L isoform data described above, the relative content of DNMT3L in sperm from Dnmt3L heterozygous males was reduced by 28% compared to sperm from wild type males (P = 0.027, n = 3 biological replicates with 2 mice per replicate) (Fig. 6C).

Discussion
These data show for the first time, that DNMT3L is a regulator of both XY body compaction and ultimately XY chromosomal segregation. The XY bodies from Dnmt3L heterozygous males are longer than those from Dnmt3L and C57Bl/6 wild type males, and this finding supports a mechanism whereby a partial failure of chromatin compaction in the Dnmt3L heterozygous males leads to paternal effects [10]. Further, our data reveals that loss of a It is established that much of the XY body becomes condensed and silenced during pachytene [12][13][14]. Our results suggest that in the heterozygous males, the chromatin of the XY body is not condensing as tightly as in the wild type males. As a result of this, some genes that should become inactived may remain accessible for transcription. Our data shows that DNMT3L is involved in regulating some of the chromatin changes occurring on the XY body. These changes may be important for the silencing of certain genes.
Previous studies have shown that DNMT3L is essential during the early phases of male meiosis. The absence of Dnmt3L results in a failure to silence retrotransposon expression in gonocytes, and a failure of homologous chromosome condensation and pairing in leptotene and zygotene spermatocytes [4,6]. The subsequent   ''meiotic catastrophe'' is a consequence of a failure of chromosomes to progress from a euchromatic state to a heterochromatic state [6]. Recently it was also shown that in addition to a failure of synapsis during meiosis, Dnmt3L null males have an impairment of the meiotic silencing by unsynapsed chromatin (MSUC) response. As such this resulted in a failure of XY body silencing which may contribute to the Dnmt3L mutant phenotype [20]. The data presented here suggest that a similar, but more subtle, mechanism is occurring on the XY bodies of Dnmt3L heterozygous males and that this ultimately leads to a three fold increased risk of XY sperm.
Sex chromosome aneuploidies are common and potentially increasing in humans. Recent data suggest that the birth incidence of Klinefelter's syndrome (XXY) is 1.09-1.75 in 1,000 [18] and 1 in 1,000 for Turner's syndrome (XO) [18,21]. It has been shown that approximately 10% of sperm from normal healthy and fertile men have structural and numerical chromosomal abnormalities (reviewed in [22]). Sperm aneuploidies almost always arise during meiosis and defective recombination between the X and Y pseudoautosomal region is believed to be the most common cause of XY sperm disomy [23,24]. Proposed mechanisms for their aetiology include both genetic and environmental factors. Men exposed to potential mutagens, including pesticides [25,26] and chemotherapy agents [27,28], have an increased risk of sperm aneuploidies. Similarly, sub-fertile and infertile men [29,30], and mice [31], have an increased frequency of sperm chromosomal abnormalities. The data contained in this study demonstrates that decreased levels of Dnmt3L in germ cells could lead to a significant increase in XY sperm aneuploidy and as we have shown previously, an increased frequency of XO offspring [10].
Studies on the parental origin of the X chromosome in Turner's syndrome patients have found that in the majority of cases the remaining X chromosome is of maternal origin i.e. the paternal X or Y chromosome is lost [32][33][34]. Furthermore, a study on fathers of Turner's syndrome patients (with a single X chromosome of maternal origin) showed that these men had an increased incidence of XY sperm, suggesting an increased frequency of non-disjunction of the X and Y chromosomes during meiosis [35]. Collectively, these data raise the possibility that decreased expression of DNMT3L in humans may also lead to Klinefelter's or Turner's syndromes in some children.
The precise mechanisms leading to decreased XY body compaction are yet to be determined. It is possible that there are changes in the methylation status throughout the genome of Dnmt3L heterozygous germ cells compared to cells from wild type animals. Alternatively, or in addition, the aberrant expression of specific genes involved in spermatogenesis contributed to abnormalities. For example, Tsga8, which was significantly up-regulated in Dnmt3L heterozygous spermatocytes compared to cells from wild type males (Fig. 3A), codes for a protein involved in DNA condensation during spermatogenesis [36].
As an extension of the partial failure of XY body compaction, we anticipated that Dnmt3L heterozygosity would lead to an upregulation in the transcription of X linked genes in meiotic and potentially post-meiotic germ cells. Some support for this comes from the microarray data, however, the absolute number of differentially regulated genes was small. In contrast, Dnmt3L heterozygosity did result in a large number of autosomal genes being differentially expressed in both spermatocytes and spermatids, revealing a previously unrecognized role of DNMT3L in the regulation of autosomal gene transcription.
Dnmt3L heterozygosity results in both the over-and underexpression of genes. It was anticipated that DNMT3L deficiency would predominantly be associated with increased transcription as a consequence of a looser chromatin state and increased access of transcription factors to gene promoters. While this was the case for many genes, including Aire and Xrn1, some genes, including Ddx50 and EPB41L2, were significantly down-regulated. While it is clear that altered DNMT3L production results in significant changes in the expression of many genes, including Dnmt3L itself, it is not as yet possible to determine whether each individual change is as a direct consequence of DNMT3L deficiency, or is secondary to changes in other epigenetic regulators or transcription factors ie. these data suggest that DNMT3L sits at the start of a cascade of gene regulation networks which ultimately impact on many aspects of a cells phenotype.
Further, our microarray data show that contrary to expectations, the overall level of Dnmt3L transcript, based on a probe against exon 13, was unchanged. This prompted us to analyse Dnmt3L expression more carefully. Quantitative PCR data show that in germ cells from Dnmt3L heterozygous males there was upregulated transcription of the full Dnmt3L splice variant from the remaining wild type allele, and there was increased use of an alternative promoter and transcription of mRNA species encoding the alternative Dnmt3L variants, Forms 1 and 2 in spermatids. The allele(s) from which these variants were transcribed was not analysed, however, it may have occurred from both the knockout allele and the wild type allele.
A previous study has shown that members of the DNMT3 group i.e. DNMT3L, DNMT3A and DNMT3B, can regulate methylation of the Dnmt3L promoter region [11]. In ES cells and embryos, a deficiency in Dnmt3a and Dnmt3b resulted in hypomethylation of the Dnmt3L promoter region. Similarly, a deficiency in Dnmt3L resulted in hypomethylation of its own promoter. The authors proposed that the DNMT3 members are involved in an auto-repression mechanism, whereby if methyltransferase activity is reduced, increased transcription of Dnmt family members occurs to compensate [11]. Our analysis of Dnmt3L expression in germ cells from wild type versus heterozygous males showed a similar mechanism exists in male meiotic and haploid germ cells. In addition, our data suggest that within at least spermatids, this mechanism includes altered transcription of the non-coding alternative Dnmt3L transcripts.
While not directly tested here, the simplest explanation for increased transcription of Dnmt3L in germ cells from heterozygous males, is hypomethylation of the CpG promoter region of Dnmt3L. In support of this hypothesis, we detected a dramatic increase in Aire expression in both spermatocytes and spermatids ( Fig. 3B and C). The Dnmt3L and Aire promoters are tightly linked on chromosome 10 in the mouse, with only 6.5 kb separating the first exon of each gene in opposite transcriptional directions [37]. AIRE (Autoimmune Regulator gene) is a transcription factor expressed in antigen presenting cells in the thymus [38], and is involved in establishing and maintaining immune tolerance [39]. Recent work has shown that AIRE is also found in spermatogonia and spermatocytes [40] and that an absence of Aire leads to male infertility most likely at the level of the epididymis [41]. Our data shows that Aire is also expressed in spermatids and strongly supports a co-regulatory relationship between Aire and Dnmt3L. As both Aire and Dnmt3L promoters are known to be regulated by CpG island methylation, the compensatory up-regulation of the full length Dnmt3L splice variant from the remaining wild type allele, and the prominent over-expression of Aire in both spermatocytes and spermatids from Dnmt3L heterozygous males, strongly supports a mechanism whereby hypomethylation of the shared promoter has lead to increased promoter access. These data also raise the possibility that at least part of the phenotype observed herein is a consequence of increased Aire transcription.
Despite mRNA data showing Dnmt3L expression in the testis, the detection of DNMT3L in the adult testis or in sperm has not previously been reported. Our data confirmed the expression of all four Dnmt3L splice variants in spermatocytes and round spermatids, and we have localized DNMT3L protein in round and elongating spermatids and in mature sperm. Further, the presence of DNMT3L in wildtype sperm raise the possibility that DNMT3L is carried through to the zygote/embryo where it may impact on health. The potential role of DNMT3L in the embryo was beyond the scope of this study.
Collectively, we have shown that the loss of a Dnmt3L allele results in a partial failure of X chromosome condensation during male meiosis and subsequently, a three fold increase in XY chromosome non-disjunction. We have shown that altered Dnmt3L expression results in changes in spermatocyte and spermatid gene expression. We have shown that there is a previously unrecognised compensatory up-regulation of both the Dnmt3L full length and alternative transcripts in germ cells from Dnmt3L heterozygous males. We have shown that DNMT3L protein is found within both haploid male germ cells and mature sperm. Furthermore, these data raise the clinically important prospect that deregulated DNMT3L production may contribute to the aetiology of human Klinefelter's and Turner's syndromes.

Animals
The Dnmt3L knockout mice were maintained in the C57B6 background as described previously [6]. Briefly, the Dnmt3L knockout allele was generated using a BamHI genomic fragment containing coding exons 1-7 isolated from a PAC library (RPCI-21, http://bacpac.chori.org/ femmouse21.htm) derived from a 129S6/SvEvTac mouse. The Mouse Genome Informatics database (http://www.informatics. jax.org/) showed no single nucleotide polymorphisms between 129S6/SvEvTac and C57BL/6J. The LacZ reporter gene including three stop codons and the loxP flanked PGK-neomycin cassette, were sub-cloned in-frame into coding exon 5, thus disrupting the PHD-like zinc finger domain. The targeting vector was electroporated into C57BL/6 ES cells and clones selected in G418-containing medium. Targeted ES cells were microinjected into Balb/C blastocysts to generate chimeras. Chimeric males were mated with C57BL/6 females to generate C57BL/6 offspring. The PGK-Neo cassette was removed via Cre-loxP mediated excision by breeding the mice to a cre transgenic strain on a C57BL/6 background. Specifically, the cre transgenic strain was created in BALB/cJ and was backcrossed to C57BL/6J for at least 10 generations (http://jaxmice.jax.org/strain/006054.html). The cre (and thus BALB/cJ) -ve Dnmt3L colony has been maintained by crossing heterozygotes or crossing to wildtype C57BL/6 for well over 10 generations. Thus the wildtype and heterozygous mice used in these experiments are essentially C57BL/6.

Meiotic Spreads and immunofluorescent labelling
Methods for meiotic spreads were prepared as described previously [42]. Following fixation, slides were washed in 10% antibody dilution buffer (ADB) (PBS, 10% horse serum, 3% BSA, 0.05% Triton-X) three times for 15 minutes each. In order to measure XY body length, meiotic spreads were incubated with a 1/100 dilution of anti-SCP3 serum (ab15092, Abcam) and a 1/ 2,000 dilution of anti-phospho-Histone H2AX serum (05-636, Upstate) in 10% ADB. Slides were incubated at 37uC for 2 hours or overnight at 4uC in a dark box. Alexa Fluor secondary antibodies (Molecular Probes) were used at a 1/200 dilution in 10% ADB and applied to each section. Slides were incubated for 1 hour at 37uC or for 2 hours at room temperature in a dark box and then washed in PBS three times for 15 minutes each. Nuclear DNA was stained with DAPI and slides were mounted under coverslips with DAKO antifade mounting media.

XY body length measurements and analysis
Immunolabelled meiotic spreads were visualized and photographed using fluorescence microscopy. Each photograph was visually assessed, then XY body lengths were traced and measured using the Image Pro Plus software program. All visual assessments and measurements were conducted without knowledge of genotype. In order to graph (but not analyse) data, measurements were grouped into 2 mm increments between 6-40 mm. To analyze the XY body length data, a general mixed linear model was used for statistical analysis using the Stata software program. This statistical analysis accounted for clustering within the mice by treating mice as random-effects and estimated differences in mean lengths between the genotypes.

Sperm FISH for X and Y chromosome content
Methods for sperm decondensation and FISH were adapted from Whyte et al [43]. To enable FISH probes to easily pass into the sperm nucleus, sperm heads needed to be decondensed. Sperm were collected from dissected cauda epididymides by snipping at several locations and placed into sterile pre-warmed PBS at 37uC to allow the sperm to swim out. The sperm solution was filtered through a 100 mm mesh filter and spotted onto Superfrost plus microscope slides (Menzel-Glä ser). Using the edge of a coverslip the drops were smeared along the slide as a thin film and slides were dried on a slide warmer.
The sperm smears were dehydrated in 80% methanol at 220uC for 20 minutes and air dried. A microwave safe plastic container with a divider in the middle was filled with 100 ml of distilled water and two slides were placed on the top section of the container above the water. A combination of microwaving while incubating slides in a reducing agent was conducted. 200 ml of 10 mM Dithiothreitol (DTT) was applied to the slides and the samples were covered with a coverslip. The slides were microwaved in the container for 15 seconds at 550W, then drained and 200 ml of 10 mM Lithium diiodosalicylate/1 mM DTT buffer was applied. The slides were covered with a coverslip and microwaved for 90 seconds at 550 W, then rinsed twice with 2X SSC and air dried for 30 minutes and after fixing with methanol:acetic acid (3:1) for 10 minutes, they were air dried and used immediately for FISH.
To prevent excessive chromosome denaturation, slides were placed on a slide warmer at 50uC for 1 hour. 1 ml of each of the X (directly labelled with FITC) and Y (directly labelled with CY3) probes (Cambio Ltd.) were diluted together in 3 ml of hybridization buffer (Cambio Ltd.). The probes were denatured at 65uC for 10 minutes, then pre-annealed at 37uC for 30 minutes and applied to the slides and the area was covered with a 10 mm 2 coverslip and sealed with rubber glue. The probes, together with the sperm, were denatured by placing the slides on a slide warmer at 75uC for 3 minutes, and then incubated overnight at 37uC in a humidified chamber.
Following removal of the rubber glue, post-hybridization washes were conducted as follows: (a) 1X SSC solution at 70uC for 2-3 minutes, (b) 1X SSC/50% formamide at 45uC twice for 5 minutes each, (c) 1X SSC at 45uC twice for 5 minutes each, (d) 4X SSC/ 0.05% Tween 20 at 45uC twice for 5 minutes each. Nuclear DNA was stained with DAPI and slides were mounted with DAKO antifade mounting media.

Analysis of sperm FISH slides
Sperm were photographed using a fluorescence microscope. The number of sperm carrying X, Y or XY chromosomes were counted without prior knowledge of the genotype. A negative binomial regression model was fitted to the data using Stata statistical software. This model determined the relative risk of sperm having X, Y or XY chromosome content, and allowed us to test for differences between wild type and heterozygous mice while allowing for heterogeneity between mice.

Germ cell isolation
Germ cell sub-types were purified using centrifugal elutriation as described previously [44,45]. Testes were dissected out of 10 adult mice per replicate. Briefly, tubules were removed from the tunica and chopped well and placed into a flask containing 20 ml PBS with 1 mg/ml trypsin, 1 mg/ml collagenase, 0.5 mg/ml hyaluronidase and 8 mg/ml DNase solution and shaken in a water bath at approximately 50 cycles/minute for 25 minutes. The solution was resuspended and 20 ml of PBS was added to the flask. The flask was shaken again at approximately 60 cycles/minute for 10 minutes and the solution was filtered through an 80 mm mesh and the cells were centrifuged at 430 g for 10 minutes. Cells were resuspended in 50 ml of PBS and separated by centrifugal elutriation [44,45]. The two fractions of cells containing the round spermatids and pachytene spermatocytes were kept. Cell counts were conducted to obtain the purity of the cell fractions and only preparations with a purity of .80% were used for RNA isolation. The cells were snap-frozen in dry ice and stored at 280uC.

RNA isolation from germ cells
For cell counts of ,10,000, the Qiagen RNeasy mini kit with on column Dnase digestion was used to purify RNA. For .10,000 cells, TRIzol reagent (Invitrogen) was used. RNA extracted by the TRIzol method was Dnase treated using the Qiagen RNeasy mini kit on column Dnase digestion.

Microarray analysis of purified germ cell gene expression
RNA from Dnmt3L heterozygous and wild type spermatocytes and spermatids were hybridized at the Australian Genome Research Facility Ltd (Melbourne, Australia) using their standard methods. The microarrays used in this project were Illumina Mouse Whole-Genome WG-6 Version 1.1 BeadChips. Three preparations of purified wild type spermatocytes and three preparations of purified heterozygous spermatocytes were hybridized onto one BeadChip. Similarly, three preparations of purified wild type spermatids and three preparations of purified heterozygous spermatids were hybridized onto a separate BeadChip. Each preparation of purified cells was made up of cells isolated from 10 mice. The sample probe profile data and control gene profile data were output from BeadStudio for subsequent statistical analysis in the Bioconductor software environment [46].
The lumi software package was used for the pre-processing and normalization. Quality control plots of intensity values showed good consistency between BeadChips and individual arrays. Probe intensities were background corrected by subtracting the mean of the negative controls for each array. A variance-stabilising transformation was applied to the corrected intensities for each array [47]. Finally the log2-expression values were quantile normalized between arrays [48]. Raw and normalized expression data and associated information is MAIME compliant and has been submitted to the GEO repository as series GSE20241. (Manuscript reviewers can access the GEO series prior to public release using the following link: http://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?token=vdilhuomcocumvk&acc=GSE20241.) Probe annotation, including gene symbol and chromosome location was obtained from the University of Cambridge (www. compbio.group.cam.ac.uk/Resources/Annotation). Negative controls on each array showed that approximately 40% of probes were not expressed in each sample. Therefore probes with the lowest 40% of mean log2-expression levels across all arrays were filtered out. This left 27,994 probes for downstream analysis.
Statistical analysis used the limma software package [49]. A linear model was fitted to the normalized log2-expression values for each probe. Comparisons were made between Dnmt3L heterozygous and wild type spermatocytes, between Dnmt3L heterozygous and wild type spermatids, and between the two cell types for each genotype. The standard errors were smoothed across probes using an empirical Bayes method to obtain moderated t-statistics for each comparison and each probe [50]. The false discovery rate was controlled globally across all contrasts using the method by Benjamini and Hochberg [51]. As well as the genome-wide analysis, a separate analysis was conducted using only the 855 probes on the X chromosome.
These analyses yielded log2 fold changes, average log2expression, moderated t-statistics and false discovery rates for each probe. The comparisons were calibrated such that probes upregulated in the Dnmt3L heterozygous samples relative to the wild type cells have negative log-fold changes and down-regulated probes have positive log-fold changes. In this way, fold change can be interpreted as direct responses to Dnmt3L dose.

Quantitative Real-time PCR (QPCR)
For QPCR analyses, three preparations each of purified Dnmt3L heterozygous and wild type spermatocytes and spermatids were used. Each preparation was made up of cells isolated from 10 mice. SYBR Green primers were ordered from Sigma Aldrich and Taqman probes were ordered from Applied Biosystems. Primer sequences are listed in the Supplementary Information File S1.
QPCR was performed with samples from the heterozygous and wild type males with three replicates each in a 7900 HT Fast Real Time PCR System (Applied Biosystems). SYBR Green reactions were 10 ml reactions containing 1X SYBR Green PCR Master Mix (Applied Biosystems), primer mix, and cDNA. Conditions for QPCR with SYBR included an initial step of 10 minutes at 95uC, followed by 40 thermal cycles of 15 seconds at 95uC, 30 seconds at 60uC or 62uC, 30 seconds at 72uC. At the end, samples were subjected to a slow temperature-ramping dissociation stage of 15 seconds at 95uC, 15 seconds at 60uC, 15 seconds at 95uC to monitor the specificity of amplification.
Taqman reactions were 10 ml containing 1X Taqman Universal PCR Master Mix (Applied Biosystems), Taqman probe (which included forward and reverse primers) and cDNA. QPCR conditions for QPCR with Taqman included an initial step of 10 minutes at 95uC, followed by 40 thermal cycles of 15 seconds at 95uC, 30 seconds at 60uC.
The relative standard curve method was used to quantify gene expression levels for both SYBR Green and Taqman. Samples were normalized to b-actin values. Specificity was confirmed by running QPCR products on a 1% agarose gel after each QPCR experiment. All expression values were divided by the mean of the wild type samples for that gene. One-sided t-tests on the logexpression values were used to test for differential expression, using the direction of change observed from the microarrays as the alternative hypothesis.
For Western blot analyses, three samples each of Dnmt3L heterozygous and wild type sperm were used. Each sample was made up of sperm lysates isolated from 2 mice. The Kodak Imagestation 4000 MM Pro was used to quantitate the amount of DNMT3L and ACTIN within each sample. All quantity values were divided by the mean of the wild-type samples. Two-sided ttests on these values were used to test for statistical significance.

Supporting Information
File S1 (in file inventory) (DOC)