The Histone H3K79 Methyltransferase Dot1L Is Essential for Mammalian Development and Heterochromatin Structure

Generation of Dot1L Conditional and Null Alleles in Mice

To target the Dot1L gene, we constructed a targeting vector in which a 2.3-kb genomic region containing exons 5 and 6 and a promoterless β-geo selection cassette were flanked, respectively, by three loxP sites (Figure 1A). Exons 5 and 6 encode 108 amino acids that form several conserved motifs in the Dot1L catalytic domain, including the SAM-binding motif and motifs X, I, and II [6]. Since mutations of conserved residues within motif I abolish the methyltransferase activity of Dot1L [4], we predicted that deletion of exons 5 and 6 would inactivate Dot1L.

ES cells were transfected with the targeting vector and selected in G418-containing medium. Clones with homologous recombination were identified by Southern blot analysis with a 5′ external probe (Figure 1B). Three of these clones, referred to as Dot1L^3lox/+, were injected into blastocysts to generate chimeric mice, which subsequently transmitted the mutant allele to their offspring. Deletion of the β-geo cassette plus exons 5 and 6 was achieved by breeding the Dot1L^3lox allele into mice expressing Cre recombinase in the germ line. The resulting null allele is referred to as Dot1L^1lox (Figure 1A). Genotypes were determined using PCR (Figure 1C).

Dot1L Is Essential for Embryonic Development

We first determined the expression of Dot1L during embryonic development, taking advantage of the fact that cells containing the Dot1L^3lox allele express lacZ under the control of the endogenous Dot1L promoter. We conducted X-gal staining on Dot1L^3lox/+ heterozygous embryos and wild-type littermates at different stages of development. Dot1L expression is ubiquitous as early as 7.5-dpc (the earliest time point tested, Figure 2A). At 9.5-dpc, Dot1L expression remains ubiquitous and areas of elevated expression are apparent. Tissues that demonstrate high levels of lacZ staining include the optic vesicle, the first branchial arch, the limb buds, the heart, the otic pit, and the neural ectoderm (Figure 2B). Dot1L is also expressed at high levels in extra-embryonic tissues, including the visceral endoderm and visceral mesoderm of the yolk sac, and in primitive erythrocytes (Figure 2C). Similar lacZ staining patterns are observed in embryos harvested at 10.5-dpc, 11.5-dpc, and 12.5-dpc (data not shown), suggesting that Dot1L is broadly expressed during embryonic development.

Dot1L^1lox/+ mice were grossly normal and fertile. However, intercrosses of Dot1L^1lox/+ mice produced no viable Dot1L^1lox/1lox homozygous offspring, suggesting embryonic lethality (Table 1). Dot1L^1lox/1lox embryos harvested at 8.5-dpc were indistinguishable from wild-type and Dot1L^1lox/+ littermates (data not shown). At 9.5-dpc, Dot1L^1lox/1lox embryos were smaller than littermates, had enlarged hearts and stunted tails on gross observation when viewed under a dissecting microscope (Figure 2D, center). Approximately 15% of the Dot1L^1lox/1lox embryos demonstrated a severe phenotype, exhibiting developmental arrest at E8.5 (Figure 2D, right). Histological examination of 9.5-dpc Dot1L^1lox/1lox embryo sections revealed focal areas of extensive apoptosis, but no obvious structural defects (Figure S1). At 10.5-dpc, the percentage of viable Dot1L^1lox/1lox embryos was substantially below the expected Mendelian ratio (Table 1), suggesting that many of the Dot1L^1lox/1lox embryos die during this time interval. The few that survived to this stage exhibited developmental arrest at E9.5 and severe cardiac dilation (Figure 2E). No Dot1L^1lox/1lox embryos survived beyond 10.5-dpc (Table 1).

As stunted growth and enlarged heart are phenotypes that often occur as a result of defects in extraembryonic tissues, we examined the yolk sac and placenta of 9.5-dpc Dot1L^1lox/1lox embryos. While the placenta showed no obvious defects, the yolk sac exhibited abnormal vascular morphology. The yolk sac vasculature was present and contained primitive erythrocytes, but was frequently underdeveloped and disorganized when compared to control littermates (Figure 2F). These observations indicate that, in the absence of Dot1L, vasculogenesis took place in the yolk sac but angiogenesis was defective.

Dot1L Deficiency in ES Cells Results in Growth Defects

To investigate the cellular function of Dot1L, we derived Dot1L mutant ES cells from blastocysts produced from intercrosses of Dot1L^1lox/+ mice. Two Dot1L^1lox/1lox and multiple Dot1L^1lox/+ and Dot1L^+/+ lines were established. As expected, H3K79 di- and tri-methylation was greatly reduced in Dot1L^1lox/1lox cells compared to Dot1L^+/+ cells (Figure 3A). Dot1L^1lox/+ cells had intermediate levels of H3K79 di- and tri-methylation, indicating haploinsufficiency of Dot1L (Figure 3A). Surprisingly, Western blot analysis using a “mono methyl H3K79” antibody (ab2886, Abcam) detected no change in signal intensity in Dot1L mutant ES cell lines (data not shown). To verify the results, we carried out mass spectrometry. In wild-type ES cells, ∼11% of histone H3 showed K79 methylation, among which mono-, di-, and tri-methylation accounted for ∼70%, ∼30%, and less than 1%, respectively. In Dot1L^1lox/1lox ES cells, H3K79 mono- and tri-methylation was absent although trace amount of di-methylation was detected (Figure 3B and Figure S2). We therefore concluded that the Western blot result showing no alteration in H3K79 mono-methylation in the absence of Dot1L was an artifact due to nonspecific recognition of histone H3 by the “mono methyl H3K79” antibody. The low level of H3K79 di-methylation detected in Dot1L^1lox/1lox samples could be from feeder cells present in the culture or due to incomplete inactivation of Dot1L. Taken together, these results indicated that Dot1L is most likely the sole H3K79 methyltransferase in mice.

Dot1L mutant ES cells maintained an undifferentiated state, as judged by morphology and ES cell markers such as Oct4 and Nanog (data not shown). We investigated whether Dot1L deficiency affects ES cell growth. We plated 3×10⁵ Dot1L^+/+, Dot1L^1lox/+, and Dot1L^1lox/1lox ES cells in standard ES cell medium and monitored proliferation. By 24 hours, the number of Dot1L^+/+ ES cells (7.3×10⁵) was significantly higher than the number of Dot1L^1lox/+ or Dot1L^1lox/1lox cells (4.2×10⁵ and 3.6×10⁵ respectively, P<0.05, Figure 3C). Over the 5 days examined, Dot1L^+/+, Dot1L^1lox/+, and Dot1L^1lox/1lox ES cells had average doubling times of 16, 22, and 26 hours, respectively. The fact that both Dot1L^1lox/+ and Dot1L^1lox/1lox cells exhibited growth defects showed the importance of Dot1L gene dosage in cellular function. The reduction of H3K79 methylation in Dot1L^1lox/+ cells (haploinsufficiency) suggested that Dot1L level is relatively low. Interestingly, Dot1L^1lox/+ mice were apparently normal despite the defects in Dot1L^1lox/+ ES cells. It is possible that 50% of Dot1L can barely maintain normal cellular function under ideal conditions (e.g. in vivo) but is not sufficient to do so under suboptimal conditions (e.g. in culture).

We next examined apoptosis and cell cycle status of the Dot1L mutant ES cells. Annexin V staining revealed that 4.0% of the Dot1^1lox/1lox ES cells and 3.1% of the Dot1L^1lox/+ ES cells were annexin V positive, while only 1.2% of the Dot1L^+/+ ES cells were annexin V positive (Figure 3D). This indicates that more than twice as many of the Dot1L mutant ES cells were undergoing apoptosis compared to wild-type ES cells. Furthermore, cell cycle analysis by propidium iodide staining revealed an elevated percentage of G2 cells and a reduced percentage of G1 cells among the Dot1L mutant ES cells when compared to the wild-type ES cells (Figure 3E). These results suggest that both elevated apoptosis and G2 cell cycle arrest contribute to the reduced growth rate of Dot1L mutant ES cells.

Dot1L-Deficient ES Cells Show Telomere Elongation and Aneuploidy

Dot1-deficient S. cerevisiae show telomere elongation and defects in telomere silencing [3]. We therefore evaluated the effect of Dot1L inactivation on telomere length. First, we used Southern blot terminal restriction fragment (TRF) analysis to estimate telomere length in two ES cell lines of each Dot1L genotype. Both Dot1L^1lox/1lox lines and one of the Dot1L^1lox/+ lines showed telomere elongation, as evidenced by the presence of high molecular weight (MW) TRFs and the increase in the lengths of bulk TRFs compared to wild-type controls (Figure 4A). Next, we carried out quantitative fluorescence in situ hybridization (Q-FISH) using a telomere-specific probe to determine the mean telomere length (mtl) and the distribution of telomere lengths for each cell line (Figure 4, B and C). Consistent with the TRF results, both Dot1L^1lox/1lox lines had higher mtl, greater percentages of elongated (>100 kb) telomeres, and reduced percentages of short (<50 kb) telomeres compared to Dot1L^+/+ lines (Figure 4C). The heterozygous ES cells again showed an intermediate phenotype (Figure 4C).

Examination of the Q-FISH samples revealed frequent aneuploidy in Dot1L-deficient cells (Figure 4B). To further investigate this phenotype, we prepared metaphase chromosome spreads from Dot1L^1lox/1lox and Dot1L^+/+ ES cells and examined them for chromosomal defects. Dot1L^+/+ cells were karyotypically stable, as the vast majority had 40 chromosomes. In contrast, over 40% of metaphase Dot1L^1lox/1lox cells were aneuploid. Most of the aneuploid cells showed gain of chromosomes and some ended up being tetraploid (Figure 4, B and D). Aside from aneuploidy, no obvious chromosomal abnormalities were frequently observed in Dot1L-deficient cells (Figure 4B). These results point to defects in chromosome segregation in the absence of Dot1L.

Aberrant Telomere Elongation in Dot1L-Deficient Cells Correlates with Activation of the ALT Pathway

Our data suggested a role for Dot1L in the homeostasis of telomere length. Two main mechanisms have been described for the maintenance of mammalian telomeres: the addition of telomeric repeats by telomerase and the so-called alternative lengthening of telomere (ALT) mechanism that relies on homologous recombination between telomeric sequences [15],[16]. Dot1L mutant ES cells showed increased telomere heterogeneity (Figure 4), which is a hallmark of ALT cells [17]. To determine whether the ALT pathway is activated in Dot1L-deficient cells, we assessed the presence of ALT-associated PML bodies (APBs, colocalization of PML and telomeres), another hallmark of ALT [17]. Dot1L^+/+, Dot1L^1lox/1lox, as well as Dnmt3a−/−3b−/− (positive control) ES cells were immunostained with antibodies against TRF1 (a telomere-binding protein) and PML. In the absence of Dot1L, both the frequency of cells showing APBs and the number of APBs per cell were significantly increased compared to wild-type cells (Figure 5, A–C, χ² tests, P<0.001), suggesting that aberrant elongation of telomeres in Dot1L-deficient cells was due, at least in part, to activation of the ALT pathway.

Dot1L Deficiency Results in Loss of Heterochromatin Marks at Telomeres and Centromeres

Aneuploidy and telomere elongation can result from defects in the chromatin structure at centromeres and telomeres, respectively [18]–[23]. To evaluate changes in chromatin structure in Dot1L mutant cells, we used chromatin immunoprecipitation (ChIP) to examine histone modifications at major satellite repeats (present at pericentric regions), minor satellite repeats (present at centromeric regions), telomeric repeats, and subtelomeric regions (Figure 6). H3K79 di-methylation was detected in all these heterochromatin regions in Dot1L^+/+ cells (Figure 6, A and B). As expected, this modification was reduced in Dot1L^1lox/+ cells and almost absent in Dot1L^1lox/1lox cells (Figure 6, A and B), validating our experimental procedures. As further controls, the levels of centromere- and telomere-bound histone H3 were similar in wild-type and Dot1L mutant cells (Figure 6, A and B), and the telomere-binding protein TRF1 associated with telomeric repeats, but not with major satellite repeats (Figure 6B).

In Dot1L^1lox/1lox cells, H4K20 tri-methylation, a hallmark of constitutive heterochromatin, was greatly reduced at minor satellite repeats and sub-telomere regions, and moderately reduced at major satellite repeats (Figure 6A). Consistent with this observation, immunofluorescence analysis revealed the loss of enrichment of H4K20 tri-methylation at pericentric heterochromatin in the absence of Dot1L (Figure S3). H3K9 di-methylation, but not H3K9 tri-methylation, was reduced in all regions examined in Dot1L^1lox/1lox cells (Figure 6, A and B). Concomitantly, H3K9 mono-methylation showed marked increases at major satellite repeats and minor satellite repeats (Figure 6A), and H3K9 acetylation, a mark of euchromatin, was elevated in all regions examined (Figure 6, A and B). These changes appeared to be heterochromatin-specific, as all histone modifications examined, except H3K79 methylation, showed no global changes in Dot1L mutant cells (Figure S4). Dot1L deficiency did not cause alterations in DNA methylation at major satellite repeats and minor satellite repeats as well as other genomic regions such as the intracisternal A-particle (IAP) retroviral elements (Figure S5). Altogether, these results suggest that loss of H3K79 methylation results in a less compacted (or more open) chromatin state at centromeres and telomeres.

Construction of the Gene Targeting Vector

The Dot1L conditional targeting vector, in which a 2.3-kb genomic region containing exons 5 and 6 was flanked by loxP sites, was constructed by sequentially subcloning Dot1L genomic fragments and a floxed βGeo cassette into pBluescript SK (Stratagene). The Dot1L genomic fragments were generated by PCR using mouse genomic DNA as the template. The primer pairs used were: 5′-TTC ACT AGT CCC CAC CTT TGG ATT G-3′ and 5′-GGC ACT AGT GTC ACA CAC CTT TA-3′ for the 5′ arm, 5′-CAT GTC GAC ACC GTG TAG TCC TGG TGG GA-3′ and 5′-CTC GGC CGG CCT TGC CTG TGG CTG ACG-3′ for the 3′ arm, and 5′-GAC ACC GGT GCC TGG CAA CCT TTT GG-3′ and 5′-CTG GGC GCG CCA CCA GGA ACA CAC AGG TAC-3′ for the floxed region (underlined are the restriction sites used for cloning). The identity of the vector was verified by DNA sequencing.

Generation of Dot1L Mutant Mice

The Dot1L conditional targeting vector was transfected into ES cells via electroporation, and transfected cells were selected with G418. Clones with homologous recombination (Dot1L^3lox/+) were identified using Southern blot. Genomic DNA was digested with EcoRI and hybridized with a 5′ external probe (The probe was generated by PCR using the following primers: 5′-CTC TGG TAC CTT TGT TGT TAT ACA G-3′ and 5′-CTC TCA AGT CGA CTG TAA GAT GAA G-3′). Multiple Dot1L^3lox/+ clones were used to generate chimeric mice and F1 heterozygotes. Deletion of exons 5 and 6 as well as the βGeo cassette was achieved by crossing Dot1L^3lox/+ mice with Zp3-Cre transgenic mice, which express the Cre recombinase in the germline. Mutant mice were maintained on a C57BL/6 inbred or a C57BL/6-129Sv hybrid background. Primers used for PCR genotyping were: DF1: 5′-GGA ACT CAA GCT ATA GAC AG-3′, DR1: 5′-CAC TGC CCA GGT CGA CAA ACA G-3′, and DR2: 5′-ATC CTC TCT CCT GAG GAG GCA GC-3′ (Figure 1).

Embryo Collection, X-Gal Staining, and Histology

Female mice in Dot1L^1lox/+ intercrosses were examined for plug formation to establish the timing of copulation. Deciduas were isolated from euthanized females at various time points following copulation, and embryos were examined under a dissecting microscope. DNA from the yolk sac was used for genotyping by PCR using the primers described above. X-gal staining was performed on 7.5- to 12.5-dpc Dot1L^3lox/+ embryos and littermates as previously described [30]. Embryo, yolk sac, and placental tissue specimens, which were harvested at 9.5-dpc, were fixed in Bouin's solution, washed extensively in 70% ethanol, processed routinely for paraffin embedding, sectioned at 5 µm, stained with hematoxylin and eosin, and then evaluated by bright field microscopy.

ES Cell Derivation and Culture

Dot1L mutant ES cells were derived from blastocysts produced from intercrosses of Dot1L^1lox/+ mice, as previously described [31]. Established ES lines were maintained in ES cell medium [32]. Apoptosis was analyzed using an Annexin V-PE apoptosis detection kit (BD Pharmingen). Cell cycle analysis was done using a PI/RNase Staining Buffer (BD Pharmingen).

Immunofluorescence and Immunoblot Analyses

Immunoblot and indirect immunofluorescence analyses were carried out using standard procedures. The following antibodies were used: anti-H3K79Me1 (Abcam), anti-H3K79Me2 (Abcam), anti-H3K79Me3 (Abcam), anti-H3 (Millipore), anti-H3K4Me2 (Millipore), anti-H3K4Me3 (Millipore), anti-H3K9Me1 (Millipore), anti-H3K9Me2 (Millipore), anti-H3K9Me3 (Millipore), anti-H3K27Me1 (Millipore), anti-H3K27Me3 (Millipore), anti-H3K9Ac (Millipore), anti-H4K20Me3 (Millipore), anti-H4Ac (Millipore), anti-Suv39h1 (Upstate), anti-ESET (Upstate), anti-G9a (Cell Signaling), anti-TRF1 (Abcam), anti-PML (Chemicon), Alexa 488-conjugated goat anti-rabbit IgG (Molecular probes), Alexa 555-conjugated goat anti-mouse IgG (Molecular Probes), and peroxidase-conjugated goat anti-rabbit and goat anti-mouse IgG (Jackson ImmunoResearch Laboratories).

Mass Spectrometry Analysis

Histone H3 purified from ES cells was digested with trypsin, and the resulting peptides were analyzed using a LTQ-FT mass spectrometer (Thermo Fisher Scientific Inc.) hyphenated with an Agilent 1200 HPLC system (Agilent). Identification of the peptides was performed by searching the MS/MS fragmentation data against the histone H3 sequence using MASCOT search software (Matrix Science, version 2.1). All identifications were manually inspected for correctness. The abundance of each identified and validated peptide was calculated from its peak intensity using extracted ion chromatogram (XIC) of LC/MS spectra. Relative quantification of different forms of H3K79 methylation was performed by comparing the signal intensities of the tryptic peptide EIAQDFKTDLR at m/z 668.35 ([MH₂]²⁺), EIAQDFKmeTDLR at m/z 675.36 ([MH₂]²⁺), EIAQDFK2meTDLR at m/z 682.35 ([MH₂]²⁺), and EIAQDFK3meTDLR at m/z 689.35 ([MH₂]²⁺).

Telomere Length Analysis

To analyze telomere length, we performed Q-FISH and TRF analyses according to procedures described previously [22].

Metaphase Spread Analysis

To prepare metaphase spreads, cells were incubated with 0.1 µg/ml of colcemid for 4 hours and then harvested and resuspended in 200 µl PBS. 10 ml of 75 mM KCl solution was added dropwise with constant gentle agitation. Cells were fixed by slow addition of 3∶1 methanol/acetic acid solution, and then dropped onto a microscope slide. Slides were washed in 70% acetic acid, stained with DAPI and mounted. Chromosome spreads were observed using a Zeiss fluorescence microscope.

Chromatin Immunoprecipitation Assays

ChIP was performed using 20×10⁶ ES cells as described in the online protocol provided by Upstate. Antibody sources are described above. Purified DNA was either analyzed with quantitative real-time PCR (qPCR) using Applied Biosystems SYBR PCR mastermix or used in a dot blot assay as described [22]. qPCR primers used were specific for major satellite repeats, minor satellite repeats [33], and subtelomeric regions of chromosome 1 (forward: 5′-TTA GGA CTT CTG GCT TCG GTA G-3′, reverse: 5′-AGC TGT GGC AGG CAT CGT GGC-3′) and chromosome 2 (forward: 5′-GAA TCC TCC CTG TAG CAG GG-3′, reverse: 5′-GTA CAT AAC CGA TCC AGG TGT G-3′). Relative enrichment was calculated as 2ˆ (C_T (control CHIP) - C_T (experimental CHIP)), where C_T is equal to the C_T (immunoprecipitated sample) - C_T (input) and normalized so that the wild-type value was 1, with the exception of H3K9Me1 at major satellite repeats where the Dot1L^1lox/1lox value was 1. Each sample used in the dot blot contained DNA precipitated from 2.5×10⁶ cells. Probes used were ³²P-labelled oligonucleotides specific for telomeric repeats ((TTAGGG)_x11) and major satellite repeats (5′-TAT GGC GAG GAA AAC TGA AAA AGG TGG AAA ATT TAG AAA TGT CCA CTG TAG GAC GTG GAA TAT GGC AAG-3′), respectively.

DNA Methylation Assay

Genomic DNA isolated from ES cells was digested with methylation-sensitive restriction enzymes and analyzed by Southern hybridization using probes specific for the major satellite repeats, the minor satellite repeats, and the intracisternal A particle retrovirus [32],[34].