Mammalian DNA methylation generally shows limited global dynamics except during pre-implantation and primordial germ cell development. The observed global demethylation of the paternal genome upon fertilization is mediated by the oocyte and is critical for establishment of totipotency and developmental competence1
. An enucleated oocyte can reprogram the epigenetic and transcriptional identity of somatic cells through a procedure known as somatic cell nuclear transfer (SCNT), though successful reprogramming occurs at low efficiency and is likely affected in part by the retention of somatically-conferred epigenetic lesions2,3
. To date, the oocyte’s intrinsic capacity to erase somatic DNA methylation patterns to a true zygotic signature remains incompletely characterized. Here, we generated genome-scale single basepair resolution maps of DNA methylation from donor fibroblasts and SCNT reconstructed mouse embryos using reduced representation bisulfite sequencing (RRBS) 4
(Supplementary Fig. 1a
). We conducted two independent rounds of nuclear transfer into enucleated BDF1 (C57/DBA F1) oocytes, each consisting of two biological replicates. We used both BDF1 x Cast hybrid and 129X1 inbred tail tip fibroblasts (referred to as Cast and X1, respectively) to serve as controls for low input RRBS by capturing over 10,000 hybrid SNPs and to ensure that dynamics observed were consistent across strain identity (Supplementary Fig. 1b–e
). Using our stringently collected samples and the first genome-scale measurement in any nuclear transfer experiment, we detected a low level (~15%) of the host oocyte genome (see Supplementary Methods
). This affected ~35% of loci in the X1 samples and likely ~13% of loci in the embryos using the hybrid donors. Given the complexity inherent to the protocol and the number of cells required, the presence of residual host DNA may be unavoidable; however, the low frequency with which it is sampled is unlikely to consistently capture the same locus across experiments. Accordingly, we present the SCNT embryo data without compensation for residual host DNA methylation.
We compared methylation profiles between the donor cells and reconstructed embryos to the methylation dynamics observed during fertilization 5
. DNA methylation patterns of donor fibroblasts and sperm exhibit a conventional somatic bimodality that depends upon relative CpG density. After nuclear transfer, a shift in the fibroblast methylation landscape resembles the demethylation that occurs within the paternal genome upon fertilization (). Though many regions are affected in both processes, demethylation occurs at a smaller magnitude after SCNT (). Globally, reconstructed embryos more closely resemble donor fibroblasts than the paternal genome after fertilization or the pre-implantation embryo (Supplementary Fig. 1f
). Importantly, SCNT embryos are more similar to each other regardless of experimental round or donor strain than they are to either fibroblasts or the early embryo, suggesting the majority of methylation changes conferred by SCNT are consistent across experiments (Supplementary Fig. 1g
). The difference in reprogramming response appears to be partly due to genomic context. For instance, we find that methylation levels of different repetitive element classes change by different magnitudes (, top). While SINE elements appear similarly demethylated in both processes, methylation levels at LINEs and LTRs only slightly decrease or do not change after nuclear transfer (, bottom, Supplementary Fig. 2a
). Upon closer inspection of LINE families, we found that methylation at L1Md_A elements remains almost completely static, whereas the evolutionarily younger L1Md_T/Gf families appear slightly more dynamic6,7
(). The predominance of LINE and LTR element classes in the genome and their recalcitrance to demethylation could explain the striking retention of somatic methylation patterns after SCNT. Taken together, we conclude that repetitive elements, which account for a large proportion of demethylation events during fertilization, appear more resistant to change when the ooplasm is confronted with a somatic nucleus. By the nature of the experiment we cannot rule out the possibility that some of the global dynamics observed in SCNT embryos may be due to the presence of residual host oocyte DNA. However, we observed similar demethylation at CpGs associated with Cast alleles compared to C57 in our hybrid fibroblast experiments, which can only result from reprogramming of the donor fibroblast genome (Supplementary Fig. 2b
). It is technically not possible to extend this SNP analysis to repetitive elements, but the reasonable association between Cast exclusive demethylation dynamics and the C57 haplotype argue that demethylation is in fact an observable and measurable event during nuclear transfer in our data.
Classifying common and distinct DNA methylation dynamics during fertilization and nuclear transfer
We next compared methylation dynamics in promoters during both processes. Given the likely stochastic models for demethylation during SCNT and the potential effect of residual host DNA, we applied stringent criteria for identifying changing promoters such that hypermethylated promoters (>0.5) were required to (1) change by > 0.2 in at least three replicates, (2) change by >0.2 in donor-normalized X1 replicates, and (3) could not have contradictory dynamics in Cast-tracked CpGs (see Supplementary Methods
). This strategy identifies the most consistently changing promoters during the nuclear transfer procedure but excludes promoters that change either less efficiently or are targeted less frequently for DNA demethylation. We then classified promoter dynamics as being either SCNT specific, shared with fertilization, or unique to fertilization (, Supplementary Table 1
). We identified a stringent set of 15 SCNT specific promoters, which include several genes that function during meiosis and are typically hypomethylated in both gametes and the early embryo (). The specific demethylation observed in the SCNT embryos suggests the presence of defined, targeting factors within the ooplasm that ensure the unmethylated status of these sites () 8
. We observed that approximately two thirds of fertilization specific demethylated targets retained considerable DNA methylation after nuclear transfer, indicating that only regions in certain contexts are equivalently demethylated in both processes (Supplementary Fig. 2c,d
). These likely represent somatically retained epigenetic information that could affect embryonic development. We also investigated 102 previously identified differentially methylated promoters that are hypermethylated in the oocyte and transiently remain methylated on the maternal allele during pre-implantation (Supplementary Fig. 2e
. The set is unmethylated in fibroblasts and remains so after nuclear transfer, representing another example of the inequivalence between SCNT reconstructed embryos and true embryos in which maternally encoded regulatory information is not conferred in SCNT.
Promoter dynamics during SCNT include demethylation of gamete-specific genes
The demethylation events observed after nuclear transfer are similar to those at fertilization, though often at lower magnitudes, which may indicate less efficient or stochastic targeting. In turn, this may affect the number of SCNT embryos that successfully reactivate sufficient developmental loci upon zygotic activation9
. Certain repeats, such as LTRs and L1Md_As, appear completely refractory to demethylation in nuclear transfer and may be protected within the epigenetic context of somatic cells. Demethylation of the paternal genome at fertilization is accompanied by global rechromatinization1
, which may provide a unique window of opportunity for parasitic genomic elements to either initiate demethylation or escape methylation machinery. It has been observed, for instance, that cytoplasmic injection of chromatinized round spermatids generates live embryos at equivalent rates to injection of terminally differentiated, protamine-compacted spermatozoa but does not co-occur with an apparently equivalent global erasure of DNA methylation10
. Our genome-scale profiling strategy confirms previous locus specific bisulfite sequencing and global immunostaining data which have shown that DNA demethylation after nuclear transfer is not as dramatic as that observed in the paternal genome7
. Presumably, the interplay between histone exchange, transcription, and active demethylation may limit the magnitude and targets for DNA demethylation observed after nuclear transfer or lead to aberrant signatures11
. Tet3 mediated hydroxymethylation may also be more robustly targeted upon fertilization than after nuclear transfer, where it is not specifically recruited to a single pronucleus but rather distributed at restricted levels across the entire diploid genome12,13
. It is interesting that demethylated loci after nuclear transfer resemble the genomic features and promoter classes enriched for hydroxymethylation within mouse embryonic stem cells, namely repetitive LINE families and germ-line gene promoters, which suggests that at least a portion of the demethylation events involved with induced pluripotency may be specifically targeted14,15
. Technical improvements to other genomic profiling strategies should soon identify the targets and dynamics of histone deposition, as well as the distribution of hydroxymethylcytosine15
, and should comprehensively define the full reprogramming capacity of the mammalian ooplasm.