As detailed previously, epigenetic processes are fundamental for the differentiation and development of multicellular organisms by controlling chromatin accessibility and transcription38. Although phenotypic variation between individuals is mainly driven by genetic traits, there is also evidence that epigenetic mechanisms may contribute to inter-individual phenotypic differences in mammals89. Examples for epigenetic differences between individuals are comparatively rare and mostly, but not exclusively, confined to the level of DNA methylation, which represents a relatively stable epigenetic modification that can often be measured in an allelic context. There is well-documented evidence that epigenetic states can be inherited across generations. For example, agouti viable yellow (Avy) mice display inheritance of yellow fur as a result of incomplete erasure of the methylation signal associated with a retrotransposon insertion96,97, and kinked-tail mice
transmit phenotype through multiple generations due to the loss of the silent epigenetic state at the Axin gene196 as well as a heritable white-tail phenotype associated with Kit-specific microRNAs197. It was also reported, that allelic variation at certain epigenetic modifier genes in mice, like DNA methyl-transferases or chromatin remodeling factors may influence the inheritance of CpG methylation patterns in trans98,99. Differences at the level of DNA methylation have also been demonstrated in supposedly genetically identical, monozygotic twins that are acquired during the lifetime of each individual through largely unknown mechanisms90,91.
In addition to trans-regulation, there are several reports demonstrating that DNA sequence variants associate with specific epigenetic states. Differential methylation states of individual alleles were first described for the in c-Ha-ras-1 gene in human cells198,199 More recently, a study by Murrell for example showed that loss of imprinting of the IGF2 gene, which is frequently observed in Beckwith-Wiedemann syndrome (BWS), correlated with loss of maternal allele-specific methylation (LOM) for certain haplotypes94. This observation was supported by another study, showing that variation in DNA methylation of the IGF2/H19 locus is amongst others mainly determined by single nucleotide polymorphisms (SNPs) in cis93. In addition, Flanagan demonstrated allele-specific methylation patterns in the CDH13 gene92, and Yamada reported a CGI, which is methylated in an allele-specific but parental-origin-independent (non-imprinted) manner200. Along this line, a recent study in humans identified several cases of allele-specific DNA-methylation at non-imprinted gene loci95, where the methylation status of each allele was likely controlled in cis by the local DNA sequence.
Taken together, the published data suggests that three types of inheritance of DNA methylation patters may exist in vivo: methylation patterns at non-imprinted loci may be inherited across generations based on genetic mechanisms (in cis and in trans) or based on epigenetic mechanisms (e.g. inheritance of DNA methylation, or other epigenetic marks).
As part of this thesis a broad analysis of differential DNA methylation in two inbred mouse strains was performed utilizing a modified protocol of DNA methylation-dependent genome fractionation (MCIp) in combination with locus-wide tiling arrays. Stringent hybridization conditions for microarray analyses of CpG-methylated and unmethylated genome fractions allowed the simultaneous detection of several hundred differentially
differences). The results suggest that the chosen approach is highly reproducible and correctly identified both DMRs and DNA sequence variations.
Why were inbred mice chosen as model organisms to investigate allele-specific DNA methylation? The house mouse is one of the most successful mammals and the premier research animal in mammalian biology. A strain is defined as inbred when it has been mated brother versus sister for 20 or more consecutive generations. Such inbred strains are therefore isogenic (genetically identical) allowing offspring to posses both genetic and phenotypic uniformity. Each inbred strain has a unique genotype and consequently a unique phenotype. The normal pigmented C57BL strain was originally derived by Little in 1921, whereas the strain BALB/c, which is albino and small in size, originated in 1923 by McDowell. C57BL is probably the most widely used of all inbred strains. Its sub strain C57BL/6, which developed before 1937, alone accounts for more than 14% of occasions on which an inbred strain is used.
Concerning immunology various differences between these two strains are known. The C57BL/6-BALB/c model has been intensively used to study the immunopathogenesis of several intracellular infections such as listeriosis201, leihsmaniasis202,203, yersiniosis204,205 and mycobacterial infections206,207. For example, the BALB/c strain is highly susceptible to the infection of Leishmania major, in contrast to resistant C57BL/6208,209. Susceptible mice as BALB/c fail to control parasite proliferation and develop progressive lesions and systemic disease, which is associated with Th2 response manifested by high levels of 4, 13, 10 and antibody production. In contrast, resistant C57BL/6 mice develop IL-12-driven, interferon-gamma (INF-gamma) dominated Th1 response that promotes healing and parasite clearance210,211. It is likely that epigenetical differences are found in immune-cells derived from the two inbred stains especially in regions containing genes which are differentially expressed without or upon stimulation. Macrophages are widely distributed immune system cells and beyond that involved in all stages of the immune response. They were chosen because they actually contribute to the clearance of an intracellular pathogen. The leishmanial parasites for example live inside the macrophage, but the macrophage is also the host‟s major effector cell in the elimination of this pathogen212. Macrophages can be phenotypically polarized by the local cytokine pattern to trigger specific functional programs. Polarized macrophages are broadly classified in two main groups: classically activated macrophages (or M1), and alternatively activated macrophages (or M2). M1 exhibit potent microbicidal properties and promote strong Th1 responses, whilst M2 support Th2-associated effector functions. An activation of inappropriate effector mechanisms can lead to enhanced pathogen growth, rather than effective pathogen clearance. For example, classically activated macrophages protect
from L. major (Th1 response) in C57BL/6 mice, because they contribute to the clearance of the bacterium. On the contrary, in susceptible BALB/c mice, macrophages with a wound-healing phenotype (Th2 response) allow for its persistence210,211. It is also important, that macrophages can be grown in vitro, without possible environmental impacts.
Both mice strains used in this study (C57BL/6 and BALB/c) belong to the most widely studied inbred mice. Especially the underlying genetic background of the C57BL/6 strain has been sequenced in detail and serves as the reference sequence for murine genetic research. It was conceivable that differences between inbred mice strains are present. In particular, since it was proposed that the genomes of inbred strains are a mosaic of regions of different sub specific origins. The fine structure of such mosaic variation was described by Wade, reporting extremely high variation spanning one-third of the genome and low variation in the remaining part213. This bimodal distribution was thought to represent regions having different, or the same, sub specific origins, respectively.
Following studies raised questions about the haplotype structure214,215, the effect of ascertainment biases in subspecific assignment216,217 and the contributions of intersubspecific versus intrasubspecific variation218. In addition, the presence of substantial intrasubspecific variation, ancestral polymorphisms and secondary introgression after the divergence of the subspecies was reported218,219. Frazer and colleagues recently identified 8.3 million SNPs by resequencing of 15 different mouse strains indicating that modern inbred strains are composed of at least four mus musculus substrains (domesticus, molossinus, musculus and castaneus)151. In contrast to the mosaic model, another recent work found that most of the genome has intermediate levels of variation of intrasubspecific origin152. However the true extent of genetic differences between these two specific mice strains still remains elusive.
In order to account for both, the reported genetic differences between strains and the unbalanced hybridization behavior seen in our experiments, we applied the virtual CGH analysis. The amount of microarray probes showing preferential hybridization only with the reference strain (C57BL/6) was unexpectedly high (>15%) in the present experiments.
Although the analysis focused on regions where differential gene expression was observed, a huge number of array probes were affected. Subchromosomal alterations
insertions and deletions within one strain and single nucleotide polymorphisms (SNPs).
The accumulation of these sequence variations correlated well with probes exhibiting unequal hybridization behavior. Since probes were designed according to the genomic sequence of the reference strain C57BL/6, the affected ones were only conspicuous for BALB/c. Most previously known SNPs, as identified by Frazer151, correlated with affected probes. In addition, a large number of previously unknown SNPs were found in exemplary sequenced regions showing unequal hybridization behavior. The genomic sequencing data yielded also that the position of a number of repetitive elements was altered in BALB/c as compared to C57BL/6. However, it is unclear if the BALB/c strain contains more or less repetitive elements in general.
Taken together, it is therefore conceivable that the genetic backgrounds of these two mouse strains vary to a greater extent than previously known and expected. These variations may in part contribute to the differential immune answer in both strains.
After correction by vCGH our experiments defined a large number of allele-specific methylation events. In the following a representative set of these genomic regions were traced in the germ line of parental animals. Interestingly, most DMRs that were studied in detail, were only differentially methylated within the somatic tissues studied (BMM and spleen) and showed equal levels of methylation in the male germ line, suggesting that the observed differences are established post-fertilization and DNA methylation itself is not inherited across generations.
Furthermore, the exemplary mass spectrometry validation of DMRs and allele specific examination of DNA methylation in F1 Hybrids using traditional sequencing of bisulfite treated genomic DNA suggests, that the propensity for acquiring DNA methylation during development mainly depended on the local sequence context in cis. It is formally possible that another, yet uncharacterized epigenetic mark at these sites determines the allelic methylation status in a heritable fashion. However, since a common association of DMRs with sequence variations was found, it is much more likely that the establishment of differential methylation patterns during embryonic development may involve cis-acting regulators, including specific DNA-binding proteins or other sequence-dependent regulatory mechanisms. Several publications accentuate the contribution of cis-acting regulators as observed in this thesis. The findings of this work are in line with a number of previous studies linking allelic DNA-methylation and genetic variation which were already discussed in the beginning of this chapter92-95,200 and share an obvious corollary with the recent observation that the genetic sequence is largely responsible for directing species-specific transcription in mice carrying the human chromosome 21220. Moreover, investigating the same human chromosome, Zhang and colleagues very
recently indentified three cases of allele-specific DNA methylation showing strong sequence dependence similar to our results221. However, our analysis contrasts a lately published study that did not find significant variation in DNA methylation between inbred or outbred mice and concluded that the impact of DNA polymorphisms on DNA methylation would not seem to be common91. The reasons for the deviant findings are not clear but may relate to the different technologies and readout systems used in each study.
For example, the above study used a CpG island microarray covering 2176 unique CpG island regions of the mouse genome that are likely more conserved, frequently contain gene promoters and are often unmethylated. Based on our findings that differential DNA methylation is associated with genetic variation and promoter-distal sites, we would predict that the microarray platform utilized by Kaminski et al. is less likely to detect allele-specific DNA methylation than the considerably larger and locus-wide microarray used in our study.
It is conceivable that trans-acting mechanisms contribute to the epigenetic status of a specific region as well. It has recently been published that the absence or functional inactivation of the few known epigenetic modifiers that mediate DNA methylation in trans show profound developmental abnormalities222. However, our data suggests, that only a relatively small proportion of strain-specific DMRs (like the ones identified in Isoc2b, Sfi1 pseudogene, or Eps8l1) depend dominantly on trans-acting mechanisms.
Taken together, the presented findings suggest that genetic variability has a major impact on epigenetic variability, likely due to its influence on the propensity of individual alleles to become methylated (or demethylated) during development. This relationship is likely not restricted to early embryonic development in mice, but may also apply to other developmental (including disease- and age-related) processes in mammals. For example Zhang et al. demonstrated allele-specific DNA methylation in human samples as well221. In essence, the results imply that any given DNA sequence (at a non-imprinted locus) may carry intrinsic information that recruits cis-acting factors and determines its DNA methylation status during mammalian development.
What is the relevance of the present findings? The systematic screening of the human genome for genetic variants that affect gene regulation was thought to advance the fundamental understanding of phenotypic diversity and lead to the identification of alleles that modify disease risk. It becomes more and more obvious, that the allele-specific
(chromatin immunoprecipitation)223. These allelic differences in chromatin may provide an explanation for the observed allelic variation in gene expression224. Evidence for the medical importance of cis-acting polymorphisms has been provided by positional cloning of susceptibility genes for common diseases such as stroke225 and type 2 diabetes226. The polymorphisms in these examples were not associated with protein coding or splice-site variations. However, diverse cis-acting mechanisms might also be important for malignant diseases. Successful epigenetic control of gene expression is necessary and aberrant DNA methylation, which has been recognized as an important feature in cancer formation, leads to the altered gene expression seen in almost all human cancers. Originally only genome-wide DNA hypomethylation was linked with cancer and was the focus of tumor research37,227,228. But during the mid 1980s regional hypermethylation of specific cancer-associated gene promoter sites attracted greater attention and the focus relocated to the participation of hypermethylation to cancer formation229-231. The present study strongly suggests, that individual alleles can exhibit different probabilities to become methylated (regulated in cis). Interestingly, a recent study investigated the regulation of allele-specific expression (ASE) by cis-acting elements in acute lymphoblastic leukemia (ALL)232. They found that specific CpG site methylation is one of the factors that regulate allele-specific gene expression in ALL cells. Besides, an independent study, done by our lab, recently obtained evidence of allele-specific methylation of certain CGIs in acute myeloid leukemia (AML) samples (unpublished data). However, further studies are certainly required, in order to shed more light on a possible connection between cis-acting mechanisms and cancer formation.
Taken together, this study describes the first comprehensive identification of allele-specific DNA-methylation in mice, and provides a resource for further mechanistic studies. The findings presented in this thesis imply that the genetic context of a specific region is strongly regulating the epigenetic status in terms of DNA methylation. It seems that genetic variability has a major impact on epigenetic variability and that this variability contributes to the shaping of different individual phenotypes. However, this might not only have an impact for early developmental steps, but also for disease-related processes in mammals.