Feature: Epigenetics and diabetes

Even many years after successful therapy, lifestyle and dietary changes to reduce high blood-glucose levels, some Type 2 (non-insulin-dependent) diabetics continue to suffer diabetes-like health problems and poor quality of life. It is as if diabetics’ cardiovascular responses become habituated to the individual’s formerly excessive blood-glucose levels.

The phenomenon of ‘glycaemic memory’ has long puzzled diabetes clinicians and researchers. Now, researchers at the Baker Heart and International Diabetes Institute (Baker IDI) have made a crucial advance that explains how exposure to high blood-glucose levels drives epigenetic changes that underlie the syndrome. The new understanding of the mechanisms involved suggests the syndrome could be treated with novel therapies to reverse the epigenetic changes.

Glycaemic memory syndrome involves a variety of cardiovascular problems normally associated with chronically high blood glucose levels and insulin resistance. They include heart attack, stroke, atherosclerosis, kidney failure, vision loss due to diabetic retinopathy, memory loss, cognitive problems and mood disorders. But the underlying cause of the syndrome has only recently begun to come to light.

In 2008, Baker IDI’s scientific executive officer, physician Dr Mark Cooper, and Professor Assam El-Osta, head of human epigenetics, published a paper in the Journal of Experimental Medicine showing that transient high blood glucose levels induced persistent methylation-mediated epigenetic changes in the cells lining the aorta. It was the first indication that epigenetic mechanisms caused glycaemic memory.

As a complement to their human aortic tissue study, Cooper and Dr Ana Calkin took plaque samples from 30 atherosclerotic mice. Half had been exposed to high blood glucose levels to induce Type 2 diabetes, while the rest had been fed a high-fat diet but were maintained at normal blood-glucose levels. Gene expression patterns in atherosclerotic plaques and plasma lipid from the aortas of the diabetic mice were strongly concordant with the in vitro glucose-response patterns in epithelial cells from human aortas and diabetic patient lipid patterns, respectfully. Haviv believes the lipids capture some of the signalling cascades that lead to the epigenetic changes we observe in the human patients.

Methylation types

Two types of methylation occur, says Haviv. One involves methylation of CpG islands – sites with extended C-G base-pair repeats – that are relatively unresponsive to environmental cues. “CpG island methylation is more tightly regulated during replication and causes relatively irreversible changes – the cell commits to some function and doesn’t lose this silencing mark again. The other mechanism is gene silencing by post-translational modification of histone proteins, particularly H3 and H4.”

The basic structural unit of chromatin is the nucleosome, a segment of DNA wound around a core of histone protein. According to Haviv, up to 70 amino acids in the amino terminals of histones H3 and H4 are actually redundant to the nucleosome’s structure, but essential for regulating the genes in that region of the genome.

---PB---

“The redundant segment flaps outside the nucleosome, acting as a barcode marker.” Haviv said. “Enzymes, such as Methyl-transferase or Acetyl-transferase, and conversely, Histone deacetylase or demethylase, modify particular amino acids, and depending on where the modification occurs, the chromatin will either be open and poised for activation, or condensed and silenced [refractory to activation]. We can interrogate the DNA from the nucleosome for the presence of epigenetic markers, which complements our work to identify transcription factors which in turn conduct the orchestrated cellular responses to environmental cues.”

Haviv says the epigenetic modifications increase expression of Nuclear Factor kappa beta (NFkB) p65 subunit, encoded by the RELA gene. Through its involvement in inflammatory responses, p65 influences cell growth, maturation and programmed cell death – apoptosis.

In their Journal of Experimental Medicine paper, Cooper and El-Osta, suggested glycaemic memory involves persistent changes in DNA packaging in chromatin in the nuclei of the cells of the blood vessel endothelium. “We imagine DNA as a linear molecule, and visualise distances between genes and regulatory elements in terms of their linear separation on a DNA strand. In fact, at any given time, the chromosomes in the nucleus of a non-dividing cell are loose, and chromatin density varies across the genome,” says Haviv.

Haviv says genomic regions not required for the specialised functions of a particular cell type remain condensed, rendering genes within them inaccessible to transcription complexes. Other chromatin regions containing genes required for cell function are open and poised for gene activation. Specialised homeobox (pioneer) molecules that open up condensed chromatin then unpack the DNA-protein complexes, making the genes accessible to transcription factors.

When DNA is unpacked after mitosis, it is not randomly distributed in the nucleus. “The active regions are positioned closer to the nuclear membrane,” says Haviv. “The arrangement makes sense, because a high degree of coordination is required to produce a primary RNA transcript from a gene, splice it into its mature form, transport it to the nuclear membrane, then export it into the cytoplasm. Positioning active regions of the genome close to the nuclear membrane reduces the distance and time involved in transporting and exporting the mature mRNA.

“It changes the way we visualise genomic distance. Researchers at the Genome Institute of Singapore recently published a paper in Nature describing a way of demonstrating that two seemingly distant points in a linear DNA sequence could be very proximal in the nucleus and interact.”

Haviv says a structure as simple as a loop could bring two distant sequences close enough to interact directly. One sequence might be a binding site for regulatory proteins, while the opposite end of the loop might harbour a complementary sequence spanning the start site of regulated gene.

---PB---

“This interaction at a distance can thus influence the activity levels our genes,” he says. “Closer inspection of these loops reveals the presence of a novel code – the epigenetic code – that is hardwired in our genome, and reflects periodicity within the ordered, three-dimensional structure of chromatin.”

Chromatin consists of DNA wound around ‘beads’ of histone proteins. The histone-wound strands then undergo progressive rounds of coiling and supercoiling that minimises packing volume, allowing an impressive three metres or so of DNA to be accommodated within the confines of the cell nucleus.

“The chances that periodicity would bring any two sequences close enough to interact directly are essentially random,” says Haviv. “So distant interactions are mediated largely by protein-protein interactions that physically bridge the proximal sequences, such as a transcription element at one locus and its target gene at another. The system needs an additional component to work: proteins that act as insulators by interposing themselves to prevent bridging proteins reaching over between the DNA strands.”

Epigenetic elves of the genome

In a complementary project, in one of his research projects Dr Haviv is exploring how non-coding RNA molecules (ncRNAs) mediate epigenetic changes to chromatin structure and gene activity. NcRNAs – the ‘elves of the genome’ – coordinate gene activity. They are transcribed from the vast tracts of non-protein-coding DNA that form more than 98 per cent of the human genome, and which were once dismissed as ‘junk DNA’.

The National Human Genome Research Institute’s Encyclopaedia of DNA Elements (ENCODE) project has catalogued functional genetic elements according to their expression patterns in common cell types. Each of the 200-plus specialised cell types that make up the human body exhibits a distinctive pattern of non-coding RNA molecules that shape its identity and regulate its functions. Haviv says a gene’s promoter contains marker DNA sequences complementary to sequences within the non-coding RNAs (ncRNAs) that regulate its activity. Conversely, ncRNA elements contain sequences corresponding to promoter sequences in their target genes.

Search algorithms exploit this complementarity to identify the genes regulated by a particular element, or conversely, to identify non-protein-coding DNA sequences harbouring the RNA elements that regulate a given gene.

Using microarrays and next generation sequencing to compare gene-expression patterns in endothelial cells from the aortas of cadavers, El-Osta’s group identified two classes of candidate genes for glycaemic memory: genes up-regulated by glucose, and genes down-regulated by glucose. In the glycaemic memory project, identifying glucose-responsive genes, and their associated CpG DNA methylation and histone acetylation, provided the means to interrogate epigenetic markers that may have chronically modified gene activity in response to environmental influences: specifically, exposure to elevated blood concentrations of glucose in Type 2 diabetes.

Haviv’s group has employed the same methodology to investigate the involvement of epigenetic mechanisms in certain cancers, including breast cancers (See “Distant SNPs and cancer”).

---PB---

Epigenetic adaptation?

Haviv says the elusive mechanism linking lifestyle factors to phenotype may involve relatively stable, heritable epigenetic marks. A decade of research has shown that exposure to certain environmental influences, including in utero factors, can cause epigenetic ‘reprogramming’ of genes, that the individual may then transmit to their own offspring – in effect, pre-programming their genomes to meet the challenges of the environment into which they will be born.

A mother’s exposure to stress, malnutrition or other environmental influences during pregnancy can induce permanent epigenetic changes in her baby during foetal development. Some pregnant women develop gestational diabetes, raising the possibility that their own exposure to high blood glucose levels could influence their children’s susceptibility to Type 2 diabetes.

Assam El Osta’s group at Baker-IDI has initiated a number of projects to investigate glucose-induced epigenetic changes in rodents and rabbit models. The animals are fed a high-fat diet before and during pregnancy, and their offspring are studied for epigenetic changes. Haviv’s team performs the data analysis for these studies. “The data show that epigenetic control of gene expression is not a simple matter,” says Haviv.

He said the epigenetic effects of glucose exposure on gene expression were completely unknown, and even before they started, they were challenged to explain a deep mystery: earlier genetic studies had shown that regulatory elements influencing gene expression were sometimes located up to a million base pairs upstream or downstream of the gene of interest.

“Expression quantitative trait loci [eQTL] analyses have revealed that around 90 per cent of transcription factors bind to sequences within the promoters of target genes, or nearby,” Haviv said. “But in about 10 per cent of cases, the eQTL signal emanates from the middle of nowhere, like a smile without the Cheshire Cat.”

To explain such ‘actions at a distance’, they turned to the vast amount of empirical data generated in ENCODE, which allowed them to link information about a target gene’s expression to the binding of specific regulatory transcription factors.

“We now have a pipeline of analysis that makes it relatively easy to identify the guiding principles and themes in epigenetic control, even over great distances,” Haviv said. “Our approach is proving particularly useful in studies of allelic variation, where eQTL data is unavailable, but a remote SNP [single nucleotide polymorphism] is strongly associated with a disease.”

SNPs that influence gene expression sometimes occur within neighbouring genes, but others turn up in non-protein-coding DNA as much as 2 megabases away. “When you submit the chromatin to precipitation with specific immunoprecipitation factors, you find that some classical transcription factors bind at great distance from the genes they regulate, so functionally, they ‘leap over’ intervening genes,” Haviv says.

---PB---

Why is the linear distance between a gene and its regulatory elements significant? When meiotic recombination forms new gene combinations, or haplotypes in gametes, any regulatory element lying within or near the gene it controls will be co-inherited with it. But the chances of recombination increase with distance, and when a gene’s regulatory element lies one to two megabases away, the chances of recombination are significantly increased. A recombination event can create a new haplotype that links the gene to different SNP variant of its control element, which can alter its expression.

Haviv summarises the problem thus: “When known disease-associated germline variation occurs at high frequency within an affected population, and it falls within a protein-coding exon, altering its amino acid sequence, it’s easy to determine what’s going on. But when the associated variation lies in a non-protein coding sequence a million base pairs away, you can spend millions of dollars investigating the problem without gaining any insight into what is going on.

“The researcher knows there must be a transcription factor that binds at the site of the variation, but cannot determine how the base-pair change results in a loss or gain in function in the gene at such a distance.”

Haviv’s findings with collaborators at the University of Melbourne school of computer science and the national information and communication technology (NICTA) have now removed this impasse. “Because certain rules apply, for every transcription factor we identify and characterise, we can now predict the specific base-pair spectrum of the gene sequences it will recognise,” Haviv said.

“For example, if the SNP substitutes a base pair (A to C, G or T, etc.), we can calculate its thermodynamical effect on the binding interaction, and then predict which transcription factors will be activated or de-activated by a particular germline change.

“Of 140 disease-associated SNPs in the middle of nowhere, reported in the NHGRI GWAS [genome-wide association study] catalogue, more than half turn out to occur on binding sites of transcription factor known to be involved in the disease being investigated, such as diabetes. We further zeroed in on 11 SNPs, and used expression profiling to identify the genes that the transcription factor was up-regulating. One of those SNPs led us to identify the transcription factor involved in diabetes control, and potentially in glycaemic memory.

“That gives us information we need to manage Type 2 diabetes. With J. Jowett, M. Febbraio and Graeme Lancaster, we are now screening possible mechanisms to reverse the epigenetic memory, by using short interfering RNA molecules [siRNAs] to regulate the level of non-coding RNA. If the approach turns out to be feasible, we would use the allelic variations and SNPs to predict which patients who are likely to respond to specific treatments. There are opportunities for therapeutic intervention – it’s an important step towards the age of personalised medicine.”