Altered states

UWA’s Professor Ryan Lister’s research into the epigenome suggests the future clinical use of induced pluripotent stem cells should proceed with great caution.

When a specialised skin or liver cell is coaxed into reverting to a more primitive pluripotent state, the question is not what is happening to its genome, says Professor Ryan Lister. The real question to ask is: what is happening to its epigenome?

The University of Western Australia’s new Winthrop Professor of Computational Systems Biology, an invited speaker at this year’s Lorne Genome conference, says that understanding cellular reprogramming is all about epigenetics. “The reprogramming of a somatic cell to a pluripotent state is a process of epigenetic reprogramming,” he stresses.

List returned in July last year to his alma mater from the Salk Institute for Biological Studies in San Diego, California, where he secured his first postdoctoral appointment in 2006. At the Salk, he developed new techniques for whole-genome mapping of one layer of the epigenome at single-base resolution throughout the Arabidopsis genome. The epigenome consists of methyl ‘tags’ placed upon cytosine bases in the genomic DNA via DNA methylation.

DNA methylation has long been known to play an essential role in transcriptional regulation, says Lister. Determining precisely where a cell places epigenetic modifications within its genome is an essential first step towards a comprehensive understanding of the epigenome’s role in regulating genes throughout the genome, in both normal and perturbed states.

“Arabidopsis is an ideal model for methylation studies in both plants and animals,” he says. “This is because it comes with excellent genetic resources and genetic control systems with which you can both develop new experimental procedures and test the biology. It’s a powerful model for exploring epigenetic processes.”

According to Lister, one key advantage with Arabidopsis is that any null mutant created by targeted gene knockout technology remains viable, including knockouts of genes for methyltransferase and demethylase enzymes. In rodent models, on the other hand, knockouts of the same enzymes are lethal during embryonic development. Plant and animal DNA methyltransferase enzymes are conserved, so the tools developed to study methylation patterns in Arabidopsis, and the biological mechanisms that they probe, were potentially applicable to mammalian cells.

Soon after Lister arrived at the Salk Institute, new DNA sequencing technologies became available that drove a phenomenal increase in sequencing speed and output. Sequencing costs plummeted as a result. Simultaneous advances in computing meant researchers could now compare methylation patterns not only in the Arabidopsis genome (~120 megabases), but also in the gigabase-sized genomes of mammalian cells.

Epigenomic roadmap

The advent of the NIH Roadmap Epigenomics Project in 2008 saw the Salk Institute’s epigenetics group join a collaborative epigenome mapping centre, where Lister, working in the laboratory of Professor Joseph Ecker, begin applying the sequencing techniques for DNA methylome profiling that they had developed for Arabidopsis to the human genome. Their first aim was to compare the epigenomes of a differentiated fibroblast cell and a human embryonic stem cell, a pluripotent cell that has the capacity to differentiate into a wide variety of distinct cell types.

When it comes to human cells, a longstanding research challenge was to map DNA methylation throughout the entire genome at single-base precision. The established dogma in the field held that DNA methylation in mammalian cells occurred only in the CG context - that is, at a cytosine followed by a guanine. While there had been a handful of studies reporting DNA methylation in a non-CG context in mammalian cells, the papers were largely overlooked in the field.

The dogma was wrong, says Lister. By precisely determining which cytosines were methylated throughout the genome, Lister and colleagues found that DNA methylation also occurs in the non-CG context, but only in pluripotent embryonic stem cells. Within the stem cell genome, a substantial fraction of methylated cytosines were identified in the non-CG context, that is, a methylcytosine followed by an A, T or C.

“With previous approaches to study DNA methylation, technological limitations imposed significant trade-offs. You could get single-base resolution identification of DNA methylation sites, but only in a small region of the genome, on the order of kilobases,” Lister said.

“To look through whole genomes, you had to employ techniques such as immunoprecipitation of methylated fragments of genomic DNA and hybridisation to tiling microarrays, which required multiple arrays for the genome of interest, and limited detection of methylation to a resolution of approximately a kilobase at best.

“The methylation of a single cytosine has been reported to modulate interaction of a protein with the genomic DNA, so to fully understand these processes it’s essential to know where the DNA methylation is at single-base resolution.”

Where methylome mapping had previously been a hard slog, the DNA sequencing revolution enabled a well-established chemical technique for identifying methylated cytosine bases, known as bisulfite conversion, to be coupled to massively parallelised shotgun sequencing.

Routine exercise

Now methylome sequencing became almost a routine exercise: isolate genomic DNA from a sample of interest - normal, mutant or genetically modified - then treat with sodium bisulfite under denaturing conditions. The treatment selectively converts all unmethylated cytosines to uracil. But, crucially, it does not convert methylated cytosines. When the converted DNA is subsequently sequenced, a C indicates that the base was methylated in the genome, whereas a T sequenced at a cytosine position in the reference genome indicates that the genomic cytosine was unmethylated.

According to Lister, high-coverage sequencing with a single-base resolution yields a strand-specific map of exactly which cytosines were methylated for over 90% of all cytosines in a genome. This comprehensive methylome sequencing delivers reference epigenomes necessary for future comparisons with methylome patterns in distinct differentiated cell types, laboratory-induced pluripotent cells and disease states.

Researchers can now modify cell lines to determine how methylation patterns change in disease or during embryonic development, or what happens as resident pools of partially differentiated adult stem cells undergo terminal differentiation during tissue renewal and repair. Furthermore, it provides the means to explore genome-environment interactions, particularly during embryonic development.

Over the past decade, evidence has accumulated that environmental influences transduced via the placenta may modify normal methylation patterns in the embryo, giving the genome the capacity to make dynamic, and potentially adaptive, changes to the individual’s genome, which may be transmitted to the next generation.

But don’t invoke the ghost of Lamarck just yet, Lister cautions. Research into transgenerational heritability of variable DNA methylation, or ‘epialleles’, is very much ongoing. If heritable epigenetic changes do indeed occur, they may offer a means by which additional information is appended to the DNA code.

In 2009, Ryan was lead author on a Nature paper that presented the first genome-wide, single-base-resolution maps of methylation in a mammalian genome, comparing the methylomes of human embryonic stem cells (ESCs) and foetal fibroblasts. In addition, the study compared the messenger RNA and small RNA components of the transcriptomes of the two cell types and mapped several histone modifications genome-wide, as well as sites of DNA-protein interactions for several key regulatory factors.

The paper identified significant, genome-wide differences in the composition and patterning of methylation. Nearly 25% of all methylated sites in the ESCs occurred in a non-CG context, suggesting ESCs and differentiated cells employ different methylation mechanisms to affect their particular patterns of gene regulation. Non-CG context methylation was concentrated within the transcribed region of highly expressed genes, and depleted in protein-binding sites and enhancers.

“The presence of non-CG methylation in the body of highly transcribed genes offers a tantalising model whereby the transcriptional activity of genes may be affected by non-CG methylation,” Lister says. In contrast, these non-CG methylation events were completely absent in differentiated fibroblast cells.

Tantalising model

He and his colleagues have identified hundreds of differentially methylated regions in the CG context, close to genes known to be involved in pluripotency and differentiation. Moreover, the fibroblast genomes featured widespread regions of reduced methylation associated with reduced transcriptional activity. It turns out that these partially methylated domains likely represent regions of the genome that are physically associated with the nuclear lamina.

In their synopsis, they concluded that “these reference epigenomes provide a foundation for future studies exploring this key epigenetic modification in human disease and development”.

Lister was also lead author on another Nature paper in March 2011 that described hotspots of aberrant epigenetic reprogramming in induced pluripotent stem cells. The paper was the most highly cited biology paper in the world in 2011.

He says much of the promised land of regenerative medicine rests on the recent technological advances led by Nobel Laureate Shinya Yamanaka, who showed that the introduction of a handful of defined factors to somatic cells forces them to revert back into a pluripotent state, creating induced pluripotent stem cells, or iPSCs.

According to Lister, the process is fundamentally an epigenetic one. However, before iSPCs can be used in a therapeutic role, researchers must determine how closely iPSCs resemble ES cells and whether the progeny of iPSCs that follow display the same methylation patterns when they are differentiated into the desired cell types. If the technology is to be viable, no rogues should sprout among the cloned roses.

After identifying large-scale differences in the methylome between ESCs and differentiated cells, Lister and colleagues began to investigate how completely the methylome is reset when a differentiated cell is reprogrammed into an iPSC. They set about sequencing the methylomes of a panel of independent iPSC lines. These were produced by different methodologies, in different collaborating laboratories, from distinct and differentiated parental cell types.

They compared the resulting whole genome methylome maps to see how they differed from undifferentiated embryonic stem cells, the parental somatic cell lines and cells differentiated from the iPSCs and ESCs.

“In general, the methylomes of the embryonic stem cells and induced pluripotent stem cells looked very similar,” he says. “It is quite remarkable that the reprogramming factors can induce such widespread changes. But when we looked closer, using special algorithms to search for small-scale changes at high resolution, we found that each induced pluripotent stem cell line had aberrant methylation states in hundreds of distinct regions throughout the genome.

“These differentially methylated regions we identified in the iPS cell genomes fell into two broad categories. In the first, we see DNA methylation of a region that differs from ES cells, but is the same as the parental somatic cell from which the iPS cell line was derived. This can be thought of as persistent epigenetic memory of the somatic cellular state.

“In the second category, we see DNA methylation patterns that differ from both ES cells and the somatic progenitor cells, that we refer to as iPS-specific differential methylation. And since we had sequenced several independent iPS cell lines, we were able to look at how frequently they arose in independent reprogramming events. Most differentially methylated regions were present in multiple independent iPS cell lines, and a substantial fraction of them were present in all our iPS cell lines.

Genomic hotspots

“So it appears there are hotspots in the human genome that frequently tend to be aberrantly reprogrammed in iPS cell lines, either by the failure to reset the patterns found in the differentiated parental cells or inappropriate resetting to a state reflective of neither ES nor differentiated cells.

“About half of these aberrant DNA methylation patterns were transmitted through differentiation when the iPS cells were triggered to differentiate in vitro, indicating they may transmit an inappropriate regulatory state to somatic cells derived from the iPS cell lines.

“So we have identified a characteristic epigenetic signature of induced pluripotent stem cells that distinguishes them from embryonic stem cells. Ideally, for use in regenerative medicine, we would like to have iPS cells identical to ES cells, unencumbered by this aberrant reprogramming and somatic epigenetic memory.

“We think we can now use these aberrant methylation patterns as markers to test the efficacy of a whole range of reprogramming conditions and develop protocols that may deliver completely reprogrammed iPS cell epigenomes.”

One potentially advantageous implication of the discovery is that any epigenetic memory that is found not to affect the function of derived somatic cell types might be used as a non-genetic molecular marker of cells that were derived from an iPS cell. These markers could be used to track the origin of cells involved in the repair process, in order to distinguish a patient’s cells from cells derived from patient-specific iPSC lines.

He expects single-base, genome-wide comparisons of DNA methylation patterns between individuals may reveal substantial, inter-individual variation which may be involved in differential activity of genomic regulatory elements.

“However, the question would be what specifies these different epigenetic patterns? Currently, we do not have a good understanding of whether they are simply derived from underlying genetic variants, as well as how specific epigenetic patterns influence gene expression.

“Very little is known about causality in the relationship between variable DNA methylation, DNA-protein interactions and changes in transcriptional activity. Examples have been described where DNA methylation state modulates protein-DNA interactions; however, the inverse has also been documented,” he says.

Recent research has identified at least two instances where DNA-binding proteins alter DNA methylation patterns when they bind to the genomic DNA. Overwhelmingly, researchers assumed that DNA methylation was causal for altering gene expression, not vice versa, so there is now a huge challenge to systematically understand causality with respect to these epigenetic changes and gene activity.

“If we don’t understand the basic biology in order to progress beyond the current assumptions, we’re going to head down a lot of dead ends when investigating the role of the epigenome in development and disease.

“It’s also important to remember that there are multiple flavours of epigenetic modification and complex relationships with non-coding RNAs. We need to understand how these all interact and their combined effect upon the complex and dynamic three-dimensional structure of the genome.

“We tend to envisage the genome as a set of linear structures, but some exquisite experiments recently have begun to map the higher-order structures of our genome. These experiments show that the three-dimensional structure of DNA is critical to understanding how the genome functions, and it will be important to integrate with these structures the maps of the epigenome that we can now generate.”

While iPSCs are touted to have tremendous potential in research and medicine, we clearly still have a lot to understand about the detail of their function and how the various epigenetic factors can influence their ultimate fate.

Professor Ryan Lister acquired a PhD in molecular bioscience at the University of Western Australia in 2004. In 2006 he accepted a postdoctoral appointment at the Salk Institute Biological Studies in San Diego, California, where he studied the epigenomes of Arabidopsis and human cells. He returned to UWA to take up an appointment as Winthrop Professor of Computational Systems Biology last July.

Image credit ©iStockphoto.com/Sebastian Kaulitzki