Feature: Cancer is a lean, mean, epigenetics machine

This feature appeared in the November/December 2010 issue of Australian Life Scientist. To subscribe to the magazine, go here.

Science’s capricious handmaiden, Serendipity, was close at hand when Susan Clark’s epigenetics research group at the Garvan Medical Research Institute went searching for hypermethylated genes in colorectal cancer tissue samples in 2006.

Hypermethylation had long been thought to be a localised process, operating at the level of the individual gene. However, Clark’s team overturned that assumption when it identified a hypermethylated region extending over a four megabase block on the long arm of chromosome 2.

This region contained what could be called three discrete ‘suburbs’, in which every gene within each ‘suburb’ had been silenced by hypermethylation. They had found evidence for a novel mechanism that appeared to act globally rather than locally, triggering wholesale changes in normal gene expression in cells.

Aberrant methylation is a widespread phenomenon in cancerous cells. It occurs in gene promoters in regions with CpG ‘islands’ – DNA sequences highly enriched in cytosine-guanine base pairs. DNA methyltransferase enzymes attach methyl groups to C-G base pairs within the gene’s promoter, blocking transcription, resulting in the equivalent of a loss-of-function mutation or gene deletion.

The Garvan team’s discovery pointed to a gene-silencing mechanism that operates at a distance to suppress expression of blocks of neighbouring genes. On the assumption that the phenomenon was not restricted to colorectal cancers, Clark’s team went hunting for hypermethylated blocks in prostate cancer – and hit the jackpot.

The Garvan team described its findings in a paper published in Nature Cell Biology in March. Marcel Coolen and Clare Stirzaker were the lead authors.

The paper provided an unprecedented view of the complex, multi-layered processes involved in tumorigenesis and, later, in metastasis. In addition to methylation, long range epigenetic silencing (LRES) involves multiple, complementary silencing mechanisms, sometimes in ‘overkill’ combinations.

They include histone deacetylation, and the involvement of histone-associated polycomb proteins, which play a central role in regulating cell plasticity during embryogenesis.

The paper notes that CpG island-associated genes have been linked to pluripotency in human embryonic stem cells, while progenitor stem cells commonly have histone H3 proteins that feature a trimethylatated lysine residue, called a H3K27me3 mark, identifying that locus as a target for polycomb proteins that regulate gene activity.

The authors write: “Intriguingly…these polycomb target genes… constitute a significant fraction of genes that are hypermethylated in cancer cells, suggesting that H3K27me3 may trigger aberrant DNA methylation by recruitment of DNA methylation machinery.”

---PB---

Reduced expression

Over multiple rounds of cell division, these gene-silencing mechanisms progressively produce highly reduced patterns of gene expression. Concordant methylation-induced silencing had previously been noted in the HOXA gene cluster, whose members switch on in order of their linear arrangement to establish the body plan in the earliest stages of embryogenesis in all animals.

But nobody seems to have been alert to the possibility that en bloc gene-inactivation it might be a generic phenomenon, involving large clusters of genes with disparate functions.

The Garvan team found that LRES regions in prostate cancer cells range between 0.2 and 5.1 megabases in size, with an average size of 1.9 Mb. ‘Silent suburbs’ contain between five and 28 genes, with 12 being the average; a majority of the silenced genes have CpG island-associated promoters and they collectively span 2.9 per cent of the genome. The region of suppression is typically broader in metastatic than localised cancers, indicating that LRES regions expand between cell generations.

Thus, the cancerous cell line’s ‘silent suburbs’ are not static, but progressively extend from their point of origin, suppressing more genes with each round of cell division. Eventually, they create what Clark describes as a “lean, mean machine”, cells that focus on their own growth, dispensing with any genes that might get in their way.

H3K27me3 markers aside, what type of marker might cue DNA methyltransferases to initiate hypermethylation at a particular locus? It’s an open question says Clark, but a likely explanation is that the methylation enzymes home in on ‘scars’ where double-stranded DNA repair enzymes have repaired – or failed to repair – old point mutations.

The Garvan team was intrigued by the fact that about a third of ‘silent suburbs’ harbour tumour-suppressor genes. It is as if some malign genie were systematically knocking out the genomic watchdogs that guard against cancer.

The tumour-suppressor p53, dubbed the ‘guardian of the genome’ for its dominant role in initiating programmed cell death in cells suffering genomic stress, is inactivated in up to 70 per cent of cancers – most frequently by mutation, but also by epigenetic modification. In the latter case, epigenetic silencing involves either methylation or acetylation.

Whatever the nature of the start point, the fact that multiple tumour-suppressor genes are inactivated in cancerous cells, in the absence of any knockout mutation, hints at how the process may work. It may explain why hypermethylated regions expand during successive cycles of cell division, adding more inactivated genes to the ‘silent suburbs’ that the Garvan researchers identified.

A tumour-suppressor gene need not contain a methylation-initiation site within its sequence. Any ‘scar’ in a nearby gene would suffice to as a starting point, and as hypermethylation extends along the chromosome, it will silence any tumour-suppressor gene in its vicinity.

There is nothing supernatural about the process: it is as random as the occurrence of mutation sites. Inevitably, random hypermethylation carries a high risk that genes essential for cell growth will be silenced; that cell line will leave no daughters.

---PB---

Going critical

But survival of cancerous cells is a numbers game, and mutant cell lines that manage to remain viable as their expressed genomes are progressively downsized, will live on to spawn daughter lineages with an every-increasing complement of impotent tumour-suppressor genes.

Like a nuclear chain reaction, the accumulation of mutations and methylated chromosomal neighbourhoods with neutered tumour suppressors will eventually cause the cell to ‘go critical’ and spiral into cancerous growth – and in time, to metastasis. With the shackles off, cellular activity would be increasingly directed towards rapid growth and division.

The Garvan prostate cancer study shows that metastatic tumours, founded by roaming cells that have detached from primary tumours, are the endpoint of the process: gene activity is stripped back to the bare minimum required to sustain growth and division.

“We were surprised to find that many of these silenced regions exhibit common genetic changes,” says Clark. “We don’t know what comes first, but it raises the question of whether the genetic changes predispose the cell to epigenetic changes, or vice versa.”

Clark says the complexity of these mechanisms requires a reappraisal of potential epigenetic-based therapies to reactivate silenced tumour suppressors, to reinstate programmed cell death.

Even if drugs like demethylases or histone deacetylase inhibitors (HDIs) can be targeted to selectively reactivate silenced tumour-suppressor genes without disrupting normal epigenetic regulation of gene activity, it may be necessary to design drug combinations keyed to the LRES profile of the individual’s cancer.

Clark says LRES “switches off everything” non-essential to the process of reducing and remodelling gene activity for rapid cell growth and division. LRES thus provides a mechanism for eliminating a healthy cell’s fail-safe systems, by silencing backup genes that build multiple layers of redundancy into the gene networks that regulate normal growth and division.

“In cancerous cell lines, at any given time, LRES would create multiple variables that, at any given time, offer an expanding pool of combinations of silenced genes for natural selection,” Clark says.

“We’re finding this with metastatic tumours. These changes progressively accumulate, and the silenced regions are not static – they extend along the chromosome with each round of cell division. It keeps going. Some lineages die out, others keep reducing genomic activity. We think this will eventually reveal the minimum genomic DNA that needs to be available for expression to keep cells alive.”

---PB---

Clark says the study has gone some way to resolving a key question in the cancer literature: do cancers arise from cancerous embryonic stem cells, or from cancerous multipotent, cancerous adult stem cells?

“We were interested to see whether the epigenome of prostate cancer cells is more similar to that of embryonic stem cells or cancerous cells that have differentiated from progenitor cells,” she says.

“Prostate cancer develops from normal epithelial cells, and in this case at least, this particular prostate cancer did not seem to have originated from an embryonic stem cell. No doubt there will be exceptions, but our findings very much support a non-stem-cell origin for cancers. Mistakes can occur along the differentiation pathway that divert lineages towards cancerous growth.”

Clark says a cancer’s aggressiveness may reflect how far the cell has progressed towards terminal differentiation; mistakes early in the differentiation process, when progenitor cells are more plastic, are likely to produce more aggressive cancers. The Garvan researchers confirmed that LRES occurs in clinical samples, as well as their laboratory prostate cancer cell lines.

Epigenetic profiling focused on LRES domains is likely to become a valuable prognostic tool that will help oncologists to devise personalised therapies for individual patients.

“We first reported LRES in colorectal cancer through therapy, and we chose prostate cancer to see if it might be a common mechanism in many cancers,” Clark says. “We’ve now gone on to show the identical phenomenon occurs in breast and ovarian cancers, so it’s now fairly safe to say it’s a generic mechanism in oncogenesis.”

Epigenetic triggers

The key question now becomes: what initiates long-range epigenetic silencing, and when, where and why? Clark’s team says that in prostate and other cancers LRES genomic locations tend to overlap with regions of genomic deletion, that result in loss of heterozygosity (LOH).

LOH occurs when one allele of a gene has already ceased to function, and the ‘backup’ allele on the other member of the chromosome pair is deleted or silenced. “It’s happening across the genome, but where does it start?” she said. “It’s a difficult question, but the answer could lie in deregulation of the architecture of chromatin.”

Clark says her data suggest silenced chromosome domains appear to correspond to loops within the higher-order structure of chromatin, which consists of sequences of DNA wound around histone protein ‘beads’. The primary coils then undergo multiple rounds of coiling and supercoiling to create the highly compact structure of condensed chromosomes.

---PB---

“LRES might also involve deregulation of DNA repair processes, affecting histone structure, which may change gene-expression patterns,” she says. “We’ve defined the features of the epigenome map in prostate cancer, and we’re now using it to assess the events downstream of gene suppression.”

In her team’s Nature Cell Biology paper, they suggest that LRES in cancer causes further consolidation of the genome to a more definitive epigenetically-induced repressive state across large domains. The process involves a large variety of epigenetic marks, and results in reduced transcriptional plasticity.

Clark says this supports the idea that many local chromosome regions are under coordinated epigenetic control, and the LRES regions they observed in prostate are merely a subset of a more generalised phenomenon.

The pattern of gene silencing in prostate cancer cells closely resembles that produced by X-inactivation in the cells of female mammals. X-inactivation traces to the X-inactive specific transcript (XIST) locus, which harbours a 17 kb, non-protein coding RNA that initiates wholesale suppression of genes across most regions of the X-chromosome.

“It would be wonderful if we found that an RNA-based mechanism also initiates LRES,” she says. “That would provide a possible target for a relatively simple RNA-based therapy to reactivate genes in suppressed domains, avoiding the need for combination therapies to reverse the various modes of silencing.”

This feature appeared in the November/December 2010 issue of Australian Life Scientist. To subscribe to the magazine, go here.