The Paper in brief

  • The authors1 analyse genomic variation among a large group of patients who respond differently to metformin — the most commonly used drug for treating type 2 diabetes.

  • This genome-wide association study (GWAS) involved 1,024 Scottish patients, as well as a further 2,896 patients to check for replication of the results.

  • Treatment success was significantly linked to the presence of a single nucleotide-base variation designated rs11212617.

  • This variation occurs in a genomic region (locus) that also contains the gene ATM. ATM encodes a tumour-suppressor protein involved in DNA repair and cell-cycle control that is mutated in ataxia telangiectasia, a neurodegenerative disease associated with a predisposition to cancer.

  • The authors further show, in a rat cancer cell line, that inhibition of ATM weakens metformin-mediated activation of a metabolic enzyme, AMPK.

A clue to metformin action

Morris J. Birnbaum

Let me start on a positive note. The genetic association between ATM and metformin sensitivity1 represents a triumph of modern pharmacogenetics, and it is reasonable to hope that it will lead to fundamental insights into metformin's mechanism of action and the regulation of carbohydrate metabolism.

You might ask how a rare variation linked to ATM could possibly be related to the treatment of a common disease such as diabetes. After all, given the prevalence of diabetes, it would be reasonable to assume that it takes only slight disruptions in nutritional intake or energy expenditure to shift the balance between health and metabolic disease.

This is in fact far from the truth. To contract an illness with the serious consequences of type 2 diabetes, not only must there be disturbances in energy balance, but the normal homeostatic mechanisms that regulate metabolism must also be impaired. In other words, if the regulatory system fails to respond to caloric overload, disease ensues. With the relative uniformity of the Western — and increasingly global — lifestyle, a major determinant of susceptibility to type 2 diabetes must therefore be an individual's genetic make-up, which dictates the response to nutritional overload or therapeutic intervention.

Metformin works mainly by reducing glucose production by the liver, but there is still uncertainty about its mechanism of action at a molecular level. The drug blocks a step in the aerobic production of the cellular energy molecule ATP, activating a signalling pathway in which the enzyme AMPK senses energetic stress within the cell. Nonetheless, despite activating AMPK, metformin actually works independently of the enzyme2,3. The discovery of a role for ATM in modulating metformin responsiveness might provide a clue to the mechanism of action of this drug.

Unfortunately, however, it could equally well be a false lead. Classic genetic screens have taught us that some candidate genes can exert very indirect effects, providing little information about crucial signalling pathways. For example, variations in another gene also affect metformin responsiveness, but that gene's product, OCT1, influences the rate of uptake of metformin by cells, rather than any major signalling pathway4.

Another possibility is that ATM influences blood glucose levels through pathways parallel to — but not the same as — those modulated by metformin, and that its effects become apparent only with synergistic input from the drug. Indeed, 40 years ago, it was noted5 that patients with ataxia telangiectasia often display a type-2-diabetes-like syndrome characterized by an insulin resistance too severe to be caused just by changes in liver glucose production.

With the genetic clues now to hand, careful biochemical and cell-biological studies should be performed to figure out the nature of the interaction between ATM and the beneficial effects of metformin. Most crucial will be to find out whether there are other molecules, apart from AMPK, that both control metabolism and are influenced by ATM.

The cancer connection

Reuben J. Shaw

A tumour-suppressor protein that mediates DNA repair and has ties with a metabolic disorder1 — this might sound far-fetched. But in fact the reported link between responsiveness to metformin and a cancer gene is not without precedent.

Previous work has shown that activation of AMPK by metformin requires the activity of the kinase enzyme LKB1. The gene encoding LKB1 was originally identified for causing an inherited cancer disorder, and is one of the most commonly mutated genes in human lung cancer6. Animal studies also point to a role for this gene in a variety of spontaneously arising cancers. Notably, deletion of Lkb1 in mouse liver leads to loss of AMPK activity in that organ and to the development of metabolic dysfunction, including hyperglycaemia and hepatic steatosis — symptoms resembling those of type 2 diabetes7. Furthermore, a genetic survey revealed that DNA-base variations in LKB1 affect how women with polycystic ovary syndrome respond to treatment with metformin8.

The fact that the present study (the first GWAS to find a locus that 'dictates' metformin response) has identified a possible role for ATM — which is also a kinase enzyme — could fit in with several earlier observations. For instance, patients with ataxia telangiectasia show insulin resistance and are at a higher risk of developing diabetes, and mice with defective ATM activity show insulin resistance and abnormal glucose homeostasis9.

How might ATM be involved at a molecular level? The present work1 hypothesizes that this enzyme modulates patients' responsiveness to metformin by affecting the drug's ability to activate AMPK. Indeed, ATM is known to phosphorylate LKB1 — AMPK's key activator10,11 — thereby affecting various cellular processes. ATM might also regulate AMPK independently of LKB1. Furthermore, it may affect metformin responsiveness by regulating other relevant targets that are independent of AMPK (Fig. 1). Indeed, ATM is known to phosphorylate other components of the insulin signalling pathway12,13.

Figure 1: Possible mechanisms of the anti-diabetic effects of ATM.
figure 1

In response to metformin, ATM could mediate the phosphorylation (P), and so the activation, of AMPK by phosphorylating LKB1. Alternatively, ATM might activate AMPK independently of LKB1, or reduce blood glucose levels through pathways entirely independent of AMPK.

In light of these intriguing connections, it is essential to rigorously examine whether the rs11212617 variant serves to modulate ATM activity, towards AMPK activation or that of other downstream targets. From the present data, it is not clear whether ATM activity is even perturbed by this variant.

As for metformin's relevance to cancer therapy, patients with diabetes who take this drug have a lower risk of developing cancer than those on other anti-diabetes medications14. Whether metformin is a general activator of ATM and its targets in the DNA-damage-response pathway should therefore be thoroughly investigated. Although LKB1 and AMPK are good candidates for mediating some of the beneficial effects of metformin on cancer risk, the involvement of the broader tumour-suppressor pathways controlled by ATM is an intriguing possibility. Future studies dissecting the relationship between metformin action, ATM, LKB1 and AMPK should shed light on the intersection between suppression of the risk of cancer and that of diabetes.