John Seymour, Britain's self-sufficiency guru, could have learned a lot from cancer cells: by producing growth factors and their receptors, they lose reliance on their neighbours. Thomas Graeber and David Eisenberg now describe a method for cataloguing these autocrine loops.

The authors reasoned that if a cancer cell expresses both a ligand and its receptor, their transcription might be coupled. So they compiled a database of ligand–receptor pairs from the literature, and then mined five collections of publicly available gene-expression data from human cancers and cancer-cell lines for the expression profiles of ligand–receptor pairs. They then calculated whether there was a correlation between the expression profiles of each ligand and its receptor.

This identified several autocrine loops known to occur in cancer. One of these was tumour necrosis factor (TNF) and its receptor TNFR1 in diffuse large B-cell lymphoma. Expression of this pair correlated only in tumour cells from patients who had not survived beyond 4 years after diagnosis and treatment, illustrating the prognostic — and potential therapeutic — relevance of autocrine loops. CC chemokines and their receptors, which are involved in cell movement and might be expected to confer a metastatic phenotype, were also overrepresented.

Some pairs indicated that cancer cells might also have strategies for blocking autocrine signalling: expression of some 'decoy' receptors correlates with their ligands, and inverse correlations between ligand and receptor expression also occur. For example, in several cancer types ERBB3 expression is downregulated whenever its ligand, neuregulin 1, is upregulated.

This study has generated a long list of hypotheses that now need testing to determine whether the loops identified are operative. But it could provide rich pickings for the identification of prognostic markers and therapeutic targets. Might it be possible to set a pharmaceutical fox among the chickens in some of these self-sufficient systems?