News & Views | Published:


The rules of attraction

The puzzle of how a drug that binds to a protein found in normal cells as well as cancer cells preferentially kills tumours is now solved — the target protein exists in a drug-binding complex in tumour cells.

Targeting a specific protein or a single signalling pathway that is required for the survival of tumour cells but not normal cells would seem to be a promising anticancer strategy. Unfortunately, few such unique targets exist, and it is becoming clear that inhibiting a single pathway might not be enough to tackle cancers that result from several genetic abnormalities. Instead, attention is turning to proteins such as heat-shock protein 90 (Hsp90) that regulate many signalling pathways in cancer cells. One feature of Hsp90 has concerned investigators — although cancer cells can produce high levels of the protein1, it is also abundant in normal cells. This might mean that drugs targeting Hsp90 prove to be unacceptably toxic. Surprisingly, however, the first Hsp90 inhibitor to be tested in clinical trials, the drug 17-AAG, has been well tolerated by patients. On page 407 of this issue, Kamal et al.2 provide data that begin to explain this apparent paradox. These authors report that Hsp90 found in tumour cells has a much higher affinity for 17-AAG than does Hsp90 from normal cells.

In 1962, while looking at the salivary-gland chromosomes of the fruitfly Drosophila, Ferruccio Ritossa noticed that certain regions of the chromosomes puffed out in response to a sudden increase in temperature3. The gene products encoded on these chromosome puffs were later isolated and termed heat-shock proteins, or Hsps. The production of Hsps accelerates in response to temperature stress, but these proteins are abundant even in unstressed cells. Hsps have been more accurately called 'molecular chaperones', because they protect other cellular proteins from becoming misshapen as a result of high temperature or other environmental insults, and certain Hsps also enable newly synthesized proteins to attain the correct conformation.

One particular chaperone, Hsp90, has been implicated in the survival of cancer cells4. Hsp90 regulates the function and stability of many key signalling proteins that help cancer cells to escape the inherent toxicity of their environment, to evade the effects of chemotherapy, and to protect themselves from the results of their own genetic instability. So inhibitors of Hsp90 could mount a multi-pronged assault on cancer cells that, if not lethal itself, might leave them sufficiently debilitated to allow control by chemotherapy or radiotherapy.

Despite this compelling rationale for Hsp90-directed anticancer therapy, the abundance of the protein in normal cells raised concerns that an Hsp90 inhibitor, such as 17-AAG, would be unacceptably toxic to patients. But unaccountably, preclinical data and the results of an initial clinical trial showed that this drug seemed to target tumour cells in preference to normal cells. Why? Kamal et al.2 provide an answer. They find that Hsp90 derived from tumour cells binds to 17-AAG up to 100 times more tightly than does Hsp90 isolated from normal cells. Intriguingly, Hsp90 from normal cells binds the drug with nearly the same low affinity as does purified Hsp90 (refs 2, 5).

To begin to investigate the higher affinity of 17-AAG for tumour Hsp90, the authors looked at whether the chaperone interacted with other proteins inside tumour cells and normal cells. To stabilize its 'client' proteins, Hsp90 assembles with other chaperones and associated proteins to form a 'super-chaperone machine'6. Kamal et al. found that in tumour cells, the bulk of Hsp90 exists in such an assembly, whereas most of the Hsp90 in normal cells exists in a free form. The Hsp90 in tumour cells also had higher ATPase activity (required for its chaperone function) — a finding that supports the view that tumour Hsp90 is present in fully active chaperone complexes.

So the affinity of 17-AAG for Hsp90 seems to depend on the incorporation of the chaperone into a multi-protein machine. Strikingly, Kamal et al. were able to increase the weak affinity of purified Hsp90 for 17-AAG roughly 50-fold by adding components of the chaperone machine. Data on drug levels achieved in patients support these findings — intravenously administered 17-AAG is present in the circulation for many hours at concentrations far exceeding its apparent affinity for tumour Hsp90, but only transiently reaches the high concentration that would allow binding to the free form of the protein found in normal cells7. This might partly explain the relative lack of toxicity of 17-AAG in patients.

What makes Hsp90 better able to bind 17-AAG when the protein is part of the super-chaperone complex? At present we can only speculate. An intriguing possibility is that the chaperone complex might catalyse a conformational change in the drug. Indeed, it has been suggested8,9 that the structurally similar drug geldanamycin (the 'parent' of 17-AAG) must undergo a conformational change from an open, planar structure to a more compact 'C-clamp' shape before it can bind to Hsp90 (Fig. 1). The energetics of the spontaneous conversion between these two forms is highly unfavourable8 (Y.-S. Lee, M. G. Marcu and L. Neckers, unpublished observations). So one or more components of the super-chaperone machine might be much more efficient than free Hsp90 at catalysing a conformational change in drugs such as 17-AAG; alternatively, the complex might potentiate the ability of Hsp90 itself to do so. Both possibilities are feasible. For example, association with other chaperones has been shown to stimulate the normally weak ATPase activity of Hsp90 (ref. 10). And other components of the Hsp90 multi-chaperone complex possess an intrinsic ability to modulate client-protein conformation11. Does the Hsp90 super-chaperone machine view 17-AAG as a protein in need of refolding? Further experiments will be needed to answer this question.

Figure 1: Catalysts of change?

The Hsp90 inhibitor geldanamycin exists in two forms. In solution it exists in an extended conformation, but it must switch to a 'C-clamp' conformation before it can bind to the Hsp90 protein. The energy barrier between these two conformations is too great to allow spontaneous conversion, so the change must be catalysed. Kamal et al.2 show that the drug 17-AAG, which is related to geldanamycin, binds more tightly to Hsp90 when the protein is part of a 'super-chaperone machine' that actively modulates the shape of 'client' proteins. It is possible that Hsp90 itself catalyses the conversion of drugs like 17-AAG — and that the super-chaperone machine might be a more efficient catalyst of this process than uncomplexed Hsp90.

Meanwhile, modifications of 17-AAG and other geldanamycin derivatives are being developed that spontaneously form the C-clamp conformation in solution. It will be interesting to compare their affinity for tumour Hsp90 with that for Hsp90 from normal cells — if they do not retain preferential binding to tumour Hsp90, will they also lose their tumour-cell-specific toxicity? The study by Kamal et al.2 suggests that subverting the natural rules of attraction that determine 17-AAG binding to Hsp90 might prove to be counter-productive. Instead of developing drugs with a higher affinity for Hsp90, what might be needed are compounds that have bigger differences in the affinities with which they bind the two forms of Hsp90.

Kamal and colleagues' findings will undoubtedly affect the future design of Hsp90 drugs, but the study also has general implications for anticancer drug development. It suggests that it is not enough to identify a potential molecular target — the drugs directed against that target must also be assessed in an appropriate cellular context.


  1. 1

    Ferrarini, M., Heltai, S., Zocchi, M. R. & Rugarli, C. Int. J. Cancer 51, 613–619 (1992).

  2. 2

    Kamal, A. et al. Nature 425, 407–410 (2003).

  3. 3

    Ritossa, F. Cell Stress Chaperones 1, 97–98 (1996).

  4. 4

    Maloney, A. & Workman, P. Expert Opin. Biol. Ther. 2, 3–24 (2002).

  5. 5

    Chiosis, G. et al. Chem. Biol. 8, 289–299 (2001).

  6. 6

    Scheibel, T. & Buchner, J. Biochem. Pharmacol. 56, 675–682 (1998).

  7. 7

    Banerji, U. et al. Proc. 93rd Ann. Meet. Am. Assoc. Cancer Res. 43, A1352 (2002).

  8. 8

    Jez, J. M., Chen, J., Rastelli, G., Stroud, R. M. & Santi, D. V. Chem. Biol. 10, 361–368 (2003).

  9. 9

    Rinehart, K. L. & Shields, L. S. Fortschr. Chem. Org. Naturstoffe 33, 231–307 (1976).

  10. 10

    Panaretou, B. et al. Mol. Cell 10, 1307–1318 (2002).

  11. 11

    Riggs, D. L. et al. EMBO J. 22, 1158–1167 (2003).

Download references

Author information

Correspondence to Len Neckers.

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.