Our results show a clear correlation between NQO1 levels and sensitivity to 17-AAG as expected [5]. Although NQO1 activation is considered necessary for 17-AAG activity [3], even OSCC cell lines without detectable NQO1 showed considerable sensitivity to 17-AAG, with IC50 concentrations around 1μM. Although this might lead one to overestimate the promise of 17-AAG as a chemotherapeutic option for OSCC, one must be mindful of the limitations of this in vitro study. Firstly, we have not measured the negative side effects attributed to the drug. Indeed, a recent clinical trial of 17-AAG found that the severity of the side effects outweighed the clinical benefit to patients with solid tumours [13]. However, this clinical trial did not investigate NQO1 levels in the enrolled patients, nor has any trial testing 17-AAG done so, to the best of our knowledge. This may have important implications as we describe below. Secondly, there are indications in the literature that cultured cell lines may express higher levels of NQO1 than lung and colon patient tumour tissue [14]. We have not been able to directly compare NQO1 levels in cell lines and OSCC tumour tissue, but in vivo expression may well not be as high as that observed in the cultured cell lines.
The findings of this report suggest that if patients could be stratified on the basis of NQO1 protein levels, then OSCC patients expressing NQO1 could potentially benefit from administration of low doses of 17-AAG, possibly in combination with other chemotherapeutics. This is because NQO1 positive patients would likely be responsive to much lower concentrations of the drug. The low dose of 17-AAG would limit the extent of toxic side effects experienced, as observed in clinical trials, where at the six lowest doses administered (6- 125 mg/m2), only one out of 20 patients experienced dose-limiting toxicity, compared with eight out of fifteen patients on the two highest doses (175- 225 mg/m2) [13]. Since severe hepatotoxicity resulting from 17-AAG treatment is reported to be due to metabolism by a different family of reductases [4], this is unlikely to correlate with NQO1 expression, although this would need to be confirmed in vivo. Furthermore, very low concentrations of 17-AAG would likely have minimal effect on normal cells, even those expressing NQO1, due to their much lower reliance on HSP90. However, there is a clear need for further in vivo testing to confirm that the presence or absence of NQO1 does not affect hepatotoxicity, and that severe side-effects can be mitigated by administration of sufficiently low doses.
An alternative possibility may be the approach proposed by Karkoulis and co-workers [15] for the treatment of bladder cancer. These authors propose that the negative side effects of BA chemotherapeutics (in this case geldanamycin) may be mitigated by orthotopic administration of drug. In the case of OSCC, similar to bladder cancer, the tumour site is relatively accessible; therefore an orthotopic delivery may also be feasible. This would allow exposure to dosages that effectively target the tumour, without reaching systemic concentrations that cause hepatotoxicity.
We noticed that 17-AAG treatment resulted in a dose-dependent decrease in endogenously and exogenously expressed NQO1. This is similar to the effect reported by Gaspar and co-workers [16] who suggested that this down-regulation of NQO1 by 17-AAG may play a role in acquisition of resistance to the drug. We found that there was no down-regulation of NQO1 mRNA levels (Additional file 2: Figure S2), suggesting a post-transcriptional mechanism of control. It is not clear what this mechanism may entail, since disruption of HSP90 activity affects a wide range of cellular functions. Although there is no evidence supporting a direct interaction between NQO1 and HSP90, NQO1 levels are reported to depend heavily on FAD levels [17], which may be disrupted by HSP90 inhibition.
It was interesting to note that the absence of detectable NQO1 in two of the cell lines (WHCO1 and WHCO6) could not be accounted for by the presence of the C609T SNP, but rather seemed to correlate with low expression of the NQO1 gene. Further investigation in tumour samples could shed light on whether this accurately reflects NQO1 expression in patients, or whether it is an artefact of a subset of cultured cell lines. The possibility therefore exists that expression of NQO1 could be induced in these two cell lines under particular environmental circumstances, such as those which may be experienced in cells of a solid tumour, e.g. the presence of reactive oxygen species or hypoxia. We postulate that due to the possibility of induction of the gene in a tumour setting, it will be necessary to specifically investigate NQO1 protein levels in biopsies, in order to estimate potential sensitivity to 17-AAG. This could be done using protein detection (Western blot/immunohistochemistry), or using an NQO1 enzyme activity assay. However, the SNP could be used as a rapid test to exclude patients with a TT genotype, who would not express NQO1 and would therefore be poor candidates for 17-AAG treatment.
The relevance of NQO1 levels in the clinical setting has been discussed by Siegel et al. [18]. The authors make the point that NQO1 levels and activity may not remain stable over the course of the treatment, limiting the predictive value of a protein assay, and supporting use of the SNP as a better biomarker of 17-AAG responsiveness. If the SNP were used as a biomarker for responsiveness, patients with the homozygous null mutation, who will certainly not express active NQO1 could easily be excluded from 17-AAG treatment. While SNP analysis could provide a relatively simple tool for elimination of non-expressors, some patients with the wild-type genotype may also express low levels of the protein, and also be less sensitive to 17-AAG treatment. Thus we propose that 17AAG may still hold promise as a chemotherapy, under certain conditions. These include that the drug either be administered orthotopically, or at low concentrations, using the C609T SNP as a screen to exclude non-expressors of NQO1 who would be poor responders.