Renal cell carcinomas (RCCs) account for 2–3% of all malignancies (1,2). Surgical resection is the first choice of treatment for RCC at an early stage. However, when patients present at an advanced stage or have local recurrence or distant metastasis to other organs, immunotherapy, chemotherapy and radiation therapy are applied, although the response rates are poor. Recently, new molecular targeted agents, such as sunitinib and sorafenib, have been developed and are widely used (3–7). Although these drugs have demonstrated a better clinical response than previously used treatments, serious adverse reactions, such as fatigue, mucositis, hand-foot syndrome, diarrhea and hypertension, are often observed and become the cause of discontinuation of the drug administration.

Monotherapy or combination therapy of interleukin (IL)-2 and interferon (IFN)-α has been relatively widely applied for the treatment of advanced kidney cancers. In Japan, to reduce the risk of adverse reactions, the dose of IL-2 is lower than that used in other countries (8–11). While the response rate of monotherapy is as low as 10–20%, that of combination therapy is slightly better, 20–25% (12–17). However, the rate of adverse reactions associated with this combination therapy appears to be lower than recently developed molecular targeting drugs (18,19). Hence, we aimed to define a subset of patients who expect to show a favorable response to this therapy through gene expression profiles of metastatic RCCs, after enrichment of the cancer cells with laser microbeam microdissection technology. In the present study, two antigen-presentation-associated molecules were identified that may predict response to IL-2 and IFN-α combination therapy for metastatic RCC. In addition, this finding may be useful for improving the drug response rate, for contributing to the improvement of the quality of life and prognosis of patients, and reducing unnecessary medical costs to non-responders.

Laser microbeam microdissection (LMM) was carried out to enrich cancer cell populations from each of 42 tumor tissues from patients that had met the inclusion criteria of this clinical trial as reported previously (21). To attempt to establish a prediction method for clinical responses to IL-2 and IFN-α combination therapy, we analyzed gene expression profiles of microdissected renal cancer cells using an oligoDNA microarray consisting of 38,500 genes or ESTs. Analyzing 10 responders (CR or PR) and 18 non-responders (PD or NC), we identified 24 genes that showed significantly different levels of expression between the responder and non-responder groups according to the two following definitions: i) expression information based on signal intensities higher than the cutoff level in >60% of samples of at least one group; ii) random permutation test P-values <0.0005 (Table III). Among the 24 genes selected, the expression levels of 14 genes were higher in the responder group, including two immunologically important genes, HLA-DQA1 (P=2.83E-05) and HLA-DQB1 (P=2.66E-04), that are known to be HLA class II molecules having critical roles in antigen presentation. Of the remaining 22 genes, three genes have also been implicated in several roles in immunological responses: MAP3K5 (ASK1) is suggested to be related to the natural immunity to the stress response. TFPI is suggested to function as a negative regulator of cytokine expression. In addition, NR4A2 has been identified as a candidate target molecule for the treatment of multiple sclerosis, one of the autoimmune diseases of the central nervous system.

Since IL-2 and IFN-α treatment is expected to enhance patient immunity with the subsequent attack by immune cells on cancer cells, these two HLA class II molecules were considered to likely play key roles in clinical response. Hence, we focused on these two molecules measuring the expression levels quantitatively and comparing them to clinical response. We confirmed the relativity of expression level and clinical response using the receiver operating characteristic (ROC) curve, and the best cutoff value for the expression level of each gene was determined (Fig. 1). The results demonstrated that IL-2 and IFN-α combination therapy showed poor response rates of 85.7 and 83.3%, respectively, when the expression level of HLA-DQA1 or HLA-DQB1 was lower than the cutoff value (Fig. 2). If patients having a lower expression of these genes were excluded, than the rate of effectiveness of this therapy is expected to be 57.1% in the case of HLA-DQA1 and 61.1% in the case of HLA-DQB1, respectively. However, since HLA-DQA1 and HLA-DQB1 are known to form a heterodimer in the antigen presentation process, we hypothesized that if either one was expressed lower than the cutoff level, the tumor cells were unlikely to produce an HLA-class II molecule sufficiently and were then unable to present the antigen(s) effectively. Considering this hypothesis, if this treatment was withheld to the patients whose tumors expressed a lower level of either HLA-DQA1 or HLA-DQB1 than each cutoff value, than the response rate could be expected to reach as high as 66.7% (Fig. 2, Table IV).

To screen genes that may be associated with the clinical response of advanced RCCs to IL-2 and IFN-α combination therapy, oligo DNA microarray analysis was applied in combination with LMM to obtain precise expression profile data of thousands of genes simultaneously. The following random permutation test of expression data identified 24 candidate genes that exhibited significant differential expression between the ‘responder’ and ‘non-responder’ groups (Table III). Since two immunologically important genes, HLA-DQA1 and HLA-DQB1, were included in the set of 24 genes, the relative expression levels were measured and then compared to clinical responses by applying the ROC curve for these two genes (Figs. 1 and 2; Table IV). As HLA-DQA1 and HLA-DQB1 are known to form a heterodimer, we hypothesized that when either gene exhibited a low level of expression in tumors, the antigen presentation from the tumor cells for immunotherapy may be insufficient. When such patients were excluded from this particular treatment, the rate of effectiveness of this therapy was expected to improve to 65–70% (Table IV).

To further apply our prediction system in clinical use, we attempted to establish the quantitative RT-PCR method. As a result, the mRNA levels measured by the quantitative RT-PCR method were found to be quite consistent to those obtained from the microarray analysis with Pearson’s correlation coefficient and Spearman’s rank-correlation coefficient of 0.89 and 0.83 for HLA-DQA1, respectively, and 0.78 and 0.65 for HLA-DQB1, respectively (Table V, Fig. 3).

In conclusion, the expression levels of HLA-DQA1 and HLA-DQB1 are reliable candidate markers for predicting the response to IL-2 and IFN-α combination therapy for RCC and provide useful information for the establishment of personalized treatment.