The Promise of Pharmacogenomics: Gemcitabine and Pemetrexed

Host-specific factors modulate susceptibility to tobaccoinduced carcinogenesis, including variations in DNA repair that may influence the rate of removal of both DNA damage and fixation of mutations. In a seminal study by Wei et al,[1] DNA repair capacity (DRC) was measured in peripheral blood lymphocytes (PBLs) with the host-cell reactivation assay that measures cellular reactivation of a reporter gene from a plasmid damaged by exposure to 75 μM benzo(a)pyrene diol epoxide, and transfected into cultured cells. The plasmids undergo intracellular repair, leading to increased gene expression. Because benzo(a)pyrene-DNA lesions are repaired via nucleotide excision repair (NER), the degree of reporter gene expression is believed to reflect the activity of this repair pathway. The mean level of DRC in lung cancer cases (3.3%) was significantly lower than in controls (5.1%; P < .01). Younger cases (< 65 years old) and smokers were more likely than controls to have reduced levels of DRC.[1] The investigators confirmed these findings in a case-control study of 316 newly diagnosed lung cancer patients and 316 cancer-free controls. Case patients who were younger at diagnosis, female, lighter smokers, or with a family history of cancer exhib ited the lowest DRC, suggesting that these subgroups are especially susceptible to lung cancer.[2] Reduced DRC and increased DNA adduct levels are thus associated with an increased risk of lung cancer. NER is a principal pathway for repair of DNA adducts that are induced by smoking-related carcinogens, and also the primary mechanism for removing cisplatin(Drug information on cisplatin) adducts.[3] The NER molecular machinery includes proteins that are mutated in xeroderma pigmentosum (XP) and Cockayne syndrome patients. In the global genome repair pathway, the protein complex XPC-HHR23B, which appears to be essential for the recruitment of all subsequent NER factors in the preincision complex, binds to damaged DNA. Then, the multicomponent transcription factor TFIIH that is responsible for unwinding the damaged region of the DNA is recruited. Next, XPG nuclease cleaves the DNA on its 3' end. Following DNA cleavage, XPA/RPA proteins join the complex and recruit the ERCC1-XPF complex, which cleaves the 5' end.[3] Polymorphisms of a number of DNA repair genes that are involved in the NER pathway potentially affect protein function and, subsequently, DRC. In lung cancer patients, reduced levels of XPG and CSB expression have been observed in peripheral lymphocytes.[ 4] Moreover, ERCC1 and XPD mRNA levels in lymphocytes correlate with DRC, and could thus be useful surrogates of DRC.[5] Extensive reviews of DRC-related genes have been presented elsewhere.[6-11] Prognostic Factors In a retrospective analysis of 2,531 non-small-cell lung cancer (NSCLC) patients with extensive disease (either distant metastases or locoregional recurrence after definitive radiotherapy) treated in a Southwest Oncology Group (SWOG) trial, Albain et al performed Cox modeling and recursive partitioning and amalgamation (RPA) to determine independent, predictive outcome factors.[12] Patients received treatment between 1974 and 1988. Performance status (PS) was defined as good (SWOG 0-1, no symptoms or with symptoms but fully ambulatory) or poor (SWOG 2-4, nonambulatory). Good PS, sex (female), and age ≥ 70 years were significant independent predictors. In a second Cox model for patients with good PS, hemoglobin levels ≥ 11 g/dL, normal calcium, and a single metastatic site were significantly favorable factors. The use of cisplatin was also an additional independent predictor of improved outcome. An RPA performed in 904 patients from more recent SWOG trials, almost all of whom were treated with cisplatin, revealed three distinct subsets based on PS, age, hemoglobin, and lactate dehydrogenase (LDH) in which 1-year survivals were 27%, 16%, and 6%. Until 1980, additional variables, such as weight loss (< 10 or ≥ 10 pounds), were not included in final analysis. In the multivariate survival analyses by prognostic variables and therapy discriminants, the significantly favorable SWOG predictive factors were PS 0-1, sex (female), age ≥ 70 years, (≥ 45 years in female patients), a single metastatic lesion, < 10 pounds of weight loss, normal LDH, normal alkaline phosphatase, and hemoglobin ≥ 11 g/dL. Median survival was 1 to 3 months better for the good PS, female, single-lesion, and cisplatinbased therapy categories.[12] Intriguingly, LDH was important in the poor PS subset; patients with a normal LDH level and poor PS had a survival outcome similar to other subsets with a good PS. In an analysis of 1,052 patients included in clinical trials conducted by the European Lung Cancer Working Party (ELCWP), Paesman and coworkers used a Cox regression model to determine following predictive variables: Karnofsky PS (≥ 80 = SWOG 0-1; ≤ 70 = SWOG 2-4), neutrophil counts, metastatic involvement of skin, serum calcium level, age and gender, as well as disease extent; patients with stage I-III disease were included in the analysis.[13] An RPA defined the best subgroup of patients as females with limited disease and a Karnofsky PS ≥ 80. In a third, smaller, analysis that included a homogenous group of stage III unresectable or inoperable patients receiving cisplatin at 120 mg/m² plus vinca alkaloid combination chemotherapy, a multivariate analysis by O'Connell and colleagues highlighted the following outcome parameters (and associations): initial PS, with patients with good PS (increased objective response/survival); bone metastases (decreased response rate and survival); elevated LDH and sex (male) (shortened survival); and ≥ 2 extrathoracic metastatic organ sites (shortened survival).[14] When objective response following chemotherapy was included in their analysis, the authors reported a strong association with increased survival. Multiple clinical characteristics influence prognosis, such as PS, hemoglobin level, and bone, liver, or skin metastases. Gender, LDH, and weight loss also influence survival. New pieces of information should shed light on these prognostic factors. First, there are interindividual differences in DRC which are accentuated according to age and gender, with females having a reduced DRC.[2] Based on DRC, females would thus have greater chemosensitivity than males. Also, data indicate that in vitro intrinsic cisplatin resistance is associated with elevated levels of DRC in NSCLC cells.[15] DRC is a surrogate of the NER pathway that eliminates cisplatin adducts,[3,5] and NSCLC patients with effective systemic (host) DRC are reported to achieve poorer survival than patients with suboptimal DRC.[16] In a study (N = 375) by Boskin et al, patients in the top DRC quartile of the group (DRC > 9.2%) had a risk of death exceeding two times that of subjects in the bottom quartile (DRC < 5.8%; P = .01).[16] Median survival was 8.9 months for patients in the top DRC quartile, compared with 15.8 months for those in the bottom quar-tile (P = .04). Earlier findings suggest that the formation and persistence of cisplatin or carboplatin(Drug information on carboplatin) adducts either in buccal cells or leukocytes predict better response.[17,18] Schaake-Kong et al examined cisplatin-DNA adducts in the nuclei of buccal cells in 27 patients with unresectable NSCLC receiving radical radiotherapy and daily administration of low-dose cisplatin.[ 19] Nuclear staining was performed in buccal cells collected 1 hour after cisplatin on the fifth treatment day (after five daily doses of cisplatin at 6 mg/m²). Cisplatin-DNA adduct staining remained a significant independent predictor of survival: patients with low levels of induced DNA adducts had a meager median survival of 5 vs 30 months for patients with elevated DNA adduct levels.[20] Thus, beyond stratification for gender in future clinical trials, measuring DRC by either functional assays or surrogates like ERCC1 or XPD mRNA in peripheral lymphocytes[5] might help to predict treatment responders, as reported previously.[14] Alberola and colleagues reported in a prospective Spanish Lung Cancer Group (SLCG) phase III study including 557 NSCLC patients treated with cisplatin-containing regimens that PS, gender and weight loss were significant prognostic factors.[21] Note that weight loss in lung cancer patients was also previously reported to be predictive of both reduced outcome and survival.[12] Upregulation of Genes in the Glycolysis and Krebs Cycle Pathway Unlike normal mammalian cells that require oxygen in energy-generating pathways, cancer cells rely on anaerobic pathways of glycolysis. Lung cancer patients with weight loss have elevated levels of 3-phosphoglycerate and phosphoenolpyruvate- both components of the glycolytic pathway (glyconeogenesis).[22] Furthermore, c-Myc and HIF-1 overexpression deregulates glycolysis through the activation of the glucose metabolic pathway, which regulates lactate dehydrogenase and induces lactate overproduction.[23] Note that O'Connell et al reported that elevated serum LDH was associated with shortened survival and remission duration.[14] Elevated mRNA levels of phosphofructokinase, glyceraldehyde- 3-phosphate dehydrogenase, phosphoglycerate kinase, and enolase were also reported, indicating the activation of several components in the glucose metabolic pathway.[23] Chen et al[24] reported the systematic identification of lung adenocarcinoma proteins using twodimensional polyacrylamide gel electrophoresis (PAGE) and mass spectrometry, and found that at least four proteins (phosphoglycerate kinase 1, phosphoglycerate mutase, alpha enolase, and pyruvate kinase M1-all components of glycolysis) had increased expression levels. They were also associated with poor survival in resected lung adenocarcinoma.[24] Expression of phosphoglycerate kinase 1, the sixth enzyme of the glycolytic pathway, reflects increased glycolysis in tumor cells, and is related to the induction of a multidrug resistant (MDR) phenotype distinct from MDR1. The hypoxic nature of the solid tumor triggers vascular endothelial growth factor (VEGF) expression that both stimulates angiogenesis and activates glycolytic enzymes, including phosphoglycerate kinase, thus facilitating anaerobic production of adenosine(Drug information on adenosine) triphosphate (ATP).[25] However, cancer cells also maintain high aerobic glycolytic rates and produce high levels of lactate and pyruvate, the end products of glycolysis (Warburg effect); thus, preferential reliance on glycolysis is correlated with disease progression as reported by Lu et al.[26] The selective adaptation of cancer cells to hypoxia is also mediated via HIF-1, which upregulates a series of genes involved in glycolysis, angiogenesis, cell survival, and erythropoiesis. Glycolytic metabolism subsequent to glyceraldehydes- 3-phosphate dehydrogenase leads to pyruvate in cells with active pyruvate kinase.[26] Similarly, elevated expression of hexokinase and phosphofructokinase is a hallmarks of glycolytic pathways of cancer cells. Interestingly, phosphofructokinase activity is stimulated by insulin, insulinlike growth factor-1, and epidermal growth factor receptor (EGFR), all of which induce HIF-1 under normoxia involving the PI3K signaling pathway.[26] Phosphoglycerate mutase and enolase activities have also been found elevated in lung cancer (as opposed to colon, liver, and nonendocrine lung tumors).[27] Accordingly, HIF-1-alpha (and HIF-2-alpha) overexpression is a common event in the natural history of NSCLC, also related to the upregulation of various angiogenic factors and with poor prognosis.[ 28] Interestingly, signal transduction from HER2 to PI3K, Akt, and FKBP-rapamycin-associated protein (FRAP) increases the rate of HIF-1- alpha synthesis, whereas hypoxia and loss of p53 activity decrease the rate of HIF-1-alpha degradation.[29] Akt/ protein kinase B (PKB) is constituitively active in NSCLC cells and promotes chemoresistance (and resistance to radiation-induced apoptosis).[30] Brabender et al reported that in 83 surgically resected NSCLC patients, high levels of HER2-neu mRNA expression as well as high EGFR/HER2- neu mRNA coexpression were significant, independent prognostic factors, ie, their expression defined low- and high-risk groups for treatment failure in curatively resected NSCLC.[31] In Figure 1 we postulate a model illustrating the significance and potentially close relationship of increased mRNA or protein expression of glycolytic genes and HIF-1-alpha/VEGF to clinical outcome, beyond the influence of serum LDH and weight loss. Clearly, this molecular-based model and its hypotheses require validation in clinical trials. A surrogate marker of the cancer glycolytic pathway could be positron emission tomography (PET), whose imaging capability can detect the bio- chemical and physiologic processes that occur in tissues. The most frequently used positron emitting radiopharmaceutical is 18-fluor, labeled 2-deoxy-D-glucose (₁₈F-FDG), a radioactively labeled glucose analog. Because cancer cells presumably exhibit a higher glycolytic rate than nonneoplastic types, the clinical application of ₁₈F-FDG-PET is promising and potentially warranted. It is reasonable that higher tumor uptake of the radiolabeled glucose analog might be a surrogate of the glycolytic pathway. Using this rationale/technology, clinicians have employed PET imaging to assess changes in intracellular tumor glucose utilization during the course of chemotherapy. In the NSCLC arena (n = 57), Weber et al reported that median time to progression (P = .0003) and overall survival (P = .005) were significantly longer for 18FDG6-PET metabolic responders (vs nonresponders) in the interval before and after the first chemotherapy cycle.[32] One-year survival rates were 44% and 10%. Overexpression of ERCC1 mRNA and other NER genes has been associated with repair of cisplatin-induced DNA damage and clinical resistance to cisplatin. In a small study of advanced NSCLC patients (n = 56 stage IIIb/IV) receiving a cisplatin/gemcitabine regimen, Lord and collaborators[ 33] reported that PS, weight loss, and low ERCC1 mRNA expression were independent prognostic factors. Indeed, ERCC1 mRNA levels were more significant factors than PS in a Cox proportional hazards multivariable analysis.[33] Median survival for patients with low ERCC1 mRNA expression was significantly longer (61.6 weeks; 95% confidence interval [CI] = 42.4-80.7 weeks) compared to those exhibiting high expression (20.4 weeks, 95% CI = 6.9-33.9 weeks). Ribonucleotide Reductase and Thymidylate Synthase mRNA Expression Ribonucleotide reductase (RR), which is increased in cancer cells, is another potentially new predictive and prognostic marker. The M2 (or RRM2) subunit is directly involved in many cellular signaling pathways.[ 34] For example, the active chemotherapeutic agent gemcitabine(Drug information on gemcitabine) (Gemzar) decreases ribonucleotide reductase activity. In a retrospective analysis of tissue using quantitative polymerase chain reaction (PCR) analysis (n = 75) as part of a randomized trial, our group reported better time to progression (P = .05) and survival (P = .0028) in gemcitabine/cisplatintreated metastatic NSCLC patients (n = 22) who had low ribonucleotide reductase subunit M1 (or RRM1) mRNA expression; the other study arms received vinorelbine (Navelbine)/ cisplatin (n = 25) and paclitaxel(Drug information on paclitaxel)/ carboplatin (Paraplatin) (n = 28).[35] Interestingly, retinoblastoma protein is sequentially phosphorylated by cyclin D-CDK4/6 and cyclin E/CDK2 during G1/S cell-cycle transition phase. This modification leads to the dissociation of retinoblastoma from E2F/DP heterodimers, leaving them in a transcriptionally active state that regulates several DNA synthesis enzymes also involved in chemotherapy response, such as dihydrofolate reductase (DHFR), thymidylate synthase (TS), and RR.[36] In small-cell lung cancer (SCLC) and NSCLC, p16/ INK4A and retinoblastoma are reciprocally inactivated, resulting in the inactivation of the same p16/INK4A/ RB pathway.[37] Therefore, in prospective studies, the predictive and/or prognostic value of certain transcripts should be kept in mind. For example, in ribonucleotide-dependent chemotherapy combinations, the role of RRM1 and TS mRNA levels might influence response and survival, as proposed in the model illustrated in Figure 1. Several studies have shown that the status of TS is a predictive marker for response to 5-FU in different primary cancers. The pioneering study by Lenz et al[38] reported that TS mRNA levels were associated with response and survival in 65 gastric cancer patients treated with 5-FU and cisplatin. Thymidylate synthase was examined by quantitative PCR (QPCR), with beta-actin gene expression as an internal control. Tumor tissue was obtained from endoscopic biopsies of gastric adenocarcinomas, and TS expression was examined from fresh tissue. The mean gastric cancer TS mRNA level in responding and resistant patients was statistically significant (P < .001). Patients with high TS mRNA levels (> 4.6) had a median survival of 6 months, while for those with lower levels (< 4.6), median survival was not yet reached at 43 months of follow- up.[38] The mean pretherapy TS mRNA level in all 65 patients was 4.6. When TS values were broken down by the ethnicity of the patients, the mean was 4.3 for Hispanics, 5.6 for Asians, 3.8 for Caucasians, and 6.1 for African-Americans. Due to the small number of patients, these differences were not statistically significant, but they provided an indication that racial differences should be taken into account in these and future analyses. Differences in response and survival according to TS levels have also been observed in colorectal, lung, and breast cancer (reviewed elsewhere[ 39]). More recently, TS mRNA levels have been examined from microdissected paraffin(Drug information on paraffin)-embedded tissue samples, facilitating the routine use of QPCR. Since the combination of pemetrexed(Drug information on pemetrexed) (Alimta) plus cisplatin has become the standard treatment in malignant pleural mesotheliomas (also approved by the US Food and Drug Administration for that indication),[ 40] one potential approach to improve therapeutic outcome would be to identify pemetrexed target geneexpression changes in cancerous mesothelial cells. The diagnosis of mesothelioma often requires a thoracoscopic or open pleural biopsy that can provide a rich source of tumor tissue for RNA isolation. TS is the principal marker of pemetrexed sensitivity, although pemetrexed also inhibits dihydrofolate reductase and the purine biosynthetic enzyme glycinamide ribonucleotide formyltransferase. In an MDA 435 breast cancer- derived cell line, Longley et al reported that TS upregulation highly inhibited pemetrexed while the growth inhibitory effects of irinotecan, cisplatin, oxaliplatin, and paclitaxel were unaffected by TS upregulation.[41] (The pemetrexed IC₅₀ dose was unobtainable when TS was overexpressed, indicating that pemetrexed toxicity was highly sensitive to increased TS expression.) In vitro and in vivo studies have also demonstrated a strong association between increased TS expression and the development of resistance to fluorouracil(Drug information on fluorouracil) (5-FU) and raltitrexed (Tomudex). Metzger and coworkers reported that TS mRNA levels and ERCC1 plus TS mRNA levels strongly predicted survival in 5-FU/cisplatin-treated gastric cancer patients (n = 38).[42] When both ERCC1 and TS mRNA levels were below their medians, 85% (11 of 13) of patients responded; with both ERCC1 and TS mRNA levels above, responses were 20% (2 of 10) (P = .003). Other upstream determinants of 5-FU chemosensitivity include the 5-FU-degrading enzyme dihydropyrimidine dehydrogenase and 5-FU-anabolic enzymes, such as orotate phosphoribosyl transferase. However, it is likely that events downstream of TS inhibition, such as activation of DNA damage response pathways (eg, ERCC1), also play key roles in determining the cellular response to TS inhibitors.[42] Customized Chemotherapy In addition to the individual predictive role of ERCC1 mRNA [33] and RRM1 mRNA [35], in a third study, we examined both ERCC1 and RRM1 mRNA expression in pretreatment bronchial biopsies from gemcitabine/ cisplatin-treated advanced NSCLC patients who were part of a large randomized trial.[21] There was a strong correlation between ERCC1 and RRM1 mRNA expression levels (R = 0.4; P < .001). Patients with low RRM1 mRNA levels had significantly longer median survival than those with high levels (13.7 vs 3.6 months; P = .009). There were no significant differences according to ERCC1 mRNA levels, although there was a tendency toward longer median survival among patients who expressed low ERCC1 levels.[43] Finally, in a fourth study of neoadjuvant gemcitabine/cisplatin-treated stage III NSCLC patients, those patients with the lowest RRM1 levels (in the bottom quartile) had a decreased risk of death compared with those in the top quartile (risk ratio = 0.30; P = .033). Median survival for these 17 patients in the bottom quartile was 52 months, whereas the 15 in the top quartile had a median survival of 26 months (P = .018).[44] These observations have been recently confirmed at the preclinical level. From expression profiling of two gemcitabine- resistant NSCLC cell lines, increased expression of RRM1 mRNA was detected in both. Quantitative PCR analysis demonstrated that resistant cells had a greater than 125-fold increase in RRM1 mRNA expression.[ 45] The accumulated evidence indicates that RRM1 may serve as a predictive marker of gemcitabine and cisplatin response. We have also analyzed BRCA1 mRNA expression in processed formalin- fixed, paraffin-embedded tissues from resected lung cancer patients and demonstrated that BRCA1 expression can be accurately assessed. BRCA1 gene expression was detectable in all 55 samples analyzed in this study. Patients in the bottom quartile of BRCA1 mRNA levels (< 0.61) obtained the maximum benefit of neoadjuvant gemcitabine/cisplatin chemotherapy, while those in the top quartile (> 2.45) had the poorest outcome.[ 46] These findings support the hypothesis that BRCA1 mRNA expression levels could be an indicator of differential cisplatin sensitivity in NSCLC, which is consistent with findings in preclinical models in breast cancer.[47] For instance, the HCC1937 cell line from a primary breast carcinoma with a germline BRCA1 mutation was transfected with either wild-type BRCA1 or an empty vector to test response to antimicrotubule drugs (paclitaxel and vinorelbine) and DNAdamaging drugs (cisplatin, bleomycin(Drug information on bleomycin) [Blenoxane], and etoposide(Drug information on etoposide)). Reconstitution of wild-type BRCA1 function into HCC1937 resulted in a 1,000-fold increase in sensitivity to paclitaxel and a 10,000-fold increase in sensitivity to vinorelbine. Conversely, it resulted in a 2-fold increase in resistance to bleomycin, a 20-fold increase in resistance to cisplatin, and a greater than 100-fold increase in resistance to etoposide.[47] Interestingly, BRCA1 failed to modulate resistance or sensitivity to the antimetabolite 5-FU, perhaps reflecting the distinct mode of action of antimetabolites.[ 47] Several cisplatin-based doublets have demonstrated similar survival in several randomized studies of advanced NSCLC. Furthermore, other studies have found no survival differences between cisplatin alone and cisplatin/paclitaxel or between docetaxel(Drug information on docetaxel) [Taxotere] alone and docetaxel/ cis-platin. Based on our results[ 46] and on preclinical data[47], we can reason that patients with low BRCA1 mRNA levels can benefit from gemcitabine/cisplatin, whereas those with high levels could benefit from single-agent docetaxel or paclitaxel. In contrast, high BRCA1 levels may diminish the synergism between the taxanes and cisplatin or carboplatin. While sensitivity to antimetabolites, such as gemcitabine, may not be affected by BRCA1 levels, gemcitabine/cisplatin synergism may be partially abrogated in tumors with high BRCA1 mRNA levels; on the other hand, these tumors may benefit from the synergism observed between the taxanes and gemcitabine. To date, no other clinical study has assessed BRCA1 mRNA expression as a predictive marker of chemotherapy response in lung cancer. If further research validates our findings, BRCA1 mRNA assessment will provide an important tool for customizing NSCLC chemotherapy in order to improve survival. Figure 2 illustrates the schema of the new Spanish Lung Cancer Group customized chemotherapy trial in advanced NSCLC. Conclusions Customized chemotherapy opens new avenues of translational research, where the main difficulty is the availability of tumor tissue for gene expression analyses. In vitro and clinical evidence point out that customized chemotherapy is a promising approach to be validated.