Lustgarten, Daniel E. Schwed MD*; Deshpande, Charuhas MD†; Aggarwal, Charu MD, MPH†; Wang, Liang-Chuan PhD†; Saloura, Vassiliki MD‡; Vachani, Anil MD†; Wang, Li-Ping MD†; Litzky, Leslie MD†; Feldman, Michael MD†; Creaney, Jeanette PhD§; Nowak, Anna K. MD, PhD§‖; Langer, Corey MD†; Inghilleri, Simona PhD¶; Stella, Giulia MD, PhD¶#; Albelda, Steven M. MD†
Malignant pleural mesothelioma (MPM) is a rare, deadly cancer associated with exposure to asbestos. Without treatment, patients with MPM have a median survival of approximately 9 months. In a landmark phase III trial by Vogelzang et al.,1 pemetrexed (Pem) and cisplatin improved outcomes compared with cisplatin alone, with commensurate improvement in progression free survival. Overall survival (OS) improved by 11 weeks when compared with cisplatin alone. This doublet has become the standard of care when systemic therapy is used.
Pem is an antifolate agent that inhibits multiple folate-dependent enzymes. The main target of Pem is thought to be thymidylate synthase (TS); to a lesser extent, glycinamide ribonucleotide formyltransferase, dihydrofolate reductase and 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase are also considered targets.2 As a folic acid analog, Pem enters the cell through the reduced folate carrier and the folate receptor-α. Once inside the cell, its activity is dependent on glutamation. Pem is poly-glutamated by the key enzyme folyl-polyglutamate synthase (FPGS). This step increases the intracellular retention of Pem and the affinity for its intracellular targets by 50- to 100-fold.2
Although the use of Pem in combination with platinum has fostered a major advance in the treatment of MPM, the objective radiological response rate is only approxi mately 40%.1 Hence, the majority of patients who receive the drug are exposed to its toxic side effects (and consi derable costs) without radiographic or clear clinical benefit. Therefore, finding biomarkers that will identify the population of patients most likely to benefit from Pem is an important goal as we enter the era of personalized therapeutic oncology.
In vitro data support a connection between TS expression and Pem response. In colon, breast and lung cancer cell lines, higher gene and protein expression of TS has been associated with resistance to Pem.3,4 Induction of resistance to Pem in a colon cancer cell line was associated with TS overexpression.5 Three phase III clinical trials have shown Pem to be more efficacious in patients with adenocarcinoma of the lung as opposed to those with squamous cell histology.6–8 TS expression has been found to be lower in adenocarcinoma than squamous cells.9 In addition, low TS levels have been shown to predict better outcomes in patients with non-squamous non–small-cell lung cancer treated with Pem.10
Three recent retrospective studies showed a statistical association between TS expression measured by immunohistochemistry (IHC) and response to Pem in patients with MPM.11–13 The later study also showed that patients with high FPGS had better tumor responses and improved disease control rate. However, the clinical utility of measuring TS expression or FPGS to select patients for treatment with Pem has yet to be verified in prospective clinical trials.
The goal of our retrospective study was to validate these findings in a well-characterized set of MPM patients treated with Pem. As it is postulated that TS and FPGS have an inverse relationship with regard to Pem sensitivity (decreased TS levels have been proposed to predict response to Pem therapy,14 although increased FPGS levels have been proposed to predict sensitivity to Pem therapy),15,16 we also tested whether the ratio of FPGS/TS expression could predict clinical responses to Pem in patients with MPM.
PATIENTS AND METHODS
Patient samples were obtained from three sites: the Fondazione IRCCS Policlinico San Matteo in Pavia, Italy (69 cases); the Sir Charles Gairdner Hospital in Perth, Australia (12 cases); and the Hospital of the University of Pennsylvania in Philadelphia (four cases). Patients were included if they had a diagnosis of MPM, received at least one cycle of a frontline Pem-containing chemotherapy regimen, and had tissue available for immunohistochemical analysis. Patients receiving adjuvant or neoadjuvant Pem in the setting of surgical resection were excluded. All patients were followed with a computed tomography of the chest every 3 months during treatment and up to 2 years after chemotherapy was completed. Patients who survived greater than 2 years had imaging every 6 months.
The primary end point was the association between TS, FPGS, and FPGS/TS expression measured by IHC and: (1) time to progression (TTP), (2) OS, and (3) best-achieved radiographic response. Radiographic response was measured according to the modified response evaluation criteria in solid tumors (Response Evaluation Criteria in Solid Tumors) for MPM.17 Patients were classified as having “disease control” if the best response was complete response, partial response, or stable disease. Otherwise, they were classified as having progressive disease.
IHC was performed on formalin-fixed, paraffin-embedded tissue. Samples were first deparaffinized with xylene and rehydrated in serial dilutions of ethanol. Antigen unmasking was performed by heat-induced epitope retrieval method with a 10 mM sodium citrate solution buffered at pH 6.0. The samples were then incubated with the primary antibody [antithymidylate synthase Ab clone TS106 (Invitrogen, Carlsbad, CA) at a dilution of 1:100 for 30 minutes at room temperature, and an anti-FPGS monoclonal Ab18 provided by Eli-Lilly and Company (Indianapolis, IN) at a dilution of 1:1500 at 4°C overnight. Appropriate treatment with 3% hydrogen peroxide and DAKO serum-free protein block were performed before incubation with the primary antibody. DAKO antimouse horseradish peroxidase polymer solution was used for the detection of the primary antibody and diamino benzidine was used as substrate. The germinal centers of normal human tonsils were used as a 3+ positive control for TS (Fig. 1A) and normal human kidney (Fig. 1C) was used as a 3+ positive control for FPGS. To assess nonspecific staining, negative controls were stained with isotype control monoclonal antibodies for the anti-TS (Fig. 1B) and anti-FPGS antibodies (Fig. 1D) then processed as above. Both antibodies showed high specificity on immunoblotting (i.e., see Fig. 4, FPGS data not shown). As they were monoclonal antibodies, blocking peptides were not available.
To determine the H score, the intensity of the cytoplasmic and nuclear staining was compared with a strong positive control and given an intensity score: 0 – no staining, 1 – mild intensity, 2 – moderate intensity, 3 – intensity equal to or greater than the positive control. The percentage of cells with each intensity score was estimated. The H score was obtained by adding the product of the intensity score multiplied by the percentage of cells staining at that intensity, thus the H score ranged from 0 to 300.19 All samples were scored by a single lung pathologist (C.D.) blinded to the patient’s clinical outcome. The H-score was verified by a second pathologist (L.L.) in the first 40 cases and showed high intraobserver concordance.
In Vitro Experiments with Cell Lines
Nine human mesothelioma cell lines were examined for sensitivity to Pem using a colorimetric MTT assay. To perform MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide] assays to estimate cell numbers, cells were plated in triplicate on 96-well plates (5000 cells per well in their respective culture medium) and were treated with various concentrations (ranging from 0.001 to 100 µM) of pemetrexed. Cell viability was assessed 72 hours after treatment by developing the reaction assay per manufacturer’s instruction (Promega, Madison, WI). The IC50 values of Pem (the concentration of Pem inhibiting tumor cell growth by 50% in comparison with the untreated control) in our human mesothelioma cell lines were determined from these MTT results.
The expression level of TS and FPGS was determined by immunoblotting. Human mesothelioma cell lines were grown in T75 tissue culture flask until they reached 90 to 100% confluence. The cells were then harvested and lysed in ice-cold lysis buffer (50 mM Tris–HCL, pH 7.4, 150 mM NaCL, 5 mM ethylenediaminetetraacetic acid, 1% Nonidet P-40, and protease inhibitor cocktail tablets [Roche Diagnostics, Indianapolis, IN] with 1 mM phenylmethyl sulfonyl fluoride). Equal amounts of protein from whole cell lysate from each sample were electrophoresed on 4 to 12% NuPAGE Novex Tris Acetate Mini Gels (Invitrogen) and transferred onto nitrocellulose membranes. Monoclonal anti-β-actin antibody (purchased from Sigma, St. Louis, MO). The Anti-TS antibody (clone TS106) was used at a concentration of 2 µg/ml (1:100 dilution). The FPGS antibody (see above) was used at a concentration of 0.25 µg/ml (1:5000 dilution).
The human OK4, OK5, OK6, and OK7 MPM cell lines were provided by Dr. Claire Verschaegen (University of Vermont). The I-45 and MSTO-211H MPM lines were purchased from the American Tissue Type Collection (Manassas, VA). The human Pt108, REN, LRK, and M30 MPM lines were derived from patient samples at the University of Pennsylvania.
OS was calculated from the time the first cycle with a Pem-containing regimen was initiated to the time of death. TTP was calculated from the time the first cycle was initiated to the time disease progression and was documented according to modified Response Evaluation Criteria in Solid Tumors criteria. OS and TTP were evaluated using standard Kaplan–Meier curves, and a log-rank test was used to compare survival between groups. Hazard ratios were obtained using a Cox proportional hazards regression model. A χ2 test was used to analyze the association between categorical variables. STATA (version 11; STATA Corp LP, College Station, TX) was used for all analysis, and statistical significance was set at a p value of 0.05 or lesser for all tests.
Eighty-five patients with MPM who were treated with Pem were included; 75 received Pem plus a platinum agent and 10 received Pem alone. Samples with fewer than 100 malignant cells were excluded from the analysis. The characteristics of the study population are summarized in Table 1.
The clinical outcomes are summarized in Table 1 and are similar to other studies.1,12 The overall disease control rate was 69%, progressive disease was documented in 31%. Median TTP was 6.8 months. Median OS time was 13 months.
The effects of clinical characteristics on outcomes are also shown in Table 1. Gender was not predictive of TTP or OS. A nonepithelioid histology was associated with shorter survival (hazard ratio: 3.03; 95% confidence intervals: 1.64–5.62; p < 0.001) but not with shortened TTP. Age above 65 years was associated with shorter TTP (p = 0.02), but not with worse survival.
Relationship of Thymidylate Synthase to Clinical Outcomes
Of 85 patients, 78 samples were available for TS staining and analysis (some sections were not included due to technical reasons). TS staining was mostly nuclear with variable cytoplasmic expression (Fig. 1). In virtually all patients, some degree of staining was seen with a range of H-scores from 10 to 190. The level of TS expression detected by IHC followed a normal distribution (Fig. 2A). The median H-score was 105.
We plotted the H-score from patients with disease progression versus disease control. As shown in Figure 2C, there was extensive overlap between groups with both having the same median score of 105.
We next used the median H-score of 105 to classify samples as having “high” or “low” TS expression (Table 2). When comparing high TS levels to low TS levels, we observed no differences in disease control rates (DCR) (p = 0.73) (Table 2). We performed the same analysis in the subgroup of patients with epithelioid histology and found similar results. Although there was a numerically longer median TTP (8 versus 6 months; p = 0.93) and OS (14 versus 11.25 months; p = 0.59) in the patients with low TS scores compared with higher TS scores these differences were not statistically significant. The Kaplan–Meier curves are shown in Figure 3A,B (on-line only).
Relationship of Folyl-Polyglutamate Synthase to Clinical Outcomes
Of 85 patients, 74 had adequate samples available for FPGS analysis due to technical reasons. FPGS was detected in almost all patients (72 out of 74), again with a normal distribution of H-scores (Fig. 2B). Staining was both nuclear and cytoplasmic (Fig. 1C). The median H-score was 120 (ranges, 0–285).
We plotted the H-scores from patients with disease progression versus disease control. As shown in Figure 2D, extensive overlap existed between the groups: the median H-score was 120 in patients with disease progression and 130 in those with disease control.
Using the median H-score of 120, samples were categorized as having “high” or “low” FPGS expression (Table 2). We did not find a significant association between high and low FPGS expression and median TTP (p = 0.77), median OS (p = 0.43), or DCR (p = 0.95). The Kaplan–Meier curves are shown in Figure 3C,D (on-line only).
The ratio of FPGS/TS was calculated by dividing the FPGS H-score by the TS H-score for each patient. The data appeared normally distributed. The median ratio was 1.17 and this value was used to classify samples as having high or low FPGS/TS ratios (Table 2). In patients with a high FPGS/TS ratio there was a higher TTP (7.5 versus 6 months; p = 0.10), OS (13 versus 12 months; p = 0.43), and DCR (76.3 versus 64.9%; p = 0.27), but these differences were not statistically significant.
Correlation of TS and FPGS Expression with In Vitro Chemosensitivity of Human Mesothelioma Cells to Pemetrexed
Given the lack of correlation between TS expression and the clinical response to Pem, we tested whether such a correlation existed in in vitro. Nine human mesothelioma cell lines were treated with varying concentrations of Pem, ranging from 0.001 to 100 µM, and cell viability was assessed 72 hours after treatment. The 50% inhibitory concentration (IC50) values were then determined in each cell line and correlated with the TS expression as determined by immunoblotting, (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/JTO/A388 and Fig. 4). Consistent with our clinical data, we did not observe any correlation between TS expression and chemosensitivity of mesothelioma cells to Pem in vitro. We performed similar in vitro studies evaluating FPGS expression by immunoblotting and again observed no correlation with IC50 values of Pem for each cell line (data not shown).
The clinical data supporting the relationship between TS and Pem sensitivity has largely been indirect; based on relatively small retrospective studies that showed an association between lower levels of TS expression and Pem efficacy in mesothelioma and lung cancer.20 Three recent observational studies11–13 demonstrated a statistical association between intratumoral TS expression and Pem efficacy in patients with MPM. Table 3 compares these studies to ours. In all studies, TS levels were semi-quantitatively measured by IHC using the same H-score metric that we used. In the Righi study,11 the median H-score was 90 and when used as a cutoff, a significant correlation between TS levels and both TTP and OS was seen. In this study, TS but not ERCC1 expression was associated with TTP and OS. No data on response rates were provided. It should be noted that the overall survival of the entire group (20.9 months) was much longer than the 13 months reported in the large phase III Pem trial by Vogelzang,1 the Zucali study,12 and our cohort, suggesting the selection of a “healthier” and likely less representative cohort of patients than is usually treated. In the Zucali study,12 the median H-score was very low (a value of 21). The progression-free survival (PFS) and OS survival were 7 and 13 months respectively, with a disease control rate of 78%, much more similar to our data and to other data cited in the literature. When the median H score was used as a cutoff, statistically significant differences in disease control rates (92 versus 73%), median PFS (7 versus 6 months) and OS (18 versus 9 months) were observed based on TS levels. In the Christoph study,13 the median H-score was very high (a value of 210). Their OS was much longer (23.3 months) compared with other studies, but similar to that seen in the Righi study. They found that median TS was associated with statistically significant differences in OS, but not PFS, objective response rate (OR), or DCR.
Our median H-score was 105, similar to that seen by Righi, but clearly different from that appreciated by Zucali or Christoph11–13 (Table 3). We observed an overall DCR of 69%, a median TTP of 6.8 months, and a median OS of 13 months (outcomes similar to those of Zucali and previous studies, and the established literature.1 Although the median TTP (8 versus 6 months) and OS (14 versus 11 months) were longer in our low TS versus high TS group, these differences did not reach statistical significance. Other cutoffs, beside the median score, were also analyzed (including the highest tertile and highest quartile), but statistical significance was still not reached. The reason why we did not observe statistically significant differences in our study is not entirely clear. One potential explanation could be differences in antibodies or technique. We view this as unlikely for the TS results, as all four studies (see Table 3) used the same anti-TS antibody clone (TS106) and three of the four studies used the antibody at the same dilution (1:100). It is possible we would have reached significance with larger numbers of patients.
Given this limitation in using TS expression alone, we also explored the utility of assessing the expression of FPGS. In vitro studies15,16 have suggested that loss of FPGS activity may lead to Pem resistance in cell lines because of decreased intracellular availability of the active poly-glutamated form of Pem. A recent study suggests there may be some predictive value in FPGS levels in MPM.13 In that study, the median FPGS H-score was 230 – a relatively high value because 300 is the maximal score. High FPGS (greater than median) expression was associated with response rate and DCR, but not OS. However, as shown in Table 2, we detected no association between FPGS expression and Pem efficacy. Of note, our median H-score was much lower at 120. We also determined if a higher FPGS/TS ratio (i.e., high(er) FPGS with low(er) TS) would correlate with improved outcome on Pem. We noted a trend towards increased TTP (p = 0.10) with higher FPGS/TS ratios, but not with OS (p = 0.43). Again, other cutoffs beside the median score were analyzed, but statistical significance was not achieved. Consistent with our clinical findings, we did not find an association between thein vitro Pem sensitivity (measured as the IC50 value) of several mesothelioma cell lines with TS or FPGS expression as measured by immunoblotting. One explanation for the discrepant results may be that we used an antibody different from that used in the Christoph study13.
The main limitation of this study was its retrospective observational design. Although stage is predictive of survival, this information was not available in the majority of patients and therefore could not be included in our survival analysis as a possible confounder. We also did not have baseline data on performance status or weight loss. The retrospective nature of our study made measurements of radiographic response and TTP less accurate because these were obtained as part of routine clinical care and not as part of a formal study protocol. Although we limited our analysis to patients who received a Pem-containing regimen in the first-line setting, some patients went on to receive second line therapies after disease progression; although benefits from second line therapy have never been proven in randomized phase III studies, it is conceivable that subsequent treatment could have influenced OS. We were also careful to exclude patients who received “adjuvant” chemotherapy for similar reasons, because these patients, by and large, were likely to have earlier stage disease, better overall survival, and no “measurable” tumor after resection.
Even with these limitations in mind, the most important implication of this study relates to the issue of “statistical” versus “clinical” utility of a test. Specifically, even if TS levels ultimately prove to have a statistical association with response and survival in patients with MPM treated with Pem, it seems doubtful this association would have any true clinical utility. This translates to a broader clinical question: should TS levels be used to decide whether Pem should be administered to a specific MPM patient? This is an important issue, because a number of companies are already actively marketing tests to clinicians to measure TS mRNA or protein expression levels in tumors with the obvious, but unstated, implication that this information should be used in therapeutic decision-making. In addition, many clinicians have already made the “leap of faith” that TS levels do, in fact, predict benefit to Pem, both in MPM and in NSCLC, and are using these tests accordingly.
We believe our data indicate that TS expression levels should not be used in such decisions. First, the measurement of TS using IHC is inherently imprecise and rather subjective, even for trained academic lung pathologists. Each of the three studies found a different median H-score value for TS ranging from 21 (an extremely low value) to 210 (an extremely high value). It is unclear which value should be chosen as a cutoff. Carefully standardized positive and negative controls must be used with each staining run. Even with these controls, detecting small differences is difficult. Reliably and reproducibly differentiating an H-score of 90 (a “low” value) and 130 (a “high” value) is challenging (Fig. 1). As a second example, in the Christoph study,13 the median FPGS H-score for responders was 255 and for non-responders was 220, again a very small difference to discern. Secondly, and perhaps even more importantly, even if an H-score can be reliably determined, the sensitivity and specificity of the test appears to be poor. This is shown in Figure 2, where we have plotted individual H-score values versus disease control status. Clearly, the overlap between groups precludes any meaningful prediction at the individual patient level. Accordingly, there were some patients with very low H-scores who did not respond to therapy, just as there were some patients with very high H-scores whose disease was controlled. This key representation of the frequency distributions of H-scores has not been shown in any of the other manuscripts.
In summary, although other data suggest that TS and FPGS expression might be potential markers of Pem efficacy in patients with MPM in a “statistical setting”, our study indicates that these markers lack sufficient predictive value in an individual patient to be used to guide clinical decisions. In fact, basic demographic factors, such as histology and age, are more predictive. Before TS and FPGS expression are used to make treatment selections, additional data are needed. Ideally, a prospective observational study with predetermined diagnostic criteria and standardized treatment, outcome measurements and follow-up would need to be performed. If positive, a prospective trial randomizing patients to treatment based on TS and/or FPGS levels would need to be conducted to show increased efficacy. Until then, we do not recommend the use of these markers as determinants of therapy in the clinical setting.
The authors thank Guanjun Cheng, MD, Jennifer Wheaton, BS, Jing Sun, MD, and Patrizia Morbini MD PhD. This work was partially supported by an NIEHS funded Environmental Health Sciences Core Center grant P30-ES013508 (S.M.A.) and NHLBI T32 HL07586 (D.S.L.).
1. Vogelzang NJ, Rusthoven JJ, Symanowski J, et al. Phase III study of pemetrexed in combination with cisplatin versus cisplatin alone in patients with malignant pleural mesothelioma. J Clin Oncol. 2003;21:2636–2644
2. Rollins KD, Lindley C. Pemetrexed: a multitargeted antifolate. Clin Ther. 2005;27:1343–1382
3. Giovannetti E, Backus HH, Wouters D, et al. Changes in the status of p53 affect drug sensitivity to thymidylate synthase (TS) inhibitors by altering TS levels. Br J Cancer. 2007;96:769–775
4. Longley DB, Ferguson PR, Boyer J, et al. Characterization of a thymidylate synthase (TS)-inducible cell line: a model system for studying sensitivity to TS- and non-TS-targeted chemotherapies. Clin Cancer Res. 2001;7:3533–3539
5. Sigmond J, Backus HH, Wouters D, Temmink OH, Jansen G, Peters GJ. Induction of resistance to the multitargeted antifolate Pemetrexed (ALIMTA) in WiDr human colon cancer cells is associated with thymidylate synthase overexpression. Biochem Pharmacol. 2003;66:431–438
6. Scagliotti G, Hanna N, Fossella F, et al. The differential efficacy of pemetrexed according to NSCLC histology: a review of two Phase III studies. Oncologist. 2009;14:253–263
7. Scagliotti G, Brodowicz T, Shepherd FA, et al. Treatment-by-histology interaction analyses in three phase III trials show superiority of pemetrexed in nonsquamous non-small cell lung cancer. J Thorac Oncol. 2011;6:64–70
8. Ciuleanu T, Brodowicz T, Zielinski C, et al. Maintenance pemetrexed plus best supportive care versus placebo plus best supportive care for non-small-cell lung cancer: a randomised, double-blind, phase 3 study. Lancet. 2009;374:1432–1440
9. Ceppi P, Volante M, Saviozzi S, et al. Squamous cell carcinoma of the lung compared with other histotypes shows higher messenger RNA and protein levels for thymidylate synthase. Cancer. 2006;107:1589–1596
10. Sun JM, Han J, Ahn JS, Park K, Ahn MJ. Significance of thymidylate synthase and thyroid transcription factor 1 expression in patients with nonsquamous non-small cell lung cancer treated with pemetrexed-based chemotherapy. J Thorac Oncol. 2011;6:1392–1399
11. Righi L, Papotti MG, Ceppi P, et al. Thymidylate synthase but not excision repair cross-complementation group 1 tumor expression predicts outcome in patients with malignant pleural mesothelioma treated with pemetrexed-based chemotherapy. J Clin Oncol. 2010;28:1534–1539
12. Zucali PA, Giovannetti E, Destro A, et al. Thymidylate synthase and excision repair cross-complementing group-1 as predictors of responsiveness in mesothelioma patients treated with pemetrexed/carboplatin. Clin Cancer Res. 2011;17:2581–2590
13. Christoph DC, Asuncion BR, Mascaux C, et al. Folylpoly-glutamate synthetase expression is associated with tumor response and outcome from pemetrexed-based chemotherapy in malignant pleural mesothelioma. J Thorac Oncol. 2012;7:1440–1448
14. Hanauske AR, Eismann U, Oberschmidt O, et al. In vitro chemosensitivity of freshly explanted tumor cells to pemetrexed is correlated with target gene expression. Invest New Drugs. 2007;25:417–423
15. Liani E, Rothem L, Bunni MA, Smith CA, Jansen G, Assaraf YG. Loss of folylpoly-gamma-glutamate synthetase activity is a dominant mechanism of resistance to polyglutamylation-dependent novel antifolates in multiple human leukemia sublines. Int J Cancer. 2003;103:587–599
16. Cho RC, Cole PD, Sohn KJ, et al. Effects of folate and folylpolyglutamyl synthase modulation on chemosensitivity of breast cancer cells. Mol Cancer Ther. 2007;6:2909–2920
17. Byrne MJ, Nowak AK. Modified RECIST criteria for assessment of response in malignant pleural mesothelioma. Ann Oncol. 2004;15:257–260
18. Dotzlaf J, Carpenter J, Luo S, et al. Derivation and characterization of monoclonal antibodies against human folypolyglutamate synthetase. Hybridoma (Larchmt). 2007;26:155–161
19. Ishibashi H, Suzuki T, Suzuki S, et al. Sex steroid hormone receptors in human thymoma. J Clin Endocrinol Metab. 2003;88:2309–2317
20. Chen CY, Chang YL, Shih JY, et al. Thymidylate synthase and dihydrofolate reductase expression in non-small cell lung carcinoma: the association with treatment efficacy of pemetrexed. Lung Cancer. 2011;74:132–138
Thymidylate synthase; Malignant pleural mesothelioma; Pemetrexed