Thymidylate synthase protein expression by IHC and gene copy number by SISH correlate and show great variability in non-small cell lung cancer

Abstract

Introduction

Increased expression of thymidylate synthase (TS) is thought to be associated with resistance to TS-targeting drugs, e.g. pemetrexed.

Methods

TS protein expression (PE) and gene copy number (GCN) were assayed using immunohistchemistry (IHC) and silver in situ hybridization (SISH), respectively, on primary tumors of 189 resected non-small cell lung (NSCLC) patients. Associations with pathological and clinical features and prognosis were explored.

Results

Median IHC H-score was 220 (range 10–380) on a 0–400 scale; 17% of patients had TS expression ≥300. TS PE expression did not significantly differ by histology and did not significantly associate with DFS or OS, however there was a tendency for increased DFS (p=0.12) and OS (p=0.12) in PE positive (>median) squamous-cell carcinoma (SCC) patients. Median GCN was 2.5 genes/nucleus (range 1.4–9.6); 29% of patients had GCN ≥3, 7% ≥4 and 0.8% amplification. GCN differed by histology (p=0.015); 50% of SCCs having GCN >2.5 versus 32% of adenocarcinomas (ADCs). There was no significant relationship between TS GCN and DFS or OS, however a trend towards better DFS (p=0.18) and OS (p=0.10) with increased GCN in SCCs was observed. TS GCN significantly correlated with PE (r=0.30, p=0.0009).

Conclusions

TS PE and GCN vary widely in NSCLC and correlate significantly to each other. TS GCN is higher in SCCs, whereas TS PE does not associate with histological subtypes, clinical features or survival. Variability of TS PE and GCN may indicate potential benefit from pemetrexed therapy in selected SCC patients.

INTRODUCTION

Lung cancer is the leading cause of cancer-related mortality worldwide, causing more than one million deaths annually. Despite advances in surgical techniques, radiation and systemic treatments, prognosis is still very poor, with cure rates around 15% ^{1, 2}.

A promising approach to improve outcomes is customization of chemotherapy according to individual phenotype of the tumor, including its histology or molecular features. Pemetrexed is a cytotoxic agent that targets at least three folate-dependent enzymes, thymidylate synthase (TS), dihydrofolate reductase (DHFR) and glycinamide ribonucleotidyl formyl transferase (GARFT)^{3, 4}. In the landmark phase III randomized trial chemonaïve patients with advanced non-squamous lung cancer (non-SCC) had improved survival following cisplatin and pemetrexed compared to cisplatin and gemcitabine (median of 11.8 vs. 10.4 months; HR=0.81; p=0.005), whereas the opposite was true in patients with SCC (median of 9.4 vs 10.8 months; HR=1.23; p= 0.05) ⁵. Likewise, a randomized phase III study examining maintenance pemetrexed vs. best supportive care following four cycles of platinum-based chemotherapy showed improved survival with pemetrexed in patients with non-SCC histology (HR 0.70; p=0.002), but not in patients with SCC (HR 1.07; p=0.678) ⁶. In consequence, pemetrexed is currently approved in patients with advanced non-SCCs as first-line, second-line, and maintenance therapy. However, while histology is the labeled indication for pemetrexed therapy in NSCLC, more specific molecular markers could potentially give a better prediction for sensitivity or resistance to this agent.

The higher efficacy of pemetrexed in non-SCC vs. SCC is believed to be attributed to differential expression of TS, the main inhibitory target of this agent. In a study involving early-stage NSCLC patients (n=56), TS mRNA and protein levels were significantly lower in the samples of adenocarcinoma (ADC) compared to SCC ⁷. In a study including 11 NLCLC cell lines, resistance to pemetrexed was significantly associated with increasing TS gene expression ⁸. In a phase II clinical trial evaluating a combination of gemcitabine and pemetrexed in stage IB-III NSCLC patients, tumor response was inversely associated with TS expression ⁹. TS gene copy number (GCN) gain evaluated by Ooyama et al. appears to be associated with resistance to antifolates ¹⁰. Amplification of TS gene, as determined by PCR, was shown to better predict outcome than protein expression (PE) assessed by immunohistochemistry (IHC) ¹¹. Thus, TS seems to be a strong candidate predictive marker for pemetrexed sensitivity/resistance, but the optimal assay method remains to be established.

The aims of our study were to profile surgically resected NSCLC for TS PE using IHC and for TS GCN using a new technique, silver in situ hybridization (SISH). Both assays are easily clinically applied and may potentially serve as predictive assays for pemetrexed therapy. We specifically aimed to determine correlation between the two variables and associations with pathological and clinical features including histology, disease-free and overall survival.

MATERIALS AND METHODS

Patient population

Primary tumor samples were obtained from 189 consecutive NSCLC patients initially clinically staged I–III (Union for International Cancer Control, TNM classification of malignant tumours, 6^th edition) who underwent curative pulmonary resection at the Medical University of Gdansk, Poland between 2001 and 2004 (Table 1). Histological classification was verified according to WHO 2004 criteria by a board-certified pathologist. Since positron emission tomography was not available during the collection period, a few patients were upstaged at surgery, thus resulting in a portion of patients with pathological stage III (32%) and stage IV (4%), all of which were included in this study. Induction or adjuvant chemotherapy was used in a small fraction of patients (4% and 2%, respectively), as those treatments were not routinely used per institutional guidelines in the period of this study. Disease-free survival (DFS) was calculated from the date of surgery to the date of relapse (distant or local), death of any cause or last follow-up. Overall survival (OS) was defined as the time between surgery and death of any cause or last follow-up. Median follow-up was 5.3 years (range: 1.1–6.9 years). Institutional review boards from both the Medical University of Gdansk and the University of Colorado approved research conduct for this study.

Table 1.

TS Protein Expression and Gene Copy Number vs Patient Characteristics

	Total Population	TS Protein Expression			TS Gene Copy Number
Characteristics		≤220	>220	p-value	≤2.50	>2.50	p-value
n	189	89	79		64	60
Age, years
Mean (SD)	63 (9)	61 (9)	65 (6)	0.013 0.011	64 (8)	62 (9)	0.147 0.038
≤ 60, n (%)	118 (62)	41 (46)	21 (27)		16 (25)	26 (43)
> 60, n (%)	71 (38)	48 (54)	58 (73)		48 (75)	34 (57)
Gender, n (%)
Female	45 (24)	20 (22)	17 (22)	0.999	10 (16)	11 (18)	0.812
Male	144 (76)	69 (78)	62 (78)		54 (84)	49 (82)
Smoking Status, n (%)
Ever	180 (95)	82 (92)	78 (99)	0.068	60 (94)	0 (0)	0.120
Never	9 (5)	7 (8)	1 (1)		4 (6)	60 (100)
Histology, n (%)
AC	55 (29)	26 (29)	21 (27)	0.168	21 (33)	10 (17)	0.015
SCC	103 (55)	54 (61)	40 (51)		37 (58)	37 (62)
LCC	5 (3)	1 (1)	4 (5)		2 (3)	1 (2)
NOS/Mixed	24 (13)	8 (9)	12 (15)		3 (5)	12 (20)
Missing	2 (1)	0 (0)	2 (2)		1 (2)	0(0)
Pathological Stage, n (%)
I	75 (40)	31 (35)	33 (42)	0.372	22 (34)	26 (43)	0.277
II	42 (22)	25 (28)	15 (19)		14 (22)	17 (28)
III/IV	69 (36)	32 (36)	29 (37)		27 (42)	17 (28)
Missing	3 (2)	1 (1)	2 (2)		1 (2)	0 (0)
Grade, n (%)
G1	20 (11)	10 (11)	8 (10)	0.245	9 (14)	9 (15)	0.174
G2	81 (43)	43 (48)	28 (35)		31 (48)	19 (32)
G3	63 (33)	25 (28)	30 (38)		16 (25)	22 (37)
Missing	25 (13)	11 (12)	13 (17)		8 (13)	10 (17)

Tissue microarray preparation

Tissue microarrays (TMA) were prepared as described previously ¹². Briefly, three cores of 1.5 mm diameter from representative tumor areas were obtained manually from each formalin-fixed paraffin-embedded tumor block. These cores were taken from different areas with different tumor differentiation pattern if heterogeneity was present. Therefore, it is unlikely that choice of particular histological pattern for core selection biases the results of this analysis. Tissue cores were processed using MaxArray customized TMA service (Invitrogen, South San Francisco, CA). Sections of 4 µm were cut from TMA master blocks, mounted on slides and processed as described below.

TS protein expression

TS PE was evaluated using automated IHC on the BenchMark XT autostainer (Ventana Medical Systems, Inc, Tucson, AZ). Tissue sections were heated in a 60°C dry oven for 1 hr prior to transfer to the autostainer where they were then treated with cell conditioning solution 1 for 1 hr. The thymidylate synthase primary antibody (4H4B1, Invitrogen, Carlsbad, CA) was added manually at a 1:10 dilution and incubated for 1 hr at 25°C. The UltraView DAB detection kit was used in conjunction with the amplification kit. Following washing, the slides were counterstained with hematoxylin for 4 min. The slides were removed from the autostainer and washed with mild soapy water and dehydrated in a graded series of ethyl alcohol baths and xylene baths. Positive (colon carcinoma) and negative (colon carcinoma with non-immune IgG as primary antibody) controls were included in all series. A certified pathologist (BRA) scored the slides using the hybrid (H) scoring method: H-score = % cells of 0 intensity + (% cells of 1+ intensity × 1) + (% cells of 2+ intensity × 2) + (% cells of 3+ intensity × 3) + (% cell of 4+ intensity × 4). The intensity score was defined as: 0 is no appreciable staining in the tumor cells, 1+ is barely detectable staining in the cytoplasm and/or nucleus compared with the stromal elements, 2+ is readily appreciable brown staining distinctly marking the tumor cell cytoplasm and/or nucleus, 3+ is dark brown staining in tumor cells obscuring the cytoplasm and/or nucleus, and 4+ is very strong dark staining (Figure 1).

Figure 1.

Representative photomicrographs of the varying range of TS protein expression detected by IHC. A) H-score of 70, B) H-score of 150, C) H-score of 210, D) H-score of 280, E) H-score of 355.

TS gene copy number

TS GCN was assayed using automated silver in situ hybridization (SISH) on the BenchMark XT autostainer (Ventana Medical Systems, Inc, Tucson, AZ). The dinitrophenol labeled TS DNA repeat-free probe, generously provided by Ventana Medical Systems, Inc, is designed against a 1 Mb amplicon on 18p11.32 and optimally formulated for use with ultraView SISH Detection Kit and accessory reagents. The TMA was probed according to the manufacturer’s instructions. Briefly, the TS probe was denatured at 95°C for 12 min and hybridization was performed at 52°C for 4 hrs. Hybridization was followed by 3 stringency washes at 72°C. The probe was visualized using a rabbit anti-DNP primary antibody and a detection kit containing a goat anti-rabbit secondary antibody conjugated with horseradish peroxidase (HRP). The sequential addition of silver A (silver acetate), silver B (hydroquinone) and silver C (hydrogen peroxide) resulted in the silver ions being reduced by hydroquinone to metallic silver atoms as a consequence of HRP oxidation of hydrogen peroxide. The silver precipitate was deposited in the nuclei, and a single copy of the TS gene was visualized as a single discrete black dot, whereas a tight cluster of black dots stacked so closely together that individual signals could not be resolved were considered amplified TS genes. The slide was then counterstained with hematoxilin II and bluing reagent. All tumor specimens were visualized with brightfield microscopy, evaluated for tumor adequacy and scored by a certified pathologist (SS). SISH detection system allows the observer to simultaneously examine TS GCN and the tumor nuclei morphology and count GCN only within tumor nuclei (Figure 2). Signals were counted by scanning the entire core and focusing on regions that appeared to have the highest copy numbers. Fifty non-overlapping nuclei were counted, less if the TMA core was partially depleted, with a minimum of 20 tumor cells required for analysis. Individual signals were given a score of one, small clusters a score of 6 and big clusters a score of 12, similarly to published reports for HER2 gene copy number evaluation in breast cancer ¹³. The scores were analyzed to determine the mean TS GCN per nucleus per core.

Figure 2.

Representative photomicrographs of TS gene copy number detected by SISH. A) 1.8 genes per nucleus, B) 3.2 genes per nucleus, C) 3.7 genes per nucleus, D) 5.0 genes per nucleus, E) 9.6 genes per nucleus.

Statistical analysis

Pearson’s correlation coefficient was used to assess correlation between continuous variables. Associations between categorical variables were explored using Fisher’s exact test. Survival curves were plotted using the Kaplan-Meier method and compared by log-rank test. Cox proportional hazard models were used for multivariate comparisons. All reported p-values were two-sided with a significance level of 0.05, with no adjustment made for multiple comparisons.

RESULTS

TS protein expression

TS PE was evaluable in 168 (89%) of the 189 patient specimens on the TMA with the most common cause for non-evaluability being depleted tissue, followed by lack of tumor tissue within the specimen. No statistically significant difference in the distribution of patient characteristics was observed between the TS IHC evaluable and non-evaluable cases. The average H-score of the triplicate cores per patient was highly correlated to the core with the highest score from the same patient (Pearson’s correlation coefficient 0.97), and therefore further analyses were performed using the maximum H-score. For the entire evaluable population the PE level of TS ranged from a H-score of 10 to 380 with a median of 220, a mean of 219 with a standard deviation of ± 75 (Figure 3A). Tumors from 29 patients (17%) had very high (≥300) expression of TS. The distribution of TS expression was similar in adenocarcinomas (median 220, range 10–350; mean 212, SD 77) and squamous cell carcinomas (median 210, range 50–380; mean 213, SD 72) (Figures 3B and C). A group including large cell carcinomas and not otherwise specified carcinomas tended to have higher values, with a median of 280 (115–365) and mean of 256 ±73 (Figure 3D).

Figure 3.

Histograms showing the distribution, frequency, median, range, mean, and standard deviation for TS protein expression in NSCLC. A) entire evaluable population, B) adenocarcinoma, C) squamous cell carcinoma, D) large cell carcinoma or not otherwise specified histology.

Patients with high TS PE were 3.5 years older on average and were 20% more likely to be over 60 years of age TS (Table 1). No significant associations were found with the other clinical features: gender, histology, stage, grade and smoking status.

The relationship between TS PE and disease-free survival (DFS) as well as overall survival (OS) was evaluated using a H-score of 220, the median for the population, as the cut-off point. Ten patients with an exact H-score of 220 were assigned to the ≤ 220 group (n = 89) patients, and the remaining 79 patients had the H-score of >220. There was no association between TS PE and either DFS (log-rank p=0.26) or OS (log-rank p=0.38) in the entire series (Figures 4A and B). Similarly, when examining PE as a continuous variable and adjusting for age, gender, smoking, stage and grade, there was no relationship between TS and either DFS (HR 1.00 p=0.36) or OS (HR 1.00 p=0.37). However, in patients with SCC histology, while it was not statistically different, there was a tendency for improved DFS (log-rank p=0.12) and OS (log-rank p=0.12) in those who expressed higher levels of TS protein in the tumors (Figures 4C and D). The median DFS for those with high and low TS expression was 1.7 years and 1.1 years, respectively, and the median OS in both groups was 2.8 years vs 1.5 years, respectively. There was no prognostic impact of TS PE for either DFS (log-rank p=0.77) or OS (log-rank p=0.54) in patients with non-squamous NSCLC (Figures 4E and F).

Figure 4.

Kaplan-Meier survival probability curves. Disease-free survival (A) and overall survival (B) according to TS protein expression for the entire evaluable population stratified at the median (H-score 220). Disease-free survival (C and E) and overall survival (D and F) for those with squamous cell carcinoma and non-squamous carcinoma, respectively, stratified at H-score 220.

We also retrospectively combined the 4+ intensity category with the 3+ category and recalculated the H-scores. The new median was 210 instead of 220, 76 of the 79 patients that were in the positive group using the 0–400 scale were also in the positive group using the 0–300 scale. The three that were no longer positive on the 0–300 scale had a score of 210 (new median), but since the cut point was >210 they moved to the negative group. Only 1 patient, of 89, who was part of the negative group (0–400 scale) was added to the positive group (0–300 scale). The H-score for this patient was 220 using both scales. There was still no difference in the overall protein expression between histological subtypes (p=0.29) and no association with DFS (log-rank p=0.38) or OS (log-rank p=0.51) when 3+ and 4+ were combined and the new median (210) was evaluated; in fact there was less separation of the curves (data not shown).

TS gene copy number

TS GCN was evaluable in 124 (66%) of the 189 patient specimens on the TMA; the remaining non-depleted samples had insufficient signals or inadequate tumor cell number for the analysis. The evaluable population was representative of the entire cohort of NSCLC patients and there was no statistical difference between the two subsets. The average GCN of the triplicate cores per tumor was compared to the maximum core of the same tumor. The highest scored core was used for further analyses due to exceptional correlation between the two (r=0.97). In the entire evaluable NSCLC population TS GCN ranged widely from 1.4 to 9.6 genes per nucleus (Figure 5A), with a median of 2.5 genes/nucleus and a mean of 2.7 (SD ±1.0). Tumors from 36 patients (29%) had a GCN ≥3 genes/nucleus, 9 (7%) had a GCN ≥4, and one (0.8%) had true gene amplification (Figure 2, Panel E).

Figure 5.

Histograms showing the distribution, frequency, median, range, mean, standard deviation and amplification for TS gene copy number in NSCLC. A) the entire evaluable population, B) those with adenocarcinoma, C) squamous cell carcinoma, or D) large cell carcinoma or not otherwise specified histology.

In the 31 evaluable ADCs the median and mean TS GCN were 2.2 (range 1.6–4.5) and 2.4 (SD ±0.7), respectively (Figure 5B). Six ADCs showed GCN ≥3, 2 had GCN of ≥4, but there was no gene amplification. Median and mean TS GCN in the 74 SCCs was 2.5 (range 1.4–9.6) and 2.8 (SD ±1.2), respectively (Figure 5C). Twenty-three SCCs had ≥3 genes/nucleus, 7 had a GCN ≥4, and one had gene amplification. In the subset of LCC and NOS cancers (n=19) a median and mean TS GCNs were 2.9 (range 1.5 – 3.4) and 2.8 (SD ±0.6), respectively (Figure 5D). There were no samples with 4 or more TS gene copies, 7 tumors had ≥3 copies and none had gene amplification in this subset of histologies.

TS GCN significantly differed by histology (p=0.015). TS GCN above 2.5 genes/nucleus, the median of the entire evaluable population, was found in 10/31 ADCs, 37/74 SCCs and 13/19 of the combined large cell carcinoma (LCC) and not otherwise specified (NOS) histologies (Table 1). Additionally, TS GCN was higher in tumors from patients younger than 60 years (p=0.038). No associations were found with any of the other clinical features examined: gender, smoking status, stage, or grade.

In the univariate analysis of all patients using the median GCN for the population (2.5 genes / nucleus) as the cut-point, TS GCN was not associated with DFS (log-rank p=0.28) (Figure 6A) or OS (log-rank p=0.40) (Figure 6B). Three cases with 2.5 genes/nucleus were assigned to the group of GCN ≤2.5 (n = 64), and the remaining 60 cases had GCN >2.5. Likewise, no differences were found when TS GCN was considered as a continuous variable and after adjustment for age, gender, smoking status, histology, stage and grade: the HRs for DFS and OS were 0.96 (p=0.77) and 0.94 (p=0.70), respectively. In the SCC subset, similar to the PE findings, there was no significant difference but a tendency for association between high GCN and increased DFS (log-rank p=0.18) (Figure 6C) and OS (log-rank p=0.10) (Figure 6D). Although these associations were not significant, the median DFS was 1.8 yrs vs. 1.4 yrs and the median OS was 4.0 years vs 1.5 years for high vs. low GCN, respectively. No such associations were found for non-SCCs (log-rank for PFS and OS p=0.68 and p=0.55, respectively) (Figures 6E and 6F).

Figure 6.

Kaplan-Meier survival probability curves. Disease-free survival (A) and overall survival (B) according to TS gene copy number for the entire evaluable population stratified at the median (GCN 2.5). Disease-free survival (C and E) and overall survival (D and F) for those with squamous cell carcinoma and non-squamous carcinoma, respectively, stratified at GCN 2.5.

Relationship between TS protein expression and gene copy number

The relationship between TS PE and GCN was examined in the 119 patients for whom both values were available. There was a significant correlation (Pearson’s r=0.30, p=0.0009) between the TS H-score and GCN (Figure 7). This population had one patient with very high GCN and when this outlier was excluded from the analysis the correlation coefficient increased to r=0.38, p<0.0001. In the combined analysis including both TS PE and GCN, DFS tended to be positively associated with increased PE and GCN (median 5.1 years vs. 1.3 years; log-rank p=0.11) and OS (median OS not reached vs. 1.3 yrs, log-rank p= 0.05; data not shown).

Figure 7.

Dot plot showing the correlation between TS protein expression, represented by the H-score, and TS gene copy number.

DISCUSSION

This study is to our knowledge the first report on TS PE analyzed and compared to TS GCN in the primary tumors from a large consecutive series of surgically treated NSCLC patients. We demonstrated similar distribution of TS PE in tumors with SCC and ADC histology, but higher GCN by SISH in SCC, although, there was a great variability in this subtype of NSCLC. Neither the TS PE nor GCN correlated significantly to the prognosis. Furthermore, this study is the first to evaluate the SISH as a new method for TS GCN, and the study demonstrated this assay to be easily clinically applied. The technology is today used for HER2 GCN assessment ¹⁴ and TS SISH could be a potential assay for prediction of response and outcome to pemetrexed.

The lack of TS PE association with histology in our study is in contrast with some other published data showing higher TS PE in SCCs as compared with ADCs ⁷. However, similar to our findings, one study using automated quantitative fluorescence microscopy (AQUA) found no association to histology for TS protein expression levels ¹⁵ and another study using IHC likewise found no statistical association ¹⁶. Association with histology has been documented for TS mRNA level ^{7, 17} and, taken together with in vitro data ¹⁸, has provided the hypothesis for higher pemetrexed efficacy in patients with ADCs due to lower levels of TS expression. However, these aforementioned studies also showed that a proportion of SCCs have low TS protein level, similar to our findings.

Our study found no association between TS PE and DFS or OS in the analysis of the whole population. However, for patients with SCC histology, there was a separation of the survival curves and a trend for improved DFS and OS with higher TS PE, which may reach statistical significance in a larger cohort. The prognostic significance of TS PE in operable NSCLC patients remains debatable. One study showed a positive prognostic impact of high TS protein expression determined by AQUA, when stratified at the 25 percentile ¹⁵. However, in that study TS protein expression was not associated with OS when dichotomized at the median. Other studies focusing exclusively on ADC in Japanese patients showed association between high TS PE and high proliferative indices, and adverse impact of this feature on prognosis^{16, 19–21}. Shintani et al showed that, while low TS mRNA associated with improved DFS, there was no association between DFS and protein expression ²². Inconsistent results of the above studies may be attributed to prognostic differences of TS PE according to histology or ethnicity, differences in clinical cohort characteristics and/or lack of standardized assay methods and data analysis. We cannot definitively explain our finding of a non-significant trend for better outcome of the subset of SCC patients with higher TS levels. This finding may be of chance alone. It is also possible that the biological significance of TS is of less importance for tumor proliferation and prognosis than other molecular events, such as protein over expression and/or gene amplification of other genes.

We quantified our PE using the previously published H-score method on a 0–400 scale ^{12, 23–30}, while other studies with other markers have used the range of 0–300³¹. We are well aware of the potential differences between 0–300 and 0–400 score, but by using the broader semi-continuous 0–400 scale we are consistent with our previous work published on this and other cohorts ^{12, 23–30}. However, we did combine 3+ with 4+ and recalculate the H-scores. There was no difference between histologies with regard to protein expression and there was no association with DFS or OS; in fact there was less separation of the curves on the 0–300 scale compared to the 0–400.

The current study demonstrated a broad range of TS GCN by SISH, with increased values in a substantial proportion of early stage tumors. In 29% of the evaluable tumors a mean GCN was ≥3 genes/nucleus, and in 7% GCN was ≥4. Thus, considering that in non-tumor disomic tissue the typical average GCN is in the order of 1.8 genes per nucleus due to nuclear truncation ¹³, increased TS GCN appears frequent in NSCLC. Furthermore, we showed that ADCs tended to have lower levels of TS GCN, as might be expected given the results of some earlier studies using TS mRNA levels or protein expression ^{7, 15}. Importantly, however, in many SCCs there were also low levels of TS GCN, suggesting that a proportion of SCC patients may potentially benefit from pemetrexed therapy if low TS GCN is considered a predictive marker for sensitivity.

Gene amplification in cancer cells is one of the most frequent mechanisms leading to protein overexpression. Other investigators have shown that amplification in the loci mapping to TS correlates with mRNA levels by qRT-PCR and protein expression in a variety of tumor xenografts or colon cancer samples ¹⁰. Our results for TS GCN are consistent with the previous mRNA studies, which demonstrated increased mRNA expression in SCCs ^{7, 17}. Closer analysis of Ceppi et al. data ⁷ shows that while TS expression was generally higher in SCCs, there were still a proportion of SCCs with low mRNA expression and negative TS immunohistochemical staining. In another study including stage I resected NSCLC, the median TS mRNA expression in SCC was twice that of ADC ¹⁵. In the latter study the range of expression according to histology was not given and therefore the proportion of SCCs with low mRNA expression levels is unknown.

In the present study TS GCN either divided by the median cut-off point or considered as a continuous variable was not found to be prognostic for DFS or OS in the entire population of early stage NSCLC patients. Furthermore, in patients with SCCs increased TS GCN had a tendency to associate with increased survival, whereas this was not the case in the subset of ADCs and in other histologies. However, as these results were not statistically significant, it is unclear whether this trend occurred by chance or reflects a true biological relationship that might be verified in a larger sample size.

Aberrations in TS GCN are thought to result in gene overexpression and resistance to drugs that target TS. TS amplification, as determined by qPCR, was observed in lung cancer cell lines with acquired resistance to pemetrexed but not in the parental sensitive cell lines ⁸. Likewise, intrinsic resistance to pemetrexed in NSCLC cell lines correlated with increased TS gene expression ^{8, 32}. In freshly explanted tumor cells from multiple cancer types including lung, low levels of TS mRNA expression significantly correlated with sensitivity to pemetrexed ³³. SNP array assays and gene dosage analysis in 7 lung cancer and 20 other cancer xenografts showed that gains in the 18p11.32 locus mappings to TS correlate with increased resistance to pyrimidine analog based drugs ¹⁰. These drugs target TS but via a mechanism different than pemetrexed. In a clinical study examining neoadjuvant gemcitabine plus pemetrexed, tumors with low TS mRNA expression were more sensitive to chemotherapy ⁹. Taken together, these results suggest that increased TS GCN in tumors may be involved in resistance mechanisms to TS-targeting drugs.

In the current study a number of specimens were not evaluable for TS GCN due to insufficient signals. One of the reasons for failure to determine GCN includes nuclear truncation, which results in signal loss during the sectioning process. Another reason may be that our specimens, like many reported in the literature, were collected over several years and inevitability were fixed under variable conditions. This might have deleterious effects on the assay, even though substantial optimization was done prior to performing the experiment. Fixative or other processing conditions might have resulted in variable denaturation of the genomic DNA and thus affect hybridization efficiency. Necrotic tissue, which is unsuitable for hybridization due to degraded DNA, was intentionally avoided during construction of the TMA. However, some contamination with necrotic tissue, which is typical for lung tumors, cannot be fully prevented.

When the relationship between TS PE and GCN was examined, there was a significant correlation between the TS H-score and GCN and the correlation increased substantially when the one patient with very high GCN was excluded from the analysis. Our data shows that there are a number of patients with normal (disomic) GCN that have high to very high PE. Suggesting that the increase in PE in these particular patients may be due to a mechanism other than an increase in GCN. Less common, but still observed, are those tumors with increased GCN (>3) but modest protein expression. Hereto suggesting that in some instances PE may be regulated by mechanisms in addition to GCN. These data are highly consistent with our other published data comparing GCN and PE for additional genes, for example EGFR²⁷, MET³⁰ and IGF1R¹² in NSCLC and IGF1R²⁴ in SCLC. In these previous studies the results were similar in that some cases had normal GCN but elevated PE and a few number of cases with increased GCN but low PE.

In conclusion our study showed that TS PE and GCN vary widely in NSCLC and that there is a significant correlation between PE and GCN. We observed no association of TS PE with histology or prognosis. TS GCN is frequently increased in surgically resected primary NSCLCs and differs by histology, with low GCN in most ADCs and a large variability of GCN in SCCs. We did not find any significant association between TS GCN and outcome. However, a broad range of TS GCN, particularly in SCCs, raises some questions not addressed in the phase III studies that evaluated pemetrexed as first-line and maintenance therapy. Future prospective studies should address the question of potential benefit from pemetrexed therapy in a subset of SCC patients with low TS protein levels or low GCN because only then can one definitively determine the suitability of PE and/or GCN for predicting response and/or outcome to pemetrexed therapy. We also demonstrated that a new assay method for TS GCN assessment i.e. TS SISH, could easily be applied to the clinical setting. The potential predictive value of this assay for pemetrexed therapy warrants further validation.