Journal of Cancer Research and Clinical Oncology

, Volume 137, Issue 2, pp 201–210

Thymidilate synthase expression predicts longer survival in patients with stage II colon cancer treated with 5-flurouracil independently of microsatellite instability

Authors

  • Marisa Donada
    • ACADEM Department, Cattinara HospitalUniversity of Trieste
  • Serena Bonin
    • ACADEM Department, Cattinara HospitalUniversity of Trieste
    • International Centre for Genetic Engineering and Biotechnology
  • Ermanno Nardon
    • ACADEM Department, Cattinara HospitalUniversity of Trieste
    • International Centre for Genetic Engineering and Biotechnology
  • Alessandro De Pellegrin
    • ACADEM Department, Cattinara HospitalUniversity of Trieste
  • Giuliana Decorti
    • Department of Biomedical SciencesUniversity of Trieste
    • ACADEM Department, Cattinara HospitalUniversity of Trieste
    • International Centre for Genetic Engineering and Biotechnology
Original Paper

DOI: 10.1007/s00432-010-0872-1

Cite this article as:
Donada, M., Bonin, S., Nardon, E. et al. J Cancer Res Clin Oncol (2011) 137: 201. doi:10.1007/s00432-010-0872-1
  • 151 Views

Abstract

Background

5-Fluorouracil (5-FU) is the most commonly used therapeutic agent for colon cancer treatment. Several studies have evaluated in patients with colon cancer, either the role of genes involved in the 5-FU pathway, such as thymidylate synthase (TS), thymidine phosphorylase (TP) and dihydropyrimidine dehydrogenase (DPD) or the role of microsatellite instability (MSI) as prognostic or predictive markers for adjuvant chemotherapy efficacy, with discordant results. In this study we investigated the combined effect of TS, TP, DPD mRNA expression and MSI status in primary tumors of patients with colon cancer, all treated with 5-FU adjuvant therapy.

Methods

TS, TP and DPD expression levels were investigated by real-time quantitative RT-PCR on RNA extracts from formalin-fixed and paraffin-embedded tissues of 55 patients with colon adenocarcinoma. In the same case study MSI status was assessed on DNA extracts.

Results

A higher TS expression was significantly associated with a longer survival for patients with cancers of stage II (P < 0.01), but not for those with stage III (P = 0.68). In addition, in multivariate analysis, a higher TS expression was significantly associated with a decreased risk of death (HR 0.13, 95% CI 0.03–0.59, P < 0.01), while the MSI status did not have effects on patients’ survival.

Conclusions

This retrospective investigation suggests that TS gene expression at mRNA level can be a useful marker of better survival in patients (especially of those with cancers of stage II) receiving 5-FU adjuvant chemotherapy, independently of the MSI status.

Keywords

Colon cancerAdjuvant chemotherapy5-fluorouracil pathway genesMicrosatellite instabilityFormalin-fixed and paraffin-embedded tissues

Introduction

Colorectal cancer is the third most frequent tumor and the third leading cause of cancer-related death in United States (Jemal et al. 2008). In Italy, as recorded by the Italian tumor registries, it is the fifth most frequently diagnosed cancer among men and the third among women and it is the fourth cause of cancer-related death among both men and women (Tumori and Rapporto 2006).

It has been reported that two major distinctive molecular pathways are related to colon cancer development: the chromosomal instability (CIN), which is related to ~80% of all sporadic colon cancers; and the microsatellite instability (MSI), which accounts for a minor part of the total cases. Recently, a third possible pathway related to serrated adenomas has been reported (East et al. 2008). Cancers with CIN are characterized by gross chromosomal rearrangements, aneuploidy and loss of heterozygosity. Conversely, cancers showing MSI display an increased rate of replication errors in the highly repetitive short DNA sequences called microsatellites. MSI is indeed the consequence of the failure of the human DNA mismatch repair system (MMR), which, in sporadic cancers, is nearly always due to the hypermethylation of the hMLH1 gene promoter (Cunningham et al. 1998). Cancers referred to CIN or MSI pathway display some differences. From a pathological point of view, cancers exhibiting MSI are usually sided in the right colon, they are diploid, poorly differentiated and often with lymphocytic infiltrates, whereas cancers with CIN are left-sided, aneuploid, highly differentiated and often without lymphocytic infiltrates. From a clinical point of view, patients with MSI cancers have a more favorable prognosis and a lower sensitivity to the 5-fluorouracil (5-FU) agent than stage-matched patients with cancers exhibiting CIN (Soreide et al. 2006). Five-FU-based chemotherapy seems indeed to give a survival benefit to patients with tumors exhibiting CIN, but apparently not to those with tumors showing MSI (Ribic et al. 2003), even though discordant results are reported (Hemminki et al. 2000; Storojeva et al. 2005).

Five-FU represents the principal therapeutic agent used against colon cancer, both in mono- or poly-chemotherapeutic regimen. It is a uracil analog, which irreversibly inhibits the thymidylate synthase (TS) enzyme, which catalyzes the conversion of dUMP to dTMP, the latter being necessary for DNA synthesis. The cytotoxic effect of this drug occurs through the action of FdUMP, a 5-FU metabolite competing with dUMP for binding to TS protein. The key enzyme responsible for 5-FU conversion to FdUMP is the thymidine phosphorylase (TP), which has been shown to be identical to the molecule platelet-derived endothelial cell growth factor (PD-ECGF) (Usuki et al. 1992) and has exhibited angiogenic properties in colorectal tumors (Takebayashi et al. 1996). The cytosolic dihydropyrimidine dehydrogenase (DPD) enzyme catalyzes the conversion of 5-FU to 5-fluoro-dihydrouracil. More than 80% of an administered dose of 5-FU is eliminated by the catabolism via DPD. For their role in 5-FU mechanism of action, these three enzymes (TS, TP and DPD) have been studied, individually or in reciprocal association, as prognostic or predictive markers of therapeutic response in patients with colorectal cancer (Ciaparrone et al. 2006; Lassmann et al. 2006; Soong et al. 2008). At the present time, the principal investigations are about TS gene and its protein expression. Although different clinical studies indicate an inverse correlation between the intratumoral TS protein or mRNA expression and patients’ survival (Ciaparrone et al. 2006; Tsourouflis et al. 2008), in other studies, opposite results have been observed (Inoue et al. 2005). The DPD levels have also been shown to correlate with possible response to 5-FU (Jensen et al. 2007; Tsuji et al. 2004). Several studies have indeed suggested a better survival in 5-FU-treated patients with tumors showing low DPD levels (Ciaparrone et al. 2006; Jensen et al. 2007). On the other hand, it has also been observed that patients with DPD deficiency can suffer severe toxicity after 5-FU treatment (van Kuilenburg et al. 2003). The meaning of TP expression is more difficult to grasp, because of its pro-angiogenetic activity besides its function in 5-FU mechanism of action. For its angiogenic properties, an elevated expression of this gene is generally predictive of poor clinical outcome (Metzger et al. 1998).

TS, TP, DPD and MSI status may have prognostic and predictive significance to outcome, which may explain some of the discrepancies in the results among different studies. For this reason, we decided to simultaneously analyze all of them, together with the MSI status. These genes were chosen because they are the key enzymes involved in 5-FU mechanism of action, and this suggests a potential role for them in the prediction of the response to the drug. In addition, it is not still well clarified whether the microsatellite instability could be related to their deregulation.

We simultaneously evaluated the expression levels of TS, TP and DPD, by means of real-time quantitative RT-PCR, in a retrospective case study of 55 patients with colon adenocarcinomas treated with 5-FU for adjuvant therapy. These data were subsequently integrated with the MSI status of the tumors and the clinical-pathological features.

Materials and methods

Patients and specimens

FFPE cancer tissues were obtained from 55 patients whose diagnosis was primary sporadic colon adenocarcinoma. These patients were selected among 536 patients, diagnosed with colorectal cancer and treated at the Trieste University Hospital between 1998 and 2000. Of these, only those patients diagnosed at an age of less than 80 years with exclusively colonic tumors and in stages II and III were selected. Moreover, only those patients for whom clinical follow-up and archival material were available were included in the study. From each patient, normal distal colon tissue was also recovered.

None of the patients had received any anticancer treatment prior to surgery, whereas after surgery all patients were treated with 5-fluorouracil-based adjuvant chemotherapy. The regimen was the following: 5-fluorouracil (370–420 mg/m2) plus leucovorin (20–200 mg/m2) on day 1–5, repeated every 4–5 weeks for 6 months. The adjuvant therapy to stage II patients was based on a mutual accordance between the oncologist and the patients. These patients were submitted to adjuvant treatment because of “high risk” to develop recurrences, presenting one of the following characteristics: inadequately sampled lymph nodes, T4 lesions, perforation or tumors with a poorly differentiated histology (Benson et al. 2004).

The cohort of patients was followed through the population-based Friuli–Venezia Giulia Cancer Registry from diagnosis of cancer to death or until 31 December 2007, whichever came first.

DNA isolation and MSI analysis

DNA was extracted from FFPE tissues of each patient’s tumor and distal normal mucosa. The areas of extraction were identified on a reference H&E-stained section by a pathologist (G. S.) and then mechanically microdissected, ascertaining the presence of adequate neoplastic or normal tissue. The dissected specimens were deparaffinized with xylene, and then DNA was extracted according to a proteinase K digestion and phenol/chloroform protocol as previously reported (Pauluzzi et al. 2004).

For MSI determination, PCR amplification was performed according to the panel of five microsatellite markers (BAT25, BAT26, NR-21, NR-27 and NR-24) as previously recommended (Buhard et al. 2006; Suraweera et al. 2002). PCR products from paired normal and tumor samples were analyzed on a 10% polyacrylamide gel by silver nitrate staining. Tumors were classified as MSI-high (MSI-H, presence of at least three markers showing allelic imbalances) or non-MSI-H (presence of one or no markers with allelic imbalances).

RNA isolation and qRT-PCR

Total RNA was extracted from FFPE specimens of primary colon adenocarcinomas after microdissection, as described elsewhere (Stanta et al. 1998). For each sample, 8 μg of total RNA were DNase digested for 15′ at 25°C in 20 μl final volume containing 5 U of DNAse I (GE Healthcare, USA) and 1 × DNase buffer (GE Healthcare, USA). The enzyme was blocked with 2 μl of 25 mM EDTA and heat inactivated at 65°C for 10′. DNase-treated RNA was reverse-transcribed into cDNA. The RT reaction was performed using Moloney Murine leukemia virus (M-MLV) reverse transcriptase and random hexamers as reported elsewhere (Nardon et al. 2009). Expression levels of TS, TP and DPD were measured by real-time quantitative RT-PCR. RNA isolated from a colon cancer cell line (HT29) was used as positive control, whereas RNA isolated from eight normal colon mucosa were pooled and used as calibrator.

The choice of the proper reference gene was performed analyzing the expression levels of four candidates housekeepings (ß-actin, glyceraldehyde-3-phosphate dehydrogenase-GAPDH, ß2-microglobulin and hypoxanthine phosphoribosyl transferase-HPRT), by means of real-time PCR, in eight paired malignant and normal samples selected from the case study. Primer sequences were created using the PRIMER3 software, and they are reported in Table 1. Amplification was performed using a Mastercycler® ep realplex (Eppendorf, Hamburg, Germany). All samples were run in duplicate using the RealMasterMix SYBR ROX 2.5x (5Prime GmbH, Hamburg, Germany) according to the manufacturer’s instructions. For each PCR reaction, 40 ng of cDNA was used in a final volume of 20 μl. Cycling conditions were as follows: 1 min 30 s at 95°C for polymerase activation and 40 cycles of denaturation for 30 s at 95°C, primer annealing for 30 s at specific annealing temperature, extension for 30 s at 72°C and fluorescence detection for 20 s. The detection temperature was set very close to that of amplicon’s melting (Table 1), in order to avoid the detection of aspecific products. Uniqueness of amplification products was checked by melting curve analysis and by 10% polyacrylamide gel electrophoresis. TS, TP and DPD expression levels were normalized against the chosen housekeeping gene (among the four analyzed) and expressed in relation to a RNA pooled from eight normal colon tissues. Relative quantification was performed according to a ∆∆Ct model as previously reported (Pfaffl 2001).
Table 1

Primer sequences and analytical parameters

Gene

Primer sequences

Gene bank accession No

Product size (bp)

Tfa (°C)

TS

 Sense

5′-CACATCGAGCCACTGAAAAT-3′

NM_001071.1

69

78.5

 Antisense

5′-CAAAGCTCAGGATTCTTCGA-3′

TP

 Sense

5′-ATCACAGCCTCCATTCTCAGT-3′

NM_001953.2

69

80.0

 Antisense

5′-GAACTTAACGTCCACCACCA-3′

DPD

 Sense

5′-GCCTATTCCTACCATCAAGGA-3′

NM_000110.3

69

75.0

 Antisense

5′-CTCAATTCACCAAATGTTCCA-3′

GAPDH

 Sense

5′-CCCTCAACGACCACTTTGTCA-3′

NM_002046

75

80.0

 Antisense

5′-GGTCCACCACCCTGTTGCT-3′

β-Actin

 Sense

5′-CGGCCCCTCCATCGT-3′

NM_001101.3

66

80.0

 Antisense

5′-AAAGGGTGTAACGCAACTAAGTCAT-3′

β2-Micro-globulin

 Sense

5′-AGTTAAGTGGGATCGAGACA-3′

NM_004048.2

74

77.5

 Antisense

5′-GGAATTCATCCAATCCAAAT-3′

HPRT

 Sense

5′-GTGTCATTAGTGAAACTGGAAAAGCA-3′

NM_000194.2

91

76.5

 Antisense

5′-CGATGTCAATAGGACTCCAGATGTT-3′

Tfa temperature of fluorescence acquisition

Statistical analysis

The selection of the reference gene for normalization was made by analyzing the raw expression data obtained by real-time PCR with “NormFinder” (Andersen et al. 2004) and “GeNorm” (Vandesompele et al. 2002) algorithms.

The relationship between TS, TP and DPD expression levels, clinical-pathological parameters and MSI status was analyzed by using the Chi-square or the Fisher exact test. Real-time qRT-PCR normalized values were dichotomized with respect to the median value. In particular, tumors with gene expression levels lower or higher than the median value of the analyzed gene were classified as low or high status of expression, respectively.

Kaplan–Meier test was performed to evaluate the faint effect of the covariates on patients’ survival. Statistical significance was assessed by long-rank test. Cox proportional hazard method was applied to evaluate the independent predictive effect on survival of TS, TP and DPD expression levels, MSI status, age at the diagnosis, sex, tumor location, stage and histological grade. Statistical significance was considered as P < 0.05. Statistical analyses were performed with the Stata/SE 9.0 package (Stata, College Station, TX).

Results

Fifty-five patients with colon cancer were included in the study: 28 patients were men and 27 women with an average age of 62.9 ± 9.1 years (±SD; range 35–79 years). The tumors were staged according to the AJCC classification (AJCC 1997). Thirty-one cancers (56.4%) were of stage II and 24 (43.6%) of stage III. Sixteen were right-sided tumors, 5 were located in transverse colon and 34 were left-sided. Regarding tumor differentiation, 21 specimens were classified as G1, 31 as G2 and 3 as G3. The 55 cases had no other cancer, had been diagnosed at an age less than 80 years old and their clinicopathological details are reported in Table 2. The patients were followed up until 31 December 2007 or death, whichever came first. During the follow-up period, nineteen patients (34%) developed metastasis and sixteen patients (29%) died for disease progression. One patient was lost in 1999 after 1 year of follow-up.
Table 2

Relationship between clinicopathological features, MSI status, TS, TP and DPD expression levels and development of metastasis in the follow-up period

Variables

Num: 55

No relapse (n = 36)

Relapse (n = 19)

P

Agea

 ≤63

25

18/36

7/19

0.35

 >63

30

18/36

12/19

Sex

 Male

27

16/36

11/19

0.34

 Female

28

20/36

8/19

Stage

 II

31

22/36

9/19

0.33

 III

24

14/36

10/19

Histological grade

 G1

21

12/36

9/19

0.72

 G2

31

22/36

9/19

 G3

3

2/36

1/19

Location

 Right sided

16

9/36

7/19

0.62

 Transverse

5

4/36

1/19

 Left-sided

34

23/36

11/19

MSI statusb

 MSI-H

7

5/36

2/19

0.54

 Non-MSI-H

48

31/36

17/19

TSc

 Low

28

14/36

14/19

0.01

 High

27

22/36

5/19

TPc

 Low

28

18/36

10/19

0.85

 High

27

18/36

9/19

DPDc

 Low

28

20/36

8/19

0.34

 High

27

16/36

11/19

aAge Dichotomization according to median value

bMSI status: MSI-H: High-degree microsatellite instability; Non-MSI-H: Absence of microsatellite instability

cTS, TP and DPD expression levels were defined high or low with respect to their median value as described in the Material and Methods section

The median duration of follow-up was 7.6 years (25th–75th percentile = 3.7–8.5 years).

MSI status determination and relationship with clinical-pathological features

To determine the MSI status, the five mononucleotide markers NR-21, NR-24, NR-27, BAT25 and BAT26 were examined in each patient included in this study. Of the 55 patients, 7 were classified as MSI-H (about 13%), while the remaining 48 were included in the non-MSI-H group.

The associations between the MSI status and clinical parameters, such as age, sex, stage and histological grade were analyzed as reported in detail in Table 3. The presence of a high-degree of microsatellite instability was not associated with age at the diagnosis, sex, tumor stage or histological grade. However, when analyzing the association between MSI status and tumor location we found that 6 out of 7 of the MSI-H patients had a tumor in the right colon, versus 10 out of 48 of the MSI-negative patients (P < 0.01).
Table 3

Relationship between MSI status and clinical parameters

Variables

Num: 55

MSI status

MSI-Ha (n = 7d)

Non-MSI-Hb (n = 48)

P

Agec

 ≤63

25

3/7

22/48

0.60

 >63

30

4/7

26/48

Sex

 Male

27

3/7

24/48

0.52

 Female

28

4/7

24/48

Stage

 II

31

3/7

28/48

0.36

 III

24

4/7

20/48

Histological grade

 G1

21

2/7

19/48

0.48

 G2

31

4/7

27/48

 G3

3

1/7

2/48

Location

 Right sided

16

6/7

10/48

<0.01

 Transverse

5

0/7

5/48

 Left-sided

34

1/7

33/48

aMSI-H: High-degree microsatellite instability

bNon-MSI-H: Absence of microsatellite instability

cAge: Dichotomization according to median value

dAll patients with MSI-H but one were younger than 70 years of age. The older patient was aged 71

Choice of the reference gene

The choice of the gene to use as reference for normalization was performed by evaluating the expression levels of 4 candidate housekeepings genes (ß-actin, GAPDH, ß2-microglobulin and HPRT) in a sample set of eight paired normal and tumoral samples taken from the case study. The proper reference gene is the one characterized, in the sample set under analysis, by the highest expression stability. “NormFinder and GeNorm” algorithms evaluate the expression stability by a stability index (the lower the value the higher the stability). GAPDH resulted the best reference gene according to both “NormFinder” and “GeNorm”, as reported in Table 4.
Table 4

Candidate reference genes for normalization and their expression stability values as calculated by “NormFinder” and “GeNorm” algorithms

Gene

Stability value according to NormFinder

Stability value according to GeNorm

GAPDH

0.168

1.02

β-Actin

0.383

1.27

β2-Micro-globulin

0.491

1.32

HPRT

0.675

1.70

A low stability value as an estimate of the combined intra- and intergroup variation of the respective gene corresponds to a high expression stability of the respective gene between the matched malignant and non-malignant samples

Relationship between clinical-pathological features and TS, TP and DPD expressions

TS, TP and DPD expression levels were investigated in the cohort of 55 patients with colon adenocarcinoma. For all the analyzed genes, there were no significant differences related to patients’ sex and age, tumor location, stage, histological grade and MSI status as reported in detail in Table 5, even though a trend toward an association between higher levels of TS and MSI-H was observed (P = 0.05).
Table 5

Relationship between TS, TP and DPD expressions and clinical features

Variables

Num: 55

TSa

TPa

DPDa

Low (n = 28)

High (n = 27)

P

Low (n = 28)

High (n = 27)

P

Low (n = 28)

High (n = 27)

P

Ageb

 ≤63

25

12/28

13/27

0.69

12/28

13/27

0.69

14/28

11/27

0.49

 >63

30

16/28

14/27

16/28

14/27

14/28

16/27

Sex

 Male

27

13/28

14/27

0.69

17/28

10/27

0.08

12/28

15/27

0.35

 Female

28

15/28

13/27

11/28

17/27

16/28

12/27

Stage

 II

31

14/28

17/27

0.33

17/28

14/27

0.51

19/28

12/27

0.08

 III

24

14/28

10/27

11/28

13/27

9/28

15/27

Histological grade

 G1

21

13/28

8/27

0.24c

12/28

9/27

0.53c

11/28

10/27

0.96c

 G2

31

14/28

17/27

15/28

16/27

16/28

15/27

 G3

3

1/28

2/27

1/28

2/27

1/28

2/27

Location

 Right sided

16

7/28

9/27

0.20d

7/28

9/27

0.43d

10/28

6/27

0.59d

 Transverse

5

1/28

4/27

4/28

1/27

 

7/28

3/27

 Left-sided

34

20/28

14/27

17/28

17/27

 

16/28

18/27

MSI status

 MSI-H

7

1/28

6/27

0.05d

4/28

3/27

0.52d

5/28

2/27

0.23d

 Non-MSI-H

48

27/28

21/27

 

24/28

24/27

23/28

25/27

aTS, TP and DPD expression levels were defined high or low with respect to their median value as described in the Material and Methods section

bDichotomization according to median value

cThe P-value was determined excluding the G3 histological grade patients from the analysis. This was done because of their small number

dThe statistical significance of the association was determined by the Fisher’s exact test

Recurrence and survival analysis

MSI status and TS, TP and DPD expression levels were associated with the development of recurrences and survival. As reported in Table 2, the presence of a high-degree of microsatellite instability or gene expression levels of TP and DPD were not associated with the development of metastasis during the follow-up period. On the contrary, TS expression was significantly higher in patients who did not recur in the follow-up period (P = 0.01). In particular, most of the patients showing a high TS level (22 of 27) were clustered among non-recurrent patients, while of the 19 recurrent patients 14 displayed a low TS level in their primary tumor.

Patients’ overall survival was also investigated during the follow-up period. The effect of molecular variables, MSI status and conventional clinical parameters was studied in detail by log-rank test. Among the reported variables TS seemed to be directly correlated with patients’ survival. The group of patients with a high expression status of TS, indeed, showed a higher overall survival in comparison with those characterized by low TS status (P < 0.01) (Fig. 1a). In particular, 89% of patients characterized by high TS survived within the first 5 years of follow-up versus the 54% of those showing low TS levels. The other molecular variables (TP, DPD and MSI status) did not seem to be associated with patients’ overall survival (P = 0.16, P = 0.72 and P = 0.51, respectively) (Fig. 1b, c, d).
https://static-content.springer.com/image/art%3A10.1007%2Fs00432-010-0872-1/MediaObjects/432_2010_872_Fig1_HTML.gif
Fig. 1

Kaplan–Meier survival curves in patients with colon cancer according to their TS (a), TP (b) and DPD (c) expression levels (straight line: low expression; dotted line: high expression) and their MSI status (d; straight line: no MSI; dotted line: presence of MSI). P value according to the log-rank test is reported. Censoring dates are reported on the curves by cross-marks

In the analyzed group of patients, clinical parameters were not correlated with patients’ overall survival.

Cox regression analysis confirmed the results obtained with the log-rank test for TS as reported in detail in Table 6. In particular, for TS the hazard ratio was 0.15 (95% confidence interval, 0.03–0.76), indicating that higher expression levels of this gene are associated with a better prognosis. According to the multivariate analysis of prognostic factors, sex, age, tumor location, histological grading, stage, TP and DPD expression levels and MSI status were not correlated with patients’ overall survival as reported in detail in Table 6.
Table 6

Results of Cox multivariate analysis

Variables

Hazard ratio (HR) for cancer specific death (P)

95% CIa

Ageb (>63 vs. ≤63)

1.36 (0.61)

0.41–4.50

Sex (female vs. male)

0.86 (0.81)

0.24–2.96

Tumor location (left-transverse-right)

0.51 (0.06)

0.25–1.03

Stage (III vs. II)

2.74 (0.15)

0.70–10.74

Grading (G3–G2–G1)

1.19 (0.76)

0.39–3.58

TS expression (high vs. low)

0.15 (0.02)

0.03–0.76

TP expression (high vs. low)

0.46 (0.23)

0.13–1.66

DPD expression (high vs. low)

0.99 (0.99)

0.33–2.97

MSI status (MSI-H vs. non-MSI-H)

1.34 (0.75)

0.22–8.05

aConfidence interval

bDichotomization according to the median value

Since the use of adjuvant therapy for stage II patients is controversial and the identification of reliable prognostic factors for therapeutic decision could be challenging (Andre et al. 2006), we analyzed separately TS expression levels in cancers of stage II and III. We found that a higher TS expression was predictive of longer overall survival only in patients with stage II (P < 0.001). On the contrary, a higher TS expression did not improve survival in patients with stage III (P = 0.68) (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00432-010-0872-1/MediaObjects/432_2010_872_Fig2_HTML.gif
Fig. 2

Kaplan–Meier survival curves in patients with colon cancer of stage II (a) or stage III (b) according to their TS expression levels (straight line: low expression; dotted line: high expression). P value according to the log-rank test is reported. Censoring dates are reported on the curves by cross-marks

Discussion

The possibility to predict response to 5-FU adjuvant therapy in patients with colon cancer is a basic prerequisite for the optimization of this therapy. The identification of molecular characteristics, which may better predict the efficacy of 5-FU adjuvant therapy, could prevent useless medication in patients unlikely to respond to this drug and enable to place these subjects on alternative therapies. Several studies have been addressed to analyze the roles of MSI or of TS, TP and/or DPD expression in patients with colon cancer, with controversial results. To our knowledge, this is the first study in which all these markers have been analyzed simultaneously in the same cohort of patients.

In the present case study, about 13% of patients exhibited MSI-H cancers. This datum is consistent with previous studies, which report 10–15% of sporadic colon adenocarcinomas to be MSI-H (Muller et al. 2004). Our results showed that around 85% of MSI-H cancers were located in the right colon versus around 20% of non-MSI-H tumors. This observation agrees with previous studies on sporadic colon cancers, which showed that MSI-H cancers are usually right-sided (Ribic et al. 2003). Typically, MSI-H cancers are reported to be more frequent in women and associated with a higher histological tumor grade (Soreide et al., 2006), but in this study no associations were found between MSI and gender or tumor grade.

It has already been reported that MSI-H colon cancers have a better prognosis compared to those with functional MMR (Popat et al. 2005). However, the role of 5-FU adjuvant therapy remains to be fully clarified in these tumors. There are some suggestions that patients showing non-MSI-H cancers may benefit from adjuvant therapy, while those exhibiting MSI-H tumors may not (Ribic et al. 2003). Our data questioning the benefit of adjuvant 5-FU therapy to patients showing MSI-H cancers suggest that this treatment may even be harmful to this subset of patients. In this study, indeed, in which all patients received adjuvant 5-FU chemotherapy, we found that patients with MSI-H cancers showed almost a 1.5-fold risk increase in cancer-related death than those with cancers displaying a stable microsatellite status (P = 0.75). This result, however, is not statistically significant, presumably because of the small number of MSI-H patients.

Regarding the expression levels of TS, TP and DPD in relation to MSI status, in this study no significant differences were detected, even though a trend toward an association between higher levels of TS and MSI-H was observed. This association has been already reported (Ricciardiello et al. 2005), although discordant data are published (Sinicrope et al. 2006). We suppose that in our case study this association was due to the small number of patients displaying MSI-H tumors. Our results agree with Sinicrope et al., who reported that MSI status and TS expression are unrelated, suggesting that TS levels cannot explain therapeutic resistance to 5-FU in MSI-H colon cancers (Sinicrope et al. 2006). Our results suggest also that DPD and TP levels cannot predict response to 5-FU in MSI-H colon adenocarcinomas. This is in line with recently published findings (Jensen et al. 2009), on the absence of association between DPD expression and MSI status. No reports are available to confirm or rebut our results on TP expression and MSI status.

Patients’ overall survival was analyzed by log-rank test and by Cox proportional hazard regression method. Both survival analyses identified that TS expression at mRNA level was involved in patients’ overall survival. The results on DPD are in contrast with previous studies that associated lower levels of this enzyme to longer survival in 5-FU-treated patients (Jensen et al. 2007; Tsuji et al. 2004). However, in other studies no significant associations were detected between the marker and survival (Soong et al. 2008), confirming our results.

Regarding TP expression, controversial data are available in literature about its correlation with patients’ survival. Some evidence related high TP expression levels with poorer prognosis because of its association with angiogenesis (Peters et al. 2008; Takebayashi et al. 1996). Our results do not support these findings, but are consistent with Soong et al. (Soong et al. 2008), who did not find any association between TP expression and survival both in 5-FU-treated and not-treated patients.

For thymidylate synthase, our results showed that patients with a higher expression level of this enzyme seemed to have a better prognosis. This finding is discordant with previous studies that have generally associated high TS expression levels with worse overall survival in colorectal cancers (Aguiar et al. 2005; Edler et al. 2002; Popat et al. 2004). Nevertheless, this association was restricted to patients who did not received 5-FU therapy. Our results are also in contrast with recent studies that did not report a significant improvement of the overall survival in 5-FU-treated stage III patients, characterized by high TS levels (Popat et al. 2006; Soong et al. 2008).

The reasons for the abovementioned discrepancies among our study and the others could be various. First of all, the differences in the techniques used to determine the genes expression levels (mainly immunohistochemistry) could play a main role. Immunohistochemistry is indeed the commonest used method to determine genes’ status (high or low) for subsequent survival analyses (Jensen et al. 2007; Popat et al. 2006; Soong et al. 2008). However, we chose of using qRT-PCR because, to our experience, immunohistochemistry is a method heavily operator-dependent and not easily reproducible in different laboratories for quantitative evaluations. Differences in the assessment of staining patterns by different observers, severe heterogeneity in staining intensities between cells and differences in antibodies and protocols used may all cause an inaccurate reflection of the protein content in the specimen. On the other hand, qRT-PCR is a more accurate method for quantitative measurements, even though some biases can come from the choice of the thresholds that are used to classify patients as high or low expressers. It has indeed been reported that these thresholds are often defined as those most likely to predict response (Popat et al. 2004). For this reason we chose as threshold the median value. In addition, the different results obtained could be related to differences in the cohorts of patients analyzed in different studies (colon and rectal cancers and different disease stage). In Popat’s case study indeed 59% of tumors were of rectal origin and in both Popat’s and Soong’s studies the survival analysis was restricted only to patients with stage III cancers. In the present study, we analyzed only colon adenocarcinomas of both stage II and III. Although, as reported in the results, analyzing separately TS expression levels in cancers of stage II and III, we found that a higher TS expression was predictive of longer survival only in patients with a stage II. On the contrary, a higher TS expression did not improve survival in patients with stage III, confirming the previously reported findings (Popat et al. 2006; Soong et al. 2008).

Our results also show that TS gene expression has a different predictive significance in stage II patients with respect to the stage III ones. This could come from the fact that lymph node–negative tumors exhibit molecular differences with respect to lymph node-positive tumors (Grade et al. 2007). It has indeed already been reported that the acquisition of a metastatic potential by the primary tumor is accompanied by specific changes in endogenous transcription and in the tumor microenvironment (Jorissen et al. 2009). Above this, however, the true reason of this difference is still unknown. Nevertheless, the finding that TS expression may be a predictive factor of longer overall survival in patients with a stage II could be of extreme importance because adjuvant chemotherapy in stage II patients is a controversial question (Andre et al. 2006). The identification of TS expression as a possible predictive marker of response to 5-FU could have clinical implications on the choice of patients to submit to 5-FU treatment.

In conclusion, in this study a high TS expression predicts a longer overall survival in a cohort of patients treated with a 5-FU-based regimen. Since the complete case study is related to treated patients, we could hypothesize that a better survival is related to patients’ response to the chemotherapeutic treatment, suggesting that higher TS expression could predict a good response to 5-FU treatment, especially in cancers of stage II. Moreover, TS expression could also be a marker of better overall survival in patients receiving 5-FU adjuvant therapy: in our case study, about 70% of patients who developed metastasis exhibited a low TS level.

Nonetheless, these results are preliminary, since the real predictive value of TS in evaluating the 5-FU response in stage II colon cancers could result only from larger retrospective and prospective randomized studies.

Acknowledgments

We would like to thank IMPACTS group for technical discussion in using archive tissues (ref. group involved in the European Project No 37211). This work was supported by grants from MIUR-Prin 2006 (project number 2006063220_003) and FIRB-MIUR. The authors also thank Dr. Valentina Melita for the English revision of the manuscript.

Conflict of interest statement

There are no conflicts of interest to declare.

Copyright information

© Springer-Verlag 2010