Thymidylate Synthase as a Predictor of Response

August 1, 1998

It has been hypothesized that intratumoral thymidylate synthase (TS) gene expression might be used to select therapy for patients with disseminated colorectal cancer. We recently

ABSTRACT: It has been hypothesized that intratumoral thymidylate synthase (TS) gene expression might be used to select therapy for patients with disseminated colorectal cancer. We recently reported the results of a clinical trial in 46 patients with disseminated or recurrent colorectal cancer testing whether expression of TS within the primary tumor, as assessed by quantitative polymerase chain reaction (PCR) methodology, would predict the responsiveness of that cancer to fluoropyrimidine-based therapy. This trial demonstrated that intratumoral TS/beta-actin messenger RNA (mRNA) ratio can accurately predict which metastatic colorectal tumors will be resistant to a leucovorin-modulated 5-FU infusion and which have a high likelihood of responding to such a regimen. Results of other studies of adjuvant therapy in gastric cancer and colorectal cancer also indicated that TS expression within the tumor is predictive of response to 5-FU-based therapy. It may be possible to use this parameter prospectively to decide which patients should receive fluorinated pyrimidine therapy: Patients whose tumors express low TS levels would be likely to benefit from such therapy, whereas limited preliminary data suggest that patients whose tumors express high TS levels may benefit from irinotecan (CPT-11 [Camptosar]). [ONCOLOGY 12(Suppl 6):43-47, 1998]


If a molecular marker for a particular tumor can predict response or resistance to specific treatment, therapy could focus not only on drugs targeting that marker but also on newer agents with different molecular targets. Specifically, it has been hypothesized that intratumoral thymidylate synthase (TS) gene expression might be used to select therapy for patients with disseminated colorectal cancer. We recently reported the results of a clinical trial testing the hypothesis that expression of TS within a metastatic colorectal cancer, as assessed by quantitative polymerase chain reaction (PCR) methodology, would predict the responsiveness of that cancer to fluoropyrimidine-based therapy.[1]

This article summarizes the design and results of that trial, as well as other data suggesting that TS may be useful as a predictor of response to adjuvant therapy for gastric and colorectal cancer.

Clinical Trial: Correlation of TS and Response to 5-FU

Patients and Methods

Eligibility criteria were typical for enrollment in a phase II colon cancer trial. Patients were required to have a diagnosis of disseminated or recurrent colorectal cancer that was measurable, a Southwest Oncology Group (SWOG) performance status of 0 to 2, and adequate hematologic, hepatic, and renal function. The exception to standard eligibility criteria was that each patient was required to have a measurable lesion accessible for biopsy. Prior therapy with fluorouracil (5-FU) was allowed, but patients were excluded if they had received prior infusional 5-FU.

All patients signed an institutional review board (IRB)-approved written informed consent for a core-needle biopsy of their tumor, performed specifically for the purpose of measuring TS expression prior to the initiation of chemotherapy. The lesion assayed for TS expression had to be bidimensionally measurable either by physical or radiologic examination.

Treatment Regimen--The treatment regimen used in this trial, developed in previously reported phase I and II studies conducted at the University of Southern California, consisted of 5-FU, 200 mg/m²/d administered as a continuous infusion by ambulatory infusion pump (Pharmacia Deltec CADD pump), with leucovorin, 20 mg/m² given by intravenous push weekly.[2,3] The initial treatment cycle was 4 weeks, followed by a 1-week rest, with subsequent cycles consisting of 3 weeks of continuous therapy and 1 week of rest.

After two cycles (8 weeks) of treatment, measurable disease was reassessed for tumor response by the same test used at baseline. Response criteria for this trial were the standard definitions used for national cooperative group trials.[4] Upon disease progression, additional tumor biopsies were taken from the measurable disease site for quantitation of TS expression when patients consented to a second procedure.

Laboratory Methods--In this trial, all specimens analyzed for TS messenger RNA (mRNA) expression were obtained by core-needle biopsy. The technique employs a coaxial system in which fine-needle aspiration is used to confirm cytologic evidence of cancer within moments of the aspiration, with the core-needle material being used for TS analysis.

Our laboratory’s methodology for determining a relative TS mRNA ratio, which has been extensively detailed in previous publications,[5-7] involves the use of an internal standard gene (beta-actin) for the measurement of relative gene expression by determining a ratio between the amplified internal standard complementary DNA (cDNA) and target cDNA within a linear range.

Response Definitions and Statistical Methods--Bidimensional measurement of the lesion sampled for TS/beta-actin mRNA ratio, referred to as the "indicator lesion," had to meet established, standard criteria for the determination of response or progression after administration of two cycles of protocol therapy. A complete response (CR) was defined as disappearance of the indicator lesion, as well as all other clinically and radiologically detectable disease. A partial response (PR) required a 50% reduction in the sum of the products of the perpendicular diameters of the indicator lesion, as well as other disease measured at baseline, without growth of other disease or appearance of new lesions.

Failure of therapy was defined as a 25% increase in the size of the sum of the product of the perpendicular diameters of the indicator lesion or other disease, or appearance of new disease, during the initial 8 weeks of therapy. Patients who showed a response or stable disease continued on protocol therapy until progression was documented.

Investigators were blinded to the results until complete data were available for each patient: Clinicians were not informed of TS mRNA levels until response data were known, and laboratory investigators were not informed of response data until after patients’ TS mRNA levels were determined.

In a preliminary analysis (with fewer patients), a TS/beta-actin mRNA ratio of 3.5 ×10-3 was identified as the "optimal" cut-off for predicting patients who might respond and those who definitely would not respond; this ratio was significantly associated with response (P < .01).[8] In the current cohort, although a TS/beta-actin mRNA ratio of 3.5 ×10-3 was the sample median, and a TS/beta-actin mRNA ratio of 4.1 ×10-3 was found to be the "optimal" cut-off, a TS/beta-actin mRNA ratio of 3.5 ×10-3 was still statistically significant (P < .01) when compared to the simulated maximal chi-square distribution.

Demographics--A total of 46 patients (29 men and 17 women) were entered into the trial. Median age was 61 years (range, 34 to 80 years). Liver metastases represented the most common site of disease that was biopsied and assessed for response (28 patients). Nodal metastases were the disease sites evaluated in six patients and peritoneal or retroperitoneal metastases were evaluated in eight.


Ratios of TS/beta-actin mRNA were available for 42 patients (12 responders, 30 nonresponders); RNA extraction on the remaining 4 specimens was insufficient in quantity to perform the PCR assay. The range of TS/beta-actin mRNA ratios measured in tumor biopsies for the 42 patients was 0.3 × 10-3 to 18.2 ×10-3, with a median TS/beta-actin mRNA ratio of 3.5 × 0-3 . Analyses revealed no significant associations of TS/beta-actin mRNA ratios with age, sex, ethnicity, or tumor site.

Of the 46 patients, 12 (26%) had partial responses to protocol therapy; 34 patients demonstrated resistance (disease progression) to treatment. For patients responding to therapy, the range of tumor TS/beta-actin mRNA ratios was 0.5 ×10-3 to 4.1 × 10-3, with a median value of 1.9 ×10-3. Nonresponders had TS/beta-actin mRNA ratios ranging from 0.3 ×10 -3 to 18.2 ×10 -3, with a median value of 5.6 ×10 -3 . The difference between responders and nonresponders in baseline TS/beta-actin mRNA ratios was statistically significant (P < .001, based on a two-sided Wilcoxon test.)


TS/Beta-Actin mRNA Ratios as Response Predictors--The median TS/beta-actin mRNA ratio (3.5 ×10-3) significantly distinguished responders and nonresponders based on the distribution of a maximal chi-square statistic (P = .001). Of the 21 patients with TS/beta-actin mRNA ratios of 0.3 ×1-3 to 3.5 ×10-3 (below the median), 11 (52%) responded, while only 1 (5%) patient with a TS/beta-actin mRNA ratio > 3.5 ×10-3 showed a response. The TS/beta-actin mRNA ratios for the 30 nonresponding patients spanned the entire range of values seen, although none of the 19 patients with a TS/beta-actin mRNA ratio of > 4.1 ×10 -3 responded.

Six patients who responded initially consented to a second tumor biopsy after their disease progressed during therapy. All of these patients had higher TS/beta-actin mRNA ratios in their cancers at progression than at baseline, with increases ranging from 3- to 25-fold.

TS/Beta-Actin mRNA Ratios and Survival--Patients with a relatively low TS/beta-actin mRNA ratio treated with the 5-FU infusion had a better median survival than those with a relatively high TS/beta-actin mRNA ratio treated with the same dose and schedule of 5-FU. The median survival for the whole cohort of 46 patients was 8.9 months.

When survival is stratified by median TS/beta-actin mRNA ratios, patients with tumors expressing a TS/beta-actin mRNA ratio £ 3.5 ×10-3 have a median survival of 13.6 months, whereas patients whose tumors have a TS/beta-actin mRNA ratio > 3.5 ×10-3 have a median survival of 8.2 months (P = .02, based on the log-rank test) (Figure 1).


The data from this trial demonstrate that intratumoral TS/beta-actin mRNA ratio can accurately predict which metastatic colorectal tumors will be resistant and which have a high likelihood of responding to a leucovorin-modulated 5-FU infusion. Although TS/beta-actin ratios for responders and nonresponders overlapped, responding patients had lower baseline TS/beta-actin ratios (median, 1.9 × 10-3 for responders vs 5.6 ×10-3 for nonresponders). Specifically, a TS/beta-actin mRNA ratio > 4.1 ×10-3 in disseminated colorectal cancer was associated with clinical resistance to the 5-FU infusional regimen employed in this trial.

Conversely, a TS/beta-actin mRNA ratio £ 3.5 ×10-3 was associated with a 52% response rate. This rate is approximately twice that expected in chemotherapy-naive patients randomly allocated to 5-FU treatment, and may be somewhat underestimated, in that one-quarter of the responders were previously treated with 5-FU and would have been expected to have an even lower response rate.[9] Within the cohort of previously treated patients, the intratumoral TS/beta-actin mRNA ratio was a more accurate predictor of initial resistance than was prior exposure to 5-FU.


Whether dose schedule and modulation of fluorinated pyrimidine therapies are associated with different TS "cut-off points" remains unknown; this may also be an important issue for the newer antifolate TS inhibitors. In a small cohort of patients with liver metastases from colorectal cancer who were treated with hepatic arterial infusion of floxuridine (FUDR), the median TS/beta-actin mRNA ratio was 5.4 ×10-3, which is higher than the cut-off for resistance to systemic infusion of 5-FU.[10]

Clearly, TS expression is only one mechanism of resistance to 5-FU-based chemotherapy, as 11 patients in this trial had TS/beta-actin mRNA ratios < 3.5 ×10 -3 but did not respond to therapy. Lenz et al have demonstrated that p53 status of the tumor also correlates with outcome to 5-FU therapy in a cohort of disseminated colorectal cancer patients, with tumors demonstrating wild-type p53 having a higher rate of response and those with mutated p53 being more poorly responsive. Patients whose tumors expressed wild-type p53 tended to have lower TS expression, but when the cohort was stratified by TS expression, p53 was no longer predictive.[11]

TS as Response Predictor in Adjuvant Therapy

Gastric Cancer

It would be unreasonable to assume that colorectal cancer is the only tumor type for which TS expression is predictive of response and resistance to fluorinated pyrimidine therapy. In a previous publication, we described a similar relationship in primary gastric cancer.[12] We found that the TS mRNA level in a biopsy specimen of a primary gastric cancer is inversely related to response and survival in patients receiving neoadjuvant 5-FU-based chemotherapy.

This analysis employed the same PCR methodology as the colorectal trial described above; the beta-actin gene was used as an internal standard for TS gene expression. Preoperative treatment consisted of two cycles of protracted-infusion 5-FU, 200 mg/m²/d administered for 3 weeks, with leucovorin, 20 mg/m² weekly. Cisplatin, 100 mg/m², was administered on days 1 and 29. The second cycle of the same treatment started on day 29. When possible, a second specimen for TS quantitation was obtained after preoperative chemotherapy.

The median TS/beta-actin ratio in the 65 patients with primary gastric adenocarcinoma was 4.6 × 10 -3 (range, 0.9 ×10-3 to 20.1 ×10-3). Of the 45 patients, 19 (42%) with measurable primary gastric cancer responded to neoadjuvant chemotherapy. Of the 22 patients evaluable for response whose TS/beta-actin quantitations were less than the median, 14 (63%) responded to 5-FU-based chemotherapy. In contrast, only 5 (22%) of the 23 evaluable patients with TS/beta-actin quantitations above the median responded to neoadjuvant chemotherapy (P = .004). The median survival for 31 patients with gastric tumor TS quantitations < 4.6 ×10-3 was 14 months, as compared with 6 months for the 34 patients with tumor TS quantitations ³ 4.6 ×10-3 (P = .001).

Colorectal Cancer

Using immunohistochemical methodology that employs a monoclonal antibody to TS, Johnston et al analyzed TS content of paraffin-embedded tumor tissue obtained from patients enrolled in the National Surgical Adjuvant Breast and Bowel Project R01 trial.[13] In this trial, patients were randomized to surgical follow-up only vs postoperative pelvic radiation vs postoperative chemotherapy using a regimen of 5-FU with semustine (methyl-CCNU) and vincristine. Overall, this trial showed a survival benefit for patients randomized to the chemotherapy arm.

In the report by Johnston et al, results of the TS analysis indicated that patients whose tumors expressed high levels of TS had worse survival than those whose tumors expressed low levels of TS. Within the cohort that received chemotherapy, those with high TS expression fared better than those with low expression, suggesting a greater benefit in a setting of higher relative risk (Table 1). Although no significant benefit was seen among patients with low TS expression who received chemotherapy, there was a trend in this direction. This high TS-expressing group had outcomes similar to patients in the untreated group who were low TS expressors, with the untreated, high TS-expressing cohort having the worst overall outcome.

Although it is somewhat counterintuitive that patients with high TS expression in their primary rectal cancers would have enjoyed any benefit from 5-FU therapy, several other points have to be considered: The treatment regimen used in this trial contains two other chemotherapy agents with mechanisms of activity that are not TS-directed. As the TS analysis was performed retrospectively, there was a slight imbalance in cohorts, with a larger number of patients having low TS expression in the surgery-only group and a higher number of high TS expressors in the chemotherapy group.

Furthermore, although the immunohistochemical staining methodology used in this trial has a relative positive correlation with PCR quantitation, it is not as specific. Therefore, some of the "high" expressors by immunohistochemistry may have fallen into the upper range of the low TS group. Since the immunohistochemical method can be associated with both focal positivity and interobserver variability, this remains a possibility.

The Eastern Cooperative Oncology Group (ECOG) performed a similar analysis of colon and rectal cancer patients enrolled in INT 0089 (adjuvant colon cancer trial; randomized 5-FU regimens; stage III), INT 0035 (adjuvant rectal cancer trial; randomized 5-FU regimens with postoperative radiation; stages II and III), and ECOG/CALGB (Cancer and Leukemia Group B) 2290 (advanced colorectal cancer; randomized modulated 5-FU regimens).[14] Low TS in the primary tumor was associated with better disease-free and overall survival, as well as an enhanced advantage from adjuvant chemotherapy. Patients in the advanced disease trial who responded had high TS scores. Although TS expression in the primary tumor may serve as an indicator of prognosis in an untreated population, variability between primary and metastatic clones with respect to this parameter may make interpretation of therapeutic benefit based on expression in the primary tumor less exact.

Future Directions

The results of the investigations described above suggest that expression of a molecular parameter, TS, within a given metastatic colorectal cancer is predictive of outcome to current front-line 5-FU therapy. We should be able to use this parameter prospectively to decide who should receive fluorinated pyrimidine therapy with a high likelihood of deriving benefit. What conclusions can be drawn about patients whose tumors express high levels of TS and therefore are expected to be resistant to 5-FU? Limited preliminary data suggest that this cohort has a high likelihood of responding to irinotecan (CPT-11 [Camptosar]).[15]

As we learn more about the expression, interaction, and correlation of individual genetic parameters with therapeutic response, we may be able to place cancer treatment, in both the disseminated and adjuvant settings, on a more rational basis. This knowledge also should allow us to move forward more quickly with new drug development.


1. Leichman CG, Lenz Hi, Leichman L, et al: Quantitation of intratumoral thymidylate synthase expression predicts for disseminated colorectal cancer response and resistance to protracted infusion 5-fluorouracil and weekly leucovorin. J Clin Oncol 15 (10):3223-3229, 1997.

2. Leichman CG, Leichman L, Spears CP, et al: Biochemical modification of protracted infusion of 5-fluorouracil with weekly leucovorin: A dose-seeking clinical trial in patients with disseminated gastrointestinal cancers. Cancer Chemother Pharmacol 26:57-61, 1990.

3. Leichman CG, Leichman L, Spears CP, et al: Prolonged continuous infusion of fluorouracil with weekly bolus leucovorin: A phase II study in patients with colorectal cancer. J Natl Cancer Inst 85:41-44, 1993.

4. Green S, Weiss GR: Southwest Oncology Group standard response criteria, endpoint definitions and toxicity criteria. Invest New Drugs 10:239-253, 1992.

5. Horikoshi T, Danenberg KD, Stadbauer THW, et al: Quantitation of thymidylate synthase, dihydrofolate reductase, and DT-diaphorase gene expression in human tumors using the polymerase chain reaction. Cancer Res 52:108-116, 1992.

6. Lenz H-J, Danenberg KD, Leichman L, et al: Quantitation of thymidylate synthase gene expression using PCR and centrifugal Sephadex columns. PCR Methods Applic 4:305-308, 1995.

7. Lenz H-J, Danenberg KD, Danenberg PV: Quantitative measurement of relative gene expression in human tumors, in Reverse Transcriptase PCR, New York, Simon and Schuster, 1994.

8. Leichman L, Lenz HJ, Leichman CG, et al: Quantitation of intratumoral thymidylate synthase expression predicts for resistance to protracted infusion of 5-fluorouracil and weekly leucovorin in disseminated colorectal cancers: Preliminary report from an ongoing trial. Eur J Cancer 31A(7-8):1306-1310, 1995.

9. Izzo J, Zarba J, Rougier PH, et al: Low dose continuous infusion (FUCI) in advanced colorectal cancer: Clinical evidence for reversal of acquired/intrinsic resistance to 5 FU or 5 Fufolinic (FuFo) (abstract). Proc Am Assoc Can Res 3:1298, 1992.

10. Xiong Y-P, Danenberg K, Metzger R, et al: Thymidylate synthase (TS) m-RNA levels predict for survival and response to hepatic artery infusion (HAI) with FUDR in patients with colorectal liver metastases. Proc Am Soc Clin Oncol 16:918, 1997.

11. Lenz HJ, Danenberg KD, Leichman CG, et al: p53 status and thymidylate synthase levels are predictors of chemotherapy efficacy in patients with advanced colorectal cancer. Proc Am Soc Clin Oncol 15:504, 1996.

12. Lenz HJ, Leichman CG, Danenberg KD, et al: Thymidylate synthase mRNA level in adenocarcinoma of the stomach: A predictor for primary response and overall survival. J Clin Oncol 14:176-182, 1996.

13. Johnston PG, Fisher ER, Rockette HE, et al: The role of thymidylate synthase expression in prognosis and outcome of adjuvant chemotherapy in patients with rectal cancer. J Clin Oncol 12:2640-2647, 1994.

14. Benson AB III, Catalano PJ, Rao S, et al: Thymidylate synthase expression as a predictor for response to 5-fluorouracil-based therapy and survival for patients with resected colon or advanced colorectal cancer: An ECOG study. Proc Am Soc Clin Oncol 16:917, 1997.

15. Saltz L, Danenberg P, Paty D, et al: High thymidylate synthase (TS) expression does not preclude activity of CPT-11 in colorectal cancer (CRC). Proc Am Soc Clin Oncol 17:1080, 1998.