Is the Measurement of Thymidylate Synthase to Determine Suitability for Treatment with 5-Fluoropyrimidines Ready for Prime Time?¹

Several studies have been published during the past few years relating measurements of the expression of the enzyme, TS,³ to respond to 5FU and FdUrd, as well as to overall outcome of patients with colorectal carcinoma. These investigations have given rise to hope that, like tumor measurements of the estrogen receptor, or Her2-neu levels in breast cancer, by assessing tumor TS levels, the physician may decide if the patient has a reasonable chance of responding to treatment to fluoropyrimidines. The purpose of this minireview is to summarize and critically evaluate these reports and stimulate new studies. The reader is also referred to recent reviews of this subject (1, 2, 3, 4, 5).

Early studies showed that the sensitivity of colorectal carcinoma cell lines to FdUrd correlated with the level of TS; cells with high levels of TS were less sensitive to this drug than cell lines with low enzyme activity (6). Tumor cell lines transfected with wild-type TS cDNA, and as a consequence overexpress this enzyme protein, are also more resistant to continuous exposure to 5FU or FdUrd (8). Finally, as in the case of many tight binding enzyme inhibitors, which FdUMP, an anabolite of 5FU and FdUrd is, a mechanism of resistance to 5-fluoroyrimidines is overexpression of this protein (9). Thus, unless cells lack 5–10 methylene tetrahydrofolate, the cofactor for TS that is necessary for tight binding of fdUMP in ternary complex with the inhibitor and TS, sensitivity to 5-fluoropyrimidines would be expected to reflect the level of TS in the cell. Differences in transport, anabolism, and catabolism of this drug from one tumor to another and apoptotic and antiapoptotic mechanisms operative in the tumor also impact on the effects of this drug on cytotoxicity (7, 8). In particular, given the complexity of the metabolism of 5FU, it is perhaps not surprising that the greatest value of measuring TS levels alone may be to eliminate those patients who are unlikely to respond to this drug, because patients with low levels of TS may not be sensitive to 5FU for other reasons than TS activity. Another important variable that must be considered is the dose schedule. Pulse 5FU with or without LV may be more of an RNA-directed drug than continuous infusion schedules of 5FU (10); thus, the predictive value of TS measurements may be less when pulse treatments are used.

Several studies of patients with advanced colorectal cancer have recently been published (11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23), indicating that patients with high tumor levels of TS are unlikely to respond to infusional treatment with 5FU, whereas patients with low levels have response rates higher than that expected (>50%). As noted in Table 1, except for a recent study by Etienne et al. (21), the numbers of patients studied with advanced disease have been relatively small. Nevertheless, the agreement between the studies has been remarkably good, despite differences in assays used (mRNA expression, Western blotting, enzyme activity, FdUMP binding, and IHC). Not yet clear is whether measurements of TS predict the response of patients receiving pulse doses of 5FU/LV (weekly or 5-day monthly schedules). One study of 41 patients, treated with the 5-day Mayo regimen (Table 1), indicated that low levels of TS, measured by IHC, also predicted response to this dose schedule (17). In this regard, it would follow that low levels of TS would also predict response to capecitabine, an p.o. drug that simulates infusional 5FU treatment. This prediction is complicated, however, by the requirement for sufficient levels of thymidine phosphorylase in tumor cells, needed for conversion to a metabolite of this drug to 5FU (24). Thus, both a high level of TP and a low level of TS may be required to predict response to this drug. It also should be pointed out that high levels of TP, by also converting 5FU to its riboside, may decrease available 5FU for conversion to FdUMP. Indeed, high levels of TP alone or together with high levels of TS are reported to be associated with poor response to 5FU (18, 25).

What are the causes for the widely disparate levels of TS noted in colorectal cancer? Karyotype analysis has shown that some colorectal cancers, the monosomic type, have loss of one chromosome 18, the location of the TS gene, and this loss is associated with lower levels of TS expression (26, 27). It would be of interest to correlate response to 5FU or FdUrd with tumors having this karyotype.

Increased expression of TS in colorectal tumors has been correlated with polymorphisms of the repeated sequences in the enhancer region of the TS gene promoter (28). Individuals homozygous for the triple repeat variant had higher tumor levels of TS protein than the double repeat variant and were less likely to respond to 5FU treatment compared with the other genotypes (28). A recent study from the Danenberg lab concluded that the increase in TS expression associated with the three repeat sequence was attributable to greater translational efficiency of this TS mRNA, although the in vitro studies were carried out with luciferase reporter assays rather than TS itself (29).

Fresh tumor samples are required for Western Blotting, enzyme assays, and, until recently, mRNA quantitation. An advantage of IHC is that paraffin sections of tumor may be used to identify tumor cells and semiquantitate protein expression. Recently, methods to microdissect tumor from paraffin sections and extract sufficient RNA suitable for PCR quantitation have been published, and success rates of ≥80% have been reported (18). What is the best assay? Limited studies show that measurement of mRNA levels of TS in fresh tumor samples correlated well with protein expression measured by Western blotting or fdUMP-binding assays (11, 30). There are also limited data comparing results of mRNA measurements in fresh tumors and paraffin sections using quantitative PCR. As mentioned, IHC has the advantage that paraffin sections may be used, and tumor cells that express TS may be specifically identified. Thus far, both polyclonal and monoclonal antibodies have been used (16, 17, 20, 21, 23); for future studies, it would be important to use a single monoclonal antibody, available to all for this purpose. In the studies where IHC has been used (Table 1), various criteria have been used to grade tumor expression. In some studies, tumor expression has been graded from 0 (no staining) to 4 (maximum staining). Other studies take into account the percentage and nature (focal or diffuse, nuclear or cytoplasmic) of staining (23, 32). Extremes of expression appear easy to distinguish, but as in the case of Her-2/neu overexpression, the intermediate grades of expression may be more difficult to correlate with outcome. Clearly, there is not yet a gold standard for assay of TS expression, and comparative assays on tumors are needed to settle this issue. One idea is the establishment of a central referral laboratory for the measurement of TS levels both at the mRNA and protein levels so that variability between the assay methods can be avoided. Whether cDNA microarrays and tissue microarrays will be better than quantitative reverse transcription-PCR and laser microdissection and IHC, respectively, will have to await method validation studies and acceptance by other laboratories engaged in such studies.

Two studies have recently reported that TS levels in hepatic metastasis of colorectal cancer also may predict response to intra-arterial administration of FdUrd (31, 32). This is perhaps not surprising, because FdUrd administered as a constant infusion over 14-day cycles with or without LV or decadron would be expected to specifically target TS, with little or no incorporation into RNA. Of interest is an adjuvant study in which systemic 5FU/LV was administered with intraarterial FdUrd/decadron, compared with intraartererial FdUrd/decadron alone, after resection of hepatic metastasis (33). The IHC grading system used to measure TS was different from previous studies in that patients were classified as high expressers if >20% of the cells stained for nuclear TS. In this study of 144 patients, TS levels and resection markers were found to be independent predictors of survival and progression, whereas p53 and EGFR, as determined by IHC, were not independent predictors. Of interest, treatment with intra-arterial FdUrd/decadron plus systemic 5FU/LV significantly improved the survival of patients with high TS levels, suggesting that high intracellular levels of FdUMP achieved by intra-arterial drug administration together with systemic treatment, in patients with minimal residual disease, may effectively inhibit the high levels of TS present (33).

It would be of benefit to patients who develop metastatic disease if the TS level of the primary tumor predicted response to 5FU, thus obviating the need for biopsy of the metastatic tumor. However, several studies have shown that there is no correlation between levels of TS in the primary tumor as compared with the metastatic tumor, the former tending to have higher levels of TS (34, 35, 36, 37). Furthermore, levels of TS vary, depending on the site of metastasis, e.g., lung metastasis and abdominal metastasis express higher levels of TS mRNA than do liver metastasis, providing a possible explanation for the relative lack of sensitivity of patients with lung and abdominal metastasis to 5FU as compared with liver metastasis (38, 39). Whether the soil or genetic make-up of the tumor influence growth at the metastatic site and TS levels are important areas for further study.

Several studies have now been reported that TS levels predict for overall outcome in patients with early stage colorectal cancer (34, 40, 41, 42). Patients with relatively low levels of TS have a better survival than patients with high levels of TS. It would also be of great benefit if one could predict which patients might benefit from adjuvant treatment with 5FU/LV by measuring TS levels in the primary tumor. Elder et al. (41) reported, in fact, that patients whose tumors expressed the highest expression of TS, as determined by IHC, had a significantly longer disease-free survival if they were treated with adjuvant therapy compared with surgery alone. Possible reasons for these surprising results are that the dose schedule used for adjuvant chemotherapy has to date usually used weekly pulse doses of 5FU/LV, which as noted above may result in more RNA-directed antitumor effects, rather than targeting TS. Alternatively, patients with high TS levels are destined to relapse, and these are the patients that adjuvant treatment may help. Because capecitabine is now being tested as adjuvant therapy for this disease, alone and in combination, it will be important to determine whether levels of TS and TP predict outcome in these patients.

Tumors, including colorectal tumors, often overexpress the transcription factor, E2F1 (43, 44). E2F1 exists as a heterodimer with another protein, DP1 or DP2. Release of free E2F1 from its binding to the retinoblastoma protein, as a result of phosphorylation of retinoblastoma protein by cyclin-dependent kinases 4 and 2, initiates S phase as a consequence of the transcriptional activation by E2F1 of many S phase enzymes, including TS. A direct correlation between E2F1 and TS mRNA levels in colorectal tumors has been reported by this laboratory and Kasahara et al. (43, 44). In cell lines with forced overexpresson of E2F1, TS is also overexpressed, resulting in decreased sensitivity to 5FU (45). Indeed, the frequency of high levels of E2F in tumors has led to a strategy that uses the E2F1 promoter to target suicide genes to the tumor, thus providing selectivity as normal cells have lower expression of E2F1 (46, 47).

Measurement of TS in metastatic tumors has promise to help decide which patients are likely to respond to infusional 5FU or intraarterial FdUrd. Additional measurements of anabolic and catabolic enzymes involved in 5FU metabolism may further refine the prediction of which patients will respond to treatment (18, 48). It will be of importance to develop a standard assay for TS expression either, mRNA or protein, which is widely available. Important issues that need to be resolved are the predictive value of TS and/or TP in the response to capecitabine (both for advanced disease and adjuvant chemotherapy) and a determination of the value of measurements of TS when pulse schedules of this drug are used, in particular, the weekly schedule. The value of measuring levels of not only TS but also E2F1 in predicting outcome should also be explored. Large prospective trials using TS measurements to demonstrate the usefulness of this assay to save patients from ineffective treatment with 5-fluoropyrimidines are needed. These studies will be complicated by the new standard(s) of treatment that now combines 5FU with irinotecan or oxaliplatin. Strategies to target tumors with high levels of TS, such as the use of drugs that are converted by TS into toxic metabolites, are promising approaches for the treatment of these patients (49).