|Home | About | Journals | Submit | Contact Us | Français|
The major determinants of 5-flurouracil response would appear, based on accumulated literature, to be thymidylate synthase (TYMS, TS) expression levels, TS gene modifications, and TP53 status. We tested 5-fluorouracil sensitivity in yeast and human cancer cell models in which TS or TP53 alleles and expression were varied. Polymorphic TS tandem repeat status, TS expression levels reported, TS intragenic mutations, and TP53 status in outbred and experimental cancer cell lines did not predict 5-FU sensitivity or resistance. Novel observations included a dose-resistant persistence of unbound TS protein in many cancers and, upon 5-FU treatment of the colon cancer cell line, HCT116, evidence of allelic switching favoring transcripts of the mutant TS allele. The reported alleles having an intragenic mutation could not be causally associated with major degrees of 5-FU sensitivity. In yeast, TS protein was altered upon treatment with fluoro-deoxyuridine monophosphate, but 5-FU toxicity appeared largely to be RNA-based, being rescued by uridine rather than by thymidine. Cancer cell lines were also rescued from 5-FU toxicity with uridine rather than thymidine. Additionally, a TS (CDC21) knockout yeast strain, obviating any potential role for TS protein as a target, was hypersensitive to 5-FU. When denatured proteins from cancer cells treated with radio-labeled 5-FU were, labeled species with alternative molecular weights other than TS were visualized, providing further evidence for alternative 5-FU protein targets. These data emphasize that TS and TP53 status do not consistently explain the variance in responses of fluoropyrimidine-treated cancer cells, in part due to RNA-based toxicity.
5-fluorouracil (5-FU) is one of the most widely used chemotherapeutics against gastrointestinal malignancies. Heidelberger and colleagues created 5-FU over 50 years ago by substituting a fluorine atom for a hydrogen in the pyrimidine ring to create an antagonist of uracil (1). Administered primarily intravenously either by bolus or prolonged continuous infusion schedules, 5-FU is metabolized both in the liver and then intracellularly by a number of enzymes. The prevalent view in the literature represents 5-FU's primary mode of action as being through its metabolite, fluoro-deoxyuridine monophosphate (FdUMP), which covalently binds to thymidylate synthase (TS or TYMS)(2). Since cancer cells are highly replicating cells compared to normal cells, inhibition of TS is believed to cause thymidylate depletion and eventual ‘thymidineless death’ of tumor cells (2).
Even though 5-FU has multiple metabolites that were shown previously to induce cell death by incorporation of fluorinated nucleotides into DNA and RNA, investigators have focused primarily on TS as the major target (3). Santi and others first showed that TS forms a stable complex (termed the ternary complex or classic complex) from the 5-FU metabolite, FdUMP, and 5,10-methylenetetrahydrofolate (MTHF)(4-6). Since these early biochemical experiments exploring the FdUMP relationship with TS, a number of investigations focused on 5-FU's relationship with TS expression levels in clinical settings (7). Johnston et al pioneered this field, describing a direct relationship between increased TS expression levels (mRNA and protein) and decreased 5-FU response (8, 9). Recently, Johnston and colleagues have modified their initial thesis (10). These investigators now propose TS expression in the metastatic lesions to be more predictive of 5-FU response than TS levels in the primary tumor (7).
Others tried to utilize the classic complex formation to obtain informative clinical information. Peters et al showed an inhibition of thymidylate synthase enzymatic activity upon the formation of classic complexes in cancer cells and proposed this to be more predictive of response than TS expression levels (11). Brody et al advanced a technique put forth by Drake (12) in detecting classic complexes in vivo, thus promoting the idea that the tissue detection of classic complexes may be predictive, potentially allowing more appropriate dosing for 5-FU cancer cell targeting and monitoring of cellular effects of 5-FU in patients (6).
Other molecular determinants evolved from the TS/5-FU connection. TS tandem repeat (TS-TR) genotyping was proposed as a means to predict levels of TS expression in patients. Three main TS genotypes are prevalent in the population: (1) homozygous for two tandem repeats (2R/2R), (2) homozygous for three tandem repeats (3R/3R), and (3) heterozygous (2R/3R) (13, 14). The molecular basis for using the TS repeat sequence as a marker for protein expression, however, is ambiguous. These investigations presumably concern the efficiency of ribosome translation, which relies on initiation at the Kozak consensus sequence. The studies performed by Yawata and Kawakami used the Kozak sequence of the firefly luciferase gene, and not the exact Kozak sequence of the human TS gene (13, 15). Thus, the laboratory experiments did not completely establish that the tandem repeat sequence is a clinically relevant marker for TS protein expression and the related 5-FU response. Furthermore, Brody et al recently reported the difficulty of using this prediction, since gastrointestinal tumors are typically unstable at chromosome 18 where the TS gene resides (14). Hence, even if it were a marker for response, it would be difficult to unambiguously determine the TS-TR status in tumor cells (14).
We recently reported a trend to exist between TS expression levels and 5-FU clinical response in the literature (7). This trend was modest, being for practical purposes insignificant, even when stratifying the data using a number of relevant variables (7). Besides TS expression and TS-TR status, endogenous TS mutations have been identified and described previously as conferring resistance to 5-FU. Hughey, C.T. et al in 1993 reported that the HCT116 cell line harbored a heterozygous TS mutation (16). A PubMed search combining the terms of “HCT116” and “5-fluorouracil” retrieved over 80 articles, underscoring the utilization of this cell line by investigators studying 5-FU, yet most of the studies did not cite nor take into account this prior work. In laboratory investigations exploring the functional significance of this mutation, the mutation affected the binding capacity of FdUMP to the TS protein that harbored the mutated amino acid change (tyrosine to histidine at residue 33) (16, 17). In another study, the colorectal cancer cell line HCT C was mutagenized by EMS and a sub-clone, C 18, was identified by selecting for cells having reduced TS enzymatic activity. The C 18 sub line had artificially acquired a homozygous mutation of TS; a new point mutation (C→T) at nucleotide 647 converted Ser216 to leucine, yielding a nonfunctional allele as tested in autotrophic TS null E. coli (18). This type of intragenic null TS mutation is not reported to exist naturally in human cancers. The authors determined the effect of these mutations when TS protein was expressed in hamster cells or prokaryotes, yet did not directly determine whether they altered sensitivity towards 5-FU in cancer cells.
Besides TS, other variables have been proposed to affect 5-FU response. Bunz et al showed that TP53 disruption in a clone of cancer cells was associated with resistance to high concentration of 5-FU (19). A number of studies followed this report: over 450 citations to date. Other predictors have also been proposed for 5-FU response and sensitivity. These are either integral parts of the 5-FU metabolic pathway that produce FdUMP or are thought to regulate TS expression, including the YES gene, thymidine kinase, thymidine phosphorylase, dihydropyrimidine dehydrogenase (DPD), orotate phosphoribosyl-transferase, or 5,10 methylenetetrahydrofolate reductase (20-26). Ribic et al found that colorectal cancer patients whose tumors had the classic form of microsatellite instability (MSI) did not benefit from 5-FU therapy, while other patients did (27). This finding was significant since roughly 15% of colorectal tumors have the MSI form of genetic instability. MSI cancers harbor a greater number of subtle nucleotide sequence changes than the majority of colorectal cancers; usually, this cancer types has gross chromosomal changes (known as chromosomal instability or CIN) without high levels of sequence instability (28).
In addition to DNA-based toxicity, 5-FU induced RNA-based toxicity has been described. For example, Cory et al reversed the toxicity of 5-FU in Novikoff hepatoma cells by adding uridine to the medium (29). Pritchard and colleagues showed that in TP53-deficient mice, 5-FU induced apoptosis was due to a disruption in RNA rather than in DNA metabolism (30), and Klubes et al found uridine to rescue 5-FU toxicity in vivo (31). Bunz et al showed that 400μM uridine, but not 400μM thymidine, rescued TP53 wild type cancer cells from apoptosis otherwise induced by exposure to 5-FU (19).
Non-cancer cell models were also used to study the mechanism of 5-FU toxicity. Previously, we used yeast to study the TS/5-FU relationship, finding that yeast TS is covalently modified upon 5-FU exposure (6). Yeast models can be used to distinguish between DNA- and RNA-based mechanisms of toxicity (32). Recently, Gustavsson and colleagues showed that tRNA-modifying enzymes can be targeted by 5-FU in yeast cells (33). In the current study, we evaluated a number of cancer cell lines and yeast experimental models to explore 5-FU sensitivity and resistance.
Colorectal cancer cell lines RKO, HCT116 cells, the TP53 knockout cell line (derived from HCT116), and DLD-1 were either collected from ATCC or donated by Dr. B. Vogelstein (Johns Hopkins University). We generated a heterozygous TS knockout clone in RKO cells and a derivative of the cell line (“add-back”) in which exogenous TS expression was introduced, as previously described (Table 1) (14). Here, we further generated stable cell lines of the RKO parental, TS+/-, and HCT116 cell lines that included overexpression of wild type TS and of a reported HCT116-derived mutant allele changing tyrosine at amino acid 33 to a histidine (16) (see description and names of cell lines in Table 1). The HCT C parental line and a sub-clone (HCT C 18) containing the TS homozygous mutation was donated by Dr. S. Berger (18). The parental and genetically engineered yeast TS knockout strains were provided to us by Dr. J.F. Diffley and maintained as previously described (34) except as indicated otherwise. We generated drug-resistant HCT116 cell lines by continuously passaging parental clones in increasing amounts of 5-FU (i.e. 1μM for one month, followed by 1.5μM, up to 3μM or 10μM). Cells were maintained in culture with ≥3μM 5-FU for over one year, with fresh media and 5-FU provided twice weekly.
We plated roughly 1 × 103 cells per well in a 96-well plate. The next day, cells were treated with 5-FU and then media not changed (Sigma, St. Louis, MO). Five to seven days after a single treatment, cells were washed, lysed in water, and the DNA labeled with 0.5% Picogreen (Molecular Probes, Eugene, OR), whereby fluorescence was subsequently measured to assess changes in the surviving cell population.
PCR-amplified products from genomic DNA templates were produced from the indicated cell lines and samples. Primers used for the amplification of serine hydroxymethyltransferase (SHMT) and TS are available upon request. PCR products were purified and analyzed by automated capillary sequencing. The mutations in cell lines HCT C18, HCT116, and TS+/-.RKO were confirmed by sequencing.
HCT116 has a heterozygous mutation of TS, Y33H (16). We treated HCT116 cells at various concentrations overnight. RNA of individual cultures was converted to bulk cDNA for sequencing. Additionally, we inserted PCR products into a TA vector (pCR 2.1) to examine sequences of a consecutive series of plasmid clones to enumerate the relative expression of the two alleles. A mutated clone was subcloned into the pCDNA3.1 Zeo vector (Invitrogen, Carlsbad, CA) for follow-up functional studies (14).
We treated two cancer cell lines with C14-labeled 5-FU for over 48 h and then made protein lysates, which were electrophoretically separated on a 10% bis-tris gel (Invitrogen). The gel was fixed and dried. Autoradiography exposure was performed after more than 2 months at -80°C.
For each cell line, roughly equal cell numbers were plated and then lysed. Protein lysates were separated by electrophoresis on 10% Bis-tris gels (Invitrogen) and TS immunoblotting performed as previously published (6).
We performed PCR with previously published primer sets (35) using genomic DNA of cell lines and tumor samples. PCR-amplified products were separated using 3% LB® agarose gels (Faster Better Media, LLC, Hunt Valley, MD) (36).
RNA was extracted using the RNeasy Mini Kit (Qiagen, Valencia, CA). RNA (1 μg) was converted to cDNA using SuperScript II Reverse Transcriptase (Invitrogen). Real-time (RT) PCR was performed on the iCycler (Bio-Rad, Hercules, CA) with TS- and glyceraldehyde-3-phosphate dehydrogenase (GAPDH)–specific primer sets (available upon request). The relative fold change was normalized using GAPDH as a control and fold change was calculated by comparing TS expression of treated and untreated cultures. For each primer set, a standard curve, melting curve, and measures of PCR efficiency were obtained (14).
Using RKO cells engineered to have a heterozygous knockout of the TS gene (14) and parental RKO cells, we further manipulated TS expression in cancer cell lines using exogenous overexpression of wild type and mutant TS gene sequences. Figure 1A shows manipulated TS protein overexpression in the RKO.TS; 116.HCTmut; TS+/-.add back (14); TS +/-.HCTmut; and TS+/- (see Figure 1A and Table 1 for labels and descriptions of the generated cell lines).
Selected cell lines were treated with 5-FU for 48 hrs with a concentration ten times the IC50 for each cell line (Figure 1B). On immunoblots, bound and unbound TS bands (both an upper and a lower band) were present, the distinction between these two bands have been previously described (6, 12)(Figure 1B). In brief, the upper band represents the formation of the classic complex (a ternary complex of drug, thymidylate synthase, and a methylene group), and the lower band is unmodified TS protein (6, 12). TS protein expression was similar among these cell lines, even though they had differing TS-tandem repeat (TS-TR) status (Table 1, Figure panels 1B and 1C). We extended and verified previous work identifying the TS-TR status in these cancer cell lines (Table 1 and Figure 1C)(14).
Examining expression levels, TS protein modification, TS-TR status, or mutational status of TS or TP53 genes, we did not find a notable association with 5-FU sensitivity (Table 1 and Figure 2A). Figure 2A shows the 5-FU dose-response of the majority of publicly available pancreatic cancer cell lines. As indicated from Table 1 and Figure 1C these cell lines have different TS-TR status and TP53 status, yet all have broadly similar sensitivity to 5-FU.
The survival data of three MSI cancer cell lines (RKO, HCT116 and DLD-1 cells) were determined (Figure 2A and Table 1). RKO and HCT116 have similar sensitivity, while DLD-1 cells are resistant to 5-FU. We further sequenced the coding sequence of the TS gene in a number of colon MSI cell lines (RKO, DLD-1, HT 29, SW48). Functionally, SHMT catalyzes the reversible interconversion of serine and THF to methyleneTHF. Because the MSI line, DLD-1, was an outlier, resistant cell line in our experiments, we thus sequenced the repetitive nucleotide tracts in the coding sequence of SHMT in DLD-1 and the other lines. No mutations were found in SHMT or TS in the above-listed cell lines (other than the reported mutations of the TS gene in HCT116 and HCT C18, data not shown).
The IC50 of parental HCT116 cells was below 5μM. We generated two drug-resistant HCT116 cell lines (116.R1 and 116.R2). These cultures were passaged in 10μM (116.R1) or 3μM 5-FU (116.R2), for over a year for the R1 line and 2 years for the R2 line (Figures 2A and 2B), with an average of two passages per week in 5-FU-supplemented medium. Figure 2A confirms the resistant status of these evolved HCT116 cell lines as compared to the parental line and to the resistant DLD-1 line. The RKO.HCTmut line (the RKO cell line engineered to stably overexpress the HCT116-derived mutation of TS, Figure 1A and Table 1) was slightly resistant as compared to the RKO.TS cells (overexpressing wild type TS) and the parental line (Figure 2B and Table 1), although perhaps within the range of variation one expects from testing solitary clones. In contrast, the TS+/- knockout cells had striking differences in 5-FU sensitivity (Figure 2B and Table 1)(14). The engineered 116.HCTmut cells were not more resistant than the adaptively created resistant lines (116.R1 and 116.R2) (Figure 2B and Table 1). These data suggest that the HCT116-derived mutant form of the TS gene is somewhat similar to wild type TS in governing the degree of 5-FU sensitivity when exogenously overexpressed, but that other rather familiar forms of variation (such as engineered decreases or increases in TS gene expression, adaptive changes after drug exposure, and choice of a particular cell line) can considerably affect the level of 5-FU sensitivity of a given cell population. The magnitude of these changes in sensitivity provides a useful perspective against which to compare other potential influences upon the 5-FU sensitivity of cells, examined below.
Figure 2C shows the consequences of disrupting p53 in a previously developed model (116.-/-p53)(19). Although reported to have relative resistance at 400μM 5-FU (19) (confirmed in Figure 2C), in the cell survival assay we did not find differences in sensitivity between the isogenic p53 wild type and the 116.-/-p53 cells when testing lower, clinically relevant 5-FU concentrations (19). In another experiment, HCT116 derivative clones showed no resistance irrespective of whether p53 was disrupted or the HCT116 mutant allele was stably overexpressed (Figure 2C). Further, we did not find any systematic differences in 5-FU sensitivity in unrelated cell lines known to be TP53-wild type or TP53-mutated (Figure 1A and Table 1).
We confirmed the homozygous nature of the HCT C mutation by sequencing (data not shown). The TS-mutated protein was not readily detected by the TS-106 antibody, perhaps due to mutational disruption of the epitope or by protein instability (Figure 1D)(18). Figure 2C shows that the HCT C isogenic paired cell lines, differing in TS status, did not differ in 5-FU sensitivity.
We confirmed the heterozygous mutation in HCT116 cells (T>C mutation in the coding region of genomic DNA and cDNA) (data not shown and Figure 3A)(16). We extracted RNA at different time points from HCT116 cell cultures exposed to various 5-FU concentrations, detecting predominantly the mutated-TS allele in 25 of 26 separate cultures by sequencing of cDNA (Figure 3A and Table 2). As a control, 4 out of 5 untreated cultures expressed predominantly wild type TS cDNA (Table 2, p≤ .001). cDNA cloning confirmed a similar pattern of TS allele-specific expression. All twelve cloned PCR products from an untreated culture were wild type in sequence, while 4 of 5 clones from a 5-FU treated culture expressed the mutant allele (Figure 3A and Table 2, p≤ .002).
Real-time PCR analysis of TS mRNA expression in HCT116 cells revealed an upregulation of TS expression in 5-FU treated cells versus untreated cells (Figure 3B). This induction in HCT116 cells occurred at 5-20μM 5-FU, while RKO cells did not induce TS expression as strongly (Figure 3B).
Drug-resistant HCT116 lines (116.R1 and 116.R2) were evolved through constant drug exposure. They had a predominant expression of the mutated TS allele upon the sequencing of bulk cDNA (Table 2), yet the termination of 5-FU exposure reduced the TS expression (indicated as 116.R2, comparing 10 uM 5-FU and no 5-FU, Figure 3B). HCT116 is homozygous for the TS tandem-repeat polymorphism (2R/2R) (Table 1 and Figure 1C), thus an allelic difference at the polymorphic site could not explain our findings. Also, in vitro studies showed that the mutated TS allele in this cell line was not directly responsible for 5FU-resistance (Figures 2D-F), and thus a different survival of cells selectively overexpressing the mutant allele seems implausible. Thus, the observed induction of TS mRNA expression, which appears to be allele-specific in HCT116, indicates autoregulation through a transcriptional activation of the TS gene most likely exerted at the level of genomic DNA. This is distinct from reports supporting or detracting from a proposed autoregulation system in which TS protein binds the TS mRNA to control mRNA translation (37). Such mechanisms of negative feedback autoregulation would tend to mitigate the role of TS as a major determinant of 5-FU toxicity, enhancing interest in alternative targets of the drug.
In RKO cells having different TS levels (Figure 1), uridine (100μM) rescued 5-FU toxicity while thymidine did not (Figure 2D). Even when treated at high doses of 5-FU (200μM 5-FU), uridine (100μM) rather than thymidine (100μM) rescued RKO parental cultures (Figure 2D). This confirms many prior reports in which mammalian and human cancer cells were rescued from 5-FU toxicity preferentially by uridine (19, 29, 38-41).
We previously showed that TS was altered upon FdUMP treatment of yeast cells (6). TS-null yeast were more sensitive to 5-FU (Figure 4A) (34). TS-null yeast grown in the presence of uridine, but not those in thymidine, avoided toxicity (Figure 4B). Figure 4C confirms the absence of TS protein in lysates of the TS knockout strain. Uridine rescue of yeast treated with 5-FU further supports the importance of RNA-related 5-FU toxicity, and use of a TS-null strain removes the possibility of an effect from having targeted the TS protein.
C14-labeled 5-FU (10μM) was added to overnight cultures. Proteins were then electrophoretically separated. Three sizes of proteins were detected (50, 36, and 20 kDa) by autoradiography. One protein was presumably TS (at 36 kDa), while the others were unknown (Figure 4D).
TS is known to be targeted by FdUMP, and hence the field has been largely focused on the TS/5-FU relationship (7). However, when analyzing both experimental models and the literature, the time-honored connection between TS and 5-FU is ambiguous (7). Logically, TS, an essential enzyme for replicating cells, could serve as a good therapeutic target in cancer cells. Perhaps in the future, more specific, novel or yet-to-be-developed TS inhibitors (42) could utilize TS expression levels as a predictive marker for drug response.
We conclude that at clinically relevant concentrations of 5-FU, we observe no relation between TP53 status and 5-FU sensitivity among either outbred cell lines or the paired parental and TP53 knockout cell line (19). From the time of the TP53 knockout publication, over 400 reports have investigated a link between TP53 and 5-FU (19). Before this seminal publication, fewer than 80 publications studied this connection.
Although TP53 and TS status have enjoyed a dominant literature position in relation to 5-FU sensitivity and resistance, other investigators have explored alternative molecular determinants of 5-FU sensitivity, (43-45). These studies, including the results of gene expression profiling (46), further warn of the complexity that is obscured when attempting to correlate a single molecular variable to 5-FU response. Optimistically, however, the search for alternative 5-FU targets could help better define practical clinical predictors for fluoropyrimidine-based therapy (Figure 4D). RNA-based toxicity has been described recently and in the past as a plausible target, but many investigators choose to ignore it (19, 29-32). Plausible candidate 5-FU targets related to RNA-mediated toxicity would include tRNA (33) and DNA primase (Figure 4D) (47).
An unsuspected finding was that the variant and wild type TS alleles had differential induction of transcripts upon 5-FU treatment in HCT116 cells (Figure 4 and Table 2). The variant allele served as a “tag”, permitting the observation, but need not be causally related. Renewed attention to feedback control of TS gene transcription at the genomic level seems warranted. Perhaps such studies would be especially attractive for certain cell types in which rescue by thymidine rather than uridine can place a dominant role on TS protein as the major determinant of toxicity, or in which the use of antifolates largely removes the possibility of a competing RNA-based mechanism of toxicity.
In summary, our work shows that variables other than TS or TP53, some being novel and others being known but often overlooked, serve as major factors altering the experimental and clinical sensitivity of cancer cells to 5-fluorouracil exposure. Therefore, many proposed biomarkers cannot be expected to readily serve as predictors of treatment response in real-life clinical settings where many salient variables, including those that are not yet discovered, operate simultaneously. The effectiveness of 5-FU may lie in the fact that multiple metabolites and targets exist (2), which also makes predicting a patient's response to this drug potentially impossible. Nevertheless, a genetic understanding of cancer and the molecular aspects that drive tumorigenesis prepares us for a new era of modern drug design that can produce agents that target limited numbers of target molecules (including TS) in cancer cells without off-target effects (48, 49), allowing clinicians to stratify patients based on a limited number of variables. An improved understanding of the targets of 5-FU may permit the further development of derivative drugs that inactivate only a select subset of the 5-FU “target universe.” Folate antagonists are but one example of this line of reasoning.
The authors (J.R.B. and S.E.K.) are governed by the Conflict of Interest policies of the Johns Hopkins University School of Medicine. The authors’ conflict of interest derives from rights they may have regarding intellectual property concerning new electrophoretic conductive media and as owners of Faster Better Media LLC, to which these rights have been assigned.