Folylpoly-γ-glutamate Synthetase Gene mRNA Splice Variants and Protein Expression in Primary Human Leukemia Cells, Cell Lines, and Normal Human Tissues¹

Folate antimetabolites play a central role as therapeutic agents in the treatment of human cancers. Natural folates and classical antifolates like MTX³ are retained in cells as polyglutamates, a process catalyzed by the enzyme FPGS (1, 2, 3). In mammalian tissues, intracellular folates exist as poly-γ-glutamate derivatives with typical peptide chains ranging from five to nine residues (4). Polyglutamation results in increased intracellular antifolate drug concentrations and cytotoxicity (reviewed in Ref. 5). Furthermore, when polyglutamated, some antifolates (e.g., Tomudex and Lometrexol) increase their inhibitory activity against their target enzymes by >100-fold (6, 7). The FPGS gene appears to be controlled by at least two mechanisms: (a) a tissue/lineage-specific (i.e., lymphoid versus myeloid normal hematopoietic cells and human leukemias) and a proliferation-dependent mechanism (8, 9, 10, 11). Among other biochemical parameters, leukemia blasts’ sensitivity to antifolate drugs has been correlated to the intracellular accumulation of drug polyglutamates and the level of FPGS activity (5, 12). In addition, transfection of the FPGS gene into human glioma cells resulted in their increased sensitivity to MTX (13), whereas absent or decreased FPGS activity have been described as mechanisms of resistance to MTX (14, 15).

FPGS enzymatic activity is distributed to both the cytosolic and mitochondrial compartments of mammalian cells. In humans, it was shown that these two FPGS forms are encoded by a single locus and differ by the use of two alternative translational start sites within exon 1 (16, 17). The alternate use of these start sites results in translation of the FPGS protein with or without the addition of a mitochondrial leader sequence. Recently, Turner et al. (18) have reported the existence of two distinct functional murine FPGS isoenzymes. FPGS protein isolated from mouse liver and kidney had different amino terminal sequence compared with FPGS expressed in murine tumors, bone marrow, and intestine. These two isoforms resulted from the exclusive use of two alternative coding regions within exon 1. These proteins exhibited different kinetics when aminopterin and 5,10-dideza tetrahydrofolate (lometrexol) were studied as substrates. In addition, Chen et al. (19) have identified four alternative exon 1 variants using the hepatoma cell line HepG2 (exons 1, 1A, 1B, and 1C), each of which is spliced to exon 2. The exon 1 variant contains the two ATG codons necessary for the initiation of translation of the mitochondrial and cytosolic FPGS forms, and thus appears to encode a functional FPGS protein. Exons 1B and 1C are generated by alternative splicing of intron 1, and exon 1A is spliced upstream of the 5′ terminal region of exon 1. The ATG initiation codon in exon 1B is out of reading frame with codons in exon 2, and exon 1C lacks an in-frame initiation codon. Thus, exons 1B and 1C appear unlikely to encode functional FPGS protein. Although the transcription start site for exon 1A has not yet been identified, its deduced encoded sequence is characteristic of a mitochondrial leader sequence (19). It is thus possible that expression of exon 1A also results in functional FPGS protein.

To date, the existence of these human FPGS splice variants has not been characterized in normal and cancer cells, nor has their significance in gene expression been established. In the present study, we have investigated their expression in normal and transformed human hematopoietic cells using RT-PCR technology. Oligonucleotides within exons 1, 1A, 1B, 1C, and 2 were designed on the basis of published nucleotide sequences, and their mRNA expression was examined in the human leukemia cell lines [AML-193 (AML), CCRF-CEM (T-lineage ALL), HL60 (AML), K562 (erythroleukemia), and Nalm-6 (B-lineage ALL)]. More importantly, we also examined their expression in primary cells from patients with ALL or AML, as well as in human peripheral blood mononuclear cells and normal bone marrow progenitors from normal volunteer donors. Expression of full-length FPGS mRNA was determined by the amplification of cDNA encompassing exons 14 and 15 at the 3′ end of the gene. In this study, we demonstrate that all four alternative exon 1 variants are expressed in all transformed primary cells and cell lines (ALL and AML) as well as in normal human hematopoietic progenitor cells. Using Northern blotting, we demonstrated that human FPGS exon 1 and 1C variants are expressed throughout fetal development and during adulthood. We also confirmed at the molecular level that the FPGS gene is controlled by at least two mechanisms, one tissue-specific and a second that responds to proliferation. Furthermore, comparisons of RT-PCR analysis and Western blotting with enzyme activity suggest that posttranslational modifications may be required for the expression of a functional FPGS protein.

The following human leukemia cell lines were used in this study: Nalm-6 (ALL; B-cell precursor), CCRF-CEM (ALL, T-cell), AML-193 (AML), HL60 (acute promyelocytic leukemia), and K562 (chronic myeloid leukemia; erythroleukemia). The Nalm-6 cell line was obtained from DSMZ (Braunschweig, Germany). AML-193, CCRF-CEM, HL60, and K562 cell lines were obtained from the American Type Culture Collection (Rockville, MD). All cell lines were grown in RPMI 1640 supplemented with 10% fetal bovine serum at 37°C under a 5% CO₂ atmosphere. Culture medium was changed according to standard tissue culture techniques to insure cellular integrity. Trypan blue exclusion methodology was used to assess cell viability.

After Institutional Review Board approved informed consent was signed, leukemia blasts were obtained from children with leukemia at diagnosis. A total of 13 samples were studied: 6 B-precursor ALL (P1, P2, P4, P10, P11, and P12), 5 T-ALL (P5, P6, P7, P8, and P9), and two AML (P3 and P13). Following the same protocol, normal hematopoietic progenitor cells were collected from normal adult volunteer donors. Leukemic blasts and normal hematopoietic cells were separated and purified using a Ficoll-Hypaque gradient before RNA extraction, as described previously (8). Using standard morphological, immunohistochemical, and flow cytometric techniques, leukemia blasts’ phenotype was established as part of the initial diagnostic work-up.

Total RNA was isolated from transformed cell lines, clinical specimens, and samples from normal volunteers using the RNeasy kit from Qiagen, Inc. (Valencia, CA). Purified RNA was subjected to PCR and no product were detected for all RNA extractions, indicating that the RNA preparations were free of DNA contamination. First-strand cDNA was synthesized using 1 μg of total RNA (DNase-treated) in a 20 μl reverse transcriptase reaction mixture containing 1 × avian myeloblastosis virus buffer [50 mm Tris-HCl (pH8.3), 50 mm KCl, 10 mm MgCl₂, 0.5 mm spermidine, and 10 mm DTT], 10 mm each of deoxynucleotide triphosphate, 0.5 μg of random hexamers, 40 units of RNasin RNase inhibitor (Promega Corporation, Madison WI), and 16 units of avian myeloblastosis virus reverse transcriptase (Promega Corporation). The reverse transcriptase reaction was carried out at 42°C for 60 min before a 5-min denaturation step at 96°C. All PCR reactions were performed in a 50-μl volume containing 1 × Qiagen PCR buffer [Tris-HCl, KCl, (NH₄)₂SO₄, and 1.5 mm MgCl₂ (pH 8.7)], 1 × Qiagen Q-solution, 0.2 μm of forward and reverse primers, 0.2 mm each of deoxynucleotide triphosphate and 1.25 U of HotStartTaq DNA polymerase from Qiagen, Inc. PCR conditions for exon 1 amplification were as follows: initial heat denaturation/HotStartTaq activation at 95°C for 15 min and then 36 cycles of denaturation at 94°C for 30 s, annealing at 68°C for 30 s, and elongation for 7 min at 72°C for 35 s. Annealing temperatures and the number of cycles were modified accordingly to optimize PCR amplification of splice variants 1A, 1B, and 1C. Also, a common region encompassing exons 14 and 15 was chosen to amplify the 3′-end nonvariable region of the human FPGS gene (exons 14/15). Thus, amplification of this region is expected to detect the sum of all described 5′-end FPGS mRNA splice variants. Annealing temperatures and number of cycles were also modified accordingly to optimize PCR amplification of exons 14/15. Expected amplification products are as follows: 115 bp (exons 1A/2); 226 bp (exons 1/2); 139 bp (exons 1B/2); 180 bp (exons 1C/2); and 136 bp (exons 14/15). Primers BA67 and BA68 were used to amplify β-actin as described by Lenz et al. (20). RT-PCR products were analyzed on a 2% agarose gel in Tris-acetate-EDTA buffer. Identity of each amplified RT-PCR fragment was confirmed by nucleotide sequence. DNA fragments were purified from 5% polyacrylamide gels and sequenced at the Biotechnology Resource Laboratory of the Medical University of South Carolina (Charleston, SC).

The relative expression of FPGS exon 1 splice variants was analyzed using a modification of an RT-PCR-based quantitation procedure described previously (20). Briefly, serial dilutions of FPGS cDNA were amplified using exon 1 variant-specific primers. RT-PCR products were separated on a 2% agarose gel and stained with ethidium bromide. The expression of β-actin was used as control and was determined for each cDNA preparation. The densitometry of each band was determined using the Gel-Pro analyzer software. The integration densitometry values were plotted against the dilutions of cDNA used in the RT-PCR, and the slope of the linear regression curve was determined using CA-Cricket Graph III software. The relative gene expression of FPGS splice variants was calculated by the densitometry ratio between each exon 1 variant and β-actin.

Oligonucleotides within the exons 1, 1A, 1B, 1C, 2, 14, and 15 were designed based on published human FPGS nucleotide sequences (GenBank accession no. U24252, U24253, U14938, and U14939) and synthesized by Genosys (Fisher Scientific). The primer sequences used in this study were the following:

• (a) 1EXONF, 5′-GGGGGCGCCGGGACTATGTCG-3′ (21 mer);

• (b) 1EXONR-1005, 5′-CAGACGCCGCTGCCAGGAATAGAG-3′ (24 mer);

• (c) 1AEXONF-769, 5′-ATGGTGCACGCGGAAGGG-3′ (18 mer);

• (d) 1BEXONF-1142, 5′-TGGGCGCGACGACACGTGG-3′ (19 mer);

• (e) 1CEXONF-1333, 5′-CCCGGGTTGGGAAGTGGGAAG-3′ (21 mer);

• (f) 1CEXONF-1347, 5′-TGGGAAGTGGCACAGGAGCTAGG-3′ (23 mer);

• (g) 1CEXONR-1422, 5′-CCCACTCTCACTGCCGGTTCAGAC-3′ (24 mer);

• (h) 2EXONR-2402, 5′-CCGCTGGCGCTTCACCTGCTC-3′ (21 mer);

• (i) 2EXONR-2408, 5′-GTCACCCCGCTGGCGCTTCAC-3′ (21 mer);

• (j) 14EXONF, 5′-GCCGTCTTCTGCCCTAACCTGA-3′ (22 mer);

• (k) 15EXONR, 5′-TGCTCTTCGTCCAGGTGGTTCC-3′ (22 mer);

• (l) BA67, 5′-GCGGGAAATCGTGCGTGACATT-3′ (22 mer);

• and (m) BA68, 5′-GATGGAGTTGAAGGTAGTTTCGTG-3′ (24 mer).

Nylon membranes (MTN Blots) containing poly(A)+ RNA from human adult or fetal tissues were obtained from Clontech Laboratories, Inc. (Palo Alto, CA). Total poly(A)+ RNA (1–2 μg) from leukemia cell lines was size fractionated on a 1% glyoxal/DMSO-agarose gel and transferred to BrightStar-Plus nylon membrane (Ambion, Inc., Austin, TX). Splice variants-specific probes were labeled with [α-³²P]dATP using the 3′-end labeling kit (Amersham Pharmacia Biotech, Inc., Piscataway, NJ). Exons 1 (70 bp), 1C (113 bp), 14/15 (136 bp), and β-actin (231 bp) probes were RT-PCR amplified using the oligonucleotide sets 1EXONF/1EXONR-1005, 1CEXONF-1333/1CEXONR-1422, 14EXONF/15EXONR, and BA67/BA68, respectively. Unincorporated nucleotides were removed using a Sephadex G-25 column (Amersham Pharmacia Biotech, Inc.). Hybridization and rehybridization were performed using the Northern Max-Gly Northern blotting kit (Ambion, Inc.). Before rehybridization, the probes were stripped off membranes using boiling 0.5% SDS solution. The absence of the signal was confirmed by exposing the membranes to X-ray BioMax MS film (Kodak, Rochester, NY) for 4 days. The densitometry of each band was determined using the Gel-Pro analyzer software. Relative quantitation of exons 1, 1C, and 14/15 was normalized to β-actin expression.

Cytoplasmic cell extracts were prepared from transformed cell lines (CCRF-CEM and Nalm-6), clinical specimens (P4 and P13), and samples from normal volunteers (BM and PBMNCs) using the NE-PER nuclear and cytoplasmic extraction reagents (Pierce Chemical Company, Rockford, IL). Protein concentration was determined using the micro BCA protein assay reagent kit (Pierce Chemical Company). Western immunoblotting was performed as described by McGuire and Russell (21). Briefly, 50 μg of proteins/well were separated in a 1.5-mm thick NuPAGE 4–12% Bis-Tris minigel (Invitrogen, Carlsbad, CA) using 1× NuPAGE 3-(N-merpholino)propane sulfonic acid-SDS running buffer. Electrophoresis was performed at room temperature at 150 V constant. The gel was electrotransferred to polyvinylidene difluoride membrane (Invitrogen) at 50 V for 2 h 30 min using the 1× NuPAGE Transfer Buffer. The polyvinylidene difluoride membrane was blocked for 4 h in 10 ml of SuperBlock Blocking Buffer-Blotting in PBS (Pierce Chemical Company) containing 0.25% Tween 20. The membrane was incubated overnight at room temperature in 10 ml of SuperBlock Blocking Buffer-Blotting solution containing a 1:10,000 dilution of rabbit polyclonal antibody to FPGS, kindly provided by Dr. John J. McGuire (Grace Cancer Drug Center, Buffalo, NY). The blot was washed twice for 15 min with 12 ml PBS/0.25% Tween 20 and shaken for 90 min at room temperature in 12 ml of SuperBlock Blocking Buffer-Blotting solution containing a 1:5,000 dilution of alkaline phosphatase-conjugated goat antirabbit IgG (Bio-Rad, Hercules, CA). The membrane was washed three times for 15 min each with 12 ml of PBS/0.25% Tween 20 and then once with 12 ml of Tris-buffered saline. The alkaline phosphatase was developed using the chemiluminescent substrate CDP-Star according to the manufacturer’s instructions (Roche Molecular Biochemicals, Indianapolis, IN). The signal was detected after exposure to X-ray BioMax MS film (Kodak).

Four alternative exon 1 variants of the human FPGS gene (1, 1A, 1B, and 1C), each of which is spliced to exon 2 (Fig. 1), have been identified in a hepatoma cell line using 5′-rapid amplification of cDNA ends and primer extension analysis (19). In this study, we have analyzed the expression of these four FPGS exon 1 splice variants in primary human hematopoietic cells (normal and transformed) and human leukemia cell lines using RT-PCR technology. Splice variants 1, 1A, 1B, and 1C were RT-PCR-amplified using the following oligonucleotide sets: 1EXONF/2EXONR-2408, 1AEXONF-769/2EXONR-2402, 1BEXONF-1142/2EXONR-2408, and 1CEXONF-1347/2EXONR-2408, respectively. Total FPGS mRNA expression was analyzed by amplification of the 3′ nonvariable region within exons 14 and 15 using the oligonucleotides 14EXONF and 15EXONR. RT-PCR products of 226 bp, 115 bp, 139 bp, 180 bp, and 136 bp were amplified from alternative FPGS exon 1 variants (1, 1A, 1B, and 1C) and exon 14/15, respectively (Fig. 1). The sequence identities and splicing junctions of each alternative FPGS exon 1 variant and exons 14/15 were confirmed by nucleotide sequence (data not shown).

As shown in Fig. 2, these four alternative FPGS exon 1 variants are expressed in all five human leukemia cell lines analyzed (Nalm-6, CCRF-CEM, AML-193, HL60, and K562). No significant qualitative differences in expression patterns were found among them regardless of their differences in immunophenotype. More important, FPGS exon 1 splice variants are expressed in all 13 primary leukemia samples obtained from patients with B-lineage ALL (6 patients), T-lineage ALL (5 patients), and AML (2 patients). In these clinical samples, no significant qualitative differences were observed with regards to their cell lineage of origin. These exon 1 splice variants are also expressed in human BM progenitors and PBMNCs obtained from normal volunteer donors. In all primary cells analyzed, we found no detectable qualitative differences in the expression of exon 1, 1A, 1B, or 1C variants between normal BM progenitors, mature circulating mononuclear cells, and acute leukemia blasts. Our data demonstrate the expression of all four known human FPGS exon 1 splice variants in cells of the human lymphohematopoietic system. Thus, we have established that alternative splicing occurs at the 5′ end of the human FPGS gene during in vivo transcription of its mRNA.

As reported, no qualitative differences in expression patterns were observed between FPGS exon 1 splice variants in normal and transformed cells. To determine whether semiquantitative differences in the expression of these FPGS exon 1 splice variants exist, we used a modification of a semiquantitative RT-PCR technique described previously (20). Our results indicate that alternative spliced exons 1 and 1C are the more abundant variants in normal and transformed human hematopoietic cells (Table 1). When we analyzed the semiquantitative level of expression of each FPGS exon 1 variant between normal and leukemia cells, no statistically significant differences were found. We were also unable to demonstrate any clear differences when comparing cells according to their hematopoietic lineage of origin. Most notably, no direct correlation was found between the semiquantitative level of expression of exon 1 variant, which contains the two ATG start codons for translation of functional mitochondrial and cytosolic FPGS, and enzyme activity. Similar results were obtained when comparing total FPGS mRNA (exons 14/15) and the level of FPGS activity. A more limited Northern blot analysis of these cells also failed to reveal a direct correlation between FPGS mRNA expression and enzyme activity (data not shown).

To investigate the effect of proliferation on FPGS mRNA expression, we compared rapidly dividing BM progenitors with nondividing PBMNCs. The mean level of FPGS activity decreased from 350 pmol/mg/hr in normal BM progenitors to levels under the detection limits of the assay in circulating PBMNCs (Table 1). In contrast, the level of all FPGS exon 1 splice variants and total mRNA (exons 14/15) decreased by only 20–30% in these same cell populations. These results indicate that FPGS mRNA levels are not predictive of FPGS enzymatic activity in primary human leukemia blasts and normal hematopoietic progenitors. It is intriguing that Nalm-6 cells expressed the highest level of FPGS enzyme activity while expressing the lowest levels of FPGS exon 1 variant mRNA. Nalm-6 cells were also the only cell line in which expression of FPGS exon 1A splice variant was consistently detected. This finding could be explained if exon 1A, or another yet unidentified mRNA splice variant(s), encodes for additional functional FPGS protein.

It has been shown that expression of FPGS activity is regulated by a proliferation-dependent mechanism in which rapidly dividing cells express higher enzymatic activity than quiescent cells (8, 10). During the fetal development of rat brain and liver, higher levels of FPGS activity were observed compared with those same tissues obtained from adult animals (8). In addition, the FPGS levels in liver remain substantially higher than those in brain throughout embryogenesis and postnatal life (8). Thus, expression of FPGS activity in rat tissues appears to be controlled by two mechanisms, one that is tissue-specific and a second one that is linked to cell proliferation. To compare the levels of expression of these alternative exon 1 splice variants and total FPGS mRNA in human fetal versus adult tissues, we have performed Northern blot analysis using specific probes to detect alternative FPGS splice variants 1 and 1C and the region encoding exons 14/15. Exons 1 and 1C were chosen to represent splice variants encoding functional and nonfunctional FPGS protein, respectively. Their level of expression was normalized to the expression of β-actin mRNA in each tissue. As shown in Fig. 3, FPGS mRNA for exons 14/15 was detected in all normal human adult (Fig. 3,a) and fetal (Fig. 3 b) tissues analyzed. We also detected mRNA for exon 1 and 1C splice variants in all tissues studied, except for adult peripheral blood leukocytes. Because exon 14/15 represents the sum of all splice variants, the absence of exon 1 and 1C mRNA could be attributable to levels under the detection limits of our methodology. Instead, a 9.5-kb band was detected in these PBMNCs (data not shown). The identity of this band remains unclear.

A quantitative analysis of these data are presented in Table 2. In adult tissues, FPGS mRNA expression of the region encompassing exons 14/15 was the highest in human skeletal muscle, followed by heart, liver and kidney. By comparison, the highest expression of exon 1 FPGS mRNA was detected in skeletal muscle, whereas liver expressed the highest level of exon 1C mRNA. In contrast, human brain showed the lowest levels of expression of all three mRNA species analyzed. More important, our results establish that the levels of expression of exons 1, 1C, and 14/15 are higher in all human fetal tissues analyzed when compared with their adult counterparts. When analyzing fetal tissues, human brain exhibited the lowest level of FPGS mRNA expression, whereas mRNA levels were higher, but comparable, in liver, kidney, and lung. These findings are consistent with our previous report comparing FPGS activity in rat brain and liver (8). In these same tissues, we did not detect any significant tissue-specific difference between the pattern of expression of exon 1 and 1C mRNA splice variants. We have thus demonstrated at the molecular level that the expression of the human FPGS gene is controlled by at least two mechanisms: (a) one responsible for an increase in mRNA expression in rapidly proliferating cells compared with nonproliferating cells of the same lineage (fetal versus adult tissues); and (b) one that determines FPGS mRNA expression in a tissue-specific manner.

We have previously determined the baseline levels of FPGS enzyme activity expressed in normal human BM progenitors (both lymphoid and myeloid) and mature PBMNCs (9). Cells of lymphoid origin consistently express higher levels of enzyme activity compared with cells of myeloid lineage. In addition, our data demonstrated no detectable FPGS activity in normal mature PBMNCs, whereas normal BM progenitor cells expressed FPGS activity in the expected lineage-specific manner (9). Our RT-PCR (Fig. 2.) and Northern blot (data not shown) findings show that FPGS exon 1 splice variants and total mRNA are detected in both circulating PBMNCs and their BM progenitors. Thus, circulating PBMNCs, in which no detectable FPGS activity was found, expressed all FPGS exon 1 mRNA splice variants. This indicates a lack of correlation between FPGS mRNA and enzyme activity in primary cells of human hematopoietic origin. Further, these findings suggest either a prolonged FPGS mRNA half-life compared with that of the FPGS protein, or that posttranscriptional and/or posttranslational modification(s) may be required for the expression of functional FPGS enzyme.

To determine whether posttranscriptional or posttranslational modifications played a role in determining the expression of functional FPGS protein, we have conducted a Western blot analysis of normal human hematopoietic cells, primary leukemia, and leukemia cell lines. Expression of FPGS protein was detected in all samples analyzed, whether normal or transformed, primary cells or tissue culture cell lines (Fig. 4). Higher protein expression was found in leukemia blasts and cell lines of B-lineage compared with T- or non-lymphoid lineage. Both normal BM progenitors (with detectable enzyme activity) and circulating PBMNCs (with under-detectable enzyme activity) demonstrated FPGS protein expression. Thus, normal circulating PBMNCs express FPGS mRNA and protein, but have levels of enzymatic activity under the detection limits of our sensitive microassay. These results offer additional evidence that, in human lymphohematopoietic cells, translational or posttranslational modifications may be required for the expression of functional FPGS.

FPGS catalyzes the sequential addition of glutamic acid residues onto the γ-carboxyl group of naturally occurring folates and classical folate antimetabolites. The expression of functional FPGS protein has been implicated in the sensitivity of human leukemias to this class of drugs (6, 7, 22). Conversely, its absence has been shown to be a mechanism of drug resistance (14, 15). A single human FPGS gene has been mapped to chromosome region 9q34.1 (19). Roy et al. (23) have reported that certain murine FPGS mRNA splice variants are preferentially expressed in mouse tumors compared with normal tissues. This finding raises the possibility of murine “tumor” versus “normal” tissue FPGS proteins. To date, this is not known to occur during transcription of the human FPGS gene. Four alternative splice variants of exon 1 of the FPGS gene have been described in vitro in HepG2 cells using 5′ rapid amplification of cDNA ends, but their existence and role in primary human cancer cells in vivo have not been established previously (19). The data presented herein is the first report to demonstrate the existence of these FPGS mRNA splice variants in primary human hematopoietic cells. Furthermore, our results establish that alternative splicing of the 5′ end of the human FPGS gene does occur in primary cells in vivo, and in these cells it is not an artifact of in vitro cloning strategies or tissue culture techniques.

It has been suggested that carcinogenesis could induce changes in alternative mRNA processing (24, 25, 26). In the present study we have investigated the expression of the four known human FPGS exon 1 splice variants in normal and transformed cells of the human lymphohematopoietic system. Previously we reported that FPGS activity in normal human hematopoietic progenitors and leukemia cells responds to a lineage-specific control, and not to leukemic transformation (8). Our current RT-PCR data show no striking differences in the pattern of alternative splicing in normal hematopoietic progenitor cells when compared with primary blasts from patients with acute leukemia, thus indicating that leukemogenesis does not influence patterns of alternative splicing either. In addition, the pattern of expression of these variants was comparable between primary blasts and their representative leukemia cell line model. Thus, this study validates these cell lines as suitable in vitro models for future studies aimed at understanding the role of alternative splicing in FPGS gene expression.

The control of functional FPGS protein appears to respond to at least two distinct mechanisms, one that is proliferation-dependent and a second one that determines its expression in a tissue-specific manner (8, 10). In cells of the human lymphohematopoietic system, FPGS activity obeys a lineage-specific control with higher levels of enzyme activity present in normal and leukemia cells of lymphoid versus myeloid origin (9). Our analysis of these representative human leukemia cell models shows that all four alternative exon 1 variants are expressed in cells of both lymphoid and myeloid origin (see Fig. 2). These were also expressed in all primary human leukemia blasts from patients with ALL and AML, and in normal BM cells from volunteer donors. Expression of FPGS splice variants 1 and 1C appears greater when compared with exons 1A and 1B. Interpretation of these data must take into consideration that semiquantitation by conventional RT-PCR poses significant limitations, making direct comparisons between different variants somewhat difficult to interpret. Nonetheless, our data are conclusive in establishing that all four FPGS exon 1 alternative splice variants are present across all cell lineages within the human lymphohematopoietic system.

The semiquantitative amplification of the region encompassing exons 14/15, expected to detect total mRNA including the sum of all 5′-end splice variants, did not correlate with enzyme activity levels (Table 2). A more limited Northern blot analysis (data not shown) confirmed these results. Thus “total” FPGS mRNA levels as determined by conventional RT-PCR were not predictive of enzyme activity. In addition to the technical limitations of RT-PCR, this finding could be expected when using primers that detect the sum of several mRNA splice variants, considering that some may not encode for a functional protein. We thus proceeded to analyze the expression of exon 1 variant, a single mRNA species known to encode for functional cytosolic and mitochondrial FPGS. No correlation with the level of enzyme activity was found either. Our results contradict a previous study in a small number of human leukemia samples indicating that FPGS mRNA expression levels correlated with enzymatic activity (20). Nonetheless, they are in agreement with other reports that have established decreases in FPGS activity without any changes in mRNA levels during selection of antifolate-resistant cells (27). We thus conclude that FPGS mRNA expression should not be used as a surrogate method to estimate enzymatic activity levels or to predict the likelihood of response to treatment with antifolate drugs.

Our data analyzing the expression of total FPGS mRNA (exons 14/15) and splice variants 1 and 1C by Northern blotting, definitively establishes the tissue-specific expression of this gene. As shown in Table 2, fetal and adult brain express low levels of mRNA, whereas fetal and adult liver and kidney express 2- to 3-fold higher levels of mRNA. Previously, we have reported using a sensitive enzymatic microassay that liver is among the tissues which express higher levels of mRNA, whereas brain expresses very low levels (8). What is intriguing is that skeletal muscle, heart, and kidney also express high levels of FPGS mRNA, whereas in our experience (data not shown) and that of others (28), these tissues demonstrated low levels of enzymatic activity. Whether this discrepancy reflects a lack of correlation between FPGS mRNA levels and functional FPGS protein or results from underestimation of enzymatic activity because of instability of the FPGS protein remains to be determined. Our data also establish that proliferating and less differentiated fetal tissues express higher FPGS mRNA compared with their adult counterparts. This is true when analyzing the 3′-end region encompassing exons 14/15 (total FPGS mRNA) or the 5′-end alternative splice variants 1 and 1C. Whether these differences result exclusively from differences in proliferative rates between fetal and adult tissues or also reflect differences between immature fetal and mature adult tissues remains unclear.

The specific elements that control the expression of the FPGS gene are not yet fully understood. The use of alternative transcriptional start sites has been reported to result either in mitochondrial or cytosolic FPGS protein. Nevertheless, it appears that human FPGS responds both to transcriptional control and to posttranscriptional and/or posttranslational modifications that play a role in the expression of functional protein, at least in some tissues. The overall lack of correlation between mRNA levels and enzyme activity suggests the latter. Our Western immunoblotting data indicating the presence of FPGS protein in fully differentiated PBMNCs in which no enzymatic activity was detected indicate that at least posttranslational modifications play a role in controlling enzymatic activity. In analyzing the expression of the FPGS protein, FPGS gene copy, and mRNA size and level from MTX-resistant cell lines, McGuire and Russell (21) also suggested that FPGS is regulated by translational and/or posttranslational mechanisms. In contrast, our data show that undifferentiated and rapidly growing hematopoietic cells (i.e., Nalm-6 cells, ALL and AML blasts, and normal BM progenitors) expressed protein levels consistent with their levels of enzymatic activity. The role of cellular proliferation, cell cycle dependency, and degree of cellular differentiation have not been established. These cellular events and other specific control mechanisms of the human FPGS gene in lymphohematopoietic cells are currently under investigation in our laboratory.

We thank Drs. Gilles M. Leclerc for helpful discussion and assistance with the Western immunoblotting and John J. McGuire for providing the rabbit polyclonal antibody to FPGS. We also thank the Medical University of South Carolina, Biotechnology Resource Laboratory, Charleston, SC, for their assistance with DNA sequence analysis.