Molecular Diagnosis of Thiopurine S-Methyltransferase Deficiency: Genetic Basis for Azathioprine and Mercaptopurine Intolerance

Methods

Human Patients and Determination of Phenotype

Through methods described elsewhere [15], erythrocytes and leukocytes were isolated from the peripheral blood of healthy volunteers and children who had acute lymphoblastic leukemia. The volunteers were unselected blood donors who had been identified during a 2-month period, as described elsewhere [15]. The children were being treated at St. Jude Children's Research Hospital or had been referred for evaluation because they could not tolerate chemotherapy. Genotype was determined for all unrelated white patients who had TPM activity that indicated heterozygous or deficient genotypes and for the same number of unrelated persons who had high activity that indicated a homozygous wild-type genotype. We focused our initial studies on white patients because they belong to the ethnic group in which we have identified the largest number of TPM-deficient and heterozygous persons. The activity of TPM in erythrocytes was determined by the radiochemical assay of Weinshilboum and colleagues [22], whose methods we modified, as described elsewhere [15]. The TPM phenotype was assigned on the basis of TPM activity in erythrocytes and according to the criteria of Weinshilboum and Sladek (that is, patients who had <5.0 U/mL of packed red blood cells were considered TPM deficient, those who had 5 to 10 U/mL were considered heterozygous, and those who had >10 U/mL were considered homozygous wild-type) [14]. We used the lowest value of TPM activity in erythrocytes that was measured in each person. We extracted RNA from leukocytes by using the method of Chomczynski and Sacchi [23], and genomic DNA was isolated by chloroform-phenol extractions. The studies were approved by the institutional review board for clinical trials at St. Jude Children's Research Hospital, and informed consent was obtained from the patients or their guardians.

Determination of Intronic Sequences

The presence of a TPM-processed pseudogene [24] that could confound PCR-based genotyping methods and the absence of data on the genomic structure of the human TPM gene led us to initially use PCR primers that were complementary to TPM exon sequences to amplify genomic DNA by Expand PCR (Boehringer Mannheim, Indianapolis, Indiana) and thereby identify intronic sequences in the human TPM gene. The final volume for all PCR assays was 50 micro L. Through use of 1 µg of placental genomic DNA (Clontech Laboratories, Inc., Palo Alto, California) as a template, PCR was done with primers A (5′-GAGTTCTTCGGGGAACATTTCATTG-3′) and B (5′-CACCTGGATTAATGGCAAC TAATGC-3′) in buffer D (Invitrogen, San Diego, California). The buffer contained Tris hydrochloride (pH 8.5), 60 mmol/L; ammonium sulfate, 15 mmol/L; and magnesium chloride, 3.5 mmol/L. The primers had been developed to amplify a fragment of genomic DNA (which included nucleotide 460) for detection of the G460A mutation. The concentration of each oligonucleotide was 0.1 OU/mL (about 0.5 µmol/L), and 0.2 µL Taq polymerase (Perkin Elmer Cetus, Norwalk, Connecticut) was used. With a Hybaid OmniGene thermocycler (Woodbridge, New Jersey), amplification was done for 30 cycles consisting of denaturation at 94 °C for 1 minute, annealing at 55 °C for 2 minutes, and extension at 72 °C for 1 minute. A final extension step at 72 °C for 7 minutes was also done. For the initial cycle, 5 µL of deoxynucleoside triphosphates (dNTP, 10 mmol/L) was added after the temperature reached 80 °C (following the “hot start” protocol). An amplified fragment of 138 base pairs was anticipated in the absence of intron sequences; the resulting fragment of 746 base pairs showed the presence of an intervening intron. This fragment was directly cloned into the plasmid pCR-II (Invitrogen). The recombinant plasmid was purified with Qiagen plasmid kits (Chatsworth, California) and sequenced with an automated sequencer using the cycle sequencing reaction and fluorescence-tagged dye terminators (Prism, Applied Biosystems, Foster City, California). The resulting intron sequence and the intron-exon boundary was then used to develop intron-specific primer P460F.

Through a similar strategy, Expand PCR was used to amplify intron sequences that flanked the exons containing the G238C mutation (intron 4) and the A719G mutation (intron 9). The resulting intron-containing fragments were directly cloned into the plasmid pCR-II; the plasmid was purified and sequenced as described above. These sequences permitted the development of intron-specific PCR primers P2C and P719F for the detection of G238C and A719G mutations.

Detection of TPM Mutations by Polymerase Chain Reaction

Detection of G238C

We used PCR amplification to determine whether the G238C transversion was present at the TPM locus. Genomic DNA, 400 ng, was amplified under conditions similar to those discussed for the intronic sequence except that 2 µL of primer P2W (5′-GTATGATTTTAT GCAGGTTTG-3′) or P2M (5′-GTATGATTTTATGCAGGTTTC-3′) was used with primer P2C (5′-TAAATAGGAACCATCGGACAC-3′) (0.1 OU/mL) in each amplification. Unpurified PCR products were analyzed by electrophoresis in 2.5% MetaPhor gels (MetaPhor Agarose, FMC Bioproducts, Rockland, Maine) stained with ethidium bromide. A DNA fragment was amplified with P2M and P2C primers when C238 (mutant) was present, whereas a DNA fragment was amplified with P2W and P2C primers when G238 (wild-type) was present (Figure 2).

Detection of G460A

To detect the G460A mutation, a PCR assay using primers P460F (5′-ATAACAGAGTGGGGAGGCTGC-3′) and P460R (5′-CTAGAACCCAGAAAA AGTATAG-3′), 0.1 µL (10 OU/mL) of each primer per reaction tube, was done under conditions similar to those discussed above. However, 250 ng of patient DNA was used as a template, and buffer J (Invitrogen) (which contained Tris hydrochloride [pH 9.5], 60 mmol/L; ammonium sulfate, 15 mmol/L; and magnesium chloride, 2.0 mmol/L) was used. The PCR product was desalted by filtration through a Centricon 30 filter (Amicon, Inc., Beverly, Massachusetts) and then digested with Mwo I (New England Biolabs, Beverly, Massachusetts) for 1 hour at 60 °C. Digested products were analyzed by gel electrophoresis. Mwo I digestion of wild-type DNA yields fragments of 267 and 98 base pairs, whereas DNA containing the G460A mutation is not digested and yields an uncleaved fragment of 365 base pairs (Figure 2).

Detection of A719G

To detect the A719G mutation, a PCR assay using primers P719R (5′-TGTTGGGATTACAGGTGTGAGCCAC-3′) and P719F (5′-CAGGCTTTAG CATAATTTTCAATTCCTC-3′) was done under the same conditions as those used for the 460 mutation except that we used buffer N (Invitrogen), which contained Tris hydrochloride (pH 10), 60 mmol/L; ammonium sulfate, 15 mmol/L; and magnesium chloride, 2.0 mmol/L. The PCR products were desalted by filtration, digested with Acc I (New England Biolabs) for 2 hours at 37 °C, and analyzed by gel electrophoresis. The A719G mutation introduces an Acc I restriction site in the amplified fragment and yields fragments of 207 and 86 base pairs. Wild-type DNA yields an uncleaved fragment of 293 base pairs (Figure 2).

Synthesis of Complementary DNA

First-strand cDNA was synthesized [12] from 2 µg of total cellular RNA. The reaction mixture (100 µL) contained Tris hydrochloride (pH, 8.3 at 20 °C), 10 mmol/L; potassium chloride, 50 mmol/L; magnesium chloride, 1.5 mmol/L; 0.001% (weight in volume) gelatin; dNTP, 0.2 mmol/L; RNasin Ribonuclease Inhibitor (Promega, Madison, Wisconsin), 20 U; random hexamers, 200 ng; and Moloney murine leukemia virus reverse transcriptase (Super-Script, GIBCO/BRL, Gaithersburg, Maryland), 200 U. We incubated the reaction mixture at 42 °C for 60 minutes as described elsewhere [19]. The PCR fragments were blunted and cloned into the Sma I site of plasmid pGEM-7Zf(+) (Promega) or were directly cloned in pCR-II (Invitrogen). Resulting plasmids were purified with Qiagen plasmid kits and sequenced.

Data Analysis

The University of Wisconsin Genetics Computer Group (Madison, Wisconsin) software package was used to analyze sequence information.

Results

The TPM phenotype was determined in an unrelated population of 282 white persons that comprised 209 healthy adult blood donors and 73 children with acute lymphoblastic leukemia. Genotype was then determined in all patients who had intermediate TPM activity (n = 21; 7.4% [95% CI, 4.7% to 11.2%]), an equal number of unrelated persons who had high activity (n = 21), and all TPM-deficient patients who were referred to our center (n = 5) or were identified among the 209 healthy volunteers (n = 1). It should be noted that 6 TPM-deficient patients would be expected in an unselected population of about 1800 unrelated persons. The TPM genotype was determined in 23 men and 25 women.

We isolated RNA from the leukocytes of three patients who had TPM deficiency. We then cloned and sequenced their TPM cDNA using methods that have been described elsewhere [19, 20]. In these patients, the TPM genotype that was determined by cDNA sequencing (TPM*2/*2; TPM*2/*3A; TPM*3A/*3A) agreed perfectly with the genotype that was determined from genomic DNA through our PCR methods.

Among the 21 patients who had heterozygous phenotypes, TPM*3A was the most prevalent nonfunctional mutant allele (18 of 21 patients [85%]); TPM*2 (5% of patients) and TPM*3C (5% of patients) were relatively rare. All 6 TPM-deficient patients had two nonfunctional alleles (TPM*3A/*3A [n = 3], TPM*3A/*3C [n = 1], TPM*3A/*2 [n = 1], and TPM*2/*2 [n = 1]). Several patients had a silent mutation in exon 7 (T474C in TPM*1S) (Figure 1) that did not alter the amino acid encoded but did affect the selection of PCR primers for the genotyping method. In other words, primer 460R was selected so that it did not include nucleotide 474 and thus would amplify in the presence of T or C at this position.

As Figure 3 shows, all 21 patients who had high TPM activity (>10 U/mL of packed red blood cells) had a wild-type/wild-type genotype (for example, TPM*1/TPM*1). Twenty of 21 patients who had intermediate activity (5 to 10 U/mL of packed red blood cells) had one nonfunctional mutant allele (TPM*3A [n = 18], TPM*3C [n = 1], or TPM*2 [n = 1]), and all 21 had at least one functional allele (TPM*1 or TPM*1S). Thus, as a diagnostic test for intermediate TPM activity (heterozygous TPM phenotype), this method had a sensitivity of 95.2% (CI, 76.2% to 99.9%) based on data from 20 of 21 patients who had intermediate activity and were successfully identified by genotype. The test specificity was 100% (CI, 83.9% to 100%) based on data from 21 of 21 patients who had high activity and homozygous wild-type genotypes (Table 1).

Discussion

Treating TPM-deficient patients with standard doses of mercaptopurine, thioguanine, or azathioprine can be fatal [9], but such patients can be successfully treated without severe toxicity if the dose is properly adjusted [11-13]. Although TPM deficiency can be diagnosed by measuring TPM activity in erythrocytes, such measurements do not reflect constitutive activity if patients have received transfusions of red blood cells within the past 2 to 3 months. This effect is shown by one of the TPM-deficient patients in our study (genotype TPM*3A/*3A). This patient had a TPM activity of 9.8 U/mL of packed red blood cells 12 days after receiving a transfusion of two units of packed red blood cells and had undetectable activity 4 months later. This patient therefore seemed to be a TPM heterozygote on the basis of TPM activity in erythrocytes (she had received a transfusion of red blood cells from a person with homozygous wild-type TPM activity).

Because transplant recipients or patients who have recently received a diagnosis of leukemia occasionally receive allogeneic transfusions of red blood cells, we have developed molecular genetic methods that are not affected by donor erythrocytes. As a result, these methods provide a more robust method with which to diagnose patients with TPM deficiency or heterozygosity at the TPM gene locus. These PCR-based methods require less than 1 µg of DNA, which is the amount contained in approximately 100 µL of whole blood. The cost of reagents to determine TPM genotype using these methods is less than $100, and the test can be done in a few hours. Because genotype does not change, it only needs to be determined once; thus, these methods are clinically relevant and relatively inexpensive. Furthermore, so-called DNA-chip technology has the potential to completely automate the determination of TPM genotype after genomic DNA has been isolated from a patient [25, 26].

These molecular genetic methods for detecting TPM*2 and TPM*3 alleles produced 98% (47 of 48) concordance of genotype and phenotype in our population of white persons, and all patients who had two of these mutant alleles were TPM deficient. The lack of concordance in a small percentage of patients (2%) may have been caused by the existence of undiscovered, rare mutant alleles at the human TPM locus; this possibility is currently under investigation. For the usefulness of these methods in other populations to be assessed, it is important that future studies determine whether these mutant alleles are the most prevalent alleles in other ethnic groups. Preliminary data from a small group of black persons (n = 9) with intermediate TPM activity indicates that the same TPM mutations (G238C, G460A, and A719G) are present in most heterozygotes in this ethnic group. However, the TPM*3C allele was found in four of nine black persons who had heterozygous phenotypes, which suggests that this may be a more prevalent allele in black persons. Future studies must include more heterozygous and deficient patients to fully elucidate the molecular mechanisms of TPM polymorphism in black persons and other groups.

The A719G mutation was present in 28 of 32 mutant alleles (88%) that were detected in our population of white persons, whereas the G460A mutation was present in 26 alleles (81%) and the G238C mutation was present in 4 alleles (12.5%). Although screening for the A719G mutation alone would detect most mutant alleles, this strategy cannot be recommended because it would not detect the TPM*2 and TPM*3B alleles or the estimated 5% of mutations that have not been isolated in white persons. Moreover, heterologous expression of mutant TPM cDNAs showed less catalytic activity when the G460A and A719G mutations were both present in the same allele [20]. It should be noted that a genotype of TPM*3A/*1 and TPM*3B/*3C would yield the same results with these PCR assays. However, we have not found evidence of the TPM*3B allele (with only the G460A mutation) in any patients or volunteers whom we have studied. The TPM*3C allele (with only the A719G mutation) has been found in two white persons and four black persons and was associated with loss of function in each. Because little additional effort and expense are associated with doing all three PCR-based tests and because these tests should soon be available on automated or multiplex systems, we recommend screening for all three mutations to accurately determine TPM genotype.

Our study elucidates the polymorphic nature of the TPM gene and establishes novel PCR-based methods for detecting the major nonfunctional mutant alleles in white persons, thus providing the first method for making a molecular diagnosis of TPM deficiency. These genotyping methods provide a simple and reliable DNA-based strategy to prospectively identify patients who require a substantial reduction in thiopurine dose to avoid life-threatening hematopoietic toxicity.

From St. Jude Children's Research Hospital and the University of Tennessee, Memphis, Tennessee.

Dr. Loennechen: Department of Pharmacology, University of Tromso, N-9037 Tromso, Norway.

Dr. Pui: Hematology/Oncology Department, ALSAC Tower, Room C-6073, St. Jude Children's Research Hospital, 332 North Lauderdale Street, Memphis, TN 38105.