A basis for differentiating among the multiple human Mu-glutathione S-transferases and molecular cloning of brain GSTM5.

Specific cDNA probes and antisera were employed to interpret genetic polymorphisms of human Mu-class glutathione S-transferases and to provide a basis for identifying individual forms in human tissues. A cDNA probe that cross-hybridized with various human and rodent Mu-glutathione S-transferase transcripts, hybridized with at least three discrete components by Northern analysis of RNA from human tissue. The smallest (1.3 kb) transcript was identified as the one that encodes GSTM3-3 subunits. A form designated GSTM5, was cloned from a human brain cDNA library and its sequence determined. The open reading frame of GSTM5 shared a high degree of homology with the sequences of other Mu-class glutathione S-transferases, but its 846-nucleotide 3'-noncoding region was unique and considerably larger than that of any of the other Mu forms. Specific synthetic peptide antigens were utilized to distinguish among Mu-class glutathione S-transferases in different tissues of representative individuals. The primary hepatic transcript was that encoding GSTM1-1 with much lesser amounts of GSTM3-3, but livers were devoid of GSTM2-2, and GSTM5-5. Immunoblots confirmed that null-phenotype individuals lacked the GSTM1 gene rather than its GSTM2 homologue that is nearly identical in its exon sequences. The null phenotype therefore was conspicuous in liver, where GSTM1-1 ordinarily was the predominant Mu transcript, but brain and testis contained all four forms. A general strategy was devised to distinguish among and assign primary structures to individual glutathione S-transferases from human tissue.

A Basis for Differentiating among the Multiple Human Mu-Glutathione &Transferases and Molecular Cloning of Brain GSTM5* ( Specific cDNA probes and antisera were employed to interpret genetic polymorphisms of human Mu-class glutathione S-transferases and to provide a basis for identifying individual forms in human tissues. A cDNA probe that cross-hybridized with various human and rodent Mu-glutathione S-transferase transcripts, hybridized with at least three discrete components by Northern analysis of RNA from human tissue. The smallest (1.3 kb) transcript was identified as the one that encodes GSTM3-3 subunits. A form designated GSTM5, was cloned from a human brain cDNA library and its sequence determined. The open reading frame of GSTMS shared a high degree of homology with the sequences of other Mu-class glutathione S-transferases, but its 846-nucleotide 3'-noncoding region was unique and considerably larger than that of any of the other Mu forms. Specific synthetic peptide antigens were utilized to distinguish among Mu-class glutathione S-transferases in different tissues of representative individuals. The primary hepatic transcript was that encoding GSTM1-1 with much lesser amounts of GSTM3-3, but livers were devoid of GSTMZ-2, and GSTM5-5. Immunoblots confirmed that null-phenotype individuals lacked the GSTMl gene rather than its GSTMZ homologue that is nearly identical in its exon sequences. The null phenotype therefore was conspicuous in liver, where GSTM1-1 ordinarily was the predominant Mu transcript, but brain and testis contained all four forms. A general strategy was devised to distinguish among and assign primary structures to individual glutathione S-transferases from human tissue.
By virtue of their capacity to bind certain lipophilic metabolites, hormones, and xenobiotics, and catalyze biotransformations of a wide variety of compounds (1)(2)(3)(4)(5)(6), glutathione Stransferases (GSTs)' are considered to function in cellular *This work was supported in part by Grant CA42448 of the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The abbreviations used are: GST, glutathione S-transferase; GSH, glutathione. Gene loci have been italicized (GSTM5), and gene products or phenotypes are identified by subunit type designations (i.e. GSTM5-5) according to nomenclature in Ref. 40. An additional form GSTM4, has been cloned in other laboratories, but its sequence has not been published at the time of this writing. adaptation to these substances (7). Multiple forms of GSTs provide diversity in ligand binding and catalytic specificities. A widely accepted, species-independent scheme for classifying major mammalian cytosolic GSTs into Alpha, Mu, and Pi categories (8) is based on sequence homologies, subunit assembly patterns, immunological and other common properties. The human Mu class has attracted considerable attention because of notable genetic polymorphisms. In particular, a frequently occurring null phenotype of a human liver and leukocyte Mu-GST has been identified (9)(10)(11)(12)(13). Extensive efforts have been made to correlate the apparent gene deletion with differences among individuals in susceptibility to effects of chemical carcinogens and compounds that cause cytogenetic damage, and in predisposition to certain cancers (6,(14)(15)(16)(17)(18)(19)(20)(21)(22). Differences in tolerance to drugs and influences on the action of some hormones, are other possible physiological consequences of the genetic polymorphisms (7).
To date the complete primary structures of only three human Mu-class GSTs have been determined; these are GSTM1-1 from liver (two alleles) (11,23), GSTM2-2 from muscle (24), and GSTM3-3 from testis (25). The existence of additional Mu-GSTs may be predicted from genomic Southern hybridization patterns (26) and from multiple species of the protein that have been detected in tissues. The Mu-GST counterparts in rats have been studied in detail and found to be differentially regulated in a tissue-specific manner (1). It has become a vexing problem for investigators in this field to distinguish between the multiple Mu-GSTs resolved from different human tissues and to correlate them with GSTs of known primary structure. Indeed, positive identifications of Mu-class isoenzymes have been limited. In this report the cDNA of a new human Mu-class GST from a human brain library has been cloned, and shown to be expressed in a characteristic tissue-specific manner. The identities of other individual Mu-class GSTs have been verified by use of specific antisera, and a sound molecular basis provided for studying physiological consequences of their genetic polymorphism.

Materials
Restriction endonucleases and most of the other enzymes and reagents as well as cloning vectors used for the molecular biology studies were obtained from Boehringer Mannheim, Promega, Sigma, or Pharmacia LKB Biotechnology Inc. Other reagents were of analytical grade. Specimens of testis, liver, and cerebral cortex were obtained from apparently normal human subjects with no evidence of pre-existing disease. The tissues obtained 3-12 h postmortem were snap frozen in liquid nitrogen and stored at -70".

Methods
Molecular Cloning and DNA Sequence Analysis-mRNA isolated from the frontal cortex of a 2-year-old female, was used to prepare a XZAPII cDNA library (Strategene Inc., La Jolla, CA). Twelve 150-8893 mm diameter plates seeded with a total of approximately 2.4 X lo5 phage that developed plaques, were screened using a 0.4-kb EcoRI-KpnI fragment of a rat Y b 3 cDNA clone that cross-hybridizes with other rodent and human Mu-class GST cDNAs (27). cDNA probes used in this study were all uniformly labeled with a multiprime DNAlabeling system (Amersham) using [ c Y -~~P I~C T P (3,000 Ci per nmol) to specific activities of greater than 10' dpm/pg DNA. Hybridizations were performed a t 42 "C in 50% formamide, 5 X Denhardt's solution, 5 X SSPE, and 0.1% SDS. Following hybridization, filters were washed with 0.1 X SSC and 0.1% SDS at 65 "C, Mu-GST-positive plaques were isolated and plaque-purified, and their DNA were isolated (28). cDNA inserts were subcloned into pUC19 plasmids (Bethesda Research Laboratories), and plasmid DNA was purified using NACS-52 PREPAC columns. Both strands were sequenced by the dideoxy chain termination method (29) using modified T 7 DNA polymerase and [ c Y -~~S I~A T P (United States Biochemical). Northern Hybridizations-The following cDNA probes were used for northern blotting: 1) the 0.4-kb fragment of the rat Yb3 clone (27, 30) described above, which was used as a common probe for all Mu-GSTs; 2) a 0.3-kb HindIII-EcoRI fragment from the 3'-end of a human testis clone specific for GSTM3-3 (25); and 3) a 0.35-kb SauI-EcoRV fragment from the 3'-end of GSTM5 (see Fig. 1). Probes were all labeled with [LU-~'P]~CTP as described above.
A multiple tissue Northern blot on a charge modified nylon membrane which contained approximately 2 pg of poly(A)+ mRNA from each of eight human tissues, was obtained from Clontech. The membrane was hybridized with the 32P-labeled probes described above in 5 X SSPE, 10 X Denhardt's solution, 100 pg/ml salmon sperm DNA, 50% formamide, 2% SDS at 42 "C. The membrane was washed with 2 X SSC containing 0.05% SDS at room temperature and then at 50 "C and prepared for autoradiography.
Purification and Analysis of Human GSTs-Cytosolic GSTs were purified from frozen tissue by procedures described previously using GSH-affinity methods (25). Enzymatic activities were measured using 1 mM GSH and l-chloro-2,4-dinitrobenzene as substrates (31). SDSpolyacrylamide gel electrophoresis was performed in 12% acrylamide gels. The gels were either stained with Coomassie Blue or prepared for immunoblotting. For immunoblotting, the proteins were trans- to a nylon membrane. Approximately 2 pg of RNA was in each lane, and the blot was tested with a 2-kb fragment of &actin cDNA as a control. A, a cDNA probe that cross-hybridizes with many human and rat Mu-class GST transcripts (27) was labeled with 32P and hybridized with the membrane (see "Experimental Procedures"). The autoradiography was after a 4day exposure. B. hybridization using the same blot and a cDNA probe specific for GSTM3-3 prepared from the 3"noncoding region of a cDNA GSTM3 derived from testis (25). Exposure was after 7 days. Hybridization conditions and other details are outlined under "Experimental Procedures." Pa, pancreas, from a 30-year-old Caucasian male; K , kidney, from a 19-year-old black female; M, skeletal muscle, from a 35-year-old Caucasian male; L, liver, from a 33-year-old Caucasian female; Lu, lung, from a Caucasian male; P , placenta (unidentified donor); B, brain, from a 57-year-old Caucasian male; H, heart from a 43-year-old Caucasian male. The numbers on the ordinate refer to kilobases of size markers. ferred to nitrocellulose membranes in 12 mM Tris, 96 mM glycine buffer, pH 8.3, containing 20% methanol for 2 h a t 30 V using a Novex semidry trans-blotting apparatus. The membranes were blocked with 5% w/v non-fat dry milk in 20 mM Tris, 150 mM NaCl buffer, pH 7.6 (TBS), and incubated with the indicated antisera for 2 h. After washing with TBS-0.05% Tween 20, blots were incubated with a goat anti-rabbit IgG horseradish peroxidase conjugate, and the color was developed with 0.06% (w/v) 4-chloro-1-naphthol and 0.01% Hz02 in TBS-methanol, pH 7.6.
Peptide Synthesis and Antisera-The indicated oligomeric peptides were synthesized by solid-phase methods using t-butyloxycarbonyl chemistry (32) with an Applied Biosystems Model 430A automated synthesizer. After cleavage of the peptides from the resin by hydrogen fluoride, the synthetic peptides were purified by HPLC, and their structures were verified by amino acid analysis and fast atom bombardment mass spectrometry.
The peptides synthesized with cysteine residues a t their N termini were coupled to keyhole limpet hemocyanin by methods described previously (33). New Zealand White rabbits were immunized with the conjugates in Freund's complete adjuvant and boosted periodically over 3 months.

RESULTS
Tissue Distribution of Mu-GST Transcripts-A cDNA fragment that cross-hybridized with various rodent and human Mu-class GST mRNAs (27) was used to probe steady-state levels of their transcripts in eight human tissues. At least three discernible bands of mRNA in the size range of 1.3-1.6kb were distinguished by Northern hybridization (Fig. 1). The same blot was probed with a sequence-specific fragment from the 3"untranslated region of GSTM3 cDNA that was cloned from human brain and testis expression vector libraries (25). The GSTM3-specific probe hybridized with the smallest of the mRNA species (about 1.3 kb), and was detected in all of the tissues with the exception of placenta. Although heart and brain were particularly rich in this form, evidently GSTM3 gene expression is not so limited as that for the brain Yt,3 subunit in rat (30, 34). In view of evidence for the existence of additional human Mu-class GST genes (11,23,26,35) and for the expression of multiple Mu forms in brain (25, 36-38), a search was undertaken for additional GSTs in human brain.
Cloning and Sequence Analysis-A positive cloned sequence of nearly 1600 nucleotides was isolated from a human frontal cortex cDNA library. This insert, which appeared to be much larger than previously described human Mu-class GST cDNAs, also had a distinct restriction endonuclease digestion pattern. For instance, it featured a Hind111 site (at position 840) and a BglII site (at position 578), neither of which were found in previously described human Mu-class GST cDNAs (11,(23)(24)(25). The complete 1557 nucleotide sequence determined for both strands of the insert, is shown in Fig. 2.
An open reading frame beginning with an ATG initiation codon consisted of 654 nucleotides encoding a protein of 217 amino acids (minus the initial methionine). The 846 nucleotides corresponding to the 3"noncoding region of this cDNA had a polyadenylation signal sequence and is the largest 3'noncoding region for any human Mu-class GST described thus far. An additional 57 nucleotides of the 5"noncoding region contained a consensus translation initiation sequence (CACC) (39) immediately preceding the open reading frame. Although the nucleotide sequence in the coding region is more than 85% homologous to that of the human liver Mu-GSTs (11,23), the 5'-and 3"noncoding regions are largely divergent. It is therefore likely that this brain Mu-GST is the product of a separate gene.
Primary Structures-The deduced amino acid sequence of 217 residues, with an N-terminal proline residue and a Cterminal lysine residue, is consistent with most other human and rodent Mu-class GSTs described thus far. This protein is The primary structure of GSTM5-5 is compared to sequences of other human and rat Mu-class GSTs in Fig. 3. Only three of the first 102 residues of GSTM5-5 differ from those of human liver GSTM1-1 (Thr3, Valz8, and Arg77). Thr3 and Arg77 are common substitutions and are found in the muscle form (GSTM2-2), but Valz8 is unique among human and rodent Mu-class GSTs. In addition, Val"', Met'66, Lys'=, and Leu'78 also distinguish GSTM4-4 from the other GSTs. On the other hand, Lys17' is replaced by Asn and Ser in GSTM1-1 and GSTM2-2, respectively, but is found in GSTM3-3 and the rat isoforms. GSTM5-5 has five additional charged residues compared to the liver GSTM1-1 (1 Arg, 3 Lys, 3 Asp, and 2 fewer Glu residues), and on the basis of its amino acid composition, is expected to be more basic than the other human Mu-class GSTs (calculated p r of 7.1 (PC-Gene, IntelliGenetics, Inc.)). There are also discrete sequence stretches invariant in all the Mu-class GSTs (Fig. 3).
Identification of Mu-class GSTs in Human Tissue-The greatest sequence divergence among the human Mu-GSTs occurs near their C termini (Fig. 3). The exon-intron organization of human Mu-GST genes is probably similar to that of the rat counterparts (35) so the divergent region is likely to be encoded by the last of the 8 exons. Accordingly, synthetic peptides corresponding to unique sequences of GSTM1-1, GSTM3-3, and GSTM5-5 (shown in brackets in Fig. 3) were selected in an effort to raise antibodies specific for each form. The sequence of GSTM2-2 is virtually identical to that of the  GSTM3-3 (25), and rat Ybl, Yb2, and Yb3 (27). Synthetic peptides corresponding to regions shown in brackets were used to raise antibodies specific for each form (see "Experimental Procedures"). GSTM1-1 peptide (one substitution of S for T), and thus the GSTM1-1 antibodies would be expected to cross-react with GSTM2-2. It is also possible that other GSTs, as yet not characterized, may cross-react with these antisera and be grouped with these classes.
To study expression of the GSTs, specimens of different tissues were procured from the same individuals, and GSTs were purified by GSH-affinity chromatographic methods. SDS-polyacrylamide gel electrophoretic patterns and immunoblot analysis of matched brain, liver, and testis GSTs for three representative individuals are shown in Fig. 4. a-class GSTs were abundant in testis and liver, whereas brains were enriched in Pi-class forms (Fig. 4A). Mu-class GSTs were resolved into two well defined bands. Both Mu-bands crossreacted with antisera produced against peptide antigens specific for GSTM1-1 or GSTM2-2 (Fig. 4B). GSTs from liver specimens L1 and L3 exhibited only the upper band, but no immunoreactive component at all was detected from Lp. Evidently Lz in the figure represents tissue from the frequently occurring null-phenotype that was observed with 10 of 18 other liver samples (data not shown). Both brain and testis exhibited two immunoreactive bands with GSTM1-1 or "2-2 antisera; these bands occurred in almost equivalent amounts in B1 and B3, but the lower band (of faster mobility) was a relatively minor form in testis. Only the lower band was observed for the brain and testis GSTs of the "null-phenotype" individuals (see Bz and Tz).
All specimens of brain, testis, and liver, including those from individuals of Mu-null phenotype, had GSTs that crossreacted with antisera specific for the GSTM3-3 peptide (Fig.  4C). GSTM3-3 was a prominent component in brain and testis, but only a minor form in liver (Fig. 4C). GSTM5-5 was also a major GST in brain and testis, but was not detected in liver (Fig. 4 0 ) . Both GSTM3-3 and GSTM5-5 comigrated with the GSTM1-1 band; varying ratios of the two Mu-protein bands in brain and testis indicated whether the tissues were derived from individuals of the null phenotype (see Fig. 4A).
A cDNA probe was prepared from a SauI-EcoRV fragment of the unique 3"noncoding region of the cDNA of GSTM5 (Fig. 2). Northern hybridizations using the GSTM5-specific cDNA probe (Fig. 5A) revealed a single component primarily found in brain and lung transcripts, with lesser amounts in heart. GST5-5 transcripts were not detected in liver, kidney, pancreas, placenta, or skeletal muscle. In contrast to multiple fragments in restriction digests of genomic DNA that hybridized with liver Mu-class cDNA probes (11,23) the GSTM5specific probe hybridized primarily with only a single fragment (Fig. 5B).

DISCUSSION
Multiple forms of mammalian GSTs are necessary to provide diversity in binding of ligands and catalytic specificities; even relatively minor forms probably have selective functions in the cell types in which they are expressed (7). Several fundamental principles concerning the human Mu-family of GSTs are implicit from cloning the GSTM5 cDNA and use of specific peptide antigens described here. Although multiple Mu forms are expressed in many tissues, tissue-specificity is determined by differences in relative amounts of each. For instance, the GST1-1 null phenotype is particularly conspicuous in liver, because it ordinarily is the primary hepatic Muclass transcript for individuals without the gene deletion; liver is virtually devoid of GSTM5-5 and GSTM2-2 and has low levels of GSTM3-3. On the other hand the Northern blots (Figs. 1 and 5) and immunoblots (Fig. 4) show that GSTM1-1, GSTM2-2, GSTM3-3, and GSTM5-5 are all expressed in brain. Therefore, the consequence of the GSTMl gene deletion on detoxification or protective functions associated with GSTs is likely to vary considerably in different tissues.
Northern blots probed with a cDNA fragment from the unique 3"untranslated region, and immunoblots with the specific antisera, show that brain, testis, and lung are relatively rich in GSTM5-5. In earlier studies, the N-terminal amino acid sequence of a Mu-class GST of brain (named GST6) (38) was found to be identical to that of the GSTMB-2 of muscle, but its isoelectric point as well as electrophoretic and immunological properties differed so that its identification was uncertain. In lung two distinct Mu-class GST subunits were resolved by SDS-polyacrylamide gel electrophoresis (41); one of these did not show genetic polymorphism (42). Even though those GSTs were not assigned primary structures, it is likely that GSTM5-5 is a prominent pulmonary Mu form and GSTM3-3 is expressed in lung as well. The tissue distribution of GSTM5-5, however, appears to be more limited than that of the other human GSTs.
The GSTM2 gene was originally cloned from muscle but common reference to this form as the "muscle enzyme" evidently is a misnomer, since GSTM2-2 is a major Mu-subunit in brain and other extrahepatic tissues. Peptide-specific antisera (Fig. 4) may be used to distinguish between GSTM2-2 and GSTM1-1 on the basis of differences in electrophoretic mobilities, even though there is extensive sequence homology of both proteins and of their mRNAs (including the 3'nontranslated region) (24). Moreover, the presence of GSTM2-2 in brain, testis (Fig. 4) and other extrahepatic tissues of individuals lacking the major liver Mu form, confirms the supposition that GSTMl rather than its GSTM2 Identification of Mu-class GSTs in liver, brain, and testis. GSTs were purified from tissues of three individuals and resolved by SDS-polyacrylamide gel electrophoresis. Ll, B1, and TI represent total GSTs purified from liver, brain, and testis of the same individual ( 4 B2, T2 and La, T3, Ta are from two other individuals). Approximately 5 pg of protein was applied to each lane and the indicated electrophoretic mobilities of Alpha, Mu, and Pi forms were determined using specific antisera for each class (data not shown). A, a Coomassie Blue-stained gel; B, immunoblot of the same GST samples with antisera produced using the GSTM1-1 peptide antigen (see Fig. 3); C, immunoblot with GSTM3-3 specific antisera; and D, immunoblot using GSTM5-5-specific antisera.
homologue is deleted in these individuals.
GSTM3-3 is the most distinctive of the Mu-class GSTs because of a lesser degree of sequence homology and because of the additional residues at its C and N termini; nevertheless it was not clearly resolved from GSTM1-1 and GSTM5-5 on SDS gels (Fig. 4C). GSTM3-3 is encoded by the smallest mRNA readily distinguished from the others (Fig. 1). The cDNA for this GST was originally cloned from testis (25) but is widely distributed in human tissue; thus testis, brain, and heart appear to be enriched in this form (Figs. 1 and 4).
Multiple Mu-class GSTs resolved in many tissues probably originate from homo-and heterodimeric combinations of the  Fig. 1 using a Sad-EcoRV fragment from the 3"untranslated region of the cDNA (Fig. 2) as a probe. Autoradiography was after a 10-day exposure. Labels are the same as those in Fig. 1. B, Southern hybridizations with the same probe with restriction digest fragments of human genomic DNA. The DNA was digested with BglII ( B ) , PstI (P), Hind111 ( H ) , and EcoRI ( E ) .
four or more subunits. If assembly is random, there are 10 possible combinations of four subunits, however additional types of subunits probably will be identified (25). Although Mu-class GSTs isolated from human tissue are readily resolved on the basis of differences in isoelectric points, chromatographic elution patterns, and electrophoretic mobilities, it is often difficult to assign individual components to GSTs of known primary structures. For instance, only two Mu-class subunits were detected by HPLC analysis of testicular GSTs (43); the current study suggests that at least four are expressed in testis. Chromatofocusing methods have been used to resolve five Mu-GSTs from heart and aorta (44) and multiple forms from skeletal muscle (45), but assignments of those forms to GSTs of known primary structures remained uncertain. In that respect N-terminal sequence analyses frequently reported for Mu-isoenzymes that are not blocked, are inadequate because of largely invariant sequences in this region (see Fig. 3). However, use of the peptide-specific antisera described here, combined with inherent differences in electrophoretic mobilities on SDS gels alone, permit identification of these four GSTS (Fig. 3) in any tissue.