Molecular Psychiatry (2002) 7, 508–514  2002 Nature Publishing Group All rights reserved 1359-4184/02 $25.00 www.nature.com/mp ORIGINAL RESEARCH ARTICLE Determination of the genomic structure and mutation screening in schizophrenic individuals for five subunits of the N-methyl-D-aspartate NM Williams*, T Bowen*, G Spurlock, N Norton, HJ Williams, B Hoogendoorn, MJ Owen and MC O’Donovan

Department of Psychological Medicine, University of Wales College of Medicine, Heath Park, Cardiff, CF14 4XN, UK

The glutamatergic system is the major excitatory neurotransmitter system in the CNS. Gluta- mate receptors, and in particular N-methyl-D-aspartate (NMDA) receptors, have been proposed as mediators of many common neuropsychiatric phenotypes including cognition, psychosis, and degeneration. We have reconstructed the genomic structure of all five encoding NMDA receptors in silico. We screened each for sequence variation and estimated the allele frequencies of all detected SNPs in pooled samples of 184 UK Caucasian schizophrenics and 184 UK Caucasian blood donor controls. Only a single non-synonymous polymorphism was found indicating extreme selection pressure. The rarity of non-synonymous changes suggests that such variants are unlikely to make a common contribution to common phenotypes. We found a further 26 polymorphisms within exonic or adjacent intronic sequences. The minor alleles of most of these have a relatively high frequency (63% above 0.2). These SNPs will therefore be suitable for studying neuropsychiatric phenotypes that are putatively related to NMDA dysfunction. Pooled analysis provided no support for association between any of the GRIN genes and schizophrenia. Molecular Psychiatry (2002) 7, 508–514. doi:10.1038/sj.mp.4001030 Keywords: candidate ; NMDA; polymorphism; glutamate

Introduction transmission, these include neuronal migration, pro- liferation and excitability, synapse formation, stability The glutamatergic system is the major excitatory neuro- and plasticity, and long term potentiation.4–6 Because transmitter system in the CNS. Glutamatergic trans- of these wide-ranging functions, altered glutamatergic mission is mediated by receptor families that are neurotransmission has been implicated in many differ- classed ionotropic (iGluRs) and metabotropic ent CNS processes, physiological and pathological. (mGluRs). iGluRs are ligand-gated ion channels, which These include learning and learning disability, mem- can in turn be sub-classified into the following groups ory, epilepsy, CNS recovery after trauma or ischaemia, based upon their ligand binding properties: N-methyl neurodegenerative diseases, schizophrenia, affective D-aspartate receptors (NMDA), alpha-amino-3-hyd- disorder, and alcohol dependence.6–9 While more than roxy-5-methyl-4-isoxazole propionate receptors 30 glutamatergic related cDNAs have been cloned in (AMPA), kainate (KA) and more recently delta. The humans, the genomic structure of most has not been majority of iGluRs are thought to be tetrameric or pen- determined experimentally. This has obviously been tameric heteromultimers although the stoichiometry an obstacle to molecular genetic tests of the glutamate and precise composition of native receptors is at hypotheses of the neuropsychiatric disorders men- present unknown.1 In addition to a multiplicity of tioned above. genes, the diversity of iGluRs is increased by the poten- All glutamatergic related genes are good candidates tial for numerous sub-unit combinations of heteromul- for several neuropsychiatric disorders, but region- timers, by alternative splicing, and by RNA editing.1–3 selective knock out studies in animals suggest that Given their diversity, it is not surprising that gluta- NMDA sub-units have probably the strongest a priori matergic neurotransmission has a role in many basic case for several phenotypes including learning and neuronal functions. In addition to fast synaptic memory.7,8 Moreover, reduced expression of GRIN1 leads to behavioural abnormalities in mice that are similar to those observed in pharmacologically Correspondence: M O’Donovan, Department of Psychological induced models of schizophrenia.9 We have therefore Medicine, University of Wales College of Medicine, Heath Park, Cardiff CF14 4XN, UK. E-mail: odonovanmcȰcardiff.ac.uk targeted NMDA receptors as a priority for molecular *These authors contributed equally to the paper. genetic analysis. Five genes encoding NMDA sub-units Received 27 September 2001; accepted 12 November 2001 have been identified in humans. NMDAR1 sub-units Polymorphisms in NMDA subunit genes NM Williams et al 509 are encoded by GRIN1 and are believed to be common Because we were concerned that reagents in the GC- to all native NMDA receptors.10 NMDA receptors are Rich kit might alter the melting characteristics of DNA, also thought to contain one or more NMDAR2 sub- the optimal DHPLC conditions for fragments amplified units from the family of NMDAR2A-D , which using this method were determined empirically as are encoded by GRIN2A-D respectively. To facilitate described elsewhere.16 Where DHPLC analysis sug- candidate gene studies of NMDA receptors, we have gested that a subject was heterozygous, the PCR pro- reconstructed in silico the genomic structure of the 4 ducts from that individual were sequenced to deter- GRIN2 genes to complement the known structure of mine the nature of the polymorphism using the Big Dye GRIN1, and subjected all the genes to mutation analysis Terminator Cycle Sequencing Kit (Perkin Elmer using denaturing high performance liquid chromato- Applied Biosystems, Cheshire, UK) as described by the graphy (DHPLC) and sequencing. manufacturer. All variants were confirmed by allele specific primer extension using SNaPshot chemistry (Perkin Elmer Applied Biosystems) using PCR pro- Materials and methods ducts from a homozygote and a putative heterozygote Genomic structure and PCR as template. The allele frequency of each polymor- The genomic sequence of GRIN1 was previously phism was estimated in 184 UK Caucasian blood reported11 and is available (Acc No: Z32772; Z32773; donors and 184 UK Caucasian schizophrenics by an Z32774). However, the annotation is inaccurate and the accurate method of pooled genotyping based upon sequences specified as exons do not include either 5Ј HPLC analysis of primer extension products17 or by or 3Ј UTRs. The genomic structure for GRIN1 including SNaPshot assay. The latter method has in our hands both the 5Ј and 3Ј UTRs was determined as described proved equally as accurate as the HPLC method below. For GRIN2A-D we identified genomic clones by (manuscript in preparation). aligning each of their respective reference cDNA sequences with the genomic sequence data deposited Subjects in the ‘non-redundant’ and ‘high-throughput sequen- All subjects used in this study were unrelated Caucasi- cing’ GenBank databases using BasicBLAST search12 ans born in the UK or Ireland. All cases met the DSM- (http://www.ncbi.nlm.nih.gov/blast/blast.cgi). BLAST2 IV18 criteria for schizophrenia with diagnosis being sequences13 (http://www.ncbi.nlm.nih.gov/blast/bl2seq/ determined using OPCRIT version 3.3119 following a bl2.html) was then used to perform a gapped alignment semi-structured interview, SCAN or PSE,20,21 and between each cDNA and its corresponding genomic examination of case notes. The sample for mutation clone in order to determine its intron/exon structure. screening consisted of 14 unrelated cases who also had Genomic clones were only selected if they mapped to at least one first degree relative who suffered from the same region as the gene and if each exon had at schizophrenia and were selected randomly from famil- least 95% homology with the reference cDNA ies ascertained for linkage analysis. Allele frequencies sequence. The derived genomic sequences were used were estimated in DNA pools constructed from 184 to design primers using Primer3 (http://www-genome. controls and 184 cases meeting DSM-IV criteria for wi.mit.edu/cgi-bin/primer/primer3Fwww.cgi). In order schizophrenia. Local Research Ethics Committee not to compromise the sensitivity of mutation approval was obtained, and subjects gave written infor- screening, large exons were amplified using sets of med consent to participate. Control individuals were amplimeres of no more than 600 bp that overlapped by group matched to cases for age, sex, and ethnicity from no less than 50 bp. All PCRs were performed on MJ more than 1400 blood donors recruited from the local thermocyclers in a volume of 24 ␮l containing 40 ng of branch of the National Blood Transfusion Service genomic DNA, 10 pmol of each primer, 100 ␮M dNTPs, (Wales). and 0.5 units of Hotstar Taq Polymerase (Quiagen, UK) with a standard touchdown protocol previously Results described.14 Genomic regions that proved particularly difficult to amplify due to their high GC content were cDNA and genomic structure amplified in a volume of 50 ␮l using a GC-Rich PCR According to the published data, GRIN1 has 21 kit (Roche) using reaction and cycling conditions exons.11 Exons 1 and 21 were reported to span 258 and according to the manufacturer’s instructions. Details of 186 bp respectively but this does not include 1093 the primer sequences and PCR conditions can be bases of 5Ј UTR and 1227 bases of 3Ј UTR. With the obtained from our web site (http://psychmed.uwcm. inclusion of UTR sequences, we conclude that exons 1 ac.uk/psychosis/ publications/nmda.html). and 21 actually span 1351 and 1344 bases respectively. We were unable to retrieve genomic clones correspond- Mutation detection ing to the terminal 484 bases of the 3Ј UTR of GRIN1 Mutation detection was performed by screening PCR and therefore cannot exclude the possibility that there products by DHPLC as described elsewhere.14 To achi- are further 3Ј exons. Sequence alignment of cDNA and eve maximum sensitivity, DHPLC analysis was perfor- genomic clones revealed that all GRIN2 genes have 13 med at a minimum of two temperatures determined by exons (Table 1). A similar conclusion regarding the DHPLCMelt software. In previous analyses in our lab- structure of GRIN2B was recently reported by Ohtsuki oratory, this protocol had a sensitivity of 100%.15 et al.22

Molecular Psychiatry Polymorphisms in NMDA subunit genes NM Williams et al 510 Table 1 Predicted genomic structure of GRIN2 genes

Exon

12345678910111213

GRIN2A Size (bp) 138 431 593 115 206 169 154 126 230 161 188 239 3387 UTR (bp) 134 17 ––––––––––1590 GRIN2B Size (bp) 161* 429 599 115 203 172 154 126 230 161 188 239 3431 UTR (bp) 161* 18 ––––––––––1577 GRIN2C Size (bp) 173 414 599* 115 212 166 154 126 230 161 188 233 1569 UTR (bp) 173 15 ––––––––––444 GRIN2D Size (bp) 62 491* 620 115 212 169 154 126 230 161 188 233 1538* UTR (bp) 62 26* ––––––––––203*

The size of each exon of GRIN2 genes, and, where appropriate, the number of bases corresponding to UTR. Each cDNA reference sequence was aligned to its respective genomic clone(s) as follows: GRIN2A – cDNA ref seq: U09002: genomic clones gi͉4878070 and gi͉4235137; GRIN2B – cDNA ref seq: U88963; genomic clones gi͉8567523, gi͉5597031, gi͉5668754 and gi͉4914349; GRIN2C – cDNA ref seq: U77782; genomic clone AC068874.2; GRIN2D – cDNA ref seq: U77783; genomic clones AC011527.1, gi͉701588 and gi͉9690314. Regions that could not be amplified and were not screened for mutations are indicated as*.

Mutation analysis trols as estimated by DNA pooling are presented in In order to perform DHPLC analysis of each gene we Table 3. As part of a comprehensive analysis of the glu- designed 109 PCR fragments. We could not amplify tamatergic system, we propose to fully report all eight of these fragments, and we were therefore unable association data on more than 100 related SNPs in a to screen these mutations. These fragments were all in single publication. In the interim, complete allele fre- GRIN2 genes and are indicated in Table 1. quency data are available at our website We have screened a total of 32 440 bp of genomic (http://psychmed.uwcm.ac.uk/psychosis/publications/ sequence of which 24 201 bp was exonic, 17 374 bp nmda.html) but we note here that none of the alleles was coding, 1516 and 4838 bp were 5Ј UTR and 3Ј UTR even displayed a trend (P Ͻ0.1) for association with respectively, and 8712 bp was flanking intronic schizophrenia. This included the GRIN2B 122CϾG sequence (Table 2). We found a total of 27 polymor- polymorphism which was previously reported to be phisms in the five NMDA receptor genes all of which associated with schizophrenia in a Japanese sample.22 have been deposited in HGBASE (http://hgbase. In our sample the estimated allele frequencies for the interactiva.de/). Ten of these were intronic, 17 were least common allele were 0.43 and 0.44 (P = 0.99) in exonic of which 10 were within coding sequence. The the controls and affecteds respectively. only non-synonymous polymorphism was found in Based upon the mutation data, we have calculated exon 6 of GRIN2C (1656GA;V490I). The details of each the number of variant sites normalised for the number polymorphism together with allele frequencies in con- of and base pairs screened23 (␪). We have

Table 2 Details of mutation screening for each GRIN gene

GRIN1 GRIN2A GRIN2B GRIN2C GRIN2D

Reference cDNA (bp) 5137 6137 6208 4340 4299 CDS (bp) 2817 4392 4452 3708 4008 DNA screened Total exonic (bp) 2817 4392 4452 3109 2604 5ЈUTR (bp) 1093 155 18 188 62 3ЈUTR (bp) 1227 1590 1577 444 0 Intronic sequence (bp) 2312 1960 1814 1012 1614 Total screened (bp) 7449 8097 7861 4753 4280 Nucleotide diversity (␪) and heterozygosity per coding nucleotide (⌸) ␪ 1.82 × 10−4 1.82 × 10−4 2.88 × 10−4 8.26 × 10−5 0 ⌸ 3.06 × 10−4 1.85 × 10−4 4.66 × 10−4 3.05 × 10−5 0

Molecular Psychiatry Polymorphisms in NMDA subunit genes NM Williams et al 511 UTR) UTR) UTR) UTR) UTR) UTR) UTR) Ј Ј Ј Ј Ј Ј Ј (5 (5 (3 (3 (3 (3 (3 0.21 INTRON 11 gctaagggcctacattccc[c]acatcacagacaagggtcct gatttccacagcggagagg tgctcggctttacctttgtc 0.05 INTRON 3 gtgagtggggctggaatggg[a]agggtgtgggagggctccca ggctacgtctggttcatggt ccatttgagccactcattca 0.32 EXON 70.30 gcccacatcagcgacgccgt[g/a]ggcgtggtggcccaggccgtgc accccacgggctctgagt EXON 5 ccacctccccaagagcag gtggaagacatagaccccct[g/a]accgagacgtgtgtgaggaa ttgcctctccagaaatcagc catgctatttcaaagggttgg 0.40 EXON 13 ttaagagaaatgagcttgac[c/a]tttaagagaaatgagctgca attcttttgcccacacttgg cactctgggctcaagtcaca 0.26 INTRON 7 cacccagtgcttctgggcct[g/a]gggcagggaacatgccgagg gtctccacccafgtgcttctg gggacctcccaaaggtattc 0.22 INTRON 1 cagaggaagttggcaagaag[c/a]ctcgtggagggagggggttg ccagggacagacaggaggt gagctaggtgggggtcaag 0.31 EXON 6 acgccctgcgctacgcccc[a/g]gacggtgagtgctgggctt tgggagtgctggagtcct agagcccgtggggtcctc 0.05 INTRON 11 gcacacaggagcgggtaggct[a/g]gacggcgggggtggggacca0.27 gtctggagcccagcagttac cctgcaggcccctctcac EXON 100.44 agcacggagagaaacattcg[g/a]aataactatccctacatgcat cccctcctcccttctcttt EXON 2 gcatgccgagagtcaatttc tcagcacagactctcacccc[c/g]atcctgggcatccacggggg gaatgagaccgacccaaaga tgcaatctggttaccttcca 0.23 INTRON 4 gggtaagacatgtctttgaca[a/g]ttcttgagcaaagtagagta ctgactgccttgttcatggt tagggacaaaagccaaagga 0.38 EXON 13 ccggaccccgcctggagcagc[g]tcctgcgccccctggttctggag cctcgagtccgaggtatgac aaaagcccttgtcgggttta 0.19 EXON 1 ctattcctcttagcccgagga[g/c]gggggtcccaagttacatggc cccgcgggtccacctcag acggagccacagacacaag 0.110.47 EXON 13 EXON 13 gacgaccagtgcttgctcca[t/c]ggcagcaaatcctacttctt aaaggtagcttttcccaaac[t/g]gatcttttcatttaggtgag tgccaacaacaagtcctcag cagcaatgggcatgtttatg agccttaccctcccgtacc gatgcttttgcttcctcacc 0.42 EXON 130.49 gaggcactcttggatcta[c/t]cctgagtatcctccaaactg acatgcaccaacagccgtat EXON 13 ggctccccatgcataagtatt gccctgtaccaacaggtctca[c/t]atcaagcacgggacgggcg gggagtttgacgagatcgag ttcttgcaagcctcacacc 0.360.31 EXON 8 EXON 13 ccttcctcagagccattcag[c/t]gctgacgtatgggtgatgat tcttaatgaactcccccac[c/t]gcaaccatgaacaacacaca0.13 ggtcatttctagcctctctgga gttcccaccccctaaaata atactatggggccggtggt cacctgagggttccttttca INTRON 3 aatgccacctgctgtcggtgg[c/t]ggcggtgaggcagaggcagg aacacaatcaatgccacctg aggcggggacaaaataagag 0.31 INTRON 8 tgagggctgggctcaggggcaa[c/t]gggtccttcagagaatacct gtctccacccagtgcttctg gggacctcccaaaggtattc 0.05 EXON 1 gcgccccactgcatcctcga[c/a]cttctcgggctacaggtacgt accctgcctctccttctctc ccaaggagccctgatct 0.050.05 INTRON 9 gtggttgtcatacacaaaca[g/a]gcatttttttcagaaaatgct EXON 13 ctgctgcgtggttgtcatac gagattaggaagtttggctc[g/a]aacagtttcagctttcttgt tcctcctgaaaggagagcag catgcctactgggtatgttgg tgtagctctccccacctgag 0.05 EXON 6 atggcaagcgggtgcgcggc[g/a]tatggaacggcatgattggg gggccacgacatactctgac agggcatctgagagccacat 0.05 INTRON 2 cgcgggtccacctcagccc[a/g]ccgtgcccccgcctcccgca cccgcgggtccacctcag acggagccacagacacaag = Ͻ ======Ͻ Ͻ Ͻ Ͻ Ͻ frequency GG GG GG GA TT TT AA AA TT GT Ͼ Ͼ Ͼ Ͼ AA Ͼ Ͼ Ͼ Ͼ TG AA AA 65insC Cins Ͼ Ͼ CC Ͼ Ͼ Ͼ – 8A Ͼ Ͼ 70C 69C 19delA Adel 41C + 22A 47G 40A 48A T (S555S) T T (T888T) T A (V285V) A G (P263P) G G (P122P) G A (V4901) A 64 − + C (H1399H) C T (H1178H) C − + − − + − A (L325L) A Ͼ Ͼ − Ͼ A (R695R) G Ͼ Ͼ Ͼ Ͼ Ͼ 243 Ͼ 5765C 5041T 122C Ͼ 5985C 4231G 5077G # # # IVS3 IVS8 IVS3 IVS1 IVS2 IVS11 IVS9 IVS4 IVS7 Polymorphism Allele Location Flanking sequence PCR primer F PCR primer R 544C 960A 1430 1844C 2843C 1026G # # 1656G IVS11 4376T 3717C # # # # # 2240G Polymorphisms in GRIN genes Indicates that the SNP has been previously deposited in dbSNP. Table 3 GRIN1 # GRIN2A GRIN2B GRIN2C GRIN2D

Molecular Psychiatry Polymorphisms in NMDA subunit genes NM Williams et al 512 also estimated average heterozygosity per site (⌸) sian subjects in that study.25 For synonymous changes, assuming all loci are in Hardy–Weinberg equilibrium. ␪ = 1.2 × 10−4 and ⌸ = 2.8 × 10−4, for non-synonymous Total ␪ for all sequence is 2.1 × 10−4 (⌸ = 2.7 × 10−4), changes, both ␪ and ⌸ are zero. The results are there- ␪ for introns is 3 × 10−4 (⌸ = 2.8 × 10−4), ␪ for exons is fore similar to our own analysis of NMDA genes. This 1.8 × 10−4 (⌸ = 2.6 × 10−4), ␪ for synonymous changes implies that NMDA receptors are under particularly is 1.3 × 10−4 (⌸ = 2.2 × 10−4) and ␪ for non-synonymous strong selection pressure to maintain conserved amino changes is 1.5 × 10−5 (⌸ = 5.5 × 10−6). The values of ␪ acid sequences. This is reflected in the high degree of and ⌸ for the coding sequences of each gene are given amino acid between human and in Table 2. mouse, with homology rates of 99%, 95%, 97%, 89% and 89% for NMDAR1 and NMDAR2A-D respectively. Perhaps this is not surprising given their pivotal role Discussion in so many neurophysiological processes. All GRIN2 subtypes have a fairly conserved genomic The design of this study was powered upon the structure of 13 exons, 12 of which contain coding widely held (but unproven) common-disease common- sequence. This is in marked contrast to NMDAR1, polymorphism hypothesis. Thus the number of sub- which has been shown to consist of 21 exons.11 We jects subjected to mutation analysis (chromosomes = were able to design primers for most of the genomic 28) gives power of just less than 0.8 to detect polymor- sequence corresponding to GRIN cDNAs, but were phisms with a minor allele frequency of 0.05, and unable to obtain specific PCR amplification for eight power of 0.95 to detect polymorphisms with a minor fragments (Table 1), despite designing three different allele frequency of 0.1. With regard to the full range of sets of primers and also attempting nested PCR. Poss- putative NMDA related disorders, we would point out ible explanations include pseudogenes, errors present that the use of schizophrenic probands for mutation in ‘working draft’ genomic clone sequences, or difficult screening only reduces power to detect relevant vari- secondary structure. Whatever the explanation, as a ants if the susceptibility alleles for these disorders are consequence, we have failed to screen 1920 bases (or protective against schizophrenia, leading to under-rep- 7.35%) of the genomic sequence corresponding to resentation in our screening set. As far as we are aware, GRIN cDNA. there is no reduction in the prevalence of any common Our data reveal that the sequences of GRIN genes are neuropsychiatric disorders in schizophrenic probands well conserved. Based upon other data,23,24 ␪ for Euro- and therefore this caveat is unlikely to apply. peans has been estimated at around 5 × 10−4 in both Our finding of only a single non-synonymous con- coding and perigenic non-coding sequences. This com- servative polymorphism (VϾI) in GRIN genes suggests pares with our estimate of ␪ for GRIN receptors of 2.1 that the common-disease common-polymorphism × 10−4. This may in part reflect reduced diversity in our hypothesis is unlikely to be applicable to GRIN recep- screening population, which is UK Caucasian. How- tors, at least with regard to variants that change ever, the effect of this is likely to be small as there is sequence structure, and more particularly with regard no reason to regard UK Caucasians as an isolated popu- to schizophrenia. However, our study does not exclude lation. A second explanation may be that our mutation the possibility that there are rare non-synonymous detection protocol is less sensitive than that used by SNPs in GRIN genes that cumulatively contribute to a the other groups. However, evaluation of DHPLC in our small proportion of common neuropsychiatric pheno- laboratory15,16 suggests this is not so. Thus, the third types. If the rare allele hypothesis is correct, generic explanation is most likely correct, that is, GRIN genes programmes of SNP identification in subjects unselec- are under strong selection pressure. ted by phenotype are not likely to detect the relevant In keeping with this explanation, the most dramatic susceptibility alleles. difference in diversity from published values was that While non-synonymous polymorphisms in GRIN for non-synonymous changes. Cargill and colleagues23 genes do not appear to make a common contribution found ␪ for non-synonymous SNPs approximately to common phenotypes, we did find 27 polymorphisms equal to that for synonymous SNPs while ⌸ for non- within GRIN exonic or adjacent intronic sequences. synonymous SNPs was around two thirds that of syn- With the increase in the number of SNPs available on onymous. In contrast, we find that ␪ for non-synony- databases, one question that arises is whether there is mous SNPs in GRIN receptors is almost 10-fold less any point in individual groups continuing to pursue than for synonymous SNPs, while ⌸ is 40-fold less. studies of this nature. For the time being at least, we Others who have recently studied individual members believe that it is. Only around one third of the SNPs of the GRIN family have also found a low rate of non- we report were previously lodged in public databases, synonymous SNPs.22,25 For example when the GRIN1 and without validation and allele frequencies, the gene was screened for mutations in 142 chromosomes SNPs in databases are of limited value.26 We have vali- from the Caucasian population, Rice and colleagues25 dated and determined allele frequencies of all our found no non-synonymous changes and only two syn- SNPs. Given their location and the relatively high onymous changes. A third synonymous change was allele frequency of most of the minor alleles (63% found by genotyping a further 72 Caucasian chromo- above 0.2), these SNPs will be suitable for molecular somes. For comparison with our data, we have re-cal- screening of GRIN receptors.26 Moreover while they are culated the coding sequence diversity values for Cauca- not obviously functional, we cannot exclude the possi-

Molecular Psychiatry Polymorphisms in NMDA subunit genes NM Williams et al 513 bility that they might alter transcriptional efficiency, References mRNA processing, editing, stability and translational efficiency. Thus, the SNPs we report in this paper will 1 Hollmann M, Heinemann S. Cloned glutamate receptors. Annu Rev be of general interest to groups studying a range of neu- Neurosci 1994; 17:31–108. 2 Pin J-P, Duvoisin R. Review: neurotransmitter receptors I. The met- ropsychiatric interest. abotropic glutamate receptors: structure and functions. Neurophar- With regard to schizophrenia specifically, our data macology 1995; 34:1–26. provide no support for the hypothesis that DNA vari- 3 Seeburg PH, Higuchi M, Sprengel R. RNA editing of brain gluta- ation in genes encoding NMDA related proteins con- mate receptor channels: mechanism and physiology. Brain Res Rev 1998; 26:217–229. tribute to susceptibility to schizophrenia, given that 4 Sugiura N, Patel RG, Corriveau RA. N-methyl-D-aspartate receptors none of the 27 SNPs provided even suggestive (P Յ0.1) regulate a group of transiently expressed genes in the developing evidence for association (http://psychmed.uwcm.ac. brain. J Biol Chem 2001; 276: 14257–14263. uk/psychosis/publications/nmda.html). There are 5 Constantine-Paton M, Cline HT. LTP and activity-dependent syn- aptogenesis: the more alike they are, the more different they three main caveats. Our case-control sample does not become. Curr Opin Neurobiol 1998; 8:139–148. have power to exclude very small genetic effects from 6 Cull-Candy S, Brickley S, Farrant M. NMDA receptor subunits: the polymorphisms we did detect. Thus, assuming diversity, development and disease. Curr Opin Neurobiol 2001; 11: effect sizes OR = 1.5, we have power of 0.8 to detect 327–335. ␣ 7 Tsien JZ, Huerta PT, Tonegawa S. The essential role of hippocam- trends for association at our specified cut-off point ( pal CA1 NMDA receptor-dependent in spatial = 0.1) with alleles with a frequency between 0.25–0.7, memory. Cell 1996; 87: 1327–1338. but power of only 0.5 for susceptibility alleles with a 8 McHugh TJ, Blum KI, Tsien JZ, Tonegawa S, Wilson MA. Impaired frequency of 0.1. The second is that very rare schizo- hippocampal representation of space in CA1-specific NMDAR1 knockout mice. Cell 1996; 87:1339–1349. phrenia susceptibility alleles may not be detected by 9 Mohn AR, Gainetdinov RR, Caron MG, Koller BH. Mice with our screening set. It should be noted that the estimates reduced NMDA receptor expression display behaviours related to of power for detecting polymorphisms of given fre- schizophrenia. Cell 1999; 98: 427–436. quencies refer to frequencies in the schizophrenic 10 Moriyoshi K, Masu M, Ishii T, Shigemoto R, Mizuno N, Nakanishi S. Molecular cloning and characterization of the rat NMDA recep- population, not the general population. Our screening tor. Nature 1991; 353:31–37. sample is drawn from families with at least two affec- 11 Zimmer M, Fink TM, Franke Y, Lichter P, Spiess J. Cloning and ted individuals and therefore, under an oligo or poly- structure of the gene encoding the human N-methyl-D-aspartate genic model, is likely to be even more strongly receptor (NMDAR1). Gene 1995; 159: 219–223. 12 Altschul SF, Madden TL, Scha¨ffer AA, Zhang J, Zhang Z, Webb M enriched for schizophrenia susceptibility alleles. et al. Gapped BLAST and PSI-BLAST: a new generation of Therefore our power estimate of 0.8 will actually be database search programs. Nucl Acids Res 1997; 25: 3389–3402. maintained for detecting schizophrenia susceptibility 13 Tatusova TA, Madden TL. Blast 2 sequences — a new tool for com- alleles with even lower frequencies in the general paring protein and nucleotide sequences. FEMS Microbiol Lett 1999; 174:247–250. population than we have estimated above. It is very 14 Austin J, Hoogendoorn B, Buckland P, Speight G, Cardno A, Bowen likely therefore that provided the sum of the fre- T et al. Comparative sequencing of the proneurotensin gene and quencies of all susceptibility alleles in GRIN genes is association studies in schizophrenia. Mol Psychiatry 2000; 5: ෂ0.05, our study should have detected at least one of 208–212. 15 Jones AC, Austin J, Hansen N, Hoogendoorn B, Oefner PJ, Cheadle them. Clearly, until all susceptibility alleles at a locus JP et al. Optimal temperature selection for mutation detection by are known, their number, frequencies, and the total denaturing HPLC and comparison to single-stranded conformation contribution of that locus to disease susceptibility can- polymorphism and heteroduplex analysis. Clin Chem 1999; 45: not be known either. It follows then that no locus can 1133–1140. 16 O’Donovan MC, Oefner PJ, Roberts SC, Austin J, Hoogendoorn B, be fully excluded from contributing to a disease. How- Guy C et al. Blind analysis of denaturing high performance liquid ever, recent modelling provides reassurance for studies chromatography as a tool for mutation detection. Genomics 1998; such as our own.27 Providing the cumulative frequency 52:44–49. of susceptibility alleles at a locus is not small (Ͻ0.01), 17 Hoogendoorn B, Norton N, Kirov G, Williams N, Hamshere ML, Spurlock G et al. Cheap, accurate and rapid allele frequency esti- a few alleles are predicted to account for a high pro- mation of single nucleotide polymorphisms by primer extension portion of the total frequency of susceptibility alleles and DHPLC in DNA pools. Hum Genet 2000; 107:488–493. at that locus.27 The third caveat is that none of the 18 American Psychiatric Association. 1994. Diagnostic and Statistical GRIN cDNAs is guaranteed in GenBank to be full Manual of Mental Disorders, 4th edn. American Psychiatric Associ- length, nor is experimental evidence concerning the ation: Washington DC. 19 McGuffin P, Farmer AE, Harvey I. A polydiagnostic application of genomic sequences of their promoters and other operational criteria in studies of psychotic illness: development important regulatory elements available. Accordingly, and reliability of the OPCRIT system. Arch Gen Psychiatry 1992; we have been unable as yet to screen sequences 48:643–647. involved in regulating the expression of NMDA sub- 20 Wing JK, Cooper JE, Satorius N. The Measurement and Classi- fication of Psychiatric Illness. Cambridge University Press: Cam- units. bridge, 1974. 21 Wing JK, Babor T, Brugha T. SCAN: schedules for the clinical assessment in neuropsychiatry. Arch Gen Psychiatry 1990; 47: 589–593. 22 Ohtsuki T, Sakurai K, Dou H, Toru M, Yamakawa-Kobayashi K, Acknowledgements Arinami T. Mutation analysis of the NMDAR2B (GRIN2B) gene in schizophrenia. Mol Psychiatry 2001; 6:211–216. This study was supported by the MRC (UK). 23 Cargill M, Altshuler D, Ireland J, Sklar P, Ardlie K, Patil N et al.

Molecular Psychiatry Polymorphisms in NMDA subunit genes NM Williams et al 514 Characterization of single-nucleotide polymorphisms in coding flanking regions of the NMDAR1 receptor gene in schizophrenic regions of human genes. Nature Genet 1999; 22:231–238. patients. Mol Psychiatry 2001; 6:274–284. 24 Halushka MK, Fan J-B, Bentley K, Hsie L, Shen N, Weder A et al. 26 Marth G, Yeh R, Minton M, Donaldson R, Li Q, Duan S et al. Single- Patterns of single-nucleotide polymorphisms in candidate genes for nucleotide polymorphisms in the public domain: how useful are blood-pressure homeostasis. Nature Genet 1999; 22: 239–247. they? Nature Genet 2001; 27: 371–372. 25 Rice SR, Niu N, Berman DB, Heston LL, Sobell JL. Identification of 27 Reich DE, Lander ES. On the allelic spectrum of human disease. single nucleotide polymorphisms (SNPs) and other sequence Trends in Genetics 2001; 17:502–510. changes and estimation of nucleotide diversity in coding and

Molecular Psychiatry