Molecular Psychiatry (2009) 14, 621–630 & 2009 Nature Publishing Group All rights reserved 1359-4184/09 $32.00 www.nature.com/mp ORIGINAL ARTICLE Evidence for HTR1A and LHPP as interacting genetic risk factors in major depression CD Neff1, V Abkevich1, JCL Packer2, Y Chen1, J Potter1, R Riley1, C Davenport1, J DeGrado Warren1, S Jammulapati1, A Bhathena2, WS Choi2, PE Kroeger2, RE Metzger2, A Gutin1, MH Skolnick1, D Shattuck1 and DA Katz2 1Myriad Genetics, Salt Lake City, UT, USA; 2Abbott Laboratories, Abbott Park, IL, USA

The HTR1A À1019C > G genotype was associated with major depression in the Utah population. Linkage analysis on Utah pedigrees with strong family histories of major depression including only cases with the HTR1A À1019G allele revealed a linkage peak on chromosome 10 (maximum HLOD = 4.4). Sequencing of all known genes in the linkage region revealed disease-segregating single-nucleotide polymorphisms (SNPs) in LHPP. LHPP SNPs were also associated with major depression in both Utah and Ashkenazi populations. Consistent with the linkage evidence, LHPP associations depended on HTR1A genotype. Lhpp or a of a collinear brain-specific transcript, therefore, may interact with Htr1a in the pathogenesis of major depression. Molecular Psychiatry (2009) 14, 621–630; doi:10.1038/mp.2008.8; published online 12 February 2008 Keywords: genetic linkage; genetic association; major depression; serotonin receptor; epistasis

Introduction Currently available antidepressants work primarily by increasing the activity of the neurotransmitters, Mood disorders, of which major depressive disorder serotonin and norepinephrine, in synapses.5 Some (MDD) is the most common, affect one person in five medications inhibit the degradation of these mole- during their lifetime. The World Health Organization cules, others decrease removal of neurotransmitters estimates that depression is currently the fourth most from the synaptic space, and some medications important worldwide cause of disability-adjusted life stimulate norepinephrine release or inhibit negative year loss, and that it will become the second most feedback of serotonin signaling. Because these med- important cause by 2020.1 Pharmaceutical treatment ications are all based on a single principle, the of depression is frequently inadequate. Using the strength and range of their efficacy is similar. The current best treatments, one-third of patients or more improvements of the last half-century have involved do not achieve remission even after several months of the development of safer and more tolerable drugs. treatment.2 When today’s drugs do help patients However, despite this, today’s drugs are neither achieve remission from their depression, the onset completely safe nor completely tolerable for many of action is over a period of weeks, and there appears patients. There is considerable need for new drugs to be an increased risk of suicide during initial that are effective in a broader range of patients antidepressant therapy, although this risk may be less (particularly for patients whose depression is resis- than that just prior to therapy initiation.3 Further- tant to available pharmaceuticals), that have a faster more, there are high recurrence rates—approximately onset of action, that are safer and more tolerable or 85% of patients who achieve remission will suffer that complement the efficacy of existing drugs. It is another episode of MDD.4 Finally, currently available unlikely that a substantial further improvement in antidepressants are associated with side effects that any of these dimensions will be achieved through the lead some patients to stop taking their medications at development of additional serotonergic or noradre- risk of sinking back (further) into depression, and to nergic agents. morbidity in others.5 Identifying the underlying genetic components of depression can open new avenues for the develop- ment of novel depression drugs. Specifically, if a Correspondence: Dr D Shattuck, Myriad Genetics, 320 Wakara genetic variant segregates with MDD in families, the Way, Salt Lake City UT 84108, USA. gene in which that variant occurs is likely to be E-mail: [email protected] involved in the pathobiology of disease. Such a gene Dr DA Katz, Abbott Laboratories, 100 Abbott Park Road, R48B/ can be a target for development of novel antidepres- AP6C-6, Abbott Park IL 60064, USA. E-mail: [email protected] sants, or lead to identification of previously unknown Received 9 July 2007; revised 14 September 2007; accepted 23 physiological pathways that can be modulated for September 2007; published online 12 February 2008 effective therapy of depression. Interacting genetic risk factors in MDD CD Neff et al 622 Materials and methods used to amplify LHPP exons are shown in Supple- mentary Materials and methods. All samples were Subjects and linkage analysis amplified with Taq Platinum (Invitrogen, Carlsbad, The ascertainment and characteristics of a majority of CA, USA) DNA polymerase. PCR cycles included an the linkage pedigrees has been described.6 Informed initial denaturation at 96 1C (12 s), annealing at 57 1C consent was obtained from each subject. The institu- (15 s) and extension at 72 1C (30–60 s). Excess tional review boards of Intermountain Health Care primers and deoxynucleotide triphosphates from and Valley Mental Health approved the research M13-tailed PCR products were digested with exonu- protocol. Two meaningful differences between the clease I (United States Biochemicals, Cleveland, present study and our previously published work are OH, USA) and shrimp alkaline phosphatase as following: first, that the definition of affected (Amersham, Piscataway, NJ, USA). PCR products status in this study did not include bipolar disorder; were sequenced with M13 forward and reverse second, that 22 additional pedigrees were ascertained fluorescent Big Dye-labeled primers (Applied using the previous criteria. The average pedigree size Biosystems, Foster City, CA, USA) on Applied was 16 individuals (range 6–55) including a mean of Biosystems 3730 sequencers. 8 affected individuals (range 4–31). The analysis leading to discovery of the chromosome 10 linkage Statistical analysis region included 1054 individuals (744 female subjects Based on several prior associations with MDD or and 310 male subjects) with MDD (DSM-IV-TR related phenotypes,8–15 the one-tailed Fisher’s exact sections 296.2x or 296.3x). Affected individuals were test (a = 0.05) was used to test the association of major genotyped and genome-wide linkage analysis was depression in the Utah population with HTR1A performed as described,6 except that only individuals –1019G allele or –1019GG genotype. carrying one or two HTR1A À1019G alleles were To establish the evidence that LHPP acts as a assigned affected status. depression gene in the chromosome 10 linkage region, allele frequencies on the 22 chromosomes Polymorphism discovery in linked pedigrees that cosegregated with MDD were compared, using Genomic DNA from members of the MDD pedigrees the two-tailed Fisher’s exact test (a = 0.05), to frequen- with the strongest linkage evidence was resequenced. cies in three sets of control chromosomes. A first set From each pedigree, two female subjects with MDD, of 60 control chromosomes included the 12 non- carrying one or two HTR1A À1019G alleles, and segregating chromosomes from genotyped, affected sharing a chromosome 10 haplotype that cosegregates pedigree members and 48 chromosomes from 24 with the disease were selected. This allowed direct unrelated individuals from Utah CEPH pedigrees. determination of whether each variant found during The second set of 180 control chromosomes was from resequencing cosegregated with the disease. In the 90 additional Utah CEPH pedigree grandparents. The case of common variants, using this strategy to predict third set of 708 control chromosomes was from 354 cosegregation could overestimate the actual cosegre- unrelated samples collected in Utah. Replication was gation of the variant with disease, but for such used, rather than a-level adjustment, to avoid an variants resequencing of additional family members inflated type I error rate. was undertaken to verify that the variant was indeed For population association analyses, cases were on the disease associated haplotype. Haplotype individuals with confirmed diagnosis of MDD sharing was determined using the program MCLINK7 (DSM-IV-TR sections 296.2x or 296.3x). Ashkenazi during linkage analysis. samples were individuals who reported four grand- Thirty-six genes (Supplementary Table S1) were parents of Ashkenazi descent. These samples were identified within the minimal recombinant region collected by PrecisionMed (Solana Beach, CA, USA). using both publicly available NCBI human genome Utah and Ashkenazi individuals ascertained without assemblies and assembly of publicly available se- respect to depression status were used as controls. quence data using a proprietary algorithm (Compu- Initial logistic regression models included LHPP gen, Israel). Several genes were extended using rapid genotype, HTR1A genotype, gender and both geno- amplification of cDNA ends. type-by-genotype and genotype-by-gender interaction PCR amplification was used to generate products to terms. LHPP and HTR1A genotypes were each screen for segregating variants in all exons and dichotomized such that risk allele carriers (hetero- proximal promoter regions of the identified genes. zygous or homozygous) were compared to non- When possible to conserve DNA, a primary amplicon carriers. Final logistic regression models were was created using genomic DNA (10–20 ng) and a 25- derived through a stepwise backward elimination cycle amplification. The primary amplicon was then process using a = 0.10. The main effect of LHPP diluted 45-fold and used as template to amplify genotype was the primary comparison of interest. several secondary amplicons using nested M13-tailed No a-level adjustment was made for testing several primers for an additional 25 cycles. However, markers within a single gene. Logistic regression odds many amplicons were sequenced after only a single ratios (OR) and 95% confidence intervals (CI) were round of amplification with M13-tailed primers calculated using JMP 5.1 (SAS Institute, Cary, using 10 ng genomic DNA and 35 cycles. All primers NC,USA).

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 623 Phylogenetic tree construction Two exon-bridging reverse transcription-PCR ex- A database of human Refseq protein sequences16 was periments were conducted to learn whether the searched using a hidden Markov model that describes unique exon of EF151005 also occurred naturally in the gene family (Pfam00702) and the Hmmersearch alternatively spliced forms of LHPP. In a first experi- algorithm.17 Sequences matching the hidden Markov ment, PCR was conducted using a forward primer model were aligned using the same model and within an upstream exon of LHPP (TGCAAGCGA Hmmeralign.18 A phylogenetic tree (Supplementary TAGGAGTGGAA) and a reverse primer within the Figure S1) was constructed from gap-free regions of unique exon of EF151005 (CCACCCCATGCCATCAA). the mutiple sequence alignment by a protein distance In a second experiment, PCR was conducted using the matrix method using the Protdist and Fitch algo- same forward primer and a reverse primer in the exon rithms from the Phylip software.19 Sixty amino-acid common to LHPP and EF151005 (CACGTACCCAT positions from the alignment were used. The tree was CAGCCTTCAC). Amplification products were se- visualized using Treeview.19 quenced as above to determine whether any unique sequence from EF151005 was included. Each of these Northern blot experiments was conducted on cDNA prepared from A pre-made poly(A) RNA Northern blot (Ambion, each of three separate lots of human spinal cord Austin, TX, USA) was probed with amplified DNA mRNA (BD Biosciences Clontech). produced by PCR using the following primers. Forward: GAATCTCCCAAATCCCAGAACTCA Results Reverse: ACACCGGGCATGACACCTTCAAGT Amplified DNA was labeled using an AmbionStrip- Serotonin receptor 1A-conditional genetic analysis EZ DNA kit (Ambion) and a-32P-labeled dATP. The The serotonin receptor 1A (Htr1a) is a therapeutic blot was hybridized overnight at 42 1C in ULTRAhyb target in the management of depressive and anxiety Ultrasensitive Hybridization Buffer (Ambion), disorders.21 A common promoter polymorphism, washed twice for 15 min at low stringency (2 Â SSPE, HTR1A À1019C > G, has been described.22 Results of 0.1% SDS) and twice for 15 min at high stringency in vitro experiments suggest that the minor allele (0.1 Â SSPE, 0.1% SDS). All procedures were carried (À1019G) prevents binding of a transcriptional repres- out according to the manufacturer’s instructions. sor, resulting in enhanced receptor expression.8 The À1019G allele or the homozygous À1019GG genotype Quantitative reverse transcription PCR has been associated with depression, suicide, bipolar Human total RNAs were purchased from either disorder, panic disorder with agoraphobia, neuro- 8–15 Ambion or BD Biosciences (Franklin Lakes, NJ, ticism and decreased antidepressant response. USA). Reverse transcription and PCR were conducted We tested the hypothesis that the HTR1A À1019G using the Invitrogen Platinum Thermoscript One Step allele increases depression risk in a Utah population. System qRTPCR kit following the manufacturer’s Observed frequencies of À1019G allele (350/688) and instructions. Fifty nanograms of DNase-treated total À1019GG genotype (89/344) were 1.10- and 1.34-fold RNA were used as a template for each reaction. All higher among unrelated individuals affected with threshold cycle readings were normalized to 28S MDD compared to unaffected individuals (312/672 rRNA. and 65/336; one-tailed Fisher’s exact test P = 0.05 and A schematic of LHPP transcripts is shown in Figure 4. 0.02, respectively). There was no significant differ- The primers and probes used are provided as ence in allele frequencies between genders in cases or Supplementary Materials and methods. controls. Thereafter, we conducted genome-wide linkage analysis conditioned on carriage of the Characterization of EF151005 HTR1A À1019G allele. Only individuals with MDD IMAGE clones h3175509, h5194531, h5197955 and also carrying one or two copies of the HTR1A and h4565014 were obtained from the American À1019G risk allele were considered affected. Type Culture Collection (Manassas, VA, USA). DNA In this HTR1A-conditional linkage analysis using a sequencing was performed with M13 forward dominant genetic model, we observed evidence of and reverse fluorescent (Big Dye, ABI) dye-labeled linkage on chromosome 10 with a maximum HLOD of primers. 2.9 at D10S1222 and D10S1676 (Figure 1a). Using a RNA -mediated rapid amplification of cDNA gender specific model that considered only affected ends was performed using the FirstChoice female subjects, the maximum HLOD was 3.1 at RLM-RACE kit (Ambion) on 10 mg human spinal D10S1222 (Figure 1b). There was no linkage evidence cord total RNA (BD Biosciences Clontech, Palo Alto, in this region when only affected male subjects were CA, USA) according to the manufacturer’s instruc- considered in the analysis (Supplementary Figure tions. Nested PCR used the following gene-specific S2). The linkage evidence on chromosome 12 in primers: analyses considering only affected male subjects, Outer: TCTCCCACTGTATGCTCCTTCCA with a maximum HLOD of 3.3 at 12-MYR0332, Inner: CTCTGCCACTTCATCTGCAGGT corresponds to our previously described linkage Products were separated by electrophoresis and in this population of MDD to apoptosis protease sequenced as above. activating factor 1.6,23 We observed an even higher

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 624 3.5 4.5 4 3 3.5 2.5 3 2.5 2

HLOD 2

HLOD 1.5 1.5 1 1 0.5

0.5 0 140 150 160 170 0 Position (cM) 1234516677 8 9 10 1112 131415 1 181920 22 X Chromosome 21 Figure 2 Linkage analysis of chromosomal region 10q26.2–10qter with additional markers (HTR1A-condi- 3.5 tional, dominant model and female-only). Black dots correspond to the position of markers used in the analysis. 3 The dashed line indicates the region with boundaries determined by recombination events in the pedigrees with 2.5 the strongest linkage evidence. The arrow indicates the approximate position of LHPP. 2

HLOD 1.5 but still using a female-only dominant genetic model we observed to an extent lesser linkage evidence 1 with a peak HLOD of 3.4 at D10S214 (Supplementary

0.5 Table S3). When the affected individuals included only female subjects carrying at least one HTR1A 0 À1019G allele, the HLOD increased and the linkage 1234516677 8 9 10 1112 131415 1 181920 22 X 21 region narrowed (as defined by a drop of HLOD of Chromosome either one or two from the peak value), better Figure 1 Results of genomic search for loci linked to MDD. localizing the disease gene. More importantly, Multipoint HLOD scores for the dominant, HTR1A À1019G- conditioning on HTR1A revealed linkage evidence conditional model are plotted on the y axis, and marker in a distinct set of pedigrees. Further investigation of positions (in cM) are plotted on the x axis. Vertical dashed pedigrees that showed linkage evidence dependent lines delimit the chromosomes. (a) HLOD for gender-neutral on the HTR1A À1019G allele was crucial to the linkage. (b) HLOD for females-only linkage. MDD, major discovery of single-nucleotide polymorphism (SNPs) depressive disorder. segregating with major depression in LHPP. Each gene in the linkage region (Supplementary HLOD of 4.6 at 12-MYR0332 without considering an Table S1) was resequenced using genomic DNA from interaction with HTR1A À1019G, thus we assume 32 affected female subjects—two representatives from that there is no interaction between apoptosis each of the 16 pedigrees. These pedigrees were protease activating factor 1 and HTR1A. No other selected on the basis of a familial HLOD of at least significant or suggestive linkage evidence was ob- 0.4. Among these pedigrees, six had not shown served using a dominant model restricted to male linkage evidence without conditioning the model on subjects. There was no significant or suggestive the HTR1A À1019G allele. The frequencies of variant conditional linkage evidence using gender-neutral or alleles of the chromosome 10 linkage region among gender-specific recessive genetic models (Supple- the 22 chromosomes that segregated with MDD mentary Figures S3–S5). Genome scan results are (‘linked chromosomes’) within these pedigrees were provided in Supplementary Table S2. compared to the frequencies among 60 control On inclusion of data from a denser marker set chromosomes (Table 1). A risk allele significantly across the central 26 cM interval (D10S1237- (two-tailed Fisher’s exact test, a = 0.05) overrepre- D10S1700), linkage evidence increased to a peak sented on linked chromosomes was identified for HLOD of 4.4 at D10S575 when the model was seven SNPs within LHPP and six SNPs in other genes restricted to female subjects (Figure 2; Supplementary within the region. A statistical trend was observed for Table S3). Linkage evidence did not improve with an eighth LHPP SNP but not for any other SNP in additional marker data in a gender-neutral model another gene. Of the LHPP SNPs, two pairs (rs2459213 (Supplementary Table S3). We, therefore, chose to and rs2491156, rs10794134 and rs3824810) were each continue gene discovery efforts focusing on the in complete linkage disequilibrium, and so only one of pedigrees contributing to the female-only linkage each pair was analyzed further. Variant allele frequen- evidence. Without considering the HTR1A genotype, cies among the chromosomes that segregated with

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 625 MDD were next compared to the frequencies in a families that showed evidence of linkage on chromo- separate set of 180 control chromosomes (Table 1). some 10. Logistic regression models included LHPP Variant alleles of only three SNPs, all in LHPP,were and HTR1A genotypes, gender and interaction terms significantly overrepresented on the chromosomes that (genotype-by-genotype and genotype-by-gender). To segregated with MDD in this comparison. These three reflect the dominant genetic model of LHPP linkage LHPP SNPs, and two others, were significantly over- on the HTR1A À1019G allele, both genotypes were represented on the disease chromosomes compared to grouped into dichotomous variables such that carriers a third set of 708 control chromosomes. This is the of a risk allele (heterozygous or homozygous) were primary evidence supporting LHPP as the gene under- compared to non-carriers. lying the MDD linkage evidence on chromosome 10. Unrelated cases were selected from the Utah LHPP haplotypes on the chromosomes segregating families used for linkage analysis that did not show with disease were quite diverse. Fifteen distinct SNP evidence of linkage on chromosome 10, or that were haplotypes segregated with MDD among affected too small to be considered in the linkage analysis. In female subjects in the 16 pedigrees, and no haplotype this Utah sample set, two markers were associated was common to more than 3 pedigrees (Supplementary with MDD: rs10794134 (adjusted OR for the T allele Table S4). This pattern is consistent with multiple 1.40, 95% CI 1.00–1.94, P = 0.05) and ss68074662 origins or allelic heterogeneity of functional variants. (adjusted OR for the A allele 2.03, 95% CI 0.99–4.48, To further support the relationship between LHPP P = 0.05). For each marker, the frequency of LHPP risk genotypes and MDD, genetic association studies allele carriage was highest among HTR1A À1019G- comparing genotype frequencies between individuals positive cases and approximately equal among all affected with MDD and healthy controls were per- other groups (Table 2). Additionally, the same LHPP formed in two populations, neither including the alleles were both linked to and associated with MDD

Table 1 Linkage evidence for LHPP on chromosome 10

Gene SNP Linked Control Set 1 P Control Set 2 P Control Set 3 P

LHPP rs2491156 10/22 5/60 0.0004 45/176 0.07 105/356 0.15 LHPP rs4246100 10/22 12/58 0.02 32/178 0.01 121/684 0.003 LHPP rs10794134 13/22 17/60 0.02 61/178 0.03 184/686 0.002 LHPP rs12265012 14/22 25/60 0.09 73/178 0.07 251/696 0.01 LHPP ss68074662 4/22 1/60 0.02 4/178 0.006 17/696 0.003 LHPP rs35878051 3/22 0/60 0.02 6/174 0.07 20/708 0.03

ADAM12 rs1466361 17/22 34/58 0.05 103/176 0.11 DOCK1 rs3740013 14/22 20/60 0.003 75/176 0.07 PTPRE rs4462251 12/22 18/58 0.05 86/178 0.65 MK167 rs1063536 6/22 2/60 0.004 21/174 0.09 CTBP2 rs3781412 15/22 25/60 0.007 106/178 0.50 FLJ32938 rs1436803 7/22 6/60 0.009 25/176 0.06

Abbreviation: SNP, single nucleotide polymorphism.

Table 2 Joint distributions of HTR1A and LHPP genotypes among MDD patients and controls in the Utah and Ashkenazi populations

Population HTR1A Status rs10794134 T þ rs12265012 A þ ss68074662 A þ

Number % Number % Number %

Utah G þ Case 197/342 58 122/347 35 29/347 8 Control 85/182 47 75/183 41 7/183 4 CC Case 33/77 43 27/77 35 3/77 4 Control 28/62 45 27/63 43 3/63 5

Ashkenazi G þ Case 51/132 39 71/132 54 10/132 8 Control 35/64 55 49/65 75 4/65 6 CC Case 29/50 58 34/50 68 1/50 2 Control 13/21 62 17/22 77 3/22 14

Abbreviation: MDD, major depressive disorder. Bold signifies values that are statistically different from other groups.

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 626 in the Utah population. There was an rs10794134 genotype-by-gender interaction, mainly reflecting different LHPP genotype distributions between male Heart Kidney Pancreas Lung Colon Placenta Muscle Liver Spleen subjects and female subjects of HTR1A À1019CC Brain genotype. In the Ashkenazi population, two markers were associated with MDD: rs10794134 (adjusted OR for the T allele 0.59, 95% CI 0.35–0.99, P = 0.05) and rs12265012 (adjusted OR for the A allele 0.43, 95% CI 0.24–0.75, P = 0.004). For each marker, one LHPP allele was underrepresented among HTR1A À1019G- positive cases, and was approximately equally 2.0 frequent among all other groups (Table 2). There was no association with MDD for either rs12265012 in 1.0 the Utah population, or for ss68074662 in the Ashkenazi population (Table 2).

Characterization of LHPP transcripts LHPP encodes an known as phospholysine phosphohistidine inorganic pyrophosphate phospha- tase (Lhpp), which was originally purified from swine EF151005 24 BG759116 brain. and subsequently purified from several LHPP additional mammalian organs.25–28 A human LHPP LHPP-SV1 LHPP-SV2 cDNA has been cloned, and functional human Lhpp LHPP-SV3 LHPP-SV4 has been purified following heterologous expression BX346339 in Escherichia coli.29 Lhpp has been characterized AK127935 in vitro as efficiently catalyzing the hydrolysis of P–N AW867792 bonds in phosphohistidine and phospholysine, and less efficiently catalyzing the hydrolysis of P–N or P–O bonds in imidodiphosphate and pyrophosphate, Non-CNS CNS respectively.30 Expression of a 1.8 kb LHPP mRNA was previously Figure 3 Tissue expression patterns of LHPP transcripts. (a) Northern blot using a probe from the 30 exon shared by observed in brain, liver and kidney.29 Using a probe 0 four transcripts (EF151005, BG759116, LHPP and LHPP- from the 3 end of LHPP mRNA, we detected full- SV1). (b) Heatmap of reverse transcription PCR results. length LHPP mRNA as an approximately 1.7 kb Tissues were clustered using the hierarchical clustering transcript in multiple tissues. In addition, an approxi- algorithm in JMP 5.1 software (SAS Institute) with default mately 1.1 kb transcript was very abundant in the settings. Scale: threshold cycle 17.5 (blue)–40 (red). Non- brain, observed at low levels in skeletal muscle and CNS tissues, left to right: adrenal gland, uterus, kidney, spleen but absent in other tissues (Figure 3a). lung, liver, testis, heart, skeletal muscle, heart, small Affymetrix U133 Plus and U133Av2 microarrays each intestine, liver, prostate, trachea, thyroid gland, placenta, include probe sets from the 30 end of LHPP mRNA. kidney, salivary gland, peripheral leukocytes, thymus, fetal Among a panel of tissue-specific mRNA libraries liver, lung and spleen. CNS tissues, left to right: amygdala, globus pallidus, whole brain, orbital frontal cortex, spinal hybridized to these microarrays, statistically signifi- À12 cord, hippocampus, hypothalamus, basal ganglia, caudate cant signals (a =10 ) for the LHPP probe sets were nucleus, medulla, pons, prefrontal cortex, thalamus, whole observed using 18/19 central nervous system samples brain and spinal cord. (the exception was a fetal brain sample) and 0/49 other samples (Supplementary Tables S5 and S6). These data suggested existence of a collinear mRNA species specifically or selectively expressed in threshold cycle: very high ( < 20), high (20–25), the brain. moderate (25–30), low (30–35), very low/questionable Potential collinear transcripts were identified by (35–40) or not detected (no signal above background mining the Genecarta transcriptome database after 40 cycles). Because of the linkage and associa- (Compugen, Israel), and experimentally. Nineteen tion of LHPP variants to MDD, there was particular transcripts (including full-length LHPP) were identi- focus on measuring expression of these transcripts in fied. Nine were eliminated from further analysis: the the brain (Figure 3b). evidence for eight was limited to a single expressed Two transcripts are each comprised of a unique 50 sequence tag each and we were unable to develop a exon spliced to the 30 exon of LHPP (Figure 4), neither reliable assay to detect expression of the other. The of which could encode an Lhpp-like protein. The 30 tissue distributions of the remaining 10 transcripts shared exon contains two potential open reading were determined using quantitative reverse transcrip- frames of at least 100 amino acids. Neither predicted tion PCR. These transcripts are shown in Figure 4. protein shares meaningful homology with any known Expression levels were categorized according to protein. EF151005 (1.2 kb) was detected as a high or

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 627 rs10794134 rs3824810 rs35878051 rs2459213 rs2491156 D10S2322 rs12265012 ss68074662 rs4246100 10 kb EF151005 BG759116 LHPP LHPP-SV1 LHPP-SV2 LHPP-SV3 LHPP-SV4 BX346339 AK127935 AW867792

Figure 4 Genetic markers, transcripts and LD structure in the LHPP gene. Red lines indicate genetic marker positions. D10S2322 delineated the centromeric end of the linkage region, based on recombinants observed in each of the four pedigrees having high familial LOD. Blue lines indicate exon positions; thinner or thicker lines indicate partial skipping due to alternative splicing or transcriptional read-through, respectively. LD was visualized using Haploview software20 accessed through the International HapMap Project website (http://www.hapmap.org) on 13 May 2005. LD magnitude is represented by red squares without numbers (D0 = 1, LOD > 2), light blue squares (D0 = 1, LOD p2) and D0-values. LD, linkage disequilibrium.

very high abundance transcript in all central nervous Full-length LHPP was detected as a high or system samples except for fetal brain (moderate), and moderate abundance transcript in all samples except as a moderate or low abundance transcript in other fetal liver (not detected). Of note, the first 137 tissues except for lung (not detected). These results nucleotides of the published sequence29 do not were replicated using a second quantitative reverse match chromosome 10 and also were not found in transcription PCR assay (not shown). Based on its any LHPP expressed sequence tag. Three splice expression pattern and length, EF151005 is a very variants of LHPP each encode truncated proteins strong candidate to explain our gel and microarray containing one or both haloacid dehydrogenase hybridization results in the brain. We sequenced domains that may be the catalytic regions of Lhpp several clones of EF151005 to confirm the sequence protein (Figure 4). LHPP-SV1, which arises from predicted from database mining, performed RNA skipping of LHPP exon 6, was detected as a high ligase-mediated rapid amplification of cDNA ends abundance transcript in all samples. Expression on a pool of human spinal cord RNA to determine its patterns of full-length LHPP and this splice variant 50ends (annotated in Entrez Nucleotide), and con- were generally similar. The upper band observed by ducted exon-bridging reverse transcriptase PCR ex- gel hybridization probably reflects both species. periments to determine that the unique exon of LHPP-SV2, which results from read-through tran- EF151005 was not observed in transcripts that also scription of approximately 111 nucleotides after contain upstream LHPP exons (data not shown). LHPP exon 6, was detected as a moderate abundance BG759116 (1.1 kb, we have not confirmed the com- transcript in most central nervous system samples, plete sequence of this species) was detected as a low as well as in several other tissues. LHPP-SV3, abundance transcript in a few samples. The strongest which results from read-through transcription of signal was observed in spleen, consistent with the approximately 190 nucleotides after LHPP exon 5, hematopoietic sources of the transcriptome database was detected as a moderate abundance transcript support for this transcript. It may explain the lower in a few samples, and as a low abundance transcript band observed in spleen by gel hybridization. in all others.

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 628 LHPP-SV4 contains LHPP exon 3, a cryptic exon a reasonable independent criterion. The present work from within LHPP intron 6 and the first 39 and last was based on a supposition that previously unknown 367 nucleotides of LHPP exon 7. It has no apparent gene–gene interactions could be discovered by con- open reading frame. This species was detected as a ditioning familial linkage analysis on a genotype that moderate abundance transcript in all central nervous is associated with disease in the population from system samples, as well as in several other tissues. which the linkage pedigrees were ascertained. Several LHPP-SV2, LHPP-SV3 and LHPP-SV4 are not linkage regions for complex diseases have been known to include the probe sequences used in reported using similar analytical approaches. The gel or microarray hybridization. In addition, three IDDM4 locus for type I diabetes mellitus was transcripts arise from LHPP intronic or proximal identified by stratifying linkage analysis according sequences. BX346339 (we have not confirmed the to HLA DR3 and DR4 genotypes.32 Several novel complete sequence of this species) was detected only linkage regions for MDD were observed when pedi- as a low abundance transcript. AK127935 and grees were dichotomized based on linkage evidence at AW867792 were detected as moderate abundance D2S2208, the marker that produced the maximum transcripts in most samples. multipoint LOD score in that set of families.33 One of these novel regions peaked at D10S1656, a marker Discussion about 300 kb centromeric of LHPP. A schizophrenia linkage region was revealed when analysis was Following observation of increased frequency of the conditioned on TNFA promoter haplotypes.34 Re- HTR1A À1019G allele among MDD cases in a Utah cently, a novel obesity linkage region was discovered population, we conditioned genome-wide linkage by conditioning analysis on a SNP in TBC1D1, itself analysis for MDD in that population on carriage of recently linked to obesity in the same pedigrees.35 the G allele. That analysis suggested the presence of a However, none of these conditional linkage analyses genetic risk factor for MDD on chromosome 10 have led to a gene discovery. The HTR1A-conditional (Figures 1 and 2). Several SNPs in LHPP were linked linkage of LHPP to MDD is the first for which a linked to MDD (Table 1), particularly in pedigrees in which gene in an interacting region has been identified, and at least four affected female subjects carried the the first for which the gene–gene interaction has been HTR1A À1019G allele. Some of these SNPs were also further supported by showing conditional association associated with MDD in each of Utah and Ashkenazi in population studies. This work demonstrates that populations (Table 2). Both familial linkage and novel epistatic risk genes for common diseases can be population association of LHPP SNPs to MDD discovered using a direct genetic approach. depended on HTR1A genotype. The combined genetic Genetic evidence supports a pathogenetic role for evidence suggests that a LHPP product may interact LHPP in MDD, perhaps in some sort of functional with Htr1a in a pathogenetic pathway of MDD. interaction with HTR1A. Additional attempts to Alternatively, these two gene products may play roles replicate our findings and analysis of LHPP genetic in two different compensatory pathways, requiring association with other psychiatric diseases are neces- disruption of both to result in disease. sary to define the robustness, sensitivity and specifi- The LHPP polymorphisms associated with MDD city of the relationship between LHPP and MDD. Our differ between Utah and Ashkenazi populations, and work has not identified a functional polymorphism, opposite alleles at rs10794134 were associated with and does not even unequivocally identify the pro- MDD between the two populations. This sort of duct(s) of LHPP involved in MDD. Expression of 10 situation is not unusual in psychiatric genetics, in LHPP transcripts has been detected (Figures 3 and 4). fact it has been observed for most of the genes that All must be considered potentially responsible for the have been linked to schizophrenia.31 One parsimo- genetic evidence. None are well characterized. nious explanation for our association study results is Indeed, little is known about Lhpp, the only certain that functional alleles of LHPP arose on different protein product of an LHPP transcript, beyond its haplotypes in the Utah and Ashkenazi populations. ability to catalyze P-N and P-O bond hydrolysis Other factors may lead to observation of association in vitro. We can only speculate as to the physiological between a complex disease and different markers or function of Lhpp. It may be a protein histidine or alleles within the same gene. For example, a parti- lysine phosphoamidase, that is, an enzyme that cular marker or allele may have a variable phenotype modifies the N-linked phosphorylation state of depending on genetic and environmental back- other proteins. Roles for N-linked phosphorylation ground. in prokaryotic signal transduction functions are Common genetic risk factors for diseases such as established,36 but histidine or lysine phosphorylation MDD will lead to only modest increases of disease has been less studied in eukaryotes.37,38 Since Htr1a prevalence, largely due to dependence on interaction is a G-protein-coupled receptor, it is of interest that a with other genes and non-genetic risk factors. The G-protein b subunit is one of the few human protein complexity of these diseases partially explains the for which a regulatory role of N-linked phosphoryla- difficulty in identifying susceptibility genes. One tion has been described.39–43 LHPP does not share approach to gene discovery within a heterogenous meaningful sequence homology with either of the disease is to condition genetic analysis on the basis of identified human phosphoamidases,44,45 although

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 629 these also are not homologous to each other. Among patients treated with citalopram. J Psychopharmacol 2005; 19: human proteins, Lhpp shares close sequence homo- 166–172. logy only with Hdhd2, also of undetermined function 11 Huang Y-Y, Battistuzzi C, Oquendo M, Harkavy-Friedman J, Greenhill L, Zalsman G et al. Human 5-HT1A receptor (Supplementary Figure S1). C(À1019)G polymorphism and psychopathology. Int J Neuropsy- EF151005 is selectively expressed in the brain, and chopharmacol 2004; 7: 441–451. hence is a strong candidate to explain the genetic 12 Lemonde S, Du L, Bakish D, Hrdina P, Albert P. Association of the relationship of LHPP to MDD. However, limited C(À1019)G 5-HT1A functional promoter polymorphism with linkage disequilibrium between EF151005 exons and antidepressant response. Int J Neuropsychopharmacol 2004; 7: 501–506. the markers linked and associated to MDD (Figure 4) 13 Rothe C, Gutknecht L, Freitag C, Tauber R, Mo¨ssner R, Franke P is less supportive of a role for that transcript. This et al. Association of a functional À1019C > G 5-HT1A receptor transcript contains two open reading frames of at least gene polymorphism with panic disorder with agoraphobia. Int 100 amino acids, and neither is homologous to any J Neuropsychopharmacol 2004; 7: 189–192. known functional protein. It is possible, nonetheless, 14 Serretti A, Artioli P, Lorenzi C, Pirovano A, Tubazio V, Zanardi R. The C(-1019)G polymorphism of the 5-HT1A gene promoter and that one or both of these produce a functional protein, antidepressant response in mood disorders: preliminary findings. that some smaller polypeptide is functional, or that Int J Neuropsychopharmacol 2004; 7: 453–460. the RNA itself is biologically active. Learning which 15 Strobel A, Gutknecht L, Rothe C, Reif A, Mo¨ssner R, Zeng Y et al. LHPP product is involved in MDD, and in what Allelic variation in 5-HT1A receptor expression is associated with anxiety- and depression-related personality traits. J Neural Trans manner, if any, it interacts with the Htr1a pathway, 2003; 110: 1445–1453. are important next steps in fulfilling the potential of 16 Pruitt K, Tatusova T, Maglott D. NCBI reference sequence (RefSeq): this depression gene discovery to lead toward a curated non-redundant sequence database of genomes, tran- improved treatment through development of novel, scripts and proteins. Nucl Acids Res 2005; 33: D501–D504. targeted therapies. 17 Finn R, Mistry J, Schuster-Bo¨ckler B, Griffiths-Jones S, Hollich V, Lassmann T et al. Pfam: clans, web tools and services. Nucl Acids Res 2006; 34: D247–D251. 18 Felsenstein J. PHYLIP—phylogeny inference package (Version Acknowledgments 3.2). Cladistics 1989; 5: 164–166. 19 Page R. TreeView: an application to display phylogenetic trees on We thank Drs James P Sullivan, Brian B Spear, Donald personal computers. Comp Appl Biosci 1996; 12: 357–358. N Halbert and Jerry Lanchbury for the guidance and 20 Barrett J, Fry B, Maller J, Daly M. Haploview: analysis and critical review of our work. This work was funded by visualization of LD and haplotype maps. Bioinformatics 2005; 21: Abbott Laboratories and Myriad Genetics. We are 263–265. 21 Barnes N, Sharp T. A review of central 5-HT receptors and their indebted to the individuals who agreed to participate function. Neuropharmacol 1999; 38: 1083–1152. in this study. 22 Wu S, Comings D. A common C-1018G polymorphism in the human 5-HT1A receptor gene. Psychiatr Genet 1999; 9: 105–106. 23 Harlan J, Chen Y, Gubbins E, Mueller R, Roch J-M, Walter K et al. References Variants in Apaf-1 segregating with major depression promote apoptosome function. Mol Psychiatry 2006; 11: 76–85. 1 Murray C, Lopez A. The Global Burden of Disease: A Comprehen- 24 Seal U, Binkley F. An inorganic pyrophosphatase of swine brain. sive Assessment of Mortality and Disability From Disease, Injuries, J Biol Chem 1957; 228: 193–199. and Risk Factors in 1990 and Projected to 2020. Harvard 25 Felix R, Fleisch H. Properties of inorganic pyrophosphatase of pig University Press: Cambridge MA, 1996. scapula cartilage. Biochem J 1975; 147: 111–118. 2 Fava M, Davidson K. Definition and epidemiology of treatment- 26 Hachimori A, Fujii T, Ohki K, Iizuka E. Purification and properties resistant depression. Psychiatr Clin North Amer 1996; 19: of inorganic pyrophosphatase from porcine brain. J Biochem 1983; 179–200. 93: 257–264. 3 Mann J, Emslie G, Baldessarini R, Beardsley W, Fawcett J, 27 Smirnova I, Baykov A. Reversible inactivation of rat liver Goodwin F et al. ACNP Task Force report on SSRIs and suicidal inorganic pyrophosphatase by and its analogs. Arch behavior in youth. Neuropsychopharmacol 2006; 31: 473–492. Biochem Biophys 1991; 287: 135–140. 4 Mueller T, Leon A, Keller M, Solomon D, Endicott J, Coryell W 28 Yoshida C, Shah H, Weinhouse S. Purification and properties of et al. Recurrence after recovery from major depressive disorder inorganic pyrophosphatase of rat liver and hepatoma 3924A. during 15 years of observational follow-up. Am J Psychiatry 1999; Cancer Res 1982; 42: 3526–3531. 156: 1000–1006. 29 Yokoi F, Hiraishi H, Izuhara K. Molecular cloning of a cDNA for 5 Mann J. Drug therapy: the medical management of depression. the human phospholysine phosphohistidine inorganic pyropho- New Engl J Med 2005; 353: 1819–1834. sphate phosphatase. J Biochem 2003; 133: 607–614. 6 Abkevich V, Camp N, Hensel C, Neff C, Russell D, Hughes D et al. 30 Hiraishi H, Yokoi F, Kumon A. 3-phosphohistidine and Predisposition locus for major depression at chromosome 12q22– 6-phospholysine are substrates of a 56-kDa inorganic pyropho- 12q23.2. Am J Hum Genet 2003; 73: 1271–1281. sphatase from bovine liver. Arch Biochem Biophys 1998; 349: 7 Thomas A, Gutin A, Abkevich V, Bansal A. Multilocus linkage 381–387. analysis by blocked Gibbs sampling. Stat Comp 2000; 10: 259–269. 31 Harrison P, Weinberger D. Schizophrenia genes, gene expression, 8 Lemonde S, Turecki G, Bakish D, Du L, Hrdina P, Bown C et al. and neuropathology: on the matter of their convergence. Mol Impaired repression at a 5-hydroxytryptamine 1A receptor gene Psychiatry 2005; 10: 40–68. polymorphism associated with major depression and suicide. 32 Hashimoto L, Habita C, Beressi J, Delepine M, Basse C, Cambon- J Neurosci 2003; 23: 8788–8799. Thomsen A et al. Genetic mapping of a susceptibility locus for 9 Arias B, Arranz M, Gasto C, Catalan R, Pintor L, Gutierrez B et al. insulin dependent diabetes mellitus on chromosome 11q. Nature Analysis of structural polymorphisms and C-1018G promoter 1994; 371: 161–164. variant of the 5-HT1A receptor gene as putative risk factors in 33 Zubenko GS, Maher B, Hughes 3rd HB, Zubenko WN, Stiffler JS, major depression. Mol Psychiatry 2002; 7: 930–932. Kaplan BB et al. Genome-wide linkage survey for genetic loci that 10 Arias B, Catala´n R, Gasto´ C, Gutie´rrez B, Fan˜ ana´s L. Evidence for a influence the development of depressive disorders in families combined genetic effect of the 5-HT1A receptor and serotonin with recurrent, early-onset, major depression. Am J Med Genet transporter genes in the clinical outcome of major depressive 2003; 123B: 1–18.

Molecular Psychiatry Interacting genetic risk factors in MDD CD Neff et al 630 34 Saviouk V, Chow E, Bassett A, Brzustowicz L. Tumor 40 Hohenegger M, Mitterauer T, Voss T, Nanoff C, Freissmuth M. necrosis factor promoter haplotype associated with schizophrenia Thiophosphorylation of the G protein b subunit in human platelet reveals a linked locus on 1q44. Mol Psychiatry 2005; 10: membranes: evidence against a direct phosphate transfer reaction 375–383. to Ga subunits. Mol Pharmacol 1996; 49: 73–80. 35 Stone S, Abkevich V, Russell D, Riley R, Timms K, Tran T et al. 41 Kowluru A. Regulatory roles for small G proteins in the pancreatic TBC1D1 is a candidate for a severe obesity gene and evidence for a b-cell: lessons from models of impaired insulin secretion. Am gene/gene interaction in obesity predisposition. Hum Mol Genet J Physiol Endocrinol Metab 2003; 285: E669–E684. 2006; 15: 2709–2720. 42 Randazzo P, Northrup J, Kahn R. Activation of a small GTP- 36 Stock J, Stock A, Mottonen J. Signal transduction in bacteria. binding protein by nucleoside diphosphate kinase. Science 1991; Nature 1990; 344: 395–400. 254: 850–853. 37 Klumpp S, Krieglstein J. Phosphorylation and dephosphorylation 43 Wieland T, Nu¨rnberg B, Ulibarri I, Kaldenberg-Stasch S, Schultz G, of histidine residues in proteins. Eur J Biochem 2002; 269: Jakobs K. Guanine nucleotide-specific phosphate transfer by 1067–1071. guanine nucleotide-binding regulatory protein b-subunits: char- 38 Matthews H. Protein kinases and phosphatases that act on acterization of the phosphorylated amino acid. J Biol Chem 1993; histidine, lysine, or arginine residues in eukaryotic proteins: a 268: 18111–18118. possible regulator of the mitogen-activated protein kinase cascade. 44 Chi H, Tiller G, Dasouki M, Romano P, Wang J, O’Keefe R et al. Pharmacol Ther 1995; 67: 323–350. Multiple inositol polyphosphate phosphatase: evolution as a 39 Cuello F, Schulze R, Heemeyer F, Meyer H, Lutz S, Jakobs K et al. distinct group within the histidine phosphatase family and Activation of heterotrimeric G proteins by a high energy chromosomal localization of the human and mouse genes to phosphate transfer via nucleoside diphosphate kinase (NDPK) B chromosomes 10q23 and 19. Genomics 1999; 56: 324–336. and Gb subunits: complex formation of NDPK B with Gbg dimers 45 Ek P, Pettersson G, Ek B, Gong F, Li J-P, Zetterqvist O¨ . Identification and phosphorylation of His-266 in Gb. J Biol Chem 2003; 278: and characterization of a mammalian 14-kDa phosphohistidine 7220–7226. phosphatase. Eur J Biochem 2002; 269: 5016–5023.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)

Molecular Psychiatry