The Pharmacogenomics Journal (2014) 14, 182–191 & 2014 Macmillan Publishers Limited All rights reserved 1470-269X/14 www.nature.com/tpj

ORIGINAL ARTICLE -expression differences in peripheral blood between lithium responders and non-responders in the Lithium Treatment-Moderate dose Use Study (LiTMUS)

RD Beech1, JJ Leffert1, A Lin2, LG Sylvia3, S Umlauf4, S Mane4, H Zhao5, C Bowden6, JR Calabrese7, ES Friedman8, TA Ketter9, DV Iosifescu10, NA Reilly-Harrington3, M Ostacher9, ME Thase11 and A Nierenberg3

This study was designed to identify whose expression in peripheral blood may serve as early markers for treatment response to lithium (Li) in patients with bipolar disorder. Although changes in peripheral blood gene-expression may not relate directly to mood symptoms, differences in treatment response at the biochemical level may underlie some of the heterogeneity in clinical response to Li. Subjects were randomized to treatment with (n ¼ 28) or without (n ¼ 32) Li. Peripheral blood gene-expression was measured before and 1month after treatment initiation, and treatment response was assessed after 6 months. In subjects treated with Li, 62 genes were differentially regulated in treatment responders and non-responders. Of these, BCL2L1 showed the greatest difference between Li responders and non-responders. These changes were specific to Li responders (n ¼ 9), and were not seen in Li non-responders or patients treated without Li, suggesting that they may have specific roles in treatment response to Li.

The Pharmacogenomics Journal (2014) 14, 182–191; doi:10.1038/tpj.2013.16; published online 14 May 2013 Keywords: BCL2L1; bipolar disorder; ; lithium-response; microarray

INTRODUCTION significance. Thus, despite intensive work by multiple groups Lithium (Li) is a well-established treatment in the management spanning several decades, the need for biomarkers that can be of bipolar disorder. It was the first medication to effectively used to monitor and predict response to treatment with Li 18 treat mania1 and, 460 years later, it is still considered first line remains as great as ever. treatment for acute and maintenance phase treatment of bipolar Treatment with Li has been shown to affect the transcription of disorder.2,3 It is also the only medication to be consistently a large number of genes in both neuronal and non-neuronal 19 associated with a reduction in suicidal ideation or attempts in cells. These changes may influence a variety of processes patients with bipolar disorder.4–6 However, clinical response to including neuroplasticity and neurogenesis, which are 20,21 Li is heterogeneous, and the molecular basis for this variability is hypothesized to be responsible for the clinical effects of Li. unknown. Similarities in gene expression between central nervous system Clinical factors, including the pattern of manic and depressive and peripheral lymphoid tissue, combined with the inability episodes, age of illness onset, number of previous hospitalizations to directly sample gene expression in the living brain, have and continuous cycling between episodes, have been studied in led investigators to examine the feasibility of using blood as a attempts to identify subsets of patients who might respond more surrogate tissue to identify biomarkers related to specific 22 or less favorably to treatment with Li. However, a metaanalysis7 psychiatric disorders. A microarray study evaluating the 23 concluded that none of the potential predictors had a very comparability of gene expression in blood and brain found strong impact on response. DNA polymorphisms associated with a that whole blood shares significant gene expression similarities number of candidate genes have also been proposed as possible with multiple regions of the central nervous system. However, predictors of response to treatment with Li.8–12 However, these to our knowledge, there are no studies where effects of Li on results have been difficult to reproduce and other studies have gene-expression on brain and blood have been directly compared. 24 found no association between polymorphisms at these same sites, In a previous study, we examined gene-expression changes in and response to Li.13–16 More recently, genome-wide association 20 patients with bipolar depression who were treated with Li. We studies have been conducted to identify common gene variants found that in Li responders (subjects with an at least 50% decrease 25 that may be associated with Li response.17 Although some loci in Hamilton Depression Rating Scale ), 1 month after starting with suggestive evidence for linkage were found, no single- treatment with Li several antiapoptotic genes including BCL2L1 nucleotide polymorphisms met the threshold for genome-wide (BCL2-like 1; also known as B-cell lymphoma-like X) were

1Department of Psychiatry, Yale University School of Medicine, New Haven, CT, USA; 2Keck Foundation Biotechnology Biostatistics Resource, Yale University School of Medicine, New Haven, CT, USA; 3Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA; 4Center for Genome Analysis, Yale University School of Medicine, New Haven, CT, USA; 5Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT, USA; 6Departments of Psychiatry and Pharmacology, University of Texas Health Science, San Antonio, TX, USA; 7Department of Psychiatry, Case Western Reserve University, Cleveland, OH, USA; 8Department of Psychiatry, University of Pittsburgh Medical Center, Pittsburgh, PA, USA; 9Department of Psychiatry and Behavioral Science, Stanford University School of Medicine, Stanford, CA, USA; 10Departments of Psychiatry and Neuroscience, Mount Sinai Medical Center, New York, NY, USA and 11Department of Psychiatry, University of Pennsylvania, Philadelphia, PA, USA. Correspondence: Dr RD Beech, Department of Psychiatry, Yale University School of Medicine, 34 Park St., 4th Floor, New Haven, CT 06519, USA. E-mail: [email protected] Received 13 November 2012; revised 15 February 2013; accepted 18 March 2013; published online 14 May 2013 Gene-expression differences in peripheral blood RD Beech et al 183 upregulated, and pro-apoptotic genes were downregulated, while Collection of blood samples and storage before RNA isolation. Blood the reverse pattern was seen in Li non-responders. However, as all samples were collected directly into PAXgene blood RNA tubes (QIAGEN, subjects had received Li, it could not be determined whether Valencia, CA, USA) and shipped by an overnight delivery service (FedEx) to these changes were specifically associated with response to Li. Dr Beech’s laboratory at Yale University. To minimize technical variability In the present study, we recruited subjects who had already due to batch-related variation in processing condition, blood samples were stored frozen at À 80 1C until all samples had been collected, and were agreed to participate in the Lithium Treatment-Moderate dose processed as close in time as possible. Use Study (LiTMUS).26,27 The LiTMUS study was a multisite clinical trial comparing the effectiveness of optimized treatment (OPT) Sample preparation and microarray analysis. Total RNA was isolated from versus OPT þ Li for the treatment of bipolar disorder. Patients 10 cc whole blood using the PAXgene Blood RNA Isolation kit per the participating in LiTMUS were randomized to OPT or OPT þ Li and manufacturer’s instructions, and depleted of globin mRNA message using followed clinically for 6 months. The goal of the present study was GLOBINclear hybridization capture technology (Ambion, Austin, TX, USA). to identify changes in the expression of specific genes occurring 1 Globin-reduced total RNA underwent complementary DNA synthesis and month after treatment initiation, which were predictive of clinical overnight in vitro transcription utilizing the Illumina TotalPrep RNA Amplification Kit (Ambion). Biotinylated complementary RNA (1.5 mg) was outcomes at the 6-month time point (study exit). hybridized onto an Illumina Sentrix Beadchip (HumanHT-12 v3), and then scanned on a BeadArray Reader, Illumina, Inc. (San Diego, CA, USA). Microarray hybridization and scanning were carried out at the Yale Center SUBJECTS AND METHODS for Genome Analysis. At the time of publication, all data will be deposited into the NCBI-GEO repository. Subjects All procedures involving human subjects were approved by the Yale Data analysis. Illumina BeadStudio software, Illumina, Inc. (San Diego, CA, Human Investigation Committee and the relevant institutional review USA) was used to generate probe and gene expression profiles of each board at each of the clinical sites participating in LiTMUS. Subjects sample. Quantile normalization was carried out using the package incor- included in the current study had already provided informed consent to porated in the Illumina BeadStudio software package.30 Further statistical participate in LiTMUS.26,27 For this study, a separate and optional genetic analysis was carried out on genes with a detection P-valueo0.01, as consent was presented to subjects allowing additional tubes of blood to determined using the Illumina BeadStudio software (that is, a 99% be drawn for this study. Inclusion and exclusion criteria for subjects with probability that expression was above background), in 90% of the samples. bipolar disorder were identical to those being used in LiTMUS. Fold-changes were calculated for each gene by dividing the level Participating sites included: Case Western Reserve University (Joseph R of expression at 1 month by the baseline level of expression of that gene in Calabrese, MD, site PI), Massachusetts General Hospital (Dan V Iosifescu, the same subject. To identify genes whose expression changed MD, site PI), Stanford University (Terence A Ketter, MD, site PI), University of differentially in Li responders and non-responders after initiation of Pittsburgh (Edward Friedman, MD, site PI), the University of Texas Health treatment, we performed a mixed model analysis of variance of the Science Center at San Antonio (Charles L Bowden, MD, site PI) and the complete microarray data set (after log transformation) with treatment University of Pennsylvania (Michael E Thase, MD, site PI). group (Li þ OPT or OPT) and outcome (responder or non-responder) as a between-subjects factors, and age and gender as covariates, using Partek Genomics Suite Software (Partek, St. Louis, MO, USA). Correction for Treatments multiple testing was done using estimated groupwise false discovery Optimized treatment. The foundation of OPT is to maintain treatment rates.31,32 Network analysis was carried out using the DAVID Functional with at least one Food and Drug Administration-approved mood stabilizer Classification tool.33 other than Li (for participants randomized to the OPT-only group) and to follow the most recent recommendations summarized in the evidence- 28 qRT-PCR analysis. Quantitative real-time reverse transcriptase-PCR based stages of Texas Implementation of Medication Algorithm. Thus, (qRT-PCR) was carried out using the TaqMan Universal PCR Master Mix the OPT-only group could receive all available treatments, except Li, that Protocol (Applied Biosystems, Inc., Foster City, CA, USA) and real-time PCR complied with these guidelines. This control group was named ‘OPT’, in probes listed on the NCBI Probe Database (http://www.ncbi.nlm.nih.gov/ contrast to ‘standard treatment’ or ‘treatment as usual’ because treatment sites/?db=probe). Relative quantitation of gene expression was done was tailored for each participant informed by systematic diagnostic by comparing the efficiency of amplification of each gene of interest with assessments, tracking of symptoms and side effects, and therapeutic blood that of a control gene (either peptidyl prolyl cis-trans isomerase B or heat levels of medications (when appropriate) in the context of specialized shock 75 kDa, mitochondrial (TRAP1)) using the DDCt method, as bipolar disorder clinics. described in User Bulletin number 2 for the ABI Prism 7700 Sequence Detection System (Applied Biosystems). Similar results were obtained using OPT þ Li. Participants randomized to Li þ OPT started at 300 mg of Li at either control genes. bedtime. After 3 days, participants were advised to increase the dosage to two capsules, or 600 mg of Li. Li doses were maintained at 600 mg per day for the first 8 weeks of the study. After that time, further adjustments were RESULTS made at the discretion of the treating clinician, considering patient tolerability, response and other medications included in treatment as part A total of 74 subjects were recruited for this study. Nine subjects of OPT. With the exception of maintaining the Li dose for the first 8 weeks, dropped out before collection of the 1-month blood draw. Five as described above, treating psychiatrists managed the Li þ OPT group, as additional subjects were dropped because of insufficient yield or they would any other patient with OPT (for example, adjusting medications poor quality of the RNA obtained. The remaining 60 subjects as needed in response to clinical exacerbations or side effects). Patients included 28 subjects randomized to Li þ OPT and 32 subjects who were unable to tolerate Li continued to receive OPT, but were not randomized to OPT. Among the subjects randomized to Li þ OPT, included in this analysis. there were 9 responders and 19 non-responders. Among subjects randomized to OPT, there were 4 responders and 28 non- Treatment response. The primary measure of response was the Clinical responders. However, this difference did not reach statistical 29 Global Impression Scale for Bipolar Disorder–Severity (CGI-BP-S) score at significance (w2 (1, N ¼ 60) ¼ 3.39, P ¼ 0.065). Subjects randomized the 6-month follow-up visit. The CGI-BP-S is a widely accepted, reliable and to Li þ OPT did not differ from those randomized to OPT by valid assessment of global illness severity in bipolar disorder, and has been gender (Li þ OPT: 9 male, 19 female; OPT: 10 male, 22 female, w2 shown to reflect clinically meaningful degrees of treatment response. ± Treatment response was defined as a 6-month follow-up CGI-BP-S visit (1, N ¼ 60) ¼ 0.01, P ¼ 0.941) or age (Li þ OPT: 39.1 10.6; OPT: rating of 1 ‘Normal, not ill’ or 2 ‘Minimally ill’. For subjects who completed 39.8±13.8; P ¼ 0.83). the 1-month blood draw, but dropped out of treatment before the In subjects randomized to Li þ OPT, there were 62 genes that 6-month follow-up visit, responder status was based on the last showed differential induction between treatment responders observation carried forward. and non-responders (fold-difference41.3, false discovery rates-

& 2014 Macmillan Publishers Limited The Pharmacogenomics Journal (2014), 182 – 191 Gene-expression differences in peripheral blood RD Beech et al 184

Table 1. Genes showing altered expression in Li þ OPT responders 1 month after treatment initiation

Fold- Induction difference Induction in Li þ OPT (responders FDR Gene in Li þ OPT Non- vs non- corrected symbol responders responders responders) P-value REFSEQ_ID Definition

(A) Genes showing selective upregulation in Li þ OPT responders BCL2L1 1.57 0.93 1.63 0.06 NM_138578.1 Homo sapiens BCL2-like 1 (BCL2L1), nuclear gene encoding mitochondrial protein, transcript variant 1, mRNA. FBXO9 1.55 1.06 1.47 0.06 NM_033480.2 Homo sapiens F-box protein 9 (FBXO9), transcript variant 2, mRNA. MICAL2 1.49 1.03 1.45 0.06 NM_014632.2 Homo sapiens microtubule associated monoxygenase, calponin and LIM domain containing 2 (MICAL2), mRNA. MAP2K3 1.27 0.98 1.43 0.10 NM_002756.3 Homo sapiens mitogen-activated protein kinase kinase 3 (MAP2K3), transcript variant A, mRNA. GTF2IP1 1.18 0.97 1.42 0.06 NR_002206.1 Homo sapiens general transcription factor II, i, pseudogene 1 (GTF2IP1) on 7. LTF 1.30 0.94 1.42 0.06 NM_002343.2 Homo sapiens lactotransferrin (LTF), mRNA. CEACAM8 1.27 0.93 1.40 0.08 NM_001816.2 Homo sapiens carcinoembryonic antigen-related cell adhesion molecule 8 (CEACAM8), mRNA. PIK3R2 1.45 0.97 1.39 0.06 NM_005027.2 Homo sapiens phosphoinositide-3-kinase, regulatory subunit 2 (beta) (PIK3R2), mRNA. TMCC2 1.55 0.95 1.39 0.08 NM_014858.2 Homo sapiens transmembrane and coiled-coil domain family 2 (TMCC2), mRNA. UBE2O 1.34 0.96 1.38 0.09 NM_022066.2 Homo sapiens -conjugating enzyme E2O (UBE2O), mRNA. SLC1A5 1.55 1.01 1.35 0.06 NM_005628.1 Homo sapiens solute carrier family 1 (neutral amino-acid transporter), member 5 (SLC1A5), mRNA. SLC7A5 1.45 1.05 1.33 0.10 NM_003486.5 Homo sapiens solute carrier family 7 (cationic amino-acid transporter, y þ system), member 5 (SLC7A5), mRNA. GLUL 1.28 1.00 1.32 0.06 NM_001033056.1 Homo sapiens glutamate-ammonia ligase (glutamine synthetase) (GLUL), transcript variant 3, mRNA. MAP2K3 1.21 0.98 1.32 0.08 NM_002756.3 Homo sapiens mitogen-activated protein kinase kinase 3 (MAP2K3), transcript variant A, mRNA. DDB1 1.16 0.84 1.31 0.06 NM_001923.3 Homo sapiens damage-specific DNA-binding protein 1, 127 kDa (DDB1), mRNA. HCG4 1.16 1.02 1.31 0.06 NR_002139.1 Homo sapiens HLA complex group 4 (HCG4), non-coding RNA. TCP11L2 1.30 1.01 1.31 0.08 NM_152772.1 Homo sapiens t-complex 11 (mouse)-like 2 (TCP11L2), mRNA. SLC6A9 1.43 1.04 1.30 0.08 NM_001024845.1 Homo sapiens solute carrier family 6 (neurotransmitter transporter, glycine), member 9 (SLC6A9), transcript variant 3, mRNA.

(B) Genes showing selective downregulation in Li þ OPT responders GRIPAP1 0.77 1.00 À 1.30 0.08 NM_207672.1 Homo sapiens GRIP1-associated protein 1 (GRIPAP1), transcript variant 2, mRNA. COX19 0.86 1.03 À 1.30 0.07 NM_001031617.2 Homo sapiens COX19 cytochrome c oxidase assembly homolog (S. cerevisiae) (COX19), mRNA. LEP 0.86 0.97 À 1.31 0.07 NM_000230.1 Homo sapiens leptin (obesity homolog, mouse) (LEP), mRNA. TDRD1 0.82 0.97 À 1.31 0.08 NM_198795.1 Homo sapiens tudor domain-containing 1 (TDRD1), mRNA. SHROOM4 0.79 0.99 À 1.31 0.06 NM_020717.2 Homo sapiens shroom family member 4 (SHROOM4), mRNA. SLC4A5 0.87 1.05 À 1.31 0.09 NM_133478.2 Homo sapiens solute carrier family 4, sodium bicarbonate cotransporter, member 5 (SLC4A5), transcript variant c, mRNA. SIVA1 0.91 1.18 À 1.31 0.07 NM_006427.3 Homo sapiens SIVA1, apoptosis-inducing factor (SIVA1), transcript variant 1, mRNA. WDR74 0.81 0.99 À 1.31 0.08 NM_018093.1 Homo sapiens WD repeat domain 74 (WDR74), mRNA. ALS2CR14 0.86 1.09 À 1.31 0.07 NM_178231.1 Homo sapiens amyotrophic lateral sclerosis 2 (juvenile) chromosome region, candidate 14 (ALS2CR14), mRNA. LOC401152 0.85 1.11 À 1.31 0.08 NM_001001701.1 Homo sapiens HCV F-transactivated protein 1 (LOC401152), mRNA.

The Pharmacogenomics Journal (2014), 182 – 191 & 2014 Macmillan Publishers Limited Gene-expression differences in peripheral blood RD Beech et al 185 Table 1. (Continued )

Fold- Induction difference Induction in Li þ OPT (responders FDR Gene in Li þ OPT Non- vs non- corrected symbol responders responders responders) P-value REFSEQ_ID Definition

HSD17B7 0.80 0.99 À 1.31 0.06 NM_016371.2 Homo sapiens hydroxysteroid (17-beta) dehydrogenase 7 (HSD17B7), mRNA. LOC100128288 0.78 0.99 À 1.31 0.06 NR_024447.1 Homo sapiens hypothetical protein LOC100128288 (LOC100128288), non-coding RNA. C9orf130 0.87 1.02 À 1.31 0.06 XM_939697.1 PREDICTED: Homo sapiens chromosome 9 open reading frame 130 (C9orf130), mRNA. GPR1 0.94 1.09 À 1.31 0.06 NM_005279.2 Homo sapiens G protein-coupled receptor 1 (GPR1), mRNA. C15orf63 0.77 0.95 À 1.32 0.06 NM_016400.2 Homo sapiens chromosome 15 open reading frame 63 (C15orf63), mRNA. C5orf28 0.82 1.07 À 1.32 0.08 NM_022483.3 Homo sapiens open reading frame 28 (C5orf28), mRNA. GZMM 0.82 1.03 À 1.32 0.08 NM_005317.2 Homo sapiens granzyme M (lymphocyte met-ase 1) (GZMM), mRNA. RBM3 0.85 1.07 À 1.32 0.06 NM_001017430.1 Homo sapiens RNA-binding motif (RNP1, RRM) protein 3 (RBM3), transcript variant 2, mRNA. LOC647389 0.82 0.98 À 1.32 0.06 XM_936461.1 Predicted: Homo sapiens hypothetical protein LOC647389 (LOC647389), mRNA. PHAX 0.78 0.96 À 1.32 0.06 NM_032177.2 Homo sapiens phosphorylated adaptor for RNA export (PHAX), mRNA. CCBE1 0.81 1.07 À 1.32 0.06 NM_133459.1 Homo sapiens collagen and calcium binding EGF domains 1 (CCBE1), mRNA. LOC100132585 0.79 0.99 À 1.33 0.06 XM_001722111.1 Predicted: Homo sapiens hypothetical protein LOC100132585 (LOC100132585), mRNA. HSPC268 0.83 0.99 À 1.33 0.06 NM_197964.1 Homo sapiens hypothetical protein HSPC268 (HSPC268), mRNA. LOC644934 0.76 0.97 À 1.33 0.06 XM_930344.2 Predicted: Homo sapiens similar to 40S ribosomal protein S26, transcript variant 1 (LOC644934), mRNA. MAFF 0.85 1.24 À 1.33 0.09 NM_012323.2 Homo sapiens v-maf musculoaponeurotic fibrosarcoma oncogene homolog F (avian) (MAFF), transcript variant 1, mRNA. LOC100129269 0.80 0.97 À 1.33 0.07 XM_001719843.1 Predicted: Homo sapiens hypothetical protein LOC100129269 (LOC100129269), mRNA. LOC100128460 0.85 1.11 À 1.34 0.09 XR_037866.1 Predicted: Homo sapiens misc_RNA (LOC100128460), miscRNA. FLJ25363 0.84 1.04 À 1.35 0.06 XM_001720720.1 PREDICTED: Homo sapiens similar to hypothetical protein FLJ25976 (FLJ25363), mRNA. LMOD3 0.79 0.97 À 1.36 0.06 NM_198271.2 Homo sapiens leiomodin 3 (fetal; LMOD3), mRNA. LOC389765 0.83 1.01 À 1.37 0.06 XM_001720643.1 Predicted: Homo sapiens similar to KIF27C (LOC389765), mRNA. FLJ36131 0.82 0.99 À 1.38 0.06 XM_001722366.1 Predicted: Homo sapiens hypothetical protein FLJ36131, transcript variant 2 (FLJ36131), mRNA. LOC100134273 0.81 0.91 À 1.39 0.08 XM_001724343.1 Predicted: Homo sapiens similar to mCG7602 (LOC100134273), mRNA. ZNF682 0.79 1.03 À 1.39 0.06 NM_033196.2 Homo sapiens zinc-finger protein 682 (ZNF682), transcript variant 1, mRNA. IL17RD 0.81 1.04 À 1.39 0.06 NM_001080973.1 Homo sapiens interleukin 17 receptor D (IL17RD), transcript variant 1, mRNA. PRO1853 0.77 1.00 À 1.40 0.06 NM_144736.3 Homo sapiens hypothetical protein PRO1853 (PRO1853), transcript variant 1, mRNA. VPS41 0.80 1.00 À 1.40 0.07 NM_014396.3 Homo sapiens vacuolar protein sorting 41 homolog (S. cerevisiae; VPS41), transcript variant 1, mRNA. LOC100130764 0.66 1.06 À 1.40 0.06 XM_001723713.1 Predicted: Homo sapiens p150-like (LOC100130764), mRNA. ZNF483 0.73 0.96 À 1.41 0.06 NM_001007169.1 Homo sapiens zinc-finger protein 483 (ZNF483), transcript variant 2, mRNA. LOC388556 0.83 1.06 À 1.42 0.08 XR_019595.2 Predicted: Homo sapiens misc_RNA (LOC388556), miscRNA. LOC648852 0.76 0.97 À 1.43 0.06 XM_940430.1 Predicted: Homo sapiens hypothetical protein LOC648852 (LOC648852), mRNA.

& 2014 Macmillan Publishers Limited The Pharmacogenomics Journal (2014), 182 – 191 Gene-expression differences in peripheral blood RD Beech et al 186 Table 1. (Continued )

Fold- Induction difference Induction in Li þ OPT (responders FDR Gene in Li þ OPT Non- vs non- corrected symbol responders responders responders) P-value REFSEQ_ID Definition

DPM3 0.82 1.15 À 1.45 0.06 NM_018973.3 Homo sapiens dolichyl-phosphate mannosyltransferase polypeptide 3 (DPM3), transcript variant 1, mRNA. FLJ38717 0.78 0.98 À 1.45 0.06 NM_001004322.1 Homo sapiens FLJ38717 protein (FLJ38717), mRNA. LOC148430 0.79 1.22 À 1.48 0.07 XR_038750.1 Predicted: Homo sapiens misc_RNA (LOC148430), miscRNA. LOC100130229 0.82 1.14 À 1.48 0.06 XM_001717158.1 Predicted: Homo sapiens hypothetical protein LOC100130229 (LOC100130229), mRNA.

Gene expression in peripheral blood was measured for all subjects before and 1 month after treatment initiation with Li þ OPT. Treatment responders and non-responders were classified on the basis of Clinical Global Impression Scale for Bipolar Disorder–Severity (CGI-BP-S) rating at the 6-month follow-up visit. Level of induction is the ratio of expression for a given gene at 1 month/its expression at baseline (i.e., onefold induction ¼ no change). Fold-difference is the ratio of the level of induction in treatment responders/treatment non-responders. P-values were calculated for the between group difference in the mixed model analysis of variance of the complete microarray data set (after log transformation), and were corrected for multiple testing using estimated groupwise false discovery rates (FDR).31,32

corrected P-valueo0.1). Of these, 18 were upregulated in Li þ OPT Depression Rating Scale25 in the prior study), there were several responders relative to non-responders (Table 1a), and 44 were overlapping findings between the two studies. Of 18 genes that downregulated (Table 1b). To determine whether these differ- were upregulated in treatment responders to Li þ OPT in the ences were specific to responders within the Li þ OPT group, we current study, 9 were upregulated by 1.3-fold or more in Li compared induction levels of these genes of responders in the responders versus non-responders in the previous study. Among Li þ OPT group with those in the OPT-only group. We also the 44 genes that were downregulated in Li þ OPT responders in compared induction levels between responders and non-respon- the present study, 11 were also downregulated in Li responders in ders regardless of treatment assignment, and between all subjects the previous study, although not all of these differences were randomized to Li þ OPT versus those randomized to OPT alone. statistically significant in the previous study. Genes showing These comparisons are shown in Table 2. overlapping changes in the two studies are listed in Table 3. As shown in Table 2, Li þ OPT reponders showed a distinctive pattern of changes that was not seen in the OPT responders, who appeared more similar to the Li þ OPT non-responders. Thus, of DISCUSSION the 18 genes that were upregulated in Li þ OPT responders In this study, we examined peripheral blood gene expression compared with non-responders, all 18 were also upregulated in profiles in subjects (n ¼ 60) participating in LiTMUS at baseline and Li þ OPT responders relative to OPT responders, and all 44 of the 1 month after treatment initiation. We found that treatment genes that were downregulated in Li þ OPT responders (relative to responders in the Li þ OPT group showed a distinctive pattern of non-responders) were also downregulated in Li þ OPT responders gene-expression changes 1 month after initiation of treatment, compared with OPT responders. which was not seen in treatment non-responders (in either group) We also performed quantitative real-time RT-PCR of selected or in responders to OPT alone. Pathway analysis using the DAVID genes (Figure 1). In general, the results of the qRT-PCR analysis Functional Classification tool33 identified neutral amino-acid were similar to those obtained by microarray hybridization, with transport and regulation of antiapoptosis as the most affected Li þ OPT responders showing greater induction of these genes pathways. These pathways, and their potential role as mediators of than that in the other groups tested. However, owing to the Li response, are discussed in more detail below. greater variability in the results obtained using qRT-PCR, with the exception of BCL2L1, these differences were not statistically significant. Consistent with the results of the microarray studies, Neutral amino-acid transport and Li response the greatest difference was seen for BCL2L1. In Li þ OPT responders, we observed upregulation of four genes To identify functional relationships among the differentially related to neutral amino-acid transport: SLC1A5 (solute carrier expressed genes in Li þ OPT responders, we carried out pathway family 1, member 5), SLC7A5 (solute carrier family 7, member 5), analysis using DAVID Functional Classification tool.33 The SLC6A9 (solute carrier family 6, member 9) and GLUL (glutamate- functional pathways identified using that software centered on a ammonia ligase). SLC1A5 is expressed in astroglial and glutama- small number of genes with roles in regulation of apoptosis and tergic cells and mediates glutamine uptake as well as glutamine amino-acid transport. These pathways, and their potential role as efflux by obligatory exchange with extracellular amino acids.34 Cell mediators of Li response are diagrammed in Figure 2 and are surface expression of SLC1A5 is regulated by the activity of high- discussed in more detail below. affinity glutamate transporter and GLUL, suggesting that these We also compared the results obtained here with those seen in are part of a tightly coupled system that regulates our previous study of depressed bipolar subjects treated with Li.24 extracellular levels of glutamine.35 SLC7A5 is expressed in both Despite numerous differences in study design, including neurons and astrocytes, and is responsible for the import of large differences in inclusion criteria (any symptomatic bipolar patient branched and aromatic neutral amino acids into growing cells and in the current study versus only bipolar depression in the prior across the blood-brain barrier.36,37 SCL1A5 and SLC7A5 are jointly study), length of time before assessment (6 months in the current involved in the regulation of autophagy through their effects study versus 8 weeks in the prior study), and rating scale used to on mammalian target-of-rapamycin (mTOR), which is dependent assess response (CGI-BP-S in the present study versus Hamilton on cellular uptake of glutamine.38 Induction of autophagy, via

The Pharmacogenomics Journal (2014), 182 – 191 & 2014 Macmillan Publishers Limited Gene-expression differences in peripheral blood RD Beech et al 187

Table 2. Specificity of gene expression changes in Li þ OPT responders

Li þ OPT: responders Li þ OPT responders Responders vs Li þ OPT vs vs non-responders vs OPT-responders non-responders (combined) OPT (combined)

Fold- P-value Fold- P-value Fold- P-value Fold- P-value Symbol difference (uncorrected) difference (uncorrected) difference (uncorrected) difference (uncorrected)

BCL2L1 1.63 0.01 1.98 0.05 1.27 0.11 1.16 0.23 FBXO9 1.47 0.02 1.76 0.14 1.25 0.02 1.19 0.03 MICAL2 1.45 0.01 1.85 0.00 1.15 0.21 1.15 0.12 MAP2K3 1.43 0.07 1.76 0.02 1.13 0.37 1.13 0.28 GTF2IP1 1.42 0.01 1.39 0.11 1.18 0.13 1.09 0.35 LTF 1.42 0.02 1.94 0.03 1.14 0.13 1.07 0.36 CEACAM8 1.40 0.05 1.94 0.06 1.17 0.11 1.15 0.08 PIK3R2 1.39 0.01 1.80 0.05 1.12 0.17 1.13 0.08 TMCC2 1.39 0.04 1.67 0.22 1.23 0.05 1.17 0.07 UBE2O 1.38 0.06 1.97 0.06 1.08 0.48 1.16 0.11 SLC1A5 1.35 0.03 1.80 0.10 1.21 0.13 1.25 0.03 SLC7A5 1.33 0.07 2.14 0.02 1.10 0.39 1.21 0.02 GLUL 1.32 0.02 1.57 0.02 1.15 0.08 1.12 0.09 MAP2K3 1.32 0.04 1.69 0.00 1.07 0.48 1.05 0.55 DDB1 1.31 0.02 1.50 0.06 1.06 0.38 1.00 0.98 HCG4 1.31 0.01 1.61 0.01 1.00 0.99 1.04 0.65 TCP11L2 1.31 0.04 1.63 0.08 1.14 0.06 1.14 0.02 SLC6A9 1.30 0.05 1.94 0.03 1.09 0.35 1.15 0.07 GRIPAP1 À 1.30 0.05 À 1.44 0.05 À 1.09 0.32 À 1.08 0.26 COX19 À 1.30 0.03 À 1.67 0.01 À 1.06 0.53 À 1.09 0.28 LEP À 1.31 0.04 À 1.51 0.10 À 1.11 0.33 À 1.12 0.22 TDRD1 À 1.31 0.05 À 1.65 0.06 À 1.10 0.32 À 1.15 0.06 SHROOM4 À 1.31 0.01 À 1.66 0.00 À 1.07 0.43 À 1.09 0.19 SLC4A5 À 1.31 0.06 À 1.54 0.07 À 1.12 0.27 À 1.13 0.15 SIVA1 À 1.31 0.03 À 1.33 0.19 À 1.10 0.22 À 1.03 0.69 WDR74 À 1.31 0.04 À 1.86 0.01 À 1.06 0.55 À 1.13 0.13 ALS2CR14 À 1.31 0.04 À 1.48 0.10 À 1.09 0.38 À 1.05 0.55 LOC401152 À 1.31 0.05 À 1.94 0.04 À 1.03 0.73 À 1.03 0.73 HSD17B7 À 1.31 0.02 À 1.69 0.02 À 1.10 0.24 À 1.11 0.11 LOC100128288 À 1.31 0.00 À 1.75 0.00 À 1.07 0.40 À 1.11 0.11 C9orf130 À 1.31 0.01 À 1.66 0.01 À 1.09 0.38 À 1.10 0.22 GPR1 À 1.31 0.02 À 1.40 0.15 À 1.13 0.31 À 1.12 0.26 C15orf63 À 1.32 0.02 À 1.71 0.02 À 1.11 0.26 À 1.14 0.08 C5orf28 À 1.32 0.05 À 1.39 0.10 À 1.08 0.40 1.01 0.88 GZMM À 1.32 0.05 À 1.36 0.30 À 1.12 0.22 À 1.12 0.14 RBM3 À 1.32 0.02 À 1.49 0.11 À 1.10 0.28 À 1.05 0.52 LOC647389 À 1.32 0.02 À 1.77 0.00 À 1.04 0.70 À 1.09 0.26 PHAX À 1.32 0.02 À 1.76 0.01 À 1.09 0.33 À 1.14 0.06 CCBE1 À 1.32 0.00 À 1.40 0.09 À 1.11 0.09 À 1.02 0.64 LOC100132585 À 1.33 0.01 À 1.48 0.04 À 1.12 0.16 À 1.10 0.15 HSPC268 À 1.33 0.02 À 1.80 0.00 À 1.06 0.46 À 1.11 0.13 LOC644934 À 1.33 0.02 À 1.75 0.01 À 1.10 0.34 À 1.15 0.10 MAFF À 1.33 0.06 À 1.72 0.01 À 1.02 0.86 1.06 0.62 LOC100129269 À 1.33 0.03 À 2.03 0.01 À 1.06 0.55 À 1.12 0.15 LOC100128460 À 1.34 0.06 À 1.41 0.21 À 1.13 0.29 À 1.07 0.49 FLJ25363 À 1.35 0.01 À 1.77 0.01 À 1.08 0.42 À 1.12 0.15 LMOD3 À 1.36 0.02 À 1.69 0.05 À 1.10 0.33 À 1.12 0.14 LOC389765 À 1.37 0.01 À 1.84 0.02 À 1.07 0.47 À 1.12 0.15 FLJ36131 À 1.38 0.01 À 1.66 0.02 À 1.09 0.36 À 1.12 0.14 LOC100134273 À 1.39 0.04 À 1.82 0.03 À 1.12 0.33 À 1.15 0.14 ZNF682 À 1.39 0.00 À 1.64 0.01 À 1.11 0.21 À 1.08 0.25 IL17RD À 1.39 0.01 À 1.75 0.00 À 1.09 0.37 À 1.08 0.32 PRO1853 À 1.40 0.01 À 1.54 0.07 À 1.13 0.17 À 1.12 0.14 VPS41 À 1.40 0.03 À 1.53 0.08 À 1.16 0.23 À 1.19 0.08 LOC100130764 À 1.40 0.00 À 1.67 0.02 À 1.17 0.05 À 1.05 0.47 ZNF483 À 1.41 0.01 À 1.93 0.01 À 1.07 0.48 À 1.09 0.30 LOC388556 À 1.42 0.05 À 1.82 0.05 À 1.12 0.32 À 1.14 0.17 LOC648852 À 1.43 0.01 À 1.89 0.01 À 1.12 0.35 À 1.20 0.05 DPM3 À 1.45 0.00 À 1.81 0.01 À 1.11 0.32 À 1.08 0.36 FLJ38717 À 1.45 0.02 À 1.88 0.00 À 1.17 0.23 À 1.24 0.05 LOC148430 À 1.48 0.03 À 1.67 0.08 À 1.11 0.41 À 1.04 0.72 LOC100130229 À 1.48 0.00 À 2.10 0.00 À 1.12 0.29 À 1.09 0.33 To determine whether differences between treatment responders and non-responders were specific to the Li þ OPT group, we compared induction levels of the genes listed in Table 1 between responders in the Li þ OPT group and those in the OPT-only group. We also compared induction levels between treatment responders and non-responders regardless of treatment assignment, and between all subjects randomized to Li þ OPT vs those randomized to OPT alone. Fold-differences in the level of induction and P-values (uncorrected) are given for each of the comparisons listed. These results show that differences observed in the Li þ OPT responders were unique to that group, and did not occur in OPT-responders or in all subjects randomized to Li þ OPT considered as a group.

inhibition of mTOR has also been implicated in the mechanism of synapses containing N-methyl-D-aspartate receptors, which action of Li.39 SLC6A9 is responsible for clearance of synaptic require glycine as a co-agonist.40 Polymorphisms in the SLC1A5 glycine from both inhibitory glycinergic synapses and excitatory and SLC6A9 genes have been associated with the risk for

& 2014 Macmillan Publishers Limited The Pharmacogenomics Journal (2014), 182 – 191 Gene-expression differences in peripheral blood RD Beech et al 188 schizophrenia,41 and inhibitors of SLC6A9 are currently being (phosphoinositide-3-kinase), and mitogen-activated protein kinase tested as possible therapeutic agents for the treatment of kinase 3 (MAP2K3). BCL2L1 is a member of the BCL-2 (B-cell schizophrenia (ClinicalTrials.gov Identifier: NCT01251055). CLL/lymphoma 2) . BCL2 gene family members Previous studies have shown that SLC6A9 is inhibited by Li,42 are key regulators of apoptotic cell death and include both suggesting that these mechanisms may be relevant to the pro- and antiapoptotic genes.43,44 Previous studies have shown treatment of bipolar disorder as well. that chronic treatment of animals45,46 with Li increases expression of the antiapoptotic gene BCL2 and decreases the expression of the pro-apoptotic gene BAX. Treatment with Li has also been Regulation of antiapoptosis and Li response In Li þ OPT responders, we observed upregulation of three well-known negative regulators of apoptosis: BCL2L1, PI3K Table 3. Overlapping findings between the current study and our previous study24

Current study: Bipolar depression Li þ OPT (responders (responders vs vs non-responders) non-responders)

Fold- Fold- REFSEQ_ID Symbol difference P-value difference P-value

NM_138578.1 BCL2L1 1.63 0.01 1.35 0.37 NM_033480.2 FBXO9 1.47 0.02 1.42 0.07 NM_014632.2 MICAL2 1.45 0.01 1.56 0.03 NM_002756.3 MAP2K3 1.43 0.07 1.66 0.06 NR_002206.1 GTF2IP1 1.42 0.01 1.47 0.01 NM_002343.2 LTF 1.42 0.02 1.46 0.05 NM_001033056.1 GLUL 1.32 0.02 1.36 0.22 NM_002756.3 MAP2K3 1.32 0.04 1.66 0.06 NR_002139.1 HCG4 1.31 0.01 1.43 0.13 NM_001031617.2 COX19 À 1.30 0.03 À 1.41 0.01 NM_198795.1 TDRD1 À 1.31 0.05 À 1.37 0.08 NM_178231.1 ALS2CR14 À 1.31 0.04 À 1.56 0.05 XM_939697.1 C9orf130 À 1.31 0.01 À 1.66 0.03 Figure 1. Confirmation of selected gene expression changes in NM_005317.2 GZMM À 1.32 0.05 À 1.59 0.03 Li þ OPT responders using quantitative real-time reverse transcrip- XM_936461.1 LOC647389 À 1.32 0.02 À 1.43 0.08 tion (qRT-PCR). qRT-PCR was performed using RNA isolated at XM_930344.2 LOC644934 À 1.33 0.02 À 1.31 0.37 baseline and 1 month after treatment initiation for each of the NM_198271.2 LMOD3 À 1.36 0.02 À 1.41 0.05 genes shown. Fold-induction was calculated at the ratio of gene XM_940430.1 LOC648852 À 1.43 0.01 À 1.63 0.04 NM_018973.3 DPM3 À 1.45 0.00 À 1.73 0.00 expression at 1 month divided by the baseline for each subject NM_001004322.1 FLJ38717 À 1.45 0.02 À 1.49 0.02 individually. **Po0.01, *Po0.1.

Figure 2. Relationship of observed changes in gene expression to hypothesized mechanisms of action of lithium (Li). Genes depicted in green were selectively upregulated in Li þ optimized treatment (OPT) responders, whereas SIVA1, depicted in red, was selectively downregulated. Genes shown in blue were not differentially regulated in this study, but are included to show the functional relationships among the differentially expressed genes. GLUL (glutamate-ammonia ligase), SLC1A5 (solute carrier family 1, member 5) and SLC7A5 (solute carrier family 7, member 5) mediate the conversion of glutamate to glutamine, and the flux of glutamine into and out of both neuronal and non-neuronal cells. These three genes are also jointly involved in the regulation of autophagy through their effects on mammalian target-of-rapamycin (mTOR), which is dependent on cellular uptake of glutamine. Induction of autophagy (via inhibition of mTOR) has been implicated in the mechanism of action of Li. Mitogen-activated protein kinase kinase 3 (MAP2K3), phosphoinositide-3-kinase (PI3K) and BCL2-like 1 (BCL2L1) are negative regulators of apoptosis, whereas SIVA1 is a positive regulator of apoptosis. The net effect of the changes observed in Li þ OPT responders (increased expression of MAP2K3, PI3K and BCLL1, and decreased expression of SIVA1) would therefore be an increase in the balance of antiapoptotic to pro-apoptotic gene expression. Negative regulation of apoptosis (via inhibition of glycogen synthase kinase-3b (GSK3b)) has also been implicated in the mechanism of action of Li. Additional information on these pathways is included in the Discussion.

The Pharmacogenomics Journal (2014), 182 – 191 & 2014 Macmillan Publishers Limited Gene-expression differences in peripheral blood RD Beech et al 189 shown to block the reduction of the BCL2/BAX ratio in animals majority of differences observed in the microarray study could not treated with methamphetamine.47 These findings have led to the be confirmed using a different method (qRT-PCR). This may reflect suggestion that changes in the expression of BCL2 and related the fact that qRT-PCR is inherently an exponential process, with genes may mediate the therapeutic effects of Li.20,21,45 each cycle of amplification corresponding to a twofold difference PI3K is a well-known activator of the Akt cell survival in concentration, whereas the differences observed between pathway,48,49 which has been implicated in the actions of Li in groups in the present study were all less than twofold. multiple studies.50,51 Activation of the PI3K/Akt pathway leads to An additional limitation is that the changes described here were phosphorylation of I k-B kinase and activation of the nuclear assayed at the level of gene expression rather than protein or factor of kappa light polypeptide gene enhancer in B functional assays. Thus, they may not correspond to the expected cells (nuclear factor-kB) pathway, which in turn upregulates functional changes at either the cellular or systems level. In BCL2L1.52 Activation of PI3K has synergistic effects with BCL2L1 addition, gene induction levels were measured by comparing in preventing apoptosis.53 MAP2K3-mediated phosphorylation expression levels at only two time points (baseline and 1 month of p38-mitogen-activated protein kinase also leads to after treatment initiation). Thus, we may have missed significant phosphorylation of I k-B kinase and activation of the nuclear changes that occurred either before or after this time point. It is factor-kB pathway,54 and has a synergistic effect with Akt on this also worth noting that subjects in both groups could receive same pathway.55 treatment with any approved medication for the treatment We also saw downregulation of SIVA1 (apoptosis-inducing of bipolar disorder, the sole difference being the inclusion (for factor), a pro-apoptotic protein that binds to and inactivates patients randomized to Li þ OPT) or exclusion (for patients BCL2L1, and thus increases the probability of apoptosis.56,57 SIVA1 randomized to OPT) of Li. Thus, it is unclear whether the observed also negatively regulates the nuclear factor-kB pathway through changes represent direct effects of Li, or the effects of Li in its effects on the adapter protein TRAF2.58 The net effect of the combination with one or more other agents. However, as these changes observed in Li þ OPT responders (increased expression of changes were not observed in patients who responded to MAP2K3, PI3K and BCL2L1, and decreased expression of SIVA1) treatment without Li, it seems likely that treatment with Li is would therefore be an increase in the ratio of antiapoptotic to pro- necessary for the observed effects. Lastly, given the likelihood that apoptotic gene expression. This is consistent with the results of at least some degree of the clinical benefit in the Li-treated group our previous study,24 as well as the proposed antiapoptotic and was attributable to nonspecific factors, such as spontaneous neurotrophic effects of Li.20,21,45 remission, our sample was not large enough to permit meaningful subgroup analyses beyond the partitioning into responder/non- Intersections between neutral amino-acid transport and regulation responder groups. of apoptosis SLC1A5 and SLC7A5 have been shown to jointly regulate the mTOR pathway through their effects on amino-acid flux.38,59 CONCLUSIONS The mTOR pathway has been implicated in the antidepressant In this study, we identified a number of genes that were properties of Li39 as well as in other antidepressants.60 The mTOR specifically induced in treatment responders among subjects pathway is involved in regulation of autophagy,38 as well as randomized to treatment with Li þ OPT. These changes were neurite outgrowth, through its effects on glycogen synthase specific to response in the Li þ OPT group, and were not seen in kinase-3b phosphorylation61 and the regulation of mitochondrial responders to OPT alone or in treatment non-responders in either apoptosis.62 We previously found that mitochondrial gene treatment group. Pathway analysis revealed that these changes expression differed between untreated subjects with bipolar were related to a number of mechanisms previously implicated in depression and healthy controls,63 suggesting that this may be the therapeutic actions of Li, including regulation of autophagy, a primary feature of the illness. Regulation of autophagy,64 neurite neurite outgrowth and mitochondrially mediated apoptosis. The outgrowth65 and mitochondrially mediated apoptosis51 have been fact that these changes were seen only in treatment responders suggested as possible mediators of the therapeutic actions of Li. and not all subjects randomized to Li þ OPT, suggests that individual differences in the response to treatment with Li occur at the level of gene induction, and are clinically relevant. Limitations If validated in larger studies, such changes could be useful Limitations of the current study include the fact that differences in clinically as surrogate outcome markers, allowing treatment gene expression were assessed in peripheral blood rather than in decisions (including whether to continue or discontinue treatment the brain. Offsetting this limitation is the fact that peripheral with Li) to be made earlier, and thus facilitate recovery in patients markers, including peripheral blood gene expression, can be with bipolar disorder. assessed repeatedly over time, and thus compared directly with changes in clinical status, as in the present study. An additional limitation is the relatively small sample size, and in particular the CONFLICT OF INTEREST small number of responders in each treatment group. Power analyses based on the current sample indicate that to confirm a The authors declare no conflict of interest. 1.6-fold difference in induction levels (as seen for BCL2L1) with 80% power would require a sample size of 10 subjects per group, whereas confirming a 1.3-fold difference would require 39 ACKNOWLEDGEMENTS subjects per group. Thus, despite the encouraging overlap in This work was supported by grants from the California Bipolar Foundation (RDB) and findings with our previous study, the results presented here are Donaghue Foundation Grant number DF08-009 (RDB). LiTMUS was sponsored by best considered as a hypothesis-generating exercise that will National Institute of Mental Health (Contract number NO1MH80001) and is registered require confirmation in a larger independent sample. with the National Library of Medicine’s ClinicalTrials.gov website (Record number: The decision to consider gene expression differences with an NCT00667745). false discovery rates-corrected P-valueo0.1 as significant represents a compromise between the desire to include as many NOTE ADDED IN PROOF real biological changes as possible (reducing type II errors), Raw and normalized data can be accessed at the National Center for Biotechnology while accepting that this entails a higher proportion of potential Information Gene Expression Omnibus (NCBI-GEO) repository (http://www.ncbi.nlm. false-positives (type I errors). A further limitation is that the nih.gov/geo/). The GSE accession number is GSE 45484.

& 2014 Macmillan Publishers Limited The Pharmacogenomics Journal (2014), 182 – 191 Gene-expression differences in peripheral blood RD Beech et al 190 REFERENCES 27 Nierenberg AA, Sylvia LG, Leon AC, Reilly-Harrington NA, Ketter TA, 1 Cade JF. Lithium salts in the treatment of psychotic excitement. Med J Aust 1949; Calabrese JR et al. Lithium treatment -- moderate dose use study (LiTMUS) for 2: 349–352. bipolar disorder: rationale and design. Clin Trials 2009; 6: 637–648. 2 APA. Practice guideline for the treatment of patients with bipolar disorder 28 Suppes T, Dennehy EB, Hirschfeld RM, Altshuler LL, Bowden CL, Calabrese JR et al. (revision). Am J Psychiatry 2002; 159(4 Suppl): 1–50. The Texas implementation of medication algorithms: update to the algorithms for 3 Grof P, Muller-Oerlinghausen B. A critical appraisal of lithium’s efficacy and treatment of bipolar I disorder. J Clin Psychiatry 2005; 66: 870–886. effectiveness: the last 60 years. Bipolar Disord 2009; 11: 10–19. 29 Spearing MK, Post RM, Leverich GS, Brandt D, Nolen W. Modification of the 4 Tondo L, Hennen J, Baldessarini RJ. Lower suicide risk with long-term lithium Clinical Global Impressions (CGI) Scale for use in bipolar illness (BP): the CGI-BP. treatment in major affective illness: a meta-analysis. Acta Psychiatr Scand 2001; Psychiatry Res 1997; 73: 159–171. 104: 163–172. 30 Illumina. Normalization and Differential Analysis. BeadStudio Gene Expression 5 Ernst CL, Goldberg JF. Antisuicide properties of psychotropic drugs: a critical Module v3.2 User Guide vol. 2009. Illumina Technical SupportSan Diego, CA, 2009. review. Harv Rev Psychiatry 2004; 12: 14–41. 31 Reiner A, Yekutieli D, Benjamini Y. Identifying differentially expressed genes using 6 Goodwin FK, Fireman B, Simon GE, Hunkeler EM, Lee J, Revicki D. Suicide risk in false discovery rate controlling procedures. Bioinformatics 2003; 19: 368–375. bipolar disorder during treatment with lithium and divalproex. JAMA 2003; 290: 32 Storey JD, Xiao W, Leek JT, Tompkins RG, Davis RW. Significance analysis 1467–1473. of time course microarray experiments. Proc Natl Acad Sci USA 2005; 102: 7 Kleindienst N, Engel R, Greil W. Which clinical factors predict response to 12837–12842. prophylactic lithium? A systematic review for bipolar disorders. Bipolar Disord 33 Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis 2005; 7: 404–417. of large gene lists using DAVID bioinformatics resources. Nat Protoc 2009; 4: 8 Serretti A, Lilli R, Lorenzi C, Gasperini M, Smeraldi E. Tryptophan hydroxylase gene 44–57. and response to lithium prophylaxis in mood disorders. J Psychiatr Res 1999; 33: 34 Broer A, Brookes N, Ganapathy V, Dimmer KS, Wagner CA, Lang F et al. The 371–377. astroglial ASCT2 transporter as a mediator of glutamine efflux. 9 Turecki G, Grof P, Cavazzoni P, Duffy A, Grof E, Ahrens B et al. Evidence for a role J Neurochem 1999; 73: 2184–2194. of phospholipase C-gamma1 in the pathogenesis of bipolar disorder. Mol 35 Gegelashvili M, Rodriguez-Kern A, Pirozhkova I, Zhang J, Sung L, Gegelashvili G. Psychiatry 1998; 3: 534–538. High-affinity glutamate transporter GLAST/EAAT1 regulates cell surface 10 Steen VM, Lovlie R, Osher Y, Belmaker RH, Berle JO, Gulbrandsen AK. The expression of glutamine/neutral amino acid transporter ASCT2 in human fetal polymorphic inositol polyphosphate 1-phosphatase gene as a candidate for astrocytes. Neurochem Int 2006; 48: 611–615. pharmacogenetic prediction of lithium-responsive manic-depressive illness. 36 Abbott NJ, Ronnback L, Hansson E. Astrocyte-endothelial interactions at the Pharmacogenetics 1998; 8: 259–268. blood-brain barrier. Nat Rev Neurosci 2006; 7: 41–53. 11 Rybakowski JK, Suwalska A, Skibinska M, Szczepankiewicz A, Leszczynska- 37 Verrey F, System L. heteromeric exchangers of large, neutral amino acids involved Rodziewicz A, Permoda A et al. Prophylactic lithium response and polymorphism in directional transport. Pflugers Arch 2003; 445: 529–533. of the brain-derived neurotrophic factor gene. Pharmacopsychiatry 2005; 38: 38 Nicklin P, Bergman P, Zhang B, Triantafellow E, Wang H, Nyfeler B et al. Bidirec- 166–170. tional transport of amino acids regulates mTOR and autophagy. Cell 2009; 136: 12 Serretti A, Lilli R, Mandelli L, Lorenzi C, Smeraldi E. Serotonin transporter gene 521–534. associated with lithium prophylaxis in mood disorders. Pharmacogenomics J 2001; 39 Cleary C, Linde JA, Hiscock KM, Hadas I, Belmaker RH, Agam G et al. 1: 71–77. Antidepressive-like effects of rapamycin in animal models: implications for mTOR 13 Cavazzoni P, Alda M, Turecki G, Rouleau G, Grof E, Martin R et al. Lithium- inhibition as a new target for treatment of affective disorders. Brain Res Bull 2008; responsive affective disorders: no association with the tyrosine hydroxylase gene. 76: 469–473. Psychiatry Res 1996; 64: 91–96. 40 Betz H, Gomeza J, Armsen W, Scholze P, Eulenburg V. Glycine transporters: 14 Lovlie R, Berle JO, Stordal E, Steen VM. The phospholipase C-gamma1 gene essential regulators of synaptic transmission. Biochem Soc Trans 2006; 34(Pt 1): (PLCG1) and lithium-responsive bipolar disorder: re-examination of an intronic 55–58. dinucleotide repeat polymorphism. Psychiatr Genet 2001; 11: 41–43. 41 Deng X, Sagata N, Takeuchi N, Tanaka M, Ninomiya H, Iwata N et al. Association 15 Michelon L, Meira-Lima I, Cordeiro Q, Miguita K, Breen G, Collier D et al. study of polymorphisms in the neutral amino acid transporter genes SLC1A4, Association study of the INPP1, 5HTT, BDNF, AP-2beta and GSK-3beta GENE SLC1A5 and the glycine transporter genes SLC6A5, SLC6A9 with schizophrenia. variants and restrospectively scored response to lithium prophylaxis in bipolar BMC Psychiatry 2008; 8:58. disorder. Neurosci Lett 2006; 403: 288–293. 42 Perez-Siles G, Morreale A, Leo-Macias A, Pita G, Ortiz AR, Aragon C et al. Molecular 16 Masui T, Hashimoto R, Kusumi I, Suzuki K, Tanaka T, Nakagawa S et al. Lithium basis of the differential interaction with lithium of glycine transporters GLYT1 and response and Val66Met polymorphism of the brain-derived neurotrophic factor GLYT2. J Neurochem 2011; 118: 195–204. gene in Japanese patients with bipolar disorder. Psychiatr Genet 2006; 16: 49–50. 43 Danial NN, Korsmeyer SJ. Cell death: critical control points. Cell 2004; 116: 17 Perlis RH, Smoller JW, Ferreira MA, McQuillin A, Bass N, Lawrence J et al. 205–219. A genomewide association study of response to lithium for prevention of 44 Adams JM, Cory S. Bcl-2-regulated apoptosis: mechanism and therapeutic recurrence in bipolar disorder. Am J Psychiatry 2009; 166: 718–725. potential. Curr Opin Immunol 2007; 19: 488–496. 18 Belmaker RH, Agam G. Bipolar disorder: neurochemistry and drug mechanisms. 45 Chen G, Zeng WZ, Yuan PX, Huang LD, Jiang YM, Zhao ZH et al. The Discov Med 2005; 5: 191–198. mood-stabilizing agents lithium and valproate robustly increase the levels of the 19 Seelan RS, Khalyfa A, Lakshmanan J, Casanova MF, Parthasarathy RN. Deciphering neuroprotective protein bcl-2 in the CNS. J Neurochem 1999; 72: 879–882. the lithium transcriptome: Microarray profiling of lithium-modulated gene 46 Manji HK, Moore GJ, Chen G. Lithium up-regulates the cytoprotective protein expression in human neuronal cells. Neuroscience 2007; 151: 1184–1197. Bcl-2 in the CNS in vivo: a role for neurotrophic and neuroprotective effects in 20 Manji HK, Moore GJ, Chen G. Clinical and preclinical evidence for the neurotrophic manic depressive illness. J Clin Psychiatry 2000; 61(Suppl 9): 82–96. effects of mood stabilizers: implications for the pathophysiology and treatment of 47 Bachmann RF, Wang Y, Yuan P, Zhou R, Li X, Alesci S et al. Common effects of manic-depressive illness. Biol Psychiatry 2000; 48: 740–754. lithium and valproate on mitochondrial functions: protection against metham- 21 Einat H, Manji HK. Cellular plasticity cascades: genes-to-behavior pathways in phetamine-induced mitochondrial damage. Int J Neuropsychopharmacol 2009; 12: animal models of bipolar disorder. Biol Psychiatry 2006; 59: 1160–1171. 1–18. 22 Gladkevich A, Kauffman HF, Korf J. Lymphocytes as a neural probe: potential for 48 Cantley LC. The phosphoinositide 3-kinase pathway. Science 2002; 296: studying psychiatric disorders. Prog Neuropsychopharmacol Biol Psychiatry 2004; 1655–1657. 28: 559–576. 49 Vasudevan KM, Garraway LA. AKT signaling in physiology and disease. Curr Top 23 Sullivan PF, Fan C, Perou CM. Evaluating the comparability of gene expression in Microbiol Immunol 2010; 347: 105–133. blood and brain. Am J Med Genet B Neuropsychiatr Genet 2006; 141: 261–268. 50 Detera-Wadleigh SD. Lithium-related genetics of bipolar disorder. Annf Med 2001; 24 Lowthert L, Leffert J, Lin A, Umlauf S, Maloney K, Muralidharan A et al. Increased 33: 272–285. ratio of anti-apoptotic to pro-apoptotic Bcl2 gene-family members in lithium- 51 Chuang DM. The antiapoptotic actions of mood stabilizers: molecular mechan- responders one month after treatment initiation. Biol Mood Anxiety Disord 2012; 2:15. isms and therapeutic potentials. Ann NY Acad Sci 2005; 1053: 195–204. 25 Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry 1960; 23: 52 Mora AL, Corn RA, Stanic AK, Goenka S, Aronica M, Stanley S et al. Antiapoptotic 56–62. function of NF-kappaB in T lymphocytes is influenced by their differentiation 26 Nierenberg AA, Friedman ES, Bowden CL, Sylvia LG, Thase ME, Ketter T et al. status: roles of Fas, c-FLIP, and Bcl-xL. Cell Death Differ 2003; 10: 1032–1044. Lithium Treatment Moderate-Dose Use Study (LiTMUS) for bipolar disorder: 53 Qian J, Zou Y, Rahman JS, Lu B, Massion PP. Synergy between phosphatidylino- a randomized comparative effectiveness trial of optimized personalized treat- sitol 3-kinase/Akt pathway and Bcl-xL in the control of apoptosis in adenocarci- ment with and without lithium. Am J Psychiatry 2013; 170: 102–110. noma cells of the lung. Mol Cancer Ther 2009; 8: 101–109.

The Pharmacogenomics Journal (2014), 182 – 191 & 2014 Macmillan Publishers Limited Gene-expression differences in peripheral blood RD Beech et al 191 54 Wada T, Penninger JM. Mitogen-activated protein kinases in apoptosis regulation. 60 Duman RS, Li N, Liu RJ, Duric V, Aghajanian G. Signaling pathways underlying the Oncogene 2004; 23: 2838–2849. rapid antidepressant actions of ketamine. Neuropharmacology 2012; 62: 35–41. 55 Fan S, Meng Q, Laterra JJ, Rosen EM. Role of Src signal transduction 61 Jin Y, Sui HJ, Dong Y, Ding Q, Qu WH, Yu SX et al. Atorvastatin pathways in scatter factor-mediated cellular protection. J Biol Chem 2009; 284: enhances neurite outgrowth in cortical neurons in vitro via up-regulating the 7561–7577. Akt/mTOR and Akt/GSK-3beta signaling pathways. Acta Pharmacol Sin 2012; 33: 56 Chu F, Borthakur A, Sun X, Barkinge J, Gudi R, Hawkins S et al. The Siva-1 putative 861–872. amphipathic helical region (SAH) is sufficient to bind to BCL-XL and sensitize cells 62 Fulda S. Modulation of mitochondrial apoptosis by PI3K inhibitors. Mitochondrion to UV radiation induced apoptosis. Apoptosis 2004; 9: 83–95. 2012; 13: 195–198. 57 Xue L, Chu F, Cheng Y, Sun X, Borthakur A, Ramarao M et al. Siva-1 binds to and 63 Beech RD, Lowthert L, Leffert JJ, Mason PN, Taylor MM, Umlauf S et al. Increased inhibits BCL-X(L)-mediated protection against UV radiation-induced apoptosis. peripheral blood expression of electron transport chain genes in bipolar Proc Natl Acad Sci USA 2002; 99: 6925–6930. depression. Bipolar Disord 2010; 12: 813–824. 58 Gudi R, Barkinge J, Hawkins S, Prabhakar B, Kanteti P. Siva-1 promotes K-48 64 Sarkar S, Floto RA, Berger Z, Imarisio S, Cordenier A, Pasco M et al. Lithium polyubiquitination of TRAF2 and inhibits TCR-mediated activation of NF-kappaB. induces autophagy by inhibiting inositol monophosphatase. JCellBiol2005; 170: J Environ Pathol, Toxicol Oncol 2009; 28: 25–38. 1101–1111. 59 Fuchs BC, Finger RE, Onan MC, Bode BP. ASCT2 silencing regulates mammalian 65 Wang Z, Wang J, Li J, Wang X, Yao Y, Zhang X et al. MEK/ERKs signaling is essential target-of-rapamycin growth and survival signaling in human hepatoma cells. for lithium-induced neurite outgrowth in N2a cells. Int J Dev Neurosci 2011; 29: Am J Physiol Cell Physiol 2007; 293: C55–C63. 415–422.

& 2014 Macmillan Publishers Limited The Pharmacogenomics Journal (2014), 182 – 191