Identification of Novel Signal Transduction, Immune Function, and Oxidative Stress Genes and Pathways by Topiramate for Treatment of Methamphetamine Dependence Based on Secondary Outcomes

Background Topiramate (TPM) is suggested to be a promising medication for treatment of methamphetamine (METH) dependence, but the molecular basis remains to be elucidated. Methods Among 140 METH-dependent participants randomly assigned to receive either TPM (N = 69) or placebo (N = 71) in a previously conducted randomized controlled trial, 50 TPM- and 49 placebo-treated participants had a total 212 RNA samples available at baseline, week 8, and week 12 time points. Following our primary analysis of gene expression data, we reanalyzed the microarray expression data based on a latent class analysis of binary secondary outcomes during weeks 1–12 that provided a classification of 21 responders and 31 non-responders with consistent responses at both time points. Results Based on secondary outcomes, 1,381, 576, 905, and 711 differentially expressed genes at nominal P values < 0.05 were identified in responders versus non-responders for week 8 TPM, week 8 placebo, week 12 TPM, and week 12 placebo groups, respectively. Among 1,381 genes identified in week 8 TPM responders, 359 genes were identified in both week 8 and week 12 TPM groups, of which 300 genes were exclusively detected in TPM responders. Of them, 32 genes had nominal P values < 5 × 10−3 at either week 8 or week 12 and false discovery rates < 0.15 at both time points with consistent directions of gene expression changes, which include GABARAPL1, GPR155, and IL15RA in GABA receptor signaling that represent direct targets for TPM. Analyses of these 300 genes revealed 7 enriched pathways belonging to neuronal function/synaptic plasticity, signal transduction, inflammation/immune function, and oxidative stress response categories. No pathways were enriched for 72 genes exclusively detected in both week 8 and week 12 placebo groups. Conclusion This secondary analysis study of gene expression data from a TPM clinical trial not only yielded consistent results with those of primary analysis but also identified additional new genes and pathways on TPM response to METH addiction.

Background: Topiramate (TPM) is suggested to be a promising medication for treatment of methamphetamine (METH) dependence, but the molecular basis remains to be elucidated.
Methods: Among 140 METH-dependent participants randomly assigned to receive either TPM (N = 69) or placebo (N = 71) in a previously conducted randomized controlled trial, 50 TPM-and 49 placebo-treated participants had a total 212 RNA samples available at baseline, week 8, and week 12 time points. Following our primary analysis of gene expression data, we reanalyzed the microarray expression data based on a latent class analysis of binary secondary outcomes during weeks 1-12 that provided a classification of 21 responders and 31 non-responders with consistent responses at both time points.
results: Based on secondary outcomes, 1,381, 576, 905, and 711 differentially expressed genes at nominal P values < 0.05 were identified in responders versus non-responders for week 8 TPM, week 8 placebo, week 12 TPM, and week 12 placebo groups, respectively. Among 1,381 genes identified in week 8 TPM responders, 359 genes were identified in both week 8 and week 12 TPM groups, of which 300 genes were exclusively detected in TPM responders. Of them, 32 genes had nominal P values < 5 × 10 −3 at either week 8 or week 12 and false discovery rates < 0.15 at both time points with consistent directions of gene expression changes, which include GABARAPL1, GPR155, and IL15RA in GABA receptor signaling that represent direct targets for TPM. Analyses of these 300 genes revealed 7 enriched pathways belonging to neuronal function/synaptic plasticity, signal transduction, inflammation/immune function, and oxidative stress response categories. No pathways were enriched for 72 genes exclusively detected in both week 8 and week 12 placebo groups. transcriptomic profiling of peripheral blood (most accessible tissue) has been shown to identify new diagnostic markers for psychiatric disorders (e.g., major depression) (26)(27)(28), because of similarities between receptor expression and mechanisms of transduction processes of cells in CNS and in peripheral blood (29). In a previous gene expression study based on the primary outcome of the clinical trial (i.e., weekly abstinence from METH during weeks 1-12) (30), we identified a set of differentially expressed (DE) genes associated with the treatment of TPM on METH dependence, as well as enriched biochemical pathways. However, as shown in the clinical trial, TPM did not significantly increase abstinence from METH use (25), so the numbers of responders with valid gene expression data in the TPM and placebo groups were relatively small, and also were not well balanced (e.g., only four and two responders were included in the placebo group for weeks 8 and 12, respectively) (30). On the other hand, an earlier latent variable analysis study reported by our group (31) showed that the heterogeneity of treatment responses could be better characterized based on six non-longitudinal binary secondary outcomes of METH use, which helped to identify more robust responder and non-responder groups to TPM or placebo, providing larger and more balanced samples. The main objective of this study was to continue a primary analysis of gene expression data based on only primary outcome (30) by conducting a secondary analysis study using the same gene expression data of the same TPM randomized clinical trial (25) but with more samples classified as either responders or non-responders by applying a latent class analysis (LCA) of binary secondary outcomes during weeks 1-12 (31) to obtain a better understanding of the molecular mechanisms associated with the differences between responders and non-responders specifically for TPM.

MaTerials anD MeThODs study Participants and Blood sample collection
A detailed description of this double-blind, multicenter, placebocontrolled, randomized, parallel-group trial has been presented elsewhere (25). In brief, after a potential participant provided written informed consent, he or she was screened for up to 14 days based on the inclusion and exclusion criteria reported previously (25). If any participant met the eligibility criteria, he or she would start a 14-day baseline assessment. Exclusion criteria included serious medical illness, psychiatric conditions requiring ongoing medication, pregnancy or lactation, nephrolithiasis or renal impairment, and court-mandated drug abuse treatment. For a detailed delineation of screening and baseline assessments, please see our previous report (25). A total 140 Diagnostic and Statistical Manual of Mental Disorders Fourth Edition (DSM-IV)-diagnosed inTrODUcTiOn Methamphetamine (METH), a synthetic derivative of amphetamine, is a highly addictive psychostimulant, with escalating rates of use worldwide (1), particularly in East and Southeast Asia, Australia, various regions of Great Britain, as well as Western and Midwestern areas of United States (1). Due to an additional methyl group in its chemical structure, METH has a higher lipid solubility than the unsubstituted amphetamine, allowing a more rapid transport of the drug across the blood-brain barrier than its parent drug (2). Thus, compared with amphetamine, METH exerts more profound effects on central nervous system (CNS) (3). Approximately 45% of METH is metabolized into amphetamine, and both highly addictive drugs are mostly excreted in renal system (4). The half-life of METH is approximately 8-12 h, and during this time, acute effects of METH take place, which include an ecstatic rush that is nearly immediate in those who either smoke or inject this drug (5). Chronic effects of METH include significant functional impairments in a range of cognitive processes, particularly in attention/psychomotor speed, verbal learning and memory, and fluency-based measures of executive systems functioning (6). Cardiovascular effects of METH are manifested by an elevated heart rate and hypertension, which can result in palpitations, arrhythmias, cardiomyopathy, valvular disease, angina, myocardial infarctions, and cerebral vascular events (7), while psychological effects are manifested by increased anxiety, insomnia, aggression and violent tendencies, paranoia, and visual and auditory hallucinations (8).
To study the efficacy of TPM in treating METH addiction, a multi-center, placebo-controlled randomized trial of 140 METH-dependent subjects was performed (25). This clinical trial generated mixed results such that TPM treatment did not achieve significant METH abstinence (primary outcome), but did decrease weekly median urine METH levels significantly. Genome-wide

classification of responders and nonresponders Based on secondary Outcomes
Subjects were classified by LCA as responders or non-responders according to six non-longitudinal, binary secondary outcomes of METH use for weeks 1-12 (31): (i) secondary outcome C for weeks 1-12: ≥21 consecutive days of METH abstinence during weeks 1-12 based on urine drug screens only; (ii) secondary outcome D for weeks 1-12: ≥21 consecutive days of METH abstinence during weeks 1-12 based on urine drug screens and self-report; (iii) secondary outcome E for weeks 1-12: ≥25% reduction in proportion of METH use days during weeks 1-12 compared with self-reported METH use during 14-day baseline period; (iv) secondary outcome F for weeks 1-12: ≥50% reduction in proportion of METH use days during weeks 1-12 compared with self-reported METH use during 14-day baseline period; (v) secondary outcome G for weeks 1-12: ≥25% reduction in median quantitative METH urine concentration during weeks 1-12 compared with median quantitative METH urine concentration during 14-day baseline period; and (vi) secondary outcome H for weeks 1-12: ≥50% reduction in median quantitative METH urine concentration during weeks 1-12 compared with median quantitative METH urine concentration during 14-day baseline period. A detailed description of these six binary secondary outcomes is shown in Supplementary Text S1 in Supplementary Material. These six binary secondary outcomes measured different aspects of both attainment of METH abstinence [(i) and (ii)] and attainment of METH use reduction [(iii), (iv), (v), and (vi)], which could be more robust than primary outcome defined only based on weekly METH abstinence during weeks 1-12 (30) to more sensitively define a responder as a study participant who had decreased METH use during the clinical trial in response to either TPM or placebo. LCA was performed using Mplus (34) to identify clinically distinct classes based on observed response patterns according to the above six secondary outcomes (31).

statistical and Bioinformatics analysis
The procedures used for outlier array detection, quality control, preprocessing, normalization, and probe set filtering to decrease noise have been described before (30). To adjust for individual variations of gene expression at baseline, each individual's gene expression level at either week 8 or week 12 time point was first baseline-normalized before identification of DE genes and pathways. Then, significantly modulated genes and enriched biological pathways were detected by the following methods.

Pathway Analysis
Ingenuity Pathway Analysis (IPA) 1 and Onto-Tools Pathway-Express 2 software tools were applied. IPA determines the probability that a given gene set is associated with predefined pathways beyond what would be expected by random chance (37). Further, this software tool computes a right-tailed Fisher's exact test P value and a Benjamini-Hochberg false discovery rate (FDR) (38) for each relevant pathway in the Ingenuity Pathways Knowledge Base (IPKB). The list of gene identifiers and their fold changes (FCs) was uploaded to the IPA, such that each gene identifier was mapped to its corresponding gene object in the IPKB. Molecular interaction networks were constructed for both direct and indirect interactions using default parameters. Onto-Tools Pathway-Express (39, 40) is a web-based application that automatically finds significantly impacted Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. For a given pathway in KEGG that includes L input genes, denoted as g1, …, gL, Onto-Tools Pathway-Express first calculates a perturbation factor (PF) for gene i, denoted as PF(gi), for I ∈ (1, …, L), where PF(gi) shows relative importance of gene gi for that pathway. Then, an impact factor of that pathway is calculated, which includes PF(g1), PF(g2), …, PF(gL), and a probabilistic term that takes into account the proportion of input genes of that pathway (39). For a given input gene list, Onto-Tools Pathway-Express calculates a gamma P value and a Benjamini-Hochberg FDR for each relevant KEGG pathway. (6 responders and 11 non-responders), and week 12 placebo group (2 responders and 13 non-responders), and had consistently greater total sample sizes for these 4 respective groups.

identification of De genes in Treatment responders
We applied two different tests, i.e., an ordinary Student's t-test and an empirical Bayes moderated t-test, for detecting DE genes ( , and these results are better than the results obtained from primary analysis (30), which had an increasing order of week 8 TPM (0.009), week 12 placebo (0.033), week 8 placebo (0.027), and week 12 TPM (0.113), because week 12 TPM group of primary outcome had fewer (total: 17) and more imbalanced (6 responders versus 11 non-responders) samples than week 12 TPM group of secondary outcomes (total: 27, 15 responders versus 12 non-responders). Further, 298, 0, 112, and 10 of the above genes with nominal P values < 0.01 also had FDRs < 0.15 for the above four groups, showing that many more genes' expressions were significantly changed by TPM than by placebo at each time point, with control for multiple testing. Of 359 genes shared between week 8 TPM and week 12 TPM groups, 300 genes were exclusively detected in TPM responders. Of 106 genes shared between week 8 placebo and week 12 placebo groups, 72 genes were exclusively detected in placebo responders. There is no overlap between these two gene sets. Among 300 genes exclusively detected in week 8 and week 12 TPM responders, 34 genes had nominal P values < 5 × 10 −3 at  IL15RA in GABA receptor signaling represent direct targets for TPM. By contrast, none of 72 genes exclusively detected in week 8 and week 12 placebo responders contain direct targets for TPM. By comparing these above DE genes detected at nominal P values < 0.01 based on secondary outcomes with those DE detected at same statistical significance threshold based on primary outcome for week 8 TPM, week 8 placebo, week 12 TPM, and week 12 placebo groups, respectively (30), 55, 11, 11, and 15 genes were shared between primary analysis and secondary analysis for the above four groups, respectively (     Table 3). It is noteworthy that 9 of the 55 genes identified by primary analysis and secondary analysis for week 8 TPM-including 4 upregulated genes, i.e., CD164, AKAP11, FGFR1OP2, and PTEN, and 5 downregulated genes, i.e., EMILIN2, DGCR14, BCR, GANAB, and NAGK, were among the 93 (48 up-and 45 downregulated) representative genes selected based on primary analysis (30), and the directions of gene expression changes were all consistent between these two analyses.
To pinpoint top genes that were changed in TPM responders compared with TPM non-responders, volcano plots, where the log10(P value)'s are plotted versus log2(FC)'s, were generated for week 8 TPM ( Figure 1A) and week 12 TPM group (Figure 1B Table S4 in Supplementary Material). Among them, none was significantly changed in week 12 placebo group at a nominal P value < 0.01. Four genes (blue color) were significantly changed in TPM responders compared with TPM nonresponders at a nominal P value < 5 × 10 −3 with a |FC| > 1.

identification of enriched Pathways in Treatment responders
Both IPA and Onto-Tools Pathway-Express were applied to detect enriched pathways for 300 and 72 genes uniquely for TPM and placebo responders, respectively. Together, at nominal i.e., PTEN, was upregulated, which was also the gene shared between the three genes detected for phosphatidylinositol signaling system and the seven genes detected for PI3K/AKT signaling in the Signal Transduction category ( Table 4). Because PTEN, a central negative regulator of the PI3K pathway (41), is required for modulating synaptic activity during plasticity (42), a 36-node PTEN-centered molecular interaction network was generated for each of week 8 and week 12 TPM groups (Figures 3A,B)

DiscUssiOn
The advent of global gene expression profiling has generated unprecedented insight into our molecular understanding of drug addiction and treatment. We previously have identified genes involved in glutamate receptor and GABA receptor signaling are changed among TPM responders compared with nonresponders based on primary outcome (i.e., METH abstinence) and a set of crucial pathways involved in neuronal function/ synaptic plasticity, signal transduction, cardiovascular function, and inflammation/immune function are significantly enriched among TPM responders (30). However, the primary analysis was limited to only METH abstinence phenotype (30), and certain significantly modulated genes and pathways could be missed because of the limited sample sizes. This study differed from the primary analysis in several aspects. First, the secondary analysis applied the LCA classifications based on the six non-longitudinal binary secondary outcomes of weeks 1-12 into "responder" and "non-responder" classes, and those in the "responder" class scored consistently better on each of the six secondary outcomes than those in the "non-responder" class, as shown in Ma et al. (31). Such a classification provided greater sample sizes and more balanced comparisons for week 8 TPM, week 8 placebo, week 12 TPM, and week 12 placebo groups, compared with the primary analysis. Second, this secondary analysis applied both ordinary Student's t-test and the empirical Bayes moderated t-test, which is distinct from the more traditional association analyses utilizing only the ordinary Student's t-test, and is particularly suitable for small-to-moderate samples sizes. Although the results obtained from this study are from peripheral blood and need further validation and examination in CNS, it is quite striking that our secondary analyses based on six non-longitudinal binary secondary outcomes revealed a rather consistent and meaningful pattern that TPM changed more genes than placebo at both week 8 and week 12, which not only confirmed important genes and pathways revealed by primary analysis based on primary outcome (30) but also extended the results by identifying additional new genes and pathways on TPM response to METH addiction.  At individual gene level, because gene sets detected by both ordinary Student's t-test and empirical Bayes moderated t-test had relatively high proportions of overlap (64.50 ± 7.43%, range 56.55-74.31%), they were merged together at each time point for TPM and placebo groups, respectively ( Table 1). Among 300 genes detected exclusively in TPM responders at weeks 8 and 12, 34 genes had nominal P values < 5 × 10 −3 at either time point and FDRs < 0.15 at both time points. After excluding two genes with inconsistent directions of gene expression changes, i.e., FNIP2 and TOX4, 32 genes were selected based on the above statistics criteria with consistent directions of gene expression changes at both week 8 and week 12 ( Table 2). FDR is expected proportion of erroneously rejected null hypotheses among rejected ones. FDR threshold is determined from observed P value distribution, and hence is adaptive to actual data. An FDR threshold of 0.15 was chosen as significance threshold, which has been used in previous gene expression studies in choosing significant genes, e.g., Ref. (43)(44)(45)(46)(47)(48). FDR is defined as the expected number of discoveries that are not truly DE divided by the total number of discoveries. An overly stringent control for FDR can result in a large number of false negatives (49)(50)(51). Therefore, determination of an appropriate FDR threshold is critical for effectively identifying truly DE genes, while minimizing both false positives and false negatives. By applying a cross-validation approach, an optimal selection of FDR threshold is shown to provide a good performance on model selection and prediction (52). Twelve of them, i.e., CASP4, COX19, CUX1, GABARAPL1, GNG2, GPR155, HSF1, IL15RA, NLRP1, SIL1, SLC25A19, and UBAP2, could have critical functions for drug addiction. Of them, GABARAPL1 encodes a protein (87% identical, 94% similar) very similar to GABAA-receptor-associated protein (GABARAP) (53) and is expressed at higher levels than GABARAP in CNS (54). Further, GABARAPL1 is suggested to be a major protein interacting with GABAA receptors (54). GPR155, which encodes an integral membrane protein related to G-protein-coupled receptors (GPCRs), is highly expressed in lateral part of striatum and hippocampus (55). Many neurons that are identifiable as GABAergic might express GPR155, implicating its pivotal role in GABAergic neurotransmission (55). GPR155 is dysregulated in lymphoblastoid cells in males with autism spectrum disorders (ASDs) relative to their non-affected siblings, suggesting that the gene is associated with ASD (56). IL15, an important cytokine in immune function, is essential to maintain neurochemical homeostasis (57). IL15RA, which encodes IL15 receptor alpha subunit, has a regulatory function during inflammation. Il15ra knockout mice have deficits in hippocampal-dependent memory and GABA transmission (58). Thus, GABARAPL1, GPR155, and IL15RA in GABA receptor signaling could be direct targets for TPM. By contrast, no genes were identified in placebo responders using same criteria.
In this study, we applied both an ordinary Student's t-test and an empirical Bayes moderated t-test. The ordinary Student's t-test is the most commonly used method for comparing the expression levels of genes between two groups, and the computation of the P value for this test is straightforward as long as the assumptions of the test are satisfied. This test has been used in our previous gene expression study based on the primary outcome (30). The empirical Bayes moderated t-test, available in LIMMA package of Bioconductor, is also a popular method for two-group comparisons of gene expressions, which reduces estimated sample variances toward a pooled estimate, producing more stable result when the sample sizes are small (59). Each of ordinary Student's t-test and empirical Bayes moderated t-test has distinct advantages and disadvantages. The ordinary Student's t-test is easy to apply, but variance estimates could be skewed by those genes having a very low variance (60). The empirical Bayes moderated t-test is an innovative method that borrows information between genes using an empirical Bayes method to obtain posterior variance estimators, and compute a moderated t statistic that follows a t distribution with augmented degrees of freedom (61), but in this study, this method produces overall higher FDRs compared with ordinary Student's t-test, indicating a lower statistical power. Nevertheless, the two analyses by ordinary Student's t-test and empirical Bayes moderated t-test demonstrated largely consistently detected genes. Similar to our study, Uusküla et al. (35) applied two different statistical approaches, i.e., ANOVA and empirical Bayes moderated t-test, and genes identified by these two approaches were shown to be relevant for the clinical outcome. We did not choose to apply nonparametric Wilcoxon rank-sum (i.e., Mann-Whitney U test) because such test has a reduced statistical power and typically detects fewer genes compared with parametric test (62). By comparing these above DE genes detected at nominal P values < 0.01 based on secondary outcomes with those DE detected at same statistical significance threshold based on primary outcome (30), 55, 11, 11, and 15 genes were shared between primary outcome and secondary outcomes for the week 8 TPM, week 8 placebo, week 12 TPM, and week 12 placebo groups, respectively ( Table 3). Among the 55 genes shared between primary outcome and secondary outcomes for week 8 TPM, 9 genes, i.e., 4 upregulated genes, i.e., CD164, AKAP11, FGFR1OP2, and PTEN, and 5 downregulated genes, i.e., EMILIN2, DGCR14, BCR, GANAB, and NAGK, were also among the 93 (48 up-and 45 downregulated) representative genes selected based on primary outcome with consistent directions (30). Therefore, these nine DE genes could be particularly related to TPM treatment response for METH dependence. Of them, PTEN, which encodes a protein that functions as a protein tyrosine phosphatase as well as a lipid phosphatase, is of particular interest. PTEN plays a critical role in both CNS development and maintenance of CNS circuit structure and function (63). PTEN is shown to be a mediator of synaptic plasticity in the adult brain (42,64) that is required for NMDAR-dependent long-term depression, and alternations of PTEN at synapses could lead to behavioral and cognitive dysfunctions (65).  (67). GAS2L1 is a susceptibility locus for schizophrenia (68), and SASH1 gene expression was affected by a history of substance dependence/abuse (69). TCF4, which encodes a basic helix-turn-helix transcription factor, regulates gene expression in immune system and in brain development (70) and is significantly associated with schizophrenia (71).
At pathway level, seven biologically important pathways (i.e., Protein Ubiquitination Pathway, Phosphatidylinositol Signaling System, PI3K/AKT Signaling, Antigen Presentation Pathway, fMLP Signaling in Neutrophils, Role of PKR in Interferon Induction and Antiviral Response, and NRF2-mediated Oxidative Stress Response) were significantly enriched in TPM responders compared with non-responders for secondary outcomes (Figure 2). Of them, PI3K/AKT signaling pathway appears to be an important pathway for neuronal survival (72). Six genes (i.e., GYS1, HSP90B1, NFKBIE, PPP2R5D, RRAS, and TP53) were downregulated by TPM in this pathway, and HSP90B1, which encodes an endoplasmic reticulum chaperone gene, is altered in the postmortem brain of bipolar disorder patients (73), and is associated with bipolar disorder (74) and schizophrenia (75)(76)(77)(78). PTEN, which encodes a dual-specificity protein phosphatase that negatively regulates the PI3K/AKT signaling pathway (79)(80)(81), was upregulated by TPM. PTEN-centered networks of 36 interactors for week 8 and week 12 TPM responders included 19 dysregulated genes (Figures 3A,B). Of them, CSNK1A1, CTNND1, CXCR4, DTX1, MAPK14, PLEKHF2, PSMB2, PSMD1, and PTEN could be important for TPM responses. Besides PTEN, MAPK14, upregulated at both time points, encodes p38 MAPK, which plays an essential role in ROS formation and oxidative stress (82), production of inflammation mediators (83,84), and neuronal apoptosis (85,86). Further, CXCR4, also upregulated at both time points, encodes a chemokine receptor critical for anti-inflammatory response (87) and p53-mediated neuronal survival (88). Ubiquitin-mediated proteolysis is involved in the turnover of many short-lived regulatory proteins. Timed destruction of cellular regulators by the ubiquitin-proteasome pathway plays a critical role in ensuring normal cellular processes. Genetic approaches or pharmacological intervention that alters the half-lives of these cellular proteins may have wide therapeutic potential (89). It is worthy of noting that PI3K/AKT signaling and PTEN signaling pathways were detected to be enriched pathways for gene modulated by nicotine (90), and therefore, PI3K/AKT signaling appears to be a crucial pathway affected by various psychoactive drugs.
An integrative model based on enriched pathways has been proposed to explain the molecular mechanisms of TPM's effect on METH addiction (Figure 4). METH could lead to increased oxidative stress by altering PI3K/AKT signaling pathway, by interacting with vesicular monoamine transporter 2, which leads to accumulated cytoplasmic dopamine with resultant free radical formation (91). The transcription factor NRF2, a guardian of redox homeostasis, regulates a coordinated induction of a set of cytoprotective, antioxidant, and anti-inflammatory genes in response to oxidative stress and inflammation (92,93). NRF2-mediated oxidative stress response pathway is one of the major intrinsic antioxidant response of the brain, and NRF2 is a therapeutic target for treatment of neurodegenerative diseases, e.g., Alzheimer's disease and Parkinson's disease (94). TPM provides neuroplasticity by modulating protein ubiquitination and PI3K/AKT signaling pathways, which in turn, could decrease oxidative stress and increase neuroplasticity, which then could lead to abstinence and reduction of METH use. The integrative model proposed based on primary outcome [i.e., Figure 2 of Ref. (30)] included "PI3K/Akt/GSK-3 signaling" and "Mitochondrial oxidative stress" pathways downstream of "Dopamine receptor signaling" pathway, and the integrative model for this study (i.e., based on secondary outcomes) has substantiated the previous models branch downstream of "Dopamine receptor signaling," and further the "Protein Ubiquitination" pathway is a newly added pathway downstream of "GABA Receptor Signaling" pathway of the previous integrative model, which then could affect both neuroplasticity and neuronal apoptosis.
In this study, by analyzing gene expression profiling of whole blood, we attempted to define transcriptional patterns that differentiate TPM responders from non-responders. Whole blood has been increasingly used as a more accessible tissue for identifying proxy gene expression biomarkers for CNS, e.g., brain's circadian phase (95) and ASD (96,97). As shown in Vawter et al. (98), the use of whole blood for studying gene expression could avoid several important confounding variables associated with postmortem brain studies, e.g., hypoxia, pyrexia, postmortem interval, mRNA integrity, cellular heterogeneity of subcortical and cortical tissues that need to be controlled for in subsequent gene expression analyses. Further, recent studies demonstrate that blood cell-derived RNA could be used to distinguish schizophrenia, bipolar disorder, and control samples with high accuracies (99). The consistency of peripheral gene expression data and the overlap with brain expression has also been evaluated by Rollins et al. (100), which demonstrated that postmortem subjects' brain and peripheral blood mononuclear cell (PBMC) profiles showed co-expression levels of summarized transcripts for 4,103 of 17,859 (22.9%) RefSeq transcripts. However, because of the concern that whole blood gene expression is not completely correlated with brain gene expression and has a heterogeneous composition (e.g., T-and B-lymphocytes, PBMCs, and other cell types), gene expression patterns identified by this study shall be interpreted with caution and required to be further validated in CNS studies. One of the major limitations of this study is that no quantitative real-time PCR was applied to validate gene expression changes of the 32 gene candidates presented in Table 2, which was primarily due to a lack of sufficient high quality RNA for us to conduct such analysis.
In conclusion, this study of the transcriptome of secondary outcomes provided additional biological insights into TPM treatment response for METH dependence beyond the previous gene expression study based the primary outcome (30). Analyses based on ordinary Student's t-test and empirical Bayes t-test have not only identified novel sets of genes consistently for week 8 and week 12 TPM responders, but also detected several unique pathways, particularly protein ubiquitination and PI3K/ AKT signaling pathways, and also a novel PTEN-centered gene interaction network. Therefore, TPM treatment could lead to a decreased METH dependence by reducing oxidative stress and inflammation and enhancing neuroplasticity, which have extended the integrative model based on primary outcome. Combining results obtained from this study with those of the previous study (30), TPM response in METH-dependent subjects is a highly complex process encompassing a diverse spectrum of biological pathways that can be classified into Neuronal function/ Synaptic plasticity, Signal transduction, Cardiovascular function, Inflammation/Immune function, and Oxidative Stress Response categories.

eThics sTaTeMenT
The Institutional Review Boards of each clinical site and the VA Human Rights Committee approved the protocol for and conduct of the study.  The clinical and expression data used in this study were from a clinical trial, which was conducted through the Department of Veterans Affairs Cooperative Studies Program (Interagency Agreement No. Y1-DA4006) with the financial support from National Institute on Drug Abuse of NIH. The clinical trial data used for this study are registered in www.Clinicaltrials. gov with the identifier: NCT00345371. The authors express our gratitude to invaluable contributions of clinical information and blood samples by all participants in this clinical trial, as well as the dedicated work of many research staff at the clinical sites who participated in the trial. All statistical and bioinformatics analyses of expression data were in part supported by NIH grant DA-137873 to ML and an NIH contract to Information Management Consultations (IMC, Reston, VA).

FUnDing
Statistical and bioinformatics analyses of expression data were in part supported by National Institutes of Health grant DA-137873 to MDL and an NIH Contract to IMC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.