Inclusion and exclusion criteria
Quality of the studies
Meta-analysis of association of the NRXN1 variants with PSD
Quality of studies
NRXN1 coding region (exon)
NRXN1 no coding region (intron)
NRXN1 promoter region
III. Análisis unitario
IV. Explicación de variables
V. Variables de sujeto
VI. Variables de la enfermedad
VII. Características del marcador genético
VIII. Variables metodológicas
IX. Variables resultado
Objective: To perform a meta-analysis of all published genetic association studies of NRNX1 variants and psychotic spectrum disorders (PSD).
Methods: Potential studies were identified through PubMed/MEDLINE, EMBASE, HuGeNet, GeneCard and WoS up from 1980 until June 2015 (week 5). Published observational studies reporting NRXN1 variants in PSD cases and in non-PSD controls were all considered eligible for inclusion in this systematic review. Two reviewers selected studies for possible inclusion and extracted data independently following a standardized protocol. Different meta-analyses were performed for each NRXN1 variant and PSD and controls, with a random-effect model to calculate the pooled OR and its corresponding 95% CI. Forest plots and Cochran’s Q-Statistic and I2 index were calculated to check for heterogeneity. Subgroup analyses and chi-square test were carried out to analyze potential moderators. Publication bias and quality of reporting were also analyzed.
Results: 19 studies met our inclusion criteria, providing a total sample of 25208 patients with PSD and 56971 controls. Meta-analyses of NRXN1 deletions (OR = 3.36), exons (OR = 3.53) and introns (OR = 2.02) showed evidence of an association between NRXN1 gene and PSD. Only one study influed in a meta-analysis (NRXN1 deletions) and no quality criterion affected as moderating variable. There was evidence of potential publication bias in NRXN1 deletions, exonic and promoter meta-analyses.
Conclusions: We found association between some variants of the NRXN1 gene and PSD. It´s important to develop a greater effort for the study of the NRXN1 gene and bigger number of studies that will clarify it´s association.
Keywords: NRXN1; neurexin 1; 2p16.3; meta-analysis; schizophrenia; psychotic disorder; schizoaffective disorder; psychosis.
The appendix is in Spanish.
Schizophrenia is the most studied psychotic disorder. It is characterized by important alterations with acute repercussion in the social functioning. It affects approximately 1% of the world population and accounts for around 2.5% of global health expenditure (Meltzer, 1999). On the other hand, aberrant synaptic connectivity is a feature of the neuropathology of schizophrenia (Stephan et al., 2006) and it is documented that both heterozygous partial deletions and other mutations and disruptions of the NRXN1 gene (such as the promoter area or the N-terminal coding regions of neurexin 1α) are associated with susceptibility to neurocognitive disorders, such as schizophrenia (Duong et al., 2012; Guilmatre et al., 2009; Rujescu et al., 2009; Walsh et al., 2008).
Kirov et cols. found a deletion of the NRXN1 gene in the 2p16.3 region of a mother and two affected siblings of the same family, as well as in two identical twins concordant for childhood-onset schizophrenia (Kirov et al., 2008). In same direction, Yue studied a Chinese population (Yue et al., 2011) and investigated the possible association between NRXN1 polymorphisms and schizophrenia using a sample of 768 cases and 738 healthy controls. As a result of the study they obtained three haplotypes were associated with schizophrenia.
Rujescu and colleagues performed DNA chip analysis (microarrays) on 2977 European patients with schizophrenia and 33,746 European controls and identified 66 deletions and 5 replications in the NRXN1 gene: 12 deletions and 2 duplications in patients with schizophrenia (0.47%) compared to the 49 deletions and 3 duplications in controls (0.15%) (Rujescu et al., 2009). No breakpoints were found in common, and CNVs ranged from 18 to 420 Kb. When they restricted the association analysis to CNVs that altered the exons, they identified a significant association with a high odds ratio (0.24% of cases vs. 0.015% of controls, p = 0.0027, OR = 8.97). The research team suggested that deletions of NRXN1 that affect exons may pose a risk for the development of schizophrenia.
It is important to assume the complexity of schizophrenia and the psychotic spectrum, so the susceptibility of various genes is probably associated with these diseases. In order to explain the association of variants with complex disorders, to recall we have two theoretical models: the "common disease / common variant" (CDCV) model and the "common disease / rare variant "(CDRV) (GWAS Consortium, 2009). The CDCV model suggests that relative genetic variants (often> 5%) in the population should confer a lower or a medium risk (eg OR = 1.1-1.5). On the other hand, the CDRV model postulates that the complex features could derive from several rare mutations of individuals (usually <1%), but with strong relative effects (eg OR> 10).
Furthermore, the possibility of variable expression or reduced penetrance together with other genetic or environmental factors that may influence the expressed phenotype of neurocognitive disorders has been suggested (Dabell et al., 2013). Mouse models have found that non-expression of NRXN1ɑ has been associated with a decrease in excitatory synapses, altered sensory-motor suppression, increased toilet behaviors, nest construction problems and parental skills, without learning impairment, Memory and social interactions (Etherton et al., 2009; Geppert et al., 1998).
There are differences in the published findings of the genetic association of the NRXN1 gene with psychotic disorders, when observed in several studies that do not obtain significant differences for the alterations of this gene, between the groups with patients with schizophrenia and the healthy control groups (Walsh et al., 2008: Kirov et al., 2008; Ikeda et al., 2010). As mentioned above, the genetic association studies referred to have investigated the different areas of NRXN1 (promoter, coding and non-coding) (Guilmatre et al., 2009; Todarello et al., 2014; Vrijenhoek et al ., 2008) and in other different types of genetic variants (insertion, deletion, duplication and polymorphisms) (Gauthier e t al., 2011; Levinson et al., 2012; Shah et al., 2010; Van Den Bossche et al., 2012). These inconsistencies may be the result of multiple factors that need to be systematically evaluated (eg homogeneous populations, type of genetic polymorphism, related individuals, sample size...), as well as to analyze which specific variants of the NRXN1 gene can influence actually in the possible association between said gene and psychotic disorders. Therefore, the objectives of the study are to evaluate the general association of the NRXN1 gene and Psychotic Spectrum Disorders (PSD) and besides that, study the specific association of the different types of alterations (insertion, deletion, duplication and polymorphisms) and the areas (promoter, coding and non-coding) of the NRXN1 gene.
A search was carried out from 1980 to June 2015 (week 5) in different bibliographic databases: PubMed / MEDLINE, EMBASE, HuGeNet, GeneCard and WoS using the search terms: "NRXN1" OR "neurexin 1" OR "2p16.3" AND "schizophrenia" OR "psychotic disorder" OR "schizoaffective disorder” OR “psychosis". Lists of included original studies and review articles were reviewed to identify other potentially selectable studies. To minimize potential publication bias, restrictions related to the study period, sample size, population, publication´s language or study type were removed. Authors of the original studies were contacted by e-mail when it was considered necessary to collect additional data not included in the studies.
Inclusion and exclusion criteria
The inclusion criteria established for the selection of articles were: criterion 1) observational studies (prospective, cross-sectional and case-control), criterion 2) to analyze the NRXN1 gene in patients diagnosed with PSD (cases), criterion 3) to analyze the NRXN1 gene in healthy subjects (controls) without PSD. The case was defined as having, at the time of the study, the diagnosis of the following PSD: schizophrenia, psychotic disorder, schizoaffective disorder, psychosis; such as DSM-III, DSM-IIIR, DSM-IV or DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders) diagnosis and established by a validated diagnostic instrument or psychiatric interviews. The study population consisted of outpatients, patients admitted or treated in a rehabilitation unit in different stages of the disease (first episode, acute phase or stability period). Studies that met these criteria were chosen in this systematic review. In this meta-analysis, studies of any gender and ethnic group were included. The results of the studies analyzed were classified according to the type of genetic alteration (insertion, deletion, duplication and polymorphisms) and the affected area (promoter, coding and non-coding) of the NRXN1 gene.
However, other research designs (as abstracts, single case studies, designs based on family studies and population studies with healthy individuals excluded) and phenotypes other than those included in the PSD, in addition to studies describing the genetic effects in other phenotypes with mental pathology not mentioned in the inclusion criteria such (as personality disorders, affective disorders, autistic spectrum disorders, development disorders, eating disorders, substance abuse and dependence disorders, or neurological syndromes), were excluded.
Likewise, items with duplicate case samples or controls and if the same samples were used in another article already included in the present meta-analysis, were excluded. We also excluded those items from which the genotyping data needed for the analyzes after contacting the authors were not included, as well as those studies that measured allelic frequencies of the NRXN1 gene in the sample of cases and controls, rather than measuring different types of alterations of said gene (i.e. deletions, insertions, duplications or polimorphisms).
Two investigators (PGM and TBC) independently reviewed the title and study summaries to check the inclusion criteria. The complete articles were obtained from the bibliographic databases of the Internet whenever possible or by contacting the authors. First, the titles and summaries of the articles identified in the search were independently reviewed. Each review author then examined the full text of all the studies they considered to be relevant. Each review author compiled a list of studies, which were thought to meet the inclusion criteria. Interobserver reliability was assessed using a correlation analysis (Cohen kappa interobserver coefficient). Data from each of the selected studies following the application of the inclusion / exclusion criteria were also independently extracted blindly by the same researchers using a previously elaborated data extraction protocol (Supplementary Appendix, Supplemental digital content 1) and introduced them into independent databases. Likewise with the inclusion / exclusion criteria, after the concordance analysis, in case of discrepancy a consensus decision was made with the involvement of two senior researchers (FNM and JSM) in the case of not reaching it.
In case that several psychiatric diagnoses were described in one study, only data from the samples with PSD included in the inclusion criteria were extracted. The analysis unit used was the studies to ensure that the data were not duplicated or repeated, so that in case of multiple articles from the same study, only the results of the publication with the largest number of participants were included.
Information on the data extraction form included: author(s), journal and year of publication; methodology details (study design, sample size for both cases and controls, diagnostic tools for determination of case status and definition of case status); and sample characteristics (gender ratio, mean age, ethnic background, country, environment-gene interaction study, NRXN1 insertions, NRXN1 deletions, NRXN1 duplications, NRXN1 polimorphisms, coding region alterations, The number of affected controls, number of controls affected, validation method [MLPA, qPCR, Affymetrix, DosageMiner, QuantiSNP...]).
Quality of the studies
The quality of each of the studies selected for inclusion was analyzed through a checklist with 12 relevant methodological aspects. East List was obtained from different sources such as the STREGA statement (Strengthening the Reporting of Genetic Association Studies (Little et al., 2009) and for what was analyzed Different aspects (Hirschhorn et al., 2002; Navarro-Mateu et al., 2013; Sagoo et al., 2009). Specifically, the quality criteria applied were: (1) representativeness of cases; (2) representativeness controls; (3) same diagnostic instrument used for in cases and controls; (4) same test/scales in both case and control groups; (5) assessment of ethnicity; (6) genotyping is done blindly of the phenotype and vice versa; (7) inform quality control procedures of the methods of genotyping; (8) Hardy-Weinberg equilibrium test (HWE); (9) stratification of population; (10) possible interactions gene-gene and gene-environment; (11) analysis adjusted for potential confounding factors; (12) control for multiple comparisons. For each study to which the quality criteria were applied, it was also performed a calculation of the Total Quality Score (TQS). This consisted of the sum of the corresponding criteria that the study complied with (rank between 0 and 12, with highest score in case of higher quality). The discrepancies in the quality evaluation of each study were resolved by consensus. We did not exclude studies with low-quality scores.
This meta-analysis examined the association of variants of the NRXN1 gene with PSD to calculate the risk. For all studies, the OR and the corresponding 95% confidence interval (95% CI). Concordance interevaluator was measured by Cohen's kappa coefficient in the inclusion and exclusion criteria. Random effects model was applied in the statistical analysis because a high heterogeneity was expected between a priori studies. The same model assumes genuine diversity in the results of several studies and incorporates a variance between the studies in the calculations.
In each meta-analysis, we calculated the combined OR and its corresponding CI 95%. In addition, the statistical significance of the combined OR was assessed using the Z test (Sanchez-Meca et al., 2008). The sensitivity analysis was carried out to assess how our results were substantially influenced by the presence of an individual study by withdrawing each of the studies systematically and recalculating the significance of the outcome.
To assess the heterogeneity between studies, three methods were used: Cochran’s Q-statistic, I2 index and visual inspection of the forest plots, with their corresponding 95% confidence intervals (95% CI). Forest plots are constructed to represent the estimated individual and pooled effects. The Q-statistic is a weighted sum of squared deviations of estimated individual ORs of the studies in regard to the combined estimator (the magnitude of the effect of each individual study is compared with the combined estimator), so that assesses the heterogeneity of the studies in the meta-analysis. When the ORs are homogeneous, Q follows a distribution X2 with r-1 (r is the number of studies) degrees of freedom (d.f.). If p<0.05, heterogeneity is considered statistically significant. The inconsistency between studies is quantified with the index I2 (I2=Q-d.f./Q), which can be interpreted as the total percentage of variation among several studies due to heterogeneity. The I2 takes values between 0 and 100%, with the higher values denoting a greater degree of heterogeneity (0-25%: no heterogeneity; 25-50%: moderate heterogeneity; 50-75%: large heterogeneity and 75-100%: extreme heterogeneity). In cases of heterogeneity, an attempt was made to find reasons explaining it by conducting sensitivity analyzes with the characteristics of most important studies.
To explore heterogeneity, different analyzes were designed by subgroups which allow Chi-square tests among group differences, taking as potential moderators the different quality elements previously described. In order for the subgroup analysis to be useful and applied properly, the meta-analysis met the following two criteria: 1) degree of heterogeneity I2≥25%, 2) minimum of 10 studies included in the analysis.
In order to assess whether publication bias was a threat to the validity of ORs combined, funnel plots were applied using the 'trim-and-fill' method of 'Duval and Tweedie's' (Duval and Tweedie, 2000), as well as Egger's test (Steme and Egger, 2005). When the asymmetry is observed in the funnel graph, the estimated corrected effect for small study effects, such as publication bias, is calculated by the trim-and-fill method. It uses the available data to calculate the number and results due to the lack of (unpublished) studies and recalculates the overall effect that would be obtained with its inclusion. Egger test is an unweighted regression consisting of taking the accuracy of each study as an independent variable (the term accuracy being defined as the inverse of the standard error of each effect size) and the effect size divided by its standard error as the dependent variable. A non-statistically significant result of the t-test for the intercept hypothesis equal to zero allows the discard of publication bias as a threat to the validity of the pooled effect (Steme and Egger, 2005).
All statistical tests were interpreted assuming a level of bilateral significance of 5% (α = 0.05). The main statistical analysis was carried out using the RevMan 5.3 program (The Cochrane Collaboration, 2014). Funnel plots with the trim-and-fill method and the Egger test were calculated using the Comprehensive Meta-analysis 2.0 program (Biostat, 2005). Methods of analysis and inclusion and exclusion criteria were first specified and documented in the protocol. Published recommendations for systematic reviews of genetic association studies (Sagoo et al., 2009; Bray et al., 2006; Hirschhorn et al., 2002) were followed. Because we only used previously published data, it was not considered necessary to approve the study by an Ethics Committee.
Fig. 1 presents a flow chart summarizing the results of the research process and selection of studies. From a total of 41 potentially selectable studies, nine were excluded, six of them due to the case-based design or case series (Bradley et al., 2010; Cristino et al., 2014; Enggaard Hoeffding et al., 2014; Luykx et al., 2014; Novak et al., 2009; Souza et al., 2010), and three due to family-based design (Duong et al., 2012; Levinson et al., 2012; Van Den Bossche et al., 2013).
Abbildung in dieser Leseprobe nicht enthalten
Flow chart of the process of search and evaluation of selection of studies. (Adapted from: Sagoo GS et al. 2009, PLoS Medicine 6:e1000028. Moher D et al. and PRISMA Group 2009, PLoS Med 6:e1000097. Navarro-Mateu F et al. 2013, PLoS ONE 8:e66227.)
Cohen's kappa coefficient inter-observer of the three inclusion criteria, ranged from 0.69 (criteria 1 and 2) to 0.83 (criterion 3); for two exclusion criteria were 0.69 (criterion 1) and 1 (criterion 2). Subsequently, of the 31 studies selected for inclusion, three were excluded by duplicate samples, namely they had the same sample for several studies (Chen et al., 2013; Ivorra et al., 2012; Purcell et al., 2009). Thus, 28 studies were included in the review and data were extracted from them (Guilmatre et al., 2009; Rujescu et al., 2009; Kirov et al., 2008; Yue et al., 2011; Ikeda et al., 2010; Todarello et al., 2014; Vrijenhoek et al., 2008; Gauthier et al., 2011; Shah et al., 2010; Van Den Bossche et al., 2012; Levinson et al., 2011; Giegling et al., 2011; Gratacòs et al., 2009; Greenwood et al., 2013; International Schizophrenia Consortium, 2008; Ivorra et al., 2014; Jenkins et al., 2011; Jenkins et al., 2014; Kenny et al., 2014; Kirov et al., 2009; Lett et al., 2011; Magri et al., 2010; Muhleisen et al., 2011; Need et al., 2009; O'Dushlaine et al., 2011; Rees et al., 2014; Stewart et al., 2011; Ye et al., 2012). Nine of these studies were excluded because of the inability to extract data in order to obtain the frequencies of the necessary genetic abnormalities in NRXN1: on the one hand, the authors of seven of these studies were contacted, without obtaining the required data (Giegling et al., 2011; Greenwood et al., 2013; Ivorra et al., 2014; Jenkins et al., 2011; Jenkins et al., 2014; Need et al., 2009; O'Dushlaine et al., 2011). On the other hand, two studies (Yue et al., 2011; Lett et al., 2011) measured the allelic frequencies of the NRXN1 gene in the whole sample of cases and controls, rather than measuring the prevalences of the type of alterations or altered regions (deletions, insertions, duplications and polymorphisms; or exon, intron and promoter region), making it impossible to extract the information of interest. Finally, 19 studies were included in the meta-analysis.
The characteristics of the studies selected for inclusion are described in table 1, including year of publication, study design, number of cases with psychosis, healthy controls and the total of both, the percentage figure of the total number of men in the sample, mean age, ethnicity, the diagnostic tool used to assess psychosis, whether genetic inheritance has been assessed, the type of genetic alteration in NRXN1, the genetically altered region in the same gene and whether the possible gene-environment interaction has been analyzed. The total of the sample with psychosis was 25208, and that of the healthy controls of 56971. Table 2 shows the frequencies of types of alteration and altered genetic region of the NRXN1 gene from all studies included in the meta-analysis.
Meta-analysis of association of the NRXN1 variants with PSD
Initially, a meta-analysis was performed according to the different types of genetic alterations of the NRXN1 gene: i) deletions (Fig. 2); ii) duplications (Fig. 3); and iii) polymorphisms (Fig. 4). It is important to clarify that no meta-analysis was possible for the "insertions" of the NRXN1, because we only obtained results for one study : OR = 4.05; 95% CI = 0.42, 39.35. The overall association between genetic alterations and PSD reached statistical significance for deletions, not for duplications or polymorphisms (Figs. 2A, 3 and 4, respectively): i) OR = 3.36; CI 95% = 2.07, 5.46; p < 0.00001; I2 = 0%; ii) OR = 3.74; CI 95% = 0.96, 14.57; p = 0.06; I2 = 0%; iii) OR = 1.19; CI 95% = 0.71, 1.99; p = 0.51; I2 = 18%. Subsequently, a meta-analysis of the different genetic regions altered in NRXN1 obtained only significant association for the coding region (exon) and the non-coding region (intron). The results for specific analysis of the exon, intron and promoter area (Figs. 5a, 6a and 7, respectively) were as follows: i) OR = 3.53; CI 95% = 1.58, 7.86; p = 0.002; I2 = 58%; ii) OR = 2.02; CI 95 % = 1.10, 3.69; p = 0.02; I2 = 0%; iii) OR = 1.76; CI 95% = 0.31, 9.85; p = 0.52; I2 = 25%.
Table 1. Characteristics of association studies eligible for inclusión in Meta-Analysis of NRXN1 polymorphisms and Psychotic Spectrum Disorders (PSD).
Abbildung in dieser Leseprobe nicht enthalten
Cas-con: Case-control; CA: Caucasian; AA: Afro-American; OT: Other ethnics; AS: Asian; HI: Hispanic; Mh: Medical history; SCAN: Schedules for Clinical Assessment in Neuropsychiatry; DIGS: Diagnostic Interview for Genetic Studies; LPDS: Lifetime Dimensions of Psychosis Scale; FIGS: Family Interview for Genetic Studies; SCID-(CV,P,I,II): Structured Clinical Interview for DSM-(Clinical Version, Patient Edition, Axis I Disorders, Axis II Personality Disorders); Ent clin: Entrevista clínica; SADS-(L): Schedule for Affective Disorders and Schizophrenia-(Lifetime Versión); SCAN: Schedules for Clinical Assessment in Neuropsychiatry; OPCRIT: Operational Criteria Checklist; MINI: Mini -International Neuropsychiatric Interview; PRISM: Psychiatric Research Interview for Substance and Mental Disorders; SIS: Suicide Intent Scale; SAPS: Scale for the Assessment of Positive Symptoms; Del: Deletions; Ins: Insertions; Pol: Polimorphisms; Dup: Duplications; Cod: Coding; NoC: No Coding; Pro: Promoter.
Table 2. Frequencies of the different altered type and altered genetic region of the NRXN1 gene in included studies.
Abbildung in dieser Leseprobe nicht enthalten
No studies have been found that alter the results when they were extracted individually from the meta-analyzes during the sensitivity analysis, except for a specific article (Rujescu et al., 2009). It is interesting the possible effect of the article Rujescu et al. (2009) in the meta-analysis of the deletions (Fig. 2a), since its specific weight in meta-analytical calculations was 58.6%. When we proceed to the same analysis, apart from this study (Fig. 2b), we did not obtain a significant alteration of the values of the results, maintaining the statistical significance (OR = 4.39, 95% CI = 2.07, 9,32, p = 0.0001, I2 = 0%).
Abbildung in dieser Leseprobe nicht enthalten
Forest plot of NRXN1 deletions and PSD.