Association of a haplotype in the NRG1 gene with schizophrenia: a case-control study

Objectives: Schizophrenia (SZ) is a severe multifactorial disease. NRG1 is a gene acting in the development of SZ. A number of NRG1 single nucleotide polymorphisms (SNPs) and their haplotypes are associated with SZ. In the present study, we investigated the association of a NRG1 haplotype (G-C in rs6988339-rs3757930 frame) which was reported to be associated with SZ, and two other SNPs in the same gene (rs74942016, rs80127039) whose rare missense alleles were found in SZ patients. Also, we analyzed disease associations of potential new haplotypes constructed by the variants of these SNPs. Methods: We genotyped 4 SNPs in a sample consisting of 302 SZ patients and 333 controls from a local Turkish population. We tested the disease associations of these variants at single SNP, haplotype and diplotype levels in case-control design. Results: At single SNP level, the CC genotype of rs3757930 was associated with SZ (p=0.038). The previously reported association of G-C haplotype in rs6988339-rs3757930 frame was absent (p=0.416), but we found another haplotype (C-G in rs3757930-rs74942016, p=0.018) and three diplotypes (A-C/G-C diplotype of rs6988339-rs3757930 frame, C-G/C-G diplotype of rs3757930-rs74942016 frame, and A-C-G/G-C-G diplotype of rs6988339-rs3757930-rs74942016 frame) associated with schizophrenia in our sample. Conclusions: Our study indicated the associations of a SNP, a haplotype, and a diplotype of NRG1 with schizophrenia and supported the involvement of NRG1 gene in the development of the disease. Since our sample was collected from a limited geographic area, the associations we have reported need to be supported by further studies in di ﬀ erent populations.


Introduction
Schizophrenia (SZ) is a severe psychotic disease (MIM181500) which affects 0.5-1 % of general human population.The major clinical characteristics of SZ are positive symptoms, negative symptoms, and cognitive deficits [1].Genetic, environmental, epigenetic, and stochastic factors contribute to development of SZ phenotype due to its complex-multifactorial nature [2].Family, twin, and adoption studies revealed the strong effects of genetic factors on schizophrenia [3,4].The heritability of disease is estimated to exceed 80 % [5].During last couple of decades, many studies in different designs (SNP association case-control studies, genome wide association studies-also known as GWAS-, whole exome sequencing-WES-and whole genome sequencing-WGS-studies and meta-analyses) uncovered hundreds of genes with their risk alleles acting in development of schizophrenia [6].Among those risk alleles of various genes, common variants known as SNPs (Single Nucleotide Polymorphisms) with moderate influences [7], and rare variants including CNVs (Copy Number Variations) and SNVs (Simple Nucleotide Variations) with intermediate or severe effects on the disease phenotype have been reported [8,9].Nevertheless, the molecular pathology behind the disease has never been defined in detail [5][6][7][8].One of the several reasons for this indistinctness is the inconsistency between the results of genetic association studies done with the samples from different populations.In practice, the association between a multifactorial disease and a genetic factor (a SNP or a rare variant) found in a population may not appear in another one.Due to this fact, to produce more robust data for determining the effects of certain genetic factors on the disease phenotypes, genetic association studies are preferred to be replicated in different populations [10][11][12].
A number of genetic risk factors revealed by the early association studies done with the schizophrenia samples were the polymorphisms of NRG1 gene which encodes for the protein called neuregulin-1 [13,14].Neuregulin 1 protein have essential functions in the development of white matter [15].In the central nervous system, oligodendrocyte migration, differentiation and myelination are induced by the effect of neuregulin 1 [16,17].Also, when considered as the whole pathway instead of only one protein, the changes in neuregulin 1 signaling pathway are known to cause structural and functional abnormalities in the oligodendrocytes altering the dopaminergic function which underling behavioral defects seen with neuropsychiatric disorders [18].
The involvement of NRG1 in development of schizophrenia was supported by a recent functional study in which behavioral phenotypes similar to deficits of schizophrenia was appeared in NRG1 haplo-insufficient mouse models [19].Also, a lot of other studies revealed a number of new variants associated with SZ in the NRG1 gene in addition to many other ones in different genes [20][21][22][23].Additionally, epistasis and interactions of different genes that participate neural development can affect the phenotype of schizophrenia when particular combinations of their variants take place [22].
A list of SNPs with potential risk alleles have been brought out by a study in which the exons of the genes encoding for the proteins that function in the neuregulin signaling pathway (NSP) were sequenced in a cohort of SZ patients (a case-only study) [24].Two of these SNPs (rs74942016 and rs80127039) were located in the NRG1 gene as well.These two SNPs were suspected to affect the phenotype because of the amino acid changes caused by their minor alleles [24].In addition to the alterations in the base sequence of the NRG1 gene that cause amino acid changes in the peptide sequence (non-synonymous alterations), there are base alterations in the introns reported to be associated with schizophrenia as well.Two of those intronic alterations are the SNPs rs6988339 and rs3757930 which are located in the fifth and sixth introns respectively.These two SNPs did not show evidence for a single-SNP association, but the A-G haplotype they constitute was reported to be associated with schizophrenia in a Scottish sample [25].Besides this, the screening of rs6988339 as single SNP in a Han Chinese case-control group did not reveal a positive association with SZ [26].Except for the papers mentioned above, when we screened the scientific literature databases, we did not encounter any studies investigating the associations of those four SNPs with SZ.Hence, more studies confirming their disease associations seems essential to uncover their contributions to development of SZ.
In one of our studies, we have sequenced the whole exomes of 30 unrelated schizophrenia patients to find genetic factors which may be associated with SZ in our population.Even though we have found a lot of other variations (data not shown), we did not observe the suspected missense alleles of rs74942016 and rs80127039 with considerable counts in the cohort.To increase the possibility of catching more carriers of rare alleles in the counts sufficient for making more reliable observations, we increased the sample size (302 patients).Also, for comparison, we set a group of controls (333 samples).Additionally, we screened rs6988339 and rs3757930 SNPs which were the components of risk haplotype reported in Scottish sample [25].To find more evidence for the contributions of all those four SNPs to development of SZ by confirming their disease associations, we carried out this case-control study.
According to literature database searches we have done, despite the presence of a few studies investigated some other NRG1 SNPs in the Turkish population [27,28] the present study seems to be the first one to investigate genetic associations of rs74942016, rs80127039, rs6988339 and rs3757930 SNPs in the NRG1 gene with schizophrenia in a Turkish sample.

Patients
We carried out all applications and procedures in accordance with the Declaration of Helsinki guidelines.The local Ethics Committee has granted ethical approval with Protocol # 2019/55.All individuals who volunteered to join our study either in the case or control groups provided written informed consents forms.The clinical diagnoses and evaluations of all volunteers in the case and control groups have been done by the second author who is a senior psychiatrist in the Psychiatry Department of School of Medicine at Inonu University (Malatya, Turkey).The criteria declared in The Structured Clinical Interview for DSM-V (SCID-I) [29] were used for diagnosis of schizophrenia.All participants in the group of cases were also evaluated according to Global Assessment of Functioning (GAF) criteria [30] and The Scales for The Assessment of Positive and Negative Symptoms (SAPS and SANS) [31].The evaluated positive symptoms (SAPS) were bizarre behavior, positive formal thought disorder, hallucinations, and delusions.In other respects, anhedoniaasociality, affective flattening or blunting, alogia, attention, avolitionapathy were the negative symptoms (SANS) that we have considered.In this study, we set our group of cases to cover 302 unrelated schizophrenia patients (213 males and 89 females) from Malatya, a city located in Eastern Anatolian Geographical Region of Turkey with a population of 810,000.The age average in the patient's group was 35.71 ± 10.08.In the group of cases, the averages of SAPS, SANS and GAF grades were 51 ± 13.3; 49 ± 19.4 and 50 ± 8.3 respectively.The average of time for the volunteers in the group of cases have been followed was 15 years (SD: 10 years, min.3 years).The average for age of Sözen and Kartalcı: A NRG1 haplotype and schizophrenia diagnosis was 24.5 with a standard deviation of 8.5.All volunteers in the group of cases declared that they had Turkish ethnic origin.We excluded the patients with mood disorders showing psychotic features, schizophrenia form disorder, schizoid disorder, schizoaffective disorder, schizotypal disorder, paranoid personality disorder, psychotic disorder due to a general medical condition and substance-induced psychotic disorder from our sample.
Our control group consisted of 333 individuals (157 males and 176 females) selected from healthy people living in Malatya-Turkey and who identified themselves to have Turkish ethnic origin.All volunteers in the group of controls have been evaluated by the second author to confirm the absence of any symptoms of Axis I psychotic disorders.The control samples have also reported the absence of any psychotic disorders among their first-degree relatives.

Extraction and preparation of DNA samples
The peripheral blood samples of volunteers have been drawn into EDTA coated tubes and used for extraction of genomic DNA with Invitrogen's (California, USA) Purelink DNA mini kit.The confidentiality of participants has been protected using a coding system which we applied to all blood and DNA samples.The quality and quantity of DNA samples were determined measuring the absorbances at 260 and 280 nm wave lengths in a UV spectrophotometer (Epoc-BioTek, Vermont, USA) and gel quantification using a gel documentation system (G:Box, SynGENE, Cambridge, UK) with the image analysis software (Gene Tools version: 4.3.10.0) from the same producer.All DNA samples were standardized to 20 ng/μL concentration and arranged in 96 well plate standard format.

SNP genotyping
We determined the genotypes of each subject in the case and control groups for each SNP (rs6988339, rs3757930, rs74942016, and rs80127039) on a Step One Plus Real -Time PCR system from Applied Biosystems (California, USA) using Applied Biosystems TaqMan ® Real-Time PCR Genotyping Master Mix (Catalog number: 4,304,437) and specific genotyping assays containing the PCR primers and allele specific probes for the target SNP designed by the same producer.The catalog numbers of TaqMan ® assays were "C___2870426_10" for rs6988339, "C__30843280_10" for rs3757930, "C_105335380_10" for rs74942016 and "C___105335415_10" for rs80127039.The PCR tests were carried out under the conditions suggested by the manufacturer (In each well of 96 well PCR plates, in a total volume of 10 µL completed with water, 5 µL of TaqMan ® Genotyping Universal PCR Master Mix, 0.5 µL of appropriate TaqMan ® genotyping assay for the SNP to be genotyped, and 20 ng of genomic DNA).We applied a hot start step (95 °C for 10 min) prior to PCR thermal cycles to activate DNA polymerase contained in the master mix.PCR was carried out as 40 cycles of: 15 s at 95 °C and 1 min at 60 °C.At the end of PCR, the real-time PCR system detected the SNP genotypes by endpoint plate reads.
The data produced by the real-time PCR genotyping tests were used for genetic association analyses at single SNP, haplotype and diplotype levels.

Genetic and bioinformatic analysis
Single SNP association: The accordance of case and control groups with Hardy-Weinberg equilibrium was tested using Pearson's chi-squared method.The disease association of each SNP were first tested as genotypic and allelic associations [32].Following that, we tested the associations of each SNP under common dominant and common recessive models [33].Haplotype association: We used Haploview software (version:4.2) for estimating the haplotypes that constructed by the SNPs we have screened [34].We also compared the haplotypes distributions in the case and control groups with the same software to reveal the haplotypes associations [34].The haplotype associations were tested by Pearson's chi-squared method applied by the software.Diplotype association: We define a diplotype as the combination of two haplotypes which were constituted by same SNPs on a pair of homologous chromosomes in an individual's genome [35].We used Microsoft Excel software's "filter" function to determine the diplotypes of the samples whose genotypes are homozygote for all SNPs in the inspected haplotype or who had maximum one SNP with heterozygote genotype in the haplotype.The counts of diplotypes in the groups were determined using "count", "count if" and "countifs" functions of Microsoft Excel.No estimations were done for diplotypes.We considered only the subjects whose diplotypes we were adequate to determine based on the homozygous SNP genotypes.We excluded the subjects who were heterozygote for two or more SNPs from diplotype association analyses.To test the association of a diplotype, we compared the case and control groups for the counts of carriers and non-carriers of that diplotype.The odds ratios were calculated by odds ratio calculator [36].

Statistical analysis
All statistical methods we have used and how their results should be evaluated have been explained in detail by Balding et al. [32] and Clarke et al. [33].We analyzed the data using Microsoft Excel and Haploview software [34].The results are presented as counts (n) and frequencies.Pearson's chi-squared goodness of fit method was used to test the statistical significances of the differences between the case and control groups for the counts and distributions of genotypes, alleles, haplotypes and diplotypes.The differences were accepted significant if the p-value was less than 0.05 (p<0.05).The statistical power of the study was calculated using G*Power software [37].

Single SNP association Genotypic and allelic associations
The general information about the SNPs screened in this study is presented in Table 1 and the comparison of our case and control groups for the distributions of genotypes and alleles of each SNP is given in Table 2.The statistical significance of differences between two groups were tested by Pearson's chi-squared test ("p" line in Table 2).As seen in the table, the distributions of genotypes and alleles of the SNPs rs6988339 and rs74942016 did not show a significant difference between our case and control groups (p>0.05).On the other hand, the differences observed in two other SNP loci (rs3757930 and rs80127039) were statistically significant (p<0.05).At rs3757930 locus, the C allele (p=0.01) and CC (p=0.038)genotype were more frequent in the group of cases.For rs80127039, the frequency of minor allele "T", was low in our total sample (0.023) and no homozygotes of this allele were observed in either group.Two patients and 13 controls were carriers of T allele as heterozygotes causing a significant difference between two groups (p=0.007,corrected: 0.015).Also, a significant difference was seen in the allelic distributions (p=0.0076,corrected: 0.016) suggesting a negative association and potential of T allele to have a protective effect against the disease.The case and control groups were both in accordance with Hardy-Weinberg equilibrium for each SNP (p>0.01)[34].

Single SNP association tests under common dominant and common recessive models
The single SNP associations were also tested under commondominant and common-recessive models (Table 3).The disease association of rs3757930 was more obvious in the recessive models.The CC genotype of rs3757930 was more common in the patients and the presence of at least one T allele in this locus was more frequent in the group of controls (p=0.01).

Haplotype association
The relations of four SNPs to each other were evaluated using Haploview software.The distances between each SNP pair as well as the numeric presentations of linkage disequilibrium between each SNP pair expressed as LOD scores, D′ and r 2 values are given in Table 4.The distributions of estimated haplotypes were compared between the case and control groups to find their possible disease associations and to have an idea about the potential effects of SNP-SNP interactions.At this step, each haplotype was taken as an allelic unit without considering their zygousities in the individuals.Twenty-five haplotypes in various frames constructed by different numbers and combinations of SNPs were estimated and their counts and frequencies were predicted by Haploview software (Table 5).Among these, the C-G haplotype in the rs3757930 -rs74942016 frame was seen more frequent in the group of patients (p=0.0176) with an odds ratio of 1.32 (95 % confidence intervals between 1.05 and 1.65).The same allele combination was encountered in the three SNP haplotype C-G-C constructed by the SNPs rs3757930 -rs74942016 -rs80127039 (Table 5).Whereas 6 haplotypes which seemed related to each other were more common in the group of controls (A-T haplotype of rs6988339 -rs3757930; T-G of rs3757930 -rs74942016 and the three SNP haplotype A-T-G of rs6988339 -rs3757930 -rs74942016 which is the combination of first two haplotypes).The rare and potentially protective T allele of rs80127039 seems linked to all these potentially protective haplotypes.

Diplotype association
We were able to determine 70 different diplotypes (homozygosities or the heterozygote combinations of haplotypes) in most of our samples.The diplotypes which were seen significantly different between the case and control groups are given in Table 6.Eight diplotypes constructed by two SNPs rs6988339 and rs3757930 (Haplotype Frame: rs6988339-rs3757930) have been determined in 244 cases (81 %) and 253 controls (76 %).The heterozygousity of A-C and G-C haplotypes (diplotype DT-1: A-C/G-C) was significantly more frequent in the group of cases (p=0.003) with an odds ratio of 1.9 (C.I. 95 % between 1.24 and 2.90).The second haplotype frame in which we have seen a potentially risk diplotype was rs3757930-rs74942016.We were able to determine haplotypes of this frame in 297 (98 %) patients and 317 (95 %) controls.The homozygousity of "C-G" haplotype of this frame (DT-2: C-G/C-G) was more frequent in the group of patients (p=0.03,odds ratio: 1.43 with C.I. 95 % between 1.03 and 1.99).When the haplotype frame was extended to include rs6988339 (haplotype frame: rs6988339-rs3757930-rs74942016) the A-C-G/G-C-G diplotype was also seen more common in the patients (p=0.003) with an odds ratio of 1.98 (C.I. 95 % between 1.259 and 3.118).
Apart from these, the distributions of rs6988339 genotypes were compared between the sub-groups of cases and controls constructed only from the carries C-G/C-G diplotype to confirm the effect of rs6988339 genotype (Table 7).While the A/G genotype of rs6988339 was more frequent in the subgroup of patients carrying the risk diplotype (p=0.031); the G/G genotype was more common in the sub-group of controls (p=0.011).In the same way, we tested the other possibilities of

Statistical power
The statistical power of this study was found to be greater than 80 % at a significance level of 0.05 for detecting a locus with an effect size of 0.3 by G*Power software [37].

Discussion
In the present study, we investigated disease associations of four SNPs (rs6988339, rs3757930, rs74942016 and rs80127039) located in the NRG1 gene at single SNP, haplotype and diplotype levels.Genetic factors have strong effects on the development of schizophrenia.Depending on its polygenic and multifactorial characteristics, schizophrenia has been reported to be associated with variants of many genes including NRG1, ERBB4, NRG3, COMT, BDFN, DAOA, DRD2, DRD3, DRD4, and TLR2 [39][40][41][42][43].The NRG1 gene is one of the earliest genetic factors found to be linked to the disease.To date, a number of alleles and haplotypes of NRG1 have been reported to be associated with schizophrenia.Among those variants, the G-C haplotype constructed by rs6988339 and rs3757930 SNPs [25] and missense variants of rs74942016 and rs80127039 SNPs [24] were mentioned in previous studies, but their disease associations were not confirmed in different populations.The disease association reported for the haplotype and the predicted damaging effects of two missense variants on the protein structure indicated the possibility of these four SNPs to participate the development of schizophrenia.To confirm the associations of these variants with schizophrenia we screened them in a case-control sample from Turkey.The group of controls was in accordance with Hardy-Weinberg Equilibrium (HWE) for all four SNPs (p>0.05)approving the reliability of all our genotyping tests.Even though deviations were seen in the group of patients for two SNPs (rs3757930 and rs74942016) when we set the significance level of p-value to 0.05, we ignored that since the threshold of p-value can be accepted as 0.01 for HWE tests [39].Even if we consider it significant, the deviation from HWE in rs3757930 which was seen only in the group of cases, but not in the controls provides evidence to support the disease association of this SNP.
Following the HWE confirmation, we tested the genotypic and allelic associations of each SNP with schizophrenia (Single SNP Association).The minor differences observed . . a p-Genotype: the presence of genotype tested against its absence.b p-Distribution: the difference of two groups was tested for the distributions of  genotypes.
between our case and control groups in rs6988339 and rs74942016 were not statistically significant (p>0.05) and there was no sufficient evidence to support disease associations of these two SNPs in our sample indicating the absence of their disease association, in accordance with Han Chinese sample [26].In general, to accept a to be associated with a disease, the distribution of its genotypes and alleles are compared between the case and control groups.If there is a difference, and that difference is statistically significant, the association is approved.The statistical significance of difference is expressed as p-values calculated by statistical tests like chi-squared test or Fisher's exact test.If the difference between two groups is significant, the p-value is less than the value we set (for the case-control studies, it is usually set to 0.05).If the p-value is above the set value (0.05) that means there is no significant difference between the case and control groups for the distributions of the alleles or genotypes of that SNP, hence that SNP is not associated to the disease.Even if there are differences in the counts, those are attributed to the chance [32,33].The other SNP, rs74942016 was mentioned by Hatzimanolis et al. [24] in a case-only study done with European Caucasian families.The six families had at least one member carrying the variant T allele.The presence of the minor allele in the 6 out of 24 families suggested the possibility of this SNP to be associated with the disease.Whereas that was a caseonly study which did not report about the presence and the frequency of variant T allele in the healthy population.The missense T allele of rs74942016 is known to cause the alteration of the valine at 228th position to leucine (Val228Leu) in the transmembrane domain of Type III Nrg1 β1A variant of Nrg1 protein (in different variants of Nrg1 protein, this amino acid position corresponds to 228, 232, 266, 299 or 321st positions).This amino acid change was reported to prevent the normal cleavage of Nrg1 protein by the enzyme γ-secretase [44].That explanation of functional mechanism by which the variant allele effects the protein expression makes rs74942016 seem to have a high potential to act in development of disease phenotype.In addition, the rare T allele of rs74942016 was reported to be present in the Hirschsprung disease in a Spanish sample supporting its effect on the phenotype [45].In the present study, we detected the T allele of rs74942016 in 22 (7 %) patients and 29 (9 %) controls in heterozygous state.Only 2 patients, but none of the controls were homozygous for the T allele.The frequencies of T allele in both of our groups (case and control) were 0.044 which was the allele frequency reported for the global population in dbSNP database [38].The genotype frequencies to be so close and the allele frequencies to be the same in the case and control groups caused no genotypic or allelic associations of rs74942016 to appear in our sample.One interesting point is the homozygousity of the T allele (TT genotype) to be present only in the group of patients, but not in any control samples.One of the possibilities is the T allele to have a recessive character, and its appearance in the phenotype may require its presence in the homozygous state.This may explain the absence of a significant difference in the frequency of heterozygote genotype between our cases and control groups.Nevertheless, only 2 homozygote subjects were not sufficient to make a reliable conclusion.Therefore, replicating the study with a larger sample size may help to uncover the effect of rs74942016 on the phenotype of SZ.On the other hand, when we investigate the other two SNPs rs3757930 and rs80127039, the differences between case and control groups were obvious and statistically significant (p<0.05)indicating their potential associations with schizophrenia.At the rs3757930 locus, the frequency of the major "C" allele was greater in the group of cases when compared with the values obtained from the group of controls (p=0.038for the CC genotype and p=0.01 for the C allele) indicating the positive association of this SNP with schizophrenia in our sample.When we consider the rs80127039 locus, despite the low frequency (0.023) and the absence of homozygotes of minor allele T in our total sample; two patients, but thirteen controls were found to carry this allele in heterozygous state (CT genotype).Accordingly, the frequency of T allele in the group of cases was much lower (0.003) than the group of controls (0.02) (p=0.0076,corrected: 0.016).This may suggest the negative association of the T allele with schizophrenia in our sample, or the protective effect of T allele against the disease.Previously, the presence of rs80127039 variant was reported in the Hirschsprung patients from Spain [41].Following this, the variant was mentioned by Hatzimanolis et al. after its appearance in a schizophrenia patient, and because of the amino acid change which it causes (R512W) was predicted to damage protein structure [24].In our study, the T allele of rs80127039 was present in the patients, but it was also present in the control group.Moreover, the allele frequency of T in our case group (0.003) was lower than it was in our control group which was same as the allele frequency in global population (0.02) in dbSNP database [38].
In addition to the allelic and genotypic association analyses, we have tested the associations of each SNP under common dominant and common recessive models which clarified disease associations of two SNPs (rs3757930 and rs80127039).As seen in the genotypic association analyses, the CC genotype of rs3757930 was more common in the patients.Correspondingly, the presence of at least one T allele in this locus (TT and CT genotypes) was significantly more frequent in the group of controls (p=0.01).The recessive model for rs80127039 produced the same result as genotypic association test, since there were no homozygotes for T allele in either group.
We estimated 25 different haplotypes when we analyzed the genotyping results by Haploview software.One of these SNPs (HT-5) was more frequent in the group of patients (p=0.0176)indicating its positive association with schizophrenia.HT-5 was C-G haplotype of the rs3757930-rs74942016 frame.The same association was seen with HT-16 (C-G-C haplotype of the rs3757930-rs74942016-rs80127039 frame) which was constructed by adding the downstream SNP rs80127039.Both haplotypes had the same frequencies since the C allele of rs80127039 was linked to HT-5 in our group.Apart from these, six haplotypes (HT-3, HT-6, HT-10, HT-12, HT-19, HT-25) were more frequent in the group of controls compared to the patients, suggesting their negative associations with the disease.The minor T allele of rs80127039 was common in three of those potentially protective haplotypes (HT-10, HT-19, HT-25) which appeared with the same counts as the heterozygotes of rs80127039 (2 patients and 13 controls).This was possibly caused by the presence of T allele on the chromosomal segment which carries A-T-G haplotype of rs6988339-rs3757930-rs74942016 SNP frame.In our sample, we were not able to observe the disease association of rs6988339-rs3757930 (G-C) haplotype reported by Walker et al. [25].This can be explained by the differences between populations.In the case of genetic association studies, it is not unusual a genetic factor to appear associated with a disease in one population, but not in another one [6,7].Because of that, disease associations of the genetic factors including the SNPs and haplotypes need to be approved by studies carried out using samples from different human populations [32,33].
In the previous studies, diplotypes of another gene were reported to be associated with schizophrenia [42].To find the potential risk diplotypes in our sample, we determined diplotypes of every subject.Since we did not have access to DNA samples or genotype data of the parents, we were not able to determine the exact diplotypes of the subjects who had heterozygous genotypes at two or more SNP loci.Despite this, nearly in 90 % of the whole cohort, we were able to determine more than 70 different diplotypes.To understand their disease associations, we compared the counts and frequencies of diplotypes between the case and control groups.Three diplotypes were more frequent in the group of cases.Especially the greater frequency of the C-G/C-G diplotype of rs3757930-rs74942016 frame (DT-2) was obvious in accordance with the positive association of the C-G haplotype with the disease which appeared in our haplotype analyses.Even though the case and control groups differed significantly, the C-G/C-G diplotype was seen in 30 % of the controls.To explain its presence in the controls, we separated the carriers from the rest of their groups and compared these sub-groups for the other two SNPs rs6988339 and rs80127039.While the C allele at rs80127039 was monomorphic, three possible genotypes (A/A, A/G and G/G) were all present at the rs6988339 locus in both subgroups (Carriers of C-G/C-G diplotype in the case and control groups).Interestingly, the G/G genotype was much more frequent in the subgroup of patients compared to the sub-group of controls (p=0.011).Controversially, the A/G genotype was more frequent in the subgroup of controls (p=0.031).The frequencies of A/A genotype did not show a significant difference between two subgroups.
In addition to the C-G/C-G diplotype, the A-C/G-C diplotype of rs6988339-rs3757930 frame and the A-C-G/G-C-G diplotype of rs6988339-rs3757930-rs74942016 frame were two other diplotypes which showed higher frequencies in the patients.On the other hand, the G-C-G/G-C-G diplotype of rs6988339-rs3757930-rs74942016 frame appeared to have a protective effect since it was more frequent in the controls compared to the patients.

Conclusions
In conclusion, we found one single SNP, one haplotype and one diplotype positively associated with schizophrenia in our sample.Except for these, we have observed negative associations of a list of variants at single SNP and haplotype levels.The CC genotype of rs3757930 was a risk factor positively associated with schizophrenia while the T allele of rs80127039 was protective allele and it was negatively associated.On the other hand, one haplotype (HT-5) and three diplotypes (DT-1, DT-2 and DT-3) showed positive associations with the disease.The genotype of rs6988339 also seemed effective in the carriers of diplotype DT-2.In the subgroup of DT-2 carriers, AG genotype of rs6988339 was more common in the patients while GG genotype was more common in the controls.Since we have used the case and control samples from a single population, the associations of these genetic factors (SNPs, haplotypes and diplotypes) needs to be confirmed with the replication studies which will be done in different populations.
The genetic factors that we have found associated with SZ might be the causative variants directly affecting the phenotype, or they might be seem associated because of their linkage with the other genetic factors which are the actual causative variants.This may be uncovered by analyzing the NRG1 gene and its interactions with the other genes more intensively with more comprehensive methods like genome sequencing or genome wide association studies.Additionally, the functional studies to investigate the mechanisms in which the SNP alleles and the haplotypes act may help to explain the molecular pathology behind the development of schizophrenia phenotype.

Table  :
Information about four screened SNPs.

Table  :
The distributions (and frequencies) of genotypes and alleles of four screened SNPs in the case and control groups.

Table  :
Comparison of case and control groups for recessive models of alternate alleles.Numbers of carriers of each allele are given as "Present" and non-carriers as "Absent".The frequencies are given in parentheses.

Table  :
Relations between the SNPs in pairs.

Table  :
Haplotypes and their frequencies in the case and control groups estimated by Haploview software.

Table  :
Diplotypes showed significant differences in frequencies between the case and control groups.

Table  :
Distributions of rs genotypes in the group of risk diplotype DT-.