Investigation of rare variants in LRP1, KPNA1, ALS2CL and ZNF480 genes in schizophrenia patients reflects genetic heterogeneity of the disease

Schizophrenia (SCZ) is a neurodevelopmental psychiatric disorder that is characterized by severely impaired cognitive processes causing hallucinations, delusions and altered emotional reactivity that disturb social behavior. This disorder is highly prevalent and according to the National Institute of Mental Health affects 1.1% of the U.S. adult population. The genetic factors predisposing to SCZ have not been fully elucidated but twin, adoptee and family studies jointly suggest that genetics is important with a heritability estimated to be up to 80% [1, 2]. Multiple approaches have been used to identify common and rare SCZ-predisposing variants using candidate gene and whole genome-scale studies [3, 4]. However neither large-scale sequencing projects, looking for rare penetrant variants, nor genome wide association studies, looking for common variants, have accounted for a significant fraction of the heritability of SCZ. In light of this limited success, it was hypothesized that deleterious de novo mutations in any of several different genes could explain the high global incidence of SCZ despite a reduced reproductive fitness. Our group first reported the presence of a significant excess of potentially deleterious de novo mutations in 401 synaptic genes using Sanger sequencing in a cohort of SCZ and autism patients [5]. We have later confirmed an excess of exonic de novo mutations and more particularly of nonsense variants in 15 SCZ trios (probands and parents) using exome sequencing. Interestingly, 4:11 ratio of nonsense to missense mutations is significantly higher than the expected 1:20 ratio (P = 0.005467), and according to the Human Gene Mutation Database, the expected ratio of nonsense to missense among all mutations reported to cause Mendelian diseases is 1:4 which in line with what we found (P > 0.05, not significant). This identified four candidate genes (LRP1, KPNA1, ALS2CL and ZNF480) possibly involved in SCZ [6]. The present report describes the mutation screening of these four genes in additional 475 SCZ cases and 189 controls.

Probands were individually interviewed and their diagnosis was based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. Exclusion criteria included patients with psychotic symptoms caused by alcohol, drug abuse or other clinical diagnosis. All samples were collected through informed consent following the approval of their respective institutional ethics review committees. An initial cohort of 189 SCZ patients of European Caucasian ancestry was used in this study. An additional cohort of 285 SCZ subjects, also of European Caucasian ancestry, was used for the screening of ALS2CL and ZNF480 genes. All coding regions and splice site junctions of LRP1, KPNA1, ALS2CL and ZNF480 genes were amplified and then sequenced using Sanger sequencing technology. Variant detection analysis was done using Mutation Surveyor (v. 3.23, Softgenetics) and rare exonic variants were confirmed by re-amplification of the relevant fragment in both probands and parents (when available) using forward and reverse oligos.

The first gene, LRP1 (NM002332, 89 exons) encodes for Homo Sapiens low density lipoprotein receptor-related protein 1 and is located at 12q13-14. We identified four rare LRP1 variants (S278I, R379H, M1795I and G3948D) absent from public databases (dbSNP [7], 1000 genome project [8] and Exome Variant Server (EVS) [9]) and two additional variants (G169D and A2160T) only reported in EVS at a very low frequency (Table 1). Interestingly, the missense G169D is predicted to affect protein function by SIFT (0.01) and Polyphen (2.337) predicting softwares [10, 11]. Unfortunately the parents’ DNAs were not available for an inheritance study. We also identified eight other LRP1 variants that were previously deposited in EVS database and their minor-allele frequencies ranged from 1 to 159 / 10 755 to 10 599 total alleles. Two of these, G3725E and R1993W, are predicted to affect protein function despite minor-allele frequencies of 12/ 10 746 and 25/ 10 733, respectively. However, the fact that G3725E is transmitted by an unaffected mother, suggests it is unlikely to play a major role in SCZ. Unlike the original published findings, we did not find any highly damaging variant (ie. nonsense or insertion/deletion) in LRP1 gene. The KPNA1 gene screening (NM002264; 14 exons), which encodes for karyopherin alpha 1 protein and is located at 3q21, did not lead to the identification of rare variants in our SCZ cohort. The sequencing of ALS2CL (NM147129; 26 exons), encoding for ALS2 C-terminal like protein and located at 3p21.31, revealed one nonsense mutation (E65X) in one SCZ patient. Because of this ALS2CL variant, we opted to screen the full gene in 286 additional SCZ patients and 189 control individuals. We observed a minor-allele frequency of 4/475 SCZ patients and of 1/189 controls for E65X. This variant has been lately added to dbSNP (rs139496961) and EVS databases with a minor allele-frequency of 20/10 738. The presence of E65X in a control individual from our ethnically-matched control cohort makes ALS2CL gene less likely to predispose to SCZ. We also observed three additional rare ALS2CL variants in our 189 initial cases (T268S, T460M and P580S) but their pathogenicity is unlikely since none of the prediction software predicted them to be deleterious. Finally, the screening of ZNF480 (NM144684; 4 coding exons) a zinc finger protein which is located at 19q13.41, led to the identification of two nonsense variants in one SCZ patient and one control individual (R276X and R500X, respectively). During our gene screening process, five nonsense variations (R304X, R360X, R416X, K434X and Q528X) were deposited in the EVS database by other groups with frequencies ranging from 1/10 755 to 2/10 756. Although we cannot rule out that schizophrenic or borderline personalities can be found in public databases such as EVS, the fact that five additional nonsense variations were found in the C-terminal region of this gene is not in favor of a deleterious effect of a de novo nonsense mutation in this region. For this reason we did not further investigate this gene in additional SCZ cases and control individuals and concluded that ZNF480 was a poor candidate for SCZ.

While the results of this mutation screening effort did not lead to the identification of additional potentially deleterious de novo mutations in LRP1, KPNA1, ALS2CL and ZNF480 in a larger cohort of SCZ patients, it was a necessary first step to assess their possible contribution to disease. One limitation of this study is that we have focused only on the coding regions. This does not exclude the involvement of disease predisposing variants in non-coding regions, which could affect allelic expression or splicing. Given the known heterogeneity of SCZ and the frequencies of variants reported thus far, the contribution of these particular genes may only emerge after the progressive sequencing of coding and non-coding regions of these genes in much larger cohorts of SCZ cases and control individuals. It’s likely that an additional mechanism (like the polygenic mode) is involved in the genetics of sporadic SCZ. However, we still believe that deleterious de novo mutations play an important role in a proportion of SCZ patients as demonstrated by our group and others [14‐16]. It is well known that SCZ is genetically heterogeneous and hundreds and probably thousands of genes are involved. Therefore, to date, very few cases with de novo mutation have been reported to be associated with the same genes. Addressing the complete genetic picture of a polygenic disease such as SCZ is still a major challenge and will require further independent replication studies to clarify the role of these genes.