Many evidences, such as the monozygotic twins have three times higher heritability rates than dizygotic twins, led to an established facts that consider the autism as one of the most heritable neurodevelopmental disorders. However, the molecular diagnostic yield is lower than expected due to high degree of heterogeneity and multifactorial etiology. The challenge is the involvement of hundreds of gene variants that affect several pathways, have highly variable risk impact and involved in other conditions. In addition to the interplay between genetic and epigenetic factors which make the task of identifying specific autism gene variants very difficult [11]. Single gene analysis has identified many SNPS that are associated with autism [12, 13]. Herein, we analyzed, for the first time, the involvement of PCSK9 in autism susceptibility.
However, the main role of PCSK9 is controlling blood cholesterol, it involved in regulating neuronal development and apoptosis. It mediates its proapoptotic function by modulation level of ApoER2 independent of N-methyl-D-aspartate (NMDA) receptor function by cooperation with extracellular-regulated kinase (ERK) and JNK signaling pathways to regulate c-Jun and caspase-3 activation [14].
Based on its role in regulating neuronal apoptosis, we suggested a possible involvement of PCSK9 in autism etiology. First, we have sequenced the 12 exons of PCSK9 gene and the flanking regions of intronic sequence to identify intronic variants which may be related to autism. This is because previous reports showed that pathogenic variants can locate deep within the introns in many gene which are associated with the diseases. These variants, which can occur more than 100 bp away from exon-intron junction, have the ability to cause non-canonical splice sites, modify splicing regulatory elements, or alter transcription regulatory motifs and non-coding RNA genes [15].
In our study, 43 variants were existed in studied subjects, 11 exonic, 25 intronic, 2 in 5 ̀ UTR, and 3 in 3 ̀ UTR. From them, 16 variants are rare and 27 are more recurrent. The rare variants were documented to be important in ASD and a significant proportion of ASDs are resulted from rare genetic variants. According to common disease/rare variant (CD/RV) model, a large number of rare, and even very rare, variants underlie the genetic disease [16]. Herein, out of the detected 16 rare variants, three pairs are existed together (rs371488778 and rs28362287, rs557622245 and rs28385701, rs28362263 and rs72646526) as one time frequency for each pair.
Two novel variants were identified, g.23809C > T in intron11 and g.24071T > G in 3 ̀ UTR. g.23809C > T is identified as deleterious by DANN and FunSeq2. g.24071T > G is classified as neutral by all used tools, however, considering that g.24071T > G is located in 3 ̀ UTR, we used bioinformatic tools to determine whether this variant locate in the recognition site of any microRNA. According to Target Scan, we found that this variant, which locate at the position 34 of 3‵ UTR of PCSK9, can alter the recognition element of miR-4725-5p and miR-4780.
Out of the more-recurrent 27 variants, seven variants (group 1) displayed differential frequency and distribution between cases and controls. All of these seven variants locate in non-coding regions. One variant (rs45448095) locate in 5‵ UTR and six variants in intronic regions; they include: rs11800231 (intron 3), rs529500286 (intron 4), rs45613943 (intron 5), rs483462 and rs45439391 (intron 10) and rs630431 (intron 11). According to Golovina et al., the majority of SNPs related to ASD are existed in the non-coding regions of the genome [4]. The role of intronic variants is documented in various reports; for example, a study used CRISPR/Cas9 methods to identify the role of ASD-associated CHD7 intronic variant in neural development found that the presence of this variant led to decreased level of CHD7 mRNA and developmental delay and maturity defects of neural cells [17].
Out of the differentially-existed seven variants (group1), five variants are displayed in higher frequency in cases than controls (rs45448095 in codominant and dominant model; rs45613943 and rs630431 dominant model; rs529500286 in which deletion is observed in 45.8% of cases compared to 17.9% of controls; and rs45439391in which the heterozygous genotype is exhibited in 12.3 of cases and absent in control subjects). On the other hand, in the remaining two variants (rs11800231 and rs483462) the minor allele is protective from autism susceptibility. For rs11800231, the minor (A) allele frequency is 0.11 in cases and 0.24 in controls, odd ratio (OR) = 2.5. while the minor allele (A) frequency of rs483462 is 0.7 in cases and 0.88 in controls, OR = 3.1. The global minor alle frequency (MAF) of rs11800231 and rs483462 is 0.083 and 0.326, respectively.
Worth mention, five of the differentially-existed seven variants are in concordance with Hardy-Weinberg equilibrium: including, rs45448095, rs11800231, rs483462, rs45439391 and rs630431. While rs529500286 and rs45613943 are deviated from it (P < 0.05), this deviation could be associated with the disease. Disease-associated deviation from Hardy-Weinberg equilibrium is previously reported in some SNPs, for example rs7249246 in CACNA1A gene [18], rs2285543 in WNT2 gene [19].
Using predictSNP2 which is platform for in silico analysis of SNPs by CADD, DANN, FATHMM, FunSeq2, and GWAVA, five of the differentially-existed seven variants are identified as deleterious by at least one tool (including: rs45448095, rs45613943, rs45439391, rs483462 and rs630431). One variant (rs11800231) is neutral by all predicted tool. The remaining indel. variant (rs529500286) is classified as benign by mutation taster.
According to ClinVAR; five of the differentially-existed seven variants are studied in familial hypercholesterolemia, four of them classified as benign (rs45448095, rs45439391, rs11800231 and rs483462) and one have uncertain significance (rs529500286 that merged to rs397735050). rs45613943 and rs630431 not reported in ClinVAR. Considering that this is the first time to correlate PCSK9 variants with autism, no studies reported in ClinVAR mentioned the pathogenic significance of these variants in autism. None of the non-synonymous variants in this study showed significance change in genotype distribution between cases and controls and none of them identified as deleterious whether by predictSNP2 or by other tools concerned with nonsynonymous variants; including: PolyPhen-2, Mutation assessor, PROVEN, REVEL and SIFT.
Additionally, none of the variants identified in the study affect the stability of PCSK9 mRNA or cause change in MFE. This result support our exhibited increase of PCSK9 concentration and suggest that the variants are gain of function not loss of function. This is because the decreased stability is a reason for mRNA degradation and, subsequently, decreased concentration of resulted protein, but in our study the protein concentration is significantly elevated in cases than control group. Previous studies analyzed circulating PCSK9 concentration in association with LDL cholesterol and documented positive correlation between them and implication of PCSK9 in susceptibility to coronary heart disease [20]. In patients with acute coronary syndrome and myocardial infarction, level of plasma PCSK9 is reported to be associated with the severity of coronary lesions [21, 22]. A study on patients with severe multiple trauma revealed that increased plasma PCSK9 is a late biomarker of severe trauma [23]. PCSK9 gain-of-function mutations are found to be associated with autosomal dominant hypercholesterolemia and premature atherosclerosis [24], while loss-of-function mutations are associated with a reduced rate cardiovascular disease [25]. In our study, the increased plasma PCSK9 concentration in autistic patients could be related to its pro-apoptotic function and its role in neuronal differentiation [7].
Epigenetic regulation, including DNA methylation, is implicated in autism [5]. Genome-wide analysis of DNA methylation for 50 pairs of monozygotic twins found numerous differentially methylated regions associated with ASD [26]. Another study analyzed over 485,000 CpG loci in post-mortem brain samples of ASD patients and has determined four differentially methylated regions, one in 3’ UTR and three in promoter regions of ASD-associated genes [27]. Herein, we analyzed the methylation density of PCSK9 promoter and we found a significant decrease in methylation percentage in autistic individuals compared to control subjects.
Collectively, the study concluded a possible involvement of PCSK9 in autism pathogenesis and recommend further studies with a greater number of subjects.