Epigenetic age acceleration (EAA) changes in ASD blood
DNAm Age was predicted using Horvath’s method, which employs 353 CpG sites (32). EAA was then estimated as the residual resulting from a multivariate model regressing the DNAm Age estimate on chronological age. Firstly, we correlated chronological age and DNAm Age in non-ASD (n = 59, combining non-ASD and non-ASD with 16.p11.2 del) and ASD (n = 72, combining ASD, ASD with 16p11.2 del, and ASD with CHD8 variants). The results show that DNAm Age was significantly correlated with chronological age in non-ASD (r = 0.72, p < 0.0001), but the correlation was lower in the ASD cohort (r = 0.49, p < 0.0001) (Figure 1A and 1B). The expected correlation found in the non-ASD group indicated that DNAm Age in our analyses can be determined EAA. The chronological age difference between the two cohorts was not significant (ASD: 9.94, non-ASD: 9.68, p = 0.723) (Figure 1C), while the DNAm Age of the ASD cohort was considerably decreased (ASD: -0.84, non-ASD: 0.84, p = 0.00078) (Figure 1D). Next, we classified non-ASD and ASD cohorts by ≤ 10 years old and ≥ 11 years old. As for previous comparisons, chronological age did not differ between ASD and non-ASD in both groups (Figure 1E). In the ASD cohort older than 11 years old, EAA shows a significant deceleration compared with an age-matched non-ASD cohort but this deceleration was not found in ASD younger than 11 years old (Figure 1F). These findings suggest that people with ASD were epigenetically younger than their chronological age which may be occurring in response to various environmental risk factors.
DNAm profile of LINE-1 and Alu elements in the heterogeneous ASD
A total of 22,352 probes mapping to LINE-1 and Alu elements were identified on the Infinium 450K platform for differential DNAm analysis. The analyses were performed for heterogeneous ASD, ASD with 16p11.2del, and ASD with CHD8 variants, versus non-ASD. Firstly, we measured the global methylation by combining all positions mapping to LINE-1 and Alu elements as the total of CpGs. In the comparison of heterogeneous ASD against non-ASD (Figure 2A), there was no significant difference in global methylation between these cohorts (Δβ = 0.003, p = 0.098). However, when we performed the methylation profile of REs by which LINE-1 and Alu positions were analyzed separately, we found that 2,802 (LINE-1) and 4,363 (Alu) DMPs were significantly differentially methylated (PFDR < 0.05) in the heterogeneous ASD compared to non-ASD (Figure 2B and 2C, Figure 3). All these loci included 2,471 hypomethylated loci (LINE-1: 1,437 loci, Alu: 1,304 loci; PFDR < 0.05) and 4,424 hypermethylated loci (LINE-1:1,365 loci, Alu: 3,059 loci; PFDR < 0.05). Due to the different activity of subfamilies of RE, LINE-1 and Alu elements were clustered by evolution age into three categories including old age (L1M, AluJ), intermediate age (L1P, L1PB, AluS), young age (L1HS, L1PA, AluY), and related (HAL1, FAM, FLAM, FRAM). The methylation of LINE-1 and Alu elements were changed in a subfamily-specific manner. We discovered that LINE-1 was considerably hypermethylated in young and intermediate age families, including L1H (∆β = 0.013, p = 0.00001) and L1P (∆β = 0.005, p = 0.027), but HAL1 was hypomethylated (∆β = -0.003, p = 0.03) (Figure 2D). Methylation of Alu elements was significantly hypermethylated in the old and intermediate age families: AluJ (∆β = 0.006, p = 0.004) and AluS (\(\varDelta\)β = 0.005, p = 0.016), respectively (Figure 2E). These findings indicated that methylation of LINE-1 and Alu elements in the heterogeneous ASD was altered in family- and locus- specific manner rather than globally.
LINE-1 and Alu methylation signatures in the homogeneous ASD (16p11.2del and CHD8 variants)
Due to the heterogeneity in the ASD population, we also intended to investigate the methylation profile of LINE-1 and Alu elements in genetically homogeneous ASD, as identified in the original article of GSE113967, including ASD individuals with 16p11.2del (n = 7) and CHD8 variant (n = 15). As for the results of ASD with 16p11.2del compared with non-ASD, we found no significant changes in the global methylation compared with non-ASD (∆β = -0.002, p = 0.771, Figure 4A). However, the analyses identified 70 significantly locus-specific DMPs of REs in ASD with 16p11.2del including 27 DMPs at LINE-1 (5 hypomethylated loci, 22 hypermethylated loci, Figure 4B) and 43 DMPs at Alu elements (23 hypomethylated loci, 20 hypermethylated loci, Figure 4C). When LINE-1 and Alu positions were categorized into families, there was no significant difference in methylation of LINE-1 and Alu elements by family (Figure 4D and 4E).
Subsequently, we analyzed data for ASD with CHD8 variants by using the same approach. We found that global methylation was exclusively hypomethylated in the ASD with CHD8 variants (∆β = -0.006, p = 0.042, Figure 5A). Analyzing by the position, the majority of DMPs at LINE-1 and Alu elements were hypomethylated (616 loci or 88.63%, PFDR<0.05) of the total identified 695 DMPs observed (Figure 5B and 5C). Among all significant DMPs in the ASD with CHD8 variants, 528 DMPs were mapped to LINE-1, while 167 DMPs were Alu elements. Moreover, changes in LINE-1 and Alu methylation regarding to their families were observed in ASD with CHD8 variants. In contrast to the differences found in the heterogeneous ASD, young and intermediate age LINE-1 families were significantly hypomethylated including L1H (∆β = -0.015, p = 0.0038) and L1P (∆β = -0.010, p = 0.0186) (Figure 5D). Hypomethylation of Alu elements was also observed in old age, intermediated age, and related families: AluJ (∆β = -0.010, p = 0.0154), AluS (∆β = -0.008, p = 0.0443), FAM (∆β = -0.008, p = 0.0399), FRAM (∆β = -0.007, p = 0.02551) respectively (Figure 5E). These findings suggest that DNAm signatures were a widespread reduction in LINE-1 and Alu regions which occurred at a specific family in the ASD with CHD8 variants but not in ASD with 16p11.2del.
Genomic distribution of LINE-1 and Alu methylation in heterogenous and homogenous ASD.
To determine the differential DNAm of LINE-1 and Alu elements by genomic features, we performed enrichment analysis using Fisher’s exact test. CpG positions at LINE-1 and Alu elements were categorized to 1500 and 200 within the transcriptional start site (TSS1500 and TSS200, respectively), the 5’ untranslated region (5’UTR), the first exon (1st exon), gene body (Body), and 3’ untranslated region (3’UTR). In the heterogeneous ASD signatures, CpG sites at LINE-1 were significantly enriched in TSS1500 (p = 0.0005) and Body (p < 0.0001) (see Additional file 2A). Whereas Alu elements were significantly enriched in TSS1500 (p < 0.0001), 5’UTR (p < 0.0001), Body (p < 0.0001), and 3’UTR (p = 0.0086) (see Additional file 3A). However, DNAm across all retrotransposons by genomic location did not significantly differ between non-ASD and heterogenous ASD (see Additional file 2B and 3B). DNAm signatures of the ASD with 16p11.2del and CHD8 variants were significantly enriched in Body (p = 0.04) and TSS1500 (p < 0.0001) respectively (see Additional file 4: Figure S5B, S5C). This result shows that the changes of probes mapping to TSS1500 and gene bodies are more likely to have a functional impact on gene expression in ASD in both heterogenous and homogenous ASD.
Biological functions and pathways of LINE-1 and Alu methylation signatures in ASD and ASD variants.
To determine the biological significance of LINE-1 and Alu methylation signatures identified in each ASD cohort, we predicted the biological function and pathway of genes located nearby DMPs of LINE-1 and Alu elements using IPA software. We found that neurological diseases were significantly enriched among genes associated with LINE-1 and Alu methylation signatures in the heterogeneous ASD (p range: 0.00495 - 3.33E-26, 2274 genes) and ASD with CHD8 variants (p range: 0.0258 – 0.000117, 302 genes) as shown in Additional file 5 and 6. The categories ASD and intellectual disability were exclusively associated with LINE-1 and Alu methylation signatures in the heterogeneous ASD (p = 2.56E-06, 253 genes). Whereas Huntington's disease, familial encephalopathy, and brain lesion were commonly associated with both ASD signatures. For ASD with 16p11.2del variant, LINE-1 and Alu methylation signatures in this cohort were significantly associated with developmental disorders (p range: 0.0393 – 0.00222, 9 genes) (see Additional file 7). However, only one gene was associated with the disease, possibly caused by a small number of genes associated with LINE-1 and Alu methylation of this ASD variant. Additionally, we discovered that several canonical pathways linked to ASD were associated with genes located nearby LINE-1 and Alu methylation signatures in each ASD cohort. More precisely, we found that the α-adrenergic signaling pathway was significantly associated in the heterogeneous ASD (p = 0.00269, 28 genes) and ASD with CHD8 variants (p = 0.00646, 7 genes). Axonal guidance signaling pathway involved in nervous system development was significantly associated with LINE-1 and Alu methylation signatures of ASD with 16p11.2del and CHD8 variants. These results indicate that genes associated with LINE-1 and Alu methylation signatures in ASD were involved with neurological diseases and ASD-comorbid disorders as well as canonical pathways known to be implicated in ASD. The list of all significant biological functions and pathways in each ASD variant is shown in Additional file 5-7.
Interactome networks or gene regulatory networks revealed the interaction of genes located nearby LINE-1 and Alu methylation signatures of each ASD variant. The functions and pathways implicated in ASD were highlighted in the networks. The interactome of the heterogeneous ASD was associated with ASD and mental retardation, as well as canonical pathways implicated in ASD such as retinoic acid receptor (RAR) and AMP-activated protein kinase (AMPK) signaling (Figure 6). In ASD with 16p11,2del, we found that the interactome related to axonal guidance and sirtuin signaling pathway (see Additional file 8: Figure S8A). The interactome of ASD with CHD8 was related to familial encephalopathy and movement disorder which conditions found in ASD individuals (43, 44). The interactomes were also associated with neuronal function including axonal guidance and synaptogenesis signaling pathways (see Additional file 8: Figure S8B and S8C).
Identification of unique target loci located nearby LINE-1 and Alu signatures in heterogeneous ASD
To investigate the functional impact of locus-specific LINE-1 and Alu methylation to target gene or neighboring gene expression in the ASD, we identified DEGs from multiple ASD studies obtained from the GEO DataSets. This approach reflected the heterogeneity of the ASD population because these studies were compiled from a different ASD cohort. There were 12,419 DEGs identified from seven datasets including four studies (one study used peripheral blood samples and three studies used post-mortem brain tissues from ASD individuals) (see Additional file 1). We subsequently overlapped the list of DEGs with differentially methylated genes (DMGs: genes located nearby LINE-1 and Alu signatures). The overlapping revealed 1,847 DMGs in the heterogeneous ASD that were differentially expressed in several ASD studies, with 155 of them being autism-related genes in the SFARI database. We identified 43 top DMGs, |Δβ| ≥ 5%, inversely related to gene expression, and differentially expressed in at least two studies (Table 2). Interestingly, two of the top DMGs, potassium voltage-gated channel subfamily Q member 3 (KCNQ3) and ubiquitin conjugating enzyme E2 H (UBE2H), were genes in the SFARI database and were enriched in the gene regulatory network related to ASD and mental retardation (Figure 6).
The genomic regions of LINE-1 and Alu methylation signatures within the DMGs are shown in Figure 7. We identified DMRs by mapping all probes located nearby LINE-1 and Alu signatures using the UCSC genome browser. The findings revealed that AluSg7 (cg16926147), which is located on the gene body of the KCNQ3 gene (Figure 7A), was hypermethylated and KCNQ3 expression level was significantly reduced in blood and post-mortem brain tissues. Interestingly, we discovered that several probes in this region, including those in the promoter region were not changed. This result suggests that LINE-1 and Alu methylation at DMRs may facilitate gene expression indicated by the inverse relationship between LINE-1/Alu methylation and gene expression. As well as AluY (cg08998414) within UBE2H gene (Figure 7B) and L1PA3 (cg24094412) within hyperpolarization activated cyclic nucleotide gated potassium channel 1 (HCN1) (Figure 7C), we also observed that AluY and L1PA3 methylation were inversely related to the gene expression levels in both blood and brain tissues of ASD cohort. Moreover, we found several DMGs that were not reported in the SFARI database but the expression of these DMGs in the blood and post-mortem brain tissues was inversely related to LINE-1 and Alu methylation such as N-deacetylase and N-sulfotransferase 1 (NDST1) (cg12611243: L1MC1), ubiquitin specific peptidase 6 (USP6) (cg23416909: L1M5), and formin binding protein 1 (FNBP1) (cg13916261: AluSg). These results suggest that DMPs at LINE-1 and Alu elements may affect the expression of genes located nearby these DMPs in the heterogeneous ASD cohort.
Table 2
Utilized gene expression microarrays/RNA-sequencing for the differential gene expression analysis of the target genes of heterogeneous ASD.
Methylome data
|
Transcriptome data
|
Probe ID
|
Elements
|
Delta
|
PFDR
|
GSE
|
Gene ID
|
Gene
|
log2FC
|
q-value
|
cg02571470
|
L1MC5
|
0.061
|
1.61E-05
|
GSE18123
|
1563708_at
|
SFXN5
|
-0.348
|
0.0164
|
cg02571470
|
L1MC5
|
0.061
|
1.61E-05
|
GSE18123
|
241999_at
|
SFXN5
|
-0.407
|
0.0386
|
cg21314304
|
AluJb
|
0.054
|
2.68E-05
|
GSE25507
|
226298_at
|
RUNDC1
|
-0.141
|
0.0272
|
cg21314304
|
AluJb
|
0.054
|
2.68E-05
|
GSE28521_FC
|
ILMN_1733875
|
RUNDC1
|
-0.234
|
0.0319
|
cg21314304
|
AluJb
|
0.054
|
2.68E-05
|
GSE59288
|
146923
|
RUNDC1
|
-0.452
|
0.0001
|
cg04668642
|
AluSx
|
0.051
|
2.68E-05
|
GSE64018
|
ENSG00000166780
|
C16orf45
|
-0.224
|
0.0421
|
cg04668642
|
AluSx
|
0.051
|
2.68E-05
|
GSE28521_FC
|
ILMN_1687821
|
C16orf45
|
-0.256
|
0.0287
|
cg02616069
|
AluJb
|
0.069
|
3.02E-05
|
GSE18123
|
215584_at
|
HECW1
|
-0.259
|
0.0068
|
cg02616069
|
AluJb
|
0.069
|
3.02E-05
|
GSE59288
|
23072
|
HECW1
|
-0.805
|
<0.0001
|
cg15531814
|
AluJo
|
0.059
|
4.42E-05
|
GSE89594
|
A_23_P310257
|
KLK2
|
-0.161
|
0.0176
|
cg15531814
|
AluJo
|
0.059
|
4.42E-05
|
GSE18123
|
1555545_at
|
KLK2
|
-0.138
|
0.0254
|
cg11204311
|
AluJr
|
0.060
|
4.54E-05
|
GSE28521_FC
|
ILMN_1679796
|
TOMM20
|
-0.506
|
0.0110
|
cg11204311
|
AluJr
|
0.060
|
4.54E-05
|
GSE59288
|
9804
|
TOMM20
|
-0.176
|
0.0438
|
cg11204311
|
AluJr
|
0.060
|
4.54E-05
|
GSE64018
|
ENSG00000173726
|
TOMM20
|
-0.231
|
0.0459
|
cg04027778
|
AluSc
|
0.054
|
4.66E-05
|
GSE28521_TC
|
ILMN_1712705
|
RAB40C
|
-0.258
|
0.0313
|
cg04027778
|
AluSc
|
0.054
|
4.66E-05
|
GSE18123
|
227269_s_at
|
RAB40C
|
-0.429
|
0.0133
|
cg04027778
|
AluSc
|
0.054
|
4.66E-05
|
GSE18123
|
1569396_at
|
RAB40C
|
-0.309
|
0.0025
|
cg04027778
|
AluSc
|
0.054
|
4.66E-05
|
GSE59288
|
57799
|
RAB40C
|
-0.322
|
0.0131
|
cg06719602
|
AluSx1
|
0.050
|
5.57E-05
|
GSE28521_FC
|
ILMN_1706238
|
CSE1L
|
-0.202
|
0.0384
|
cg06719602
|
AluSx1
|
0.050
|
5.57E-05
|
GSE28521_FC
|
ILMN_1665797
|
CSE1L
|
-0.184
|
0.0449
|
cg23935361
|
AluJb
|
0.066
|
5.82E-05
|
GSE42133
|
ILMN_1738093
|
RNFT2
|
-0.102
|
0.0090
|
cg23935361
|
AluJb
|
0.066
|
5.82E-05
|
GSE59288
|
84900
|
RNFT2
|
-0.259
|
0.0467
|
cg17429234
|
AluSp
|
0.061
|
6.88E-05
|
GSE59288
|
1780
|
DYNC1I1
|
-0.852
|
<0.0001
|
cg17429234
|
AluSp
|
0.061
|
6.88E-05
|
GSE28521_FC
|
ILMN_1690397
|
DYNC1I1
|
-0.516
|
0.0081
|
cg17429234
|
AluSp
|
0.061
|
6.88E-05
|
GSE64018
|
ENSG00000158560
|
DYNC1I1
|
-0.324
|
0.0410
|
cg23376467
|
AluSq
|
0.077
|
7.13E-05
|
GSE18123
|
1556907_at
|
ZNF474
|
-0.292
|
0.0346
|
cg23376467
|
AluSq
|
0.077
|
7.13E-05
|
GSE89594
|
A_33_P3360565
|
ZNF474
|
-0.205
|
0.0200
|
cg06471678
|
AluYc
|
-0.065
|
7.85E-05
|
GSE25507
|
233694_at
|
HSPA1L
|
0.111
|
0.0307
|
cg06471678
|
AluYc
|
-0.065
|
7.85E-05
|
GSE42133
|
ILMN_1654566
|
HSPA1L
|
0.107
|
0.0300
|
cg15074424
|
AluSx1
|
0.083
|
1.01E-04
|
GSE28521_FC
|
ILMN_2212354
|
WDR46
|
-0.201
|
0.0364
|
cg15074424
|
AluSx1
|
0.083
|
1.01E-04
|
GSE28521_TC
|
ILMN_2212354
|
WDR46
|
-0.303
|
0.0335
|
cg02747612
|
AluSq
|
-0.068
|
1.02E-04
|
GSE42133
|
ILMN_1802053
|
ZNF91
|
0.162
|
0.0158
|
cg02747612
|
AluSq
|
-0.068
|
1.02E-04
|
GSE18123
|
236128_at
|
ZNF91
|
0.306
|
0.0474
|
cg13916261
|
AluSg
|
-0.100
|
1.05E-04
|
GSE59288
|
23048
|
FNBP1
|
0.354
|
0.0128
|
cg13916261
|
AluSg
|
-0.100
|
1.05E-04
|
GSE42133
|
ILMN_1797342
|
FNBP1
|
0.156
|
<0.0001
|
cg02827046
|
L1PA16
|
0.099
|
1.17E-04
|
GSE89594
|
A_33_P3420900
|
PATE2
|
-0.221
|
0.0142
|
cg02827046
|
L1PA16
|
0.099
|
1.17E-04
|
GSE42133
|
ILMN_2133784
|
PATE2
|
-0.157
|
0.0236
|
cg07930329
|
AluSx3
|
0.056
|
1.21E-04
|
GSE59288
|
10055
|
SAE1
|
-0.295
|
0.0064
|
cg07930329
|
AluSx3
|
0.056
|
1.21E-04
|
GSE28521_TC
|
ILMN_1657204
|
SAE1
|
-0.194
|
0.0417
|
cg07930329
|
AluSx3
|
0.056
|
1.21E-04
|
GSE64018
|
ENSG00000142230
|
SAE1
|
-0.194
|
0.0395
|
cg16926147
|
AluSg7
|
0.062
|
1.31E-04
|
GSE18123
|
206573_at
|
KCNQ3
|
-0.093
|
0.0182
|
cg16926147
|
AluSg7
|
0.062
|
1.31E-04
|
GSE59288
|
3786
|
KCNQ3
|
-0.634
|
0.0003
|
cg08499057
|
AluJo
|
0.057
|
1.63E-04
|
GSE28521_FC
|
ILMN_1781999
|
ABCF2
|
-0.164
|
0.0426
|
cg08499057
|
AluJo
|
0.057
|
1.63E-04
|
GSE28521_FC
|
ILMN_1669201
|
ABCF2
|
-0.168
|
0.0426
|
cg08499057
|
AluJo
|
0.057
|
1.63E-04
|
GSE28521_TC
|
ILMN_1669201
|
ABCF2
|
-0.202
|
0.0372
|
cg08499057
|
AluJo
|
0.057
|
1.63E-04
|
GSE18123
|
207623_at
|
ABCF2
|
-0.321
|
0.0053
|
cg08998414
|
AluY
|
-0.106
|
1.85E-04
|
GSE64018
|
ENSG00000186591
|
UBE2H
|
0.130
|
0.0352
|
cg08998414
|
AluY
|
-0.106
|
1.85E-04
|
GSE42133
|
ILMN_1757644
|
UBE2H
|
0.126
|
0.0258
|
cg08998414
|
AluY
|
-0.106
|
1.85E-04
|
GSE25507
|
222419_x_at
|
UBE2H
|
0.171
|
0.0230
|
Identification of unique target loci located nearby LINE-1 and Alu signatures in ASD variants.
To investigate the effects of unique LINE-1 and Alu methylation signatures to target gene or neighboring gene expression in the genetically homogeneous ASD, we obtained 39 and 101 DMPs that were found exclusively in the ASD with 16.p11.2 del and CHD8 variants, respectively (see Additional file 4: Figure S4A). We re-analyzed them for ASD variant versus the heterogenous ASD. Next, we conducted the same strategy used for the heterogeneous ASD to select the candidate DMPs by overlapping with the transcriptome data. The overlapping of unique DMPs with transcriptome data revealed 11 and 31 unique DMGs in the ASD with 16.p11.2 del and CHD8 variants, respectively (Table 3 and 4). Among the unique DMGs, we found several genes linked to neurodevelopmental disorder and ASD, including XK related 6 (XKR6) (Figure 8A), zinc finger protein 107 (ZNF107) (Figure 8B), and myeloma-overexpressed gene 2 protein (MYEOV2) (Figure 8C) in the ASD with 16.p11.2 del. The significant DMPs at AluY (cg21300361) within XKR6 was hypermethylated, while as AluSq (cg01772945) within ZNF107 and L1MB3 (cg13749477) within MYEOV2 were hypomethylated. Interestingly, these genes were differentially expressed in the blood transcriptome of multiple ASD cohorts, and their expression was inversely relative to LINE-1 and Alu methylation levels.
For ASD with CHD8 variants, we found that all LINE-1 and Alu elements located on candidate genes were markedly hypomethylated, as expected from global and family-specific methylation levels. These DMPs consist of L1MC5 (cg22706070) within Euchromatic Histone Lysine Methyltransferase 2 (EHMT2) (Figure 9A), AluJo (cg06421197) within caspase 1 (CASP1) (Figure 9B), and AluSx (cg18699242, cg01963623, cg02169692) within ubiquitin-specific peptidase 18 (USP18) (Figure 9C). EHMT2 was significantly increased in the blood of ASD, while CASP1 was increased in both the blood and brain of multiple ASD cohorts (one probe was decreased). These changes were inversely relative to LINE-1 and Alu methylation levels within that gene. We found that the expression of USP18 was not inversely relative to AluSx methylation located on the gene. Additionally, the DMRs of XKR6, ZNF107, MYEOV2, EHMT2, and CASP1 genes revealed LINE-1 and Alu probes as well as non-LINE-1/Alu probes located in the same DMRs (Figure 8A-C and Figure 9A-B).
Table 3
Utilized gene expression microarrays/RNA-sequencing for the differential gene expression analysis of the target genes of ASD with 16p11.2 deletion.
Methylome data
|
Transcriptome data
|
Probe ID
|
Elements
|
Delta
|
PFDR
|
GSE
|
Gene ID
|
Gene
|
log2FC
|
q-value
|
cg22062537
|
L1MB3
|
-0.079
|
0.0002
|
GSE25507
|
1553515_at
|
MYEOV2
|
0.107
|
0.0496
|
cg13749477
|
L1MB3
|
-0.038
|
0.0004
|
GSE25507
|
1553515_at
|
MYEOV2
|
0.107
|
0.0496
|
cg07628769
|
L1MB3
|
-0.101
|
0.0028
|
GSE25507
|
1553515_at
|
MYEOV2
|
0.107
|
0.0496
|
cg01772945
|
AluSq
|
-0.048
|
0.0003
|
GSE18123
|
205739_x_at
|
ZNF107
|
0.382
|
0.0017
|
cg09168728
|
HAL1
|
0.042
|
0.0039
|
GSE18123
|
202651_at
|
LPGAT1
|
0.160
|
0.0038
|
cg21300361
|
AluY
|
0.024
|
0.0139
|
GSE18123
|
1553640_at
|
XKR6
|
-0.949
|
0.0206
|
cg21300361
|
AluY
|
0.024
|
0.0139
|
GSE25507
|
1553640_at
|
XKR6
|
0.134
|
0.0112
|
cg01270736
|
AluJb
|
-0.037
|
0.0167
|
GSE25507
|
207289_at
|
MMP25
|
0.193
|
<0.0001
|
cg26620682
|
L1PA2
|
-0.058
|
0.0206
|
GSE28521_FC
|
ILMN_1765641
|
SEMA3A
|
-0.200
|
0.0343
|
cg26620682
|
L1PA2
|
-0.058
|
0.0206
|
GSE59288
|
10371
|
SEMA3A
|
-0.590
|
0.0139
|
cg05073382
|
L1MA7
|
0.153
|
0.0290
|
GSE89594
|
A_23_P258912
|
MYOM2
|
-1.611
|
0.0114
|
cg11438448
|
AluSx4
|
-0.014
|
0.0364
|
GSE18123
|
233429_at
|
SPEF2
|
-0.285
|
0.0484
|
cg10059324
|
L1MC4a
|
0.050
|
0.0368
|
GSE59288
|
8863
|
PER3
|
0.296
|
0.0269
|
cg11859345
|
AluSp
|
-0.031
|
0.0474
|
GSE18123
|
228766_at
|
CD36
|
0.171
|
0.0249
|
cg11859345
|
AluSp
|
-0.031
|
0.0474
|
GSE18123
|
206488_s_at
|
CD36
|
0.273
|
0.0076
|
cg11859345
|
AluSp
|
-0.031
|
0.0474
|
GSE25507
|
209554_at
|
CD36
|
0.099
|
0.0131
|
cg25283432
|
L1MA8
|
0.004
|
0.0482
|
GSE25507
|
232421_at
|
SCARB1
|
0.116
|
0.0189
|
Table 4
Utilized gene expression microarrays/RNA-sequencing for the differential gene expression analysis of the target genes of ASD with CHD8 variants.
Methylome data
|
Transcriptome data
|
Probe ID
|
Elements
|
Delta
|
PFDR
|
GSE
|
Gene ID
|
Gene
|
log2FC
|
q-value
|
cg10071848
|
AluJb
|
-0.033
|
0.0439
|
GSE59288
|
5826
|
ABCD4
|
0.39
|
0.0078
|
cg07730946
|
AluJb
|
-0.027
|
0.0128
|
GSE42133
|
ILMN_1684585
|
ACSL1
|
-0.24
|
0.0344
|
cg12934569
|
AluSx
|
-0.043
|
0.0319
|
GSE42133
|
ILMN_1684585
|
ACSL1
|
-0.24
|
0.0344
|
cg11111835
|
L1MA3
|
-0.020
|
0.0359
|
GSE18123
|
1553603_s_at
|
ATL2
|
0.28
|
0.0248
|
cg11111835
|
L1MA3
|
-0.020
|
0.0359
|
GSE18123
|
237968_at
|
ATL2
|
-0.25
|
0.0492
|
cg04154142
|
AluSx1
|
-0.063
|
0.0110
|
GSE42133
|
ILMN_1672596
|
BCAR1
|
-0.09
|
0.0127
|
cg04154142
|
AluSx1
|
-0.063
|
0.0110
|
GSE18123
|
223116_at
|
BCAR1
|
-0.26
|
0.0226
|
cg14115346
|
L1MB3
|
-0.038
|
0.0201
|
GSE18123
|
205750_at
|
BPHL
|
0.28
|
0.0097
|
cg10211626
|
AluSx1
|
-0.059
|
0.0028
|
GSE59288
|
79640
|
C22orf46
|
0.56
|
0.0002
|
cg17129519
|
AluSg
|
0.097
|
0.0360
|
GSE64018
|
ENSG00000204564
|
C6orf136
|
-0.18
|
0.0430
|
cg17129519
|
AluSg
|
0.097
|
0.0360
|
GSE28521_TC
|
ILMN_1813236
|
C6orf136
|
-0.30
|
<0.0001
|
cg06421197
|
AluJo
|
-0.034
|
0.0266
|
GSE18123
|
206011_at
|
CASP1
|
0.14
|
0.0122
|
cg06421197
|
AluJo
|
-0.034
|
0.0266
|
GSE42133
|
ILMN_2326509
|
CASP1
|
-0.16
|
0.0258
|
cg06421197
|
AluJo
|
-0.034
|
0.0266
|
GSE42133
|
ILMN_2326512
|
CASP1
|
-0.17
|
0.0213
|
cg06421197
|
AluJo
|
-0.034
|
0.0266
|
GSE42133
|
ILMN_1727762
|
CASP1
|
-0.25
|
<0.0001
|
cg06421197
|
AluJo
|
-0.034
|
0.0266
|
GSE59288
|
834
|
CASP1
|
0.61
|
0.0096
|
cg13606720
|
L1ME3F
|
-0.044
|
0.0125
|
GSE42133
|
ILMN_3248676
|
CBWD3
|
-0.11
|
0.0180
|
cg11379605
|
AluJb
|
-0.033
|
0.0359
|
GSE59288
|
965
|
CD58
|
0.47
|
0.0159
|
cg11379605
|
AluJb
|
-0.033
|
0.0359
|
GSE18123
|
222061_at
|
CD58
|
0.36
|
0.0256
|
cg11379605
|
AluJb
|
-0.033
|
0.0359
|
GSE64018
|
ENSG00000116815
|
CD58
|
0.50
|
0.0328
|
cg09425611
|
L1MB7
|
-0.028
|
0.0360
|
GSE18123
|
209616_s_at
|
CES1
|
0.44
|
0.0025
|
cg00053536
|
AluSx
|
-0.028
|
0.0462
|
GSE25507
|
206274_s_at
|
CROCC
|
0.12
|
0.0458
|
cg09604414
|
AluSx
|
-0.067
|
0.0336
|
GSE25507
|
229079_at
|
EHMT2
|
0.09
|
0.0471
|
cg22706070
|
L1MC5
|
-0.008
|
0.0496
|
GSE25507
|
229079_at
|
EHMT2
|
0.09
|
0.0471
|
cg14570121
|
FLAM_A
|
-0.031
|
0.0375
|
GSE59288
|
2068
|
ERCC2
|
-0.26
|
0.0149
|
cg24803614
|
AluSc8
|
0.051
|
0.0341
|
GSE59288
|
2495
|
FTH1
|
0.42
|
0.0045
|
cg03085932
|
L1PA6
|
-0.043
|
0.0459
|
GSE25507
|
220249_at
|
HYAL4
|
0.12
|
0.0179
|
cg08005007
|
L1HS
|
-0.031
|
0.0095
|
GSE18123
|
235111_at
|
LSAMP
|
-0.23
|
0.0386
|
cg08005007
|
L1HS
|
-0.031
|
0.0095
|
GSE18123
|
229244_at
|
LSAMP
|
-0.25
|
0.0303
|
cg22993878
|
L1MB3
|
-0.021
|
0.0080
|
GSE25507
|
1553515_at
|
MYEOV2
|
0.11
|
0.0496
|
cg03514928
|
L1PA17
|
-0.022
|
0.0470
|
GSE59288
|
4897
|
NRCAM
|
-0.31
|
0.0104
|
cg20723844
|
AluY
|
-0.061
|
0.0163
|
GSE42133
|
ILMN_1789616
|
NUPL2
|
-0.09
|
0.0221
|
cg00984715
|
AluSq
|
-0.049
|
0.0224
|
GSE59288
|
5465
|
PPARA
|
0.51
|
0.0149
|
cg00984715
|
AluSq
|
-0.049
|
0.0224
|
GSE18123
|
1560981_a_at
|
PPARA
|
-0.24
|
0.0256
|
cg00984715
|
AluSq
|
-0.049
|
0.0224
|
GSE18123
|
1558631_at
|
PPARA
|
-0.46
|
0.0091
|
cg24351819
|
AluSp
|
-0.026
|
0.0474
|
GSE59288
|
55170
|
PRMT6
|
-0.33
|
0.0057
|
cg09521141
|
AluSg7
|
-0.045
|
0.0263
|
GSE28521_FC
|
ILMN_1765641
|
SEMA3A
|
-0.20
|
0.0343
|
cg09521141
|
AluSg7
|
-0.045
|
0.0263
|
GSE59288
|
10371
|
SEMA3A
|
-0.59
|
0.0139
|
cg26620682
|
L1PA2
|
-0.041
|
0.0346
|
GSE28521_FC
|
ILMN_1765641
|
SEMA3A
|
-0.20
|
0.0343
|
cg26620682
|
L1PA2
|
-0.041
|
0.0346
|
GSE59288
|
10371
|
SEMA3A
|
-0.59
|
0.0139
|
cg18221988
|
L1PA2
|
-0.043
|
0.0499
|
GSE18123
|
210804_x_at
|
SLC8A1
|
-0.30
|
0.0392
|
cg18221988
|
L1PA2
|
-0.043
|
0.0499
|
GSE25507
|
1565306_a_at
|
SLC8A1
|
0.18
|
0.0189
|
cg18221988
|
L1PA2
|
-0.043
|
0.0499
|
GSE59288
|
6546
|
SLC8A1
|
-0.41
|
0.0245
|
cg18221988
|
L1PA2
|
-0.043
|
0.0499
|
GSE18123
|
235518_at
|
SLC8A1
|
0.29
|
0.0057
|
cg18221988
|
L1PA2
|
-0.043
|
0.0499
|
GSE64018
|
ENSG00000183023
|
SLC8A1
|
-0.26
|
0.0430
|
cg07020453
|
L1MDa
|
-0.009
|
0.0457
|
GSE18123
|
217968_at
|
TSSC1
|
-0.21
|
0.0185
|
cg08428949
|
AluSz
|
-0.028
|
0.0124
|
GSE64018
|
ENSG00000128881
|
TTBK2
|
-0.15
|
0.0352
|
cg01810763
|
AluSx1
|
-0.053
|
0.0442
|
GSE64018
|
ENSG00000198431
|
TXNRD1
|
0.30
|
0.0443
|
cg11191744
|
AluY
|
-0.050
|
0.0457
|
GSE64018
|
ENSG00000198431
|
TXNRD1
|
0.30
|
0.0443
|
cg02169692
|
AluSx
|
-0.109
|
0.0275
|
GSE42133
|
ILMN_3240420
|
USP18
|
-0.31
|
0.0250
|
cg02169692
|
AluSx
|
-0.109
|
0.0275
|
GSE42133
|
ILMN_3240420
|
USP18
|
-0.31
|
0.0250
|
cg18699242
|
AluSx
|
-0.106
|
0.0448
|
GSE42133
|
ILMN_3240420
|
USP18
|
-0.31
|
0.0250
|
cg18699242
|
AluSx
|
-0.106
|
0.0448
|
GSE42133
|
ILMN_3240420
|
USP18
|
-0.31
|
0.0250
|
cg15726387
|
AluSz6
|
-0.044
|
0.0950
|
GSE18123
|
220079_s_at
|
USP48
|
0.10
|
0.0053
|
cg15726387
|
AluSz6
|
-0.044
|
0.0950
|
GSE18123
|
225925_s_at
|
USP48
|
0.15
|
<0.0001
|
cg06385000
|
AluSc
|
-0.041
|
0.0216
|
GSE18123
|
227434_at
|
WBSCR17
|
-0.46
|
0.0392
|
cg06385000
|
AluSc
|
-0.041
|
0.0216
|
GSE28521_FC
|
ILMN_1701557
|
WBSCR17
|
-0.19
|
0.0402
|
cg06385000
|
AluSc
|
-0.041
|
0.0216
|
GSE28521_TC
|
ILMN_1701557
|
WBSCR17
|
-0.23
|
0.0485
|
cg03870862
|
L1MB3
|
-0.063
|
0.0318
|
GSE59288
|
23144
|
ZC3H3
|
0.27
|
0.0193
|
Sensitivity and specificity of unique LINE-1 and Alu signatures in ASD variants.
To predict diagnosis of the genetically homogenous ASD by using LINE-1 and Alu methylation signatures, we subsequently conducted ROC curves analysis of these loci and other probes within unique DMRs to distinguish each homogenous ASD variant from non-ASD and ASD with non-specific variants. For ASD with 16.p11.2 del, AluY within XKR6 (cg21300361) exhibited high sensitivity and specificity (AUC = 0.905, 95%CI = 0.83-0.98) to distinguish ASD with 16.p11.2 del from non-ASD and ASD with CHD8 variants as shown in the ROC curves (Figure 8D). In addition, the ROC curves of AluSq within ZNF107 (cg01772945) and L1MB3 within MYEOV2 (cg13749477) also exhibited high AUC value (AluSq: AUC = 0.900, 95%CI = 0.83-0.97 and L1MB3: AUC = 0.841, 95%CI = 0.74-0.95) (Figure 8E and 8F). In the ASD with CHD8 variants, LINE-1 and Alu methylation signatures within candidate DMGs showed moderate sensitivity and specificity as demonstrated by AUC values (AUC range: 0.712-0.819) compared with the specificity of unique loci in ASD with 16.p11.2 del, including L1MC5 (cg22706070) within EHMT2 (Figure 9D), AluJo (cg06421197) within CASP1 (Figure 9E), and AluSx (cg18699242, cg01963623, cg02169692) within USP18 (Figure 9F). Our findings suggest that these novel DMPs at the LINE-1 and Alu elements could be used for clinical purposes. However, an independent cohort is required for validation, as we were limited by the percentage of ASD individuals affected by these genetic variants.