Paternal males with children affected by autism (case) or without (control) were recruited, and paternal sperm samples were collected at the Andrology Laboratory of IVIRMA Clinic in Valencia, Spain. The sperm sample was collected upon enrollment. Thirty-six patients were enrolled, which included thirteen in the control group, thirteen in the autism case group, and eight or ten for the blinded test groups. The differences (mean ± SD) between the semen analysis for both control and case group are shown in Table 1. Observations from the groups showed no significant difference in age, fathers age at pregnancy, fathers age upon sperm collection, sperm volume, concentration, or sperm concentration between the groups. Progressive sperm motility was greater in the autism case group, with no difference in non-progressive sperm motility, as shown in Table 1a. The motile percentage was higher in the control group, and no difference was observed in the total motile sperm count. One of the control samples, IVI 14, had a very high sperm count of 396.62 million that was outside two standard deviations of the mean (2 ± SD), so the analysis was redone without this sample. When the IVI 14 sample was not used in the analysis, the total sperm number was increased in the autism case group (p < 0.02), and the total motile sperm count (time) was increased in the autism case group (p < 0.017), as well as the progressive spermatozoa (%) (p < 0.019) and immotile % (p < 0.019) parameters. In addition to the case and control male age and sperm analysis parameters, Table 1a, all the blinded test set males, Table 1b, c, age and sperm parameters were analyzed and found to also be within the mean ± SD of the case and control samples presented (Additional file 2: Figure S1). Therefore, the blinded test set of individuals was appropriate comparisons with the same clinical parameters.
The participant demographics and clinical information were similar between the case and control population participants. The ethnicity of all the fathers was Caucasian. No major comorbidities were observed within either the control or case populations. The date of the patient sperm collection, age of the father upon collection and age of the father at conception of child are all not statistically different and provided in Table 1. Although age can impact sperm DNA methylation, the mean age upon sperm collection, which required a 3-year collection period, for the case and control was not statistically different, as given in Table 1. In addition, no statistical difference was observed in the age of the fathers at conception of child, as given in Table 1. All the autistic children were males. Since the focus was on paternal sperm, and due to IRB restrictions, the offspring ASD spectrum severity was not considered. The human subjects approval and informed consent were obtained from all participants prior to the initiation of the study and approved by the Ethics Committee of Valencian Infertility Institute—Reproductive Medicine Associates (IVIRMA) Valencia, Spain, with code, #1311-VLC-136-FC.
Individual patient sperm samples from the collection upon enrollment were prepared for sperm analysis, and an aliquot was taken and flash frozen with liquid nitrogen and stored at − 20 °C until shipment on dry ice for the epigenetic analysis. The samples were thawed, and prior to DNA isolation, the sperm were sonicated to destroy and remove any contaminating somatic cells, as previously described [16]. Due to the sperm nuclei being resistant to sonication, any contaminating somatic cells are removed following sonication. The DNA was extracted from the sperm and then fragmented for a methylated DNA immunoprecipitation (MeDIP) procedure to obtain methylated DNA for analysis to identify differential DNA methylated regions (DMRs). The MeDIP is a genome-wide analysis examining 95% of the genome, which is comprised of low-density CpG regions in comparison with the less than 5% of the genome of high-density CpG regions such as CpG islands. The MeDIP DNA libraries were prepared for next-generation DNA sequencing by creating individual patient sequencing libraries. Samples were then sequenced for bioinformatic analysis, as described in Additional file 1 section. A comparison of the sequences between the control (non-autism children) and case (autism children) participant sperm samples identified DMRs, as shown in Fig. 1a. At a p value of p < 1e−05 there were 805 DMRs identified with the majority being a single 1-kb window with fewer (i.e., six) having multiple adjacent 1-kb windows involved. The DMRs at a number of p values are presented for p < 001 to p < 1e−07, Fig. 1a. The DMRs at EdgeR p < 1e−05 all had false discovery rate (FDR) of p < 0.05 and were used for subsequent data analysis. The p < 1e−05 was used to optimize DMR numbers and statistical considerations. A list of these DMRs with various genomic features (e.g., CpG density and chromosomal location) are presented in Additional file 3: Table S1. Observations suggest that males with autistic children have a sperm DMR signature that is distinct from males without autistic children (control).
a DMR identifications. Autism case versus control sperm DMR analysis. The number of DMRs found using different p value cutoff thresholds. The all window column shows all DMRs. The multiple window column shows the number of DMRs containing at least two adjacent significant windows and the number of DMRs with each specific number of significant windows at a p value threshold of 1e−05. b Autism case versus control patient DMR analysis. The DMR locations on the individual chromosomes are identified. All DMRs at a p value threshold of p < 1e−05 are shown with the red arrowheads and clusters of DMRs with the black boxes. c DMR CpG density in the autism case versus control patient DMRs. The number of DMRs at different CpG densities is indicated. All DMRs at a p value threshold of p < 1e−05. d Autism case versus control patient DMR lengths in kilobases. All DMRs at a p value threshold of 1e−05 are shown
The genomic features of the offspring autism susceptibility DMRs were investigated. The chromosomal locations of the DMRs at p < 1e−05 within the human genome are presented in Fig. 1b. The red arrowheads indicate the individual DMRs, and the black boxes represent clusters of DMRs. The DMRs are present on all chromosomes. The CpG density of the DMRs is generally less than 10 CpG per 100 bp with 1–3 CpG predominant for the paternal offspring autism susceptibility DMRs, as shown in Fig. 1c. The size of the DMRs was predominantly 1–3 kb for the sperm DMRs, as shown in Fig. 1d. Additional genomic features are presented in Additional file 4: Table S1. The log-fold change (LFC) in DMA methylation in Additional file 3: Table S1 demonstrated for the 805 DMR in the autism group that 303 have an increase in DNA methylation and 502 have a decrease in DNA methylation. The autism DMRs involved a 38% increase or 62% decrease in DNA methylation. Therefore, the majority of the sperm DMRs had low CpG density, termed a CpG desert, and were 1 kb in length with either an increase or decrease in DNA methylation.
The paternal offspring autism susceptibility sperm DMR-associated genes and corresponding gene functional categories were determined, as presented in Additional file 4: Table S1. The total autism 805 DMRs had 193 with no DMR gene associations (24%), and the DMRs were intergenic and not associated with genes. From the 612 DMR with gene associations (76%), there were 493 DMR that overlapped with annotated genes. There were 17 DMR in the 1–1000 bp and 62 DMR in the 1–5 kb proximal promoter regions. There were 40 DMRs in the 5–10 kb distal promoter region. Therefore, approximately 20% of the DMR are in the proximal and distal promoter region and 80% overlapping the gene annotation regions. There were 193 DMRs that were intergenic. These DMRs are intergenic and not proximal to genes, but can influence gene expression events for megabase distances through ncRNA and chromatin structure alterations, as previously described [19, 20]. Genes within 10 kb of a DMR were identified, which has previously been shown to be optimal for both proximal and distal promoter regions and epigenetic associations [21]. The functional categories corresponding to each DMR-associated gene are summarized in Fig. 2a. The signaling, transcription and metabolism functional categories are predominant. This reflects that these gene functional categories have the highest number of genes within them. A comparison of previously identified genes associated with neurodegeneration, neurodevelopment and autism with the DMR-associated genes of this study is summarized in Fig. 2b. These autism-associated genes have previously been shown to be regulated or involve genetic mutations within autism patients, and the gene symbols, descriptions and associated references are presented in Additional file 5: Table S2. The DMR-associated genes were also used in a gene pathway or gene set analysis to identify associated pathways. Interestingly, the top pathway or gene set identified was autism, and the majority of the subsequent pathways with greater than three genes were all neurodevelopmental- or neurobiology-associated pathways, as given in Table 2. All those gene sets were found to be significant, and a list of the specific DMR-associated genes is provided, as given in Table 2. As shown with all the DMR-associated genes, Additional file 5: Table S2, the associated genes in Table 2 also had approximately a 50% mixture of genes with an increase or decrease in DNA methylation. Therefore, the DMR-associated genes did correlate well with previously identified autism- and neurodevelopment-associated genes. Since the sperm DMRs will impact the embryo epigenomes and transcriptomes of all subsequent somatic cells, this dynamic cascade of developmental epigenetics needs to be considered in potential links in sperm epigenetics and potential neurological impacts on autism.
a DMR-associated gene categories. DMRs at a p value threshold p < 1e−05 are shown. b Autism case versus control DMR PCA. PCA for DMRs at p < 1e−05. The first two principal components used and samples color code index indicated. The underlying data are the RPKM read depth for all DMRs. c DMR-associated genes and autism. The paternal offspring autism-susceptible DMRs previously shown to correlate with autism and associated neurodegenerative disease are presented. DMR-associated genes from the current study were compared to genes associated with autism in the published literature using Pathway Studio software (Elsevier, Inc.). Those that were in common are depicted
The final analysis examined the statistical significance and validation of the DMRs identified. A permutation analysis was performed on the DMRs to determine whether the DMRs were due to background variation in the data or randomly generated. The permutation analysis shows the number of DMRs generated from the control versus autism case comparison was significantly greater than the DMRs generated from random subset analysis, Additional file 3: Figure S2. The red line to the right indicates the comparison DMRs versus the low numbers from the random subset comparison. Another analysis involved a cross-validation of the DMRs and demonstrated approximately 80% accuracy in the confirmation of the DMRs to assess autism susceptibility [22]. A principal component analysis (PCA) of the control male sperm without an autistic child versus the male sperm with an autistic child is presented in Fig. 2c. A clear separation of the DMR principal components is seen between the groups. This demonstrates a distinction between the DMR principal components.
The final validation involved using blinded test sets of samples for analysis to identify and assess the accuracy to determine actual case or control samples. Three different molecular analyses of the original 13 cases and 13 controls were performed and combined for this test set analysis. The test set analysis involved four independent analyses that were combined for the final analysis. The first test set involved eight blinded samples selected from the main analysis samples and reanalyzed (BS1–BS8) and identified as actual case or control sample, as given in Table 1b. All were accurately identified, except one false negative which was identified as a control, but actually was a case sample. A second set of ten blinded test samples (BS9–BS18) was provided by IVI-RMA clinical collaborators and was also identified in an independent analysis as case or control, as given in Table 1c. All were accurately identified, except one false negative which was identified as a control, but was actually a case sample. Therefore, the blinded test set analysis indicated all but one in each test set were accurately identified for approximately a 90% accuracy in the analysis. Since multiple analysis was used for this blinded test set analysis, random batch effect outlier DMRs identified were removed to optimize the analysis. Although significantly more validation with larger clinical test sets is needed, the current study provides the proof of concept that epigenetic biomarkers potentially exist and may be used to diagnose that a father may potentially have a child with a susceptibility for autism.
from Hacker News https://ift.tt/3xemeCt
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.