It is known that neurodevelopmental disorders, including ID, have a sex-biased presentation. Vicoso and Charlesworth (2006)  proposed that beneficial variants are better fixed on the X chromosome than on the autosomes because they will always be expressed in hemizygous males under selection. Therefore, the X chromosome would accumulate beneficial variants at a higher rate than the autosomes. Thus, one hypothesis raised was that X-linked genes carrying deleterious variants under negative selection would evolve more slowly than X-linked genes under positive selection [22,23]. This would explain, as discussed by Turner et al. (2019) , the discrepancy of ID/NDD between sexes and the enrichment of ID/NDD genes on the X chromosome, stating that it may be caused by the deleterious or lethal status of these genes in a hemizygous state, as well as a higher frequency of variants in escape genes, such as DDX3X, in females affected by ID.
Even with the protection of the XCI process, heterozygous females carrying XLID P variants can still manifest ID through several mechanisms, such as the stochastic inactivation process, with a random chance that the X chromosome carrying an alteration will become preferentially active [25–27]; the advancement of age, but in this case a pathogenic variant would only have phenotypic expression in late-onset diseases [28,29]; a positive selection of the mutated X chromosome, due to a growth advantage from the variant; and a negative selection that favors the wild-type chromosome, with escape expression of the mutated allele [26,27,30]. Women usually show milder phenotypes with respect to X-linked variants, even if important inactivation drift occur. A crucial aspect of discussion is the occurrence of disease phenotype in women with pathogenic variants located on the preferentially inactivated chromosome. Remarkably, this indicates that XCI is not always complete (100:0) and may result in residual expression of the mutated allele.
Amos Landgraf et al. (2006)  showed that approximately 1.8% of women in the general population would have extremely or completely skewed XCI deviation (≥90:10) by chance due to the stochastic process of XCI. Additionally, in a sample of 118 Brazilian females, it was demonstrated that moderately skewed XCI (80:20 – 89:11%) was present in 2.6% of the investigated subjects . Our study investigated whether there was an increased frequency of extremely or completely skewed XCI in a cohort of patients with ID. Eleven patients (8%) exhibited skewed XCI ≥90%, of whom four presented with completely skewed XCI (100:0). Considering the ~2% extreme/complete deviation frequency found by Amos Landgraf and colleagues (2006)  in the general population, 3/136 women in our study would be expected to have extreme or complete XCI deviation. However, we observed a value that was four times higher with an enrichment of females with ID and skewed XCI.
These findings are supported by previous studies [7; 13], which observed skewed XCI in 7.6-13% of the females with ID. Considering only the 98 informative sporadic cases investigated by our study, eight (8.2%) showed extreme or total XCI skewing. This is smaller than the frequency that was found in the study by Vianna et al. (2020 - 13.2%), which involved only unrelated patients. The smaller frequency observed in our study compared to the study conducted by Vianna et al. (2020)  with Brazilian patients could be explained by the fact that we previously excluded patients with a positive CMA result, and chromosomal rearrangements involving the X chromosome can result in extremely skewed XCI.
The most plausible hypothesis to explain the increased frequency of skewed XCI compared to the healthy female population would be the presence of a causative variant on the X chromosome, simultaneously associated with the condition of ID. For all cases of variants in genes with a dominant pattern of inheritance, XCI can be considered a contributing factor to the large heterogeneity of phenotypes and degrees of involvement among those affected.
WES of the 11 patients in our cohort with ID and extremely/completely skewed XCI revealed pathogenic variants in eight of them (~73%), which were all mapped to genes with dominant inheritance patterns, even though only four patients had X-linked variants (DDX3X, WDR45, PDHA1); different pathogenic variants in the DDX3X gene were identified in two patients. Of the three X-linked genes identified as mutated, only one of them, DDX3X is a gene described to escape XCI, having biallelic expression in females [31,32]. DDX3X microduplications were also reported in ID patients, indicating that the disturbance of gene dosage leads to pathological phenotypes . This finding of skewed X-chromosome inactivation in females with DDX3X variants was previously described in the study by Fieremans and collaborators (2016)  and in our recent work (Fonseca et al., 2021 ). In theory, DDX3X would have different expression levels in males and females, and skewed expression would not be expected in women carrying mutations. However, the pattern for these genes that escape X inactivation is more complex, as many of them are known to present lower expression in the inactive chromosome X than in the active X, as well as variable expression among different females and, in some cases, within different tissues of the same woman. These genes would essentially also be "dosage compensated", like genes subjected to inactivation, but with a more heterogeneous expression pattern. DDX3X has a homologous paralog on the Y-chromosome – DDX3Y gene – and escape genes with retention on the Y chromosome are considered of great evolutionary importance . The conservation of DDX3X gene and the high intolerance of DDX3X and DDX3Y to loss-of-function variants (pLi= 1 and 0.96, respectively, according to gnomAD) indicate great functional relevance: while variants in DDX3X are associated with cognitive disease, variants in DDX3Y result in infertility [34–36].
Unlike DDX3X, WDR45 and PDAH1 are subject to XCI. WDR45 (OMIM *300526) is associated with the regulation of cellular autophagy processes. Therefore, the expression of P/LP variants may result in the accumulation of intracellular debris, affecting the functioning of the affected tissue. Such accumulations tend to primarily impair the central nervous system due to iron assembling in brain tissue and may impact global development, causing ID and cognitive degeneration [37,38]. Skewed X-chromosome inactivation in women carrying WDR45 variants have been previously reported, in which only the mutated alleles were detected [7,40,41]. The PDAH1 gene (OMIM *300502) acts in the energy pathway of ATP production, and its mutations have been associated with the accumulation of intracellular pyruvate and lactic acid, resulting in an acidosis state that causes neurological damage [39,40]. Willemsen and colleagues (2006)  performed an investigation of the XCI pattern in fibroblasts in four women with PDAH1 variants who were affected by pyruvate dehydrogenase deficiency, observing a 90% inactivation shift in all patients.
Fieremans et al. (2016)  observed that among ID patients who had skewed inactivation ≥90:10, ~45% carried pathogenic variants, nine of which were in X-linked genes, including DDX3X and WDR45 (DDX3X, SMC1A, WDR45, NHS, MECP2, MED12, HDAC8, and TAF9B), and two on autosomes (EP300 and SYNGAP1). In a Brazilian cohort, Vianna et al. (2020)  also identified pathogenic variants in seven patients with skewed XCI, two with structural chromosomal rearrangements (chromosome 3 deletion and unbalanced translocation t(X;2)), and four variants mapped to three X-linked genes were identified (NLGN4X, USP9X, and TAF1), in addition to one variant in an autosomal gene (DVL1).
An interesting finding in our study was the detection of variants in four autosomal genes (KCNB1, CTNNB1, YY1, ANKRD11), not documented in previous studies using the same approach. In these cases, the skewed XCI may have occurred by chance. However, we cannot exclude the possibility that variants in autosomal ID-related genes may have a crosslink with skewed XCI [7;13]. We did not identify a robust connection between the CTNNB1, ANKRD11, and KCNB1 genes and the XCI. Liu and colleagues (2016)  observed in hepatocellular carcinoma that the lncRNA for the FTX gene (lnc-FTX), which acts in XCI as an activator of Xist in mice , interferes with the activity of beta-catenin, which is encoded by CTNNB1. However, the reverse effect—that is, a beta-catenin alteration interfering with lnc-FTX levels—was not documented, and it is an interesting hypothesis that could be experimentally evaluated. It would therefore be beneficial to investigate the pattern of XCI in females with CTNNB1 deleterious variants to assess the hypothesis that this autosomal gene could be related to the whole XCI process.
On the other hand, the autosomal gene YY1 (OMIM *600013) is directly linked to XCI. YY1 (OMIM *600013) is a transcription factor involved in embryogenesis, differentiation and cell proliferation and acts as a transcriptional repressor . According to several studies [45–47], YY1 plays an essential role in XCI, participating via the association of Xist RNA with DNA on the inactive X chromosome and acting as a bridge. The YY1 protein binds to the active Xist RNA, presenting a monoallelic attachment restricted to the inactive X chromosome. In a functional study conducted by Makhlouf et al. (2014) , it was reported that YY1 was also essential in the regulation and maintenance of Xist transcription and expression. YY1 knockdown decreases Xist levels by 80% in mouse somatic cells after the establishment of X inactivation, as well as the number of cells with the Xist condensation cloud during cell differentiation. This work concluded that YY1 is the first identified factor controlling the monoallelic expression of Xist during the initiation and maintenance of X-inactivation . Therefore, we can hypothesize that a de novo YY1 pathogenic variant could be the cause of the primary extreme XCI deviation found in P6; only one patient with YY1 pathogenic variant was previously investigated regarding XCI and she also presented extreme skewing . The investigation of the XCI pattern of other females with pathogenic YY1 could clarify this issue.
The present study presents limitations, such as the use of a single locus (AR gene) to investigate XCI patterns. The use of more than one marker would increase our ability to identify informative females and would also confirm the identified patterns. A second limitation was the use of genomic DNA samples only from peripheral blood, which should not be completely representative of the XCI pattern in the central nervous system; however, a recent study showed that blood results can be used as a proxy for other tissues . Moreover, the AR gene methylation study performed to determine the inactivation status of the X chromosome may represent an indirect method, considering that it is not a fully predictive character of the inactivation status of the investigated locus, since does not determine which allele is preferentially inactive in a patient. A more direct and consistent alternative would be to use RNA expression analysis by RTqPCR to study the XCI pattern . Notwithstanding, using this approach, we did identify a representative number of patients with skewed XCI and P/LP variants, in four patients with variants mapped to X-linked genes and in one patient with a variant mapped to an autosomal gene associated with the XCI process itself, supporting our XCI data and the hypothesis that an extremely skewed XCI in females with ID is indicative of the presence of pathogenic variants in X-linked or autosomal genes linked to the XCI process.
Based on these findings, we can conclude that the frequency of XCI deviation in the peripheral blood of females with ID is significantly higher than that in the general population. We also found that the analysis of the XCI pattern in females with ID, a cost-effective molecular procedure, is useful for detecting extreme or complete deviation (≥90%), which was demonstrated as a factor indicative of causative variants located on the X chromosome or even in autosomes. In the scenario of a limited budget, the investigation of XCI patterns would indicate female patients with ID who carry pathogenic monogenic variants to be selected for further genetic tests and genetic counseling.