Heritability of skewed X-inactivation in female twins is tissue-specific and dependent on age

To balance the X-linked transcriptional dosages between the sexes, one of the two X-chromosomes is randomly selected to be inactivated in somatic tissues of female placental mammals. Non-random, or skewed X-chromosome inactivation (XCI) toward one parental X has been observed in female somatic tissues, and this skewing effect has been associated with several complex human traits. However, the extent of the influence of genetic and environmental factors on XCI skewing is largely unknown. Here, we use RNA-seq and DNA-seq data taken from a large cohort of female twins to quantify the degree of skewing of XCI (DS) in multiple tissues and to study the relationship of XCI with age, genetic factors and complex traits. We show that the XCI patterns are highly tissue-specific with a higher prevalence of skewed XCI in blood-derived tissues than in fat or skin tissue. We also show that the DS in blood-derived tissues is associated with age and that the acquired DS occurs uniquely in blood-derived tissues with an inflection point at approximately 55 years of age. Heritability analysis indicates that the heritability of DS is both age and tissue specific; DS is heritable in blood tissue of females >55 years-old (h2 = 0.34) but is not heritable in blood tissues of females <55 years-old (h2 = 0), nor in skin and fat tissues at any age. We find a positive association between the DS and smoking status in blood tissues of older females (P = 0.02). The high tissue specificity of XCI patterns in human indicates the existence of tissue-specific mechanisms influencing XCI patterns, including genetic and environmental factors. We conclude that the heritability of XCI skewing in blood-derived tissues is dependent on age, representing a Gene x Age interaction that can shift the functional allelic dosage of an entire chromosome in a tissue-restricted manner.


Introduction 27
To balance the X-linked transcriptional dosages between the single X chromosome of males and the 28 two X chromosomes of females, one X chromosome is silenced in female placental mammals 1 . The 29 X-chromosome inactivation (XCI) process starts during preimplantation phases of human embryonic 30 development, presumably at around the 8-cell stage 2 . XCI is initiated by the transcription of XIST, a 31 17 kb, alternatively spliced long non-coding RNA mapped to Xq13.2 and exclusively expressed on 32 the inactive X (Xi) 3 . Once transcribed, XIST molecules spread in cis along the X chromosome 4,5 33 inducing a progressive epigenetic silencing through the recruitment of chromatin remodelling 34 enzymatic complexes, which impose repressive histone and DNA changes on the Xi chromosome 6,7 . 35 Within each cell, the parental X chromosome selected for inactivation seems to occur at random, 36 and the Xi is mitotically inherited to future somatic daughter cells. This random inactivation results 37 in a mosaic of cells within an individual, where overall, a balanced expression (50:50) of both 38 parental X-linked alleles is expected. Asymmetric selection of the X chromosome to inactivate 39 causes the predominance of one parental Xi in a population of cells, unbalancing the X-linked 40 transcriptional and allelic dosages toward one parental X chromosome. This phenomenon, known 41 as skewed XCI (or non-random XCI), occurs when at least 80% of cells within a tissue inactivate the 42 same parental X chromosome. The factors underlying primary skewed XCI are varied and several 43 mechanisms are possible (reviewed in 8 ). Secondary (or acquired) skewed XCI can result from 44 positive selection of cells that after having inactivated a particular parental X, acquire a survival 45 advantage over cells who inactivated the other parental X chromosome. Skewed XCI patterns can 46 also be generated by the stochastic overrepresentation of cell clones in a given tissue, due for 47 instance, to depletion of stem cell populations. 48

49
Comprised of 155 MB and containing >800 protein-coding genes, the X chromosome represents 50 approximately 8% of the human genome. In heterozygous females with skewed XCI, the X-linked 51 homozygous. Skewed XCI is a major cause of discontinuity of dominance and recessiveness, as well 53 as penetrance and expressivity of X-linked traits. How skewed XCI patterns modulate phenotypes in 54 females, and whether they are a cause or a consequence of associated phenotypes is not fully 55 understood. Skewed XCI patterns have been observed in females with X-linked diseases 9,10 , 56 autoimmune disorders 11,12 , as well as in breast 13 and ovarian cancer 14 . In autoimmune diseases with 57 higher prevalence in females, including rheumatoid arthritis and systemic lupus erythematosus, XCI 58 is hypothesized to play a role. Chromosome X is enriched for immune-related genes and skewed XCI 59 patterns cause the breakdown of thymic tolerance induction processes 15 conferring a high 60 predisposition to develop autoimmunity (reviewed in 16 ). XCI skewing levels in blood-tissues have 61 been associated with ageing, with multiple studies indicating an increase after 50-60 years of age 17-62 22 . To date, the mechanisms underlying skewed XCI in humans remain to be determined and 63 hypotheses are still controversial. Several twin studies have reported that genetic factors contribute 64 to XCI skewing in blood derived cells 21,23 , while other evidence indicated that most of the XCI 65 skewing levels in human are acquired secondarily 24 . 66 67 Nearly all studies of XCI skewing levels in humans have been carried out in peripheral blood samples 68 or in very small sample sizes 25 , while XCI patterns in other tissues have not been studied in great 69 detail 19,26,27 . In this study, we comprehensively assessed XCI patterns in a multi-tissue sample of 70 nearly 800 female twins from the TwinsUK cohort 28 . We quantified the degree of skewing of XCI 71 using a metric based on XIST allele-specific expression from paired RNA-seq and DNA-seq data in 72 four tissues. We examined the tissue-specific prevalence of skewed XCI patterns, compared the XCI 73 skewing levels between tissues and evaluated the association between XCI skewing and lifestyle 74 traits. In order to investigate the factors underlying the skewed XCI, we utilized classical twin models 75 to characterize the extent of the influence of genetic and environmental factors on the tissue-76 basis. 78 79

81
We assessed XCI patterns in multi-tissue samples from female twin volunteers from the TwinsUK 82 cohort aged 38-85 years old (median age = 60; Figure S1) 28,29 . We quantified the degree of skewing 83 of XCI using a metric based on XIST allele-specific expression (ASE) from paired RNA-seq and DNA-84 seq data. XIST is uniquely expressed from the Xi 3 , so the relative expression of parental alleles within 85 We assessed the robustness of our estimates of the degree of skewing with an alternative DNA-100 based measure of XCI, the Human Androgen Receptor Assay (HUMARA) 34 . HUMARA was and is still 101 exhibited skewed XCI (Table 1). In order to examine the extent of similarities of XCI patterns 127 between tissues, we compared the tissue-specific XCI skewing levels in a pairwise manner (Fig 3). 128 For each tissue-tissue comparison, we included individuals with XISTASE calls in both tissues (Table  129 2). We found the strongest correlation on XISTASE calls between LCLs and whole-blood (N = 59, r = 130 0.78, P = 2 x 10 -13 ), indicating that blood-derived tissues share highly similar XCI skewing levels. We 131 also found a good degree of similarity between the XCI skewing levels in fat and skin tissues (N = 132 252, r = 0.47, P = 2 x 10 -15 ; Fig 3). However, low concordance was observed between skin and whole-133 blood (N = 47, r = 0.3, P = 0.04) and fat and whole-blood (N = 57, r = 0.33, P = 0.02). Our data 134 demonstrate that tissue-specific XCI skewing within an individual is common in the population, 135 indicating that XCI patterns are partially controlled by tissue-specific regulatory mechanisms. 136

137
The active or inactive state of each X chromosome in a cell is clonally passed on to daughter cells.
In a pool of cells derived from a single clone (or patch), the XCI patterns are expected to be completely skewed. Patch size refers to the amount of cell clones in a pool of cells (e.g. in a tissue biopsy). We considered the possibility that patch size might bias our quantification of XCI patterns in fat and skin samples. This is likely to occur in biopsies that are smaller than the tissue patch size.
However, several considerations led us to exclude the possibility that patch sizes in fat and skin biopsies might confound our XISTASE calls. First, the biopsies included skin samples of 8mm 3 in size, which were cut into 2 skin and 3 fat samples. As reported in another study, this size is large enough to measure the XCI ratio without being confounded by patch size 37 . Second, most individuals exhibit random XCI patterns in fat and skin tissues, which is unlikely if patch size was larger than the biopsies. We therefore conclude that the biopsies used in this study are large enough to accurately assess the XCI patterns without being biased by patch size.
Lymphoblastoid Cell Lines (LCLs) generated by Epstein-Barr virus mediated transformation of B 139 lymphocyte cells have been and are widely used in gene expression studies. However, the possibility 140 that the cell lines are monoclonal and/or polyclonal due to selection in the transformation process 141 or clonal expansion in cell culture, and hence not be representative of the in-vivo XCI skewing levels, 142 is a potential problem when using LCLs to assess XCI skewing 38 . As the profiled RNA in this study was 143 extracted from the LCLs very shortly after transformation with limited passaging or time in culture 144 we expected this effect to be minimal, however, to address the possibility we performed the 145 following analyses. First, as described above and shown in Figure 1a, the degree of skewing in LCLs 146 were highly correlated with the HUMARA-based quantifications of XCI patterns in paired whole-147 blood samples (r = 0.71, N = 10). We would not expect such high similarity between the two 148 quantifications if clonal propagation had occurred in LCLs samples after preparation. This was 149 confirmed by the high correlation between LCLs and whole-blood XISTASE values (r = 0.78, N = 59; 150 Fig 3) and overall similarity in the prevalence of skewed XCI in LCLs and whole-blood (Table 1). 151 Finally, we assessed the degree of skewing in monocytes, B, T-CD4 + , T-CD8 + and natural-killer (NK) 152 cells purified from two monozygotic twins exhibiting skewed XCI patterns in LCLs and from one 153 individual exhibiting random XCI patterns in LCLs. We found that in both monozygotic twins showing 154 skewed XCI in LCLs, the majority of immune cell types exhibited skewed XCI patterns. Conversely, 155 none of the immune cell types purified from the non-skewed individual exhibited skewed XCI 156 patterns (Table 3). We conclude that the XCI skewing levels of LCLs in this study are representative 157 of XCI skewing in-vivo in blood tissues. 158 159 XCI skewing levels are positively associated with age in blood-derived tissues 160 XCI skewing levels in peripheral blood have been shown to increase with age in multiple studies 17-161 21,23,35,39,40 . The age-related increase of XCI skewing levels continues throughout life, since 162 limited knowledge on the relationship between XCI patterns and ageing in tissues other than blood. 164 In order to explore this, we investigated the association between age and degree of skewing in LCLs, 165 fat and skin. Our whole-blood estimates were excluded from analysis due to low sample size (N = 166 72). Age was positively associated with XCI skew in LCLs (N = 422, P < 0.01), but we did not detect 167 any association between XCI skew and age in skin (N = 336, P = 0.4) or in fat (N = 378, P = 0.7). 168

169
We next explored the dynamics of DS and age progression in each tissue, using the lowess 170 procedure. Lowess curve detected an increase of DS beginning at around 55 years-old in LCLs (Fig  171   4), in agreement with what was found in other studies 19,21 . Since the increase of DS starts at around 172 55 years, we divided LCLs samples into a younger group (N = 141, age < 55) and an older group (N = 173 281, age ≥ 55). We found that the mean DS in LCLs was significantly higher in older than in younger 174 females (DSyounger = 0.2, DSolder = 0.24, P = 0.03; Figure 4). Accordingly, we found that the frequency 175 of skewed XCI in LCLs was significantly higher in older (38%) than in younger (28%) females (c $ test, 176 P = 0.04; Fig 4). In agreement with the lack of association between the DS and age, we did not detect 177 significant differences between the mean DS in young and older females in fat (DSyounger = 0.15, 178 DSolder = 0.15) or in skin tissues (DSyounger = 0.16, DSolder = 0.17). To acquire a more detailed view of 179 the tissue-specific prevalence of skewed XCI in different groups of age, we categorized the samples 180 into four age groups (40-50, 50-60, 60-70, >70) and calculated the frequency of skewed XCI in each 181 category (Fig 4). We found that the frequency of skewed XCI increased with age in LCLs, with 41% 182 of individuals >65 years-old demonstrating skewed XCI patterns. We did not observe any increase 183 in the skewed XCI frequencies with age in fat and skin tissues. Overall, these data further confirm 184 that XCI skewing levels increase with age in blood-derived tissues, supporting previous 185 investigations. However, we find that there is no increase in XCI in fat and skin tissue from the same 186 individuals, suggesting that acquired XCI skewing with age is a distinctive feature of blood-derived 187 phenotype variance. In order to investigate whether heritability varies with age, we stratified the 198 twin pairs into a younger group (age < 55) and an older group (age ≥ 55; Table S1). Age 55 was 199 chosen as it was identified as the inflection point at which XCI skew begins to increase in the lowess 200 analysis above. We found that XCI skewing is heritable in LCLs of older females (h 2 = 0.34 , P = 9.6e-201 07), but not younger females (h 2 = 0, P = 1). There was no evidence of heritability of XCI skew in fat 202 or in skin tissues at any age (Table 4). The highest proportion of variance was explained by unique 203 environmental factors in all tissues of both younger and older females (E 2 LCLs_younger = 0.99, E 2 LCLs_older 204 = 0.66, E 2 Fat_younger = 0.73, E 2 Fat_older = 0.92, E 2 Skin_younger = 1, E 2 Skin_older = 1). As a complement to the 205 heritability analysis, we calculated intraclass correlation (IC) of XCI skew within MZ and DZ twin pairs 206 of all ages, and within younger and older MZ and DZ twin pairs (Table 5). IC analyses of twin pairs is 207 often used to demonstrate the existence of genetic effect in smaller sample sizes. The IC of XCI skew 208 within MZ twins pairs was positive and statistically significant (ICMZ_allAges = 0.31, P = 0.02). We found 209 significant IC of XCI skew within older MZ twin pairs (ICMZ_older = 0.42, P = 0.005), but not within 210 young MZ twin pairs (ICMZ_younger = 0.06, P = 0.8). We did not detect significant IC within DZ twin pairs 211 pairs compared with DZ twin pairs indicates the involvement of genetic determinants in the 213 regulation of XCI skew in blood-derived tissues. The increase of IC in older compared to younger MZ 214 twin pairs and the fact that the heritability of XCI skew is observed only in females older than 55, 215 confirm a role for genetic variants as age-dependent regulators of the acquired XCI skew in blood-216 derived tissues. Presumably, genetically-determined secondary cell selection processes act in 217 haematopoietic cell lineages, with the high mitotic rates contributing to the manifestation of their 218 effects in blood-derived tissues. Results also highlight an age-independent role for environmental 219 factors as regulators of XCI skew in blood, fat and skin tissues. 220

222
Tobacco smoking has been reported to induce epigenomic changes including DNA methylation 223 variation (reviewed in 41 ). Smoking is a well characterized risk factor in cancer 42 and, as more recently 224 discovered, in the aetiology of autoimmunity 43 . Although smoking-related X-linked DNA methylation 225 sites have been discovered 44 , no previous studies, to our knowledge, have investigated the 226 relationship between smoking and XCI patterns. We reasoned that changes of XCI patterns may 227 result from smoking, and affect in turn short-term and long-term health. In order to test our 228 hypothesis, we used the 270 individuals in our dataset for which we had smoking status at the time 229 of sample collection, including 233 never smokers and 37 current smokers 45 . We found no difference 230 in the frequency of skewed XCI patterns between never and current smokers (36% and 35% 231 respectively) in LCLs. To take into account the effects of age on the degree of skewing in blood-232 derived tissues and to examine the relationship between smoking status and degree of skewing at 233 different ages, we split the dataset into a younger (age < 55) and older group (age ≥ 55; Table S2). 234 While the frequencies of skewed XCI were very similar between young smokers and young never 235 smokers (27% and 28% respectively), we detected a higher prevalence of skewed XCI in older 236 overall positive association between XCI skew and smoking status in older (P = 0.02), but not in 238 younger individuals (P = 0.5). The data suggest a role for smoking as a modulator of XCI skew in 239 blood-derived tissues of females older than 55. Presumably, the association between smoking and 240 XCI skew changes is complex, and further investigations are needed to characterize the genetic and 241 molecular mechanisms underlying this phenomenon. 242 243

244
In this study, we used multi-tissue transcriptomic data from twins to comprehensively characterize 245 XCI patterns in LCLs, whole-blood, fat and skin tissues from a healthy twin cohort. We show XCI 246 patterns to be tissue-specific and that blood-derived tissues exhibited the highest prevalence of 247 skewed XCI and share the highest similarity of XCI patterns. These findings indicate that XCI patterns 248 are partially driven by tissue-specific mechanisms, and that the XCI skew measured in blood is not a 249 reliable proxy for the skew in other tissues. Skewed XCI patterns limited to disease-relevant tissues 250 and cells have been observed in multiple conditions 9,10,13,14,46,47 but except for several cases of X-251 linked diseases, their roles in disease aetiology and predisposition remain largely unknown. Our 252 results demonstrate that tissue-specific XCI patterns within an individual is common in this healthy 253 population. 254

255
We show that XCI skewing levels in blood tissues increase with age, with an inflection point at 256 around 55, confirming previous reports [17][18][19][20][21]23,35,39,40 . In this study, more than 41% of females >65 257 years-old demonstrate skewed XCI patterns in blood-derived tissues, indicating that acquired 258 skewed XCI is a highly prevalent phenotype in ageing populations. We show age-related increase in 259 XCI skew is a distinctive feature of blood-derived tissues, with no evidence for an age-related 260 increase in fat or skin. Age-related increase in XCI skew partially explains the higher incidence of 261 ageing remain largely unknown, but may have a broad impact on the immune system. 263 Hematopoietic stem cells and the immune system continue to develop throughout life. Presumably, 264 imbalanced X-linked immune-related gene expression toward one parental haplotype leads to a 265 reduced molecular diversity, which may translate in a decline of immune repertoire as well as poor 266 sustenance of the immunological memory. 267

268
Previous twin studies have reported that XCI patterns in blood have a genetic component 21,23 . To 269 our knowledge, this is the first study to investigate heritability of XCI skewing levels in other tissues. 270 We found that the heritability of XCI skewing level is limited to blood-derived tissues of females >55 271 years-old (h 2 = 0.34), with no evidence of heritability in fat or skin or younger individuals in any 272 tissue. The restriction of heritability to blood of older individuals is of interest given the link between 273 skewed X-inactivation and clonal haematopoiesis. Positive selection of cells carrying an 274 advantageous somatic mutation will lead to clonal haematopoiesis and skewed XCI patterns as the 275 selected cells will carry the same inactivated parental X. Somatic mutation-driven clonal 276 haematopoiesis is now known to be common in blood of healthy older individuals and is often 277 referred to as clonal haematopoiesis of indeterminate potential (CHIP) 31,48-50 . CHIP is associated 278 with increased risk of both cancer and all-cause mortality 51,52 . The increase in XCI skew in older 279 smokers in our study is consistent with the increase in clonal haematopoiesis observed in 280 smokers 50,53,54 . It is unknown to what extent CHIP accounts for age-acquired XCI skew, however if it 281 is a major driver this would suggest that like age-related XCI skew, CHIP has a significant germline 282 genetic component. Stochastic selection of cells could also contribute to the variance of XCI skewing 283 levels, but, in agreement with previous works 21,23 , we reason that their contribution is minimal. If 284 stochastic selection of cells was a dominant mechanism, the correlation of XCI patterns between 285 twin pairs would decrease with age. 286 Overall, the data presented in this study indicate a Gene x Age interaction that shifts the functional 288 allelic dosages of chromosome X in a tissue-restricted manner. The high prevalence of skewed XCI 289 and tissue-restricted XCI in a healthy population could complicate discovery of Chromosome X  Whole-Genome Sequence data (WGS) were generated within the UK10K project as previously 312 described 56 . 557 individuals had both X-chromosome sequence data and RNA-seq data for at least 313 one tissue. For individuals with unavailable X chromosome sequence data, X-linked genotypes data 314 were retrieved from the TwinsUK genotypes previously imputed into the 1000 Genomes Project 315 phase 1 reference panel 57,58 , as described 59 . Haplotypes of X-linked SNPs with a MAF >5% were then 316 phased using shapeit v2.r837 60,61 , with the --chrX flag to set up all functionalities for the phasing of 317 non pseudo-autosomal (non-PAR) regions of the X-chromosome, along with a phasing window of 318 2Mb and 1000 conditional states. Phasing was performed using the genetic map b37 and the 1000 319 Genomes Project phase 3 reference panel of non-PAR X-linked haplotypes 61 . Phased X-linked SNPs 320 were used for further analysis.  Monocytes, B, T-CD4 + , T-CD8 + and NK cells were purified using fluorescence activated cell sorting 337 (FACS) from two monozygotic twins exhibiting skewed XCI patterns in LCLs and from 1 individual 338 exhibiting random XCI patterns. Total RNA was isolated and cDNA libraries for sequencing were 339 generated using the Sureselect sample preparation protocol. Samples were then sequenced in 340 triplicates on an Illumina HiSeq machine and 126 bp paired-end reads were generated. Adapter and 341 polyA/T nucleotide sequences were trimmed using trim_galore and PrinSeq tools 64 respectively. 342 Human and prokaryotic rRNAs were identified using sortmerna v.2.1 66 and removed. Reads were 343 aligned to the UCSC GRCh37/hg19 reference genome using STAR v.2.5.2a 65 . Alignments containing 344 non-canonical and unannotated splice junctions were discarded. Properly paired and uniquely 345 mapped reads with a MAPQ of 255 were retained for further analysis. 346

347
To eliminate mapping biases, all RNA-seq data were re-aligned within the WASP pipeline for 348 mappability filtering 67 . The WASP tool has an algorithm specifically designed to identify and correct 349 mapping biases in RNA-seq data. In each read overlapping a heterozygous SNP, the allele is flipped 350 to the SNP's other allele (generating all possible allelic combinations) and the read is remapped. 351 Reads that did not remap to the same genomic location indicate mapping bias and were discarded. 352 Reads overlapping insertions and deletions were also discarded. Properly paired and uniquely 353 mapped reads were retained for analysis. 354

355
Allelic read counts at heterozygous SNPs within XIST were quantified from paired RNA-seq and X-356 linked genotypes data with GATK ASEReadCounter 68 . Reads flagged by ASEReadCounter as having 357 low base quality were discarded. To increase the confidence that genotypes were truly 358 heterozygous, only X-linked SNPs with both alleles detected in RNA-seq data and with a read depth 359 of least 10 reads were retained for analysis. X-chromosome allelic read count tables of samples with 360 at least 1 XIST-linked SNP passing all quality filters were retained as informative of XCI skewing 361 levels. To quantify the allele specific expression (ASE) of each heterozygous X-linked SNP, the read 362 count at the major allele was divided by the read depth at the site. 363

364
In each sample, the XCI skewing levels were quantified by averaging the ASE values of heterozygous 365 SNPs within XIST. All SNPs were phased prior to averaging as detailed above. The measure, called 366 The relative contributions of additive genetic factors (A), shared (C) and unique environmental 386 factors (E) to the tissue-specific variance of DS, were calculated using the twinlm() function in the 387 mets R package 69 . For each tissue, samples were split into a young (< 55) and an older (≥ 55) group 388 according to their ages (Table S1). Due to the low number of MZ and DZ twin pairs in each group, 389 whole-blood was excluded from heritability analysis. To further assess the contribution of genetic 390 effects, the intraclass spearman's correlation (IC) of DS in blood-derived tissues of young and older 391 MZ and DZ twin pairs was also calculated. 392

393
Association between the degree of skewing in LCLs and self-reported smoking status was tested in 394 the 270 individuals with reliable smoking status recorded 45 . Dataset included 270 females classified 395 either as current smokers (N = 37) or never smokers (N = 233; Table S2). To examine the association 396 between DS and smoking status, the smoking status was converted into a binary trait (0 = no smoker, 397 1 = smoker). A linear model of the DS as a function of the smoking status was then implemented for 398 younger (age < 55) and older (age ≥ 55) individuals separately. Age was used as covariate. A P value 399 ≤ 0.05 was considered to be statistically significant.  184-193, doi:10.1152/physiolgenomics.00163.2003 494 (2004).  9-16, doi:10.1182/blood-2015-03-631747 526 (2015). 527 52 Steensma, D. P. Clinical consequences of clonal hematopoiesis of indeterminate potential. 528 Blood Adv 2, 3404-3410, doi:10.1182/bloodadvances.2018020222 (2018 A.Z. and K.S.S. wrote the manuscript. All authors read and approved the manuscript. The authors 584 also thank Julia El-Sayed Moustafa and Amy Roberts for providing feedback on the manuscript. 585 586

Competing interests 587
The authors declare that they have no competing interests. 588 589

Materials and Correspondence 590
Correspondence should be addressed to Kerrin S. Small (kerrin.small@kcl.ac.uk) 591 592

Data Availability 593
TwinsUK RNAseq data is available from EGA (Accession number: EGAS00001000805). 594 TwinsUK genotypes are available upon application to TwinsUK (www.twinsuk.ac.uk).    Table 3. XCI-skew in LCLs in this study is representative of XCI in purified primary immune cells.
Degree of skewing of XCI in immune cell types purified from 2 monozygotic twins (Twin A, Twin B) exhibiting skewed XCI patterns in LCLs, and from 1 individual (Individual C) exhibiting random XCI pattern in LCLs. Degree of skewing ≥ 0.3 indicates skewed XCI patterns.   Table 5. Intra-class correlations of XCI skew in age stratified twin pairs. Twin pairs < 55 years-old are classified as young, twin pairs ≥ 55 years-old are classified as old.