Age acquired skewed X chromosome inactivation is associated with adverse health outcomes in humans

  1. Amy L Roberts  Is a corresponding author
  2. Alessandro Morea
  3. Ariella Amar
  4. Antonino Zito
  5. Julia S El-Sayed Moustafa
  6. Max Tomlinson
  7. Ruth CE Bowyer
  8. Xinyuan Zhang
  9. Colette Christiansen
  10. Ricardo Costeira
  11. Claire J Steves
  12. Massimo Mangino
  13. Jordana T Bell
  14. Chloe CY Wong
  15. Timothy J Vyse
  16. Kerrin S Small  Is a corresponding author
  1. Department of Twin Research & Genetic Epidemiology, King’s College London, United Kingdom
  2. Department of Medical and Molecular Genetics, King’s College London, United Kingdom
  3. NIHR Biomedical Research Centre, Guy's and St Thomas' Foundation Trust, United Kingdom
  4. Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, United Kingdom

Abstract

Background:

Ageing is a heterogenous process characterised by cellular and molecular hallmarks, including changes to haematopoietic stem cells and is a primary risk factor for chronic diseases. X chromosome inactivation (XCI) randomly transcriptionally silences either the maternal or paternal X in each cell of 46, XX females to balance the gene expression with 46, XY males. Age acquired XCI-skew describes the preferential selection of cells across a tissue resulting in an imbalance of XCI, which is particularly prevalent in blood tissues of ageing females, and yet its clinical consequences are unknown.

Methods:

We assayed XCI in 1575 females from the TwinsUK population cohort using DNA extracted from whole blood. We employed prospective, cross-sectional, and intra-twin study designs to characterise the relationship of XCI-skew with molecular and cellular measures of ageing, cardiovascular disease risk, and cancer diagnosis.

Results:

We demonstrate that XCI-skew is independent of traditional markers of biological ageing and is associated with a haematopoietic bias towards the myeloid lineage. Using an atherosclerotic cardiovascular disease risk score, which captures traditional risk factors, XCI-skew is associated with an increased cardiovascular disease risk both cross-sectionally and within XCI-skew discordant twin pairs. In a prospective 10 year follow-up study, XCI-skew is predictive of future cancer incidence.

Conclusions:

Our study demonstrates that age acquired XCI-skew captures changes to the haematopoietic stem cell population and has clinical potential as a unique biomarker of chronic disease risk.

Funding:

KSS acknowledges funding from the Medical Research Council [MR/M004422/1 and MR/R023131/1]. JTB acknowledges funding from the ESRC [ES/N000404/1]. MM acknowledges funding from the National Institute for Health Research (NIHR)-funded BioResource, Clinical Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London. TwinsUK is funded by the Wellcome Trust, Medical Research Council, European Union, Chronic Disease Research Foundation (CDRF), Zoe Global Ltd and the National Institute for Health Research (NIHR)-funded BioResource, Clinical Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London.

Editor's evaluation

XCI skewing is affected by age but how this may affect a person's health is not known. Roberts et al., demonstrate that these changes result in an increased risk of cardiovascular disease and cancer. These findings will be of interest to researchers studying the impact of age on health.

https://doi.org/10.7554/eLife.78263.sa0

Introduction

Ageing is a heterogenous process characterised by cellular and molecular hallmarks and can manifest clinically as frailty and multimorbidity (Clegg et al., 2013; López-Otín et al., 2013). Ageing is a primary risk factor for diseases such as cardiovascular disease and cancer, and a better understanding of the biomarkers of ageing promises to reduce the burden of chronic disease which significantly impacts the human healthspan (López-Otín et al., 2013).

X chromosome inactivation (XCI) evolved in placental mammals to compensate for the X-linked gene dosage between XX females and XY males. XCI transcriptionally silences either the maternal or paternal X in each cell to equalise the gene expression between 46, XX females and 46, XY males (Lyon, 1961). The selection of which X is silenced is a random process that occurs during development, with the XCI status then clonally inherited by all daughter cells. Therefore, mammalian female tissues are mosaics with respect to XCI status, with an expected ratio of 1:1.

However, some individuals display a skewed pattern of XCI (XCI-skew), which is defined as a deviation from the expected 1:1 ratio. Examples of primary XCI-skew have been identified, including stochastic events resulting in XCI-skew across all tissues (Tukiainen et al., 2016) or the preferential selection of cells expressing functioning alleles in heterozygous females with X-linked recessive traits (Busque and Gilliland, 1998; Nyhan et al., 1970). However, secondary or age acquired XCI-skew is more common and refers to increasing XCI-skew with age, particularly in mitotically active blood tissue (Busque et al., 1996; Gale et al., 1997). Within individuals, the correlation of XCI ratios between blood and other tissues diminishes over the life course as the XCI ratios in blood continue to skew with age (Bolduc et al., 2008; Zito et al., 2019).

The stability of XCI-skew in blood has been demonstrated over 18–24 months and is thought to be a gradual process affecting the whole haematopoietic stem cell population rather than representing fluctuations in the active stem cell pool (Tonon et al., 1998; van Dijk et al., 2002). Therefore, though XCI-skew is a sex-specific measurable phenotype, it is a potential marker of stem cell depletion or polyclonal expansion of haematopoietic stem cells, which are age-associated traits irrespective of chromosomal sex (Busque et al., 1996; Busque et al., 2012; Gale et al., 1997). Age acquired XCI-skew has previously been linked to autoimmunity, which presents with a stark sex-imbalance (Chabchoub et al., 2009), as well as breast and ovarian cancers, albeit with inconsistent findings (Kristiansen et al., 2002; Lose et al., 2008; Manoukian et al., 2013; Struewing et al., 2006). Yet the consequences of XCI-skew on chronic disease risk in an unselected population have largely been unexplored.

Clonal expansion of haematopoietic stem cells is also measurable by somatic mutations shared across blood cells, indicating a common stem cell precursor (Xie et al., 2014). Clonal haematopoiesis of indeterminate potential (CHIP) is a cellular phenotype describing a pre-malignant state in which ≥4% of blood cells harbour the same somatic mutation (Jaiswal and Ebert, 2019), thus representing monoclonal expansion. CHIP is robustly associated with all-cause mortality (Jaiswal et al., 2014), haematological cancers (Genovese et al., 2014), and cardiovascular disease (Jaiswal et al., 2017). XCI-skew can sometimes be a marker of CHIP: XCI-skew was previously used to determine clonality (Busque and Gilliland, 1998), and exome sequencing of females with XCI-skew identified TET2 mutations as drivers of pre-malignant clonal haematopoiesis. However, XCI-skew and CHIP are not completely mutually inclusive (Busque et al., 2012).

Given XCI-skew is potentially tagging changes to the haematopoietic stem cell pool, we hypothesised that XCI-skew may be a marker of biological ageing and a risk factor for chronic disease. We tested this hypothesis by assaying XCI-skew in 1575 females from the TwinsUK cohort and employed prospective, cross-sectional, and intra-twin study designs to characterise the relationship of XCI with molecular and cellular measures of ageing, cardiovascular disease risk, and cancer diagnoses.

Methods

TwinsUK cohort

Archival DNA samples derived from whole blood (collected 1997–2017) were selected from individuals of the TwinsUK population cohort (Verdi et al., 2019). Twin pairs were date matched and the final dataset of 1575 samples comprised 423 monozygotic (MZ) twin pairs (nindividuals = 846), 257 dizygotic (DZ) twin pairs (nindividuals = 514), and 215 singletons (Figure 1—figure supplement 2). The age range of the XCI cohort is 19–99, with a median age of 61 (Figure 2A). All samples and information were collected with written and signed informed consent, including consent to publish within the TwinsUK study. TwinsUK has received ethical approval associated with TwinsUK Biobank (19/NW/0187), TwinsUK (EC04/015) or Healthy Ageing Twin Study (HATS) (07 /H0802/84) studies from NHS Research Ethics Service Committees London – Westminster.

Human Androgen Receptor Assay (HUMARA)

The HUMARA method combines methylation-sensitive restriction enzyme digest and amplification of a highly polymorphic (CAG)n repeat in the first exon of the X-linked AR gene, allowing for the differentiation of the active and inactive chromosomes in heterozygous individuals (Cutler Allen et al., 1992). Here, 625 ng of genomic DNA was divided into three aliquots and incubated for 30 min at 37 °C with (i) the methylation-sensitive enzyme HpaII, (ii) the methylation-insensitive enzyme MspI, or (iii) water (mock digest) in 1 × New England Biolabs CutSmart Buffer. The HpaII digest was followed by an additional 20 min at 80 °C to avoid residual enzymatic activity. Fluorescently labelled PCR primers (FAM, VIC, NED, or PET; Forward primer 5’-dye-GCTGTGAAGGTTGCTGTTCCTCAT-3’, Reverse primer 5’-TCCAGAATCTGTTCCAGAGCGTGC-3’) were used in New England BioLabs One Taq Master Mix to amplify 1.5 μl of digested PCR product. The Mock and HpaII digested DNA were amplified in triplicate (using FAM, VIC, and NED), and the MspI digest, used as control of digestion efficiency, was amplified once (using PET). All PCRs were amplified with an initial denaturation step at 94 °C for 5 min, followed by 30 cycles of 94 °C for 30 s, 60 °C for 1 min, and 72 °C for 2 min, and a final elongation step of 72 °C for 15 min. To minimize technical bias and batch effects, the labelled amplified products were diluted 1:15 with nuclease-free ddH2O and pooled together with the GeneScan 500 LIZ size standard and analysed on an ABI 3730xl. Twin pairs were assayed on the same plate and plates contained a mix of both MZ and DZ pairs. Two replicates were included on each plate and a within-plate correlation of 0.99 was measured. A total of 2382 DNA samples were assayed, with 194 failed samples, and 601 samples were homozygous for the CAG repeat and were therefore uninformative (Figure 1—figure supplement 2).

Calculation of XCI

Data were analysed using the Microsatellite Analysis Software available on the Thermo Fisher Cloud. The XCI status was calculated in each of the triplicates as follows:

  • Allele Ratio Mock Digestion (Rm)=allele 1 peak height / allele 2 peak height

  • Allele Ratio HpaII Digestion (Rh)=allele 1 peak height / allele 2 peak height

  • Normalized Ratio (Rn)=Rh/Rm

  • XCI percentage = [Rn/(Rn +1)] * 100

A coefficient of variation (CV) was calculated across the triplicates and samples with CV >0.15 were excluded from downstream analysis (n=12; Figure 1—figure supplement 2). A mean XCI percentage (0–100%) was calculated for each sample, where 50% is perfectly balanced XCI and the directionality of XCI away from 50% is uninformative (e.g., both 0% and 100% are considered equal). Therefore, the XCI values are collapsed to a range of 50–100% when XCI is the dependent variable in analyses.

XCI-skew categorical variable

The XCI percentage data were normalised, and a categorical XCI-skew variable was created from the absolute values of the normalised distribution as follows: standard deviation (SD)<1 from the mean = random XCI (0); 1≤SD<2 = skewed XCI (1); and SD≥2 = extreme skew (2). As such, XCI-skew equated to ≥75% XCI, and extreme XCI-skew equated to ≥91% XCI (Figure 1—figure supplement 1). These thresholds are very similar to previous studies (Busque et al., 1996; Gale et al., 1997), and allowed for linear associations to be tested using the XCI-skew categorical variable.

Statistical analysis

In all regression models, a linear mixed effects model was used with relatedness and family structure fitted with a random intercept using the lme4 package (Bates et al., 2015). The relevant fixed effects are described in each section below and were specific to each test. All analyses were carried out using R version 4.1.1, and all plots were generated using ggplot (Wickham, 2016).

Chronological and biological ageing

Datasets were matched to be within 1 year of the XCI DNA sample and the significance threshold after Bonferroni correction was p<0.007 to account for multiple testing across the seven tests in this section. Chronological age was calculated at time of DNA sampling and the association was tested using XCI as the dependent variable. Body Mass Index (BMI) measures were taken during clinical visits and obesity was defined as BMI ≥30. Smoking status was classified based on longitudinal questionnaire answers (Christiansen et al., 2021). The associations were tested with XCI as the dependent variable and obesity (obese/not obese) or smoking (ever/never smoker) as the independent variable, controlling for age as a fixed effect. A Frailty Index (Searle et al., 2008) was calculated based on longitudinal questionnaire data and used as the dependent variable, with XCI-skew as the independent variable and age and BMI as fixed effects. Leukocyte Telomere Length was measured using qPCR as previously described (Codd et al., 2010), and the normalised measures were used as the dependent variable and XCI-skew as the independent variable, with age and smoking as fixed effects. DNA methylation (DNAm) GrimAge was calculated using 450 K methylation data and GrimAge epigenetic age acceleration measures were obtained from regressing epigenetic age on chronological age (Costeira et al., 2021). GrimAge Acceleration was used as the dependent variable and XCI-skew as the independent variable, with no additional covariates included in the model.

Whole blood count data

Automated whole blood count data were date-matched to the XCI DNA sample, and each of the 10 blood count variables was normalised. The significance threshold after Bonferroni correction was p<0.005 to account for multiple testing across the 10 tests. In addition, Monocyte-to-Lymphocyte Ratio (MLR) and Neutrophil-to-Lymphocyte Ratio (NLR) were calculated by dividing the total monocyte or neutrophil count, respectively, by total lymphocyte count, and a Bonferroni-corrected significance threshold of P<0.025. Associations were tested with XCI-skew as an independent variable after controlling for age, BMI, seasonality, and smoking status as fixed effects in a linear mixed effects model.

Cytokines levels and C-reactive protein

Serum IL-1β, IL-10, IL-6, and TNF were measured simultaneously using the bead-based high sensitivity human cytokine kit (HSCYTO-60SK, Linco-Millipore) according to the manufacturers’ instructions. CRP concentrations from serum were measured with the Human Cardiovascular Disease Panel 2 LINCOplex Kit (HCVD2-67BK, Linco-Millipore) and with the Extracellular Protein Buffer Reagent Kit (LHB0001, Invitrogen). CRP concentrations were diluted 1:2000 prior to analysis and assayed in duplicate, as previously described (Ligthart et al., 2018). Date-matched data with batch effects regressed out were normalised and associations were assessed using linear mixed models controlling for seasonality and age as fixed effects. The significance threshold after Bonferroni correction was p<0.01 to account for multiple testing across five tests.

Atherosclerosis and cardiovascular disease (ASCVD) risk score

The ASCVD risk score was calculated for a subset of 228 individuals with date-matched data on age, total cholesterol, HDL cholesterol, smoking, diabetes, systolic blood pressure, and hypertension medication, as previously described (Goff et al., 2014). A linear mixed effects model was used to control for BMI and monocyte abundance as fixed effects (Madjid et al., 2004). Twin pairs discordant for XCI-skew but matched for date of visit and age (n=34 pairs) were used for the intra-twin study, and ASCVD risk scores were compared using a one-sided paired samples Wilcoxon test.

Cancer and all-cause mortality

Anonymised data were obtained from the National Disease Registration Service. Study entry was the date of DNA sampling and follow-up occurred through to January 2020 (study end date). For the cancer analysis, individuals who had not experienced the event were censored at 10 years, study end date, or date-of-death. All participants with a history of cancer before sampling, or within 6 months of sampling, were excluded from analyses, and reports of non-melanoma skin cancer were filtered, leaving a sample size of 1417. For all-cause mortality, individuals who had not experienced the event were censored at 10 years or study end date. For both analyses, age, relatedness, and zygosity were controlled for in the Cox regression model using R package Survival (Therneau, 2021). Proportional hazards assumptions were assessed using the cox.zph function of the Survival package. Kaplan-Meier plots were used for graphical representation of years until diagnosis. XCI-skew and extreme XCI-skew groups were combined due to the limited number of events.

Results

XCI-skew in the TwinsUK population cohort

We measured XCI in DNA derived from whole blood using the methylation-sensitive PCR-based Human Androgen Receptor Assay (HUMARA), which differentiates between alleles from the active and inactive X (Cutler Allen et al., 1992; Hatakeyama et al., 2004). HUMARA is an extensively used assay which correlates well with transcription-based methods (Bolduc et al., 2008; Zito et al., 2019). The output of the HUMARA assay is a continuous XCI variable from 0–100%, where 50% is perfectly balanced XCI and the directionality of XCI away from 50% is uninformative (e.g., both 0% and 100% are considered equal). We normalised the distribution of the continuous XCI values across the cohort, and defined XCI-skew as measures 1 SD from the mean, corresponding to XCI score ≥75%, and extreme XCI-skew as measures 2 SD from the mean, corresponding to XCI score ≥91% (Figure 1—figure supplement 1). Of note, these values are extremely similar to thresholds previously used in the literature (Bolduc et al., 2008; Busque and Gilliland, 1998). Figure 1 shows representative samples of how the HUMARA assay relates to the three categorical XCI-skew statuses.

Figure 1 with 2 supplements see all
Measuring XCI with the HUMARA assay.

The Human Androgen Receptor Assay (HUMARA) uses methylation-sensitive restriction enzyme digest and PCR to measure skewed X-inactivation. The assay estimates XCI-skew by comparing the relative abundance of allele specific fragments from a mock digest to a methylation-sensitive HpaII digest in which only the alleles from the inactive X are amplified. Representative examples are displayed of fragment analysis of the PCR products for samples with random XCI (top), skewed XCI (middle), and extreme skewed XCI (bottom). The x-axis shows the size, and the y-axis represents the abundance, of the PCR products, respectively. The left panel shows the PCR products after a mock digest with water, resulting in amplification of both alleles regardless of chromosomal inactivation. The right panel shows the PCR products after a restriction enzyme digest with methylation-sensitive enzyme HpaII, resulting in amplification of only the alleles deriving from inactive chromosomes. For each sample, the ratio of the HpaII digested allele products (Rh = allele 1/allele 2) is divided by the ratio of the Mock digest allele products (Rm = allele 1/allele 2) to create a Normalized Ratio (Rn). The XCI percentage is then calculated using the formula [Rn/(Rn +1)] * 100. Images were generated using the Microsatellite Analysis Software on the Thermo Fisher Cloud.

We defined XCI-skew in 1575 participants (median age = 61; Figure 2A) unselected for chronic disease status from the TwinsUK population cohort (Verdi et al., 2019), which comprised of 423 MZ pairs, 257 DZ pairs, and 215 singletons. In line with previous studies (Vickers et al., 2001), we see increased concordance of XCI-skew within MZ twin pairs compared with DZ twin pairs: 27% of MZ twin pairs (114/423), and 45.5% of DZ twin pairs (117/257), were discordant for their categorical XCI-skew status. We date-matched the XCI data with existing phenotypes from TwinsUK (e.g., blood count data, molecular markers) in the subsets of individuals on whom each phenotype was available (Figure 1—figure supplement 2).

Age acquired XCI-skew across age groups and time.

(A) A histogram displaying the age distribution of the TwinsUK HUMARA cohort (age range: 19–99; median age = 61). (B) The proportions of individuals (y-axis) in each of three XCI-skew categories across increasing age groups (x-axis) are shown (N=1575). (C) A Sankey plot shows the longitudinal changes to XCI in 31 individuals across two measures 15–17 years apart. Colours indicate XCI at visit 1, axis 1 displays the age group of individuals at visit 1, and axis 2 displays XCI at visit 2.

Figure 2—source data 1

XCI-skew across age groups.

Categorical variable of XCI-skew and corresponding age group of each individual in the study (n=1575).

https://cdn.elifesciences.org/articles/78263/elife-78263-fig2-data1-v1.txt

Cross-sectional and longitudinal changes to XCI-skew with age

We assessed changes in frequency of XCI-skew across increasing age groups and identified 12% (9 of 75) of individuals under 40 years old (yrs) displaying XCI-skew (≥75% XCI); 28% (183 of 652) of 40–59 yrs; 37% (185 of 498) of 60–69 yrs; and 44% (132 of 303) of those over 70 yrs (Figure 2B). Proportions of individuals displaying extreme XCI-skew (≥91% XCI) remains consistent at ~3–4% below the age of 60 but increases to 7% of 60–69 yrs and 9% of those over 70 yrs. These results suggest a stepwise increase in prevalence of XCI-skew happening after 40 years of age, then again after 60 years of age, where we also see the first increase in prevalence of extreme XCI-skew (Figure 2B). As expected, after controlling for relatedness and zygosity, we find a significant positive association between age and XCI skewing (p=2.8 × 10–9, N=1575). This result replicates the many existing studies on age acquired XCI-skew and acts as a validation of the TwinsUK HUMARA dataset (Busque et al., 1996; Gale et al., 1997; Hatakeyama et al., 2004; Zito et al., 2019).

We assessed change in XCI-skew over time using 31 individuals on whom we had an additional second sample available from 15 to 17 years prior to the main study (Figure 2C. median age at visit 1=55.5; median age at visit 2=72.1). The two individuals who had extreme XCI-skew at visit 1 still displayed extreme XCI-skew at visit 2. Of the eight individuals who had XCI-skew at visit 1, seven remained skewed and one progressed to extreme XCI-skew at visit 2. Of the 21 who had a random pattern of XCI at visit 1, 15 (71.4%) remained the same, and 6 (28.6%) progressed to XCI-skew at visit 2. These longitudinal data indicate that XCI-skew categorisation persists over extended periods of time and increases over the life course.

XCI-skew is independent of known markers of biological ageing

Ageing is a heterogenous process in which an individual’s biological age can differ from their chronological age. Smoking and obesity are risk factors for accelerated ageing (Tam et al., 2020). Accelerated ageing can be estimated through measures of frailty (Clegg et al., 2013; Chen et al., 2014), and on a molecular level using measures of leukocyte telomere length shortening (Blackburn et al., 2006) and epigenetic ageing clocks such as DNA methylation (DNAm) GrimAge (Lu et al., 2019); all these measures are associated with adverse health outcomes (Blackburn et al., 2006; Hewitt et al., 2020; Lu et al., 2019).

Given the robust association with chronological age, we sought to establish whether XCI-skew was associated with biological ageing using measures taken within 1 year of the XCI DNA sample and using linear regression mixed effects models, controlling for relatedness and zygosity as random effects (Figure 3). We observed no association with smoking status (p=0.33, Nnever_smoker=879; Never_smoker = 673; Figure 3A), nor obesity (p=0.88, Nnot_obese=726; Nobese = 165; Figure 3B), after correcting for age. We also found no association with a robust frailty index (p=0.59, Nranodm XCI=398; nSkewed XCI = 177; Nextreme skew = 36; Figure 3C) after correcting for age and BMI, nor with leukocyte telomere length shortening (p=0.9, Nranodm XCI=278; nSkewed XCI = 103; Nextreme skew = 16; Figure 3D), after correcting for age and smoking status. Finally, we see no association with DNAm GrimAge acceleration (p=0.22, Nranodm XCI=101; nSkewed XCI = 30; Nextreme skew = 6; Figure 3E), however, we believe the relationship between XCI-skew and epigenetic ageing would benefit from a follow-up study with a larger sample size particularly given the limited number of samples with extreme XCI-skew in our analysis. Together, these data, ranging from the molecular to organismal level, suggest age acquired XCI-skew is independent of many known markers of biological ageing and is potentially a unique biomarker with unexplored utility.

Figure 3 with 1 supplement see all
Age acquired XCI-skew and markers of biological ageing and blood cell counts.

Box plots representing the results of the linear regression mixed effects models to assess (A) Smoking status (p=0.33, Nnever_smoker = 879; Never_smoker = 673) and (B) obesity (p=0.88, Nnot_obese = 726; Nobese = 165) after correcting for age with XCI as the dependent variable, and (C) frailty index (p=0.59, Nranodm XCI = 398; nSkewed XCI = 177; Nextreme skew = 36), after correcting for age and BMI, (D) Leukocyte telomere length shortening (p=0.9, Nranodm XCI = 278; nSkewed XCI = 103; Nextreme skew = 16) after correcting for age and smoking status, and (E) DNAm GrimAge acceleration (p=0.22, Nranodm XCI = 101; nSkewed XCI = 30; Nextreme skew = 6), with XCI-skew as the dependent variable. All boxplots display the median and IQR, and have the residuals of the models on the y-axis. (F) A forest plot of associations with data-matched Complete Blood Count data (top panel) and Myeloid-to-lymphoid ratios (bottom) with effect size and lower and upper confidence intervals indicated. Associations were tested with XCI-skew as an independent variable after controlling for age, BMI, seasonality, and smoking status as fixed effects in a linear mixed effects model (Nranodm XCI = 445; nSkewed XCI = 183; Nextreme skew = 43). The significance threshold after Bonferroni correction was p<0.005 and p<0.023 to account for multiple testing across the 10 tests and 2 tests, respectively.

Figure 3—source data 1

XCI-skew and DNAm GrimAge Acceleration.

Normalised DNAm GrimAge Acceleration variable and categorical XCI-skew data (n=137).

https://cdn.elifesciences.org/articles/78263/elife-78263-fig3-data1-v1.txt

XCI-skew is associated with increased monocyte abundance and decreased IL-10 levels

Changes in blood cell composition can be indicative of ill health or systemic inflammation (Kabat et al., 2017; Madjid et al., 2004; Patel et al., 2009), and a haematopoietic stem cell bias towards the myeloid lineage is observed with ageing (Pang et al., 2011). We tested for associations between XCI-skew and whole blood count data, including white cell differentials, in a subset of individuals with matched data (Nranodm XCI = 445; nSkewed XCI = 183; Nextreme skew = 43, median age = 63). After controlling for age, seasonality, BMI, smoking, relatedness and zygosity in a linear regression mixed effects model, XCI-skew is association with increased monocyte abundance after multiple testing correction (p=0.0038), and we observed nominal increases in abundance across other myeloid cells (Figure 3F). We next tested the hypothesis that XCI-skew was associated with a myeloid lineage bias using the Monocyte-to-Lymphocyte Ratio (MLR) (Chen et al., 2019) and Neutrophil-to-Lymphocyte Ratio (NLR) (Arbel et al., 2012), and detect an association between XCI-skew and MLR (p=0.019), and a nominal association with NLR (p=0.042) (Figure 3F). Though myeloid cells show a greater degree of skewing, thought to be due to the shorter lifespan of these cells (Gale et al., 1997), we do not believe the association seen here between XCI-skew and increasing monocyte numbers is causal: XCI-skew is defined as ≥25% shift in cell mosaicism, whereas monocytes account for only ~10% of white blood cells.

‘Inflammageing’ is the chronic pro-inflammatory phenotype observed in ageing and is considered an altered state of intercellular communication (López-Otín et al., 2013). Markers of inflammageing include cytokines produced by immune cells and C-reactive protein (CRP) produced by liver cells (Salminen et al., 2012). We date-matched the XCI data with serum levels of CRP (Nranodm XCI = 121; nSkewed XCI = 38; Nextreme skew = 6) and a more modest cytokine dataset of interleukin (IL)–6, IL-1B, IL-10, and TNF (Nranodm XCI = 23; nSkewed XCI = 4) and used linear regression mixed effects models to control for age, seasonality, relatedness and zygosity (Figure 3—figure supplement 1). We see no association with primary markers of inflammageing CRP (p=0.41), IL-6 (p=0.41), or TNF (p=0.61), though a nominal association with IL-1β is observed (p=0.02) which does not pass multiple correction, but warrants follow up analysis with a larger sample size. However, we observe a strong negative association with IL-10 (p=0.0008, Figure 3—figure supplement 1). IL-10 is a broadly expressed anti-inflammatory cytokine which can inhibit the proinflammatory responses of both innate and adaptive immune cells (Saraiva and O’Garra, 2010). Though we note that due to minimal overlap in the datasets, we were unable to control for cell type composition, which may impact these findings.

Atherosclerotic cardiovascular disease risk is increased in individuals with XCI-skew

Cardiovascular disease (CVD) is the leading cause of death worldwide, and monocytes are an innate immune cell type known to be mediators in CVD disease progression and are found in atherosclerotic lesions (Libby et al., 2016). Given the association of XCI-skew with increased monocyte abundance, and the association of CHIP with CVD, we hypothesised that XCI-skew could also be associated with CVD. The atherosclerotic cardiovascular disease (ASCVD) risk score (Goff et al., 2014) (see Methods) captures traditional risk factors and gives a predicted risk of a major CVD event in the next 10 years, with an ASCVD risk score >7.5% representing intermediate risk, and an ASCVD risk score >20% representing high risk. In a cross-sectional study of 228 individuals (Nranodm XCI = 155; nSkewed XCI = 56; Nextreme skew = 17; median age = 62) with matched health data available, XCI-skew was associated with increased ASCVD risk score after controlling for BMI, monocyte abundance, relatedness and zygosity using a linear regression mixed effects model (p=0.01, Figure 4A). 23.5% (4 of 17) and 35.3% (6 of 17) of individuals with extreme XCI-skew have high and intermediate ASCVD risk, respectively, compared to 4.5% (7 of 155) and 21.9% (34 of 155) of individuals with random XCI.

Age acquired XCI-skew and cardiovascular disease risk score.

(A) Using a linear regression mixed effects model to control for BMI, monocyte count, relatedness and zygosity, XCI-skew is associated with increased atherosclerotic cardiovascular disease risk score (Nranodm XCI = 155; nSkewed XCI = 56; Nextreme skew = 17; p=0.01). The boxplot displays the median and IQR. (B) Using age-matched twin pairs discordant for their XCI-skew status (Npairs = 34), XCI-skew is associated with increased atherosclerotic cardiovascular disease risk score in the intra-twin analysis (one-sided paired samples Wilcoxon test; p=0.025).

Figure 4—source data 1

XCI-skew and ASCVD Risk Score.

ASCVD Risk Score, log ASCVD Risk Score, and categorical XCI-skew data (n=228).

https://cdn.elifesciences.org/articles/78263/elife-78263-fig4-data1-v1.txt

To ensure the observed association wasn’t spuriously driven by the age component of the ASCVD risk score (see Methods), we took age matched MZ and DZ twin pairs (npairs = 34) discordant for their XCI-skew status and tested intra-twin ASCVD risk scores. The intra-twin analysis validated the association between higher XCI-skew and increased ASCVD risk score (one-sided paired samples Wilcoxon test p=0.025; Figure 4B), adding further support to the finding.

XCI-skew is predictive of future cancer diagnosis in 10-year follow-up

The association of XCI-skew with cancer has largely been assessed in case-control studies focusing on cancers of female reproductive organs, with limited replicated findings (Buller et al., 1999; Kristiansen et al., 2002; Struewing et al., 2006). We conducted a prospective 10 year follow-up study (median follow-up time 5.65 years) from time of DNA sampling in 1417 individuals (Nrandom = 948, Nskewed = 469, median age = 60) who were cancer-free at baseline to assess the association between XCI-skew and future cancer diagnoses (cancer events = 58; Supplementary file 1). Each cancer event represents the first cancer diagnosis (excluding non-melanoma skin cancer) in each individual and subsequent diagnoses are not included. Using multivariate Cox regression analysis controlling for age, relatedness and zygosity, XCI-skew was associated with increased probability of cancer diagnosis (p=0.012; Hazard Ratio (HR)=1.95 (95% Confidence Interval (CI)=1.16–3.28); Figure 5). Though we were underpowered to run associations for specific cancers, we were interested to explore the potential that blood cancers were partly driving the association given the robust association of CHIP with haematological cancers. However, running the analysis without the Haematopoietic/Lymphoid Tissue cancers (see Supplementary file 1) strengthened the observed association (p=0.009; HR = 2.04 [1.20–3.49]) suggesting a crucial difference in cancer risk between XCI-skew and CHIP.

Prospective study of XCI-skew and future cancer diagnosis.

Using a Cox regression analysis and controlling for age, relatedness and zygosity, XCI-skew is predictive of future cancer incidence in a 10 year follow-up (p=0.012; HR = 1.95). Each cancer event represents the first cancer diagnosis (excluding non-melanoma skin cancer) in each individual. A Kaplan-Meier plot (top) and cumulative events (bottom) of cancer diagnosis in individuals with XCI-skew (N=469) and random XCI (N=948) are shown. 2.9% (28/948) of individuals with random XCI, and 6.4% (30/469) of individuals with XCI-skew, developed cancer in the 10 years follow-up.

It has been suggested that XCI-skew could have a protective effect against all-cause mortality in cohorts selected for longevity (Mengel-From et al., 2012). However, in our 10 year follow-up study in a population cohort (median follow-up time = 5.71 years; deaths = 41), we see a non-statistically significant trend towards a positive association with all-cause mortality (p=0.29; HR = 1.39 [0.75–2.58]).

Discussion

The age association of XCI-skew in blood tissue has long been established, with increased prevalence of females displaying XCI-skew after middle age, which is a critical time for the development of chronic disease (Busque et al., 1996; Gale et al., 1997; Zito et al., 2019). Although XCI-skew is measurable in a sex-specific manner, age acquired XCI-skew is potentially a marker of stem cell depletion or clonal expansion of haematopoietic stem cells (Busque et al., 1996; Busque et al., 2012; Gale et al., 1997; van Dijk et al., 2002), and could therefore have a role in age-related chronic disease, which has been robustly established for CHIP (Jaiswal and Ebert, 2019). In our cross-sectional study of 1575 females, we identify associations of blood-derived age acquired XCI-skew with an increased atherosclerotic cardiovascular disease risk score and increased probability of future cancer diagnosis.

Intriguingly, XCI-skew appears independent of other known markers of biological ageing, including leukocyte telomere length shortening, inflammageing, and a robust measure of frailty, making XCI-skew a unique biomarker with clinical potential. We note that though we also see no association with the mDNA GrimAge measure of accelerated epigenetic ageing, our sample size was limited, and this perhaps warrants a follow-up study to better establish the lack of correlation. Importantly, we also demonstrate that smoking and obesity are not risk factors for age acquired XCI-skew, and with a limited longitudinal dataset, that XCI-skew persists and increases over extended periods of time (median 16 years).

We find an association with increased probability of future cancer diagnosis in a 10 year follow-up study. Previous studies assessing the role of XCI-skew in cancer have typically used case-control studies of breast and ovarian cancers and have not been robustly replicated. The lack of replication is potentially explained by the heterogeneity in study design, including age-of-onset and therapy timing (Kristiansen et al., 2002; Struewing et al., 2006), BRCA1 mutation stratification (Kristiansen et al., 2005; Lose et al., 2008; Manoukian et al., 2013), and threshold level for definition of XCI-skew (Buller et al., 1999). We controlled for potential confounding effects of cancer treatment on blood cells by excluding all individuals with a cancer diagnosis prior to study entry. Due to limited cancer events, our Cox regression analysis combined all individuals with XCI-skew (≥75% XCI) and extreme XCI-skew (≥91% XCI), which suggests even modest levels of skewing represents a greater risk of cancer.

Though CHIP is also predictive of future cancer diagnosis, the association is limited to haematological cancers (Desai et al., 2018). Whereas here we see blood-derived measures of XCI-skew is predictive of all future cancer diagnoses, even when the haematological cancers are removed from the analysis. Follow-up studies to assess the risk of cancer of specific tissues in individuals with XCI-skew are needed, but we hypothesise that the relationship between XCI-skew measures in blood tissue and cancer will not be causal. Instead XCI-skew is likely a marker of chronic inflammation, which can predispose to the development of cancer through increased mutagenesis and can promote tumorigenesis by shaping the tumour microenvironment to stimulate tumour growth (Greten and Grivennikov, 2019). However, it is interesting that common environmental factors that induce chronic inflammation, such as smoking and obesity, have not been found to be risk factors for XCI-skew in our study.

We also present an association with increased atherosclerotic cardiovascular disease risk score, which captures traditional risk factors (age, total cholesterol, HDL cholesterol, smoking, diabetes, systolic blood pressure, and hypertension medication) and estimates the risk of developing CVD in the next 10 years (Goff et al., 2014). We were also able to utilise the powerful discordant twin design to validate this finding, thus excluding the possibility that the cross-sectional association was driven by the age component of the ASCVD risk score. Following our finding of a correlation between blood-derived XCI-skew and risk of CVD, ascertaining whether XCI-skew is associated with incident CVD is of upmost importance. CHIP has been robustly associated with incident CVD and this association is thought to be causal (Jaiswal et al., 2017). Furthermore, the association of CHIP and all-cause mortality is partly driven by CVD (Jaiswal et al., 2014). We lacked statistical power to demonstrate an association with all-cause mortality, though the trend was towards greater risk (HR = 1.39 [0.75–2.58]). A higher-powered study able to focus on specific causes of mortality will be of great interest and importance.

Given the existing links between CHIP and XCI-skew as two age acquired blood traits, and that CHIP mutations can be found in individuals with XCI-skew (Busque et al., 2012), could the association between XCI-skew and CVD risk be partly driven by CHIP? It is expected that some of the individuals with XCI-skew in our study will also harbour CHIP mutations. However, a better understanding of the co-occurrence of XCI-skew and CHIP within individuals, and the mutational burden within individuals with and without XCI-skew, is an important area of future work. Whether the co-occurrence of XCI-skew and CHIP represents an amplified risk of disease, or helps better define risk categories, will be of clinical significance to establish. As the inherited genetic risk of CHIP is fast becoming better understood, deciphering the genetic predisposition to XCI-skew will also enable the assessment of potential shared genetic susceptibility between the two traits (Kar et al., 2022). However, many of the negative results in our study are crucial findings given they expose differences between the risk factors of XCI-skew and CHIP. Notably, recent work has demonstrated smoking and telomere length are causal risk factors for CHIP (Kar et al., 2022), whereas we see no association with either trait in our study, suggesting CHIP and XCI-skew have some distinct aetiologies.

On a cellular level, we observe that increased abundance of monocytes correlates with increased XCI-skew, which is of particular interest given that monocytes/macrophages are involved in the inflammatory pathophysiology of CVD (Libby et al., 2016). It is important to note however that changes in monocyte counts alone do not explain the observed levels of XCI-skew. Monocytes account for ~10% of white blood cells whereas XCI-skew is defined here as ≥25% shift in cell mosaicism. Furthermore, age acquired XCI-skew has previously been shown across isolated neutrophils, monocytes, and T cells, with correlations between these fractions (Tonon et al., 1998; van Dijk et al., 2002), albeit with lower levels of skewing in lymphoid cells (Gale et al., 1997). Instead, XCI-skew is likely associated with the age-related haematopoietic bias toward the myeloid lineage (Pang et al., 2011), as we see nominal associations across other myeloid cell types in addition to the monocyte- and neutrophil-to-lymphocyte ratios. Our study also demonstrates an association of reduced levels of IL-10 with increased XCI-skew. IL-10 is an anti-inflammatory cytokine produced by a broad range of immune cells (Saraiva and O’Garra, 2010) and subsets of monocytes differ in their capacities to secrete IL-10 (Skrzeczyńska-Moncznik et al., 2008). Follow-up work on inflammatory profiles linked to XCI-skew may reveal mechanistic insights.

We derived our threshold for XCI-skew from the normalised distribution of the continuous XCI values across the cohort, and defined XCI-skew as measures 1 SD from the mean, corresponding to XCI score ≥75%, and extreme XCI-skew as measures 2 SD from the mean, corresponding to XCI score ≥91%. These values are very similar to thresholds used in the literature (Busque et al., 1996; Gale et al., 1997; Zito et al., 2019), but allow us to test for linear associations across the increasing thresholds of XCI-skew and demonstrate that individuals with lower levels of XCI-skew, which affects >37% of females over 60, are still at elevated risk of ASCVD and future cancer diagnoses. However, to avoid Type I error inflation, we did not run additional tests to compare the extreme XCI-skew group with the random XCI group.

There are some limitations to our study. Despite the high sample size of individuals with measured XCI, analyses are carried out in subsets of these individuals where date-matched phenotype data were available. In particular, the cytokine analyses have a low sample size and though we detect significant associations with IL-10, we may be underpowered to rule out the possibility of detecting a weaker association with other cytokines, particularly IL-1β. We were also unable to control for cell type composition in the cytokine analyses due to minimal overlap with the date-matched whole blood count data, which may have a significant impact on these findings. Though most of the measures of biological ageing were well-powered, the mDNA GrimAge analysis may benefit from replication in a larger sample size to more robustly establish the relationship between XCI-skew and accelerated epigenetic ageing. Also, with only 41 deaths across the cohort, our all-cause mortality analysis is underpowered, and a study able to focus on specific causes of mortality is warranted.

In summary, we demonstrate XCI-skew is a highly prevalent cellular phenotype in females and is associated with elevated cardiovascular disease risk and predictive of future cancer incidence. Further investigations are needed to translate the biological value of XCI-skew into clinical applications for studying the association of age-related haematopoietic changes and chronic disease associations, regardless of chromosomal sex. Understanding the mechanisms underlying this phenomenon, whether XCI-skew is reflective of other ageing markers that increase disease risk, and whether it is therapeutically actionable, are areas of particular interest.

Data availability

Source data are provided for Figure 2B, Figure 3E and Figure 4A. All data in the manuscript have been deposited to the TwinsUK BioResource data management team and are available by application to the Twin Research Executive Access committee (TREC) at King's College London. The TwinsUK BioResource is managed by TREC, which provides governance of access to TwinsUK data and samples. This excludes the National Disease Registration Service data which are only available to internal departmental members who are ONS accredited due to the terms of the data linkage. TwinsUK data users are bound by data sharing agreement set out in the data access application form (https://twinsuk.ac.uk/wp-content/uploads/2018/11/DTR_DataAccess_Policy_0318.pdf). This includes responsibilities with respect to third party data sharing and maintaining participant privacy. Further responsibilities include a responsibility to acknowledge data sharing.

References

    1. Cutler Allen R
    2. Zoghbi HY
    3. Moseley AB
    4. Rosenblatt HM
    5. Belmont JW
    (1992)
    Methylation of hpaii and hhai sites near the polymorphic CAG repeat in the human androgen-receptor gene correlates with X chromosome inactivation
    American Journal of Human Genetics 51:1229–1239.
    1. Ligthart S
    2. Vaez A
    3. Võsa U
    4. Stathopoulou MG
    5. de Vries PS
    6. Prins BP
    7. Van der Most PJ
    8. Tanaka T
    9. Naderi E
    10. Rose LM
    11. Wu Y
    12. Karlsson R
    13. Barbalic M
    14. Lin H
    15. Pool R
    16. Zhu G
    17. Macé A
    18. Sidore C
    19. Trompet S
    20. Mangino M
    21. Sabater-Lleal M
    22. Kemp JP
    23. Abbasi A
    24. Kacprowski T
    25. Verweij N
    26. Smith AV
    27. Huang T
    28. Marzi C
    29. Feitosa MF
    30. Lohman KK
    31. Kleber ME
    32. Milaneschi Y
    33. Mueller C
    34. Huq M
    35. Vlachopoulou E
    36. Lyytikäinen LP
    37. Oldmeadow C
    38. Deelen J
    39. Perola M
    40. Zhao JH
    41. Feenstra B
    42. Amini M
    43. Lahti J
    44. Schraut KE
    45. Fornage M
    46. Suktitipat B
    47. Chen WM
    48. Li X
    49. Nutile T
    50. Malerba G
    51. Luan J
    52. Bak T
    53. Schork N
    54. Del Greco M. F
    55. Thiering E
    56. Mahajan A
    57. Marioni RE
    58. Mihailov E
    59. Eriksson J
    60. Ozel AB
    61. Zhang W
    62. Nethander M
    63. Cheng YC
    64. Aslibekyan S
    65. Ang W
    66. Gandin I
    67. Yengo L
    68. Portas L
    69. Kooperberg C
    70. Hofer E
    71. Rajan KB
    72. Schurmann C
    73. den Hollander W
    74. Ahluwalia TS
    75. Zhao J
    76. Draisma HHM
    77. Ford I
    78. Timpson N
    79. Teumer A
    80. Huang H
    81. Wahl S
    82. Liu Y
    83. Huang J
    84. Uh HW
    85. Geller F
    86. Joshi PK
    87. Yanek LR
    88. Trabetti E
    89. Lehne B
    90. Vozzi D
    91. Verbanck M
    92. Biino G
    93. Saba Y
    94. Meulenbelt I
    95. O’Connell JR
    96. Laakso M
    97. Giulianini F
    98. Magnusson PKE
    99. Ballantyne CM
    100. Hottenga JJ
    101. Montgomery GW
    102. Rivadineira F
    103. Rueedi R
    104. Steri M
    105. Herzig KH
    106. Stott DJ
    107. Menni C
    108. Frånberg M
    109. St. Pourcain B
    110. Felix SB
    111. Pers TH
    112. Bakker SJL
    113. Kraft P
    114. Peters A
    115. Vaidya D
    116. Delgado G
    117. Smit JH
    118. Großmann V
    119. Sinisalo J
    120. Seppälä I
    121. Williams SR
    122. Holliday EG
    123. Moed M
    124. Langenberg C
    125. Räikkönen K
    126. Ding J
    127. Campbell H
    128. Sale MM
    129. Chen YDI
    130. James AL
    131. Ruggiero D
    132. Soranzo N
    133. Hartman CA
    134. Smith EN
    135. Berenson GS
    136. Fuchsberger C
    137. Hernandez D
    138. Tiesler CMT
    139. Giedraitis V
    140. Liewald D
    141. Fischer K
    142. Mellström D
    143. Larsson A
    144. Wang Y
    145. Scott WR
    146. Lorentzon M
    147. Beilby J
    148. Ryan KA
    149. Pennell CE
    150. Vuckovic D
    151. Balkau B
    152. Concas MP
    153. Schmidt R
    154. Mendes de Leon CF
    155. Bottinger EP
    156. Kloppenburg M
    157. Paternoster L
    158. Boehnke M
    159. Musk AW
    160. Willemsen G
    161. Evans DM
    162. Madden PAF
    163. Kähönen M
    164. Kutalik Z
    165. Zoledziewska M
    166. Karhunen V
    167. Kritchevsky SB
    168. Sattar N
    169. Lachance G
    170. Clarke R
    171. Harris TB
    172. Raitakari OT
    173. Attia JR
    174. van Heemst D
    175. Kajantie E
    176. Sorice R
    177. Gambaro G
    178. Scott RA
    179. Hicks AA
    180. Ferrucci L
    181. Standl M
    182. Lindgren CM
    183. Starr JM
    184. Karlsson M
    185. Lind L
    186. Li JZ
    187. Chambers JC
    188. Mori TA
    189. de Geus E
    190. Heath AC
    191. Martin NG
    192. Auvinen J
    193. Buckley BM
    194. de Craen AJM
    195. Waldenberger M
    196. Strauch K
    197. Meitinger T
    198. Scott RJ
    199. McEvoy M
    200. Beekman M
    201. Bombieri C
    202. Ridker PM
    203. Mohlke KL
    204. Pedersen NL
    205. Morrison AC
    206. Boomsma DI
    207. Whitfield JB
    208. Strachan DP
    209. Hofman A
    210. Vollenweider P
    211. Cucca F
    212. Jarvelin MR
    213. Jukema JW
    214. Spector TD
    215. Hamsten A
    216. Zeller T
    217. Uitterlinden AG
    218. Nauck M
    219. Gudnason V
    220. Qi L
    221. Grallert H
    222. Borecki IB
    223. Rotter JI
    224. März W
    225. Wild PS
    226. Lokki ML
    227. Boyle M
    228. Salomaa V
    229. Melbye M
    230. Eriksson JG
    231. Wilson JF
    232. Penninx B
    233. Becker DM
    234. Worrall BB
    235. Gibson G
    236. Krauss RM
    237. Ciullo M
    238. Zaza G
    239. Wareham NJ
    240. Oldehinkel AJ
    241. Palmer LJ
    242. Murray SS
    243. Pramstaller PP
    244. Bandinelli S
    245. Heinrich J
    246. Ingelsson E
    247. Deary IJ
    248. Mägi R
    249. Vandenput L
    250. van der Harst P
    251. Desch KC
    252. Kooner JS
    253. Ohlsson C
    254. Hayward C
    255. Lehtimäki T
    256. Shuldiner AR
    257. Arnett DK
    258. Beilin LJ
    259. Robino A
    260. Froguel P
    261. Pirastu M
    262. Jess T
    263. Koenig W
    264. Loos RJF
    265. Evans DA
    266. Schmidt H
    267. Smith GD
    268. Slagboom PE
    269. Eiriksdottir G
    270. Morris AP
    271. Psaty BM
    272. Tracy RP
    273. Nolte IM
    274. Boerwinkle E
    275. Visvikis-Siest S
    276. Reiner AP
    277. Gross M
    278. Bis JC
    279. Franke L
    280. Franco OH
    281. Benjamin EJ
    282. Chasman DI
    283. Dupuis J
    284. Snieder H
    285. Dehghan A
    286. Alizadeh BZ
    287. Alizadeh BZ
    288. Boezen HM
    289. Franke L
    290. van der Harst P
    291. Navis G
    292. Rots M
    293. Snieder H
    294. Swertz M
    295. Wolffenbuttel BHR
    296. Wijmenga C
    297. Benjamin E
    298. Chasman DI
    299. Dehghan A
    300. Ahluwalia TS
    301. Meigs J
    302. Tracy R
    303. Alizadeh BZ
    304. Ligthart S
    305. Bis J
    306. Eiriksdottir G
    307. Pankratz N
    308. Gross M
    309. Rainer A
    310. Snieder H
    311. Wilson JG
    312. Psaty BM
    313. Dupuis J
    314. Prins B
    315. Vaso U
    316. Stathopoulou M
    317. Franke L
    318. Lehtimaki T
    319. Koenig W
    320. Jamshidi Y
    321. Siest S
    322. Abbasi A
    323. Uitterlinden AG
    324. Abdollahi M
    325. Schnabel R
    326. Schick UM
    327. Nolte IM
    328. Kraja A
    329. Hsu YH
    330. Tylee DS
    331. Zwicker A
    332. Uher R
    333. Davey-Smith G
    334. Morrison AC
    335. Hicks A
    336. van Duijn CM
    337. Ward-Caviness C
    338. Boerwinkle E
    339. Rotter J
    340. Rice K
    341. Lange L
    342. Perola M
    343. de Geus E
    344. Morris AP
    345. Makela KM
    346. Stacey D
    347. Eriksson J
    348. Frayling TM
    349. Slagboom EP
    (2018) Genome analyses of > 200,000 individuals identify 58 loci for chronic inflammation and highlight pathways that link inflammation and complex disorders
    The American Journal of Human Genetics 103:691–706.
    https://doi.org/10.1016/j.ajhg.2018.09.009

Decision letter

  1. Carlos Isales
    Senior and Reviewing Editor; Augusta University, United States
  2. Lambert Busque
    Reviewer; Hôpital Maisonneuve-Rosemont, Canada

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Age acquired skewed X Chromosome Inactivation is associated with adverse health outcomes in humans" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Carlos Isales as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Lambert Busque (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. Was XCI measured on whole blood, versus fraction (PMN, Mononuclear cells, buffy coat, other)?

2. Did the authors look at the correlation between skewing in MZG vs DZ twins. This is known but would be interesting to document.

3. Was there enough discordant MZG twin pairs (best controls) for the ASCV risk comparison?

4. Age is the major confounding factor in all analyses. Although the information is present in the text and methods, I think controlling for age (or age-matched) should also be indicated in all figures.

5. The hypothesis that cardiovascular and cancer risk could be driven in part by CHIP should be added to the discussion.

6. The XCI skewing data is gathered using the HUMARA test. This is a classical test to score skewing in human cells. Since it is the central method of this manuscript, illustrative examples of the data (random, skew, and extreme skew, as well as non-informative samples) should be provided. This will be helpful to the reader and of broad interest to the eLife audience, especially for (epi)geneticists.

7. When mentioning karyotype, the most common nomenclature is 46, XX, and 46, XY instead of XX,46 or XY,46 – in lines 50 and 51 and maybe elsewhere in the manuscript.

8. Figure 1B in line 154 and 157 is actually Figure 1C.

9. Table 1. Table S1, Figure S1 – It is unclear what N means. It should be clarified how many women with or without XCI-skew were analysed.

10. Please provide in the legend, the numbers of individuals analysed in the three categories in Figure 2A.

11. Please clarify the methodology performed in the sentence stated on lines 181-183: “XCI-skew was associated with increased ASCVD risk score after controlling for BMI and monocyte abundance”.

12. In Figure 3, the cumulative number of events table is misleading (e.g. at 10 years, we have 28 events in the random XCI group and 30 events in the XCI-skew group, and therefore they do not look different, while they are in the graph presenting percentages). It should be indicated the total population analysed per group. Also whether the events occur in distinct individuals or in one individual more than one time should be stated.

13. One aspect that would be important to analyse in the future is the relationship between XCI skewing and mutation burden in blood cells. For example, XCI skew could be a bystander of genetic mutation(s) that would cause clonal expansion of that population. Or XCI skew could determine the fixation of an X-linked mutation in the cell populations. These aspects could be addressed in the discussion.

14. In the general introduction, dosage compensation by XCI and XCI-skewing is not explained in sufficient detail. Several important features, such as the difference between primary and secondary XCI skewing are completely omitted, as well as a clear overview of the clinical relevance of skewed XCI. The main technology employed to assay XCI skewing is also not described. As a consequence, a general readership is likely to not fully appreciate the open questions addressed by authors, and their relevance in light of the current knowledge and available technologies.

15. The authors claim to characterize the relationship of skewed XCI with “molecular, cellular and organismal measures of ageing and cardiovascular disease risk and cancer diagnosis”. However, the correlation between XCI skewing in blood-derived DNA and increased age has been already extensively described (i.e. Bolduc et al., 2008, Busque et al., 1996, Hatakeyama et al., 2004; Zito et al., 2019, and many more). Importantly, with the exception of the assessment of XCI skewing which was obtained for >1.500 individuals, the molecular/cellular data presented by the authors is rather limited. Furthermore, when describing the association between skewed XCI and increased monocyte abundance, the authors omit to discuss that within the blood, different populations of hematopoietic cells show different degrees of XCI skewing with myeloid cells showing a high level of skewed XCI (Gale et al., 1997). The association described by the authors is therefore probably reinforcing these observations, and can likely be explained by biases in cell representation within samples. Furthermore, given the limited number of individuals tested for serum levels of CRP and cytokine interleukin, and the negative association observed for IL-10, these results are difficult to interpret.

16. The relationship between skewed XCI and future cancer diagnosis is also limited by the low numbers of diagnoses in 10-year follow-up and does not allow to take into account cancer type and tissue-specificity. In this context, the authors do not discuss (nor speculate on) how skewed XCI as measured in the blood can be a clinical marker of risk for diseases occurring in other organs. This is a very important point to consider for data interpretation, in light of the existing research addressing the degree of XCI skewing in different tissues and cell types, and not only in blood.

17. To improve the quality of their manuscript to possibly target a broader audience, the authors should more carefully discuss their findings while including relevant references that are omitted in the current version.

18. Lines 7-9.

Age acquired XCI-skew describes the preferential inactivation of one X chromosome across a tissue, which is particularly prevalent in blood tissues of ageing females, and yet its clinical consequences are unknown.

This sentence is misleading. The preferential inactivation of one of the two X chromosomes only occurs at the onset of XCI and results in primary skewed XCI, which influences the choice of which allele becomes inactivated. Age-acquired XCI skewing is a result of post-XCI selection. The authors should include a clear description of the differences between primary and secondary XCI skewing in the introduction, and they should also describe their relevance to human diseases.

19. Line 48-49

X chromosome inactivation (XCI) evolved in placental mammals to compensate for the differences in size between the X and Y sex chromosomes.

The authors should rephrase this sentence. XCI has evolved to compensate for the X-linked gene dosage between XX females and XY males.

20. Lines 85-89

We measured XCI in blood-derived DNA from 1,575 participants (median age = 61) unselected for chronic disease status from the TwinsUK population cohort(Verdi et al., 2019), which comprised of 423 MZ pairs, 257 DZ pairs, and 215 singletons. Using the normalised distribution of XCI, we derive a categorical variable in which XCI-skew equated to 75% XCI, and extreme XCI-skew equated to 91% XCI.

The authors should describe how the degree of XCI skewing was assayed in their experimental settings as a more general readership is not necessarily familiar with the HUMARA assay. A representative image should be shown in a figure, ideally, a main figure but the Supplementary would also work. A detailed description of the method is also required to allow the reader to follow the considerations related to data analysis in the discussion. Indeed the authors are generally far too vague about their approach, even in the Methods section.

21. Line 336-337

The assay failed in 194 samples, and 601 samples were homozygous for the CAG repeat and were therefore uninformative.

The authors should perhaps comment/speculate on why the assay fail on 194/2382 (almost 10%) of samples. Is this a sample-intrinsic issue? e.g. too-low quality/concentration. If not, 10% failure rate is pretty poor.

22. Figures quality and data description in legends should be significantly improved. For example, the authors very often talk about "association" in the manuscript with reference to a suboptimal data representation (e.g. a table) – or without showing the data at all. In my opinion, every such statement should be supported be at least a basic scatter plot to illustrate the trends the authors are talking about (and especially the spread of the data). Especially considering that their figures are extremely small in their current version. On a related note, the authors should move the data from the Supplementary to the main figures. They show so little data in the main figures that they can afford to do it.

23. In figure 2B the labels "lower XCI" and "higher XCI" are misleading.

24. In figure 3, the authors should plot "probability of cancer diagnosis" as referred to in line 209 of the text and provide a comprehensive figure legend.

25. When they described the association between skewed XCI and future cancer diagnosis the authors should break the data down into constituent cancer/organ types. It would be interesting to specifically look into data related to blood cancer – especially considering that they are in many cases sex-biased and also more relatable to their blood-derived XCI-skew measurements. Furthermore, the authors should referred to the well-established links between inflammation and many kinds of cancer if this is the reason why it is meaningful to make more general inferences about cancer from blood measurements.

Reviewer #1 (Recommendations for the authors):

Amy L. Roberts et al., investigated the health outcomes impacts of age acquired skewing of X-chromosome inactivation (XCI) ratios occurring in blood cells in 1575 females comprised of 423 monozygotic twins pairs, 257 dyzygotic twins pairs and 215 singleton. They demonstrate: (i) associating between skewing and age; (ii) skewing was independent of other biological markers of aging such as smoking, telomere length, or DNAm Grim Age acceleration or frailty index (iii). (iv) that skewing was associated with a myeloid bias with an increase monocyte to lymphocyte and neutrophil to lymphocyte ratio; (v)a strong negative association with il-10 level and skewing; (vi) skewing was associated with increase cardiovascular risk; and that (vi) skewing was predictive of cancer.

General comments.

Age-associated XCI skewing has been described 25 years ago, yet this phenomenon remains enigmatic in terms of etiology and consequences. The study demonstrates the age-effect on skewing (which is known), but the evolution over time in a sub cohort of subject that were re-sampled. The originality of this study resides in the demonstration that age-associated skewing is associated with health outcomes. The particular strength of the study is the twin component (always aged-matched) and the iterative re-sampling of subgroup of the population study. Despite starting with a large cohort of 1552 individuals, most of the critical observations are made on smaller subsets of subjects: for ASCV risk score the total of subject is 231; IL-10 n=27. Fortunately, for the ASCV score this observation is re-enforced by the twin pairs comparison (N=34). The association with cancer risk is performed on a larger cohort (1417) an controlled for age. However, despite increased risk of Cardiovascular event and cancer, there is no significant association with overall mortality. This indicate that the effect of age-associated skewing is present but modest. The observation that there is a myeloid bias, which is associated with aging, that is more pronounced in subjects with skewing is interesting.

One of the major unknown of this study is how much of the effect attributed to XCI skewing is in fact related to true clonal hematopoiesis or CHIP that is associated with both outcomes described in this study and will mascarade by changes in XCI.

Reviewer #2 (Recommendations for the authors):

The manuscript by Roberts et al., is an important piece of work regarding the potentiality of XCI skewing as a biomarker of increase risk for age-driven chronic diseases and cancer. However, the data presenting is exclusively correlative using cohorts of patients.

– The relationship between XCI skewing and ageing in sound (Figure 1A-B). This is the most robust data from this study that confirms previous results (Busque et al., 1996; Gale et al., 1997; Tonno et al., 1998; van Dijk et al., 2002; Zito et al., 2019). It is unquestionable the importance of this result, but as pointed out by the authors, it is not a novel finding.

– The link between XCI skewing and atherosclerotic cardiovascular disease risk or cancer incidence is an interesting one. However, the data is not totally convincing in my view. First, the association with cardiovascular disease (CVD) is based on a risk score and not in actual CVD events. On the other hand, the prospective study of XCI-skew and future cancer diagnosis is based on 58 cancer events. As the authors pointed out, the link between XCI skewing and cancer has been inconsistent in the literature, perhaps because all these studies, including this one, do not reach enough cancer events to draw definitive conclusions. While the prospect of XCI skewing as a biomarker of chronic disease risk and cancer incidence is important, larger cohorts might be needed to ascertain the usefulness of XCI skew as a prognosis marker.

– The data presented is exclusively correlative. Although out of scope of this manuscript, functional experiments addressing the effect of XCI skewing in disease outcome would give insights about a potential causal role of XCI skewing in disease onset/progression/prognosis. These types of experiments will increase the impact of the findings using cohort of patients.

https://doi.org/10.7554/eLife.78263.sa1

Author response

Essential revisions:

1. Was XCI measured on whole blood, versus fraction (PMN, Mononuclear cells, buffy coat, other)?

The DNA used in this study was extracted from whole blood. This has been added to the abstract, method, and results at lines 12-13, 100, and 239.

2. Did the authors look at the correlation between skewing in MZG vs DZ twins. This is known but would be interesting to document.

We have now added the concordance rates to the Results section of the manuscript along with the relevant references:

“In line with previous studies (Vickers et al., 2001), we increased concordance of XCI-skew within MZ twin pairs compared with DZ twin pairs: 27% of MZ twin pairs (114/423), and 45.5% of DZ twin pairs (117/257), were discordant for their categorical XCI-skew status.”

3. Was there enough discordant MZG twin pairs (best controls) for the ASCV risk comparison?

Of the 34 pairs of discordant twins, 17 were DZ and 17 were MZ. The direction of effect is the same across both sets, though of course the power is lower. Using MZs only, we see P = 0.095; and using DZs only, we see P = 0.08.

4. Age is the major confounding factor in all analyses. Although the information is present in the text and methods, I think controlling for age (or age-matched) should also be indicated in all figures.

We have updated all figure legends throughout the manuscript to be clearer. Please see Lines 275-291, 305-310, 340-355, 418-424, 447-454 for all relevant legends, but here we give an example from part of the legend for Figure 3:

“Box plots representing the results of the linear regression mixed effects models to assess (A) Smoking status (P=0.33, Nnever_smoker = 879; Never_smoker = 673) and (B) obesity (P=0.88, Nnot_obese = 726; Nobese = 165) after correcting for age with XCI as the dependent variable, and (C) frailty index (P=0.59, Nranodm XCI = 398; nSkewed XCI = 177; Nextreme skew = 36), after correcting for age and BMI, (D) Leukocyte telomere length shortening (P=0.9, Nranodm XCI = 278; nSkewed XCI = 103; Nextreme skew = 16) after correcting for age and smoking status, and (E) DNAm GrimAge acceleration (P=0.22, Nranodm XCI = 101; nSkewed XCI = 30; Nextreme skew = 6), with XCI-skew as the dependent variable.”

5. The hypothesis that cardiovascular and cancer risk could be driven in part by CHIP should be added to the discussion.

We have added a new paragraph to the discussion which addresses this point and the overlap between CHIP and XCI-skew more generally:

“Given the existing links between CHIP and XCI-skew as two age acquired blood traits, and that CHIP mutations are found in individuals with XCI-skew, could the association between XCI-skew and CVD risk be partly driven by CHIP? It is expected that some of the individuals with XCI-skew in our study will also harbour CHIP mutations. However, a better understanding of the co-occurrence of XCI-skew and CHIP within individuals, and the mutational burden within individuals with and without XCI-skew, is an important area of future work. Whether the co-occurrence of XCI-skew and CHIP represents an amplified risk of disease, or helps better define risk categories, will be of clinical significance to establish. As the inherited genetic risk of CHIP is fast becoming better understood, deciphering the genetic predisposition to XCI-skew will also enable the assessment of potential shared genetic susceptibility between the two traits (Kar et al., 2022). However, many of the negative results in our study are crucial findings given they expose differences between the risk factors of XCI-skew and CHIP. Notably, recent work has demonstrated smoking and telomere length are causal risk factors for CHIP (Kar et al., 2022), whereas we see no association with either trait in this study, suggesting CHIP and XCI-skew have some distinct aetiologies.”

6. The XCI skewing data is gathered using the HUMARA test. This is a classical test to score skewing in human cells. Since it is the central method of this manuscript, illustrative examples of the data (random, skew, and extreme skew, as well as non-informative samples) should be provided. This will be helpful to the reader and of broad interest to the eLife audience, especially for (epi)geneticists.

We have added a new Figure 1 to the manuscript with illustrative examples of the raw data from the HUMARA assay representing a random, skewed, and extreme skew sample. We agree this will make the manuscript easier to interpret for a broader readership.

7. When mentioning karyotype, the most common nomenclature is 46, XX, and 46, XY instead of XX,46 or XY,46 – in lines 50 and 51 and maybe elsewhere in the manuscript.

Thank you, the abstract and introduction have now been updated with the correct nomenclature.

8. Figure 1B in lin 154 and 157 is actually Figure 1C.

Thank you for spotting this error. This has been updated. Given the new figures included, this is now Figure 3F.

9. Table 1. Table S1, Figure S1 – It is unclear what N means. It should be clarified how many women with or without XCI-skew were analysed.

We have reorganised the manuscript such that Table 1 and Table S1 are now replaced with more informative figures, as per point 22 below. In each figure legend, as well as the accompanying text, we now specified the sample size with regards to each XCI-skew subgroup. We have left the Supplementary flowchart unchanged for ease of interpretation of overall numbers across the study as we believe the data are now readily available through the manuscript.

10. Please provide in the legend, the numbers of individuals analysed in the three categories in Figure 2A.

The sample sizes have been added to this specific figure legend, as well as to all other legends and Results sections throughout the manuscript.

11. Please clarify the methodology performed in the sentence stated on lines 181-183: "XCI-skew was associated with increased ASCVD risk score after controlling for BMI and monocyte abundance".

We have included the methodology in the ASCVD Results section:

“In a cross-sectional study of 228 individuals (Nranodm XCI = 155; nSkewed XCI = 56; Nextreme skew = 17; median age = 62) with matched health data available, XCI-skew was associated with increased ASCVD risk score after controlling for BMI and monocyte abundance using a linear regression mixed effects model (P=0.01, Figure 4A).”

12. In Figure 3, the cumulative number of events table is misleading (e.g. at 10 years, we have 28 events in the random XCI group and 30 events in the XCI-skew group, and therefore they do not look different, while they are in the graph presenting percentages). It should be indicated the total population analysed per group. Also whether the events occur in distinct individuals or in one individual more than one time should be stated.

To carry out this analysis, we excluded any individual with a cancer diagnosis before or within 6 months of the XCI-skew measure to ensure they were cancer free at baseline. We then took the first cancer diagnosis only for each person within up to 10 years of the XCI measure. Therefore, each event in the Cox regression analysis is an independent cancer diagnosis in one person. These details were included in the methods, but we have added them to the figure legend and the results for clarity. The results now read:

“We conducted a prospective 10-year follow-up study (median follow-up time 5.65 years) from time of DNA sampling in 1,417 individuals (Nrandom = 948, Nskewed = 469, median age = 60) who were cancer-free at baseline to assess the association between XCI-skew and future cancer diagnoses (cancer events = 58; Supplementary Table S1). Each cancer event represents the first cancer diagnosis (excluding non-melanoma skin cancer) in each individual and subsequent diagnoses are not included.”

We have also added the numbers in each group as well as the percentages of each group with a cancer diagnosis, to the figure legend:

“A Kaplan-Meier plot (top) and cumulative events (bottom) of cancer diagnosis in individuals with XCI-skew (N=469) and random XCI (N=948) are shown. 2.9% (28/948) of individuals with random XCI, and 6.4% (30/469) of individuals with XCI-skew, developed cancer in the 10-year follow-up.”

We have also updated Table S1 to represent the number of cancer diagnoses in each organ/tissue for the Random XCI and XCI-skew group which we hope adds clarity to the data.

13. One aspect that would be important to analyse in the future is the relationship between XCI skewing and mutation burden in blood cells. For example, XCI skew could be a bystander of genetic mutation(s) that would cause clonal expansion of that population. Or XCI skew could determine the fixation of an X-linked mutation in the cell populations. These aspects could be addressed in the discussion.

We agree that this is an important piece of future work. We have added the following paragraph to the discussion which addresses the need for this work:

“Given the existing links between CHIP and XCI-skew as two age acquired blood traits, and that CHIP mutations can be found in individuals with XCI-skew (Busque et al.,. 2012), could the association between XCI-skew and CVD risk be partly driven by CHIP? It is expected that some of the individuals with XCI-skew in our study will also harbour CHIP mutations. However, a better understanding of the co-occurrence of XCI-skew and CHIP within individuals, and the mutational burden within individuals with and without XCI-skew, is an important area of future work. Whether the co-occurrence of XCI-skew and CHIP represents an amplified risk of disease, or helps better define risk categories, will be of clinical significance to establish. As the inherited genetic risk of CHIP is fast becoming better understood, deciphering the genetic predisposition to XCI-skew will also enable the assessment of potential shared genetic susceptibility between the two traits (Kar et al., 2022). However, many of the negative results in our study are crucial findings given they expose differences between the risk factors of XCI-skew and CHIP. Notably, recent work has demonstrated smoking and telomere length are causal risk factors for CHIP (Kar et al., 2022), whereas we see no association with either trait in this study, suggesting CHIP and XCI-skew have some distinct aetiologies.”

14. In the general introduction, dosage compensation by XCI and XCI-skewing is not explained in sufficient detail. Several important features, such as the difference between primary and secondary XCI skewing are completely omitted, as well as a clear overview of the clinical relevance of skewed XCI. The main technology employed to assay XCI skewing is also not described. As a consequence, a general readership is likely to not fully appreciate the open questions addressed by authors, and their relevance in light of the current knowledge and available technologies.

We have expanded the introduction to include examples of primary XCI-skew before introducing secondary XCI-skew, as well as a broadening the topic more generally:

“However, some individuals display a skewed pattern of XCI (XCI-skew), which is defined as a deviation from the expected 1:1 ratio. Examples of primary XCI-skew have been identified, including stochastic events resulting in XCI-skew across all tissues (Tukiainen et al., 2016) or the preferential selection of cells expressing functioning alleles in heterozygous females with X-linked recessive traits (Busque and Gilliland, 1998; Nyhan et al., 1970). However, secondary or age acquired XCI-skew is more common and refers to increasing XCI-skew with age, particularly in mitotically active blood tissue (Busque et al., 1996; Gale et al., 1997).

We have also provided a new Figure 1 which details how XCI-skew is derived from the HUMARA assay (as per point 6 above), and included the following statement in the first Results section regarding the correlation of HUMARA with transcription-based methods:

“We measured XCI in DNA derived from whole blood using the methylation-sensitive PCR-based Human Androgen Receptor Assay (HUMARA) (Cutler Allen et al., 1992; Hatakeyama et al., 2004) which differentiates between alleles from the active and inactive X (Cutler Allen et al., 1992; Hatakeyama et al., 2004). HUMARA is an extensively used assay which correlates well with transcription-based methods (Bolduc et al., 2008; Zito et al., 2019).

15. The authors claim to characterize the relationship of skewed XCI with "molecular, cellular and organismal measures of ageing and cardiovascular disease risk and cancer diagnosis". However, the correlation between XCI skewing in blood-derived DNA and increased age has been already extensively described (i.e. Bolduc et al., 2008, Busque et al., 1996, Hatakeyama et al., 2004; Zito et al., 2019, and many more). Importantly, with the exception of the assessment of XCI skewing which was obtained for >1.500 individuals, the molecular/cellular data presented by the authors is rather limited. Furthermore, when describing the association between skewed XCI and increased monocyte abundance, the authors omit to discuss that within the blood, different populations of hematopoietic cells show different degrees of XCI skewing with myeloid cells showing a high level of skewed XCI (Gale et al., 1997). The association described by the authors is therefore probably reinforcing these observations, and can likely be explained by biases in cell representation within samples. Furthermore, given the limited number of individuals tested for serum levels of CRP and cytokine interleukin, and the negative association observed for IL-10, these results are difficult to interpret.

We hope it was clear in our manuscript that the association seen with ageing is a replication of the many studies that predate ours, and not a novel finding. See lines 463-465 of discussion, which are unchanged from the original version, but which introduce the topic in reference to the long-established association with increasing age. We feel that no study on XCI-skew would be complete without a summary of the association with age seen in this cohort. However, we have amended the manuscript throughout to ensure its as clear as possible that the age association is an expected result, and acts as a validation of our dataset in line with the existing literature. The results now read as follows:

“As expected, after controlling for relatedness and zygosity, we find a significant positive association between age and XCI skewing (P=2.8x10-9, N=1,575). This result replicates the many existing studies on age acquired XCI-skew and acts as a validation of the TwinsUK HUMARA dataset (Busque et al., 1996; Gale et al., 1997; Hatakeyama et al., 2004; Zito et al., 2019).”

We have also amended the statement in our introduction so that it now reads:

“We tested this hypothesis by assaying XCI-skew in 1,575 females from the TwinsUK cohort and employed prospective, cross-sectional, and intra-twin designs, to characterise the relationship of XCI with molecular and cellular measures of ageing, cardiovascular disease risk, and cancer diagnoses.”

We have amended the discussion include reference to the Gale 1997 study, in which there is lower levels of skew in T lymphocytes compared to monocytes, thought to be due to the longer life span of these lymphoid cells. This is appended to our existing discussion point in which we reference two other studies which show correlation between the skew of different cell types.

“Furthermore, age acquired XCI-skew has previously been shown across isolated neutrophils, monocytes, and T cells, with correlations between these fractions(Tonon et al., 1998; van Dijk et al., 2002), albeit with lower levels of skewing in lymphoid cells (Gale et al., 1997).”

We agree that the cytokine data are limited, and though we had addressed this in the discussion (see Lines 566-568), we have now updated the results to better reflect this too:

“We date-matched the XCI data with serum levels of CRP (Nranodm XCI = 121; nSkewed XCI = 38; Nextreme skew = 6) and a more modest cytokine dataset of interleukin (IL)-6, IL-1B, IL-10, and TNF (Nranodm XCI = 23; nSkewed XCI = 4) and used linear regression mixed effects models to control for batch effects, age, seasonality, family structure and zygosity (Supplementary Figure S3). We see no association with primary markers of inflammageing CRP (P=0.41), IL-6 (P=0.41), or TNF (P=0.61), though a nominal association with IL-1β is observed (P=0.02) which does not pass multiple correction, but warrants follow up analysis with a larger sample size.”

16. The relationship between skewed XCI and future cancer diagnosis is also limited by the low numbers of diagnoses in 10-year follow-up and does not allow to take into account cancer type and tissue-specificity. In this context, the authors do not discuss (nor speculate on) how skewed XCI as measured in the blood can be a clinical marker of risk for diseases occurring in other organs. This is a very important point to consider for data interpretation, in light of the existing research addressing the degree of XCI skewing in different tissues and cell types, and not only in blood.

With 28 cancer events in 948 individuals with random XCI, and 30 cancer events in 469 individuals with XCI-skew, we believe this is a well-powered study for total cancer risk. However, we do agree that we were not sufficiently powered to take into account the tissue-specificity, and it would be interesting to explore cancer risk in specific tissues. Of note, see point 25 below in which removing Haematopoietic / Lymphoid Tissues cancers slightly strengthens the association observed.

Notwithstanding the above, we have added a new paragraph to the discussion in which we speculate on these associations and how they may likely be driven by inflammation and are unlikely to be causally driven by XCI-skew:

“Though CHIP is also predictive of future cancer diagnosis, the association is limited to haematological cancers (Desai et al., 2018). Whereas here we see blood-derived measures of XCI-skew is predictive of all future cancer diagnoses, even when the haematological cancers are removed from the analysis. Follow-up studies to assess the risk of cancer of specific tissues in individuals with XCI-skew are needed, but we hypothesise that the relationship between XCI-skew measures in blood tissue and cancer will not be causal. Instead XCI-skew is likely a marker of chronic inflammation, which can predispose to the development of cancer through increased mutagenesis and can promote tumorigenesis by shaping the tumour microenvironment to stimulate tumour growth (Greten and Grivennikov, 2019). However, it is interesting that common environmental factors that induce chronic inflammation, such as smoking and obesity, have not been found to be risk factors for XCI-skew in our study.”

17. To improve the quality of their manuscript to possibly target a broader audience, the authors should more carefully discuss their findings while including relevant references that are omitted in the current version.

We have extended the discussion (see points 13 and 16 above) and included additional references throughout. We hope the reviewer agrees this makes the manuscript more relevant to a broader audience.

18. Lines 7-9.

Age acquired XCI-skew describes the preferential inactivation of one X chromosome across a tissue, which is particularly prevalent in blood tissues of ageing females, and yet its clinical consequences are unknown.

This sentence is misleading. The preferential inactivation of one of the two X chromosomes only occurs at the onset of XCI and results in primary skewed XCI, which influences the choice of which allele becomes inactivated. Age-acquired XCI skewing is a result of post-XCI selection. The authors should include a clear description of the differences between primary and secondary XCI skewing in the introduction, and they should also describe their relevance to human diseases.

We have amended this sentence in the abstract:

“Age acquired XCI-skew describes the preferential selection of cells across a tissue resulting in an imbalance of XCI, which is particularly prevalent in blood tissues of ageing females and yet its clinical consequences are unknown.”

We have also expanded the introduction to include primary XCI skew too (see also point 14 above).

“However, some individuals display a skewed pattern of XCI (XCI-skew), which is defined as a deviation from the expected 1:1 ratio. Examples of primary XCI-skew have been identified, including stochastic events resulting in XCI-skew across all tissues (Tukiainen et al., 2016) or the preferential selection of cells expressing functioning alleles in heterozygous females with X-linked recessive traits (Busque and Gilliland, 1998; Nyhan et al., 1970). However, secondary or age acquired XCI-skew is more common and refers to increasing XCI-skew with age, particularly in mitotically active blood tissue (Busque et al., 1996; Gale et al., 1997). Within individuals, the correlation of XCI ratios between blood and other tissues diminishes over the life course as the XCI ratios in blood continue to skew with age (Bolduc et al., 2008; Zito et al., 2019).”

19. Line 48-49

X chromosome inactivation (XCI) evolved in placental mammals to compensate for the differences in size between the X and Y sex chromosomes.

The authors should rephrase this sentence. XCI has evolved to compensate for the X-linked gene dosage between XX females and XY males.

We agree and have changed the sentence on lines 52-53 as suggested.

20. Lines 85-89

We measured XCI in blood-derived DNA from 1,575 participants (median age = 61) unselected for chronic disease status from the TwinsUK population cohort(Verdi et al., 2019), which comprised of 423 MZ pairs, 257 DZ pairs, and 215 singletons. Using the normalised distribution of XCI, we derive a categorical variable in which XCI-skew equated to 75% XCI, and extreme XCI-skew equated to 91% XCI.

The authors should describe how the degree of XCI skewing was assayed in their experimental settings as a more general readership is not necessarily familiar with the HUMARA assay. A representative image should be shown in a figure, ideally, a main figure but the Supplementary would also work. A detailed description of the method is also required to allow the reader to follow the considerations related to data analysis in the discussion. Indeed the authors are generally far too vague about their approach, even in the Methods section.

We have now included a new Figure 1 in the manuscript which represents how the raw HUMARA assay data is used to calculate XCI skew (see also point 6 above). We have also included an additional Supplementary Figure S1 which demonstrates how the categorical variables were derived from the continuous data, as well as added the following text to the Results section for clarity:

“The output of the HUMARA assay is a continuous XCI variable from 0-100%, where 50% is perfectly balanced XCI and the directionality of XCI away from 50% is uninformative (e.g., both 0% and 100% are considered equal). We normalised the distribution of the continuous XCI values across the cohort, and defined XCI-skew as measures 1s.d. from the mean, corresponding to XCI score ≥75%, and extreme XCI-skew as measures 2s.d. from the mean, corresponding to XCI score ≥91% (Supplementary Figure S1).”

We have also expanded the HUMARA methods section to allow a fuller understanding of the assay:

“The HUMARA method which combines methylation-sensitive restriction enzyme digest and amplification of a highly polymorphic (CAG)n repeat in the first exon of the X-linked AR gene, allowing for the differentiation of the active and inactive chromosomes in heterozygous individuals (Cutler Allen et al., 1992). Here, 625ng of genomic DNA was divided into three aliquots and incubated for 30 minutes at 37°C with (i) the methylation-sensitive enzyme HpaII, (ii) the methylation-insensitive enzyme MspI, or (iii) water (mock digest) in 1x New England Biolabs CutSmart Buffer. The HpaII digest was followed by an additional 20 minutes at 80°C to avoid residual enzymatic activity. Fluorescently labelled PCR primers (FAM, VIC, NED, or PET; Forward primer 5’-dye-GCTGTGAAGGTTGCTGTTCCTCAT-3’, Reverse primer 5’-TCCAGAATCTGTTCCAGAGCGTGC-3’) were used in New England BioLabs One Taq Master Mix to amplify 1.5μl of digested PCR product. The Mock and HpaII digested DNA were amplified in triplicate (using FAM, VIC, and NED), and the MspI digest, used as control of digestion efficiency, was amplified once (using PET). All PCRs were amplified with an initial denaturation step at 94°C for 5 mins, followed by 30 cycles of 94°C for 30 secs, 60°C for 1 min, and 72°C for 2 mins, and a final elongation step of 72°C for 15 mins.”

21. Line 336-337

The assay failed in 194 samples, and 601 samples were homozygous for the CAG repeat and were therefore uninformative.

The authors should perhaps comment/speculate on why the assay fail on 194/2382 (almost 10%) of samples. Is this a sample-intrinsic issue? e.g. too-low quality/concentration. If not, 10% failure rate is pretty poor.

We do indeed suspect most assay failures were due to underlying low DNA concentrations of the archived, biobanked samples used in the study. We did not observe any batch effects with the failure rates that would suggest experimental issues. Furthermore, samples typically failed across all three PCRs (FAM, VIC, and NED) also suggesting intrinsic issues which typically impact PCRs, such as DNA concentration. Though the best effort was taken to standardise the DNA concentration of the samples prior to amplification, we suspect DNA concentrations will be a major cause of sample failures.

22. Figures quality and data description in legends should be significantly improved. For example, the authors very often talk about "association" in the manuscript with reference to a suboptimal data representation (e.g. a table) – or without showing the data at all. In my opinion, every such statement should be supported be at least a basic scatter plot to illustrate the trends the authors are talking about (and especially the spread of the data). Especially considering that their figures are extremely small in their current version. On a related note, the authors should move the data from the Supplementary to the main figures. They show so little data in the main figures that they can afford to do it.

We have included two new main figures (see Figure 1 and Figure 3), and two new supplementary figures (see Supplementary Figure S1 and Supplementary Figure S3) to better present the data, particularly replacing Table 1 and Table S1 with figures (see also point 9 above). We have also re-written all figure legends to ensure the methodology is clear to the reader.

23. In figure 2B the labels "lower XCI" and "higher XCI" are misleading.

The discordant twins can be any combination of the 3 categorical variables: random XCI and skewed XCI, random XCI and extreme skewed XCI, or skewed XCI and extreme skewed. Therefore, we found it clearest to label the x-axis as “lower” and “higher” as the groups are somewhat heterogenous. However, we have amended these labels to read “twin with lower skewed XCI” and “twin with higher skewed XCI”, which we hope adds clarity. We would be very happy to use a different term here if the reviewer has other suggestions.

24. In figure 3, the authors should plot "probability of cancer diagnosis" as referred to in line 209 of the text and provide a comprehensive figure legend.

The axis label for Figure 5 (note, previously figure 3) has been changed, and the figure legend (as per comments 4 and 12) has been improved.

25. When they described the association between skewed XCI and future cancer diagnosis the authors should break the data down into constituent cancer/organ types. It would be interesting to specifically look into data related to blood cancer – especially considering that they are in many cases sex-biased and also more relatable to their blood-derived XCI-skew measurements. Furthermore, the authors should referred to the well-established links between inflammation and many kinds of cancer if this is the reason why it is meaningful to make more general inferences about cancer from blood measurements.

The origin of the cancers is included in “Supplementary Table S1: Cancer diagnoses recorded in 10-year follow-up by organ/site.”. We have extended this table to show the cancer diagnoses per XCI random and XCI skewed group too. We are underpowered to run the association on subsets of cancers, however, re-running the analysis after first excluding the Haematopoietic / Lymphoid Tissues cancers slightly strengthens the association observed (P=0.009; HR=2.04 (1.20-3.49)) compared to the original analysis (P=0.012; HR = 1.95 (1.16-3.28)). This suggests the association is not driven by the small number of Haematopoietic / Lymphoid Tissues in the dataset. We have added this to the results (Lines 440-443) and the discussion (Lines 496-498).

“Though we were underpowered to run associations for specific cancers, given the robust association of CHIP with haematological cancers we were interested to explore the potential that blood cancers were partly driving the association. However, running the analysis without the Haematopoietic/Lymphoid Tissue cancers (see Supplementary Table S1) strengthened the observed association (P=0.009; HR = 2.04 (1.20-3.49)) suggesting a crucial difference in cancer risk between XCI-skew and CHIP.”

We have also extended the discussion to include the role of inflammation in cancer, and how this may be the link between XCI-skew in blood and cancer across various tissues.

“Though CHIP is also predictive of future cancer diagnosis, the association is limited to haematological cancers (Desai et al., 2018). Whereas here we see blood-derived measures of XCI-skew is predictive of all future cancer diagnoses, even when the haematological cancers are removed from the analysis. Follow-up studies to assess the risk of cancer of specific tissues in individuals with XCI-skew are needed, but we hypothesise that the relationship between XCI-skew measures in blood tissue and cancer will not be causal. Instead XCI-skew is likely a marker of chronic inflammation, which can predispose to the development of cancer through increased mutagenesis and can promote tumorigenesis by shaping the tumour microenvironment to stimulate tumour growth (Greten and Grivennikov, 2019). However, it is interesting that common environmental factors that induce chronic inflammation, such as smoking and obesity, have not been found to be risk factors for XCI-skew in our study.”

Reviewer #1 (Recommendations for the authors):

Amy L. Roberts et al., investigated the health outcomes impacts of age acquired skewing of X-chromosome inactivation (XCI) ratios occurring in blood cells in 1575 females comprised of 423 monozygotic twins pairs, 257 dyzygotic twins pairs and 215 singleton. They demonstrate: (i) associating between skewing and age; (ii) skewing was independent of other biological markers of aging such as smoking, telomere length, or DNAm Grim Age acceleration or frailty index (iii). (iv) that skewing was associated with a myeloid bias with an increase monocyte to lymphocyte and neutrophil to lymphocyte ratio; (v)a strong negative association with il-10 level and skewing; (vi) skewing was associated with increase cardiovascular risk; and that (vi) skewing was predictive of cancer.

General comments.

Age-associated XCI skewing has been described 25 years ago, yet this phenomenon remains enigmatic in terms of etiology and consequences. The study demonstrates the age-effect on skewing (which is known), but the evolution over time in a sub cohort of subject that were re-sampled. The originality of this study resides in the demonstration that age-associated skewing is associated with health outcomes. The particular strength of the study is the twin component (always aged-matched) and the iterative re-sampling of subgroup of the population study. Despite starting with a large cohort of 1552 individuals, most of the critical observations are made on smaller subsets of subjects: for ASCV risk score the total of subject is 231; IL-10 n=27. Fortunately, for the ASCV score this observation is re-enforced by the twin pairs comparison (N=34). The association with cancer risk is performed on a larger cohort (1417) an controlled for age. However, despite increased risk of Cardiovascular event and cancer, there is no significant association with overall mortality. This indicate that the effect of age-associated skewing is present but modest. The observation that there is a myeloid bias, which is associated with aging, that is more pronounced in subjects with skewing is interesting.

One of the major unknown of this study is how much of the effect attributed to XCI skewing is in fact related to true clonal hematopoiesis or CHIP that is associated with both outcomes described in this study and will mascarade by changes in XCI.

Thank you for your helpful comments. Though we agree that sample sizes are low in some analyses (which we now address more carefully in the revised manuscript), we would also like to draw attention to the many analyses with robust sample sizes, such as smoking (n=1,552), telomere length (n=397), frailty (n=611) and obesity (n=891). These negative findings are vital in our understanding of the causes and consequences of XCI-skew, suggesting XCI-skew as an independent biomarker.

We completely agree that it is of upmost importance to establish the link with CHIP more robustly, and we hope to do this in follow-up work. However, based on existing literature and our current study, there are some key differences between XCI-skew and CHIP. Firstly, telomere length and smoking are both been identified as causal risk factors for CHIP (Kar, 2022), yet we see no such association in our work. And secondly, CHIP is exclusively associated with blood cancers, whereas we see an association with cancer across all tissues/organs even when removing the small number of haematopoietic cancers from the dataset (n=4 cases). Together these observations suggest that robustly defining the relationship between XCI-skew and CHIP is of vital importance as we do not believe one is fully explained by the other. We have updated the manuscript to draw attention to these important differences. What will be of great interest is determining if the presence of both CHIP and XCI-skew within an individual poses a greater disease risk than either molecular trait occurring independently. We have added these points to the discussion of the paper (see Lines 522-537) which we believe improves the manuscript and we thank you once again for your comments.

Reviewer #2 (Recommendations for the authors):

The manuscript by Roberts et al., is an important piece of work regarding the potentiality of XCI skewing as a biomarker of increase risk for age-driven chronic diseases and cancer. However, the data presenting is exclusively correlative using cohorts of patients.

– The relationship between XCI skewing and ageing in sound (Figure 1A-B). This is the most robust data from this study that confirms previous results (Busque et al., 1996; Gale et al., 1997; Tonno et al., 1998; van Dijk et al., 2002; Zito et al., 2019). It is unquestionable the importance of this result, but as pointed out by the authors, it is not a novel finding.

– The link between XCI skewing and atherosclerotic cardiovascular disease risk or cancer incidence is an interesting one. However, the data is not totally convincing in my view. First, the association with cardiovascular disease (CVD) is based on a risk score and not in actual CVD events. On the other hand, the prospective study of XCI-skew and future cancer diagnosis is based on 58 cancer events only. As the authors pointed out, the link between XCI skewing and cancer has been inconsistent in the literature, perhaps because all these studies, including this one, do not reach enough cancer events to draw definitive conclusions. While the prospect of XCI skewing as a biomarker of chronic disease risk and cancer incidence is important, larger cohorts might be needed to ascertain the usefulness of XCI skew as a prognosis marker.

– The data presented is exclusively correlative. Although out of scope of this manuscript, functional experiments addressing the effect of XCI skewing in disease outcome would give insights about a potential causal role of XCI skewing in disease onset/progression/prognosis. These types of experiments will increase the impact of the findings using cohort of patients.

Thank you for your helpful comments. We agree that assessing the risk of XCI-skew on incidence CVD is of upmost importance to establish the true risk of this age acquired trait and this has now been emphasised further in our discussion.

However, we believe our cancer analysis to be more robust than previous studies for three main reasons: 1) we have corrected for age at study entry in our model; 2) all individuals were cancer-free at baseline therefore removing confounding factors of treatments; 3) we did not focus on cancers of the female reproductive system only as we do not believe this is a robust a priori hypothesis. However, we fully agree with the reviewer that larger studies would ideally be required to confirm the usefulness of XCI-skew as a prognosis marker.

We also agree with the need for functional experiments, and we hope our study demonstrating these correlations will enable this crucial follow-up work to take place. We would also like to emphasise once more that many of the robust analyses in our study with negative results are of great importance when considering follow-up work. For instance, we have shown that XCI-skew does not correlate with other cellular and molecular markers of biological ageing, nor indeed that frailty, smoking nor obesity are risk factors, and these results we feel are highly significant findings.

https://doi.org/10.7554/eLife.78263.sa2

Article and author information

Author details

  1. Amy L Roberts

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing - original draft, Project administration
    For correspondence
    amy.roberts@kcl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5704-9249
  2. Alessandro Morea

    1. Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    2. Department of Medical and Molecular Genetics, King’s College London, London, United Kingdom
    Present address
    The FIRC Institute of Molecular Oncology, Italian Foundation for Cancer Research, Milan, Italy
    Contribution
    Data curation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7919-1932
  3. Ariella Amar

    Department of Medical and Molecular Genetics, King’s College London, London, United Kingdom
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Antonino Zito

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Present address
    1. Department of Molecular Biology, Massachusetts General Hospital, Boston, United States
    2. Department of Genetics, Harvard Medical School, Boston, United States
    Contribution
    Data curation, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1931-984X
  5. Julia S El-Sayed Moustafa

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Max Tomlinson

    1. Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    2. Department of Medical and Molecular Genetics, King’s College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Ruth CE Bowyer

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Xinyuan Zhang

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Colette Christiansen

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
  10. Ricardo Costeira

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
  11. Claire J Steves

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Resources, Writing – review and editing
    Competing interests
    Consultant for Zoe Ltd
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4910-0489
  12. Massimo Mangino

    1. Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    2. NIHR Biomedical Research Centre, Guy's and St Thomas' Foundation Trust, London, United Kingdom
    Contribution
    Resources, Data curation, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2167-7470
  13. Jordana T Bell

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  14. Chloe CY Wong

    Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, London, United Kingdom
    Contribution
    Supervision, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  15. Timothy J Vyse

    Department of Medical and Molecular Genetics, King’s College London, London, United Kingdom
    Contribution
    Resources, Supervision, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1123-1464
  16. Kerrin S Small

    Department of Twin Research & Genetic Epidemiology, King’s College London, London, United Kingdom
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing – review and editing
    For correspondence
    kerrin.small@kcl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4566-0005

Funding

Medical Research Council (MR/M004422/1)

  • Kerrin S Small

Medical Research Council (MR/R023131/1)

  • Kerrin S Small

Economic and Social Research Council (ES/N000404/1)

  • Jordana T Bell

National Institute for Health Research

  • Massimo Mangino

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors acknowledge use of the research computing facility at King’s College London, Rosalind (https://rosalind.kcl.ac.uk), which is delivered in partnership with the National Institute for Health Research (NIHR) Biomedical Research Centres at South London & Maudsley and Guy’s & St. Thomas’ NHS Foundation Trusts, and part-funded by capital equipment grants from the Maudsley Charity (award 980) and Guy’s & St. Thomas’ Charity (TR130505). This work uses data that has been provided by patients and collected by the NHS as part of their care and support. The data are collated, maintained and quality assured by the National Disease Registration Service, which is part of NHS Digital. KSS acknowledges funding from the Medical Research Council [MR/M004422/1 and MR/R023131/1]. JTB acknowledges funding from the ESRC [ES/N000404/1]. MM acknowledges funding from the National Institute for Health Research (NIHR)-funded BioResource, Clinical Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London. TwinsUK is funded by the Wellcome Trust, Medical Research Council, European Union, Chronic Disease Research Foundation (CDRF), Zoe Global Ltd and the National Institute for Health Research (NIHR)-funded BioResource, Clinical Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London.

Ethics

Human subjects: All samples and information were collected with written and signed informed consent, including consent to publish within the TwinsUK study. TwinsUK has received ethical approval associated with TwinsUK Biobank (19/NW/0187), TwinsUK (EC04/015) or Healthy Ageing Twin Study (HATS) (07/H0802/84) studies from NHS Research Ethics Service Committees London - Westminster.

Senior and Reviewing Editor

  1. Carlos Isales, Augusta University, United States

Reviewer

  1. Lambert Busque, Hôpital Maisonneuve-Rosemont, Canada

Publication history

  1. Received: February 28, 2022
  2. Preprint posted: April 5, 2022 (view preprint)
  3. Accepted: October 11, 2022
  4. Version of Record published: November 22, 2022 (version 1)

Copyright

© 2022, Roberts et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 445
    Page views
  • 45
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Amy L Roberts
  2. Alessandro Morea
  3. Ariella Amar
  4. Antonino Zito
  5. Julia S El-Sayed Moustafa
  6. Max Tomlinson
  7. Ruth CE Bowyer
  8. Xinyuan Zhang
  9. Colette Christiansen
  10. Ricardo Costeira
  11. Claire J Steves
  12. Massimo Mangino
  13. Jordana T Bell
  14. Chloe CY Wong
  15. Timothy J Vyse
  16. Kerrin S Small
(2022)
Age acquired skewed X chromosome inactivation is associated with adverse health outcomes in humans
eLife 11:e78263.
https://doi.org/10.7554/eLife.78263

Further reading

    1. Epidemiology and Global Health
    Bingyi Yang, Bernardo García-Carreras ... Derek A Cummings
    Research Article

    Background: Over a life-course, human adaptive immunity to antigenically mutable pathogens exhibits competitive and facilitative interactions. We hypothesize that such interactions may lead to cyclic dynamics in immune responses over a lifetime.

    Methods: To investigate the cyclic behavior, we analyzed hemagglutination inhibition titers against 21 historical influenza A(H3N2) strains spanning 47 years from a cohort in Guangzhou, China and applied Fourier spectrum analysis. To investigate possible biological mechanisms, we simulated individual antibody profiles encompassing known feedbacks and interactions due to generally recognized immunological mechanisms.

    Results: We demonstrated a long-term periodicity (about 24 years) in individual antibody responses. The reported cycles were robust to analytic and sampling approaches. Simulations suggested that individual-level cross-reaction between antigenically similar strains likely explain the reported cycle. We showed that the reported cycles are predictable at both individual and birth-cohort level and that cohorts show a diversity of phases of these cycles. Phase of cycle was associated with the risk of seroconversion to circulating strains, after accounting for age and pre-existing titers of the circulating strains.

    Conclusions: Our findings reveal the existence of long-term periodicities in individual antibody responses to A(H3N2). We hypothesize that these cycles are driven by pre-existing antibody responses blunting responses to antigenically similar pathogens (by preventing infection and/or robust antibody responses upon infection), leading to reductions in antigen specific responses over time until individual's increasing risk leads to an infection with an antigenically distant enough virus to generate a robust immune response. These findings could help disentangle cohort-effects from individual-level exposure histories, improve our understanding of observed heterogeneous antibody responses to immunizations, and inform targeted vaccine strategy.

    Funding: This study was supported by grants from the NIH R56AG048075 (D.A.T.C., J.L.), NIH R01AI114703 (D.A.T.C., B.Y.), the Wellcome Trust 200861/Z/16/Z (S.R.) and 200187/Z/15/Z (S.R.). This work was also supported by research grants from Guangdong Government HZQB-KCZYZ-2021014 and 2019B121205009 (Y.G. and H.Z.). D.A.T.C., J.M.R. and S.R. acknowledge support from the National Institutes of Health Fogarty Institute (R01TW0008246). J.M.R. acknowledges support from the Medical Research Council (MR/S004793/1) and the Engineering and Physical Sciences Research Council (EP/N014499/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

    1. Epidemiology and Global Health
    2. Medicine
    Sonali Amarasekera, Prabhat Jha
    Insight

    Individuals recently diagnosed with a cardiovascular disease are at higher risk of developing a mental illness, with mortality increasing when both conditions are present.