Influence of polygenic risk on height and BMI in adults with a 22q11.2 microdeletion

doi:10.21203/rs.3.rs-5050172/v1

Background Adult height and body mass index (BMI) are highly heritable traits that can be influenced by both common variants, captured collectively in a polygenic risk score (PRS), and by rare variants of large effect. In this study, we aimed to investigate the modifying effect of PRSs for height and BMI in individuals where there is a baseline increased risk for short stature (<3rd percentile height) and obesity (BMI ≥30) conferred by the 22q11.2 microdeletion. Methods Genetic analyses were performed in 259 adults of European ancestry with a typical 22q11.2 microdeletion and available genome sequencing and phenotypic data. We applied published PRSs for height and BMI and tested for association using multivariable linear regression models, adjusting for clinical/demographic variables and 22q11.2 microdeletion extent. Results In multivariable linear regression models, the PRSs for height and BMI were significantly associated (p<0.001) with their respective traits, explaining 25.8% and 5.7% of the variance (ΔR2). The overall models (p<0.001) explained 33.9% and 14.7% of the variance (R2) for height and BMI, respectively. Significant additional covariate predictors of shorter sex-standardized height were longer (LCR22A-D) deletion extent, congenital heart disease, intellectual disability, and the first principal component of ancestry; for BMI, they were female sex and older age. When applying the height-PRS to stratify risk for short stature, we found that significantly more of those in the lowest PRS quintile had short stature, compared to those in the highest quintile (odds ratio=11.46, p=1.74E-05). Using logistic regression models to predict short stature, a receiver operating characteristic curve analysis showed that a model combining height-PRS and clinical/demographic covariates achieved an area of the curve of 0.78, performing significantly better than a covariate-only model. Conclusions The results demonstrate that adult height and BMI can be influenced by the effects of both a 22q11.2 microdeletion and genome-wide common variants. The findings contribute to growing evidence that polygenic risk together with other standard factors can shape adult outcomes of a high-impact rare variant. For height in 22q11.2 microdeletion, PRS may be approaching clinical utility.

polygenic risk score

22q11.2 microdeletion

height

short stature

body mass index

obesity

copy number variation

SNPs

Adult height and body mass index (BMI) are two clinically relevant and highly heritable traits [1,2]. Approximately 40% of the variance in height [2] and 5% of the variance in BMI [3], can be attributed to common genetic variation, captured collectively in a polygenic risk score (PRS). As well, rare copy number variants (CNVs) have been associated with shorter height[4–6] and both higher and lower BMI [7].

Previously, we demonstrated that typical 22q11.2 microdeletions are associated with an approximately 7-fold increased risk for short stature (<3^rd percentile height)[8] and 2-fold increased risk for obesity (BMI ≥30) [9], in comparison to population norms. The 1.5–2.5 Mb microdeletion that defines 22q11.2 deletion syndrome (22q11.2DS)[10] is among the most common of rare CNVs with an estimated live birth prevalence of 1 in 2148 [11]. While associated congenital, neurodevelopmental and psychiatric conditions are well-recognized [10], more recent evidence has emerged demonstrating an elevated risk for metabolic and endocrinological conditions in adulthood [9,12–14].

The variable presentation of individual features provide the opportunity to investigate how additional genome-wide variation may shape expression, in the presence of a large-effect yet incompletely penetrant variant, such as the 22q11.2 microdeletion [15]. For example, previous studies have found that genome-wide rare variants and common variant-derived PRSs can modify the risk for expression of schizophrenia [16–19] and hypertriglyceridemia [20], in the context of an elevated baseline risk conferred by the 22q11.2 microdeletion. In the current study, we hypothesized that PRSs derived from general population studies may influence height and BMI in the presence of a 22q11.2 microdeletion. Using a deeply phenotyped cohort of individuals with a typical 22q11.2 microdeletion (Table 1, Supplementary Figure 1), we tested the association of previously published PRSs [2,21] with height and BMI, adjusting for potential effects of other phenotypic variables and of 22q11.2 microdeletion extent. We subsequently assessed the performance of the height-PRS for stratifying risk for short stature in 22q11.2DS.

22q11.2DS cohort and phenotypic and demographic variables

The study cohort consisted of individuals aged ≥17 years with a typical 22q11.2 microdeletion (i.e., the common low copy repeat (LCR)22A–LCR22D extent or proximal nested LCR22A–LCR22B or LCR22A–LCR22C microdeletions extents) ascertained from a specialized adult 22q11.2DS clinic in Toronto, Canada. Prior to filtering and applying exclusion criteria, there were 360 individuals; 259 individuals met criteria for inclusion in primary genetic analyses (Table 1, Supplementary Figure 1). The presence of a typical 22q11.2 microdeletion was identified through standard clinical laboratory methods [13,14] and breakpoints were confirmed using genome sequencing data.

To be included in the study, participants had to have at least one of either a height or BMI (height and weight) measurement obtained from clinical examination. For individuals with multiple height and weight measurements, the most recently available measurement was used, along with the corresponding age at measurement. All but four individuals in this adult cohort were 18 years or older. There were four individuals in the primary analyses for whom the most recently available height and weight measures available were taken at age 17 years. None of these four individuals had short stature. Additionally, we included other phenotypic or demographic variables that may affect height and/or BMI based on their inclusion in the phenotype-only studies of these traits [8,9]. These were sex, age, moderate-severe congenital heart disease (CHD), intellectual disability (ID), psychotic illness, and clinically documented ancestry that was verified using genetic principal component analysis (PCA) (Supplementary Figure 2). As before, we defined “psychotic illness” as individuals diagnosed with schizophrenia or schizoaffective disorder [20].

For details on genome sequencing methods and variant annotation, quality control of common variants, and PCA for ancestry assignment, see Supplementary Methods.

Polygenic risk score regression analyses

We used previously published gold-standard PRSs for height [2] and BMI [21]. Genotype positions and effect sizes were retrieved from the PGS catalog [22] (height: PGS002804, BMI: PGS000027). After performing standard common variant quality control, 1,004,205 variants were used in the height-PRS (91.4% of the 1,099,005 variants in PGS002804) and 2,062,833 variants were used in the BMI-PRS (98.2% of the 2,100,302 variants in PGS000027). The PRS for each individual in the cohort was calculated using PRSice-2 [23]. Both height and BMI PRSs were derived from entirely European cohorts, and upon testing these PRSs within the European and non-European subsets of the cohort in this study, we observed dramatically diminished effect sizes in the non-European subset (Supplementary Figure 3). We therefore opted to perform our primary analyses restricted to individuals of European ancestry.

We tested for associations between the respective PRS and the height or BMI measures using linear regression (lm() function in R) in 1) a univariable model and 2) a multivariable model that adjusted for phenotypic/demographic variables, 22q11.2 deletion extent (LCR22A–D vs LCR22A–B/LCR22A–C), sequencing platform/batch (categorical variable with four levels: TCAG HiSeqX vs TCAG Novaseq6000 vs CIDR NovaSeq6000 vs IBBC HiSeq2500 or HiSeqX), and the first four principal components (PCs) of ancestry.

1) BMI or sex-standardized height ~ corresponding PRS

2) BMI or sex-standardized height ~ corresponding PRS + sex^a,b + age + deletion extent^a + CHD^a + ID^a + psychotic illness^a+ sequencing platform/batch^a + PC1 + PC2 + PC3 + PC4

^aCategorical variable

^bOnly included in the regression model for BMI

To adjust for the large height difference between sexes, height was represented in the regression analyses as the difference from the individual’s height to the mean height of the individual’s sex (i.e., “sex-adjusted height”), a method previously used in the study generating height-PRS [2]. Sex was therefore not included as a covariate in the multivariable regression model for height. Categorical variables were treated as factors (as.factor() in R) and all continuous variables were standardized using the scale() function in R to produce beta coefficients for the regression analyses. For the sequencing platform/batch variable, CIDR NovaSeq6000 was used as the reference level in the regression model (and is thus not shown in Table 2). The increase in height or BMI per standard deviation increase in PRS was obtained from the “estimate” coefficient ± standard error from the lm() function in R, where only the PRS values were standardized and the raw values for sex-adjusted height (cm) or BMI (kg/m²) were used. The variance in height or BMI that explained the multivariable model was quantified by the multiple R² metric in R. The variance explained by the PRS variable alone in the multivariable model was represented by the difference in multiple R² (ΔR²) between the full multivariable model and the multivariable model without the PRS variable. An interaction between sex and PRS for height and BMI was tested using a sex*BMI interaction term in linear regression models.

Risk stratification for short stature using polygenic risk score

To assess the capacity of the height PRS to stratify risk for short stature, we first stratified our cohort by quintiles of height-PRS. Short stature was defined as less than the third percentile height, stratified by sex, based on World Health Organization growth curves at age 18 (https://www.dietitians.ca/Advocacy/Interprofessional-Collaborations-(1)/WHO-Growth-Charts/WHO-Growth-Charts-Set-2). This corresponds to a height cut-off of less than 163 cm for males and 151 cm for females. A Fisher’s exact test was used to compare the proportion of individuals with short stature in the lowest vs highest quintiles of height-PRS.

Furthermore, we constructed receiver operator characteristic (ROC) curves using logistic regression models predicting short stature to calculate sensitivity and specificity of these models without the use of arbitrary cut-offs. We compared three logistic regression models:

1) Covariate only: short stature* ~ sex* + age + deletion extent* + CHD* + ID* + psychotic illness*

2) PRS only: short stature* ~ PRS

3) PRS + covariates: short stature* ~ height-PRS + sex* + age + deletion extent* + CHD* + ID* + psychotic illness*

*binary variable

ROC curves were created using the R package “pROC”. The difference between the area under the curve (AUC) of two ROC curves was compared using Delong’s test for two correlated ROC curves and the optimal sensitivity and specificity of each ROC curve was determined using Youden’s J statistic.

All statistical analyses were performed using R version 4.0.3. Statistical significance was defined as p<0.05. All test performed were two-tailed. P-values were not adjusted for multiple testing.

Cohort description of individuals with a 22q11.2 microdeletion

Table 1 summarizes the clinical, demographic, and 22q11.2 microdeletion extent features of 259 individuals (median age=30.9 years, interquartile range=23–41 years) of European ancestry with height and/or BMI data, along with genome sequencing data that passed quality control and other exclusion criteria (Supplementary Figures 1 and 2). Similar to previous phenotypic studies of overlapping cohorts, 19.4% had short stature (median height: 169.0 cm in males, 157.5 cm in females), and 45.7% had obesity (median BMI=29.0 kg/m²), representing large increases over the expected population prevalence for short stature of 3% [8] and an age-matched Canadian general population prevalence of obesity of ~25% [9]. The only sex differences in the 22q11.2 microdeletion cohort studied related to obesity, with females having significantly greater BMI and proportion with obesity than males, and the expected shorter height in females than males (Table 1).

Polygenic risk score and other predictors of height and BMI

We first assessed whether PRSs for height and BMI developed in large general population cohorts [2,21] would be associated with their corresponding traits in individuals with a 22q11.2 microdeletion. The results showed that both PRSs were significant predictors of their respective traits, with each standard deviation increase in PRS corresponding to a 3.49 ± 0.40 cm (beta= 0.4804, p=2.66E-16) increase in sex-adjusted height and a 1.54 ± 0.49 kg/m² (beta=0.1900, p=0.0018) increase in BMI, in univariable linear regression models (Table 2).

The PRSs for both sex-adjusted height (beta=0.5444, p<2.00E-16 ) and BMI (beta=0.2429, p=8.01E-05) remained significant predictors of their respective traits in multivariable linear regression models that adjusted for sex (covariate only in the BMI model), age, 22q11.2 deletion extent (LCR22A–D vs LCR22A–B/A–C), CHD, intellectual disability, psychotic illness, sequencing platform/batch, and the first four principal components of ancestry (Table 2). Other significant predictors (all p<0.05) of shorter sex-adjusted height included longer (LCR22A–D) microdeletion extent, CHD, and intellectual disability. The first PC of ancestry was also significantly associated with height. Other significant predictors of higher BMI included female sex and older age (Table 2). The multivariable models explained 33.9% and 14.7% of the variance (R²) in height and BMI, respectively, with the PRS variables alone (ΔR²) explaining 25.8% and 5.7% of the variance (Table 2). There was no interaction between sex and PRS identified for either height (p= 0.218) or BMI (p= 0.702) with both PRSs having a similar effect size within sex (Supplementary Figures 5A and 6A).

Using height polygenic risk score to stratify risk for short stature

Although the PRSs for height and BMI were both significantly associated with their respective traits in linear regression models, the effect size for the height-PRS (beta=0.480, ΔR²=0.258) was substantially greater than for the BMI-PRS (beta=0.190, ΔR²=0.057) (Table 2). Therefore, we decided to focus on assessing the capacity of the height-PRS to stratify risk for clinically defined adult short stature.

We first stratified our cohort by quintiles of height-PRS (Figure 1). Across each increase in height-PRS quintile, we observed an increase in median height (Figure 1A) and a corresponding stepwise decrease in the prevalence of short stature (Figure 1B). Among those in the lowest height-PRS quintile, 42.3% (n=22/52) had short stature compared to 5.9% (n=3/51) in the highest height-PRS quintile (odds ratio=11.46, Fisher’s exact p=1.74E-05).

Furthermore, we constructed ROC curves using logistic regression models predicting short stature (Figure 2). The addition of height-PRS to a covariate-only model (i.e., PRS + covariates model) produced a significant increase in the area under the curve (AUC) compared to the covariate-only model (AUC 0.78 vs 0.62, p=0.0011) (Supplementary Table 1). This combined (PRS + covariates) model was the best performing model with an optimal sensitivity and specificity of 0.82 and 0.65, respectively. As well, the PRS-only model (AUC=0.73) had a higher AUC than the covariate-only model (AUC=0.62) (Figure 2), although this difference did not reach statistical significance (p=0.1124) (Supplementary Table 1).

Adults with a 22q11.2 microdeletion have a baseline predisposition to shorter stature and obesity [8,9]. The results of this study demonstrate that adult height and BMI are further modified by genome-wide polygenic risk, while accounting for other clinical and demographic variables, and that the addition of height-PRS significantly improves risk prediction for short stature compared to a model with only clinical/demographic covariates. Those in the lowest quintile of height-PRS were at approximately 11 times the odds of having short stature than those in the top height-PRS quintile.

Although height has not previously been investigated, the ability of genome-wide polygenic risk to modify the expression of this high-impact variant identified in this study is consistent with some previous studies of 22q11.2 microdeletion [24]. PRS has been reported to modify the baseline risk conferred by the 22q11.2 microdeletion for expression of schizophrenia [18], lower IQ/intellectual disability [19], and elevated blood triglyceride levels/hypertriglyceridemia [20]. Also, the results for the BMI-PRS on BMI for 22q11.2 microdeletion are consistent with a previous report of PRS modifying the BMI-lowering effect of a 16p11.2 duplication and the BMI-raising effect of a 16p11.2 deletion [25].

Collectively, these results support the notion that PRS may eventually have potential utility for estimating the penetrance of a pathogenic rare variant at an individual level [24]. For example, individuals with a pathogenic LDLR mutation who are represented in the highest percentiles of coronary artery disease (CAD)-PRS have an exceedingly high risk for CAD, whereas those with an LDLR mutation in the lowest percentiles of CAD-PRS have a risk for CAD that is similar to the general population average [26]. Similarly, in the current study, individuals with a 22q11.2 microdeletion in the lowest quintile of height-PRS had a prevalence of short stature (42%) about 14 times the defined general population norm of 3%, while the prevalence of short stature for those in the highest height-PRS quintile (6%) is closer to the population expectation.

In addition to the influence of PRS, there were significant effects of other clinical, demographic or genetic covariates (Table 2, Supplementary Figures 5 and 6). Consistent with results from an overlapping sample in our phenotype-only study [8], female sex, LCR22A–D 22q11.2 microdeletion (vs the proximal nested LCR22A–B or A–C microdeletions), CHD, and intellectual disability were associated with shorter height. Of note, ~85% of individuals with a 22q11.2 microdeletion have the longer LCR22A–D deletion [10], which has previously been associated with a slightly more severe impact on cognitive functioning compared to the proximal nested deletion extents [27]. For BMI, older age was associated with higher BMI, as expected. We also found that female sex was associated with higher BMI (Table 2, Supplementary Figure 6A), consistent with general population trends in most countries [28], but not previously reported in 22q11.2DS [9].

Although these general population-derived PRSs remained associated with height and BMI in individuals with a 22q11.2 microdeletion, it is possible that the effect size of the PRS differs from when it is applied to general population samples (i.e., there may be a potential interaction effect between the 22q11.2 microdeletion and PRS). For height, the variance explained by the PRS in this study (ΔR²) of 25.8% is notably lower than the most closely comparable R² value of ~40% reported in the source study [2], suggesting a possibly reduced effect size of the PRS in those with a major risk already conveyed by a 22q11.2 microdeletion. However, these R² values are not directly comparable for many reasons, including the use of different covariates in the multivariable regression models, different genotyping platforms, non-identical statistical analysis methodologies and programs, and the use of a PRS that is not entirely identical in SNP composition to the published PRS (see PRS section in Methods). To more reliably assess for a potential interaction, a direct comparison to a general population sample should be performed using matched genotyping, variant calling, and analysis pipelines [18].

Advantages and Limitations

To our knowledge, this study involves contains the largest cohort available of individuals with a 22q11.2 microdeletion with available adult height, BMI, and genome sequencing data. The deep phenotyping enabled identification of significant clinical and demographic predictors and a comparison of the relative predictive value of these variables with the PRS. This study also demonstrates the value of studying a clinical cohort ascertained using a genotype-first method for the purpose of jointly studying effects of a pathogenic rare variant and common genetic variants [29]. Such a study that investigates the genomic and phenotypic heterogeneity for a single pathogenic CNV would not be possible using current large-scale biobank data due to the recognized selection bias against individuals with rare high-impact CNVs (e.g., the UK BioBank has just 10 individuals with a pathogenic 22q11.2 deletion) [29,30].

Due to the rarity of the 22q11.2 microdeletion, particularly in adults, our sample size is still relatively small [2,21]. This was most limiting when stratifying risk by PRS quintiles (with each bin containing only 51–52 individuals), and for establishing reliable associations for covariates with more borderline associations (Table 2). As for all current PRS studies, a critical limitation is the reduced effect size of the available PRS for non-European populations [31] (Supplementary Figure 3), which unfortunately compelled us to exclude the non-European subset of our sample from the primary analyses (Supplementary Figure 1). We also note that there was no notable improvement among the non-European subset when we used the multi-ancestry height-PRS (PGS002802). Additionally, given the preliminary nature of these results, still limited predictive accuracy, and absence of any consensus on potential clinical actionability, the findings do not support any clinical use of these PRSs at this time. Future studies are also needed to assess for the contribution of additional factors such as other genome-wide rare variants and lifestyle factors in 22q11.2DS.

In summary, PRSs for height and BMI are associated with expression of their respective traits in individuals with a 22q11.2 microdeletion where a priori risks for short stature and obesity are elevated. The results demonstrate that genome-wide common variants can modify the major effects of a rare CNV on adult height, even while accounting for other contributory factors. While promising for height, further research will be needed before polygenic risk achieves individual-level clinical utility as a predictor of this adult outcome for this high-impact rare variant.

Ethics approval and consent to participate

All participants provided written informed consent and the study was approved by the local research ethics boards of hospitals affiliated with the University of Toronto (the Centre for Addiction and Mental Health and University Health Network) where participants were ascertained (e.g., the Dalglish Family 22q Clinic at the Toronto General Hospital), with referrals originating from generalists and specialists. All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Availability of data and materials

The datasets generated and/or analysed during the current study, including the raw genome sequencing and clinical data, are not publicly available due to the sensitive nature of this data. The polygenic risk scores used in this study are listed in the Methods and are publicly available in the PGS Catalog (https://www.pgscatalog.org/). The code and calculated individual-level PRSs used for analyses are available from the corresponding author upon reasonable request.

Competing interests

None of the authors have any conflicts to report.

Funding

This work was supported by the Canadian Institutes of Health Research (CIHR) (PJT-169161, PJT-148924). ASB holds the Dalglish Chair in 22q11.2 Deletion Syndrome at the University Health Network and University of Toronto. Some genome sequencing was originally done as part of the International 22q11.2DS Brain and Behavior Consortium and National Institute of Mental Health grant U01MH0101720, with recalling of sequencing data supported by a McLaughlin Centre Accelerator Grant.

Authors' contributions

SY and ASB conceived the study. SY performed all data analyses and wrote the manuscript with contributions from ASB and TH. BT performed all pre-processing of the genome sequencing data and wrote part of the Supplementary Methods. TH retrieved and curated the phenotype data. SY, TH, BT, BM, RKCY, and ASB revised the manuscript. ASB managed, recruited, and examined all participants.

Loos RJF, Yeo GSH. The genetics of obesity: from discovery to biology. Nat Rev Genet. 2022;23:120–33.
Yengo L, Vedantam S, Marouli E, Sidorenko J, Bartell E, Sakaue S, et al. A saturated map of common genetic variants associated with human height. Nature. 2022;610:704–12.
Huang J, Huffman JE, Huang Y, Do Valle Í, Assimes TL, Raghavan S, et al. Genomics and phenomics of body mass index reveals a complex disease network. Nat Commun. 2022;13:7973.
van Duyvenvoorde HA, Lui JC, Kant SG, Oostdijk W, Gijsbers AC, Hoffer MJ, et al. Copy number variants in patients with short stature. Eur J Hum Genet. 2014;22:602–9.
Zahnleiter D, Uebe S, Ekici AB, Hoyer J, Wiesener A, Wieczorek D, et al. Rare Copy Number Variants Are a Common Cause of Short Stature. PLOS Genetics. 2013;9:e1003365.
Dauber A, Yu Y, Turchin MC, Chiang CW, Meng YA, Demerath EW, et al. Genome-wide Association of Copy-Number Variation Reveals an Association between Short Stature and the Presence of Low-Frequency Genomic Deletions. The American Journal of Human Genetics. 2011;89:751–9.
D′Angelo CS, Koiffmann CP. Copy Number Variants in Obesity-Related Syndromes: Review and Perspectives on Novel Molecular Approaches. Journal of Obesity. 2012;2012:845480.
Heung T, Conroy B, Malecki S, Ha J, Boot E, Corral M, et al. Adult Height, 22q11.2 Deletion Extent, and Short Stature in 22q11.2 Deletion Syndrome. Genes. 2022;13:2038.
Voll SL, Boot E, Butcher NJ, Cooper S, Heung T, Chow EWC, et al. Obesity in adults with 22q11.2 deletion syndrome. Genet Med. 2017;19:204–8.
McDonald-McGinn DM, Sullivan KE, Marino B, Philip N, Swillen A, Vorstman JAS, et al. 22q11.2 deletion syndrome. Nat Rev Dis Primers. 2015;1:1–19.
Blagojevic C, Heung T, Theriault M, Tomita-Mitchell A, Chakraborty P, Kernohan K, et al. Estimate of the contemporary live-birth prevalence of recurrent 22q11.2 deletions: a cross-sectional analysis from population-based newborn screening. CMAJ Open. 2021;9:E802–9.
Blagojevic C, Heung T, Malecki S, Ying S, Cancelliere S, Hegele RA, et al. Hypertriglyceridemia in young adults with a 22q11.2 microdeletion. Eur J Endocrinol. 2022;187:91–9.
Van L, Heung T, Malecki SL, Fenn C, Tyrer A, Sanches M, et al. 22q11.2 microdeletion and increased risk for type 2 diabetes. EClinicalMedicine. 2020;26:100528.
Malecki SL, Mil SV, Graffi J, Breetvelt E, Corral M, Boot E, et al. A genetic model for multimorbidity in young adults. Genet Med. 2020;22:132–41.
Gur RE, Bassett AS, McDonald-McGinn DM, Bearden CE, Chow E, Emanuel BS, et al. A neurogenetic model for the study of schizophrenia spectrum disorders: the International 22q11.2 Deletion Syndrome Brain Behavior Consortium. Mol Psychiatry. 2017;22:1664–72.
Bassett AS, Lowther C, Merico D, Costain G, Chow EWC, van Amelsvoort T, et al. Rare genome-wide copy number variation and expression of schizophrenia in 22q11.2 deletion syndrome. Am J Psychiatry. 2017;174:1054–63.
Ying S, Heung T, Zhang Z, Yuen RKC, Bassett AS. Schizophrenia risk mediated by microRNA target genes overlapped by genome-wide rare copy number variation in 22q11.2 deletion syndrome. Front Genet. 2022;13:812183.
Cleynen I, Engchuan W, Hestand MS, Heung T, Holleman AM, Johnston HR, et al. Genetic contributors to risk of schizophrenia in the presence of a 22q11.2 deletion. Mol Psychiatry. 2021;26:4496–510.
Davies RW, Fiksinski AM, Breetvelt EJ, Williams NM, Hooper SR, Monfeuga T, et al. Using common genetic variation to examine phenotypic expression and risk prediction in 22q11.2 deletion syndrome. Nat Med. 2020;26:1912–8.
Ying S, Heung T, Thiruvahindrapuram B, Engchuan W, Yin Y, Blagojevic C, et al. Polygenic risk for triglyceride levels in the presence of a high impact rare variant. BMC Medical Genomics. 2023;16:281.
Khera AV, Chaffin M, Wade KH, Zahid S, Brancale J, Xia R, et al. Polygenic Prediction of Weight and Obesity Trajectories from Birth to Adulthood. Cell. 2019;177:587-596.e9.
Lambert SA, Gil L, Jupp S, Ritchie SC, Xu Y, Buniello A, et al. The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation. Nat Genet. 2021;53:420–5.
Choi SW, O’Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scale data. GigaScience. 2019;8:giz082.
Lewis CM, Vassos E. Polygenic risk scores: from research tools to clinical instruments. Genome Medicine. 2020;12:44.
Oetjens MT, Kelly MA, Sturm AC, Martin CL, Ledbetter DH. Quantifying the polygenic contribution to variable expressivity in eleven rare genetic disorders. Nat Commun. 2019;10:4897.
Fahed AC, Wang M, Homburger JR, Patel AP, Bick AG, Neben CL, et al. Polygenic background modifies penetrance of monogenic variants for tier 1 genomic conditions. Nat Commun. 2020;11:3635.
Zhao Y, Guo T, Fiksinski A, Breetvelt E, McDonald-McGinn DM, Crowley TB, et al. Variance of IQ is partially dependent on deletion type among 1,427 22q11.2 deletion syndrome subjects. American Journal of Medical Genetics Part A. 2018;176:2172–81.
Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19·2 million participants. The Lancet. 2016;387:1377–96.
Bassett AS, McDonald-McGinn DM, Boot E, Óskarsdóttir S, Yuen RKC. Approaches to studying the impact of 22q11.2 copy number variants. Am J Hum Genet. 2023;110:1216–8.
Crawford K, Bracher-Smith M, Owen D, Kendall KM, Rees E, Pardiñas AF, et al. Medical consequences of pathogenic CNVs in adults: analysis of the UK Biobank. Journal of Medical Genetics. 2019;56:131–8.
Martin AR, Kanai M, Kamatani Y, Okada Y, Neale BM, Daly MJ. Clinical use of current polygenic risk scores may exacerbate health disparities. Nat Genet. 2019;51:584–91.

Table 1. Height, BMI, and other phenotypic and genetic variables in individuals of European ancestry with a 22q11.2 microdeletion.

	Total sample^a,b			Males^b			Females^a			p-value^c
*Continuous variables*	n	Mean	SD	n	Mean	SD	n	Mean	SD
Height (cm)^a	258	162.7	9.5	119	169.3	7.2	139	157.0	7.4	2.20E-16
BMI (kg/m²)^a,b	258	29.8	7.9	118	28.8	6.9	140	30.8	8.5	0.0387
Age (years)	259	33.5	12.0	119	33.2	12.1	140	33.7	12.0	0.7037
*Categorical variables*	Total sample (n=259)^a,b			Total Males (n=119)^b			Total Females (n=140)^a			p-value^c
Individuals with	n	%		n	%		n	%
Short stature^a,d	50	19.4		22	18.5		28	20.1		0.7548
Obesity^b,e	118	45.7		43	36.4		75	53.6		8.31E-03
LCR22A–D deletion extent^f	236	91.1		112	94.1		124	88.6		0.1306
Congenital heart disease^g	35	13.5		17	14.3		18	12.9		0.8556
Intellectual disability^h	19	7.3		11	9.2		8	5.7		0.3414
Psychotic illnessⁱ	94	36.3		49	41.2		45	32.1		0.1541

SD Standard deviation, BMI body mass index. P-values reaching statistical significance (p<0.05) are bolded.

^aOne female (128.0 cm) was excluded from height analyses due to being an outlier in the height distribution (Supplementary Figure 4). This individual had a BMI of 20.5 kg/m² and was not an outlier in the BMI distribution.

^bOne male (63.6 kg/m²) was excluded from BMI and obesity analyses due to being an outlier in the BMI distribution (Supplementary Figure 4).

^cStatistical analyses comparing males and females; t-test for continuous variables; Fisher’s exact test for categorical values. All tests were two-tailed.

^dDefined as less than the third percentile for height, by sex, at age 18 years, based on World Health Organization growth curves (cut-offs: 163 cm for males, 151 cm for females).
^eDefined as BMI ≥ 30.

^fThose without an LCR22A–D deletion had either a LCR22A–B or LCR22A–C deletion.

^gDefined as individuals with major (moderate–severe) congenital heart disease.

^hDefined as individuals with moderate–severe intellectual disability.

ⁱDefined as individuals with schizophrenia or schizoaffective disorder.

Table 2. Linear regression analyses testing polygenic risk score (PRS) as a predictor of height or BMI, using a univariable model and using a multivariable model adjusting for phenotypic, 22q11.2 microdeletion extent, sequencing platform/batch, and ancestry variables.

	Height (n=158)^a			BMI (n=158)
	beta	std error	p	beta	std error	p
*Univariable model*
PRS	0.3492	0.0570	3.27E-09	0.1900	0.0601	1.76E-03
*Multivariable model*
PRS	0.5444	0.0557	2.00E-16	0.2429	0.0605	8.01E-05
Female sex^a	–	–	–	0.2787	0.1189	0.0199
Age	-0.1071	0.0572	0.0623	0.1342	0.0633	0.0351
LCR22A–D deletion extent^b	-0.5260	0.1871	0.0053	-0.2895	0.2113	0.1720
Congenital heart disease	-0.5096	0.1605	0.0017	-0.1507	0.1758	0.3920
Intellectual disability	-0.3971	0.2012	0.0495	0.0166	0.2227	0.9406
Psychotic illness	-0.0336	0.1199	0.7798	0.2050	0.1329	0.1243
IBBC_HiSeq2500/X^c	0.0633	0.2918	0.8284	0.2919	0.3247	0.3696
TCAG_HiSeqX^c	0.1352	0.1144	0.2384	0.1613	0.1277	0.2076
TCAG_NovaSeq6000^c	0.3597	0.2084	0.0857	-0.1263	0.2267	0.5781
Ancestry PC1	-0.1681	0.0640	0.0092	0.0348	0.0694	0.6159
Ancestry PC2	-0.0817	0.0715	0.2543	0.1029	0.0752	0.1724
Ancestry PC3	0.0462	0.0656	0.4820	-0.0337	0.0724	0.6417
Ancestry PC4	-0.1332	0.0709	0.0614	-0.0560	0.0805	0.4875
	R²	p		R²	p
Model	0.3386	4.27E-16		0.1468	3.05E-04
	ΔR²			ΔR²
PRS variable	0.2583			0.0565

Beta coefficients represent the standard deviation change in the dependent variable (height or BMI) per one standard deviation increase for continuous independent variables (PRS, age, and ancestry PC1–4), or when compared to the reference category for categorical independent variables (sex, deletion extent, intellectual disability, and sequencing platform/batch^c). ΔR² is an estimate of the variance explained by the PRS variable alone within a multivariable model (see Methods). P values reaching statistical significance (p<0.05) are bolded.

PRS polygenic risk score, BMI body mass index, std standard, PC principal component

^aTo adjust for the large height difference between sexes (Table 1), height was represented in the regression model as the difference (cm) between the individual’s height and the mean height of the individual’s sex. Therefore, sex was not included in the multivariable regression model for height[2].

^bThose who did not have the most common LCR22A–D 22q11.2 microdeletion were categorized as having a proximal nested (i.e., LCR22A–B/LCR22A–C) 22q11.2 microdeletion.

^cTo adjust for sequencing platform/batch, one categorical variable with four levels (IBBC_HiSeq2500/X, TCAG_HiSeqX, TCAG_NovaSeq6000, and CIDR_NovaSeq6000) was added to the linear regression model. CIDR_NovaSeq6000 was used as the reference category, and is thus not shown.

No competing interests reported.

Influence of polygenic risk on height and BMI in adults with a 22q11.2 microdeletion

Status:

Version 1

Abstract

Figures

INTRODUCTION

METHODS

RESULTS

DISCUSSION

Conclusion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Version 1