Advertisement

Copy Number Variants increasing risk for schizophrenia: shared and distinct effects on brain morphometry and cognitive performance

Open AccessPublished:October 28, 2022DOI:https://doi.org/10.1016/j.bpsgos.2022.10.006

      Abstract

      Background

      CNVs conferring risk for mental disorders are associated with brain changes and cognitive deficits. However, whether these effects are shared or distinct across CNVs remains untested. Here we compared the effects on brain morphometry and cognitive performance across CNVs with shared psychiatric liability.

      Methods

      Unaffected and unrelated participants of white British and Irish ancestry were drawn from the UK Biobank. After quality control we retained 31,941 participants not carrying any damaging CNV and 202 carrying one CNV increasing risk for schizophrenia. Using regression analyses we tested the association between brain morphometry and cognitive performance with CNV carrying status and compared these effect sizes across CNVs using Z-test for the equality of regression coefficients. Equation modelling was used to examine the mediation of brain phenotypes on the association between CNVs and cognitive performance.

      Results

      We detected different patterns of association between CNVs and brain morphometry and cognitive abilities. Comparing across CNVs, 1q21.1 deletion showed the strongest association with surface area in frontal lobe (beta=-1.03, p=4x10-8; beta=-0.81, p=.00001) and performance in Digit Memory (beta=-1.58, p=.00003); while 1q21.1duplication did with volume on the putamen (beta=-0.70 p=.0004) and Reaction Time (beta=-1.14, p=.000002). We also showed that even when two CNVs were associated with performance in the same cognitive ability, these associations were mediated by different brain changes.

      Conclusions

      Despite sharing similar psychiatric liability, the CNVs under study appeared to have different effects on brain morphometry and on performance in cognitive abilities, suggesting the existence of distinctive neurobiological pathways into the same clinical phenotypes.

      Keywords

      Introduction

      Several recurrent Copy Number Variation (CNV) – i.e. deletions or duplications of large DNA segments resulting from non-allelic homologous recombination at meiosis – have been identified as risk factors for neurodevelopmental disorders(
      • Cooper G.M.
      • Coe B.P.
      • Girirajan S.
      • Rosenfeld J.A.
      • Vu T.H.
      • Baker C.
      • et al.
      A copy number variation morbidity map of developmental delay.
      ), with a small subgroup of these (i.e. thirteen to date) showing to also increase risk for schizophrenia(
      • Rees E.
      • Kendall K.
      • Pardiñas A.F.
      • Legge S.E.
      • Pocklington A.
      • Escott-Price V.
      • et al.
      Analysis of Intellectual Disability Copy Number Variants for Association With Schizophrenia.
      ,
      • Marshall C.R.
      • Howrigan D.P.
      • Merico D.
      • Thiruvahindrapuram B.
      • Wu W.
      • Greer D.S.
      • et al.
      Contribution of copy number variants to schizophrenia from a genome-wide study of 41,321 subjects.
      ). Improving our understanding of the shared and distinct phenotypic effects among these CNVs provides a unique opportunity to advance our knowledge of the neurobiological mechanisms behind mental disorders(e.g.(
      • Marshall C.R.
      • Howrigan D.P.
      • Merico D.
      • Thiruvahindrapuram B.
      • Wu W.
      • Greer D.S.
      • et al.
      Contribution of copy number variants to schizophrenia from a genome-wide study of 41,321 subjects.
      ,
      • Kirov G.
      • Pocklington A.J.
      • Holmans P.
      • Ivanov D.
      • Ikeda M.
      • Ruderfer D.
      • et al.
      De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia.
      ,
      • Pocklington A.J.
      • Rees E.
      • Walters J.T.
      • Han J.
      • Kavanagh D.H.
      • Chambert K.D.
      • et al.
      Novel Findings from CNVs Implicate Inhibitory and Excitatory Signaling Complexes in Schizophrenia.
      ,
      • Raznahan A.
      • Won H.
      • Glahn D.C.
      • Jacquemont S.
      Convergence and Divergence of Rare Genetic Disorders on Brain Phenotypes: A Review.
      )).
      Previous research has shown these CNVs to have a significant effect on brain morphometry and cognition, with carriers showing an apparent diversity of effects on subcortical and cortical morphometry(
      • Warland A.
      • Kendall K.M.
      • Rees E.
      • Kirov G.
      • Caseras X.
      Schizophrenia-associated genomic copy number variants and subcortical brain volumes in the UK Biobank.
      ,
      • Caseras X.
      • Kirov G.
      • Kendall K.M.
      • Rees E.
      • Legge S.E.
      • Bracher-Smith M.
      • et al.
      Effects of genomic copy number variants penetrant for schizophrenia on cortical thickness and surface area in healthy individuals: analysis of the UK Biobank.
      ,
      • Sønderby I.E.
      • van der Meer D.
      • Moreau C.
      • Kaufmann T.
      • Walters G.B.
      • Ellegaard M.
      • et al.
      1q21.1 distal copy number variants are associated with cerebral and cognitive alterations in humans.
      ,
      • Modenato C.
      • Kumar K.
      • Moreau C.
      • Martin-Brevet S.
      • Huguet G.
      • Schramm C.
      • et al.
      Effects of eight neuropsychiatric copy number variants on human brain structure.
      ), but a relatively common general decrease in cognitive abilities(
      • Sønderby I.E.
      • van der Meer D.
      • Moreau C.
      • Kaufmann T.
      • Walters G.B.
      • Ellegaard M.
      • et al.
      1q21.1 distal copy number variants are associated with cerebral and cognitive alterations in humans.
      ,
      • van der Meer D.
      • Sønderby I.E.
      • Kaufmann T.
      • Walters G.B.
      • Abdellaoui A.
      • Ames D.
      • et al.
      Association of Copy Number Variation of the 15q11.2 BP1-BP2 Region With Cortical and Subcortical Morphology and Cognition.
      ,
      • Kendall K.M.
      • Rees E.
      • Escott-Price V.
      • Einon M.
      • Thomas R.
      • Hewitt J.
      • et al.
      Cognitive Performance Among Carriers of Pathogenic Copy Number Variants: Analysis of 152,000 UK Biobank Subjects.
      ,
      • Thygesen J.H.
      • Presman A.
      • Harju-Seppänen J.
      • Irizar H.
      • Jones R.
      • Kuchenbaecker K.
      • et al.
      Genetic copy number variants, cognition and psychosis: a meta-analysis and a family study.
      ,
      • Guyatt A.L.
      • Stergiakouli E.
      • Martin J.
      • Walters J.
      • O'Donovan M.
      • Owen M.
      • et al.
      Association of copy number variation across the genome with neuropsychiatric traits in the general population.
      ). Importantly, previous studies have also shown some of the brain changes reported in carriers to partly mediate the association between carrier status and cognitive performance(
      • Warland A.
      • Kendall K.M.
      • Rees E.
      • Kirov G.
      • Caseras X.
      Schizophrenia-associated genomic copy number variants and subcortical brain volumes in the UK Biobank.
      ,
      • Sønderby I.E.
      • van der Meer D.
      • Moreau C.
      • Kaufmann T.
      • Walters G.B.
      • Ellegaard M.
      • et al.
      1q21.1 distal copy number variants are associated with cerebral and cognitive alterations in humans.
      ,
      • van der Meer D.
      • Sønderby I.E.
      • Kaufmann T.
      • Walters G.B.
      • Abdellaoui A.
      • Ames D.
      • et al.
      Association of Copy Number Variation of the 15q11.2 BP1-BP2 Region With Cortical and Subcortical Morphology and Cognition.
      ), informing about potential neurobiological pathways linking genetic risk and clinical phenotypes. However, the extend to which these CNVs show distinctive effects on brain and/or cognition indicating potential different neurobiological pathways into the same clinical phenotypes, remains unresolved. So far, studies comparing across different CNVs included small sample sizes(
      • Warland A.
      • Kendall K.M.
      • Rees E.
      • Kirov G.
      • Caseras X.
      Schizophrenia-associated genomic copy number variants and subcortical brain volumes in the UK Biobank.
      ,
      • Caseras X.
      • Kirov G.
      • Kendall K.M.
      • Rees E.
      • Legge S.E.
      • Bracher-Smith M.
      • et al.
      Effects of genomic copy number variants penetrant for schizophrenia on cortical thickness and surface area in healthy individuals: analysis of the UK Biobank.
      ), descriptively collated data from independent single studies or used methodologies not optimised to allow these comparisons(
      • Warland A.
      • Kendall K.M.
      • Rees E.
      • Kirov G.
      • Caseras X.
      Schizophrenia-associated genomic copy number variants and subcortical brain volumes in the UK Biobank.
      ,
      • Caseras X.
      • Kirov G.
      • Kendall K.M.
      • Rees E.
      • Legge S.E.
      • Bracher-Smith M.
      • et al.
      Effects of genomic copy number variants penetrant for schizophrenia on cortical thickness and surface area in healthy individuals: analysis of the UK Biobank.
      ,
      • Modenato C.
      • Kumar K.
      • Moreau C.
      • Martin-Brevet S.
      • Huguet G.
      • Schramm C.
      • et al.
      Effects of eight neuropsychiatric copy number variants on human brain structure.
      ,
      • Sønderby I.E.
      • Ching C.R.K.
      • Thomopoulos S.I.
      • van der Meer D.
      • Sun D.
      • Villalon-Reina J.E.
      • et al.
      Effects of copy number variations on brain structure and risk for psychiatric illness: Large-scale studies from the ENIGMA working groups on CNVs.
      ).
      In this study we aimed to investigate the association between carrying damaging CNVs and changes in brain morphometry and cognitive abilities in a sample of unaffected participants, placing the emphasis on cross-CNV comparisons. To this aim we focused on the subset of 13 CNVs that to date have shown to significantly increase risk for neurodevelopmental disorders but also for schizophrenia(
      • Rees E.
      • Kendall K.
      • Pardiñas A.F.
      • Legge S.E.
      • Pocklington A.
      • Escott-Price V.
      • et al.
      Analysis of Intellectual Disability Copy Number Variants for Association With Schizophrenia.
      ,
      • Marshall C.R.
      • Howrigan D.P.
      • Merico D.
      • Thiruvahindrapuram B.
      • Wu W.
      • Greer D.S.
      • et al.
      Contribution of copy number variants to schizophrenia from a genome-wide study of 41,321 subjects.
      ), potentially representing a more homogeneous group with regards of their psychiatric liability. We were also interested in examining whether the potential association between these CNVs and specific cognitive abilities could be mediated by different brain phenotypes (i.e. neurobiological pathways).

      Methods

      Participants

      We used a subsample of UK Biobank participants (www.ukbiobank.ac.uk) from whom brain MRI anatomical images (T1) were available at the time of the analyses (n= 47,927). From these we only retained unrelated participants (estimated kinship coefficient < 0.0442; i.e. 3rd degree relative, coefficient of relatedness < 12.5%, n= 1,489 lost to relatedness) of British/Irish ancestry (based on the first 3 population components compared to 1000G phase 3 super-population ancestries, n=3,998 lost to ancestry). We then excluded participants who self-reported to have received a diagnosis of schizophrenia or bipolar disorder by a doctor, or for whom hospital records for such diagnoses plus schizoaffective or ‘other-psychosis’ were documented (n= 472). Further exclusion criteria included evidence (self-reported or hospital records) of learning disability or of neurological conditions that could affect brain morphometry (n= 346) (see Supplemental table 1). After all the above genetic and health exclusion criteria, we retained 34,862 participants for whom the brain metrics of interest were available.
      All participants provided informed consent to participate in UK Biobank. Ethical approval was granted to the UK Biobank project by the North West Multi-Centre Ethics committee. Data were released to us after application project reference 17044.

      Genotyping and CNV calling

      Genotyping was performed using the Affymetrix UK BiLEVE Axiom array on an initial 50,000 participants, and the Affymetrix UK Biobank Axiom® array for the remaining participants. The two arrays are extremely similar (with over 95% common content). Sample processing at UK Biobank is described in their documentation (https://biobank.ctsu.ox.ac.uk/crystal/docs/genotyping_sample_workflow.pdf).
      CNV calling was conducted following the same procedure as described in a previous study(
      • Kendall K.M.
      • Rees E.
      • Escott-Price V.
      • Einon M.
      • Thomas R.
      • Hewitt J.
      • et al.
      Cognitive Performance Among Carriers of Pathogenic Copy Number Variants: Analysis of 152,000 UK Biobank Subjects.
      ). Briefly, normalised signal intensity, genotype calls and confidences were generated using ∼750,000 biallelic markers that were further processed with PennCNV-Affy software(
      • Wang K.
      • Li M.
      • Hadley D.
      • Liu R.
      • Glessner J.
      • Grant S.F.
      • et al.
      PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data.
      ). Individual samples were excluded if they had >30 CNVs, a waviness factor >0.03 or <-0.03 or call rate <96%. Individual CNVs were excluded if they were covered by <10 probes or had a density coverage of less than one probe per 20,000 base pairs.
      The list of CNVs considered in this study, their prevalence in our sample, the genomic coordinates of their critical regions along with their breakpoints can be found in supplemental table 2. These were manually inspected to confirm that they met our CNV calling criteria: we required a CNV to cover more than half of the critical interval and to include the key genes in the region (if known), or in the case of single gene CNVs, the deletions to intersect at least one exon and the duplications to cover the whole gene.
      Nine carriers of one of our target CNVs also carried at least one other damaging CNV (the criteria for defining damaging CNV has been previously fully described(
      • Kendall K.M.
      • Rees E.
      • Escott-Price V.
      • Einon M.
      • Thomas R.
      • Hewitt J.
      • et al.
      Cognitive Performance Among Carriers of Pathogenic Copy Number Variants: Analysis of 152,000 UK Biobank Subjects.
      )) and were excluded from our analyses. Only 6 of the preselected CNVs were present in more than 5 participants and were taken forwards into our analyses (see Table 1). As a control comparison group, we used individuals that carried none of the 90 CNVs defined as damaging due to their association with neurodevelopmental disorders(
      • Dittwald P.
      • Gambin T.
      • Szafranski P.
      • Li J.
      • Amato S.
      • Divon M.Y.
      • et al.
      NAHR-mediated copy-number variants in a clinical population: mechanistic insights into both genomic disorders and Mendelizing traits.
      ,
      • Coe B.P.
      • Witherspoon K.
      • Rosenfeld J.A.
      • van Bon B.W.
      • Vulto-van Silfhout A.T.
      • Bosco P.
      • et al.
      Refining analyses of copy number variation identifies specific genes associated with developmental delay.
      ) (non-CNV carriers, n= 31,941).
      Table 1Sociodemographic characteristics in CNV carriers and non-CNV carriers
      CNVAgeSexQualificationIncomeTownsend
      nmean(sd)%femalemean(sd)mean(sd)mean(sd)
      1q21.1del

      1q21.1dup

      NRXN1del

      15q11.2del

      16p13.11dup

      16p12.1del

      non-carriers
      10

      17

      11

      106

      45

      13

      31,941
      58.5(6.4)

      62.3(6.4)

      63.4(6.2)

      63.9(7.2)

      64.6(6.2)

      59.7(6.6)

      63.7(7.5)
      50.0%

      70.6%

      54.5%

      50.9%

      46.7%

      69.2%

      52.7%
      2.59(1.54)

      1.89(0.92)

      2.44 (1.33)

      2.98(1.80)

      2.98(1.94)

      2.62 (1.89)

      2.24(1.54)
      1.82(1.51)

      2.90(0.99)

      2.64(1.28)

      2.17(1.55)

      1.82(2.28)

      1.54(2.84)

      2.54(1.88)
      -0.28(3.94)

      -1.31(2.79)

      -2.80(2.16)

      -1.93(2.52)

      -2.02(2.50)

      -1.71(3.15)

      -1.98(2.66)
      *non-carriers: participants not carrying any of the 90 CNVs identified as damaging by previous research
      Qualification: highest educational qualification achieved (lower number indicates higher qualification with 1 equivalent to university or college degree); Income: Household annual income categorised by the UK Biobank in five bands, as follows: <£18,000, £18,000–30,999, £31,000–51,999, £52,000–100,000 and >£100,000; Townsend: Townsend Deprivation index assigned to participants based on their postal code at the time of recruitment (higher values indicate higher deprivation, with negative values related to affluent areas).

      Brain imaging and cognitive data

      Brain images were acquired using Siemens Skyra 3T scanners in UK Biobank’s imaging centres using identical acquisition protocols. T1-weighted brain images were processed using different procedures and a basic quality control was run on the raw images (documentation on data acquisition and processing is freely available from UK Biobank at https://biobank.ctsu.ox.ac.uk/crystal/ukb/docs/brain_mri.pdf). For our project we focused on estimates of subcortical volumes in mm3, mean cortical thickness in mm and surface area in mm2 for each gyrus based on the Desikan-Killiany (D-K) atlas parcellation obtained via the FreeSurfer v.5.3 software (https://surfer.nmr.mgh.harvard.edu). Average cortical thickness and total surface area were calculated for the frontal, parietal, temporal and occipital lobes, and the cingulate cortex. To avoid error values due to deficient segmentation of tissue types or parcellation into gyri, extreme values defined as ±3 standard deviations from the group mean were removed from the analyses.
      On the date of their scan, UK Biobank participants were invited to complete a cognitive test battery assessing different cognitive skills involving reasoning, memory and speed processing (https://biobank.ndph.ox.ac.uk/showcase/label.cgi?id=100026) that have shown moderate-to-high indices of validity(
      • Fawns-Ritchie C.
      • Deary I.J.
      Reliability and validity of the UK Biobank cognitive tests.
      ). Based on their distribution being closer to normality in our sample, from these we selected: Associative learning (number of words paired correctly), Card pairs matching (errors on second round), Digit memory (maximum number of digits remembered correctly), Fluid intelligence (total score), Reaction time (mean time to correctly identify matches), Trail making numeric (duration to complete), Trail making alphanumeric (duration to complete), Symbol digit matches (correct responses), Matrix patterns (number correct), Tower rearrangement (number correct).

      Analyses

      Regression analyses were used to test the association between CNV carrier status (predictor) and brain and cognitive phenotypes (outcome), while controlling for the effect of several confounders (i.e. age, sex, age*sex, testing centre, date attended and the first 10 population stratification PCs; regression models for brain phenotypes also included: intracranial volume, x, y and z brain position in the scanner, scanner table position, and head motion estimated from the rest functional MRI scan). The outcome variables were z-transformed prior to the analyses, and carrier status was coded as 0 ‘non-carrier v 1 ‘carrier’. Therefore, the regression coefficients (betas) represent the change of the outcome variable in standard deviations associated with carrying CNV under analysis. Bonferroni correction for the number of independent tests was applied for subcortical volumes (42 tests, p<.001), cortical measures (72 tests, p<.0007) and cognitive skills (60 tests, p<.0008).
      To compare the effects on brain and cognition across CNVs, the beta coefficients from the above regression analyses were compared using Z-test for the equality of regression coefficients.
      In order to examine the ability of brain morphometry to account for the association between CNV carrier status and cognitive performance, mediation analyses were conducted in a structural equation modelling framework using the Lavaan package in R(20). We applied this analysis to cognitive tests that showed significant association (p<.05) with at least two CNVs. For each of these, we entered as mediators those brain measures more strongly associated with performance on that cognitive test. To identify those, we first ran regression models for each cognitive test as outcome and all brain phenotypes and covariates as predictors; we retained for the mediation analyses those brain phenotypes that were significant in these models (supplemental table 3).

      Results

      Brain morphometry

      Subcortical volumes

      The 15q11.2deletion and 1q21.1duplication showed several significant associations with subcortical volumes at p<.05, in all cases indicating reduced volume in carriers. The 1q21.1deletion and 16p13.11duplication carriers showed increased hippocampal volume at p<.05. Only the associations of 15q11.2deletion and 1q21.1duplication with the volume of the putamen, and of 15q11.2deletion with the volume of the pallidum, remained significant after correction for multiple testing (p<.001) (Figure 1, Table 2). The comparison of beta coefficients across these associations showed the 1q21.1duplication to have the largest negative effect on putamen (v. 15q11.2deletion [p= .04], v. 16p13.11duplication [p=.01], and v. 16p12.1deletion [p=.07]) and hippocampus (v. 15q11.2deletion [p=.03] and v. 16p12.1deletion [p= 0.09]) (Figure 2). As per 15q11.2deletion, none of the betas found were significantly larger than those of any other CNV.
      Figure thumbnail gr1
      Figure 1Association (beta from regression analyses) between carrier status for each SZ-CNV present in the sample and subcortical volumes. Positive values indicate increased volume in carriers, negative values decreased volume in carriers. Error bars represent the 95% confidence interval of the beta values. Asterisks indicate significant association after correction for multiple testing (p<.001).
      Table 2Association (beta [p]) between CNV carrier status and subcortical volumes
      1q21.1del (n=10)1q21.1dup (n=17)NRXN1del (n=11)15q11.2del (n=106)16p13.11dup (n=45)16p12.1del (n=13)
      Thalamus0.33 (>.1)-0.30 (.07)0.01 (>.1)0.047 (>.1)0.03 (>.1)0.06 (>.1)
      Caudate0.20 (>.1)-0.44 (.03)0.21 (>.1)-0.20 (.01)0.09 (>.1)-0.27 (>.1)
      Putamen0.36 (>.1)-0.70 (<.001)*0.08 (>.1)-0.27 (<.001)*-0.16 (>.1)-0.18 (>.1)
      Pallidum0.19 (>.1)-0.56 (.003)0.16 (>.1)-0.31 (<.001)*-0.10 (>.1)-0.22 (>.1)
      Hippocampus0.53 (.03)-0.60 (.002)0.15 (>.1)-0.16 (.04)0.26 (.03)-0.11 (>.1)
      Amygdala0.27 (>.1)0.06 (>.1)0.08 (>.1)-0.14 (.06)0.10 (>.1)-0.17 (>.1)
      Accumbens-0.21 (>.1)-0.12 (>.1)-0.11 (>.1)-0.19 (.01)-0.07 (>.1)-0.19 (>.1)
      * Significant after Bonferroni correction based on 7 brain factors explaining >95% of the
      variance x 6 SZ CNVs = 42 tests
      Figure thumbnail gr2
      Figure 2Effect sizes (beta) for the effect of carrying a single SZ-CNV on brain morphometric markers that showed significant association with carrying status (carrier v non-CNV carrier). This figure allows direct comparison across different SZ-CNV. Betas indicate the difference in standard error between carrying a specific CNV v non carrying any pathological CNV. Error bars represent the standard error of the beta.
      Carriers of the 1q21.1deletion and duplication showed the opposite direction of effect (positive for deletion, negative for duplication) in all subcortical volumes but amygdala and accumbens. These differences reached significance in hippocampus (beta=-1.04, p=.0003), the putamen (beta=-1.02, p=.0005), the thalamus (beta=-0.60, p=.02) and the pallidum (beta=-0.65, p=.03) (supplemental table 4).

      Cortical thickness and surface area

      Carrying the 15q11.2deletion was associated with widespread thicker cortex in all lobes but the occipital; however, only the association with both parietal lobes and right temporal lobe remained significant after correction for multiple testing (p<.0007) (Figure 3, Table 3). The comparison of beta coefficients across CNVs, though, showed the effects of 15q11.2deletion to not be significantly different to those of most other CNVs.
      Figure thumbnail gr3
      Figure 3Association (beta from regression analyses) between carrier status for each SZ-CNV present in the sample and average cortical thickness (top) and total surface area (bottom) in left and right frontal, parietal, temporal and occipital lobes and cingulate cortex. Warmer colour indicates increased thickness/area in carriers, cooler colour indicates reduced thickness/area in carriers. Only brain areas with at least nominal significant results are coloured, asterisks indicate associations where the significant result survives correction for multiple testing (p<.0007).
      Table 3Association (beta [p]) between CNV carrier status and cortical thickness and surface area
      1q21.1del (n=10)1q21.1dup (n=17)NRXN1del (n=11)15q11.2del (n=106)16p13.11dup (n=45)16p12.1del (n=13)
      ThicknessLeftRightLeftRightLeftRightLeftRightLeftRightLeftRight
      Frontal.25(>.1)-.06(>.1)-.01(>.1).25(>.1).16(>.1).35(>.1).26(.006).24(.012)-.04(>.1)-.13(>.1).04(>.1)-.29(>.1)
      Parietal.22(>.1)-.05(>.1)-.08(>.1)-.15(>.1).39(>.1).31(>.1).39(<.0007)*.40(<.0007)*-.03(>.1)-.04(>.1).07(>.1)-.09(>.1)
      Temporal-.19(>.1)-.06(>.1)-.29(>.1).15(>.1).52(.08).35(>.1).17(.07).38(<.0007)*.14(>.1)-.01(>.1)-.24(>.1)-.25(>.1)
      Occipital-.34(>.1)-.67(.03)-.17(>.1).29(>.1).34(>.1)-.11(>.1).08(>.1).11(>.1)-.04(>.1)-.002(>.1)-.17(>.1).03(>.1)
      Cingulate-.36(>.1).13(>.1).27(>.1).13(>.1).24(>.1).24(>.1).01(>.1)-.01(>.1).04(>.1)-.21(>.1).01(>.1)-.08(>.1)
      Area
      Frontal-1.03(<.0007)*-.81(<.0007)*.16(>.1).49(<.0007)*-.21(>.1)-.38(.02)-.15(.007)-.13(.02).04(>.1).02(>.1)-.29(.07)-.22(>.1)
      Parietal-.47(.02)-.31(>.1)-.02(>.1).01(>.1)-.18(>.1)-.24(>.1)-.15(.01)-.08(>.1)-.08(>.1)-.005(>.1)-.03(>.1)-.02(>.1)
      Temporal-.69(<.0007)*-.84(<.0007)*.12(>.1).19(>.1)-.54(.004)-.34(.06)-.09(>.1).02(>.1)-.11(>.1)-.03(>.1)-.16(>.1)-.12(>.1)
      Occipital-.37(>.1)-.41(>.1).02(>.1).19(>.1)-.13(>.1).16(>.1)-.17(.03)-.17(.02)-.21(.09)-.11(>.1)-.01(>.1)-.08(>.1)
      Cingulate-.47(.04)-.66(.007).02(>.1).08(>.1)-.15(>.1)-.06(>.1)-.10(>.1)-.18(.01).11(>.1)-.16(>.1)-.28(>.1)-.43(.04)
      * Significant after Bonferroni correction based on 12 brain factors explaining >95% of the
      variance x 6 SZ CNVs = 72 tests
      Table 4Association (beta [p]) between CNV carrier status and cognitive skills
      1q21.1del (n=10)1q21.1dup (n=17)NRXN1del (n=11)15q11.2del (n=106)16p13.11dup (n=45)16p12.1del (n=13)
      Word pairs-.71 (.06)-.42 (>.1)-.75 (.03)-.07 (>.1)-.19 (>.1)-.28 (>.1)
      Card match.-.62 (.05).10 (>.1)-.65 (.03)-.14 (>.1)-.16 (>.1).44 (>.1)
      Digit memory-1.5 (<.0008)*-.33 (>.1)-.56 (>.1)-.23 (.05).12 (>.1)-.68 (.04)
      Fluid intel.-.65 (.04)-.18 (>.1)-.44 (>.1)-.34 (<.0008)*-.47 (.002)-.72 (.01)
      Reaction time.08 (>.1)-1.1 (<.0008)*-.59 (.04)-.23 (.01)-.10 (>.1).12 (>.1)
      Trial num.-.08 (>.1)-.71 (.01)-.28 (>.1).06 (>.1).12(>.1).07 (>.1)
      Trial alph..21 (>.1).009 (>.1)-.25 (>.1)-.36 (.001)-.12 (>.1)-.21(>.1)
      Symbol/Digit-.56 (>.1)-.41 (>.1)-.62 (.06)-.14 (>.1).01 (>.1)-.55 (.07)
      Matrix-.35 (>.1)-.001 (>.1)-.42 (>.1)-.33 (.005)-.33 (.06)-.01 (>.1)
      Tower-.43 (>.1)-.43 (>.1)-.32 (>.1)-.44 (<.0008)*-.28 (>.1)-.33 (>.1)
      * Significant after Bonferroni correction based on 10 cognitive tests explaining >95% of the
      variance x 6 SZ CNVs = 60 tests
      Several SZ-CNV showed nominal associations with surface area indicating reduced area in carriers, except for 1q21.1duplication that showed the opposite direction of effect. However, only the effects of 1q21.1deletion on frontal and temporal lobes bilaterally, and of 1q21.1duplication on the right frontal lobe, survived correction for multiple testing (p<.0007) (Figure 3, Table 3). The comparison of beta coefficients showed that the negative effects of 1q21.1deletion on left and right frontal lobes were significantly stronger than those of any other CNVs: v. 15q11.2deletion (p=1x10-5 and p=.0005; left and right respectively), v. 16p12.1deletion (p=.002 and p=.014; left and right respectively), and v. NRNX1deletion on the left frontal lobe (p=.001) but not on the right (p= .1) (Figure 2). Likewise, the negative effects of 1q21.1deletion on left and right temporal lobe were among the largest detected: v. 16p12.1deletion (p= .04 and p= .003; left and right respectively), and v. NRXN1 deletion on the right (p=.06) but not the left (p> .1) (Figure 2).
      Results from a more fine-grain parcellation of the cortex in gyri based on the Desikan-Killiany atlas (Supplemental table 5) confirmed the results above: widespread positive association of 15q11.2deletion with thickness predominantly in frontal and parietal cortices, albeit with effect sizes not larger than those of other CNVs. Also, 1q21.1deletion showing its stronger effects in giry within the frontal and temporal lobes, these being among the strongest effects found across the board.
      Finally, 1q21.1deletion and duplication carriers showed againt mostly opposing directions of effects on frontal and temporal surface areas (reductions in deletion carriers and increases in duplication carriers) (Supplemental table 6).

      Cognitive performance

      Multiple significant associations at p<.05 were found between CNVs and cognitive abilities (Figure 4), in all cases indicating poorer performance in carriers. However, only four of these associations survived correction for multiple testing: 1q21.1del with Digit memory (beta=-1.58, p=.00003), 1q21.1dup with Reaction time (beta=-1.14, p=.000002), and 15q11.2del with Fluid intelligence and Tower rearrangement (beta=-0.34, p=.0005 and beta=-0.44, p=.0001; respectively). The comparison of beta coefficients (Figure 5) showed the 1q21.1deletion to have the largest effect on Digit memory (v. 1q21.1duplication [p=.008], v. NRXN1deletion [p=.05], v. 15q11.2deletion [p=.001], and v. 16p12.1deletion [p=.07]); the 1q21.1duplication one of the largest on Reaction time (v. 15q11.2deletion [p=.0004] and v. 16p13.11duplication [p=.0001], but compared to NRXN1deletion [p>.1]). The effects of 15q11.2deletion on Fluid intelligence and Tower rearrangement were not statistically larger than those shown by any other CNVs (all p>.2).
      Figure thumbnail gr4
      Figure 4Association (beta from regression analyses) between carrier status for each SZ-CNV present in the sample and performance in cognitive tests. In all cases positive values indicate higher performance in carriers and negative values lower performance in carriers. Error bars represent the 95% confidence interval of the beta values. Asterisks indicate significant association after correction for multiple testing (p<.008).
      Figure thumbnail gr5
      Figure 5Effect sizes (beta) for the effect of carrying a single SZ-CNV on performance in cognitive tests that showed significant association with carrying status (carrier v non-CNV carrier). This figure allows direct comparison across different SZ-CNV. Betas indicate the difference in standard error between carrying a specific CNV v non carrying any pathological CNV. Error bars represent the standard error of the beta.

      Mediation analyses

      For Digit memory, three mediation models were independently tested for 1q21.1deletion, 15q11.2deletion and 16p12.1deletion (Supplemental figure 1). None of these resulted in any significant mediation effects.
      Mediation models for Fluid intelligence were tested for 15q11.2deletion, 16p13.11duplication and 16p12.1deletion (Supplemental figure 2). Hippocampal volume appeared to mediate the association between 15q11.2deletion (beta= -.009, p= .05) and performance in this cognitive test, also showing a trend to mediate this association for 16p13.11duplication (beta= .012, p=.06). Interestingly, these effects were in the opposite direction.
      Mediation models for Reaction time were tested for 1q21.1duplication, NRXN1deletion and 15q11.2deletion (Supplemental figure 3). For 1q21.1 duplication, volume of the thalamus (beta=0.019, p=.003), the pallidum (beta=0.019, p=.020) and the hippocampus (beta=0.018, p=.011) showed significant mediation effects. For 15q11.2deletion, volume of the pallidum (beta=0.010, p=.018) and area of the right cingulum (beta=-0.006, p=.039) mediated the association with Reaction time. No significant effects were found for the model including NRXN1deletion.

      Discussion

      The main aim of this study was to compare the effects of carrying different CNVs on brain morphometry and cognitive performance, and to examine the potential existence of different brain morphometric pathways mediating the association between CNVs and cognitive performance. Our results show that despite selecting a homogenous group of CNVs with regards of their psychiatric liability, these present with distinctive patterns of associations with brain morphometry and cognitive abilities. Moreover, we also show that even when two CNVs are associated with the same behavioural phenotype (i.e. performance in a cognitive test), these associations are mediated by different brain changes (i.e. neurobiological pathways).
      The methods in this study were optimised in several ways to allow for comparison of the effects on brain morphometry and cognitive performance across CNVs. Fist, since individual CNVs have different penetrance, which would have resulted in different ratios of affected/unaffected participants, by only including healthy participants we avoided confounding the comparison by reverse causation. Second, by focusing on a large sample drawn from a single cohort population (i.e. participants of British/Irish ancestry living in the UK) we minimised socio-demographic variability (Table 1). Third, variability in our phenotypes was minimised by applying the same acquisition/processing MRI protocol on data obtained in three identical scanners (Siemens Skyra 3T) and a unique cognitive testing protocol; furthermore, testing site, date of data acquisition and other potential MRI sources of noise were accounted for in our analyses(
      • Alfaro-Almagro F.
      • McCarthy P.
      • Afyouni S.
      • Andersson J.L.R.
      • Bastiani M.
      • Miller K.L.
      • et al.
      Confound modelling in UK Biobank brain imaging.
      ). Finally, we kept the non-CNV carrier group unchanged across the analyses for individual CNVs, allowing a direct comparison of their effect sizes. Whereas some recent studies have compared the effects on brain morphometry and cognition across CNVs(
      • Modenato C.
      • Kumar K.
      • Moreau C.
      • Martin-Brevet S.
      • Huguet G.
      • Schramm C.
      • et al.
      Effects of eight neuropsychiatric copy number variants on human brain structure.
      ,

      Sønderby IE, Ching CRK, Thomopoulos SI, van der Meer D, Sun D, Villalon-Reina JE, et al. (2021): Effects of copy number variations on brain structure and risk for psychiatric illness: Large-scale studies from the ENIGMA working groups on CNVs. Hum Brain Mapp.

      ), to our knowledge this is the first study applying an optimised design to support such comparison.
      Our results show important differences in the association of individual CNVs with brain morphometry. Whereas most rare variants showed mainly small non-significant effects on subcortical volumes, 1q21.1duplication showed its largest associations with volume of the striatum and hippocampus, these effects being statistically larger than those of most other CNVs. Moreover, 1q21.1duplication and the 15q11.2deletion showed the opposite direction of effect over hippocampal volume (reduction) compared to 1q21.1deletion and 16p13.11duplication (increase). This diversity was also evident in the cortex, where 15q11.2deletion was associated more strongly with cortical thickness, whereas 1q21.1deletion was with surface area; most other CNVs showing weak sparce associations with either measure. Reductions in surface area associated with 1q21.1deletion in frontal and temporal lobes were significantly larger than those of most other CNVs. Despite 15q11.2deletion showing several significant associations with subcortical and cortical measures, the effect sizes were not significantly larger than those of most other CNVs. The fact that 15q11.2deletion was the most prevalent CNV in our sample and, therefore, carried the largest statistical power, explained the larger number of significant results despite the lack of differences in effect size. This concurs with recent literature showing that 15q11.2deletion is one of the most prevalent rare variants in the general population, but with a lesser negative effect on health (
      • Crawford K.
      • Bracher-Smith M.
      • Owen D.
      • Kendall K.M.
      • Rees E.
      • Pardiñas A.F.
      • et al.
      Medical consequences of pathogenic CNVs in adults: analysis of the UK Biobank.
      ,
      • Calle Sánchez X.
      • Helenius D.
      • Bybjerg-Grauholm J.
      • Pedersen C.
      • Hougaard D.M.
      • Børglum A.D.
      • et al.
      Comparing Copy Number Variations in a Danish Case Cohort of Individuals With Psychiatric Disorders.
      ).
      We also investigated potential dose-effects of the deletion and the duplication at 1q21.1, finding several instances where deletion and duplication carriers showed the opposite direction of effect. However, only for hippocampal volume and surface area in right frontal lobe, dose-effects were confirmed by both carrier groups being statistically different from non-CNV carriers; replicating previous results from a multicentric study with clinical and non-clinical samples(
      • Sønderby I.E.
      • van der Meer D.
      • Moreau C.
      • Kaufmann T.
      • Walters G.B.
      • Ellegaard M.
      • et al.
      1q21.1 distal copy number variants are associated with cerebral and cognitive alterations in humans.
      ).
      We also found different association profiles across CNVs with cognitive performance: 15q11.2deletion was mostly associated with tasks involving executive function and reasoning (i.e. Fluid intelligence, Tower rearrangement, Matrix patterns and Trial making alphanumerical), 1q21.1duplication with performance on tasks tapping into speed processing (i.e. Reaction time and Trial making numerical), and 1q21.1deletion with working memory (i.e. Digit memory, Card matching and Words pairs). It is also interesting to note that at nominal level (p<.05), all CNVs showed some association with cognitive performance, which supports the idea that deeper/more detailed phenotyping of samples should allow a better understanding of the effects of rare variants on observable phenotypes. In this respect, the use of single clinical diagnoses might be limiting progress in the field. The deletion and the duplication CNVs at 1q21.1 showed the strongest detrimental effects on cognition: the deletion on working memory and the duplication on speed processing. The 15q11.2deletion benefited again of a higher statistical power (i.e. larger number of carriers), but none of the effect sizes found for this CNV were statistically different to the effects of most other CNVs. Interestingly, unlike for brain morphometry, no dose-effects were observed between the 1q21.1deletion and duplication for any cognitive ability.
      Finally, we showed that different brain morphometric measures (i.e. potentially distinct neurobiological mechanisms) could mediate the association between individual CNVs and cognitive abilities. For example, we found that hippocampal volume partly contributed to the lower performance in Reaction time of 15q11.2deletion carriers, whereas it showed a protective role in carriers of the 16p13.11duplication. This result highlights the fact that some of the brain changes seen in unaffected carriers of these CNVs – most likely those discordant with changes shown in affected participants - could be protective against developing mental disorders and warrant further investigation.
      Some limitations of this study should be highlighted. First, due to our strict selection criteria, the number of carriers for some individual CNV was limited, and consequently the statistical power to detect significant effects. As such, future research with larger samples of CNV carriers should be able to identify more subtle differences across CNVs. Second, also due to our selection criteria, rarer and potentially more penetrant CNVs were not present in our sample (eg. 22q11.2deletion). In this case, the joined effort of consortium like ENIGMA is important in accessing these participants. Third, despite that UK Biobank participants are in general healthier, wealthier and more educated than the UK general population(
      • Fry A.
      • Littlejohns T.J.
      • Sudlow C.
      • Doherty N.
      • Adamska L.
      • Sprosen T.
      • et al.
      Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population.
      ) and that our carriers did not differ from non-carriers in these measures, there is the possibility that subthreshold symptoms not accounted for could have partly explained some of the group differences found here. Fourth, we limited our brain measures to metrics of macrostructure and found some CNVs to show very little, if any, association with these. This should not be taken as a suggestion that these rare variants are unrelated to brain biomarkers, but rather that research focusing on other brain phenotypes such are neurite density, myelin content or network connectivity, should shed further light on the effects of these CNVs.
      In conclusion, we showed that CNVs with a rather homogenous psychiatric liability have important differential effects on brain phenotypes, with 1q21.1duplication and 1q21.1deletion showing the strongest effects respectively on the volume of striatum and hippocampus, and surface area in fronto-temporal cortices. All CNVs investigated showed negative effects on cognitive performance, although the strength of the effects differed across CNVs and cognitive abilities. Even when two CNVs were associated with the same cognitive phenotype, this association appeared mediated by different brain changes, suggesting the existence of different neurobiological pathways to the same phenotype.

      Uncited reference

      Rosseel Y (2012): lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software. 48:1 - 36.

      .

      Acknowledgements

      This research was conducted using the UK-Biobank resource under project ref. 17044 and was supported by the Medical Research Council Programme grant ref. G08005009.
      Disclosures
      The authors report no biomedical financial interests or potential conflicts of interest.

      References

        • Cooper G.M.
        • Coe B.P.
        • Girirajan S.
        • Rosenfeld J.A.
        • Vu T.H.
        • Baker C.
        • et al.
        A copy number variation morbidity map of developmental delay.
        Nat Genet. 2011; 43: 838-846
        • Rees E.
        • Kendall K.
        • Pardiñas A.F.
        • Legge S.E.
        • Pocklington A.
        • Escott-Price V.
        • et al.
        Analysis of Intellectual Disability Copy Number Variants for Association With Schizophrenia.
        JAMA Psychiatry. 2016; 73: 963-969
        • Marshall C.R.
        • Howrigan D.P.
        • Merico D.
        • Thiruvahindrapuram B.
        • Wu W.
        • Greer D.S.
        • et al.
        Contribution of copy number variants to schizophrenia from a genome-wide study of 41,321 subjects.
        Nat Genet. 2017; 49: 27-35
        • Kirov G.
        • Pocklington A.J.
        • Holmans P.
        • Ivanov D.
        • Ikeda M.
        • Ruderfer D.
        • et al.
        De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia.
        Mol Psychiatry. 2012; 17: 142-153
        • Pocklington A.J.
        • Rees E.
        • Walters J.T.
        • Han J.
        • Kavanagh D.H.
        • Chambert K.D.
        • et al.
        Novel Findings from CNVs Implicate Inhibitory and Excitatory Signaling Complexes in Schizophrenia.
        Neuron. 2015; 86: 1203-1214
        • Raznahan A.
        • Won H.
        • Glahn D.C.
        • Jacquemont S.
        Convergence and Divergence of Rare Genetic Disorders on Brain Phenotypes: A Review.
        JAMA Psychiatry. 2022;
        • Warland A.
        • Kendall K.M.
        • Rees E.
        • Kirov G.
        • Caseras X.
        Schizophrenia-associated genomic copy number variants and subcortical brain volumes in the UK Biobank.
        Mol Psychiatry. 2019;
        • Caseras X.
        • Kirov G.
        • Kendall K.M.
        • Rees E.
        • Legge S.E.
        • Bracher-Smith M.
        • et al.
        Effects of genomic copy number variants penetrant for schizophrenia on cortical thickness and surface area in healthy individuals: analysis of the UK Biobank.
        Br J Psychiatry. 2021; 218: 104-111
        • Sønderby I.E.
        • van der Meer D.
        • Moreau C.
        • Kaufmann T.
        • Walters G.B.
        • Ellegaard M.
        • et al.
        1q21.1 distal copy number variants are associated with cerebral and cognitive alterations in humans.
        Transl Psychiatry. 2021; 11: 182
        • Modenato C.
        • Kumar K.
        • Moreau C.
        • Martin-Brevet S.
        • Huguet G.
        • Schramm C.
        • et al.
        Effects of eight neuropsychiatric copy number variants on human brain structure.
        Transl Psychiatry. 2021; 11: 399
        • van der Meer D.
        • Sønderby I.E.
        • Kaufmann T.
        • Walters G.B.
        • Abdellaoui A.
        • Ames D.
        • et al.
        Association of Copy Number Variation of the 15q11.2 BP1-BP2 Region With Cortical and Subcortical Morphology and Cognition.
        JAMA Psychiatry. 2020; 77: 420-430
        • Kendall K.M.
        • Rees E.
        • Escott-Price V.
        • Einon M.
        • Thomas R.
        • Hewitt J.
        • et al.
        Cognitive Performance Among Carriers of Pathogenic Copy Number Variants: Analysis of 152,000 UK Biobank Subjects.
        Biol Psychiatry. 2017; 82: 103-110
        • Thygesen J.H.
        • Presman A.
        • Harju-Seppänen J.
        • Irizar H.
        • Jones R.
        • Kuchenbaecker K.
        • et al.
        Genetic copy number variants, cognition and psychosis: a meta-analysis and a family study.
        Mol Psychiatry. 2021; 26: 5307-5319
        • Guyatt A.L.
        • Stergiakouli E.
        • Martin J.
        • Walters J.
        • O'Donovan M.
        • Owen M.
        • et al.
        Association of copy number variation across the genome with neuropsychiatric traits in the general population.
        Am J Med Genet B Neuropsychiatr Genet. 2018; 177: 489-502
        • Sønderby I.E.
        • Ching C.R.K.
        • Thomopoulos S.I.
        • van der Meer D.
        • Sun D.
        • Villalon-Reina J.E.
        • et al.
        Effects of copy number variations on brain structure and risk for psychiatric illness: Large-scale studies from the ENIGMA working groups on CNVs.
        Hum Brain Mapp. 2022; 43: 300-328
        • Wang K.
        • Li M.
        • Hadley D.
        • Liu R.
        • Glessner J.
        • Grant S.F.
        • et al.
        PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data.
        Genome Res. 2007; 17: 1665-1674
        • Dittwald P.
        • Gambin T.
        • Szafranski P.
        • Li J.
        • Amato S.
        • Divon M.Y.
        • et al.
        NAHR-mediated copy-number variants in a clinical population: mechanistic insights into both genomic disorders and Mendelizing traits.
        Genome Res. 2013; 23: 1395-1409
        • Coe B.P.
        • Witherspoon K.
        • Rosenfeld J.A.
        • van Bon B.W.
        • Vulto-van Silfhout A.T.
        • Bosco P.
        • et al.
        Refining analyses of copy number variation identifies specific genes associated with developmental delay.
        Nat Genet. 2014; 46: 1063-1071
        • Fawns-Ritchie C.
        • Deary I.J.
        Reliability and validity of the UK Biobank cognitive tests.
        PLoS One. 2020; 15e0231627
      1. Rosseel Y (2012): lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software. 48:1 - 36.

        • Alfaro-Almagro F.
        • McCarthy P.
        • Afyouni S.
        • Andersson J.L.R.
        • Bastiani M.
        • Miller K.L.
        • et al.
        Confound modelling in UK Biobank brain imaging.
        Neuroimage. 2021; 224117002
      2. Sønderby IE, Ching CRK, Thomopoulos SI, van der Meer D, Sun D, Villalon-Reina JE, et al. (2021): Effects of copy number variations on brain structure and risk for psychiatric illness: Large-scale studies from the ENIGMA working groups on CNVs. Hum Brain Mapp.

        • Crawford K.
        • Bracher-Smith M.
        • Owen D.
        • Kendall K.M.
        • Rees E.
        • Pardiñas A.F.
        • et al.
        Medical consequences of pathogenic CNVs in adults: analysis of the UK Biobank.
        J Med Genet. 2019; 56: 131-138
        • Calle Sánchez X.
        • Helenius D.
        • Bybjerg-Grauholm J.
        • Pedersen C.
        • Hougaard D.M.
        • Børglum A.D.
        • et al.
        Comparing Copy Number Variations in a Danish Case Cohort of Individuals With Psychiatric Disorders.
        JAMA Psychiatry. 2022; 79: 59-69
        • Fry A.
        • Littlejohns T.J.
        • Sudlow C.
        • Doherty N.
        • Adamska L.
        • Sprosen T.
        • et al.
        Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population.
        Am J Epidemiol. 2017; 186: 1026-1034