Opposing Brain Differences in 16p11.2 Deletion and Duplication Carriers

Cognitive and behavioral phenotypes

Cognitive and behavioral phenotypes of the Simons VIP group of 16p11.2 deletion and duplication carriers described elsewhere (Hanson et al., 2014; Zufferey et al., 2012, a portion of their data). Here we focus on the specific sample that contributed quality-controlled data for the present study. These data represent a subset of the original sample.

Social impairment was common in both the deletion and duplication carriers (Table 1). The SRS and Vineland-II are nonspecific screening measures of social impairment. SRS ≥60 reflects “at risk” individuals, and similarly Vineland-II scores ≤85 reflect a “moderately-low to low” adaptive level. Child deletion carriers (16/25 and 16/25) met screening cutoff for broad social impairment according to SRS and Vineland-II, respectively. Adult duplication carriers (3/17 and 8/15) also met these criteria. No child controls met these cutoffs, whereas one adult control had a Vineland-II score of 79. The SCQ is a screening test for ASD. A score ≥15 is the cutoff for “possible ASD.” Six of 25 child deletion carriers met this screening test for ASD, whereas no child controls did. SCQ was not tested in adults. The ADI and the ADOS are specific tests for ASD. Child deletion carriers (11/19 and 7/25) tested positive on ADI and ADOS, respectively. Of these child deletion carriers, a total of 14/25 were either positive for ADI or ADOS. In adult duplication carriers, 1/2 and 2/17 tested positive for ASD on ADI and ADOS, respectively.

The behavioral testing in Table 1 shows a progression from broad features of social impairment to those specific to ASD. SRS and Vineland-II are the broadest of the set, then SCQ screens for possible ASD, and finally ADI and ADOS are specific to the deficits characteristic of ASD. Furthermore, abnormal tests were most frequent on the broad level of SRS and/or Vineland-II, whereas only a few individuals tested positive for both ADI and ADOS. Consistent with Hanson et al. (2014), our subsample of deletion and duplication carriers shows variability in social impairment.

The study aimed to recruit controls who matched cases in nonverbal IQ (NVIQ). Despite best efforts, there was a 16-point difference between the child groups (t = 5.06, p < 0.001; Table 1). No significant difference was seen in adults. Full scale IQ was not controlled for and showed a similar pattern. A modest positive correlation with IQ and brain volume, and in particular cortical gray matter volume, has been observed in healthy children (Reiss et al., 1996), as well as healthy adults (Andreasen et al., 1993). Additionally, NVIQ differences between autism cases and controls can bias results (Piven et al., 1992). NVIQ was regressed from the results to address this potential confound.

A significant difference in age was found between child deletion carriers and age-matched controls (t = 2.96, p < 0.01), whereas no significant difference was found in adults. The study also aimed to match for handedness; however, an excess of left-handed individuals were recruited in the deletion group (∼1:2 vs ∼1:6 in child controls). Within the deletion group, left- and right-handed individuals did not significantly differ in volume in respect to ipsilateral and contralateral morphometry of cortical gray matter volume, cortical white matter volume, surface area, cortical thickness, or thalamus volume. Right-hemisphere cortical volume was increased in deletion carriers compared with the left-hemisphere cortical volume, but this was also the case for child controls. In contrast, right thalamus volume was smaller than the left in deletion carriers. Handedness did not affect the overall results. There were no significant differences in sex or scanner site between groups. Results were covariance-corrected for the nuisance factors of age, sex, scanner site, and NVIQ.

There is a potential for height to covary with head size. After controlling for age and sex, height was not significantly different between groups (children: t = 1.01, p = 0.32; adults: t = −1.17, p = 0.25). On the other hand, child deletions had significantly higher BMI than controls (t = −2.68, p < 0.01), whereas no significant difference was seen in adults. Higher BMI can be expected based on prior studies (Jacquemont et al., 2011; Zufferey et al., 2012), but the mirrored phenotype BMI phenotype was not seen in adults as was previously reported in the U.S. cohort (Zufferey et al., 2012; t = −0.07, p = 0.94). Height and BMI were not correlated in the sample as a whole (r = −0.01), and mildly correlated within groups (r = 0.17–0.38). Finally, if ICV and whole brain volume are normalized for height, results are not significantly affected.

Clinical diagnoses

All participants were evaluated for DSM-IV diagnoses (Table 2). A clinical diagnosis of ASD was based not only on ADI and ADOS scores, but clinical criteria and overall clinical evaluation. Four of 25 (16.0%) child-deletion carriers met clinical criteria for ASD, and 1/17 (5.9%) in adult duplication carriers. The prevalence in the child cohort was similar to the larger cohort evaluated at the phenotyping core sites from which the participants were recruited (Zufferey et al., 2012). Additionally, only those participants from the phenotyping core sites who could best tolerate imaging were included, which is a selection bias toward higher functioning individuals who can lie still in the scanner. The lower prevalence of ASD in adults compared with children may be due to duplications being less impactful than deletions, or to a selection difference (adults were often identified as carriers after clinical identification of a proband and then subsequent cascade genetic testing in the family such that carrier status in the proband would trigger testing in first degree relatives, then second, etc.). The deletion and duplication carriers were given a broad range of DSM-IV diagnoses (Table 2). Of note, none of the 16p11.2 carriers was given a diagnosis of schizophrenia. Schizophrenia has been associated with decreased brain volume (Crespi and Badcock, 2008), which could have potentially confounded adult duplication data if such diagnoses were present.

Radiographic abnormalities were also noted for deletion and duplication carriers (Table 3). Controls did not show radiologic abnormalities that would exclude them from analysis. Minor findings included occasional white matter hypointensities, small mega cisterna magnas, small arachnoid cysts, ectopic tonsils, empty sella syndromes, and mild cortical atrophy in some older adults.

Structural estimates were reliable

Participants were imaged across two sites (UC and CHOP) that were matched for scanner hardware and sequences. Effort was made to align details of imaging including placement of the head within the scanner. Nonetheless, reliability across sites was an open question. For this reason, five individuals not participating in the study were scanned at both sites with data analyzed independently for each acquisition using the same procedures as for the target cohort (mean age = 34.6; 4 male). Mean group differences between sites were marginal for these subjects in terms of ICV (UC_mean = 1687 cm³, CHOP_mean = 1682 cm³, t = −0.04, p = 0.97), cortical gray matter volume (UC_mean = 540 cm³, CHOP_mean = 549 cm³, t = 0.49, p = 0.64), surface area (UC_mean = 189 cm², CHOP_mean = 191 cm², t = −0.22, p = 0.83), and average cortical thickness (UC_mean = 2.53 mm, CHOP_mean = 2.56 mm, t = −1.52, p = 0.29). Although correlation is not properly estimated with such a small sample size, the intraclass correlation (ICC) was calculated: intracranial volume ICC = 0.97, cortical gray matter volume ICC = 0.93, surface area ICC = 0.99, and cortical thickness ICC = 0.85.

Deletion and duplication carriers display opposing effects on global measures of brain volume

Prior studies have shown that deletion carriers have increased head circumference, whereas duplication carriers have decreased head circumference (McCarthy et al., 2009; Shinawi et al., 2010; Jacquemont et al., 2011; Zufferey et al., 2012). Clinicians use head circumference as a rough measure for intracranial volume. Estimated ICV in this study corroborated prior results. ICV was increased for deletion carriers compared with their controls (t = −3.05, p < 0.01), and decreased for duplication carriers compared with their controls (t = 3.78, p < 0.001; Fig. 1A). If the results are also normalized for height the t test value increases slightly in children (t = −3.19, p < 0.01), and adults (t = 4.11, p < 0.001). Whole-brain volume was also revealed to be larger in child deletion carriers compared with controls (t = −3.55, p < 0.001), and smaller in adult duplication carriers compared with controls (t = 5.67, p < 0.001; Fig. 1B). If the results are also normalized for height the t test value increases slightly in children (t = −3.68, p < 0.001), and remain stable for adults (t = 5.53, p < 0.001). Because height has a minimal influence on global measures, subsequent results are only covariance-corrected for age, sex, scanner site, and NVIQ.

The CNV effects on brain volume did not appear to be driven by a few individuals who were diagnosed with clinical micro/macrocephaly. Instead it appears that the deletion and duplication broadly shift the distributions of brain volumes consistent with a highly penetrant effect on brain volume observable on the individual participant level before (Fig. 2A) and after correction for nuisance variables (Fig. 2B). This is an important observation as diagnosis of clinical macrocephaly requires an extreme deviation from the group central tendency (head size >2 SDs above the mean) which may not fully capture the effects on all individuals as a brain volume increase or decrease in an individual may still leave him or her within the “normal” range of the brain volume distribution simply because there is considerable normal variation (Moreno-De-Luca et al., 2013).

Relevant to this point, although diagnosis of clinical macrocephaly has been associated with idiopathic autism in only ∼15–20% of cases (Lainhart et al., 1997; Stigler et al., 2011), increased mean brain volume is a robust phenomenon in autism in particular at early ages (Redcay and Courchesne, 2005). For instance, Courchesne et al. (2001) found that 90% of 2- to 4-year-old autistic boys had whole-brain volumes above the mean for controls, and only one normal boy had a brain volume that exceeded the autistic mean. One meta-analysis estimated a 5–10% increase in total brain volume for children ages 1.5–4 years old (Amaral et al., 2008). Our cohort of 16p11.2 deletion carriers was similar. Twenty-four percent (6/25) of child-deletion carriers had clinical macrocephaly. More impressive is the fact that almost all child-deletion carriers were above the mean for their controls (mean = 1277 cm³), and had an effect size of a mean whole brain volume 9.7% larger than controls. Duplication carriers possessed a mirror opposite pattern; 23.5% (4/17) had clinical microcephaly and a decrease in mean whole-brain volume by 9.3% below that of controls (mean = 1204 cm³). Thus, it is likely that carrier status, age, gender, and genetic background all play significant roles in the observed head size in the 16p11.2 cohort, but that carrier status is the common element and hence most observable.

The effect on brain volume was also observed when cortical gray matter volume (Fig. 3A) and white matter volume (Fig. 3B) were analyzed separately. Child-deletion carriers showed increased volume compared with controls (cortical gray matter: t = −3.06, p < 0.01; cortical white matter: t = −3.54, p < 0.001), whereas adult duplications showed decreased volume compared with controls (cortical gray matter: t = 4.03, p < 0.001; cortical white matter volume: t = 5.07, p < 0.001). The percentage change in 16p11.2 carriers compared with controls was ∼11.9% in white matter volume and ∼8.5% in gray matter (Table 4). A few studies with young children (1.5–4 years) with idiopathic autism found greater increases in white mater than gray matter (Courchesne et al., 2001; Herbert et al., 2003; Hazlett et al., 2005; for review, see Amaral et al., 2008). In the present study, the age-associated increase in white matter volume in controls is difficult to interpret because the two cohorts were normalized within their age groups, nonetheless it may be a real effect consistent with observed increases in estimated volume that occur in adolescence (Matsuzawa et al., 2001; Barnea-Goraly et al., 2005; Perrin et al., 2008).

Opposing effects extend broadly to multiple subcortical volumes and the cerebellum

The effect of 16p11.2 on brain volume is not limited to cerebral cortex. Cerebellum volume is significantly increased in child deletion carriers (t = −2.33, p < 0.05) and significantly decreased in duplication carriers (t = 3.64, p < 0.001; Fig. 4A). Subcortical measures also show this same pattern, including the thalamus (children: t = −4.12, p < 0.001; adults: t = 4.73, p < 0.001; Fig. 4B) and the hippocampus in adults (adults: t = 2.90, p < 0.01; Fig. 4C). A trend toward significance was observed for the amygdala in adults (t = 1.95, p = 0.06) and hippocampus in children (t = −1.90, p = 0.06). Amygdala volume in children was not significant (t = −1.25, p = 0.22). In idiopathic autism total cerebellum volume has been found to be increased (Hardan et al., 2001; Herbert et al., 2003; Palmen et al., 2005). The thalamus (Hardan et al., 2008), hippocampus (Aylward et al., 1999; Sparks et al., 2002; Schumann et al., 2004), and amygdala (Aylward et al., 1999; Pierce et al., 2001; Schumann et al., 2004, 2009) have revealed divergent findings (for review, see Stigler et al., 2011).

To better understand the relationships across the various measures of brain structure, Table 4 displays the differences in structural brain volumes estimated as a percentage difference from the age-matched control group. In this manner, for each structure, the effect of CNV can be appreciated as a quantitative value. The effects are pervasive and roughly of similar magnitude across most structures. Of all individual structures, the thalamus shows the greatest difference, although it is difficult to interpret small relative differences between structures given differences in baseline structure size as well as idiosyncratic features that arise in templates used as the target for estimation.

As another way to display relationships across measures, Table 4 also displays the percentage differences after correction for ICV. This removes the global effect of head size and estimates the differences that are present that are beyond those accounted for by global head size differences. Head size normalization attenuated most of the differences suggesting that a single global effect accounts for much of the results. An exception is the thalamus, which remains significantly larger in deletion carriers (t = −2.37, p < 0.05) and smaller in duplication carriers (t = 2.46, p < 0.05) after accounting for the global effect of head size. Duplication carriers also were significantly smaller than controls in brain volume (t = 3.84, p < 0.001), cortical white matter (t = 2.55, p < 0.05), and in cortical thickness (t = 2.18, p < 0.05), but a significant difference was not found in deletion carriers. Note that the results reported throughout this paper do not correct for ICV.

Allometric scaling plots suggest brain differences are due to genetic effects not age

The above results all converge to suggest that brain volumes are larger for deletion carriers and smaller for duplication carriers. However, the two groups, in our study, confound age. It could be the case that during childhood both deletion and duplication carriers have increased brain volumes compared with controls, and as adults both deletion and duplication carriers have decreased brain volumes compared with controls. This would be consistent with the age-specific anatomical changes that have been reported for idiopathic autism, in which brain overgrowth occurs in early life, and later life is characterized by arrested growth or decline below normal levels (Courchesne et al., 2011). If this effect is present for our dataset then a single nonuniform process might be present in 16p11.2 carriers, instead of two opposing processes. To control for age, Figure 5 plots gray matter volume in relation to ICV. Within an age group, the two values show allometric scaling such that values tend to fall along a line (Davis and Wright, 1977). This plot demonstrates that well-behaved allometric scaling is observed here within the age cohorts. The effect of age can be seen as a downward shift between the child controls (light green circles) and adult controls (dark green circles). Critically, 16p11.2 deletion and duplication carriers shift along their own age-appropriate lines. Deletion carriers (red triangles) cause a rightward shift, and duplication carriers (blue squares) cause a leftward shift. In other words, the effect of age is dissociable from the effect of the CNV.

Analysis of clinically acquired MRI suggest brain differences are due to genetic effects not age

In an effort to further explore the inferences drawn from allometric scaling plots, clinically acquired MRI images were obtained from the phenotyping core sites. This superset provided a sufficient number of child deletion carriers, unaffected siblings as controls, and the group missing from our primary analysis, child duplication carriers. Although a larger sample was available in the superset only a portion of these MRI images met quality control standards (Table 5).

Whole-brain volumes of child duplication carriers were found to be significantly smaller than controls (t = 3.25, p < 0.01), and significantly smaller than child deletion carriers (t = 6.36, p < 0.001). Additionally, brain volumes of child deletion carriers were significantly larger than controls (t = −2.71, p < 0.01; Fig. 6). This cohort allowed for direct comparison of child deletion with child duplication carriers, which was not available in the primary dataset. The results further support that reciprocal 16p11.2 CNVs lead to opposing effect on brain volume, instead of a singular process of brain overgrowth. Courchesne et al. (2011) report of “age-specific anatomical changes in autism” noted larger brain volumes in 2- to 4-year-olds, and smaller brain volumes as early as 8- to 16-year-olds. The age sampling in our cohorts are not sufficiently powered to explore effects between these age ranges.

Cortical volumetric differences are greater in surface area than thickness

Both surface area and cortical thickness contribute to cortical gray matter volume, thereby affecting whole brain volume. Expanded cerebral surface area across species is hypothesized to be due, in part, to an increase in the size of the progenitor cell pool through an extended growth phase that disproportionately affects cortical surface area over thickness, i.e., from mouse to human there is a 1000-fold increase in surface area and only a approximately twofold increase in cortical thickness (Rakic, 1988). For this reason, although there are likely multiple factors that affect cortical surface area and thickness across individuals (Hill et al., 2010 for analysis of postnatal effects), differential effects of genetic dosage on thickness versus surface area are particular informative. Prior studies have found these two measures to be genetically and phenotypically distinct (Panizzon et al., 2009; Winkler et al., 2010).

To explore differential effects on the cerebral cortex, we estimated differences in cortical thickness separately from cortical surface area. No significant difference in cortical thickness was observed in children (t = −1.10, p = 0.28). A significant, but modest, difference was observed in adults (t = 2.28, p < 0.05). A significant difference was observed in surface area for both children (t = −2.40, p < 0.05) and adults (t = 4.22, p < 0.001). Quantitatively, the effect size was considerably greater for surface area. Put into context, in children, 89.9% of the gray matter volume differential was accounted for by surface area, whereas only 10.1% was accounted for by differences in thickness. Similarly, in adults, 79.4% of the gray matter volume differential was accounted for by surface area, whereas 20.6% was accounted for by thickness. These differences could be appreciated at the individual subject level in the scatter plots (Fig. 7C,D). Increased cortical thickness has been observed in autism (Hardan et al., 2006), but at least in 16p11.2 the reciprocal differences in cerebral gray matter volume are preferentially accounted for by differences in cortical surface area.

16p11.2 influences brain volume in a dose-dependent manner

Broadly speaking, a general symmetry was observed in the deviation from the mean value of controls. The percentage difference was approximately equal on the most robust structural brain measures between deletion and duplication carriers (Table 4): ICV (∼7%), brain volume (∼9%), surface area (∼7%), cortical thickness (∼1–2%), etc. This suggests that 16p11.2 influences brain structures in a dose-dependent manner: one copy of the CNV (deletion) is associated with abnormally large brain volume (+9%), whereas two copies (controls) lead to typical development (reference), and three copies (duplication) are associated with abnormally small brain volume (−9%). Dosage-dependence implies a linear relationship between copy-number and gene expression. This is consistent with the dosage-dependent gene expression observed in a mouse model of 16p11.2. Horev et al. (2011) found mRNA expression to be significantly altered in all brain regions evaluated such that gene expression was directly correlated with dosage. Furthermore, gene expression was affected much more by dosage compared with the change in gene expression by brain region effect (comparing the cortex, cerebellum, brainstem, and optic nerves; Horev et al., 2011). Their analysis may explain the general symmetry in percentage change from controls observed in the present study for a given structure, while at the same time allowing for some variation across brain structures.

The effects are pervasive across brain structures

Explorations of the 16p11.2 CNV and the included genes in animal models have reported involvement of widespread brain structures. Cerebellum, basal ganglia, brainstem, and optic nerve involvement has been reported in mouse models of 16p11.2 (Horev et al., 2011; Portmann et al., 2014). A study using zebrafish embryos systematically altered gene expression for each of the 29 genes encompassed by 16p11.2, and found KCTD13 expression levels to be a primary driver of the head size phenotype (Golzio et al., 2012). Systematic overexpression of the mRNA transcript exclusively of KCTD13 caused microcephaly, whereas splice-blocking the KCTD13 message with morpholinos produced macrocephalic embryos (with no differences in somatic trunk size). Early in development KCTD13 transcript was found to be strongly expressed at the anterior forebrain, midbrain, and hindbrain 24 h postfertilization (Golzio et al., 2012).

A major aim of the current study was to provide greater detail as to which brain structures are affected in humans with the 16p11.2 CNV. Brain volume in the present study is a composite measurement of all brain structures: cortical gray matter, cortical white matter, thalamus, basal ganglia, limbic structures, cerebellum, and brainstem, without the voxels in the ventricles (https://surfer.nmr.mgh.harvard.edu/fswiki/MorphometryStats). A significant difference in brain volume between carriers and controls is analogous to the results found in the animal work. The detailed morphometric analyses allowed us to unpack this observation further.

Both cortical gray matter and white matter volume were significantly larger/smaller than controls. Subcortical structures, such as the thalamus and hippocampus, were also significantly different from controls. The amygdala and striatum (caudate and putamen) did not reach significance, which may be related to inadequate power in the study as they trended in the same direction as the other structures with approximately the same magnitude of volume differences. Cerebellum and brainstem volumes also reached significant levels. These results highlight the pervasive effect of 16p11.2 across brain structures. This in turn suggests involvement of 16p11.2 early in development, as the prosencephalon (cortex and thalamus), mesencephalon, and rhomboencephalon (brainstem and cerebellum) are all influenced. The thalamus was the only structure that retained significance in both deletion and duplication carriers after regressing ICV as a covariate (Table 4). This suggests that thalamic volume increased in deletion carriers and decreased in duplication carriers to a greater degree than the overall head size differences.

Collectively, the effects across brain structures are pervasive.

The brain volume phenotype may share the same genetic underpinnings as ASD in 16p11.2

Case reports provide some evidence that KCTD13 may also be a primary driver of the ASD phenotype (Crepel et al., 2011; Golzio et al., 2012). If brain volume and ASD phenotypes share a primary driver on 16p11.2, then this CNV may be best modeled as a “simplex cis-epistatic” CNV, where dysfunction of a primary driver is necessary and sufficient for multiple phenotypes but it is subject to epistatic modulation by other genes on the CNV. Additionally, epistatic modifiers heavily influence penetrance and expressivity (Golzio and Katsanis, 2013). MVP and MAPK3 modify KCTD13's effect on head size (Golzio et al., 2012), whereas a different set of epistatic modifier genes could contribute to the ASD phenotype. This model provides one explanation for the disparity in our dataset between the high penetrance and expressivity seen in brain volume versus the reduced penetrance and variable expressivity seen with the ASD phenotype. The disparity may be representative of the complex nature of 16p11.2 CNV with its phenotypes, and not an apparent sample bias toward fewer carriers with ASD. Thus, changes in brain volume itself may provide insight into the core pathobiology of ASD in 16p11.2 carriers despite differences in penetrance and expressivity.

Preferential influence on surface area suggests a candidate mechanism for abnormal brain size

An informative finding in this study was the preferential influence of 16p11.2 on surface area in contrast to a minimal effect on cortical thickness. This preferential effect on surface area provides insight into when in development the 16p11.2 CNV derails typical development. Alterations in cortical thickness can occur either during embryonic formation or postnatally as a result of synaptic exuberance and pruning (Stiles and Jernigan, 2010). The expansion of cortical surface area is strongly affected in the early embryo via increased progenitor cell production (Rakic, 1988) consistent with the effects of manipulating KCTD13 in the zebrafish (Golzio et al., 2012). In particular, augmentation of the large pool of self-renewing progenitor cells in the outer subventricular zone could account for differences in surface area in 16p11.2 deletion and duplication carriers (Hansen et al., 2010; Lui et al., 2011). Moreover, the involvement of MAPK3 is also consistent with findings in patients, as alterations of the RAS-MAPK pathway can lead to macrocephaly in Costello syndrome (Gripp et al., 2010, 2011). The differential effect on cortical surface area is not a definitive marker of an early embryonic mechanism as postnatal growth of human surface area has also been observed in particular for association regions (Hill et al., 2010). Nonetheless, the convergence between the mechanistic work in the zebrafish suggesting an early embryonic mechanism affecting cell proliferation and the pervasive, highly penetrant effect on the human brain that is preferential to cortical surface area over thickness is intriguing.

Limitations

A limitation of the present study is that the sample in unlikely representative of the full population of 16p11.2 deletion and duplication carriers. Only probands with deficits severe enough to warrant clinical evaluation and carrier parents well enough to seek recruitment were ascertained, which selected for more severely impaired children and less affected parents. Furthermore, only the subset of carriers that were capable of lying still in an MRI scanner proceeded to the neuroimaging core sites, which may have precluded the most severely affected carriers. Inadequate recruitment of child duplication and adult-deletion carriers prevented direct analysis of all six groups for all brain measures; however, the core result (reciprocal increase and decrease in brain volume) was observed in a subset of age-matched child deletion and duplication carriers (Fig. 6). The study was not powered for subgroup analysis between de novo and inherited cases. Hanson et al. (2014) have noted significantly greater impairment on SRS and NVIQ in de novo versus inherited cases. Moreover, many of our analyses of deletion and duplication carriers used matched controls in the general population (Fig. 6 is an exception, as it used a family based control group). Thus, other genetic effects transmitted by noncarrier parents were not controlled for in the analyses of brain region volumes. Despite this heterogeneity, we were still able to see robust changes in both groups of carriers, independent of the clinical severity of the groups.

As with all morphometric measures of brain structure, subtle differences in data quality and image contrast (which covaries with age) can affect measurements. These technical concerns should be kept in mind especially when interpreting regional (spatially specific) effects and differential effects on tissue compartments (white matter versus gray matter). We screened the data carefully for data quality and generally found pervasive effects that could be observed at the level of individuals across the age spectrum. That said, the preferential effect on the thalamus (for both deletion and duplication carriers) and the differential effect on surface area over thickness should be explored with independent methods (e.g., manual assessment and postmortem analysis). Of the various measures, although still showing high reliability, cortical thickness showed the least stable reliability (ICC = 0.85).

Conclusions

Deletion and duplication carriers of 16p11.2 show opposing and pervasive effects on brain structure. Differences in cortical gray matter volume were due to the preferential influence of 16p11.2 copy number variation on surface area. This finding suggests that the mechanism underlying differences in brain volume in 16p11.2 may involve early stages of development.