Participants

Participants with nested 22q11.2 duplications or deletions

Participants included 43 individuals with a nested duplication (n = 13) or deletion (n = 30) of 22q11.2 that lay entirely within LCR-A to LCR-D but was not completely inclusive of LCR-A to LCR-D (see Table 1). The only exception to this was one participant who carried a duplication of LCR-B to LCR-D and also a very small duplication between LCR-E and LCR-F. Participants were recruited from a specialty clinic at The Children’s Hospital of Philadelphia (CHOP) or were referred from a similar specialty clinic at another institution. The CHOP “22q and You” clinic represents the largest single-site 22q11.2 clinic in the world and maintains a large catchment area across the eastern US, with patients concentrated within a few hundred mile radius of CHOP. The sample includes probands who came to clinical attention, as well as their affected siblings (n = 2 with duplication and n = 3 with deletion) and parents (n = 2 with duplication and n = 2 with deletion) whose 22q11.2DS or 22q11.2DupS was identified after the proband’s diagnostic process. The duplication or deletion was confirmed using single nucleotide polymorphism (SNP) microarray or Multiplex Ligation Probe Amplification (MLPA).

Region	Number	% de novoa	Age mean (sd)	Age range (years)	% male
Total	46	60%	10.8(10.1)	0.8–39	52%
AB/AC deletion group	15	86%	8.9(4.2)	2–15	53%
A-B deletion	11	80%	7.8(3.8)	2–15	55%
A-C deletion	4	100%	11.8(4.4)	5–14	50%
BD/CD deletion group	18	69%	11.1(10.7)	1–38	50%
B-D deletion	14	77%	10.4(9.0)	1–38	43%
C-D deletion	4	33%	13.6(16.9)	0.8–36	75%
BD/CD duplication group	10	13%	14.5(15.2)	1–39	60%
B-D duplication	8	0%	16.5(16.6)	1–39	63%
C-D duplication	2	50%	6.5(2.0)	5–7	50%
A-B duplicationb	1	Unknown	7.0	–	0%
E-F duplicationb	2	0%	6.6(2.8)	4–8	50%

Participant characteristics for all individuals with a nested deletion or duplication of 22q11.2, including three case studies with atypical nested duplications noted with superscript b

^aPercentage of individuals with known inheritance

^bCase studies not included in statistical analysis, medical chart review, or AB/AC and BD/CD group totals

Samples whose copy number variations (CNVs) were tested by MLPA were examined using the SALSA P250 DiGeorge diagnostic probe kit (MRC-Holland, Amsterdam, The Netherlands). Commercially available software, Gene Marker from SoftGenetics (State College, PA), was used to analyze the data. Gene Marker has developed a completely integrated application for MLPA analysis with integrated functions specific for the analysis of data derived from MLPA reactions. Samples whose CNVs were identified by SNP array were analyzed using the Affymetrix SNP Array 6.0 platform following the manufacturer’s instructions (Affymetrix, Santa Clara, CA, USA). Quality control values were calculated in Affymetrix Genotyping Console (Affymetrix), and any samples with contrast QC greater than 0.4 or mean absolute pairwise difference (MAPD) greater than 0.35 were excluded from further analysis. The B allele frequency and log R ratio plots were visualized using the Affymetrix Chromosome Analysis Suite to support CNV calls.

Three additional patients who carried very small and rare atypical duplications are included in this paper in a descriptive manner (in the case study sections), but are not combined with the other groups in tables, figures, or statistical analyses. One patient carried a very small duplication within LCR-A to LCR-B. The other two patients (who were related to three patients in the main LCR-B to D duplication group) carried a small duplication nested between LCR-E and LCR-F.

All 43 participants were included in the medical history chart review. Nine participants were excluded from the ASD and psychiatric symptom analyses (n = 34; see Fig. 2) for two types of reasons: (1) ASD classification could not be determined (n = 2; see below) or (2) if they presented with another medical issue likely to affect brain development (n = 2 extreme prematurity and/or birth weight < 5th centile, n = 2 with CEDNIK syndrome, n = 1 with 16p11.2 deletion which is independently associated with ASD, n = 2 history of hypoxic brain injury) [31–34]. Participant characteristics of the sample excluding these nine cases are described in Table 2. Please note that some ages differ from those in the medical record review (Table 1) because a review of updated records pertinent to ASD classification, when available, was conducted 3 years later to allow for infants to reach the age (3 years) at which ASD symptoms would be present.

Region	Number	% de novoa	Age mean (sd)	Age range (years)	% male
AB/AC deletion group	13	83%	10 (4.2)	5–18	54%
A-B deletion	10	78%	8.9 (3.6)	5–15	60%
A-C deletion	3	100%	13.7 (4.8)	9–18	33%
BD/CD deletion group	12	50%	14.2 (12.7)	3–42	50%
B-D deletion	8	57%	13.4 (12.1)	4–42	38%
C-D deletion	4	33%	15.9 (15.5)	3–37	75%
BD/CD duplication group	9	14%	16.9 (14.7)	5–39	56%
B-D duplication	7	0%	19.4 (15.9)	5–39	57%
C-D duplication	2	50%	8 (4.1)	5–11	50%
Total	34	55%	13.3 (11.0)	3–42	53%

Participant characteristics for the subset of individuals with a nested deletion or duplication of 22q11.2 included in description of psychiatric diagnosis rates

^aPercentage of individuals with known inheritance

Rates of autism were analyzed separately for individuals with nested deletions and duplications (see Table 3). Only one individual per family (the proband) was included to avoid confounding autism rates with risk factors shared by related individuals. In one family with B-D duplication, we included an affected family member instead of the proband because the proband harbored a 16p11.2 deletion. For deletions, 20 individuals were included after excluding five parents and younger siblings (2 B-D, 2 C-D, 1 A-B). For duplications, five individuals were included after excluding four parents and younger siblings (4 B-D). No individuals presented with nested duplications involving LCR-A to LCR-B or C.

Region	Number	n ASD (male)	% de novoa	Age mean (sd)	Age range (years)	% male
AB/AC deletion group	12	5(3)	90%	10.0 (4.4)	5–18	58%
A-B deletion	9	4(3)	90%	8.7 (3.7)	5–15	67%
A-C deletion	3	1(0)	100%	13.7 (4.8)	9–18	33%
BD/CD deletion group	8	0	60%	10.5 (4.8)	5–18	25%
B-D deletion	6	0	70%	10.2 (3.8)	6–17	17%
C-D deletion	2	0	0%	11.6 (9.1)	5–18	50%
Classic AD duplication	29	7(5)	67%	7.1 (3.4)	2–13	75%
BD/CD duplication group	5	1(0)	30%	12.7 (10.4)	5–31	60%
B-D duplication	3	1(0)	0%	15.8 (13.1)	7–31	67%
C-D duplication	2	0	50%	8.0 (4.1)	5–11	50%

Participant characteristics and autism diagnosis for all probands with a nested deletion of 22q11.2. Individuals harboring an AB or AC deletion presented with ASD at a rate of 41.6% (5 of 12). Case studies are excluded from this table

^aPercentage of individuals with known inheritance

Procedures

We collected data from record review, questionnaires administered remotely, and, for a subset, an autism-specific evaluation. Record review included the participant’s electronic health record at CHOP whenever possible, as well as external medical and educational records (e.g., IEP evaluations) provided by families for individuals who did not receive routine medical care at our institution.

Higher rates of ASD when LCR-A to B involved

We observed a trend toward a higher rate of ASD among probands with deletions in the AB/AC group (41.7%, or 5 in 12 individuals with LCR-A to B, or LCR-A to C) compared to the BD/CD group (0%, or 0 in 8 individuals with LCR-B to D, or LCR-C to D; χ = 4.4, p = 0.051, CI 0.99, Inf; see Table 3). In a more conservative analysis that matched groups on approximate size of deleted region, we continued to observe similar rates of ASD within each group (44.4%, or 4 of 9 individuals with deletions of LCR-A to B, and 0%, or 0 in 6 individuals with deletions of LCR-B to D; χ = 3.64, p = 0.092, CI 0.702, Inf). The rate of ASD did not change meaningfully when related individuals were included to increase sample size; the increased sample size provided more statistical power and revealed significant results (n = 25; 38.5% rate in AB/AC group, 0% in BD/CD group; χ = 5.77, p = 0.024, CI 1.39, Inf). Thus, the LCR-A to B region may confer increased risk of ASD diagnosis, but a larger sample without related individuals is needed to confim.

Among duplications, individuals with the classic and BD/CD duplications showed similar rates of ASD (24.1% rate or 7 of 29 in classic group, 20% rate or 1 of 5 in BD/CD; OR = 0.79, p = 0.764, CI 0.03, Inf). Results did not change meaningfully when related individuals were included to increase sample size (21.4% rate in classic group, 11.1% rate or 1 of 9 in BD/CD; OR = 0.40, p = 0.65, CI 0.02, Inf), but this analysis in particular would benefit from a larger sample.

Our categorical analysis was supported by quantitative reports of autistic symptoms in the SRS-2 and SCQ (see Fig. 3). A subset of each group (BD/CD deletions, AB/AC deletions, BD/CD duplications, classic duplications, classic deletions) completed the SCQ, including both individuals with and without ASD diagnoses. For deletions, the BD/CD group showed less autistic symptoms than the AB/CD group with large effect sizes (d’s of 1.01 and 1.20). For duplications, the difference was small-to-medium (d’s of 0.27 and 0.50) between the BD/CD group and the classic group. No effects reached statistical significance (see Table 4)

.

Moderately lower adaptive and social functioning when AB region involved

We computed effect sizes for differences in autistic symptoms, psychiatric symptoms, and adaptive behavior skills (see Figs. 3 and 4, Table 4, Additional file 2). For duplications, the differences were usually small between the BD/CD group and the classic duplication group (see Table 4, “Classic Duplication” rows). For deletions, the BD/CD group showed less impairment than the AB/AC group across most measures with medium or large effect sizes that did not reach statistical significance. We also calculated effect sizes for group differences between the AB/AC deletions and classic AD deletion groups and observed small or medium differences (see Table 4, "Classic Deletion" rows). We observed negligible differences between these two groups on most adaptive functioning scales. The classic deletion group showed slightly lower levels of autistic symptoms compared to the AB/AC group––small to medium effect sizes on the SRS-2 and SCQ––that were not statistically significant.

	Number	Mean (SD)	d	95% confidence interval
SRS-2 T score		50 (10)
BD/CD deletion	9	53.4 (13.6)
AB/AC deletion	10	68.5 (15.7)	1.01	(− 0.07, 2.11)
Classic deletion	61	63.0 (12.6)	− 0.41	(− 1.11, 0.27)
BD/CD duplication	6	59.5 (16.6)
Classic duplication	28	64.0 (16.3)	0.27	(− 0.67, 1.22)
SCQ raw total		Cutoff 15
BD/CD deletion	7	5.7 (4.9)
AB/AC deletion	8	15.0 (9.3)	1.20	(− 0.11, 2.52)
Classic deletion	52	10.8 (7.3)	− 0.54	(− 1.32, 0.23)
BD/CD duplication	4	7.2 (9.5)
Classic duplication	22	12.4 (10.3)	0.50	(− 0.67, 1.68)
Vineland composite		100(15)
BD/CD deletion	5	103.8 (19.6)
AB/AC deletion	10	85.7 (24.0)	− 0.79	(− 2.11, 0.52)
Classic deletion	57	87.1 (15.9)	0.08	(− 0.61, 0.78)
BD/CD duplication	6	92.6 (18.7)
Classic duplication	27	89.4 (19.4)	− 0.16	(− 1.11, 0.78)
Vineland communication		100(15)
BD/CD deletion	5	106.0 (21.7)
AB/AC deletion	10	83.0 (18.5)	− 1.17	(− 2.55, 0.20)
Classic deletion	58	89.7 (18.1)	0.36	(− 0.32, 1.06)
BD/CD duplication	6	92.1 (20.2)
Classic duplication	27	91.2 (18.5)	− 0.04	(− 0.99, 0.90)
Vineland daily living		100(15)
BD/CD deletion	5	96.4 (14.6)
AB/AC deletion	10	86.9 (26.2)	− 0.40	(− 1.69, 0.87)
Classic deletion	57	88.4 (15.1)	0.08	(− 0.60, 0.78)
BD/CD duplication	6	91.8 (23.0)
Classic duplication	28	93.6 (21.2)	0.08	(− 0.86, 1.02)
Vineland socialization		100(15)
BD/CD deletion	5	107.6 (19.6)
AB/AC deletion	10	91.0 (27.4)	− 0.65	(− 1.96, 0.65)
Classic deletion	57	90.1 (16.5)	− 0.04	(− 0.74, 0.64)
BD/CD duplication	6	98.1 (18.3)
Classic duplication	27	90.8 (21.2)	− 0.35	(− 1.30, 0.60)
CASI ADHD
BD/CD deletion	2	2.5 (1.3)
AB/AC deletion	7	3.4 (1.4)	0.63	(− 1.56, 2.83)
Classic deletion	43	2.7 (1.2)	− 0.52	(− 1.36, 0.32)
BD/CD duplication	4	2.7 (2.1)
Classic duplication	20	2.8 (1.3)	0.09	(− 1.09, 1.27)
CASI anxiety
BD/CD deletion	2	1.3 (0.7)
AB/AC deletion	7	1.9 (1.1)	0.49	(− 1.68, 2.67)
Classic deletion	43	2.3 (1.3)	0.30	(− 0.53, 1.14)
BD/CD duplication	4	1.9 (1.5)
Classic duplication	20	1.8 (1.6)	− 0.05	(− 1.24, 1.12)
CASI ASD
BD/CD deletion	1	n/a
AB/AC deletion	4	1.0 (1.0)	n/a	n/a
Classic deletion	37	0.5 (0.4)	− 0.83	(− 1.94, 0.27)
BD/CD duplication	4	0.6 (0.4)
Classic duplication	16	0.8 (0.8)	0.26	(− 0.98, 1.50)
CASI Schizoaffective
BD/CD deletion	1	n/a
AB/AC deletion	3	0.5(0.5)	n/a	n/a
Classic deletion	6	0.5(0.2)	− 0.09	(− 0.78, 0.61)
BD/CD duplication	0	n/a
Classic duplication	4	0.5(0.2)	n/a	n/a
CASI behav. regulation
BD/CD deletion	2	0.9 (0.4)
AB/AC deletion	7	0.9 (0.6)	0.06	(− 2.08, 2.21)
Classic deletion	43	1.0 (0.6)	0.16	(− 0.67, 1.00)
BD/CD duplication	4	0.3 (0.3)
Classic duplication	20	1.0 (0.8)	0.78	(− 0.42, 2.00)
CASI Depression
BD/CD deletion	2	0.4 (0)
AB/AC deletion	7	0.6 (0.7)	0.33	(− 1.82, 2.50)
Classic deletion	43	0.5 (0.7)	− 0.15	(− 0.99, 0.67)
BD/CD duplication	4	0.2 (0.4)
Classic duplication	20	0.4 (0.4)	0.48	(− 0.70, 1.68)

Group means on neuropsychiatric questionnaires. We show 95% confidence intervals of effect sizes as Cohen’s d, which can be interpreted as follows: 0.2 as small, 0.5 as medium, and 0.8 as large [54]. Means and standard deviations for each group are presented, as well as the mean and SD for each measure to aid in interpretation. We derived SRS T-scores using the updated SRS-2 norms for all participants, regardless of the version the participant completed. We averaged CASI-4R raw item scores on similar subscales into composites instead of using T-scores because we encountered a strong ceiling effect when using CASI-4R T-scores because CASI-4R norms collapse all high raw scores into a T-score of 70. Thus, population-normed means and standard deviations are not available for comparison for these averaged composites. For example, all items from the dysthymia subscale and major depression subscales were averaged into a “Depression” composite, after accounting for the number of items in each subscale so that both scales were weighted equally in the composite. The composites are interpreted as “3” indicating that on average, the parent endorsed symptoms in the domain as occurring “very often,” 2 as “often,” 1 as “sometimes,” and 0 as “never”

Vineland Vineland Adaptive Behavior Scales, 2nd Edition; CASI Child and Adolescent Symptom Inventory-4R; SCQ Social Communication Questionnaire; SRS-2 Social Responsiveness Scale, 2nd Edition; n/a not applicable

Increased rates of psychiatric disorders

Higher rates of medical comorbidities

Case study 1

Case study 2

Implications for medical screening

Prior studies have suggested that individuals with nested deletions have similar types of medical problems to those with classic deletions and should receive similar clinical treatment. The medical chart review of our patients supported this hypothesis. It also suggested that our patients are representative of other previously reported patients with nested deletions with regard to the frequency and types of medical problems. It is notable that there appeared to be fewer medical problems in individuals with LCR-C to D. However, this region is much smaller, encompassing fewer genes than the other regions. In size and total number of genes, LCR-A to LCR-B and LCR-B to LCR-D are roughly equivalent, and the rates of medical comorbidities are similar. We also observed higher rates of some medical comorbidities in several of the nested groups as compared to individuals with full LCR-A to LCR-D deletions (e.g., cervical spine anomalies in 100% of screened individuals with LCR-A to LCR-B deletion), but our sample sizes are too small to determine if this is due to chance or truly represents a higher risk subgroup. We were somewhat surprised to find that many patients had not completed portions of the recommended medical screening for individuals with 22q11.2-related disorders. It is unclear if this is due to a perception by providers that individuals with nested deletions do not need as aggressive screening as those with full deletions or duplications. Overall, we observed rates of each of the medical comorbidities in the LCR-A to LCR-B and LCR-B to LCR-D subgroups that are comparable to rates in individuals with full LCR-A to LCR-D deletions or duplications. Although the rate of medical problems appears lower in the LCR-C to LCR-D deletion and duplication groups, the sample sizes are extremely small, and therefore, no strong conclusions can be made about the validity of an altered screening protocol for these patients.

RANBP1 as a potential ASD candidate gene

The LCR-A to B region associated with ASD risk in our sample involves approximately 25 genes, including COMT, PRODH, and TBX1. Prior research implicates the interaction of low activity COMT and PRODH alleles in ASD risk [28, 29]. Other genes in the region may also confer ASD risk, and indeed, the risk could be additive. We propose another possible candidate gene, Ran-binding protein 1 (RANBP1), which could not be examined given our study design with multiple candidate genes in the LCR-A to B region, warranting further investigations. We base this speculation on five circumstantial pieces of evidence.

First, we cite the involvement of RANBP1 in the metabotropic glutamate receptor (mGluR) gene network [44], which is disrupted in two other syndromic forms of ASD, fragile X syndrome and tuberous sclerosis complex [45]. Second, we previously observed a 10-fold increase in ASD rate among individuals with 22q11.2DS with a “second hit” in an mGluR network gene compared to individuals without a “second hit” (5 affected in 25 individuals with 22q11.2 compared to 1 in 50) [30]. Third, two teratogens associated with increased rates of ASD––valproate and thalidomide––both decrease expression of RANBP1 [46, 47, 48]. Fourth, the important link between RANBP1 and expression in human brains was demonstrated by Meechan et al. [49], who showed higher RANBP1 expression in developing fetal brains compared to adult brains during a peak in neurogenesis. Finally, several studies in the 22q11.2 animal literature highlight RANBP1 as important for neural development in 22q11.2 (e.g., [49–51]). Taken together, these disparate pieces of literature converge on a role of RANBP1 in brain development, and potentially in ASD. Like other genes and gene families recently associated with ASD, RANBP1 serves a general function within the cell (metabolizing GTP and regulating material transport to the nucleus [52]). RANBP1 has not been identified previously as an ASD candidate gene in large ASD studies; of the approximately 25 genes in the 22q11.2 LCR-A to LCR-B region, previous genome-wide association studies or whole exome sequencing studies have identified PRODH as a candidate gene with suggestive evidence and TBX1 and GNB1L as candidate genes with minimal evidence at this time (SFARI gene database https://gene.sfari.org/database/human-gene/). It is not yet clear whether genes in this region modify ASD risk in the general population or in the context of 22q11.2 syndromes alone.

Insights from two case studies involving TBX1 and RANBP1

Individuals with very small nested duplications and deletions offer a unique method of studying the associations between isolated regions or genes and individual features of the 22q11.2DS phenotype. In the present study, we could not tease apart the contributions of individual genes to portions of the phenotype, as the LCR-A to B region includes approximately 25 genes. Here, we contrast two case studies, case study 1 and a prior case study by Weisfeld-Adams and colleages [53], each with a very small duplication including either TBX1 or RANBP1, but not both, to provide some insight into the possible relative contributions of TBX1 and RANBP1 to the phenotype in a descriptive fashion. Weisfeld-Adams et al. described a patient and sibling with duplication of six genes including TBX1 but not RANBP1. This proband showed complex medical problems, but neither the 19-month-old proband nor the 3-year-old sibling showed any symptoms of autism or neurodevelopmental delay besides mild motor delay. (Although no concern for ASD was noted at 19 months of age, we caution against over-interpretation because ASD can be missed in toddlers when symptoms are not severe. However, by 19 months of age most children with 22q11.2DS show significant delays, little speech, and aloof social behavior, so the lack of delay suggests social development was on course.) In contrast, in case study 1, we described an individual with microduplication involving RANBP1 but not TBX1 who had ASD but no medical comorbidities. Both our patient, who had a purely psychiatric phenotype and duplication that does involve RANBP1, and the case presented by Weisfeld-Adams et al.––a purely medical phenotype that does not involve RANBP1––provide preliminary suggestive evidence that RANBP1, not TBX1, specifically might confer risk for ASD and other psychiatric diagnoses. Both microduplications include COMT and exclude PRODH, so we cannot speculate about the roles of these genes based on case studies.

Limitations

The two primary limitations of our study lie in the phenotyping and the sample size. This single-site study relied primarily on questionnaires and chart review, supplemented by in-person evaluation when feasible for the family. Thus, the phenotyping, while accurate, could be improved with systematic prospective evaluations. Our sample size was small, owing to the rarity of individuals with nested duplication or deletions in the 22q11.2 region. Our study would benefit from replication with a multi-site study that combines clinics around the world to improve statistical power.

Another limitation includes the unknown role of background genetics. We were unable to account for other contributors to ASD risk, such as common variants or known pathogenic variants occurring outside 22q11.2 that would be identified with whole-exome sequencing, not clinical genetic testing with MLPA and SNP arrays. However, this risk is likely to affect all groups equally. Furthermore, we believe this unknown potential risk is likely to be small compared to the known, larger ASD risk of carrying 22q11.2DS or DupS.

Future directions might involve whole-exome sequencing of 22q11.2 samples to identify other factors that contribute to ASD risk. Such a study should include an analysis leveraging the sequencing of PRODH, COMT, RANBP1, and TBX1 in individuals with nested 22q11.2 deletions and duplications to isolate the influence of these mutations on the ASD phenotype.