We examined the genomic structure of the 1q21.1 region to determine the relationships among the various low-copy repeats (LCRs) (Fig. 1b). As previously indicated², this region contains numerous LCRs that could mediate genomic rearrangements. The reference sequence used in our analysis (Human Genome Build 36.1, from the University of California Santa Cruz (UCSC) genome browser) contains 15 gaps within sub-band 1q21.1, many of which are adjacent to LCRs, thus limiting the analysis, and revision of the reference sequence for this region is likely to alter the overall structure of the LCRs.

We carried out FISH analyses to refine the breakpoint regions. BAC clone RP11-769J20 was used to examine whether CNVs extend through the TAR syndrome region. BAC clones RP11-115G11 (which maps within the telomeric repeat region but in a region of relatively unique sequence) and RP11-15H9 (which maps to sub-band 1q21.2, distal to the telomeric repeats and the cluster of histone genes) were used to determine the telomeric extent of the CNVs (summarized in Fig. 1c).

Combined array-CGH and FISH results indicated that deletions cluster into two classes. Class I deletions are defined as involving only the distal 1q21.1 region, with telomeric breakpoints mapping within clone RP11-115G1. Class II deletions are defined as larger deletions including both the TAR syndrome and distal 1q21.1 regions, with telomeric breakpoints mapping within RP11-115G1 or further within the telomeric LCR (Fig. 1c).

Among microduplication cases, 16 of 17 involved only the distal 1q21.1 region (class I). In one case (clinical information not available), the microduplication extended from the TAR region through the distal 1q21.1 region (class II). In all microduplication cases, clone RP11-115G11 showed three signals whereas RP11-15H9 showed two signals, indicating that the breakpoints were located within or distal to RP11-115G11. One individual (subject 36) with microduplication of the distal 1q21.1 region was found to have an intriguing complex rearrangement involving microdeletion of the TAR region, which was inherited from the mother and also present in two similarly affected siblings. Both parental samples were available for analysis in 17 of 21 microdeletion cases and 12 of 15 microduplication cases, and the CNVs were inherited from a parent in most cases (Tables 1 and ​and2).2). There were three de novo microdeletions and one de novo microduplication.

Certain facial dysmorphic features such as frontal bossing, deep-set eyes and bulbous nose in individuals with 1q21.1 microdeletion (Fig. 2), and frontal bossing and hypertelorism in individuals with 1q21.1 microduplication (Fig. 3) seem to be common. Otherwise, no distinctive pattern of facial dysmorphic features was noted. A wide range of congenital anomalies was present, although without a clearly apparent pattern (Tables 1 and ​and2).2). Developmental delay and/or learning disabilities were reported in most cases. Finally, behavioral abnormalities were frequently observed, including attention deficit hyperactivity disorder (ADHD), autism, anxiety/depression, antisocial behavior, aggression and even hallucinations. It should be noted that many of the affected individuals are young and may not yet have manifested atypical behaviors at the time of examination. Furthermore, clinically relevant abnormalities including congenital cataract, learning disabilities, ADHD, depression/anxiety and antisocial behavior were present in some but not all parents carrying the 1q21.1 CNVs (Tables 1 and ​and2),2), and these parents had not previously been suspected of having a chromosomal abnormality.

Individuals with microdeletion and microduplication differed significantly in their mean head sizes (unpaired Student's t-test, P < 0.0001). The mean Z score for microdeletion cases (probands, parents and siblings carrying the microdeletion) was −2.53 (95% confidence interval (c.i.) = −2.96; −2.11), significantly different from the population mean, which is zero by definition (one sample t-test, P < 0.0001). Among these individuals 21 of 29 had a head circumference Z score less than −2. There was greater variation among microduplication cases, with 10 of 24 having head circumference Z scores greater than +2. Nevertheless, the mean Z score for microduplication cases (probands, parents and siblings carrying the microdeletion) was +0.95 (95% c.i. = 0.06; 1.83), which was statistically different from the population mean (one sample t-test, P < 0.05). Because of the possible effect of shared genetic background on head size among parents and siblings, we also carried out the statistical analyses on probands only, finding that the mean Z score for probands with the microdeletion was −2.55 (95% c.i. = −3.12; −1.98) and the mean Z score for probands with the microduplication was +1.15 (95% c.i. = 0.03; 2.28). Both the microdeletion and microduplication groups were still significantly different from each other (unpaired t-test, P < 0.0001) and from the population mean (deletion probands, one sample t-test, P < 0.0001; duplication probands, one sample t-test, P < 0.05). To rule out ascertainment bias for head size abnormalities within the referred population, which is enriched for individuals with clinically relevant phenotypes, we obtained head circumference measurements for 50 randomly selected individuals referred for array-CGH testing during the same time period in which the index cases were received. These control individuals had a much wider range of head circumference measurements than either the microdeletion or microduplication cases. The mean Z score among control individuals was −0.62 (95% c.i. = −1.33; 0.08), which was not statistically different from the population mean (one sample t-test, P = 0.083) (Fig. 4a). In addition, we found statistically significant differences between microdeletion cases and controls (unpaired t-test, P < 0.0001) and between microduplication cases and controls (unpaired t-test, P < 0.01), thus strengthening the finding that 1q21.1 CNVs are important in determining head size (Fig. 4a).

It was previously reported that a large interchromosomal duplication containing a paralog of the HYDIN gene on chromosome 16q22.2 was inserted into 1q21.1 during primate evolution, but this contig is not accounted for in the current build of the human genome⁷. Homozygous inactivating mutations of HYDIN cause hydrocephalus in mice⁶, probably as a result of impaired ciliary function of ependymal cells⁹. Notably, a conserved protein domain found several times in the HYDIN protein and designated as ASH (ASPM, SPD-2, Hydin) was identified at the N terminus of abnormal spindle-like microcephaly-associated protein (ASPM)^10,11, which is implicated in control of cerebral cortex size¹². Moreover, individuals with cytogenetically visible deletions on 16q22.2 that presumably involve the HYDIN gene have been reported to have microcephaly^13–15.

The 1q21.1 HYDIN paralog is exclusively expressed in the brain⁶, and we suggest that it could be involved in the head size abnormalities seen in our affected individuals. Accordingly, the 1q21.1 HYDIN paralog is deleted or duplicated in all individuals with deletion or duplication of distal 1q21.1, respectively (Fig. 4b–d). These results also indicate that this locus maps in one of the two gap segments between the TAR syndrome region and the distal 1q21.1 region. Indeed, it was predicted on the basis of sequence data that the duplicated paralog extends into a contig (NT_034398) mapping to 1q21.1 (^{ref. 7}). Given that the LCRs flanking the distal 1q21.1 region most likely mediate reciprocal deletion or duplication (Fig. 1b, blue blocks labeled A and A′), the most parsimonious placement of the HYDIN paralogous segment is within the gap immediately centromeric to the distal 1q21.1 region (Fig. 1a,b, gray triangle bounded by dashed lines). Consistent with this proposal, we have also identified a proband with an atypical deletion beginning approximately at nt 144253465 but not involving the distal 1q21.1 region, most likely mediated by a pair of small segmental duplications in this region (Fig. 1b, light blue blocks, and Fig. 1c). This atypical deletion does not include the HYDIN paralog by FISH (data not shown), and the head circumference of this individual is within the normal range. Our results are particularly of note in light of the recent finding of an association between 1q21.1 deletions and schizophrenia^4,5, and the recognition of smaller head circumference and facial dysmorphic features among some schizophrenic individuals^16–18.

The group of affected individuals described herein represents the largest collection of individuals with microdeletions or microduplications within chromosome 1q21.1 reported in the literature. We have clearly demonstrated a dosage effect of 1q21.1 copy number on head size. Most of the probands in our cohort presented with developmental delay, and many of them also manifested a range of behavioral abnormalities, nonspecific dysmorphic features and congenital anomalies. These findings are similar to the broad range of referring diagnoses among people undergoing clinical array-CGH testing, thus posing a challenge to the delineation of specific syndromic features other than head size abnormalities. We found only one individual with microdeletion who had a bicuspid aortic valve, so the initial report of an association between distal 1q21.1 deletions and congenital heart defects¹ is probably a result of ascertainment bias.

CNVs within 1q21.1 can be found in apparently normal individuals (as was also observed for unaffected carriers of the TAR syndrome microdeletion⁸) or associated with psychiatric phenotypes such as schizophrenia^4,5. The presence of parents carrying 1q21.1 CNVs but not manifesting clinically evident phenotypes raises the issue of whether 1q21.1 microdeletions and microduplications are benign CNVs or are pathogenic variants with incomplete penetrance. Indeed, CNVs within distal 1q21.1 were previously reported in normal individuals in population screens of genomic variation^4,19 but only at a frequency of ~0.02% in two large control populations in studies of schizophrenia^4,5. Moreover, while this manuscript was under review, Mefford et al. reported a series of 4,737 controls in which no 1q21.1 microdeletions and only one microduplication case were found²⁰. Furthermore, no 1q21.1 CNVs were detected in 550 normal parental cases analyzed in our laboratory. Therefore, these CNVs occur at a higher frequency among individuals referred for clinical array-CGH testing and are unlikely to be benign.

It is clear that an important consideration in genetic counseling for these syndromes is the potential for reduced penetrance and variable expressivity, which has also been reported for well-characterized syndromes such as 22q11.2 microdeletion, where phenotypically mild deletion carriers have escaped clinical recognition until they had children with more severe manifestations²¹. Findings such as these and the recently described 16p13.11 and 15q13.3 microdeletion syndromes^22,23 raise difficult questions in the context of genetic counseling for newly diagnosed cases and particularly for prenatal diagnosis.

In conclusion, our results demonstrate that recurrent reciprocal microdeletions and microduplications within 1q21.1 represent novel genomic disorders consisting of microcephaly or macrocephaly, respectively, and can manifest with a range of developmental delay, neuropsychiatric abnormalities, dysmorphic features and a variety of other congenital anomalies. These features seem to be subject to incomplete penetrance and variable expressivity. We propose that the 1q21.1 HYDIN paralog (which was deleted or duplicated in all of our cases and in several deletion cases reported by Mefford et al.²⁰) is dosage sensitive and is important in determining head size. Other genes within the deleted or duplicated region could mediate other phenotypes associated with 1q21.1 CNVs, perhaps in association with other genetic or genomic variants.