Clinical Genetic Testing for Patients With Autism Spectrum Disorders

DISCUSSION

Our findings indicate that CMA with whole-genome coverage detects more abnormalities than G-banded karyotype and fragile X DNA testing in patients with ASD, and suggest that CMA should be a first-tier test in this patient population. CMA could not entirely replace a G-banded karyotype in this patient population because of the inability of CMA to detect balanced rearrangements, but these are a small proportion of abnormal results. We identified 10 patients with a balanced rearrangement representing 1.2% of all patients tested (n = 852). If these patients had only been tested by CMA, then it is possible that a pathogenic change would be missed.

Although CMA does not detect balanced rearrangements, a significant proportion of balanced rearrangements are probably not clinically significant. The balanced pericentric inversions on chromosome 2 (patient ASD-09-002) and chromosome 9 (patients ASD-09-017 and ASD-09-018) could also occur in healthy individuals and likely are not related to ASD. In fact, the chromosome 2 inversion was maternally inherited. Chromosome 9 inversions are known polymorphisms, and also likely inherited, but parental samples were not available for testing.

We found 6 cases of balanced translocations, but they are also not necessarily pathogenic. They may be inherited from an unaffected parent, making the child a balanced carrier like the parent. Among 6 balanced translocations in our cohort, 1 patient (ASD-09-004) had the identical result as the parent, 2 cases had no parent data, and 3 cases were de novo. The de novo balanced translocations are not necessarily pathogenic, either. Balanced rearrangements are known to occur in healthy individuals, even when they interrupt a known gene. In a recent study of balanced rearrangements that interrupt a gene, approximately half (16 of 31) were found in healthy individuals.¹⁵

Pathogenic balanced rearrangements are likely to account for only a small number of ASD cases. Studies of cytogenetically balanced rearrangements in large cohorts of patients with ASD are not available, but such studies have been done for patients with MR and should be comparable. Balanced rearrangements make up only ∼10% of cytogenetically visible abnormalities in patients with developmental disabilities such as MR, meaning that only ∼0.3% of patients would have such changes.^16,–,18 Although traditional karyotyping could detect these events, they represent a similarly small proportion of cases in our cohort and may or may not be related to ASD.

Patient ASD-09-009 had mosaicism for a marker chromosome that is probably of little clinical significance because (1) the level of mosaicism is low, and (2) small marker chromosomes typically contain gene-poor repetitive DNA. CMA does not contain probes from these repetitive DNA regions, and the failure of whole-genome CMA to detect this anomaly is actually evidence that the marker is repetitive DNA. Karyotype testing and CMA can detect mosaicism at the level of ∼5% to 10% abnormal cells and 30% abnormal cells, respectively. We only found 1 such example of low-level mosaicism, demonstrating that these events also occurred at low frequency in our ASD cohort.

The proportion of patients with positive results for any of the 3 tests in this study was similar to other studies on ASD, some of which were performed on research samples.^3,10,11,19 Our yield for CMA was <10%, perhaps for several reasons. Whole-genome scans for copy-number variation have identified large de novo CNVs in 7% to 10% of simplex ASD families (1 child affected), 2% to 3% of multiplex families, and only 1% of control families.^10,19 Our patients were added through clinical care and were not selected on the basis of simplex versus multiplex families and are, therefore, not enriched for simplex cases. Diagnostic yield of CMA may have been limited by technical factors. Some tests (∼17%) were performed on platforms that have coverage below the ability to detect all 500-kb copy-number changes. However, most of our samples (83%) were tested by Agilent 244k or Affymetrix 500k and v5.0 whole-genome arrays. The trend toward higher yield with whole-genome arrays as compared with targeted arrays has been reported by authors of other studies.⁷

We might have expected to find higher numbers of definite abnormal results for CMA on the basis of yields for patients with generalized MR, which are ≥10%.^20,21 Our yield was lower, but our cohort of patients with ASD almost certainly contains more high-functioning individuals than a cohort of patients with MR, including 31 individuals with Asperger disorder in whom no clinically significant CNVs were identified. This suggests that yield from CMA may be lower in patients with high functioning autism, and this is consistent with other reports.²² Our cohort had a relatively low proportion of patients with secondary diagnoses known to have a high rate of abnormalities on CMA. Only 54 of 461 patients (11.7%) in the AC cohort were diagnosed with MR by medical record review. Similarly, only 16 of 461 patients (3.5%) in the AC cohort had a secondary diagnosis of multiple congenital anomalies, which was reported to have CMA abnormalities in 19.9% of patients.⁸ Our yield of abnormal results for fragile X testing was also lower than expected but may represent a selection bias against patients with fragile X syndrome, as has been suggested in similar studies.²³ Two of these patients with fragile X syndrome were premutation carriers, but their results were included as abnormal because recent studies revealed that there may be a higher incidence of neuropsychiatric conditions, including autism, among fragile X premutation carriers.²⁴

Our study has potential limitations. Our patients were diagnosed by clinical evaluation using DSM-IV-TR criteria. The gold standard for research studies of ASD would include the Autism Diagnostic Observation Schedule (ADOS) and the Autism Diagnostic Interview-Revised (ADI-R) in addition to meeting criteria for a pervasive developmental disorder as defined by the DSM-IV-TR. Some of the patients included in this study may not have met full research criteria for an ASD diagnosis if tested with the ADOS and ADI-R. Removing some patients from our sample on the basis of failure to meet criteria for an ASD diagnosis because of ADI-R/ADOS may actually increase the proportion of patients with an abnormality by removing patients with a milder phenotype. We cannot exclude the possibility of bias based on ascertainment of patients through tertiary care centers. These patients may be more likely to have abnormal genetic test results because they were referred because of other complicating factors such as specific family history or dysmorphic features. We did not observe a high rate of such issues, but we cannot rule out underreporting of complex features at the time of ascertainment.

The causal relationships between many of the abnormal CNVs identified in these patients with ASD and the clinical symptoms will require further study. Our conclusions about pathogenicity are based on the best current knowledge but could evolve over time.

In general, sporadic cases of autism may be more likely caused by de novo mutations.²⁵ Inherited CNVs may also contribute to autism or autistic symptoms but may have more mild effects that could vary among family members. It is ironic that many apparently common recurrent pathogenic copy-number changes may not be de novo but exhibit decreased penetrance and variable expressivity, such as 16p11.2,15q13.2q13.3 and 1q21.^26,–,29 This has important implications for recurrence risk counseling. Identifying rare de novo copy-number changes is equally important for genetic counseling.

The increased yield of CMA, especially in comparison with G-banded karyotype testing, has important clinical impact. Genetic testing can be expensive, and payers may not be willing to reimburse for 2 tests that provide similar information. In such cases, CMA would be an appropriate choice despite a small number of balanced rearrangements that would be undetectable. Although we identified slight differences in the rate of abnormal CMA results based on gender and specific ASD category, these should not influence clinical decisions about offering CMA given the small magnitude of differences and also the potential variability of diagnosis over time, particularly in young children.^30,–,32 Also, other genetic testing may be indicated in select populations of patients with ASD (eg, testing for MECP2 mutations among girls with ASD and microcephaly or testing for PTEN mutations among boys or girls with ASD and macrocephaly).^33,34

Establishing a clear diagnosis may lead to earlier initiation of services and consequently improve outcome.^35,–,38 In most cases of ASD, some clinical symptoms are apparent before the age of 3 years, but in many cases children may not be diagnosed until they are much older.³⁹ ASD will remain a clinical diagnosis, but identifying a clear genetic etiology is advantageous in several ways. A clear genetic diagnosis can affect patient management decisions, availability of developmental services, and accuracy of genetic counseling about recurrence risks, which may range from <5% to as high as 50% depending on the cause. A clear genetic diagnosis also spares the patient and family a diagnostic odyssey involving multiple rounds of diagnostic testing.

Specific clinical recommendations for including CMA as a first-tier test in the evaluation of patients with ASD have not kept pace with this rapidly evolving technology. Considerations for including CMA in the evaluation of children with ASD have been outlined elsewhere^4,40,41 but have stopped short of recommending that CMA be offered as a first-tier genetic diagnostic test for ASD. On the basis of our results, genetic diagnosis will be missed in at least 5% of ASD cases without CMA, and our results suggest that CMA with whole-genome coverage should be adopted as a national standard of care for genetic testing among patients with ASDs.