Using ES and CMA, we identified or confirmed a molecular diagnosis in 134 of 332 pediatric participants (40.4%) with suspected or known IEI. CNVs contributed to 18 of 134 diagnoses (13.4%), increasing the overall diagnostic yield by 15.5%. Nondiagnostic CNVs were found in one-third of all probands (109/332, 32.8%). Notably, half the participants who received CMA diagnoses had diagnostic variants in at least one gene not currently listed by the IUIS as the cause of an IEI, highlighting the importance of comprehensive genotypic and phenotypic evaluation of patients with known or suspected IEI. Including only known immune genes in CNV analysis would have reduced CMA contribution to diagnostic yield by half. The complexity of these cases is further emphasized by data indicating that nearly one-third of participants (107/332, 32.2%) had phenotypes involving ten or more top-level HPO categories, roughly corresponding to organ systems.
The diagnostic yield of CNV analysis varies based on cohort characteristics [15, 23]. Studies examining the contribution of CNVs to diagnostic yield for IEI also vary [14, 15], which may be explained by different targets for computational methods calling CNVs in immune genes from ES data and the specific characteristics of different research cohorts. Unlike prior analyses, here we used CMA rather than computational methods to call CNVs and limited our analyses to pediatric participants. Overall, our finding of a 15.5% increase in diagnostic yield from CMA is comparable to but slightly higher than other studies [14, 23, 28].
The reasons for performing CMA in the participants we describe were varied: some were referred for CMA based on their clinical phenotypes, particularly those with neurological involvement; others based on findings from the ES data; still others received CMA without specific clinical or genetic indications for further testing beyond the lack of a diagnosis from ES and young age. As a result, we predict that the diagnostic contribution of CMA would be lower if this test were performed in an unbiased manner. Conversely, we would expect the diagnostic contribution of CMA to be greater when CMA is conducted on a targeted basis.
Fifty-three CMAs returned only copy number VUS. While VUS may be reclassified over time as scientific understanding of the human genome increases, the practical logistics of reclassification at a large scale may limit implementation in practice. Genetic counseling serves an important role in contextualizing these findings for both patients and providers, since misinterpretation of these variants can lead to unnecessary emotional distress or medical mismanagement [15].
At least one gene not associated with an IEI was implicated in half of the participants who received diagnoses from CMA (9/18, 50.0%). In some cases, these non-IUIS findings explained the participant’s immunological phenotype. For instance, a known 1 kb deletion was identified in trans with a pathogenic nonsense variant in DNAH5. Recessive loss of function variants in DNAH5 are associated with primary ciliary dyskinesia (PCD). While the root cause of the associated phenotype lies outside the immune system, PCD manifests with several of the same symptoms as many IEIs, including recurrent bacterial respiratory infections, chronic cough, and bronchiectasis. Similarly, while TGFBR1- and TGFBR2-related Loeys-Dietz syndrome are included in the IUIS gene list for IEI, SMAD3-related Loeys-Dietz syndrome is not, despite manifesting with many of the same symptoms. Given the complexity of these participants’ phenotypes, a common situation in IEI clinical practice, we were not surprised to find relevant variants outside the current list of IEI-associated genes when performing comprehensive genetic evaluation. These cases highlight the value of a molecular diagnosis in clarifying the root cause of conditions that may be clinically indistinguishable from one another and illustrate the limitations of relying on narrow analysis.
In other cases, non-IUIS findings contributed to a condition unrelated to the participant’s immunological phenotype. None of these CNVs involved genes listed as secondary findings by the ACMG; instead, they explained specific components of the participants’ phenotypes, even when they did not provide answers for the primary reason for referral. The majority (6/9) of the clinically significant non-immunological CNVs were neurological diagnoses. For example, P0002492 presented with coccidioidomycosis and various neurological symptoms that were presumed related to infection of the brain and spine. However, CMA showed a CNV causative of Charcot-Marie-Tooth Disease Type 1A. This rare nervous system disorder can lead to weakness, atrophy, and sensory loss in the limbs beginning in adolescence. This CMA finding was found to be consistent with and more likely to explain her neurological phenotype. This case demonstrates the potential for unexpected diagnoses that comprehensive genetic evaluation can provide. Particularly for patients with multiple diagnoses, which are a common occurrence in IEI clinical practice, it can be difficult to distinguish the root cause of a given phenotype, and careful consideration is required for the return of results and clinical decision-making.
More generally, the high incidence of non-IUIS diagnostic CNVs emphasizes the importance of a comprehensive approach including both genotypic and phenotypic evaluation in patients with IEI. Nearly one-third of participants’ phenotypes involved ten or more top-level HPO categories. It can be difficult to rule out the possibility of multiple diagnoses without comprehensive evaluation. Further, since immune dysfunction can impact every organ system, these complex cases can expand our understanding of the phenotypic spectrum in rare immune conditions.
In some cases, a contiguous gene deletion that encompassed multiple genes contributed to a complex phenotype. For instance, P0007442 presented with recurrent respiratory infections, nausea and vomiting, low circulating ornithine levels, and global developmental delay. CMA identified a 3.801 Mb heterozygous deletion on Xp21.1p11.4 affecting 17 genes, including CYBB and OTC. Variants in CYBB are associated with X-linked chronic granulomatous disease (CGD) and immunodeficiency 34. While CYBB-associated conditions are classified as X-linked recessive, female carriers, like this participant, often exhibit less severe immune phenotypes, including recurrent infections and gastrointestinal inflammation. Similarly, female heterozygotes carrying defects in OTC have been known to manifest symptoms of ornithine transcarbamylase deficiency under metabolic stress. This case highlights that diagnostic CNVs may involve multiple genes in combination that contribute to a participant’s overall phenotype, as opposed to sequence variants that primarily manifest as monogenic disorders. These patients are often clinically complicated and may require heightened attention to medical management. (Interestingly, more complex phenotypes were associated with greater likelihood of genetic diagnosis by ES, but not by CMA, despite the larger genetic area encompassed by CNV. However, these findings may be due to the small sample size of CMA diagnoses relative to ES diagnoses in our cohort.)
In other cases, multiple types of variants contribute to a single genetic diagnosis, necessitating combined evaluation of data from multiple types of genetic testing. As discussed above, in P0007523, ES identified a heterozygous c.6763C > T (p.Arg2255Ter) nonsense variant in DNAH5 and subsequent CMA detected a 1 kb copy number loss affecting DNAH5, thus explaining the participant’s phenotype and establishing the molecular diagnosis of PCD. Cases like this that require multiple tests to reach a diagnosis highlight the utility of combining ES and CMA for comprehensive analysis.
This study has several limitations. Participants referred to this study constitute a unique research cohort with extensive prior workup, which may select for particularly complicated cases. For instance, some classic IEIs such as complement deficiency were conspicuously underrepresented or absent. Moreover, these participants are generally referred to specific protocols for the expertise of particular investigators, which biases the cohort toward certain phenotypes. Additionally, as discussed above, the varied reasons for CMA may affect the diagnostic yield relative to cohorts sending CMA on an unbiased or more selective basis. Previous molecular diagnoses also informed the decision to perform CMA for some participants.
Consistent with previous studies [14], nearly two-thirds of participants with clinical features suggestive of IEI did not receive a molecular diagnosis, highlighting the need for a greater understanding of the complex factors that can lead to IEI, including mosaicism, genetic modifiers, epigenetic regulation, environmental factors, and the stochastic nature of rearrangements leading to the repertoire of B and T cell receptors in each individual [28, 40, 41]. Moreover, the relevance of CMA in identifying CNVs is evolving with the increasing ability to detect CNVs from genome sequencing (GS) data. At this time, GS remains an inaccessible option for many patients outside of a research setting, the ability of the predominant short-read sequencing methodologies to identify structural variants seems to be incomplete [42], and direct comparisons of the clinical performance of array-based versus GS-based CNV analysis in IEI have not been performed. Further research is also needed to elucidate the overall CNV burden in IEI and the potential association of CNVs with young age of onset. Age of onset data were not available for this cohort.
Molecular diagnosis of IEI can be complex due to the potential for many overlapping factors contributing to patient phenotypes. Establishing a molecular diagnosis has substantial implications for medical management and genetic counseling. We describe the molecular diagnostic contribution of CMA as a supplement to ES in children with IEI, highlighting the role of CNV detection in the diagnosis of IEI. CMA accounted for 13.4% of all diagnoses in this cohort, a 15.5% increase in diagnostic yield. Notably, half of CMA diagnoses at least partially involved non-immune genes, which would not typically appear on commercial panels for IEI. We observed that this two-pronged approach to genetic testing helped untangle complex phenotypes, not only by clarifying the differential diagnosis, but, in some cases, by identifying multiple diagnoses that contributed to the participant’s overall presentation. For children with unexplained IEI, coupling CMA and ES can provide a comprehensive evaluation that clarifies the complex factors contributing to both immune and non-immune phenotypes.