The global prevalence and ethnic heterogeneity of primary ciliary dyskinesia gene variants: a genetic database analysis

Abstract

Background: Primary ciliary dyskinesia (PCD) is a motile ciliopathy characterised by otosinopulmonary infections. Inheritance is commonly autosomal recessive, with extensive locus and allelic heterogeneity. The prevalence is uncertain. Most genetic studies have been done in North America or Europe. The aim of the study was to estimate the worldwide prevalence and ethnic heterogeneity of PCD.

Methods: We calculated the allele frequency of disease-causing variants in 29 PCD genes associated with autosomal recessive inheritance in 182 681 unique individuals to estimate the global prevalence of PCD in seven ethnicities (African or African American, Latino, Ashkenazi Jewish, Finnish, non-Finnish European, east Asian, and south Asian). We began by aggregating variants that had been interpreted by Invitae, San Francisco, CA, USA, a genetics laboratory with PCD expertise. We then determined the allele frequency of each variant (pathogenic, likely pathogenic, or variant of uncertain significance [VUS]) in the Genome Aggregation Database (gnomAD), a publicly available next-generation sequencing database that aggregates exome and genome sequencing information from a wide variety of large-scale projects and stratifies allele counts by ethnicity. Using the Hardy-Weinberg equilibrium equation, we were able to calculate a lower-end prevalence of PCD for each ethnicity by including only pathogenic and likely pathogenic variants; and upper-end prevalence by also including VUS. This approach was similar to previous work on Li-Fraumeni (TP53 variants) prevalence. We were not diagnosing PCD, but rather estimating prevalence based on known variants.

Findings: The overall minimum global prevalence of PCD is calculated to be at least one in 7554 individuals, although this is likely to be an underestimate because some variants currently classified as VUS might be disease-causing and some pathogenic variants might not be detected by our methods. In the overall cohort, Invitae data could be included for variants without gnomAD data for a primary ethnicity. When using only gnomAD allele frequencies to calculate prevalence in individual ethnicities, the estimated prevalence of PCD was lower in each ethnicity compared with the overall cohort. This is because the overall cohort includes additional data from the Invitae database such as copy number variants and other variants not present in gnomAD. With gnomAD we found the expected PCD frequency to be higher in individuals of African ancestry than in most other populations (excluding VUS: 1 in 9906 in African or African American vs 1 in 10 388 in non-Finnish European vs 1 in 14 606 in east Asian vs 1 in 16 309 in Latino; including VUS: 1 in 106 in African or African American vs 1 in 178 in non-Finnish European vs 1 in 196 in Latino vs 1 in 188 in east Asian). In addition, we found that the top 5 genes most commonly implicated in PCD differed across ethnic ancestries and contrasted commonly published findings.

Interpretation: PCD appears to be more common than has been recognised, particularly in individuals of African ancestry. We identified gene distributions that differ from those in previous European and North American studies. These results could have an international impact on case identification. Our analytic approach can be expanded as more PCD loci are identified, and could be adapted to study the prevalence of other inherited diseases.

Funding: None.

Figure 1:. VUS (variants interpreted by Invitae as variant of uncertain significance), PLP (variants interpreted by Invitae as either pathogenic or likely pathogenic), pLoF (gnomAD presumed loss-of-function variant).

Figure 1 demonstrates the filtering process of variants included in the analysis. We aggregated all PLP variants classified at Invitae in 29 genes in the course of genetic testing. We used these variants to calculate the allele frequency in two presumably unaffected populations: gnomAD and Invitae patients who did not have testing ordered of PCD genes. Because patients who had testing ordered for evaluation of PCD are likely enriched for PLP variants, we excluded 3,824 PLP and VUS that were only observed in this population (or if they were in gnomAD but were not present in any of the major ethnicities (only “Other” ethnicity) and absent from the Invitae control population). gnomAD pLoF variants with an allele frequency that did not exceed the expected compared to PLP variants interpreted at Invitae were also aggregated. 10,688 variants seen in gnomAD or control individuals with testing at Invitae were present after excluding the 3,824 variants. 2703 variants were present in Invitae data but were either absent in gnomAD or were not seen in any of the major ethnicities; 23 of these VUS were excluded due to the high allele frequency, and 2680 of these variants were included in the analysis for calculation of an “Overall” prevalence. In total,713 VUS were excluded due to prevalence as detailed in the manuscript. A total of 9,975 variants were included in the analysis. 2,133 PLP were analyzed (of which 53 were CNVs and present only in the Invitae dataset). 7,842 VUS were analyzed. It is important to note that for some variants, there are not data available from gnomAD for every ethnicity. When variants interpreted at Invitae were absent from gnomAD, allele frequencies could be calculated using data from control individuals with testing ordered at Invitae (testing ordered for indications other than PCD); these data were only used in calculations for the “overall” cohort since information was not available regarding allele frequency in different ethnicities. There were data present in the Invitae control dataset and/or gnomAD for 9,975 variants in the overall cohort, 6,919 variants in the South Asian cohort, and 7,295 variants in the remaining ethnicities.

Figure 2a and 2b:

VUS (variant of uncertain significance). In each ethnicity, the portion of PCD due to each of the five genes most commonly implicated in PCD is shown with each of the five genes assigned a color in the graphs. The portion of PCD due to all other evaluated PCD genes is shown in gray. Figure 2a includes only the pathogenic/likely pathogenic variants. Figure 2b includes all variants. As noted in the text, the “other” ethnic groups are not displayed in Figure 2a and 2b as these individuals are included in the overall cohort and cannot be considered a separate ethnic grouping. As can be seen in the tables, the other group unsurprisingly most closely mimics the overall prevalence.

Figure 2a and 2b:

VUS (variant of uncertain significance). In each ethnicity, the portion of PCD due to each of the five genes most commonly implicated in PCD is shown with each of the five genes assigned a color in the graphs. The portion of PCD due to all other evaluated PCD genes is shown in gray. Figure 2a includes only the pathogenic/likely pathogenic variants. Figure 2b includes all variants. As noted in the text, the “other” ethnic groups are not displayed in Figure 2a and 2b as these individuals are included in the overall cohort and cannot be considered a separate ethnic grouping. As can be seen in the tables, the other group unsurprisingly most closely mimics the overall prevalence.