Trio-based whole exome sequencing in patients with suspected sporadic inborn errors of immunity: A retrospective cohort study
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This study reports on the diagnostic utility of TRIO-based whole-exome sequencing (WES) in a cohort of 123 unrelated patients with suspected monogenic inborn errors of immunity. The authors further explored the diagnostic rate in this cohort by focusing their analyses on the identification of de novo variants (DNVs). This manuscript will be of interest to medical geneticists, immunologists, and physicians working with patients with primary immunodeficiencies.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
De novo variants (DNVs) are currently not routinely evaluated as part of diagnostic whole exome sequencing (WES) analysis in patients with suspected inborn errors of immunity (IEI).
Methods:
This study explored the potential added value of systematic assessment of DNVs in a retrospective cohort of 123 patients with a suspected sporadic IEI that underwent patient-parent trio-based WES.
Results:
A (likely) molecular diagnosis for (part) of the immunological phenotype was achieved in 12 patients with the diagnostic in silico IEI WES gene panel. Systematic evaluation of rare, non-synonymous DNVs in coding or splice site regions led to the identification of 14 candidate DNVs in genes with an annotated immune function. DNVs were found in IEI genes ( NLRP3 and RELA ) and in potentially novel candidate genes, including PSMB10 , DDX1 , KMT2C, and FBXW11 . The FBXW11 canonical splice site DNV was shown to lead to defective RNA splicing, increased NF-κB p65 signalling, and elevated IL-1β production in primary immune cells extracted from the patient with autoinflammatory disease.
Conclusions:
Our findings in this retrospective cohort study advocate the implementation of trio-based sequencing in routine diagnostics of patients with sporadic IEI. Furthermore, we provide functional evidence supporting a causal role for FBXW11 loss-of-function mutations in autoinflammatory disease.
Funding:
This research was supported by grants from the European Union, ZonMW and the Radboud Institute for Molecular Life Sciences.
Article activity feed
-
-
Author Response
Reviewer #1 (Public Review):
The current manuscript examined patients with inborn errors of immunity (IEI) using whole exome sequencing (WES) and identified de novo variants (DNVs) associated with the disease. They found 14 genes associated with DNVs, including four novel genes - PSMB10, DDX1, KMT2C, and FBXW11, and conducted a systematic assessment of affected genes.
Given the level of heterogeneity underlying IEI, the sample size is limited. Although the authors clearly stated this, the analysis of the current manuscript does not add much value to describing genes affected by DNVs. The sample size is small to perform exome-wide evaluation (authors described they did "exome-wide evaluation" in Abstract - line 10 but there is no statistical evaluation to prioritize effect genes). They could go with systems biology …
Author Response
Reviewer #1 (Public Review):
The current manuscript examined patients with inborn errors of immunity (IEI) using whole exome sequencing (WES) and identified de novo variants (DNVs) associated with the disease. They found 14 genes associated with DNVs, including four novel genes - PSMB10, DDX1, KMT2C, and FBXW11, and conducted a systematic assessment of affected genes.
Given the level of heterogeneity underlying IEI, the sample size is limited. Although the authors clearly stated this, the analysis of the current manuscript does not add much value to describing genes affected by DNVs. The sample size is small to perform exome-wide evaluation (authors described they did "exome-wide evaluation" in Abstract - line 10 but there is no statistical evaluation to prioritize effect genes). They could go with systems biology approaches, explaining the biological pathway of affected genes or underlying cell types from immune single-cell datasets. As the authors stated that IEI constitutes a large group of heterogeneous disorders, there should be some analysis to explain the functional convergence of affected genes in disease development.
We believe the term ‘exome-wide evaluation’ might have led to misinterpretation. We used it in the context of reviewing each DNV found in a single patient’s exome outside the diagnostic IEI gene panel (i.e. ‘exome-wide’), instead of reviewing DNVs across all exomes. We have rephrased the sentences containing this term. The main purpose of this manuscript was to identify ‘all’ coding DNVs in each case, and explore whether they include any pathogenic or novel candidate DNVs. Our main purpose was to urge the IEI field to apply trio-based WES more systematically, and share candidate DNVs with the field for further validation.
As the reviewer points out, our sample size would be too limited to perform systems biology approaches for variant prioritization. The signal-to-noise ratio would be very high, because many genes causing inborn errors of immunity remain to be discovered and the studied group of patients with inborn errors of immunity is very heterogeneous. This means that we would not have the power to investigate potential enrichment or burden of DNVs in specific genes nor the functional convergence of affected genes or pathways in specific phenotypes. In this study, we aimed to show the additional value of the systematic DNV analysis as a method to identify and prioritize candidate variants in individual cases, but ideally we would like to answer other important research questions using computational/statistical approaches in a larger cohort in the future, as has been performed in other rare disease fields. The suggestion of the reviewer is helpful, and this approach has been shown to implicate novel pathways enriched in disease for various forms of neurodevelopmental diseases for which ten-thousands of trio-based WES have been performed [9, 10].
For DNV identification, the authors filtered out variants with ExAC & gnomAD AF > 0.1% or GoNL AF > 0.5%. I think this is too lenient a cutoff for filtering for DNV. For example, gnomAD AF 0.1% is approximately ~200 individuals in population. Given the filtering parameters (<5 variation reads, <20% variant allele frequency, or low coverage DNVs), they did not use specific filtering metrics to find DNV and there might be false-positive variants in the final DNV set. As far as I can find in the manuscript, they used the GATK pipeline from the previous study (REF 29). The GATK unified genotype generates a range of filtering metrics to increase specificity in variant filtering. It is very surprising that the authors seem to use three parameters (variation reads → FORMAT:AD[1]; variant allele frequency → FORMAT:AB? and low coverage → FORMAT:DP? but the authors did not state the cutoff) to filter de novo variants, which are fragile to false-positive variant calling.
The chosen population database fraction cut-offs align with DNV filtering strategies in literature. We have not chosen a stricter cut-off to avoid missing true positives, since patients with IEI can exhibit late-onset disease, variable penetrance and have postzygotic mutations, while limiting the chance of false-positive findings. For instance, we have reduced local false-positives by filtering on allele frequencies in our in-house database and Dutch population database. Moreover, automated DNV calling required >2% alternate reads in either parent and variants were prioritized based on prediction scores and annotated immune function. Additionally, and in accordance with this expert reviewer, we have now put a stricter cut-off in place for variation reads (from 5 to 10) to further minimize false-positive findings. Lastly, we visually inspected the final 14 candidate DNVs in IGV and/or Alamut, which supports the validity of the findings. The DNVs reported in our final DNV list (Table 2B) are therefore unlikely to contain falsepositive findings.
Reviewer #2 (Public Review):
The manuscript by Hebert et al., reports on the utility of TRIO-based whole-exome sequencing (WES) in patients who presented as sporadic cases and are suspected of having inborn errors of immunity (IEIs). The authors developed an in-house pipeline for data analysis and used a set of known algorithms to prioritize the impact of genetic variants located mostly in the coding region of proteins. The data analysis was done in two steps; the first step involved the routine WES diagnostic analysis that led to the identification of pathogenic (P) and likely pathogenic variants (LP) in genes already associated with IEIs. The authors claim that this analysis resulted in a likely molecular diagnosis in 19 (~15%) of patients, while an additional 14% of cases were carriers for VUSs or other risk factors in the disease causal genes. As many of these variants are either inherited from one parent or are present as heterozygous (monoallelic) variants in genes associated with recessive diseases, their clinical significance is unclear.
In the second step, the authors focused on the identification of de novo variants (DNVs), including SNVs, CNVs, and small indel, since these variants are more likely to be deleterious on protein function. The authors identified 136 non-synonymous DNVs, which were then filtered down to 14 best candidate variants using various in silico tools and database searches. These 14 variants included DNVs in genes previously associated with autoinflammatory diseases, such as CAPS and RELA haploinsufficiency. Three patients are found to carry de novo copy number variants (CNVs) of unknown clinical significance. Finally, several de novo loss-of-function (LoF) variants have been identified in genes that are not yet associated with any IEIs but are good functional candidates. Their potential pathogenicity is further supported by the observation that they are found in genes intolerant to loss of function. Functional validation has been performed only for the patient carrier of the novel FBXW11 splice variant. The authors state that the maximum solve rate (i.e., probable molecular diagnosis) in this cohort might be as high as 23%, which is comparable to similar reports of patients with IEIs, however, the reported results do not yet support this conclusion.
The main conclusion of this study is that TRIO-based WES analysis for DNVs could improve the diagnostic rate and can result in the identification of novel disease-causing genes. TRIO-based sequencing is also preferable when analyzing patients from populations underrepresented in gnomAD and ExAC. As the cost of WES has come down, WES has been increasingly used in the clinical diagnosis of many human disorders. Despite the major progress in the development of novel sequencing technologies and new in silico tools, the diagnostic rate is still below 50%. In summary, this study suggests that despite the identification of over 400 genes associated with IEIs, there are many more genes to be identified and that the heritability of these diseases is very complex.
We thank the reviewer for the elaborate summary of our study and the suggestions that have helped to further improve the manuscript.
-
Evaluation Summary:
This study reports on the diagnostic utility of TRIO-based whole-exome sequencing (WES) in a cohort of 123 unrelated patients with suspected monogenic inborn errors of immunity. The authors further explored the diagnostic rate in this cohort by focusing their analyses on the identification of de novo variants (DNVs). This manuscript will be of interest to medical geneticists, immunologists, and physicians working with patients with primary immunodeficiencies.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)
-
Reviewer #1 (Public Review):
The current manuscript examined patients with inborn errors of immunity (IEI) using whole exome sequencing (WES) and identified de novo variants (DNVs) associated with the disease. They found 14 genes associated with DNVs, including four novel genes - PSMB10, DDX1, KMT2C, and FBXW11, and conducted a systematic assessment of affected genes.
Given the level of heterogeneity underlying IEI, the sample size is limited. Although the authors clearly stated this, the analysis of the current manuscript does not add much value to describing genes affected by DNVs. The sample size is small to perform exome-wide evaluation (authors described they did "exome-wide evaluation" in Abstract - line 10 but there is no statistical evaluation to prioritize effect genes). They could go with systems biology approaches, explaining …
Reviewer #1 (Public Review):
The current manuscript examined patients with inborn errors of immunity (IEI) using whole exome sequencing (WES) and identified de novo variants (DNVs) associated with the disease. They found 14 genes associated with DNVs, including four novel genes - PSMB10, DDX1, KMT2C, and FBXW11, and conducted a systematic assessment of affected genes.
Given the level of heterogeneity underlying IEI, the sample size is limited. Although the authors clearly stated this, the analysis of the current manuscript does not add much value to describing genes affected by DNVs. The sample size is small to perform exome-wide evaluation (authors described they did "exome-wide evaluation" in Abstract - line 10 but there is no statistical evaluation to prioritize effect genes). They could go with systems biology approaches, explaining the biological pathway of affected genes or underlying cell types from immune single-cell datasets. As the authors stated that IEI constitutes a large group of heterogeneous disorders, there should be some analysis to explain the functional convergence of affected genes in disease development.
For DNV identification, the authors filtered out variants with ExAC & gnomAD AF > 0.1% or GoNL AF > 0.5%. I think this is too lenient a cutoff for filtering for DNV. For example, gnomAD AF 0.1% is approximately ~200 individuals in population. Given the filtering parameters (<5 variation reads, <20% variant allele frequency, or low coverage DNVs), they did not use specific filtering metrics to find DNV and there might be false-positive variants in the final DNV set. As far as I can find in the manuscript, they used the GATK pipeline from the previous study (REF 29). The GATK unified genotype generates a range of filtering metrics to increase specificity in variant filtering. It is very surprising that the authors seem to use three parameters (variation reads → FORMAT:AD[1]; variant allele frequency → FORMAT:AB? and low coverage → FORMAT:DP? but the authors did not state the cutoff) to filter de novo variants, which are fragile to false-positive variant calling.
-
Reviewer #2 (Public Review):
The manuscript by Hebert et al., reports on the utility of TRIO-based whole-exome sequencing (WES) in patients who presented as sporadic cases and are suspected of having inborn errors of immunity (IEIs). The authors developed an in-house pipeline for data analysis and used a set of known algorithms to prioritize the impact of genetic variants located mostly in the coding region of proteins. The data analysis was done in two steps; the first step involved the routine WES diagnostic analysis that led to the identification of pathogenic (P) and likely pathogenic variants (LP) in genes already associated with IEIs. The authors claim that this analysis resulted in a likely molecular diagnosis in 19 (~15%) of patients, while an additional 14% of cases were carriers for VUSs or other risk factors in the disease …
Reviewer #2 (Public Review):
The manuscript by Hebert et al., reports on the utility of TRIO-based whole-exome sequencing (WES) in patients who presented as sporadic cases and are suspected of having inborn errors of immunity (IEIs). The authors developed an in-house pipeline for data analysis and used a set of known algorithms to prioritize the impact of genetic variants located mostly in the coding region of proteins. The data analysis was done in two steps; the first step involved the routine WES diagnostic analysis that led to the identification of pathogenic (P) and likely pathogenic variants (LP) in genes already associated with IEIs. The authors claim that this analysis resulted in a likely molecular diagnosis in 19 (~15%) of patients, while an additional 14% of cases were carriers for VUSs or other risk factors in the disease causal genes. As many of these variants are either inherited from one parent or are present as heterozygous (monoallelic) variants in genes associated with recessive diseases, their clinical significance is unclear.
In the second step, the authors focused on the identification of de novo variants (DNVs), including SNVs, CNVs, and small indel, since these variants are more likely to be deleterious on protein function. The authors identified 136 non-synonymous DNVs, which were then filtered down to 14 best candidate variants using various in silico tools and database searches. These 14 variants included DNVs in genes previously associated with autoinflammatory diseases, such as CAPS and RELA haploinsufficiency. Three patients are found to carry de novo copy number variants (CNVs) of unknown clinical significance. Finally, several de novo loss-of-function (LoF) variants have been identified in genes that are not yet associated with any IEIs but are good functional candidates. Their potential pathogenicity is further supported by the observation that they are found in genes intolerant to loss of function. Functional validation has been performed only for the patient carrier of the novel FBXW11 splice variant. The authors state that the maximum solve rate (i.e., probable molecular diagnosis) in this cohort might be as high as 23%, which is comparable to similar reports of patients with IEIs, however, the reported results do not yet support this conclusion.
The main conclusion of this study is that TRIO-based WES analysis for DNVs could improve the diagnostic rate and can result in the identification of novel disease-causing genes. TRIO-based sequencing is also preferable when analyzing patients from populations underrepresented in gnomAD and ExAC. As the cost of WES has come down, WES has been increasingly used in the clinical diagnosis of many human disorders. Despite the major progress in the development of novel sequencing technologies and new in silico tools, the diagnostic rate is still below 50%. In summary, this study suggests that despite the identification of over 400 genes associated with IEIs, there are many more genes to be identified and that the heritability of these diseases is very complex.
-
