Diverse and unselected adults with clinically relevant ACADS variants lack evidence of metabolic disease

Abstract

Introduction:

The clinical significance of Short-chain acyl CoA dehydrogenase deficiency (SCADD), caused by biallelic variation in the ACADS gene, is contested. Clinically ascertained individuals have a range of reported metabolic and physical symptoms. Conversely, individuals identified through newborn screening remain overwhelmingly asymptomatic. Two common ACADS variants, c.511C>T (p.Arg171Trp) and c.625G>A (p.Gly209Ser) are known to reduce enzymatic activity with undetermined clinical correlate. We applied a genome-first approach to evaluate the prevalence and clinical consequences of ACADS variants in an ancestrally diverse and unselected patient population.

Material and Methods:

We used exome sequence data linked to electronic health records (EHRs) to identify clinically relevant ACADS variants, and estimate their prevalence and clinical implications in 27,447 ancestrally diverse and unrelated adults from the BioMe Biobank in New York, NY. We extracted International Classification of Diseases, ninth (ICD-9) and tenth (ICD-10) revision codes corresponding to eight SCADD-associated phenotypes relevant to adults from participants’ EHRs. Phenotypes included intellectual disability, behavioral disorders with onset in childhood, epilepsy or seizure disorders, hypoglycemia, muscle weakness, metabolic acidosis, fatty liver, and a diagnosis of SCADD or disorder of fatty acid oxidation. We performed manual chart reviews for individuals homozygous for rare pathogenic variants. Multivariate logistic regression was used to determine the association between clinically relevant ACADS variants and phenotypes of interest.

Results:

1 in 10,000 BioMe participants were homozygous for rare pathogenic variants (PVs) in ACADS, 1 in 20 were homozygous or presumed compound heterozygous for common variants (CVs), and 1 in 300 harbored both a PV and a CV. Of the 2,035 variant positive individuals, none had a documented diagnosis of SCADD. We identified five PV/PV positive individuals, none of whom had evidence of symptomatic SCADD on manual chart review. CV/CV positive and CV/PV positive individuals did not have increased odds of any of the eight ACADS phenotypes evaluated compared to variant negative individuals (OR for CV/CV 0.99, 95% CI 0.86 – 1.1, p = .88; OR for CV/PV OR 1.49, 95% CI 0.87– 2.6, p = .15).

Conclusions:

The prevalence of clinically relevant ACADS variants in an unselected population was higher than previously reported SCADD prevalence of 1 in 35,000 in the United States. Clinically relevant variants in ACADS were not associated with evidence of metabolic disease in a large and ancestrally diverse adult population. These findings support the assertion that SCADD is more likely a biochemical entity without clinical correlate, in particular when caused by one or more common variants.

1. Introduction

Short-chain acyl-CoA dehydrogenase deficiency (SCADD, MIM 2014740) is an autosomal recessive condition with an estimated prevalence of 1:35,000 in the United States¹. Deficiency of short chain acyl-CoA dehydrogenase (SCAD, MIM 606885), the first enzyme in mitochondrial short-chain beta-oxidation, results in accumulation of butyryl-CoA (C4-CoA) byproducts, primarily ethylmalonic acid and butyrylcarnitine (C4)². SCADD is caused by biallelic variation in the ACADS gene. To date, more than 80 disease-associated variants have been reported³.

SCADD was first recognized as a clinical entity in 1987 in two neonates with metabolic acidosis and ethylmalonic aciduria⁴. In the years since, the clinical significance of this biochemical phenotype has been questioned due to a discordance in symptomatology between clinically ascertained patients and those identified through newborn screening. Patients diagnosed after symptomatic presentation have shown a range of indications for metabolic testing, including developmental delay, seizures, and susceptibility to metabolic acidosis and hypoglycemia^4–8. Conversely, patients identified through newborn screening programs have remained overwhelmingly asymptomatic on follow up, raising suspicion for ascertainment bias in clinically diagnosed patients^1,9. Nevertheless, newborn screen reporting for SCADD is recommended as a secondary target on the United States Recommended Uniform Screening Panel (RUSP)¹⁰.

Further complicating our understanding of the clinical significance of SCADD is the presence of two common variants, which reduce enzymatic activity with uncertain clinical correlate. These variants, c.511C>T (p.Arg171Trp) and c.625G>A (p.Gly209Ser), have an estimated allele frequency of 3–8% and 22–43%, respectively. Approximately 14% of the general population is homozygous or compound heterozygous for these variants^2,11–13. These variants have been proposed to confer susceptibility to SCADD in the homozygous state or in trans with a rare pathogenic variant, particularly in the presence of other environmental or physiologic stressors¹³. Conversely, other studies propose that individuals who are compound heterozygous or homozygous for these variants have SCADD deficiency without clinical disease¹⁴. Most patients ascertained clinically and through newborn screening harbor at least one of these common variants^5,15.

To address if SCADD represents an isolated biochemical finding or a clinical disease entity, we identified individuals harboring clinically relevant variants in ACADS in the electronic health record (EHR)-linked BioMe Biobank in New York, NY, and performed association analyses using EHR data to evaluate for evidence of manifest disease in these individuals.

2. Materials and Methods

2.1. Patient samples and data

The BioMe Biobank is an electronic health record (EHR)-linked biorepository with over 70,000 participants from the Mount Sinai Health System in New York, NY, USA. Patients are enrolled non-selectively, predominately through ambulatory care practices. Self-reported race and ethnicity data obtained from survey questionnaires administered at enrollment into BioMe were mapped into eight population groups, as previously described^16,17. BioMe participants provide blood samples that are linked to their de-identified EHRs. Both genotype array (Illumina Global Screening Array) and exome sequence data is available for over 30,000 BioMe participants through a collaboration with Regeneron Genetic Center^16,18. This study was approved by the Icahn School of Medicine Institutional Review Board.

2.2. Study participants

This study included BioMe Biobank participants aged 18 or over at enrollment and with exome sequence data available. We restricted our cohort to an unrelated subset of 27,794 participants, which included only one individual in every first- and second-degree relationship¹⁶. We then excluded 333 individuals with missing sex, 14 individuals with missing self-reported race and ethnicity data. Prevalence data was estimated in the remaining 27,447 individuals with all necessary demographic and clinical variables. For association analyses, we excluded 9,487 individuals heterozygous for ACADS PVs or CVs (i.e., carriers). The sample size for association analyses therefore included 17,960 unrelated ancestrally diverse adult participants with all necessary demographic and clinical variables (Figure 1). This included individuals identifying as European American (n=4,160), African/African-America (n=5,292), East/Southeast Asian (n=539), Hispanic/Latin American (n=5,515), South Asian (n=353), Native American (n=40), more than one race selected (Multiple selected; N=683), and did not select any of these race or ethnicity groups (Other; n=1,378). Genetic ancestry estimates used in downstream analysis were derived via principal component analysis using genotype array data and have been previously described¹⁷.

Figure 1.

Flow chart showing inclusion/exclusion of study participants and analytic methods by genotype. Sample size is designated as (n).

Abbreviations: CV: common variant; PV: pathogenic variant; CV/CV: Homozygous or presumed compound heterozygous for CVs; CV/PV: Presumed compound heterozygous for a CV and a PV, PV/PV: homozygous for a pathogenic variant.

2.3. Variant Curation

Exome sequence data were analyzed for clinically relevant variants in ACADS, including pathogenic or likely pathogenic variants, variants with conflicting interpretations and those with predicted loss-of-function (pLOF). Variants were manually curated utilizing the clinical genetics database ClinVar ¹⁹. Pathogenic or likely pathogenic variants, including pLOF variants, were collectively termed PVs. The two common variants linked to moderately reduced SCAD enzymatic activity, c.625G>A and c.511C>T, are referred to here as CVs.

Variant positive individuals were defined as individuals harboring ≥2 clinically relevant ACADS variants. Except for homozygous individuals, all individuals with two variants were presumed to be in trans as phasing studies were not available.

Variant positive individuals included:

• CV/CV positive: Homozygous (n=1,610) or presumed compound heterozygous (n=333) for CVs (total n=1,943)

• CV/PV positive: Presumed compound heterozygous for a CV and a PV (n=87)

• PV/PV positive: Homozygous for a PV (n=5)

Variant negative individuals were defined as individuals not harboring any clinically relevant ACADS variants (n=15,925).

2.4. EHR phenotypes

Literature review was performed to identify a comprehensive list of SCADD-associated phenotypes relevant to adults. These included: intellectual disability, behavioral disorders with onset in childhood, epilepsy or seizure disorders, hypoglycemia, muscle weakness, metabolic acidosis, fatty liver, and a diagnosis of SCADD or disorder of fatty acid oxidation. International Classification of Diseases, ninth (ICD-9) and tenth (ICD-10) revision codes corresponding to these eight phenotypes were extracted from BioMe participants’ EHRs (Supplementary Table 1).

Manual chart review was performed for the 5 PV/PV positive individuals, as well as for 21 CV/PV positive individuals who were found to have diagnoses warranting additional investigation (convulsions, acidosis, hypoglycemia, fatty liver).

2.5. Statistical Analysis

Bivariate associations were performed between all variables, variant positive and variant negative individuals. For continuous data, the Wilcoxon rank-sum test was used due to violation of the normality assumption. Categorical variables were compared using Fisher’s exact test.

Next, multivariate logistic regression was performed to determine the association between clinically relevant ACADS variants and phenotypes of interest. PV/PV positive individuals were excluded from this analysis due to small sample size (n =5); manual chart review was conducted for these individuals instead. Two of the eight phenotypes were excluded from these analyses due to low case counts: SCADD diagnosis (zero cases identified) and behavioral disorder with onset in childhood (2 cases identified).

We examined the association between CV/CV individuals (n=1,943) and CV/PV positive individuals (n=87) and a composite of any SCADD-associated phenotype, compared to variant negative individuals (n=15,925). We then performed serial logistic regression modeling for each of the individual SCADD phenotypes among CV/CV positive individuals. CV/PV individuals were excluded from this analysis due to smaller sample size.

Biologically relevant variables (population group, age, sex, and the first five principal components of ancestry) were included in all models. We assessed for first order interactions with ACADS genotype group and self-reported population group and no meaningful interactions were identified. Statistical analysis was performed using STATA software, version 16.

3. Results

3.1. Study Population

We evaluated the prevalence and impact of clinically relevant ACADS variants in adult BioMe Biobank participants with exome sequence data (Figure 1). We estimated prevalence in 27,447 individuals, after excluding 3,347 individuals who were first- or second-degree relatives or who had missing data. Association analyses were performed in 17,960 individuals, after also removing 9,487 ACADS heterozygous carriers.

3.2. Identification of clinically relevant ACADS variants

We identified 23 clinically relevant ACADS variants in BioMe. These included 16 variants with a pathogenic or likely pathogenic assertion in ClinVar, and 5 pLOF variants (together termed PVs), and the two CVs, c.625G>A and c.511C>T, both of which had conflicting interpretations of pathogenicity in ClinVar (Supplementary Table 2).

3.3. Prevalence and characteristics of individuals harboring ACADS variants

Among 27,447 individuals, we identified 2,035 variant positive individuals. This included 1,610 (83%) individuals homozygous and 333 (17%) compound heterozygous for CVs, totaling 1,943 CV/CV positive individuals (1:20). In addition, we identified 5 PV/PV positive individuals (1:10,000) and 87 CV/PV positive individuals (1:300). All compound heterozygotes were presumed to be in trans. There were no individuals harboring three or more variants.

Median age of variant negative individuals was 64 (IQR 50–74), compared to 65 (IQR 51–77) in CV/CV, 64 (IQR 54–77) in CV/PV and 66 (IQR 60 – 68) in PV/PV positive individuals (Kruskal-Wallis, p <.01) (Table 1). Of CV/CV positive individuals, 44% self-identified as European American and 37% as Hispanic/Latino. CV/PV positive individuals were also predominantly European American and Hispanic/Latino (55 and 29%, respectively). Of the 7,536 self-reported European American individuals, 3,376 were carriers for either a CV or a PV (nearly 1 in 2), and 618 were homozygous for a CV (nearly 1 in 10).

Table 1:

Characteristics of ACADS variant negative and variant positive participants

Characteristic	Variant Negative (n=15,925)	Variant Positive*			P-value
		CV/CV (n=1,943)	CV/PV (n=87)	PV/PV (n=5)
Sex					.02
Male	6,453 (40.5)	858 (44)	36 (58)	3 (60)
Female	9,472 (59.5)	1,085 (56)	51 (42)	2 (40)
Age, years - median (IQR)	64 (50 – 74)	65 (51–77)	64 (54 – 77)	66 (60 – 68)	<.01
Population Category ‡					<.01
European American	3,253 (20)	855 (44)	48 (55)	4 (80)
African/ African American	5,250 (33)	33 (2)	8 (9)	1 (20)
East/ Southeast Asian	528 (3)	11 (1)	0 (0)	0 (0)
Hispanic/ Latin American	4,770 (30)	720 (37)	25 (29)	0 (0)
South Asian	293 (2)	60 (3)	0 (0)	0 (0)
Native American	40 (0.25)	0 (0)	0 (0)	0 (0)
Multiple selected	576 (4)	103 (5)	4 (4.6)	0 (0)
Other	1,215 (7.75)	161 (8)	2 (2)	0 (0)

Categorical variables reported as n (%) and continuous variables reported as mean ± standard deviation, unless otherwise noted.

^‡Population category is based on participant self-report.

^*Variant positive individuals include: CV/CV: Homozygous or presumed compound heterozygous for two common susceptibility variants (CVs); CV/PV: Presumed compound heterozygous for a CV and a pathogenic variant (PV), PV/PV: homozygous for a PV

3.4. Association of clinically relevant ACADS variants with SCADD phenotypes

Of the 2,035 variant positive individuals, none had a documented diagnosis of SCADD or related disorder of fatty acid oxidation. Manual chart review of the 5 PV/PV positive individuals did not show evidence of symptomatic SCADD (Table 2). We also conducted manual chart review of 21 CV/PV positive individuals with phenotypes suspicious of SCADD, including convulsions, acidosis, hypoglycemia, fatty liver, and metabolic disorder. In all cases, these phenotypes were explained by primary diagnoses unrelated to SCADD (Supplementary Table 3).

Table 2:

Clinical characteristics and evidence of disease in five individuals harboring two pathogenic variants in ACADS (PV/PV positive) based on manual chart review

Age (years)	Sex	ACADS variant (chr:pos:ref:alt)	ACADS associated phenotypes	Other notable phenotypes	Evidence of SCADD? (Yes/No)**
38	Female	12_120737094:C:T	None	Crohn’s disease, B12 deficiency, Pre-ventricular contractions	No
56	Male	12_120737094:C:T	None	Hyperlipidemia, Reactive airway disease, Tubular Adenoma	No
63	Male	12_120737094:C:T	Hypoglycemia due to diabetes mellitus type 2	Hypogonadism	No
82	Female	12_120737893:T:C	Hypoglycemia due to diabetes mellitus type 2	Hyperlipidemia	No
61	Female	12_120737893:T:C	None	Hypertrophic cardiomyopathy, obesity	No

Abbreviations: CHR: chromosome; POS: chromosomal position based on hg19 build of the human reference genome, REF: reference allele; ALT: alternate allele, PV/PV: PV/PV: homozygous for a pathogenic variant (PV)

^**Evidence of SCADD was defined by the presence of an SCADD associated phenotype (included intellectual disability, behavioral disorders with onset in childhood, epilepsy or seizure disorders, hypoglycemia, muscle weakness, metabolic acidosis, fatty liver, and a diagnosis of SCADD or disorder of fatty acid oxidation) not explained by a secondary diagnosis

In total, 17% of variant negative, 16% of CV/CV positive, and 24% of CV/PV positive individuals were found to have at least one of the ascertained SCADD phenotypes (intellectual disability, epilepsy, hypoglycemia, muscle weakness, metabolic acidosis, and fatty liver). There were no differences in the odds of each phenotype among CV/CV positive and variant negative individuals (Table 3), adjusting for age, sex, population group, and first five principal components of ancestry.

Table 3:

SCADD-associated phenotypes in individuals harboring two common variants (CV/CV) or a common variant and pathogenic variant (CV/PV) in ACADS compared to variant negative individuals.

		CV/CV (n=1,943) vs variant negative			CV/PV (n=87) vs variant negative
	Variant negative (n=15,925)	N (%)	OR (95% CI)	P Value	N (%)	OR (95% CI)	P Value
Intellectual Disability	454 (.03)	54 (.03)	1.04 (.77 – 1.4)	.77
Epilepsy	386 (.02)	25 (.01)	.71 (.47 – 1.1)	.12
Hypoglycemia	361 (.02)	42 (.02)	1.2 (0.86 – 1.7)	.27
Muscle Weakness	715 (.04)	61 (.03)	0.9 (.68 – 1.2)	.47
Metabolic Acidosis	284 (.02)	29 (.01)	1.0 (.67 – 1.5)	.97
Fatty liver	613 (.04)	98 (.05)	1.1 (.86 – 1.4)	.53
Composite Phenotype	2,422 (.15)	267 (.14)	0.99 (.861.1)	.88	17 (.20)	1.49 (.87 – 2.6)	.15

Values are n (%) unless otherwise indicated. OR and P values based on logistic regression model adjusted for age, sex, self-reported population group, and principle components 1–5.

^*CV/CV: Homozygous or compound heterozygous for Common Variants (CV)

^**CV/PV: Compound heterozygous for a CV and a pathogenic variant (PV)

In a composite model, CV/CV positive individuals did not have increased odds of having any ACADS phenotype (OR 0.99, 95% CI 0.86 – 1.1; p = .88 compared to variant negative individuals, after adjusting for age, sex, self-reported population group, and first five principal components of ancestry. Similarly, CV/PV positive individuals did not have increased odds of having any SCADD phenotype (OR 1.49, 95% CI 0.87– 2.6; p = .15) compared to variant negative individuals.

4.0. Discussion

Since its description in 1987, the clinical significance of SCADD has drawn controversy. Although clinically ascertained patients have exhibited a range of neurologic and biochemical symptoms, patients identified through newborn screening programs have been largely asymptomatic^1,2,6,8–10. Nonetheless, SCADD remains a secondary target on RUSP¹⁰. Substantial anxiety has been reported in the parents of SCADD screen positive newborns, despite the absence of disease associated symptoms²⁰.

In this study, exome sequence data in an ancestrally diverse, unrelated, and unselected clinical cohort identified ACADS variants at a higher than previously reported frequency. 1 in 10,000 individuals were homozygous for a rare pathogenic variant and an additional 1 in 20 were homozygous or presumed compound heterozygous for common variants in ACADS. This distinguishes SCADD as either the most common inborn error of metabolism or a clinically inconsequential metabolic derangement.

To clarify the clinical implication of clinically relevant biallelic ACADS variants, we utilized linked EHRs to detect phenotypes that have been attributed to SCADD in the literature. EHR-extracted diagnosis codes, with or without manual chart review, did not reveal evidence of a metabolic disease entity in ACADS variant positive individuals.

Prior studies have been limited to sample sizes of less than 115 and involved predominantly pediatric patients^1,5,21, which restricted the ability to detect rare and late-onset consequences of SCADD. Additionally, these studies focused on patients who were either ascertained based on clinical symptoms or biochemical profile (in the case of newborn screening). In contrast, the present study involved a large and ancestrally diverse adult patient cohort and applied a genome-first approach to reduce the risk of ascertainment bias. While it is plausible that the older study population introduced a sampling bias by missing severe early onset cases resulting in early demise, newborn screening data has not supported the existence of an early-lethal phenotype¹.

Despite its strengths, this study has limitations. Although EHR-linked biobanks offer tremendous opportunities for genetics and genomics research, the data in EHRs may be incomplete, inaccurate, and biased²². For example, ICD codes are not exhaustive and may be used inconsistently between providers^22–24. This study was limited to sequence variants and did not identify copy number variants in the ACADS gene; however to date large copy number variants have not been described as disease-causing for SCADD²⁵. Another important limitation was the inability to perform phasing studies to confirm compound heterozygosity. Finally, this study did not allow for biochemical profiling to supplement molecular findings. However, prior studies have shown that the common ACADS c.511C>T and c.625G>A variants can cause biochemical evidence of SCADD, including C4 elevations sufficient for a positive newborn screen¹.

5.0. Conclusions

This genome-first study evaluating clinically relevant ACADS variants in a diverse and unselected patient population revealed a high prevalence of common variants and little evidence of manifest disease among individuals considered to be variant positive. This indicates that SCADD is unlikely to represent a clinically significant disease entity in adulthood, in particular when caused by the presence of one or two common variants. Our findings support prior assertions that SCADD is a biochemical entity without clinical correlate and favors the removal of SCADD from newborn screening programs^8,26.