ABHD16A deficiency causes a complicated form of hereditary spastic paraplegia associated with intellectual disability and cerebral anomalies

Summary

ABHD16A (abhydrolase domain-containing protein 16A, phospholipase) encodes the major phosphatidylserine (PS) lipase in the brain. PS lipase synthesizes lysophosphatidylserine, an important signaling lipid that functions in the mammalian central nervous system. ABHD16A has not yet been associated with a human disease. In this report, we present a cohort of 11 affected individuals from six unrelated families with a complicated form of hereditary spastic paraplegia (HSP) who carry bi-allelic deleterious variants in ABHD16A. Affected individuals present with a similar phenotype consisting of global developmental delay/intellectual disability, progressive spasticity affecting the upper and lower limbs, and corpus callosum and white matter anomalies. Immunoblot analysis on extracts from fibroblasts from four affected individuals demonstrated little to no ABHD16A protein levels compared to controls. Our findings add ABHD16A to the growing list of lipid genes in which dysregulation can cause complicated forms of HSP and begin to describe the molecular etiology of this condition.

Main text

Hereditary spastic paraplegias (HSPs) comprise a broad and diverse group of inherited neurodevelopmental or neurodegenerative disorders in which the main neurological symptom is progressive lower limb spasticity and weakness. HSPs are genetically heterogeneous with approximately 80 suspected genetic subtypes and more than 60 genes identified to date; autosomal-dominant, autosomal-recessive, and X-linked modes of inheritance have all been described.^1,2 HSPs were historically clinically classified as isolated or complicated; complicated HSP presents with additional neurological or extra-neurological signs such as intellectual disability, cerebellar dysfunction, neuropathy, or ophthalmological anomalies.³ Advances in genomic technologies have revealed significant clinical and genetic overlap within different HSP subtypes and between HSP and other neurodegenerative diseases, such as Charcot-Marie-Tooth neuropathy.^4,5

Many HSP-associated genes have been found to alter lipid metabolism and homeostasis, highlighting the importance of these processes in neuronal physiology. Pathogenic variants in genes encoding various enzymes involved in the metabolism and the inter-organellar trafficking of different lipid classes, including cholesterols, phospholipids, sphingolipids, and fatty acids, have been associated with HSP (reviewed in Darios et al.⁶). Alterations of lipid metabolism are postulated to be contributing factors that lead to the degeneration of axons and, thus, are involved in the underlying pathophysiology of HSP.

In this study, we present a cohort of individuals affected with a progressive neurological condition who all harbor bi-allelic variants in abhydrolase domain-containing protein 16A, phospholipase (ABHD16A [MIM: 142620]). ABHD16A encodes the major phosphatidylserine (PS) lipase in the brain, which synthesizes the signaling lipid lysophosphatidylserine (lyso-PS).⁷ To our knowledge, ABHD16A has not previously been linked to any genetic disorder.

ABHD16A was initially identified as a compelling novel candidate gene in affected sibling pairs from two French-Canadian families (individuals P1 and P2 from family 1 and individuals P3 and P4 from family 2, Figure 1A). The affected individuals presented with intellectual disability and progressive spasticity of the upper and lower limbs (Tables 1 and S1). Cerebral MRIs revealed a thin corpus callosum and white matter anomalies (Figure 2). Three of these individuals also presented with an axonal neuropathy, and one had myoclonic seizures. Although the two families were not known to be related to each other, their exome sequencing data were investigated together because of their ethnicity and similar clinical presentation. Exome sequencing of the two affected sibling pairs, performed as part of the Care4Rare Canada research program,⁸ identified a homozygous c.353G>A (p.Arg118His) variant in ABHD16A (GenBank: NM_021160.2) in all four affected individuals (P1, P2, P3, and P4). Interestingly, this variant was not called by an older version of the Care4Rare bioinformatics pipeline. The c.353G>A variant occurs at a highly conserved site (PhyloP score of 4.56). This variant is rare in gnomAD v.2.1.1, having only two alleles identified (2/246,600 alleles) and an allele frequency of 0.0008%.⁹ In silico prediction programs, including CADD, MutationTaster, SIFT, and PolyPhen-2, all predict this missense change to have a deleterious effect on ABHD16A. Sanger sequencing confirmed that each parent in families 1 and 2 was heterozygous for the c.353G>A variant and that the variant was present in the homozygous state in the four affected individuals. Also, the two unaffected children in family 1 were shown to not be homozygous for the c.353G>A variant. ABHD16A is the major phospholipase in the brain that converts PS to lyso-PS (Figure 1B).⁷ Lyso-PS is present in all tissues but is most abundant in the central nervous system and immune cells where it is postulated to regulate numerous immunological and neurological processes.¹⁰ For instance, lyso-PS is involved in the regulation of macrophage activation, clearance of apoptotic cells, human cancer cell migration, and glucose metabolism in the brain and skeletal muscles.11, 12, 13, 14, 15 Given the evidence above, the homozygous variant in ABHD16A was deemed to be a strong candidate for causing the HSP phenotype in the two sibling pairs.

Figure 1.

Pedigree and summary of ABHD16A variants

(A) Pedigree of the six families with variants in ABHD16A, showing consanguinity in three families. Black symbols represent affected individuals. Symbols containing a dot represent a heterozygous carrier status confirmed in an adult. Unaffected siblings were all confirmed to not be homozygous for the variant present in the family.

(B) Schematic of phosphatidylserine (PS) lipase and lysophosphatidylserine (lyso-PS) lipase activities by ABHD16A and ABHD12.

(C) Representation of the six variants in ABHD16A (GenBank: NM_021160.2) present in the 11 affected individuals. This transcript encodes a protein that is 558 amino acids in length.

Table 1.

Clinical features of 11 individuals from six families with bi-allelic variants in ABHD16A

Family	1		2		3		4	5		6
Variant in ABHD16Aa	c.353G>A		c.353G>A		c.1226T>G		c.1333C>T; c.573delG	c.362A>T		c.755G>A
Protein change	p.Arg118His		p.Arg118His		p.Leu409Arg		p.Arg445∗; p.Thr192Glnfs∗25	p.Asn121Ile		p.Arg252Gln
Zygosity	homozygous		homozygous		homozygous		compound heterozygous	homozygous		homozygous
Ethnicity	French-Canadian		French-Canadian		Armenian		Mixed European	Egyptian		Pakistani
Consanguinity	−		−		+		−	+		+
Individual	1	2	3	4	5	6	7	8	9	10	11
Sex (M/F)b	M	F	F	F	F	M	F	F	F	F	M
Age (years)c	12	11	14	16	21	10	5	21	12	16	12
ID/GDDd	+	+	+	+	+	+	+	+	+	+	+
Upper limb spasticity	+	+	+	+	+	+	−	−	−	−	+
Joint contractures	+	+	+	+	+	+	−	−	−	+	+
Neuropathy	−	+	+	+	Ue	Ue	Ue	Ue	Ue	Ue	Ue
Seizures	+	−	−	−	−	+	−	−	−	−	−
Thin corpus callosum	+	+	+	+	+	+	+	+	+	−	−
White matter anomalies	+	+	+	+	+	+	+	+	−	+	+
Behavioral anomaliesf	−	−	−	−	−	+	+	−	−	−	−

^aGenBank: NM_021160.2.

^bM, male; F, female.

^cAge at time of publication.

^dID, intellectual disability; GDD, global developmental delay.

^eU, unknown (EMG/NCS was not performed in these individuals).

^fAutism spectrum disorder, self-injurious behaviors.

Figure 2.

Brain MRI characteristics of individuals P1–P7

The MRIs were performed at the following ages: P1, 3 years 3 months; P2, 2 years 3 months; P3, 12 years 3 months; P4, 9 years 4 months; P5, 19 years 3 months; P6, 3 years 9 months; and P7, 26 months. From left to right, sagittal midline, axial T1 (except P6, coronal), axial T2 (except P7, coronal) at the level of basal ganglia, axial T2 (except P7, FLAIR) at the level of the lateral ventricles, and axial FLAIR at the level of the basal ganglia. Affected individuals with bi-allelic variants in ABHD16A have a thin corpus callosum (first column), with or without cavum vergae and/or cavum septum pellucidum (columns 2, 3, and 5), and non-specific periventricular and/or deep white matter abnormalities (columns 3, 4, and 5).

We next sought to identify additional affected individuals by using the internal Care4Rare database, Matchmaker Exchange (MME),¹⁶ and Geno2MP. Within the internal Care4Rare database, we identified potentially deleterious variants in ABHD16A in three additional individuals (P5, P6, and P7) from two unrelated families with a similar HSP phenotype (Table 1, Figure 1A, and Table S1). The first two affected individuals (P5 and P6, Figure 1A) were siblings from a consanguineous Armenian family (family 3). Similar to the affected individuals in families 1 and 2, individuals P5 and P6 presented with intellectual disability, progressive spasticity of upper and lower limbs, white matter anomalies, and a thin corpus callosum (Figure 2). SNP array revealed several long contiguous stretches of homozygosity greater than 5 Mb in size, estimated to encompass approximately 5.7% of the autosome genome, which is consistent with the parents’ being reported as first cousins. Exome sequencing identified a homozygous c.1226T>G (p.Leu409Arg) missense variant in ABHD16A in both individuals P5 and P6. Similar to the exome data for families 1 and 2, the variants in ABHD16A were only called in the updated version of the Care4Rare bioinformatics pipeline. This c.1226T>G variant is absent from gnomAD⁹ and occurs at a highly conserved site (PhyloP score of 4.16), and in silico prediction software predict this missense change to have a deleterious impact on ABHD16A. Sanger sequencing confirmed that each parent was heterozygous for the c.1226T>G variant and that the variant was present in the homozygous state in the two affected siblings. The final individual identified in the Care4Rare database was individual P7 from family 4 (Figure 1A), who was compound heterozygous for a c.1333C>T (p.Arg445^∗) variant and a c.573del (p.Thr192Glnfs^∗25) variant in ABHD16A. This patient presented with hypotonia, severe global developmental delay, and progressive lower limb spasticity (Tables 1 and S1). Her cerebral MRI revealed a thin corpus callosum and white matter anomalies (Figure 2). The paternally inherited c.1333C>T variant was a nonsense variant predicted to result in a premature stop codon at amino acid position 445 located in exon 16. The c.1333C>T variant is rare in gnomAD v.2.1.1, having only one allele identified (1/247,968 alleles and an allele frequency of 0.0004%).⁹ The c.573del variant was maternally inherited and absent from gnomAD.⁹ The deletion of the single G nucleotide is predicted to cause a frameshift at amino acid position 192 located in exon 7 and a premature stop codon at amino acid position 217.

The Geno2MP database was interrogated for ABHD16A variants linked to individuals with the Human Phenotype Ontology (HPO) term “abnormality of the nervous system” (Geno2MP accessed in April 2020). Through this one-sided matchmaking strategy, two additional sets of siblings (individuals P8 and P9 from family 5 and individuals P10 and P11 from family 6) were identified as having potentially deleterious variants in ABHD16A. Collaborative discussions revealed that the affected individuals from both families had a remarkably similar clinical presentation to the aforementioned individuals (Tables 1 and S1). Individuals P8 and P9 (family 5, Figure 1A). were affected sisters from a consanguineous Egyptian family. A homozygous c.362A>T (p.Asn121Ile) missense variant in ABHD16A was identified by exome sequencing as a potential novel candidate gene and submitted to the Geno2MP database. The c.362A>T variant is absent from gnomAD⁹ and occurs at a moderately conserved site (PhyloP score of 3.76), and in silico programs predict this missense change to have a deleterious effect on protein function. Family segregation studies confirmed that individuals P8 and P9 were homozygous for the c.362A>T variant and indicated that each parent was heterozygous for the c.362A>T variant and that the unaffected sibling did not carry the variant in the homozygous state. Finally, individuals P10 and P11 (family 6, Figure 1A) were affected siblings from a consanguineous Pakistani family. A homozygous c.755G>A (p.Arg252Gln) missense variant was identified by exome sequencing in the two affected individuals as a potential novel candidate gene and submitted to the Geno2MP database. The c.755G>A variant is absent from gnomAD⁹ and occurs at a highly conserved site (PhyloP score of 5.61), and in silico programs predict this missense change to have a deleterious effect on protein function. Follow-up studies by Sanger sequencing confirmed that individuals P10 and P11 were homozygous for the c.755G>A variant, their mother was heterozygous for the variant, and the unaffected sibling did not carry the variant in the homozygous state.

In summary, exome sequencing combined with a one-sided matchmaking strategy resulted in the assembly of a cohort of six unrelated families with 11 affected individuals having bi-allelic variants in ABHD16A and presenting with a complicated form of HSP. No other variants in disease-associated or novel genes were retained as plausible candidates by exome analysis for any of the affected individuals. The phenotypic overlap between the 11 affected individuals is striking because they all present with variable degrees of developmental delay/intellectual disability and progressive spasticity affecting predominantly the lower limbs. The mean age of onset of the lower limb spasticity in individuals from this cohort was 3 years old, and the spasticity progressively became worse over time (Table S1). In addition, almost all individuals presented with corpus callosum and white matter anomalies. The identified variants in ABHD16A were very compelling given their deleterious in silico predictions, their rarity in population data, and the involvement of ABHD16A in lipid metabolism, a key process in neuronal physiology.

Next, we performed functional studies to investigate for additional evidence of causality. We began by evaluating the impact of the ABHD16A variants on overall ABHD16A expression in primary fibroblast cells from the affected individuals P3, P4, P7, and P9. Real-time PCR data revealed cells from P3 and P9 did not have significantly different ABHD16A mRNA transcript abundance compared to control fibroblasts, while fibroblast cells from P4 had a significant increase compared to controls but was not significantly different from P3 and P9 (Figure 3A). On the other hand, P7 had a significant decrease in ABHD16A mRNA transcript abundance compared to all other samples, which was not unexpected because this individual is compound heterozygous for a nonsense and a frameshift variant; all other affected individuals harbor missense variants (Figure 3A and Table 1). Immunoblot analysis of whole fibroblast cell extracts demonstrated little to no ABHD16A in all tested affected individuals compared to controls (Figure 3B). The ABHD16A antibody used for this study targets the C-terminal end of ABHD16A and is thus predicted to not detect the p.Thr192Glnfs^∗21 mutant protein. Also, because it is not known whether the epitope is present before or after amino acid position 445, the ability of this antibody to detect the p.Arg445^∗ mutant protein is not known. Nevertheless, on the basis of the real-time PCR results, which showed significant reduction of ABHD16A mRNA in P7 fibroblast cells, the mutant protein levels in this affected individual would be predicted to also be significantly reduced. Finally, we investigated the effect of drastically reduced ABHD16A protein levels on cellular proliferation. We determined that fibroblasts from the affected individuals have a broad range of proliferation compared to controls (Figure 3C).

Figure 3.

Bi-allelic ABHD16A variants result in decreased ABHD16A expression in fibroblast cells from affected individuals

(A) Real-time PCR analysis on cDNA from affected individual fibroblast cells showing ABHD16A transcript levels in affected cells compared to controls. Letters above error bars denote statistical significance (p < 0.05) as determined by one-way analysis of variance (ANOVA) and Tukey’s multiple-comparison test; samples with the same letter are not significantly different.

(B) Immunoblot analysis on extracts from affected individual fibroblast cells showing little to no ABHD16A expression in affected cells compared to controls.

(C) Cell proliferation of control and affected fibroblast cells was measured via assessment of cell confluency on each of the specified days after initial plating. For each of the panels in this figure, five control cell lines were used and the mean data were graphed; all error bars depict standard error of the mean.

Given the role of ABHD16A in synthesizing lyso-PS, the effects of the ABHD16A variants on lipid metabolism were investigated via cultured patient fibroblasts and plasma samples. Lipidomic studies in fibroblasts unexpectedly did not reveal a clear change in lyso-PS levels in affected individual fibroblasts (P3, P4, P7, and P9) relative to control fibroblasts; PS levels were elevated, but lyso-PS levels were not decreased (Figure S1). A similar observation was made in plasma samples collected from patients (P1, P2, P3, and P4) when compared to unaffected heterozygous family members (parents and unaffected sibling; data not shown). Other lipidomic changes in fibroblasts were observed, suggesting that ABHD16A may be involved in regulating the lipidome through other pathways (Figure S1). To date, the characterization of ABHD16A has largely focused on confirming its enzymatic function in converting intracellular PS to lyso-PS in the brain and spinal cord.⁷ Therefore, although ABHD16A PS lipase activity has been detected in vitro in some non-neuronal human cell lines, mainly lymphoblast and cancer cell lines, its activity in most non-neuronal cells, including fibroblasts, is currently unknown.⁷ The presence of other PS lipases in fibroblasts could also be a reason that no disturbances were found in lyso-PS levels. Moreover, the biological function of lyso-PS synthesized by ABHD16A in non-neuronal cells is currently unknown. Of note, lipidomic studies on brain tissue from a deceased individual with Sjögren-Larsson syndrome, another neurometabolic disorder, have been shown to give invaluable insight into the lipid profile of an affected tissue and to differ significantly from the profile seen in plasma and fibroblasts.¹⁷ The lipidomic analyses performed in the current study were limited to skin and plasma, raising the possibility that lipidomic profiling of central nervous system tissues may give very different results.

As shown in Figure 1B, ABHD12 encodes a lipase that degrades lyso-PS and, thus, functions downstream of ABHD16A. Bi-allelic pathogenic variants in ABHD12 are associated with the neurological disorder PHARC (polyneuropathy, hearing loss, ataxia, retinitis pigmentosa, and cataract [MIM: 612674]).^18,19 In PHARC, dysfunction of the ABHD12 lyso-PS lipase leads to an accumulation of lyso-PS in the brain, which is postulated to result in microglial activation and motor defects.¹⁹ Surprisingly, Abhd16a-null mice do not appear to have postnatal central nervous system anomalies despite having reduced levels of lyso-PS.⁷ Because the Abhd16a-null mice were assessed until 10 weeks after birth, the development of central nervous system anomalies after this time point remains a possibility. However, there appears to be an increased incidence of prenatal lethality, and the Abhd16a-null mice that were born were 30% smaller in size compared to their wild-type littermates.⁷ More research is required to elucidate the different roles that lyso-PS plays in different species and tissues as well as during different stages of development. In addition, differences in several lipid species were observed between affected individual and control fibroblasts, including phosphatidylcholine (PC), akylphosphatidylethanolamine (PC[O]), and alkenylphosphatidylcholine (PC[P]), and to a lesser extent their ethanolamine counterparts (Figure S1), raising the possibility that ABHD16A is involved in additional routes of lipid metabolism. Nevertheless, the association of ABHD16A and ABHD12 with distinct neurological disorders in humans highlights the importance of this pathway in maintaining the normal function of the post-natal human central nervous system.

ABHD16A is located on chromosome 6p21.33, and its most abundant transcript (GenBank:NM_021160.2) consists of 20 exons that encode a protein that is 558 amino acids in length. ABHD16A belongs to the ABHD superfamily of proteins that share an α/β-hydrolase fold.²⁰ Collectively, four missense variants, one nonsense variant, and one frameshift variant were identified through exome sequencing in affected individuals from this study. Interestingly, the six variants are located throughout the gene and do not cluster to any particular region (Figure 1C). In fact, none of the four missense variants localize to the α/β-hydrolase domain or to any of the predicted conserved nucleophilic or acetyltransferase motifs.^21,22 Although the variants identified in this study occur throughout ABHD16A, the c.353G>A variant was identified in the two unrelated sibling pairs of French-Canadian descent, raising the possibility that this particular variant is a founder mutation in the French-Canadian population. A shared haplotype on chromosome 6 ((chr6: 29,709,714–32,522,379(hg19)) with the variant in ABHD16A located at chr6: 31,664,801(hg19)) was identified in the four affected individuals and the two sets of parents from families 1 and 2, supporting that this variant was inherited from a common ancestor.

Interestingly, the ABHD16A variants in families 1, 2, and 3 were only called after the exome data were rerun with an updated version of the Care4Rare bioinformatics pipeline (version 2.0) and had been missed during the first exome analyses performed prior to 2017. Thus far, we have been unsuccessful in determining why they were previously missed. We know that the variants were not included in the original variant calling files, but we did not identify any overt issues with either our alignment or our calling that would explain why this was the case. ABHD16A has no known pseudogenes that might increase the risk of mapping errors. This gene currently has ample coverage in public databases such as gnomAD, and our updated pipeline, as well as the Broad Institute pipeline (which sequenced families 5 and 6 in this cohort), had no issues identifying variants in ABHD16A in their most recent iteration. In addition, on the basis of our 5 years of experience with MME, we were very surprised that no matches were made via this platform, which again suggests variants in this gene might have been missed in the past by other pipelines. ABHD16A-associated complicated HSP might thus be a more frequently occurring condition than is currently acknowledged. We recommend that molecular geneticists confirm the inclusion of ABHD16A in their variant calling files, particularly when investigating individuals with complex HSP.

In conclusion, our findings support that bi-allelic deleterious variants in ABHD16A cause an autosomal-recessive subtype of complicated HSP. ABHD16A PS lipase deficiency results in intellectual disability, progressive spasticity, abnormal corpus callosum, and white matter anomalies, most likely through the dysregulation of lipid species including lyso-PS. The identification of ABHD16A as an HSP gene again demonstrates the crucial role that lipid metabolism and homeostasis have in maintaining the nervous system. Importantly, because this disorder is an inborn error of lipid metabolism and has a progressive course, research is needed to investigate whether affected individuals could benefit from pharmacological, or other, intervention. Finally, this multi-center study highlights the importance of sharing genomic data for the identification of novel genes in rare disorders such as HSP.

Published: September 28, 2021

Data and code availability

The ABHD16A variants were submitted to ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) (GenBank: NM_021160.2; accession numbers SCV001787122, SCV001787123, SCV001787124, SCV001787125, SCV001787126, and SCV001787127). The exome datasets supporting this study have not been deposited in a public repository because of ethical restriction but are available from the corresponding author on request.

Web resources

CADD, https://cadd.gs.washington.edu/

Geno2MP, https://geno2mp.gs.washington.edu/

gnomAD, https://gnomad.broadinstitute.org/

Mutation Taster, http://www.mutationtaster.org/

OMIM, https://omim.org

PolyPhen2, http://genetics.bwh.harvard.edu/pph2/

SIFT, https://sift.bii.a-star.edu.sg/