Novel compound heterozygous NMNAT1 variants associated with Leber congenital amaurosis

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: The gene encoding nicotinamide nucleotide adenylyltransferase 1 (NMNAT1) was recently found to be mutated in a subset of patients with Leber congenital amaurosis (LCA) with macular atrophy. The aim of this study was to determine the occurrence and frequency of NMNAT1 mutations and associated phenotypes in different types of inherited retinal dystrophies.

Methods: DNA samples of 161 patients with LCA without genetic diagnosis were analyzed for variants in NMNAT1 using Sanger sequencing. Variants in exon 5 of NMNAT1, which harbors the majority of the previously identified mutations, were screened in 532 additional patients with retinal dystrophies. This cohort encompassed 108 persons with isolated or autosomal recessive cone-rod dystrophy (CRD), 271 with isolated or autosomal recessive retinitis pigmentosa (RP), and 49 with autosomal dominant RP, as well as 104 persons with LCA in whom the causative mutation was previously identified.

Results: Compound heterozygous alterations were found in six patients with LCA and in one person with early-onset RP. All except one carried the common p.E257K variant on one allele. Macular atrophy was absent in one patient, who carried this variant in combination with a truncating mutation on the other allele. The p.E257K alteration was also found in a heterozygous state in five individuals with LCA and one with RP while no mutation was detected on the other allele. Two individuals with LCA carried other NMNAT1 variants in a heterozygous state, whereas no NMNAT1 variants in exon 5 were identified in individuals with CRD. The p.E257K variant was found to be enriched in a heterozygous state in individuals with LCA (0.94%) compared to Caucasian controls (0.18%), although the difference was statistically insignificant (p=0.12).

Conclusions: Although macular atrophy can occur in LCA and CRD, no NMNAT1 mutations were found in the latter cohort. NMNAT1 variants were also not found in a large group of patients with sporadic or autosomal recessive RP. The enrichment of p.E257K in a heterozygous state in patients with LCA versus controls suggests that this allele could act as a modifier in other genetic subtypes of LCA.

Inherited retinal dystrophies (IRDs) represent a heterogeneous group of disorders characterized by the degeneration of photoreceptor cells. Variable ages of onset and diverse grades of involvement of rod and cone photoreceptor cells divide these disorders into various clinical subsets [1]. The most severe form of retinal degeneration is Leber congenital amaurosis (LCA, OMIM #204000), with severely impaired visual function from birth and strongly reduced or undetectable electroretinogram (ERG) responses within the first year of life [2]. Another form of retinal dystrophy, involving initially rod photoreceptor degeneration which is followed by cone cell death, is retinitis pigmentosa (RP, OMIM #268000) [3]. RP is characterized by night blindness, progressive visual field deterioration, and eventual loss of central vision. In persons with cone-rod dystrophy (CRD, OMIM #120970) [4], cones are primarily involved whereas rods are concomitantly or later affected. One of the additional symptoms that may occur in LCA or CRD is macular atrophy (also called pseudocoloboma), which is not related to embryonic fissure closure [5]. Until now, mutations in 204 genes were found to be involved in the pathogenesis of IRDs (RetNet). Recently, NMNAT1, a gene involved in nicotinamide adenine dinucleotide (NAD) synthesis, was found to be responsible for a subset of LCA cases, mainly with macular lesions [6-9].

Earlier studies described three nicotinamide mononucleotide adenylyltransferases catalyzing the reaction of nicotinamide mononucleotide with ATP to form NAD⁺ [10], a crucial agent in many redox reactions especially important in the central nervous system. Whereas NMNAT2 and NMNAT3 localize to the Golgi apparatus and mitochondria, respectively, NMNAT1 is a nuclear isoform that was previously studied in the Wallerian degeneration, slow (Wlds) mouse model. In these mice, a fusion protein (Wlds) was identified, consisting of Ube4b and the complete coding sequence of Nmnat1. Subsequently, this fusion protein was found to have a protective effect on axonal degeneration after neuronal injury [11,12], and Nmnat1 activity was found to be required for this phenomenon [13-15]. Homozygous null mutations in this gene in Drosophila melanogaster and murine models result in early lethality. Heterozygous knockout mice displayed normal development [16], and a retinal knockout in the fruit fly resulted in progressive retinal degeneration [17]. NAD⁺ levels rapidly decreased in axons before their degeneration. However, retinal degeneration cannot probably be attributed to enzymatic activity of NMNAT1. An inactive Nmnat1 in Drosophila prevented photoreceptor degeneration [17]. Therefore, Perrault et al. hypothesized that this protein may have an additional chaperone function [9].

The NMNAT1 mutations reported thus far were found throughout the entire coding sequence. Most of the mutations (58%), however, are clustered in exon 5. Most variants identified in patients with LCA (31/39) are missense mutations. Only six mutations result in C-terminal truncations of NMNAT1, one extends the protein, and one abolishes the start codon. The most common mutation, p.E257K, was reported in 27 cases (in one case homozygously) and proven via an NAD/NADH assay to impede NMNAT1 activity when present in homozygous state in vivo and in a mutant construct in vitro [8]. The purpose of this study was to assess the prevalence of NMNAT1 mutations in a mixed ethnicity LCA cohort and to investigate the involvement in other progressive retinal degenerations.

Among all patients in this study, compound heterozygous mutations were found in six families. All patients carrying NMNAT1 mutations were of European origin. Six of the identified mutations, five missense and one nonsense variant, had not been previously reported (Table 1). Two other variants were described earlier: p.R207W and p.E257K. The pathogenicity assessment is presented in Table 2. Single heterozygous c.769G>A (p.E257K) alleles were present in seven probands with LCA. None of the patients with RP or CRD, as well as genetically diagnosed patients with LCA, carried NMNAT1 exon 5 mutations. The control population did not show the presence of any of these alleles, with the exception of p.E257K, which was present in a heterozygous state in 1/271 persons. In all cases containing two mutations in exon 5, segregation was confirmed with allelic cloning. All sequencing results are presented in Table 1.

All patients with compound heterozygous variants were legally blind and displayed only light or hand movements perception, depending on their age. Patient 4, the only one who did not have macular atrophy, did not retain any light perception when he was 6 years old. His fundus pictures are unavailable, since the right eye was enucleated because of corneal pain, and examination of the left eye was impossible due to corneal damage. In patient 6, macular atrophy was suspected but not possible to determine, since the degeneration was too far advanced; on funduscopy, panretinal chorioretinal atrophy was visible. In sibship 3, the boy (3.1) had distinct macular atrophy at the age of 6 months, while his sister (3.2) developed macular atrophy within the first two years of life. In all patients in whom electroretinography was performed, the signals were not detectable. Abnormal peripheral fundus pigmentation was visible in six patients. Patient 1 had subcapsular cataracts, whereas patient 6 already underwent cataract extraction. The results of all clinical examinations are presented in Table 3.

Due to high energy consumption, the retina may be sensitive to deficiencies in substrates associated with cellular energy supply. NAD⁺, the product of a reaction catalyzed by NMNAT1, is such a substrate. NAD⁺ is also indirectly responsible for posttranslational covalent modifications of key proteins (such as mono- and poly-ADP ribosylation) [19]. Therefore, it is not surprising that NMNAT1 mutations cause a severe form of hereditary retinal degeneration (Figure 1). In this study, none of the patients with RP or CRD carried NMNAT1 exon 5 mutations. A separate question is the origin of the macular atrophy. Atrophic macular lesions may arise secondary to photoreceptor death, as it has been described in CRD [4], but also in blue cone monochromacy [20] and achromatopsia [21]. In LCA, the macular lesions may evolve in time: patient 3.2 showed abnormal central pigmentations at the age of 4 months, which developed into macular atrophy at the age of 2 years. In our study, one of the seven patients (patient 4) in whom we identified NMNAT1 mutations did not exactly display this feature, but parafoveal atrophy. The genetic data gathered from our study, as well as other research, suggest that patients with LCA with NMNAT1 variants have significant residual NMNAT1 activity. Most of the variants represent combinations of nonsense mutations or missense mutations predicted by in silico prediction programs to be severe, with other alterations that are presumed to exert a milder effect on the protein. Only one patient with a homozygous null mutation (p.W169*) has been reported [9]. The lethality of Nmnat1 knockout mice suggests that human embryos with such a genotype would not be viable [16]. It is not clear how the p.W169* variant would retain any NMNAT1 activity; however, it was suggested that this phenomenon can be ascribed to NMNAT1 chaperone function, which is predicted to remain intact [9].

The p.E257K mutation is of particular interest. Deemed to be neutral by prediction programs SIFT and PolyPhen2, this missense alteration also displays a low level of nucleotide conservation, and the physicochemical difference between these amino acids, measured with the Grantham score is only marginal. Nevertheless, a previous study showed that this mutation not only disrupts NMNAT1 activity but also has the most profound impact of all five mutations identified in that study [8]. The mechanism of this negative effect is yet unknown. Perrault and colleagues speculated that the p.E257K variant is in linkage disequilibrium with a yet undiscovered regulatory variant in the same gene, causing mRNA to be expressed at a low level. However, the experiments in question were performed by expressing recombinant NMNAT1 protein in cell lines, using constructs that did not contain regulatory parts of the gene. Another peculiar finding is the enrichment of heterozygotes for this mutation in patients with LCA compared to control groups, which was previously found to be present in 4.4% of the patient alleles versus none in 400 control alleles in one study [9], which was a statistically significant difference. Nonetheless, the considered group consisted only of prescreened, unsolved LCA probands. In this study, we also screened our LCA cohort with patients in whom causal mutations had been previously identified. In this cohort, the p.E257K variant was present heterozygously in five LCA proband alleles (0.94%; Table 2) versus only one in 271 Caucasian control alleles (0.18%; Fisher’s exact test, p=0.12). Moreover, when the LCA cohort was compared to European American frequency in Exome Variant Server data, a statistically significant difference was observed between the patients with LCA and the controls (0.13%, Fisher’s exact test, p=0.002). It can be inferred that this variant is specific for the Western European population; however, the enrichment in patients with LCA may indicate that this variant acts as a modifier in one or several other genetic subtypes of LCA. Whole exome sequencing of these cases may further shed light on a potential modifier effect of this NMNAT1 allele on other genetic subtypes of LCA.

In addition to the p.E257K change, one other mutation, p.R207W, was detected in multiple patients. The combination of p.R207W and p.E257K variants was by far (8/311 prescreened cases from our and two other studies) the most prevalent among NMNAT1-mutated patients with LCA. Eight individuals of French origin were found to carry this combination of variants [8,9], which may indicate that p.R207W is a founder mutation in this population.

Two additional, potentially harmful mutations were identified only heterozygously in patients with LCA: c.12dup (p.E5Rfs*4) and a putative splice site mutation c.115+3A>G (p.?). Since they were identified in cases with LCA and RP, these mutations may be present in a compound heterozygous state with a missing “second hit” on the other allele; for example, an unidentified alteration either deep within an intron or in a regulatory sequence. The c.115+3A>G alteration, although predicted to diminish the strength of the splice site, may however be a benign alteration, since it is rather frequent in the Caucasian population (Table 2). All other mutations were predicted to have a deleterious effect. Functional studies are required to evaluate the potential effect of these alterations on the protein. The indirect assessment of NMNAT1 activity by NAD measurement in patients’ erythrocytes and lymphoblast cells was previously used for evaluating the pathogenicity of a given mutation. Since other tissues appear to be unaffected by NMNAT1 mutations, it cannot be certain whether the results of these measurement reflect the in vivo situation in the retina.

In conclusion, there is a strong association between NMNAT1 mutations and macular atrophy, although this feature is not always present. NMNAT1 seems to be involved in a small subset of LCA cases, since it is responsible for 2.3% of the cases in our cohort of 265 patients with LCA. Although the difference was not significant, the p.E257K mutation was enriched in a heterozygous state in individuals with LCA compared to ethnically matched controls, suggesting that this allele potentially acts as a modifier in other genetic subtypes of LCA.