Broadening the phenotypic and molecular spectrum of FINCA syndrome: Biallelic NHLRC2 variants in 15 novel individuals

FINCA syndrome [MIM: 618278] is an autosomal recessive multisystem disorder characterized by fibrosis, neurodegeneration and cerebral angiomatosis. To date, 13 patients from nine families with biallelic NHLRC2 variants have been published. In all of them, the recurrent missense variant p.(Asp148Tyr) was detected on at least one allele. Common manifestations included lung or muscle fibrosis, respiratory distress, developmental delay, neuromuscular symptoms and seizures often followed by early death due to rapid disease progression. Here, we present 15 individuals from 12 families with an overlapping phenotype associated with nine novel NHLRC2 variants identified by exome analysis. All patients described here presented with moderate to severe global developmental delay and variable disease progression. Seizures, truncal hypotonia and movement disorders were frequently observed. Notably, we also present the first eight cases in which the recurrent p.(Asp148Tyr) variant was not detected in either homozygous or compound heterozygous state. We cloned and expressed all novel and most previously published non-truncating variants in HEK293-cells. From the results of these functional studies, we propose a potential genotype-phenotype correlation, with a greater reduction in protein expression being associated with a more severe phenotype. Taken together, our findings broaden the known phenotypic and molecular spectrum and emphasize that NHLRC2-related disease should be considered in patients presenting with intellectual disability, movement disorders, neuroregression and epilepsy with or without pulmonary involvement.


INTRODUCTION
The NHL repeat containing 2 (NHLRC2) gene (HGNC: 24731) encodes a ubiquitously expressed and well conserved protein consisting of an N-terminal thioredoxin domain (TRX) and a large NHL repeat domain. The exact function remains elusive and only few functional studies have been conducted [1]. Nishi and colleagues showed that NHLRC2 plays a role in reactive oxygen species (ROS)-induced apoptosis and that the loss of NHLRC2 results in increased sensitivity to ROS-induced cell death [2]. Using genome-wide CRISPR screening, Haney and colleagues identified NHLRC2 as a regulator of phagocytosis involved in actin polymerization and filopodia formation [3]. Recently, an NHLRC2 knockout mouse model revealed embryonic lethality due to gastrulation failure [4].
In 2018, Uusimaa and colleagues identified the same compound heterozygous variants in NHLRC2 in three individuals from two nonconsanguineous families of Finnish descent as the cause of a cerebral-pulmonary disorder [MIM: 618278] [5]. They named this novel syndrome FINCA disease based on manifestations observed at tissue level: fibrosis, neurodegeneration and cerebral angiomatosis. All patients developed feeding difficulties and growth retardation within the first two months of life. The disease progressed rapidly and all died before the age of two, probably due to severe respiratory distress. Brain magnetic resonance imaging (MRI) scans and post-mortem histopathology revealed brain atrophy, vacuolar white matter degeneration and interstitial lung fibrosis. While further studies in patients carrying biallelic NHLRC2 variants confirmed this severe multisystem phenotype [6], cases with predominantly neurological involvement and survival into the second decade of life have also been reported [7,8]. In all 13 individuals described so far, the recurrent missense variant c.442G>T, p.(Asp148Tyr) was present either in homozygous or compound heterozygous state. Respiratory defects and altered lung morphology were observed in nine individuals, highlighting pulmonary findings as a common feature of NHLRC2-related disease.
Here, we present nine novel NHLRC2 variants detected in 15 additional individuals with various neurological symptoms, thereby expanding the allelic series and broadening the phenotypic spectrum of NHLRC2-related disease.

MATERIALS AND METHODS
This study adheres to the principles set out in the Declaration of Helsinki and was approved by institutional Ethics Committees of Charité -Universitätsmedizin (EA2/177/18). We recruited the affected individuals 1 and 2 via the TRANSLATE-NAMSE project, an Undiagnosed Disease Program at Charité -Universitätsmedizin Berlin [9]. Probands 3-15 were identified through matches within GeneMatcher and across the Matchmaker Exchange, and by checking the ClinVar database for recently submitted variants [10][11][12]. All NHLRC2 variants were detected by exome sequencing (ES) and classified according to the ACMG guidelines [13] and refer to MANE transcript (NM_198514.4). Informed consent was obtained for each participant.

Lymphoblastoid cell lines (LCLs)
B cells were isolated from patients' heparin whole blood samples and immortalized by Epstein-Barr virus (EBV) transfection as previously described [14]. Established LCLs were cultured at 37°C and 5% CO 2 in RPMI-1640 with L-Glutamine supplemented with 10% fetal calf serum (FCS) and Penicillin/Streptomycin (all purchased from: Gibco™, Thermo Fisher Scientific, Waltham, Massachusetts, USA).

RT-qPCR and cDNA sequencing of LCLs, Western blots
A detailed description of RT-qPCR and cDNA sequencing of LCLs and Western blots can be found in Supplementary File 1.

Generation of expression constructs for NHLRC2 non truncating variants
Sequences of wildtype and mutant NHLRC2 with an added C-terminal Flagtag and XhoI and XbaI restriction sites were amplified from cDNA of LCLs.
Amplified fragments were either cloned directly into pcDNA.3 via XhoI and XbaI or into pJETeasy and subsequently subcloned from pJETeasy into pcDNA.3. Final pcDNA.3 constructs were checked for correct integration of the respective NHLRC2 variant by Sanger sequencing.
For generation of non-truncating variants where no prior LCL had been available, the respective mutation was introduced via primers and a 5' and 3' partial fragment carrying the respective mutation in a 30 bp common overlap were amplified from wildtype NHLRC2 cDNA. Fragments were assembled via overlap extension PCR and either directly cloned into pcDNA.3 or first assembled into pBluescript(-) via overlap extension PCR and subsequently subcloned into pcDNA.3. A complete list of the generated variants, fragments and used primers can be found in Supplementary Table 1.

In silico protein modelling
Computational predictions of the protein structure were generated for wildtype NHLRC2 and all variants with an expected altered amino-acid sequence. Predictions were computed using a simplified AlphaFold model (v2.2.4) using Colaboratory by Google. The computations were performed with default configuration parameters and a randomized seed value. Structural dissimilarity between mutant and wildtype structure was measured by the predicted local-distance difference test (pLDDT), a score of local structural configuration differences predicted from the model loss [15]. Protein-protein interaction (PPI) site predictions were computed for each protein structure using MaSIF (commit: 2a37051) and visualized with pymol [16].

Homozygosity mapping
Homozygosity mapping was performed for families A and F using the AutozygosityMapper [17] as described previously [18].

Clinical spectrum
The clinical findings of the 15 individuals are summarized in Table 1. Comparison of the clinical presentation of this cohort with previously reported cases is summarized in Table 2. Pedigrees are shown in Fig. 1a and detailed clinical reports are provided in Supplementary File 2 and Supplementary Figure 1.
Our cohort of new cases comprised 15 individuals (ten females and five males) from 12 unrelated families. Consanguinity was reported in seven of them. All documented birth measurements were within the normal range. Postnatal adaptation and development in the first weeks of life were unremarkable in 14 of the 15 children. Only one child (individual 5) developed postnatal complications in the form of respiratory distress and pulmonary hypertension. In the following months, recurrent upper respiratory tract infections, anemia and hepatomegaly were noted, and she died prematurely of viral pneumonia at two years of age. All other individuals were aged between 30 months and 19 years at the last follow-up. Thirteen of the 15 individuals had no respiratory or pulmonary symptoms. At the last clinical examination, four      individuals (ages five, six, eight and 19) had microcephaly (SD below −2). Short stature and decreased body weight were observed in seven and nine individuals, respectively, with five individuals having both. All 15 individuals showed global developmental delay and later intellectual disability (ID), which in most cases was classified as moderate or severe. Nine children were non-verbal at the last assessment, and five could speak only a few rudimentary words. In four cases, speech regression occurred after 1-3 years of age, and the siblings (individual 1 and 2) had motor regression starting from the age of six. Behavioral abnormalities (hyperactivity, impulsive, aggressive, anxious and autistic behavior) were reported in seven individuals. All 15 individuals had variable neuromuscular involvement: 12 had truncal hypotonia, five individuals were never able to walk and four developed hyperkinetic movement disorder, spasticity or hyperreflexia. Two depended on a wheelchair for long distances and five had an unsteady, wide gait. Seizures occurred in 11 individuals, with age of onset ranging from nine months to 13 years. In individual 5, epileptiform discharges were documented but no clinical seizure was observed. In six individuals, multiple antiepileptic drugs were tried due to intractable seizures and two children were fitted with a vagus nerve stimulation (VNS) device. Brain abnormalities (dilated ventricles, corpus callosum hypoplasia, mild cortical atrophy and delayed myelination) were diagnosed in five individuals, while six had unremarkable brain MRI scans. Other less common findings included strabismus (n = 3), diarrhea and/or malabsorption (n = 3), recurrent infections (n = 2), anemia (n = 2), congenital heart defect (n = 2) and unilateral renal reflux (n = 1).

Molecular findings
In addition to the recurrent missense variant p.(Asp148Tyr), we identified nine novel variants, including one start-loss variant, three missense variants, two single amino acid in-frame deletions, two nonsense and one C-terminal frameshift variant. The localization of the variants, their level of conservation and the pathogenicity prediction using different in silico tools are summarized in Fig. 1b- Supplementary Fig. 2. While RT-qPCR showed no clear difference in NHLRC2 expression levels between patient and control LCLs (Fig. 2a), Western blot showed a clear reduction in NHLRC2 protein levels in all patient samples tested compared to control samples (Fig. 2c, d).
Consistent with the RT-qPCR results, Sanger sequencing of cDNA from patients' LCLs confirmed all variants at the mRNA level ( Fig. 2b and Supplementary Fig. 3).
In one individual with early pulmonary distress and severe multisystem involvement (individual 5), the recurrent missense variant p.(Asp148Tyr) was detected in trans with an N-terminal nonsense variant p.(Gln50*). In seven individuals with overall less severe neurological manifestations and without pulmonary symptoms, ES detected the known pathogenic variant p.(Asp148Tyr) in homozygous state. The homozygous NHLRC2 start-loss variant c.1A>G was detected in four individuals from two unrelated families (A and F) in association with a severe and progressive neurological phenotype without pulmonary disease. The variant is absent in controls according to the gnomAD database and the closest in-frame alternative translation start codon is located at c.433, p.145. By analyzing the individual vcf files we confirmed that they share the same disease haplotype ( Supplementary Fig. 4). NHLRC2 protein levels, extracted from LCLderived cells of individuals 1 and 2, were strongly reduced compared to control samples (Fig. 2c, d).
In individual 4, presenting with neuroregression and epileptic encephalopathy, a nonsense variant p.(Leu584*) was detected in combination with a rare missense variant p.(Asp692Tyr) affecting a moderately conserved residue located in the ß-strand domain. The nonsense variant is predicted to undergo nonsensemediated decay (NMD) and Western blotting showed a strong reduction in NHLRC2 protein levels compared to control samples (Fig. 2c, d).
Two missense variants p.(Gln365Pro) and p.(Phe462Ser), affecting highly conserved residues within the ß-propeller domain, were detected in compound heterozygous state in individual 6, in association with speech regression, mild gait disturbance and medication-responsive epilepsy. While the former variant is listed once in heterozygosity in gnomAD, the latter has an allele frequency of 0.1% in the non-Finnish European population.
In individual 10, presenting with a severe disease including respiratory distress, malabsorption, anemia and recurrent infections, ES detected a single amino acid in-frame deletion affecting the highly conserved residue p.Glu33. Individual 11 was compound-heterozygous for an in-frame deletion p.(Gln333del) and a C-terminal frameshift variant that is unlikely to undergo NMD. She had severe neurological manifestations including epileptic encephalopathy and was previously reported with USP19 as a candidate gene [19].
In vitro studies of missense and in-frame deletion variants Firstly, to test the impact of the non-truncating variants identified in patients from our study (n = 5), secondly to compare them with previously reported missense variants associated with pulmonary symptoms (n = 4) and thirdly to critically review the strongly reduced NHLRC2 protein levels observed in LCLs, we cloned all nine non-truncating NHLRC2 variants together with a wildtype control into pcDNA3. For reliable detection, we added a C-terminal Flag-tag (Fig. 3a, b, Supplementary Table 1).
Western blot of HEK293 cells transfected with these different NHLRC2 constructs (Fig. 3c, d)  To investigate for a possible genotype-phenotype correlation, we compared these findings to variants and their combinations identified in individuals with and without pulmonary disease (Fig. 3e). We observed a correlation of remaining NHLRC2 protein levels with phenotype severity: higher reduction in total NHLRC2 protein level correlated with more severe phenotypes, ranging from milder over more severe neurological symptoms to additional lung disease (Fig. 3f).  Supplementary  Fig. 3) c Western Blot of NHLRC2 in whole cell lysates from patients' or healthy control lymphoblastoid cell lines. β-Actin is shown as loading control. d Quantification of NHLRC2 intensity relative to β-Actin with Image J.
In silico modeling of non-truncating variants To further investigate how the NHLRC variants that still result in stable NHLRC2 protein, as deduced from the in vitro overexpression assays, affect the folding of the protein, we modeled the missense variants p.(Asp148Tyr), p.(Asp692Tyr) and p.(Phe462Ser) and the p.(Gln333del) variant with the AlphaFold tool (Fig. 4). In the recurrent missense variant p.(Asp148Tyr), the replacement of the negatively charged amino acid aspartate by the polar, uncharged tyrosine is predicted to result in the loss of several hydrogen bonds (Fig. 4a, b). Analysis of the internal model accuracy estimates as    measured by the pLDDT score at the p.148 position in the wildtype vs. the mutant model revealed a prediction accuracy reduction at the position of the exchange. Notably, this position has previously been predicted with high certainty in the wildtype (Fig. 4c, d) and is also predicted with certainty in other models not affecting the same amino acid position ( Supplementary Fig. 5). All four other variants analyzed also lead to the loss of at least one hydrogen bond formed at the wildtype position of the respective variant (Fig. 4e). Further analysis of the amino acid positions affected by the missense variants with the MaSIFtool suggested possible protein binding sites in the vicinity to amino acids Gln333, Phe462 and Asp692. We therefore modeled the respective variants and their potential effect on these binding sites. While we observed a decrease in the probability of protein-protein interaction at Gln333 and Phe462 for the p.(Gln333del) and p.(Phe462Ser) variants, respectively, we could not detect such an effect for the p.(Asp692Tyr) variant ( Fig. 4e-h, Supplementary Figure 5).

DISCUSSION
Only 13 individuals from eight unrelated families of different geographical origin with biallelic pathogenic NHLRC2 variants have been published. Therefore, NHLRC2 has only a limited evidence class assignment in the curation database GenCC. In all previously published cases, the recurrent missense variant p.(Asp148Tyr) was detected. When FINCA syndrome was first characterized, this missense variant was reported in three individuals in trans with the truncating variant p.(Arg201Glyfs*6) [5]. All of these first patients presented with a severe multisystem phenotype including progressive respiratory disease. In total, 11 out of 13 individuals with NHLRC2-related disease showed respiratory symptoms and six died of respiratory failure before the age of three years. In contrast, only one child from our cohort (individual 5) had an early fatal course with pulmonary involvement. Notably, the combination of the recurrent missense variant p.(Asp148Tyr) and a truncating variant p.(Gln50*) was also detected here. In three previously published families, the recurrent missense variant occurred in combination with a second missense variant p.(Asp75Val), p.(His143Arg) or p.(Pro338Leu) in a total of five individuals with pulmonary disease [6,7].
Only a few children without progressive pulmonary symptoms have been reported so far. In four of them, the recurrent variant was present in the homozygous state and once it was detected in trans with the missense variant p.(Gly326Val) [7,8]. In our cohort, homozygosity for the recurrent variant p.(Asp148Tyr) was found in six children from five unrelated families. In agreement with the cases mentioned above, these individuals did not have respiratory symptoms and were overall less severely affected.
In addition, we describe the first eight individuals with an NHLRC2-associated disease, in whom the recurrent variant p.(Asp148Tyr) was not detected. All of those presented with a variable progressive clinical course, comprising global developmental delay, and various neuromuscular symptoms. Only one child had a more pronounced multisystem phenotype including respiratory distress, recurrent infections, malabsorption and anemia (individual 10, p.(Glu33del)). In two individuals with severe neurological manifestation including epileptic encephalopathy, the combination of a nonsense/frameshift variant and a nontruncating variant was detected (  Fig. 4). It is currently unclear whether a rescue mechanism is involved, or whether a significantly truncated protein is formed that cannot be detected with the antibody used here. To our knowledge, no individuals with biallelic complete LoF variants have been described. Notably, a complete knockout also results in embryonic lethality in mice [20].
Overall, there appears to be variability in the time course and severity of clinical manifestations in individuals with biallelic NHLRC2 variants. Studying different non-truncating variants and linking their effects on NHLRC2 protein levels revealed a putative genotype-phenotype correlation: variants leading to severely reduced protein levels (either in homozygous or in compoundheterozygous state with another severe missense or frameshift/ nonsense variant) were associated with an early onset multisystem phenotype including pulmonary disease. When the sum of NHLRC2 protein levels from both alleles appears to exceed a certain critical level, a phenotype without progressive respiratory symptoms was observed (Fig. 3e, f). This may explain why the recurrent p.(Asp148Tyr) variant, resulting in reduced but still detectable remaining protein levels, is associated with a more severe phenotype with pulmonary involvement when in trans with a LoF variant (e.g. nonsense, frameshift or one of the severe missense variants p.(Asp75Val), p.(Pro338Leu), p.(His143Pro)), whereas homozygosity for this variant results in a milder phenotype without pulmonary disease. Interestingly, we also observed a tendency in the group without reported lung involvement. The lower the total protein level, the earlier and more pronounced the neurological manifestations occurred (e.g., intractable seizures and epileptic encephalopathy seen in individuals 4 and 11). Notably, in silico modeling of variants predicted to result in stable NHLRC2 protein levels revealed possible effects on proper NHLRC2 function via alteration of protein binding sites (Fig. 4). Since we had only cloned the previously published missense variants, which were associated with pulmonary involvement, it would be interesting to know whether overexpression of the p.(Gly326Val) variant (detected in trans with the recurrent missense variant in a patient without pulmonary disease) [8] also results in a stable but presumably reduced product. Under this assumption, in silico models of the p.(Gly326Val) variant showed a reduction in prediction accuracy at the affected amino acid position ( Supplementary Fig. 5), potentially implying interference of the variant with the local structural context.
Currently, it is difficult to classify non-truncating variants as (likely) pathogenic according to the ACMG scoring framework, especially in autosomal recessive disorders when the second variant is not known to be pathogenic. In this situation, the criteria PM2 and PP3 are likely to be assigned to most NHLRC2 missense variants and subsequently classified as a variant of uncertain significance (VUS). The small number of pathogenic NHLRC2 variants described so far and the lack of knowledge about the function of NHLRC2 are likely to hinder a molecular diagnosis in unsolved cases. In this regard, we consider it useful to apply the protein modeling approach presented here and to perform a comprehensive clinical characterization of the affected individual. We recommend that AlphaFold mutant protein modeling and MaSIF prediction be calculated to check whether the variant is localized to a potentially functionally relevant region or substantially alters the tertiary structure of the protein. The criterion PM1 could then be applied at the level of supporting evidence (PM1_sup). However, we urgently need basic knowledge of the physiological function and interaction partners or substrates of NHLRC2 to validate the predicted effects and to establish a possible functional readout.
In addition, the specificity of the phenotype criterion (PP4) could be upgraded to moderate level of evidence, if additional symptoms besides NDD/ID are present and comprehensive genetic testing has been performed. In our opinion, this should include symptoms from different systems, such as neurological, pulmonary, gastrointestinal or hematological.
Taken together, our data broaden the hitherto known phenotypic spectrum, extend the allelic series and emphasize that rare biallelic NHLRC2 variants should be considered relevant in patients with NDD/ID, movement disorders, neuroregression and epilepsy even in the absence of pulmonary findings. To explain this variable phenotypic spectrum, we propose a genotype-phenotype correlation of residual NHLRC2 protein level and function with phenotype severity. Follow-up studies reporting the clinical course of affected individuals would be important to understand whether all affected individuals have a progressive multi-organ disease with variability in age of onset and severity of clinical manifestations, or whether there is indeed a distinct genotype-phenotype correlation. In this respect, our proposed model is limited and needs to be critically evaluated in further publications.

CODE AVAILABILITY
The source code for AlphaFold is publicly available at https://github.com/deepmind/ alphafold. The source code for MaSIF is publicly available at https://github.com/LPDI-EPFL/masif. All analysis scripts for processing AlphaFold models and MaSIF predictions are available on request.