Genetic and Clinical Studies of Peripheral Neuropathies with Three Small Heat Shock Protein Gene Variants in Korea

Small heat shock proteins (sHSPs) are ATP-independent chaperones that help correct the folding of denatured proteins and protect cells from stress. Mutations in HSPB1, HSPB8, and HSPB3 are implicated in inherited peripheral neuropathies (IPNs), such as Charcot-Marie-Tooth disease type 2 (CMT2) and distal hereditary motor neuropathies (dHMN). This study, using whole exome sequencing or targeted gene sequencing, identified 9 pathogenic or likely pathogenic variants in these three sHSP genes from 11 Korean IPN families. Most variants were located in the evolutionally well conserved α-crystallin domain, except for p.P182S and p.S187L in HSPB1. As an atypical case, a patient with dHMN2 showed two compound heterozygous variants of p.R127Q and p.Y142H in HSPB1, suggesting a putative case of recessive inheritance, which requires additional research to confirm. Three HSPB8 variants were located in the p.K141 residue, which seemed to be a mutational hot spot. There were no significant differences between patient groups, which divided by sHSP genes for clinical symptoms such as onset age, severity, and nerve conduction. Early-onset patients showed a tendency of slightly decreased sensory nerve conduction values compared with late-onset patients. As a first Korean IPN cohort study examining sHSP genes, these results will, we believe, be helpful for molecular diagnosis and care of patients with CMT2 and dHMN.


Introduction
Inherited peripheral neuropathies (IPNs) are a group of genetically and clinically heterogeneous neuromuscular disorders characterized by distal muscle weakness and loss of sensation in the upper and lower extremities. They include hereditary motor and sensory neuropathy (HMSN), conventionally called Charcot-Marie-Tooth disease (CMT), distal hereditary motor neuropathies (dHMN), hereditary sensory autonomic neuropathies (HSAN), and hereditary neuropathy with pressure palsies (HNPP). As the most common IPN, CMT is usually divided into three categories according to the peripheral nerve conduction velocities (NCVs): the demyelinating type (also called CMT1), the axonal defective type (also called CMT2), and the intermediate type.
Mutations in more than 130 genes with variable functions have been reported to be implicated in the pathogenesis of IPNs. Among these genes, three small heat shock protein (sHSP) genes belonging to the ATP-independent molecular chaperones family have been reported to cause several types of IPNs, such as dominant CMT2 and dHMN. Mutations in

Clinical Examinations
Clinical symptoms of motor and sensory impairments, muscle atrophies, and deep tendon reflexes (DTRs) were measured as per standard methods. Strengths of the flexor and extensor muscles were manually evaluated by the standard Medical Research Council (MRC) scale. Physical disability was determined by dual methods of the CMT neuropathy score ver. 2 (CMTNSv2) and functional disability scale (FDS). Sensory impairments were measured by severity and level of pain, vibration, temperature, and position. Onset ages were estimated by asking patient's ages when a physical symptom first appeared [21].

Nerve Conduction Studies
Motor and sensory conduction velocities were measured by surface stimulation. Motor nerve conduction velocities (MNCVs), and compound muscle action potentials (CMAPs) of the median, ulnar, and radial nerves were measured by stimulation of the elbow and wrist. The peroneal and tibial MNCVs and CMAPs were measured by stimulation of the knee and ankle. Amplitude of CMAP was determined from baseline to negative peak value. Sensory nerve conduction velocities (SNCVs) of the median, ulnar, radial, and sural nerves were determined using the finger-wrist segments by the orthodromic method. Amplitude of sensory nerve action potential (SNAP) was determined from positive to negative peaks.

DNA Extraction and Genetic Screening
Total DNA was extracted from blood using the HiGene Genomic DNA Prep Kit (Biofact, Daejeon, Korea). All samples were first examined for PMP22 duplication, and then, proband samples with no PMP22 duplication were screened by whole exome sequencing (WES; 512 samples) or targeted gene panel sequencing (TS; 166 samples) based on the next generation sequencing (NGS). Exome capture and NGS were carried out using the SureSelect Human All Exon 50M kit (Agilent Technologies, Santa Clara, CA, USA) and the HiSeq 2000 or 2500 Genome Analyzers (Illumina, San Diego, CA, USA), respectively. TS was performed by the method of Nam et al. [22]. As the reference sequence, the UCSC assembly GRCh37/hg19 was used for mapping (http://genome.ucsc.edu/ accessed on 1 November 2021). Rare or unreported variants with the minor allele frequencies (MAFs) of <0.01 were isolated from IPN-related genes; their MAFs were obtained from human genome databases of the 1000 Genomes Project (1000G; http://www.1000genomes.org/ accessed on 1 November 2021) and Genome Aggregation Database (gnomAD; https://gnomad. broadinstitute.org/ accessed on 1 November 2021). For Korean frequency data, Korean Reference Genome Database (KRGDB; http://coda.nih.go.kr/coda/KRGDB/index.jsp/ accessed on 1 November 2021) was used. Pathogenicity of variants was evaluated basically by the American College of Medical Genetics and Genomics (ACMG) guideline (https:// wintervar.wglab.org/ accessed on 1 November 2021) and previous studies. For pathogenic candidate variants, Sanger sequencing was performed to confirm their presence using the SeqStudio or 3130XL Genetic Analyzers (Life Technologies-Thermo Fisher Scientific, Carlsbad, CA, USA).

Statistical Analysis
Non-parametric Mann-Whitney U test and parametric unpaired t-test were used to compare the clinical data of patients grouped by sHSP genes. Parametric Pearson r or non-parametric Spearman r values were used to analyze the correlation between clinical values and onset ages. Significant differences were determined at p < 0.05. Statistical analyses were performed using GraphPad Prism (ver. 8.00, GraphPad Software, San Diego, CA, USA).

Causative Variants in sHSP Genes
We identified nine pathogenic or likely pathogenic sHSP gene variants in 11 families from the Korean IPN cohort study (Table 1); their pathogenicity was basically determined using the guideline of ACMG (Table S1). Five HSPB1 variants were observed in six families ( Figure 1A) and three HSPB8 variants were observed in four families ( Figure 1B). Only one variant was found in HSPB3 ( Figure 1C). These results include the sHSP gene variants in Koreans previously reported by our research group [7,[18][19][20]. The genotypes are provided at the bottom of all the examined individuals in the pedigrees in Figure 1. All the causative variants were confirmed in the extended family members by Sanger sequencing (Figure 2A).
Amino acids of most variant sites were highly conserved from worm to mammal ( Figure 2B). Moreover, variants located in the α-crystallin domain were also fairly conserved among human sHSP paralogues, while the variants located in the variable C-terminal domain were less conserved ( Figure 2C). Most variant sites were located in the evolutionally well conserved α-crystallin domains, while the p.P182S and p.S187L in HSPB1 were located in the C-terminal domain ( Figure 3A). Pathogenic effects were predicted for all the causative variants by at least one of three in silico analyses (Table 1).      Five pathogenic or likely pathogenic mutations were identified in HSPB1 from six CMT2 or dHMN families ( Figure 1A). A c.404C>T (p.S135F) was observed three times in one large dHMN family (FC189) and two CMT2 families (FC522 and FC567); many studies have reported this mutation as a genetic cause of CMT2F and dHMN2B [1,2,18,19,23,24].

Variants in HSPB1
Five pathogenic or likely pathogenic mutations were identified in HSPB1 from six CMT2 or dHMN families ( Figure 1A). A c.404C>T (p.S135F) was observed three times in one large dHMN family (FC189) and two CMT2 families (FC522 and FC567); many studies have reported this mutation as a genetic cause of CMT2F and dHMN2B [1,2,18,19,23,24]. A 6-year-old boy (VI-2) in the FC189 family and two girls (6 and 7 years old, III-1 and III-2) in the FC522 family had the p.S135F variant, but they were still not affected. As a same site mutation, c.404C>A (p.S135Y) was once reported as an underlying cause of CMT2 [28]. A variant of c.544C>T (p.P182S) was observed in a CMT2F family (FC313). Neither of the unaffected parents (I-1 and I-2) of the proband showed that variant, suggesting a de novo mutation. This variant was once reported by Kijima et al. [25]. A c.560C>T (p.S187L) was observed in a CMT2F family (FC1150), which was reported by Echaniz-Laguna et al. [23]. Two heterozygous variants of c.380G>A (p.R127Q) and c.424T>C (p.Y142H) were interestingly found in a sporadic patient with dHMN (FC1005). The p.R127Q was reported as a variant of uncertain significance in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/ accessed on 5 December 2021), while two mutations at the p.R127 residue have been reported several times to be pathogenic in Belgian, French, Norwegian, and Japanese CMT2F or dHMN2B patients with a dominant manner: p.R127W [1,23,26,29,30] and p.R127L [31,32]. The p.Y142H was an unreported novel variant. Both variants were evaluated as "likely pathogenic" by the guidelines of the ACMG. Since three children had only one of the two variants (III-1: p.Y142H; III-2 and -3: p.R127Q), each variant seemed to be inherited from each of the proband's parents. Both parents were normal by history taking and three children showed no abnormal phenotype at the recent neurological examination, including motor and sensory function, and DTR. Thus, it was suggested that the dHMN phenotype might be due to the biallelic effect of the two compound heterozygous variants.

Variant in HSPB3
One variant of c.352T>C (p.Y118H) was observed from HSPB3 in a CMT2 family (FC702) ( Figure 1C). The ACMG guideline evaluated the p.Y118H as a variant of uncertain significance (VUS). However, this study classified it as a likely pathogenic variant, because it was previously reported as an underlying cause of CMT2 by Nam et al (7). In HSPB3, only a p.R7S mutation was reported as a pathogenic variant causing dHMN2C, except for p.Y118H [6].

Simulation of 3D Protein Structure
Three-dimensional conformational changes of mutant proteins were simulated to predict the effects of the sHSP gene mutations (Figure 4). In HSPB1, the p.R127 residue of the p.R127Q mutation in the α-crystallin domain forms salt bridges with negatively charged p.D107 and p.E125 located in the β4 and β5 strands. However, the salt bridge with p.D107 was destroyed in the p.R127Q mutant, predicting a conformational modification. The p.S135 forms hydrogen bonds with p.H124 and p.Y133 within the β-strands arranged in parallel. However, both hydrogen bonds were disappeared in the p.S135F mutant. The HSPB1 p.Y142, which is located at the β6-β7 strand in an α-crystallin domain, has a hydrogen bond with a near-located positively charged p.R140 residue. However, it was predicted to break and the β-strand structure was distorted by the p.Y142H mutation. The p.P182S mutation located in the C-terminal IxI/V motif newly forms two hydrogen bonds with an adjacent hydrophilic p.Q175 residue, which might result in a conformational change of the protein quaternary structure. In the p.S187L mutation of the C-terminal domain, hydrogen bonds of p.S187 with p.F185 and p.Q190 were destroyed and new bonds with p.T184 and p.A189 were formed. The p.S187L mutation was previously reported to cause protein aggregation [23]. In HSPB8, p.K141 residue in the α-crystallin domain forms hydrogen bonds with p.N101 and p.V121, and forms a cation-π interaction with p.F139. Amino acid substitution with a negatively charged residue (p.K141E) caused the breakage of these hydrogen bonds and formed new hydrogen bonds with N-terminal residues. In the p.K141N mutation, a new hydrogen bond was similarly formed between p.K141N and N-terminal p.R86. In the p.K141T mutant, a hydrogen bond to p.N101 was broken, which was predicted to cause a structural alteration of the β-sheets.
In HSPB3, p.Y118 of wild type forms two hydrogen bonds with p.L94, and p.W93 formed a cation-π interaction with p.K119. However, in the p.Y118H mutant, the cation-π interaction was destroyed, predicting an unstable structural opening.  In HSPB3, p.Y118 of wild type forms two hydrogen bonds with p.L94, and p.W93 formed a cation-π interaction with p.K119. However, in the p.Y118H mutant, the cationπ interaction was destroyed, predicting an unstable structural opening.

Rare Variants in sHSP Genes with Uncertain Significance
We observed three rare sHSP variants with MAFs less than 0.01 in the 1000G and KRGDB in the sporadic IPN patients who showed phenotypes of CMT1 or HNPP instead of CMT2 or dHMN: p.R27P and p.Q128X in HSPB1 and p.F79C in HSPB8. The p.R27P in HSPB1 and p.F79C in HSPB8 were found in each patient with the HNPP phenotype (Table  S2). The p.Q128X stop-gain variant in HSPB1 was found in a patient with CMT1. These three variant were evaluated by VUS by the ACMG guideline. Several HSPB1 stop-gain or frameshift mutations, such as p.W16X, p.W45X, p.L58Cfs*52, and p.L191Qfs*36, have

Rare Variants in sHSP Genes with Uncertain Significance
We observed three rare sHSP variants with MAFs less than 0.01 in the 1000G and KRGDB in the sporadic IPN patients who showed phenotypes of CMT1 or HNPP instead of CMT2 or dHMN: p.R27P and p.Q128X in HSPB1 and p.F79C in HSPB8. The p.R27P in HSPB1 and p.F79C in HSPB8 were found in each patient with the HNPP phenotype (Table S2). The p.Q128X stop-gain variant in HSPB1 was found in a patient with CMT1. These three variant were evaluated by VUS by the ACMG guideline. Several HSPB1 stop-gain or frameshift mutations, such as p.W16X, p.W45X, p.L58Cfs*52, and p.L191Qfs*36, have been reported to VUS by ClinVar, but some variants (such as p.L58Afs*105, p.A61Rfs*100, and p.Q175X) have been reported to be pathogenic [23,32,33].

Clinical Characterization of Patients with sHSP Gene Mutations
The clinical phenotypes of 29 patients with sHSP gene mutations are shown in Table S3. The clinical phenotypes of the patients with three gene mutations were largely similar. Weakness and atrophy of muscles began and predominated in the distal portions of legs, and were less severe in the distal upper limbs. Disability in the distal lower limbs varied from mild weakness to complete paralysis. The mean onset age of the patients with the HSPB1 mutations was 23.6 ± 8.2 years, with a wide range of 11 to 45 years; that of the patients with the HSPB8 mutations was 18.6 ± 3.9 years, with a relatively narrow range of 13 to 26 years. In the HSPB1 group, the mean onset ages of the dHMN2B and CMT2F patients were similarly 23.8 ± 7.1 years and 23.1 ± 10.1 years, respectively. In the HSPB8 group, the mean onset age of the dHMN2A patients was 19.5 ± 3.3 years, and that of the CMT2L patient was 13 years. The mean FDS and CMTNSv2 were 2.7 ± 1.8 and 13.4 ± 8.2 in the HSPB1 group, and 3.3 ± 1.5 and 15.7 ± 6.0 in the HSPB8 group, respectively.
Sensory loss was present in 7 of 20 patients with the HSPB1 mutations, in 1 of 7 patients with the HSPB8 mutations, and in all patients with the HSPB3 mutations. Pes cavus was present in 17 of 20 patients of HSPB1 and in all patients of HSPB8 and HSPB3. As additional symptoms, one CMT2L patient with p.K141T in HSPB8 (FC031: II-1) showed scoliosis, and a 62-year-old dHMN2B woman with p.S135F in HSPB1 (FC189: IV-13) showed the most severe physical disability of being wheelchair bound (FDS: 7, CMTNSv2: 31), while a dHMN2B man with compound heterozygous HSPB1 mutations of p.R127Q and p.Y142H (FC1005: II-5) revealed late onset (45 years) and mild clinical symptoms (FDS: 1, CMTNSv2: 4) compared with other affected individuals.
When clinical phenotypes were compared between HSPB1 and HSPB8 patients, patients with HSPB8 mutations showed slightly earlier onset and more severe physical disability than those with HSPB1 mutations, but no significant differences were observed in the onset age (U = 39.5, p = 0.096), FDS (U = 51, p = 0.293), and CMTNSv2 (t = 0.695, p = 0.493) ( Figure 5A-C). In addition, there were no significant differences in onset and physical severity between male and female patients (onset: U = 97.5, p = 0.859; FDS: U = 93, p = 0.699; CMTNSv2: t = 0.655, p = 0.518). When the correlations between onset ages and physical disability levels were examined for all the examined patients, no significant correlations were observed in the values of FDS and CMTNSv2 (FDS: p = 0.418, CMTNSv2: p = 0.571) ( Figure 6A,B).

Electrophysiological Findings of Patients with sHSP Gene Mutations
The electrophysiological values of 28 patients with sHSP gene mutations are shown in Table S4. The nerve conduction findings of the patients with three gene mutations were largely similar, and there was no significant difference among the three gene groups. All patients showed predominantly decreased motor CMAP amplitudes in lower limbs compared to upper limbs. The mean median CMAP in upper limbs was 11.1 ± 7.4 mV and the mean peroneal CMAP in lower limbs was 0.8 ± 1.7 mV. In the follow-up studies of some patients, the median MNCV was found to decrease at an average of 1.56 ± 2.20 m/s per year, but the decrease rate of the median SNCV was much lower, with an average of 0.04 ± 1.63 m/s per year.
When phenotypes were compared between HSPB1 and HSPB8 patients, no significant differences were observed in the median MNCV (U = 47, p = 0.213) or median SNCV (U = 55, p = 0.422) ( Figure 5D,E). In addition, there were no significant differences in nerve conduction values between male and female patients (median MNCV: U = 77.5, p = 0.403; median SNCV: U = 89, p = 0.763). When the correlations between onset age and electrophysiological features were analyzed, early-onset patients showed a tendency of slightly decreased sensory nerve conduction values compared with late-onset patients (median SNCV: p = 0.024 and median SNAP: p = 0.031); however, other comparisons of motor nerve conduction values showed no significant correlations (median MNCV: p = 0.319, and median CMAP: p = 0.308) ( Figure 6C,D).
When clinical phenotypes were compared between HSPB1 and HSPB8 patients, patients with HSPB8 mutations showed slightly earlier onset and more severe physical disability than those with HSPB1 mutations, but no significant differences were observed in the onset age (U = 39.5, p = 0.096), FDS (U = 51, p = 0.293), and CMTNSv2 (t = 0.695, p = 0.493) ( Figure 5A-5C). In addition, there were no significant differences in onset and physical severity between male and female patients (onset: U = 97.5, p = 0.859; FDS: U = 93, p = 0.699; CMTNSv2: t = 0.655, p = 0.518). When the correlations between onset ages and physical disability levels were examined for all the examined patients, no significant correlations were observed in the values of FDS and CMTNSv2 (FDS: p = 0.418, CMTNSv2: p = 0.571) ( Figure 6A, 6B).

Electrophysiological Findings of Patients with sHSP Gene Mutations
The electrophysiological values of 28 patients with sHSP gene mutations are shown in Table S4. The nerve conduction findings of the patients with three gene mutations were largely similar, and there was no significant difference among the three gene groups. All patients showed predominantly decreased motor CMAP amplitudes in lower limbs compared to upper limbs. The mean median CMAP in upper limbs was 11.1 ± 7.4 mV and the mean peroneal CMAP in lower limbs was 0.8 ± 1.7 mV. In the follow-up studies of some patients, the median MNCV was found to decrease at an average of 1.56 ± 2.20 m/s per year, but the decrease rate of the median SNCV was much lower, with an average of 0.04 ± 1.63 m/s per year.

Discussion
Mutations in the HSPB1, HSPB8, and HSPB3 genes have been reported to cause autosomal dominant IPNs of CMT2 and dHMN. This study identified nine pathogenic or likely pathogenic sHSP gene mutations from 11 families in the Korean IPN cohort study. As expected, HSPB1 mutations were most frequently observed in six families with five mutations. HSPB8 mutations were observed in four families with three mutations, and HSPB3 mutation was observed only in a family. Three mutations (p.R127Q in HSPB1, p.K141T in HSPB8, and p.Y118H in HSPB3) were unreported in other ethnic groups, although they were registered in dbSNP. Two paternal-originated de novo mutations were observed in each HSPB1 and HSPB8. The de novo mutation was suggested to have originated from the father by haplotyping analysis using adjacent SNPs. The frequencies of IPN families with sHSP gene mutations was determined to be 1.41% in the total unrelated IPN patients and 1.62% in the patients excluding CMT1A (Table 2). When the frequency in the patients excluding CMT1A was compared with other study groups, it was similar with 1.59% of Japanese [10]. It was higher than the frequencies of Han Chinese (0.91%), British (0.83%), American (0.00%), and Brazilian (0.00%), while lower than those of Italian (4.80%) and Spanish (3.94%) [8,9,[34][35][36].
As an unusual case, a male patient with dHMN2 interestingly showed two heterozygous mutations of p.R127Q and p.Y142H in HSPB1. The p.R127Q has been registered as a variant of uncertain significance in ClinVar, while p.R127W and p.R127L in the p.R127 residue have been reported to be pathogenic for CMT2F or dHMN2B [1,23,26,[29][30][31][32]. Both of this patient's parents were unaffected, although genetic and neurological testing was not performed, and three children having alternatively one of each variant were also unaffected. Thus, the patient seems to be a recessive type of dHMN2B by bialleles of two compound heterozygotes. However, we could not exclude a possibility that any of two variants may role as a dominant genetic cause. As the reasons, (a) the ages of the three children (29,28, and 22 years old, respectively) may not have reached the onset age yet, considering the proband's onset of 45 years old, and (b) it may possible that a variant was inherited by a de novo event. In spite of these two mutations, our patient showed late onset and mild symptoms. So far, a few cases of autosomal recessive CMT2 or dHMN2 patients with HSPB1 mutations have been reported. As a first report, a p.L99M homozygous mutation was reported in a consanguineous Pakistani patient with dHMN [2]. Then, a p.R140G homozygous mutation was reported in an Indian patient with distal vacuolar myopathy and motor neuropathy [3], and pS315F and p.R316L homozygous mutations were identified in the Republic of Cabo Verde and Iranian families with CMT2, respectively [24]. Abbreviations: sHSP, small heat shock protein; CMT1A, Charcot-Marie-Tooth disease type 1A.
All three HSPB8 mutations (c.421A>G;p.K141E, c.422A>C;p.K141T, and c.423G>T; p.K141N) observed in this study were, interestingly, located in the p.K141 residue. Including this result, four mutations at the p.K141 residue have been reported as the genetic causes of CMT2 or dHMN: p.K141E, p.K141T, p.K141M, and p.K141N [4,5,20,23,27]. Thus, the "AAG" sequence coding p.K141 seems to be a mutational hot spot, which is predicted to locate a stem of single-stranded DNA hairpin structures. Most patients with HSPB8 defects showed mutations in the p.K141 site, with the exception of two families with p.P90L and p.N138T. The p.K141 was well conserved with basic amino acids of lysine or arginine among homologues of paralogues as well as orthologues.
The clinical features of the patients with three gene mutations were largely similar, and there were no significant differences for onset age, severity, nerve conduction, or muscle atrophy between patient groups between HSPB1 and HSPB8 mutations. Actually, the clinical features of HSPB1 and HSPB8 patients were difficult to distinguish in both clinical examination and nerve conduction studies. Therefore, when sHSP patients are suspected, genetic testing can be said to be the most useful. Physical severities measured by FDS and CMTNSv2, electrophysiological values measured by CMAP, and MNCV showed no significant difference by onset age except for SNCV and SNAP. The mean onset age was earlier in CMT2 patients than in dHMN patients, but with no significant difference.

Conclusions
This IPN cohort study identified nine sHSP gene mutations in 11 families as the underlying causes of CMT2 or dHMN phenotypes. In particular, this study is the first report of a putative patient with autosomal recessive HSPB1 mutations in Korea. We carefully analyzed genotype-phenotype correlations and also compared the clinical phenotypes according to genes and onset ages. As a first Korean IPN cohort study analyzing sHSP genes, we believe that this study will be helpful for the molecular diagnosis and care of patients with CMT2 and dHMN.