1 INTRODUCTION

Charcot–Marie–Tooth neuropathy (CMT) is the most common inherited peripheral neuropathy, affecting 1/2500 individuals.¹ CMT is characterized by progressive length-dependent loss of peripheral motor and sensory nerves, resulting in distal muscle weakness and sensory symptoms.² Patients are broadly divided into subtypes based on whether nerve conduction studies (NCS) indicating demyelinating (CMT1) or axonal (CMT2) forms of the disease. Recessive loss-of-function variants in the membrane metalloendopeptidase (MME) gene have been previously reported to cause axonal Charcot–Marie–Tooth neuropathy type 2T (CMT2T; OMIM: #617017).^3-11 MME encodes for neprilysin, a widely expressed membrane-bound metallopeptidase that has a key role in neuropeptide processing.¹² A significant portion of patients with axonal CMT remain genetically undiagnosed,^13-17 indicating that further disease-causing genes and pathogenic variants in known genes are yet to be identified.

Splicing variants are increasingly being recognized as a cause of Mendelian disease.^{18, 19} The precise removal of introns (non-coding regions) and inclusion of exons (coding regions) in the final mature mRNA relies on the spliceosome and auxiliary splicing factors recognizing specific sequence motifs, such as the 5′ donor splice site and 3′ acceptor splice site. Splice-altering variants can weaken or abolish recognition of the correct splice sites, or alternatively strengthen or create cryptic splice sites that mimic consensus splicing sequences.²⁰ These variants typically lead to one or more mis-splicing events that result in the skipping of partial or complete exons, and/or the retention of partial or complete introns.^21-42 Pathogenic splicing variants have been found in several CMT genes including MPZ,^{21, 30, 39, 40} MFN2,^{22, 30, 31} LRSAM1,²³ IGHMBP2,²⁴ INF2,²⁵ MCM3AP,²⁶ SH3TC2,^{30, 32, 38} GDAP1,^{27, 42} SBF1,²⁸ NDRG1,³⁷ and FGD4.²⁹ Variants affecting canonical splice donor and acceptor sites have also been described in MME.^{4, 41}

Exon-trapping, also known as a mini-gene assay, is an in vitro technique used to identify exons in a genomic region of interest.⁴³ This is of particular use when relevant patient tissue is unavailable or when relevant transcripts may be unstable or degraded by nonsense mediated decay (NMD).⁴⁴ The genomic region of interest is cloned into an exon-trapping vector between two known exons. This exon-trapping vector is transfected into a cell line where it is transcribed and undergoes a series of post-transcriptional processes that include pre-mRNA splicing to create mature mRNA. This mRNA consists of the ‘trapped’ exons of the genomic region of interest flanked by the known exons, which can then be Sanger sequenced to characterize the ‘trapped’ exons. Comparison of ‘trapped’ exons between wild type and variant genomic sequences can also indicate if a candidate variant affects splicing.^45-51 The well-validated exon-trapping vector pSpliceExpress⁵² consists of known exons of the rat insulin gene, Ins2, and has been used previously to determine the splicing impacts of multiple pathogenic variants in Mendelian disease.^45-51

Here we report a deep intronic variant in MME [chr3:155142758A>G (hg38); MME c.1188+428A>G], found in a recessive state in two Australian families. Exon-trapping revealed that this variant creates a novel splice donor site in MME intron 12 (NM_000902.5) which, along with a preceding existing cryptic splice acceptor site, results in the incorporation of an 83 bp pseudoexon in the MME transcript [chr3:155142675-155 142 757 (hg38); r.1188_1189ins[1188+345_1188+427]]. The coding frameshift caused by this pseudoexon leads to a PTC in exon 14 and likely NMD of the MME transcript (p.Ala397ProfsTer47), resulting in a loss of MME function.

2 MATERIALS AND METHODS

3 RESULTS

3.1 Clinical phenotypes

3.1.1 Family 1

Family 1 consists of four affected siblings from a consanguineous family of European (non-Finnish) background (Figure 1A). The phenotype was consistent with a generalized sensorimotor axonal neuropathy and sensory ataxia without cerebellar signs (Table 1), and was confirmed with NCS for individuals V:1 and V:6 (Table S2). Affected individuals had previously undergone diagnostic and research whole exome sequencing with negative results.

TABLE 1. Phenotypic characterization of individuals with recessive CMT MME variants reported in this manuscript.

Family #	Family 1				Family 2
Individual	V:1	V:3	V:4	V:6	II:1
Sex	M	F	M	M	F
Onset	Fourth decade	Childhood	Childhood	Childhood	Early-adult
Age at evaluation (years)	62	67	69	67	54
Presenting symptom	Gait unsteadiness	Difficulty running	Difficulty sitting on crossed legs during childhood	Difficulty running, multiple ankle sprains	Bilateral leg weakness and sensory disturbance
Atrophy UL/LL	Yes/yes	Yes/yes	Yes/yes	Yes/yes	No/Yes (left calf only)
Foot deformity	None	Pes cavus/hammer toes	-	Pes cavus	Pes cavus
Tone UL/LL	Normal/reduced	Normal/normal	Normal, increased		Normal/normal
Shoulder abduction	5	5	3	5	4
Elbow flexion	4	5	3	4	5
Elbow extension	4	5	3	4	5
Finger abduction	2	4	3	1	4
Hip flexion	4	3	3	3	4
Knee extension	4	3	3	4	4
Knee flexion	4	3	3	4	4
Ankle dorsiflexion	0	0	3	1	3
Ankle plantarflexion	0	0	3	0	4
Deep tendon reflexes UL/LL	Absent/absent	Absent ankle	Hyperreflexia knee, absent ankle	Absent/absent	Normal/ reduced ankle reflexes
Plantars	Absent	-	Absent	Flexor	Flexor
Proprioception UL/LL	Normal/reduced to the MTP	-	-	Normal/reduced to the ankle	Normal/reduced to ankle
Vibration UL/LL	Normal/reduced to ankle	Absent in toes	Reduced at the knee	Reduced to wrist/reduced to knees	Reduced to mid-shin
Pinprick UL/LL	Normal/reduced to ankle	Normal/reduced to the mid shin	Reduced to elbows/reduced to knees	Reduced to PIP joint/reduced to knees	Reduced to mid-shin Reduced dorsum hand
Temperature sensation UL/LL	Normal/reduced to ankle	-	-	Reduced to wrist/reduced to above knee	Reduced to mid-shin
Other symptoms/signs/results	Cough	GERD and cough	Corticospinal tract signs	Nil	Fatty infiltration on left gastrocnemius on MRI
Gait	High steppage/ataxic	High steppage/ataxic	Steppage gait	High steppage/ataxic	High steppage
Romberg's sign	Positive	Positive		Positive	Mild sway
Mobility aids	AFO, walking stick and mobility scooter	AFO, walker	AFO, walker, mobility scooter	Wheelchair	AFO, elbow crutch
CMTNS (version 2)	31	-	28	36	15

• Abbreviations: AFO, ankle foot orthoses; CMTNS, Charcot–Marie-Tooth Neuropathy Score; GERD, gastroesophageal reflux disease; UL/LL, upper limb/lower limb.

Individual V:1

This 62-year-old man developed symptoms from age 38 years with gait unsteadiness and progressive weakness of the upper and lower limbs. The balance and weakness progressed with multiple falls, and he required walking aids (ankle foot orthoses, walking stick and mobility scooter). Additionally, he reported loss of sensation in the feet and a dry cough. Neurological examination revealed a distal pattern of weakness and sensory loss in the upper and lower limbs, absent deep tendon reflexes, unresponsive plantars, and an ataxia gait with a positive Romberg's test. NCS showed a severe, generalized, sensorimotor axonal polyneuropathy (Table S2).

Individual V:3

This 67-year-old woman presented with difficulty running at school, weakness of the ankles, and recurrent ankle sprains. She developed numbness and soreness of the feet and a slowly progressive gait disturbance with recurrent falls. She is now only just able to get around the house using a walker and fixed AFO splints. She also had a distal pattern of weakness and sensory loss.

Individual V:4

This 69-year-old man first experienced symptoms as a child with difficulty sitting on crossed legs. He then noticed weakness in the ankles and legs and sensory loss in the feet, which slowly progressed to severe walking difficulty with falls requiring a walker, mobility scooter and AFOs. On examination, he had global muscle weakness, and sensory loss to the knees and elbows, with subtle corticospinal tract signs (increased tone and hyperreflexic knee jerks).

Individual V:6

This 67-year-old man first developed symptoms in childhood with difficulty running, multiple ankle sprains and difficulty with sport. The symptoms progressed and by his mid-twenties, he developed overt distal weakness and had difficulty with balance, with multiple falls. He currently has severe weakness in the upper and lower extremities with sensory deficits and balance problems and mobilizes with a wheelchair. Neurological examination revealed global weakness in the upper and lower limbs (distal greater than proximal), absent deep tendon reflexes throughout, downgoing plantars, glove and stocking sensory loss, and a positive Romberg's sign. There were no cerebellar or extrapyramidal signs, and the head impulse test was negative. NCS revealed a severe generalized axonal sensorimotor polyneuropathy (Table S2).

3.1.2 Family 2

Family 2 consists of one affected female born to a healthy, non-consanguineous couple (Figure 1B). There is no history of neuromuscular disorders in the family. The 60-year-old proband presented with slowly progressive bilateral leg weakness and sensory disturbances. NCS showed evidence of an axonal sensorimotor neuropathy (Table S2). Neurological examination revealed mild distal upper limb weakness and moderate distal lower limb weakness, with muscle atrophy only in the left gastrocnemius muscle (Table 1). Bilateral MRI of the thighs and calves showed multifocal calf muscle T2 hyperintensity, muscle atrophy and T1 hyperintense fatty replacement of the left medial gastrocnemius muscle (Figure S1). The patient had reduced ankle reflexes and showed mild muscle cramps and muscle fatigue. Creatine kinase (CK) levels were consistently elevated (>400 U/L).

3.1.3 Genetic analysis

WGS screening using the seqr platform in Family 1 revealed a single homozygous variant in two affected individuals (V:1 and V:3), MME c.1188+428A>G (NM_000902.5). This variant was reported in dbSNP build 155⁶⁶ (rs61758195), with a low MAF in gnomAD v4.0.0⁶⁷ (10/152120), All of Us (35/490748),⁶⁸ and TOPMED⁶⁹ (15/264290). No homozygous individuals were reported in gnomAD, All of Us, or TOPMED. SpliceAI predicted that MME c.1188+428A>G could create a strong novel splice donor site (Score 0.97; where a SpliceAI score above 0.8 is considered a ‘high precision’ predicted splice variant⁵⁶). This in turn strengthened the SpliceAI prediction for a cryptic splice acceptor site 83 bp upstream of the novel splice donor site (Score 0.99).

Segregation analysis in Family 1 confirmed the biallelic inheritance of the MME c.1188+428A>G variant segregated with the CMT phenotype (Figure 1A). Sanger sequencing of DNA from available individuals showed that all affected individuals were homozygous for the MME c.1188+428A>G variant and unaffected individuals were carriers (Figure S2).

WGS screening using the seqr platform and AIP was conducted in the proband of Family 2 (II:1). The MME c.1188+428A>G variant was detected in a compound heterozygous state with a second MME variant (MME c.467del; p.Pro156Leufs*14), which has previously been reported as pathogenic⁴¹ (Figure 1B). Sanger sequencing validated the presence of each MME variant in the index individual (Figure S3). As parental DNA was unavailable, ONT long-read sequencing was conducted to phase the heterozygous MME variants in the proband (Figure 1C,D). ONT long-read sequencing confirmed that MME c.1188+428A>G (boxed red in Figure 1C) and c.467del (boxed red in Figure 1D) were present on alternative haplotypes (hap-1: red, hap-2: blue) and therefore were in trans in the proband.

3.1.4 In vitro exon-trapping of MME-pSpliceExpress vectors

Two separate exon-trapping pSpliceExpress vectors were generated to assess the in vitro splicing impact of the MME c.1188+428A>G variant: pSpliceExpress-MME_WT (wild-type) and pSpliceExpress-MME_{c.1188+428A>G} (variant). A schematic of the constructs and relevant SpliceAI scores are shown in Figure 2A. RT-PCR products produced following transfection of these vectors were analyzed using gel electrophoresis (Figure 2B), which revealed a visible size difference between the wild-type (381 bp) and MME c.1188+428A>G amplicons (464 bp). The gel-purified amplicons were Sanger sequenced and the sequencing was aligned to the WT MME mRNA sequence. The sequenced products showed that the pSpliceExpress-MME_WT produced a transcript that was correctly spliced between MME exon 12 and 13 (Figure 2C). In contrast, exon-trapping of the pSpliceExpress-MME_{c.1188+428A>G} vector showed that an 83 bp pseudoexon had been spliced between MME exon 12 and 13 (Figure 2D). A BLAT search⁷⁰ using the sequence of the trapped pseudoexon revealed alignment to the intronic region directly upstream of the MME c.1188+428A>G variant [chr3:155142675-155 142 757 (hg38)]. This suggests that the MME c.1188+428A>G variant creates a novel splice donor site leading to the aberrant inclusion of 83 bp of intronic MME sequence in the final spliced transcript, as predicted by SpliceAI. Prediction of the novel MME coding sequence caused by the introduction of the pseudoexon showed that a PTC was generated in exon 14 (p.Ala397ProfsTer47) at genomic position chr3:155144368 (hg38) (Figure S4). This PTC likely leads to NMD of the MME transcript.

3.1.5 In silico splicing analysis of MME variants

All reported MME variants in gnomAD v.4.0.0 were assessed using the integrated SpliceAI scores to determine if splicing variants in MME were a likely underrecognized cause of disease. There were 37 264 total variants reported in MME in gnomAD, of which 35 673 variants had a MAF <0.01. Of these, 88 variants had a maximum SpliceAI ∆score above 0.8 (Table S1). There were no homozygotes reported for any of the 88 MME putative splicing variants. The majority of these predicted splicing variants were predicted to directly change either the canonical splice donor sites (26/88) or canonical splice acceptor sites (30/88) of MME, including an inframe deletion (c.1317_1317+2del) and a frameshift variant (c.957+1del). An additional nine variants were predicted to affect a splicing region, including one which is also annotated as a synonymous variant (c.1188G>A; p.Lys396Lys). Nine missense variants (p.Asp209Gly, p.Ile217Ser, p.Glu282Val, p.Arg365Ile, p.Ser436Gly, p.Asp533Gly, p.Ile553Val, p.Val554Phe, p.Gln692Arg) and a synonymous variant (p.Gly417Gly) were also predicted to alter splicing. Twelve variants were annotated as ‘intron variants’, of which two could be considered ‘deep intronic’ variants (c.1188+428A>G, described in this manuscript, and c.197-9871A>G). Interestingly, further analysis using SpliceAI demonstrated that MME c.197-9871A>G was predicted to create a splice donor site (Score 0.81) and strengthen an upstream cryptic splice acceptor site (Score 0.81) in a similar manner to c.1188+428A>G, thereby possibly creating an in-frame 96 bp pseudoexon.

4 DISCUSSION

Here we report a deep intronic variant, MME c.1188+428A>G causing recessive CMT2T in two unrelated Australian families. This variant results in a pseudoexon that likely leads to NMD of the MME transcript, resulting in a loss-of-function. This is in keeping with previously reported MME variants which have broadly been characterized as ‘loss-of-function’ variants. To our knowledge, this is the first report of a pathogenic deep intronic variant in MME. Given that a significant portion of patients with axonal CMT remain genetically undiagnosed,^13-17 it's possible that deep intronic variants in MME explain a portion of this diagnostic gap.

Individuals from Family 1 previously underwent both diagnostic WES and research WES, and the proband in Family 2 previously underwent targeted gene panel testing (PathWest neuro v3). These testing methods did not capture deep intronic regions and returned negative results. However, deep intronic regions have previously been shown to have a higher prevalence of variants than coding regions and canonical splice sites.⁷¹ Our findings suggest that intronic SNPs should be analyzed to determine if they impact splicing before they are dismissed as benign. Detection and functional validation of deep intronic variants has previously been shown to increase diagnostic rates in other Mendelian diseases including X-linked Alport syndrome,⁷² inherited retinal disorders,^{73, 74} and dystrophinopathy.^{75, 76} Pathogenic deep intronic variants, such as described here, are likely to be underreported amongst CMT-causing genes due to a lack of detection by WES and targeted gene panels, and a lack of functional investigation upon detection.

The MME c.1188+428A>G variant was reported in multiple different genetic ancestry groups in gnomAD v.4.0.0,⁶⁷ including in the European (Non-Finnish) (8/68008), African/African American (1/41432), and ‘remaining' (1/2092) ancestry groups. This was also reflected in the ‘All of Us’ Research Program,⁶⁸ which reported the MME c.1188+428A>G variant in the African (2/107888) and European populations (33/256804). Whilst it is possible that MME c.1188+428A>G represents a recurrent de novo variant, it is also likely that this variant has persisted at low levels in the global population. As this variant is missed by WES and may not be prioritized by variant-filtering approaches focusing on coding variants, this variant may therefore represent an underappreciated cause of recessive CMT. This is further supported by its detection in two Australian CMT families who are not known to be related.

Whilst the predicted splicing variants reported in MME are individually rare, we have described 88 variants in gnomAD that are predicted by SpliceAI to alter splicing. Nine of these variants were annotated as missense variants, and one was a synonymous change. This is of note as these ‘missense’ variants are often assumed to result in a single amino acid change, and synonymous variants are often considered functionally neutral, with their effect on splicing typically not assessed.²⁰ Our results suggest that the discovery of any of these 88 variants in a homozygous or compound heterozygous state in an individual with CMT should prompt further functional investigation of their effect on splicing of the MME transcript.

Prior to functional validation, MME c.1188+428A>G was considered a variant of uncertain significance (VUS) according to American College of Medical Genetics and Genomics (ACMG) criteria⁷⁷ (BP4, PM2, PM3, PP1), However, the functional evidence generated by the splicing assay now allows for the addition of PVS1 (null variant in a gene where loss-of-function is a known mechanism of disease) and PS3 (well-established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product) criteria. This allows reclassification of the variant as ‘pathogenic’. Therefore, this work demonstrates the importance of functional validation to confirm the effect of candidate variants on splicing to increase diagnostic rates for those with inherited disease.

Previously described biallelic MME patients typically have a phenotype consistent with late-onset axonal neuropathy. In contrast, three of the four affected siblings in Family 1 self-described childhood onset and the proband of Family 2 self-described symptom onset in early adulthood. This may reflect an expansion of the phenotype associated with biallelic MME variants in CMT2T. It has also previously been reported that heterozygous MME variants can cause spinocerebellar ataxia type 43 (SCA43)⁷⁸ as well as autosomal dominant CMT.³ However, the individuals in Family 1 who were MME c.1188+428A>G heterozygotes were all clinically assessed as neurologically normal, including an individual who is in their seventh decade. Homozygous affected individuals in Family 1 were noted to have a sensory ataxia rather than cerebellar ataxia, although MRI brain studies were not conducted. Therefore, our findings here do not support a role for MME c.1188+428A>G to cause SCA43 or autosomal dominant CMT, and further expand the phenotype of recessive CMT2T.

Understanding the specific effects of splice-affecting variants is crucial for developing potential therapeutic strategies. Antisense oligonucleotides (ASOs) that modulate splicing are an active area of research, including individualized approaches to treat rare genetic diseases.^79-81 Several antisense nucleotides that modify splicing have been approved by the United States Food and Drug Administration and have resulted in marked improvements in clinical outcomes in those with genetic diseases.^{79, 82-91} An FDA-approved ‘n-of-1’ ASO, milasen, successfully blocked pathogenic pseudoexon inclusion in MFSD8 in a patient with neuronal ceroid lipofuscinosis type 7.⁹¹ ASOs which block pathogenic pseudoexons, such as that created by MME c.1188+428A>G, have also been described in in vivo preclinical models.^{79, 91-95} As such, individuals with the MME c.1188+428A>G variant may represent a form of CMT that is treatable through personalized ASO therapy and warrants further investigation.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.