Pathomechanisms of Monoallelic variants in TTN causing skeletal muscle disease

Abstract

Pathogenic variants in the titin gene (TTN) are known to cause a wide range of cardiac and musculoskeletal disorders, with skeletal myopathy mostly attributed to biallelic variants. We identified monoallelic truncating variants (TTNtv), splice site or internal deletions in TTN in probands with mild, progressive axial and proximal weakness, with dilated cardiomyopathy frequently developing with age. These variants segregated in an autosomal dominant pattern in 7 out of 8 studied families. We investigated the impact of these variants on mRNA, protein levels, and skeletal muscle structure and function. Results reveal that nonsense-mediated decay likely prevents accumulation of harmful truncated protein in skeletal muscle in patients with TTNtvs. Splice variants and an out-of-frame deletion induce aberrant exon skipping, while an in-frame deletion produces shortened titin with intact N- and C-termini, resulting in disrupted sarcomeric structure. All variant types were associated with genome-wide changes in splicing patterns, which represent a hallmark of disease progression. Lastly, RNA-seq studies revealed that GDF11, a member of the TGF-β superfamily, is upregulated in diseased tissue, indicating that it might be a useful therapeutic target in skeletal muscle titinopathies.

Titin, TTNtv, TTN deletion, Dominant titinopathy, titinopathy pathomechanisms

Introduction

Titin is the largest known human protein and a critical component of the sarcomere, providing passive stiffness to striated muscle by acting as a molecular spring [1]. The human titin gene (TTN) comprises 364 exons (363 exons in the meta transcript) and produces a range of different cardiac and skeletal isoforms through extensive alternative splicing of its molecular spring region [2]. Pathogenic variants in TTN are known to cause an increasingly broad range of cardiac, skeletal and cardioskeletal disorders with onset across the lifespan [3].

Currently, three autosomal dominant titinopathies are recognized: tibial muscular dystrophy (TMD), hereditary myopathy with early respiratory failure (HMERF), and TTN-related dilated cardiomyopathy (TTN-DCM). TMD is a late-onset distal myopathy caused by an 11 bp insertion/deletion in exon 363 [4]; HMERF is a proximodistal myopathy with respiratory involvement, typically caused by my missense changes in exon 344 [5]. Until recently, all other skeletal muscle titinopathies were considered to be recessive [6]. However, recent reports have shown dominant inheritance in several families with TTN deletions [7, 8] and skeletal muscle involvement has been reported in patients with dilated cardiomyopathy due to monoallelic TTNtvs [9].

Titin’s N-terminus is embedded into the Z-disk of the sarcomere and is followed by an elastic I-band region that functions as a molecular spring. The C-terminally located M-band is thought to have a signalosome function through its kinase domain and various protein interactions [10]. Located in between titin’s I- and M-bands, titin’s A-band segment is important for axial alignment along the thick filament and regulation of thick filament length [11, 12]. Monoallelic pathogenic A-band variants have frequently been associated with cardiomyopathy [13] and have much higher odds of causing disease than pathogenic variants in the Z-disk or I-band [14]. This can be explained by the lack of alternative splicing in the A-band region and its presence in most major titin isoforms [15]. Pathogenic variants include nonsense-, frameshift- and canonical splice site variants, and possibly missense variants [16]. Nonsense- and frameshift variants can give rise to premature termination of translation and production of truncated peptides. Premature stop codon formation can also be caused by splicing variants through non-canonical splice site usage or intron retention [17]. Alternatively, splice site variants can lead to production of shortened titin isoforms if alternative splice site usage produces in-frame transcripts through partial or complete skipping of exons. Exon skipping as a result of a splice donor variant has been demonstrated in skeletal muscle of titinopathy patients [18].

Current knowledge of TTN molecular disease mechanisms is primarily based on studies of cardiac tissue. Although truncating variants (TTNtv) are the leading genetic cause of dilated cardiomyopathy (DCM), its underlying mechanism of disease is not well resolved and has not been investigated thoroughly in skeletal muscle. Proposed pathways leading to cardiac disease include haploinsufficiency, accumulation of truncated “poison peptides” and perturbation of autophagy and cardiac metabolism [14, 19, 20]. In recent studies, Fomin et al. [21], McAfee et al [22], and Kellermayer et al [23] demonstrated the presence of truncated titin protein in a subset of end-stage DCM hearts, with some evidence for incorporation of truncated titin isoforms into the sarcomere [23, 24]. Experiments on human induced pluripotent stem cell-derived cardiomyocytes containing TTNtvs do not support that truncated proteins are integrated into the sarcomere, but that their cytosolic accumulation impairs protein quality control pathways. Thereby, a poison peptide effect might play a role as a pathogenic mechanism [19, 25], in addition to haploinsuffiency caused by reduced translation and altered protein degradation. The presence of truncated titin protein has been shown to increase intracellular aggregate formation, which can be counteracted by inhibition of the ubiquitin-proteasome-system (UPS, 13). It has also been suggested that the autophagy-lysosomal pathway plays an important role in clearing TTNtv aggregates when the UPS is saturated by high expression of TTNtvs [25].

Although nonsense-mediated mRNA decay (NMD) typically causes degradation of aberrant mRNAs harboring premature termination codons (PTCs) [26], reduction of truncating titin transcripts has previously only been demonstrated in rat models [14] and in a subset of hearts of end-stage DCM patients [21]. Reduction of total titin transcript levels as a consequence of mutant allele degradation has so far not been detected in human patients. Furthermore, NMD is not considered to be a major contributor to haploinsufficiency because significant allelic imbalance has not been observed in TTNtvs in the heart [22], especially for TTNtvs in the A-band region [27].

NMD is a translation-dependent process which needs the central factor UPF1 to function, in addition to proteins of the exon-junction complex and release factors [28]. Apart from its basal role in degrading aberrant RNAs, UPF1 also acts as an E3 ubiquitin ligase to repress human skeletal muscle differentiation [29]. NMD can also target endogenous genes for degradation, for example splice factors of the serine-arginine (SR) rich family. These SR proteins act as splice factors [30] and can also activate NMD by including PTC-containing exons in their RNA targets. Because alternative splicing and NMD are frequently coupled [31], both TTNtvs and splice variants could trigger disease through either pathway.

In order to investigate the mechanism by which monoallelic TTN variants result in skeletal muscle disease, we performed clinical, genetic and molecular analyses in a cohort of patients with mild, progressive skeletal myopathy with monoallelic truncating, splice, or deletion variants in TTN. We characterized gene expression, alternative splicing, protein levels and structural and mechanical features in skeletal muscle of 11 affected patients. Our results show dominant transmission of this previously unrecognized titinopathy phenotype, and illuminate the underlying mechanism of disease, with implications for diagnosis, genetic counseling, and potential therapy.

Results

Clinical and genetic studies

Review of case medical records identified a pattern of mild but progressive axial and proximal or proximodistal weakness, with frequent complaints of myalgia and fatigue. Muscle weakness was often first noted in childhood, but most cases did not present for medical evaluation until adulthood. Creatine kinase (CK) levels were normal to moderately elevated, and electromyogram (EMG) was normal or indicated non-irritable myopathy. Muscle biopsy findings were nonspecific. Many cases developed dilated cardiomyopathy in middle age, often preceded by atrial fibrillation, tachycardia, bradycardia or other cardiac symptoms. The affected brother of case 5 required heart transplant at the age of 39. Case clinical features are summarized in Table 1.

Table 1

Summary of clinical data.

ID	Age Sex	TTN Variant, (band, exon)	ACMG Class	Pattern of Weakness	CK	EMG	Muscle biopsy	Cardiac Phenotype
1	53 M	c.29024C > A (p.Ser9675*) (I, exon 100)	VUS	DUL, PUL, PLL, DLL, neck extensor, facial	154	Non-irritable myopathy	Myopathic and dystrophic changes with fibrosis, fatty replacement, internal nuclei	Afib (30s), low normal EF, LV noncompaction [48]
3	62 M	c.76717C > T (p.Arg25573*) (A, 326)	LP	PLL, DLL, facial. decreased muscle bulk in the calves.	386, 276	Non-irritable myopathy, cramp fasciculation	n/a	SVT, Afib (51y), NICM, ICD (60y)
5	30 M	c.96076_107490del, (A-M, exons 346–362, in frame)	P	PLL, DL, neck flexor, decreased muscle bulk.	n/a*	n/a*	n/a*	Bradycardia. Mild LV dilatation with global hypokenesis, LVEF 46%
7	49 M	c.79793 T > G (p.Leu26598*), (A, 326)	LP	PLL, myalgia	1351	Normal	n/a*	Afib, DCM, LVEF 30%
8	49 F	c.31426 + 1G > C, (I, intron 117, splice donor)	VUS	PLL	220	Chronic denervating neurogenic process	n/a	Normal
11	52 F	c.77646_77662delinsAGA (p.Ile25883Aspfs*3), (A, exon 326)	P	PLL, DLL	600–1200	Myopathic	internal nuclei, rounded fibers, some moth-eaten fibers	Normal
15	50 F	c.44816-1G > A (I, intron 242 Splice acceptor)	P	PLL, neck flexor	normal	Normal	n/a	DCM (52y), LVEF 35% (improved to 45% with Rx)
16	47 F	c.44816-1G > A (I, intron 242 Splice acceptor)	P	PLL, neck flexor, myalgia	normal	n/a*	n/a*	DCM (45y), LVEF 45%
18	22 F	c.44816-1G > A (I, intron 242 Splice acceptor)	LP	DUL, DLL, neck flexor	n/a	n/a*	n/a*	Tachycardia. Low normal systolic function, LVEF 54%
22	48 F	c.71583_96801del (A, deletion exons 326–347, out of frame)	LP	PUL, PLL, neck flexor	normal	n/a*	n/a*	Arrhythmia, LVEF 55%, slightly enlarged LV
23	38 F	c.71583_96801del (A, deletion exons 326–347, out of frame)	LP	PUL, DUL, PLL, DLL, neck flexor	normal	Normal	occasional mild atrophic muscle fibers	PFO, Arrhythmia (23y) requiring ablation, LVEF 50–55%

Note: All variants correspond to transcript NM_001267550.2. Age, age at time of biopsy. Abbreviations: M, male; F, female; LP, likely pathogenic; P, pathogenic; PUL, proximal upper limb weakness, DUL, distal upper limb weakness; PLL, proximal lower limb weakness; DLL, distal lower limb weakness; LV, left ventricular; LVEF, left ventricular ejection fraction; SVT, supraventricular tachycardia; DCM, dilated cardiomyopathy; Afib, atrial fibrillation; ICD, implantable cardiac defibrillator; Rx, medical treatment, PFO, patent foramen ovale. ^*study performed on family member with same genotype, see text. Patients 12 and 19 are not listed because they were included as disease controls.

Review of family history data revealed apparent autosomal dominant transmission of cardioskelatal or skeletal muscle disease in 7/8 pedigrees (see Fig. 1). In the family of case 5, the TTN deletion was demonstrated to be de novo in the case’s father. The family of cases 15, 16, 18, showed evidence for dominant transmission of a mild, cardioskeletal phenotype associated with the monoallelic splice variant, with a more severe phenotype associated with biallelic TTN variants in case 15’s son. Alphamissense scoring and segregation analysis of identified secondary TTN variants revealed likely recessive inheritance in Case 1 and confirmed dominant inheritance in the remainder of cases (see Table 2).

Case 1

Case 1 (Fig. 1A, II-2) had a history of muscle weakness since childhood which worsened in his 30s with foot drop and falls. He developed atrial fibrillation in his 30s. Physical examination at age 53 identified proximodistal weakness, scapular winging, facial weakness, and bilateral foot drop with a Trendelenburg gait. Muscle biopsy revealed chronic myopathic and dystrophic changes with fibrosis, fatty replacement, internalized nuclei, and core-like areas on NADH staining. CK was 154. EMG identified non-irritable myopathy. Echocardiogram showed low normal ejection Fraction (EF); cardiac MRI revealed left ventricular (LV) noncompaction. Muscle MRI revealed symmetric fatty replacement/infiltration of the muscles of the body wall, paraspinals, buttocks, thighs and legs.

He was found to have a nonsense variant in exon 100 of the I-band of TTN, c.29024C > A (p.Ser9675^*) on neuromuscular panel testing, initially interpreted as a variant of unknown significance (VUS) but recently reclassified to likely pathogenic [32]. No other family members were reported to have muscle weakness or cardiac disease.

The p.Ser9675^* variant was also identified in his brother but not present in his mother or two daughters. Review of sequence data from Case 1 identified the presence of p.Val31745Gly, (distal tandem Ig segment, exon 242), absent from Gnomad, in trans with p.Ser9675^*. This missense variant has a pathogenicity score of 0.901 in Alphamissense [17], indicating high pathogenic potential (Table 2). The p.Val31745Gly was present in both his daughters. Inheritance pattern and pathogenicity scores are consistent with a recessive inheritance in this case.

Case 3

Case 3 (Fig. 1B, II-6) developed proximal-predominant muscle weakness, fasciculations, and cramping at age 49, leading to a diagnosis of limb-girdle muscular dystrophy. At age 55, he was diagnosed with DCM with an LVEF of 40%–45% and nonischemic fibrosis and myocardial edema/inflammation on MRI. An implantable cardiac defibrillator (ICD) was placed at age 60. CK was measured at 386 and 276. EMG identified short duration motor unit action potentials. He was found to have a nonsense variant in exon 326 of the A-band of TTN (c.76717C > T; p.Arg25573^*) on cardiomyopathy panel testing, interpreted as likely pathogenic. Review of sequence data identified no rare secondary variant in TTN. Two of the proband’s sisters were also identified to have the p.Arg25573^* variant, one of whom was diagnosed with DCM at 38 (II-3). An older, untested brother (II-1) died at 68 from sudden cardiac death. His mother (not tested) developed cardiac arrhythmia in her 60s and died from congestive heart failure at age 88.

Figure 1

Pedigrees of titinopathy cases included in this study. Studied cases are indicated by arrows. Individuals with known symptoms are shaded blue (for skeletal muscle) or red (for cardiac muscle). Individuals who underwent genetic testing are denoted with a ‘+’ to indicate the TTN variant was found, or a ‘-’ to indicate the TTN variant was not found. Individuals not marked with ‘+’ or ‘-’ were not available for testing. Cases 12 and 19 are not listed because they were included as disease controls.

Table 2

Secondary TTN variants(^*) identified in TTNtv, splice variant, and deletion cases.

ID	Primary TTN Variant, (band, exon, ACMG interpretation)	Secondary Variant(s)(**) (band, exon, ACMG interpretation)	Secondary Variant gnomAD Frequency	MAF Lower Confidence Interval Limit(***)	Alpha-missense Pathogenicity Score(****)	Phase to Primary Variant	Segregation	Phenotype	Suspected pattern of transmission of phenotype
1	c.29024C > A (p.Ser9675*) (I band, exon 100, LP)	c.95234 T > G (p.Val31745Gly) (A band, exon 242, VUS)	absent	not present in gnomAD	0.901 (pathogenic)	trans	Unaffected daughters do not have (p.Ser9675*), both have p.Val31745Gly	cardioskeletal	recessive
3	c.76717C > T (p.Arg25573*) (A band, exon 326, P)	None identified	n/a	n/a	n/a	n/a	n/a	cardioskeletal	dominant
5	c.96076_107488del, (Bands A-M, exons 346–362, in frame, P)	c.81617 T > C (p.Ile27206Thr) (A band, exon 326, VUS)	0.01443%	0.0009546%	0.1256 (benign)	trans	Affected father and 2 affected brothers have c.96076_107488del, none have p.Ile27206Thr or c.11312-5174del	cardioskeletal	dominant
		c.11312-174del (Intron 47, VUS)	absent	not present in gnomAD	n/a	trans
7	c.79793 T > G (p.Leu26598*), (A band, exon 326, LP)	None identified	n/a	n/a	n/a	n/a	n/a	cardioskeletal	dominant
8	c.31426 + 1G > C, (I band, intron 117, splice donor, LP)	c.51377A > G (p.Lys17126Arg) (Exon 271, VUS)	0.001612%	0.002358%	0.1074 (benign)	trans	Affected mother has c.31426 + 1G > C, does not have p.Lys17126Arg	skeletal	dominant
11	c.77646_77662delinsAGA (p.Ile25883Aspfs*3), (A band, exon 326, P)	None identified	n/a	n/a	n/a	n/a	n/a	skeletal	dominant
15	c.44816-1G > A (I band, intron 242 Splice acceptor, P)	c.13262A > G (p.Asn4421Ser) (Exon 48), VUS	0.01464%	0.02052%	0.1281 (benign)	trans	Affected sister and niece have p.Lys3286Arg (in cis with c.44816-1G > A), neither has p.Asn4421Ser	cardioskeletal	dominant and recessive
		c.9857A > G (p.Lys3286Arg) (Exon 42), VUS	0.009914%	0.009914%	0.2376 (benign)	cis
16		c.9857A > G (p.Lys3286Arg) (Exon 42), VUS	0.009914%	0.01376%	0.2376 (benign)	cis	Affected sister and daughter have p.Lys3286Arg (in cis with c.44816-1G > A)	cardioskeletal	dominant and recessive
18		c.9857A > G (p.Lys3286Arg) (Exon 42), VUS	0.009914%	0.01376%	0.2376 (benign)	cis	Affected mother and aunt have p.Lys3286Arg (in cis with c.44816-1G > A)	skeletal	dominant and recessive
22	c.71583_96801del (A band, deletion exons 326–347, out of frame, LP)	None identified	n/a	n/a	n/a	n/a	Affected sister and mother carry c.71583_96801del	cardioskeletal	dominant
23		c.82448A > G (p.Lys27483Arg) (exon 326), VUS	absent	not present in gnomAD	0.0755 (benign)	trans	Affected sister and mother carry c.71583_96801del, neither carry p.Lys27483Arg	cardioskelatal	dominant

Patients 12 and 19 are not listed because they were included as disease controls.

a

All variants correspond to transcript NM_001267550.2. Abbreviations: VUS, variant of uncertain significance; LP, likely pathogenic; P, pathogenic, per ACMG variant interpretation criteria.

b

A secondary variant is defined as a variant with a minor allele frequency of < 0.038%, with uncertain significance per ACMG criteria.

c

Because subpopulation cohorts in gnomAD vary significantly in size, 95% confidence intervals around the reported allele frequencies were used for calculations, rather than raw allele frequency (AF). The lower limits shown indicate 95% confidence that the true population AF may be as rare as the displayed value.

d

Cheng et al., 2023. All alpha missense prediction relative to Protein PB Q8WZ42.

Case 5

Case 5 (Fig. 1C, III-3) had facial weakness, neck flexor weakness, proximodistal weakness, and decreased muscle bulk on examination at age 30. Echocardiogram at age 33 showed LVEF of 46% with global hypokinesis. His father, the original proband (II-2, Fig. 1c) reported muscle weakness since childhood, with difficulty running and lifting, and eventually developed a unique gait marked by profound external rotation of bilateral hips. He had facial weakness, normal CK, and a myopathic EMG. He was diagnosed with DCM and died from heart failure at 54 years old. Case 5 has two affected brothers: one (III-4) with childhood-onset of muscle weakness (proximal>distal) with low ventricular systolic function (LVEF 50%) at age 19, and another (III-1) with dilated cardiomyopathy and chronic systolic heart failure requiring heart transplant at age 39. All affected family members were initially found to carry a 16.430 kb deletion spanning part of the A- and M-bands of TTN (c.96076_107488del), classified as pathogenic. The deletion was determined to be de novo in the father of Case 5. During the course of this study, the deletion was found to be in frame and spans 2 more bases than initially reported (c.96076_107490del, see Supplementary Material, Fig. S8). Review of sequence data from case 5 identified two rare variants in TTN (p.Ile27206Thr, Alphamissense score 0.1256 (benign) and c.11312-5174del), neither of which were present in his affected father or two affected brothers (see Table 2). This pedigree has been previously reported in detail.

Case 7

Case 7 (Fig. 1D, III-4) developed non-ischemic cardiomyopathy and proximal upper extremity weakness at age 38. He had previously been discharged from military service due to inability to perform push-ups and CK elevation (1000). His father (II-3, Fig. 1c) was diagnosed with limb-girdle muscular dystrophy at age 60 and subsequently DCM with LVEF of 30% and atrial fibrillation at age 73. His CK was 200 and muscle biopsy showed fiber size variability, internal nuclei and occasional split fibers. Case 7’s daughter (IV-4) has proximal muscle weakness and tachycardia. Each of these family members was found to have a nonsense variant in exon 326 of the A band of TTN, c.79793 T > G, p.Leu26598^*, interpreted as likely pathogenic [32], first identified in case 7’s father. Review of sequence data from case 7 identified no rare secondary variant in TTN. Case 7 also has a family member (IV-1) who died at 25 from uncharacterized muscular dystrophy (no information on cardiac status), as well as another distant relative (V-3) with an uncharacterized form of muscular dystrophy and no cardiac involvement. Medical records on these individuals are not available.

Case 8

Case 8, (Fig. 1E, II-2) now 49 years old, developed proximal weakness in her late 20s which progressed leading to falls and difficulty climbing stairs. CK was 220; EMG identified a chronic denervating neurogenic process. Echocardiogram was normal. Her mother (I-2) also had proximal weakness; two sons are not affected. Neuromuscular panel testing identified a variant in intron 117 of the I-band of TTN, c.31426 + 1G > C (splice donor), initially interpreted as VUS but recently reclassified as likely pathogenic, also present in her mother. Review of sequence data in case 8 identified a rare missense change, p.Lys17126Arg, Alphamissense score 0.1074 (benign), not present in her affected mother. See Table 2.

Case 11

Case 11 (Fig. 1F, II-1), now 52 years old, developed progressive proximal muscle weakness in her 40s. She had a history of myalgia with teen onset. CK was 600–1200. EMG showed spontaneous activity with myopathic potentials. Echocardiogram was normal. Her sister (II-3) also has muscle weakness, and her son (III-2) was discharged from military service due to CK elevation and inability to exercise. Neuromuscular panel testing identified a frameshift variant in exon 326 of the A band of TTN, c.77646_77662delinsAGA (p.Ile25883Aspfs^*3), interpreted as pathogenic. The p.Ile25883Aspfs^*3 TTN variant was also identified in her affected sister (II-3), and 14 year old son (III-4) and 12-year-old nephew (III-5) with no current symptoms. Son III-2, with CK elevation, declined genetic testing. Review of sequence data from case 7 identified no rare secondary variant in TTN.

Cases 15, 16, and 18

Cases 15, 16, and 18 are members of the same family (Fig. 1G, individuals II-3, II-4, and III-2, respectively). Case 15 first noted muscle weakness after the birth of her son, with myalgia and difficulty walking or standing for long periods. On examination at age 44, she had weakness in the hip flexors and neck. CK and muscle MRI was normal; pulmonary function was decreased when supine. She was diagnosed with DCM at age 52. Case 16 (sister of case 15) had myalgia, hip and neck flexor weakness on examination at age 45. Muscle MRI identified mild fatty infiltration of the paraspinals. She developed DCM at age 45. Case 18 (daughter of case 16) has mild neck flexor and proximodistal weakness with foot drop at age 22, with tachycardia and low normal systolic function, LVEF 54%. The mother of cases 15 and 16 (I-2) had onset of arrhythmias in her 40s requiring ablation. On examination at age 69, she had neck flexor and proximal leg weakness, with fibrofatty replacement in the paraspinal and gastrocnemius muscles. Post-mortem muscle biopsy identified internal nuclei and fiber size variability.

Each of these family members were identified to carry an I-band splice variant, c.44816-1G > A, intron 242 (Splice acceptor), interpreted as pathogenic. Review of sequence data from cases 15, 16, and 18 revealed that all three carried a rare missense change, p.Lys3286Arg (Alphamissense score 0.2376, benign), determined to be in cis with the c.44816-1G > A splice variant. In addition, case 15 carried p.Asn4421Ser (Alphamissense score 0.1281, benign) in trans with the splice variant, but her affected sister and niece did not carry it (Table 2).

Case 15’s son (III-1) has centronuclear myopathy as the result of biallelic c.44816-1G > A/c.40558G > C TTN variants and was previously reported [33]. His father, II-2, also has the c.40558G > C variant and no known skeletal muscle weakness or cardiac symptoms; his paternal uncle, II-1, also has the c.40558G > C variant and atrial fibrillation.

Cases 22, 23

Cases 22 and 23 are members of the same family (Fig. 1H, II-1 and II-2). Case 22 had proximal leg and hip flexor weakness with an abnormal gait marked by excessive pelvic rotation on examination at age 48. Echocardiogram showed mild LV enlargement. Case 23 developed arrhythmia requiring ablation after childbirth at age 23. She experienced multiple transient ischemic attacks in her early 30s (attributed to a patent foramen ovale which was repaired). She noted muscle weakness later in her 30s with difficulty lifting her arms and carrying objects. On examination at age 38, she had proximodistal and neck flexor weakness and pectus excavatum. Echocardiogram showed borderline LVEF at 50%–55%. CK was normal; muscle biopsy showed mild variation in fiber diameter with occasional atrophic fibers. Their mother, (I-2) has proximal muscle weakness, and DCM with chronic systolic heart failure (LVEF 41%). Case 22’s teen son (III-2) has a bicuspid aortic valve. Each of these family members were found to carry an out-of-frame partial deletion of part of exon 326, exons 327–346, and part of exon 347 in TTN, (c.71583_96801del), interpreted as likely pathogenic. Review of sequence data from case 22 identified no rare secondary variant in TTN. Case 23 carried a rare missense change, p.Lys27483Arg (Alphamissense score 0.0755, benign) in trans with the deletion, but her affected sister and mother did not carry it (Table 2).

Titin transcript levels and splicing

To analyze gene expression in biopsy samples, RNA sequencing studies were performed. Initially we focused on how TTNtvs and splice site variants impact TTN transcript levels and individual exon inclusion. TTN reads were normalized to nebulin, a major skeletal muscle protein and a robust myofiber marker. Truncating variants had significantly reduced amounts of titin transcript compared to controls (P = 0.0024, Fig. 2A). Given the established association of TTNtvs with nonsense-mediated mRNA decay (NMD), allelic balance was assessed for all samples (Fig. 2B). Similar to TTN transcript levels, allelic balance ratios were significantly reduced in truncating variants, suggesting nonsense-mediated decay as a mechanism to degrade transcripts carrying these pathogenic variants. In contrast, all splice-site variants had normal transcript levels and allelic balance ratios (Table 3). Reduced TTN transcript amounts without allelic imbalance were observed in TTN^del346–362 (in-frame deletion, case 5) and TTN^HMERF (case 12). Unexpectedly, we observed contrary results for two siblings with an identical large out-of-frame deletion (TTN^del326–347, cases 22 & 23). Allelic imbalance and normal transcript levels were detected in one sibling (TTN^{del326–347-1}, case 22), while the other sibling had reduced titin levels along with a strongly reduced allelic balance ratio. This difference might be explained by a difference in cell type composition between the two samples. TTN^{del326–347-1} had very low levels of myofiber markers, and an increased content of adipocyte and fibroblast markers (Supplementary Material, Fig. S3, group titinopathy B) compared to TTN^{del326–347-2}.

Figure 2

Titin transcript levels, allelic balance and exon inclusion levels. Titin transcript levels, determined by RNA sequencing analysis and normalized to nebulin levels (A). Truncating variants displayed reduced transcript levels, while splice site variants were similar to controls. Analysis of biallelic sites confirmed a shift in the allelic balance ratio for this group of variants (B). Numbers represent p-values from unpaired t-tests. Exon 117 was skipped in almost half TTN^splice117 transcripts (C), while only low percentages of exon skipping (1.5%–7.5%) were observed in TTN^splice242 (patients 15, 16 & 18, D). In-frame deletion TTN^del346–362 had the affected exons missing in half of the transcripts (E). Two patients with a large out-of-frame deletion (TTN^del326–347) showed low degrees if exon skipping over the deletion site (F). One PEVK-domain is absent in TTN^splice117 due to exon skipping (G) while the last C-zone super-repeat and most of the M-band (except for M10) are deleted in TTN^del346–362 (H).

Table 3

Titin transcript levels and allelic balance ratio.

Case	Group	Exon	Identifier	AB Ratio	ttn/neb
1	truncating	Exon 100	TTNstop100	0.23	3.81
2	control			0.49	7.22
3	truncating	Exon 326	TTNstop326–1	0.24	3.46
4	control			0.49	7.62
5	deletion	346–362	TTNdel346–362	0.46	3.97
6	control			0.48	5.87
7	truncating	Exon 326	TTNstop326–2	0.30	4.09
8	splice donor	Intron 117	TTNsplice117	0.48	6.50
9	control			0.42	7.02
10	control			0.49	6.70
11	truncating	Exon 326	TTNstop326–3	0.34	3.23
12	HMERF		TTNHMERF	0.49	3.87
13	control			0.49	5.80
14	control			0.49	5.03
15	splice acceptor	Intron 242	TTNsplice242–1	0.47	6.27
16	splice acceptor	Intron 242	TTNsplice242–2	0.43	6.99
17	control			0.48	6.88
18	splice acceptor	Intron 242	TTNsplice242–3	0.47	7.53
19	truncating/missense	Exon 358/Exon 363	TTNstop358-mis363	0.20	6.82
20	control			0.39	5.96
21	control			0.47	6.18
22	deletion	Exon 326–347	TTNdel326–347-1	0.47	8.00
23	deletion	Exon 326–347	TTNdel326–347-2	0.23	3.87

An observed shift in allelic balance correlated with a reduction of titin transcript, except for TTN^{stop358-mis363}. In an in-frame deletion and an HMERF patient (cases 5 and 12), lower amounts of titin transcripts were measured despite unchanged allelic balance.

In addition to total TTN mRNA levels, we also determined individual exon inclusion levels and performed a splicing analysis. While transcript levels in splice variants were normal, exon 117 was excluded in about half of TTN^splice117 transcripts (Fig. 2C, case 21). Instead of the mutated donor site, the majority of splice events were initiated from intron 116, and to a low degree from intron 115. This skipping event would not be expected to have an effect on titin translation because the reading frame would stay intact and only one PEVK domain would be skipped (Fig. 2G). In TTN^splice242 samples (cases 15, 16 & 18), we detected a low degree of exon 243 skipping (Fig. 2D) and frequent alternative splicing events downstream of the mutated site. Alternative splicing events start around exon 289 and increase towards titin’s C-terminus (Supplementary Material, Fig. S1B).

Splicing analysis revealed exclusion of exons 347 to 361 in about 50% of transcripts for TTN^del346–362 (in-frame deletion), which confirms monoallelic deletion of this region. This deletion was previously published as an out-of-frame change [8], but RNA data and Sanger sequencing confirmed the deletion to be in-frame (Supplementary Material, Fig. S8), resulting in a shortened titin protein lacking the final C-zone super-repeat and the M-band except for domain M10 (Fig. 2H). For TTN^{del326–347-1} (case 22), we determined a low degree of skipping from exon 326 to 347 (Fig. 2F). Lack of alternative splicing events in this region would most likely trigger nonsense-mediated decay of mutant transcripts, which was the case for TTN^{del326–347-2} (case 23). Interestingly, NMD could not be observed in TTN^{del326–347-1} (case 22, Table 3), but genome-wide splicing changes were prevalent and myofiber markers in this sample were very low (Figs. 3 and Supplementary Material, Fig. S3).

Figure 3

A subset of titinopathy samples shows differential splicing patterns. A: An initial analysis of differential splicing between three patient samples with an intron 242 splice acceptor variant [&, 15, 16, 18] and controls revealed a total of 1333 differentially spliced exons (p < 0.05 & PSI change > 5%). Together with two other patient samples [19 & 22], they form a distinct group with splicing patterns that are different from all other control & patient samples during column-wide clustering. Shown are inclusion levels (PSI) of differentially spliced exons, as determined during the initial analysis. B: One or more exons of genes known to be important for muscle development and function are differentially spliced in titinopathy patients (red). Controls are plotted in blue and titinopathy patients without genome-wide splicing changes in black. PSI values are plotted against exon numbers.

To identify other mechanisms that could potentially be disease-relevant in pathogenic variants, we analyzed splicing on a genome-wide level. Initially, we were particularly interested in splicing differences between intron 242 variants and controls, because the presence of multiple samples from the same family allowed significance testing of identical variants versus controls. This analysis revealed a total of 1333 differentially spliced exons. When this subset of exons was evaluated for splicing in our complete dataset, hierarchical clustering showed similar splicing patterns of TTN^splice242, TTN^{stop358-mis363} and TTN^{del326–347-1} (cases 15,16,18,19 & 22, Fig. 3A).

Significant splicing differences were detected in many genes that are essential for muscle structure and function, such as ANK2, DMD, TPM3 and PDLIM5, as well as transcriptional regulators like MEF2C and LRRFIP1 (Fig. 3B). Many other genes encoding for sarcomeric proteins were affected, some of which are known to be differentially regulated in muscular dystrophies (Supplementary Material, Fig. S2). Because these alternative splicing events were shared between different genetic variants (splice acceptor, truncating variant and large deletion), we investigated how these different variant types could lead to similar splicing patterns. Different isoforms are known to be cell-type specific, which led us to quantify known marker genes for different cell types. Due to significantly differential splicing patterns, we grouped titinopathy samples into two groups: Titinopathy A, which was more comparable to the control group, and Titinopathy B, in which we had determined differential splicing. This group had very low levels of marker genes for myofibers, myoblasts and satellite cells, while markers for adipocytes, fibro-adipogenic progenitor cells (FAPs), fibroblasts and immune cells were increased (Supplementary Material, Fig. S3). Differences in exon inclusion of sarcomeric genes might therefore reflect the different cell composition of biopsies.

To confirm that the observed splicing patterns reflect a disease state with loss of muscle fibers and fatty and fibrous tissue replacement, we determined genome-wide splicing patterns of another RNA-seq dataset obtained from patients with Facioscapulohumeral muscular dystrophy (FSHD, Supplementary Material, Fig. S4) [34]. The subset of titin patients with differential splicing patterns (cases 15, 16, 18, 19, 22) clustered with another subset of FSHD patients after splicing analysis (Supplementary Material, Fig. S4, first four columns), which confirmed that these splicing differences were disease-related, but not specific to a particular genetic variant or type of muscle disease. Instead, they are most likely a reflection of different disease states and degree of muscle degradation in a particular biopsy. Comparing differentially spliced exons between titinopathy and FSHD samples, there are distinct subgroups with PSI changes in both muscle diseases, but not in controls (Supplementary Material, Fig. S5). For titinopathy samples in this study, differential splicing did not correlate with observed disease severity during clinical evaluation.

Despite the observed heterogeneity of biopsy samples, transcript-specific gene expression analysis revealed consistent upregulation of GDF11, which has been suggested as a pathological effector of skeletal muscle, in titinopathy samples (Fig. 4A). GDF11 also represented the most significantly upregulated transcript comparing truncating- and deletion variants to controls (Fig. 4B). Because GDF11 is known to increase in mouse skeletal muscle with age, we compared GDF11 levels with age and did not find significant differences in age between patient- and control groups. While GDF11 levels slightly decreased with age in the control group, this correlation was not found in the patient population (Fig. 4C).

Figure 4

GDF11 levels are increased Titinopathy patients. Transcript-specific expression revealed increased levels of GDF11 in titinopathy patients (A). Sample grouping based on variant type revealed a high level of significance in truncating and deletion variants versus controls (B). Significance was calculated from transcript abundances and represents multiple-testing corrected p-values. GDF11 levels did not increase with age of the patient population and correlated negatively in controls (C). Average age did not differ between patients and controls (patients:45.45, controls: 45).

Titin protein levels

Gene expression analysis had demonstrated significantly reduced amounts of titin transcript in patients with TTNtvs. Therefore, we investigated whether this would also affect the amount of full-length titin protein and/or result in the presence of truncated titin proteins. For the majority of biopsy samples, we only detected a single major titin band which was comparable in migration distance to that of controls, thus, likely reflecting full-length titin (Fig. 5A). In-frame deletion TTN^del346–362 produced a full-length titin and a shortened titin protein (Fig. 5A, case 5), which would be expected to be 3805 amino acids smaller than control. Reactivity of antibodies specific for titin’s N- and C-termini demonstrated that both ends of titin are intact (Fig. 5B). Domains M8 and M9 were confirmed to be missing in protein from the deletion allele (Fig. 5C), as only regular-sized titin could be detected with an M8M9-specific antibody. The level of shortened titin protein was slightly lower than that of full-length titin in case 5 (Fig. 5D). Overall, full-length titin amounts were comparable between control and patient samples (Fig. 5E). Differences between controls and TTNtvs/deletions did not reach statistical significance (Fig. 5F).

Figure 5

Titin protein analysis. Titin protein was separated on agarose gels and detected with antibodies against its N-termini (Z1Z2, (A), top panels) and C-termini (M10, bottom panels). (A) Shortened TTN protein could only be detected in-frame deletion TTN^del346–362 (patient 5). Overlays of Z1Z2 and M10 images (B) demonstrate that no C-terminally truncated proteins are present at expected protein sizes (indicated by white boxes). Domains M8 and M9 are internally deleted in sample 5 and not detected in shortened titin with an M8M9-specific antibody (C) in contrast to domain M10 (A & B). Shortened titin is less abundant in sample 5 than full-length protein when quantified on agarose gels (D, values represent intensity). Protein levels of truncating and deletion variants (E) were not significantly changed from controls when normalized to MHC (F). Five biopsy samples had protein levels that were too low to detect by gels or Western blots (samples 8, 11, 16, 19 & 22).

There was no truncated protein detected in other titin variants. With elevated brightness and contrast, a few additional bands became visible (Supplementary Material, Fig. S6), but they are neither specific for one variant nor run at the expected size (indicated by white boxes). Additionally, C-terminally truncated protein would appear as a green band in overlay images. We were not able to detect any protein for cases 8,11,16,18,19 and 22 (TTN^splice117, TTN^{splice242–2}, TTN^{splice242–3}, TTN^{stop358-mis363} and TTN^{del326–347-1}). The latter four samples had also been found to have strongly reduced levels of myofiber markers compared to controls. We also did not observe small titin truncation products specific to pathogenic variants on Western blots of solubilized biopsy proteins run on 10% SDS gels (Supplementary Material, Fig. S7). Thus, in contrast to recent studies on DCM patients with TTNtvs in which truncated titin proteins were detected (see Introduction), truncated proteins were not detected in the present study on skeletal muscle samples.

Impact of pathogenic titin variants on muscle structure

Production of titin protein of different lengths, as observed in TTN^del346–362 (case 5), could impact muscle structure. Therefore, we examined muscle structure by electron microscopy. In the TTN^del346–362 patient we observed severe abnormalities in muscle structure (case 5, Fig. 6). Both Z-disks and M-lines were irregular, adjacent myofibrils were longitudinally displaced, which could be a result of integration of shortened and regular-sized titin into the sarcomere. This was not observed for truncating variants (cases 07 & 11) or splice donor variant TTN^splice117 (case 8), in which we did not detect any major differences from controls (cases 9 & 13).

Figure 6

TTN^del346–362 muscle shows abnormalities in muscle structure. Analysis of tibialis anterior muscle by electron microscopy revealed irregular Z-disks and M-lines from in-frame deletion TTN^del346–362 (case 5, first 4 panels), which produces shortened titin protein. Normal muscle structure was observed for two patients with truncating variants (TTN^stop326–2 and TTN^stop326–3), a patient with a splice donor variant (8,TTN^splice117) and controls [9 and 13].

Contractile measurements in single muscle fibers

In order to investigate whether the skeletal muscle weakness experienced by these patients originates at the myofilament level, we performed mechanical studies on membrane-permeabilized single muscle fibers. In five of the 13 examined patient biopsies we did not find enough usable skeletal muscle fibers as they primarily consisted of fat, blood vessels and fibrotic tissue with few fibers, indicating that loss of muscle fibers is a part of the pathology of these patients, a conclusion consistent with the RNA-seq studies. In the remaining biopsies mechanical experiments were successful. After these experiments the main myosin isoform in the fibers was determined by SDS-PAGE (see Fig. 7D for examples), making it possible to segregate results according to fiber type.

Figure 7

Passive mechanics in slow fibers from titinopathy patients. (A) Left panel: Ramp-stretch protocol for determining passive tension (force divided by fiber cross-sectional area) in skeletal muscle fibers. Fibers were set at slack sarcomere length and stretched in steps of 10% of fiber length (FL) at a speed of 1% FL/s and held for 20 s until the next step. Right panel: Passive force response in a control fiber from the ramp-stretch protocol. (B) Peak tensions plotted against sarcomere length for each step in the stretch protocol. (C) Peak tension at the sarcomere length range 3.0–3.12 μm (shaded area in D). (D) Cut out gel image of silver-stained 8% SDS-PAGE used to determine fiber types. Std; standard, Ctrl; control fiber, Pat-f#; patient fiber. Individual patient fibers are displayed. Controls are the average of 3–11 fibers per individual. ^*p < 0.05, and ^**P < 0.01 vs controls.

We measured passive tension (passive force divided by fiber cross-sectional area) by stretching the fibers with a step-based ramp-hold protocol (Fig. 7B, left panel) with the developed tension showing a characteristic pattern with higher peaks that then decay after each stretch (Fig. 7A and B). Most titinopathy patients had fibers with normal passive tensions, only TTN^del346–362 (case 5) had steeper passive tension-sarcomere length relationships for slow fibers at longer sarcomere lengths (Fig. 7B). At the sarcomere length range of 3.0 to 3.1 μm, slow fibers from these patients had higher passive tensions compared to slow fibers from healthy controls (Fig. 7C). All patients had normal passive tension in fast type 2 fibers (Supplementary Material, Fig. 9A and B).

Next, we activated the fibers with increasing [Ca²⁺] at a sarcomere length of 2.6 μm and measured the myosin-based force generation. A typical tension (force normalized to fiber cross-sectional area) recording is presented in Fig. 8A. Protocols for active stiffness (Fig. 8B) and rate of force redevelopment (Fig. 8C) are also shown. Maximal tension at saturating [Ca²⁺] was decreased in fast and slow fibers from TTN^splice117 (case 8), while slow fibers from TTN^HMERF (case 12) only produced 36% of the force of slow fibers from healthy controls (Fig. 8D and Supplementary Material, Fig. S9C). Only two fast fibers from this HMERF-patient were found and they did not show a force deficit. Calcium sensitivity was increased (p = 0.05) in slow fibers from TTN^splice117 (Fig. 8E). The Hill coefficient (nH) was decreased in slow fibers from TTN^HMERF (Fig. 8F). No differences in calcium sensitivity or Hill coefficient were seen in fast fibers from any titinopathy patient (Supplementary Material, Fig. S9D and E).

Figure 8

Membrane-permeabilized muscle fiber experiments reveal myofilament dysfunction in slow fibers from some individual titinopathy patients. (A) Left panel: Recording of tension developed by a slow skeletal muscle fiber at incremental [Ca2+]. Right panel: Tension-Ca2 + −curve for the fiber in A with EC50 ([Ca2+] for 50% tension) and Hill coefficient (nH) of curve shown. (B) Active stiffness protocol. Left panel: Relative length change; 0.3% of fiber length (FL) 500 Hz sinusoid for 20 ms was used to determine active stiffness. Right panel: Force trace for a fast and a slow fiber. Stiffness is calculated by dividing the force change with the length change. (C) Rate of force redevelopment protocol. Left panel: Relative length change; the fiber is shortened to 75% of FL in a ~ 1 ms step, held for 20 ms, stretched 103% of FL and immediately returned to 100% of FL. Right panel: Force trace for a fast and a slow fiber fitted with double exponential function. (D) Active tension (force normalized to fiber cross-sectional area). (E) Calcium sensitivity as EC50. (F) Hill coefficient (nH) is slope of the steep rise segment (black line in A, left panel) in tension-Ca²⁺-curve. (G) Active stiffness, an indicator of strongly-bound crossbridges. (H) Tension-stiffness ratio. (I) Rate of force redevelopment (k_tr). Control values are average of 5–21 fibers from 8 subjects without known neuromuscular disease. Individual fibers are shown for patients. One-way ANOVA with Dunnett’s multiple comparisons post hoc test was used for statistical testing. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 and ^****P < 0.0001 vs control.

To estimate the number of force-producing cross-bridges, we measured active stiffness (protocol in Fig. 8B). Slow fiber TTN^splice117 (case 8) and TTN^HMERF (case 12) had decreased active stiffness in both slow and fast fibers (Fig. 8G and Supplementary Material, Fig. S9F). Slow fibers from a splice site variant, TTN^splice117, and TTN^HMERF patient had an increased tension-stiffness ratio (Fig. 8H). This ratio was also increased in fast fibers from TTN^splice117 (Supplementary Material, Fig. S9C). These results suggest that force deficit in some titinopathy patients could be caused by a lower number of force-generating cross-bridges.

Finally, we measured the rate of force redevelopment (k_tr), an indicator of cross-bridge cycling kinetics as it reflects cross-bridge attachment and detachment rates [35]. The protocol is shown in fig. 8C (left panel) with the characteristic force response after a brief slack-restretch for a fast and slow fiber displayed in the right panel. k_tr was decreased in slow fibers from TTN^splice117 and TTN^HMERF (Fig. 8I). Also, fast fibers from TTN^HMERF had lower k_tr-values compared to fibers from healthy controls (Supplementary Material, Fig. S9H).

Discussion

We identified a cohort of patients with mild but progressive skeletal myopathy associated with monoallelic truncating, splice, or deletion variants in TTN. Family and genetic studies indicate that these variants segregate with disease via an autosomal dominant mode of transmission in seven out of eight studied pedigrees. The possibility of secondary variants in other genes cannot be ruled out as all cases underwent gene panel testing and not whole-genome sequencing. Review of sequence data, along with Alphamissense scoring and family studies, does not support biallelic pathogenic variants in TTN. Notably however, the TTN^splice242 variant was associated with dominant, mild cardioskeletal disease in the monoallelic state (in cases 15, 16, and 18), as well as a recessive centronuclear myopathy in the biallelic state (in case 15’s son, [33]). The in-frame deletion identified in case 5 was shown to be de novo in his affected father, with subsequent transmission to 3 affected sons with two different mothers, suggesting dominant inheritance.

Topf et al. [36] recently observed skeletal muscle disease in a cohort of myopathy patients with deleterious variants in SRPK3, which manifested only when co-inherited with TTN variants, suggesting digenic inheritance. SRPK3 encodes an X-linked serine/protein kinase which can phosphorylate the splice factor RBM20, thereby impacting titin on the post-transcriptional level and resulting in reduced titin levels when monoallelic titin variants are present. Notably, the cases presented in this study have a clear myopathy phenotype, yet titin protein levels and SRPK3 transcript levels are normal, unlike in the SRPK3/TTN myopathy cohort (Supplementary Material, Fig. S10). While we cannot eliminate the possibility of secondary and digenic inheritance in all studied families, the in-frame deletion (case 5) suggests a dominant mode of action through production of shortened titin, and we did not observe a reduction in titin protein.

Phenotypic features common to the majority of cases in our cohort include axial and proximal or proximodistal weakness, often first noted in childhood or young adulthood, becoming more pronounced with age. Most but not all cases also developed cardiac disease with age, including arrhythmias and dilated cardiomyopathy. Although the phenotypic presentations of these monoallelic truncating, splice, and deletion variants were similar, the underlying genetic mechanisms were found to be distinct.

T‌TVtvs are subject to NMD.

Titin transcript levels were strongly reduced in patients with TTNtvs. Allelic balance calculation suggests that nonsense-mediated degradation of the mutant allele might be causative for reduced transcript levels. However, in contrast to recent findings in digenic titinopathy cases [36], titin protein levels were found to be normal. This might be due to increased ribosomal translation speed or changes in protein turnover, possibly through activation of the MTOR1 pathway [14]. Except for the internal in-frame deletion (case 5), we were only able to detect full-length titin proteins. These finding differ from recent studies in cardiac tissue, where the favored disease mechanism of TTNtvs has been an accumulation of truncated titin proteins as “poison peptides” and haploinsufficiency due to differences in translational read-through or activity of protein degradation pathways [21].

The biopsies in this study likely represent relatively early skeletal muscle disease stages; NMD might be more prevalent during disease onset and decrease with its progression. This could explain why several large-scale studies from end-stage DCM hearts have not found evidence for allelic imbalance [15,22,27,37]. Likewise, this might be why two recent cardiac studies clearly demonstrated accumulation of truncated proteins for a subset of pathogenic titin variants [21, 22] while we were not able to detect truncated proteins in skeletal muscle from titinopathy patients. In a recent study of patients with TTN deletions [7], NMD was not detected in skeletal muscle biopsies, but most patients carried in-frame deletions and titin protein size was not analyzed to confirm expression from deletion alleles. If production of truncated titin peptides contributes to disease, abundance of truncated proteins should increase during disease progression. In support of this hypothesis, Fomin et al. [21] found that patients with the highest levels of truncated titin protein in the heart, required heart transplantation at early ages.

Nonsense-mediated decay of truncating alleles could represent a way to counteract pathologic pathways by reducing transcription of potentially disease-causing truncated titin peptides during early stages of disease progression. In support of this hypothesis, NMD appears to be more active in TTNtv skeletal muscles than described in the literature for cardiac tissue from TTNtv DCM patients. In addition, NMD and allelic balance also differed in two patients with the same deletion (TTN^del326–347, cases 22 & 23), which likely reflects different disease stages. One of these patients showed strongly reduced titin transcripts and allelic imbalance while markers for myofiber- and satellite cell markers were normal. In contrast, the patient’s sibling had normal titin transcript level and allelic balance but showed stronger evidence for muscle disease. This was evidenced by low amount of myofibers and satellite cells and increased abundance of markers for adipocytes and fibroblasts, which indicates fibrofatty replacement. Increased fat infiltration of skeletal muscle has recently been observed in patients with TTNtvs who had previously been diagnosed with cardiac disease [9]. All other truncating variants with allelic imbalance and reduced titin transcript also did not show increased fibrofatty replacement, indicating that NMD may indeed have a protective effect for muscle during early disease stages.

Splice variants and deletions induce exon skipping.

One of the splice variants in our study, TTN^splice117 (case 8), induces complete skipping of exon 117. It has been suggested that skipping of symmetric I-band exons is unlikely to cause disease [15]. In addition, the sequence of this exon belongs to the group of PPAK repeats (Fig. 2G), which are frequently spliced out during skeletal muscle development. The skipped exon keeps titin in frame, and it is unclear how this splice donor variant would negatively impact titin function. The deleted exon is part of titin’s PEVK segment, a main source of extensibility of titin in skeletal muscle [38]. However, the PEVK segment is large and deletion of one PPAK repeat is expected to have a negligible effect on passive tension, consistent with experimental results (Fig. 7D). Further studies are required to establish the disease mechanism in TTN^splice117 patients.

Three patients carrying a splice acceptor variant of intron 242 (cases 15,16 & 18) showed only low degrees of exon 243 skipping and did not have increased intron retention. Marker gene expression points towards a high degree of fibrofatty replacement in all three patients, which was also observed during clinical evaluation of patients in paraspinal muscle by MRI. Fatty infiltration might be related to the frequent alternative splicing events that we detected in these patient’s A- and M-band sequences. Exon skipping in these regions interferes with titin’s role in thick filament length regulation and alignment and would likely cause myopathy [11, 12].

Likewise, exon skipping in out-of-frame deletion TTN^{del326–347-1} (case 22) would also interfere with protein function, but the percentage of exon-skipped transcripts is low. While about 10% of transcripts from TTN^{del326–347-2} (case 23) carry the deleted region, there’s no evidence for shortened titin protein, indicating that the mutant allele is either not translated or the mutant transcript is degraded.

Internal in-frame deletion produces shortened titin with a dominant negative effect.

An internal heterozygous large deletion of titin (TTN^{del 346–362}, case 5) resulted in expression of both regular-length and shortened titin transcripts, and allelic balance analysis confirmed that the deletion transcript does not get degraded. Amounts of shortened titin protein were ~ 48% less than WT titin (Fig. 5D), indicating that it is not efficiently integrated into the sarcomere and gets partially degraded. M-band titin is critical for integration into the sarcomere and maintaining structural integrity [39,40], which might explain the irregular M-band structure and poor thick filament alignment in stretched muscle (Fig. 6). Pathogenicity of this deletion was particularly strong with early-onset muscle weakness and early-onset cardiac phenotypes in all examined family members, which might be due to high amounts of shortened titin protein being produced. Protein from the disease allele lacks the final C-zone super-repeat and the complete M-band except for M10 (Fig. 2H), which might cause disease not only by impacting thick filament structure and function but also by interfering with M-band function including titin kinase signaling [41].

Differential splicing patterns and increased levels of GDF11 are distinctive features of diseased tissue.

Various genes associated with sarcomere function or skeletal myopathies were found to be differentially spliced in a subset of titinopathy patients with different variant types. These changes might reflect increased replacement of myofibers with fibrofatty tissue and lack of proliferation. For example, Ankyrin 2 skipping of exons 12,13 & 17 is typically found in muscle cells, but isoforms including these exons are more frequently found in other tissue types [42]. Similarly, enhanced presence of Mef2c exon 4 in titinopathy samples would repress myogenic gene expression through recruitment of HDAC5 [43]. Inclusion of FLNC exon 31 and Pdlim5 exon 5 is also more typical for embryonic tissues and decreases during differentiation [44,45].

Differences in splicing patterns are likely representative of differences in disease stages. Splicing patterns similar to those observed in titinopathy samples are also present in FSHD, a muscular dystrophy characterized by abnormal reactivation of DUX4, an embryonic transcription factor. It has been speculated that generation of truncated RNA-binding proteins caused by DUX4 could be the cause of mRNA misprocessing and disease [46]. Muscular dystrophies are often associated with increased muscle regeneration to maintain muscle function [47]. In addition, satellite cell function is essential to counteract muscle degeneration and increases in fibroadipose tissue [6,48]. Therefore, reduced presence of myofiber markers and satellite cells might indicate that the regenerative capacity is exhausted in some titinopathy samples, which is reflected in differential splicing patterns that are more typical for embryonic tissues. Adult-to-embryonic isoform switches are frequently observed in muscular dystrophies such as myotonic dystrophy type I [23]. It is not clear whether splicing changes cause pathogenic remodeling in muscle or if they are mainly the result of disease progression. Nevertheless, a similar set of genes is affected in different muscle diseases. This might be due to mis-regulation of splice factors, which could potentially provide therapeutic targets for a wide range of myopathies.

Another protein which has been suggested as a pathological effector of muscle is GDF11, which belongs to the TGF-β family of cytokines, together with myostatin [49]. Although several high-profile reports attributed rejuvenating effects to GDF11 in heart and skeletal muscle [50,51], more recent studies show that increased levels inhibit skeletal muscle regeneration and induce striated muscle atrophy and wasting by activating SMAD2/3 and MAPK signaling pathways [52–54]. Increased levels of GDF11 have also been associated with frailty in patients based on criteria such as low endurance, slow gait, and weak grip strength. High GDF11 levels are also a risk factor for post-operative health complications, re-hospitalization, and comorbidity in cardiovascular disease [55]. Considering that the levels of GDF11 are consistently increased among titinopathy patients, it might be useful as a pathological biomarker and/or therapeutic target because of its powerful effect on skeletal muscle growth.

Some TTN variants cause myofilament dysfunction at the single fiber level.

In order to investigate whether the skeletal muscle weakness experienced by these patients is caused by impaired contractility, we studied the mechanical properties of membrane-permeabilized single muscle fibers. The membrane-permeabilization process allowed investigation of the active force generation capacity of these fibers by addition of exogenous Ca²⁺. Furthermore, as one of the principal functions of titin is to act as a molecular spring that resists sarcomeric stretching and shortening above and below the sarcomere slack length, we evaluated the fibers passive tensions [56]. Interestingly, the muscle fibers from most patients had normal active and passive mechanical properties. Titin-dependent passive tension that develops as a muscle fiber is stretched was only increased in slow fibers from the TTN^del346–362 patient (case 5). This patient has a heterozygous deletion of TTN exons 346–362 that translates to a TTN gene product with greater mobility during gel electrophoresis, indicating a smaller protein. It is possible that the A-band segment of the mutant protein is not at the exact same location along the thick filament, compared to WT titin and that this would result in a larger extension of the I-band segment and increase passive tension when sarcomeres are stretched. Ultrastructural examination of fibers from this patient showed irregular A-bands, which may be a consequence of a weakened M-band that provides less connectivity between neighboring thick filaments, or perhaps due to the higher passive tension, particularly if the passive tension in the two half sarcomeres is not well balanced. Other studies from our group suggest that striated muscle can compensate for increased passive tension by adding sarcomeres in series, which induces a leftward shift in sarcomere length working range resulting in decreased active force production [57,58]. Surprisingly, at a sarcomere length of 2.6 μm, active tension in fibers from this patient appeared normal, suggesting that the misaligned A-band and shortened titin molecule had minimal effect on force production (Fig. 8D).

Although all patients had clinical muscle weakness; fiber experiments only showed that two out of seven patients had reduced maximal active tensions in the membrane-permeabilized preparations (Fig. 8D). Active stiffness measurements suggest that the cause of the weakness in these patients was a lower fraction of force-producing cross-bridges (Fig. 8G). One of these patients was diagnosed with HMERF (case 12) and another patient had a splice site variant (case 8). In contrast, two other patients with truncating variants had mostly unchanged contractile properties.

HMERF is a relatively rare disease, which is caused by amino acid substitutions in the Fn3–119 domain of titin. Patients with HMERF usually present with respiratory failure at the time of diagnosis with limb-girdle muscle weakness identified later [59]. We found that one HMERF-patient had striking impairments as seen by a ~ 65% reduction in active tension due to decreased cross-bridge formation and changes in kinetics of cross-bridge cycling (Fig. 8D, G and I). A more comprehensive study of muscle fiber function from more HMERF patients is highly warranted.

To our knowledge this is the first comprehensive study of contractile consequences of pathogenic titin variants on skeletal muscle fibers. Only one other study has investigated contractile function in a single titinopathy patient [60]. This patient had a homozygous pathogenic variant in titin and suffered from contractures, rigid spine, and muscle weakness. Muscle fibers from the patient had unaltered maximal tensions but increased calcium sensitivity, indicating increased submaximal tension which could contribute to the rigid spine and contractures in the patient.

The relatively small impact of titin variants on myofilament function in this study compared to pathogenic variants of thin filament-associated proteins [61,62] highlights the complex regulatory functions of titin in addition to its structural role.

Conclusions

Our results suggest that a range of pathogenic variants in TTN may cause a mild skeletal myopathy or muscular dystrophy in the monoallelic state. We have identified a dominant negative mechanism in an in-frame deletion which produces a shortened protein that impairs sarcomeric structure. Truncating variants lead to nonsense mediated decay, which may serve as a protective mechanism against potential harmful effects of truncated proteins. Splice variants and out-of-frame deletions induce exon skipping and may impact functional domains. Although further study is necessary to fully elucidate the dominant mode of action for the latter variant types, family and genetic studies demonstrate clear autosomal dominant segregation of these variants with skeletal muscle disease. Our work adds to the growing evidence for a new class of dominant skeletal muscle titinopathies [7–9,63], a category historically restricted to only HMERF and LGMD2J.

Incomplete understanding of genotype–phenotype relationships has been a barrier to accurate diagnosis and care of patients with TTN variants, particularly those with neuromuscular presentations [63]. Our findings indicate that monoallelic truncating, splice, and deletion TTN variants may cause both skeletal and cardiac disease, with reproductive risks associated with both dominant transmission (in the monoallelic state) and recessive transmission (in the biallelic state). Recognition of the spectrum of potential clinical manifestations and reproductive implications associated with different variants in both the biallelic and monoallelic state is necessary for appropriate medical management and genetic counseling of persons identified with TTN variants. With the recent addition of TTN to the American College of Medical Genetics and Genomics list of ‘secondary findings’ genes for clinical testing disclosure [64], carriers of TTN variants will be increasingly identified across all medical settings.

No truncated titin was detected for TTNtvs included in this study, in contrast to studies on cardiac muscle of TTNtv patients [23–23], where a poison peptide effect has been proposed as a mechanism of disease. Nevertheless, long-lasting upregulation of titin translation or reduced turnover to maintain a normal level of wildtype titin protein might have deleterious consequences, as it could reduce the potential for muscle regeneration and stress adaptation, ultimately leading to skeletal muscle disease. A common feature of diseased tissue in our studies was upregulation of GDF11. Considering that antagonizing the GDF11-STAT3 axis has been shown to inhibit proteolytic pathways in vitro [65], inhibition of GDF11-induced skeletal muscle wasting as a therapeutic approach might be worth exploring in future studies.

Materials & methods

Patient identification, consent, and biopsy

A total of 11 patients with skeletal muscle weakness and apparently monoallelic premature stop, splice, or deletion variants in TTN were included in the case cohort. Ten healthy controls and two disease controls (one with hereditary myopathy with early respiratory failure (HMERF, case 12) and one with TTN-related recessive distal myopathy (case 19)) were included. All cases and controls provided signed, informed consent under the OSU2018H0408 protocol approved by the Institutional Review Board of The Ohio State University. Five patients were identified in a prior study (The Ohio State University OSU2016H0308; 31), while 6 others were newly identified. Medical records and family history information were reviewed and documented.

Muscle tissue was obtained from each patient via a needle biopsy procedure. The skin overlying the tibialis anterior was prepped with ChloraPrep sterile antiseptic solution. The skin, subcutaneous, and muscle tissues were anesthetized with 1% lidocaine (up to 10 ml or 100 mg). A 3 mm incision was made through the fascia to pass the biopsy device into the muscle. Percutaneous muscle biopsy was performed using a 14-gauge disposable Guillotine soft tissue biopsy needle (SuperCore, Angiotech). Muscle biopsies were divided into three parts. One part was frozen in liquid nitrogen and used for protein studies, another one was membrane-permeabilized and designated for ultrastructural and mechanical studies and the third part was stored in RNAlater and used for RNA-sequencing.

Molecular analysis

Next-generation sequencing gene panels were used to simultaneously test for both sequence and exon-level copy number variants, as previously described [66–68]. All targeted regions are sequenced with ≥50x depth or were supplemented with additional analysis. Reads were aligned to a reference sequence (GRCh37), and sequence changes were identified and interpreted in the context of a single clinically relevant transcript. Enrichment and analysis focused on the coding sequence of the relevant transcripts, 20 bp of flanking intronic sequence, and other specific genomic regions demonstrated to be causative of disease at the time of assay design. Promoters, untranslated regions, and other non-coding regions were not otherwise interrogated. Exonic deletions and duplications were called using an algorithm that determines copy number at each target by comparing the read depth for each target in the proband sequence with both mean read-depth and read-depth distribution, obtained from a set of clinical samples. Variants were reported according to the Human Genome Variation Society (HGVS) guidelines. Confirmation of the presence and location of reportable variants was performed as needed based on stringent criteria using one of several validated orthogonal approaches [68]. Sequencing was performed by Invitae Corporation (1400 16th Street, San Francisco, CA 94103, #05D2040778).

Each proband was sequenced using a comprehensive panel including up to 200 genes associated with neuromuscular disorders, cardiomyopathies and/or arrhythmias. Variants were classified as pathogenic or likely pathogenic (P/LP), of uncertain significance, likely benign, or benign using Sherloc [69], a variant interpretation framework that relies on a point-based evidence scoring system built on the joint consensus guidelines from the American College of Medical Genetics and Genomics and the Association for Molecular Pathology [70]. The following TTN exons: 45–46, 147, 149, 164, and 172–201 (NM_001267550.2) were excluded from analysis because not yet identified in biologically relevant products (exons: 45–46, 147 149 and 184) or because being part of a tandem-repeat region sharing DNA sequence homology among them (exons: 172–198) thus unable to be disambiguated using the current technology. Variants are named relative to the NM_001267550.2 (meta) transcript.

In order to investigate the possibility of unrecognized ‘second hit’ variants resulting in a recessive disease mechanism, TTN sequence data from each case was reviewed. Each rare missense variant identified was scored for pathogenicity by Alphamissense [17] and available family members were tested to determine segregation.

RNA sequencing

Biopsies of titinopathy- or control samples were collected and stored in RNAlater to preserve RNA integrity. For RNA extraction, 600 μl pre-chilled buffer RLT (RNeasy Fibrous Tissue Mini Kit, Qiagen) with 1% β-Mercaptoethanol was added to muscle tissue stored in RNAlater in a 4 ml cryovial. Tissue was disrupted using a rotor-stator-homogenizer for 30 seconds. A protein digest was performed by adding 600 μl RNase-free water containing 6 mAU Proteinase K and incubating at 55°C for 10 min. Samples were transferred to a 1.5 ml microfuge tube and centrifuged for 3 min at 14000 g. The supernatant was pipetted to a 2 ml tube with 600 μl Ethanol and transferred to an RNeasy mini spin column. Thereafter, RNA extraction was performed following the manufacturer’s instructions and quantified using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific). RNA integrity was checked by running the samples on a 2100 bioanalyzer (Agilent) and all RIN scores were confirmed to be > = 7.

Sequencing was performed on an Illumina Hiseq2500 sequencer using 150-bp paired-end sequencing. Adapters and low-quality reads were removed with Trim Galore [71], and reads were mapped to the human genome (Release GRCh38.p13) using STAR [72] with default settings.

Differential gene expression and splicing analysis

Transcript levels determined by STAR were used to calculate relative titin expression levels. Because nebulin is known to be a suitable marker for myofibers, it was used to normalize titin reads and allow for comparison of titin levels across biopsies with varying cell composition. Marker gene expression was normalized by applying median of ratios normalization from DESeq2 [73] to account for differences in library size and RNA composition of samples. For analysis of transcript-specific gene expression, transcript abundances were quantified with Salmon [74] against human genome version gencode v44 (GRCh38.p14). Differential expression was analyzed with a nonparametric method using Swish [75], which is well-suited for analyzing datasets with high degrees of inferential uncertainty and heterogeneity between biopsy samples.

For genome-wide calculation of inclusion percentages of exons, inclusion reads (IRs) and exclusion reads (ERs) were counted for each exon based on MANE annotation [76]. For titin for example, isoform NM_001267550.2 was used, which contains 363 exons. IRs are reads overlapping the exon being investigated, normalized by exon length. ERs are read either upstream or downstream that support exclusions of the read. From these factors, the following equations were used to calculate the PSI index using the ASpli R-package [77].

$$ {\displaystyle \begin{array}{l}\mathrm{IRi},\mathrm{n}=\mathrm{IRi}/\mathrm{length}\ \mathrm{exoni}+\mathrm{read}\ \mathrm{length}-1\\ {}\mathrm{ERi},\mathrm{n}=\mathrm{ERi}/\mathrm{read}\ \mathrm{length}-1\\ {}\mathrm{PSIi}=\left(\mathrm{IRi},\mathrm{n}/\mathrm{IRi},\mathrm{n}+\mathrm{ERi},\mathrm{n}\right)\%\end{array}} $$

where i is the exon number and n is the normalized read count.

Differentially spliced exons between intron 242 splice acceptor variants and controls were determined by unpaired t-tests. Exons with a minimum mean PSI difference of 5% were considered to be differentially spliced. For heat map generation, differentially spliced exons were hierarchically clustered using complete linkage. Heatmaps were generated in R with pheatmap [78].

Variant detection & allelic balance calculation

To quantify allele ratios for titin variants, we used GATK v4.2.5. [79] on paired-end, 150 nt RNA-seq reads. According to GATK best practices [80], duplicate reads that might originate from the same RNA transcript were marked with Picard v3.1.0 (https://broadinstitute.github.io/picard/) to prevent counting them as additional evidence variant identification. Sequencing reads were then processed using GATK SplitNCigarReads, which removes intronic sequences and thereby reduces false positive detections.

For allele ratio calculations, SNVs from biallelic sites with high coverage were counted and ratios were calculated as follows: Allele 1 (lower abundance)/Allele 1 (lower abundance) + Allele 2 (higher abundance). Following this calculation, equal expression from both alleles would give a ratio of 0.5. Any allelic imbalance would shift the ratio closer to zero.

Muscle mechanics

One part of the obtained biopsy was placed in a membrane-permeabilizing ‘skinning’ buffer containing 1% Triton X-100 detergent (v/v) in relax solution (in mM: 40 BES, 10 EGTA, 6.56 MgCl₂, 5.88 Na-ATP, 46.35 K-Proprionate, 15 Creatine Phosphate, 1 DTT, 0.03 E64, 0.141 Leupeptin and 0.25 PMSF) and shipped overnight on regular ice to the University of Arizona. On arrival, the buffer was exchanged with fresh skinning buffer and the tubes were placed on a rocker at 4°C. After ca 3 h the buffer was replaced with relax solution and stored on ice in 4°C. Single muscle fibers were isolated from the biopsy and each fiber end was placed into a 27G stainless steel trough attached to a force transducer (403B, Aurora Scientific, Aurora, ON, Canada) and a motor-controlled length arm (322C-I, Aurora Scientific, Aurora, ON, Canada) on the 1400A Permeablizied Fiber System (Aurora Scientific, Aurora, Canada). A 4–0 nylon monofilament was placed on top of the fiber and ~ 9-0G loops prepared from dental floss were tightened around the monofilament and trough to fix the single fiber. The fiber apparatus was placed on the stage of an inverted microscope. Sarcomere length of the fiber was measured using a high-speed camera and 901D software (Aurora Scientific, Aurora, ON, Canada). Fiber diameter and depth (determined with a built-in prism that allowed for side view of the fiber) were measured at four points along the fiber and the cross-sectional area was calculated assuming an elliptical fiber shape. The temperature was maintained at 15°C.

Protocols: For measuring passive force, single fibers were set at their slack sarcomere length and stretched 90% using a ramp-hold protocol in a relaxing solution. At slack length, the fiber dimensions were measured. The ramp-hold protocol consisted of 9 ramps, each stretching the fiber 10% of the initial fiber length (FL) at a speed of 1% FL per s. After each 10% stretch the fiber was held for 20 s and the reached sarcomere length was recorded. The sarcomere length was recorded from the same area each time. If the sarcomere length was > 3.6 μm, the protocol was aborted, to avoid overstretching the fiber.

To determine myosin-based force, the fiber was lengthened to a sarcomere length of 2.6 μm and its dimensions were measured. The fiber was moved to a bath containing pre-activating solution (in mM: 40 BES, 1 EGTA, 6.32 MgCl₂, 5.82 Na-ATP, 81.71 K-Proprionate, 15 Creatine Phosphate, 1 DTT, 0.01 E64, 0.047 Leupeptin and 0.25 PMSF) for ~ 30 s, and then to activating solution (in mM: 40 BES, 10 CaCO3-EGTA, 6.29 MgCl₂, 6.12 Na-ATP, 45.3 K-Proprionate, 15 Creatine Phosphate, 1 DTT, 0.01 E64, 0.047 Leupeptin and 0.25 PMSF) until the fiber reached steady state force at which it was moved back to relaxing solution. To determine calcium sensitivity the fiber was activated in solutions containing incremental concentrations of free calcium. The solutions were created by mixing relaxing and activating solutions taking into account the K_d of Ca²⁺ according to the model developed by Fabiato & Fabiato [81]. Calcium sensitivity was calculated by fitting the force-[Ca²⁺] to the Hill equation: normalized tension = (F_max^*[10^-Ca2+]^nH)/(K^nH+[10^-Ca2+]^nH), where nH is the Hill coefficient, F_max is force at max [Ca²⁺] and pCa50 = −log K.

Active stiffness, an indicator of the number of strongly bound cross-bridges, was measured using 20 ms 0.3% FL 500 Hz sinusoidal length changes in fully activated fiber [82]. Stiffness is calculated by dividing the force change with the length change.

Rate of tension redevelopment (k_tr) was determined with protocol as in Brenner [35]. At steady state force, the activated fiber was rapidly shortened (<1 ms) to 75% of FL for 20 ms, then stretched to 103% of FL and immediately after the stretch moved to the initial length. This procedure caused all bound cross-bridges to detach lowering force to zero. Then the force started to redevelop. The force rise was fitted with double exponential function: F = F_ss^*((1—e^-ktr*t) + (1—e^k*t)) + c, where F is force at time t, F_ss is steady state force, k is the time constant of the slow phase of force redevelopment; ktr is the time constant of the fast phase of force redevelopment, which reflects kinetics of cross-bridge turnover, where the fast component corresponds to the rate constant [83].

After the mechanical experiments the fiber was removed from the setup and placed in a tube containing 25 μl sample buffer (62.5 mM Tris–HCL, 2% SDS, 15% Glycerol, 5% 2-mercapto-ethanol, 2.5 mM Bromophenol blue, and pH 6.8) for future fiber type determination as in Lindqvist et al [83]. In brief, 8% sodium dodecyl sulfate polyacrylamide gels were run for 24 h at 275 V. The gels were silver stained according to manufacturer’s instructions (Bio-Rad, Hercules, CA, USA).

Electron microscopy

The ends of single fibers or small fiber bundles were attached to aluminum T-clips and stretched ~ 20%. The preparations were then thoroughly washed in relaxed buffer and fixed with 3.7% paraformaldehyde, 3% glutaraldehyde, and 0.2% tannic acid in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄, pH 7.2) for 1 hour at 4°C. Then, muscles were postfixed in 1% OsO₄ in PBS for 30 min at 4°C, and then the samples were dehydrated in an ethanol graded series, infiltrated with propylene oxide, and transferred to a mix of 1:1 propylene oxide:Araldite 502/Embed 812 (Epon-812, EMS). Subsequently, samples were transferred to a pure Araldite 502/Embed 812 resin and polymerized for 48 h at 60°C. 50-nm longitudinal sections were obtained with a Reichert-Jung ultramicrotome and contrasted with 1% potassium permanganate and lead citrate. Observations were made with a TECNAI Spirit G2 TEM (FEI, Hillsboro, OR), and images were acquired with a side-mounted AMT Image Capture Engine V6.02 (4Mpix) digital camera operated at 100 kV.

Acknowledgement

We thank Natalie Nichols for assistance with Fig. 1.

Conflict of Interest statement. AM and MV are employees and stockholders of Invitae Corporation. BE received research funding from Biogen, Genentech, Alexion, Pharnext, and Viela Bio, and served as a consultant for Biogen, Genentech, and Argenx. The other authors declare no conflict of interest.

Funding

This work was financially supported by National Institutes of Health grants R01AR083233 and R35HL144998 to HG, Team Titin, the Foye family, and the Tse family.