The importance of an integrated genotype-phenotype strategy to unravel the molecular bases of titinopathies

Next generation sequencing (NGS) has allowed the titin gene (TTN) to be identified as a major contributor to neuromuscular disorders, with high clinical heterogeneity. The mechanisms underlying the phenotypic variability and the dominant or recessive pattern of inheritance are unclear. Titin is involved in the formation and stability of the sarcomeres. The effects of the different TTN variants can be harmless or pathogenic (recessive or dominant) but the interpretation is tricky because the current bioinformatics tools can not predict their functional impact. Moreover, TTN variants are very frequent in the general population. The combination of deep phenotyping associated with RNA molecular analyses, western blot (WB) and functional studies is often essential for the interpretation of genetic variants in patients suspected of titinopathy. In line with the current guidelines and suggestions, we implemented for patients with skeletal myopathy and with potentially disease causing TTN variant(s) an integrated genotype-transcripts-protein-phenotype approach, associated with phenotype and variants segregation studies in relatives and confrontation with published data on titinopathies to evaluate pathogenic effects of TTN variants (even truncating ones) on titin transcripts, amount, size and functionality. We illustrate this integrated approach in four patients with recessive congenital myopathy.


Introduction
TTN encodes titin, a giant sarcomeric protein [1] that plays important functional and structural roles in the sarcomere. Mutations in TTN are associated with a broad range of skeletal and/or cardiac phenotypes with autosomal dominant or recessive inheritance. Since the widespread use of Next Generation Sequencing (NGS), the exhaustive analysis of the 364 coding exons of TTN revealed an increasing number of reported TTN variants. Nevertheless, assessing the pathogenicity of the identified variants, in particular missense ones, is often difficult, due to the clinical heterogeneity of the pathology, the large size of the gene , the frequency of TTN variants in the general population (including 0, 36% of truncating variants affecting all transcripts [2] ) and the complexity of gene expression (many different isoforms are expressed over time and tissue location) [3] . Different types of mutation have been reported as pathogenic, but their consequences on transcripts and protein, in particular loss or gain of titin function, are not fully elucidated. In myopathic titinopathies, several studies demonstrated that TTN truncating mutations are associated with loss function leading to recessive titinopathies, including congenital titinopathies. Interpretation of non-truncating (mainly missense) variants remains more difficult. Their localization in important functional domains (such as the titin kinase domain) and interaction domains (with proteins such as calpain-3, actin, myosin, nebulin, or one of the numerous interacting proteins (for review, see [4] )) may suggest a potential pathogenicity. Specific variants in the last exon (including missense variants) and all the truncating variants causing the loss of the Cterminal portion (i.e. M-band truncating variants) causes the loss of the interaction with calpain-3 and its reduction at western blot, may contribute to disease pathogenesis, raising the possibility that the decrease of calpain-3 level may contribute to disease pathogenesis [5][6][7][8][9] .
Expert laboratories have worked on the establishment of specific recommendations for TTN variants interpretation [10] , based on familial and genetic data (number and type of variant(s)), histopathology [11] , magnetic resonance imaging (MRI) [12] or clinical features [13] . In line with the current guidelines and suggestions [2 , 3 , 14] the combination of deep phenotyping associated with RNA molecular analyses, western blot (WB) and functional studies is often essential for the interpretation of genetic variants in patients suspected of titinopathy. Namely, pattern featuring a prominent semitendinous muscle involvement on MRI, prominent nuclear internalization, peculiar alteration of oxidative reactions (such as cores, minicores and inverted cores) and presence of cytoplasmic bodies on muscle biopsy are suggestive and common, although non-specific, features found in all titinopathies .
For these reasons, we implemented a workflow based on deep phenotyping, comprehensive TTN variants assessment including bioinformatic predictions and evaluation of their effects on skeletal muscle titin transcripts and protein. We hereby describe our strategy for TTN variant interpretation in four patients with recessive congenital myopathy.

Patients
The study was approved by the ethical guidelines issued by our institutions for clinical studies in compliance with the Helsinki Declaration. Patients or parents gave informed consent for the genetic analysis according to French legislation (Comité de Protection des Personnes OUEST 6-CPP1128HPS3 IDRCB-2018-AO2287-48).

Bioinformatic evaluation of TTN Variants Pathogenicity
TTN (NM_001267550.1) variants were identified by NGS targeted on exons and exon-intron junctions of 185 genes, as previously reported [15] and Juntas Morales et al, submitted.
The pathogenicity of variants was assessed through a set of criteria reported by Zenagui et al. [15] , according to the American College of Medical Genetics and Genomics (ACMG) guidelines [16] . In the present work, additional tools recently described were used: the MPA software [17] that aggregates bioinformatic prediction tools and provides a 0-10 prioritization score, accessible from https://mobidetails. iurc.montp.inserm.fr/, and the TITINdb, a database that takes into account several statistical and structural criteria for the interpretation of TTN variants ( http://fraternalilab.kcl.ac. uk/ TITINdb/ ) [18] . In addition, to take into account the complexity of skeletal transcripts of TTN, in particular the fact that some exons are only expressed during fetal development ("metatranscript-only exons"), we systematically referred to the recent study reported by Savarese et al. [19] . All variants presented in this study have been submitted in Leiden Open Variation Database (LOVD) https:// databases.lovd.nl/ shared/ genes/TTN.

Muscle morphological analyses and immunohistochemistry
Open muscle biopsies were performed, frozen and fixed following previously described standardized methods and analyzed with histochemical, histoenzymatic, techniques with light and electron microscopy [11 , 20] .

Transcripts analysis
mRNAs were extracted from 20 mg of muscle biopsy and prepared as described in [15] . 2 μg of RNA was reverse-transcribed with Invitrogen Superscript II reverse transcriptase. For conventional analysis, we performed overlapping RT-PCR reactions covering regions with TTN variants (supplementary table 1). For RNAseq analyses, sequencing was performed by IntegraGen SA (Evry, France) using the same protocol with poly-A selection of mRNA (Illumina (San Diego, CA, USA) TruSeq® Stranded Total RNA Library Prep) used in the GTEx sequencing project [21] . Paired-end 75-bp sequencing was performed on Illumina NovaSeq6000® platform as described in [21] . 88 Millions of reads were generated by RNAseq, including 687000 reads for TTN transcripts. RNA sequences alignment was performed using Ultrafast universal RNA-seq aligner (STAR) software [22] and analyzed using Integrative Genomics Viewer (IGV) [23] targeted on titin transcripts. The mean depth of RNA sequencing was 131 reads for titin transcripts NM_133378 corresponding to N2A isoform. Four exons had coverage under 30 ×: exon 11 (7 ×), exon 149 (14 ×), exon 150 (20 ×), and exon 151 (15 ×). These exons were described in [19] to be constitutively splice out for exon 11 and included in 35%, 34% and 40% of transcripts for exons 149, 150, 151 respectively. Exon 1 had a depth of sequencing of 157 ×, confirming that the 5' extremity was well covered.

Western blot analysis (WB)
Protein solubilization from muscle biopsies was prepared as previously described [24] . For full length titin WB, 5 μl of protein extract were loaded on a SDS-Agarose 0.8% gel prepared as previously described [25 , 26] and ran for 3h20 at 15mA. Protein transfer on PVDF membrane was performed using wet transfer Trans blot® cell Bio-Rad (Bio-Rad, Hercules, CA, USA) at 400mA for 3h. Membranes were subsequently incubated in Odyssey® blocking buffer, washed and incubated overnight at 4 °C in Odyssey® blocking buffer in PBS specific primary antibodies: anti-titin specific for the C-terminal part M10-1 (Rabbit; 1:1000; kindly provided by Dr Isabelle Richard [7] ) or N-terminal part (Mouse; 1:1000; Sigma SAB1400284). Membranes were washed and incubated for one hour with donkey anti-mouse IR Dye® 800CW (1:30,000) or donkey anti-rabbit IR Dye®680RD (1:30,000). Membranes were then washed, dried and scanned with the Odyssey® Infrared Imaging system.

Clinical and imaging findings
Clinical features of all the patients are summarized in Table 1 .
Patients 1 and 2 (P1, P2) were twin sisters born to a non-consanguineous family. Pregnancy and delivery were normal, no neonatal hypotonia was reported. At the age of six months, they developed axial hypotonia. The motor development was delayed, and they were not able to walk before the age of 29 months. Neurological examination at the age of 4 years showed facial weakness, high arched palate, and mild proximal and axial weakness with Gowers's sign. There was no ptosis or ophthalmoparesis. Cardiological test at 9 years showed for P2 an ejection fraction at the lower limit of the normal range. P1 and P2 progressively developed hip, ankle, elbow and cervical spine contractures. Complementary exams were performed for patient 1 and results were as follow: Creatin kinase (CK) level was normal and electromyography showed myopathic pattern. In patient 1, WB studies with an antibody against calpain-3 revealed reduced protein level (data not shown). CAPN3 screening was normal. Brain MRI disclosed the presence of moderate periventricular white matter hyperintensity on Flair weighted signal ( Fig. 1 (A)). Muscular MRI was not done. Patient 3 (P3) is an adopted child from Latvia. At 3 years, parents noticed a waddling gait and difficulties in sport activities. Clinical examination at 8 years revealed facial weakness, moderate scapula alata and mild proximal weakness. There were no contractures or skeletal deformities. Cardiac assessment at 14 years was normal. There was a severe restrictive respiratory syndrome with Forced Vital Capacity (FVC) at 1.2L (33.2% of predicted). As for P1, brain MRI revealed the presence of moderate periventricular white matter hyperintensity on Flair weighted signal ( Fig. 1 (B)) Patient 4 (P4) is a sporadic case. The girl developed, from the age of 4 years, slowly progressive lower limb weakness. The patient progressively lost ambulation and was wheelchair bound at 17 years. Neurological examination at 45 years showed bilateral facial weakness, severe axial and proximal weakness of four limbs without contractures. Mild distal weakness was also present. There was restrictive respiratory syndrome with an FVC at 58%. At 55 years, a mild dilated cardiomyopathy with ejection fraction of 45% was revealed by cardiac ultrasonography. Muscle MRI of lower limbs showed bilateral soleus, lateral peroneus and tibial anterior involvement.

Muscle morphological data
A skeletal muscle biopsy was obtained from all patients except P2. There was an important percentage of centralized nuclei, from 9% in P1 to 95% in P4 ( Fig. 2 (A) and (B)) and minicore-like lesions with NADH oxidative reaction ( Fig. 2 (C)-(E)). P1, P2, P3 and P4 were diagnosed as having a centronuclear myopathy. Ultrastructural studies in P3 and P4 revealed the presence of multifocal areas of sarcomeric disruption. In P4, we observed well delimited areas containing filamentous material and cellular debris corresponding to 'Cap' structures were visualized ( Fig. 2 (F)-(H)). As a whole, excessive internal nuclei, minicore-like and cap structures were consistent with TTN related muscle morphological presentation.

Molecular and protein results
Localization of variants on corresponding titin domains is shown on Fig. 3 . The exon numbering is reported following the numbering used in LOVD. No other pathogenic variants than the TTN ones were identified in the reported patients.  ( Fig. 3 ). The missense variant in exon 360 affects the last nucleotide of the exon and is predicted to have an impact on exon 360 RNA splicing (HSF score shift at −46,1% and MaxEnt score at-29%). To confirm this prediction, we investigated the region of transcripts containing the variant (exons 359-362) by RT-PCR and Sanger sequencing ( Fig. 4 (A)). cDNA agarose gel revealed the presence of two bands of intensity difference, and sequencing of RT-PCR products identified two populations of transcripts, a major one with aberrant splicing (exon 360 skipping) and a minor one with normal splicing of exon 360. This minor transcript with normal exon 360 splicing do not carry the paternal c.106531G > C variant in exon 360, indicating that the paternal variant results in 100% of the transcripts with the exon 360 skipping. This experiment also suggests that the transcripts with the maternal nonsense variant have a reduced level, probably due to RNA degradation by Nonsense Mediated RNA Decay (NMD). Exon 360 skipping disrupts the open reading, leading to a Stop codon 6 nucleotides downstream from the frameshift. Overall, both TTN mutated alleles in P1 and P2 are predicted to produce a M-band truncated protein ( Fig. 5 (A)). WB analysis of the C-terminal part of titin showed the loss of the C-term p45 band in P1 and, thereby, the loss of titin C-terminal portion ( Fig. 5 (B)). The total absence of the C-terminal fragments is a further argument suggesting that no normal transcripts are produced from the paternal allele with the variant affecting the last nucleotide of the exon.

P3 patient
In P3, two heterozygous TTN variants, located in exon 93, c.26503 A > T p.(Lys8835 * ) and in intron 272-exon 273, c.51437-4_51444delTCAGAACCCCCT, were identified. In the absence of possible sampling of his parents (adopted child), the phasing of the variants could not be determined. They were not reported in the gnomAD database. The first variant is a nonsense in the I-band domain and the second a deletion removing 4 nucleotides of the acceptor splice site of exon 273 and the first 8 nucleotides of exon 273. The latter is predicted to affect exon 273 splicing (HSF and MaxEnt prediction scores for loss of exon 273 splicing: 100%). RT-PCR and cDNA sequencing of the region containing the deletion (exons 272-276) identified aberrant transcripts with a loss of the 21 first nucleotides of exon 273, due to the use of a cryptic acceptor splice site 13 bp downstream the genomic deletion ( Fig. 4 (B)). This deleted transcript population is not predicted to affect the reading frame but rather to lead to an internally deleted protein, lacking seven amino acids encoded by exon 273 ( Fig. 4 (B)). These seven amino acids are located in Fibronectin type-III 24, a myosin interaction domain. Titin WB was not performed for this patient due to the lack of biopsy material.

P4 patient
P4 has three TTN variants, c.-14 + 5G > C in exon 1 and c.68029A > G; p.(Thr22677Ala) in exon 321, both absent in her mother (no sample was provided for the father) and c.65575 + 2T > G in intron 313 inherited from the mother. There was no proband's siblings to complete the segregation study. All these variants were absent from gnomAD population database. Exon 1 is reported to be expressed but not translated into amino acids. Presumably on the same allele, the missense mutation in exon 321, p.(Thr22677Ala), is localized in A band, Fibronectin type-II 64 domain. It has an MPA score at 6.7/10 and is predicted to destabilize the protein structure, according to titindb. cDNA analyses of the region containing the c.68029A > G; p.(Thr22677Ala) variant showed normal splicing and normal allelic balance with the maternal allele, not supportive of a potential reduction of TTN expression caused by the exon 1 variant presumably in cis . On the other allele, the c.65575 + 2T > G variant is predicted to have an impact on exon 313 splicing (HSF and MaxEnt predictions scores: 100%). This was confirmed by RT-PCR sequencing and RNAseq that revealed a complex effect with two abnormal populations of transcripts ( Fig. 4 (C) and (D)). A full exon 313 skipping (306 nucleotides) was found in one of the transcript population, leading to the synthesis of an internally deleted protein lacking 102 residues in A-band, in a myosin interaction domain [27] . This deletion is extremely small compared to the size of the protein and is therefore not detectable in WB. In the second population, transcripts are deleted of the 163 last nucleotides of exon 313, due to the use of a cryptic donor splice site (GTACGG) within this exon. This transcript population induces a shift in the reading frame and a Stop codon 12 nucleotides downstream from the shift. If not degraded by nonsense mediated RNA decay (NMD), this allele is predicted to produce a truncated protein with an approximate molecular weight of 2.2MDa ( Fig. 5 (C)), lacking a large part of the A-band and the Mband. RNA-seq analyses allowed to exclude elusive variants such as deep intronic variants affecting splicing. WB analysis of muscle biopsy showed the presence of normal N2A protein   [8] ) and P1 with absence of p45 peptide specific of C-term part of titin as a consequence of the two M-band truncated variants. Black star is unspecific band. Loading was normalized with GAPDH antibody. (C) Scheme of P4 predicted titin alleles. In addition to the 3.8MDa predicted titin allele with the missense variant p.(Thr22677Ala), the in frame transcript with exon 313 skipping due to c.65575 + 2T > G is predicted to lead to a 3.79 MDa titin, indistinguishable from the normal length titin. The out of frame transcript population missing the last 163 bases of exon 313 is predicted to produce a C-term truncated protein with an approximately molecular weight of 2,2MDa. (D) Control (Ctrl) and P4 skeletal muscle titin WB analysis on SDS-Agarose gel with specific antibodies against C-term part (M10.1) and N-term part (Sigma SAB1400284). C-term antibody allows to observe N2A isoform (3,8MDa) and T2 titin degradation products (around 3MDa). N-term antibody analysis revealed the presence of normal N2A protein (3,8MDa) and Novex-3 (around 1MDa) in Ctrl and P4 muscles. This antibody shows several bands in the Ctrl muscle, and only one additional band in P4 muscle with a molecular weight of 2,2MDa. This band could correspond to the truncated protein due to the specific characterized splicing effect in exon 313.
(3,8MDa) and one additional band (although at a low amount) at an approximate molecular weight of 2.2MDa, with no associated background noise ( Fig. 5 (D)). This band could correspond to the truncated protein due to the premature stop codon in transcripts induced by the splicing defect in exon 313.

Discussion
Due to the widespread use of NGS, TTN is emerging as a major causative gene of neuromuscular disorders. Interpretation of the pathogenic effect of variants is particularly challenging. In this study we present an exhaustive workflow in order to confirm the pathogenicity of TTN variants in four patients with recessive congenital myopathy with centralized nuclei, minicores and Cap structures.
With respect to the phenotype of the patients, the presence of mild diffuse weakness, facial weakness and respiratory insufficiency was concordant with titinopathies phenotypic spectrum [5 , 14] . In a recent series of 30 titin patients, most of them had mild to moderate weakness (81%), facial weakness (70%) and respiratory insufficiency (63%) [14] . Although not specific of titinopathies, some signs were previously reported on congenital myopathy patients suspected of titinopathies [14] . In Oates et al., study (2018), 19 of 26 patients had high arched palate, one of 22 had pes cavus and 3 of 12 patients had minor structural abnormalities of unknown significance on brain MRI. This sign could be better explored in the future.
Cardiomyopathy was also a frequent feature found in 46% of cases [14] . Of note, P4 developed dilated cardiomyopathy at 50 years and P2 had ejection fraction at the lower limit of the normal range. Our study confirms that a strict cardiac workup should be performed in all congenital titinopathy patients as subtle cardiac involvement might be present in infancy and an overt cardiomyopathy might develop later in life.
From a myopathologic standpoint, all patients presented centralized nuclei and minicore-like lesions. Our findings are in line with data reported by us and others [13 , 14] , where the presence of prominent nuclear centralizations and alterations in the intermyofibrillar network are prominent features of congenital titinopathies.
With respect to the interpretation of genetic variants in TTN , Savarese and colleagues have proposed that identifying two TTN truncating variants in trans plus a suggestive phenotype results in a diagnosis of titinopathy, but a complete molecular characterization of variants affecting the canonical or noncanonical splice sites by cDNA or protein studies was suggested. [3] . In our cohort, P1 and P2 have a nonsense in the first M-band exon (359) on the maternal allele and a variant affecting the last nucleotide of the following exon (360) on the paternal allele leading to an out-of-frame exon 360 skipping in 100% of the transcripts. Both TTN variants lead to a M-band truncation. In line with the current understanding, and considering that the total absence of titin is most probably incompatible with life [28 , 29] , we expect both the mutated alleles to result in a truncated protein. Of interest, WB showed calpain-3 clear reduction consistent with previous reports of calpain 3 deficiency secondary to truncation of the titin C-terminal part [7 , 8] .
P3 has a nonsense variant in exon 93 in the I-band domain and a splicing variant resulting in a small internal inframe deletion in exon 273. Considering the current understanding, the nonsense variant in the I-band most probably causes a NMD and is unlikely to result in a truncated protein. The aberrant transcripts resulting from the splicing variant lacks the 21 first nucleotides of exon 273, leading to the loss of seven residues within a titin-myosin interaction domain. Since phasing with the nonsense variant is impossible, and the effect of the splicing variant is not an out-of-phase truncation, this splice variant is then considered as probably pathogenic and P3 is most probably a titinopathy case.
For P4, the interpretation is still challenging. He harbors a missense in exon 321 and a splice variant in intron 313. Transcript analyses showed complex consequences of the splice variant in intron 313, with a population of transcripts harboring an in-frame complete exon 313 skipping associated with another population of out-of-frame transcripts resulting from activation of a cryptic splice site. If not degraded by NMD, the latter is predicted to lead to a truncated protein of an approximate molecular weight of 2.2MDa lacking a large part of the A-band and the M-band. One additional band of 2.2 Mb was detected by WB analysis on the P4 muscle. Even if the correspondence of this band with the truncated protein due to the characterized splicing defect is not certain, this observation suggests that the transcript with the premature stop codon in exon 313 would not be degraded by nonsense mediated decay. Missense variants interpretation is a real challenge and need the association of deep clinical, histopathological and radiological phenotyping, complete molecular analyses including large segregation studies, and functional studies. Our study has limitation since the presence of clinical and histological features consistent with a congenital titinopathy cannot be a stand-alone proof. We could not perform segregation studies nor functional analyses for the missense variant identified in P4, that is then considered as uncertain significance according to ACMG guidelines. With the confirmation of paternal allele inheritance, the missense variant for P4 would have been classified as "likely pathogenic", underlining the importance of segregation studies. In their proposed workflow to interpret monoallelic TTN variants pathogenicity in congenital myopathies, Savarese et al. [10] also recommend to search for the presence of a further elusive truncating variant. No such variant was identified in P4 by RNAseq analysis. Even if the phenotype is in agreement, and the consequences of the splicing variant at the cDNA and protein levels have been demonstrated, P4 should be considered as a "possible titinopathy case". Functional studies and/or identification of other cases with the same variant would be important to confirm pathogenic effects of the missense variant. This challenging case highlights the importance of sharing data using curated databases such as LOVD and publication reports using the same comprehensive approach [14 , 30] ." In conclusion, our study illustrates that deep clinical and muscle morphologic phenotyping together with segregation studies, transcripts and protein analyses are essential for the interpretation of titin variants. Analysis of the consequences on transcripts of all variants is crucial, in order to have an exhaustive understanding of their effects at the transcriptional and protein levels. Given the complexity of titin transcripts in healthy subjects and patients, RNASeq will be the technique of choice as it allows to identify TTN variants consequences on all the length of all transcript's isoforms. This study and many others [14 , 30] are essential to describe such challenging cases in publications using this approach. The next steps will include the study of protein dysfunction at the structural or interacting levels, since published studies suggest that some titin variants could affect the stability of the protein and its interaction with direct binding partners [3 , 5 , 31] .