Expanding the clinical spectrum of COL2A1 related disorders by a mass like phenotype

MASS phenotype is a connective tissue disorder clinically overlapping with Marfan syndrome and caused by pathogenic variants in FBN1. We report four patients from three families presenting with a MASS-like phenotype consisting of tall stature, arachnodactyly, spinal deformations, dural ectasia, pectus and/or feet deformations, osteoarthritis, and/or high arched palate. Gene panel sequencing was negative for FBN1 variants. However, it revealed likely pathogenic missense variants in three individuals [c.3936G > T p.(Lys1312Asn), c.193G > A p.(Asp65Asn)] and a missense variant of unknown significance in the fourth patient [c.4013G > A p.(Ser1338Asn)] in propeptide coding regions of COL2A1. Pathogenic COL2A1 variants are associated with type II collagenopathies comprising a remarkable clinical variablility. Main features include skeletal dysplasia, ocular anomalies, and auditory defects. A MASS-like phenotype has not been associated with COL2A1 variants before. Thus, the identification of likely pathogenic COL2A1 variants in our patients expands the phenotypic spectrum of type II collagenopathies and suggests that a MASS-like phenotype can be assigned to various hereditary disorders of connective tissue. We compare the phenotypes of our patients with related disorders of connective tissue and discuss possible pathomechanisms and genotype–phenotype correlations for the identified COL2A1 variants. Our data recommend COL2A1 sequencing in FBN1-negative patients suggestive for MASS/Marfan-like phenotype (without aortopathy).

The COL2A1 gene encodes procollagen type II. Three of these molecules form the triple-helical procollagen homotrimer 1 . After secretion into the extracellular matrix (ECM) and removal of procollagen's N-and C-terminal propeptides (Fig. 1a), processed collagens form a covalently cross-linked fibrillar network in the ECM, that provides tensile strength to connective tissues 2 . Type II collagen is found primarily in cartilage, the adult vitreous, the pituitary gland, the stomach, and the epididymis (proteinatlas.org and gtexportal.org, accessed 06-29-2021); it is synthesized mainly by chondrocytes. Pathogenic sequence variants in the COL2A1 gene cause clinically distinguishable type II collagenopathies with mild to lethal phenotypes, which are usually of dominant inheritance 3,4 . Main known features of COL2A1-associated disorders are skeletal dysplasia, ocular anomalies, and auditory impairment, all with different severities 3 . The skeletal manifestations are extremely heterogeneous itself and range from osteoarthritis (OA) with mild chondrodysplasia (OSCDP, MIM #604,864) to achondrogenesis type II (ACG2, MIM #200,610) and platyspondylic skeletal dysplasia, Torrance type (PLSD; MIM #151,210) with potentially fatal courses 4,5 . Ocular symptoms are diverse and include myopia, retinal detachment, lattice degeneration of the retina, cataracts and glaucoma 6 . Auditory manifestations mainly include sensorineural hearing loss; conductive hearing loss may occur, especially in children and patients with palatal defects 7 .
Various types of COL2A1 gene alterations (e.g. substitutions, deletions) primarily with missense and lossof-function consequences have been described [8][9][10][11] . Most disease-associated COL2A1 variants (i.e. 80-85%) are located in the triple-helical region of the collagen chain ( Fig. 1a) 4,11 . In this region, missense variants have been associated with severe phenotypes [e.g. ACG2, Kniest dysplasia (MIM #156,550), Spondyloepiphyseal dysplasia The right model shows the surface hydrophobicity after molecular replacement of Asp 65 by asparagine. Different colours indicate the hydrophobicity properties of amino acids. The most polar and charged residues are in cyan and the most hydrophobic residues are in tan. The surface colouring feature of the UCSF Chimera tool (version 1.2) was used 61 . VWFC, von Willebrand factor (VWF) type C repeat. (c) Structural impact of the COL2A1 p.Lys1312Asn amino acid change. Ribbon representations of the COL2A1 fibrillar collagen NC1 domain show amino acids 1312 and surrounding residues within a radius of 5 Å as well as calcium (green sphere) binding residues as sticks. Sidechains are coloured by element (hydrogen: white; carbon: beige; oxygen: red; nitrogen: blue; cysteine: yellow). The left model shows the structural environment of Lys 1312 with Van-der-Waals (VDW) overlaps ≥ -0.4 Å (contacts, green lines), whereas the right model depicts the structural environment of Asn 1312 with VDW overlaps ≥ 0.6 Å (clashes, magenta lines). (d) Structural impact of the COL2A1 p.Ser1338Asn amino acid change. Ribbon representations show details of the COL2A1 fibrillar collagen NC1 domain for wild-type (p.Ser1338, left) and mutated (p.Asn1338, right) COL2A1. Colour codes and prediction of VDW contacts and clashes are as mentioned above.
(congenital type, SEDC, MIM #183,900)], whereas loss-of-function variants (e.g. nonsense variants, splice site variants, frameshift variants) have been mainly described in patients with milder phenotypes such as Stickler syndrome, type 1 (STL1) 4,8,9 . Thus, the molecular spectrum is different across the phenotypes 8 . The relationship between phenotypes and variants in the terminal regions of COL2A1 as well as propeptide-specific pathomechanisms are poorly understood; even if variants in the N-and C-propeptides (Fig. 1a) have been associated with milder and severe/lethal phenotypes, respectively 4 .
Here, we report four patients carrying variants in the propeptide regions of COL2A1 and presenting with tall stature, arachnodactyly, dural ectasia, and varying skeletal manifestations, resembling MASS (mitral valve, myopia, aorta, skeletal and skin features) phenotype.

Material and methods
Proband recruitment and clinical examination. Probands presented at the University Heart and Vascular Center Hamburg in our specialized outpatient clinic for connective tissue diseases or in the pediatric outpatient clinic for hereditary aortopathies between 12/2017 and 09/2021. Subsequent standardized clinical examination, 200 patients were included to this study according to following criteria: clinical features were suggestive of Marfan syndrome (MFS), Loeys-Dietz syndrome (LDS), Ehlers-Danlos syndrome (EDS), congenital contractural arachnodactyly (CCA), MASS phenotype, syndromic/non-syndromic TAAD, or an unspecified heritable disorder of connective tissue with or without vascular involvement. Blood samples of these 200 probands were obtained and targeted next-generation sequencing (tNGS, i.e. gene panel sequencing) was performed. The four subjects described in this study were referred to the clinical genetics outpatient clinic of the University Medical Center Hamburg-Eppendorf for genetic counselling and further clinical examination. They underwent clinical scoring on Marfan syndrome according to the revised Ghent nosology 12 . Aortic root measurements were carried out according to the Ghent nosology guidelines (for details see Supplementary Information). Spinal MRI scans were performed in 3 individuals to screen for dural ectasia.
Genetic testing, variant prioritization and classification. In each patient, 62 genes were analyzed using a tNGS approach as previously reported by our group 13 . These genes are either associated with thoracic aortic aneurysm/dissection-spectrum disorders or with connective tissue diseases or they have crucial functions in connective tissue homeostasis (Table S1). Synonymous, missense and nonsense variants, coding indels, and intronic alterations at exon-intron boundaries ranging from −10 to + 10 were included into the analysis. Variants with an allele frequency (AF) according to the gnomAD database v.2.1.1 14 higher than the predicted maximum population frequency (MPF) of respective gene variants were excluded from further analysis. The MPF was calculated based on the respective disease prevalence, penetrance, and the genetic/allelic contribution of the respective gene/disease 15 , which resulted in an MPF of 5.15E-06 for COL2A1 variants (Table S1). Variants passing these filters were classified according to their likelihood for pathogenicity based on the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) guidelines on variant interpretation: pathogenic variant (PV), likely pathogenic variant (LPV), variant of uncertain significance (VUS), likely benign variant (LBV), and benign variant (BV) 16,17 . Individual ACMG/AMP criteria were assigned according to an automated in-silico VarSome analysis (version 10.2) 18 . Criteria based on in-silico splicing prediction and familial segregation were added manually. Reportable variants including (L)PV and VUS were communicated to the subjects. Detailed information on targeted next-generation sequencing, variant prioritization and classification, concomitant variants as well as molecular modelling is given in the Supplementary Information.

Results
Subject 1A was first examined in our clinic at the age of 11 years. He was born after uncomplicated pregnancy and underwent pyloromyotomy due to hypertrophic pyloric stenosis in the newborn period. Otherwise, he was a healthy boy with a normal development. The family history was positive for arthritis/gout but negative for aortic aneurysms/dissections and sudden cardiac deaths. At the age of 9 years, he developed painless swelling of the metacarpophalangeal joints (MCP) of the right hand. Juvenile idiopathic arthritis (JIA) was suspected by the age of 10 years. An MRI scan of the hand revealed synovitis and effusion of the MCP3 joint. Antinuclear antibodies (ANA) were positive (maximum titer 1:2,560), whereas rheumatoid factors and Human Leukocyte Antigen (HLA) B27 could not be detected. The patient showed progressive arthritis of further MCP joints of both hands and the wrist of the right hand. He was treated with nonsteroidal anti-inflammatory drugs (NSAID) and methotrexate. Further MRI scans showed necrosis of the medial sesamoid bone of the left first toe and bursitis trochanterica as well as mild synovialitis of the left hip joint. At the age of 11 years, the clinical examination revealed arachnodactyly (Fig. 2a) with positive wrist and thumb signs. X-ray of his hands detected arthritis-associated destruction of the metacarpophalangeal joints of the fingers and the distal radioulnar joint. Subluxation of the carpometacarpal joints of both thumbs was observed (Fig. 2a). In consequence, the therapy was extended by Adalimumab (TNF-α monoclonal antibody). One year later, no clinical improvement was observed and the treatment was changed to Etanercept (TNF-α inhibitor). Furthermore, at the age of 13 years the clinical examination revealed kyphoscoliosis and pectus excavatum. His height exceeded the 99th percentile in the toddler period and reached 178 cm (+ 1.8 SD, > 95P) by the age of 13 years 19 . His arm span to height ratio was 1.07. An MRI scan of the spine revealed dural ectasia, anterospondylolisthesis, and vertebral osteochondrosis. Significant aortic root dilatation was excluded by transthoracic echocardiography (Fig. 2a). Repeated ophthalmologic investigations remained unremarkable. The diagnosis of hereditary connective tissue disorder (HCTD) was considered and discussed. According to the revised Ghent nosology 12 , this patient reaches a systemic score of 7, which led to the diagnosis of a MASS-like phenotype (  Table S2). Subject 1B is the mother of subject 1A, who presented for genetic counselling at the age of 44 years. She had a medical history of swollen finger joints since the toddler age and developed contractures of the MCP and proximal interphalangeal (PIP) joints as a teenager. She was suspected of having rheumatoid arthritis and tendinitis. However, she never experienced pain in her joints nor was treated by permanent medication. In adulthood, an MRI of the feet showed bone edema at the left Os metatarsale 3 and 4 and the right Os cuniforme, as well as mild arthrosis of the metatarsophalangeal joint of the right big toe. Echocardiography as well as ophthalmologic investigations yielded normal results. At the age of 44 years her height was 172 cm (+ 1.3 SD) and her arm span to height ratio was 1.07. Examination revealed high arched palate, pectus excavatum, scoliosis, arachnodactyly and camptodacyly of fingers II, III, and V of both hands (Fig. 2b). Facial features included dolichocephaly, enopthalmus, and malar hypoplasia. According to the revised Ghent nosology 12 , she reaches a systemic score of 5, and thus, a MASS-like phenotype was diagnosed. Genotyping showed that she carries the familial heterozygous c.3936G > T p.(Lys1312Asn) variant in COL2A1 (Table 1). Genotyping did not identify the c.3936G > T p.(Lys1312Asn) variant in the unaffected sister and father of subject 1A (Fig. 2c).
Subject 2 underwent clinical examination in our center at the age of 53 years. The family history was negative for aortic aneurysms/dissections and sudden cardiac deaths, however, arthritis was suspected in the father. She presented with tall stature (179 cm, + 2.4 SD), arachnodactyly, scoliosis, hypermobility of knees and elbows, varicosis, and pes planus. Furthermore, she reported ear murmur of unknown cause. A lumbar spine MRI revealed dural ectasia. Echocardiographic investigation was normal with an aortic root Z-score < 2. The diagnosis of HCTD was considered and discussed. According to the revised Ghent nosology 12 , she reached a systemic score of 7, and a MASS-like phenotype was diagnosed. Gene panel sequencing revealed the heterozygous missense variant c.193G > A p.(Asp65Asn) in COL2A1 (Table 1), which was classified as LPV according to the ACMG/ AMP guidelines 17 . Rules, considerations and data for assigning classification criteria are outlined in the Supplementary Information (Supplementary Results and Table S2). Segregation analysis in parents was not possible because the father had been deceased and the mother was not available (Fig. 2c). www.nature.com/scientificreports/ Subject 3 underwent clinical examination in our center at the age of 22 years. The family history was negative for connective tissue disorders, aortic aneurysms/dissections, and sudden cardiac deaths. He presented with tall stature (198 cm, + 3.0 SD), dolichocephaly, malar hypoplasia, retrognathia, high-arched palate, scoliosis, pectus carinatum with asymmetric thorax, arachnodactyly and hindfoot deformity. Ophthalmological investigation revealed iris transillumination and myopia (< 3 dpt). Bicuspid aortic valve was suspected by transthoracic echocardiography. A lumbar MRI scan showed dural ectasia (Fig. 2d). The diagnosis of HCTD was considered and discussed. The diagnosis of a MASS-like phenotype was established in this patient as well, as he has an aortic root Z-score < 2 and a systemic score of 9 according to the revised Ghent nosology 12 . Gene panel sequencing revealed the heterozygous missense variant c.4013G > A p.(Ser1338Asn) in COL2A1 (Table 1), which was classified as VUS (class 3) according to the ACMG/AMP guidelines 17 . Rules, considerations and data for assigning classification criteria are outlined in the Supplementary Information (Supplementary Results and Table S2). The parents of subject 3 provided us with self-collected buccal swabs, but they were not available for clinical examination.  www.nature.com/scientificreports/ Genotyping of parental DNA identified the heterozygous COL2A1 c.4013G > A transition in the father but not in the mother of subject 3 (Fig. 2c). Taken together, we ascertained four individuals from three families with a MASS-like phenotype and putative disease causing COL2A1 variants. The mean age of these four subjects was 33 ± 18.6 years with an even female to male ratio. All patients presented with tall stature, arachnodactyly and (kypho)scoliosis. Dura ectasia could be detected in all three patients who underwent spine MRI. Other than that, patients presented with further marfanoid features including pectus and foot deformities and typical craniofacial dysmorphism. None of the patients showed aortic root dilatation or a specific eye phenotype (i.e. ectopia lentis).
A MASS or Marfan-like phenotype (here defined as arachnodactyly and scoliosis without aortic dilatation and ectopia lentis) was recognized in 23 (11.5%) of a total of 200 probands who underwent gene panel sequencing. The three probands (subjects 1A, 2 and 3) positive for LPV/VUS in COL2A1 account for 13.0% (3/23) of patients with MASS/Marfan-like phenotypes (subject 1B was genotyped separately as part of a family segregation analysis). The remaining 20 probands with MASS/Marfan-like phenotypes did not show any COL2A1 variant. However, two patients had (likely) pathogenic variants in FBN1 and one patient had a likely pathogenic variant in FBN2. In another five patients of this subset, we identified eight VUS in the genes TNXB (3), B4GALT7 (1), COL5A1 (1), BGN (1), MYLK (1), and FLNA (1). These data suggest genetic heterogeneity of MASS/Marfan-like phenotypes.
All identified COL2A1 variants are located in the N-or the C-propeptide regions (Fig. 1a), which are removed during the processing from procollagen to collagen. The p.(Asp65Asn) substitution affects the von Willebrand factor type C (VWFC) repeat in the N-propeptide region of COL2A1 and the p.(Lys1312Asn) and p.(Ser1338Asn) substitutions are located in the fibrillar collagen C-terminal non-collagenous (NC1) (fibrillar collagen NC1) domain within the C-terminal propeptide (Fig. 1a).
We explored the structural impact of the p.Asp65Asn change by using the crystallographic structure of the COL2A1 VWFC domain (amino acids 29-98) as a template (PDB ID 5NIR) 20 . Molecular replacement of Asp 65 for an asparagine did not significantly affect intramolecular interactions and no Van-der-Waals (VDW) overlaps (non-covalent clashes) of Asn 65 with adjacent residues were predicted (Fig. S1a). Modelling of hydrophobicity and electrostatic potential of COL2A1 VWFC surface uncovered a hydrophobic patch comprising Val 46 , Ile 54 , Val 56 , Leu 63 and Ile 67 opposed by the negatively charged Asp 65 on one side (Fig. 1b, Fig. S1b). This hydrophobic pocket mediates binding with bone morphogenetic protein 2 (BMP2) 20 . The substitution of the negatively charged Asp 65 by a polar, uncharged asparagine may affect the binding properties and electrostatic potential of the hydrophobic interface (Fig. 1b, Fig. S1b).
For modelling the structural environment of Lys 1312 and Ser 1338 , the crystallographic structure of COL1A1 C-terminal propeptide homo-trimer (amino acids 1218-1464) was used as template (PDB ID 5K31) 21 . SWISS-Model was applied to build a COL2A1 model based on 68.72% sequence identity with COL1A1 22 . Molecular replacement of amino acid Lys 1312 for an asparagine resulted in loss of five intramolecular interactions (VDW contacts) with Asp 1297 and Trp 1299 and the formation of several potential non-covalent (i.e. VDW) clashes between the side chains of Asn 1312 and Trp 1299 (Figs. 1c, S2a). New VDW contacts of Asn 1312 with adjacent residues were not predicted (Figs. 1c, S2a). However, p.Lys1312Asn replacement changed surface hydrophobicity and electrostatic potential (Fig. S2b). Molecular replacement of Ser 1338 for an asparagine resulted in loss of six intramolecular interactions with Lys 1340 , Trp 1336 and Gly 1471 and the formation of several possible VDW clashes of Asn 1338 with Trp 1336 (Figs. 1d, S3a). New VDW contacts of Asn 1338 with adjacent residues were not predicted (Fig. S3a) and surface hydrophobicity and electrostatic potential were unchanged (Fig. S3b). Taken together, these data suggest that both p.(Lys1312Asn) and p.(Ser1338Asn) in COL2A1 induce structural alterations which possibly interfere with COL2A1 function.

Discussion
Clinical aspects. Here we report four subjects with a clinical presentation partially overlapping with Marfan syndrome (MFS; MIM #154700) and to a minor extent with Loeys-Dietz syndrome 3 (LDS3; syn. aneurysmsosteoarthritis syndrome; MIM #613795) 23,24 (Table 2) and fulfilling the criteria of MASS-like phenotype diagnosis according to the revised Gent nosology. However, we detected no VUS, LPV or PV in the respective disease genes FBN1 and SMAD3. Instead, we identified a COL2A1 variant putatively relevant to the disease in each of the patients. It has been demonstrated that heterozygous pathogenic FBN1 variants underlie the MASS phenotype in a total of 36 unrelated individuals [25][26][27][28][29] . On the other hand, FBN1 sequencing did not reveal pathogenic variants in 17 unrelated patients with MASS phenotype [30][31][32][33] . Accordingly, it was suggested that additional disease genes for the MASS phenotype remain to be identified 34 . We detected disease-relevant FBN1 variants only in two (8.7%) out of 23 patients with MASS/Marfan-like phenotypes. Additionally, we identified putative causal COL2A1 variants in 3 (13.0%) individuals. To our knowledge, this is the first time that a MASS-like phenotype has been associated with causal variants in any gene other than FBN1.
Several syndromic disorders are caused by pathogenic variants in the COL2A1 gene, also referred to as type II collagenopathies 4 . These disorders differ in their specific combinations of clinical features 3,11,35 . The most frequent form of type II collagenopathies, Stickler Syndrome 1 (STL1), is characterized by a multisystem involvement including ocular manifestations (congenital myopia, vitreous, retinal abnormalities), craniofacial manifestations (midface hypoplasia, micrognathia), palatal abnormalities (cleft palate, highly arched palate), skeletal manifestations (joint hypermobility, arachnodactyly, spine abnormalities, pectus deformity), sensorineural hearing loss, and early-onset degenerative arthritis 3,9 . In comparison, a type II collagenopathy involving a single organ system is the COL2A1-related osteoarthritis with mild chondrodysplasia (OSCDP) that is characterized by isolated skeletal/joint involvement, i.e. OA of hips, knees, shoulders, wrists, hands, joint stiffness, spine abnormalities (irregular endplates, mild platyspondyly, anterior wedging) and manifestations of the hands (enlarged MCP, PIP  www.nature.com/scientificreports/ Taken together, in addition to the overlapping clinical manifestation with STL1, OSCDP and other type II collagenopathies, the patients of this study present with pronounced skeletal features, resembling MFS or-to a minor extent-LDS3 (Table 2). This led us to apply the revised Ghent criteria 12 , and the clinical diagnosis of a MASS-like phenotype could be established in all four patients (aortic root Z-score < 2, at least one skeletal feature [i.e. pectus deformity], and a systemic score ≥ 5). The fact that a MASS-like phenotype was considered in our clinically different patients may suggest that the diagnosis of MASS is applicable not only to FBN1-related disorders, but also to other hereditary disorders of connective tissue. Since our patients were young at the final examination with an average age of 33 years, we cannot exclude that aortic events may appear and evolve later in life 31 . Thus, we will validate our diagnosis of a MASS-like phenotype in a longitudinal examination of the subjects and, if need be, revise it in a Marfan syndrome-like phenotype. Despite the wide clinical spectrum of type II collagenopathies, a MASS-or Marfan syndrome-like phenotype associated with COL2A1 variants has not been reported to date. Therefore, our findings further expand the phenotypical spectrum of COL2A1-related disorders.
Genetic and molecular aspects. The VWFC repeat in the N-propeptide region of COL2A1 (Fig. 1a) is involved in protein-protein interaction with transforming growth factor β (TGFβ ) and bone morphogenetic protein 2 (BMP2) 20,38 , and thereby, is able to control numerous biological processes including embryonic patterning, differentiation, and homeostasis of various tissue types 39 . Amino acid asparagine 65 (p.Asp65Asn) localizes in the so-called subdomain 1 (SD1; amino acids Asp 43 -Ile 68 ) of the VWFC repeat (Fig. 1a), adjacent to Cys 64 that links two strands of a three-stranded antiparallel β-sheet via a disulfide bond with Cys 55 (Fig. S1a) 20,40 . Given the opposite orientation of Asp 65 compared to Cys 64 on the β-sheet and the only minor effects of the p.Asp65Asn molecular replacement on intramolecular contacts/clashes (Fig. S1a), it is unlikely that this change interferes with disulfide bridge formation. It has been shown, that a hydrophobic patch including the nonpolar, hydrophobic amino acids Val 46 , Ile 54 , Val 56 , Leu 63 and Ile 67 mediates the interaction with BMP2 (Fig. 1b) 20 . Asp 65 that is highly conserved across different species (Fig. S1c, Table S2) borders this hydrophobic pocket. Asp 65 is an acidic, negatively charged amino acid, and therefore, may determine the binding pocket's physicochemical properties and substrate specificity. It can be inferred that the change of Asp 65 by a polar, uncharged Asn very likely affects binding properties and BMP2 substrate recognition. Note that charged amino acids form salt bridges, whereas polar residues can participate in hydrogen bond formation. In analogy, the different substrate specificities of the proteases trypsin and chymotrypsin is determined in part by a single amino acid within their almost identical hydrophobic substrate-binding pocket: residue 189 is a negatively charged asparagine in trypsin and a polar serine in chymotrypsin 41 . BMP2 is an essential participant in skeletal homeostasis, and the osteogenic signal provided by BMP2 is required for inherent reparative capacity of bone 42 . Changes in BMP activity in adult bone are associated with osteoporosis, OA 43 , and in the reduced ability to heal fractures, all major health problems that increase in severity with age 42 . Thus, altered BMP signaling may underlie the clinical manifestation in subject 2.
The fibrillar collagen NC1 domain (Fig. 1a) facilitates intracellular trimerization by association of C-propeptides from three procollagen molecules followed by zipper-like folding of the triple-helical region towards the N-terminal end 44 . In addition, C-propeptides also control the assembly of collagen fibrils in the ECM by maintaining the solubility of trimers 44 . The structure model of COL2A1 C-terminal propeptide homotrimer (Fig. S4) based on highly homologous COL1A1 21 shows that the amino acids Lys 1312 and Ser 1338 (and also the homologous amino acids Lys 1288 and Ile 1314 in COL1A1 and Lys 1291 and Thr 1317 in COL3A1, data not shown) are not involved in the interaction between adjacent collagen alpha-chains essential for trimerization 21,45 . These amino acids are also not involved in the stabilization of intra-chain disulfide bonds relevant for protein folding 21,45 . Lys 1312 localizes in the so-called base (Fig. S4) of the COL2A1 C-terminal propeptide in proximity to the Ca 2+ binding loop (Figs. 1c, S2a). The base regions stabilize the procollagen trimer 21,45 . Lys 1312 is conserved across mammalia but not across less related species (Fig. S2c, Table S2). The pathogenic variants p.Tyr1298Asn and p.Gly1305Ala, which localize near Lys 1312 in the COL2A1 base region, have been associated with type 2 collagenopathies 11, 46 . It has been suggested that these changes destabilize the base and disrupt the Ca 2+ -binding loop, respectively 45 . Similarly, our modelling data indicate that the p.(Lys1312Asn) variant compromises the shape and function of the base and/or Ca 2+ -binding loop (Fig. 1c, Fig. S2). Amino acid Ser 1338 is located in the petal of the COL2A1 C-terminal propeptide (Fig. S4), which assures procollagen chain selectivity 21,45 . However, Ser 1338 is not directly involved in inter-chain interactions 21 . Ser 1338 is conserved across vertebrates but not across less related species (Fig. S3c, Table S2). The p.(Ser1338Asn) change induces marked structural alteration that mainly concerns Trp 1336 (Fig. 1d, Fig. S3a). A pathogenic variant affecting the homologous amino acid in COL1A1 (p.Trp1312Cys) with putative negative consequences on intramolecular interactions has been identified in a patient with a lethal form of osteogenesis imperfect 45,47 . Moreover, the pathogenic variant p.Lys1313Arg in COL3A1 (homologous to COL2A1 Lys 1334 , Fig. S3a) that probably disrupts surface interactions has been shown in a patient with vascular Ehlers-Danlos syndrome 45,48 . These data demonstrate that the molecular region around Ser 1338 is disease critical. Based on our modelling data we conclude that the p.Ser1338Asn variant rather interferes with intramolecular interactions than affect molecule surface properties (Fig. 1d, Fig. S3). The function of C-propeptides in extracellular fibril assembly is not well established. Particularly, protein interactions and signaling functions of C-propeptides need further characterization 44 . Nonetheless, it has long been known that the release of the C-terminal propeptide decreases the solubility of collagen trimers by 1000-fold, thus leading to spontaneous fibril assembly 44,49 . Bone morphogenetic protein 1 (BMP1) is the proteinase mainly responsible for the cleavage of C-propetides and it involves cofactors such as the ECM protein procollagen C-endopeptidase enhancer (PCOLCE) that binds to C-propetides 44 . Given the surface exposition of COL2A1 amino acids Lys 1312 and Ser 1338 (Fig. S4), it is possible that protein-protein binding properties (binding affinities) and/or the proteolytic release of COL2A1 C-propeptide is affected, which finally may result in impaired extracellular fibril assembly. Summarized, the lack of valid data for the biological functions of COL2A1 C-propeptides-be it fibril assembly, www.nature.com/scientificreports/ signaling properties and/or even a structural function in the ECM-makes it challenging and speculative to assess the functional consequences of amino acid changes in this region.
Genotype-phenotype correlation. Pathogenic variants affecting the COL2A1 N-terminal propeptide region cause mild phenotypes such as STL1 4,50 . In line with this, subject 2 with the p.(Asp65Asn) change in the VWFC region also presents with mild, preferentially skeletal features. Interestingly, the COL2A1 missense variant c.170G > A p.(Cys57Tyr) in the VWFC region has been identified in a patient with an ocular form of Stickler syndrome but no extraocular features 51 . Though we are unable to explain the respective underlying pathophysiology, the drastic phenotypic differences between the patient with p.(Cys57Tyr) and subject 2 with p.(Asp65Asn) impressively reflect the pronounced clinical variability of type II collagenopathies 4,8,11 . The phenotypic spectrum of pathogenic missense variants affecting the COL2A1 C-propeptide is highly variable with a focus on skeletal and ocular manifestations 4,8,11,52 . Factors such as type and position of the amino acid substitution or tissue specific alternative splicing may contribute to the clinical manifestation 4,8,11,[52][53][54][55] . Pathogenic COL2A1 C-propeptide variants are especially known to result in atypical, severe and lethal phenotypes 4,8,46,52,53,[56][57][58][59] . However, this correlation is not perfect, because mild phenotypes such as STL1 or Czech dysplasia (MIM # 609162) have also been associated with C-terminal variants 4,11,52 . Accordingly, our two patients with the missense variants affecting the C-terminal fibrillar collagen NC1 domain presented with mild clinical manifestations. Thus, our data support the observation that pathogenic variants affecting surface-located C-propeptide residues result in mild to moderate phenotypes, whereas those that interfere with inter-or intramolecular interactions are usually associated with severe phenotypes 44,45 . Limitations. This study also has limitations. Intronic variants have been identified in FBN1-related disorders 27,60 ; such variants are missed by tNGS. Thus, in cannot be ruled out that deep intronic FBN1 variants cause the clinical manifestation in our patients. Likewise, tNGS does not capture structural variants, copy number variations or non-coding variants (e.g., in promoters or enhancers), which also applies to the FBN1 gene. Finally, only 62 genes were screened in our study; however, variants in untested genes may also underlie the clinical manifestation of the presented individuals.

Conclusion
We report here for the first time that COL2A1 variants are associated with a MASS-like phenotype characterized by tall stature, arachnodactyly, spine deformity, dolichostenomelia, foot deformity, arched palate, and pectus deformity. As dural ectasia and dolichocephaly have not been described in type II collagenopathies, our data expand the clinical spectrum associated with COL2A1 variants. The delineation of known COL2A1-associated phenotypes and definition of new ones can be helpful in the establishment of a valid genotype-phenotype correlation. Patients suggestive for MASS-like phenotype or Marfan-like syndrome (without aortopathy) and negative for disease-relevant FBN1 variants should be tested for sequence alterations in COL2A1.
Ethics approval. This study was performed in line with the principles of the Declaration of Helsinki.
Approval was granted by the Medical Chamber Hamburg (vote no. PV7038).
Consent to participate. The patients or patients' parents provided written informed consent for the participation in the study, clinical data and specimen collection and genetic analysis according to the Declaration of Helsinki and the national legal regulations [e.g., the German Genetic Diagnosis Act (GenDG)].

Consent for publication.
The patients or patients' parents signed informed consent regarding publishing their data and photographs.

Data availability
All data generated or analysed during this study are included in this published article [and its supplementary information files]. All variants and associated phenotypes have been submitted to the ClinVar database (submission name SUB9514030). www.nature.com/scientificreports/