Mutational spectrum and clinical signatures in 114 families with hereditary multiple osteochondromas: insights into molecular properties of selected exostosin variants

Abstract

Hereditary multiple osteochondromas (HMO) is a rare autosomal dominant skeletal disorder, caused by heterozygous variants in either EXT1 or EXT2, which encode proteins involved in the biogenesis of heparan sulphate. Pathogenesis and genotype–phenotype correlations remain poorly understood. We studied 114 HMO families (158 affected individuals) with causative EXT1 or EXT2 variants identified by Sanger sequencing, or multiplex ligation-dependent probe amplification and qPCR. Eighty-seven disease-causative variants (55 novel and 32 known) were identified including frameshift (42%), nonsense (32%), missense (11%), splicing (10%) variants and genomic rearrangements (5%). Informative clinical features were available for 42 EXT1 and 27 EXT2 subjects. Osteochondromas were more frequent in EXT1 as compared to EXT2 patients. Anatomical distribution of lesions showed significant differences based on causative gene. Microscopy analysis for selected EXT1 and EXT2 variants verified that EXT1 and EXT2 mutants failed to co-localize each other and loss Golgi localization by surrounding the nucleus and/or assuming a diffuse intracellular distribution. In a cell viability study, cells expressing EXT1 and EXT2 mutants proliferated more slowly than cells expressing wild-type proteins. This confirms the physiological relevance of EXT1 and EXT2 Golgi co-localization and the key role of these proteins in the cell cycle. Taken together, our data expand genotype–phenotype correlations, offer further insights in the pathogenesis of HMO and open the path to future therapies.

Introduction

Hereditary multiple osteochondromas (HMO), previously known as hereditary multiple exostoses, is a rare autosomal dominant skeletal disorder with a frequency of ~1:50 000 individuals worldwide. HMO is caused by heterozygous variants in either EXT1 (MIM, # 133700) or EXT2 (MIM, # 133701) in about 90% of the cases (1). Phenotypic hallmarks of HMO are multiple osteochondromas, which are benign cartilage-capped bone tumors sprouting from the long bones metaphyses and skeletal deformities. Osteochondromas can arise everywhere, but mainly occur in the distal femur, proximal tibia, fibula and humerus (2–5).

HMO is a highly variable condition within and between families. In HMO, osteochondromas usually appear during infancy and tend to present with highly heterogeneous symptoms during adolescence and adulthood due to their age-dependent increase in size and number (6). The most severe clinical implication of osteochondromas is the risk of malignant degeneration in chondrosarcomas or osteosarcomas with a lifetime incidence ranging from 0.5% to 20% in HMO (7). Some preliminary genotype–phenotype correlations have been published and mostly include an apparently increased risk of malignant degeneration (8), a larger number of lesions (9), a more severe phenotype and male preponderance in families with variants in EXT1 (10).

EXT1 and EXT2 encode exostosin 1 (EXT1) and 2 (EXT2), respectively. These proteins are two heparan sulfate glycosyltransferases involved in heparan sulphate (HS) synthesis and elongation, and, thus, are implicated in chondrocyte proliferation and differentiation. EXT1 and EXT2 are composed by an N-terminal exostosin domain and a C-terminal glycosyl-transferase family 64 domain (11). Although EXT1 and EXT2 are ubiquitously expressed in many tissues, the most common pathogenic effect of their alteration affects the growing bones (11).

The molecular mechanism underlying such a phenomenon remains poorly defined. On physiological conditions, EXT1 and EXT2 form a Golgi-located hetero-oligomeric complex (12), which catalyzes HS biosynthesis by either the GlcNAc transferase (GlcNAcT) or GlcA transferase (GlcAT) activity. Thus, they are HS polymerases synthesizing the disaccharide-repeating region of the growing HS chains. The EXT1/2 hetero-oligomer has a higher glycosyltransferase activity compared to EXT1 and EXT2 alone. This suggests that both the integrity of EXT1/2 hetero-oligomer and its specific Golgi-localization represent highly conserved properties, which preserve EXT1 and EXT2 glycosyltransferase activity in physiological conditions (13). In addition, EXT1 alone as well as the EXT1/EXT2 complex can catalyse in vitro polymerization of the HS backbone structure on an oligosaccharide primer, whereas EXT2 harbors a lower catalytic activity (14). It has been also proposed that EXT2 is not directly involved in HS backbone elongation, but rather serves as a chaperon by delivering EXT1 to the Golgi apparatus (15,16). Maturation and elongation of HS represent the main role of EXT1 and EXT2, which also affect additional different cellular processes including cell survival, division and migration. Accordingly, it has been reported that reduced EXTs expression decreased HS biosynthesis and consequently suppressed the FGF2-dependent proliferation of mouse fibroblasts (17).

Here, we report the EXT1 and EXT2 mutational spectrum in 114 Italian families, 55 novel pathogenic variants and additional genotype–phenotype correlation data. Expression studies for selected EXT1 and EXT2 alleles were carried out to examine subcellular localization and effect on cell proliferation.

Table 1

Mutational analysis of EXT1 gene

EXT1NM_000127
ID	EX	N.P.	N.F.	DNA variants	Predicted amino acid change	Reference
EXT1_01	1	2	1	c.64_71del	p.(Tyr22Argfs*5)	Present work
EXT1_02	1	4	2	c.173del	p.(Phe58Serfs*78)	Ciavarella, (2013) Gene
EXT1_03	1	3	1	c.250C>T	p.(Gln84*)	Francannet (2001) J Med Genet
EXT1_04	1	1	1	c.262dup	p.(Ala88Glyfs*101)	Present work
EXT1_05	1	2	1	c.301del	p.(Glu101Serfs*35)	Present work
EXT1_06	1	4	1	c.351C>G	p.(Tyr117*)	Ciavarella, (2013) Gene
EXT1_07	1	1	1	c.437del	p.(Ser146Trpfs*11)	Present work
EXT1_08	1	1	1	c.500C>A	p.(Ser167*)	Wuyts (2005) Clin Genet
EXT1_09	1	1	1	c.552G>A	p.(Trp184*)	Sarrión (2013) Sci Rep
EXT1_10	1	2	2	c.60_64del	p.(Tyr22Argfs*6)	Present work
EXT1_11	1	1	1	c.600G>A	p.(Trp200*)	Present work
EXT1_12	1	2	1	c.620del	p.(Val207Glyfs*45)	Present work
EXT1_13	1	1	1	c.637C>T	p.(Gln213*)	Present work
EXT1_14	1	2	1	c.742del	p.(Arg248Glyfs*4)	Present work
EXT1_15	1	1	1	c.752del	p.(Leu251*)	Present work
EXT1_16	1	1	1	c.796_797ins	p.(Phe266Leufs*24)	Present work
EXT1_17	1	1	1	c.806_807ins	p.(Lys269Asnfs*20)	Present work
EXT1_18	1	1	1	c.808A>T	p.(Arg270Trp)	Present work
EXT1_19	1	3	2	c.838A>G	p.(Arg280Gly)	Wuyts (1998) Am J Hum Genet
EXT1_20	1	1	1	c.943_944del	p.(Asp315Glnfs*5)	Wuyts (2000) Am J Hum Genet
EXT1_21	1	2	2	c. (?-1)_(962+ 1_963-1)del	p.0?	White (2004) Hum Mutat
EXT1_22	2	2	2	c.966T>G	p.(Tyr322*)	Lonie (2006) Hum Mutat
EXT1_23	2	3	3	c.1018C>T	p.(Arg340Cys)	Philippe (1997) Am J Hum Genet
EXT1_24	2	7	5	c.1019G>A	p.(Arg340His)	Raskind (1998) Hum Mutat
EXT1_25	2	1	1	c.1031C>T	p.(Ser344Phe)	Present work
EXT1_26	3	1	1	c.1064G>T	p.(Cys355Phe)	Ishimaru(2016) BMC Genet
EXT1_27	3	2	1	c.1065C>A	p.(Cys355*)	Present work
EXT1_28	3	2	1	c.1091G>A	p.(Trp364*)	Present work
EXT1_29	1	1	1	c.1160del	p.(Leu387Tyrfs*16)	Present work
EXT1_30	3	1	1	c.1162C>T	p.(Gln388*)	Wuyts (2005) Clin Genet
EXT1_31	3	1	1	c.1164+1G>A	p.0?	Present work
EXT1_32	4	2	1	c.1213A>T	p.(Arg405*)	Dobson (2000) Eur J Hum Genet
EXT1_33	4	1	1	c.1278_1279del	p.(Leu427Argfs*14)	Present work
EXT1_34	4	1	1	c.1284+1G>T	p.0?	Present work
EXT1_35	4	1	1	c.1284+1G>A	p.0?	Present work
EXT1_36	4	2	1	c.1285-2A>G	p.0?	Present work
EXT1_37	5	2	1	c.1316C>G	p.(Ser439*)	Present work
EXT1_38	5	2	1	c.1335G>A	p.(Trp445*)	Present work
EXT1_39	5	1	1	c.1396_1399del	p.(Pro466Thrfs*6)	Present work
EXT1_40	5	1	1	c.1418-2A>G	p.0?	Present work
EXT1_41	6	2	1	c.1425del	p.(Lys475Asnfs*13)	Present work
EXT1_42	6	1	1	c.1469del	p.(Leu490Argfs*9)	Xu (1999) Hum Gen
EXT1_43	6	1	1	c.1483C>T	p.(Gln495*)	Pei (2010) Genet T M B
EXT1_44	6	1	1	c.1490del	p.(Val497Glyfs*2)	Present work
EXT1_45	6	1	1	c.1522C>T	p.(Gln508*)	Pedrini, not yet published*
EXT1_46	6	3	2	c.1536+1G>T	p.0?	Present work
EXT1_47	6	2	2	c.1537-1G>T	p.0?	Present work
EXT1_48	6	1	1	c.1537-2A>T	p.0?	Ciavarella, (2013) Gene
EXT1_49	7	1	1	c.1561A>T	p.(Lys521*)	Present work
EXT1_50	7	1	1	c.1633-2A>T	p.0?	Pedrini, (2013) Hum Mutat
EXT1_51	8	1	1	c.1642del	p.(Ser548Alafs*73)	Present work
EXT1_52	8	5	2	c.1714del	p.(Thr572Glnfs*49)	Present work
EXT1_53	9	1	1	c.1746G>A	p.(Trp582*)	Francannet (2001) J Med Genet
EXT1_54	9	1	1	c.1757_1758del	p.(Pro586Argfs*15)	Present work
EXT1_55	9	1	1	c.1759G>T	p.(Glu587*)	Present work
EXT1_56	9	4	1	c.1797G>A	p.(Trp599*)	Seki (2001) Am J Med Genet
EXT1_57	9	1	1	c.1841del	p.(Asn614Thrfs*7)	Jennes (2009) Hum Mutat
EXT1_58	10	1	1	c.1921_1922insAGCC	p.(Ser641Lysfs*14)	Pedrini, not yet published*
EXT1_59	10	2	2	c.1942C>T	p.(Gln648*)	Vink (2005) Eur J Hum Genet
EXT1_60	11	1	1	c.2071dup	p.(Arg691Profs*4)	Present work
EXT1_61	11	2	2	c.2098C>T	p.(Gln700*)	Pedrini (2005) Hum Mutat
EXT1_62	11	2	2	c. (?-1)_(*1_?)del	p.0?	Wuyts (2002) Am J Med Genet

#: number.

Results

Molecular findings

A total of 114 HMO index patients (56 females, 58 males) were identified with pathogenic EXT1 (79, 69.3%) or EXT2 (35, 30.7%) variants. In 26 families, biological samples of two or more affected members were available, for a total of 44 additional affected individuals (23 females, 21 males). In familial cases, segregation of the disease occurred from an affected mother in 21 meioses and from an affected father in 9 meioses (P = 0.02846, assuming a priori chance of 1:1). In these cases, segregation with the disease was demonstrated for all identified variants (Supplementary Material, Fig. S1). Of the remaining, we were not able to distinguish true sporadic cases from familial cases with a single available affected individual. In total, we found pathogenic variants in 158 affected individuals. Eighty-seven different variants (55 novel and 32 known) were identified: 62 (36 novel) in EXT1 and 25 (19 novel) in EXT2. A variant was defined ‘novel’ if not reported in the multiple osteochondroma mutation database (MODB) (http://medgen.ua.ac.be/LOVDv.2.0/home.php) (2). All novel variants have been submitted to MODB database, with the following code number: IDs #00208634–#00208709 for EXT1 patients, IDs #00208710–#00208749 for EXT2 patients.

From a total of 87 variants, 36 (42%, 23 in EXT1 and 13 in EXT2) were frameshift, 28 (32%, 22 in EXT1 and 6 in EXT2) nonsense, 10 (11%, 6 in EXT1 and 4 in EXT2) missense, 9 (10%, all in EXT1) splicing and 4 (5%, 2 in EXT1 and 2 in EXT2) intragenic rearrangements. A summary of the identified variants was reported in Tables 1 and 2. Of the 62 EXT1 variants, 13 (21.3%) recurred in two or more families; p.(Arg340His) was the most common and occurred in five pedigrees. Of the 25 EXT2 variants, two (8%) were recurrent; p.(Asn288Ser) occurred in 10 families and p.(Pro351Leu) in two.

Table 2

Mutational analysis of EXT2 gene

EXT2NM207122
ID	Ex	N.P.	N.F.	DNA variants	Predicted amino acid change	Reference
EXT2_01	2	15	10	c.67C>T	p.(Arg23*)	Wuyts (1998) Am J Hum Genet
EXT2_02	2	1	1	c.81del	p.(Thr28Profs*31)	Pedrini (2005) Hum Mutat
EXT2_03	2	1	1	c.151G>T	p.(Glu51*)	Present work
EXT2_04	2	1	1	c.173del	p.(Val58Glufs*54)	Ciavarella (2013) Gene
EXT2_05	2	1	1	c.244del	p.(Asp82Ilefs*30)	Present work
EXT2_06	2	1	1	c.315del	p.(Val106Cysfs*6)	Present work
EXT2_07	2	1	1	c.380del	p.(Arg128Glyfs*142)	Wuyts (2005) Clin Genet
EXT2_08	2	3	1	c.390T>A	p.(Tyr130*)	Present work
EXT2_09	2	1	1	c.448del	p.(Ala150Profs*120)	Present work
EXT2_10	2	1	1	c.454_457del	p.(Val154Profs*115)	Present work
EXT2_11	3	1	1	c.607del	p.(Leu203Trpfs*67)	Present work
EXT2_12	4	2	1	c.668G>C	p.(Arg223Pro)	Shi (2000) Hum Mutat
EXT2_13	4	4	1	c.684_688delins	p.(Ser229Alafs*41)	Present work
EXT2_14	4	1	1	c.728del	p.(Pro243Glnfs*27)	Present work
EXT2_15	5	3	1	c.863A>G	p.(Asn288Ser)	Present work
EXT2_16	5	1	1	c.863del	p.(Asn288Thrfs*44)	Present work
EXT2_17	5	1	1	c.883_887 del	p.(Ser295Profs*2)	Present work
EXT2_18	6	1	1	c.991_992ins	p.(Leu331*)	Present work
EXT2_19	6	1	1	c.1004del	p.(Leu335Tyrfs*101)	Present work
EXT2_20	6	1	1	c.1052C>T	p.(Pro351Leu)	Present work
EXT2_21	6	1	1	c.1016G>A	p.(Cys339Tyr)	Li (2009) Genet T M B
EXT2_22	8	1	1	c.1188G>A	p.(Trp396*)	Present work
EXT2_23	8	1	1	c.1286G>A	p.(Trp429*)	Present work
EXT2_24	10	2	1	c.(1495+1_1496-1)_ (1661+1_1662-1)del	p.0?	Present work
EXT2_25	16	4	2	c.(1495+1_1495-1)_ (*1_?)del	p.0?	Present work

NM_000401 followed for MLPA and QPCR analysis; #: number.

Genotype–phenotype correlations

Detailed clinical information was available for 42 EXT1 and 27 EXT2 subjects (Supplementary Material, Tables S1 and S2). Among them, there were 38 females and 31 males, with an age range from 4 to 74 years. Rough data show an excess of reported/identified osteochodroma/exostoses at the long bones of the upper and lower limbs. Chest bones (costae, scapulae, claviculae) were the second most commonly affected sites. Occasional vertebral osteochodroma/exostoses were registered in EXT1 patients, while none EXT2 patient presented similar lesions. Intellectual disability was reported in two EXT1 patients. In both cases, a single heterozygous nonsense variant was identified [#13, c.637C>T p.(Gln213^*); #27, c.1065C>A p.(Cys355^*)]. A biallelic disorder in EXT1 was excluded due to the absence of a second causative point variant. A compound phenotype with the involvement of a second gene associated with human neurodevelopmental disorders was not further investigated in this study.

The 69 patients with available clinical data were subgrouped according to involved gene in order to explore further genotype–phenotype correlations (42 and 27 patients carrying EXT1 and EXT2 variants, respectively, Supplementary Material, Tables S1–S2). No significant differences for age and sex were registered between the two groups (EXT1 versus EXT2 affected patients) (Supplementary Material, Table S3). The mean number of total osteochondromas was ~10 and ~7 in EXT1 and EXT2 patients, respectively (P = 0.0351) (Fig. 1, Supplementary Material, Table S4–S7). Such excess was more evident in distal humerus (P < 0.0001), distal ulna (P = 0.0011), ribs (P = 0.0014), proximal humerus (P = 0.0173) and distal radius (P = 0.0195) (Supplementary Material, Table S4). By considering patients with EXT1 variants, females had more lesions in ribs (P < 0.0001) and distal ulna (P = 0.0451), while males in feet (P = 0.0147) (Supplementary Material, Table S5). In EXT2 patients, males had more lesions in hands (P = 0.0026) and feet (P = 0.0058) (Supplementary Material, Table S6). Among males, those with EXT1 variants had more osteochondromas in general (P = 0.0085) (Fig. 1), and, more specifically, in distal humerus (P = 0.0007), distal ulna (P = 0.0009), ribs (P = 0.0017) and hands (P = 0.0048) (Supplementary Material, Table S7). Among females, the overall number of lesions was comparable between those with EXT1 and EXT2 variants. However, more osteochondromas were observed in distal humerus (P = 0.015) and distal radius (P = 0.0209) in EXT1 females and in proximal humerus (P = 0.045) in individuals with EXT2 variants (Supplementary Material, Table S8). By comparing the total number of osteochondromas in individuals carrying the same variant, no significant difference was observed between sexes.

Figure 1

Correlation study. Mean number of exostoses according to EXT variant and gender.

Golgi subcellular localization of EXTs mutants

We investigated the pathogenic effects of selected EXT1 and EXT2 variants by assessing the hypothesis that variants in EXTs may cause the loss of Golgi physiological localization and, consequently, impair glycosyltransferase activity. To evaluate the importance of functional domains of EXT1 and EXT2 proteins, we tested a set of truncating variants which compromising exostosin N-terminal domain or glycosyl-transferase 64 C-terminal domain, in either EXT1 and EXT2 gene (see Materials and Methods, Minigene generation). Furthermore, to understand the role played by missense variants in HMO pathogenesis, a novel missense variant in the EXT2 gene was also examined.

Wild-type and mutated EXT1/2 vectors were tested for the expression by immunoblotting (Supplementary Material, Fig. S2A–C). For EXT1, U₂OS cell lines were transiently transfected with EGFP-Myc-EXT1 or HA-EXT1 wild-type proteins and corresponding mutants EGFP-Myc-EXT1-p.(Cys355^*) and HA-EXT1-p.(Leu427Argfs^*14). For EXT2, the same experiment was carried out by transfecting FLAG-EXT2 wild-type and FLAG-p.(Pro351Leu), FLAG-p.(Leu335Tyrfs^*101) and FLAG-p.(Pro243Glnfs^*27) vectors. Wild-type EXT1 and EXT2 physiologically co-localized with GALNT2 Golgi marker (Fig. 2A–B). Conversely, all mutants showed an abnormal localization pattern. Specifically, p.(Cys355^*) EXT1 mutant assumed a diffuse localization mainly into the perinuclear region retaining a partial Golgi localization; the p.(Leu427Argfs^*14) EXT1 localized into both the Golgi and nucleus (Fig. 2A). The p.(Pro351Leu) EXT2 mutant showed a physiological Golgi localization, but was also localized partially into the perinuclear region. The p.(Pro243Glnfs^*27) EXT2 mutant entirely de-localized into the nucleus, while the p.(Leu335Tyrfs^*101) EXT2 mutant, seemed co-localize partially with Golgi marker, but also de-localized diffusely in the cytoplasm (Fig. 2B).

Figure 2

Confocal study. (A) U₂OS cells transiently transfected with EGFP-Myc-EXT1, EGFP-Myc EXT1 p.(Cys355^*) and HA-EXT1 p.(Leu427Argfs^*14) mutants, were stained with anti-GALNT2 antibody. (B) U2OS cells transiently transfected with FLAG-EXT2 and FLAG-EXT2 p.(Pro351Leu), FLAG-EXT2 p.(Pro243Glnfs^*27) and FLAG-EXT2 p.(Leu335Tyrfs^*101) mutants, were stained with anti-GALNT2 antibody. (C) U₂OS cells transiently co-transfected with FLAG-EXT2 and EGFP-Myc-EXT1 wild type or EGFP-Myc-EXT1 p.(Cys355^*) and HA-EXT1 p.(Leu427Argfs^*14) mutated protein, respectively, were stained for anti-FLAG and anti-HA antibodies. (D) U₂OS cells, transiently co-transfected with EGFP-Myc-EXT1 and FLAG-EXT2 or FLAG-EXT2 p.(Pro351Leu), FLAG-EXT2 p.(Pro243Glnfs^*27) and FLAG-EXT2 p.(Leu335Tyrfs^*101) mutants, were stained with anti-FLAG antibody.

Subcellular co-localization of EXTs variants

To address if the selected EXT1 and EXT2 mutants can dominantly loss their co-localization, we tested their sub-cellular co-localization by co-transfecting both EXT1 and EXT2. Mutant EXT1 and EXT2 vectors (described above) were co-transfected with EXT2 or EXT1 wild-type vectors and analysed for microscopy staining in U₂OS cells. Both EXT1 mutants failed the ability to co-localize entirely with EXT2. In particular, p.(Cys355^*) EXT1 mutant showed a diffuse sub-cellular localization, the p.(Leu427Argfs^*14) EXT1 mutant assumed a fractional nuclear localization; both mutants also partially co-localize with EXT2 (Fig. 2C). Concerning EXT2 variants, the p.(Pro351Leu) EXT2 and p.(Leu335Tyrfs^*101) EXT2 mutants showed a diffuse and perinuclear localization somewhat co-localizing with EXT1, while the nuclear localization of p.(Pro243Glnfs^*27) EXT2 mutant abrogated definitely EXT1 co-localization (Fig. 2D).

Cell viability report of EXTs mutants

Data from the literature shows that EXT family member proteins regulate Fibroblast Growth Factor 2 (FGF2) signalling and that reduced EXTs expression correlates with a reduction of cell growth (17). Additionally, some inherited EXTs variants accumulate in the Endoplasmic Reticulum (ER) by losing their Golgi localization (18) and this suggests an increase of ER stress. In order to investigate these aspects, the selected EXTs wild-type and mutants proteins were transfected in HEK293 cell lines and the rate of cell grown for 5 days was evaluated (Fig. 3A and B). Cells expressing EXT mutants proliferated more slowly than cells expressing the corresponding wild-type proteins (Fig. 3).

Figure 3

Viability assay on variants expressing cell lines. (A) Graphic result of viability assay in HEK293 cells transiently transfected with EGFP-Myc-EXT1, EGFP-Myc-EXT1 p.(Cys355^*) or HA-EXT1 p.(Leu427Argfs^*14) mutated proteins. The rate of proliferation viability was annotated up to 5 days. Data were analysed using a Student’s t-test, and asterisks indicate significant differences compared to wild-type EXT1 expressing cell line, ^*: P ≤ 0.05; ^**: P ≤ 0.01. (B) Proliferation study. Graphic result of viability assay in HEK293 cells transiently transfected with FLAG-EXT2 or FLAG-EXT2 mutated proteins. The rate of proliferation was been annotated up to 5 days. Data were analysed using a Student’s t-test, and asterisks indicate significant differences compared to wild-type EXT2 expressing cell line, ^*: P ≤ 0.05; ^**: P ≤ 0.01.

Discussion

In this study, we presented the molecular and clinical findings in a cohort of 114 index Italian patients with HMO and provided functional analyses on selected HMO-linked variants. The extreme allelic heterogeneity was a feature of our sample with 55 novel variants from a total of 87 identified (63.2%). We confirmed the preponderance of EXT1 variants in the etiology of HMO with a proportion EXT1/EXT2>2/1. In this cohort, truncating variants (i.e. nonsense, frameshift and splicing) occurred in 90% of the cases. The 15 recurrent variants (12 truncating and 3 missense) accounted for 38% of pathogenic alleles identified in our sample, with the two most common [p.(Arg340His) in EXT1 and p.(Arg23^*) in EXT2] explaining 15% of cases. In EXT1, we confirmed that the residue Arg340 is a mutational hotspot, as it was involved in eight families in our sample with two different amino acid changes: p.(Arg340His) in five cases and p.(Arg340Cys) in three cases (19). Of the two recurrent EXT2 variants, the most common p.(Arg23^*) was already reported (7), while the c.(1495+1_1495-1)_(^*1_?)del was novel. Taken together, these data confirm the need of a wide genetic approach for the molecular confirmation of HMO also in cohorts with a homogeneous geographical origin. Although some variants are recurrent, exon analysis prioritization may not represent an issue in the era of next generation sequencing diagnostics.

Our restricted genotype–phenotype analysis confirmed a tendency towards a more severe phenotype (i.e. a higher number of lesions) in EXT1 patients compared to EXT2. We failed to demonstrate a more severe involvement in males, as previously registered (10). A variety of biases may exist, including: (i) partially available clinical data, (ii) size of patients’ sample and (iii) different geographical origin and genetic background. Slight, but significant differences were registered in the site distribution of osteochondromas between sexes for the two genes. Data are exploratory as the patients’ sample was small and explanations are hardly identified. However, such divergences might be the result of the background interactions between the perturbed exostosins activity at the cellular level and sexual dimorphism during bone maturation. The apparently significant preponderance of transmitting mothers in our sample remains without a reasonable explanation. Although very preliminary, this evidence could be related to sex-influenced, though minor differences on prezygotic selection of gametes and/or reproductive fitness of individuals carrying and not carrying an EXT1/EXT2 heterozygous variant, in these families.

In most cases, the biological consequences of EXT1 and EXT2 variants on protein function have not been tested directly, but have been usually predicted and inferred on structural considerations (20). This could raise the question as to whether some previously reported EXT1/EXT2 variants may be really pathogenic or have a minimal or clinically insignificant detrimental effect on protein function. Here, we tried to demonstrate the pathogenic effect of selected variants in order to support previous generalizations, explore the molecular and cellular pathogenesis of HMO and propose possible confirmatory assays for a future clinical use. Early studies suggest that EXT1 and EXT2 may act as a tumor suppressor because losses of heterozygosity for markers around these genes may occur in sporadic and exostosis-derived chondrosarcomas (21). Differently, it is also evident that EXT1 and EXT2 positively respond to FGF signalling, which is a well-known anti-chondrogenetic factor (21–27). Moreover, as at least some EXT variants loss their Golgi localization (16,18) and probably accumulate into the ER, an inherited defect of EXTs might associate with an increase of ER stress and, therefore, with a reduction of cell viability. Accordingly, we reported that cells expressing EXT mutants slightly reduced cell viability compared to wild-type expressing cells and this might correlate with a reduction of HS maturation.

Golgi co-localization of EXT1 and EXT2 is a prerequisite for optimal synthesis and assembly of HS (21–23). Perturbed Golgi localization of EXT1 and EXT2 has been previously reported for selected EXT1 and EXT2 variants (12). Our confocal study in the human osteosarcoma cell line U₂OS supported these data by demonstrating that the sub-cellular localization of all selected EXT1 and EXT2 mutants only partially resumed their Golgi localization, differently from the corresponding wild-type protein which correctly localized in the Golgi apparatus. In fact, in addition to Golgi location, the different mutants showed a diffuse and/or perinuclear pattern localization. The single p.(Pro243Glnfs^*27) EXT2 mutant completely delocalized from Golgi and exhibited a total nuclear localization. We also carried out an exploratory colocalization study with the two EXT proteins and demonstrated that both EXT1 and EXT2 mutants partially or completely failed to co-localize with the wild-type paralog partner. These findings provided a rationale to explain how inherited variants in either of the two EXT genes could alter the physiological localization of EXT encoding proteins and probably resulted in an alteration of their enzymatic activity. In light of our findings, we can speculate that genomic variants in EXTs may be sufficient to cause deficiencies in the processing of proteoglycans destined for the cell surface or extracellular matrix and this provides compelling evidence that osteochondroma formation in HMO is caused by a deficiency in HS polymerase.

In conclusion, we broadened the spectrum of pathogenic variants in EXT1 and EXT2 and contributed to genotype–phenotype correlations in HMO. We further explored functional consequences of five EXT1 and EXT2 variants affecting (i) Golgi localization of the encoded proteins, (ii) EXT1/EXT2 co-localization into Golgi apparatus and (iii) cell cycle by reducing proliferation of U₂OS cells. Taken together, these data would help physicians and researchers in elucidating disease pathogenesis and exploring tailored treatments for HMO.

Materials and Methods

Patients’ selection

Patients were enrolled from different Italian clinical providers, who requested molecular diagnosis of HMO and sent patients’ samples to the Division of Medical Genetics and CSS-Mendel Institute of Fondazione IRCCS-Casa Sollievo della Sofferenza. A minimal criterion of radiological evidence of multiple (>3) lesions compatible with osteochondromas/exostoses was assured in all patients (criteria of access to molecular testing). When possible, clinical providers were asked to send all available clinical details by filling a pre-compiled questionnaire mainly exploring the distribution of skeletal lesions at the time of ascertainment. All patients signed a dedicated informed consent. Molecular studies carried out in this work were based on routine clinical care and Institutional Review Board approval was not requested.

Molecular screening

Genomic DNA was extracted from fresh and/or frozen peripheral blood leukocytes of the probands and their available relatives by using an automated DNA extractor and commercial DNA extraction Kits (EZ1, Qiagen, Denmark). Molecular analysis of EXT1 and EXT2 genes was performed by PCR amplification and direct sequencing (BigDye Terminator Cycle Sequencing Kit v 1.0 and ABI Prism 3130XL Sequencer, Applied Biosystems, US) of all coding exons and flanking exon-intron junctions sequences of EXT1 (RefSeq NM_000127.2) and EXT2 [RefSeq NM_207122.1 for Sanger sequencing; NM_000401 for multiplex ligation-dependent probe amplification (MLPA) studies, see below]. All patients resulted negative at the Sanger sequencing underwent MLPA analysis by using the Salsa MLPA Kit (code #P215-B1 EXT, MRC Holland, Amsterdam, The Netherlands), in accordance with the manufacturer’s instructions, in order to detect exon or multiexon deletions/duplications. Quantitative real-time PCR (qPCR) next confirmed positive results of MLPA analysis. qPCR was carried out as previously described (24), with the oligo pairs based on the UCSC GRCh37/hg19 assembly. Quantitative analysis of the PCR product was carried out as reported in (25). The comparative Ct method reported in (26) was used to measure relative amount.

Pathogenicity of the identified novel variants was supported by either absence or showing a frequency lower than expected by considering the disease frequency of HMO in ExAC and/or 1000 Genomes databases. The four novel missense variants [p.(Arg270Trp) and p.(Ser344Phe) in EXT1, and p.(Asn288Ser) and p.(Pro351Leu) in EXT2], were also studied by in silico analysis by the following software tools: Polyphen-2 (version 2.2.2, available at: http://genetics.bwh.harvard.edu/pph), PROVEAN (version 1.1.3, available at: http://provean.jcvi.org/index.php), SIFT (version 1.03, available at: http://sift.jcvi.org/), MutationTaster 2 (27), MetaSVM and MetaLRT (28), M-CAP (29), CADD (version 1.3,42), DANN, fathmm-MKL (30) and MutationAssessor (31). For the EXT2 p.(Pro351Leu) variant, co-segregation with the disease in three affected family members was demonstrated.

Minigene generation

Total RNA from HEK293, extracted by using RNeasy® Mini kit (Qiagen, France), was reverse transcribed by using QuantiTect® Reverse Transcription Kit (Qiagen), according to the manufacturer’s instructions. cDNA-PCR fragments spanning the exons 1–11 of EXT1 (NM_000127.2) and 1–14 of EXT2 (NM_207122.1) were generated and cloned into the pcDNA3.1-EGFP-Myc or pcDNA3.1-HA and pcDNA3-p3XFLAG vectors, respectively. Site-directed mutagenesis (QuickChange II kit, Stratagene, USA) was employed to generate the EXT1 c.1065C>A p.(Cys355X) variant in the EGFP-Myc-EXT1vector, the EXT1 c.1278_1279delAC p.(Leu427Argfs^*14) variant in the HA-EXT1vector, and the EXT2 c.1052C>T p.(Pro351Leu), c.1004delT p.(Leu335Tyrfs^*101) and c.728delC p.(Pro243Glnfs^*27) variants in the FLAG-EXT2 vector. The five selected variants were considered representative of the mutational spectrum seen in our patients’ cohort including different truncating variants, which clearly compromise one or both functional EXT domains and a novel missense variant. All minigenes carrying wild-type or variant alleles were verified by sequencing.

Cell culture and transfection

U₂OS and HEK293 cells were grown in Dulbecco’s modified Eagle’s medium (Thermo Fisher, USA), supplemented with 10% fetal bovine serum (Thermo Fisher), 100 units/ml penicillin and 100 μg/ml streptomycin sulfate and maintained in a 5% CO₂ incubator at 37°C. Plasmids transfections were performed using calcium phosphate method or Lipofectamine LTX (Life Technologies, USA) as previously described (32).

Immunoblotting

HE293 cell lines were lysed in 0.5% Nonidet P-40, 1× complete inhibitor (Roche, USA), 1.5 mm phenylmethylsulfonyl fluoride and PBS 1X. Proteins were separated on a Sodium dodecyl sulfate (SDS)-polyacrylamide gel. Western blots were prepared as previously reported in (32) and probed using a set of antibodies including anti-MYC (1:1000, # 49001 Roche, USA), anti-HA (1:1000 dilution, # 26183 Thermo Fisher Scientific, USA), anti-FLAG (1:1000 dilution, # F3165 Sigma Aldrich, Germany). Bound primary antibodies were visualized using Enhanced ChemiLuminescence (ECL) western blotting or ECL plus western blotting detection reagents (GE Healthcare, UK).

Confocal study

For immunocytochemical analysis, U₂OS cells were plated in 3 cm culture dishes and transiently transfected with expression plasmids. Twenty-four hours after transfection, cells were subjected to immunofluorescence protocol, as previously reported (32). Cells were counterstained with anti-GALNT2 antibody (1:500 dilution, #PA521541 Thermo Fisher Scientific), anti-FLAG antibody (1:200 dilution, # F3165 Sigma Aldrich) and anti-HA (1:200 dilution, # 26183 Thermo Fisher Scientific) for 2 h at room temperature followed by incubation with Alexa Fluor 568 goat anti-rabbit IgG (1:500, #A11011 Thermo Fisher Scientific) and/or Alexa Fluor 488 anti-Mouse IgG (1:500, # Cat # A-11034, Thermo Fisher Scientific), finally with DAPI (Molecular Probes, #D1306). Cells were examined on a Leica TCS SP8 confocal microscopy (Leica, Wetzlar, Germany). All confocal images were obtained using the necessary filter sets for GFP, Alexafluor 568 and Alexafluor 488 using a ×63 (1.2 numerical aperture) water immersion objective. Acquisition of data was performed with the same intensity settings.

PrestoBlue assay

Cell viability was measured using PrestoBlue reagent (Thermo Fischer, Carlsbad, CA, USA) and according to the manufacturer’s protocol. Briefly, HEK293 cells were grown for 24 h in 12 well white plates with a transparent bottom and transiently transfected with expression plasmids. PrestoBlue reagent was added to determine viability at different time points, for 5 days. Plates were incubated for 30 min at 37°C and 5% CO₂. Fluorescence at 560 excitation and 590 nm emission was determined using Quantus™ Fluorometer (Promega, UK). PrestoBlue endpoints were determined in triplicate for each replicate. T-tests were performed to compare the differences between wild-type and mutants expressing cell lines, and a P-value < 0.05 was considered statistically significant.

Statistical methods

Patients’ demographical and clinical characteristics were reported as means and standard deviations or as frequencies and percentages for continuous and categorical variables, respectively. Group comparisons were carried out using t-test or chi-square test as appropriate. The number of osteochondromas/exostoses, total and by site, was analysed and compared between groups using Poisson regression model. An exact inference was used to account for possible very low number of osteochondromas/exostoses in some locations. A two-sided P-value <0.05 was considered for statistical significance. All statistical analyses were performed using SAS Release 9.4 (SAS Institute, Cary, NC, US).

Acknowledgements

The authors thank all families that participated to the study. They also thank Dr. Giuseppe Merla for providing EGFP-Myc and pCDNA.3 HA empty vectors.

Conflict of Interest statement. All authors declare that there is no conflict of interest concerning this work.

Funding

Ricerca Corrente 2018-21 granted by the Italian Ministry of Health.

Author contributions

M.Cas. and C.F. conceived and designed the research. G.N., V.G., L.M. and L.D. carried out the molecular genetic studies. C.F. and G.N. carried out the functional studies. M.Cop. performed statistical studies. All other authors provided samples. M.Cas. and C.F. interpreted the results and wrote the manuscript. All authors discussed the results and commented on the manuscript.