The hyperbranched polyglycerol-aldehyde polymer was purchased from Flintbox (University of British Columbia, Vancouver, BC, Canada). Isobaric tag for relative and absolute quantitation (iTRAQ) labels are from AB Sciex (Concord, ON, Canada). Porcine trypsin was purchased from Promega (#V511A; Madison, WI, United States). Ultrafiltration devices were purchased from Merck Millipore. All other reagents were purchased from Sigma-Aldrich (Saint-Quentin Fallavier, France).

Fibroblasts were obtained from the dermis of healthy donors (normal fibroblasts, NF) and of a patient suffering from the dermatosparactic type of the Ehlers-Danlos syndrome (dermatosparactic fibroblasts, DF) (Nusgens et al., 1992). Cells were cultured in DMEM (Lonza) supplemented with 10% fetal bovine serum (FBS, Lonza). For collagen production, cells were cultured in DMEM supplemented with 1% FBS and 50 μg/ml 2-phospho-l-ascorbic acid (Sigma Cat#49752). The conditioned medium was collected after 48h, centrifuged for 15 min at 4,000 rpm at 4°C. The supernatant was conserved at −80°C before collagen purification.

The experiment was conducted in triplicate using twelve 8-week-old adult mice (2 males and one female of each genotype (Wt, TS2^−/−, TS14^−/− and TS2^−/−/TS14^−/−)). The investigation on mice was reviewed and approved by Ethics Committee for Animal Use and Care of the University of Liege (Belgium) (protocol N°1,109). A square (1 cm²) of shaved skin was collected from the anteroventral part of the mice and incubated 1 h at 4°C under shaking in 3 ml of 50 mM HEPES sodium salt (pH 7.5), 2 mM CaCl₂ and 150 mM NaCl containing a cocktail of inhibitors of proteases and phosphatases (MS-SAFE, Sigma). Skin pieces were harvested and crushed with ceramic beads (MagNA Lyser Green Beads, Roche, n°03358,941,001) using the MagNA Lyser Instrument (Roche) in 90% (V/V) tissue protein extraction reagent (Thermo Scientific #78510), 10% (V/V) HEPES sodium salt (pH 7.5), 2 mM CaCl₂, 150 mM NaCl containing the MS-SAFE inhibitors cocktail. Crushed samples were centrifuged 5 min at 8,000 rpm at 4°C and the protein concentrations estimated (Bradford assay (Bradford, 1976)) in the collected supernatants and adjusted to 1.0 mg/ml. Samples of 500 µg of proteins for each condition were used for iTRAQ-TAILS labeling (Kleifeld et al., 2011; Bekhouche et al., 2016). Proteomic analyses were performed at the GIGA Proteomic platform on the ESI-Q Exactive (ThermoFisher) coupled with a 2D-RP/RP liquid chromatography (2D-RP/RP NanoAcquity UPLC, Waters, Milford, United States) for the peptide fractionation in three fractions.

Proteomic data analysis was previously described (Bekhouche et al., 2016). Briefly, in a first step, peptides were identified using Mascot (version 2.2.06; Matrix Science Inc., Boston, MA, United States) and allowing non-tryptic cleavages and two missed cleavages/peptide. Carbamidomethyl cystein was set as a fixed modification, and other modifications were set as variable: N-terminal acetyl, deamidation (NQ), Pyro-glu (N-term E), Pyro-Gln (N-term Q), Oxidation (M), iTRAQ (K), iTRAQ (Y) and iTRAQ (N-term). Peptide tolerance was set at 0.02 Da.

The tandem mass spectrometry (MS/MS) data were analyzed using the TransProteomicPipeline (TPP). The PeptideProphet and ProteinProphet software programs, embedded into TPP, were used to validate protein and peptide assignment. The nontryptic model was omitted in the PeptideProphet parameters. The error rate to validate proteins or peptides was respectively set at 2 and 5%. Then, Clipper software was used to determine the upper and lower cutoffs corresponding to 3-sigma calculated from the normal distribution of the log2(P:C ratio) from natural mature N termini. A Gaussian error function was used to calculate a p value that reflects the probability of a peptide to be a false-positive. A peptide with a P:C ratio above or below the 3-sigma cutoff has 99,8% chance to be dependent of the studied protease (auf dem Keller and M Overall, 2012). The cutoffs for each experiment are reported in Supplementary Table S4. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD022179 and 10.6019/PXD022179.

The biological processes have been investigated using the Panther database (PANTHER version 14: more genomes, a new PANTHER GO-slim and improvements in enrichment analysis tools (Mi et al., 2019)). Statistical overrepresentation tests were performed using the whole Mus musculus genome as reference dataset. The p-Values are determined using the Fisher’s Exact test corrected by the determination of the false discovery rate (fdr). The fdr was below 0.05 for all the reported biological process.

The Venn diagram were drawn using the BioVenn software (BioVenn—a web application for the comparison and visualization of biological lists using area-proportional Venn diagrams (Hulsen et al., 2008)).

Recombinant ADAMTS2, its inactive mutant, or ADAMTS14 were produced in HEK293 cells, purified and quantified as previously described (Colige et al., 2005). Briefly, recombinant proteases were purified, using Concanavalin A-Sepharose and Heparin-Sepharose columns, from 1 L of serum-free medium conditioned during 48 h. Proteases were recovered in 50 mM Tris, pH 7.5, 1 M NaCl, 2 mM CaCl₂.

Collagen from dermatosparactic or normal calf skin was extracted and purified according to a reported procedure (Nusgens and Lapiere, 1979; Colige et al., 1995). Collagen from fibroblasts cultures, was concentrated by adding ethanol to the conditioned medium (see above, 2.2) at a final concentration of 33% (V/V). After overnight incubation at 4°C, precipitated collagen was recovered by centrifugation (7,000 rpm, 40 min, 4°C). Pellets were solubilized in 0.1 M acetic acid (pH 2.9) under shaking at 4°C for 18 h. Non-solubilized contaminants were excluded by centrifugation (17,000 rpm, 40 min, 4°C) and the supernatant containing the collagen was neutralized by adding 1 M Tris base solution (at 1/10 of the final volume). Collagen cleavage was analyzed by incubation of 10 µg of type I collagen with recombinant ADAMTS2 or 14 (200 nM for 16 h at 37°C) in 50 mM Tris (pH 7.5), 0.5 M NaCl, 2 mM CaCl₂, in the presence or absence of 25 mM EDTA used as inhibitor of ADAMTS metalloproteinase activity. Collagens were used in native form or after thermal denaturation at 95°C for 10 min. Digestion products were separated by SDS-PAGE and stained with Coomassie Blue.

Cell lysates from human normal fibroblasts (NF) were prepared by sonication and incubated with recombinant human ADAMTS2 and/or 14 (200 nM for 16 h at 37°C) in 50 mM Tris (pH 7.5), 0.5 M NaCl, 2 mM CaCl₂, in the presence or absence of 25 mM EDTA. Actin and vimentin cleavages were analyzed by Western blot using polyclonal anti-alpha-actin-1 (A2066, Sigma) and polyclonal anti-vimentin (VI008–01, Quartett).

Amino acid sequence logos, corrected by the natural abundance of amino acids in the human proteome, were generated using the iceLogo software package (Colaert et al., 2009). Analyses were based on the cleavage sites determined by proteomics for all the potential extracellular substrates, without type I collagen cleavage sites or without all fibrillar collagens.

ADAMTS2 and ADAMTS14 are clearly implicated in several physio-pathological processes (Colige et al., 1999; Kesteloot et al., 2007; Dubail et al., 2010; Dupont et al., 2018; Wang et al., 2020). To better understand their in vivo roles in skin physiology and extracellular matrix homeostasis, N-terminomic experiments have been performed on skin of mice, either wild type (Wt), deficient in Adamts2 (TS2^−/−), deficient in Adamts14 (TS14^−/−) and deficient in both Adamts2 and Adamts14 (TS2^−/−TS14^−/−) (Figure 1A). For each experiment, proteins extracted from the four genotypes were labeled with specific iTRAQ labels for quantification by mass spectrometry (MS/MS). This 4-plex experiment was performed in triplicate. The proteomic analysis shows that between 68 and 79% of the proteins were labeled and have therefore been used for relative quantification. Among the labeled proteins between 13 and 26% are extracellular. By using Venn diagrams as analytical tool, a total of 1,117 proteins, including 840 with an iTRAQ labeled peptide, were identified in the three experiments, but only 41 and 38% of them were common to the three experiments, for unlabeled and labeled proteins, respectively (Figure 1B). As actual substrates might have been missed during the proteomic analyses, especially when their abundance is low, the data of these three experiments can be considered as complementary. In order to take this limitation into account but to generate also high confidence data, a labeled peptide was considered to identify a candidate substrate when it was observed in at least two experiments with a similar Protease/Control (P/C) fold change (either increased or decreased).

The degradomes (proteolytic events related to a proteolytic enzyme) are determined from the Protease/Control (P/C) ratios significantly different from the normal distribution of the natural N-termini. They include the proteins cleaved directly by the studied protease and also indirectly by the activation of other proteases or regulatory pathways. In this study, any N-terminally iTRAQ labeled peptide was considered to be part of the ADAMTS2 and/or ADAMTS14 degradomes when its P/C ratio was above or below a 3-sigma cut-off from the normal distribution of natural N-termini, including those obtained after the removal of the signal peptide. The degradome of ADAMTS2 has been determined from Wt vs TS2^−/− and from TS14^−/− vs TS2^−/−TS14^−/− ratios, and that of ADAMTS14 has been determined from Wt vs TS14^−/− and from TS2^−/− vs TS2^−/−TS14^−/− ratios, while that of ADAMTS2 and ADAMTS14 together has been determined from Wt vs TS2^−/−TS14^−/− ratios (Figure 2A). In this analysis, the whole degradome of ADAMTS2 and/or ADAMTS14 was found to be composed of 437 proteins, including 68 proteins secreted or anchored at the cell surface. Forty extracellular and cell surface proteins were related to both ADAMTS2 and ADAMTS14 activities, while five are specific of ADAMTS2 and 21 (13 + 8) of ADAMTS14. Of note, two proteins were detected only when comparing Wt and TS2^−/−TS14^−/−, the coagulation factor XIIIa and the mast cell protease 4 (Figure 2B). The degradomes of ADAMTS2 and ADAMTS14 are common at 59% (40/68) when considering specifically the extracellular and cell surface proteins, while only a 40% (148/369) overlap is found when considering all the other proteins, showing an enrichment for common extracellular substrates (Figure 2B). Investigation of the biological processes (pantherdb.org) related to these degradomes clearly illustrated a fundamental role for these two enzymes in collagen fibril organization, which was expected for ADAMTS2 but was more surprising for ADAMTS14 since collagen fibrils in TS14^−/− mice appear to be normal. This analysis also shed light on their implication in lipoprotein regulation and assembly, notably by the cleavage of the proapolipoprotein A-II and apolipoprotein A-I, and in immune response by the regulation of cytokines and chemokines secretion and complement activation, notably through the cleavage of complement proteins (C3C, C4-B) and of the macrophage migration inhibitory factor (Figure 2C, Table 1, Supplementary Figure S1).

To confirm our iTRAQ data showing multiple cleavages in the Col1 triple helical domain and in the C-propeptide of type I collagens, we performed in vitro assays using recombinant enzymes and collagen purified from the skin of dermatosparactic calf which is characterized by the persistence of the aminopropeptides in about 80% of type I collagen molecules (Figure 4A, lane 1). This particular substrate was chosen because it provides the opportunity to have an internal control consisting in the processing of the aminopropeptide of Col1A1 and Col1A2. This cleavage (conversion of pNa1 and pNa2 into a1 and a2 chains) was almost complete in the presence of recombinant ADAMTS2, but much reduced in the presence of ADAMTS14 in accordance with its lower aminoprocollagen peptidase activity (Figure 4A, lanes 2 and 4, respectively). When looking at lower molecular weight products (Figures 4C,E), the released pNa1 propeptide was observed with an apparent 30 kDa MW and was mainly found in the presence of recombinant active ADAMTS2 (lanes 2 and 6). The presence of other bands and of “trails”, covering the entire migration lanes and corresponding to cleavages of collagen in multiple sites, were also identified (lane 2), including in the presence of recombinant ADAMTS14 (lane 4). Since ADAMTS14 has only a reduced aminoprocollagen peptidase activity, it shows that the two types of activity are not related.

The same type I collagen preparation was also used as substrate after heat denaturation to verify whether disruption of the triple helical folding impacts the sensitivity to cleavage. Processing of the aminopropeptides by ADAMTS2 or ADAMTS14 was still observed, but marked differences were also visible as compared to gels obtained with native collagen (compare panels a, c, e to panels b, d, f of Figure 4), such as the presence of products at 97 and 110 kDa observed in the presence of ADAMTS2 and 14 (lanes 2, 4 and 6 on each panel), and at 105 kDa specifically in the presence of ADAMTS14 (lanes four on each panel). Additional discrete degradation products of lower MW were also identified as well as more pronounced degradation trails along the entire migration lanes (Figure 4, lanes 2, 4, 6; panels d and f).

The presence of cleavage sites within the C-propeptide of Col1A1 and Col1A2 was also evidenced by our iTRAQ analyses on in vitro assays. Since collagen purified from dermatosparactic skin lacks the C-propeptide, we used collagen produced by human dermatosparactic fibroblasts in culture which is mainly secreted as complete procollagen still retaining its two propeptides (pro-alpha one and pro-alpha 2). After incubation with active ADAMTS2 (Figure 5, lane 2), proa1 and proa2 are converted into pCa1 and pCa2, respectively, as a result of the aminoprocollagen peptidase activity of ADAMTS2. Fully processed alpha1 was also observed while similar amount of pNa1 was absent in the control (lane 1) suggesting that it is produced by cleavage of the N- and C-propeptides of the proa1 chain. In line with this hypothesis, the amount of alpha2 chain recovered after incubation with ADAMTS2 exceeded the amount of pNa2 in the control. Moreover, the relative intensity of pCa2 was lower after incubation with ADAMTS2 (lane 2) than expected from the intensity of proa2 in the control. These two observations clearly suggest some cleavage of the C-propeptide of proa2. Regarding ADAMTS14, only a low aminoprocollagen peptidase activity was observed, as expected, as illustrated by the presence of low amounts of pCa2 generated from proa2 (lane 3). However, accumulation of pNa2 was clearly observed indicating that ADAMTS14 can cleave the C-propeptide of Col1A2 more efficiently than its N-propeptide.

Several proteoglycans, non-fibrillar collagens and matricellular proteins are known to be involved in the regulation of collagen fibril formation and functions. Some of them were found to be potential substrates of ADAMTS2 and/or ADAMTS14 based on P/C ratios significantly >2 (Table 1). They are therefore mentioned, although cleavages were not confirmed by other methods, since their processing could be involved in the clinical features found in dermatosparactic EDS and other connective tissue disorders.

Two sites of cleavage in type XIV collagen were found for both ADAMTS2 and ADAMTS14: at positions 630 (AQQY.LEID) within the fifth fibronectin type-III domain and 1,636 (MARY.TAIL) within the third collagen like domain. Identical sites for both enzymes strongly suggest that the cleavages are real and that type XIV collagen is a true substrate. Potential cleavages were also found in COL4A1 (G₁₄₃₈.₁₄₃₉T), at a position corresponding to the release of the arresten cryptic bioactive fragment and in type VI collagens: (1) in the first VWFA domain of COL6A1 (F₁₈₆.₁₈₇S); (2) in the VWFA1 and VWFA2 domains of COL6A2 (F₁₁₆.₁₁₇S and F₁₄₁.₁₄₂A) and (3) in the VWFA6 domain of COL6A3 (F₁₀₅₁.₁₀₅₂A). Besides collagens, potential proteolytic cleavages were also found in PCPE-1 (a co-factor stimulating the processing of the C-propeptide of fibrillar collagens by BMP-1) and in three proteoglycans regulating matrix assembly: biglycan, lumican and osteoglycin. Finally, the degradomes of ADAMTS2 and ADAMTS14 in mouse skin point out several proteins involved in the immune system such as immunoglobulins, complement proteins (C3, C4-B, factor B, factor H), the macrophage inhibitory factor or annexins A8, A1 and A2 known to be involved in leukocyte recruitment and activation (Swisher et al., 2007; Poeter et al., 2014; Sugimoto et al., 2016).

This work was primarily focused on extracellular and cell surface degradome since ADAMTS2 and ADAMTS14 are secreted proteases. Unexpectedly however, numerous cleavage sites related to the presence/absence of ADAMTS2 and/or ADAMTS14 were found in intracellular proteins (Supplementary Table S1), including actins and vimentin (Figure 6A). On average, P/C ratios for actin were higher for ADAMTS14 than for ADAMTS2, as opposed to what was seen for the N-propeptides cleavages, and even higher when comparing Wt skin and TS2^−/−TS14^−/− skins. Of note also, cleavage sites in actin-2, also known as gamma-actin, are identical to the corresponding cleavage sites in actin-1 which suggests their specificity. In order to confirm these surprising observations, fibroblasts extracts were incubated in the presence of purified ADAMTS2 and/or ADAMTS14 with or without a saturating concentration of EDTA used as inhibitor (Figure 6B). In the absence of ADAMTS and EDTA, actin was detected as a single band of 42 kDa with an antibody raised against the C-terminal decapeptide of alpha-actin-1, demonstrating absence of cleavage by intracellular endogenous proteases, while a second product of about 40 kDa was observed after incubation with ADAMTS2, which would correspond to cleavages at positions V₁₉.₂₀K and/or A₂₁.₂₂G in alpha-actin-1 (Figure 6A). This product was also seen with ADAMTS14, in addition to a second one at about 34–35 kDa which was not observed with ADAMTS2. Remarkably, only discrete bands were seen, with no trace of smear which would indicate multiple cleavages at specific positions.

Regarding vimentin, several peptides with N-terminal iTRAQ labeling were detected, some with P/C ratio slightly but significantly higher in the presence of ADAMTS2 or ADAMTS14 (Figure 6A). For confirmatory purposes Western blot analysis was performed on fibroblast extracts incubated or not with purified ADAMTS2 or ADAMTS14. In the control samples, vimentin appeared as two major bands at 48 kDa and at 58 kDa the latter likely corresponding to the full-length protein. Additional minor products at 53, 50, 46 and 39 kDa were also visible (Romano et al., 2020) (Figure 6C). Upon incubation with ADAMTS2 or 14, the intensity of bands at 58 and 53 kDa were markedly reduced, accompanied by an increased intensity of the products at 48 and 39 kDa and the presence of additional bands at 37 and 32 kDa. These data clearly suggest the existence of several cleavage sites by ADAMTS2 and ADAMTS14 but were not investigated further.

The cleavage sites observed by N-terminomics were used to evaluate cleavage sites enrichment linked to the presence of ADAMTS2 and ADAMTS14. All the potential cleavage sites were used, in a first step, to establish the privileged consensus cleavage sites (Figure 7, upper panel). This revealed an overrepresentation of G and P that could be linked, at least in part, to the abundance of these amino acids in collagens. In addition, Y, F and A were often found at P1 and A and S at P1’, which is similar to previously described cleavage sites for aminoprocollagen peptidases (Bekhouche et al., 2016; Janssen et al., 2016) (Figure 7, upper panel). Considering all the potential extracellular substrates, excluding type I collagen (middle panel) or all fibrillar collagens (lower panel), it was shown that ADAMTS2 and ADAMTS14 display common preferential cleavage sites, enriched in small nonpolar, amphipatic or slightly hydrophobic amino acids (G, P, F, Y, A, L and M). Of note also was the presence of acidic amino acids at P2’, P3’ and P4’.