Distinct biological events generated by ECM by two homologous

Inna Solomonova, Eldar Zehoraia, Dalit Talmi-Franka,b, Sharon G. Wolfc, Alla Shainskayad, Alina Zhuravleva, Elena Kartvelishvilyc, Robert Vissee, Yishai Levind, Nir Kampff, Diego Adhemar Jaitinb, Eyal Davidb, Ido Amitb, Hideaki Nagasee, and Irit Sagia,1

aDepartment of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel; bDepartment of Immunology, Weizmann Institute of Science, Rehovot 76100, Israel; cDepartment of Chemical Research Support, Weizmann Institute of Science, Rehovot 76100, Israel; dNancy and Stephen Grand Israel National Center for Personalized Medicine, Weizmann Institute of Science, Rehovot 76100, Israel; eKennedy Institute of Rheumatology, Nuffield Department of Orthopaedics, Rheumatology, and Musculoskeletal Sciences, University of Oxford, Oxford, OX3 7FY, United Kingdom; and fDepartment of Materials and Interfaces, Weizmann Institute of Science, Rehovot 76100, Israel

Edited by Christopher M. Overall, University of British Columbia, Vancouver, Canada, and accepted by Editorial Board Member Rakesh K. Jain August 8, 2016 (received for review October 4, 2015) It is well established that the expression profiles of multiple and to unique ECM–cell crosstalk. We reveal that MMP-1 and MMP-13 possibly redundant -remodeling (e.g., collagenases) cause distinct ECM degradation, bringing about significantly distinct differ strongly in health, disease, and development. Although cellular phenotypes. Our findings show the complexity and selectivity enzymatic redundancy might be inferred from their close similarity in of -associated degradation mechanisms during tissue structure, their in vivo activity can lead to extremely diverse tissue- remodeling; these mechanisms could be used as a tool for future remodeling outcomes. We observed that proteolysis of -rich therapeutic interventions. natural (ECM), performed uniquely by individual homologous proteases, leads to distinct events that eventually affect Results overall ECM morphology, viscoelastic properties, and molecular Selective Degradation of ECM by Collagenases Determines Fibroblast composition. We revealed striking differences in the motility and Behavior. We set out to characterize the specific influences of the signaling patterns, morphology, and -expression profiles of cells – highly abundant collagenases on fibroblast ECM crosstalk. In this BIOCHEMISTRY interacting with natural collagen-rich ECM degraded by different study we used natural collagen fascicles (termed “ECM”)from collagenases. Thus, in contrast to previous notions, matrix-remodeling of 6-mo-old rats (22, 23). We examined the effect of this systems are not redundant and give rise to precise ECM–cell crosstalk. ECMonrat-1fibroblasts,whichareknowntobesensitivetoECM Because ECM proteolysis is an abundant biochemical process that is composition and rigidity. Moreover, these cells, much like epithe- critical for tissue homoeostasis, these results improve our fundamen- lial cells, can exhibit different modes of migration toward different tal understanding its complexity and its impact on cell behavior. substrates and engage in cell–cell interactions (24, 25). Because fibroblasts inherently express ECM and remodeling en- ECM | MMP | proteolysis zymes, we conducted our experiments at early stages of interaction

he function and integrity of the extracellular matrix (ECM) is Significance Tvital for cell behavior as well as for whole-tissue homeostasis – (1 4). The ECM undergoes constant remodeling during health Extracellular matrix (ECM) proteolysis is an abundant bio- and disease states. Its components are regularly being deposited, chemical process. Here we describe the multilayered biological degraded, or otherwise modified. The highly stable fibrillar col- complexity generated by structurally homologous collagenases lagen type I (Col I) is abundant in many organ-derived ECMs (matrix -1 and -13) in and connective tissues (5, 6); it serves as a tissue scaffold, deter- collagen-rich, native ECM, that may prove central to tissue ho- mining ECM mechanical properties and anchoring other ECM meostasis and pathology. The events induced by these two col- proteins necessary for cell function (7). These processes are or- lagenases generate microenvironments characterized by distinct chestrated by multiple remodeling , among which the ma- chemical, biomechanical, and morphological ECM properties that trix metalloproteinase (MMP) family plays an important role. Only lead to different cellular behaviors. Our findings improve the a few members of this proteinase family, the collagenases, are able fundamental understanding of selective ECM degradation by to degrade the resistant fibrillar , i.e., Col I, as well as other homologous collagenases and its impact on cell behavior. ECM (8, 9). The collagenases have conserved amino acids in their -containing catalytic domain (8, 10) and display Author contributions: I. Solomonov, E.Z., I.A., H.N., and I. Sagi designed research; I. Solomonov, high structural similarities, as reflected by their functional domain E.Z., S.G.W., A.S., A.Z., E.K., Y.L., N.K., and D.A.J. performed research; R.V. and H.N. contrib- organization. Nevertheless, the complex effects exerted by different uted new reagents/analytic tools; I. Solomonov, E.Z., D.T.-F., A.S., Y.L., D.A.J., E.D., and MMPs on ECM and cells in vivo remain poorly understood. I. Sagi analyzed data; and I. Solomonov, E.Z., D.T.-F., S.G.W., and I. Sagi wrote the paper. The enzymatic activity of MMPs, and specifically of collage- The authors declare no conflict of interest. nases, in vivo is tightly regulated (11), with enzymatic dysregu- This article is a PNAS Direct Submission. C.M.O. is a Guest Editor invited by the Editorial lation causing irreversible damage associated with a variety of Board. diseases. Abnormally elevated levels of MMP-1 or of both MMP-1 Data deposition: MS data and /peptide identifications from in-gel analysis have been uploaded to the ProteomeXchange Consortium via the PRIDE repository, www.ebi. and MMP-13 have been associated with different types of cancers ac.uk/pride/archive/login (project accession no. PXD003796; username: reviewer59836@ andwithinflammatorydiseases(12–21). ebi.ac.uk; password: sTHj8wE8). Mass spectrometry data and protein/peptide identifica- Here, we explored the degradation patterns of natural collagen- tions from in-solution digestion have been uploaded to the ProteomeXchange Consor- rich ECM by two homologous MMPs and tested the response of tium via the PRIDE repository, www.ebi.ac.uk/pride/archive/login (project accession no. PXD003553; username: [email protected]; password: cJGfrm0Z). The RNA- cells to intact vs. selectively degraded ECM. We collectively pro- sequencing dataset has been deposited in the Omnibus database (acces- filed the unique remodeling events caused by two secreted colla- sion no. GSE79749). genases, MMP-1 and MMP-13, by using biochemical, biophysical, 1To whom correspondence should be addressed. Email: [email protected]. and proteomics tools. We show that these proteases drive morpho- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. logical, biochemical, and viscoelastic changes in the ECM leading 1073/pnas.1519676113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1519676113 PNAS Early Edition | 1of6 Downloaded by guest on September 23, 2021 (up to 4 h); in this time frame, no collagen deposition or MMP-1 adopted round, symmetrical morphologies. To validate these and MMP-13 secretion was detected (Fig. S1 A and B). morphological changes, we further quantified the aspect ratios of Cell motility was characterized using real-time optical imaging. cells that adhered to differentially degraded ECM using our Although fibroblasts demonstrated a clustered, directional motil- SEM images at 4 h after seeding (Fig. 1 D–F). This analysis ity toward native ECM (control) with a measured velocity of 40 ± yielded a trend in cell aspect ratio similar to that obtained with 14 μm/h, they showed individual modes of motility with a velocity live-cell imaging at the measured time point (Fig. 1H). Overall, of 37 ± 5 μm/h when moving toward MMP-1–degraded ECM. In these analyses provide cell-shape statistics indicating that rat-1 the presence of MMP-13–degraded ECM, the cells exhibited re- cells adopt different morphologies upon interaction with differ- duced or arrested motility with a measured velocity of 2 ± 2 μm/h entially degraded Col I-rich ECM. (Fig. 1 A–C, Fig. S1C,andMovies S1–S3). The differential phenotypes acquired by fibroblasts interacting To analyze the overall shape of the cells, we quantified the cell with degraded ECM suggested a typical and unique cellular re- aspect ratio (the length divided by the width) as a function of sponse driven by signal modulations transmitted by the ECM time using our live-cell imaging data (Fig. 1G). This analysis (26, 27). Because cells regulate migration, proliferation, and allowed us to compare precisely the shape of cells migrating on adhesion mainly through the activation of the ERK1/2 cascade differentially remodeled collagen-rich ECM. Cells imaged on (28), we examined the effect of ECM remodeling on this cas- control ECM showed aspect ratios ranging from 1.36 to 1.42, cade. Fibroblasts adhering to native ECM demonstrate a sus- tained mode of ERK1/2 activation; in contrast, a transient exhibiting flat and stable morphologies. Cells imaged on MMP- activation, peaking at 30 or 60 min, was detected in cells ad- 1–degraded ECM showed increasing cell aspect ratios over time hering to ECM degraded by either MMP-1 or MMP-13. More- ranging from 1.49 to 4.5, exhibiting an elongated cell shape. over, cells interacting with degraded ECM exhibited higher levels Remarkably, cells imaged on MMP-13–degraded ECM showed – of total protein (represented by ERK1/2), indicating improved cell aspect ratios of 1.34 1.51, exhibiting round and stable (Fig. S2 A–C). These results are in good agreement morphologies over time. In agreement with the live-cell images with published work (29, 30) demonstrating that adhesion to the A–C reported in Fig. 1 , these results demonstrate that cells mi- ECM permits efficient growth factor-mediated activation of – grating on MMP-1 degraded ECM adopted an elongated shape ERK1/2 in fibroblasts. We further quantified the Western blot – with time, whereas cells migrating on MMP-13 degraded ECM experiments and calculated the overall adherence rate by ap- plying a linear regression algorithm (Fig. S2D). The results show an increase in the number of cells adhering to degraded ECM in comparison with natural ECM. Remarkably, an increase of ∼34% in cell adherence was observed in cells that adhered to MMP-1– versus MMP-13–degraded ECM. In correlation with the observed changes in cell shape in Fig. 1, these results suggest that degradation of the ECM by MMP-1 results in improved adherence manifested through cell elongation. By profiling fibroblast transcriptional responses, we found 3,163 that were differentially expressed in cells interacting with ECM remodeled by MMP-1 or MMP-13 compared with control. The transcriptional responses of fibroblasts interacting with degraded ECM clustered into several groups of genes in- − − volved in adhesion (P < 10 9), proliferation (P < 10 5), and − morphogenesis (P < 10 5)(Fig. S2E and Datasets S1 and S2). Induction was observed in protocadherins (Pcdhga3, Pcdhga5, and Pcdhga8), which are involved in cell adhesion and have been shown to interact with a wide range of binding partners, resulting in cytoskeleton changes (31, 32) as well as transcription factor 3 (Tcf3), which is involved in and . The elevated expression of the Pcdhga3 and Pcdhga5 genes can be related to direct and indirect cellular adhesion properties corroborated by a large number of other interacting proteins, including phosphatases, kinases, and adhesion molecules. In addition, we found induction of genes involved in pro- liferation, such as cyclin-dependent kinases (Cdk2, Cdk9, and Cdk14). Induction also was found in genes that are involved in such as discoidin domain tyrosine kinase 1 (Ddr1), which is activated by various types of collagen and is in- volved in cell growth and communication (33), and tropomyosin 1 (Tpm1) and cadherins (Cdh2, Cdh3). Our gene-expression data were validated using quantitative PCR (qPCR) analysis of repre- sentative genes (Fig. S2 F–H and Dataset S3). Overall, these Fig. 1. Cell–ECM interactions. (A–C) Real-time in vivo imaging demonstrates analyses indicate that the responses of rat-1 cells differ upon their the morphological features of the cells interacting with natural ECM (A)or interactions with differentially degraded Col I-rich matrices. ECM degraded by MMP-1 (B) or MMP-13 (C) at different time points. Within the same treatment, colors identify representative cells at different time MMP-1 and MMP-13 Produce Distinct Microscale Morphologies and points. (Scale bars: 15 μm.) (D–F) SEM images of fibroblasts adhered to Viscoelastic Alterations of ECM. natural ECM (D) or to ECM degraded by MMP-1 (E) or MMP-13 (F) 4 h after Because collagenases are highly cell seeding. (Scale bars: 20 μm.) (G) Kinetics of cell aspect ratio calculated potent proteases that are capable of irreversibly cleaving and from real-time in vivo imaging; n = 50 cells per treatment. (H) Cell aspect reshaping the ECM landscape, we next focused on identifying the ratios quantified from SEM images; n = 200 cells per treatment; *P < 0.05; morphological changes exerted upon the ECM as a result of **P < 0.01, t test. specific collagenase activity. SEM images demonstrate that natural

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1519676113 Solomonov et al. Downloaded by guest on September 23, 2021 ECM consists mainly of collagen fibrils aligned along the fiber axis. Upon degradation by MMP-1 or MMP-13, the spatial orga- nization of the collagen fibrils is changed: the fibril alignment is disrupted, producing specific and robust digestion patterns. MMP-1 produces widely distributed broken and bent fibrils exhibiting multiple orientations, whereas MMP-13 causes the splitting of the native collagen fibrils into thinner ones, as op- posed to the straight and aligned intact fibrils (Fig. 2). The unique ECM microscale morphologies produced by MMP-1 and MMP-13 may lead to changes in ECM biomechanical properties on the macroscale level. By applying rheology measurements, we determined the fre- quency dependence of the elastic (G′) and viscous (G′′) moduli, measuring the stress response of the ECM with frequencies varying from 1 to 100 Hz (Fig. S3A). All samples exhibited gel- like behavior: G′ was higher than G′′, and both parameters in- Fig. 3. Structural features of fibrillar collagen degradation by MMP-1 and creased slightly with frequency. A comparative analysis of G′ ∼ MMP-13. We accumulated more than 2,500 TEM images from more than 300 values points to intact ECM as being stiffer ( 37 kPa) than de- samples per treatment and analyzed the highly reproducible images. Shown graded ECMs (G′ of ∼1.5 kPa for MMP-1 and ∼14 kPa for are TEM images of cryo-preserved and negatively stained supernatants of MMP-13). In addition, the G′′ values revealed that intact ECM control (A and D), MMP-1–treated (B and E), and MMP-13–treated (C and F) has the highest viscosity (∼1.75 kPa), whereas ECM altered by samples. All fibrils show characteristic of the Col I banding pattern of 67 nm. MMP-1 and MMP-13 is less viscous (G′′ ∼0.6 kPa). Taken to- Arrows indicate Col I fibril polarity from the C to the N termini. (Scale bars: gether, these findings demonstrate that selective degradation 100 nm in A–C; 200 nm in D–F.) results in distinct ECM microscale morphologies and viscoelastic properties that may lead to differential regulation of cell be- Collagenolysis Is Driven by Distinct Structural Mechanisms. We used havior (34, 35). transmission electron microscopy (TEM), cryoTEM, and negative

staining to visualize the degradation products present in super- BIOCHEMISTRY natants after MMP degradation. TEM images of native ECM supernatants revealed extremely low quantities of individual fibrils. The images display empty background areas around highly ordered fibrils, confirming the near absence of degradation, as expected because Col I is very stable and abundantly cross-linked (Fig. 3 A and D and Fig. S3B) (5). In contrast, samples treated with MMP-1 or MMP-13 displayed highly abundant, ruffled fibrils surrounded by unique degradation products, strongly suggesting that these fi- brils are formed during MMP degradation (Fig. 3 B, C, E,andF and Fig. S3B). The distinct banding of 67 nm observed in Col I fibrils has been used to correlate protein sequence location to the bands (36), and these assignments were correlated to the bands observed by cryoTEM (Fig. S3C). This correlation led to the identification of the N- and C-telopeptide regions and the site of MMP cleavage (Fig. 3). From these assignments, we observed structural anisotropicity of Col I cleavage in the cases of both proteases. The images reveal the “peeling” of degraded fragments, fringing off from the C-to-N terminus direction of the Col I fibrils (Fig. 4 B, C, E,andF). This directionality in MMP degradation may be dictated by the natural polarity of Col I fibrils, in which C and N termini of collagen molecules directed toward different poles of the fibrils. The anisotropicity of collagen degradation is also confirmed by comparing fibril termini, which display distinct morphologies (Fig. S4). The N-terminal ends of the degraded fi- brils are more compact than their C-terminal counterparts, sug- gesting that fibril degradation generally progresses from the C to the N terminus of the fibril. Most importantly, the cryoTEM images of degraded fibrils show that cross-linked C-telopeptides are not degraded during MMP-1 processing, as indicated by their presence in the background of proteolyzed morphologies extending out from the fibrils (Fig. 3B, green arrows). In comparison, cross- linked C-telopeptides are not present in fibrils degraded by Fig. 2. MMP-1 and MMP-13 produce distinct microscale ECM morphologies. MMP-13 (Fig. 3C), indicating the existence of a highly selective Shown are SEM images of natural ECM (A and B), ECM degraded by MMP-1 degradation mechanism on the Col I fibril. TEM images of nega- (C and D), and ECM degraded by MMP-13 (E and F). The ECM degradation tively stained samples reveal that both proteases produce hetero- was done using 500 nM MMP-1 or MMP-13 at 30 °C for 24 h. The detailed geneous populations of degraded products, with triangular analysis of more than 400 SEM images shows that normal collagen fibrils microfibril morphologies prevalent in MMP-1–treated ECM and lying underneath the degraded fibers are observed in all treatments. At – chosen conditions the ratio of degraded to undegraded fibrils was estimated rod-like fragments prevalent in MMP-13 treated ECM (Fig. 3 to be 0.13:1. Proteolysis could be observed up to ∼80 microns deep within E and F and Figs. S5 and S6). The normalized distribution of the collagen fascicles. (Scale bars: 1 μm.) fragment lengths for MMP-1 and MMP-13 showed that the most

Solomonov et al. PNAS Early Edition | 3of6 Downloaded by guest on September 23, 2021 or MMP-13. This analysis revealed distinctly different degradation patterns for MMP-1 and MMP-13, whereas the control samples, as expected, contained a minimal amount of degradation products (Fig. S8). Fig. 4 A and B and Dataset S4 list the total relative abundances of matrisome proteins released from treated ECM, in which Col I is the most abundantly degraded protein. Because trypsin digestion is highly specific (40), we correlated semitryptic peptides detected by MS with the proteolytic activity of MMPs and determined Col I (and several noncollagenous proteins) unique and common cleavage sites for MMP-1 and MMP-13 (Fig. 4C and Fig. S9). Principal component analysis (PCA) of Col I tryptic peptides resulted in three distinctly isolated, closely clustered populations (Fig. S10), indicating that each collagenase degrades Col I fibrils using a distinct mechanism. The most striking differences between the two supernatant profiles are the content and relative abundance of noncollagenous components such as proteoglycans (PG), glycoproteins (GP), ECM-affiliated proteins, and other function-related ECM regu- lators. Fig. 4 shows that selective ECM degradation impacts not only the ECM’s morphology and viscoelastic properties but also its molecular composition, thus adding complexity to the ob- served effect. For example, our data confirmed the previously reported MMP-13 specificity for the GP tenascin-C (41), and the PG (42) and showed much broader specificity of MMP- 13 to matrisome proteins. Other differentially released non- collagenous components include PGs such as decorin and Fig. 4. Mass spectrometry-based proteomics data of ECM degraded by MMP-1 fibromodulin and GPs such as fibulin 5, all known to be critical or MMP-13 determined from silver-stained SDS/PAGE and analyzed by nano- for cell signaling, spreading, and motility (43–45). Taken to- LC-ESI-MS/MS. (A) Relative abundances of matrisome proteins released during gether, these data demonstrate that both collagenases effectively ECM degradation by MMP-1 or MMP-13. (B) Zoom-in of relative abundances of degrade native collagen-rich ECM in a highly selective mode core- and matrisome-associated proteins. The color-coded proteins are (1) Col I; which consequently releases bioactive matrisome molecules and (2) collagen type VI; (3) collagen type XV; (4) decorin; (5) fibromodulin; (6) isoform 2 of aggrecan core protein; (7) proteoglycan 4; (8) fibulin 5; (9) at the same time exposes new matrix-associated epitopes and tenascin-C; (10) transforming growth factor-β–induced protein Ig-h3 precursor; protein complexes. (11) long isoform of hyaluronan and proteoglycan link protein; (12) lactad- herin; (13) myocilin; (14) procollagen C- enhancer; (15) annexin Discussion A1; (16) short isoform of annexin A2; (17) annexin A5; (18) peptidase The constant remodeling of the ECM environment in healthy and inhibitor clade F, member 1; (19) inter–α-trypsin inhibitor heavy chain H3. diseased states continuously subjects cells to a variety of stimuli. (C) Col I cleavage sites identified under proteolytic degradation of natural ECM Nevertheless, we lack an understanding of the cellular implications by MMP-1 or MMP-13. Green indicates cleavage sites common to both MMPs. of these stimuli. This study shows that cell–ECM crosstalk is Red indicates cleavage sites detected under degradation of Col I by MMP-1 or governed by specific and selective activity of remodeling enzymes MMP-13. Single and double asterisks indicate specific native triple-helical Col I cleavage sites. Spectral counts represent the absolute number identified for that exert intricate effects on the ECM, thus altering its mor- each Col I semitryptic peptide and demonstrate the differential efficiency of each phology and viscoelastic and biochemical properties. Specifically, MMP at any detected cleavage site. (Data reproducible in five experiments.) our results provide mechanistic insights into how homologous collagenases selectively degrade native collagen-rich ECM. Cells seeded onto collagen matrix may secrete their own ECM abundant lengths were 223 ± 15 nm and 82 ± 13 nm for MMP-1 proteins and remodeling enzymes over a longer time (Fig. S1). and 207 ± 15 nm and 83 ± 15 nm for MMP-13 (Fig. S7). These Therefore, we conducted our experiments at the early stages of distributions reflect the signature cleavage positions at 3/4 and cell–ECM interactions. In addition, the control and degraded 1/4 of the monomeric length of collagen, as well as other non- collagen-rich ECM was washed extensively to minimize the classical cleavage sites marked by the broad Gaussian peak. The presence of degradation products and exogenously added individual rod-like fragments resulting from MMP-13 degrada- MMPs. We found that the observed changes in cell shape, mo- tion observed in the TEM images (Fig. 3F and Figs. S5 and S6) tility, and signaling are affected most by the degraded matrix had a diameter of ∼4 nm, corresponding to the proposed di- morphology, biochemistry, and mechanical properties and less by ameters of individual microfibrils (five- bundles) from the degradation products. However, we could not completely TEM and diffraction studies (37, 38). We interpret the triangular exclude the contribution of cell-derived matrix proteins and morphologies present in MMP-1–degraded samples as being formed proteases to the effects observed on the ECM or to the activa- by bundles of microfibrils that are connected at the C-telopeptide tion of other cell-produced proteases. Nevertheless, our results terminus. Our observations strongly suggest that one microfibril is clearly demonstrate that homologous collagenases promote dif- processed as a single cleavage incident. This conclusion is sup- ferential degradation patterns on natural ECM. Although MMP- ported by degradation kinetics studies showing a processive burst 1 and MMP-13 are structurally homologous and degrade Col I of 15 ± 4 cleavage events occurring within one cut (39), corre- anisotropically from the C to the N terminus, we show that they sponding to five triple-helical molecules in a single microfibril have a different specificity and selectivity to natural collagen (5 × 3 = 15 cleavage events). fascicles: MMP-13 exhibits broader specificity than MMP-1 and produces a greater variety of degradation products. Differential Proteomic Profiles Are Generated During ECM Degradation In addition, we found that ECM degradation by either MMP-1 by Collagenases. Mass spectrometry analysis [nano-LC-electrospray or MMP-13 reveals distinct collagen-cleavage mechanisms, pro- ionization (ESI)-MS/MS] was used to examine the proteomic ducing characteristic degradation fragments as shown by both profiles of supernatants of ECM degraded either by MMP-1 TEM images and MS analysis (Figs. 3 and 4). The distribution

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1519676113 Solomonov et al. Downloaded by guest on September 23, 2021 analysis of Col I fragment lengths showed that each alterations (53, 54). Overall, the remodeling events of collagen- produced intrapopulation heterogeneity, confirming our MS rich ECM by either MMP-1 or MMP-13 contributed to fibroblast– data and indicating the existence of several cleavage sites on Col matrix communication, suggesting that such stimulation may be I. These significantly different cleavage patterns suggest that involved in cell migration in normal and pathological processes. MMP-1 and MMP-13 access different epitopes of the assembled For example, MMP-13 has been shown to enhance the remodeling or partially digested collagen fibrils. of 3D collagen matrix, to affect cell morphology, and to promote Our results are supported by recently published data indicating proliferation of dermal fibroblasts in both in vitro and in vivo MMP-13’s ability to cleave a range of dissolved collagen type II models of wound healing (44, 55). peptides (46). In addition, regions of helical instability and triple- Finally, our results confirm that ECM degradation by both helix local dissociation recently identified in native hydrated col- collagenases is accompanied by a significant loss of mechanical lagen fibrils (47) may enable MMPs to access other exposed sites. rigidity on the macroscale level, resulting from the degradation of These results are further supported by PCA demonstrating the dis- the outer layers of each fascicle. Both MMP-1 and MMP-13 bring tinct tryptic fragments of native MMP-1– and MMP-13–degraded about ECM softening, with MMP-1 having a stronger softening Col I. Furthermore, TEM and nano-LC-ESI-MS/MS analysis effect than MMP-13. Softening of polyacrylamide substrates and provided proof that the C-telopeptides remain intact in MMP- other artificial matrices is known to reduce the spread of fibro- 1–degraded Col I but are cleaved by MMP-13. Previous in vitro blasts, decrease cell motility, and induce cell rounding (56, 57). and in silico studies suggested that the cleavage of C-telopeptides is Remarkably, we observed that softening of native collagen-rich a critical initial step in collagenolysis, enabling the MMPs to access ECM by individual collagenases resulted in differential cell be- the cleavage site (48). Remarkably, our data show that MMP-1 havior. This finding implies that the biomechanical properties of collagenolysis can occur efficiently without prior C-telopeptide native ECM constitute only one of the parameters impacting cell cleavage. It is noteworthy that our experimental set-up using native motility, shape, and spreading. Our results demonstrate the im- collagen fascicles, a fibroblast cell line, and individual proteases portance of a combinatorial effect that includes the integration of may not completely mimic the complex natural action of collage- all the events driven by MMP degradation, namely ECM mor- nases in vivo during health and disease states. For example, excess phology, molecular composition, and biomechanics, that govern activity of multiple interstitial collagenases could be correlated with cell behavior. These observations suggest that the concerted action a chronically nonhealing wound (44). Nevertheless, our results of several collagenases, including MMP-8 and MMP-14, among highlight another, previously unrecognized level of complexity, other proteases detected in various pathologies, will manifest

showing that each exhibits a unique proteolytic pattern the severity of the disease by interfering with the tissue and ECM BIOCHEMISTRY within the native tissue. Although the degradation products may integrity in a combined or synergistic manner. Such concerted differ among different tissues in vivo, our mass spectrometry analysis proteolytic action also can contribute to the activation of other conducted on the model we used demonstrates the general concept proteases and their tissue inhibitors in vivo. that homologous collagenases do not exhibit redundant enzyme In conclusion, our results demonstrate the distinct roles of activity at multiple biochemical levels. We show that the degrada- ECM-remodeling enzymes in generating specific ECM proper- tion of collagenous and noncollagenous proteins, such as decorin, ties that affect cells and determine their fate. Our integrated fibromodulin, and aggrecan, which are required for the proper experimental approach identified the specific changes in mor- organization of the ECM, may also change the ECM’s spatial phology, biomechanics, and molecular composition that occur in organization and its morphology. In addition, under physiological the native ECM during degradation reactions. Our study reveals conditions the release or degradation of these matrisome molecules the exquisite specificity and selectivity of the enzymatic activity was reported to contribute to cell signaling through their interac- of two structurally homologous collagenases in the context of tions with cell-surface molecules (45, 49, 50). Finally, selective their natural microenvironment. Collectively, our results high- degradationofPGsandGPsisknowntocontributetocelladhesion light the importance of selective ECM remodeling. and activation of cell proliferation (43, 44, 51, 52). We demonstrate that ECM degradation by MMPs improves Materials and Methods fibroblasts’ ability to adhere to ECM, suggesting that ECM deg- Details of reagents, antibodies, detailed methods for ECM preparation, ex- radation leads to the exposure of adhesion sites and/or signaling pression, and purification, activation and enzyme activity assays for MMP-1 and molecules bound to the ECM scaffold. Indeed, some of the genes MMP-13, , RNA isolation, qPCR, Western blot analysis, imaging that were induced in the cells following interaction with the de- studies, proteomic analysis by mass spectrometry, rheology, and statistical graded ECM were annotated as adhesion molecules, such as analysis used in this study can be found in SI Materials and Methods. protocadherins (Pcdhga3 and Pcdhga5), belonging to the cadherin ACKNOWLEDGMENTS. We thank Joseph P. R. O. Orgel for providing the rat family, and morphogenesis/adhesion molecules (Ddr1 and Tpm1). tails and sample preparation protocols; Prof. R. Seger for his input in the These genes are known to interact with a wide range of binding design and analysis of ERK1/2 activation experiments; A. Aloshin for pro- partners regulating cell adhesion (32). The induced transcriptional viding technical help; and T. Mehlman, D. Merhav, and A. Gabashvili for responses of genes related to proliferation (Cdk9 and Cdk14) in assisting in the mass spectrometry experiments. The EM studies were sup- ported in part by the Irving and Cherna Moskowitz Center for Nano and fibroblasts interacting with MMP-degraded ECM also support the Bio-Nano Imaging at the Weizmann Institute of Science; Deutsch–Israelische observed induction of the ERK1/2-signaling cascade. We also Projektkooperation (German–Israel Foundation) Grant AD 364/2-1 FR 2190/6-1; examined the other main MAPKs signaling pathways (JNK, p38, Thompson Grant 16137; European Union Seventh Framework Programme and AKT) and found no significant activation under the applied Save Me Project Grant 263307; Wellcome Trust Grant 075473; a Weizmann UK-Making Connections Collaborative Grant; Israeli Science Foundation Grant experimental set-up. Furthermore, we show an increase in colla- 1226/13; the Geraldo Rozenkranz fund; and by the Kimmelman Center for gen receptor Ddr1, which regulates cellular functions such as cell Structural Biology at the Weizmann Institute of Science. I. Sagi is the incumbent spreading and polarity, in response to ECM–microenvironment of the Pontecorvo Professorial Chair.

1. Hynes RO (2009) The extracellular matrix: Not just pretty fibrils. Science 326(5957): 5. Kadler KE, Holmes DF, Trotter JA, Chapman JA (1996) Collagen fibril formation. 1216–1219. Biochem J 316(Pt 1):1–11. 2. Frantz C, Stewart KM, Weaver VM (2010) The extracellular matrix at a glance. J Cell Sci 6. Shoulders MD, Raines RT (2009) Collagen structure and stability. Annu Rev Biochem 123(Pt 24):4195–4200. 78:929–958. 3. Bonnans C, Chou J, Werb Z (2014) Remodelling the extracellular matrix in develop- 7. Kadler KE, Baldock C, Bella J, Boot-Handford RP (2007) Collagens at a glance. J Cell Sci ment and disease. Nat Rev Mol Cell Biol 15(12):786–801. 120(Pt 12):1955–1958. 4. Cowin SC (2004) Tissue growth and remodeling. Annu Rev Biomed Eng 6: 8. Visse R, Nagase H (2003) Matrix and tissue inhibitors of metal- 77–107. loproteinases: Structure, function, and biochemistry. Circ Res 92(8):827–839.

Solomonov et al. PNAS Early Edition | 5of6 Downloaded by guest on September 23, 2021 9. Page-McCaw A, Ewald AJ, Werb Z (2007) Matrix metalloproteinases and the regula- 40. Olsen JV, Ong SE, Mann M (2004) Trypsin cleaves exclusively C-terminal to tion of tissue remodelling. Nat Rev Mol Cell Biol 8(3):221–233. and residues. Mol Cell Proteomics 3(6):608–614. 10. Lovejoy B, et al. (1999) Crystal structures of MMP-1 and -13 reveal the structural basis 41. Knäuper V, et al. (1997) The role of the C-terminal domain of human collagenase-3 for selectivity of collagenase inhibitors. Nat Struct Biol 6(3):217–221. (MMP-13) in the activation of procollagenase-3, substrate specificity, and tissue in- 11. Gaffney J, Solomonov I, Zehorai E, Sagi I (2015) Multilevel regulation of matrix hibitor of metalloproteinase interaction. J Biol Chem 272(12):7608–7616. metalloproteinases in tissue homeostasis indicates their molecular specificity in vivo. 42. Fosang AJ, Last K, Knäuper V, Murphy G, Neame PJ (1996) Degradation of Matrix Biol 44-46:191–199. aggrecan by collagenase-3 (MMP-13). FEBS Lett 380(1-2):17–20. 12. Ito T, et al. (1999) Expression of the MMP-1 in human pancreatic carcinoma: Re- 43. Ferdous Z, et al. (2010) A role for decorin in controlling proliferation, adhesion, and lationship with prognostic factor. Mod Pathol 12(7):669–674. migration of murine embryonic fibroblasts. J Biomed Mater Res A 93(2):419–428. 13. Johansson N, et al. (1997) Expression of collagenase-3 (matrix metalloproteinase-13) 44. Preis M, et al. (2006) Effects of fibulin-5 on attachment, adhesion, and proliferation of in squamous cell carcinomas of the head and neck. Am J Pathol 151(2):499–508. primary human endothelial cells. Biochem Biophys Res Commun 348(3):1024–1033. 14. Konttinen YT, et al. (1999) Analysis of 16 different matrix metalloproteinases (MMP-1 45. Neill T, Schaefer L, Iozzo RV (2015) Decoding the matrix: Instructive roles of pro- to MMP-20) in the synovial membrane: Different profiles in trauma and rheumatoid teoglycan receptors. Biochemistry 54(30):4583–4598. . Ann Rheum Dis 58(11):691–697. 46. Howes JM, et al. (2014) The recognition of collagen and triple-helical toolkit peptides 15. Mort JS, Billington CJ (2001) Articular cartilage and changes in arthritis: matrix deg- by MMP-13: Sequence specificity for binding and cleavage. J Biol Chem 289(35): radation. Arthritis Res 3(6):337–341. 24091–24101. 16. Murray GI, Duncan ME, O’Neil P, Melvin WT, Fothergill JE (1996) Matrix metal- 47. Orgel JP, Persikov AV, Antipova O (2014) Variation in the helical structure of native loproteinase-1 is associated with poor prognosis in colorectal cancer. Nat Med 2(4): collagen. PLoS One 9(2):e89519. 461–462. 48. Perumal S, Antipova O, Orgel JP (2008) Collagen fibril architecture, domain organi- 17. Murray GI, Duncan ME, Arbuckle E, Melvin WT, Fothergill JE (1998) Matrix metal- zation, and triple-helical conformation govern its proteolysis. Proc Natl Acad Sci USA loproteinases and their inhibitors in gastric cancer. Gut 43(6):791–797. 105(8):2824–2829. 18. Nikkola J, et al. (2002) High expression levels of collagenase-1 and stromelysin-1 49. Hildebrand A, et al. (1994) Interaction of the small interstitial proteoglycans biglycan, correlate with shorter disease-free survival in human metastatic melanoma. Int J decorin and fibromodulin with transforming growth factor beta. Biochem J 302(Pt 2): Cancer 97(4):432–438. 527–534. 19. Poola I, et al. (2005) Identification of MMP-1 as a putative breast cancer predictive 50. Gubbiotti MA, Iozzo RV (2015) Proteoglycans regulate autophagy via outside-in sig- marker by global gene expression analysis. Nat Med 11(5):481–483. naling: An emerging new concept. Matrix Biol 48:6–13. 20. Sauter W, et al.; LUCY-Consortium (2008) Matrix metalloproteinase 1 (MMP1) is as- 51. Brandan E, Cabello-Verrugio C, Vial C (2008) Novel regulatory mechanisms for the sociated with early-onset cancer. Cancer Epidemiol Biomarkers Prev 17(5): proteoglycans decorin and biglycan during muscle formation and muscular dystro- 1127–1135. phy. Matrix Biol 27(8):700–708. 21. Vincenti MP, Brinckerhoff CE (2002) Transcriptional regulation of collagenase (MMP-1, 52. Midwood KS, Orend G (2009) The role of tenascin-C in tissue injury and tumorigen- MMP-13) genes in arthritis: Integration of complex signaling pathways for the re- esis. J Cell Commun Signal 3(3-4):287–310. cruitment of gene-specific transcription factors. Arthritis Res 4(3):157–164. 53. Huang Y, Arora P, McCulloch CA, Vogel WF (2009) The collagen receptor DDR1 reg- 22. Screen HR, Berk DE, Kadler KE, Ramirez F, Young MF (2015) functional ex- ulates cell spreading and motility by associating with myosin IIA. J Cell Sci 122(Pt 10): tracellular matrix. J Orthop Res 33(6):793–799. 1637–1646. 23. Franchi M, Trirè A, Quaranta M, Orsini E, Ottani V (2007) Collagen structure of tendon 54. Valiathan RR, Marco M, Leitinger B, Kleer CG, Fridman R (2012) Discoidin domain relates to function. ScientificWorldJournal 7:404–420. receptor tyrosine kinases: New players in cancer progression. Cancer Rev 24. Sixt M (2012) Cell migration: Fibroblasts find a new way to get ahead. J Cell Biol 31(1-2):295–321. 197(3):347–349. 55. Toriseva MJ, et al. (2007) Collagenase-3 (MMP-13) enhances remodeling of three-di- 25. Omelchenko T, et al. (2001) Contact interactions between epitheliocytes and fibro- mensional collagen and promotes survival of human skin fibroblasts. J Invest blasts: Formation of heterotypic cadherin-containing adhesion sites is accompanied Dermatol 127(1):49–59. by local cytoskeletal reorganization. Proc Natl Acad Sci USA 98(15):8632–8637. 56. Pelham RJ, Jr, Wang Yl (1997) Cell locomotion and focal adhesions are regulated by 26. Juliano RL, Haskill S (1993) Signal transduction from the extracellular matrix. J Cell substrate flexibility. Proc Natl Acad Sci USA 94(25):13661–13665. Biol 120(3):577–585. 57. Wells RG (2008) The role of matrix stiffness in regulating cell behavior. Hepatology 27. Lukashev ME, Werb Z (1998) ECM signalling: Orchestrating cell behaviour and mis- 47(4):1394–1400. behaviour. Trends Cell Biol 8(11):437–441. 58. Chung L, et al. (2000) Identification of the (183)RWTNNFREY(191) region as a critical 28. Keshet Y, Seger R (2010) The MAP kinase signaling cascades: A system of hundreds of segment of matrix metalloproteinase 1 for the expression of collagenolytic activity. components regulates a diverse array of physiological functions. Methods Mol Biol J Biol Chem 275(38):29610–29617. 661:3–38. 59. Knight CG, Willenbrock F, Murphy G (1992) A novel coumarin-labelled peptide for 29. Fincham VJ, James M, Frame MC, Winder SJ (2000) Active ERK/MAP kinase is targeted sensitive continuous assays of the matrix metalloproteinases. FEBS Lett 296(3): to newly forming cell-matrix adhesions by integrin engagement and v-Src. EMBO J 263–266. 19(12):2911–2923. 60. Plotnikov A, Chuderland D, Karamansha Y, Livnah O, Seger R (2011) Nuclear extra- 30. Besson A, Davy A, Robbins SM, Yong VW (2001) Differential activation of ERKs to cellular signal-regulated kinase 1 and 2 translocation is mediated by casein kinase 2 focal adhesions by PKC epsilon is required for PMA-induced adhesion and migration and accelerated by autophosphorylation. Mol Cell Biol 31(17):3515–3530. of human glioma cells. Oncogene 20(50):7398–7407. 61. Trapnell C, Pachter L, Salzberg SL (2009) TopHat: Discovering splice junctions with 31. Banerjee I, Yekkala K, Borg TK, Baudino TA (2006) Dynamic interactions between RNA-Seq. 25(9):1105–1111. myocytes, fibroblasts, and extracellular matrix. Ann N Y Acad Sci 1080:76–84. 62. Anders S, Huber W (2010) Differential expression analysis for sequence count data. 32. Halbleib JM, Nelson WJ (2006) Cadherins in development: Cell adhesion, sorting, and Genome Biol 11(10):R106. tissue morphogenesis. Genes Dev 20(23):3199–3214. 63. Vizcaíno JA, et al. (2016) 2016 update of the PRIDE database and its related tools. 33. Lu P, Takai K, Weaver VM, Werb Z (2011) Extracellular matrix degradation and re- Nucleic Acids Res 44(D1):D447–D456. modeling in development and disease. Cold Spring Harb Perspect Biol 3(12):1–24. 64. Shevchenko A, Wilm M, Vorm O, Mann M (1996) Mass spectrometric sequencing of 34. Berry CC, Campbell G, Spadiccino A, Robertson M, Curtis AS (2004) The influence of proteins silver-stained polyacrylamide gels. Anal Chem 68(5):850–858. microscale topography on fibroblast attachment and motility. Biomaterials 25(26): 65. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS (1999) Probability-based protein iden- 5781–5788. tification by searching sequence databases using mass spectrometry data. Electrophoresis 35. Lo CM, Wang HB, Dembo M, Wang YL (2000) Cell movement is guided by the rigidity 20(18):3551–3567. of the substrate. Biophys J 79(1):144–152. 66. Keller A, Nesvizhskii AI, Kolker E, Aebersold R (2002) Empirical statistical model to 36. Hodge AJ, Schmitt FO (1960) Charge profile of the tropocollagen macromolecule and estimate the accuracy of peptide identifications made by MS/MS and database search. the packing arrangement in native fibrils. Proc Natl Acad Sci USA 46(2):186–197. Anal Chem 74(20):5383–5392. 37. Ruggeri A, Benazzo F, Reale E (1979) Collagen fibrils with straight and helicoidal 67. Nesvizhskii AI, Keller A, Kolker E, Aebersold R (2003) A statistical model for identi- microfibrils: A freeze-fracture and thin-section study. J Ultrastruct Res 68(1):101–108. fying proteins by tandem mass spectrometry. Anal Chem 75(17):4646–4658. 38. Orgel JP, Irving TC, Miller A, Wess TJ (2006) Microfibrillar structure of type I collagen 68. Li GZ, et al. (2009) Database searching and accounting of multiplexed precursor and in situ. Proc Natl Acad Sci USA 103(24):9001–9005. product spectra from the data independent analysis of simple and complex 39. Sarkar SK, Marmer B, Goldberg G, Neuman KC (2012) Single-molecule tracking of peptide mixtures. Proteomics 9(6):1696–1719. collagenase on native type I collagen fibrils reveals degradation mechanism. Curr Biol 69. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical and 22(12):1047–1056. powerful approach to multiple testing. J R Stat Soc B 57:289–300.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1519676113 Solomonov et al. Downloaded by guest on September 23, 2021