Reversion‐inducing cysteine‐rich protein with Kazal motifs and MT1‐MMP promote the formation of robust fibrillin fibers

Abstract Fibrillins (FBNs) form mesh‐like structures of microfibrils in various elastic tissues. RECK and FBN1 are co‐expressed in many human tissues, suggesting a functional relationship. We found that dermal FBN1 fibers show atypical morphology in mice with reduced RECK expression (RECK‐Hypo mice). Dermal FBN1 fibers in mice‐lacking membrane‐type 1‐matrix metalloproteinase (MT1‐MMP) show a similar atypical morphology, despite the current notion that MT1‐MMP (a membrane‐bound protease) and RECK (a membrane‐bound protease inhibitor) have opposing functions. Our experiments using dermal fibroblasts indicated that RECK promotes pro‐MT1‐MMP activation, increases cell‐associated gelatinase/collagenase activity, and decreases diffusible gelatinase/collagenase activity, while MT1‐MMP stabilizes RECK in these cells. Experiments using purified proteins indicate that RECK and its binding partner ADAMTS10 keep the proteolytic activity of MT1‐MMP within a certain range. These findings suggest that RECK, ADAMTS10, and MT1‐MMP cooperate to support the formation of robust FBN1 fibers.

We previously found that RECK binds and stabilizes ADAMTS10 (Matsuzaki et al., 2018). The RECK gene, initially isolated as a transformation suppressor gene (Takahashi et al., 1998), is conserved from insects to mammals as a single gene and encodes a glycosylphosphatidylinositol-anchored glycoprotein of 125 kDa forming cowbell-shaped dimers (Omura et al., 2009). In the mouse, Reck is expressed in multiple tissues, and Reck-null mice die around Embryonic Day 10.5 with reduced tissue integrity (Almeida et al., 2015;Oh et al., 2001).
MT1-MMP, a type I transmembrane protein, can digest a broad range of extracellular matrix components, including type I collagen, and promote cell migration, invasion, and proteolytic activation of zymogens, such as pro-MMP2 and pro-MMP13 (Gifford & Itoh, 2019;Sato et al., 1994;Sternlicht & Werb, 2001). Mmp14-deficient (MT1knockout [KO]) mice exhibit craniofacial dysmorphism, arthritis, osteopenia, dwarfism, and fibrosis of soft tissues; these phenotypes have been attributed to the ablation of the collagenolytic activity essential for the development of skeletal and other connective tissues (Holmbeck et al., 1999).
In this study, we used two lines of mutant mice, one with reduced RECK expression (RECK-Hypo) and the other lacking MT1-MMP (MT1-KO), together with dermal fibroblasts derived from these mice (MDFs) to examine the effects of RECK and MT1-MMP on FBN1 fiber formation. We also employed a human cell line, MG63, to investigate the effects of complete loss of RECK expression on this process. Our data suggest that RECK, ADAMTS10, and MT1-MMP cooperate to promote the formation of robust FBN1 fibers.

| Mice
Mice carrying mutant Reck alleles (Δ, Low; summarized in Figure S2a) have been described (Almeida et al., 2015;Yamamoto et al., 2012), and the Mmp14-deficient mouse has also been described (Sakamoto & Seiki, 2009). Mice used in this study were on the C57BL/6J genetic background unless stated otherwise. All animal experiments were approved by the Animal Experimentation Committee, Kyoto University, and conducted in accordance with its regulations.

| Antibodies
Primary antibodies used in this study: Rabbit polyclonal anti-human FBN1 (HPA021057, used in immunofluorescence staining of MG63 cells; Sigma-Aldrich), mouse monoclonal anti-FN (610078; BD), mouse monoclonal anti-RECK (5B; Takahashi et al., 1998) 2.3 | Cell culture, gene transfer, and conditioned media HEK293 and MG63 cells were obtained from the American Type Culture Collection and maintained in growth medium (GM) consisting of Dulbecco's modified Eagle's medium supplemented with 10% (vol/vol) fetal bovine serum, streptomycin sulfate (100 µg/ml), and penicillin G (100 U/ml). MDFs were prepared from the back skin of mice at Postnatal Days 3 or 4 by the cold trypsin method (Freshney, 1987). In brief, the skin tissues were cut into small pieces (∼8 mm 3 ), soaked in 0.025 mg/ml trypsin in phosphate-buffered saline (PBS). After overnight incubation at 4°C, excessive trypsin was removed, and the sample was incubated at 37°C for 20 min. After addition of 1 ml trypsin-ethylenediaminetetraacetic acid (EDTA; 0.5 mg/ml trypsin and 0.53 mM EDTA), the cells were dispersed extensively by pipetting. The suspension was diluted with 6 ml GM to quench trypsin, dispersed by pipetting, passed through a cell strainer (100 µm; BD) to remove debris, and the cells were collected by centrifugation at 1,000 rpm for 10 min followed by resuspension and plating with GM. Plasmid transfection was performed using either Lipofectamine 2000 (Invitrogen) or a CalPhos Mammalian Transfection Kit (Clontech) using late passage MDFs (more than 18 passages). Retrovirus mediated transfer of RECK complementary DNA using LXSB vector was performed as described previously  transfected with pEF-Adamts10 (Ats10-CM) or parental HEK293 cells (control CM) were rinsed with PBS and incubated for 5 days in serumfree CD293 medium (Thermo Fisher Scientific). The culture supernatant was collected, filtered (pore size: 0.22 µm), and frozen at -80°C.

| Generating RECK-deficient MG63 cells
For inactivating the RECK gene in MG63 cells, we followed the protocol of Ran et al. (2013) employing the double-nicking strategy using the Cas9 nickase mutant. The following three plasmids were kindly provided by Dr. Kanako Yuki: two hSpCas9-based vectors expressing a gsRNA targeting the sequence starting from 7 (minus strand) or 54 (plus strand) bases downstream of the RECK initiation codon, and the repair template containing the puromycin resistance gene cassette flanked by the left (999 bp) and the right (1,020 bp) RECK homology sequences in a pTA2 vector (Toyobo). Primers used in these constructs are listed in Table S1.
The three plasmids were cotransfected into MG63 cells using Lipofectamine 2000 (Invitrogen), and the transfectants were selected in GM containing 2 µg/ml puromycin. Thirty-five puromycin-resistant clones were isolated, and their RECK expression examined by an immunoblot assay. RECK expression was reconstituted using retrovirus-mediated gene transfer. The RECK-deficient MG63 cell lines m1 and m2 are also known as BB9 and BE6, respectively, and the latter has been described elsewhere (Matsuzaki et al., 2018).

| Immunofluorescence staining
For double-staining FBN1 and FN, MDFs or MG63 cells (2 × 10 4 /well) were plated and incubated for 4 (MDFs) or 7 days (MG63) on eight-well chamber slides (MATSUNAMI: SCS-008). The cells were gently washed with PBS, dried at RT, fixed for 10 min in cold acetone (−80°C), and then stained as described for skin tissues (see above) using anti-FBN1 diluted 1:500 and anti-FN diluted 1:400. Images were recorded using a confocal microscope (TCS SP8; Leica), and the stacked images of Z-series optical sections are presented. Three properties (i.e., peak, valley, and width) of fluorescent signals/structures on micrographs were determined as illustrated in Figure S3. First, a linear "region of interest (ROI)" was selected manually on a multicolor micrograph using the "Linear Profile" function of LASX (Leica; Figure S1a), which yields the profile of signal intensity along with the ROI in two forms: a line graph (jpg format; Figure S3b) and its numerical dataset (csv format). The fluorescence intensity of all peaks (corresponding to highlights) and valleys (corresponding to dark spots; Figure S3c) were extracted from the numerical data using Excel (Microsoft). The width of a spike horizontally cut at a threshold intensity ( Figure S3d) was also determined using Excel (Microsoft). Threshold intensity was adjusted in each experiment by comparing several ROIs on the original image (ROI.001-ROI.005 in Figure S3a) and their corresponding fluorescence profiles (in jpg format) so as to cut across all fibers visible on the image. Note that the width of fiber thus determined can be equal or larger than the actual diameter of the fiber, depending of the angle between the ROI and the fiber ( Figure S3e). Results are presented using box-and-whisker plots with symbols indicating mean (X) and median (bold horizontal bar; Figure S3f).

| Gelatinolysis assay in vitro
where µ stands for the average of four wells containing the same sample; Ei is the fluorescence emitted from a single well of the experimental sample (containing pMT1 or aMT1) at time point i; and Bi the fluorescence at time point i emitted from a blank well (no pMT1 or aMT1), which represents spontaneous nonenzymatic degradation of the substrate.

| In situ zymography
Early passage MDFs, up to three passages, were seeded at 2 × 10 5 /well on eight-well chamber slides and cultured for 3 days.

| Co-expression of RECK and FBN1 in normal human organs
In an effort to find clues to the function of RECK in vivo, we utilized the database ONCOMINE to detect genes co-expressed with RECK in normal human organs. In a relatively large dataset (Roth normal; 353 samples from 65 organs), several genes sharing high similarity in tissue distribution with RECK were detected, including PTRF, LAMA4, TIMP3, SH3D19, CAV1, CAV2, FBN1, VGLL3, and NOTCH2 ( Figure S1a). Among these genes, FBN1 is of particular interest since RECK was top-ranked in the converse search using FBN1 as a query term ( Figure S1b). These results raise the possibility that RECK and FBN1 play roles in a common biological process so that their expressions need to be regulated in a tightly coordinated fashion.

| Abnormal FBN fibers, tissue architecture, and elastin fibers in the skin of mice with reduced RECK expression
To test whether RECK affects FBN1, we visualized and compared FBN1 fibers in the skin tissues of control and Reck mutant mice by immunofluorescence staining (Figure 1a). Since Reck-null mice die in utero, we used viable Reck mutant mice, named RECK-Hypo, with greatly reduced RECK expression (ca., 20% of the normal level; Low/Δ in Figure S2b). Control littermates (Low/+) express RECK at about 90% of the normal level ( Figure  We also stained FBN2 in these mice ( Figure S4). In immunogold electron microscopy, smoothly curved continuous threads with a typical FBN1 labeling pattern (an array of equally spaced clusters of gold particles, representing the beaded-chain structure (Keene et al., 1991) were frequently found in the skin tissues of control mice while such typical structures were rarely found in RECK-Hypo mice; instead, amorphous clusters of gold particles were frequently found in the skin tissues of RECK-Hypo mice (Figure 1a, Panels 11 and 12).
These observations suggest that RECK is required for the formation of smoothly curved FBN fibrils.
Electron microscopy also revealed that in RECK-Hypo mice, unusually large bundles of collagen fibrils are found in some regions of the papillary dermis ( Figure S5a,b, brackets) and fat cells are closer to the epidermis than in the control mice ( Figure S5d).
Since FBN fibers are often associated with elastin fibers in the skin, we visualized elastin fibers in the skin tissues of 10 week-old mice by Elastica van Gieson staining; elastin fibers were significantly shorter in RECK-Hypo mice (see Figure 1d, Panels 1-4 for typical images; Panel 5 for quantification).
These observations indicate that RECK affects the properties of FBN fibers and elastin fibers as well as tissue architecture in mouse skin.

| Altered morphology of FBN1 and FN fibers in cultured RECK-deficient cells
To facilitate further studies on the mechanisms by which RECK affects FBN-fiber formation, we employed two in vitro experimental systems (Figures 2 and 3). First, we prepared MDFs from wild-type and RECK-Hypo mice, culturing these cells under identical conditions, and performed immunofluorescence double-staining with antibodies against FBN1 (red) and FN
Moreover, in the present study, RECK was found to have a significant impact on FN-fiber formation by MG63 cells (see Figure 3). We, therefore, examined the major components of FN receptor (integrins α 5 , α v , and β 1 ) and another member of this protein family, integrin α 2 (a component of collagen receptor) in the MG63-derived cell lines by immunoblot assay (Figure 4b-d; Figure S7a). In the cases of integrins α 2 , α 5 , and β 1 , the intensity of smaller bands relative to the full-length band was decreased when RECK was expressed (Figure 4b-d, arrowheads; densitometry data in Figure 4g-i), suggesting that RECK protects these integrins from degradation. On the other hand, no prominent and consistent effects of RECK on integrin α v was found ( Figure S7a). It has been reported that ADAMTS10 cleaves FBN1 and accelerates FBN1 fiber formation (Kutz et al., 2011). We previously found that RECK binds and stabilizes ADAMTS10 (Matsuzaki et al., 2018).
We, therefore, investigated the effects of RECK and ADAMTS10 on FBN in the MG63 system (Figure 4l,m). Cells transfected with control vector (V) or RECK-expression vector (R) were exposed to condi-   Figure S9a). There was no apparent difference in the width of the FBN2 fibers in the MT1-KO mice and the control mice ( Figure S9b).
In addition, we noted that the skin of MT1-KO mice tend to be thinner than that of wild-type mice ( Figure S9c,d), another feature common between MT1-KO mice and RECK-Hypo mice (see Figure S2c,d).

F I G U R E 4 Immunoblot detection of RECK, integrins, FN, and FBN in RECK-KO MG63 mutants. Lysates of MG63 cells (Lanes 1 and 2) and
two RECK-deficient sublines, m1 (Lanes 3 and 4) and m2 (Lanes 5 and 6), transfected with either empty vector (V; Lanes 1, 3, and 5) or RECK-expression vector (R; 2, 4, and 6) were subjected to immunoblot assay with antibodies against RECK (a), integrin β 1 (b), integrin α 2 (c), integrin α 5 (d), FN (e), and FBN (f). Symbols: #, full-length protein; red arrowhead, a fragment more abundant when RECK is absent; and green arrow, a fragment more abundant when RECK is present. (f) The full-length FBN bands (312-315 kDa) are undetectable under these conditions. (g-k) Effects of RECK on fragments of integrins, FN, and FBN. (b-f) Immunoblot images together with similar data obtained using two other RECK-deficient lines (data not shown) were subjected to densitometry. The ratio of intensity between each integrin subfragment and the full-length integrin band (g-i) and the relative intensity of FN (j) and FBN (k) normalized against GAPDH were determined. The graph represents the mean ± standard error of the mean. Student's t test: *p < .05. (l) Immunoblot detection of FBN fragments in MG63 conditioned media. Confluent cultures of the indicated transfectants were exposed for 3 hr to the same volume of conditioned medium prepared from the control cells (Lanes 1, 3, 5, 7, 9, and 11) or ADAMTS10-expressing cells (Lanes 2, 4, 6, 8, and 10), and then the culture supernatants were subjected to immunoblot assay using anti-FBN antibodies. (m) The data were subjected to densitometry, and the total density of three bands (arrowheads) are presented. Note that the amount of FBN fragments released from the cells is higher in the absence of RECK (Lanes 5, 6 vs. 7, 8; Lanes 9, 10 vs. 11, 12) and that in the absence of RECK, the addition of ADAMTS10 to the cells increases the amount of FBN fragments released from the cells (Lanes 6, 10 vs. 5, 9). FBN, fibrillin; FN, fibronectin; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; KO, knockout F I G U R E 5 FBN1 fibers in the skin of MT1-KO mice and immunoblot characterization of MT1-KO MDFs. (a) Immunological visualization of FBN1 fibers. Tissue sections prepared from corresponding areas of the back skin of wild-type (WT; left panels) and MT1-KO (Mmp14 −/− ; right panels) male mice at Postnatal Day 15 were subjected to immunofluorescence staining using anti-FBN1 antibodies (red signals) followed by nuclear counterstaining (blue signals). Scale bar = 40 µm in (a1 and a2) and scale bar = 10 µm in (a3-a8). Two parts of (a3) indicated by box-A and box-B are shown at higher magnification in (a5 and a7), respectively; two parts of (a4) are similarly shown in (a6 and a8). were also found when DQ-collagen was used as a substrate (i.e., collagenolysis assay; data not shown). Note that when both RECK and ADAMTS10 were present, the gelatinolytic activity of pMT1 was significantly increased (b: Bar 4; p = .0164) while the gelatinolytic activity of aMT1 was significantly decreased (c: Bar 4; p = .0319). (d and e) In situ zymography of MDFs. Early passage MDFs (up to three passages) prepared from wild-type (WT; left panels) and RECK-Hypo (right panels) mice were cultured on eight-well chamber slides for 3 days, then overlaid with collagen gel containing DQ-gelatin (d) or DQ-collagen (e), incubated for 24 hr, and subjected to nuclear counterstaining (blue). (d and e) Nuclear blue fluorescence (Panels 1 and 6), green fluorescence representing proteolysis (Panels 2 and 7), and the merged images (Panels 3-5, 8-10) are presented; boxed areas in Panels 3 and 8 are shown at higher magnification in Panels 4 and 9, respectively; and the virtual Z-axis sections (reconstituted from Z-series images) along the white dotted lines in Panels 3 and 8 are shown in Panels 5 and 10, respectively. For DQ-collagen, the intensity profile of green fluorescence along a linear ROI is also shown (Panels 11 and 12) in which the white broken line indicates the median fluorescence and the yellow broken line the maximum fluorescence. Scale bar = 50 µm in Panels 1-4 and 6-9, and Scale bar = 20 µm in Panels 5 and 10. (f) Cumulative density of green fluorescence. Densitometric determination of green fluorescence was performed on images as shown in Panels 2 and 7 in (d) and (e). The bar represents mean ± standard error of the mean (n = 12 for DQ-gelatin; n = 8 for DQ-collagen). Welch's t test: **p < 5 × 10 −3 , ***p < 5 × 10 −7 . Note that in RECK-Hypo MDFs, the green signals associated with individual cells are decreased (Panels 9 vs. 4 in d and e) while the overall, diffuse green signals are increased (Bars 2, 4 vs. 1, 2 in f). MDF, mouse dermal fibroblast; MT1-MMP, membrane-type 1-matrix metalloproteinase; ROI, region of interest Our initial question was why are RECK and FBN1 co-expressed in human tissues. As an approach to address this question, we attempted to characterize FBN fibers in Reck mutant mice. Since mice die in utero when Reck is completely inactivated in the whole body (Oh et al., 2001), we chose to use a viable mutant with greatly reduced RECK expression (RECK-Hypo; Yamamoto et al., 2012) in this study. We found that dermal FBN fibers show atypical morphology in these mice (Figure 1a-c; Figure S4). Consistent to the known function of FBN fibers in stabilizing elastin fibers (Baldwin et al., 2013), RECK-Hypo mice also have elastin fibers of significantly shorter length than in the control mice (Figure 1d), suggesting that Reck not only affects the morphology but also supports the function, of FBN fibers in vivo. In vitro studies showed that RECK-Hypo MDFs and RECK-KO MG63 cells failed to produce FBN1-fiber network as elaborate as that produced by control cells (Figures 2 and 3). These To objectively characterize FBN and FN fibers, we attempted to quantify the intensity of fluorescence (at peaks and valleys) and the width of fluorescent fibers (see Figure S3). We chose to analyze fluorescence along with a linear ROI (i.e., one-dimensional measurement) to facilitate data acquisition and evaluation. A drawback of this simplification is that we cannot discriminate between fibers and other structures (e.g., spots, dots, and aggregates) from the numerical data obtained. We can, however, address this issue by comparing a few linear ROIs on the original image ( Figure S3a, colored lines) and their corresponding intensity profiles ( Figure S3b) to see what kind of structures are cut across by the ROIs. Nevertheless, this method was useful in most cases in extracting data regarding features of immunofluorescence images, such as the intensity and variation of fluorescent signals at both highlights (peaks) and dark spots (valleys; Figure S3c) and the width of the fibers and its variation ( Figure S3d,e). Our findings through this method in various images obtained in this study are summarized in Figure S12a. FBN1 is expressed in multiple human organs ( Figure S1b), while FBN2 is known to be mainly expressed in developing tissues (Quondamatteo et al., 2002) and upregulated in healing wounds (Brinckmann et al., 2010). In the skin tissues of both RECK-Hypo mice (at P4) and MT1-KO mice (at P15), FBN2-positive areas were expanded ( Figure S4a, Panels 2 vs. 1; Figure S8b vs. S8a) and the intensity of FBN2 signals was increased ( Figure S4b; Figure S9a; indicated by the red box in Figure S12b) The net effect would be to keep the proteolytic activity of MT1-MMP within a certain range. On the basis of these observations (summarized in Figure S12c), we propose novel roles for RECK, ADAMTS10, and MT1-MMP in FBN-fiber formation ( Figure S12d).

| Effects of RECK on gelatinolytic and collagenolytic activities associated with MDFs
This model proposes that ADAMTS10 contributes to the formation of robust FBN fibers by working together with RECK to control the activity as well as localization of MT1-MMP.
What would be the roles of FN and its receptors in this scheme?
Previous works have provided evidence indicating that FBN-fiber formation requires FN fibers (Hubmacher & Apte, 2015;Kinsey et al., 2008;Sabatier et al., 2009) and that FN-fiber formation requires mechanical tension produced by the movement of FN receptors on the cell surface (Geiger, Bershadsky, Pankov, & Yamada, 2001;Schwarzbauer & Sechler, 1999). Most of our observations in cultured cells (summarized in Figure S12a, Nos. 2-4) agree in that RECK mutations simultaneously reduce the formation of FBN1 fibers and FN fibers. On the other hand, RECK stabilizes FN receptor components (Figures 4b and 4d) and increases the amount of cell-associated FN as well as FBN1 fragments (Figure 4e and 4f).
These findings are consistent with the idea that the primary role for RECK in this system is to promote the formation of robust FN fibers competent for stimulating proper FBN-fiber biogenesis.
Our observation with RECK-Hypo MDFs that FBN1 fibers tend to Our model postulating that the protease MT1-MMP as well as the protease regulator RECK promote FBN-fiber formation ( Figure   S12d) may seem counterintuitive. Compelling evidence indicates a role for MT1-MMP in cell migration (Gifford & Itoh, 2019). RECK has also been implicated in the stability of focal adhesions required for persistent directional migration of mouse embryo fibroblasts . It is reasonable to speculate that proper cell migration requires proteases (assisting detachment from the substrate) and their inhibitors (assisting attachment to the substrate) whose activities are tightly regulated and coordinated spatially and temporally. Whether persistent directional migration is required for the formation of straight FBN fibers with uniform thickness is an interesting question to be tested in future studies. RECK and FBN1 are co-expressed in multiple tissues, but our model does not require direct interaction between RECK and FBN1. Addressing the question of whether RECK co-localize with FBN1 in the tissues may be an important step in testing the feasibility of the above model.
A single-nucleotide polymorphism (SNP) in the RECK promoter (rs10814325; from TT to TC or CC) has been associated with the risks of developing hepatocellular carcinoma and lymph node metastases of oral cancer (Chung et al., 2012). This SNP is likely to affect the level of RECK gene expression. Whether this or some other RECK SNP is a risk or aggravating factor for FBN-related disorders would be another important question to be addressed in future studies.

Increased collagen gene expression has been found in mice with
Fbn gene mutations (Manne, Markova, Siracusa, & Jimenez, 2013;Sengle et al., 2012), and this has been attributed to deregulated TGFβ signaling (Olivieri, Smaldone, & Ramirez, 2010). Interestingly, unusually large bundles of collagen fibrils were found in some regions of papillary dermis in RECK-Hypo mice ( Figure S5b), which may support the idea that the function of FBN fibers to sequester latent forms of TGF-β family proteins are compromised in these mice. In fact, our preliminary data indicate that the number of nuclei positive for phospho-Smad1/5/8 is increased in RECK-Hypo skin (unpublished), suggesting deregulated BMP signaling.
In summary, our data implicate RECK and MT1-MMP in proper FBN-fiber biogenesis and suggest the possible relevance of ADAMTS10 to this mechanism.