Type XI1 and XIV Collagens Mediate Interactions Between Banded Collagen Fibers in Vitro and May Modulate Extracellular Matrix Deformability*

Type XII and XIV collagens are very large molecules containing three extended globular domains derived from the amino terminus of each a chain and an interrupted triple helix. Both collagens are genetically and immunologically unique and have distinct distributions in many tissues. These collagens localize near the surface of banded collagen fibrils. The function of the molecules is unknown. We have prepared a mixture of native type MI and XIV collagens that is free of contaminating proteins by electrophoretic criteria. In addition, we have pu- rified the collagenase-resistant globular domains of type XI1 or XIV collagens CUI-NC-3 or XIV-NC-3). In this study, we have investigated the effect of intact type XI1 and XIV and XII-NC-3 or XIV-NC-3 on the interactions between fibroblasts and type I collagen fibrils. We find that both type XI1 and XIV collagens promote collagen gel contrac- tion mediated by fibroblasts, even in the absence of serum. The activity is present in the NC-3 domains. The effect is dose-dependent and is inhibited by denatur-ation. The effect of type MI NC-3 is inhibited by the ad- dition of anti-XI1 antiserum. To elucidate the mechanism underlying this phenomenon, we examined the effect of XII-NC-3 or Mv-NC-3 on deformability of collagen gels by centrifugal force. XII-NC-3 or XIV-NC-3 markedly pro- motes gel compression after centrifugation. The effect is also inhibited by denaturation, and the activity of type MI-NC3 is inhibited by the addition of anti-MI antiserum. The results indicate that the effect of MI-NC-3 or XIV-NC-3 on collagen gel contraction by fibroblasts is not due to activation of cellular events but rather results from the increase in mobility of hydrated collagen fibrils within the gel. These studies suggest that collagen types MI and XIV may modulate the biomechanical properties of tissues. domains at the amino terminus ex-tension Rotary shadowing im-ages globular domains se-quences al(XI1) is synthesized in two forms. The molecular mass of one is greater than the other by 90 kDa, and each globular arm of molecules containing the larger form is longer by 27 nm (3, 4), as predicted from the cDNA sequence analysis (1). The longer form is the

Type XII and XIV collagens are very large molecules containing three extended globular domains derived from the amino terminus of each a chain and an interrupted triple helix. Both collagens are genetically and immunologically unique and have distinct distributions in many tissues. These collagens localize near the surface of banded collagen fibrils. The function of the molecules is unknown. W e have prepared a mixture of native type MI and XIV collagens that is free of contaminating proteins by electrophoretic criteria. In addition, we have purified the collagenase-resistant globular domains of type XI1 or XIV collagens CUI-NC-3 or XIV-NC-3). In this study, we have investigated the effect of intact type XI1 and XIV and XII-NC-3 or XIV-NC-3 on the interactions between fibroblasts and type I collagen fibrils. W e find that both type XI1 and XIV collagens promote collagen gel contraction mediated by fibroblasts, even in the absence of serum. The activity is present in the NC-3 domains. The effect is dose-dependent and is inhibited by denaturation. The effect of type MI NC-3 is inhibited by the addition of anti-XI1 antiserum. To elucidate the mechanism underlying this phenomenon, we examined the effect of XII-NC-3 or Mv-NC-3 on deformability of collagen gels by centrifugal force. XII-NC-3 or XIV-NC-3 markedly promotes gel compression after centrifugation. The effect is also inhibited by denaturation, and the activity of type MI-NC3 is inhibited by the addition of anti-MI antiserum. The results indicate that the effect of MI-NC-3 or XIV-NC-3 on collagen gel contraction by fibroblasts is not due to activation of cellular events but rather results from the increase in mobility of hydrated collagen fibrils within the gel. These studies suggest that collagen types MI and XIV may modulate the biomechanical properties of tissues. ~~ Type XI1 and XIV collagens are structurally similar. Both contain a relative short and interrupted triple-helical domain plus three large and apparently identical globular domains at the amino terminus (NC-3 domain), each projecting as an extension of one of the subunit a1 chains. Rotary shadowing images of these globular domains appear as elongated structures, consistent with the predictions from the cDNA deduced sequences (1,2  No corresponding size collagen variant has been observed or predicted for type XIV (2,5,6). Furthermore, both type XI1 and type XIV can be glycanated (7).
Type XI1 and XIV collagens are found in many tissues. They occur together within the same region in some tissues, but ofien they show distinctive regional distributions. For example, skin type XI1 is predominately found in the papillary dermis, while type XIV predominates within the reticular dermis. In blood vessels, type XI1 is present within the intimal layer, and type XIV is in the adventitia (8).
Ultrastructural immunolocalization of type XI1 and XIV indicates that both molecules are found associated with the banded collagen fibrils within the interfibrillar space. In some micrographs, the molecules appear to associate with the fibril surface and to bridge adjacent fibrils (9). Together, these observations suggest that ( a ) type XI1 and XIV collagens associate with the surface of the banded collagen fibrils through a portion of the triple helix and ( b ) the NC-3 domain projecting from the fibril surface influences additional interactions between fibrils or between fibrils and cells. So far these hypotheses have not been directly tested, although type XIV has been shown to bind to heparan sulfate proteoglycan (10). Surprisingly, no binding of type XIV to fibrous collagens or to cells was observed when evaluated by solid-phase binding assays, although binding to type VI collagen was seen (10).
We investigated the functions of type XI1 and XIV collagens in vitro using dermal equivalents. Neonatal human skin fibroblasts were cultured within a gel made from type I collagen. In the presence of serum, the fibroblasts contract the collagen gel by a mechanism that is still only poorly understood but that directly involves integrin a201 (11)(12)(13). The majority of the evidence supports the model that a2pl mediates the transmembrane interactions of the cellular actin cytoskeleton and gelled collagen fibrils. Contraction of the actin network acts upon individual collagen fibrils to pull them through the gelled matrix, away from the periphery of the gel, and against frictional forces derived from lateral associations of the collagen fibrils. As the gel contracts, the collagen fibril concentration increases. Contraction of the gel continues until the frictional force equals the actin-generated cellular force. Further contraction will occur if the cellular force is increased or if gel friction is decreased. While the contribution of the frictional resistance to the process of gel contraction has been relatively unstudied, the role of the cell-derived force has received considerable attention. The forces driving contraction are increased by factors that increase cell numbers or the amount of a2pllcell. These factors include
The molecular description of the mechanism by which a 2 p l mediates the interaction with collagen fibrils is only partly understood. a 2 p l is believed to bind ligands through a tetrapeptide sequence (DGEA) within cyanogen bromide cleavage fragment 3 of the a1 chain of type I collagen (12,16). In collagen fibrils, these occur along the length of the collagen a chains within the triple-helical domain. This is problematic since the glycine residue is at the center of the triple helix, making the sequence unavailable to the integrin as long as the helical conformation is intact. Binding could occur during "breathing" of the triple helix, but the physical state of the collagen helix during cell binding has not been explored in detail. It is a formal possibility that another molecule may bridge a2pl and the collagen fibril surface and be responsible for cell binding to collagen fibrils in uiuo. Collagens XI1 and XIV are obvious candidates.
Since type XI1 and XIV collagens occur in proximity to the collagen fibril surface, either or both could participate in increased integrin binding to the collagen fibril, or in modulating interfibrillar interactions. Therefore, using the dermal equivalent system to assess type XI1 and XIV function appeared appropriate. In this study, we have investigated the effect of NC-3 domains of type XI1 or type XIV collagens, which is expected to influence interactions between fibrils and cells or between fibrils.

MATERIALS AND METHODS
Purification of o p e XII and XN Collagens and Their NC-3 Domains-Intact type X I and XIV collagens were extracted and purified from fetal bovine skin as described in a previous paper (3,8). The mixture of type XI1 and XIV collagens was dialyzed against Dulbecco's modified Eagle's medium (DMEM, Life Technologies, Inc.). After sterilization by filtration, type X I W collagens in DMEM were stored at -80 "C prior to use. The globular domains (collagenase-resistant domains) of type XI1 and XIV collagens (XI-NC-3 and XIV-NC-3) were extracted and purified from fetal bovine skin as described previously (8). The appropriate Mono Q fractions were stored at -80 "C. Protein concentrations were determined using the bicinchoninic acid-based assay according to the manufacturer's instructions (Pierce Chemical Co.). When added to cell cultures, XI-NC-3 and XIV-NC-3 solutions were dialyzed extensively against DMEM. After sterilizing by filtration, XI-NC-3 and XIV-NC-3 in DMEM were kept at -80 "C prior to use.
Cell Culture-Fibroblasts were isolated from human foreskin as described previously (17). The cultures were initiated as out-growths from explants of the skin. The primary cultures were grown in DMEM supplemented with 10% fetal bovine serum (FBS), 3.7 g/ml sodium bicarbonate, and 50 unitdml penicillin and 50 pg/ml streptomycin (Life Technologies, Inc.). Subconfluent cells were dispersed with 0.1% trypsin and 0.02% EDTA in phosphate-buffered saline, pH 7.4 (PBS), and propagated in the above medium. Cultures at the 3rd population doubling were frozen and kept in liquid nitrogen. Foreskin fibroblasts from the 6th-9th population doublings were used throughout these experiments. The cells grown in 10% FBSDMEM were detached by washing with PBS and subsequently treated with 0.1% trypsin and 0.02% EDTA in PBS. After detachment, trypsin activity was neutralized by the addition of soybean trypsin inhibitor (0.1%). The cells were harvested by centrifugation, resuspended twice in serum-free DMEM, and counted.
Preparation of Collagen Gels-A solution of bovine skin pepsintreated type I collagen in dilute HCl (0.012 N) at a concentration of 3.0 mg/ml (Vitrogen 100) was obtained from Collagen Corp. (Palo Alto, CA). Acid-soluble bovine tendon type I collagen was kindly provided by Dr. Jerome Gross. Three-dimensional gels of collagen fibrils were prepared in 35-mm plastic culture dishes that were coated with 0.2% bovine serum albumin (Sigma) in DMEM and then washed 3 times with DMEM as described previously (17,18j. The collagen solution (12. DMEMj were added to the collagen solution, and the mixture was warmed to room temperature. Two ml of the solution was added to precoated 35-mm plastic culture dishes, and the collagen was polymerized by incubating the solution for 3 h at 37 "C. The peripheral edge of the gels was scraped gently to release them from the dishes, and the collagen gels were allowed to contract uniformly (17).
For centrifugation experiments, collagen solutions were prepared without cells as above. The solutions were then degassed at 4 "C by aspiration, and 1 ml of the solution was added to Eppendorf plastic tubes and incubated for 24 h at 37 "C.
Measurement of Gel Contraction-The extent of collagen gel contraction by fibroblasts was determined by measuring the diameter of the gels. The area of the gels was calculated from the averaged diameters and expressed as a percentage of the original area. Each data point is an average of measurements from three separate gels. The reproducibility of these experiments is high as reported previously (17). The standard deviations of the means among different series of experiments under the same conditions are usually less than 10%.
Measurement of Gel Contraction by Centrifugation-Gels incubated with or without additions for 24 h at 37 "C were centrifuged for 10 min at the indicated relative centrifugal forces. After centrifugation, supernatant solutions were weighed. Collagen gel volume after centrifugation was represented as a percentage of original gel mass 1% of original gel mass = (1000 mgsupernatant mass)/1000 mg x 100%).
Determination of Cell Numbers Within Collagen Gels-The number of cells in collagen gels was determined by measurement of DNA content as described by Labarca and Paigen (19). Briefly, cells grown in gels were placed in a plastic tube and incubated with 0.2% bacterial collagenase in PBS containing 1 m~ CaCl, for 30 min at 37 "C. The floating cells were collected by centrifugation at 300 x g for 10 min. The cell pellets were washed twice with 0.1% trypsin plus 0.02% EDTA in PBS. DNA content was determined fluorometrically.
Cell Attachment Assays-Cell attachment experiments were performed in 24-multiwell plates (Costar, Cambridge, MA). The plastic plates were incubated with XI-NC-3 or XIV-NC-3 in DMEM for 2 h at 37 "C and then washed 3 times with DMEM. For cell attachment to type I collagen gels, the plates were incubated with a DMEM solution of type I collagen (1 mg/ml) containing XI-NC-3 or XIV-NC-3 for 2 h at 37 "C and then washed 3 times with DMEM. Suspensions of fibroblasts (1.5 x lo5 cells/well) in DMEM were incubated for 1 h at 37 "C with the indicated additions. The number of attached cells was determined by measurement of DNA as described previously (18). The amount of DNA was converted to cell number by using the factor 8.0 pg of DNMcell. A collagen gel containing cells was placed in a plastic tube and incubated with 0.2% bacterial collagenase (1 m~ CaCl, in PBS (pH 7.4)) for 2 h at 37 "C to dissolve the gels. The floating cells were then collected by centrifugation at 300 x g for 10 min. The cell pellets were washed twice with 0.1% trypsin and 0.02% EDTA in PBS. DNA was determined as described by Labarca and Paigen (19).
Heat Denaturation-XI-NC-3 or XIV-NC-3 in DMEM was heated to 60 or 100 "C for the indicated times. After heating, the solutions were rapidly cooled on ice. Following denaturation, an aliquot was analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting as described previously (20)(21)(22) to assure the presence of intact protein.
Antibodies-Polyclonal antisera to type X I and XIV collagens have been previously described (3,9). Polyclonal antibodies were purified from rabbit serum by chromatography on protein G-Sepharose. Monoclonal antibodies to type XI1 and XIV collagens were prepared as described (3,9). Antibodies were tested for cross-reactivity by enzymelinked immunosorbent assay (3,23). The monoclonal antibody to TGF-pl (TB21) was obtained from Anogen Inc. (Ontario, Canada) and the monoclonal antibody to the pl integrin subunit was obtained from Chemicon, Inc. (Temecula, CAI. Recovery of Collagen After Fibril Formation-The collagen gel with XII-NC-3 or XIV-NC-3 was incubated at 37 "C for 24 h and centrifuged at 16,000 x g for 10 min. The pellets were washed 3 times with PBS. Following this procedure, almost all of the XI-NC-3 or the XIV-NC-3, as detected by SDS-polyacrylamide gel electrophoresis and Western blot ting, was in the supernatant solutions after centrifugation. The amount of pelleted collagen in fibrils was determined using the bicinchoninic acid-based assay according to the manufacturer's instructions (Pierce Chemical Co.).
Other Materials-Human recombinant TGF-p1 was purchased from Boehringer Mannheim (Germany). Fibronectin was obtained from Sigma.

RESULTS
A mixture of type XI1 and XIV collagens was purified from other components of an extract of fetal bovine skin under nondenaturing conditions as described under "Materials and Methods." The mixture was judged to be free of contaminating proteins by Coomassie Brilliant Blue staining following gel electrophoresis (data not shown). Unfortunately, we have so far been unable to obtain completely pure preparations of either intact type XI1 or type XIV alone.
Human fibroblasts were isolated from neonatal foreskin and maintained in culture prior to incorporation into Vitrogen 100 gels (2.5 x 10'' cells/ml) as described previously. When the mixture of type XI1 and XIV collagens was added to the Vitrogen, contraction of the gel by the fibroblasts was significantly increased in a dose-dependent manner (Fig. 1) in the absence of serum. Addition of antiserum to type XI1 collagen partially reduced the rate of contraction. In order to evaluate this phenomenon more thoroughly, and since we were unable to obtain pure preparations of native type XI1 and XIV collagens separately, we continued the experimentation with isolated NC-3 domains.
The NC-3 domains of types XI1 and XIV collagens were isolated from bovine skin and digested with collagenase prior to purification. The resulting products showed only the expected electrophoretic gel hands by Coomassie Brilliant Blue staining (data not shown). There was no reactivity of monoclonal antitype XI1 antibody (ClJ) with the type XIV product and no reactivity to monoclonal anti-type XIV (1011G) with the preparation of type XI1 (data not shown). The rotary shadowed image of the molecules contained in each preparation were examined and found to he identical to those described previously for each collagen type (data not shown).
The addition of the NC-3 domains to Vitrogen solutions had no effect upon the rate of gel formation (data not shown), nor upon the percent of Vitrogen incorporated into the gel (Table I).
In the absence of serum and NC-3 domains, the gels contracted minimally (Fig. 2) during 100 h of culture. Addition of the NC-3 domains of type XI1 and XIV collagens markedly enhanced the rate of gel contraction in the absence of serum in a dose-dependent manner (Fig.  2). The addition of serum (0.34,) or

Incuballon Time (h)
A TGF-01 (10 ng/ml) also stimulated grl contraction in the absence of XII-NC-3 and XIV-NC-3 as exprctrd ( Table 11). As shown in Table 111, the rate of gel Contraction XII-NC3 and XIV-NC3 (50 pg/ml) was equal to that caused by serum ( 0 . 3 5 ) and was greater than that causrd by TCF-13 (10 ng/ml I. When serum or TGF-P1 and XII-NC-3 were added simultaneously, the rate of contraction was greater than that that occurred using either agent alone ( Table I1 ), suggrsting that the mrchanism of gel contraction promoted by srrum or by TGF-Pl was independent of that of XII-NC-3. Essentially idrntical rrsults were obtained with XIV-NC-3 fdata not shown,. Further, antibodies to TGF-01 neutralized the effect ofTGF-{jI addition, hut had no effect upon the contraction promoting activity of NC-3 domains ( Table 111).
The gel contraction-promoting activitv of XII-NC-3 could be neutralized by the addition of polyclonal anti-type XI1 collagrn antibodies (Fig. 3). The neutralization was antibody concentration-dependent. Addition of anti-type XIV collagrn antibodies had no effect upon gel contraction promoted by XII-NC-3. nor did anti-type XI1 antibodies have any effwt upon the contraction of gels by type XIV NC-3. Anti-type XI1 antihodirs also had no effect upon gel contraction in the ahsrnce of any promoting agent nor upon contraction in the prrsrncr of swum, indicating that XII-NC-3 is not responsible for the serum effrct and confirming previous observations that XII-NC-3 is not prrsrnt   in normal serum within the limits of detection of Western blotting (data not shown). These data strongly support the hypothesis that the enhancement of gel contraction by type XI1 collagen NC-3 is specific to that molecular domain and is not due to minor contaminants in the preparation. Gel contraction enhancement by type XI1 or type XIV collagens requires the native conformation of the NC-3 domain. Incubation of XII-NC-3 or XIV-NC-3 at 60 "C for 5-30 min, or 100 "C for 15 min causes a significant loss of activity (Fig. 4A). The loss of activity is not due to degradation of the proteins during the incubation since intact XII-NC-3 and XIV-NC-3 were recovered at the end of the incubation period with little or no loss as judged by immunoblotting following electrophoresis (Fig. 4B). Heating the materials caused some loss of interchain disulfide bonding with increased incubation time, particularly with type XIV NC-3 ( Fig. 4 B , lunes 6-8).
The XII-NC-3 domain does not have to be incorporated into the gel during fibril formation in order to enhance gel contraction. As shown in Fig. 5, the addition of the NC-3 domain to the medium overlying the gel causes a rate of contraction equivalent to that seen when XI-NC-3 is added a t gel formation. Identical results were obtained with XIV-NC-3 (data not shown).
The mechanism by which types XI1 and XIV collagen NC-3 domains promote gel contraction is not clear from the studies described above. Therefore, further studies were conducted to evaluate three possible mechanisms. Gel contraction could be promoted by ( a ) an effect upon the cells to up-regulate integrin a2pl either by an increase of integrin upon the cell surface or by an increase in cell number; ( b ) NC-3 domains increasing the interaction of cells with fibrils by binding the cell through a receptor, possibly a2pl or another integrin, and binding to the fibril surface; (c) NC-3 domains decreasing the resistance to contraction by minimizing interfibrillar interactions, allowing individual fibrils to slide past one another.
The gel contraction by fibroblasts is known to depend on the number of cells within the gel. In the absence of additions, no gel contraction was observed by 2.5 x lo4 cells/ml as used in all experiments reported here. However, fibroblasts a t 10 x lo4 cells/ml caused gel contraction equivalent of that seen using 50 pg/ml of NC-3 domains (data not shown). The NC-3 domain of type X I collagen has no effect upon the proliferation rates of cells grown in collagen gels as shown in Table IV. Even in the presence of 0.3% serum, no proliferative effect was observed. Similar results were obtained with XIV-NC-3 (data not shown). Therefore, the NC-3 domains do not have mitogenic effects nor are the preparations measurably contaminated with mitogens.
As shown in Fig. 6 A , mAb to the p l integrin chain completely inhibited collagen gel contraction by fibroblasts, irrespective of whether the contraction experiments were performed in medium containing 0.3% FBS, in medium containing 50 pg/ml XI-NC-3, or in the presence of larger numbers of cells (10 x lo4 cells/ml; data not shown). These results indicate that fibroblasts in serum-free condition express a collagen receptor, possibly a2pl or other p l integrins, and that p l integrins are necessary to transduce intracellular forces to the extracellular matrix. The NC-3 domains have no effect upon gel contraction in the absence of the cell-mediated contractile forces.
Culturing fibroblasts in collagen gels causes the cells to assume an elongated bipolar spindle-like shape. It is known that the process of gel contraction correlates to the elongated shape of fibroblasts (17,241. Morphology of fibroblasts in collagen gels was examined after culture for 24 h. As shown in Fig. 6B, fibroblasts in collagen gel showed an elongated shape in DMEM (serum-free), in DMEM containing XI-NC-3 (50 pg/ ml), and in 0.3% FBSDMEM. However, fibroblasts became spherical in the presence of mAb p l integrins (0.5 pg/ml), suggesting that cell elongation requires integrins. In contrast, polyclonal anti-type XI1 antibodies, which inhibit the gel contraction promoted by XI-NC-3, had no effect upon the elongated shape of fibroblasts in the gel. The results suggest that anti-type XI-NC-3 antibodies do not interfere with the attachment of cells to fibrils.
Neither XII-NC-3 nor XIV-NC-3 promotes fibroblast attachment to plastic or gel-covered substrates. As shown in Fig. 7, neither NC-3 domain has any effect upon fibroblast attachment to a collagen gel. Surprisingly, both domains decreased cell binding to plastic substrates in a concentration-dependent manner. Our interpretation of this result is that the NC-3 do-

Effect of XI-NC-3 on cell proliferation in collagen gels after 4 days of culture
Initial cell number in gel, 5.0 x lo4 celld2 ml. Data represent the mean of triplicate determinations and S.D. mains are not ligands of high affinity for fibroblast cell surface binding proteins and that coating plastic with these domains decreases the surface area of the substrate that is permissive for fibroblast attachment. However, these domains have no effect on the fibroblast attachment to collagen fibrils. We be- lieve that this reflects a low affinity of the NC-3 domains for fibrils relative to their affinity for plastic substrates.
The binding of the NC-3 domains of type XI1 and XIV collagens to monomeric and fibrillar collagen was then evaluated by enzyme-linked immunosorbent assay. No specific binding to type I, 11, or 111 collagen monomers or fibrils was observed (data not shown), confirming previous observations (10). Binding of the NC-3 domains to collagen fibrils was then evaluated. Vitrogen gels were cast in the presence or absence of NC-3 domains and subjected to low speed centrifugation (2,000 x g ) to compact the gels. The supernatant solutions were then qualitatively assayed for the amount of NC-3 domains by Coomassie Brilliant Blue staining and immunoblotting relative to the initial amount added. More than 95% of the NC-3 domains were recovered from the supernatant solutions. It is possible that the remaining 5% of the NC-3 domains bound the fibrils but that concentration of NC-3 has minimal effect upon gel contraction.
The results indicate that the NC-3 domains do not significantly partition to the fibrillar phase, supporting the previous indications that the domains have little affinity for collagen fibrils.
Although type XI1 and XN collagen NC-3 domains appear not to bind the fibril surface, we observed a marked effect upon the deformability of Vitrogen gels. As seen in Fig. &4, NC-3 of both type XI1 and XIV collagen significantly decreased the relative centrifugal force required to compress the gel. Bovine serum albumin, TGF-P1, and fibronectin had no effect. This change in the physical properties of the collagen gel was dependent upon the native structure of the NC-3 domains (Fig.  8B), as it was abolished by heating the domains. Also, anti-type Anti-XI1 antibodies (50 pg/ ml) (A) in control culture (x) had no effect. C, the effect on the deformability of the gel. 50 pg/ml type XI1 and XN collagens (0) were compared with 50 pgfml XI-NC3 (A) and to untreated control (x). In A X , each point represents the mean of triplicate determinations and S.D.
XI1 antibodies neutralized the effect observed with XII-NC-3.
Thus, gel compression is inhibited by the same manipulations that inhibit gel contraction. Intact type XI1 and XN were also evaluated under these conditions (Fig. 8C) with identical findings. The results suggest that contraction promotion by NC-3 domains is not due to cellular events but rather results from modulation of interfibrillar forces. Since collagen gels made from acid-soluble collagen (which retain both telopeptides) have greater stiffness and are more translucent than those made for the same concentration of Vitrogen (pepsin-sohbilized collagen lacking telopeptides) and since acid-soluble collagen fibrils more closely resemble collagen fibrils made in vivo, we evaluated both fibroblast-mediated gel contraction and centrifugation-driven gel compression using acid-soluble collagen gels. The effects of NC-3 domains were qualitatively identical to those seen using Vitrogen gels (data not shown). The amplitude of the changes observed were somewhat less, probably due the increased stiffness of the acidsoluble collagen-derived gel.

DISCUSSION
In these studies, we have evaluated the observation that the NC-3 domains of type XI1 and XIV collagens significantly promote the contraction of collagen gels. We believe that gel contraction is a suitable model system for studying the interaction of mesenchymal cells with a fibrous collagen substrate. We have assumed that the contraction phenomena are a dynamic equilibrium between cell-derived forces originating within the actin bundles and transmitted to the matrix through transmembrane binding proteins and the resistance of the matrix to deformation. We have been unable to demonstrate any direct stable interaction of the NC-3 domains with skin fibroblasts, with monomeric fibrillar collagens, or with collagen fibrils. If centrifugal compression is substituted for the cell-derived forces that drive contraction, the NC-3 domains promote gel compression. The sensitivity of this phenomenon is directly comparable with that of gel contraction with regards to inhibition by antibodies and NC-3 conformation dependence. Our results indicate that the contraction promotion observed can be explained by an effect of the NC-3 domains upon the resistance of a collagen matrix to applied force, and appears not to modulate cell-associated events. The effect is specific to these collagen types since it is inhibited by well characterized antibodies.
The addition of other matrix molecules such as chondroitin sulfate or hyaluronic acid have no effect upon the system (25), while heparin has an unexplained inhibitory effect.
The NC-3 domains do not effect the polymerization of collagen monomers into fibrils, nor do they cause concentration of fibrils by dehydration since the addition of the domains to solutions overlying gels has no effect in the absence of forces generated either physically or biologically. The simplest explanation of the effects of NC-3 addition is that these domains decrease interactions between collagen fibrils. Theoretically, such forces could be minimized if the domains decrease ionic or hydrophobic bonding at the sites on the fibril surface where fibrils intersect within the gel. This could be accomplished if the NC-3 domains compete for the interfibrillar interaction sites, thereby decreasing bonding between adjacent fibrils. It could also occur by an increase in the effective fibril diameter since in a fibril solution with constant collagen concentration, an increase in diameter results in a decrease in fibril surface area and number of interaction sites among fibrils. Interfibrillar interactions would be predicted to vary directly with fibril surface area and number of the sites.
While we believe that this gel contraction system is a valid model for the evaluation of the interaction of fibroblasts with collagenous substrata, its relationship to physiological events is less clear. Fibroblasts placed in a fibrous but acellular environment of the vitreous humor of the eye will cause contraction of that gel, similar to what is observed in vitro (26). Therefore fibroblasts can recognize and reorganize collagen fibers in vivo despite the increased complexity of the biological matrix. The ability of type XI1 and XN NC-3 domains to modulate interfibrillar interaction of these more complex fibrillar systems is not known. However, the dramatic in vitro results strongly support the hypothesis that types XI1 and XN mediate biomechanical aspects of the matrix.
The results shown here are consistent with the possibility that both type XI1 and X N collagens participate in extracellular matrix deformability. By decreasing interfibrillar interactions, the matrix could become locally more pliable in the absence of cellular stress, or could become progressively more rigid if the NC-3 domains facilitate cell-mediated alignment and concentration of banded collagen fibrils. This possibility might resolve the apparently contradictory finding that types XI1 and XIV are associated with both large and small-diameter collagen fibrils (9,2?, 281. The model we propose assumes that the outcome of the association of types XI1 and XIV with the banded fibrils is dependent upon the local cell activity. There i s some experimental support for this concept. Schafer et al. (29) observed that fibroblasts derived from the papillary dermis were less able to contract collagen gels than fibroblasts from the reticular dermis, In addition, mixing papillary cells with reticular cells inhibited gel contraction. These data suggest that the greater fibril density of the reticular dermis may be due to the increased activity of the reticular cells. Thus, it is possible that in concert with appropriate local cellular activity, types XI1 and XN could mediate increased or decreased matrix deformability.
These studies show no differences between the effect of type XI1 and of type X W . This is surprising since the two molecules segregate to nearly exclusive tissue regions in vivo (9). It is possible that the effects of either XI1 or XIV collagens could be modulated differentially in vivo by the presence of the larger transcript in the case of type XI1 or by the addition of glycosaminoglycan.