Collagen Unfolding Accelerates Water Influx, Determining Hydration in the Interstitial Matrix

Abstract

In the interstitial matrix, collagen unfolding at physiologic temperatures is thought to facilitate interactions with enzymes and scaffold molecules during inflammation, tissue remodeling, and wound healing. We tested the hypothesis that it also plays a role in modulating flows and matrix hydration potential. After progressively unfolding dermal collagen in situ, we measured the hydration parameters by osmotic stress techniques and modeled them as linear functions of unfolded collagen, quantified by differential scanning calorimetry after timed heat treatment. Consistent with the hypothetical model, the thermodynamic and flow parameters obtained experimentally were related linearly to the unfolded collagen fraction. The increases in relative humidity and intensity of T₂ maps were also consistent with interfacial energy contributions to the hydration potential and the hydrophobic character of the newly formed protein/water interfaces. As a plausible explanation, we propose that increased tension at interfaces formed during collagen unfolding generate local gradients in the matrix that accelerate water transfer in the dermis. This mechanism adds a convective component to interstitial transfer of biological fluids that, unlike diffusion, can speed the dispersion of water and large solutes within the matrix.

Introduction

This study focused on dermal collagen unfolding in situ to gain a better understanding of interstitial fluid transfer mechanisms. Collagen organizes and supports the extracellular matrix of blood vessels and organs, contributing ∼6% of total body weight. In the dermis, collagen bundles form a three-dimensional mesh with other fibrilar proteins and interstitial cells, including a network of contractile fibroblasts (1–3). The structure is cemented by a hyaluronic acid and proteoglycan gel that allows for viscous flux and mediates interchange of water nutrients and metabolites between the blood, the lymph, the interstitial cell network, and the epidermal layers (4–8).

The collagen molecule is composed of three polypeptide chains that wind around each other to form a 1.5-nm-thick triple helix ∼300-nm long. Each chain’s primary amino acid sequence includes both polar and apolar residues grouped in alternating hydrophilic and hydrophobic patches (9). Collagen molecules assemble into supertwisted microfibrils that interdigitate with neighboring microfibrils packing into quasihexagonal supramolecular arrangements; the individual molecules in the arrangements are cross-linked covalently within and between microfibrils (10,11). The microfibril is considered the basic structural unit in collagen’s higher order structures, fibrils, fibers, and bundles; these structures are further stabilized in situ by other supramolecular interactions and cell attachments mediated primarily by proteoglycans, fibronectins, and other connective tissue proteins (1,5,12–14). As in other proteins, structural modeling shows that many of collagen’s nonpolar residues are largely sequestered from solvent in the folded triple helix structure (11,13); unfolding in an aqueous environment exposes apolar groups to water (13,15). At physiologic temperatures, collagen’s triple helix structure fluctuates between folded and partially unfolded states (11,14–21); locally limited unfolding of collagen chains apparently facilitates specific molecular contacts with enzymes and scaffold molecules during inflammation, tissue remodeling, and wound healing (11,17–20).

We considered the possibility that collagen unfolding also plays a role in local control of interstitial fluid transfer. Pressure differences between the blood capillaries and the extracellular matrix drive flux across interstitial spaces; colloidosmosis and tension, generated and regulated by glycosaminoglycans and interstitial contractile cells, respectively, are believed to control matrix hydration (22–25). However, we found additional interstitial flow potential emerging in the dermis after explants were heat killed, and the collagen denatured (25), suggesting that collagen unfolding contributes to hydration.

To test this hypothesis, we modeled hydration potential, specific water adsorption, and adsorption rates as linear functions of the extent of collagen unfolding. Influx/efflux time distributions for live and heat-killed dermal explants were compared to identify changes associated with cell action and collagen unfolding. We also explored the participation of hydrophobic collagen surface by vapor pressure measurements and magnetic resonance imaging. This report includes 1), experimental methods to evaluate thermodynamic and dynamic parameters of interstitial water as a function of unfolded collagen in situ; 2), a thermodynamic and mechanistic framework for interstitial fluid transfer; and 3), experimental results, analyses, and discussion of the equilibrium parameters and interstitial flow dynamics related to the proposed mechanism.

Materials and Methods

Dermal samples and collagen unfolding in situ

Full-thickness pig skin samples were collected immediately after euthanasia as by-products of unrelated studies, following protocols approved by the institutional Animal Care and Use Committee. Samples were trimmed from subcutaneous fat and fascia and divided into portions of ∼25 cm², minimizing air exposure to prevent water loss by evaporation. To induce incremental collagen unfolding, replicated samples were tightly wrapped in aluminum foil heated to 60°C and maintained at that temperature for 0, 5, 10, 20, 50, and 90 min intervals. After heating, samples were equilibrated at room temperature and stored at 4°C. Control reference samples were maintained at room temperature during the heating protocol but otherwise processed and stored the same way. Additional reference samples were heated to 40–42°C to reversibly denature the glycosaminoglycan gel (23,26) without cell death, and other samples were heated to 50–52°C to kill the cells without unfolding collagen (25,27). Changes in water content and size were monitored by weighing and measuring skin samples before and after heating using a precision balance and calipers, respectively. Water content did not change significantly during heating, but the dimensions changed: thickness increased, whereas length and width decreased (25).

DSC

We measured the amount of folded collagen remaining in each sample after heating by DSC (28,29) using a Q200 calorimeter (TA Instruments-Waters, New Castle, DE). Representative subsamples, 3 mm in diameter, were punched out from each sample set, hermetically sealed in aluminum pans, and scanned from 20 to 100°C at a rate of 5°C/min. Folded collagen was determined from the heat absorbed (cal/g) during the helix-to-coil transition by integrating the tracings between 50 and 90°C. The fraction of irreversible unfolded collagen in each heated sample was expressed relative to native collagen in its paired nonheated control.

Osmotic stress technique

Hydration potentials and flow rates of water in/out of the explants were determined by an osmotic stress technique, as before (30,31). Briefly, the wrapped samples with a known proportion of folded to unfolded collagen were equilibrated at room temperature, unwrapped, and rapidly sectioned into ∼200-mg subsamples, which were then immersed in a bath with a 20-fold excess volume of culture medium (DMEM/F12, Invitrogen, diluted to 1/2 strength with phosphate-buffered physiologic saline solution, including 5% fetal calf serum, the antibiotics streptomycin and penicillin, and 15 mM HEPES buffer pH 7.2) at six colloidosmotic pressure levels, ranging from 3 to 219 mmHg, and two temperatures, 4 and 37°C. Colloidosmotic pressure was fixed with polyethylene glycol 8,000, and the volume of water adsorbed or desorbed from the explants was determined from the change in sample weight as a function of time. Maximal volumes of water adsorbed or released were determined from logistic dose-response equations fitted to volume change versus time data points (31).

Vapor pressure determination

Water vapor pressure at equilibrium in the dermal samples was measured by the chilled mirror technique using an Aqualab 3T water-activity meter (Decagon Devices, Pullman, WA). Samples were sealed in the measuring chamber that contains the mirror where the temperature is controlled by a thermoelectric cooler, and condensation detected by changes in reflectance. At equilibrium, the vapor pressure of the air in the sealed chamber is the same as in the water confined in the dermal matrix. Before each measurement, samples were progressively dehydrated by exposure to room air, and measurements repeated at 6–12 different hydration levels. Plots of the vapor pressure, vp, versus (1-hydration) were fitted with a generic power equation, and the 0.75 hydration level used for comparisons among explants with different proportions of unfolded collagen. This method detects vp value differences of <0.025.

Magnetic resonance imaging

Changes in the samples’ spin-spin relaxation times, T₂, were inferred from changes in the intensity of magnetic resonance images obtained using a CPMG (Carr-Purcell-Meiboom-Gill) multiecho protocol, as before (25,32). Complete sample sets were imaged in a 7T scanner (Bruker Biospin, Ettlinger, Germany); using a shim insert that produced a 1,000 mT/m gradient, and a quadrature volume radio frequency coil with an inside diameter of 35 mm for signal transmission and reception. Repetition and echo times were 2,000 and 8.5 ms, respectively, and the number of excitations 72. The scan time was 15 h 20 min—eight echoes were acquired—and the field of view with 512 × 512 pixels covered 2.8 cm at 55 μm/pixel in-plane resolutions. Six explants, with different proportions of unfolded collagen, from each animal were scanned together. On the basis of previous studies (15), T₁ was assumed not to change with collagen denaturation. T₂ values in the interfollicular areas of the explants were determined by fitting a single exponential to the mean intensities in the first eight echoes of the CPMG protocol.

Sample sets, curve fitting, and statistical analyses

Osmotic stress experiments included four to six sets of explants per animal with different amounts of unfolded collagen. Twelve subsets per set were equilibrated in baths at two temperatures and six colloidosmotic pressures, and influx and efflux trajectories constructed using six pairs of volume/time data points for each pressure. These data were fitted with logistic dose-response curves to determine equilibrium volume change, ΔN^e, at each level of colloidosmotic pressure in the bath. Hydration potentials were determined from linear plots of ΔN^e versus bath colloidosmotic pressure. Initial rates of in and out flow were determined from the first derivative of the fitted trajectories taken at a time interval 0–5 min. Osmotic stress and flow kinetic experiments were reproduced in dermal samples from 10 animals. Magnetic resonance imaging and vapor pressure measurements included six complete sample sets from three animals. Vapor pressure values at 75% hydration were determined by fitting vapor pressure versus hydration with a simple power equation.

The statistical significance of differences among and between sample sets was determined by analyses of variance and post hoc multiple Fisher’s protected least-significance difference (PLSD) and paired t-tests. Model-equation selection, curve-fitting and regression analyses, descriptive statistics, analysis of variance (ANOVA), and t-tests were executed using the commercially available software packages Table Curve and Stat View (SAS Institute, Cary, NC). In all comparisons, differences with a p-value < 0.05 were considered statistically significant. Coefficients of simple determination, r² were used to measure the proportion of the variation of the dependent variables (hydration potential, entropy, work of transfer, and adsorption) that is determined by a straight line relationship with the independent, experimentally controlled variables (collagen unfolding and bath water potential).

Theoretical Considerations and Experimental Rationale

Physiologic background and phenomenological description

Transcapillary flux in vivo is driven by pressure differences between the blood capillaries and the interstitium. The work of fluid exchange is ultimately generated by the heart’s contractions and local mechanisms that adapt perfusion to tissue demands. Whether and how the interstitial matrix participates in these local mechanisms remain open questions in physiology.

Flux/pressure relationships in vivo are generally rationalized according to Starling’s hypothesis (33), described by phenomenological equations of the following type:

$$F=K\left[\left(Pc\text{-}Pi\right)-\left(Cc\text{-}Ci\right)\right],$$ (1)

where the steady-state flow rate, F, to and from the interstitium and capillaries is determined by differences in the effective hydrostatic (P) and colloidosmotic (C) pressures between the capillaries and interstitial matrix, designated by subscripts c and i, respectively, in Eq. 1. Although the capillary components Pc and Cc are fairly well understood physiologically, the origin, magnitude, and regulation of the flow-driving forces of the local interstitial components Pi and Ci remain uncertain (6–8). The proportionality constant K reflects all the factors that influence fluid conductivity, including the material properties of the interstitial matrix, viscosity of the transferred biological solution, permeability of the blood and lymphatic capillaries, and all the other material and structural properties that contribute to flow resistance in the dermis. Because this resistance is regulated by local and systemic controls, in vivo studies cannot easily separate capillary from interstitial contributions.

Local thermodynamic forces and the kinetics governing fluid transfer across the interstitial matrix can be studied ex vivo using live tissue explant cultures (22,25,31). Whereas in vivo, local blood and the lymphatic capillary network provide the surface for fluid exchange with the systemic circulation, ex vivo, fluid crosses the dissected surfaces of the explants to exchange with the bath. Experimental measurements and analyses are much simpler than those in vivo; the capillary components of Eq. 1 can be replaced by the bath water chemical potential, μ_w, a single and independently adjustable variable.

When explants are dissected, the interstitial matrix is released from the constraints that maintained steady-state fluid balance. The net potential resulting from the now unbalanced interstitial forces can be measured as a hydration potential, HP, using osmotic stress techniques (25,30,31). We were interested in changes in this potential as a function of the unfolded collagen.

Thermodynamic description

The interstitial aqueous fluid in the dermal matrix is confined by a complex system of nanometer-scale pores and slits defined by a supramolecular structure of collagen and glycosaminoglycans (1,2,10,11,34). We assume that this structure is constituted of adjacent functional units, each with water/protein interfaces, water/glycosaminoglycan interfaces, and contractile cells. Water transfer in/out of the matrix in ex vivo cultures can then be imagined as an idealized thermodynamic system, where the structural/functional complexities of the real system are represented by an effective value of the implied geometric and thermodynamic quantities.

Departing from a simple model for confined fluid (35,36), we describe the changes in the idealized explant culture in terms of the differential of a grand potential:

$$d\Omega =\text{-}pdV\text{-}SdT\text{-}Nd{\mu }_w+\gamma d{A}_t-A_{t}fdL,$$ (2)

where p is the bath pressure, and V, the volume (both remain constant in our system); S is the total entropy; N, the total number of water molecules in the system, also constant; μ_w, the average free energy of water molecules; γ, the effective interfacial tension; A_t, the total interfacial area; L, the distance separating the interfaces; and f, an effective force acting between the interfaces that includes all force components normal to the interfaces that decrease water adsorption by closing interfacial gaps. Considering the differential of the grand potential in another system—the culture bath alone, with the same volume and chemical potential but no explants—

$$d{\Omega }^b=\text{-}pdV-S^{b}dT\text{-}{N}^{b}d{\mu }_w,$$ (3)

the experimentally accessible interstitial excess functions may then be defined as

$${\Omega }^e=\Omega –{\Omega }^b;S^e=S-S^{b\equiv A_{t}s};N^e=N-N^b\equiv A_t\Gamma ,$$

where superscripts b and e denote bath and excess functions, respectively; s is the interfacial entropy, and $\Gamma$, the water adsorbed per unit of interface; that is, the specific adsorption. The change in the excess potential is then derived from Eqs. 2 and 3:

$$d{\Omega }^e=-\left(A_{t}s\right)dT-\left(A_t\Gamma \right)d{\mu }_w+\gamma d{A}_t–\left(A_{t}f\right)dL,$$ (4)

and for interfaces with fixed Γ and γ, ${\Omega }^{\mathit{e}}=\gamma A_t$.

Discussion on expected results is in the Supporting Material.

Experimental Results and Discussion

Accelerated water influx coincides with cell death and collagen unfolding

To explore the role of the collagen matrix supramolecular structure and identify the relative contribution of cells and glycosaminoglycan arrangements to water transfer across the interstices of the dermis, we analyzed the distribution of flow velocities in explants either not heated or heated to a nonlethal temperature, 40–42°C (groups A and B, Fig. 1, b and c), or 50–52°C and 60–62°C (groups C and D, Fig. 1, b and c), to kill cells and unfold collagen, respectively. In all groups, the fluid influx/efflux velocity, V_f, decreased with time and for intervals <200 min, generic equations of the type dN^e/dt = V_f = a +b $\surd t$ fitted data points generally, with coefficients of determination, r² > 0.9 suggesting diffusive and/or Washburn-like dynamics (37–39).

Figure 1.

Fluid influx/efflux to/from live and heat-killed dermal explants. (A) Fluid transfer measured gravimetrically as a function of time in live (blue, unheated control group A) and heat-killed (red, heated at 60°C for 30 min, group D) dermal explants. They were immersed in physiologic solutions with a colloidosmotic pressure of either 3 mmHg (influx, left panel) or 219 (efflux, right panel). Trajectories were derived from logistic dose-response curves fitted to six volume/time data points, each normalized relative to maximal volume change and distributed in 20 equal intervals. (B) Mean influx and efflux times were determined from their distribution; the stars indicate statistically significant differences between heat-killed explants (group C, 50–52°C; and D, 60–62°C) and live controls (group A, nonheated; and B, heated to 40–42°C). Arrows point to the slowest and fastest mean times. Bars indicate 95% confidence limits (group size, n = 9 sets). (C) The proportion of unfolded collagen was determined as 1-χ, where χ is the ratio between folded collagen in paired heated and control samples. Bars indicate 95% confidence limits (n = 5 sets); differences are only significant in explants from group D, showing that heating to 60°C was required for irreversible collagen unfolding.

However, logistic dose-response curves that fit the data better, generally giving r² > 0.98, were used to compare the mean and distributions of influx/efflux times in the various experimental groups. Mean influx times in group B explants heated to 42°C, a temperature shown to induce structural transitions in hyaluronan gels (23,26), did not differ from those in undisturbed controls. In contrast, in explants with irreversibly unfolded collagen, the mean influx times were significantly shorter than in all other groups, including group C, where the cells were dead. Trajectories normalized by the maximal volume change are illustrated for groups A and D, Fig. 1, a.

These results complement previous studies showing the increased hydration potential in heat-killed explants (25).Although the increased mean influx times in groups C and D relative to live explants can be explained by loss of cell action, the relative acceleration measured in group D explants suggest that other mechanisms emerge when collagen unfolds. These findings also indicate that disturbing supramolecular interaction by reversibly denaturing the glycosaminoglycan gel does not significantly change flow velocity distribution relative to undisturbed controls.

We conclude that whereas at the supramolecular level, the spatial distribution of matrix glycosaminoglycans may change on heating, their polar interactions with water and, consequently, the colloidosmotic potential of their gel-sol phases do not change. Thus, the surface energy contributed by hydrophilic glycosaminoglycan interfaces can be considered approximately constant, and changes measured after heating ascribed to the new interfaces formed during collagen unfolding.

Hydration potential increases are determined by the extent of collagen unfolding

To determine whether the observed inflow change in group D samples is related to the extent of new surface exposed to water when collagen unfolds, equilibrium and flow parameters were measured in heat-killed explants where the extent of unfolding was controlled. Sets of explants with a known fraction of unfolded collagen were prepared by heating the dermis at 60°C for intervals ranging from 0 to 90 min. The amount of collagen per dry weight in each sample was determined by DSC and expressed as a fraction, $\chi ,$ of the amount in nonheated paired controls; the extent of unfolded collagen surface A_h was calculated as A_h = $\left(\mathbf{1}\mathbf{-}\mathit{\chi }\right),$ assuming that none of the collagen in controls was irreversibly unfolded.

Fig. 2 shows the approach we used to determine hydration potential from volume/pressure relationships in sets of explants with a known fraction of unfolded collagen. At each level of unfolded collagen, the relationship between the volume transferred and the bath’s colloidosmotic pressure was linear. The bath potential at which no net volume transfer was detected corresponds to the excess potential in the explants, Ω^e, which we termed HP (25), and the volume transferred to reach equilibrium represents the excess water, $N^e=N-N^b=A_t\Gamma$.

Figure 2.

Adsorbed volume in dermal explants as a function of osmotic pressure: effect of increasing the fraction of unfolded collagen. (A) Dermal explants were immersed in bath solution at 37°C, and the volume of water they adsorbed or released, ΔN^e, over time was measured gravimetrically until near equilibration. Equilibrium volume change, N^e, was determined by fitting logistic dose-response curves to volume/time data points. The bath potential,$\Omega$^b, was controlled with polyethylene glycol 8,000 added at six concentrations within the indicated range. The equation (inset) predicts straight lines for plots of N^e versus pressure. The pressure at which each line intercepts the level, N^e = 0, corresponds to the hydration potential, HP =$\Omega$^e per unit volume. Results shown are for four sets of explants with unfolded collagen fractions of 0 (●), 0.42 (○), 0.76 (▴), and 0.93 (▵). The slope, m, is the experimentally determined interstitial compliance reflecting the material properties of the matrix. (B) Representative tracings for explants set in panel A. Explant samples with different proportions of unfolded collagen were prepared by heating replicate samples at 60°C for various times and reequilibrating at 4°C and at room temperature before determining the proportion of unfolded collagen by DSC. The fraction of native collagen in preheated samples, H, was determined by integrating the collagen melting peak, normalized per gram of dry tissue, and expressed relative to collagen in paired, nonheated controls, χ = (cal/mol.g heated)/(cal/mol.g controls). (C) Kinetics of collagen unfolding at 60°C. Explants were maintained at 60°C for 0–90 min intervals and scanned at 5°C/min. The figure shows the fraction of folded collagen relative to unheated controls as a function of heating time for the same explants in panels A and B. The proportion of folded collagen was determined from the integrated DSC tracings (see B). (D) Distribution of unfolded collagen in 56 heated explant samples from 10 donors. Mean and standard error values for samples falling below each quintile are indicated. The mean for the first quintile is zero (per hypothesis) and includes the paired controls to each heated sample.

Plots of volume adsorbed, $N^e,$ versus unfolded collagen (1-χ) at different μ_w levels yielded y-intercepts that increased markedly with μ_w as expected for adsorption by hydrophilic surfaces and positive slopes that change little with μ_w, indicating that in regard to their interaction with water, interfaces emerging as collagen irreversibly unfolds differ qualitatively from preexisting interfaces (Table 1).

Table 1.

Changes in adsorption as a function of unfolded collagen, determined at different levels of water chemical potential

Δμw/v; (mmHg)	$N_l^e\left(\mathit{\mu }\text{l}.A_l\right)$	Γh (μl/Ah)	r2
219	−445 ± 60	316 ± 92	0.854
170	−347 ± 59	393 ± 90	0.905
135	−210 ± 58	352 ± 88	0.888
84	−89 ± 48	317 ± 74	0.902
35	−20 ± 49	356 ± 75	0.918
3	85 ± 69	299 ±106	0.801

The values of $N_l^e={\Gamma }_l$A_l and $N_h^e=$ Γ_hA_h were determined from the y-intercept and slope, respectively, of linear fits where the explants’ adsorption, $N^e=\left(N_l^e+N_h^e\right)$, was plotted as a function of the extent of unfolded collagen (1-χ) = A_h. Water chemical potential per unit volume, Δμ_w/v of explant culture was decreased by adding PEG8000 to the bath at concentrations 0–12% resulting in colloidosmotic pressures ranging from 3 to 219 mmHg. Coefficient of determination r² obtained for each (Δμ_w/v) level is indicated in the far right column and indicate the strong linear dependency of the N^e on (1-χ). The y-intercepts $N_l^e$ but not the slopes Γ_h increased significantly with μ_w.

A_l and A_h are area units representing the water-adsorbing interfaces in 1 g of tissue where the collagen is folded and the added adsorbing surface exposed on unfolding, respectively.

Hydration potential values also followed those of unfolded collagen linearly (giving least-square regression lines with r² > 0.92; Fig. 3 A), indicating that unwinding the triple helix exposes surfaces that contribute proportionally to the explants’ inflow potential, consistent with a casual relationship. The potentials, Ω = Nμ_w (calculated by integrating transferred water-volumes times the pressure, between the experimental pressure bounds 3 and 219 mmHg Fig. 3 B), and Ω^b (estimated as the product of the total volume 0.527 ± 0.030 ml by the net pressure, 216 mmHg) differed by amounts that were within error of HP values (in mmHg.ml) consistent with the definition of HP as the excess potential, Eqs. 2–4.

Figure 3.

Dermis hydration potential and entropy changes as a function of unfolded collagen. (A) The hydration potentials of explants at 37°C, plotted as a function of collagen unfolding. The linear regression line, coefficient of determination r² = 0.92 (P-value = 0.0098), was generated from the mean of values below the 2nd, 3rd, 4th, and 5th quintiles. (B) Changes in the work of water transfer, μ_wN, determined by integrating the product of transferred volume by the bath pressure over the 3 to 219 mmHg range. The linear regression line r² = 0.87(P-value = 0.0206) was generated from the mean of values below the 2nd, 3rd, 4th, and 5th quintiles. (C) Entropy changes, estimated from the difference between μ_wN values measured at 37 and 4°C, ΔT = 33°, is plotted as a function of irreversibly unfolded collagen. The relationship was also linear with r² = 0.97 (P-value = 0.0026).

If we consider the potential generated from the stable and the added unfolded components separately, ${\Omega }_t^e={\gamma }_l{A}_l+{\gamma }_h{A}_h$, the y-intercept, ∼63 mmHg A_l and slope, ∼132 mmHg/$A_h$, reflect the surface energy of preexisting interfaces and added interfaces exposed as collagen unfolds, respectively. Conservatively assuming that the surface units A_h and A_l have similar values (in nm², for example), it follows that ${\gamma }_h>{\gamma }_l$. The inequality in effective surface tension of contiguous interfaces is expected to induce flows (38–43) consistent with the increase in water adsorbed by explants with unfolded collagen. The mean potential change measured in killed explants without unfolded collagen (incubated at 52°C for 30 min) interpolated in the regression line to A_h ∼0.15, suggesting that the added potential in these samples may develop from transiently microunfolded (14) collagen.

We also expected to find changes in entropic contribution to the potential measured in explants with unfolded collagen. To examine this possibility, we measured μ_wN at 4 and 37°C, ΔT = 33°C, in paired sample sets and used the temperature-dependent difference as an estimate of the change in surface entropy with collagen unfolding. We found a linear relationship between the extent of unfolded collagen, (1-χ), and entropy in killed explants (Fig. 3 C); The value for killed controls with no irreversibly unfolded collagen, extrapolates to A_h ∼ 0.21.

We speculate that the cells regulate fluid influx by directly controlling local collagen assemblies but need more information on how interstitial fibroblasts and myofibroblasts participate. The structure, dynamics, and distribution of cell adhesion and other functional interactions between integrins, fibronectins. and supramolecular collagen arrangements in situ are not completely understood, particularly as they relate to the control of unfolding. Unfolding (11,17,19,20) is related to remodeling by metalloproteinases, several aspects of which are known to be regulated by the cells (5).

Additional evidence that interfacial energy contributes to matrix hydration potential

Magnetic resonance imaging was used to explore local changes in interfacial water when collagen unfolds. Fewer hydrogen bonds with interfaces more densely hydrophobic should prolong the time- and space-averaged spin-spin relaxation times of water protons and increase image intensity (44,45). In contrast, polar interactions at hydrophilic interfaces should shorten relaxation times and decrease image intensity. Because total water content remains constant, as the proportion of unfolded collagen increases so will the proportion of water molecules exposed to aliphatic residues (15). Fig. 4 A shows differences in the T₂-map intensity of control explants and explants with unfolded collagen, which, as predicted, increases with the extent of collagen unfolding (45,46).

Figure 4.

Magnetic resonance imaging and vapor pressure change with unfolded collagen. (A) Magnetic resonance T₂ map of one set of replicated dermal samples in which collagen is unfolded to different extents by heating at 60°C for 0, 5, 10, 20, 60, and 90 min (arranged from top to bottom). In-plane resolution was ∼55 μm, and the bar represents ∼1 mm. (B) T₂ values as a function of unfolded collagen. Data points from three independent sets linearly relate 18 paired measurements of collagen unfolding and T₂ values. Coefficient of determination r² = 0.715 and P-value < 0.0001. The T₂ values in the interfollicular areas were determined by fitting a single exponential to the mean intensities in the first eight echoes of a CPMG protocol. (C) Sample vapor pressure was determined using the chilled mirror technique in the same sets shown in B. The graph shows the mean changes in heated samples with unfolded collagen relative to nonheated controls. For each sample, 8–10 measurements were made at various dehydration levels, and data points were fitted with a generic power equation, P = A+ BC^X, where X is the dehydration level. Values at 0.75 hydration, obtained from the fitted curve by interpolation were used to calculate the vp changes. The difference between control and heated samples was statistically significant (P-value <0.004 ANOVA and Fisher’s PLSD post hoc test) but not the differences among samples with different fractions of unfolded collagen.

Similarly, the vapor pressure of explants should decrease if the excess potential results from increased hydrophilic solute and/or interfaces that lower the water’s chemical potential, although, in general, hydrophobic interfaces in pores of small radius should increase vapor pressure. Fig. 4 B shows a significant increase in vapor pressure in all groups of explants with unfolded collagen relative to control groups. Differences among groups in which collagen was more or less unfolded were not statistically significant, which may reflect sample variability and/or that the effective radius of the fluid paths in the matrix increases as a larger proportion of the collagen unfolds, with the vapor pressure reaching a plateau (47–50). The linear relationship predicted by the Kelvin equation is not expected to hold at the nanometer scale, particularly at hydrophobic interfaces. Theoretical models predict fluctuations that may smear the vapor pressure values measured at macroscopic scales (51). In any case, considered with the magnetic resonance images, these results support the hypothesis that the excess potential driving fluid transfer into the matrix is related to interfacial phenomena, consistent with the exposure of collagen hydrophobic surfaces to solvent as they unfold.

Decrease in interstitial flow resistance with extent of unfolded collagen

Osmotic stress data also allow some general analysis of water transfer dynamics. Flow velocity, V_f, in/out of the dermal explants should be directly proportional to the pressure gradient and inversely proportional to frictional resistance, 1/R_m, of the matrix to the flow: V_f = dN^e/dt = 1/R_m ΔP. The initial driving pressure, t∼0, is the difference between the pressure in the bath, P^b, and the net pressure in the tissue, P^T. By comparison, net flow rate in explants with more or less unfolded collagen can be computed using paired samples by subtracting influx from efflux rates at bath pressures below and above the hydration potential; that is, bath pressures of 3 and 219 mmHg,

$$\text{Influx}=\frac{1}{R_m}\left(P_3^b\text{-}{P}^T\right);\text{Efflux}=\frac{1}{R_m}\left(P_{219}^b\text{-}{P}^T\right);\text{Influx}\text{-}\text{Efflux}=\frac{1}{R_m}\left(P_3^b\text{-}{P}_{219}^b\right).$$ (5)

At the same driving pressure, ΔP = 216 mmHg, differences in the initial slope, 1/R_m, among sample sets with collagen unfolded to different extents represent differences in flow resistance (Table 2). We found a marked proportional decrease in flow resistance with collagen unfolding, and the controls value, 0.9 μl/g/min extrapolated in the regression line corresponds, to an unfolded collagen fraction of ∼0.16, suggesting that reversible collagen unfolding under physiologic conditions contributes to interstitial flux (Fig. 5).

Table 2.

Initial net-flow rate in dermal explants increases with unfolded collagen

Unfolded collagen		Initial flow rate μl/g.min
(1-χ)	(n)	INFLUX (At 3 mmHg)	EFFLUX (At 219 mmHg)	NET (Δ216 mmHg)
0	(11)	3.3 ± 1.4	4.2 ± 1.1	−0.9 ± 0.7
0.312 ± 0.020	(7)	4.3 ± 0.7	3.4 ± 0.3	0.9 ± 0.7
0.513 ± 0.022	(8)	4.1 ± 0.6	2.7 ± 0.7	1.4 ± 0.7
0.709 ± 0.013	(16)	4.6 ± 0.2	2.0 ±0.05	2.6 ± 0.2
0.911 ± 0.017	(14)	7.0 ± 0.2	1.2 ± 0.01	5.6 ± 0.2

Collagen in each sample was determined by DSC and normalized as a fraction, χ, relative to collagen in paired, nonheated controls, and distributed in quintiles. None of the heated controls is included in the 1st quintile.

Initial influx and efflux rates were determined at t = 5 min from the first derivative of logistic dose-response curves fitted to data points from paired explants in cultures with colloidosmotic pressure fixed at either 3 or 219 mmHg, respectively. Net flow rate for a ΔP = 216 mmHg was calculated as the difference between initial inflow and outflow rates. Rate changes with unfolded collagen reflect inversely proportional changes in flow resistance.

Data are from 56 explants samples, and values are mean and standard errors for the indicated number of sample pairs (n) below each quintile.

Figure 5.

Flow resistance as a function of collagen unfolding. (A) Net flux trajectories were determined as the difference between in- and outflow volumes, measured at 3 and 219 mmHg, respectively, for the same sample sets used in Fig. 2A. (B) Rates calculated from the first derivative of the progression curves were plotted versus the extent of unfolded collagen. A straight line fitted data with a coefficient of determination r² = 0.87 and slope = 7.9 μl g⁻¹ min⁻¹/A_h. Data are from 56 paired sets from 10 donors distributed into quintiles according to the unfolded collagen fraction in the explants; points in the graph are the means for each quintile.

Because the dermal matrix structure is dynamic and very complex across spatial scales, we cannot stipulate the physicochemical changes that determine the observed changes in frictional resistance with collagen unfolding. Several possibilities include the increases in a), hydrophobicity of the interfaces that can lubricate the flow (49); b), the radius and homogeneity of fluid paths, with larger radii and straight, simple paths decreasing hydrodynamic dispersion and frictional losses; and c), formation of preferential paths with a less viscous gel phase (50). These possibilities are not mutually exclusive and are also compatible with conclusions from the equilibrium parameters implicating forces rising from out-of-balance interfaces. It also suggests the possibility that in vivo, these suction forces may connect with lymphatic drainage mechanisms; unlike osmotic diffusion, the surface-tension mediated flows are convective flows that could transfer large molecules rapidly.

Conclusions

Collagen unfolding in the interstitial matrix of the dermis determines local increases in interfacial energy. The resulting higher hydration potential and accelerated fluid influx can be explained by emerging surface-tension gradients and decreased resistance to local flow, both consistent with a higher density of hydrophobic residues on solvent-accessible surfaces. This conclusion is based on interpretation of the experimental results in the context of available information on in situ collagen structure (10,11) and theories about hydrophobic interfaces (40,43).

Limited unfolding of the collagen triple helix in situ does not change the fibrils’ usual appearance in electron micrographs (47,52). Labile unfolded regions codistributes with more stable ones, and the fibril cores are more labile than the peripheries. This remarkable higher order stability can be explained by microfibril structure (10,11): individual collagen molecules interdigitate and form cross-links between and within the microfibrils. Local unfolding transitions from triple helix to random coil expose new, nanoscale surfaces dense with nonpolar residues (13,15) that increase the hydrophobicity of water/collagen interfaces; theory and simulation predict that these type interfaces have properties similar to those of interfaces between aqueous solutions and their vapors, including high surface tension (38–40,42,43). Experimentally, the measured increases in T₂ and equilibrium vapor pressure further indicate newly formed hydrophobic collagen/water interfaces (15,45,46,51). Interpreted as a whole, our data supports the idea that folding transitions of the collagen triple helix in situ together with the colloidosmotic forces of glycosaminoglycans gels (2,12) and the action of interstitial cells (4,22) contribute to the local regulation of interstitial flows.

According to this view, under physiologic conditions, the proportion of unfolded collagen at any given time is a local, microscopic functional property. Transient microunfolding events aid the steady-state renewal and drainage of interstitial matrix fluid, whereas subclinical injury repair may involve more sustained and extensive unfolding. Pathologic versions of this mechanism will occur either when 1), a large fraction of collagen is irreversibly unfolded, or 2), microunfolding is deficient. In the first case, fluid will flow into the connective tissues until opposing forces—for example, a general buildup of hydrostatic pressure at tissue scale—prevent it. Such conditions are seen clinically in burns and compartment syndromes with extensive acute interstitial damage. The second case may include more insidious deficiencies; for example, increased covalent cross-linkage of collagen may limit matrix hydration potential and, over time, bias some flux toward lower pressure subdermal regions, contributing to the lax appearance of old skin.