Interactive effect of chondroitin sulphate C and hyaluronan on fluid movement across rabbit synovium

Abstract

The polysaccharide hyaluronan (HA) conserves synovial fluid by keeping outflow low and almost constant over a wide pressure range (‘buffering’), but only at concentrations associated with polymer domain overlap. We therefore tested whether polymer interactions can cause buffering, using HA-chondroitin sulphate C (CSC) mixtures. Also, since it has been found that capillary filtration is insensitive to the Starling force interstitial osmotic pressure in frog mesenteries, this was assessed in synovium. Hyaluronan at non-buffering concentrations (0.50–0.75 mg ml⁻¹) and/or 25 mg ml⁻¹ CSC (osmotic pressure 68 cmH₂O) was infused into knees of anaesthetised rabbits in vivo. Viscometry and chromatography confirmed that HA interacts with CSC. Pressure (P_j) versus trans-synovial flow (Q̇_s) relations were measured. Q̇_s was outwards for HA alone (1.2 ± 0.9 μl min⁻¹ at 3 cmH₂O, mean ± s.e.m.; n = 6). CSC diffused into synovium and changed Q̇_s to filtration at low P_j (−4.1 μl min⁻¹, 3 cmH₂O, n = 5, P < 0.02, t test). Filtration ceased upon circulatory arrest (n = 3). At higher P_j, 0.75 mg ml⁻¹ HA plus CSC buffered Q̇_s to ∼3 μl min⁻¹ over a wide range of P_j, with an outflow increase of only 0.04 ± 0.02 μl min⁻¹ cmH₂O⁻¹ (n = 4). With HA or CSC alone, buffering was absent (slopes 0.57 ± 0.04 μl min⁻¹ cmH₂O⁻¹ (n = 4) and 0.86 ± 0.05 μl min⁻¹ cmH₂O⁻¹ (n = 5), respectively). Therefore, polymer interactions can cause outflow buffering in joints. Also, interstitial osmotic pressure promoted filtration in fenestrated synovial capillaries, so the results for frog mesentery capillaries cannot be generalised. The difference is attributed to differences in pore ultrastructure.

Hyaluronan is a 2000–7000 kDa polymer of the disaccharide N-acetyl-d-glucosamine-d.-glucuronic acid, and adopts a voluminous, random coil configuration in aqueous solution (Fraser & Laurent, 1996). Its concentration in joint fluid ranges from ≤ 0.5 mg ml⁻¹ in rheumatoid infusions (Dahl et al. 1985) and the dependent joints of large animals (Ogston & Stanier, 1950; Persson 1971) to 2–4 mg ml⁻¹ in healthy human and rabbit joints (Balsazs, 1982; Price et al. 1996). Its major physiological functions in joints are lubrication (Mabuchi et al. 1994) and the conservation of synovial fluid during flexion; flexion raises intra-articular pressure and drives fluid out of the joint through the synovial lining.

Regarding fluid conservation, the opposition to fluid escape from the joint cavity increases as a function of joint pressure in the presence of 2–4 mg ml⁻¹ hyaluronan, a phenomenon termed ‘outflow buffering’. This limits the outflow and creates a virtual plateau on a plot of drainage rate versus pressure (McDonald & Levick, 1995). Outflow buffering has been attributed primarily to the partial reflection of hyaluronan molecules by synovial extra- cellular matrix during leaky ultrafiltration. Ultrafiltration creates a concentration polarisation layer at the tissue- fluid interface, which raises the local hyaluronan osmotic pressure and opposes fluid drainage (Coleman et al. 1999). Partial reflection has been confirmed experimentally (Scott et al. 1998) and its dependence on polymer domain volume was demonstrated using shortened hyaluronan chains (Coleman et al. 2000) and dextrans (Scott et al. 2000a).

In contrast to the marked effects of hyaluronan at 2–4 mg ml⁻¹, concentrations of 0.2–0.5 mg ml⁻¹ fail to buffer outflow; their drainage vs. pressure relation is approximately linear (McDonald & Levick, 1995; Scott et al. 2000b). Low osmotic pressures are partly responsible for the ineffectiveness of weaker solutions, but in addition there is a decline in molecular reflection at low concentrations (Scott et al. 1998). This led to the hypothesis that chain-chain interactions at normal concentrations enhance reflection and hence outflow buffering. Hyaluronan at 2–4 mg ml⁻¹ is in the overlap regime, in which the voluminous domains of adjacent hyaluronan chains overlap and create a strong chain-chain coupling, whereas there is little to no domain overlap at 0.2–0.5 mg ml⁻¹ (Wik & Wik, 1998). One way to test the contribution of chain-chain interaction to outflow buffering would be to use a low, sub-overlap concentration of hyaluronan, and then enhance the chain-chain interactions by the addition of an interacting polymer, such as chondroitin sulphate.

Chondroitin sulphate is a polymer of N-acetylgalactosamine-glucuronic acid disaccharides. It is present at a concentration of 28 mg ml⁻¹ in rabbit knee cartilage, 0.55 mg ml⁻¹ in synovium and ∼0.04 mg ml⁻¹ in synovial fluid (Price et al. 1996). The amine residues may be sulphated at position 4 (chondroitin-4-sulphate, C4S, chondroitin sulphate A) or position 6 (chondroitin-6-sulphate, C6S chondroitin sulphate C). The C4S:C6S ratio is 0.7 in the cartilage, 7.3 in synovium and 1 in synovial fluid. An interaction between chondroitin sulphate and hyaluronan was first reported by Turley & Roth (1980), who found that hyaluronan-coated beads agglutinated with C6S-coated beads. Beads coated in impure C4S interacted less strongly and contained some C6S. Subsequent work by Scott (1992) and Scott et al. (1992) indicated that C6S forms a stable, ordered association with hyaluronan through hydrophobic patches on the polymer chains, and that the association is substantially impaired in C4S due to steric interference by the unfavourably positioned 4-sulphate. Rheological data provided further evidence of an interaction between C6S and hyaluronan (Nishimura et al. 1998). A weak interaction between hyaluronan and a nominal C4S preparation was attributed in part to contamination with C6S.

In view of the above evidence we postulated that an enhancement of the average polymer domain volume through C6S-hyaluronan association should enhance reflection, leading to outflow buffering. The primary aim, therefore, was to test whether the addition of C6S to non-buffering levels of hyaluronan led to buffering of trans-synovial flow.

A second aim was to assess the effect of chondroitin sulphate alone on trans-synovial flow. The motivation for this was the remarkable finding that the rate of filtration by continuous, mesenteric capillaries is almost insensitive to the colloid osmotic pressure of interstitial fluid (Hu et al. 2000). This is contrary to conventional interpretations of the Starling principle of fluid exchange, but conformed accurately to a mathematical model of flow through endothelial intercellular junctions. By contrast, there is evidence that filtration across synovial capillaries, which are fenestrated, increases in response to elevation of interstitial albumin osmotic pressure - although the response is less than half of that to intravascular albumin (McDonald & Levick, 1993). The structure of synovial joints is such that they offer a unique opportunity to manipulate interstitial fluid osmotic pressure within a few micrometres of synovial capillaries without any intervening, continuous cellular barrier such as mesothelium. In view of the findings of Hu et al. (2000) we investigated the effect of interstitial C6C colloid osmotic pressure on synovial capillary filtration.

METHODS

Physiological methods in vivo

Overview

The rate of escape of an infused fluid though the synovial lining of rabbit knees (trans-synovial flow, Q̇_s) was measured over a range of joint pressure (P_j). Initial experiments involved hyaluronan at 0.5 mg ml⁻¹, which does not buffer the fluid drainage. The infusate contained 0.5 mg ml⁻¹ hyaluronan alone (n = 6 joints), or chondroitin sulphate C alone (n = 5), or 0.5 mg ml⁻¹ hyaluronan with chondroitin sulphate C (n = 7). The small quantity of endogenous hyaluronan (180 μg, Coleman et al. 1997) was washed out with three flushes of chondroitin solution prior to three of the chondroitin-alone experiments. Since the outcome was similar in two unwashed joints, the chondroitin-alone results were pooled. The chosen concentration of chondroitin sulphate C, 25 mg ml⁻¹, was based on viscometric evidence of a marked interaction with hyaluronan at this concentration (see Results).

As a result of trends detected at 0.5 mg ml⁻¹ a second set of experiments was carried out using 0.75 mg ml⁻¹ hyaluronan, with and without 25 mg ml⁻¹ chondroitin sulphate C.

Measurement of intra-articular pressure and trans-synovial flow

New Zealand White rabbits weighing 2–3 kg were anaesthetized with 30 mg kg⁻¹ sodium pentobarbitone plus 500 mg kg⁻¹ urethane i.v. and tracheostomized. Anaesthesia of sufficient depth to abolish the corneal blink reflex was maintained by 15 mg sodium pentobarbitone plus 250 mg urethane i.v. every 30 min. All procedures conformed to national legislation on animal experimentation and the animals were killed by i.v. sodium pentobarbitone at the end of an experiment.

Following the method of Coleman et al. (1999) an intra-articular cannula was connected to a water-calibrated pressure transducer to record P_j (±0.1 cmH₂O). A second intra-articular cannula was connected to an infusion reservoir. Flow from the reservoir into the joint matched the trans-synovial drainage rate and held P_j constant. The inflow Q̇_in was recorded by a photoelectric drop counter of drop size 5.6 μl. P_j was increased in steps of ∼2–4 cmH₂O, beginning at ∼2–3 cmH₂O, by raising the infusion reservoir at 30 min intervals, at which times the flows are in a steady state. Experiments continued to ∼24 cmH₂O, which corresponds to a taut arthritic effusion.

Net trans-synovial drainage rate Q̇_s was calculated from Q̇_in at the end of each period by subtracting the known rate of volume creep of the cavity wall i.e. the slight increase in joint volume with time due to viscoelastic creep at a given pressure (e.g. 2 μl min⁻¹ at 10 cmH₂O) as detailed in Coleman et al. (1999). When chondroitin sulphate was present, Q̇_s reversed direction at low intra-articular pressures and was directed into the joint cavity. Since this halted flow through the drop counter, Q̇_s was then calculated as the rate of rise of pressure dP_j /dt multiplied by the joint compliance dV/dP_j taken from Knight & Levick (1982).

Measurement of polysaccharide reflection by the synovial lining

The method was as described by Scott et al. (1998). Briefly, at the end of the above experiment the intra-articular fluid was mixed by flexion-extension cycles and aspirated for chromatographic analysis. The polymer reflected fraction was defined as the mass accumulated in the cavity divided by the mass in the filtrand (cumulative volume of filtrate × infused concentration). The accumulated mass of polymer was calculated as (concentration in mixed aspirate - concentration infused) × joint fluid volume, minus the endogenous hyaluronan mass of 182 μg (Scott et al. 1998). This method was not feasible for the hyaluronan- chondroitin mixture because the osmotic influx of water evoked by the chondroitin diluted the hyaluronan.

Biochemical and biophysical methods

Materials

For joint infusions the rooster comb hyaluronan and chondroitin sulphate (Sigma Chemical Co., Poole, UK) were dissolved in Baxter Ringer solution (147 m.m. Na⁺, 4 m.m. K⁺, 2 m.m. Ca²⁺, 156 m.m. Cl⁻, pH 7.2; Baxter Healthcare Ltd., Thetford, Norfolk, UK) and adjusted to pH 7.4 with drops of NaOH solution. The average molecular mass of the rooster hyaluronan was 2.1 × 10⁶ Da by size-exclusion gel chromatography (Coleman et al. 1999). The nominal molecular mass of the shark cartilage sodium chondroitin sulphate C (lot 128 H1576) was 58 800 daltons by low angle laser light scattering and the nominal C6S:C4S ratio was 9:1. In addition, bovine tracheal sodium chondroitin sulphate A was studied in vitro (lot 55 H0306, nominal mass 20 800 dalton, C4S:C6S ratio 7:3). Samples were dissolved in phosphate-buffered saline at pH 7.4 for the biophysical studies.

Polysaccharide analysis by high performance liquid chromatography (HPLC)

Chondroitin sulphate and hyaluronan concentrations were analysed by HPLC (Waters Ltd., Watford, UK) using a size exclusion TosoHaas TSK G6000 PW_XL column (Anachem Ltd., Luton, UK). Absorbance was measured at 206 nm. The column was calibrated with known concentrations of polymer, and the minimum reliable detection level was ∼0.01 mg ml⁻¹.

Viscometry and molecular domain volume

The viscosity η of solvent (phosphate-buffered saline, pH 7.4) and chondroitin sulphate solutions at 0–35 mg ml⁻¹, with and without 0.50–0.75 mg ml⁻¹ hyaluronan, were measured in an Haake RS150 rheometer (Carl Stuart, Leek, England) (±0.01 mPa s) at 23–26 °C and shear rates 17–500 s⁻¹. Intrinsic viscosity [η] was determined by linear extrapolation of a plot of the logarithm of reduced viscosity, log((η_r − 1)/C), versus concentration C to zero concentration; η_r is viscosity relative to solvent. Intrinsic viscosity (millilitres per gram) represents the volume occupied by a gram of solute at infinite dilution and is thus a function of polymer domain volume and molecular mass (Boyd & Phillips, 1993). Domain volume is usually characterised by a sphere with a radius of gyration R_g, which is the root mean square of polymer segment distances from the molecular centre of gravity. R_g was estimated from [η] and molecular mass M using the relation R_g³ = [η]M/8.84N_A, where N_A is Avogadro's number (Flory, 1971; Boyd & Phillips, 1993). The critical concentration for molecular domains overlap, C*, was calculated as 2.1/[η] (McDonald & Levick, 1995). The relative viscosity, intrinsic viscosity and R_g of the rooster hyaluronan preparation have been published (Scott et al. 2000a, b).

Osmometry

Osmotic pressures (±0.3 cmH₂O) were measured at room temperature using a Diaflo DM30 membrane (Amicon, Lexington, USA) with a nominal molecular rejection of ≥30 kDa. The electronic osmometer and the osmotic pressure-concentration relation for rooster hyaluronan were described by Knight et al. (1988) and Coleman et al. (1999) respectively.

Statistical methods

To facilitate comparisons at identical pressure (since P_j varied a little between experiments), flows were interpolated to standard pressures at 2.5 cmH₂O intervals by linear interpolation between the bounding measurements as in previous work (Coleman et al. 1999). Sets of flow-pressure results were compared by two-way analysis of variance (ANOVA), with Bonferroni's post hoc test. Polynomials were fitted by non-linear regression as implemented in Graphpad Prism (Graphpad Software Inc., San Diego, CA, USA). Relations that did not deviate significantly from linearity by the ‘runs test’ were fitted by linear regression analysis. Regression slopes were compared by analysis of covariance (ANCOVA) as implemented in Graphpad Prism. Student's paired and unpaired t tests were used where appropriate. P < 0.05 was accepted as a significant difference. Means and slopes are followed by their standard errors throughout.

RESULTS

Pressure-flow relation at low concentrations of hyaluronan

The effects of intra-articular pressure on trans-synovial fluid drainage in the presence of 0.50 mg ml⁻¹ and 0.75 mg ml⁻¹ hyaluronan solutions are compared with published results for 2.0 and 3.6–4.0 mg ml⁻¹ hyaluronan in Fig. 1. In the presence of 0.5 mg ml⁻¹ hyaluronan the trans-synovial flow increased approximately linearly with pressure, with a regression slope of 1.16 ± 0.04 μl min⁻¹ cmH₂O⁻¹ (n = 6 joints). A slight increase in slope at >10 cmH₂O was not statistically significant (P = 0.07, runs test). With 2–4 mg ml⁻¹ hyaluronan, by contrast, the slope of the pressure-flow relation approached a quasi-plateau at ≥10 cmH₂O, with flows of 7–8 μl min⁻¹ for 2 mg ml⁻¹ hyaluronan and 3–4 μl min⁻¹ for 3.6–4.0 mg ml⁻¹ hyaluronan (Scott et al. 2000b).

Figure 1. Effect of rooster hyaluronan alone on fluid drainage across synovial lining.

Points for 0.5 mg ml⁻¹ rooster hyaluronan (•, continuous line) are means ± s.e.m. for 6 knee joints. Points for 0.75 mg ml⁻¹ hyaluronan (•, continuous line)are means for 4 joints. Previous data for higher concentrations of hyaluronan (broken lines) are reproduced for comparison (Scott et al. 2000b).

In a pilot experiment with rooster hyaluronan at 1 mg ml⁻¹, a flow plateau of 11–12 μl min⁻¹ was observed. Since 0.5 mg ml⁻¹ was a little less than the lowest plateau-generating concentration (1 mg ml⁻¹), 0.5 mg ml⁻¹ hyaluronan was used in the initial C6C-hyaluronan study. For reasons that emerge below, the hyaluronan concentration was raised to 0.75 mg ml⁻¹ in later experiments. With 0.75 mg ml⁻¹ hyaluronan alone, the fluid drainage rate increased with pressure with a slope of 0.57 ± 0.04 μl min⁻¹ cmH₂O⁻¹ (Fig. 1, 4 joints). The slope was approximately half that observed with 0.5 mg ml⁻¹ hyaluronan (P < 0.001, ANCOVA).

Effect of chondroitin sulphate C alone on pressure-flow relation in vivo

The effect of the chondroitin sulphate solution was particularly striking below 5 cmH₂O, because the net drainage from the cavity ceased entirely and the intra-articular pressure began to increase steadily (Fig. 2A). Pressure continued to rise slowly for 30 min. This indicated that net trans-synovial flow had reversed direction; filtration into the joint cavity was slowly raising its fluid content and pressure.

Figure 2. Effect of chondroitin sulphate C (25 mg ml⁻¹) on net trans-synovial flow.

A, paired records of joint fluid pressure (P_j) and flow into joint cavity (Q̇_in) after a step elevation of the infusion reservoir. With hyaluronan solution (upper pair) the joint pressure rises to a steady level, and fluid drainage is continuous. With chondroitin sulphate C solution (lower pair) the flow of infusate into the synovial cavity soon stops, and a continuing rise in pressure reveals a net filtration of fluid into the cavity by the synovial lining. Each spike on the drop record represents 5.7 μl. The initial high flow into the cavity is the filling phase, i.e. cavity expansion. B, relation between intra-articular fluid pressure and mean trans-synovial flow in presence of 25 mg ml⁻¹ chondroitin sulphate C in vivo (n = 5 joints) and after circulatory arrest (n = 3 joints; means ± s.e.m.). Dashed line shows curve for 0.5 mg ml⁻¹ hyaluronan from Fig. 1 for comparison.

When the infusion pressure was raised above 5 cmH₂O a net drainage was restored, but at lower rates than with 0.50–0.75 mg ml⁻¹ hyaluronan. The reduction in drainage rate was caused by changes in both the intercept and slope of the relation (Fig. 2B, 5 joints). The rightward shift of the pressure intercept indicated that chondroitin sulphate C was exerting an osmotic pressure that opposed net fluid drainage. The slope of the relation, 0.86 ± 0.05 μl min⁻¹ cmH₂O⁻¹ (r² = 0.98), was reduced compared with that for 0.5 mg ml⁻¹ hyaluronan (P < 0.0001, ANCOVA). Deviation from linearity was not significant by the runs test (P = 0.11).

The reduction of slope by chondroitin sulphate C was not as great as the reduction in fluidity (reciprocal of bulk viscosity). The fluidity of chondroitin sulphate C solution (0.12 mPa⁻¹ s⁻¹, viscosity 8.10 mPa s) was 24 % of the fluidity of 0.5 mg ml⁻¹ hyaluronan (0.515 mPa⁻¹ s⁻¹, viscosity 1.94 mPa s), whereas the chondroitin slope was 69 % of the hyaluronan slope. Similarly, the fluidity of the chondroitin solution was 10 % of the fluidity of Ringer solution (1.18 mPa⁻¹ s⁻¹, viscosity 0.8 mPa s), whereas the chondroitin sulphate slope was 38 % of slope observed with Ringer solution by Coleman et al. (2000). In other words, the effect of chondroitin sulphate C on slope was less marked than its effect on bulk fluidity and viscosity. A similar viscous anomaly has also been observed using intra-articular albumin solutions (McDonald & Levick, 1993; Levick & McDonald, 1994).

Pressure-flow relation for chondroitin sulphate solution post mortem

A model of trans-synovial flow predicts that the reversal of net trans-synovial flow and rightward shift of the pressure intercept should result from the enhancement of synovial capillary filtration by a rise in interstitial colloid osmotic pressure (Levick, 1994). To test whether chondroitin sulphate exerted the above effects through microvascular filtration, infusions were repeated in three animals immediately after stopping the synovial blood flow by cardiac arrest, using i.v. sodium pentobarbitone.

After circulatory arrest chondroitin sulphate C infusions no longer reversed the net trans-synovial flow; the net flows at low pressures were now directed out of the cavity (Fig. 2B). Circulatory arrest increased the net drainage rate at all pressures. The increase was entirely attributable to a leftward shift in the intercept (Fig. 2B). The slope itself was actually reduced slightly by circulatory arrest, from 0.86 ± 0.05 in vivo to 0.66 ± 0.05 μl min⁻¹ cmH₂O⁻¹ post mortem (r² = 0.97; P = 0.014 for slope difference, ANCOVA), in keeping with the functional deletion of a conducting pathway (microvascular endothelium) in parallel with the interstitial pathway (see Discussion).

Combined effect of 0.5 mg ml⁻¹ hyaluronan and chondroitin sulphate

The infusion of 0.5 mg ml⁻¹ hyaluronan plus 25 mg ml⁻¹ chondroitin sulphate C depressed the net trans-synovial drainage rate relative to hyaluronan alone, and at ≤ 5 cmH₂O the flow reversed direction from net drainage to net filtration into the joint cavity as with chondroitin sulphate C alone (Fig. 3, n = 7 joints). The right-shifted pressure intercept for the mixed infusate, 6.2 cmH₂O, was almost identical to that for chondroitin sulphate alone, 6.0 cmH₂O. At higher pressures, a comparison of the curves shows that most, but not all, of the depressed net outflow was attributable to the chondroitin sulphate C per se (Fig. 3A). There was a further depression of drainage rate at >10 cmH₂O, however, that was associated with the presence of the hyaluronan. Overall, the depression of the hyaluronan-chondroitin relation below the chondroitin relation was of marginal statistical significance (P = 0.08, two way ANOVA).

Figure 3. Effect of 0.5 mg ml⁻¹hyaluronan with chondroitin sulphate C (♦) on trans-synovial fluid exchange; means ± .s.e.m., n = 7 joints.

A, net trans-synovial flow as function of intra-articular pressure. Relations for 0.5 mg ml⁻¹ hyaluronan alone (dashed line) and chondroitin sulphate C alone (dotted line) from Figs 1 and 2 are shown for comparison. B, opposition to drainage through the synovial lining, i.e. the increase in pressure required to drive unit outflow, in presence of hyaluronan, chondroitin sulphate C and hyaluronan-chondroitin sulphate mixture.

Inspection showed that the further reduction in flow at >10 cmH₂O was caused by a small but consistent change in the shape of the pressure-flow relation. The curve became increasingly sigmoidal due to a flattening of the slope between 10.0 and 17.5 cmH₂O. Although the effect was small, it was in line with the hypothesis under investigation (see Discussion and Fig. 8). The regression slope fitted between 10.0 and 17.5 cmH₂O, namely 0.49 ± 0.05 μl min⁻¹ cmH₂O⁻¹ was significantly flatter than that for chondroitin sulphate C alone, namely 0.71 ± 0.03 μl min⁻¹ cmH₂O⁻¹ (P = 0.02, ANCOVA). Indeed, the slope in the flattened mid-region was only 1.6 times that of the quasi-plateau generated by 2 mg ml⁻¹ hyaluronan, namely 0.26 ± 0.03 μl min⁻¹ cmH₂O⁻¹ (Scott et al. 2000b).

Figure 8. Theoretical effects of hyaluronan on flow across a membrane with concentration polarisation in vitro.

Curves are calculated using a steady-state, partial-reflection concentration-polarisation model for a membrane of fixed area and resistance (Coleman et al. 1999). A, osmotic pressure π_HA at the membrane interface as a function of filtration pressure. For a bulk concentration of 0.5 mg ml⁻¹, π_HA is negligible at σ_HA = 0.47 (lower continuous line) but increases substantially if σ_HA is raised to 0.95 (upper continuous line). Dashed lines are effective osmotic pressure exerted at the surface, σ_HAπ_HA. Curves for 0.75 mg ml⁻¹ (not shown) are of similar shape but reach higher osmotic pressures. The osmotic pressure relative to the filtration pressure (dotted line) passes through a maximum that corresponds to the region of optimal buffering of filtration. B, non-linear flow-pressure relations arising from changing osmotic pressure of the concentration-polarisation layer. For 0.5 mg ml⁻¹ hyaluronan and σ_HA = 0.47 the relation is effectively linear (top continuous line) but raising σ_HA to 0.95 creates a sigmoidal curve (middle continuous line), as does addition of chondroitin sulphate in vivo. Raising hyaluronan concentration to 0.75 mg ml⁻¹ increases the osmotic buffering of the relation (bottom continuous line). The dashed line is the same relation as the bottom continuous line (0.75 mg ml⁻¹ hyaluronan) but with the intercept shifted 6 cmH₂O to the right and the effective intra-membrane fluid viscosity doubled to reproduce the effect of the chondroitin sulphate in vivo. Experimental results for 0.75 mg ml⁻¹ hyaluronan with chondroitin sulphate C (Fig. 4) are shown for comparison.

The flattening of any pressure-flow relation indicates an increased opposition to outflow (see Discussion). Opposition to outflow at a given pressure was calculated as (pressure-intercept pressure)/flow, i.e. the pressure increment needed to drive one unit of outflow (Coleman et al. 1999). In the presence of the hyaluronan-chondroitin sulphate mixture, the opposition to outflow increased by 28 % over the pressure range 10–15 cmH₂O, from 1.49 to 1.91 cmH₂O min μl⁻¹, demonstrating the existence of a weak buffering of outflow (Fig. 3B, top curve). Buffering was not present with 0.5 mg ml⁻¹ hyaluronan alone (Fig. 3B, bottom curve), or chondroitin sulphate C alone (Fig. 3B, middle curve). Opposition to outflow fell at higher pressures. The latter effect is well known and is caused by a fall in the hydraulic resistance of synovial lining due to stretch and increased matrix hydration (Levick et al. 1998).

Combined effect of 0.75 mg ml⁻¹ hyaluronan and chondroitin sulphate

To assess further the conferment of outflow buffering by chondroitin-hyaluronan interaction, a second set of experiments was carried out using a slightly higher concentration of hyaluronan, 0.75 mg ml⁻¹ (n = 4 joints). It was reasoned that, since the relation between hyaluronan concentration and osmotic pressure is curvilinear (Coleman et al. 1999), a feeder concentration of 0.75 mg ml⁻¹ should disproportionately enhance the osmotic pressures generated by a concentration-polarisation layer, while still being too low to buffer outflow on its own (Fig. 1).

The combination of 0.75 mg ml⁻¹ hyaluronan and chondroitin sulphate C had a striking effect on the shape of the pressure relation (Fig. 4A). An obvious plateau of flow at ∼3 μl min⁻¹ developed between ∼10 and ∼20 cmH₂O (slope 0.04 ± 0.02 μl min⁻¹ cmH₂O⁻¹, n = 4), followed by marginal increases in flow at >20 cmH₂O. The attenuation of drainage by the 0.75 mg ml⁻¹ hyaluronan- chondroitin sulphate mixture was substantially greater than that by the 0.5 mg ml⁻¹ hyaluronan-chondroitin sulphate mixture, or chondroitin sulphate alone (P < 0.0001, two-way ANOVA). The changing opposition to outflow is plotted in Fig. 4B. The opposition to outflow increased 5.5-fold between 7.5 and 20 cmH₂O, reaching 6.36 cmH₂O min μl⁻¹ at 20 cmH₂O.

Figure 4. Effect of 0.75 mg ml⁻¹ hyaluronan-chondroitin sulphate C mixture (•) on trans-synovial flow (A) and opposition to outflow (B); n = 4 joints, means ± .s.e.m.

A, net trans-synovial flow as function of intra-articular pressure. Relations for 0.75 mg ml⁻¹ hyaluronan alone (dotted line) and chondroitin sulphate C alone (□) from Figs 1 and 2 are shown for comparison. B, effect of 0.75 mg ml⁻¹ hyaluronan alone, chondroitin sulphate C alone, and 0.75 mg ml⁻¹ hyaluronan-chondroitin sulphate mixture on opposition to drainage through the synovial lining.

Comparison of polymer concentration in infusate and aspirate

When 0.5 mg ml⁻¹ hyaluronan alone was infused, the hyaluronan concentration in mixed, intra-articular fluid sampled at the end of the experiment increased to 2.00 ± 0.41 times the infused concentration (P = 0.02, paired t test, n = 6; Fig. 5A). The fraction of polymer molecules reflected by the synovial lining, R, was 47 %, similar to the reflection of 48 % observed with 0.2 mg ml⁻¹ rooster hyaluronan by Scott et al. (1998). For 0.75 mg ml⁻¹ hyaluronan alone the results were R = 67 % and an increase in concentration by 1.65 ± 0.14 times (n = 4). The lesser fractional increase in concentration for 0.75 mg ml⁻¹ than for 0.5 mg ml⁻¹ was due to a smaller filtered volume (Fig. 1).

Figure 5. Polymer concentration in mixed aspirate from joint cavity (filled columns) after several hours of filtration of the infusate (open columns) across the synovial lining; means ± .s.e.m.

A, hyaluronan concentration in joints infused with 0.5 mg ml⁻¹ hyaluronan (left pair of columns, n = 6) or 0.75 mg ml⁻¹ (right pair, n = 4). B, chondroitin sulphate C concentration in presence of hyaluronan (left pair, pooled results for 0.5 and 0.75 mg ml⁻¹ hyaluronan-chondroitin sulphate mixtures, n = 10) or absence of hyaluronan (right pair, n = 5).

The intra-articular concentration of chondroitin sulphate C, by contrast, fell slightly over the course of the experiment (Fig. 5B); the aspirate-infusate concentration ratio was 0.90 ± 0.03 (n = 5) for chondroitin sulphate alone. The ratio was not significantly affected by the presence of hyaluronan, being 0.88 ± 0.08 for chondroitin sulphate with 0.5 mg ml⁻¹ hyaluronan (n = 6) and 0.89 ± 0.01 for chondroitin sulphate with 0.75 mg ml⁻¹ hyaluronan (n = 4). The slight reduction in the intra-articular concentration of chondroitin sulphate during the experiment is attributed to dilution during the initial period of capillary filtration into cavity, and to diffusion out of joint cavity. It was not possible to measure hyaluronan reflection in the presence of chondroitin sulphate, owing to the initial fluid movement into the joint cavity.

Viscometric interaction between hyaluronan and chondroitin sulphates in vitro

The viscosity of chondroitin sulphate C, with and without 0.50–0.75 mg ml⁻¹ hyaluronan, in phosphate-buffered saline at pH 7.4, 25 °C and 104 s⁻¹, is plotted versus concentration in Fig. 6A. The solvent viscosity was 0.93 ± 0.04 mPa s (n = 3), slightly above that of pure water (i.e. 0.89 mPa s at 25 °C; Robinson & Stokes, 1970). Chondroitin sulphate solutions showed little shear dependence; the viscosity of 25 mg ml⁻¹ chondroitin sulphate C, namely 6.86 ± 0.08 mPa s at 104 s⁻¹ (n = 3), varied by < 4 % over the shear rate range 17–500 s⁻¹. The viscosities of 0.5 and 0.75 mg ml⁻¹ hyaluronan at 104 s⁻¹ were 2.17 ± 0.04 (n = 3) and 3.13 ± 0.02 mPa s, respectively (n = 3), and showed only minor shear thinning at these concentrations (from 2.69 to 1.82 and from 4.21 to 3.01 mPa s, respectively, at 17–500 s⁻¹).

Figure 6. Biophysical properties of the polymer solutions.

A, viscosity (η) at 104 s⁻¹ as a function of the concentration (x) of chondroitin sulphate C alone (CSC), CSC with 0.5 mg ml⁻¹ hyaluronan (HA), or CSC with 0.75 mg ml⁻¹ HA. Means of triplicate measurements; the standard error bars fall inside the symbols. For explanation of curve labelled ‘Additive’, see Results. The curves represent the second order polynomial η = A + Bx + Cx², where A, B and C are, respectively, 0.916, 0.034 and 0.0081 for chondroitin sulphate C alone, 2.081, 0.106 and 0.0095 for CSC + 0.5 mg ml⁻¹ HA and 3.111, 0.118 and 0.0111 for CSC + 0.75 mg ml⁻¹ HA. Arrow illustrates determination of size of interactive effect for the 25 mg ml⁻¹ CSC-0.5 mg ml⁻¹ HA mixture. B, magnitude of interactive viscous component for chondroitin sulphate C with 0.50 and 0.75 mg ml⁻¹ hyaluronan, and for chondroitin sulphate A with 0.5 mg ml⁻¹ hyaluronan. C, osmotic pressure of chondroitin sulphate C solutions (π, cmH₂O) in the presence and absence of hyaluronan. The fitted curves are the second order polynomials. π = (0.643 ± 0.295)C + (0.088 ± 0.007)C² for chondroitin sulphate C alone (r² = 0.96) and π = 0.026 + (0.516 ± 0.253)C + (0.104 ± 0.008)C² for chondroitin sulphate C plus 0.5 mg ml⁻¹ hyaluronan (r² =0.98).

Viscosity increased as a non-linear function of chondroitin sulphate concentration. The relations were well fitted by the second order polynomials given in the legend to Fig. 6. Hyaluronan increased the slope of the curve, indicating that its effect is more than additive. The curve that would result from a simple additive effect was calculated by adding the specific viscosity of the hyaluronan solution to each point on the chondroitin sulphate C curve; specific viscosity is the viscosity of a solution minus the viscosity of the solvent. The resulting ‘additive’ curve is shown for chondroitin sulphate C plus 0.5 mg ml⁻¹ hyaluronan in Fig. 6A. The increment in viscosity in excess of the additive viscosity, i.e. the difference between observed and additive values for the mixture (e.g. arrow in Fig. 6A), is termed the interactive viscosity component.

The interactive component is plotted in Fig. 6B. For the mixture of 25 mg ml⁻¹ chondroitin sulphate C and 0.50 mg ml⁻¹ hyaluronan the interactive component was 2.66 mPa s. The interactive component accounted for 25 % of the viscosity, which was 10.76 ± 0.11 mPa s (n = 3). For the mixture of 25 mg ml⁻¹ chondroitin sulphate C and 0.75 mg ml⁻¹ hyaluronan, which had a viscosity of 12.76 ± 0.08 mPa s (n = 3), the interactive component of 3.69 mPa s, or 39 % bigger than with 0.5 mg ml⁻¹ hyaluronan (Fig. 6B). These results support a previous report of rheological interactions between hyaluronan and chondroitin sulphate C chains (Nishimura et al. 1998).

To test the specificity of the interaction of hyaluronan with chondroitin sulphate C, the interaction with chondroitin sulphate A was also investigated by viscometry. The interaction was very small (Fig. 6B, bottom curve). The viscosity of the mixture was only marginally higher than the sum of the component viscosities, and even the small interaction observed may have been caused, in part, by the presence of chondroitin-6-sulphate in the impure commercial sample (see Methods).

Intrinsic viscosity, domain radius and overlap concentration

The intrinsic viscosities of chondroitin sulphates C and A were, respectively, 189 and 34 ml g⁻¹; that of rooster hyaluronan is 2953 ml g⁻¹ (Coleman et al. 1999). The radius of gyration R_g of chondroitin sulphate C was calculated to be 12.8 nm (see Methods). R_g for rooster hyaluronan is 101–185 nm (Coleman et al. 1999). The critical concentration C* for overlap of the chondroitin sulphate C domains was calculated to be 11 mg ml⁻¹. Hyaluronan at 0.5 mg ml⁻¹ was below its overlap concentration of 0.71–1.35 mg ml⁻¹ (Coleman et al. 1999, Scott et al. 2000b), and hyaluronan at 0.75 mg ml⁻¹ was on the borderline.

Polymer retention times during chromatography

Hyaluronan retention time during HPLC is negatively related to chain length (Coleman et al. 1997). The retention time of the rooster hyaluronan was 7.60 ± 0.10 min (n = 20). In aspirates taken at the end of hyaluronan-alone experiments in vivo, the retention time was reduced to 7.50 ± 0.09 min (P = 0.08, paired t test, n = 14). Addition of chondroitin sulphate C to hyaluronan in vitro reduced the hyaluronan retention time to 7.34 ± 0.07 min (P = 0.03, unpaired t test, n = 12). The change corresponded to an increase in average chain size by 470 kDa. The retention time for chondroitin sulphate C, 10.29 ± 0.07 min (n = 6), was not altered significantly by hyaluronan (10.46 ± 0.07 min, n = 8, P = 0.22, unpaired t test).

Osmotic pressure

The osmotic pressure of 25 mg ml⁻¹ chondroitin sulphate C varied a little between batches, averaging 67.7 ± 1.4 cmH₂O (n = 6) for one batch and 61.0 ± 1.8 cmH₂O (n = 5) for another. The osmotic pressures of 0.5 and 0.75 mg ml⁻¹ hyaluronan are trivial, 0.03 and 0.05 cmH₂O, respectively (Coleman et al. 1999). The osmotic pressure of chondroitin sulphate was not increased significantly by the addition of hyaluronan at 0.50–0.75 mg ml⁻¹ (P = 0.34, unpaired t test). The number-average molecular mass, calculated from the first virial coefficient of polynomials fitted to the osmotic pressure-concentration relations (Fig. 6C), was 39 310 dalton for chondroitin sulphate C alone and 48 960 dalton for chondroitin sulphate C with hyaluronan (van’ t Hoff's law). The difference in the virial coefficients was not statistically significant.

DISCUSSION

The principal new findings were that the elevation of interstitial oncotic pressure by chondroitin sulphate promotes filtration into the joint cavity (Fig. 2); that neither 25 mg ml⁻¹ chondroitin sulphate C alone, nor 0.50–0.75 mg ml⁻¹ hyaluronan alone, buffers trans-synovial drainage significantly (Fig. 1 and Fig. 3B); that the addition of chondroitin sulphate C to 0.5 mg ml⁻¹ hyaluronan leads to a moderate buffering of outflow (Fig. 3); and that the addition of chondroitin sulphate C to 0.75 mg ml⁻¹ hyaluronan confers a strong buffering of outflow over a wide pressure range (Fig. 4).

The importance of outflow buffering at physiological hyaluronan concentrations lies in the conservation of synovial fluid during pressure elevation by flexion; the proportionate attenuation of fluid escape with pressure prevents the flexed joint from squeezing itself dry. Conversely, the loss of outflow buffering at 0.50− 0.75 mg ml⁻¹ hyaluronan and negligible chondroitin sulphate concentration, levels typical of arthritic joints, facilitates the drainage of effusions.

Pericapillary osmotic action of chondroitin sulphate

Due to its relatively small volume domain (R_g 12.8 nm) chondroitin sulphate C should readily permeate the synovial interstitial matrix, which has an equivalent cylinder radius of ≤ 41–87 nm (Scott et al. 2000a). In support of this it was found that chondroitin sulphate C did not accumulate in the joint cavity, and that the pressure intercept at zero flow was only ∼6 cmH₂O despite an osmotic pressure of 61–68 cmH₂O. The reversal of net flow in vivo but not post mortem confirmed that chondroitin reached the microcirculation.

The microvascular filtration evoked by chondroitin sulphate is attributed to an elevation of the interstitial oncotic pressure (π_i), in qualitative agreement with the Starling principle of fluid exchange (Michel, 1997). During flow reversal the chondroitin reaches the capillaries by diffusion against the flow, a process facilitated by the shortness of the diffusion distance, ∼5 μm. Capillary filtrate enters the joint cavity because fenestrations in the walls of synovial capillaries are clustered on the side facing the joint cavity (Knight & Levick, 1984).

The importance of local, as opposed to global interstitial Starling forces in determining capillary filtration rate was emphasised by Michel (1984). Contrary to the conventional assumption, the endothelium does not ‘see’ the bulk π_i but a locally different value (Michel 1997; Michel & Curry 1999; Hu & Weinbaum, 1999; Levick & Mortimer, 1999). Local concentrations and flows within the interstitium during chondroitin sulphate infusion were computed by iteration using a spatially distributed, steady-state model of synovium of resolution 0.5 μm (Levick, 1994); model parameters are given in the legend to Fig. 7. The computed concentration field in the steady state showed local dilution of interstitial chondroitin sulphate around filtering fenestrations. The flow field showed filtration from capillary into cavity through matrix overlying capillaries and simultaneous drainage of intra-articular fluid through matrix remote from capillaries (Fig. 7). Net trans-synovial flow depended on the relative sizes of the component flows. For saline infusions under 2.5 cmH₂O pressure the model predicted a net trans-synovial drainage rate of 3.65 μl min⁻¹ (Fig. 7A). With 25 mg ml⁻¹ chondroitin sulphate in the joint cavity under the same pressure, there was a computed net filtration into the cavity (−1.70 μl min⁻¹, Fig. 7B). This is in broad agreement with the experimental observation.

Figure 7. Model predictions of steady-state trans-synovial flow patterns evoked by 25 mg ml⁻¹ chondroitin sulphate C.

Symmetrical half-unit of synovium with half-capillary on right and inter-capillary synovium on left (Levick, 1994). Joint cavity at top, subsynovium at bottom, pressures 2.5 and 1 cmH₂O, respectively (Scott et al. 2001), in A and B. Black bars show location of a plaque of endothelial fenestrae. Arrows show direction of flow in 0.5 μm microdomains, not scaled for magnitude. A, saline in joint cavity. Net flow is outwards (drainage), with a minor inflow component. B, chondroitin sulphate in joint cavity; osmotic pressure 67.7 cmH₂O. Net trans-synovial flow is reversed; filtration from capillary into cavity outweighs outflow. Fast filtration streams around the fenestral exits reduce the local, steady-state chondroitin concentration to 51% of that in joint cavity. C, chondroitin sulphate in cavity but joint pressure raised to 10 cmH₂O. Corresponding subsynovial pressure is 3 cmH₂O (Scott et al. 2001). Net flow reverts to drainage. Model parameters: extrafibrillar biopolymers 11.7 mg ml⁻¹ (Scott et al. 2001); corresponding interstitial conductivity 90.6 μm² min⁻¹ cmH₂O⁻¹: synovial endothelial conductance 0.75 μm min⁻¹ cmH₂O⁻¹ (Levick, 1994); capillary pressure 34 cmH₂O: rabbit plasma colloid osmotic pressure 20 cmH₂O: chondroitin sulphate free diffusion coefficient 4 ×10⁻⁷ cm² s⁻¹(Comper & Zamparo, 1990); effective hydrodynamic radius in matrix 2.5 nm (based on a fraction of Einstein radius, Munch et al. 1979). Steady state boundary conditions: intra-articular and intra-capillary pressures and concentrations uniform; subsynovial pressure uniform, concentrations as computed. Perifenestral values as computed.

Qualified applicability of Starling principle to pericapillary osmosis

Few studies have tested the widely accepted assumption that capillary filtration rate is a direct function of bulk π_i. A recent study showed that elevation of bulk π_i around the continuous capillaries in a frog mesentery, using albumin, had virtually no effect on capillary filtration rate (Hu et al. 2000). By contrast in the present study, filtration was clearly enhanced by interstitial chondroitin sulphate, although the magnitude of the osmotic flow was attenuated by the perifenestral dynamic dilution of chondroitin sulphate (51 % in the computed simulation in Fig. 7B). Similar conclusions were reached from the results of albumin infusions into joints (McDonald & Levick, 1993; Levick & McDonald, 1994). The current evidence indicates, therefore, that there is a major difference in the sensitivity of different capillary beds to π_i.

The difference in responsiveness of fenestrated and continuous capillaries to π_i may stem from the difference in ultrastructure. In both cases the endothelial glyocalyx is thought to be the site of the semipermeable pores, but in continuous endothelium the exit from the glycocalyx pore is located deep inside a long, narrow intercellular cleft and is not directly exposed to interstitial fluid or the bulk π_i (Michel, 1997; Michel & Curry, 1999; Hu & Weinbaum, 1999; Hu et al. 2000). In fenestrae, by contrast, the pore exits are almost directly exposed to the interstitial fluid and are thus more responsive to π_i, albeit with marked attenuation of the osmotic flow due to filtration-induced perifenestral concentration gradients (Levick & McDonald, 1994; Levick, 1994). The potential importance of solute gradients at pore exits during osmotic flow has been noted previously by Pedley & Fischbarg (1978), Tedgui & Lever (1985) and Yan et al. (1986).

In contrast to the observations of Hu et al. (2000) with continuous frog mesenteric endothelium, Smaje et al. (1970) found that filtration rate was directly proportional to bulk π_i in rabbit omental and rat cremasteric continuous capillaries. There are thus unexplained differences in the responsiveness of different types of continuous capillary to π_i. Attention is drawn to this issue here because of the potential ramifications for the current view that changes in π_i provide one of the major ‘safety margins’ against oedema formation.

Anomalous interstitial viscosity of chondroitin sulphate C

Chondroitin sulphate reduced the slope of the relation between trans-synovial outflow and hydraulic pressure. According to Darcy’ s law the slope depends on fluidity inter alia (Curry, 1984). The reduction in slope, however, was not as great as the reduction in infusate fluidity. Since the anomaly persisted post mortem, it is unlikely to be the result of dilution of interstitial chondroitin sulphate by capillary filtrate. Rather, the results indicate that the effective viscosity of chondroitin sulphate within the synovial matrix is about half of the bulk phase viscosity. A similar anomaly has been observed using albumin solutions, and the postulated mechanism was partial steric exclusion of the macromolecule from part of the water space in narrow channels (McDonald & Levick, 1993; Levick & McDonald 1994), although a formulation of this effect under predicted the observed magnitude (Levick, 1994).

Linear polymers are known to permeate narrow pores as if they have a smaller effective dimension than their hydrodynamic radius in free solution, and this has been attributed to alignment of the polymer chains during flow through pores (Laurent et al. 1975; Munch et al.1979). One may speculate, therefore, that alignment of the chondroitin chains during flow within the synovial matrix could contribute to their low effective interstitial viscosity.

Effect of hyaluronan alone

The new results with hyaluronan fill in the gap between the concentrations of 0.2 and 2.0 mg ml⁻¹explored by Scott et al. (2000b), and identify the concentration needed to generate buffering as > 0.75 and ≤ 1.0 mg ml⁻¹. The inference that outflow buffering is not mediated by a high viscosity per se (McDonald & Levick, 1994) gains additional support from the finding that chondroitin sulphate alone did not buffer outflow even though its bulk viscosity, ∼7 mPa s, is close to that of 1.6 mg ml⁻¹ rooster hyaluronan. Outflow buffering by hyaluronan at ≥1 mg ml⁻¹ is largely explicable by concentration polarisation due to partial reflection at the synovial surface (Coleman et al. 1999).

Interaction of hyaluronan and chondroitin sulphate C in vitro

Nuclear magnetic resonance data, rotary shadowing electron microscopy and models of geometric and energetic constraints led Scott et al.(1992) to propose that hyaluronan associates with chondroitin sulphate through hydrophobic bonds that form between extensive (9 CH-units), hydrophobic patches that repeat along alternating sides of the polymer chain. Chondroitin sulphate A was predicted to bind less favourably than chondroitin sulphate C due to the greater steric and electrostatic hindrance caused by sulphate groups in the C-4 position. This is supported by the marked difference between ‘A’ and ‘C’ viscous interaction with hyaluronan (Fig. 6B). The results confirm the viscous amplification reported by Nishimura et al. (1998) for chondroitin sulphate C-hyaluronan mixtures.

The amplification of chondroitin sulphate C viscosity by hyaluronan may have more than one cause. Steric exclusion may contribute to the amplification (Ogston & Phelps, 1961); but several considerations indicate that steric exclusion is unlikely to be the sole or major mechanism. (1) To generate the observed rise in viscosity, chondroitin sulphate C would have to be excluded from ∼19 % of the water space by 0.5 mg ml⁻¹ hyaluronan, whereas a polyglucose of size 33 kDa experiences an exclusion of only 8 % at ∼0.5 mg ml⁻¹ hyaluronan (Ogston & Phelps, 1961). (2) Steric exclusion increases the osmotic pressure of the excluded species (Preston et al. 1965; Wiederhielm et al. 1976; Shaw & Schy, 1977; Scott et al. 2000a), whereas hyaluronan at the concentrations used here had a negligible effect on the osmotic pressure of chondroitin sulphate (Fig. 6C). (3) The viscous interaction between chondroitin sulphate A and hyaluronan was trivial, which should not be the case if steric exclusion were the main mechanism at work.

From the work of Scott et al. (1992), in conjunction with the above considerations, we infer that much of the observed viscous amplification is caused by an associative interaction between chondroitin sulphate C and hyaluronan chains, resulting in larger polymer complexes. One may also infer that only a small fraction of the chondroitin sulphate C chains were bound to hyaluronan, since: (i) there was a large molar excess of chondroitin over hyaluronan, (ii) the osmotic pressure of chondroitin sulphate was not reduced by hyaluronan, and (iii) the mean HPLC retention time for chondroitin sulphate was not shortened significantly by hyaluronan. The interaction of a small percentage of the more abundant species, chondroitin sulphate C, with a high percentage of the hyaluronan chains will cause little change in osmotic pressure (a colligative property) but should markedly affect the viscosity due to the exquisite sensitivity of viscosity to chain length. The inference that a high proportion of the hyaluronan chains was associated with chondroitin sulphate C was supported by the reduction in HPLC retention time for hyaluronan in the presence of chondroitin sulphate C.

Mechanism of outflow buffering by hyaluronan- chondroitin mixtures

The generation of outflow buffering by a mixture of solutes that individually failed to buffer outflow is attributed to the associative interaction of hyaluronan and chondroitin sulphate C chains, resulting in a statistical increase in average particle size, increased reflection and thus the osmotic buffering of outflow. Another possibility, namely that the chondroitin chains exerted osmotic pressure at the interface due to reflection by a hyaluronan concentration polarisation layer, is rejected because chondroitin sulphate rapidly permeated the system (see above).

The chain-chain interaction envisaged above is the hydrophobic association between hyaluronan and chondroitin sulphate C described by Scott et al. (1992). An additional possibility is autologous, hyaluronan chain-chain interaction arising from the steric exclusion of hyaluronan from part of the water space by the chondroitin sulphate chains, although the absence of significant viscosity amplification in chondroitin sulphate A-hyaluronan mixtures provides little support for this. The rod-rod exclusion formula of Ogston (1970), neglecting minor end effects (Levick, 1987), predicts that 78 % of the water space should be available to hyaluronan (chain radius 0.4 nm, segment persistence length 4 nm) in a solution of 25 mg ml⁻¹ chondroitin sulphate (chain radius 0.56 nm, volume fraction 0.016). Although this would raise a nominal concentration of 0.75 mg ml⁻¹ hyaluronan to an effective concentration of 0.96 mg ml⁻¹, the observed buffering (Fig. 4) was much greater than could be accounted for by this alone.

Relation of results to partial-reflection concentration-polarisation model

According to the model of Coleman et al. (1999) buffering depends on hyaluronan concentration and the reflection coefficient σ. A 0.5 mg ml⁻¹ solution with a reflection coefficient of ∼0.47 (the observed reflected fraction) will show negligible buffering because the maximal interfacial concentration, namely bulk concentration/(1 - σ), is 1.0 mg ml⁻¹, with an osmotic pressure of only 0.1 cmH₂O (Fig. 8). If the reflection coefficient of hyaluronan (σ_HA) were increased by chondroitin-induced chain associations, e.g. to ∼0.95 (reflected fraction of hyaluronan in overlap regime; Scott et al. 1998), the maximum interfacial concentration becomes 10 mg ml⁻¹ and the osmotic pressure 6 cmH₂O. This changes a linear relation into a sigmoidal curve of the form observed experimentally (Fig. 8). The model predicts much stronger buffering of outflow, as observed experimentally, for 0.75 mg ml⁻¹ hyaluronan with a reflection coefficient of 0.95, because the predicted interfacial concentration is 15 mg ml⁻¹ and the osmotic pressure 13 cmH₂O (Fig. 8).

To summarise, increased filtration into joints upon raising the perifenestral chondroitin sulphate concentration confirmed a Starling-type response to interstitial colloid osmotic pressure. Through its effect on viscosity the chondroitin sulphate also reduced interstitial drainage, but to an anomalously small degree. Addition of chondroitin sulphate C to hyaluronan conferred the capacity to buffer outflow, supporting the hypothesis that polymer chain interactions contribute to outflow buffering, and thus to synovial fluid conservation at physiological hyaluronan concentrations.