Direct evidence for the partial reflection of hyaluronan molecules by the lining of rabbit knee joints during trans-synovial flow

Abstract

1. In synovial joints hydraulic and turnover studies indicate that the synovial lining may partially reflect large macromolecules like hyaluronan, despite discontinuities in the lining cell layer. The reflection hypothesis was tested directly in the present study.

2. Solutions of high molecular weight hyaluronan were infused at controlled pressures into the cavity of rabbit knees under anaesthesia, at concentrations of 0.2 g l⁻¹ (n = 5), 2 g l⁻¹ (n = 5) and 4 g l⁻¹ (n = 6). Time-averaged trans-synovial flows were 9.6, 4.8 and 2.9 μl min⁻¹, respectively. After 5 h infusion the intra-articular fluid was mixed and sampled. Hyaluronan concentration was determined by size-exclusion chromatography.

3. In all sixteen experiments the hyaluronan concentration in the aspirate was greater than that in the infusate (P = 0.0001, Student's paired t test). The increases averaged 2.28 ± 0.04 times at high filtration rates (0.2 g l⁻¹ infusates; mean ± s.e.m.), 1.60 ± 0.09 times at intermediate filtration rates (2 g l⁻¹ infusates) and 1.26 ± 0.08 times at low filtration rates (4 g l⁻¹ infusates).

4. Between 48 and 95% of the hyaluronan in the filtrand was retained in the joint cavity. The greater retention at 2 g l⁻¹, viz.95%, than at 0.2 g l⁻¹, viz.48%, was attributed to interactions between overlapping molecular domains in the more concentrated solution.

5. It is concluded that synovial interstitial matrix can partially reflect hyaluronan molecules, and thus conserve intra-articular lubricant.

Hyaluronan is a long, unbranched, loosely coiled polysaccharide of alternating N-acetyl glucosamine and D-glucuronic acid residues. Its concentration in the synovial fluid of rabbit knee joints is 3.6 g l⁻¹ (Price, Levick & Mason, 1996) and the endogenous molecular mass is (2.4-2.9) × 10⁶ Da (Coleman, Scott, Ray, Mason & Levick, 1997). Hyaluronan is a hydrodynamic lubricant of the articulating surfaces. In addition it has a fluid-conserving role, as indicated by its effect on fluid escape from the joint cavity (McDonald & Levick, 1994, 1995). When hyaluronan solution is present in the cavity, elevation of intra-articular pressure beyond a certain level causes very little increase in the trans-synovial outflow, i.e. a flow plateau develops. This led to the hypothesis that hyaluronan molecules are reflected by the synovial lining, creating a dynamic molecular ‘filtercake’ (concentration polarization layer) that impedes fluid escape. Reflection of hyaluronan would be physiologically useful in conserving intra-articular lubricant. Since the synovial cell layer is discontinuous, the putative reflection is attributed to the interstitial matrix that plugs the intercellular gaps.

Molecular reflection by interstitial matrix is a little-studied phenomenon. The existing evidence for partial reflection of hyaluronan by synovial interstitium may be summarized as follows. (i) The trans-synovial flow of hyaluronan solution attains an almost pressure-independent plateau, which is the characteristic behaviour of an ultrafilter with a concentration polarization layer (Blatt, Dravid, Michaels & Nelsen, 1970; Kozinski & Lightfoot, 1972; Wijmans, Nakao, van den Berg, Troelstra & Smolders, 1985). (ii) The turnover time for hyaluronan in synovial fluid is an order of magnitude longer than that for albumin or water (Coleman et al. 1997). (iii) The radius of gyration for hyaluronan of mass 2.9 × 10⁶ Da is ∼210 nm (Fraser & Laurent, 1996), whereas the porosities in the synovial interstitial matrix have an estimated mean hydraulic radius of only 15-45 nm (Levick, Price & Mason, 1996). In the present study direct evidence for molecular sieving by synovium was obtained by comparing the concentration of hyaluronan in an infused solution with the concentration in the same fluid aspirated from the cavity after a period of filtration across the synovial lining.

METHODS

New Zealand White rabbits weighing 2.0-3.5 kg were anaesthetized by 30 mg kg⁻¹ pentobarbitone plus 500 mg kg⁻¹ urethane i.v. Intra-articular pressure (P_j) and trans-synovial flow (Q̇_s) were measured as described by McDonald & Levick (1995). Briefly, two cannulae were inserted into the suprapatellar bursa of the knee. One was connected to a pressure transducer to record P_j, and the other to an infusion reservoir, whose height controlled P_j. Flow into the joint cavity was recorded by an intervening photoelectric drop counter. Measurements were taken in the steady state after 30-60 min at a stable pressure, with a correction for residual creep of the cavity walls. The intra-articular pressure was increased in seven steps from ∼2 cmH₂O to a final level of ∼25 cmH₂O, to generate the flow plateau as before (McDonald & Levick, 1995). The total duration of the infusion was 5 h.

The infusate was a solution of rooster comb sodium hyaluronate (Sigma). Its average molecular mass, determined by size-exclusion high performance liquid chromatography (HPLC), was 2.1 × 10⁶ (Da Coleman et al. 1997). Three concentrations were studied, namely 0.2 g l⁻¹ in five joints, 2 g l⁻¹ in five joints and 4 g l⁻¹ in six joints. The sub-physiological concentration of 0.2 g l⁻¹ occurs in some pathological joint effusions, and the concentrations 2-4 g l⁻¹ span the physiological range for many joints, including rabbit knee joints. The vehicle was Baxter Ringer solution, a commercial intravenous fluid containing 147 mM sodium, 4 mM potassium, 2 mM calcium and 156 mM chloride (Baxter Healthcare Ltd, Thetford, Norfolk, UK), adjusted to pH 7.4. At the end of the 5 h, two samples were taken for analysis, one from the infusion line and the other from the joint cavity. Before taking the latter, the joint was put through a mixing protocol to dissipate any intra-articular concentration gradients. The joint was flexed and extended ten times, left for 10 min and then flexed and extended a further ten times before aspiration. Fluid return was facilitated by gentle massage. Samples were stored at -80°C and analysed by HPLC using a TosoHaas TSK G6000 PW_XL exclusion column (Waters Ltd, Watford, UK), as described by Coleman et al. (1997).

Procedures conformed to UK legislation and animals were killed at the end of the experiment by an overdose of i.v. sodium pentobarbitone (Euthatal, Rhône Mérieux, Ireland). Means are followed by s.e.m. throughout.

RESULTS

Trans-synovial flows in the presence of rooster comb hyaluronan were low compared with the trans-synovial flow of Ringer or albumin solutions (Fig. 1). The average of the trans-synovial flows over the period of infusion was 9.6 μl min⁻¹ for 0.2 g l⁻¹ hyaluronan, 4.8 μl min⁻¹ for 2 g l⁻¹ hyaluronan and 2.9 μl min⁻¹ for 4 g l⁻¹ hyaluronan. The pressure-flow relations for 0.2 g l⁻¹ hyaluronan were approximately linear and did not show the increase in conductance that is a characteristic feature with Ringer infusions. The pressure-flow relations for 2-4 g l⁻¹ hyaluronan were concave towards the pressure axis, which indicates an increased opposition to outflow with pressure, and reinforces similar observations with umbilical hyaluronan of lower molecular weight (McDonald & Levick, 1995). If the concentration polarization hypothesis is correct, an intra-articular accumulation of hyaluronan should have developed in these joints.

Figure 1. Effect of intra-articular pressure on trans-synovial flow in the presence of 0.2 and 2 g l⁻¹ rooster comb hyaluronan (HA) in Ringer solution in two rabbit knee joints.

Flows of Ringer solution rise to much higher levels in the absence of hyaluronan (dashed line, from a third joint), and the slope of the relation increases due to increases in synovial lining conductance (McDonald & Levick, 1995). The converse happens in the presence of 2 g l⁻¹ hyaluronan.

In every experiment (n = 16) the concentration of hyaluronan in the aspirate was greater than that in the infusate (Fig. 2; P = 0.0001, Student's paired t test). In the five joints infused with the most dilute solution, 0.2 g l⁻¹, the aspirate hyaluronan concentration was 0.44 ± 0.07 g l⁻¹. The infusate concentration measured by HPLC was 0.20 ± 0.01 g l⁻¹ (P = 0.03, paired t test). The intra-articular hyaluronan concentration had increased 2.28 ± 0.04-fold over 5 h at an average filtration rate of 9.6 μl min⁻¹.

Figure 2. Hyaluronan concentration in synovial cavity following infusion of hyaluronan solutions.

The concentration of hyaluronan in fluid aspirated from the joint cavity after 5 h of fluid filtration across the joint lining is plotted as a function of the concentration in the infusion line, measured by high pressure liquid chromatography. The infusate concentrations covered a 20-fold range. All the results lie above the line of equality (P = 0.0001, paired t test).

At physiological concentrations the filtration rates were lower than for 0.2 g l⁻¹ and the relative increases in concentration were smaller. In five joints infused with a nominal 2 g l⁻¹ solution, the aspirate concentration was 3.47 ± 0.25 g l⁻¹ and the measured infusate concentration was 2.17 ± 0.10 g l⁻¹. The aspirate concentration was 1.60 ± 0.09 times that infused (P = 0.003, paired t test) and the filtration rate averaged 4.8 μl min⁻¹.

In six joints infused with a nominal 4 g l⁻¹ solution, the aspirate concentration was 5.48 ± 0.25 g l⁻¹ and the measured infusate concentration was 4.37 ± 0.13 g l⁻¹. For these solutions the size of the concentration increase was 1.26 ± 0.08-fold (P = 0.02, paired t test) and the filtration rate was 2.9 μl min⁻¹.

The HPLC column retention times for aspirate and infusate did not differ significantly (infusate, 7.332 ± 0.026 min; aspirate, 7.335 ± 0.021 min). This indicated that the average molecular weight of the polydisperse hyaluronan was not altered substantially.

To test whether a layer of concentrated, viscous hyaluronan solution might remain close to the synovial surface despite the mixing protocol, the joint cavity was dissected open after the mixing protocol and fluid was aspirated directly from the synovial surface in one experiment with 2 g l⁻¹ hyaluronan. The concentration of hyaluronan in the sample from the surface, 3.25 g l⁻¹, was no greater than that in the main aspirate, viz. 3.53 g l⁻¹.

DISCUSSION

The results show that hyaluronan is partially reflected by the synovial lining during fluid filtration out of the joint cavity. Other factors contributing to the increase in aspirate concentration, namely active secretion of hyaluronan by the lining and mixture with the endogenous synovial fluid, are quantitatively insufficient to explain the increases (Table 1). The secretion rate into the cavity of the rabbit knee in vivo averages 4.8-5.8 μg h⁻¹ (Coleman et al. 1997). Over 5 h this generates 24-29 μg hyaluronan. The volume of liquid in the cavity at the end of the experiment, when pressure averaged 22.5 cmH₂O, is 2.0 ml (Knight & Levick, 1982). From these values it follows that secretion accounts for 5% of the increase in intra-articular mass after 0.2 g l⁻¹ infusions, and for 1% after 2 or 4 g l⁻¹ infusions. The small mass of hyaluronan in the endogenous synovial fluid, 182 μg (Coleman et al. 1997), accounts for 38% of the ‘excess’ intra-articular hyaluronan mass after 0.2 g l⁻¹ infusions but for only 6% after 2 and 4 g l⁻¹ infusions.

Table 1.

Increase in intra-articular hyaluronan mass after filtration of hyaluronan solution through synovium in vivi; calculation of rejected fraction

(a) Infused concentration (mg ml−1)	0.20	2.0	4.0
(b) Aspirate concentration (mg ml−1)	0.442	3.474	5.477
(c) Hyaluronan mass in cavity at end (b × 2 × 1000)* (μg)	883.4	6948	10954
(d) Endogenous hyaluronan + mass secreted over 5 h† (μg)	208.5	208.5	208.5
(e) Hyaluronan in 2 ml infusate* (μg)	400.0	4000.0	8000.0
(f) Increase in hyaluronan mass due to reflection (c - d - e)	274.9	2739.5	2745.5
(g) Time-averaged trans-synovial flow (μl min−1)	9.61	4.83	2.89
(h) Volume of filtrand over 300 min (g × 300/1000) (ml)	2.883	1.448	0.867
(i) Hyaluronan mass in filtrand (h × a × 1000) (μg)	577	2896	3469
(j) Rejected fraction (f/i)	0.48	0.95	0.79

^*Volume of fluid in cavity is 2 ml at 22.5 cmH₂O, the average pressure at the end of the experiment (Knight & Levick, 1982).

^†Mass of hyaluronan in endogenous synovial fluid is 182 μg. Secretion rate is 4.8–5.8 μg h⁻¹ (Coleman et al. 1997).

It is clear, therefore, that the majority of the increase in intra-articular hyaluronan concentration was caused by the retention of infused hyaluronan molecules, while water and electrolytes continued to filter out through the intercellular spaces. The fractional increase in concentration was greatest for the 0.2 g l⁻¹ solution, viz. 2.28 ×, and least for the 4 g l⁻¹ solution, viz. 1.26 ×, because the volume of fluid filtered through the membrane was highest for 0.2 g l⁻¹ hyaluronan (high flows) and least for 4 g l⁻¹ hyaluronan (low flows).

What proportion of the hyaluronan in the filtrand was reflected in each group of experiments? This can be calculated from mass balance considerations, by dividing the increase in intra-articular hyaluronan mass, corrected for secretion and endogenous hyaluronan, by the mass of hyaluronan in the filtrand. The latter is mean trans-synovial flow × duration × infusate concentration. Values are given in Table 1. The fraction of filtrand hyaluronan that was rejected by the filter was 0.48 for the group of 0.2 g l⁻¹ infusions, which did not induce a flow plateau (Fig. 1), 0.95 for the group of 2 g l⁻¹ infusions and 0.79 for the group of 4 g l⁻¹ infusions, both of which induced flow plateaux. In a previous study of endogenous hyaluronan in vivo, where the average concentration is 3.6 g l⁻¹, a rejection fraction of 0.94-0.95 was calculated from the secretion rate and fluid turnover rate (Coleman et al. 1997). As noted in the Introduction, sieving of hyaluronan molecules by synovial interstitial matrix is dimensionally feasible, and Parker & Winlove (1984) observed sieving of hyaluronan by artificial membranes with a pore radius in excess of 45 nm.

Factors that influence the rejected fraction include (i) differences in molecular size between infused and endogenous hyaluronan molecules, (ii) increasing molecular interactions as concentration is increased (see below), (iii) dependence of sieving ratios on filtration rate (see below), and (iv) experimental errors such as failure to mix concentrated fluid close to the membrane with the bulk of the intra-articular liquid before aspiration. The last source of error may not be severe in view of the result from the dissected joint. Also, (v) Barry, Gowman & Ross-Ethier (1995) reported that the reflection of hyaluronan by a Nucleopore membrane increased with time as a concentration polarization layer developed; so the rejected fraction may have changed over the course of the experiment.

The rejected fraction of hyaluronan molecules in the experiments at 0.2 g l⁻¹, namely 0.48, was less than the rejected fractions at 2-4 g l⁻¹, namely 0.79-0.95. This may arise from factor (ii) above. At 0.2 g l⁻¹ the hydrated domains of adjacent hyaluronan molecules do not touch or overlap (the ‘dilute regime’). The critical concentration C* at which adjacent molecular domains become contiguous is ∼1 g l⁻¹ (Fraser & Laurent, 1996). At 2-4 g l⁻¹ (the ‘semi-dilute’ regime) the domains of adjacent hyaluronan molecules overlap, due to their highly hydrated, expanded coil configuration of radius of ∼100 nm. This generates strong interactions between contiguous hyaluronan molecules, which behave as a quasi-continuous network of chains at 2-4 g l⁻¹ (Fraser & Laurent, 1996). The rejection of hyaluronan in 2-4 g l⁻¹ solutions can, therefore, be expected to be greater than the rejection of the discrete, deformable hyaluronan molecules present in solutions of 0.2 g l⁻¹.

For the two solutions above C*, the rejected fraction at 2 g l⁻¹, viz. 0.95, was greater than that at 4 g l⁻¹, viz. 0.79. This may arise from factor (iii) above. In a membrane whose reflection coefficient, σ, is < 1, the fraction of the solute that is rejected by the membrane increases as a curvilinear function of filtration rate, asymptotically approaching the maximum, σ (Patlak, Goldstein & Hoffman, 1963; Curry, 1984). Since filtration rates were higher at 2 g l⁻¹ than at 4 g l⁻¹, a higher rejection fraction may be expected at 2 g l⁻¹. Two caveats apply, however. First, the results do not necessarily establish σ, because the rejected fraction was not necessarily maximal. A plot of filtration rate versus rejected fraction is needed to establish σ. Second, the molecular sieving theory of Patlak et al. (1963) is based on a well stirred compartment, whereas in an unstirred joint cavity the hyaluronan concentration at the synovial surface during filtration is greater than in the bulk phase (concentration polarization). Differences in the concentration polarization layer thickness at 2 and 4 g l⁻¹ could contribute to the difference in the apparent rejected fraction.

Although the quantitative aspects of hyaluronan ultrafiltration in vivo require further study, including exploration of the relation between rejection fraction and filtration rate, the results clearly demonstrate the primary phenomenon, namely that synovium partially reflects hyaluronan molecules. This property will help to retain hyaluronan in the joint cavity and prolong its working life there.