| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1 Division of Physiology, St George's Hospital Medical School, Cranmer Terrace, London SW17 0RE, UK
2 Division of Medicine, Imperial College, London W12 0NN, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
3 h were analysed by high-performance liquid chromatography. R was calculated as (1 – downstream/upstream concentration), using [FD] to adjust for joint lymph dilution in femoral lymph. Intra-articular HA concentration after 3 h, 0.47 ± 0.02 mg ml–1 (mean ±S.E.M., n= 31), exceeded the infusate concentration, 0.20 mg ml–1, while subsynovial and lymph [HA] were reduced relative to [FD]. The changes in [HA] demonstrated synovial molecular sieving of HA. R from cavity to lymph (Rlymph) fell monotonically from 0.93 at 6 µl min–1 to 0.14 at 89 µl min–1 (P < 0.0001, regression analysis, n= 33). R values calculated from the intra-articular HA accumulation (Rasp) or the low subsynovial concentrations (Rsyn) were similar negative functions of filtration rate. R for lymphatic capillary endothelium (Rendo), calculated from lymph/subsynovial concentration ratios, was effectively zero (–0.03 ± 0.18, n= 21), confirming that synovium, not initial lymphatic endothelium, is the reflection site. Logarithmic linearisation of the results evaluated the synovial HA reflection coefficient as 0.91. In conclusion, the existence of concentration polarisation during joint fluid drainage was supported by the demonstration of a negative relation between filtration rate and Rlymph, Rasp and Rsyn.
(Received 27 February 2004;
accepted after revision 7 April 2004;
first published online 8 April 2004)
Corresponding author J. R. Levick: Division of Physiology, St George's Hospital Medical School, Cranmer Terrace, London SW17 0RE, UK. Email: tvttfgheot63{at}hotmail.com
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
5 cmH2O produce remarkably little increase in trans-synovial fluid escape (McDonald & Levick, 1995). This physiologically important process, which we call outflow buffering, conserves the joint's lubricating fluid during periods of high intra-articular pressure, such as a sustained flexion. It also greatly prolongs the intra-articular working life of HA, which is one of the lubricating materials in joint fluid. Coleman et al. (1999) attempted to explain the outflow buffering quantitatively, starting from the hypothesis that filtration causes HA to accumulate at a partially reflecting interstitium–fluid interface. It was argued that a concentration gradient forms near the unstirred interface (‘concentration polarisation’ theory) and the concentrated HA opposes filtration through its inward-directed osmotic pressure. The problem of filtration attenuation by concentration polarisation is a general one, of concern not only to physiologists but also to chemical engineers developing membrane ultrafiltration systems (Blatt et al. 1970; Zhang & Ethier, 2001; Peppin & Elliott, 2001).
A fundamental postulate of the joint concentration polarisation hypothesis is that the synovial lining partially reflects HA molecules. On first acquaintance this seems unlikely, because there are large, micrometre-wide gaps between the discontinuous synovial lining cells. Moreover HA can permeate the interstitium, both in joints and dermis, to reach lymph (Brown et al. 1991; Fraser & Laurent, 1996; Brown et al. 1999). Nevertheless, partial HA reflection by synovium is indicated by three observations. First, the residence half-life of HA in the joint cavity is unusually long, 14–32 h (Denlinger, 1982; Brown et al. 1991). Indeed, given the known, very low rate of HA secretion and relatively rapid fluid turnover, the high physiological HA concentration can only be explained if HA is selectively retained in the cavity during fluid drainage (Coleman et al. 1997). Second, when an intra-articular HA solution is filtered experimentally through the synovial lining, HA accumulates in the joint cavity (Scott et al. 1998). Third, the HA concentration in joint lymph is much lower than in joint fluid (Sabaratnam et al. 2003). The evidence thus consistently indicates that the synovial interstitial drainage pathway can partially sieve out HA molecules from the draining liquid.
The sieving of HA is probably due mainly to steric hindrance by the biopolymer matrix in the synovial intercellular spaces. The matrix is a complex network of sulphated and non-sulphated glycosaminoglycans, proteoglycans, glycoproteins and microfibrils (review, Levick et al. 1996). The polymers create a high resistance to transport, as shown by the observation that the pressure gradient needed to generate 1 unit of flow is 20-fold higher for synovium than subsynovium (Scott et al. 2003). The subsynovium consists of loose areolar tissue and contains a network of lymphatic capillaries that do not penetrate the synovium itself (Yamashita & Ohkubo, 1993; Xu et al. 2003). The subsynovial lymphatic network drains through the femoral lymph trunks to the iliac nodes (Davies, 1946; Nagai, 1987; Reimann et al. 1989).
An algebraic expression that describes partial solute reflection and concentration polarisation by a homoporous membrane under specific boundary conditions leads to an unusual prediction, namely that the effect of fluid velocity on reflected fraction is biphasic. The reflected fraction is (1 – transmitted fraction). The transmitted fraction, or ‘sieving coefficient’ in the glomerular filtration literature, is filtrate concentration, Cout, divided by the upstream bulk concentration, Cin. In the absence of concentration polarisation it is well known that the reflected fraction is a positive, monotonic function of filtration rate; it increases curvilinearly from 0 at zero filtration rate to a plateau 
(the reflection coefficient) at high flows (Curry, 1984). This is the case for plasma protein ultrafiltration across capillary endothelium, for example (Garlic & Renkin, 1970; Taylor & Granger, 1984). By contrast, when upstream concentration polarisation is present, the predicted reflection vs. flow relation is biphasic (Coleman et al. 1999 and Fig. 1). The initial ‘conventional’ phase of increasing reflected fraction with increasing filtration rate is followed by a declining phase, because the membrane surface is presented with ever-increasing concentrations of solute. To illustrate this, let us suppose that a 1 mg ml–1 solution is filtered rapidly through a membrane of = 0.5. When the ultrafiltration rate is fast enough to raise the local concentration at the membrane surface to 2 mg ml–1, the lowest possible filtrate concentration is x 2 mg ml–1, i.e. 1 mg ml–1. The transmitted fraction Cout/Cin thus approaches 1 at high flows: and its complement, the reflected fraction, approaches zero. A negative sieving phase has indeed been observed in vitro (Blatt et al. 1970; Munch et al. 1979).
|
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The joint cavity of a rabbit knee was cannulated and infused with a mixture of HA and fluorescein–dextran (FD, reference solute) at a controlled intra-articular pressure to generate a sustained filtration across the synovial lining over several hours. The measured trans-synovial drainage rates ranged from 6 to 89 µl min–1 and the pressures from 10.5 to 43.8 cmH2O. A single filtration rate was generally studied per preparation, but two levels of filtration rate were studied in four cases. Femoral lymph was then collected at intervals over a further 3 h and analysed for HA and FD. The ratio of [HA] to freely permeating [FD] in lymph was used to calculate a cavity-to-lymph sieving coefficient (transmitted fraction). At the end, fluid was aspirated from the joint cavity and subsynovial compartment to measure the intra-articular HA accumulation and to determine the tissue responsible for the molecular sieving. The lymph method has been evaluated in detail previously (Sabaratnam et al. 2002, 2003) and is summarised briefly below.
Materials
Rooster comb HA (0.2 mg ml–1, 2 x 106 Da, radius of gyration 101–181 nm, Coleman et al. 1999) and fluorescein–dextran (30 µg ml–1, 0.02 x 106 Da, Stokes-Einstein radius 3.1 nm) were obtained from Sigma Chemical Co. (Poole, UK). The two solutes were coadministered in Baxter Ringer solution (147 mM Na+, 4 mM K+, 2 mM Ca2+, 156 mM Cl–; Baxter Healthcare Ltd, Thetford, UK).
The HA concentration was set to 0.2 mg ml–1, which is below the critical concentration for molecular domain overlap, 0.79–1.35 mg ml–1 (Coleman et al. 1999; Scott et al. 2000a), to enable us to explore a wide range of trans-synovial flows. By contrast, it is impossible to achieve high flows at >1 mg ml–1 HA because of the extreme osmotic effectiveness of outflow buffering at high bulk HA concentrations (Scott et al. 2000a,b). The HA concentration used in this study was at the bottom of the range recorded in rheumatoid arthritis (Dahl et al. 1985).
The reference solute FD was included to delineate the proportion of joint lymph in femoral lymph. In a few initial experiments Evans blue-labelled albumin (0.5 mM Evans blue, 5 mg ml–1 bovine albumin, >99% bound) was used as reference solute (Sabaratnam et al. 2002), because the coloration facilitated the detection and dissection of the lymphatic vessels. With practice, however, coloration proved unnecessary; the femoral trunk vessels (diameter 0.5 mm) were readily identified through a dissecting microscope, so the more sensitive FD method was used in most studies (n= 25).
Animal preparation, joint cannulation and trans-synovial filtration
New Zealand White rabbits weighing 2–3 kg were anaesthetised with 30 mg kg–1 sodium pentobarbitone plus 500 mg kg–1 urethane I.V. and tracheostomised. 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. Only one joint per animal was studied (n= 31) and in most cases only one trans-synovial flow was studied per preparation. Following the procedure of Coleman et al. (1999), an intra-articular cannula was connected to a pressure transducer to record intra-articular fluid pressure, Pj (± 0.1 cmH2O). A second intra-articular cannula was connected to an infusion reservoir, the height of which regulated Pj. Flow from the reservoir into the joint cavity in the steady state depends mainly on trans-synovial drainage rate and was measured using a photoelectric drop counter (5.6 µl) and chart recorder. A small correction was applied for viscoelastic creep of the cavity walls as previously described. Procedures conformed to UK legislation and animals were killed humanely by an overdose of I.V. sodium pentobarbitone at the end of the experiment.
Lymph collection
Prenodal joint lymph drains antero-medially into two to three lymphatic trunks in the femoral triangle. These were dissected clear of the femoral artery, vein and nerve under a Zeiss dissecting microscope and ligated. Joint cannulation and trans-synovial filtration were set up prior to the lymphatic dissection but the filtration rate was variable during the 1–2 h surgical preparation due to operating movement/pressure artifacts. A further 1 h priming interval was then allowed at the required, constant, undisturbed, trans-synovial filtration rate. This was followed by a 2- to 3-h period of femoral lymph collection, using the inside-out cannulation method of Sabaratnam et al. (2002). The aspirated lymph accumulated in a fluid trap. The trap was emptied and the lymph weighed (± 1 mg) every 15 min for 2–3 h. The lymph flow was well maintained over 3 h as illustrated previously (Sabaratnam et al. 2002). If necessary, the height of the infusion column was adjusted slightly over the 3 h to maintain the trans-synovial filtration rate at the desired level; such adjustments were small, of the order +1–2 cmH2O.
Estimation of joint lymph dilution factor (VV) using the reference solute
Femoral lymph (flow Lfemoral) is a mixture of joint lymph (flow Ljoint) and lymph from the rest of the leg. We have shown previously that the volume fraction of joint lymph in femoral lymph, Vv, is given by:
|
| (1) |
Calculation of joint-to-lymph HA transmitted fraction Rlymph (sieving coefficient)
If a fraction Rlymph of the HA molecules in the bulk filtrand (infusate) is reflected by the synovium-to-lymph pathway during filtration, the transmitted fraction (1 –Rlymph) is given by the relation:
|
| (2) |
Assessment of HA reflection using joint aspirate analysis (Rasp)
An additional way of demonstrating HA reflection is to measure the upstream HA accumulation. For this purpose the intra-articular fluid was mixed by 10 flexion–extension cycles at the end of the experiment and aspirated for analysis. The accumulation-based estimate of reflection, called Rasp, was calculated as the HA mass reflected and retained in the cavity divided by the HA mass in the cumulative filtrand volume, as described by Scott et al. (1998). Rasp was calculated therefore from the increase in the intra-articular HA concentration x intra-articular fluid volume (giving the HA mass reflected and retained in the joint cavity) and the cumulative volume of fluid filtered during the experiment x infusate concentration (giving the HA mass in the total filtrand volume). Rasp and Rlymph did not differ significantly in previous work (Sabaratnam et al. 2003).
Subsynovial fluid sampling (Rsyn, Rendo)
To confirm that the synovial lining, and not the lymphatic capillary endothelium, is the chief selective membrane, a sample of the intervening subsynovial fluid was aspirated at the end of the experiment and analysed. The animal was killed by I.V. pentobarbitone overdose and 1 ml Evans blue solution was injected into the joint cavity to demonstrate its boundaries. The periarticular tissue was then dissected away from the exterior to within a millimetre or so of the border of the intact cavity. A sample of 10–500 µl clear subsynovial fluid was aspirated for analysis using a flexible polythene cannula. Comparison of the HA and FD concentrations in subsynovial fluid with that in the infusion line gave an estimate of the synovial lining reflected fraction (Rsyn), while comparison with lymph gave an estimate of lymphatic capillary endothelial reflection (Rendo), as previously described (Sabaratnam et al. 2003).
Sample preparation
Samples were centrifuged at 7800 g for 5 min, diluted to 200 µl in Ringer solution and digested with 5.6 units papain (Sigma, UK) at 60°C for 1 h to prevent partial masking of the small HA band by endogenous extravascular albumin. Papain digestion does not alter the HA molecular size (Coleman et al. 1997). Samples were re-centrifuged at 7800 g for 5 min prior to high performance liquid chromatography.
HA and FD analysis by high performance liquid chromatography (HPLC)
The recovered HA was quantified using size exclusion, high performance liquid chromatography (HPLC) as described by Coleman et al. (1997). The system comprised a Waters 2690 separation module (Waters Ltd, Watford, UK), a TosoHaas TSK G6000 PWXL column (Anachem Ltd, Luton, UK) of nominal resolution 40–8000 kDa and a Waters 486 ultraviolet absorbance detector set at 206 nm for HA analysis. The chromatograms were analysed using Waters Millennium32 software. The injection volume was 50–100 µl and the column flow 1 ml min–1 Ringer solution.
A calibration curve was constructed for each sample batch using rooster HA, and was linear from 3 ng ml–1 to 400 ng ml–1 (Coleman et al. 1997). The mean retention times for HA standards of weight-average molecular mass from 210 kDa to 5500 kDa donated by Dr O. Wik (New Pharmacia, Uppsala, Sweden) were used to estimate average molecular size. The standards had been characterised by laser light scattering.
Fluorescein–dextran was analysed using the same HPLC column and an in-line Waters 474 SATIN fluorimeter at excitation wavelength 475 nm and emission wavelength 530 nm. Minimum detection level was <0.3 µg ml–1. Control studies showed that FD did not influence the HA chromatogram at 206 nm, nor did HA influence the FD chromatogram. Evans blue albumin was analysed by the Waters fluorimeter at excitation wavelength 450 nm and emission wavelength 575 nm.
Concentration polarisation theory for a partially reflecting, homoporous membrane
Coleman et al. (1999) described a steady-state model for filtration from a cross-perfused compartment of time-independent bulk concentration Cin across a membrane of reflection coefficient < 1 bounded by an upstream concentration polarisation layer of constant thickness 
. The magnitude of is set by local stirring conditions in the model. They used expressions for mass balance, convection (solvent drag) and diffusion within a concentration–polarised boundary layer to derive the relation between volume filtration rate 
and the steady-state sieving coefficient Cout/Cin, where Cout is filtrate concentration. For a concentration-independent diffusion coefficient D the sieving relation is:
|
| (3) |
(1 –)/PA and PA is the membrane permeability–area product. Equation (3) describes a biphasic relation between transmitted fraction and fluid velocity 
as illustrated in Fig. 1. The transmitted fraction is 1.0 at zero filtration rate, due to diffusive permeation; decreases to a minimum of >[1 –] as filtration rate increases, then increases towards the limit of 1.0 at high filtration rates. Reflected fraction, which is (1 – transmitted fraction), does the converse. By contrast, sieving from a perfectly stirred, cross-perfused compartment with no concentration polarisation is described by the well-validated, classic Patlack equation (Curry, 1984; Taylor & Granger, 1984):
|
| (4) |
as filtration rate increases; reflection fraction rises monotonically to a plateau equal to (Fig. 1). The curves for sieving in the presence and absence of concentration polarisation are thus strikingly different, which enables transmission measurements at high flows to distinguish between the two scenarios. As Coleman et al. (1999) emphasised, a trans-synovial HA filtration experiment in vivo entails many more complexities than the simple model represented by eqn (3), including regions of dead-end filtration (see Discussion). Such factors can affect the slope of the relation, but they do not abnegate the fundamental prediction of a negative relation at high filtration rates if concentration polarisation is present. Expression (3) was thus a key stimulus for the present study and a useful heuristic aid in the discussion and interpretation of the results.
Statistical analysis
Means ±S.E.M. are cited throughout. Student's t test was used for paired, unpaired and one-sample comparisons. One-way ANOVA with Tukey's post hoc test was used for multiple comparisons. Lines were fitted by linear regression analysis and their slopes compared by analysis of covariance (ANCOVA) as implemented in GraphPad Prism (San Diego, CA, USA). Correlations were analysed by Spearman's correlation coefficient, r. Significance was accepted at P
0.05.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Femoral lymph flow increased as a function of trans-synovial flow but the lymph flow was always substantially smaller than the corresponding trans-synovial flow (Fig. 2A). A regression line through the results from 23 preparations had a slope of 0.176 ± 0.020, i.e. the lymph flow averaged 17.6% of the trans-synovial flow (slope P < 0.0001). The y-intercept, 1.3 ± 1.0 µl min–1, represents the basal femoral lymph drainage from the leg at zero net trans-synovial filtration. The regression slope of <1 confirmed our previous finding that periarticular lymphatic drainage is incompletely coupled to interstitial fluid load in the presence of joint effusions (Sabaratnam et al. 2002). The poor coupling probably contributes to the periarticular oedema and morning stiffness that characterise swollen joints.
|
Reflected fraction for joint-to-lymph pathway (Rlymph) and effect of filtration rate
The HA transmitted fraction calculated from the lymph/infusate concentration ratio using eqn (2) was less than unity at all flows, showing that the net joint-to-lymph pathway acts as partial molecular sieve for HA at all drainage rates. The reflected fraction for the net pathway, Rlymph (eqn (2)), ranged from 0.93 at the lowest trans-synovial flow, 6 µl min–1, to 0.14 at the highest trans-synovial flow, 89 µl min–1 (Fig. 3A). The negative correlation between Rlymph and filtration rate was highly significant, with Rlymph falling by –0.0094 ± 0.0014 per µl min–1 increase in filtration rate (regression analysis, Spearman's r=–0.71, P < 0.0001, n= 33). These changes are in the direction predicted by concentration polarisation theory (Fig. 1, filled symbols), and are in sharp contrast with the plateau predicted for ultrafiltration with no concentration polarisation (Fig. 1, open symbols).
|
The HA concentration in the mixed, aspirated joint fluid increased to between 1.5 times and 4.2 times the infused concentration over several hours of trans-synovial filtration. The mean final aspirate concentration was 0.47 ± 0.02 mg ml–1 (n= 31, P < 0.0001, 1-sample t test comparison with infused 0.20 mg ml–1). The reflected fraction Rasp, calculated from the accumulated HA, showed a highly significant negative correlation with trans-synovial filtration rate (Spearman's r=–0.85, P < 0.0001) (Fig. 3B). The negative Rasp–flow relation was very similar to the negative Rlymph–flow relation, its slope of –0.0108 ± 0.0012 per µl min–1 being almost identical to that for Rlymph (P= 0.49, slope comparison by ANCOVA). The Rasp results thus provided further evidence for a concentration polarisation-dominated transport process.
The net pathway that determines Rlymph is composed of two potential barriers in series, namely synovium and lymphatic endothelium. The latter is located in the loose areolar subsynovial tissue close to the outer border of the synovium. Since only 17.6% of the trans-synovial filtrate crossed the lymphatic endothelium (Fig. 2A and above), the potential influence of an endothelial barrier on Rasp should be considerably less than on Rlymph. Yet the Rasp and Rlymph relations were almost identical. This indicates that synovium, not lymphatic endothelium, must account for most of the molecular sieving. To test this inference directly, we measured the HA concentration in subsynovial fluid, i.e. in the fluid compartment located between the synovial membrane and the lymphatic capillary endothelium, with the following results.
Subsynovial, pre-lymphatic hyaluronan concentration and effect of filtration rate
The HA concentration in the subsynovial filtrate, mean 0.073 ± 0.010 mg ml–1 (n= 22), was much lower than in the mixed intra-articular aspirate, 0.47 ± 0.02 mg ml–1, or the infusate, 0.20 mg ml–1 (P < 0.0001, one-sample t test). The subsynovial HA concentration increased with filtration rate from 0.049 mg ml–1 at the lowest sampled filtration rate (10 µl min–1) to 0.16 mg ml–1 at the highest sampled filtration rate, 72 µl min–1 (Fig. 4A) (P= 0.009, linear regression analysis). The subsynovial results thus demonstrated (a) a major reduction in HA concentration across synovium per se (cf. lymphatic endothelium) to a minimum of 5% of the infused concentration, and (b) an increase in the synovial transmitted fraction with increasing filtration rate.
|
Comparison of subsynovial and lymph concentrations; lymphatic capillary endothelial reflection (Rendo)
Although the above results indicate that synovium is the main reflection site for HA, we also investigated whether the lymphatic capillary endothelium might partially reflect HA. To assess lymphatic reflection, the ratio of HA concentration in femoral lymph to that in subsynovial fluid was calculated and corrected for the dilution of joint lymph in femoral lymph. To correct for dilution, each HA concentration ratio was divided by its corresponding FD concentration ratio, i.e. [FD] in femoral lymph/[FD] in subsynovial fluid. For example, in an experiment where the lymph/subsynovial fluid HA ratio was 0.442, the FD ratio was 0.455, giving an endothelial transmitted fraction of 0.971 and a reflected fraction Rendo of 0.029.
The femoral lymph/subsynovial fluid ratio for [HA] averaged 0.629 ± 0.112 (n= 21). The corresponding ratio for [FD] was 0.623 ± 0.065. The mean of the lymphatic endothelial transmitted fractions was 1.033 ± 0.176 and the mean Rendo was –0.033 ± 0.176 (n= 21). The endothelial reflection fraction was not significantly different from zero (P= 0.85, one sample t test). Likewise the slope of the relation between Rendo and trans-synovial flow, 0.0005 ± 0.0085 min µl–1, was not significantly different from zero (P= 0.95, n= 21, linear regression analysis).
The mean values for Rlymph, Rasp, Rsyn and Rendo are compared in Fig. 5. One way ANOVA (P < 0.0001) shows significant differences between Rendo and Rlymph (0.59 ± 0.05, n= 33) or Rasp (0.48 ± 0.05, n= 29) or Rsyn(0.59 ± 0.06, n= 22) (P < 0.001 in each case, Tukey's post hoc test). There were no significant differences between Rlymph, Rasp and Rsyn (P > 0.05, Tukey's test).
|
The membrane parameter on which ultrafiltration primarily depends is the reflection coefficient . This can be estimated from results at high filtration rates 
, when transport is dominated by convection rather than diffusion. When the Péclet number 
(1 –)/PA is >5, the exponential term 
(eqn (3)) approaches zero. The term 
then equals . Thus at high flows, with Cout/Cin re-expressed in terms of the reflected fraction R, eqn (3) gives the linear relation:
|
| (5) |
for the pooled Rlymph, Rasp and Rsyn results at filtration rates >20 µl min–1, where eqn (5) can be considered valid, the estimated Péclet number being >> 5. Linear regression analysis showed a highly significant relation (P < 0.0001), with a y-intercept of 2.27 ± 0.35 and slope –0.051 ± 0.007 min µl–1. Since the y-intercept equals ln[/(1 –)] (eqn (5)), the mean synovial is calculated to be 0.91, with 95% confidence intervals of 0.83–0.95.
|
/DA, an estimate of , the unknown boundary layer thickness, can be made in principle. Uncertainties in D and A limit this to an order-of-magnitude estimate. The HA diffusion coefficient D is probably in the range 1.38–2.10 x 10–7 cm2 s–1, based on its concentration dependence (Wik & Comper, 1982) and estimates of Cm at high flows (2.2–4.0 mg ml–1 for = 0.91–0.95). Rabbit knee synovial surface area A is 17.4 cm2 at 18 cmH2O (Levick, 1994, Table 1) but may increase at higher Pj. For the above values the estimate of is 71–111 µm, i.e. the boundary layer thickness is of the order 100 µm. Chromatogram retention times for lymph hyaluronan; the permeating chain length
The peak HA retention time tret in the HPLC size-exclusion column is related negatively to the average chain length (Coleman et al. 1997). Rooster comb HA is a polydisperse material with a Mw/Mn ratio of 2.16 (weight average/number average molecular mass; Coleman et al. 1999). To examine whether the polydisperse HA chains were differentially sorted by size during trans-synovial filtration, we compared the tret values of HA in the synovial filtrand and filtrate. The mean tret was 7.360 ± 0.018 min for infused HA (n= 48), 7.206 ± 0.021 min for intra-articular aspirated HA at the end of an experiment (n= 27), 7.257 ± 0.038 min for subsynovial HA (n= 22) and 7.293 ± 0.026 min for lymph HA (n= 139). The differences did not quite reach conventional significance by one way ANOVA (P= 0.079); the reduction in aspirate HA tret (implying selective intra-articular retention of the longer chains) compared with infusate HA tret came closest to a significant difference. There was no statistically significant reduction in the average molecular size of the HA chains reaching the lymphatic system, although the variance of the lymph HA tret was increased. The chromatogram profile (spread) likewise showed no obvious change. There was no significant relation between trans-synovial filtration rate and HA tret for intra-articular aspirate, subsynovial fluid or lymph; the slopes fitted by regression analysis were not significantly different from zero.
The lymph HA tret corresponded to mean molecular mass of >106 Da. Very large HA chains are thus capable of slowly permeating the cavity-to-lymph barrier. Lymph nodes preferentially absorb and catabolise long HA chains relative to short chains with such an avidity that chains of >106 Da are not found in postnodal lymph (Tengblad et al. 1986; Fraser et al. 1988). The finding that the average chain size of HA in the femoral lymph was >106 Da in these experiments therefore supports the anatomical evidence that the joint lymph in femoral lymph trunks is pre-nodal in nature.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Comparison with previous studies of joint hyaluronan clearance
The finding that HA reflection was always less than unity was in broad agreement with an autoradiographic study by Brown et al. (1991), who showed that radiolabelled HA penetrates slowly through the synovial lining over 2 h. More recently, Asari et al. (1998) reported histological evidence that intra-articular 2300 kDa fluorescein–labelled HA hardly penetrates the synovial lining of the dog knee whereas 840 kDa HA penetrates the lining more readily. The slow clearance of large proteoglycans and radio-colloid from the joint cavity compared with albumin further supports the view that synovium acts as a size-selective molecular sieve (Page-Thomas et al. 1987; Reimann et al. 1989), as do studies with HA of different weight-average molecular mass (Coleman et al. 2000). The present HPLC tret values for sample upstream and downstream of the membrane assessed the possible separation of short chains (the size and quantity of which are unknown) from long chains (likewise unquantified) in a polydisperse HA solution. The difference between upstream and downstream sample tret fell a little short of conventional significance (P= 0.079, ANOVA).
Site of the molecular sieve; reflection across barriers in series
The transport of HA from the joint cavity into lymph involves the permeation of two barriers in series, namely the synovial interstitial matrix, which separates the joint cavity from the subsynovial compartment, and the lymphatic capillary endothelium, which separates the subsynovial fluid from the lymph. In principle either or both barriers could reflect HA significantly, resulting in an overall reflection coefficient that is a weighted average of the component reflection coefficients 1 and 2 (Patlack et al. 1963):
|
| (6) |
C1 and C2 are the concentration differences across barriers 1 and 2. In the cavity-to-lymph barrier the concentration difference between the joint cavity and subsynovium was much larger than that between the subsynovium and lymph: and the reflected fraction across lymphatic capillary endothelium, Rendo, was not significantly different from zero (Fig. 5). The synovial component 1C1 therefore greatly exceeds the lymphatic capillary endothelial component 2C2, and average
1. Several additional observations reinforce the conclusion that synovium is the main reflection site. The quantitative agreement between Rlymph, Rasp and Rsyn and the closely similar effect of filtration rate on all three indicate reflection at the synovial surface rather than the microlymphatic endothelium. Intra-articular chymopapain, which depletes synovium of its proteoglycans and other non-collagen interstitial protein, abolishes the reflection-dependent process of outflow buffering (Coleman et al. 1999). A seemingly odd finding from biochemical analysis, namely that the HA concentration in the synovial interstitial extrafibrillar space is only 21.4% of that in the contiguous native synovial fluid, indicates a partial exclusion of HA from the matrix (Price et al. 1996a), and further excludes the microlymphatic wall as the primary reflection site.
The absence of significant HA reflection by lymphatic capillary endothelium is presumably due to its discontinuous basement membrane and the distensible intercellular gaps. The latter can be up to 1 µm wide and are capable of transmitting even fine particulate matter (Schmid-Schönbein, 1990). The radius of gyration of a HA molecule, 100–180 nm, is smaller than many lymphatic endothelial gaps, and the radius of the HA chain itself is only 0.4 nm. The ready permeation of hyaluronan from the subsynovial region into the lymphatic system is thus not unexpected.
Lymphatic endothelium uniquely possesses the recently described hyaluronan-binding receptor LYVE-1, and a slow HA endocytosis via LYVE-1 binding has been postulated (Prevo et al. 2001; Xu et al. 2003). Lymphatic endocytosis appears to be quantitatively insignificant in the present experiments, because a significant endocytotic uptake would reduce the lymph [HA] and thus create an apparently positive Rendo. Measured Rendo was not significantly different from zero, however. The quantitative agreement between Rlymph and Rsyn or Rasp likewise indicates that the low lymph [HA] cannot be caused by local intralymphatic endocytosis or catabolism. Also, since Rasp and Rsyn are not significantly different, the accumulated mass of HA in the joint cavity is not significantly different from the mass deficit in subsynovium, so we can conclude that synovial cell catabolism of HA did not significantly influence the result.
In light of the above discussion, we equate the term Rsyn with reflection across the synovial layer alone. The fact that 82% of the trans-synovial filtrate did not cross the lymphatic wall (Fig. 2A), but simply accumulated in the periarticular tissues, further justifies this view.
Molecular sieving by interstitial matrix; effective pore size versus polysaccharide radius
The synovial cell layer is discontinuous, so molecular sieving cannot be an attribute of occludens–adherens junctions as it is in epithelia and capillary endothelium. The micrometre-wide intercellular gaps are plugged by a dense matrix of glycosaminoglycans, proteoglycans, glycoproteins, collagen fibrils and microfibrils (Levick et al. 1996). The analysed glycosaminoglycan concentration is 4 mg per ml extrafibrillar water (Price et al. 1996a) and the current estimate for total non-collagen biopolymer is 11.7 mg ml–1 (Scott et al. 2003). Strand-like interstitial elements at these concentrations transect the interstitial void volume (water space) to create spaces of sufficiently small dimensions to exclude and reflect partially the very large solutes, and sustain substantial pressure gradients (Scott et al. 2003).
Although the spaces or ‘pores’ between the matrix strands are probably irregular, their scale can be characterised in a simplified, one number form by addressing the following question. If the matrix is represented by a set of uniform cylindrical pores, what pore radius would reproduce the observed membrane properties? Synovial hydraulic conductivity measurements indicate that the matrix pores have a mean hydraulic radius, rH, of 15–45 nm; for a regular cylinder rH is by definition half the cylinder radius, 30–90 nm. Molecular sieving studies indicate an equivalent sieving radius of 87 nm (dextran sieving, Scott et al. 2000b) to 70 nm (HA sieving in the dilute regime, Coleman et al. 1999). Since the radius of gyration Rg of rooster hyaluronan, 101–181 nm, exceeds the estimates of equivalent pore size, reflection can be expected on steric grounds. The fact that reflection was only partial may be a consequence of the chain flexibility. Random thermodynamic movements along the flexible chain enable many polymers to wriggle snake-like (‘reptate’) through channels that are much narrower than their radius of gyration (Munch et al. 1979; Barry et al. 1996). If there were no reptation and HA behaved as an equivalent solid sphere of radius 0.8Rg (Munch et al. 1979), then for = 0.91 the sieving pore radius is 103–113 nm (Curry, 1984). This provides an upper limit estimate for matrix pore size.
Qualitative nature of relation between reflected fraction and filtration rate
As discussed above, the negative relation between filtration rate and Rlymph, Rasp and Rsyn is due primarily to a synovial barrier to HA egress. An expression describing filtration rate-dependent molecular sieving by a cross-perfused membrane of reflection coefficient < 1 in the presence of an upstream concentration polarization layer of fixed thickness (eqn (3)) predicts that the reflected fraction is zero at zero flow, increases steeply with flow to a maximum close to at a relatively low flow, and then falls progressively towards a limit of 0 at high flows (Fig. 1). The theoretical falling limb is the direct result of concentration polarisation, for the following reason. The concentration of reflected solute at the membrane surface, Cm, increases with increasing filtration rate, as plotted out in Appendix Fig. 9 of Coleman et al. (1999). Consequently, the transmitted concentration, which equals Cm(1 –) at high Péclet numbers, increases with filtration rate, approaching a limit equal to the bulk feeder concentration Cinfusate. Thus the ratio of transmitted to infused concentration, Cm(1 –)/Cinfusate, approaches 1 as flow increases. Its complement, the bulk reflected fraction, approaches zero. The results in Figs 3 and 4 show that the HA reflected fraction falls with increasing filtration rate, thus conforming qualitatively with the concentration polarisation model. An alternative form of concentration polarisation model to the steady-state model (eqn (3)), namely non-steady state, dead-end partial ultrafiltration, likewise predicts a negative relation under the conditions of our study (Lu et al. 2004); see below.
Another possibility, namely that the negative phase is caused by a progressive fall in as the membrane is subjected to increasing filtration pressures, is negated by the extremely effective osmotic buffering of fluid drainage as filtration pressure increases. The buffering depends on increased effective osmotic pressures 
across the synovial membrane as filtration pressure is raised (Coleman et al. 1999). The effective osmotic pressures would decline if fell.
The inferred preservation of a high despite increasing pressure is intriguing because a raised pressure causes marked deformation of synovium, namely area expansion, reduced thickness, intercellular space deformation and, in saline-infused joints, increased matrix hydration (McDonald & Levick, 1988; Levick & McDonald, 1989; Price et al. 1996b). The last of these implies an increase in pore dimensions, and indeed the hydraulic permeability to saline increases more than can be accounted for by increased area and reduced thickness of membrane (Levick, 1994). It is conceivable that the effective is maintained, despite the structural changes, by the concentrated HA, which might act through its osmotic pressure and/or matrix impaction (pore fouling). In support of the latter possibility, studies of HA ultrafiltration across an artificial membrane of pore radius 7.5 nm indicate an increase in over a 20 h filtration period (Barry et al. 1996).
The model represented by eqn (3) is a steady-state model. Analyses of successive 15-min lymph collection over 3 h demonstrated that lymph flow, lymph HA concentration, [HA]/[FD] ratio and trans-synovial flow were in a steady state, within the limits of the statistical variance; regression slopes versus time were not significantly different from zero (Sabaratnam et al. 2003). Likewise, the trans-synovial flow–time curve in the presence of HA indicated a close approach to a steady-state at 60 min (Coleman et al. 1999; Fig. 3).
Quantitative aspects of relation between reflection and filtration rate
The difficulty in applying eqn (3) quantitatively to joints in vivo is that, whereas the model parameters D (diffusivity), A (membrane area), (reflection coefficient) and (boundary layer thickness) are constants, these parameters probably change in joints. The effects of such changes on the predicted curves are illustrated and compared with the experimental results in Fig. 7.
|
/DA (eqn (5)). The diffusion coefficient of HA is generally found to increase with concentration (Laurent et al. 1960; Barry et al. 1996), though for fluorescein–HA self-diffusion in a laser-bleached spot the converse applies (Gribbon et al. 1999). For mutual diffusional exchange of HA and water across a relaxing boundary, an appropriate model for a concentration polarisation layer, the relation is D= 10–7{0.393 + 0.426C} cm2 s–1 (Wik & Comper, 1982). D in the boundary layer will thus rise as increased filtration rates raise local C. The effect will be to reduce the slope, e.g. from curve A to curve B in Fig. 7.
Surface area too increases, because increased joint pressures stretch the synovial membrane (Knight & Levick, 1982; McDonald & Levick, 1988; Levick & McDonald, 1989). This reduces the fluid velocity 
/A, which attenuates the rise in Cm and reduces the slope. The area–pressure relation is unknown at high pressures but curve C of Fig. 7 shows the marked slope reduction caused by area expansion.
The concentration polarisation layer thickness, , may change with filtration rate for two reasons, each related to the flattened, elongated geometry of the joint space, which more closely resembles a slab configuration than a sphere. (1) When the velocity of the fluid spreading out from the point-source intra-articular cannula is raised, the shear rate of fluid within the cavity must increase. This may reduce and hence the slope –/DA (eqn (5)), as illustrated in curve D of Fig. 7. (2) In the ‘dead-end’ corners of the cavity, where the outer sheet of synovium meets the inner sheet, different boundary conditions will apply. Mathematical modelling shows that in unstirred dead-end filtration, increases with time due to back diffusion, until eventually a steady state is reached when Cm is high enough for Cout to equal Cin. Reflected fraction R approaches zero therefore with time. The mathematics of dead-end, time-dependent partial ultrafiltration are complex, but work in progress indicates that the time constant will be short at high filtration rates and long at low filtration rates (Lu et al. 2004). Consequently, in experiments of fixed duration, such as those described here, R will be closer to zero at high filtration rates than at low filtration rates. Thus both the steady-state and time-dependent dead-end filtration models predict a negative slope under the conditions of our experiments.
Finally, it is possible that may increase as HA chains impact in the surface matrix and clog its pores. This possibility is supported by the observation that the outflow buffering curve (the 
–Pj relation) for 2–4 mg ml–1 HA is even flatter than predicted by the concentration polarisation model (Coleman et al. 1999). Moreover, membrane fouling has been noted during HA ultrafiltration in vitro (Peppin & Elliott, 2001). The effect of a rise in is to flatten the relation (Fig. 7, curve E).
There exists therefore a large family of sieving curves, with each curve describing a particular combination of D, A, and at a given pressure and filtration rate. The shallow relation observed experimentally may be the result of shifts from one iso-parameter curve to another on the right as Pj and 
are raised.
To summarise, femoral lymph, subsynovial and intra-articular fluids were analysed to assess the hyaluronan reflected fraction across synovium over a range of filtration rates. Hyaluronan concentration was reduced in subsynovial fluid and lymph and increased in intra-articular fluid, confirming partial reflection by the synovial lining. The reflected fraction decreased as a function of filtration rate at high rates, which indicates the existence of a concentration polarisation layer at the synovial fluid/synovium interface.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Barry SI, Gowman LM & Ethier CR (1996). Obtaining the concentration-dependent diffusion coefficient from ultrafiltration experiments: application to hyaluronate. Biopolymers 39, 1–11.[CrossRef]
Blatt WF, Dravid A, Michaels AS & Nelsen L (1970). Solute polarization and cake formation in membrane ultrafiltration: causes, consequences and control techniques. In Membrane Science and Technology, ed. Flinn JE, pp. 47–95. Plenum Press, New York.
Brown TJ, Alcorn D & Fraser JRE (1999). Absorption of hyaluronan applied to the surface of intact skin. J Invest Dermatol 113, 740–746.[CrossRef][Medline]
Brown TJ, Laurent UBG & Fraser JRE (1991). Turnover of hyaluronan in synovial joints: elimination of labelled hyaluronan from the knee joint of the rabbit. Exp Physiol 76, 125–134.[Abstract]
Coleman PJ, Scott D, Mason RM, Levick JR (1999). Characterization of the effect of high molecular weight hyaluronan on trans-synovial flow in rabbit knees. J Physiol 514, 265–282.
Coleman PJ, Scott D, Mason RM & Levick JR (2000). Role of hyaluronan chain length in buffering interstitial flow across synovium in rabbits. J Physiol 526, 425–434.
Coleman PJ, Scott D, Ray J, Mason RM & Levick JR (1997). Hyaluronan secretion into the synovial cavity of rabbit knees and comparison with albumin turnover. J Physiol 503, 645–656.
Curry FE (1984). Mechanics and thermodynamics of transcapillary exchange. In Handbook of Physiology, section 2, The Cardiovascular System, vol IV, Microcirculation, ed. Renkin EM & Michel CC, pp. 309–374. American Physiological Society, Bethesda.
Dahl LB, Dahl IM, Engström-Laurent A & Granath K (1985). Concentration and molecular weight of sodium hyaluronate in synovial fluid from patients with rheumatoid arthritis and other arthropathies. Ann Rheumatic Dis 44, 817–822.
Davies DV (1946). The lymphatics of the synovial membrane. J Anat 80, 21–23.[Medline]
Denlinger JL (1982). Metabolism of sodium hyaluronate in articular and ocular tissues. Doctoral Dissertation, Université de Sciences et Techniques de Lille.
Fraser JRE, Kimpton WG, Laurent TC, Cahill RNP & Vakakis N (1988). Uptake and degradation of hyaluronan in lymphatic tissue. Biochem J 256, 153–158.[Medline]
Fraser JRE & Laurent TC (1996). Hyaluronan. In Extracellular Matrix, vol 2, Molecular Components & Interactions, ed. Comper WD, pp. 141–199. Harwood Academic Publishers, Amsterdam.
Garlick D & Renkin EM (1970). Transport of large molecules from plasma to interstitial fluid and lymph in dogs. Am J Physiol 219, 1595–1605.
Gribbon P, Heng BC & Hardingham TE (1999). The molecular basis of the solution properties of hyaluronan investigated by confocal fluorescence recovery after photobleaching. Biophys J 77, 2210–2216.
Knight AD & Levick JR (1982). Pressure–volume relationships above and below atmospheric pressure in the synovial cavity of the rabbit knee. J Physiol 328, 403–420.
Laurent TC, Ryan M & Pietruszkiewicz A (1960). Fractionation of hyaluronic acid. The polydispersity of hyaluronic acid from the bovine vitreous body. Biochim Biophy Acta 42, 476–485.[Medline]
Levick JR (1994). An analysis of the interaction between extravascular plasma protein, interstitial flow and capillary filtration; application to synovium. Microvascular Res 47, 90–125.[CrossRef][Medline]
Levick JR & McDonald JN (1989). Ultrastructure of transport pathways in stressed synovium of the knee in anaesthetised rabbits. J Physiol 419, 493–508.
Levick JR, Price FM & Mason RM (1996). Synovial matrix-synovial fluid system of joints. In Extracellular Matrix, vol. 1, Tissue Function, ed. Comper WD, pp. 328–377. Harwood Academic Publishers, Amsterdam.
Lu Y, Levick JR & Wang W (2004). Flow dependence of hyaluronan reflection in synovial joint cavities. J Physiol P, C71.
McDonald JN & Levick JR (1988). Morphology of surface synoviocytes in situ at normal and raised joint pressures, studied by scanning electron microscopy. Ann Rheum Dis 47, 232–240.
McDonald JN & Levick JR (1995). Effect of intra-articular hyaluronan on pressure–flow relation across synovium in anaesthetised rabbits. J Physiol 485, 179–193.
Munch WD, Zestar LP & Anderson JL (1979). Rejection of polyelectrolytes from microporous membranes. J Membrane Sci 5, 77–102.[CrossRef]
Nagai S (1987). The basic study of the synovial membrane of the normal rabbit knee by using the induction system of carbon particles. J Tokyo Med College 45, 488–498.
Page-Thomas DP, Bard D, King B & Dingle JT (1987). Clearance of proteoglycan from joint cavities. Ann Rheumatic Dis 46, 934–937.
Patlack CS, Goldstein DA & Hoffman JF (1963). The flow of solute and solvent across a two-membrane system. J Theor Biol 5, 426–442.[CrossRef][Medline]
Peppin SSL & Elliott JAW (2001). Non-equilibrium thermodynamics of concentration polarization. Adv Colloid Interface Sci 92, 1–72.[CrossRef][Medline]
Prevo R, Banerji S, Ferguson DJP, Clasper S & Jackson DG (2001). Mouse LYVE-1 is an endocytic receptor for hyaluronan in lymphatic endothelium. J Biol Chem 276, 19420–19430.
Price FM, Levick JR & Mason RM (1996a). Glycosaminoglycan concentration in synovium and other tissues of rabbit knee in relation to hydraulic resistance. J Physiol 495, 803–820.
Price FM, Levick JR & Mason RM (1996b). Changes in glycosaminoglycan concentration and synovial permeability at raised intra-articular pressures in rabbit knees. J Physiol 495, 821–833.
