Fluid Movement Across Synovium in Healthy Joints: Role of Synovial Fluid Macromolecules Ann Rheum Dis: First Published As 10.1136/Ard.54.5.417 on 1 May 1995

Annals ofthe Rheumatic Diseases 1995; 54: 417-423 417

Fluid movement across synovium in healthy joints: role of synovial fluid macromolecules Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from

J R Levick, J N McDonald

Synovial fluid is not a static pool, but is is affected by movement of a joint, linking joint continually being absorbed and replenished by motion to fluid transport. Active or passive the synovial lining ofthe joint cavity (synovium, flexion of a normal joint can increase IAP to synovial intima), which is approximately 20 ,m just above atmospheric pressure (much greater thick (rabbit knee) to approximately 60 pm for effusions), whereas in extension LAP is a thick (human knee). The three key elements few cm H2O less than atmospheric. At sub- for fluid turnover are the synovial capillary, atmopheric pressures, net flow is often into the synovial interstitium, and the lymphatic joint cavity, and at supra-atmospheric pressures drainage system. net flow is directed out of the healthy cavity.4 5 Synovial capillaries. Normal areolar and adi- Synovial lymphatic system. A plexus of terminal pose synovium have a row of capillaries - 5 ,m lymph vessels is found at the subsynovium- (rabbit) to -30 ,um (man) beneath the surface, synovium border and drains away fluid, macro- many bearing fenestrations (membranes of molecules, and particles that have escaped high permeability to water), often on the side from the joint cavity.6 9 10 Subsynovium com- facing the joint cavity.'3 The high capillary prises loose areolar tissue (substantial areas), density, superficial location, and fenestral and fatty and fibrous tissue. Areolar sub- orientation are well adapted for synovial fluid synovium connects with surrounding connec- formation and nutrient supply. Synovial fluid tive tissue planes and, like areolar tissue is formed primarily by ultrafiltration of plasma elsewhere, acts as a compliant, low-pressure across the fenestral membranes, driven by a net 'sink' when fluid is driven into it, as in imbalance in the 'Starling pressures' acting experiments described here. across the membrane. The Starling pressure imbalance is the pressure drop from capillary plasma to synovial interstitium, minus the Intra-articular albumin and trans- difference in effective colloidal osmotic synovial flow

pressure (COP) across the capillary wall. Edlund established that IAP promotes the http://ard.bmj.com/ Experimental studies of how capillary blood escape of saline from the synovial cavity, and pressure, plasma COP and intra-articular that the pressure-outflow relation steepens at pressure affect net trans-synovial flow have pathological IAPs because the hydraulic resis- been reviewed elsewhere.4 The influence of tance ofthe synovial lining decreases." Normal intra-articular (IA) colloids, however, has only synovial fluid, however, is more complex than recently begun to be studied and is considered saline: its most abundant macromolecule,

here. Although capillary ultrafiltrate is the raw albumin (40-45% of plasma concentration), on October 1, 2021 by guest. Protected copyright. material for synovial fluid formation, the endows the fluid with COP, while hyaluronan synovial lining cells also actively secrete the (-3 g/l) dominates the viscosity. glyosaminoglycan hyaluronan and the glyco- Albumin not only increases COP but also protein lubricin to produce the highly viscous, increases viscosity modestly. To study its lubricating synovial fluid.6 effect on trans-synovial flow in our laboratory, Synovial interstitium. The path into the joint we inserted two cannulae into the joint cavity cavity from a capillary, and from joint cavity to of a rabbit knee under anaesthesia. One a lymphatic vessel in the subsynovium, passes cannula recorded IAP and the other infused between the lining cells. These intercellular albumin solution from a gravity-driven reser- gaps, several ,um wide, contain a complex voir, the vertical height ofwhich regulated IAP. fibrous matrix (type I, III, and V collagen fibrils, An intervening drop counter recorded the rate type VI collagen microfibrils, hyaluronan, ofabsorption ofinfusate. Flows were measured chondroitin and heparan proteoglycans, in the steady state, and after a small correction possibly keratan sulphate, fibronectin) in open for wall creep this was equated with net trans- contact with IA fluid. The hydraulic conduc- synovial flow, Q. (net absorption rate). (The tivity of the matrix is about 10-` cm4 s-1 dyn-' absorbate mostly passes into subsynovium, but or less,7 8 which helps to reduce the rate of transcapillary flow is involved also-see Department of escape of IA fluid when joint pressure is below.) The procedure was to infuse known Physiology, in St George's Hospital increased (for example flexion). albumin concentrations alternately with Krebs Medical School, Effect ofjoint motion and angle. Intra-articular solution; after 15 minutes of infusion, the joint London, fluid pressure (IAP) is an important factor was flushed with the new infusate. No United Kingdom affecting net flow across synovial interstitium: was present. in presence J R Levick hyaluronan Q, the of J N McDonald it opposes capillary filtration by increasing peri- albumin was then expressed as a fraction of Qs Correspondence to: capillary interstitial pressure, and it promotes for Krebs solution; this will be referred to as the Professor Levick drainage from joint cavity to subsynovium. IAP 'ractionalflow'. 418 Levick, McDonald

Albuminl E concentration relative fluidity (reciprocal of relative viscosity) (g/l) if Darcy's law applies and pressure gradient 00 300 200 100 50 dP/dx and bed area A are constant. The dashed Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from 1.0- line of equality in figure 1 is thus a guide to -- - - whether net transynovial flow obeys Darcy's o4- <,/ law. Net Qs in vivo did not follow this line, Eso- -- however: at IAPs of 3-6 cm H2O, albumin D, 0505_ reduced net Qs more than expected from the >. 0 ,z,/reduced fluidity. At very high, non-physio- C5 logical concentrations, net flow even reversed - - direction to a net filtration into the The ,'*1/ cavity. .o.c- deviation below the Darcy line and eventual ~ 1 l * reversal is caused by capillary filtration into the cavity (see later), which is increased by the 10 colloid osmotic pressure of the IA extra- -0.2 vascular fluid (Starling's principle). Capillary Relative fluidity filtration into the cavity partially or wholely ( Valbhkrebs) offsets drainage from the cavity via intercellular spaces, reducing the net trans-synovial absorp- Figure 1 Effect of interstitial albumin on fractional trans- tion rate.nratu ralbuminlthuson- synovial flow in vivo(E) and after death(1) at 3 ton rate. Intra-artcular album thus demon- H20 intra-articular pressure, plotted as afunction of strably permeated the lining to reach the peri- solution fluidity, f. Fractionalflow is net trans-synovial capillary spaces. The above interpretation was flow of albumin solution relative to that ofKrebs solution: supported by experiments (see (alb/ 0Krebs Relative intra-articularfluiditylIIalbis 14Krebs- Arrowed lines unite asequence of results in a single joint. Dashed line of equality represents a simple porous medium features in vivo are notable. in which fractional change inflow equals change in relative Are pericapillary COP and intra-articular COP fluidity ofliquid. equal? The effect of IA albumin on netQS was less than expected from the effect of intra- The effect of albumin on fractional flow in vascular albumin onQs,"2 if one were to assume vivo is shown by the filled squares in figure 1. that pericapillary albumin concentration The x axis here is 'relative fluidity'. Fluidity equals that in the joint cavity. This suggested (f), the reciprocal of viscosity (-r), was plotted the possibility that pericapillary albumin because, according to Darcy's law of flow, flow concentration might be less than in the joint through a rigid macroscopic porous medium cavity here, despite the proximity of the (e.g. sand) is inversely proportional to viscosity: capillaries. This inference was supported + by K(AI) (dP/dx) = KA (dP/dx) modelling studies (see below). It synovial fluid protein concentrationappearsand COPthat

where K specific hydraulic conductivity, are not necessarily true measures of the values http://ard.bmj.com/ W A=bed area. hen the flow of albumin acting at the capillary wall at high filtration solution is divided by that of the solvent (Krebs rate. The difference may be small normally, solution here), fractional flow should equal the however. Inflow and outflow simultaneously across the synovial lining in vivo. In the above experi- *+-l Infusate ments, IA fluid was aspirated for analysis after the 15 minute infusion To our 2.0I inf period. initial on October 1, 2021 by guest. Protected copyright. Jointcavitysavt - surprise, IA albumin concentration was less 0 l' ;; \| Synovial | than that infused, Cu * ( d a being despite the prolonged c) \ Fenestrated , interstitium infusion period and a preceding triple flush C 3 c capillary .3' (fig 2). The aspirate/infusate concentration Ca) 15.5 0 *| I -Loose areolar ratio decreased with increasing infusate con- C., subsynovium centration, and increased closer to unity with -0 c) Lymphatic vessel increasingIAP. This led us to suspect that an internal circulation of fluid was equality occurring a)) ---equality (fig 2 inset). The COP of pericapillary albumin Cu 18 c)c \ increases the rate of capillary ultrafiltration into C 18_ the joint cavity by the Starling principle, 0) C-,C E66 leading to IA dilution. At the same time, 0 outflow occurs through the interstitial pathway 0) C. 0.5. . parallel to and some micrometres distant from the capillary. A net trans-synovial outflow C)0 (absorption) is recorded when interstitial . , outflow from the cavity exceeds capillary 0 ______50 150 ,,filtration_ rate into it. For this hypothesis to be 0 100 200 250 tenable, however, IAalbumin must diffiu s sufficiently rapidly the stream of capillary 2 Isotigfiltrate to maintain a filtration-enhancing COP Figure 2 Albumin concentration in joint cavity of rabbit knee relative to infused in the pericapillary region. Is this a realistic concentrration after continuous infusion for 15 minutes preceded by triple flush, plotted as a possibility? To address this, a mathematical function of infused concentration (mean, SEM). " Results at three different intra-articular model of trans-synovial flow and albumin fluid preessures(IAPs) shown. Inset shows proposed explanation, based on net trans- synovias1 consisting of two oppositely directedflows. diffusion was constructed, as follows. Fluid movement across synovium in healthyjoints: role ofsynovialfluid macromolecules 419

Tissue capillary wall conductance and interstitial hydraulic resistance were evaluated from -Subsynovi um published data. The tissue was subdivided into tiny cubic elements and protein flux across Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from each face of the element calculated by the non- linear membrane transport equation of Patlack *4 Cavity et all3 for combined diffusional and convective y (wash-along) transport. Each element was treated, in effect, as a leaky porous membrane and was allotted a finite concentration vL x -, difference, pressure drop, hydraulic resistance, partial albumin exclusion, albumin reflection Subsynovial space coefficient, and restriction to diffusion. All these material properties were related to max Unit cell interstitial biopolymer concentration (glycosaminoglycan, etc) by random fibre matrix Fenestral - i 2 theory (table) . `-20 Capillary hydraulic per- 9 plaque * Synovi iur meability was represented by discrete sites

0 * * (patch of fenestrations): the simplifying Capillary assumption of uniform capillary permeability to water (widely used in mathematical models) causes significant errors. The numerical solu- -x 0 /'~~~~o.' , max tions led to several useful insights into trans- Joint cavity synovial flow, as follows. ~~~~~* Theoretical support for simultaneous bidirectional flow across synovium. Figure 4 shows the solution for one particular set of experimental boundary conditions (IAP, COP, etc). Over the synovial surface as a whole, a net outflow from the cavity is predicted (absorption), due to flow interstitium remote the AL through from capillary; ri~~~~~ssuea element but there is also a smaller, localised inflow through the interstitium immediately overlying the capillaries, caused by capillary ultrafiltration. There is thus a local circulation of extravascular fluid and this explains much of A Cj *pi the IA dilution observed experimentally (fig 2). Figure 3 Two dimensional modelfor trans-synovialflow.8 The orientation is inv Similar bidirectional trans-synovial flow compared withfigure 2 in order to make the y axis positive. Synovial lining in sitiuerted http://ard.bmj.com/ (Tissue) is subdivided into identical unit cells (middle), and the unit cellffurther siubdivided patterns are computed at lower IAP-COP into very small blocks (Tissue element). J =flux ofwater or albumin in given direoction; combinations such as might occur in vivo. The Pi, C, = interstitial pressure andprotein concentration at definedpoint; A, L = su7rface area experiments and model thus reveal a novel and length ofeach tissue element. mechanism for synovial fluid turnover in a stationary joint, provided that the joint angle is Modelling trans-synovial flow and protein such that IAP is greater than subsynovial pressure.

transport on October 1, 2021 by guest. Protected copyright. The two dimensional model (fig 3) was based Pericapillary COP v intra-articular COP. Most on mean morphometric values for rab)bit knee of the hydraulic permeability of a synovial synovium:8 anatomical heterogenteity is capillary is concentrated in a tiny section of its neglected. The core of the model is a basic perimeter, namely the patch of fenestrations. repeating unit consisting of a half capil:lary with Filtration velocity (cm/s) is therefore orders of an orientated fenestral cluster and a neigh- magnitude greater than ifthe same flow (cm3/s) bouring tissue domain. Quantities such as occurred uniformly across the entire area of wall (cm2). High filtration velocities tend to Biophysical aspects ofinterstitial transport: the properties are governed primarily iby the wash interstitial plasma protein molecules concentration ofinterstitial biopolymers, especially glycosaminoglycan, the volumeefraction away from the outside of the fenestrations, ofwhich is 0 reducing their local concentrations (fig 4, Property Key biophysical relations* Referencelassunnption right). Despite this, a certain perifenestral protein concentration is preserved by diffusion Hydraulic conductance ki = K/hi.L K = 1-6 x 10- 18. -23 14 the ultrafiltration stream the Solute exclusion KAV = exp{-(0/rf')(rf + rj)} 4 = KAV/e`6 15 16 against along Reflection coefficient ¢i = (1-4)) 16 17 locally steepened concentration gradient Effective viscosity Th =fi(bWk3 4), rs) Particle in tub Water flow J. = k{APj-ATr} 19 emodel'8 (rather as a dye can diffuse up a sluggish Restricted diffusion Dr. = D.exp{-(O/rf2)0 5(rf + rj)} 20 stream), but the resulting perifenestral concen- Albumin transport by 7s =f(AC,, Pec) Non-linear tramsport tration is always less than IA concentration to diffusion and convection where Pec = J7(I-)/M,A equation'3 some degree. This helps explain why a given IA *Corrections for cell volume fraction, collagen fibril volume fraction, and tortuosity are not shown here, for simplicity. A = area; C = concentration; D,,, = restricted diffusion cofficien concentration of albumin had apparently less within matrix; J = transport rate; KAV = fraction of extrafibrillar water space availablle to solute; effect on fluid exchange than intracapillary k = hydraulic conductance of unit area of material; L = length of path; M = diffusiornal permeability; Pec = Peclet number; r = radius of solute or biopolymer chain; APJ = pressuri edifference albumin-a feature noted earlier. across a small element of tissue; 0 = volume fraction occupied by a component;; 4 = solute partition coefficient in extrafibrillar space; 'q = viscosity; K = specific hydraulic condi t In the above extravascular material; rT = colloid osmotic pressure due to plasma protein (oncotic pressure); (cr osmotic experiments, reflection coefficient; f = fibre (molecular chain); i = interstitium; s = solute; v = volumle. albumin was supplied via an infusion line. Can 420 Levick, McDonald

Flow pattern Albumin field Subsynovium Subsynovium 41.1 41.1 41.1 41.1 41.1 41.1 41.1 4.2 N.Zz 1.2 -7I1.2 1.1 I.~~~~~~~~~ 41.1 41.1 41.1 41.1 41.1 41.1 41.1 41.2 41.2 41.2 41.2 41.3 Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from ..t...... t...... ,,,...... 41.1 41.1 41.1 41.1 41.1 41.1 41.2 41.2 41.2 41.2 41.3 41.3 I. .% I t t t t * tt ttt ttI..t t 41.1 41.1 41.1 41.1 41.1 41.2 41.2 41.2 41.2 41.3 41.3 41.3 ..i*t t t t f t . t .. t .....t 41.3 41.4 I t t t t t t t * t t t.. . ..t t 41.1 41.1 41.1 41.1 41.1 41.2 41.2 41.2 41.2 41.3 _ *t t t tf tv t t t t tt t t t t t t I 41.141.1 41.1 41.2 41.2 41.2 41.2 41.3 41.3 41.4 I t t ittttttftf,.....,t 41.2 41.2 41.2 41.2 41.3 41.3 41.4 X t tt t t t t t tttf.tfI ' . >s41.2 41,2 41.2 41.3 41.3 41.4 t t t t t * * * .9 * .9 . t 9 Capillary lumen 41. 41.2 41.2 41.3 41.4 Capillary lumen )tt t' ''''' ' tt' 41.6 41.1 41.2 41.3 41.4 , f t t ? *~ t t t t tf .f t t tI t t Protein con 53 g/l j 4.l 41.6 41.1 41.3 41.4 ' i// / t ~t t t t t. t t. t t t t t '.1.9 46.4 46.6 41.1 41.3 41.5 I \ lt?Ptr ,t ...... e -1,439.34.3 4.6 41.1 41.4 41.6 O'tZN~ortsf ' f * * t t t . . I 339.5 46.1 4.5 4.6 41.2 41.5 41.? zI****WbXs-***I'*t,X t *99 41.6 41.5 41.3 41.6 46.? 46.5 4.6 46.9 41.2 41.4 41.6 41.8 41.? 41.6 41.5 41.3 41.2 41.1 41.1 41.3 41.5 41.? 41.6 42.1 41.9 41.8 41.6 41.1 41.6 41.5 41.6 41.? 41.6 41.9 42.1 42.2 42.1 42.1 42.1 42.6 42.1 41.9 42.6 42.1 42.1 42.2 42.4 42.5 X *v* \ b ..X, * I- O p P.9909999ttt9?9 42.4 42.4 42.4 42.4 42.3 42.3 42.4 42.4 42.5 42.5 42.6 42.? 42.6 42.8 42.? 42.? 42.? 42.? 42.? 42.8 42.6 4.3 42.9 43.6

I ___ 43.1 43.1 43.1 43.1 43.1 43.1 43.1 43.1 43.2 43.2 43.3 43.3 y = 43.5 43.5 43.5 43.5 43.5 43.5 43.5 43.5 43.1 43.6 3.6 436 x= ° tt \4 Joint cavitY 7 7 Joint cavity Protein concn 44 g/l

Figure 4 Predicted bidirectional pattern offlow across synovial interface withjoint cavity (left) and corresponding interstitial albumin concentrations (right: x axis expandedfor clarity).8 Dark bar on capillary wall shows site ofpatch of fenestrations. Boundary conditions: intra-articular pressure (AP) 6 cm H20; subsynovial pressure atmospheric; intra- articular albumin concentration 44 gIl; capillary pressure = plasma colloid osmotic pressure. Left: Thin arrows are flow vectors;flow into cavity 0-2 uik/min;flow out ofcavity 10 7 ,ul/min; capillaryfiltrationfraction 0 044. Larger inflow! outflow ratios occur upon increasing intra-articular albumin concentration or decreasing IAP. Right: Numbers represent albumin concentration in available, non-excluded interstitial water. Note the substantial dilution immediately around the filteringfenestral patch (thickened line). Over the capillary, albumin is diffusing out of the cavity against the inward filtration stream.

similar perifenestral gradients of plasma mental estimates of synovial hydraulic resis- protein exist normally? To answer this, we tance.8 22 Recently, the very small (mg) must consider the normal route(s) of entry of quantities of synovium in a rabbit knee have plasma protein into the synovial interstitium been analysed directly and quantitatively for and synovial fluid. The fenestral membrane certain biopolymers. The net concentration of itself has a high protein reflection coefficient chondroitin sulphate, heparan sulphate, and

(08 for albumin3) and it is generally accepted hyaluronan is -3 9 mg/ml of extrafibrillar http://ard.bmj.com/ that much of the extravascular plasma protein space (recalculated). This is substantially less mass, especially for the larger molecular than the theoretical biopolymer content of species, enters interstitium via the capillary 13-9 mg/ml. The difference is partly large pore system, which comprises the explained by the existence of unassayed bio- vesicular transport system, continuous large polymers (for example glycoproteins, proteo- pores or both.2' Thus much protein enters glycan core-present on immunohisto-

interstitium at regions spatially separate from chemistry); indeed, similar discrepancies occur on October 1, 2021 by guest. Protected copyright. the site of fenestral filtration, and perifenestral in other tissues if such components are gradients analogous to those in figure 4 may neglected.22 The attempt to match biochemical therefore occur normally. As noted above, this composition and physiological properties would imply that intra-articular COP is only an in synovium is thus incomplete at present. approximation to the COP around the capillary fenestrations. The goodness of this approximation varies with filtration rate. Trans-synovial flow in absence of The model predictions for IA dilution and capillary filtration: a viscosity anomaly fractional flow8 broadly fitted the scattered Killing an animal abolishes capillary filtration, results of figures 1 and 2. The variability of Q. and this provided direct evidence that the in rabbit joints is a consistent feature in our deviation of fractional flow below the Darcy experience. line in vivo (filled squares in figure 1) was caused by capillary filtration. When the animal was killed the low net outflows induced by Relation between synovial lining albumin increased to much greater values permeability and biopolymer (open squares in figure 1). Thus depression of concentration trans-synovial outflow below the Darcy line in The net concentration of biopolymers (glycos- vivo is caused by microvascular filtration. aminoglycans, proteoglycans and glyco- Trans-synovial flows of albumin solution proteins) in synovial interstitium is a key factor post mortem lay above the Darcy line, rather governing hydraulic resistance and permea- than along it as one might expect upon first bility to macromolecules. For the model, a consideration. Experimental design may have biopolymer concentration of -13-9 mg/ml of exaggerated the deviation,24 but there is a extrafibrillar space (interstitial space excluding fundamental theoretical reason why the fluidity collagen fibrils) was deduced from experi- of interstitial fluid should be different from that Fluid movement across synovium in healthyjoints: role ofsynovialfluid macromolecules 421

ofthe IA solution. The reason is that interstitial showed that the anomalous viscosity of a biopolymers exclude albumin from a fraction suspension of particles (radius r) is given by: of the water space.25 26Around any biopolymer = 1/iapparent V-{(Rcore/Rtube)4( 1_(1/,lcore))} Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from chain there is an annular space into which the centre of mass of a large globular solute cannot where 'lapparent and lqcore are the apparent relative penetrate, because of the solute's finite radius viscosity and the bulk viscosity of the liquid in (steric exclusion), but into which the much the non-excluded core, respectively, solvent smaller water molecule has free access. The relative viscosity is 1, and the radius ofthe non- greater the glycosaminoglycan concentration, excluded region, Rcore, is Rtube -0-735r. In the greater is this excluded volume fraction5 25 extrapolating this to a biopolymer matrix, the (table). Since the trans-synovial pressure equivalent tube radius of the irregular channels gradient must drive some water and electro- between the molecular fibres is calculated as lytes through the albumin-excluded space, in the tube radius that produces the same addition to driving albumin solution through fractional steric exclusion of particle (albumin the albumin-available space, the effective molecule here) as the matrix. The resulting relative fluidity is a flow-weighted average theoretical curves deviate above the Darcy line between that of the saline solution (relative (line of equality, figure 1), but less so than the fluidity 1 0) and the albumin solution in the observed results. Possible reasons for the non-excluded space (fluidity < 1 0). Thus the discrepancy between experiment and theory effective interstitial fluidity is greater (closer to have been assessed previously.8 1 -0) than that ofthe feeding phase, as seen post mortem (fig 1). In other words, steric exclusion of interstitial plasma proteins reduces the Effect ofintra-articular hyaluronan on effective viscosity of the permeating fluid. This trans-synovial flow Qs effect was obscured in vivo by the large, COP- The physical properties ofhyaluronan (3-4 g/l; induced changes in capillary filtration, except mol.wt > 106) contrast strongly with those of perhaps at low albumin concentrations. albumin (fig 5). The COP of hyaluronan is slight, but its effect on viscosity is great. The effect of hyaluronan on Q, is modest at low Anomalous viscosity in narrow channels IAP, but remarkable at high IAP (fig 6). In the A first-approximation treatment of anomalous experiment illustrated, a fixed concentration of viscosity was developed by analogy with hyaluronan (3 or 6 g/l) was infused and IAP laminar flow of particulate suspensions (such was increased every 15 to 20 minutes to as blood) down narrow cylindrical tubes; here, determine the IAP-Q. relation. The control steric exclusion at the walls produces a similar curve with Krebs solution shows the well anomalous low viscosity (Fahreus-Lindquist known increase in slope with IAP, indicating a effect27). Flow through the spaces between decrease in hydraulic resistance with increasing

cylindricalfibres exhibits Poiseuille-like laminar IAP (Edlund curve"). This is attributable http://ard.bmj.com/ patterns,28 because the channel is in effect an partly to stretching of the lining with pressure29 'inside-out' tube. For a narrow cylindrical tube and partly to dilution of interstitial matrix (radius Rwbe) with laminar flow, Whitmore'8 biopolymers.23 Albumin reduces absorption rate in vivo at any given pressure (see above), 80 - 80 on October 1, 2021 by guest. Protected copyright.

Tr= nm 60- v 60-100 HYALURONAN 60 E 0 i-. U) i a) 40 - ALBUMIN . r= 3.6 nm 40 t- .I 0 Co . 0) I 5 g/l hyaluronan t 0CoUn 20- 20 I

I 60 g/l albumin I

- . * *0 ,- - 0 10 20 0 100 200 300 IAP (cm H20) Colloid osmotic pressure (cm H20) Figure 6 Effect of intra-articular pressure on net trans- Figure 5 Effects ofhyaluronan and albumin on colloid osmotic pressure (at 35°C and synovialflow out ofjoint cavity (absorption) for Krebs pH 7 4) and viscosity (at 35°C and wall shear stress of29 dyn cm-2). Points represent solution (A), albumin 110 gIl solution (S) and increasing concentrations. Molecular dimension are compared in the inset, roughly to scale. hyaluronan 6 g/l solution (U) in three individual rabbit r = Molecular radius. knees. 422 Levick, McDonald

I 33% and 66% of control flow-a surprisingly .l modest effect considering that hyaluronan Joir cavity reduces the bulk fluidity to 6-10% of control ... N mrTole' fluidity.30 At greater IAPs, trans-synovial flow Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from 7 .- decreases increasingly below the control curve I...... I>1 and begins to plateau. In several joints, including that illustrated, further increases in irterstitiwr IAP then result, remarkably, in a slight decrease in the steady state QO. A reduction in steady I -Mv c ; - Basemenit rm state flow upon increasing IAP indicates an resistance to flow. If En::adothe11AM increasing subsynovial MirtevVi'Infhbof Glycocalvx pressure does not change substantially and is close to atmospheric pressure (there are grounds for such an assumption), the IAP Plasma required to drive unit flow across the synovial lining is an index of synovial lining hydraulic Col.)lagen I resistance, and this is found to increase Figure 7 Sketch showing composition ofpathway between joint cavity andiplasma, progressively as a function of IAP. With Krebs comprisingfibrous matrices ofsynovial interstitium, capillary basement mennbrane (m.) and solution, by contrast, resistance decreases with endothelial glycocalyx. The voluminous hyaluronan molecules ofsynovialfl'uid are thought increasing IAP, as has been known since the to permeate the lining lessfreely than water, leading to an accumulation of/ iyaluronan at work of Edlund. l 1 the interface when the intra-articular pressure is increased.

but it does not alter the basic shapee of the Hypothesis: hyaluronan molecular pressure-flow relation (middle curve, fig 6). filtercake atthe synovial surface? Hyaluronan, by contrast, alters the bas,ic shape A phenomenon analogous to that shown in of the pressure-flow relation. At lovw IAPs, figure 6 occurs in vitro;3' when a hyaluronan flows are reduced by hyaluronan to 1between solution is driven through a Millipore membrane (pore width 0 45 ,um), the resistance to flow increases, because a filtercake of hyal- A uronan (or 'concentration polarisation layer') Hydraulic conductivity of GAG matrix (x 10-12 cm4 s-i dyn-1 forms at the surface. The hydrated, mutually 1500 500 200 100 50 30 20 10 5 3 2 1 0.6 overlapping molecular domains of the vast hyaluronan molecules are too large to pass easily through the pores, so they accumulate 1*0 just upstream and create a high resistance to the passage of water.32 As the equivalent tube radius within synovial interstitium is much less than 045 ptm," 22 hyaluronan molecules may http://ard.bmj.com/ CD r- filtercake at the synovial surface also (fig 7). A a preliminary study using ruthenium red to stain 80. polyanionic material supports this idea: 0 electron micrographs showed an accumulation a) of ruthenium red positive material just within o0 the superficial interstitium and at the surface after infusion of hyaluronan solution under on October 1, 2021 by guest. Protected copyright. 1 5 10 50 30 pressure.30 Hyaluronan can thus affect the hydraulic GAG concentration (mg/ml available water) resistance of the joint cavity lining, and by B acting as a 'dynamic waterproofing' during 1.0 - Albumin: a = 3.55 nm periods of joint flexion and increased IAP it _0a may help the synovial lining retain the vital U --- alb synovial fluid. A great deal more work remains 0 to be done on this phenomenon, such as IgG: a= 5.6 nm exploring the importance of hyaluronan con- .0 0.5 - 0-c:-o - G centration and chain which are - - length, 9-IgG subnormal in rheumatoid arthritis. en- 2M:a = 9.1 nm 1-3 ---- C2 MG Is synovium freely and equally permeable If 0 50 100 to all macromolecules? There is as to whether the Flow (RI/min) controversy synovial lining matrix is sufficiently dense to impede the Figure 8 A: Relation between reflection coefficient (a) to a spherical macromolecule of egress ofIA macromolecules selectively (on the a and the concentration ml radius ofglycosaminoglycan (GAG) per ofextrafibrillar basis of size) as implied above. The limited space, modelled as random rigid cylindrical rods of radius 0-6 nm (see table). Dark bar shows currently estimated range for biopolymer concentration in rabbit knee synovium evidence has been reviewed previously." (see text). Small arrow shows estimated conductivity to saline at 35°C. B: Theoretical Briefly, size selectivity has not been detected in dependence ofsieving ratio (outflow concentration/inflow concentration, e.g. lymph the human knee upon comparing large and concentration/intra-articular concentration) on flow, for a matrix with aflow-independent reflection coefficient. The ratio is close to 1 for all molecules at very slowflows. Adapted small globular plasma proteins, but there is from Levick. 33 a2MG = c2-Macroglobulin. some evidence for size based selectivity in the Fluid movement across synovium in healthyjoints: role ofsynovialfluid macromolecules 423

non-inflamed rabbit knee comparing albumin 7 Levick J R. A two-dimensional morphometry-based model of interstitial and transcapillary flow in rabbit synovium. with proteoglycan or, as here, albumin with Exp Physiol 1991; 76: 905-21. hyaluronan. Size selectivity is expected in a 8 Levick J R. An analysis of the interaction between extravascular plasma protein, interstitial flow and capillary Ann Rheum Dis: first published as 10.1136/ard.54.5.417 on 1 May 1995. Downloaded from polymer matrix on the theoretical grounds filtration; application to synovium. Microvasc Res 1994; that steric exclusion is closely linked to the 47: 90-125. 9 Simkin P A, Benedict R S. Iodide and albumin kinetics in generation of reflection coefficients (table). normal canine wrists and knees. Arthritis Rheum 1990; 33: Figure 8A shows the theoretical relation 73-9. 10 Jensen L T, Henriksen J H, Olesen H P, Risteli J, between the synovial reflection coefficient to Lorenzen I. Lymphatic clearance of synovial fluid in globular macromolecules and biopolymer con- conscious pigs; the aminoterminal propeptide of type III procollagen. EurJ Clin Invest 1993; 23: 778-84. centration (modelled as randomly distributed 11 Edlund T. Studies on the absorption of colloids and fluid cylinders of radius 0-6 nm) or hydraulic from rabbit knee joints. Acta Physiol Scand 1949; 18 (suppl 62): 1-108. permeability (related to concentration22 12 Levick J R, Knight A D. Interaction of plasma colloid (table)). In the region of interest for synovium osmotic pressure and joint fluid pressure across the endothelium-synovium layer; significance ofextravascular (dark bar and arrow in figure 8A), the resistance. Microvasc Res 1988; 35: 109-21. reflection coefficients of the larger plasma 13 Patlack C S, Goldstein D A, Hoffman J F. The flow ofsolute and solvent across a two-membrane system. I Theor Biol proteins are unlikely to be zero and may exceed 1963; 5: 426-42. 0-1. While the maximum molecular separation 14 Comper W D, Zamparo 0. Hydraulic conductivity of polymer matrices. Biophys Chem 1989; 34: 127-35. is 1-cF (i.e. the smallest effluent:input ratio is 15 Ogston A G. On the interaction of solute molecules with 1-c, for example < 0 9 for or > 0 1), the actual porous networks. J Physical Chem 1970; 74: 668-9. 16 Curry F E. Mechanics and thermodynamics of separation depends equally on trans-synovial transcapillary exchange. In: Renkin E M, Michel C C, fluid velocities: as filtration velocity is reduced, eds. Handbook ofPhysiology, Section 2, The Cardiovascular System, Volume IV, The Microcirculation, ch 8. Bethesda: diffusion increasingly offsets molecular sieving American Physiology Society, 1984; 309-74. and increases the ratio towards unity (fig 8B). 17 Anderson J L, Malone D M. Mechanism of osmotic flow in porous membranes. Biophys3 1974; 14: 957-82. The view that synovium is less permeable to 18 Whitmore R L. Rheology ofthe circulation. Oxford: Pergamon hyaluronan than to albumin implies macro- Press, 1968. 19 Kedem 0, Katchalsky A. Thermodynamic analysis of the molecular selectivity. In support of this, the permeability of biological membranes to non-electrolytes. bulk turnover of IA water and protein is Biochim BiophysActa 1958; 27: 229-45. 20 Ogston A G, Preston B N, Wells J D. On the transport of estimated to be an order of magnitude faster compact particles through solutions of chain polymers. than the turnover of IA hyaluronan. The turn- Proc R Soc A 1973; 333: 297-316. 21 Rippe B, Haraldsson B. Transport of macromolecules over time for synovial fluid volume and protein across microvascular walls: the two-pore theory. Physiol is estimated to be about one hour in rabbit and Rev 1994; 74: 163-219. 22 Levick J R. Flow through interstitium and other fibrous normal human knees,5 while that for hyal- matrices. QJExpPhysiol 1987; 72: 409-38. uronan in the rabbit shoulder is 20-28 hours.34 23 Mason R M, Price F M, Levick J R. A quantitative investigation of the glycosaminoglycans of the synovium. Similarly, in the rabbit knee hyaluronan has a Trans Orthop Res Soc 1994; 19: 403. half life of 27-32 hours at normal volume,35 or 24 Levick J R, McDonald J N. Viscous and osmotically mediated changes of interstitial flow induced by extra- between 13 hours (mol.wt 6 X 106) and 10 vascular albumin in synovium. Microvasc Res 1994; 47: hours (mol.wt 0-9 x 106) in volume expanded 68-89. 25 Ogston A G. The spaces in a uniformly random suspension knees.36 Large differences between hyaluronan offibres. Trans Faraday Soc 1958; 54: 1754-7. http://ard.bmj.com/ and water/albumin turnover times seem to 26 Aukland K, Reed R K. Interstitial-lymphatic mechanisms in the control ofextracellular fluid volume. PhysiolRev 1993; imply selective retention of hyaluronan within 73: 1-78. the joint cavity and therefore a selective 27 Goldsmith H L, Cokelet G R, Gaehtgens P. Robin Fahreus; evolution of his concepts in cardiovascular physiology. capacity for the lining tissue. Am3tPhysiol 1989; 257: H1005-15. 28 Happel J, Brenner H. Low Reynolds number hydrodynamics. This work was supported by grants from the Arthritis and Englewood Cliffs: Prentice-Hall, 1965; 392-404. Rheumatism Council and the Wellcome Trust (031 333/Z/90/A, 29 Levick J R, McDonald J N. Ultrastructure of transport

039033/Z/93/Z/l -27). pathways in stressed synovium of the knee in on October 1, 2021 by guest. Protected copyright. anaesthetized rabbits. J Physiol 1989; 419: 493-508. 30 McDonald J N, Levick J R. Hyaluronan reduces fluid escape 1 Stevens C R, Blake D R, Merry P, Revell P A, Levick J R. rate from joints disparately from its effect on fluidity. Exp A comparative study by morphometry of the micro- Physiol 1994; 79: 103-6. vasculature in normal and rheumatoid synovium. Arthritis 31 Parker K H, Wmlove C P. The macromolecular basis of the Rheum 1991; 34: 1508-13. hydraulic conductivity of the arterial wall. Biorheology 2 Levick J R, Smaje L H. An analysis of the permeability of 1984; 21: 181-96. a fenestra. Microvasc Res 1987; 33: 233-56. 32 Johnson M, Kamm R, Ethier C R, Pedley T. Scaling laws 3 Knight A D, Levick J R, McDonald J N. Relation between and the effect of concentration polarization on the trans-synovial flow and plasma colloid osmotic pressure, permeability of hyaluronic acid. Physicochemical with an estimation of the albumin reflection coefficient in Hydrodynamics 1987; 9: 427-41. the rabbit knee. QJ Exp Physiol 1988; 73: 47-66. 33 Levick J R. Synovial fluid; determinants of volume and 4 Levick J R. Blood flow and mass transport in synovial joints. material concentration. In: Kuettner K E, Hascall V C, In: Renkin E M, Michel C C, eds. Handbook ofPhysiology. Schleyerbach R, eds. Articular cartilage and osteoarthritis. Section 2, The Cardiovascular System, Volume IV, The New York: Raven Press, 1992; 529-41. Microcirculation, ch 19, Bethesda: American Physiology 34 Knox P, Levick J R, McDonald J N. Synovial fluid-its Society, 1984; 917-47. mass, macromolecular content and pressure in major limb 5 Levick J R. Synovial fluid and trans-synovial flow in joints of the rabbit. QJ Exp Physiol 1988; 73: 33-46. stationary and moving joints. In: Helminen H, Kiviranta I, 35 Denlinger J L. Metabolism of sodium hyaluronate in Tammi M, Saamaren A M, Paukonnen K, Jurvelin J, eds. articular and ocular tissues [dissertation]. Lille: Universite J'oint Loading: Biology and Health of Articular Structures. de Sciences et Techniques de Lille, 1982. Bristol: Wright & Sons, 1987; 149-86. 36 Brown T J, Laurent U B G, Fraser J R E. Turnover of 6 Henderson B, Edwards J C W. The synovial lining. London: hyaluronan in synovial joints. Exp Physiol 1991; 76: Chapman & Hall, 1987. 125-34.