Factors influencing the oxygen concentration gradient from the synovial surface of articular cartilage to the cartilage–bone interface: A modeling study




There is very little information on the gradients of oxygen concentration from the synovial surface to the subchondral bone in articular cartilage. Cartilage is usually regarded as hypoxic, even though cellular metabolism is inhibited at low oxygen concentrations. We therefore measured rates of cellular consumption of oxygen and used these rates to calculate oxygen tension profiles across articular cartilage.


The rate of oxygen consumption by bovine articular chondrocytes was measured in vitro, either in intact cartilage slices or in isolated chondrocytes. The oxygen tension profile across articular cartilage was predicted by solving a 1-dimensional reaction–diffusion equation. The effect of synovial fluid oxygen concentration, cell density, cartilage thickness, and influx of oxygen from the subchondral bone on the oxygen profile in the tissue was examined.


Oxygen consumption rates were relatively independent of oxygen tension at high oxygen tensions (5–21%), where they were ∼10 nmoles/106 cells/hour for both isolated chondrocytes and for cartilage slices. Below 5% oxygen, the rate fell in an oxygen tension–dependent manner. Analysis showed that the oxygen profile across cartilage fell steeply in all but the thinnest cartilage samples but only fell to ∼1% for low oxygen tensions in synovial fluid, with no supply from the subchondral bone.


The oxygen tension in normal cartilage is not likely to fall to 1% except under abnormal conditions. Oxygen tensions within cartilage are strongly affected by a number of factors, including oxygen concentrations in synovial fluid, cartilage thickness, cell density, and cellular oxygen consumption rates. Supply from the subchondral bone may be of particular importance.

Articular cartilage is distinct from other tissues in that its cells, the articular chondrocytes, consume much less oxygen than do most other cell types (1). Chondrocytes derive energy mainly by the Embden-Meyerhof-Parnas pathway of glycolysis, which does not require oxygen (2, 3). How they use oxygen is still unclear, but the actual levels of oxygen that are present in cartilage affect cellular function. Although chondrocytes can survive for many days without oxygen (4), at low oxygen tensions (<5%), rates of matrix synthesis and energy production fall steeply in an oxygen tension–dependent manner (2, 4).

Little is known about oxygen tensions within articular cartilage. Articular cartilage is avascular, so all nutrients, including oxygen, are supplied to the chondrocytes by diffusion from surrounding sources, principally, from the synovial fluid. Oxygen tension falls with distance from the cartilage surface, and the gradient is dependent on the balance between the rate of oxygen transport through cartilage and the rate of consumption by the cells. The only reported measurements have been performed in vitro (5, 6), where it was found that oxygen tensions fell steeply with distance from the surface of cartilage plugs or constructs to levels that were 1–2.5% in the center. Based on these measurements, it is commonly asserted that cartilage is hypoxic and that oxygen tensions in the deep zone of articular cartilage are ∼1% (4, 7). However, in view of the inhibitory effect of low oxygen tensions on chondrocyte metabolism and matrix synthesis (2, 4), it would be surprising if such low oxygen tensions exist in vivo.

As a first step in determining oxygen levels in vivo, we calculated the oxygen levels through various cartilage samples. Since transport of small solutes such as oxygen through cartilage appears to occur mainly by diffusion (8, 9), we used Fick's law to evaluate oxygen tension gradients across cartilage; a similar approach has previously been used to estimate oxygen profiles across the growth plate (10, 11), the intervertebral disc (12, 13), and other avascular tissues, such as the cornea (14, 15). For this analysis, we first measured the effect of oxygen tension on the oxygen consumption rate, since such data are not available. We then examined how the synovial fluid concentration, cell density, cell density profile, cartilage thickness, and transport from the subchondral bone might affect oxygen levels in cartilage. The results of these calculations indicate that oxygen tensions in normal cartilage are unlikely to fall as low as 1%.



General chemicals, Dulbecco's modified Eagle's medium (DMEM; catalog no. D 5523), which contained L-glutamine and 1,000 mg/liter of glucose (without sodium bicarbonate) and was buffered with 50 mM HEPES, and type I collagenase were obtained from Sigma (Poole, UK). Antibiotic/antimycotic solution was obtained from Invitrogen (Paisley, UK). Zero-oxygen solution (sodium sulfite and cobalt chloride hexahydrate solution) was obtained from Oakton Instruments (Vernon Hills, IL). The Clark-type oxygen electrode and oxygen respiration chamber were obtained from Strathkelvin Instruments (Glasgow, UK).

Extraction of chondrocytes from bovine cartilage

All experimental measurements reported here were obtained from cartilage slices harvested from the metacarpophalangeal joints of 18–24-month-old bovine steers. Chondrocytes were isolated by 24-hour digestion with 0.1% collagenase in DMEM containing 1% antibiotics and antimycotics as described previously (4). Cells were counted using a hemocytometer, and their viability was determined using trypan blue exclusion. Suspensions with a viability <90% were discarded. All procedures were carried out aseptically.

Oxygen consumption chamber

The oxygen consumption rate was measured in a stirred, water-jacketed chamber maintained at 37°C and sealed by a tight-fitting, gas-tight plunger into which the Clark-type electrode was inserted. A thin groove on the side of the plunger allowed the escape of gas bubbles. To overcome oxygen leakage at low oxygen tensions, gases at appropriately low oxygen tensions (0–10%) were introduced into the interstices above the seal by a medical catheter to minimize differences between oxygen tensions inside and outside the chamber.

The oxygen electrode was calibrated before every experiment using a 2-point calibration with zero-oxygen solution and air-saturated DMEM. Oxygen consumption of the electrode at different oxygen tensions was determined by inserting the electrode into 1 ml of DMEM bubbled with gases of known oxygen concentrations and was found to be linear with oxygen tension and low compared with oxygen consumption by the cells (data not shown); this rate was subtracted from the total consumption rate.

Measurement of the oxygen consumption rate

Cell suspension

Two milliliters of a cell suspension (∼10 million cells/ml), supplemented with glucose to 3.0 gm/liter to prevent glucose exhaustion, was transferred into the respiration chamber. After sealing, the oxygen tension in the chamber was continuously recorded by a computer. At the cell densities used, the oxygen tension dropped to 0 in 1–2 hours. At the end of the experiment, the cells were removed, the cell density checked, and the pH measured to ensure that it remained above pH 7.0 (acid pH values may affect oxygen consumption rates [data not shown]). The oxygen consumption rate was calculated from the slope of the oxygen tension–time curve and reported as nanomoles per 106 cells per hour.

Cartilage slices

In order to ensure that oxygen consumption rates obtained using isolated chondrocytes represented rates in cartilage, the oxygen consumption rates were measured using small slices of cartilage at 21% oxygen tension, and the results were compared with those obtained from the cell suspension.

Freshly harvested cartilage, taken within 2–3 hours of slaughter, was chopped into ∼1-mm cubes. A random sample of the pooled cartilage pieces was weighed (∼0.5 gm) and transferred into the respiration chamber with 2 ml of DMEM that had previously been stirred in air for 10 minutes. The oxygen consumption rate was then measured as described above for the cell suspension. At the end of the experiment, all the slices were removed from the chamber. In order to determine the number of viable cells, the slices were incubated at 37°C in 20 ml of DMEM containing 1% antibiotic/antimycotic solution and 1.5% collagenase, with continuous shaking. After 24 hours, the cartilage pieces were completely digested, and the number of cells was counted. Consumption rates were then normalized to the number of viable cells.

In addition, the volume and weight of cartilage slices were measured to determine cartilage density. The cell density of the cartilage was calculated by dividing the cell number by the volume of cartilage. The water content of cartilage slices was obtained by weighing the cartilage wet and then after drying to constant weight at 70°C (for ∼24 hours). These data were used in the model.

Oxygen consumption rate as a function of oxygen concentration

The oxygen consumption rate of cartilaginous cells is thought to follow the Michaelis-Menten equation (13), equation 1:

equation image

where r is the oxygen consumption rate (nmoles/106 cells/hour), Vmax is the maximum oxygen consumption rate (nmoles/106 cells/hour), Km is the Michaelis-Menten constant (%), and C is the oxygen concentration. The values for Vmax and Km were determined by inverting equation 1 and using linear regression on the measured rates of oxygen consumption versus oxygen concentration.

Modeling oxygen gradients in cartilage

The articular cartilage was treated as a flat slab. Thus, oxygen diffusion is 1-dimensional. Three assumptions were made to simplify the analysis: 1) The synovial fluid was well mixed and of constant oxygen concentration, and hence, any oxygen gradients in the fluid were negligible. 2) Oxygen transport was at steady state; thus, the effects of mechanical loading and of cartilage deformation on oxygen transport were ignored. 3) Gradients from the cell surface to the bulk tissue were negligible (confirmed by initial calculations).

Hence, equation 2 can be used to describe the steady-state oxygen concentration profile:

equation image

where D is the oxygen diffusivity, X is the distance from the surface of the cartilage, and ρ is the cell density.

One boundary condition for equation 2 is that at X = 0 (synovial fluid), C = Csyn (oxygen concentration in synovial fluid). The other boundary condition, that X = δ (subchondral bone), is more difficult to define because the structure is complicated and its geometry and permeability are poorly defined. To overcome this, we made the simplifying assumption that oxygen tension at X = δ is uniform. The proportion of oxygen supplied from the subchondral bone was then taken as a variable and used as a boundary condition for the calculation. For example, if no oxygen is supplied by the subchondral bone, the slope of the oxygen gradient on the subchondral side:

equation image

would be 0. If 30% of the oxygen consumed by the cartilage is from the subchondral bone, then:

equation image

The equation was solved numerically by the finite element method, with the total thickness (d; in mm) divided into 40 grids, each of similar magnitude to the size of chondrocytes, and solved with the software Matlab 6 (MathWorks, Natick, MA).

Since we could not find any reported measurements of oxygen diffusivity in cartilage, we assumed that oxygen was retarded to a similar extent as other small solutes and was ∼40% of that in water at 37°C (16), giving a diffusivity (D) of 4.6 mm2/hour, similar to that found in the nucleus of the intervertebral disc (13). The solubility of oxygen in cartilage water at 37°C was taken as 1.0268 × 10–5 moles/% oxygen/liter (13), and the water content of cartilage was measured and found to be 80% (volume/volume) as described below.

Several different cartilage types were analyzed (Table 1). Where no other data were available, we used data obtained here for the oxygen consumption rates of bovine articular cartilage. The effect of synovial fluid concentration (Table 2), cell density, cartilage thickness, and the proportion of oxygen diffusing from the subchondral bone on the oxygen tension distribution across cartilage was examined.

Table 1. Cartilage thickness, cell density, and oxygen consumption rate of different cartilage types used for calculations of oxygen profiles*
Species and cartilage typeThickness, mmCell density, ×106/mlOxygen consumption rate, ml O2/ml tissue/hourReference
  • *

    Data on cartilage thickness and cell density are from references19 and43, except for the bovine ankle data, which are from the present study. For the conversion of units, tissue water content was taken as 75% and density as 1.1 gm/ml.

  • Data used for calculations were obtained from Figure 1.

Rabbit knee0.211880.04144
Dog knee0.6744.40.01545
Bovine ankle1.028.7Present study
Bovine knee1.719.8Present study
Human knee2.314.1Present study
Pig, young    
 10 mmoles of glucoseNANA0.0113
 5 mmoles of glucoseNANA0.0123
 0 mmoles of glucoseNANA0.0263
Figure 1.

Oxygen consumption as a function of oxygen tension. Results were derived from 4 independent measurements on cell suspensions obtained from 4 different bovine feet. The results were fitted by the Michaelis-Menten equation to give a maximum oxygen consumption rate of 10.81 nmoles/106 cells/hour and a Michaelis-Menten constant of 2.43% (R2 = 0.974).

Table 2. Oxygen tension in synovial fluid from the knee joints of rabbits and humans with various conditions
SourceCondition/diagnosisPO2, %Reference
RabbitAcute inflammation2.7–6.847
HumanRheumatoid arthritis0–6.418
HumanTraumatic exudates5.5–1148
HumanRheumatoid arthritis0–11.848
HumanVariety of joint diseases1–10.349


Oxygen consumption rate in cell suspensions

The oxygen consumption rates measured are shown in Figure 1. The oxygen consumption rate was relatively constant and fairly independent of oxygen tension until the oxygen tension fell below ∼5% (1% oxygen = ∼1 kPa = 9.5 μmoles/liter of solution). The mean ± SD oxygen consumption rate was 10.6 ± 1.4 nmoles/106 cells/hour at 21% oxygen and 8.6 ± 0.9 nmoles/106 cells/hour at 10% oxygen. Below 5% oxygen, the rate fell in a concentration-dependent manner.

After regression, the 2 parameters of the Michaelis-Menten equation were determined and gave the following values: Vmax = 10.81 nmoles/106 cells/hour and Km = 2.43% (R2 = 0.974).

Oxygen consumption rate in cartilage slices

The mean ± SD cartilage density was 1.11 ± 0.06 gm/ml (n = 9 samples), and the mean ± SD water content was 73 ± 1% weight/weight (n = 4) and 81% volume/volume. The mean ± SD density of viable cells measured in cartilage slices was 28.7 ± 1.5 million/ml of wet tissue (n = 5) and was similar to that previously estimated from measurements of DNA (17). The oxygen consumption rate measured in a suspension of cartilage pieces was 10.6 ± 1.0 nmoles/106 cells/hour (n = 4) at 21% oxygen. There was no significant difference between the oxygen consumption rate per cell measured in the cartilage slices and that measured in the cell suspension.

Modeling of oxygen tension profiles across cartilage

Profiles of oxygen tension across cartilage were calculated using equation 2. We evaluated the effect of synovial fluid concentration, cell density, cartilage thickness, proportion of oxygen supplied by the subchondral route, oxygen consumption rate, and cell distribution profile on the oxygen tension distribution.

Effect of oxygen tension in synovial fluid

Oxygen tensions measured in synovial fluid from rabbit and human knee joints have been reported to vary from 0 to ∼10% (see Table 2). We varied the synovial fluid concentrations from 3% to 10% to examine the effects of these concentrations on the oxygen profile. The results were calculated from equation 2 for the human knee using values for cartilage thickness and cell density from Table 1 and using consumption rates for bovine cartilage taken from Figure 1. It was assumed that cell density was uniform and that all oxygen was supplied to the cartilage from the synovial fluid. The synovial fluid concentration was assumed to be constant.

In all cases, oxygen tension fell with increasing distance from the synovial surface and was lowest in the deep zone (Figure 2). Oxygen tension throughout the tissue depended on the concentration in the synovial fluid, increasing as the synovial fluid concentration increased. With an increase in synovial fluid oxygen tension from 3% to 10%, the calculated oxygen tension at the subchondral plate rose from 0.5% to 3.4%.

Figure 2.

Effect of synovial fluid oxygen tension and oxygen concentrations in the synovial fluid (Csyn) of 3%, 6%, 8%, and 10% on oxygen profiles through the cartilage thickness. Results are for human knee cartilage (thickness 2.3 mm, uniform cell density 14.1 million/ml, as shown in Table 1). Oxygen consumption rates were taken from the fitted curve shown in Figure 1.

To provide information on the lower limits of oxygen tension that are likely to occur normally, we based our further calculations on 6% oxygen in the synovial fluid, since below this value, intraarticular acidosis occurs (18).

Effect of cartilage thickness, cell density, and oxygen consumption rate.

Cartilage thickness and cell density vary from species to species and from joint to joint. Stockwell (19) showed that, in general, thinner cartilage has higher cell density than thick cartilage and found an inverse relationship between the total thickness (d; in mm) and cell density (ρ), that is, ρ = 27.9 × d–0.88 million cells/ml. The oxygen consumption rate per cell also appeared to be highest in thin cartilage. If we consider these 2 factors and calculate oxygen tension gradients across cartilage of different thicknesses, we can estimate how oxygen tension profiles may vary between cartilage of different species.

Figure 3A shows the estimated oxygen profiles across species (rabbit, dog, bovine, and human) and across cartilage types (human femoral condylar [knee] and bovine metacarpophalangeal [ankle]), as calculated from equation 2, assuming that all oxygen was supplied by the synovial fluid, where the oxygen concentration was constant at 6%. Values for cartilage thickness and cell density (assumed to be uniform) were taken from the literature (Table 1). For dog and rabbit values, the oxygen consumption rate was fitted to the rates in Table 1 with the Michaelis-Menten equation, using the values for Km found above since no other data are available; for these thin cartilage types, the calculated oxygen profile was not sensitive to Km, since the oxygen tension did not fall below 5%. The oxygen consumption rate obtained from the fitted curve in Figure 1 was used for bovine knee cartilage as well as for human knee cartilage because data for human cartilage are not available. Calculated oxygen tensions fell with increasing distance from the synovial surface, to ∼2% in the deep zone of human and bovine knee cartilage and to 3.6% for bovine ankle cartilage, but remained above 5% for rabbit knee cartilage.

Figure 3.

A, Oxygen concentration profiles in rabbit, dog, bovine, and human femoral condylar (knee) and bovine metacarpophalangeal (ankle) cartilage samples, as calculated from equation 2. Calculations assumed uniform cell density and used data on cell density and cartilage thickness obtained from the literature and shown in Table 1, and rates of oxygen consumption taken from Table 1 and fitted to the Michaelis-Menten equation using measured values of the Michaelis-Menten constant (dog, rabbit) or from the fitted curve in Figure 1 (bovine, human). The synovial fluid concentration was assumed to be constant at 6% oxygen. B, Lowest oxygen tension in cartilage samples of different thicknesses (different species), calculated using data from Table 1 and Figure 1. The synovial fluid concentration was assumed to be constant at 6% oxygen. C, Oxygen concentration profile across human knee cartilage calculated using either the oxygen consumption rate fitted by the Michaelis-Menten equation (Figure 1) or the oxygen consumption rate at 21% oxygen (10.6 ± 1.4 nmoles/106 cells/hour). Data were taken from Table 1 (for human knees), and the synovial fluid concentration was assumed to be constant at 6% oxygen. D, Effect of cell distribution on oxygen profiles across human knee cartilage. Calculations were based on 1) measured values of the thickness and cell density profile across human knees (20), 2) uniform cell density of 9.6 million/ml, and 3) data reported by Stockwell (19), with a thickness of 2.3 mm and a cell density of 14.1 million/ml (Table 1). The oxygen consumption rate was taken from the fitted curve in Figure 1. All oxygen was assumed to be supplied by the synovial fluid (oxygen concentration 6%).

Figure 3B shows the lowest level that oxygen tension may reach in cartilage of different thicknesses (different species), assuming that all oxygen is supplied to the tissue from the synovial fluid. Calculations using data shown in Table 1 found the lowest oxygen tension in the thickest cartilage. In very thin rabbit cartilage, despite the high cell density, the lowest oxygen tension was close to that in synovial fluid because of short diffusion distances, whereas in thick human knee cartilage, it was calculated to fall to ∼1.4%. However, because of the relatively high cell density and oxygen consumption rate (Table 1) in dog cartilage, oxygen tension was similar in thin dog and thicker bovine cartilage.

All these calculations were based on the assumption that the oxygen consumption rate follows the Michaelis-Menten equation. This oxygen tension–dependent pattern of the oxygen consumption rate is very important as a protective feature. Figure 3C shows a comparison of the calculated oxygen concentration profile across human knee cartilage (thickness 2.3 mm and cell density 14.1 million/ml) using the Michaelis-Menten oxygen consumption rate with that based on a constant oxygen consumption rate (rate measured at 21% oxygen) (Table 1). In the former case, oxygen tension did not fall to 0 because of the fall in oxygen consumption rate with the decrease in oxygen tension. However, if the oxygen consumption rate was considered constant and independent of oxygen tension, oxygen was depleted at a normalized thickness of ∼0.4.

Our calculations also illustrated how the cell density distribution influences oxygen tension. In native cartilage, cell distribution is not uniform; cell density is highest in the superficial zone and decreases in the middle zone, and there is then a moderate increase in the deepest zone. Recent quantitative data found that for human femoral condyles (mean thickness 2.41 mm), cell density in the surface zone was 24.0 million/ml and in the midzone, it was 6.9 million/ml, compared with an overall average cell density of 9.6 million/ml (20). This cell distribution was incorporated into the model, assuming that all oxygen was supplied by the synovial fluid (6% oxygen), by dividing the cartilage depth into 6 zones, with the cell density for each zone taken from the data reported by Hunziker et al (20). Figure 3D shows that the lowest calculated oxygen tension in cartilage, then, was 2.6%, which is 30% higher than that based on uniform cell density (2.0% oxygen). In comparison, when we used the data reported by Stockwell (2.3 mm thick and uniform cell density of 14.1 million/ml) (Table 1), the lowest oxygen tension was 1.4%.

Effect of oxygen from subchondral bone

The effect of oxygen supplied from the subchondral bone on the calculated profile of oxygen concentrations across the cartilage thickness is shown in Figure 4. The curves were calculated from equation 2, assuming that the oxygen concentration in the synovial fluid was constant at 6% and that a proportion of the oxygen used by the cells was supplied via the subchondral bone. The curves were calculated for human knee cartilage using the data for cell density and thickness given in Table 1 and the rates of consumption taken from the fitted curve in Figure 1.

Figure 4.

Effect of subchondral oxygen supply on the profiles of oxygen tension across human knee cartilage. The effect of changing the proportion of oxygen from the subchondral bone in the range 0–50% is shown, assuming a uniform cell density of 14.1 million/ml and a cartilage thickness of 2.3 mm. Rates of oxygen consumption were taken from the fitted curve in Figure 1.

As the proportion of oxygen supplied by the subchondral route increased from 0 to 50%, the lowest oxygen tension existing in this cartilage rose from ∼1.5% to ∼4.5% (Figure 4). If oxygen was supplied by the subchondral route, then the lowest tension was not at the subchondral plate (as seen in Figure 3A). With an increase in the subchondral supply, the point where the lowest oxygen concentration was found moved from the subchondral plate (all oxygen supplied from the synovial fluid) to a normalized distance of 0.5 (50% oxygen supplied by the subchondral bone).


The modeling analysis presented here shows that oxygen tension gradient between the synovial surface and the cartilage–bone interface is regulated by a number of factors that affect either the delivery of oxygen to cartilage or the rate of oxygen consumption. These factors include cell density and its distribution, cartilage thickness, oxygen tension in synovial fluid, oxygen supply from the subchondral surface, and oxygen consumption rate per cell. It is clear that oxygen tensions within cartilage will differ from joint to joint and from species to species. In all but the thinnest cartilage, oxygen tension falls steeply with distance from the surface (Figure 3A), but we estimate that even in the deepest zone, oxygen tension is unlikely to fall as low as 1% in normal joints.

Consumption of oxygen by the cartilage depends on the rate of oxygen consumption per cell. The available information for mammalian cartilage is summarized in Table 1. Published results for the same species vary considerably, possibly because of differences in sampling site, and hence, cell density, and possibly also because of age differences (21). Nevertheless, the average rate of all measurements available is about 11 nmoles/million cells/hour, which is close to the values we measured in bovine chondrocytes either in situ or isolated from the matrix. We therefore thought it reasonable to use our data on the oxygen consumption rate for most calculations, particularly since we also measured the oxygen tension–dependence of the oxygen consumption rate (Figure 1). Figure 3C shows that this latter factor must be considered; if oxygen consumption is taken as being constant and independent of oxygen tension, the depth of penetration of oxygen into cartilage is significantly lowered. It should be noted, however, that the oxygen consumption rate also varies with the concentration of other nutrients, such as glucose (3), although at present, there are insufficient data in this regard to use such features in a model.

Another factor that significantly affects the oxygen tension gradient is the cartilage thickness (Figure 3A), which determines the diffusion distance; solutions to the diffusion equation are very sensitive to the diffusional distance (22). The effective diffusivity (D) of oxygen in the matrix would also affect the oxygen transport. We used a value for D that is consistent with the finding that effective diffusivity of all small solutes in cartilage is ∼40–50% of their diffusivity in aqueous solution (16). We did not consider possible variations in diffusivity between cartilage types because of lack of data; D is likely to be higher in younger and more hydrated tissues (23) and, so, increase, rather than limit, the oxygen supply to these tissues.

The thickness of cartilage, apart from affecting delivery, also affects the total consumption, since together with cell density, it determines the total number of cells throughout the cartilage depth and, hence, the total amount of oxygen consumed. Stockwell (19) suggested that, indeed, the cell density was limited by nutritional constraints, since he found a relationship between cell density (ρ) and cartilage thickness (d; in mm), that is, ρ = 27.9 × d–0.88 million cells/ml (19), which predicts that the total cell number under a unit area of cartilage surface is relatively constant, independently of cartilage thickness. Our calculations (Figure 3A) indicated that oxygen profiles did not normalize according to the relationship described by Stockwell. Glucose, which unlike oxygen, is essential for the survival of the chondrocyte (4, 24), is also supplied to cartilage by diffusion (25); glucose gradients across cartilage are similar to those of oxygen, whereas the gradients of lactic acid, which is produced by the cells, are reversed, with concentrations lowest at the cartilage surface (12). Since chondrocytes consume both oxygen and glucose (and other metabolites, such as glutamine) and produce lactic acid (2, 3), gradients in these nutrients and metabolites cannot occur independently in vivo.

Nutrient regulation of cell density, as suggested by Stockwell (19), may depend on the total nutrient–metabolite environment rather than the fall in concentration of 1 essential nutrient, such as oxygen, alone. Indeed, recent data indicate that cell viability is governed by interactions between levels of nutrients and metabolites. Cartilaginous cells die more rapidly under glucose deprivation if lactic acid concentrations are high and pH is acidic, for instance (24).

As well as the total cell density, the cell density distribution also appears to be important. Cell density is highest in the superficial zone, decreases in the middle zone, and there is then a moderate increase in the deepest zone (20). When this distribution was incorporated into the model, the consequent oxygen tension was higher than that based on uniform cell density (Figure 3D). This result suggests that the cell distribution is optimized so that the limited nutritional supply can support a maximal number of chondrocytes.

In regard to delivery of oxygen and other nutrients to cartilage, there is no doubt that these diffuse into the cartilage from the synovial fluid, as has been shown both in vitro and in vivo (26–28). These nutrients are supplied to the synovial fluid by the synovial capillaries (25). Diffusion from the synovial capillaries through a thin film of synovial fluid to the center of the cartilage surface is, however, insufficient to adequately supply the cartilage, and without the “stirring” induced by joint movement, synovial fluid would be depleted of nutrients (25), which possibly explains in part why cartilage degrades in immobilized joints (29). Here, we only considered the situation of constant synovial fluid concentration. Our calculations find, as expected, that synovial fluid oxygen tension directly affects the oxygen tension across cartilage, with cartilage oxygen tension falling in proportion to the decrease in synovial fluid oxygen tension (Figure 2), if we assume that synovial fluid is the only source of supply.

But, a very important question is whether oxygen is also supplied from the subchondral bone. There is strong evidence that the subchondral bone is a source of nutrients to immature cartilage (30, 31), but for adult cartilage, the answer is somewhat controversial. Physically, the subchondral bone is not completely impermeable; it is highly vascularized, and some of its terminal vessels contact directly with the deepest hyaline cartilage layer (32). Studies of normal femoral heads from adult humans found vascular channels to be in contact with the undersurface of the calcified zone of cartilage (33), while some vessels from the subchondral bone pass through the subchondral plate and terminate in the basal cartilage layer (32, 34). Although some studies have suggested that in adult animals, nutrient supply to the cartilage from the bone is minimal (35), other experimental evidence indicates that nutrition from the subchondral bone is important for articular cartilage, even in humans and large animals. Autografts with subchondral supply implanted into baboons remained intact over 3 years, while interruption of contact between articular cartilage and the vascularized subchondral bone resulted in degeneration of the cartilage over the long term (36). Another convincing piece of evidence for a subchondral supply comes from an in vivo study which found that gadolinium(diethylenetriaminepentaacetic acid)2– penetrated adult knee cartilage from the articular surface after intraarticular injection, but from both the articular surface and the subchondral bone after intravenous injection (37).

It thus seems quite likely that the subchondral region is also a source of oxygen and other nutrients for the cartilage, but is it a significant source? Some have concluded that diffusion from the subchondral region is limited, despite the presence of penetrating vessels (38, 39). Others have suggested that up to 50% of nutrients are supplied via the subchondral bone (32). Our calculations show that both the value and position of the lowest oxygen tension are significantly modified by the subchondral supply (Figure 4). If indeed the cell density profile is optimized to the prevailing nutrient supply, the finding that the cell density is similar in the midzone and toward the subchondral plate (20) indicates that the subchondral supply should not be ignored when considering cartilage nutrition.

Since synovial fluid and subchondral bone are the only 2 sources of oxygen and other nutrients for cartilage, pathologic changes in these tissues may threaten normal cartilage maintenance. Synovial oxygen tensions may fall to very low levels in chronic inflammatory arthritis or other joint disorders (Table 2), leading to corresponding decreases in cartilage oxygen tension (Figure 2), possibly affecting matrix metabolism adversely (2, 4). In osteoarthritis, changes in the subchondral bone (40) may influence its possible role in nutrient transport. The first signs of osteoarthritic damage, in hip cartilage at least, have been reported to occur in the midzone (41), the region of lowest cell density and lowest nutrient concentration if subchondral transport is involved (Figure 4).

In conclusion, although the calculations shown here can only provide estimates of oxygen tension profiles across articular cartilage in vivo, our results indicate that only under the most unfavorable conditions—thick cartilage, low synovial fluid concentration, and no transport from the subchondral bone—does oxygen tension in cartilage fall to ∼1% in the deep zone (Figures 2 and 3A). Our calculations also suggest that a significant subchondral supply is required for the lowest oxygen tension in the normal knee joint to remain close to 5%, the tension found to be optimal for the maintenance of matrix metabolism (2, 42). Finally, it should be noted that although this report discusses only oxygen, the factors which regulate oxygen profiles also regulate gradients in other essential nutrients, such as glucose, and in metabolic products, such as lactic acid. Chondrocyte metabolism is likely to be regulated by the entire nutrient–metabolite milieu, rather than the concentration of oxygen alone.