Differential cytokine response in interstitial fluid in skin and serum during experimental inflammation in rats


Corresponding author T. Nedrebø: Department of Physiology, University of Bergen, Jonas Lies vei 91, N-5009 Bergen, Norway. Email: torbjorn.nedrebo@fys.uib.no


Tumour necrosis factor-α (TNF-α) and interleukin-1β (IL-1β) are important mediators produced during inflammation. We hypothesized that the pro-inflammatory cytokine response in the interstitial fluid (IF) is different from that in serum, and we aimed at quantifying the amount of TNF-α and IL-1β in the IF. By centrifugation of rat skin at < 424 g pure IF is extracted. Using ELISA such fluid was analysed for cytokines in back and/or paw skin of pentobarbital-anaesthetized rats, after either induction of endotoxaemia or ischaemia–reperfusion (I/R) injury. During endotoxaemia, TNF-α increased in the IF from 0 in control to 640 ± 100 pg ml−1 (mean ±s.e.m.) after 90 min, with the serum concentration being 5–10 times higher at all time points. The response pattern of IL-1β after lipopolysaccharide (LPS) challenge differed greatly from that of TNF-α with a large increase in IF from 390 ± 90 to 28 000 ± 1500 pg ml−1 after 210 min, and a significantly smaller increase in serum (600 ± 45 pg ml−1). During reperfusion of the hind paw after 2 h of ischaemia, there was a gradual increase of TNF-α in both IF of the paw skin and serum after 3 min of reperfusion. Both declined after 20 min. The pattern for IL-1β differed, increasing significantly less in serum (25 ± 15 pg ml−1 after 20 min of reperfusion) than in the IF (1100 ± 200 pg ml−1). Immunostaining of the inflamed tissues showed increased expression of the two cytokines in cells of both epidermis and dermis compared to controls. Subdermal injections of TNF-α and IL-1β at the same concentrations found in IF after LPS infusion affected interstitial fluid pressure significantly. Local TNF-α production dominates after I/R injury, whereas in endotoxaemia systemic production predominates. For IL-1β local production dominates in both conditions. Thus, there is a differential pattern of cytokine production and the current method allows the study of the role of cytokines in IF during different inflammatory reactions.

Cytokines are regulatory peptides or glycoproteins that can be produced by virtually every nucleated cell type. During inflammation, pro- as well as anti-inflammatory cytokines are produced, and among the most investigated pro-inflammatory cytokines are TNF-α, IL-1β and IL-6. For many years it was believed that lipopolysaccharide (LPS) was the agent responsible for the shock and organ dysfunction during bacterial sepsis (Nishijima et al. 1973). Later work has shown that TNF-α and IL-1β are the culprits in the development of septic shock, as evidenced by increased survival during endotoxaemia (Tracey et al. 1987) in mice passively immunized with anti-TNF-α antiserum.

During an acute inflammatory reaction like ischaemia–reperfusion injuries of extremities and heart, and in endotoxaemia/sepsis, increased levels of cytokines are found (Hesse et al. 1988; Seekamp et al. 1993; Nossuli et al. 2000). Different types of investigatory methods have been used to detect cytokines in plasma/serum and in different tissues: enzyme linked immunosorbent assay (ELISA), immunostaining, polymerase chain reaction (PCR) and different blotting techniques. ELISA is by far the most used method to measure the concentration of various proteins, including cytokines; the other methods are mostly used for qualitative studies. Quantitative studies of cytokines have so far mostly been restricted to plasma/serum and supernatants of homogenized tissue samples or cell cultures, whereas the concentration of various substances in the interstitial fluid (IF), the fluid bathing the cells, has been unknown. In the human brain, microdialysis has been used to get an estimate of cytokine levels during inflammation (Winter et al. 2002). The problem has been to extract undiluted IF, without contamination from the cellular compartment.

Development of oedema is one of the cardinal signs of acute inflammation, and previously we have shown that lowering of the interstitial fluid pressure (Pif) is a major factor in oedema generation in many acute inflammatory reactions resulting from, e.g. burn, frostbite and ischaemia–reperfusion injuries (Lund et al. 1988; Berg et al. 1999; Nedrebøet al. 2003). Furthermore we have shown that different pro-inflammatory cytokines contribute to the development of oedema in skin following endotoxaemia by lowering the Pif (Nedrebøet al. 1999).

Proteins in the IF may give fundamental information on fluid exchange. Such fluid is however, not readily accessible, and thus various methods have been developed for its isolation. Recently Wiig et al. (2003) reported that centrifugation of tumours and skin at < 424 g can be used for this purpose. Application of this method gives us the possibility of quantifying the amount of cytokines present in the interstitial fluid, both in normal and pathological situations.

In this study we aimed at determining the concentration of two different cytokines in the IF in two models of inflammation, ischaemia–reperfusion (I/R) and endotoxaemia, and to show the potential role of these cytokines in regulation of Pif, and thereby oedema formation. We hypothesized that there may be a difference in cytokine response pattern between serum and IF, and a potential difference that could also be affected by the experimental model. Immunohistochemistry of skin from back or paw was performed to examine which cells might be responsible for a possible local production of cytokines both in I/R and endotoxaemia. Our findings demonstrate a different expression of TNF-α and IL-1β in experimental inflammation, where TNF-α is present in much higher concentrations in the systemic circulation, while IL-1β has much higher concentration in the interstitial fluid than in the systemic circulation. Cytokines given subdermally at the same concentrations as those found in IF during LPS inflammation lowered the Pif, suggesting a mechanistic role of these substances in oedema generation during inflammation.



All animals used were non-fasted female Wistar rats (weight 210–235 g) anaesthetized with pentobarbital intraperitoneally (50 mg kg−1). While anaesthetized the body temperature was maintained at 36.5–37.5°C with a heating lamp. The right external jugular vein was cannulated with a PE-50 catheter. The experiments described in this article have been carried out with the approval of and in accordance with the recommendations laid down by the National Animal Research Authority, and were approved by the local ethical committee.


LPS (Sigma-Aldrich, UK) was diluted in phosphate-buffered saline (PBS) containing 2% bovine serum albumin. The stock solution had a concentration of 10 mg ml−1, and was diluted in PBS containing 2% bovine serum albumin prior to injection. TNF-α and IL-1β (R & D systems, UK) were diluted in PBS containing 2% bovine serum albumin. The stock solution had a concentration of 10 μg ml−1. Prior to injection, TNF-α was diluted to 1 ng ml−1 and IL-1β to 25 ng ml−1.

Sampling of interstitial fluid

When the endotoxaemic or I/R-injury period ended, the animals were killed with an intracardiac injection of saturated potassium chloride, and were then immediately transferred to an incubator kept at room temperature and 100% relative humidity. Skin from paw and/or back was carefully excised. Any superficial blood contamination was removed by flushing with saline followed by careful blotting with tissue paper. The centrifugation method has recently been described by Wiig et al. (2003). Briefly, a skin sample was transferred to a preweighed centrifuge tube, provided with a basket of nylonmesh with pore size ∼15–20 μm, and the skin was placed with the subcutis facing the mesh. The nylon mesh will keep the skin sample up from the bottom of the tube (Aukland, 1991). The tube was immediately capped and spun at 424 g (2000 r.p.m.) in an Eppendorff 5417 R centrifuge placed in a coldroom at 4°C. After centrifugation the tube was brought back into the incubator. The fluid collected at the bottom of the tube (0.5–20 μl) was collected with a graded micropipette and diluted in PBS containing 2% bovine serum albumin. Thereafter the samples were frozen (−70°C) as soon as possible, and kept at this temperature until analysis.

Sampling of interstitial fluid in muscle

Due to the unexpected findings of IL-1β in skin of untreated animals, we analysed the interstitial fluid in hindlimb muscle for this cytokine. Interstitial fluid was sampled by inserting dry three-stranded nylon wicks (∼1 mm diam; Enkalon, the Netherlands) intermuscularly, as described in details elsewhere (Wiig et al. 1991). The wicks were prewashed in acetone, ethanol and distilled water. After washing and drying the wicks were stored in a box at 100% relative humidity. Briefly, the wicks were threaded into a 6 cm long PE-160 catheter. A 2–3 mm skin incision was made distally at the medial side of the leg, and a 1–2 mm incision was made in the muscle fascia. The wick-loaded catheter was inserted through the incision and advanced 2–3 cm intermuscularly. The loading catheter was withdrawn while the wick was pushed out with a PE-50 catheter, and both catheters were removed. At the end of a 20 min implantation period, the wicks were removed and put in preweighed vials filled with saline for elution. The next day the wicks were removed and the fluid was analysed for IL-1β with ELISA.

Blood samples

All blood samples were taken by intracardiac puncture during anaesthesia. Samples were allowed to clot for two hours at room temperature before centrifuging at 2000 g for 20 min. Serum was removed immediately and stored at below −20°C until analysis.

Analysis of TNF-α and IL-1β in interstitial fluid and serum

TNF-α and IL-1β in serum and interstitial fluid were measured with a commercial ELISA kit (R & D systems, UK). All plates were precoated with either TNF-α or IL-1β antibodies, and controls, standards and experimental samples were added. The washing and adding of antibodies, substrate solutions and stop solutions and analyses were performed according to the manufacturer's specifications. For each well the optical density was read using a microplate reader (Spectramax, Molecular Devices, USA) set to 450 nm, with wavelength correction. The detection limit was 5 pg ml−1.


An immunohistochemical staining technique, using the avidin–biotin complex (ABC) was performed on tissue sections. Skin samples from back or paw of control or experimental rats were taken after i.v. injection of saturated KCl. The samples were immediately placed in Histocon (Jonsson et al. 1978), before they were snap-frozen in isopentane cooled in liquid nitrogen, and embedded in tissue-tek (Sakura Finetek, Zeuterwounde, the Netherlands). All samples were stored at −70°C until sectioning in a cryostat (5 μm). The sections were fixed on glass microscopy slides. The slides were first fixed in 50% and thereafter 100% acetone, before they were rinsed in Tris-buffered saline (TBS). The slides were rinsed in TBS for 10 min between every new procedure. Following rinsing, the slides were treated with 0.3% H2O2 for 5 min to avoid endogenous peroxidase activity, and then incubated for 15 min in a biotin blocking solution (Vector Laboratories, CA, USA). After washing, the slides were blocked with 5% normal rabbit serum for 15 min. This was followed by adding the primary antibody (antirat IL-1β antibody or antirat TNF-α antibody (R & D systems, Burlington, UK)) to the slides, and with an incubation period of 30 min. Then a biotinylated secondary antibody (rabbit-antigoat) was added and incubated for 30 min. The slides were subsequently treated with Avidin–Biotin solution (DAKO AB, Glostrup, Denmark) including peroxidase for 30 min, before they were developed with 10 mg 3-amino-9-etyl-carbaxol (AEC) in 6 ml dimethylsulfoxide (DMSO) for 15 min. After a last rinse in TBS the slides were counterstained with Mayer's haematoxylin eosin. Evaluation of immunostaining was performed under a light microscope (Leica, Wetzler, Germany) with final magnification of 600. The tissue was evaluated for increased cytokine expression in (1) cells in epidermis, (2) cells and structures in dermis, and (3) cells and structures in hypodermis and the underlying muscle tissue.

Interstitial fluid pressure measurements

P if was measured by a micropuncture technique using sharpened glass capillaries connected to a servocontrolled counterpressure device. The method has previously been validated and the criteria for acceptance of the measurements have been defined elsewhere (Wiig et al. 1981). Measurements were performed on the dorsal side of the hind paws at a depth of 0.3–0.5 mm below the skin surface. In experiments with subdermal injection of test substance, a circle with diameter 5 mm was outlined in ink with a pen, with the centre at the injection point. Five microlitres of the test substance was injected on the dorsal side of the paw, with a 10 μl chromatography syringe (Hamilton, Bonaduz AG, Switzerland). The test volume was carefully injected subdermally, after the needle had been inserted and the tip of the needle had reached the centre of the circle. Measurements were made in the outer part of the ink circle, which corresponds to the edge of the injected volume (Reed et al. 1992). The measurements were averaged for the following time periods: 0–20, 21–40 and 41–60 min. In these experiments circulatory arrest was induced as part of the experimental protocol. During an inflammatory reaction the fluid filtration to the tissues will increase, and thereby increase the Pif. It has been shown that Pif after circulatory arrest is not different from control for a period of 90 min (Wiig et al. 1981).

Experimental protocols

Controls.  Twelve rats served as controls. After anaesthesia, a cardiac puncture was performed and an arterial blood sample taken. The blood samples were taken from 3 min to 3 h after induction of anaesthesia, at times matching those of the animals in the experimental groups. After i.v. injection of saturated KCl, skin samples were taken from the back (n= 12) and paw (n= 7).

Endotoxaemia.  Endotoxaemia was induced by intravenous injection of LPS. There were three groups receiving LPS (4 mg kg−1) i.v., each consisting of seven animals. In the first group (A) LPS circulated for 30 min, in the second group (B) it circulated for 90 min and in the third group (C), for 210 min. At the time of ending the experiment (30, 90 or 210 min), a cardiac puncture was performed and an arterial blood sample taken. After i.v. injection of KCl, samples were taken from the skin of the back. From Group C, skin was also sampled for immunostaining.

Ischaemia–reperfusion injury.  In this group the animals were subjected to 2 h of ischaemia of their left hindlimb. An occluding tourniquet, 1 cm in width, was placed proximal to the ankle joint as previously described (Nedrebøet al. 2003). The tourniquet had soft lining on the inside to avoid mechanical damage to the skin. After 2 h of occlusion the tourniquet was removed, and reperfusion started. Animals were assigned to one of three different groups, all with n= 7, according to the length of the reperfusion period: 0, 3 and 20 min. No reperfusion (0 min) was chosen for study if there was any increase of cytokines in the tissue due to ischaemia only. We used 3 min of reperfusion, since this was the time needed to cause a lowering of Pif (Nedrebøet al. 2003), which has been shown to be an important factor for oedema development during different inflammatory processes (Lund et al. 1988). A reperfusion period of 20 min was chosen based on pilot studies where no TNF-α was detectable 1 h after the start of reperfusion, an observation suggesting that the cytokine might gradually decline in the tissue. At the end of the reperfusion period, a cardiac puncture was performed and an arterial blood sample taken. This was followed by i.v. injection of saturated KCl, followed again by sampling of skin from the paw. From Group 3, skin was also harvested for immunostaining.

Interstitial fluid in muscle.  Two rats were investigated for IL-1β in the intermuscular fluid. Both rats were untreated, and intermuscular fluid was harvested as previously described. From both rats fluid from four wicks was analysed.

Statistical methods

All values are means ±s.e.m. unless otherwise stated. The data were analysed with one-way analysis of variance (ANOVA) followed by Dunn's or Bonferroni's test. P < 0.05 was considered statistically significant.



In control animals there was no detectable amount of TNF-α in either serum or IF (Fig. 1). No IL-1β could be detected in serum in control rats, but in the IF the levels (in total 21 samples) ranged from 0 (n= 4) to 1100 pg ml−1, with an average of 390 ± 90 pg ml−1 in skin fluid isolated from the back using the centrifugation method (Fig. 2). Interstitial fluid in muscle sampled by wicks had no detectable amount of IL-1β. In paw skin, detectable amounts of IL-1β were observed in one rat only; the others were negative.

Figure 1.

Concentration of TNF-α in serum (●) and interstitial fluid (○) in experimental endotoxaemia, induced by i.v. injection of lipopolysaccharide
The experiment was ended at 30, 90 and 210 min. n= 7 for each time point, except control (0 min, includes serum and IF) where n= 12. * and †, P < 0.001 compared with respective control values. Data are means ±s.e.m.

Figure 2.

Concentration of IL-1β in serum (●) and interstitial fluid (○) before and during experimental endotoxaemia (i.v. LPS)
The experiment was ended at 30, 90 and 210 min. n= 7 for each time point, except control (0 min) where n= 12. * and †, P < 0.05 and **P < 0.001 compared with respective control values. Data are means ±s.e.m.

Concentrations of TNF-α and IL-1β in serum and interstitial fluid during endotoxaemia

The levels of TNF-α (Fig. 1) and IL-1β (Fig. 2) increased compared to control values in both serum and IF, although with a different pattern. TNF-α concentrations in serum and IF were at a maximum at 90 min (3960 ± 150 and 640 ± 100 pg ml−1, respectively) (P < 0.05 compared to control), and fell after 210 min. The amount of TNF-α was significantly higher in serum compared to IF at all times (P < 0.05). IL-1β concentration in serum and IF continued to rise throughout the observation period, and did not reach a peak within the 210 min used as maximal experimental duration. In serum there was a modest increase to 25 ± 15 pg ml−1 at 90 min, but at 210 min the increase in concentration was 20 times that at 90 min, which was significantly different (P < 0.05). In IF, however, the concentration was 28 000 ± 1500 pg ml−1 at 210 min, almost 50 times higher than in serum at the same time, and was at 90 as well as 210 min significantly higher than in serum and control (P < 0.05 and P < 0.001, respectively).

Concentrations of TNF-α and IL-1β in serum and interstitial fluid during ischaemia–reperfusion injury

For TNF-α, the levels increased in all groups compared to controls (Fig. 3), with a maximum in both serum and IF at 3 min of reperfusion (20 ± 6 and 140 ± 20 pg ml−1, respectively) (P < 0.05 compared to control). Interestingly we found a slight increase in the skin of ischaemia only, to 3 ± 4 pg ml−1 in serum (with the mean below the detection limit set by the manufacturer), and 30 ± 15 pg ml−1 in IF, which for IF is significantly different from control (P < 0.05). For IL-1β, the levels after ischaemia–reperfusion injury were totally different from those of TNF-α (Fig. 4). We did not observe any increase in serum until after 20 min of reperfusion (P > 0.05), whereas for IF there was an increase (not significant) after ischaemia only, followed by a continuous rise throughout the reperfusion period. The IL-1β concentration was significantly different from control at 20 min of reperfusion, with a value of 1100 ± 500 pg ml−1 (P < 0.001).

Figure 3.

Concentration of TNF-α in serum (○) and interstitial fluid in skin of rat paw (●) during ischaemia and reperfusion injury
n= 7 for each time point. * and †, P < 0.05 compared to respective control values (before start of ischaemia, includes serum and IF). // indicates the 2 h ischaemic period. Data are means ±s.e.m.

Figure 4.

Concentration of IL-1β in serum (●) and interstitial fluid in skin of rat paw (○) before ischaemia, after ischaemia and after ischaemia–reperfusion injury
n= 7 for each time point. *P < 0.001 and †P < 0.05 compared to respective control values (before start of ischaemia). // indicates the 2 h ischaemic period. Data are means ±s.e.m.

Interstitial fluid in muscle

IL-1β was not detected in interstitial fluid isolated using intermuscular wicks in muscle in control rats.


Controls ( Fig. 5A and B).  In untreated animals there were a few cells in stratum basale of epidermis expressing IL-1β, but no cells expressing TNF-α. In dermis there was some expression of IL-1β in epithelial cells around the hair follicles, and a few fibroblast-like cells expressing TNF-α. In the deeper muscle layer there was a weak expression of IL-1β, but not of TNF-α.

Figure 5.

Microphotographs of immunohistochemical staining of backskin in untreated rats, and in rats following LPS challenge
Increased expression of either IL-1β or TNF-α is represented as red staining. A and B are from untreated/control rats. C and D are stained for IL-1β and photographs E and F are stained for TNF-α 210 min after LPS challenge.

Endotoxaemia.  After 210 min of endotoxaemia a significantly increased staining for IL-1β (Fig. 5C and D) was observed among cells of the epidermis, and also some expression of TNF-α was detected (Fig. 5E and F). In the dermis all epithelial cells surrounding the hair follicles showed a marked increased expression of IL-1β, and also some increase of TNF-α expression. In this experimental condition a lot more of fibroblast-like cells in the interstitium showed increased staining of TNF-α. In the deeper muscle layer there was more pronounced expression of IL-1β and TNF-α than in the control situation.

Ischaemia–reperfusion injury.  After 2 h of ischaemia and 3 min of reperfusion there was a major increase of expression of IL-1β in the basal layer of epidermis, but very scant expression of TNF-α. In the dermis all epithelial cells surrounding the hair follicles showed a marked increased expression of IL-1β, and also some increase of TNF-α expression. There were no fibroblast-like cells expressing IL-1β, and very few (non-specific) expressing TNF-α during the I/R experiment.

Interstitial fluid pressure ( Fig. 6).  Based on results obtained for the concentrations of the measured cytokines in IF, we wanted to test whether these concentrations could induce local oedema formation as reflected in a reduction of Pif. The experimental group (n= 6) was treated as follows. After anaesthesia, a control pressure was measured in the left paw. Then a total volume of 5 μl containing a mixture of TNF-α (2.5 μl, 1 ng ml−1) and IL-1β (2.5 μl, 25 ng ml−1) was injected subdermally in one paw. Measurements were performed for the next 60 min. Then the animals were killed with an intracardiac injection of saturated potassium chloride, and thereafter 5 μl of the TNF-α–IL-1β mixture was injected subdermally in the other paw. Measurements were again performed for 60 min. This way we could examine the effect with intact circulation, as well as after circulatory arrest. Controls (n= 8) were treated similarly to experimental animals, except that 0.9% NaCl was injected subdermally.

Figure 6.

Effects of subdermal injections of TNF-α and IL-1β (1 and 25 ng ml−1, respectively) in paw with (▼, n = 6) or without (●, n = 6) circulatory arrest

Figure 7.

Ratio of interstitial fluid and serum concentrations
A, ratio of concentration of TNF-α and IL-1β in interstitial fluid over serum in early (30 min, black bar) and late (210 min, grey bar) phase of the endotoxaemia. Data are means ±s.e.m.B, ratio of concentration of TNF-α and IL-1β in interstitial fluid over serum in early (3 min, black bar) and late (20 min, grey bar) phase of ischaemia–reperfusion injury. Early phase of IL-1β is missing due to undetectable amounts in serum. Data are means ±s.e.m.

Control animals had a pressure of −0.5 ± 0.1 mmHg (n= 8), which did not change significantly throughout the observation period (P > 0.05). Pif in rats with intact circulation and arrested circulation did not differ significantly. In the experimental paw with intact circulation the pressure increased significantly from −0.4 ± 0.05 mmHg at 0 min to 1.1 ± 0.3 mmHg at 60 min (P < 0.02). In the opposite paw, the pressure fell significantly from −0.2 ± 0.3 to −3.3 ± 0.5 mmHg after 60 min (P < 0.02). At 40 and 60 min both experimental paws were significantly different from each other as well as from controls receiving 0.9% NaCl (P < 0.02).


To the best of our knowledge, this is the first time quantification of cytokines has been done in interstitial fluid in normal and in inflamed skin. Most reports regarding elevated amounts of cytokines in inflammation are described by qualitative methods, or indirectly. The centrifugation method used in this study, and also the wick method, makes it possible to get access to uncontaminated interstitial fluid from skin and muscle. Combined with ELISA, these methods enable us to quantify the amount of different inflammatory proteins in this fluid, and thereby get a better understanding of the mechanisms involved in inflammatory processes occurring locally at the tissue level.

Methodological considerations

Various techniques have been developed for sampling tissue fluid; all of these have their different limitations (Aukland & Reed, 1993). A potential risk involved in these methods is cell damage, resulting in contamination of the extracellular fluid with intracellular fluid and proteins. In a recent paper we have developed and validated a centrifugation method to isolate interstitial fluid (Wiig et al. 2003). By centrifugation of skin at a G-force < 424 g, we found the contribution from the intracellular compartment to be negligible as examined by 51Cr-EDTA. This probe will not enter the cells (Løkken, 1970). After centrifugation the ratio of 51Cr-EDTA in centrifugate and plasma was not different from 1.0, demonstrating that intracellular fluid was not diluting the fluid obtained by centrifugation. Centrifugation will sample fluid from the whole extracellular fluid volume phase, but the centrifugate was found to contain less than 2% of a tracer confined in the intravascular fluid phase, suggesting that the contribution from this compartment was negligible. Contamination of the skin by blood during the experiment might be a problem, but if this happened (only twice during the whole experiment) the skin was washed with saline and thereafter gently dried. Given the reservations above we feel confident that the present fluid isolation method will reflect the changes taking place in the interstitial fluid during experimental inflammation.

Interstitial cytokines

Our results show that in endotoxaemia there is an increase of TNF-α and IL-1β in the interstitium, and that the cytokine level continues to increase throughout a 3 h experimental period for IL-1β, and for at least 90 min for TNF-α. The concentrations of IL-1β in IF were much larger than TNF-α, and even a detectable amount of IL-1β was found in IF in controls. When comparing TNF-α and IL-1β, the TNF-α is substantially higher in serum, while IL-1β is largely confined to the IF. In Fig. 7A this is presented as the ratio of mean concentration in interstitial fluid over serum in an early (30 min) and late (210 min) phase of endotoxaemia. The observed ratio of concentration in IF over that in serum being larger than 1 clearly demonstrates that IL-1β is produced in the interstitium, since any protein being transported across the microvasculature from plasma to IF will be present in lower concentrations in IF than in plasma (Michel & Curry, 1999). In I/R we found increased amounts of TNF-α and IL-1β in skin of paw after the 2 h ischaemic period. This is in agreement with results from ischaemic myocardium (Shames et al. 2002) where ischaemia alone increased TNF-α gene expression and peptide synthesis. Initially there was an increase of TNF-α in both serum and interstitium, where the concentration in interstitial fluid was up to 10 times higher than in the circulation. As evident from Fig. 3, the levels of TNF-α increased to a maximum at around 3 min of reperfusion, followed by a gradual reduction in concentration at continued reperfusion. After only 20 min of reperfusion, TNF-α in interstitial fluid decreased rapidly to levels measured in animals having only ischaemia of the limb, while serum concentrations declined more slowly. This observed reduction in TNF-α in the IF could be due to a washout effect of the cytokine from the interstitium, meaning that the TNF-α in the IF in the reperfused tissue is brought into the circulation via the lymphatics, again explaining why the levels of TNF-α in the circulation have a slower decline than in the IF. Compared to endotoxaemia, the results for I/R states show that the latter is more a local inflammation.

The levels of IL-1β during I/R-injury had a development different from TNF-α. In serum IL-1β did not appear until 20 min of reperfusion, whereas in IF there was an increase already after 3 min of reperfusion, an increase that continued after 20 min of reperfusion. A hypothesis that may be derived from these observations is that IL-1β seen after I/R is largely derived from the interstitium. In Fig. 7B the ratios for IF/serum in an early (3 min) and late (20 min) phase of I/R-injury are presented. The ratio for IL-1β is clearly over 1, again indicating a local production (the early phase is lacking due to undetectable amounts in serum). Similarly, the TNF-α ratios are over 1, indicating that this cytokine also is produced locally during I/R-injury.

Immunhistochemistry revealed that the source of IL-1β in the tissue, in both endotoxaemia and I/R injury, was cells in the epidermis and cells surrounding the hair follicles. The source for TNF-α in the interstitium during endotoxaemia was cells in the basal layer of epidermis, but also fibroblast-like cells in the interstitium. A somewhat different pattern was found in I/R injury, where the increased TNF-α expression was seen in epithelial cells surrounding the hair follicles.

We have previously shown a role for TNF-α and IL-1β in exchange of fluid from the capillaries and into the interstitium by lowering of the interstitial fluid pressure (Pif), and we have also seen a significant lowering of Pif and increased albumin extravasation after I/R injury in hindlimb of rats (Nedrebøet al. 2003), and during endotoxaemia in rats (Nedrebø & Reed, 2002). In the present study we observed that both these cytokines were up-regulated in the IF in both of these inflammatory reactions. Furthermore, by applying the observed concentrations directly to subcutis we were able to elicit a change in Pif mimicking that observed in previous LPS and I/R studies, a response inducing fluid flux across the capillaries. These observations strongly suggest that these cytokines take part in control of fluid exchange during inflammation. TNF-α has been reported to induce increased vascular endothelial permeability (Royall et al. 1989) and tissue oedema formation (Beutler et al. 1985). It has also been reported that alterations of the extracellular matrix, which can reduce cell-matrix contact, could be the result of TNF-α causing increased endothelial permeability (Partridge et al. 1992).

One of the most surprising findings in this study was the elevated amounts of IL-1β in control, i.e. untreated animals. That this observation was not due to a methodological artefact was shown by the lack of IL-1β in unstimulated muscle. IL-1 has been found to be present in normal human epidermis, most likely IL-1α (Hauser et al. 1986). IL-1 is not produced by unstimulated cells, with the exception of keratinocytes in skin, some epithelial cells and cells of the central nervous system (Dinarello, 1994a,b). Normal production of IL-1 is, however, critical for initiation of normal host response to injury and infection. Intracellular IL-1β consists exclusively of a 31 kDa precursor form (Hazuda et al. 1990). Extracellular IL-1β consists of a mixture of both the precursor and the mature IL-1β. There are reports that ELISA kits using mature IL-1β as a standard will detect, but considerably underestimate, the unprocessed IL-1β precursor (Dinarello, 1992; Herzyk et al. 1992). The precursor form is usually not the predominant form of IL-1β. If we compare the response of TNF-α and IL-1β, we observe that the production and large amount of IL-1β are much more confined to the interstitium than are those of TNF-α. The rapid increase of IL-1β could be explained by already existing precursor IL-1β being rapidly activated to the mature form by interleukin-1β-converting enzyme (ICE).

In conclusion, the present experiments show that quantification of inflammatory mediators in plasma/serum does not give a representative picture of concomitant changes in the interstitium. We have demonstrated a fundamental difference in the expression of the two cytokines studied, TNF-α and IL-1β. IL-1β was markedly up-regulated in the interstitium both in a local and systemic inflammatory reaction, while the TNF-α response pattern differed in these two conditions. The cytokine concentrations observed in IF during LPS stimulation were able to induce lowering of Pif, suggesting a mechanistic role of these substances in inflammatory reactions. Isolation of native IF allows us to analyse the fluid in which the communication between cells and extracellular matrix takes place and thereby get a better understanding of the molecular mechanisms involved in inflammatory processes.



Financial support from The Norwegian Research Council, The Norwegian Heart Association and Locus on Circulatory Research University of Bergen are gratefully acknowledged. We thank Gerd S. Salvesen, Marianne Eidsheim, Turid Tynning, Malin Jonsson, Olav Tenstad and Per Sakariassen for most valuable technical assistance.