Pentoxifylline inhibits the fibrogenic activity of pleural effusions and transforming growth factor-β

Physiopathology of organ fibrosis is far from being completely understood, and the efficacy of the available therapeutic strategies is disappointing. We chose pleural disease for further studies and addressed the questions of which cytokines are relevant in pleural fibrosis and which drugs might interrupt its development. We screened pleural effusions for mediators thought to interfere with fibrogenesis (transforming growth factor-β (TGF-β), tumour necrosis factor α (TNFα), soluble TNF-receptor p55 (sTNF-R)) and correlated the results with patient clinical outcome in terms of extent of pleural thickenings. We found pleural thickenings correlated with TGF-β (p < 0.005) whereas no correlations could be observed with TNFα and sTNF-R. Further, we were interested in finding out how TGF-β effects on fibroblast growth could be modulated. We found that pentoxifylline is able to inhibit both fibroblast proliferation and collagen synthesis independently of the stimulus. We conclude that, judging from in vitro studies, pentoxifylline might offer a new approach in the therapy of pleural as well as pulmonary fibrosis.


Introduction
The normal brotic response to lung tissue injury is nely controlled. Despite multifactorial origins, mesenchymal cells migrate to the site of injury, proliferate and subsequently synthesize extracellular matrix components. 1 Neither the factors that contribute to limited matrix production in wound healing nor those that control excess or sustained matrix production in brosis have been fully characterized. Cytokines have been considered highly important in the development of pulmonary brosis (e.g. platelet derived growth factor (PDGF), tumour necrosis factor a (TNFa), transforming growth factor b (TGF-b)). 2 5 A therapeutic breakthrough has, however, not been forthcoming. Corticosteroids are still the rst-line therapeutics in pulmonary brosis, even though they improve the course of disease in only about 25% of cases, and that to an often disappointing extent. 6 Alternative therapeutic regimen are either toxic, not convincingly effective or have not yet been subjected to suf cient study. 6 We chose pleural disease for further studies on the physiopathology of brosis and therapeutic intervention strategies. Mesothelial cells of the pleura, being highly susceptible to harmful events, 7 exfoliate and uncover the submesothelial connective tissue. An exudative in ammatory reaction results with the accumulation of uids in the pleural space. 8 These uids contain a variety of substances with the potential to in uence cell growth. 9,10 Pleural effusions are most intimately in contact with extracellular matrix producing submesothelial connective tissue, so we hypothesized they might re ect the complex mechanisms of inammation, wound healing, and development of pleural brosis. We addressed the question of whether mediators considered to be causally involved in brosis of the lungs by either stimulating matrix synthesis (TGF-b 11 ) or inhibiting cytokines or matrix production (soluble TNF-receptor p55 (sTNF-R), 12 are also physiopathologically relevant in pleural brosis.
Having worked out comparative mediator pro les and clinical patient outcomes, we investigated the effect of the cytokine found to be relevant (TGF-b) and of pleural effusions, a material containing a multiplicity of mediators, on in vitro broblast cultures in order to search for alternatives to steroid treatment. Screening various drugs, we chose the xanthine derivative pentoxifylline (POF) for comparative studies. Pentoxifylline is a methylxanthine initially prescribed in the therapy of peripheral vascular disease 13 which has been shown to interact with several cell types including broblasts. 14 Our rationale for this choice included the growing set of data indicating that the intracellular concentration of cyclic adenosine monophosphate (cAMP) in uences broblast activity. 15,16 Pentoxifylline augments cAMP by inhibiting phosphodiesterase activity 16 and activating adenylate cyclase by inducing prostacyclin 17 and has been shown to inhibit broblast collagen, glycosaminoglycan and bronectin synthesis and increase collagenase activity. 18 POF inhibits broblast proliferation stimulated by platelet-derived growth factor (PDGF), a further cytokine considered important in brosis. 19 The drug has also been shown to modulate in ammatory and immune reactions effectively (e.g. inhibition of leukocyte adhesion, aggregation, degranulation and superoxide release and monocyte TNF-a synthesis 16 ). Thus we assumed that pentoxifylline's mechanism of action might constitute an effective therapeutic action in brosis.
In this paper, we provide data on (1) cytokine concentrations in pleural effusions (TGF-b, TNFa, sTNF-R); (2) pentoxifylline effects on (a) broblast cell proliferation, and (b) broblast collagen synthesis stimulated by pleural effusions and pure cytokines (TNFa, TGF-b). Concluded from these in vitro investigations, data is in favour of a pentoxifylline bene t in brosis.

Subjects
Unless otherwise indicated, proteins were analysed in pleural effusions from 49 consecutive patients suffering from non-carcinomatous pleurisy. Table 1 lists the pertinent clinical features and diagnoses. In all cases, thoracentesis or thoracoscopy was performed for diagnostic reasons. Only pleural uids from the rst diagnostic or therapeutic pleural intervention were used in this study. Routine analysis included the determination of speci c weight, cell differentiation, total protein concentration and LDH concentration and subsequent classication as transudates (total protein in pleural uid , 30 g/l, LDH-ratio pleural uid:blood: , 0 6, or exudates (protein . 30 g/l, LDH-ratio . 0 6). Further analyses (cytomorphology, immunological markers, etc.) were done as appropriate. The uids were centrifuged and the supernatant devoid of cellular components was stored at 808 C until further investigation.
In 24 of the subjects we were able to do follow-up studies. Pleural thickenings were measured using chest X-ray lms or computed tomography (CT) scans of the thorax made 3± 6 months after diagnosis during follow-up examinations and classi ed according to the international classi cation of radiographs of pneumoconioses (International Labor Of ce (ILO)). 20 The results were correlated with the concentrations of TGF-b, TNFa, and sTNF-R in the pleural effusions.

Assay of transforming growth factor-b
Measurement of TGF by bioassay (growth inhibition of a mink lung epithelial cell line) did not reveal reproducible results. Thus TGF-b 1 was determined by ELISA in which studies on recovery and reproducibility showed good results (Quantikinine, R&D Systems, MN, USA). Sample preparation included acidi cation using 1 mol/l acetic acid to split the molecule off the protein binding and subsequent dialysis using Visking membranes (Serva, Germany, exclusion limit: 8000± 15000 dalton) against phosphate buffer solution (PBS, Gibco BRL, w/o calcium magnesium). Thus, both the inactive, latent, and the active form of TGF-b are detected. The assay uses a sandwich enzyme immunoassay technique with a monoclonal antibody against TGF-b and a detection system using a polyclonal antibody conjugated to horseradish peroxidase. Optical density determined in a multiwell scanning spectrophotometer (ELISA-reader, Dynatech MR 5000) at a wavelength of 450 nm allowed calculation of TGF-b of the samples by comparison with a standard curve. Lower detection limit of the assay is at 5 pg/ml.

Assay of tumour necrosis factor a
Sample preparation was not a prerequisite in this assay. The sandwich immunoassay (Medgenix Diagnostics, Belgium) consists of oligoclonal TNF-antibodies and an anti-TNF-antibody detection system conjugated to horseradish peroxidase. It detects total TNFa (i.e. monomeric, trimeric, receptor-bound, and non-bound TNF). Concentrations of TNF can be determined by comparing optical densities of samples with standard curves. No cross-reactivity has been reported with tumour necrosis factor b (TNF-b), interleukin-1 (IL-1), interleukin-2 (IL-2), interferon-a (IFN-a), interferon-b (IFN-b), or interferon-c (IFN-c) (product information Medgenix). Detection limit of the assay is at 3 pg/ml.

Assay of soluble tumour necrosis factor receptor p55
This assay was a kind gift of Dr H. Gallati (Hoffmann-La Roche, Switzerland). Principle of the assay is a sandwich enzyme immunoassay technique consisting of a monoclonal TNFreceptor p55 antibody (mouse) and peroxidaseconjugated recombinant TNF-a. Concentrations of unknowns are calculated from optical densities determined at a wavelength of 450 nm using a ELISA-reader. Detection limit is 100 pg/ ml. No cross-reactivity with IFN-a, IFN-c, IL-1a, IL-1b, PDGF-AA, PDGF-AB, PDGF-BB has been reported (product information Hoffmann-La Roche).

Assay of ® broblast proliferation
A human lung broblast cell line was obtained from the American Type Culture Collection (WI-38, derived from normal embryonic lung tissue of a caucasian female) and grown in Basal medium (Eagle) (BME) according to standard protocols. Only early passage cell cultures (days 15± 35, split ratio 1:2 once weekly) were used in these experiments. Fibroblasts (WI-38) were seeded at subcon uent density of 0 13 10 6 cells/ml (medium BME, 10% FCS added) into 96-well at bottom microtitre plates (0.1 ml per well, Greiner, Germany) and cultivated for 24 h. After a 1 h 'washout period' to reduce or remove FCS, BME medium without or with FCS 0.4%, and pleural effusions or cytokines (TGF-b1 at 3 ng/ml, human, expressed in Escherichi a coli, Sigma Chemicals, USA; TNFa at 100 ng/ml, human, recombinant, Boehringer, Mannheim, Germany) as well as 1% antibiotics (penicillin 100 U/ml, streptomycin 100 U/ml, amphotericin 0.25 mg/ml)) were added. In respective experiments, medium was supplemented with pentoxifylline (Rentylin, Dr Rentschler, Laupheim, Germany, at 50 and 100 mg/ml). Medium was removed completely after 72 h, and 100 ml of the tetrazolium salt MTT (dimethylethimazoldiphenyltetrazolium bromide, Sigma, dissolved in PBS at 5 mg/ml) was added. MTT is converted by mitochondria of living cells to the blue coloured substance formazan, and the amount of formazan produced is proportional to the number of cells present. 21 The optical density was determined in an ELISA-reader at a wavelength of 550 nm.
Experiments were done six-fold; important results have been con rmed by counting the cells in a haemocytometer (Coulter counter, after preparation of the cell nuclei).

Assay of collagen synthesis by in vitro cultured ® broblasts
Experiments using TGF-b or TNFa as stimulant were done in quadruplicate, and all others in duplicate. Pleural effusions of seven patients were investigated regarding stimulation of col- lagen synthesis and its inhibition by pentoxifylline (Rentylin, Dr Rentschler, Laupheim, Germany, at 50 mg/ml).
Principle of the assay is the incorporation of [ 3 H]-proline into proteins and its hydroxylation to [ 3 H]-hydroxyproline in collagenous proteins as outlined in Ref. 22. The method was adapted such that non-incorporated radioactivity was removed by ultra ltration instead of dialysis. Brie y, broblasts (WI-38) were grown in 24well at bottom microtitre plates to visual con uency and incubated in the presence of 50 mg/ml ascorbate, pleural effusions at a nal dilution of 1:5, or cytokines TNFa or TGF-b diluted in medium supplemented with 0.4%FCS (TNFa: human, recombinant, Boehringer, Mannheim, Germany; TGF-b1: human, expressed in E. coli, Sigma Chemicals, USA), and 10 mCi [ 3 H]-proline/ml medium (L-[2,3-3H]-proline, Dupont, USA). After 24 h, cell pellets as well as cell-free supernatants were harvested, freezethawed three times, pelleted by centrifugation, and the supernatant was washed four times and ultra ltrated (centrifugal concentrators Microsep Filtron, Karlstein, Germany, molecular weight cutoff 10 K; aqua dest. supplemented with 0.3 ml/ml phenylmethansulphon uorid (PMSF), proteinase-inhibitor, Roth, Germany). Subsequently, samples were resuspended in 6 N HCl (Merck), hydrolysed (1108 C, 24 h), dried in a vacuum-desiccator, and the amounts of [ 3 H]proline and [ 3 H]-hydroxyproline were determined by automated amino acid analysis. Collagen concentrations were calculated according to the method of Wiestner et al. 22 Collagen synthesis was expressed as percentage of collagen of generated total protein and as radiolabelled hydroxyproline per cell.

Statistics
The data is provided in means SEM unless otherwise stated. Variance signi cance was calculated by means of the Mann± Whitney Utest, Spearman's rank correlation test and Wilcoxon's signed rank test. For comparisons, P values , 0 05 were adopted as signi cant.

TGF-b, TNFa, and TNF-receptor p55 in pleural effusions
We found a signi cant positive correlation between the concentrations of TGF-b1 and the extent of pleural thickenings classi ed according to ILO as outlined in Fig. 1 (coef cient of correlation: 0.54; P , 0 005). Assuming a cutoff at TGF-b 100 ng ml and de ning pleural thickenings ILO 0 and 1a to be a restitutio ad integrum, the sensitivity of TGF-b measurements to detect risk of pleural brosis was 75% and speci city was 80%. Table 2 provides data on TGF-b, TNFa, and Total TGF-b isoform 1 (that is, both inactive protein-bound and active forms) was determined by ELISA and correlated with pleural thickening as determined by chest X-ray ® lms or computed tomography scans, then classi® ed as ILO 0± 3c. Transforming growth factor-b1 (TGF-b1), tumour necrosis factor a (TNFa), and soluble tumour necrosis factor receptor p55 (sTNF-R) were determined in non-carcinomatou s pleural effusions. The primary site of disease being the pleura, neutrophil-rich¯uids were designated empyema to differentiate them from parapneumonic neutrophil effusions. Data is given after normalization for total protein content in pleural effusions.
sTNF-R concentrations in relation to the various diagnoses in the study population. We determined that the protein amounts in the pleural transudates were lower than in the pleural exudates in general. However, results varied widely in all patients groups so that the differences did not reach signi cant levels.
Effects of pentoxifylline on in vitro ® broblast proliferation stimulated by TGF-b or TNFa TNFa, at 100 ng/ml, proved to be a stimulant of broblast proliferation (compared with the control: 38%), and addition of pentoxifylline nearly prevented the cells from proliferating altogether (Fig. 2).
TGF-b proved to be a very weak stimulant of broblast proliferation only when cells were already proliferating (i.e. already stimulated by 4% FCS (TGF-b at 3 ng/ml); stimulation was, compared with the control experiment, only 10%). There was no effect at all on quiescent (i.e. non-FCS stimulated) broblasts. Pentoxifylline reduced the TGF-b effect by 50%, P , 0 05 as shown in Fig. 2.
Results have been con rmed by counting the cells in a haemocytometer.

Effects of pentoxifylline on in vitro ® broblast collagen synthesis stimulated by TGF-b or TNFa
Collagen synthesis was clearly stimulated by TGF-b. Addition of pentoxifylline, 50 or 100 mg/ ml, inhibited collagen synthesis, reducing it by 40% when stimulated with TGF-b at 3 ng/ml (P , 0 05). Stimulation with TGF-b at 20 ng/ml was not signi cantly inhibited by POF, although a slight decrease was observed (not shown). TNFa proved to be a weak stimulant of collagen synthesis which pentoxifylline was capable of inhibiting. The results were comparable when collagen synthesis was calculated per broblast.

Effects of pleural effusions on in vitro ® broblast proliferation and collagen synthesis with and without addition of pentoxifylline
Pleural effusions in all patient groups stimulated in vitro broblast proliferation. Pentoxifylline inhibited proliferation signi cantly, P , 0 025 in experiments using pleural effusions as stimulant (see Fig. 4).
In vitro broblast collagen synthesis was also stimulated, varying between 3 and 12% of total protein synthesis. Pentoxifylline inhibited pleural effusion stimulated collagen synthesis signi cantly as shown in Fig. 5 ( P , 0 025). Similar results were obtained when collagen  synthesis was calculated per broblast. Results did not correlate with concentrations of TGF-b, TNFa, or s-TNF-R.

Discussion
Transforming growth factor-b (TGF-b) is considered a key cytokine, sustained synthesis of which underlies the development of tissue brosis. 11 Screening pleural effusions for several proteins considered to be involved in tissue remodelling (namely tumour necrosis factor a (TNFa), soluble TNF receptor p55 (sTNF-R), and TGF-b), we found TGF-b1 to be positively correlated with development of pleural thickenings in patients with non-carcinomatous pleurisy. Prevention and therapy of brosis still being a source of controversy, we searched for alternatives to steroid administration and found evidence that elevating intracellular cAMP, e.g. with pentoxifylline, might be a promising approach.
Our results also highlight the importance of TGF-b in tissue repair. There are three known isoforms of TGF-b in humans (TGF-b 1, 2 and 3) the biological properties of which are nearly identical. 23 TGF-b is strongly chemotactic for broblasts and induces these cells to secrete extracellular matrix proteins. 11 It exhibits autoinduction potency 11 and modulates the actions of platelet-derived growth factor, broblast growth factor, interleukin-1, and TNFa in such a manner that central role in orchestration of tissue brosis can probably be ascribed to it. 23 In an animal model of bleomycin-induced pulmonary brosis, neutralizing antibodies to TGFb isoforms 1 and 2 were able to attenuate the brosing processes. 24 The TGF data presented in this study were obtained from ELISA tests on TGF-b. The decision to use ELISA was methodological: bioassays to determine bioactive TGF-b in pleural effusions did not reveal reproducible data.
Since we were looking for pleural brosis markers, this turned out to be a crucial decision, since it provided an explanation of why TGF-b might indeed indicate the course of disease: the data includes both active and latent, inactive TGF-b. The physiological function of latent, inactive TGF-b is currently the subject of intensive investigations. It appears likely that not only synthesis of TGF-b, but in particular its activation and deactivation, constitutes a major controlling step in tissue remodelling and that, for instance, activated macrophages might create micro-environments at the site of disease that could contribute to activation of latent TGF-b. High concentrations of latent TGF-b are  also known to modulate immune processes. 25 The correlation between TGF-b and pleural thickenings occurring months later opens up perspectives for prevention and therapy of pleural brosis: inhibition of TGF-b might be desirable since it is likely that TGF-b is causally involved in brogenesis. Further, TGF-b might prove to be a tool for estimating pleural brosis risk, although this aspect will have to be addressed in the context of a larger study population.
The poor correlations of TNFa and its soluble receptor sTNF-R p55 with diagnosis or course of disease is not surprising when one considers their brief half-lifes. 26,27 This elucidates two general concerns of ex vivo investigations in that the time of sample-drawing after onset of disease is a matter of chance and studying a single probe does not allow for assessment of future developments unless a longitudinal disease marker is used. Although sustained TNF synthesis might be of importance in chronic brosis, TNF certainly is no marker for chronic developments, but rather a pro-in ammatory cytokine that is activated by multiple mechanisms other than chronic brosing in ammation.
In a search for brosis-modulating drugs, we found pentoxifyllin to inhibit in vitro human lung broblast (WI-38) proliferation and differentiation subsequent to various stimuli as shown in Figs 2± 5. The POF concentrations used to cause inhibition are comparable with those used to inhibit in vitro TNFa formation: oral or i.v. administration of POF in recommended doses causes a similar reduction in TNF synthesis, as do 50 mg/ml POF in the in vitro system. 28 Pentoxifylline also in uences many of the known contributors to brogenesis: acute lung injury induced by chemicals and in ammatory mediators is attenuated. 29 Pentoxifylline seems to be protective of pneumocyte function. 30 It inhibits the formation of free radicals effectively 31 as well as formation and action of TNFa, 28 and injury of lungs perfused with human neutrophils. 32 Pentoxifylline inhibits in vitro broblast proliferation driven by different stimuli (fetal calf serum, 18 platelet-derived growth factor, 19 tumour necrosis factor a 14 ), and inhibits synthesis of several broblast products like collagen, glycosaminogylcans and bronectin. 14,18 Pentoxifylline also augments in vitro collagenase production by broblasts. 18 Our own data add that (1) TGF-b is most likely a relevant cytokine in pleural brosis; (2) pentoxifylline inhibits the effect of transforming growth factor-b on brogenesis; (3) the effects, not only of selected cytokines, but even of uids containing a complex heterogeneity of mediators like pleural effusions, can be inhibited by pentoxifylline in concentrations easily attainable in vivo. This is important since, although steroids have been administered in brosis for a long time, whether they exert direct inhibitory in uence on broblasts remains controversial: the in vitro data, at least, depend heavily on the culturing methods employed. 33 Thus pentoxifylline may exhibit an advantage in that it inhibits not only in ammatory mediators (as do corticosteroids) but brogenesis as well. The inhibition of TGF-b-driven collagen production by pentoxifylline deserves special emphasis since, in a rat model of bleomycin-induced pulmonary in ammation, raised levels of TGF-b1 synthesis by alveolar macrophages was not suppressed by high-dose steroid treatment. 34 This might be one line of explanation for the limited ef cacy of steroid treatment in idiopathic pulmonary brosis.
In conclusion, we suggest that xanthinesespecially pentoxifylline-might be effective in prevention and therapy, not only of pleural brosis but of other brosing disorders as well. There are many results which seem perfectly suited to the actions of xanthines and the physiopathology of brosis. We feel that sufcient evidence now exists to propose pentoxifylline for a prospective therapeutic intervention study in human disease.