Cytokine & Growth Factor Reviews 11 (2000) 133±145
www.elsevier.com/locate/cytogfr
TGF-b in blood: a complex problem David J. Grainger a,*, David E. Mosedale a, James C. Metcalfe b a
Department of Medicine, University of Cambridge, Box 157, Addenbrookes Hospital, Hills Road, Cambridge CB2 2QQ, UK b Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge CB2 1QW, UK
Abstract The cytokine transforming growth factor-b (TGF-b) was initially puri®ed from human platelets, a rich source of this protein. In addition to platelets, TGF-b1 is also found in other blood fractions, including plasma and the circulating leukocytes. However, more than 15 years after the initial isolation of TGF-b1, there remains no consensus on how much TGF-b1 is present in normal human plasma. Here we review the diculties associated with measuring TGF-b concentrations in complex biological ¯uids, and discuss the current state of knowledge on the distribution of TGF-b isoforms in various blood fractions as well as the nature of the TGF-b-containing protein complexes. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: TGF-b; Plasma; ELISA; Wound healing; Blood clot
Contents 1.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
2.
Levels of TGF-b in normal human plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.
Initial characterisation of the TGF-b1 complexes in human plasma . . . . . . . . . . . . . . . . . . . 137
4.
Where does plasma TGF-b come from? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5.
What is the mechanism of clearance of TGF-b1 from the plasma?. . . . . . . . . . . . . . . . . . . . 139
6.
What controls the levels of plasma TGF-b? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.
Is there any active TGF-b in normal human plasma? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
1. Introduction Abbreviations: sRII, soluble extracellular domain of the TGF-b type II receptor; TGF-b, transforming growth factor type beta. * Corresponding author. Tel.: +44-1223-336812; fax: +44-1223762770. E-mail address:
[email protected] (D.J. Grainger).
The transforming growth factor-b (TGF-b) superfamily is a collection of structurally related multi-functional cytokines that have been implicated in a wide range of physiological and pathological processes,
1359-6101/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 9 - 6 1 0 1 ( 9 9 ) 0 0 0 3 7 - 4
134
D.J. Grainger et al. / Cytokine & Growth Factor Reviews 11 (2000) 133±145
including wound healing, development, oncogenesis, immunomodulation and atherosclerosis [1]. The prototypical member of this superfamily, TGF-b1, was originally identi®ed as a growth factor for transformed cells, but was ®rst puri®ed to homogeneity from human platelets [2]. There are now known to be three closely related mammalian TGF-b isoforms (TGF-b1, -b2 and -b3) which are thought to have similar functions, at least in vitro, although less is known about TGF-b2 and TGF-b3. Each of the three isoforms is produced as a pre-pro-protein which rapidly dimerises. After loss of the signal sequence the dimer is further processed by addition of sugar moieties to the propeptide region (known as the latency-associated peptide; LAP). In addition, there is proteolytic cleavage between the LAPs and the mature dimer (which, when released from the LAP dimer, is able to bind to the TGF-b signalling receptors). After cleavage, however, the LAP dimer usually remains non-covalently associated with the mature dimer forming a complex known as the small latent complex [3]. Either prior to secretion or in the extracellular milieu the small latent complex can bind to a wide range of other proteins forming a large number of higher molecular weight latent complexes. The best characterised of these proteins are the latent TGF-b binding protein family (LTBP1-4 and ®brillin-1 and 2) [4,5]. For example, in platelets, the majority of the TGF-b present is thought to exist in a latent complex consisting of the LTBP-1 covalently coupled to the small latent complex. This complex is called the large latent complex. Once in the extracellular environment (such as the extracellular matrix of solid tissues or in the blood plasma), the latent complexes, possibly associated with various accessory proteins, must be activated in order to exert their biological eects. It is assumed that this activation process involves release of the mature dimer from its association with the LAP dimer. A number of pathways are likely to participate in TGF-b activation, such as cleavage of the LAP dimer by proteases, including plasmin [6,7], and conformational changes in the latent complex induced by binding to matrix components such as thrombospondin [8,9] or certain integrins [10]. Once activated, the TGF-b is able to signal through cell surface receptors, of which the best studied are the type I and type II TGF-b receptors (TbRI and TbRII) which can form heteromultimers in response to ligand binding resulting in an active signalling complex which is thought to be responsible for many of the cellular responses to TGF-b [11]. TGF-b1 has a very wide range of activities in vitro. For example, TGF-b regulates important cellular functions such as rate of proliferation and production of extracellular matrix proteins by a wide range of cell types [1]. Thus, misregulation of TGF-b has been proposed to play a key role in the development of a num-
ber of diseases in which the normal adult tissue architecture is progressively lost, including scarring during wound repair [12], carcinogenesis [13], atherosclerosis [14], osteoporosis [15,16] and neurodegenerative diseases [17]. As a result of the wide range of activities attributed to TGF-b, a number of groups have investigated whether circulating levels of TGF-b1 might be altered in various disease states. With only one exception, all of these studies agree that TGF-b1 is found at detectable levels in plasma from healthy human subjects (see Table 1). Moreover, plasma TGF-b1 concentrations markedly diered (by as much as 10-fold) in subjects suering from various diseases, including various cancers [18±21], autoimmune disorders [22,23] and atherosclerosis [24,25], compared with control subjects. Based on these early studies, it seems likely that plasma levels of TGF-b1 may be a useful diagnostic criteria for the presence of one or more of these dieases. In addition, it remains possible, but unproven, that altered plasma levels of this multifunctional cytokine participates in disease progression, rather than simply acting as a marker for disease status. If such a pathophysiological role for plasma TGF-b1 is proven, it could become both a prognostic indicator of future risk of disease and a target for therapeutic intervention. 2. Levels of TGF-b b in normal human plasma Despite the interest in measuring levels of TGF-b1 in plasma as a result of early reports suggesting associations between plasma TGF-b1 concentration and disease, there is still no consensus on the concentration range of TGF-b1 in normal human plasma. The levels of TGF-b1 protein in plasma from various groups of normal subjects have been measured in more than 20 studies in the literature to date, yet the mean (or median) values reported range from below 0.1 ng/ml to more than 25 ng/ml (Table 1). There are a large number of factors which are likely to contribute to the greater than 100-fold range of reported values for TGF-b1 protein concentration in plasma from healthy individuals. These factors can be broadly divided into three classes; the method of selection of the normal subjects studied, the methods used to prepare the plasma and the wide variations in the methods used to assay TGF-b levels in the samples. The groups of normal human subjects who have been studied to date represent highly selected populations, usually disease-free control populations for comparison to a speci®c group of individuals under study. For example, the control subjects in a study of osteoporosis might all be female and aged over 55, while subjects in a study of type I diabetes may be of either gender and under the age of 30. Thus, variations
D.J. Grainger et al. / Cytokine & Growth Factor Reviews 11 (2000) 133±145
in plasma TGF-b1 levels with age, gender or racial background may contribute signi®cantly to the wide range of mean values reported. To date, however, in a number of unpublished studies we have measured TGF-b1 protein concentrations in plasma from over a 1000 individuals and have found no evidence for any consistent variations in plasma TGF-b1 protein levels with age, sex or geographical location (Fig. 1). Hence it seems unlikely that selection of the study populations has contributed more than a small fraction of the variation in levels of TGF-b1 reported in the literature. Preparation of the plasma samples is a more likely source of variation. During plasma preparation it is dicult to prevent a small amount of platelet degranulation from occurring, and because platelets are a rich source of TGF-b1 [2] this can lead to an overestimate of the plasma concentration of TGF-b1. Indeed, in serum, where all the platelets have been allowed to degranulate, more than 90% of the serum TGF-b1 originates from the platelets. As a result, studies which measure TGF-b1 protein concentration in serum samples are primarily measuring the platelet content of TGF-b1 and its ability to be released during clotting. Such studies are therefore unlikely to detect any variations in the concentration of TGF-b1 protein circulating in the plasma prior to drawing the blood. An exception to this general observation is for studies where the assay being used does not detect the platelet large latent complex of TGF-b1. Measuring plasma TGF-b1 levels therefore requires the use of speci®c protocols designed to minimise platelet degranulation, as recommended by Wake®eld and colleagues [26]. Speci®cally, blood should be drawn without use of a
135
tourniquet through a wide-gauge butter¯y needle into Diatube1 H plasma preparation tubes. When plasma is prepared in this manner there is typically less than 0.1% platelet degranulation (based on measurement of platelet factor-4 concentrations). In contrast, blood drawn through a narrow gauge needle into 3.8% sodium citrate as the anticoagulent may undergo as much as 10% platelet degranulation during plasma preparation (data not shown). Inadequate control of platelet degranulation may therefore contribute signi®cantly to the variation in plasma TGF-b1 levels reported. Platelets are not the only blood cells to contain TGF-b, and it is possible that both hemolysis (red cell lysis) and contamination with leukocytes could lead to an over-estimate of TGF-b protein levels in plasma. Despite these caveats, when high quality plasma (with minimal platelet degranulation or other cellular contamination) is prepared, signi®cant concentrations of TGF-b1 protein are still detected [26]. This suggests that TGF-b1 complexes of some nature do circulate in the blood plasma prior to drawing the sample and that the TGF-b1 detected is unlikely all to be an artefact of plasma preparation. The greatest potential source of variation in reported plasma TGF-b1 concentrations, however, is the selection of assay methodology. Since the ®rst assays for the measurement of TGF-b1 in plasma were reported by Danielpour and colleagues a decade ago [27±29], there have been a wide range of assays developed for measuring TGF-b1, but only a few attempts to standardise these assays against one another [30]. The assays fall broadly into two classes; bioassays and
Table 1 Plasma TGF-b1 concentrations in normal populations reported in the literaturea Plasma TGF-b1 protein (ng/ml), mean2SD
Method
Reference
< 0.1 0.520.2 0.620.1 1.420.8 1.520.2 2.5 2.621.1 3.120.4 3.822.9 4.320.3 5.323.3 6.122.0 6.522.3 9.722.6 9.7211.3 25221
ELISA (Genzyme) ELISA Bioassay (MvLu) ELISA ELISA (Quantikine) Bioassay (MvLu) ELISA and Bioassay ELISA (Quantikine) ELISA ELISA Bioassay (MvLu) Bioassay (MvLu) ELISA ELISA ELISA ELISA (Genzyme)
Snowden et al. [23] Higley et al. [69] MuÈller et al. [70] Shirai et al. [18] Hayasaka et al. [71] Zauli et al. [72] Murase et al. [73] Pfeier et al. [22] Junker et al. [52] Kong et al. [21] Shirai et al. [74] Anscher et al. [75] Szymkowiak et al. [76] Grainger et al. [35] Grainger et al. [24] Claudepierre et al. [77]
a The mean (2 standard deviation) TGF-b1 concentration in plasma from apparently healthy individuals reported in 16 dierent studies are shown. A variety of dierent ELSA techniques have been used: where the ELISA used is commercially available, the manufacturer is shown in parentheses; in all other cases the ELISA method is described in the report. Bioassays measured TGF-b1-mediated inhibition of DNA synthesis in MvLu cells are previously described.
136
D.J. Grainger et al. / Cytokine & Growth Factor Reviews 11 (2000) 133±145
antibody-based assays such as ELISAs. Bioassays for TGF-b1 have used either mouse HT-2 cells [31] or more commonly the mink lung epithelial cell line CCL64 [32,33]. TGF-b bioactivity is measured by its ability to suppress the growth of these cells lines (assessed by tritiated thymidine incorporation into DNA). Recently, however, these assays have been largely superseded by the more reproducible and potentially more speci®c 3P-Lux system developed by van Waarde and colleagues [34] in which a promoter construct from the PAI-1 gene is coupled to a luciferase reporter gene and transfected into the mink lung epithelial cells. This promoter responds to TGF-b activity by upregulating transcription of the luciferase reporter gene which can readily be assayed. Bioassays have the advantage that they report an operational de®nition of TGF-b activity which does not depend on any knowledge of the molecular composition of the active factor. However, to be used as a measure of total TGF-b1 protein concentration (active plus latent forms) all latent complexes, irrespective of composition, require the quantitative conversion to form(s) of TGF-b1 which possess bioactivity. Furthermore, any activated form(s) of TGF-b1 generated must all possess the same degree of bioactivity as the standard under the conditions of the assay. ELISA assays have also been developed because they are easier to perform on large numbers of samples when compared with bioassays. In most cases, a capture agent capable of binding TGF-b1 is coated onto a plastic well. After incubation with the plasma sample the TGF-b1 is retained while other proteins are washed away. The bound TGF-b1 is then detected, usually using a labelled antibody. However, a wide range of such assays have been described which use a variety of dierent capture reagents and detection antibodies [27,29,30,35]. How much of each TGF-b1 complex is detected by a given assay depends on the speci®city and anity of the particular reagents used. Thus, although many of these assays have been fully validated for the detection of puri®ed mature TGF-b1 dimer, it is unclear what form(s) of TGF-b1 they detect in a complex biological ¯uid such as plasma. An important ®rst indication of the diculties of measuring TGF-b1 in complex biological ¯uids came from Danielpour's earliest studies. He found that serial dilution of biological ¯uids containing TGF-b1 did not yield a linear decrease in signal in an ELISA assay unless the TGF-b1 was ®rst extracted and activated (using ethanol and acid) prior to performing the assay [28]. One possible explanation for this observation is that TGF-b1 exists in plasma in one or more complexes non-covalently bound to accessory proteins and that during dilution these complexes (which were not detected by the antibodies used in the ELISA) then dissociate resulting in anomalously high readings as
the dilution increases. The problem of non-linear dilution curves is also avoided by pre-treatment of the sample with 1 M acid and 5 M urea prior to assay, as used in the Quantikine TGF-b1 assay (R&D Systems). Both for this reason, and because the treatment with acid and urea is relatively quick and simple compared with procedures for extracting TGF-b1 from biological ¯uids, we use the Quantikine TGF-b1 assay for the majority of our studies.
Fig. 1. Variations in plasma TGF-b1 protein levels with age, sex and geographical location. The plasma concentration of TGF-b1 was determined using the Quantikine ELISA kit (R&D Systems) in accordance with the manufacturer's instructions. Data are taken from a number of dierent unpublished studies which have been combined for meta-analysis (n = 1006) of the association between plasma TGF-b1 protein concentration and age (a). There was no statistical signi®cant association (r=ÿ0.040; P = 0.26). The data were also analysed (mean2SEM) split by sex and by geographical location in which the subjects resided (b). No signi®cant dierences were observed (ANOVA; P > 0.05).
D.J. Grainger et al. / Cytokine & Growth Factor Reviews 11 (2000) 133±145
Once an antigen under study is known to exist in more than one form, there are two approaches to measuring the total amount of antigen present in a sample. The ®rst approach is to use an assay which is known to detect all of the dierent forms equally, and the second is to quantitatively convert all the complexes into a single form which can then be measured. All the published methods used to measure total TGFb1 protein in a sample (whether ELISAs or bioassays) depend on one or other of these approaches. The diculty with the ®rst of these approaches lies in identifying an assay which detects all complexes containing TGF-b. Without an assay which is known to detect all TGF-b complexes, it is dicult to characterise the array of TGF-b complexes which are present in vivo. Furthermore, without a complete characterisation of all these TGF-b complexes, it is impossible to determine if a given assay detects all complexes. This circular argument hampers progress towards a de®nitive measurement methodology once the antigen under study (in our case, TGF-b1) is known to exist in more than one form. The problem with the second approach is that all latent complexes must be converted to ones with activity. However, we do not presently know under what conditions each of the latent complexes which are present in vivo might be activated. Furthermore, under the conditions of the assay we must be sure that all complexes are converted to form(s) which have equal biological activity to the standard solution of TGF-b1 used to calibrate the assay. As a result of these fundemental problems in measuring an antigen which exists in more than one form, it may be dicult to reach a consensus on the level of total TGF-b1 protein in normal human plasma. In the interim, however, comparison of studies between groups may be substantially aided if a single methodology, such as the Quantikine assay which being a commercial reagent is available to all, is used universally unless it is inappropriate for the scienti®c question being addressed. However, this raises the spectre of a future problem Ð if the majority of studies are performed using a single assay, will some TGF-b complexes not detected by this assay remain forever hidden?
137
(assessed using the Quantikine TGF-b1 assay). Any TGF-b1 complexes which are not detected by the Quantikine TGF-b1 ELISA are, by de®nition, excluded from this analysis. The bound TGF-b1 can then be eluted by increasing the salt concentration in a gradient from 20 mM to 500 mM. The fractions collected as the salt concentration rises can then be assayed, after treatment of the sample with acid and urea, for the presence of TGF-b1 complexes using the Quantikine assay. When platelet lysate is analysed in this way, we detect a single broad peak eluted by approximately 275 mM salt (Fig. 2a and [36]). In contrast, in plasma we ®nd at least ®ve dierent peaks
3. Initial characterisation of the TGF-b b1 complexes in human plasma We have recently used column chromatography to begin to address the question of the molecular composition of the TGF-b1 complexes in plasma [36]. When plasma, serum or platelets are applied to an anion exchange column and then unbound material washed away, all the TGF-b1 is retained on the column
Fig. 2. Separation of TGF-b1 complexes in various blood fractions. Samples of platelets (a), platelet-poor plasma (b) and serum (c), prepared from the same blood sample from a single individual (male, age 45), were analysed by anion exchange chromatography, exactly as previously described [36]. Fractions containing two distinct TGFb1 complexes (labelled I and III) were subject to further analysis (see Fig. 3).
138
D.J. Grainger et al. / Cytokine & Growth Factor Reviews 11 (2000) 133±145
containing TGF-b1 protein (Fig. 2b and data not shown). As expected, analysis of serum yielded the same peaks as those seen in plasma, plus the additional platelet peak (Fig. 2c). We have begun to analyse the composition of the complexes responsible for the major peaks present in the anion exchange pro®le of normal human plasma. However, the concentration of TGF-b1 in the column fractions is very low (
