Neurobiology of Cytokines: Methods in Neurosciences, Vol. 16

Narahashi

Section I

General Aspects

1

Pharmacological, Biochemical, and Molecular Biological Studies on Cytokine Receptors

Steven K. Dower

Publisher Summary

The largest family of cytokine receptors is termed as hematopoietin or type I cytokine receptor family. It includes the receptors for interleukin 2 (IL-2), interleukin 3 (IL-3), interleukin 4 (IL-4), interleukin 5 (IL-5), interleukin 6 (IL-6), interleukin 7 (IL-7), interleukin 9 (IL-9), granulocyte-macrophage colony-stimulating factor (GMCSF), granulocyte colony-stimulating factor (GCSF), leukemia inhibitory factor (LIF), oncostatin M, ciliary neurotrophic factor (CNTF), erythropoietin (EPO), growth hormone, and the product of a receptor-like protooncogene (c-mpl) found after the isolation of its viral counterpart (v-mpl); no ligand has yet been identified for this molecule. Within this family, a subgroup of large cytokine receptors can be discerned that is composed of the receptors for GCSF, LIF, and oncostatin M. The oncostatin M receptor polypeptide, called gpl30, is also the β chain of the LIF and IL-6 receptors. All of the type I receptors share a domain structure in common, consisting of approximately 130 residues in the extracellular ligand-binding region, which is repeated two or more times.

Introduction

The molecular characterization of cytokine receptors has proceeded rapidly over the last 5 years, in large part due to the advent of expression cloning methods of sufficient sensitivity to allow the isolation of cDNAs corresponding to rare mRNAs encoding receptors expressed at low levels in tissues or cultured cell lines (1–4). The expression cloning methods in turn have been dependent on the characterization of the binding properties of the receptors and some estimation of the molecular size of receptor subunits. These data are obtained by more classic receptor biochemical techniques, which will be covered in detail in the next section of this chapter.

The primary sequence data that have been obtained from the cloning studies have identified five separate families of cytokine receptors (4). The largest family has been termed the hematopoietin or type I cytokine receptor family (5), and includes the receptors for interleukin 2 (IL-2) (6, 7), IL-3 (8), IL-4 (9), IL-5 (10), IL-6 (11), IL-7 (12), IL-9, granulocyte–macrophage colony-stimulating factor (GMCSF) (13), granulocyte colony-stimulating factor (GCSF) (14), leukemia inhibitory factor (LIF) (15), oncostatin M (16), ciliary neurotrophic factor (CNTF), erythropoietin (EPO) (17), growth hormone (18), and the product of a receptor-like protooncogene (c-mpl) (19) found after isolation of its viral counterpart (v-mpl); no ligand has yet been identified for this molecule. Within this family, a subgroup of large cytokine receptors can be discerned, composed of the receptors for GCSF, LIF, and oncostatin M. The oncostatin M receptor polypeptide, called gp130, is also the β chain of the LIF and IL-6 receptors (16, 20). In addition, one subunit of the IL-12 heterodimer is clearly related to the type I cytokine receptors (21). All of the type I receptors share in common a domain structure consisting of approximately 130 residues in the extracellular ligand-binding region, which is repeated two or more times. The membrane-proximal domain of type I receptors has a characteristic sequence WSXWS close to its C terminus and an imperfect version of this motif can also be found in the more membrane-distal regions. In addition, several of these receptors, most notably the large subset, contain fibronectin type III motifs. The solution of the X-ray structure of the complex of a soluble fragment of the growth hormone receptor with growth hormone confirms that the hematopoietin domain sequence repeats do indeed correspond to relatively independently folded structures (18).

The receptors for the interferons (IFN-α, IFN-β, and IFN-γ) form a distinct second set of two receptors (IFN-α and IFN-β share a common receptor) that are distantly related to one another, and even more distantly related to the type I receptors (4). There is indirect evidence that the functional receptors include additional subunits. For example, the human IFN-γ gene, which maps to chromosome 6, will not reconstitute signaling when transfected into murine cells unless those cells also contain a copy of human chromosome 21.

The tumor necrosis factor (TNF) receptors [type I or p60 (22, 23) and type II or p80 (24)], the low-affinity nerve growth factor (NGF) receptor (25), and the cell surface proteins CD40 (26), CD27 (27), FAS, CD30 (28), and OX40 (29) form another family characterized by an approximately 40-residue cysteine-rich repeat in the extracellular ligand-binding region. The TNF and NGF receptors were originally identified as receptors for soluble ligands, but cloning of TNF-α revealed that it is synthesized as a membrane-bound precursor of the type II receptor (30), and the soluble form is generated by proteolytic cleavage. A type II integral membrane protein ligand has been identified for the CD40 antigen (31); whether this or a soluble cleavage product is the biologically relevant form remains to be established. No ligands have been identified for the other members of this family of putative receptors.

There is also a group of cytokine receptors that have extracellular regions composed of approximately 110-residue immunoglobulin-like domains. This family includes the two IL-1 receptors (type I and type II) (2, 32), the macrophage colony-stimulating factor receptor (MCSF or CSF-1) or c-fms protooncogene product (33), and the steel factor/mast cell growth factor/stem cell factor receptor or c-kit protooncogene product (34). These last two receptors are structurally related to one another and to the two platelet-derived growth factor (PDGF) receptors (α and β). In addition, fibroblast growth factor (FGF) receptors also belong to this group of immunoglobulin-like regulatory factor receptors (35). The subgroup composed of c-Fms, c-Kit, the PDGF receptors, and some forms of FGF receptor all have large cytoplasmic domains with intrinsic protein tyrosine kinase activity. This subgroup is further distinguished from other receptor protein tyrosine kinases such as the epidermal growth factor receptor by having an insert in the kinase domain. The evolutionary relationship between them is underscored by the finding that the genes for c-Fms and PDGF receptor β subunit (PDGFRβ) map next to one another on human chromosome 5 (36).

Finally, cDNA clones have been isolated for IL-8 receptors (37). Interleukin 8 is a member of a large family of mediators termed small inflammatory cytokines. The IL-8 receptor is a member of the β-adrenergic receptor family, being composed of seven membrane-spanning regions connected by a series of short loops. Like other members of this receptor family, IL-8 receptors activate intracellular signals by coupling through heterotrimeric G proteins. Presumably, the other members of the small inflammatory cytokine family bind to receptors of similar structure.

In the preceding summary the receptors for cytokines were described as if each polypeptide were a discrete entity. This is not the case. In most instances it has become clear that these receptor polypeptides are subunits of multi-chain complexes. Indeed, in many instances individual chains are shared between receptors, so that binding of a cytokine to a cell leads to recruitment of subunits from a common pool in a combinatorial fashion to form a final structure that transduces the signal for that cytokine. Thus, for example, the receptors for IL-3, IL-5, and GMCSF are formed by association of cytokine-specific α chains with a common β subunit (38, 39), and the receptors for IL-6, LIF, oncostatin M, and CNTF are formed by combinatorial assembly from the subunits IL-6Rα, gp130, LIFR, and CNTFR [the IL-6R being IL6Rα/gp130 (20), the LIFR and oncostatin M receptors being gp130/LIFR (16), and the CNTFR being gp130/LIFR/CNTFR]. There is also evidence that as these complexes assemble, in a ligand-driven fashion, that other subunits associate with the complex. Thus when NGF binds to its receptor, a tyrosine kinase subunit, the c-trk protooncogene product, binds to the complex and is involved in transducing signals (40). There is also a growing body of evidence that following ligand binding, there are subunits with protein tyrosine kinase activity that associate with many of the hematopoietin family of receptors (41, 42). In addition to heterologous cross-linking of subunits, it is clear for many cytokine receptor systems that ligand binding leads to homologous cross-linking. For example, TNF is a trimer and has been shown to cause dimerization and trimerization of its receptors when it binds. Similarly, PGDF in all three forms (AA, AB, and BB) is a dimer and causes receptors to dimerize when it binds (43). Furthermore, in what may well be a paradigm for the hematopoietin receptor family, the binding of growth hormone to its receptor leads to receptor dimerization (18). The notion that lateral aggregation is a general mechanism for transmembrane signaling has been well established for many years. It now appears that cytokine receptors also employ this mechanism to deliver signals.

Finally, an intriguing finding in the cytokine receptor field is the existence of soluble receptors in many systems. It is clear that alternatively spliced mRNAs exist that encode soluble forms of a number of cytokine receptors. For example, soluble forms of IL-5R (44), IL-7R (12), IL-1R type II (45), IL-2Rα, IL-6R, IFN-γR, TNFRI, and TNFRII have all been found (46, 47).

In this last group of systems, no evidence for alternatively spliced mRNAs encoding these soluble receptor forms has been detected and the presumption is that they are derived from the integral membrane-bound forms by proteolysis. The physiological functions of soluble forms of cytokine receptors remain to be established. It has been suggested, on the basis of experiments in which pharmacological doses of soluble receptors administered to animals can be shown to inhibit cytokine action (48), that endogenously produced soluble receptors act as antagonists, but this remains to be proved. In at least one case this is not so; the soluble IL-6 receptor α chain when complexed with IL-6 will act as an agonist by signaling through the IL-6 receptor β chain (gp130) (49). Neither soluble IL-6 receptor α chain nor IL-6 alone are capable of this; thus the heterodimeric complex might be regarded as a ligand for gp130, allowing cells that express this protein but not the IL-6 receptor α chain to respond to IL-6. The IL-12 receptor system is a different variation on this theme. Interleukin 12 is a stable heterodimer of a chain that is homologous to IL-6 and a chain that is homologous to IL-6 receptor α chain (21). The IL-12 receptor, for which no cDNA clones have yet been isolated, has a size, as estimated by cross-linking, that suggests that it will be related to the gp130/LIFR/GCSFR subgroup of the hematopoietin receptor family.

The remainder of the chapter focuses on the methods that have been used to generate the data reviewed above.

Analysis of Binding Properties of Cytokine Receptors

Production and Radiolabeling of Recombinant Cytokines

Most of the studies of the binding properties of cytokine receptors have been carried out with recombinant cytokines. Cytokines purified from natural sources were used in early studies. Thus, for example, the original experiments with IL-1 and IL-2 receptors were done with natural forms of the molecules purified from activated human peripheral blood monocytes (IL-1β) (50) and human Jurkat T cell (IL-2) (51) supernatants, respectively. In general, this is not a practical approach as most cytokines have high specific biological activities and are hence produced in small amounts by natural sources. Thus, to generate the quantities (10- to 100-µg range) needed for chemical labeling requires large amounts of natural starting materials. Recombinant cytokines have been generated from cloned cDNAs in a variety of expression systems, yeast and Escherichia coli being the most common hosts. A description of the expression vectors, host systems, fermentation procedures, and purification schemes is beyond the scope of this review.

Purified recombinant cytokines in quantities sufficient for labeling can be obtained from a variety of commercial sources.

The most commonly employed radiolabeling methods utilize ¹²⁵I as a tracer and introduce the label into the protein with either Bolton–Hunter reagent, which labels lysine ε-amino groups, or the Enzymobead (glucose oxidase-lactoperoxidase; Bio-Rad, Richmond, CA) method, which labels tyrosine residues in the ortho position relative to the hydroxyl on the phenol ring. Typical procedures would be as follow.

Bolton–Hunter Reagent

One to 10 µg of cytokine (e.g., IL-1β) in 10–20 μg of borate (0.05 M, pH 8.5)-buffered saline (0.15 M) is labeled with 1 mCi (0.23 nM) of ¹²⁵I-labeled diiodo-Bolton–Hunter reagent (New England Nuclear, Boston, MA). The reagent is supplied as a dry benzene solution and is prepared for use by evaporation of the solvent with a stream of dry nitrogen. The reagent is hydrolyzed and hence inactivated by moisture in the air; thus, the reagent must be used immediately after the vial is opened, with minimal handling. The protein solution is introduced into the vial containing the dried reagent and the reaction is allowed to proceed at 4–8°C for 30 min or overnight. Subsequently, 30 μl of 2% gelatin is added as a carrier [the reagent will bind noncovalently to bovine serum albumin (BSA)], and the labeled protein is separated from the hydrolyzed label by gel filtration on a 1-ml bed volume column with either Sephadex G-25 or BioGel P10. The column is preblocked with protein by running 2 ml of a 10% BSA or gelatin solution through, and then washing with 10–20 ml of PBS. The labeled protein is eluted with PBS and collected in 0.1-ml fractions. The ¹²⁵I-labeled protein will elute in fractions 3 and 4. The specific activity estimated for the protein (cpm/unit of protein) will clearly be dependent on the recovery. The most convenient way to estimate this is to carry the same amount of unlabeled cytokine through the procedure after spiking it with a small amount of ¹²⁵I-labeled cytokine, and either omit the Bolton–Hunter reagent or use noniodinated Bolton–Hunter reagent [3-(p-hydroxyphenyl)propionic acid N-hydroxysuccinimide ester] in the first step. Protein recovery can be estimated as input counts/output counts. Because this is an external, not internal, correction, the procedure should be recalibrated with each new cytokine or at any time when the protocol is modified significantly (e.g., scaled up, scaled down, column material changed, or source of cytokine changed). An alternative is to measure the total recovery of protein (i.e., carrier). This is also, of course, subject to sources of error. The presumption is that the labeled cytokine will behave like the carrier in sticking to glass, the column bed, and so on. In addition, if any blocking protein leaches off the column it can lead to an overestimate of recovery. Furthermore, for gelatin carrier, absorbance at 280 nm is a poor method for estimating concentration, because the tryptophan content of this protein is low. Chemical methods, which are more sensitive and accurate, consume labeled protein, and also involve handling of relatively high quantities of label. For all these reasons, we prefer the simpler external estimation method. Finally, it is important to check that all the label in the preparation is covalently attached to protein by performing a trichloroacetic acid precipitation (10%) on a aliquot. For IL-1β, this method routinely gives 1–3 × 10¹⁵ dpm/mmol or approximately 0.5–1.5 atoms ¹²⁵I per molecule protein.

Enzymobead (Glucose Oxidase-Lactoperoxidase) Method

Protein [1–10 µg in 50 μl of sodium phosphate (0.2 M, pH 7.2)] is combined with 50 μl of Enzymobead reagent reconstituted according to manufacturer (Bio-Rad) instructions. Then 20 μl of the same buffer containing 2 mCi (0.8 nmol) of Na¹²⁵I is added and the reaction is started by adding 10 μl of β-dglucose, and allowed to proceed for 10 min at room temperature. The reaction is terminated by adding sodium azide (20 μl, 25 mM) and then sodium metabisulfite (10 μl, 5 mg/ml). Incubation is continued for a further 5 min at room temperature to allow enzyme inactivation and reduction of residual peroxide and I3+ to occur. The entire reaction mixture is applied to a small column and separated as described above for the Bolton–Hunter reaction, except that BSA (1%, w/v) may be used as a carrier. The initial characterization of the preparation is done essentially as described for the Bolton–Hunter method.

There are several other radiolabeling techniques that we have used less frequently in the laboratory: these include two other tyrosine-directed methods, a modified chloramine-T method, and the Iodogen (Pierce) method. In principle, these two methods and the Enzymobead method described above, by oxidizing I– and generating I3+, the species that gives rise to electrophilic substitution at the ortho position relative to the hydroxyl on tyrosine, should be equally effective at labeling any protein. Many cytokines that can be labeled with full retention of activity by the Enzymobead method suffer significant losses of activity with chloramine-T, which is a strong oxidizing agent that can damage proteins through side reactions; Iodogen is intermediate in harshness, as is the immobilized form of this reagent (Iodobeads). Finally, we have on occasion used Woods reagent to label the N-terminal free α-amino group.

The concern over labeling damage necessitates determination of the activity of the labeled relative to the unlabeled cytokine. Ideally, the labeled material should be 100% active relative to the unlabeled material both in a standard bioassay and in a radioreceptor assay. In practice, this may not always be possible and in that case the use for which the material is intended should be considered. The requirement for 100% bioactivity of the cytokine can be applied less than rigorously for at least two reasons. First, many bioassays are simply not quantitative enough to be able to determine with any confidence that two preparations of cytokine have relative activities that are closer than a factor of 2. Furthermore, because most bioassays are carried out for 12 hr or more at 37°C, factors such as stability to proteases and intracellular trafficking pattern might affect the activity of a molecule in the assay. It is therefore reasonable to argue that even if there is some loss of bioactivity, the reagent should be satisfactory for characterizing receptor binding if the affinity constant of the labeled protein [KA (M–1)] is the same as the inhibition constant of the unlabeled protein [KI (M–1)] under the conditions of the receptor-binding assays.

Another measure that is often used to determine the quality of a labeled preparation of cytokine is the maximum percentage of radioligand that can be bound to receptor-bearing cells. The format of the experiment is simply to choose a concentration of labeled protein that is 10–30% of the KD (i.e., 1/KA) and to titer cells bearing receptors against the labeled ligand. By plotting % cpmbound [bound cpm/(bound cpm + free cpm)] vs cell concentration or, more quantitatively, by plotting % cpmbound/cell concentration (y) vs % cpmbound (x), essentially a Scatchard plot, one can extrapolate to percentage label bound at infinite cell concentration. This is a useful parameter to measure, and clearly if percentage label bound at infinite cell concentration is significantly less than 100%, then quantitative data obtained with that material should be regarded as suspect. However, if this value is low, it is difficult to make a simple correction to data obtained with the preparation in the absence of several other pieces of information:

1. Is the failure of the labeled preparation to bind 100% a consequence of damage during the labeling or due to a preexisting inactive fraction of protein?

2. Is there a correlation between the extent of labeling and activity? In general, one might assume that more heavily labeled molecules might be less active. However, in the absence of data this remains an assumption.

3. What is the average level of labeling? If, for example, labeling does indeed decrease the binding activity of a cytokine to 10% of that of the unlabeled material, and in the simplest case there is one site on the protein that can be modified, the preparation will have an apparent KA that is the same as that of the unlabeled protein but will give an apparent maximal level of binding that is only 10% of the true value. By contrast, if the same labeling procedure yields a preparation that is 90% modified, the apparent KA will be 10% that of the unlabeled protein but the estimated maximal binding will be approximately correct.

4. At an average level of labeling, what is the distribution of label? It is possible that two different cytokines labeled by the same method might have different fractions of a labeled preparation, with zero, one, two, or more atoms of iodine per molecule of protein, with an average of one atom per molecule.

5. Are the receptors being studied themselves a homogeneous population, or are there subpopulations of receptors that can differentially bind subpopulations of the ligand preparation?

It is of course possible to write down systems of equilibria that model each of these situations, and determine how one might correct the binding data. It is frankly best to attempt to find a source of cytokine/labeling method that yields a relatively high fractional substitution, at least as estimated by atoms iodine per molecule protein, while retaining a KA that is close to the KI of the unlabeled material.

Binding and Kinetic Assays: Phthalate Oil Method

In this section, the approach used for characterizing the binding properties of receptors is summarized. The method that we routinely use is based on a phthalate oil technique developed by Segal and Hurwitz (52). A typical experiment is carried out as