Discovery of a Cytokine and Its Receptor by Functional Screening of the Extracellular Proteome

See allHide authors and affiliations

Science  09 May 2008:
Vol. 320, Issue 5877, pp. 807-811
DOI: 10.1126/science.1154370


To understand the system of secreted proteins and receptors involved in cell-cell signaling, we produced a comprehensive set of recombinant secreted proteins and the extracellular domains of transmembrane proteins, which constitute most of the protein components of the extracellular space. Each protein was tested in a suite of assays that measured metabolic, growth, or transcriptional responses in diverse cell types. The pattern of responses across assays was analyzed for the degree of functional selectivity of each protein. One of the highly selective proteins was a previously undescribed ligand, designated interleukin-34 (IL-34), which stimulates monocyte viability but does not affect responses in a wide spectrum of other assays. In a separate functional screen, we used a collection of extracellular domains of transmembrane proteins to discover the receptor for IL-34, which was a known cytokine receptor, colony-stimulating factor 1 (also called macrophage colony-stimulating factor) receptor. This systematic approach is thus useful for discovering new ligands and receptors and assessing the functional selectivity of extracellular regulatory proteins.

Complex organisms rely on an extensive system of secreted and cell-surface molecules for communication between cells, tissues, and organs. Many secreted proteins function as ligands for membrane-spanning receptors, which serve as the transducers by which extracellular signals influence intracellular physiology. Some ligands have highly specific functions because their receptors are expressed in a limited number of tissues, whereas other ligands act on a broad spectrum of tissues. There are ∼2200 human genes encoding secreted proteins, of which only a minority have functions that are well understood. We classified proteins as being secreted on the basis of one or more of the following algorithms: (i) the presence of a signal sequence; (ii) the presence of conserved domains of secreted proteins and extracellular regions of transmembrane proteins; (iii) presence of non-canonical secreted proteins and nonsecreted proteins found in human plasma (1); and (iv) a “secreted” classification by a supervised learning algorithm based on random decision trees (24).

We produced recombinant secreted proteins from ∼3400 cDNAs that encode secreted proteins and extracellular domains of transmembrane proteins (table S2). Each cDNA was individually transfected into 293T cells in a high-throughput format, and samples of conditioned medium containing each protein were collected for testing in cell-based assays within 12 hours. A parallel set of cDNAs that encode the proteins in tagged forms was similarly transfected into 293T cells and the amount of protein released into the medium was estimated by Western blot analysis (fig. S1A) and enzyme-linked immunosorbent assay (fig. S1B). More than 90% of the proteins were detected in the culture supernatant with a mean concentration of 75 ng/ml and a median of 20 ng/ml (fig. S1B).

To identify the functions and potential therapeutic utility of all human secreted proteins and their receptors, we chose a wide range of cells that represent different tissue types and various assay readouts (table S4) that measured cellular responses to secreted factors. The complete recombinant protein collection was screened in an automated high-throughput format across a wide spectrum of assays (table S4).

We focused on the subset of secreted proteins with high functional selectivity by comparing the activities of a given protein across multiple assays (table S4) and searched for those that were active in only a small number of assays. More than 201,000 data points generated from this set of assays were used for functional selectivity analysis of the secreted proteins (Fig. 1B). To identify the proteins with the highest functional selectivity, we defined a measure of activity that can be compared across multiple assays and readouts and devised a quantitative assessment of selectivity.

Fig. 1.

Identification of IL-34 using a functional screening approach. (A) Comparison of functional selectivity of IL-15, IGF-II, FGF3, and IL-34. Using P values for a Wilcoxon rank sum test, activities (–logP) for several proteins against their respective sorted assays are shown. The assays were sorted from the lowest to the highest P value, producing a protein-dependent list of sorted assays. Assay descriptions are shown in table S4. (B) Analysis of functional selectivity of secreted proteins. Functional selectivity of 1910 secreted proteins in the screening set was analyzed using the maximal differential activity s and selectivity index k.

The activity of each protein in eliciting a cellular response in a given assay was measured. Within an assay, the measurements of responses to n proteins were assigned ranks 1 to n from the lowest through highest numerical values of the assay measurements. Ranks for the responses to a protein from assays performed in replicate were summed, giving a Wilcoxon rank sum statistic. This statistic was used to test the null hypothesis that the ranks are random, versus the alternative hypothesis that the protein has some effect on the cells. The P value is the probability that the rank sum (as an activator or an inhibitor) is as strong as or stronger than observed, assuming there was only assay noise and that there was no effect of the protein. The quantity –log P, called activity in Fig. 1A, is then a positive number (a small value of P gives a large value of –logP) whose magnitude reflects the response of cells to a protein in a given assay.

For each protein, the assays were then sorted from the assay with the highest activity for that protein to the assay with the lowest activity. These relative activities across sorted assays were plotted for four examples, covering a range of selectivity (Fig. 1A). The order of the assays in Fig. 1A is different for each protein. For example, the values for interleukin-34 (IL-34) and IL-15 decrease sharply across assays compared with those for insulin-like growth factor II (IGF-II), indicating that they were active in one or two assays, whereas IGF-II was active in many; fibroblast growth factor 3 (FGF3) gave a low signal in many assays but did not maximally activate any assay (Fig. 1A).

The notion of selectivity of a protein in eliciting responses across many assays was quantified with a pair of numbers (s, k), where s is the maximal differential activity and k is the selectivity index. Suppose there are n assays, and P1P2 ≤... ≤ Pn are the corresponding sorted P values. The difference between adjacent activities (–logP1) and (–logP2) is equal to the log ratio Embedded Image Let r1,r2,...rn–1, denote the activity differences Embedded Image and define s = max{r1,r2,...,rn–1} and k = max{i:1 ≤in – 1 and ri = s}. Briefly, s is the difference and k is the index where the maximum occurs. Clones with high values of s and low values of k will be highly active in just a few assays. Figure 1B plots s for 1910 clones in n = 25 assays, showing some proteins—such as interferon α (IFN-α), FGF2, FGF3, platelet-derived growth factor D (PDGF-D), IGF-II, IL-1α, and IL-6—with low selectivity and others with high functional selectivity. Among the 100 clones with the highest selectivity (Fig. 1B), all but nine of these have k = 1 and thus are selective for one of the assays. Although this analysis is limited by the small number (25) of assays relative to all possible biological activities and by the fact that the assays were not representative of all tissues, the data provided a starting point for selecting proteins that might have a high degree of selectivity.

Because the secreted proteins used in the screening assays were expressed at various levels, we questioned whether the functional selectivity of proteins might correlate with protein concentration. Analysis of more than 700 proteins for their activity or maximal differential activity and expression levels demonstrated no correlation between activity or maximal differential activity of a protein and its expression level (fig. S2, A and B, and tables S7 and S8).

We focused on the proteins with a high selectivity score (Fig. 1B) for further analysis. The proteins that were highly selective included a number of known cytokines or growth factors, such as colony-stimulating factor 1 (CSF-1), selective for the monocyte viability assay; vascular endothelial growth factor C (VEGF-C), selective for the cardiomyocyte phosphoprotein assay; FGF19, selective for the H-4-II-E PCK1 expression assay; IL-15, selective for the T lymphocyte and natural killer (NK) cell viability assays; and FGF21, selective for the adipocyte assay (Fig. 1B and table S4). These proteins are known to be selective in their actions (510).

In the pool of highly selective proteins, we were especially interested in uncharacterized proteins. For further study, we chose a protein that had a high selectivity score and stimulated monocyte viability. This 241–amino acid protein was encoded by a cDNA clone, which differs by one amino acid residue from a hypothetical protein in the public database (GenBank accession number NP_689669). This protein increased monocyte viability (fig. S3 and table S5) and has been designated as IL-34, isoform 1 (GenBank accession number EU599219; referred to as IL-34 in this paper).

The amino acid sequence of human IL-34 shares a sequence identity of 99.6%, 72%, and 71% with that of the chimpanzee, rat, and mouse IL-34 orthologs, respectively (fig. S4), and the IL-34 gene is syntenic in the human, chimpanzee, rat, and mouse genomes. We detected no similarity of the IL-34 protein sequence to that of any other human protein, and IL-34 has no apparent consensus structural domain or motif. IL-34 mRNA is expressed in various tissues, including heart, brain, lung, liver, kidney, spleen, thymus, testes, ovary, small intestine, prostate, and colon, and it is most abundant in the spleen (fig. S5A). Consistent with its role in regulating myeloid cell growth and differentiation, IL-34 protein was also detected by immunohistochemistry in the sinusoidal endothelium in the red pulp of the spleen (fig. S5B).

We purified the human IL-34 protein (99% purity) as a homodimer consisting of 39 kD monomers. The purified protein increased the number of monocytes in an assay that uses adenosine 5′-triphosphate (ATP) content as a measure of cell number (Fig. 2A). This measurement, which we termed “viability,” does not distinguish between change in survival or proliferation. IL-34 activity also was detected in an assay measuring DNA replication (fig. S6), suggesting that an increase in cell proliferation indeed contributes to the increase in viable cells. In flow cytometry analysis, biotinylated IL-34 protein specifically bound to CD14+ monocytes in peripheral blood mononuclear cells (PBMCs), and this binding was decreased in the presence of unlabeled IL-34 (Fig. 2B). In contrast, IL-34 did not bind to CD3+ T cells or CD56+ NK cells (Fig. 2B).

Fig. 2.

Activities and cell-binding selectivity of purified IL-34 protein. (A) Dose-response curve of purified IL-34 protein in a monocyte viability assay. Recombinant IL-34 protein was purified from conditioned medium of 293T cells stably transfected with IL-34 cDNA. Relative viability (survival and/or proliferation) of primary human monocytes induced by the purified IL-34 protein was determined by the cell viability (CellTiter-Glo) assay (Promega). In this assay, the amount of ATP released from lysed cells was determined based on the luminescence signal [relative light unit (RLU)] generated from reaction with luciferin in the presence of luciferase. (B) Specific binding of IL-34 to human monocytes in PBMC. PBMC incubated with unlabeled IL-34 or bovine serum albumin (BSA) as competitors at 200-fold molar excess were incubated with biotinylated IL-34 protein and stained with streptavidin-allophycocyanin (APC). The cells also were labeled with fluorescein isothiocyanate (FITC)–conjugated antibodies against lineage markers, CD14 (monocytes), CD3 (T cells), and CD56 (NK cells) and analyzed by flow cytometry.

We searched for an IL-34 receptor in our collection of extracellular domains (ECDs) of membrane-spanning proteins. There are ∼1100 type I transmembrane proteins and ∼500 type II transmembrane proteins encoded by the human genome (table S2), as determined by literature sources and our prediction algorithms (4).

We screened our collection of ECDs for the ability to block the functional activity of IL-34 on monocytes through binding and sequestration of IL-34. Purified IL-34 was incubated with each ECD and then assayed for the effect on human monocyte viability (Fig. 3). Ten of the 858 ECDs screened inhibited IL-34 activity in the primary screen (Fig. 3 and table S6), but only one of these proteins gave reproducible inhibition in repeated testing (table S6). This protein was encoded by a cDNA for the CSF-1 receptor (CSF-1R) ECD, suggesting that CSF-1R is a receptor for IL-34. The inhibitory activity of CSF-1R ECD appeared to be specific because inhibition was not seen with other ECDs in the screening set or with granulocyte-macrophage CSF receptor (GM-CSFR) ECD or interferon-γ receptor (IFNγR) ECD (fig. S7). CSF-1R ECD blocked binding of biotinylated IL-34 to the human monocytic cell line THP-1, whereas GM-CSFR ECD or IFNγRECD hadnoeffect on IL-34 binding (Fig. 4A). Furthermore, unlabeled IL-34, CSF-1, or antibody to CSF-1R blocked binding of biotinylated IL-34 to THP-1 cells (Fig. 4A). The binding affinity of IL-34 to the immobilized CSF-1R ECD was determined by Biacore analysis with a dissociation constant (Kd) of about 1 pM (fig. S8A). CSF-1 bound to the CSF-1R ECD with a Kd of 34 pM (fig. S8B).

Fig. 3.

Identification of the receptor for IL-34 by screening of the ECDs of transmembrane proteins. IL-34 protein was individually incubated with conditioned medium of 293T cells transfected with one of the cDNAs encoding ECDs of transmembrane proteins. Human primary monocytes were then incubated with the treated IL-34 protein and analyzed by the cell viability assay. Results from a screen of the collection of 858 ECDs are shown as the standard deviation from the median (sigma from median). Purple circles (inhibitors) indicate clones for which activity was detected with greater than twofold standard deviation of the assay plate for more than one test. Cells without IL-34 treatment were used as controls (blue circles) on each assay plate. The activities of the ECD clones that inhibited IL-34 activity on monocytes in primary and secondary screens are shown in table S6. CSF-1R ECD V5-His Tag, CSF1R ECD with a V5-His epitope tag.

Fig. 4.

IL-34 functions as a specific and independent ligand of CSF-1 receptor, stimulates CSF-1R–dependent phosphorylation of ERK1/2, and promotes CFU-M colony formation. (A) IL-34 binding to CSF-1R on human monocytic THP-1 cells. Biotinylated IL-34 protein was incubated with conditioned media of 293T cells transfected with cDNA encoding CSF-1R ECD (i), GM-CSFR ECD (ii), or vector. THP-1 cells were then incubated with the treated IL-34 protein, labeled with streptavidin-APC, and analyzed by flow cytometry. THP-1 cells preincubated with BSA (iii), unlabeled IL-34 (iv), CSF-1 (v), neutralizing antibody to CSF-1R (Anti-CSF-1R Ab) and control mouse immunoglobulin G1 (mIgG1) (vi), or buffer control [phosphate-buffered saline (PBS)] were labeled with biotinylated IL-34, stained with streptavidin-APC, and analyzed by flow cytometry. (B) IL-34 function on monocytes is independent of CSF-1. Human monocytes were incubated with IL-34 or CSF-1 that was heat-inactivated (100°C for 5 min) or treated with rabbit polyclonal antibodies to IL-34 (rabbit anti-IL-34 Ab), a mouse antibody to CSF-1 [mouse anti-CSF-1 Ab (R&D Systems, no. MAB216], or control IgG [rabbit IgG (Millipore) and mouse IgG2a (R&D Systems, no. MAB003)] and assayed for viability by CellTiter-Glo assay (Promega). Relative viability is shown as luminescence (RLU). (C) Inhibition of IL-34 activity on ERK1/2 phosphorylation by a CSF-1R–specific inhibitor. Human primary monocytes were incubated with medium control, 5 nM IL-34, or 5 nM CSF-1 at 37°C for 3 min in the presence or absence of 60 nM CSF-1R–specific inhibitor, GW2580 (11). The monocyte cell lysate was analyzed by Western blotting using an antibody to phospho-ERK1/2 (p-ERK) (Santa Cruz Biotechnology, Inc., no. sc-7383). The Western blot was then stripped and probed with antibodies to ERK (Santa Cruz Biotechnology, Inc., no. sc-94) for detection of the total ERK. The results were reproducible in three experiments. (D) IL-34 promotes CFU-M colony formation from human bone marrow. Colony formation assays were performed by StemCell Technologies, Inc. as described (14, 15), using CSF-1 (R&D Systems, Inc.) or IL-34, at indicated concentrations. Representative data (n = 3 replicates for each concentration and treatment) from one of three experiments are shown. Images of the CFU-M colonies and of the May-Grunwald-Giemsa–stained cells are shown in fig. S9.

In functional studies, the activity of IL-34 on monocyte viability was blocked by antibodies to IL-34 but not by an antibody to CSF-1, whereas the activity of CSF-1 was blocked only by its own antibody and not by antibodies to IL-34 (Fig. 4B), suggesting that the IL-34 activity in monocyte viability was independent of CSF-1. The activity of IL-34 could be abolished when the IL-34 protein was heat-inactivated (Fig. 4B). IL-34, like CSF-1, stimulated phosphorylation of extracellular signal-regulated kinase-1 and -2 (ERK1/2) in human monocytes (Fig. 4C). Furthermore, the IL-34–induced phosphorylation of ERK1/2 was inhibited by a CSF-1R–specific inhibitor, GW2580 (11) (Fig. 4C). In addition, IL-34 promoted the formation of the colony-forming unit–macrophage (CFU-M), a macrophage progenitor, in human bone marrow cultures (Fig. 4D). The IL-34–promoted CFU-M colonies (fig. S9, A to C) had a size and morphology similar to those of colonies promoted by CSF-1 (fig. S9, D to F). Upon staining with the May-Grunwald-Giemsa stain, cells derived from these colonies showed vacuole-containing cytoplasm (fig. S9, G to J), confirming that they are cells of CFU-M (12).

The selectivity pattern of IL-34 across many assays was predictive of the biology of this protein. IL-34 is a new ligand for a highly selective and relatively well-studied receptor, the CSF-1 receptor. Our discovery of IL-34 illustrates that even in well-understood systems, there are functional protein regulators yet to be discovered using a systematic screening approach. CSF-1 and CSF-1R are important for monocyte formation and subsequent differentiation into the various functional phagocytes (6). Mice defective in CSF-1R have a more severe osteopetrosis in the femur and a more severe depletion of F4/80+ macrophages in the kidney than the CSF-1–deficient mice (13). Thus, the existence of a second ligand, in addition to CSF-1, for CSF-1R has been proposed (13). Here we report the identification of such a factor, IL-34.

Supporting Online Material

Materials and Methods

Figs. S1 to S9

Tables S1 to S8


References and Notes

Stay Connected to Science

Navigate This Article