Crystal Structure of the Labile Complex of IL-24 with the Extracellular Domains of IL-22R1 and IL-20R2 Free

Human IL-24 was originally described as a product of melanoma differentiation associated gene mda-7 (1). IL-24 displays a variety of physiological activities, among which is an ability to suppress proliferation of several cancer cell lines (2). Clinical trials utilizing IL-24 as a putative anticancer agent have been taking place for a number of years (3–5), although this cytokine is not yet approved as a drug.

The most widely recognized property of cytokines is their affinity for specific receptors capable of forming heterodimeric transmembrane complexes, the cytoplasmic domains of which are constitutively associated with JAK kinases. Binding of cytokines to the extracellular domains of heterodimeric receptors triggers a sequence of phosphorylation events constituting the initial steps of the JAK-STAT signaling pathway. Based on the signature sequences within extracellular domains, cytokine receptors are classified into two classes, 1 and 2.

IL-24 belongs to the family of cytokines interacting with class 2 cytokine receptors. This family additionally includes five other ILs (IL-10, IL-19, IL-20, IL-22, and IL-26), as well as IFNs from three different subfamilies, α/β, γ, and λ. The IL subset is often referred to as the IL-10 family (6). Three of these ILs, IL-19, IL-20, and IL-24, designated IL-20 subfamily cytokines (IL-20SFCs), signal through receptor complexes that share the common chain, IL-20R2. Additionally, IL-20 and IL-24 share with IL-22 another receptor chain, IL-22R1, although the latter cytokine utilizes IL-10R2 as its second receptor.

The open reading frame of IL-24 encodes a polypeptide chain composed of 206 aa residues; however, excision of the signal peptide leads to the mature form consisting of 155 residues. Interestingly, the signal peptide in IL-24 (51 aa) is twice as long as in other related human cytokines. Whereas the predicted molecular mass of IL-24 monomer is 18.3 kDa, well-established N-glycosylation of up to three sites in mammalian cells (7) leads to expression of several isoforms, with molecular mass ranging from 18.3 kDa up to 35 kDa (8). It was previously suggested that N-glycosylation of IL-24 is dispensable for induction of tumor cell apoptosis and related activities (9). Other data indicated that partial N-linked glycosylation and one disulfide bond are required for solubility and bioavailability of IL-24 and attempts to produce biologically active IL-24 in bacteria were unsuccessful (7). Although production of active IL-24 and its mutant in Escherichia coli has been reported more recently (10, 11), there is no evidence that the amounts of stable cytokine sufficient for structural studies have ever been obtained in bacteria.

IL-24 forms signaling complexes with two different receptor pairs, IL-20R1/IL-20R2 and IL-22R1/IL-20R2, which it shares with the other IL-20SFCs and IL-22. These complexes were defined as type I (with the IL-20R1/IL-20R2 pair of receptors) and type II (IL-22R1/IL-20R2) (12, 13). For both types of complexes, the signaling events are accomplished by the activation of the Tyk2/Jak1→STAT1/STAT3 pathway. Several structures of these cytokines, either alone or bound to their receptors, are available. Crystal structures of free IL-19 (14) and IL-22 (15, 16) have been previously determined. Crystal structures are also available for the binary complexes of IL-22 with its high-affinity receptor (RH) chain IL-22R1 (17, 18). Additionally, the structure of the ternary complex IL-20/IL-20R1/IL-20R2 has also been determined (13). Thus the structures of all receptors used by IL-24 are already known (but in complexes with other family members). No crystal structure of IL-24, either free or in the form of a ligand/receptor complex, has been published to date, although a model based on the close homology with IL-19, IL-20, and IL-22 has been proposed (19). In this article, we present several approaches to expressing IL-24 and its receptors and evaluating the biological activity of the cytokine. Furthermore, we describe the crystal structure of IL-24 in the type II ternary complex (IL-24/IL-22R1/IL-20R2) and compare it with the structures of closely related cytokines.

A synthetic gene, codon-optimized for E. coli expression and encoding the mature wild type IL-24 (IL-24_wt; aa 52–206) and a cleavable N-terminal affinity tag, “HHHHHHENLYFQG” (the underlined fragment corresponds to the TEV protease recognition sequence), was purchased from GeneScript. The gene was cloned into the Gateway System (Thermo Fisher Scientific). The expression vector pDEST 14 (Thermo Fisher Scientific) was used for transforming BL21(DE3) pLysS cells (Agilent Technologies), and expression was attempted under various regimens of temperature, time, and induction point. The expression experiments invariably resulted in production of an insoluble product, identified as IL-24 by Western analysis using both Penta·His (Qiagen) and anti–IL-24 (R&D Systems) Abs. Subsequently, the synthetic gene of IL-24_wt was cloned in frame into three additional vectors, pRSET A (Thermo Fisher Scientific), pET32a(+) (EMD Millipore), and pMAL C5X (New England Biolabs [NEB]), resulting in the following constructs: MRGS–(His)₆–“Xpress-epitope-tag”–“enterokinase-cleavage site”–IL-24_wt–STP, Thioredoxin–(His)₆–“TEV cleavage site”–IL-24_wt–STP, and “Maltose binding protein”–(His)₆–“TEV cleavage site”–IL-24_wt–STP. These constructs were used for transformation of the BL21(DE3) pLysS cells. The latter two constructs were aimed at production of IL-24_wt N-terminally fused to cleavable protein partners, with an expectation of improved solubility and/or folding supported by disulfide formation, compared with expression of the unfused cytokine. The correctness of these constructs was confirmed by sequencing (Macrogen). For each of the three constructs, expression experiments were conducted at two temperatures, 37 and 18°C. In the case of the construct based on pRSET A, IL-24_wt was expressed in an insoluble form with a lower yield compared with that obtained from the Gateway vector. Both fusion proteins were also accumulating mostly in the inclusion bodies (IBs), with only a small fraction remaining in the soluble form. However, following cleavage of fusion partners with TEV protease, the released IL-24_wt rapidly precipitated.

Because refolding became an unavoidable step for a preparation of soluble IL-24_wt, a new expression vector encoding Met–IL-24_wt [i.e., lacking the (His)₆–ENLYFQ sequence] was generated with a QuikChange protocol using the original Gateway-based construct and used for transformation of BL21(DE3) pLysS cells. Cells were grown at 37°C in the lysogeny broth medium supplemented with ampicillin (100 μg/ml). After cell density reached a midlog phase (OD₅₉₅ 0.5 / 0.7), expression was induced with the isopropyl-β-d-1-thiogalactopyranoside (IPTG) supplied at the concentration of 1 mM. Growth of the cell culture continued at 37°C for an additional 3–4 h, reaching OD₅₉₅ 3.5–4.5. The cell pellet was collected by a centrifugation at 6000 × g for 20 min and was subsequently suspended in 150 ml of the lysis buffer (50 mM Tris, pH 8, 150 mM NaCl) supplemented with 10 mM MgCl₂, 10 mM CaCl₂, 20 μg/ml DNase I (Roche), 20 μg/ml RNase A (Thermo Fisher Scientific), and the complete Protease Inhibitor Cocktail (Roche) and frozen at −80°C for further use.

The suspension of E. coli cells expressing insoluble IL-24_wt was thawed on ice and subjected to cell lysis using the APV-1000 homogenizer (Invensys APV Products, Albertslund, Denmark) operating at 70 MPa. Insoluble fractions (IBs) were collected by centrifugation at 35,000 × g for 30 min. IBs were washed twice with 100 ml of 50 mM Tris (pH 8) with 20 mM 2-ME, 1 mM EDTA, and 2% Triton X-100, followed by two washes with the same buffer but without Triton X-100. Subsequently, IBs were solubilized in 120 ml of the denaturing buffer (6 M guanidinium chloride, 50 mM Tris, pH 8, 20 mM DTT, 1 mM EDTA). Insoluble residue was removed by centrifugation (35,000 × g for 30 min). The cleared solution was gradually diluted with 40 ml of buffer A consisting of 10% (v/v) solution of acetonitrile, supplemented with 0.1% trifluoroacetic acid—further dilution led to precipitation of the protein. The diluted solution was applied onto the 100 × 7.5 mm PEEK column, custom-packed with the POROS R2 20 μm reverse-phase resin (Thermo Fisher Scientific), and washed extensively with buffer A. Protein was eluted at a flow rate 10 ml/min with a linear gradient (10–35% v/v) of acetonitrile. Fractions containing IL-24 (identified by SDS-PAGE, NaDodSO₄–PAGE) were pooled and lyophilized. The molecular mass, measured by electrospray ionization mass spectroscopy, was 18,268.3 Da, which is in good agreement with the theoretical value of 18,268.1 Da.

About 50 mg of lyophilized sample was dissolved in 100 ml of a buffer containing 6 M guanidinium chloride, 0.1 M Tris (pH 8), 0.5 M arginine hydrochloride, 5% (v/v) glycerol, and 1 mM EDTA, as well as 5 mM reduced and 0.5 mM oxidized glutathione. In addition to gentle rocking of the sample vial, solubilization was enhanced by several short bursts of sonication on ice. The resulting sample was diluted 20-fold with a solution containing 0.1 M Tris (pH 8), 0.5 M arginine hydrochloride, 0.5 M sodium chloride, 5% (v/v) glycerol, 1 mM EDTA, 5 mM reduced, and 0.5 mM oxidized glutathione and stirred for 48 h at 4°C. Subsequently, the solution containing soluble IL-24 was filtered through a 0.22 μm polyethersulfone membrane and concentrated using the Amicon YM10 membrane fixed in a compatible stirred cell (EMD Millipore), to the final volume of 100 ml. That solution was dialyzed overnight using a 45 mm/MWCO 3500 membrane (Spectrum Labs.) against 3.5 l of buffer B containing 0.1 M Tris (pH 8), 0.5 M sodium chloride, and 5% (v/v) glycerol. Dialysis resulted in substantial precipitation of misfolded IL-24_wt, which was removed by centrifugation. A cleared solution (∼110 ml) of refolded IL-24 was concentrated 20-fold using a stirred cell with an Amicon YM10 membrane and applied onto a HiPrep 16/60 Superdex 200 HR column (GE Healthcare Life Sciences) equilibrated with buffer B. Fractions containing the recombinant protein were identified using SDS-PAGE, pooled, and concentrated using the Amicon YM10 membrane. The final sample of IL-24 was judged to be >95% pure by SDS-PAGE. The molecular mass was 18,266.6 Da, as compared with the theoretical 18,266.2 Da.

DNA sequences encoding proteins of interest were amplified from the cDNA samples purchased from OriGene (www.origene.com). They included IL-24 (cat. no. RC216502), IL-20R1 (cat. no. RC238599), IL-20R2 (cat. no. RC213197), and IL-22R1 (cat. no. RC206447). Plasmids used for the expression of proteins in Drosophila S2 Schneider cells were generated with an aid of the commercial pMT/BiP/V5-His A vector (Thermo Fisher Scientific), utilizing the BglII and AgeI sites. The DNA fragments (inserts) encoding extracellular regions of the mature IL-20R1 (aa 30–250, Q9UHF4/UniProt; NM_001278722.1/GenBank, residue numbers correspond to native translated genes that include signal sequences), IL-20R2 (aa 30–233, Q6UXL0/UniProt; NM_144717.3/GenBank), IL-22R1 (aa 15–228, Q8N6P7/UniProt; NM_021258.3/GenBank), and the sequence corresponding to mature IL-24 (aa 52–206, Q13007/UniProt; NM_001185156.1/GenBank) were PCR amplified (using PfuUltra II Hotstart PCR Master Mix; Agilent Technologies) in the presence of appropriately designed synthetic primers (Integrated DNA Technologies) and gel purified after sequential digest with BglII and AgeI (NEB). The native pMT/BiP/V5-His A vector was subjected to a similar protocol resulting in BglII-vector-AgeI. After ligations (using Quick Ligation kit; NEB) of inserts with BglII-vector-AgeI, plasmids were transformed in E. coli DH5α and plated on lysogeny broth substrate with ampicillin (100 μg/ml), and individual clones were selected for preparation and sequencing (Macrogen) of plasmids used for expression of soluble IL-20R1 (sIL-20R1), sIL-20R2, sIL-22R1, and IL-24.

All these wild type variants carry multiple potential N-glycosylation sites, the presence of which could be detrimental for crystallization. Therefore, several constructs encoding N-deglycosylated variants of the abovementioned proteins were also generated by mutating appropriate codons (Asn→Gln). This task was accomplished with the aid of the QuikChange strategy (see www.genomics.agilent.com/files/Manual/200523.pdf; Agilent Technologies). The resulting plasmids were IL-24_Q85Q99Q126 (asparagines 85, 99, and 126 mutated to glutamines), IL-20R2_Q40Q134, and IL-20R1_K111RK113R. IL-24_Q85Q99Q126 represents the fully N-deglycosylated variant, whereas mutations introduced to the genes encoding soluble fragments of the IL-20Rs followed the earlier suggestions (13). In addition, plasmids encoding the sequences of three fusion proteins were also constructed, based on the native pMT/BiP/V5-His A vector. They included sIL-20RA–linker–IL-24_Q85Q99Q126, sIL-20RB–linker–IL-24_Q85Q99Q126, and sIL-22RA–linker–IL-24_Q85Q99Q126, where the linker sequence GGGGSETVRFQSGGGSEGGGSE included the recognition sequence ETVRFQS of the TVMV protease. To generate each of the fusion inserts flanked by the BglII and AgeI restriction sites (see above), fragments for each fusion partner were first PCR amplified with pairs of synthetic DNA primers (Integrated DNA Technologies). Whereas the upstream primers in each of the two fragments were the same as those used in earlier cloning experiments and encoded appropriate restriction sites, the downstream primers were largely complementary and, when assembled, encoded the linker sequence. Pairs of PCR-amplified fragments were assembled by PCR reaction into complete fusion inserts and treated subsequently as described above. All plasmids aimed at the protein expression in Drosophila S2 Schneider cells encoded open reading frames following a common structure, BiP-RS-insert-TG-His6-STP, in which BiP encodes signal sequence, RS and TG are cloning artifacts (representing BglII and AgeI restriction sites, respectively), insert corresponds to the protein of interest, His6 encodes the His₆ sequence of the affinity tag, and STP is the termination signal. In addition, the TEV protease recognition sequence ENLYFQG was embedded at the 3′ terminus of the insert in constructs encoding fusion proteins.

Plasmids encoding the components of the IL-24 signaling complex were used for transfection of Drosophila S2 Schneider cells (cat. no. R690-07; Thermo Fisher Scientific) to establish stably transfected cell lines. Briefly, a culture (∼5 ml) of high-viability S2 cells in the SFX Insect media (Thomas Scientific) supplemented with 10% (v/v) heat-inactivated FBS (Thomas Scientific) was used to inoculate several small cultures (in six-well plate, ∼2.4 ml/well per transection) in fresh complete medium (CM) (SFX Insect with FBS) to the cells density ∼10⁶ cells/ml. Cells were left to adhere for ∼4–6 h at 28°C. The transfection protocol used the Effectene Transfection Reagent (cat. no. 301425; Qiagen) and followed manufacturer recommendations. Cells were always cotransfected with a mixture of an expression vector and pCoBlast (Thermo Fisher Scientific) at an approximate molar ratio 1:19, with the latter vector providing blasticidin resistance. When the S2 cells were cotransfected with several expression vectors, different concentration ratios of these vectors were tested, whereas the concentration of pCoBlast was usually 19 times higher than the combined concentration of expression vectors. After 2–4 d, when the cell viability was ascertained (usually by visual inspection), most of the medium was replaced with fresh CM supplemented with blasticidin S (cat. no. A1113902; Thermo Fisher Scientific) at the concentration 10–20 μg/ml. After this step, every 2–4 d, cells were pelleted by centrifugation (100–500 × g, 5–8 min.) and resuspended in fresh CM supplemented with gradually increased concentrations of blasticidin S (maximum concentration 25–30 μg/ml). After several passages, the volumes of cell cultures were expanded. The selection process was usually complete after 4 wk. Subsequently, small aliquots of stably transfected cells were frozen following a standard protocol (see the Drosophila Expression System manual, https://assets.thermofisher.com/TFS-Assets/LSG/manuals/des_man.pdf), whereas part of each cell culture was subjected to expression trials (appropriate protocols are described in the Drosophila Expression System manual). At this stage, production and identities of proteins were confirmed by the SDS-PAGE and Western analysis using specific Abs.

Expression utilizing the pMT/BiP/His A vector was induced with Cu²⁺ ions, and proteins were secreted to the media. For large-scale expression, transfected cells were cultured in T-75 flasks in CM supplemented with blasticidin S until the culture volume reached 20–30 ml and density reached ∼5 × 10⁶ cells/ml. At this stage, cells were harvested by centrifugation (100–500 × g, 5–8 min), resuspended in fresh serum-free SFX Insect medium to assure the density >2 × 10⁶ cells/ml, and transferred to an appropriate shaker flask (assuring good aeration). Flasks were placed in the incubated shaker and agitated at 105–125 rpm. Cell density and viability were monitored daily, and gradually, the cultures were expanded to the final volumes of 2.l. Approximately 12 h prior to induction, cells were harvested and resuspended in fresh SFX Insect medium. Protein expressions were induced with 100 mM copper sulfate solution to the final concentration of Cu²⁺ ions between 500 and 750 μM. Expression usually continued for additional 4–5 d, with cell growth and viability monitored daily. A medium with the expressed protein product was harvested by centrifugation (6000 × g for 30 min), cleared using 0.22 μl filters, and immediately subjected to a purification procedure.

A typical purification protocol consisted of three steps. First, protein-containing media were applied onto a 5 ml HisTrap Excel column (cat. no. 17371206; GE Healthcare Life Sciences). Extensive wash with buffer A containing 500 mM NaCl and 20 mM sodium phosphate (pH 7.4) was followed by wash with buffer A supplemented with 30 mM imidazole. Protein was eluted with buffer A supplemented with 500 mM imidazole. At this stage, a majority of the protein was isolated, and the volume of a protein sample was usually between 20 and 50 ml; however, the purity of the preparation assessed by SDS-PAGE was inferior compared with typical Ni-affinity purification. Therefore, in the next step, protein solution was either diluted 7–10 times with buffer A or dialyzed against this buffer overnight (to lower the concentration of imidazole) and applied onto a second Ni-affinity column, 5 ml HisTrap HP (cat. no. 17-5247-01; GE Healthcare Life Sciences). The purification protocol mirrored very closely the one described earlier except that elution was usually accomplished with 250–300 mM imidazole. After concentrating the fractions containing a target protein, the sample was applied onto the size exclusion column Superdex 75 Increase (cat. no. 29148721; GE Healthcare Life Sciences) and equilibrated/eluted in the final buffer containing 50 mM Tris (pH 7.5) and 200 mM NaCl. After assuring purity, protein preparations were concentrated, frozen, and stored at −80°C for further use.

All measurements of molecular mass referred to in this manuscript were performed by means of the MALDI-TOF mass spectroscopy using Agilent 6100 series single quadrupole LC/Mass Selective Detector (Agilent Technologies).

SDS-PAGE analyses were performed using Novex NuPAGE SDS-PAGE gel system on NuPAGE Bis-Tris precast gels (Thermo Fisher Scientific). Proteins were transferred from gels to nitrocellulose membranes with the iBlot Dry Blotting System (Thermo Fisher Scientific). The following Abs were used during analyses: anti–IL-24, cat. no. AF1965; anti–IL-20R1, cat. no. AF1176; anti–IL-20R2, cat. no. AF1788; and anti–IL-22R1, cat. no. AF2770 (all from R&D Systems) as well as anti–Penta-His-tag, cat. no. 34660 (Qiagen).

Hamster cells expressing chimeric human IL-20R1/γR1 and intact human IL-20R2 were collected by trypsinization, aliquoted into tubes (10⁵ cells/200 μl), and treated with various concentrations of IL-24 samples for 15 min and used in EMSAs to detect STAT1 activation with the γ-activated sequence probe as described (20, 21). EMSAs were performed with a 22 bp sequence containing a Stat1α binding site corresponding to the γ-activated sequence element in the promoter region of the human IRF-1 gene (5′-GATCGATTTCCCCGAAATCATG-3′) (22). Two oligonucleotides, 5′-GATCGATTTCCCCGAAAT-3′ and 5′-CATGATTTCGGGGAAATC-3′, were annealed by incubation for 10 min at 65°C, 10 min at 37°C, and 10 min at 22°C and labeled with [³²P]dATP by filling-in with the Klenow fragment of DNA polymerase I in the presence of the other three dNTPs. Whole-cell extracts were prepared as followed. Untreated or IL-24–treated cells were precipitated by centrifugation at 3000 × g for 1 min, washed with 1.0 ml of cold PBS, pelleted again, and resuspended in 50 μl of lysis buffer (10% glycerol, 50 mM Tris–HCl, pH 8, 0.5% Nonidet P-40, 150 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 0.2 mM PMSF, 1 mM Na₃VO₄, 3 μg/ml aprotinin, 1 μg/ml pepstatin, and 1 μg/ml leupeptin). After 30 min on ice, the extracts were centrifuged for 5 min at full speed in a microcentrifuge, and the supernatant was recovered and used directly for EMSAs or stored at −80°C until use. EMSA reactions contained 2.5 μl nuclear extract, 1 ng ³²P-labeled probe (sp. act. 10⁹ cpm/μg), BSA (24 μg/ml), poly(dI:dC) (160 μg/ml), 20 mM HEPES (pH 7.9), 1 mM MgCl₂, 4.0% Ficoll (Sigma-Aldrich), 40 mM KCl, 0.1 mM EGTA, and 0.5 mM DTT in a total volume of 12.5 μl. Incubations were performed at 22°C for 20 min, then 4 μl of the reaction mixture were electrophoresed at 400 V for 3–4 h at 4°C on a 20 × 20 cm² 5% polyacrylamide (19:1 acrylamide/bisacrylamide) gel. The dried gel was exposed to Kodak XAR-5 film with intensifying screens for 12 h at −80°C.

Crystals of the ternary complex were grown by vapor diffusion in hanging droplets formed by mixing equal volumes of the protein solution (8 mg/ml in 200 mM NaCl and 50 mM HEPES, pH 7) and the reservoir solution (21% w/v MePEG2000, 50 mM Tris, pH 8.5, 50 mM trimethylamine N-oxide dihydrate). Crystals grew within 3–5 d. For data collection, crystals were cryoprotected by brief transfer to the reservoir solution enriched with 10% (v/v) of glycerol and frozen in liquid nitrogen. X-ray data were collected from one crystal on beamline 22-ID at the Advanced Photon Source, Argonne National Laboratory (Argonne, IL). All diffraction experiments were conducted at −173°C. The experimental images were processed and scaled using the program HKL3000 (23). Details of experimental data collection and processing statistics are presented in Table I. The structure was solved by molecular replacement with the program PHASER (24) using the coordinates of the IL-20/IL-20R1/IL-20R2 ternary complex [Protein Data Bank (PDB) identifier (ID): 4doh (13)] as the search model. Data extending to 2.5 Å were used for molecular replacement, yielding Z-score of 42.5 and the log-likelihood gain of 10,796. Subsequently, using complete data set within resolution ranges 30–2.15 Å, the solution was rebuilt, including conversion to the correct amino acid sequence, with the aid of the programs PHENIX (25) and MR Rosetta (26), with final refinement using Refmac5 (27). In addition to the sequence conversion, a total of 506 aa residues and 172 pseudo-water atoms were modeled in the electron density. At this stage, the values of the R_work and R_free were 0.265 and 0.325, respectively, and the stereochemistry of the model was mostly correct. The final steps of model improvement included a series of manual corrections, refinement, and monitoring aided by the programs Coot (28), Refmac5 (27), and Molprobity (29). The final model is characterized by R_work and R_free values 0.183 and 0.225, respectively, with acceptable stereochemistry. Detailed characteristics of the final model are shown in Table II.

The extent and the energy contribution of the interfaces between IL-24 and the receptors were evaluated with the web servers PDBePISA (http://www.ebi.ac.uk/pdbe/pisa/) and CCharPPI (https://life.bsc.es/pid/ccharppi). Whereas PDBePISA directly calculates the buried surfaces and estimates the binding energies due to interatomic interactions, CCharPPI provides access to the scores of 108 quantitative descriptors of protein–protein interactions developed by different research groups. The descriptors range from particular contributions to binding energy (e.g., van der Waals, electrostatics, hydrogen bonds, solvation energies) to the net energy of the whole interface, based on either on empirical functions, geometry of the contacts, or various statistical estimators. To estimate the free energy of binding based on descriptor scores, we have used as a reference the Affinity benchmark dataset with the same scores precalculated for 144 proteins with experimentally measured binding energy (30). For every available descriptor, we have calculated a linear fit (least squares method) of the experimental binding energy versus the descriptor score. Out of 108 descriptors available on the CCharPPI server, we were able to match and fit 106 using the database. Although for any of the individual descriptors, the Pearson correlation coefficient with the energy was estimated to be below 40%, the large number of estimates and the agreement among the predicted energy values for different approaches bring reasonable confidence in the results.

The coordinates and structure factors have been deposited in the PDB (http://www.rcsb.org, accession code 6df3).

Because some IL-10–related cytokines have been previously prepared using E. coli bacterial expression systems, confirming the feasibility of such an approach (15, 31), we explored whether a similar strategy can be used for IL-24. Deglycosylated cytokines were reported to retain their native structures and also, whenever tested, the ability to bind cognate receptors (18). Potential advantages of bacterial expression include elimination of glycosylation that may interfere with crystallization, short expression time, and low cost. Several earlier reports addressed the problem of expressing IL-24 in bacteria (10, 11), as well as the influence of posttranslational modifications on the activity of this cytokine (7, 9). In most of these reports, however, the evidence of successful production of IL-24 in E. coli was limited to characterization by SDS-PAGE and mass spectroscopy (10, 11), thus preventing reliable assessment of the practicality of such approaches. Fuson et al. (7) reported E. coli expression of IL-24, but only in an insoluble form. The protein could be refolded in the presence of decyl-maltoside yet precipitated quantitatively after removal of the detergent. These authors also found that the yield of expression of soluble GST–IL-24 fusion was very low, and IL-24 quantitatively precipitated after the cleavage of the GST-tag. They concluded that the solubility/stability, secretion levels, and (presumably) receptor-mediated ability to kill melanoma cells by IL-24 depend on proper N-glycosylation and on the presence of an intact single disulfide bond (7).

The results of our expression trials of IL-24 in E. coli largely mirror the earlier observations (7), particularly regarding the solubility of this cytokine, as well as the outcome of experiments involving fusion variants and refolding in the presence of detergents (data not shown). When it became clear that a refolding step is inevitable, we focused on preparation of samples devoid of any nonnative additions (i.e., affinity tags, cleavage sites, etc.) which could unfavorably affect the structure of refolded IL-24. Insoluble recombinant IL-24 could be expressed at moderate-to-high levels (20–35 mg/l), as illustrated in Fig. 1A. It should be noted that recombinant IL-24 migrates somewhat anomalously on polyacrylamide gel (SDS-PAGE), where its molecular mass appears 3–4 kDa smaller than expected. Subsequent Western blotting analysis confirmed the identity of insoluble recombinant IL-24 (Fig. 1B). IBs were extracted and washed as described in 2Materials and Methods. The wash step was followed by thorough purification using reverse-phase HPLC, resulting in highly pure, desalted polypeptide with the sequence corresponding to native mature IL-24 (aa 52–206), with an extra N-terminal methionine (an artifact of cloning and expression in bacterial cells).

Of all the tested refolding conditions, the most promising results were obtained when the initial sample of IL-24 was solubilized in 6 M guanidinium chloride, 50 mM Tris buffer (pH 7.5), at the concentration of ∼1 mg/ml. The resulting solution was then diluted 50-fold by a dropwise addition to the solution containing Tris buffer, NaCl, a glutathione shuttle, Arg-HCl, and glycerol (see 2Materials and Methods for details). Dilution was associated with only minor turbidity which did not increase significantly over 24–48 h of stirring at 5°C. The solubility of such refolded IL-24 was, however, highly correlated with the presence of Arg-HCl, an antiaggregation agent, present at a concentration of 0.5 M. Removal of Arg-HCl by dialysis led to significant precipitation of IL-24, although small amounts of the cytokine remained in solution. The IL-24–containing solution was concentrated using a stirred pressure cell and subjected to size-exclusion chromatography purification. The size-exclusion chromatography elution profile indicated partial aggregation of IL-24 on the column. Fractions corresponding to nonaggregated cytokine identified by SDS-PAGE were combined and concentrated to ∼3 mg/ml using a stirred pressure cell equipped with 3 kDa membrane. The protein precipitated upon any further concentration. Subsequent characterization by SDS-PAGE and mass spectrometry (see 2Materials and Methods) suggested formation of an unstable single disulfide bond in a molecule of refolded IL-24 (see Fig. 1C). The PAGE analysis of refolded IL-24, conducted under native conditions, revealed the presence of several discrete oligomeric forms (Fig. 1D). This observation, together with the results of the SDS-PAGE analysis, suggests that unstable monomers could coexist with oligomeric forms of refolded IL-24, which are stabilized by spuriously formed intermolecular disulfide bonds.

Whereas this procedure appeared to be more successful than the one previously described by Fuson and coworkers (7), as well as better than our own alternative refolding protocols, the resulting protein solution became turbid within minutes, and slowly but continuously, the bulk of IL-24 precipitated. This disappointing observation led to several conclusions. Similarly to the previous studies of refolded, deglycosylated IL-24, the cytokine aggregates in even modestly concentrated solutions, suggesting a potential role of N-glycosylation in its stability and solubility. The lack of a network of disulfide bonds, typically found in related cytokines, may also be an important contributor to low stability. In contrast to other IL-20SFCs stabilized by three disulfide bonds each, or IL-22 by two, a molecule of IL-24 contains only two cysteines and can potentially form only a single disulfide bond (Fig. 2A). However, after incubating the solution of IL-24 at 5°C for up to 5 d, the amount of the cytokine remained in solution was sufficient for completion of the JAK-STAT signaling assay. For this purpose, we used an approach developed earlier by one of us (S.V. Kotenko), briefly outlined below.

Cytokines demonstrate various degrees of species specificity. For instance, hamster cells are not responsive to human cytokines; therefore, appropriate human receptors must be expressed in these cells to render them responsive. Such modified cells allow for easier monitoring of signaling by cytokines, which is a weak event in intact cells owing to the low level of receptor expression. In the case of IL-10–related cytokines, one receptor subunit in a receptor complex determines the specificity of signaling and subsequent biological activities (32). By substituting the intracellular domain of the signaling receptor subunit with the corresponding region of IFN-γR1, it is possible to create new, chimeric receptors which retain their ligand-binding specificities yet gain the ability to trigger, in response to cytokine treatment, IFN-γ–specific biological activities that can be uniformly measured. Following this approach, reporter hamster cell lines expressing receptor complexes for IFNs, IL-10, IL-22, and IL-26 were previously generated (20, 33, 34). Using the same strategy, we created hamster cells expressing chimeric human IL-20R1/γR1 and an intact human IL-20R2, which specifically respond to IL-20SFCs. This reporter cell line was used to evaluate biological potency of recombinant IL-24. The assay was conducted in parallel for the refolded IL-24 described in this article and for a commercial sample of this cytokine (cat. no. 1965-IL-025; R&D Systems), the latter expressed in mouse myeloma cells. As can be seen in Fig. 1E, the refolded, deglycosylated IL-24 signals quite comparably to its fully glycosylated variant, indicating that in contrast to an earlier report (7), N-linked carbohydrates and other potential posttranslational modifications are not critical for the signaling properties of this cytokine. In conclusion, although IL-24 expressed in bacteria binds to and activates the IL-20R1/IL-20R2 receptor complex, in its deglycosylated form this cytokine is quite unstable and prone to aggregation and is thus unsuitable for crystallographic studies.

All cytokines from the IL-10 family signal by forming complexes with heterodimeric receptors. Usually, the affinity of a cytokine to one of the receptor chains is significantly lower than to the other one, at least for the soluble receptor domains (35). For example, the common IL-10R2 receptor chain, recognized by IL-10, IL-22, IL-26, and IFNs-λ, has very low affinity to cytokine ligands and is capable of binding to them only in the presence of the specific RH chain. The soluble ternary complexes of these cytokines are quite unstable, leading to difficulties in their isolation, crystallization, and structural characterization. Recently, Mendoza et al. (36) described a path to circumvent such a problem in the case of a receptor complex of IFN-λ3. Their approach involved extensive mutagenesis of the cytokine and was labor-intensive and time-consuming. Additionally, the resulting complex contained mutations in the ligand–receptor interfaces that require in silico extrapolation to reconstruct the interactions present in the native complex. In this report, we present a simpler approach, in which a cytokine is tethered with the aid of a long and flexible polypeptide linker to the low-affinity receptor (RL).

Employing receptor–ligand fusions in structural studies is not a novel idea, and, for example, we previously successfully used a similar approach while studying the structure of a complex between the phage protein g3p and its bacterial coreceptor, TolA (37). In complexes of cytokines with heterodimeric receptors (RH and RL), the cytokine–receptor fusion can be generally constructed in four ways, namely cytokine-RL, RL-cytokine, cytokine-RH, and RH-cytokine. In principle, stable association of a cytokine with RH does not require covalent tethering, as shown by the available structures of several cytokine-RH binary complexes. In the case of IL-24, however, an additional complication arises from the paucity of data on its affinity for the individual receptor chains. As mentioned earlier, the failure to prepare sufficient amounts of stable IL-24 prevented us from quantitatively assessing affinities to individual receptor chains; thus, we could not be certain which receptor represents RL or RH. Consequently, for each of the two complexes, type I and type II, at least two fusions needed to be prepared (four, if the order of components in a fusion is considered) (i.e., IL-20R2–IL-24 and IL-20R1–IL-24 in the case of type I complex). Furthermore, it is worth noting that covalent tethering of a cytokine to its RL does not assure formation of a cytokine-RL binary complex, and only after addition of an RH chain might the advantages of the cytokine-RL fusion be realized. A long and unstructured linker, tethering noninteracting cytokine and RL molecules, may be subject to proteolytic degradation, resulting in a significantly reduced yield. The latter possibility, however, may be circumvented by the coexpression of the cytokine-RL fusion with an appropriate RH. All these scenarios were considered in our experiments conducted for the IL-24 system. A schematic diagram of the fusion construct successfully used for structural studies is shown in Fig. 2A.

When IL-24 was expressed in the absence of a receptor or fused to IL-20R2 (in both cytokine–receptor and receptor–cytokine configurations), the yields of expression in insect cells were very low (albeit detectable by Western analysis). Some increase in the yield was observed for expression of IL-20R1–IL-24 or IL-22R1–IL-24 fusions; however, when the latter fusion was coexpressed with IL-20R2, the yield increased dramatically (see Fig. 2B), reaching tens of milligrams per liter of Drosophila cell culture. The ternary complex formed during this experiment could be readily purified and its identity was confirmed by SDS-PAGE and Western analysis.

Crystal structure of the complex of IL-24 with two receptors, IL-22R1 and IL-20R2, was determined by molecular replacement using data extending to the resolution of 2.15 Å. Diffraction data were collected from one tetragonal crystal in space group P4₃, with a single complex in the asymmetric unit (Table I). The structure was solved by molecular replacement with the program PHASER (24) with the coordinates of the IL-20/IL-20R1/IL-20R2 ternary complex [PDB ID: 4doh (13)] used as the search model. The initial solution was optimized with MR Rosetta (26), and the structure was rebuilt with Coot (28) and refined with Refmac5 (27) to R_work of 0.183 and R_free 0.225 (Table II). Unlike what was reported in the case of the complex of IL-20/IL-20R1/IL-20R2, for which a dimer was present in the asymmetric unit (13), a limited number of crystal contacts seen in the structure of the IL-24 complex does not suggest any oligomerization. Even though molecules of IL-24 and IL-22R1 were fused into a single chain, the linker residues were completely disordered; thus, in the following description these two proteins are numbered as two independent chains. Whereas a majority of quality indicators of this structure are in an acceptable range, three indicators (i.e., the individual B-factors, Z-score of the real space R value), and the overall small number of water molecules may appear unsatisfactory when compared with structures of many globular proteins determined at a similar resolution. However, this phenomenon is common for ternary complexes of cytokines and was similarly manifested, for example, in the structures of the IL-20–IL-20R1–IL-20R2 complex [PDB ID: 4doh (13)] and IFN-λ3–IFN-λR1–IL-10R2 complex [PDB ID: 5t5w (36)]. Such behavior is due to the dynamic nature of the whole complexes and, particularly, to plasticity of the receptor molecules.

The IL-24 gene encodes 206 aa, of which the first 51 represent the signal sequence. The construct used in this study starts with Q52 and ends with L206 (the numbering used throughout accounts for the residues belonging to the signal peptide), with all amino acids in between modeled in the electron density without any breaks, in contrast to some previously reported structures of related cytokines. As is also the case of IL-20, the structure of which was only determined in a complex with its receptors, no structural data are available for IL-24 alone, making the structure described in this article the first one for this cytokine, to our knowledge. The fold of IL-24, which it shares with other helical cytokines, is called a four-helix-bundle, despite actually containing six helices (Fig. 3A, 3F). The N-terminal residues Q52–K62 form a β hairpin, similar to the one previously seen in IL-20 (Fig. 3A, 3F). The α1 helix extends from P66 through D84 and is followed by an irregular chain and a very short helix α2 (Q94–Q98). Helix α3 extends from E104 through Q126, whereas R132–Q150 form helix α4. The helix α5 (S161–L180) is connected by a short linker to helix α6 (E183–F203). Although C59 is located in close vicinity of C106, both poised to form a disulfide bridge, the electron density does not support its existence. Whereas the side chain of C59 appears to be well ordered, the side chain of C106 accommodates two different confirmations (Fig. 4). When the SG atom of the latter residue points at SG of C59, the closest distance of their approach is 2.74 Å (Fig. 4A), which is significantly longer than the typical disulfide bond distance of 2.05 Å. The attempts to force the presence of this disulfide during refinement did not succeed and the electron density quite clearly supported the interpretation presented here (Fig. 4B–D). Therefore, one can conclude that this disulfide is formed under some conditions (as shown by mass spectrometry of the refolded IL-24 expressed in E. coli), although it is not seen in the current structure.

Superposition of the coordinates of molecule A of IL-20 from the ternary complex of that cytokine (PDB ID: 4doh) onto IL-24 yields a root mean square deviation 1.3 Å for 123 pairs of Cα atoms. Such a close similarity is present despite only limited amino acid sequence identity of these cytokines (29%). Both cytokines contain an N-terminal β hairpin, although insertion of G63 in IL-24 (absent in IL-20) leads to its significant distortion. Relative placement of helix α2 and the conformation of the following residues (Q98–Q99) differ because no corresponding motif is present in IL-20 (Fig. 3E). Helix α3 is longer in IL-24 than in IL-20, leading to significant structural differences for residues H125–R127 (Fig. 3B). Positions of the Cα atoms in the loop connecting helices α4 and α5 (P153–I162) differ by as much as 8.5 Å owing to the presence of an additional turn in the helix α4 of IL-20. Finally, IL-24 contains two additional residues on its C terminus (K205 and L206).

Sequence identity between IL-24 and IL-19 is 28.5% (similar to the identity with IL-20), and the RMSD is 1.4 Å between 133 equivalent Cα atoms of the two cytokines. Nevertheless, structures of the amino termini of both of them differ significantly, most certainly because of the presence of an additional α helix in IL-19 (PDB ID: 1n1f), replacing the β hairpin in IL-24. Other areas of differences include helix α2, the C terminus of helix α3, and the connector chain between helices α4 and α5 (the latter being virtually identical in IL-19 and IL-20). The C terminus is shorter in IL-24 than in IL-19 and points in a different direction.

IL-22 differs more from IL-24 than from the other two closely related cytokines, with the sequence identity of only 19.8%. When the chain I (IL-22) in the complex of IL-22 with IL-22R1 (PDB ID: 3dlq; this entry represents one of the two available structures of IL-22/IL-22R1, which is refined at higher resolution) is compared with IL-24, the RMSD is 2.23 Å for 121 equivalent Cα atoms. The linking region between helices α4 and α5 is not observed in that structure of IL-22. This region, however, is ordered in the structure of uncomplexed IL-22 (PDB ID: 1m4r), but the difference in chain length between IL-22 and IL-24 leads to deviations in positions of equivalent Cα atoms exceeding 6 Å.

Structures of both IL-22R1 and IL-20R2 have been determined in the past (13, 17, 18). Two copies of IL-22R1 are present in one of the structures of its binary complex with IL-22 (PDB ID: 3dgc), the components of which were expressed in insect cells. One of the molecules of the receptor is clearly glycosylated at N172, whereas no presence of the carbohydrate moiety can be detected in the other molecule. Only one molecule of IL-22R1 is present in the other similar complex (PDB ID: 3dlq), with both proteins expressed in E. coli and thus not glycosylated. In the structure described in this article (Fig. 5A, 5B), IL-22R1 is clearly N-glycosylated at N80 and N87, although no modification of the N172 side chain was found. Both previously determined structures indicated the presence of a disulfide formed between C71 and C79, whereas no such bond is seen in the current structure. Although the thiol groups from both cysteines point toward each other, in the refined structure, distance between Sγ atoms (3.1 Å) significantly exceeds the length of a standard disulfide bond (2.05 Å), and a negative electron density peak is present between these atoms. We interpret this result as showing that this disulfide might be labile. Such interpretation is additionally supported by the presence of significant negative electron density in the middle of the corresponding disulfide in 3dlq and in one of such disulfides in 3dgc, suggesting that their supposed presence may have been forced by the refinement procedures. Superposition of IL-22R1 in the complex discussed in this article onto the receptor (chain R in the PDB entry 3dgc) yields RMSD of 0.74 Å for 193 Cα pairs, whereas RMSD is 0.77 Å in a superposition with the equivalent receptor molecule in the PDB entry 3dlq. By comparison, superposition of the two receptor molecules in 3dgc yields RMSD of 0.89 Å for 202 equivalent Cα pairs. Despite the differences in expression protocols for both samples and nonidentical crystal contacts, the models of IL-22R1 are remarkably similar.

The only previously published structure of IL-20R2 is in a ternary complex of IL-20 [PDB ID: 4doh (13)]. The receptor was expressed in insect cells, but the potential N-glycosylation sites were removed by mutating N40 and N134 to Glu. Although only N134 was mutated in the IL-24 complex discussed in this article, there is no indication in the electron density map that N40 was N-glycosylated. Superposition of the two structures yields RMSD 0.94 Å for 190 equivalent Cα pairs, with the only significant differences (∼3 Å) in the loop around P60.

Interactions between IL-24 and its receptors, as well as between IL-22R1 and IL-20R2, were evaluated with the program Contact in CCP4 (38), as well as with servers PDBePIsa (http://www.ebi.ac.uk/pdbe/pisa/) and CCharPPI (https://life.bsc.es/pid/ccharppi). Computational estimation of the binding energy for a protein–protein interface in a crystallographic structure or a molecular model is a notoriously challenging problem. There have been developed dozens of computational tools based on different empirical or analytic approaches; however, their performance varies based on a particular kind of protein, and it is difficult to pinpoint a single one that would universally provide a reliable quantification (39).

According to PDBePIsa, the primary contribution to complex formation is provided by interactions between IL-24 and IL-20R2 (Fig. 5C). The interface area between these molecules extends over 875 Ų and involves 13 hydrogen bonds. With the exception of the carbonyl oxygen of L117 of IL-24, all other H-bonded atoms of the cytokine are contributed by amino acid side chains (Q52, Q126, K135, S138, T139, N142, N143, and Y204). Ten hydrogen bonds are made by the side chain atoms Y70, Y74, E75, H81, D102, T104, and K210 of IL-20R2; however, the main chain peptide carbonyl oxygen of T104 and amide nitrogen atoms of E73 and Y74 are also involved. Two interactions (H125–E75 and K135–D102) represent potential salt bridges. Several additional polar contacts use well-ordered water molecules residing within interfaces (i.e., Wat13 lodged between the amide nitrogen of T106 of the receptor and the carbonyl oxygen of G63, as well as OD1 of N143 of the cytokine). Another water-mediated contact involves Y109 OH of the receptor and S138 OG of IL-24 (Wat8). Finally, a number of hydrophobic contacts stabilize the interactions between the receptor and the cytokine. Examples include the side chains of I84 of IL-20R2 and L134 of IL-24 or the extensive contacts of the ring of Y74 of the receptor with Ala41 and F121 of IL-24. The energy gain computed with PDBePISA and due to the high-affinity interaction between IL-24 and IL-20R2 is significant (ΔⁱG = 9.1 kcal/mol).

Although the interface area between IL-24 and IL-22R1 (Fig. 5C) is slightly larger (961 Ų) and involves a larger number of hydrogen bonds between the cytokine and the receptor (17), the stabilization energy estimated by PDBePISA is quite modest (ΔⁱG = 1.7 kcal/mol). The hydrogen bonds within this interface engage the side chains of W74, K77, D84, T87, R90, Q94, K188, E192, and D194 of IL-24 as well as the main chain carbonyl oxygens of T87, A89, G191, and Y204. The counterparts contributed by IL-22R1 are the side chains of K58, Y60, E90, R112, Q117, and H161 as well as main chain carbonyl oxygens of T89 and T207 and the amide nitrogen atom of G61. Salt bridges are made by R90–E90 and D84–R112 of IL-24 and IL-22R1, respectively. No clearly defined water molecules mediate the receptor–cytokine interactions, although Wat67, despite its high B factor, may link the carbonyl oxygen of G61 of IL-22R1 with the amide nitrogen of R90 of IL-24. The most prominent hydrophobic interaction involves W208 of the receptor and W70 of the cytokine.

The interface between the two receptor chains (Fig. 5C) is much smaller (441 Ų, ΔⁱG = 2.4 kcal/mol), with only four hydrogen bonds involving the side chains of D148 and K181, as well as the main chain carbonyl oxygen of P189 and the amide nitrogen of H191 in IL-20R2, paired with the side chain of N172, main chain carbonyl atoms of Q176 and H178, and the amide nitrogen of Q176 of IL-22R1. The only observed water-mediated contact involves Wat58 interacting with the amide nitrogen of H178 of IL-22R1 and the carbonyl oxygen of H191 of IL-20R2.

The numerical results obtained with the CCharPPI server are somehow different from the estimates provided by PDBePISA. This server provided a “collective” estimate using 106 different scoring functions developed by different groups. We have correlated the scores with the binding energy using the benchmark dataset of the proteins with experimentally measured binding (30). The estimates were performed separately for IL-22R1/IL-24 and IL-20R2/IL-24 interfaces. For both interfaces, the binding energies were comparable, on the scale of ∼11 kcal/mol. Nevertheless, the majority of estimates predict the binding at IL-20R2 interface as slightly stronger, typically in the range of 0.1–0.4 kcal/mol, which is considerably smaller than the energy of a typical hydrogen bond. This analysis estimates the total binding energy for IL-24 to be around 22 kcal/mol, and it does not directly answer the question of which receptor is primary versus secondary, although it still indicates that IL-20R2 may play a role of the former one.

In this work, we have determined the crystal structure of the type II ternary complex between human IL-24 and the extracellular domains of two receptors, IL-22R1 and IL-20R2. This is only a third example of a ternary complex for a cytokine from the IL-10 family and, to our knowledge, the first detailed structural characterization of IL-24. IL-24 is more difficult to prepare in a stable form than most cytokines from the IL-10 family. Its main difference from the other members of this family, in particular other IL-20SFCs, is the low content of Cys residues. Whereas most of the related cytokines are stabilized by two or three intramolecular disulfide bonds, the two cysteines in IL-24 can only form a single intramolecular disulfide, which appears to be unstable. It was previously suggested that N-glycosylation and the formation of a single disulfide bond are both needed for manifestation of biological activities of IL-24 (7). The results of our attempts to prepare stable and functional IL-24 from bacterial cultures were only partially consistent with those earlier observations. Although preparations of deglycosylated IL-24 were quite unstable, sufficient remaining quantities of the properly folded, soluble cytokine could still engage productively with cognate receptors and initiate the JAK/STAT signaling pathway at concentrations comparable to fully N-glycosylated commercial sample of IL-24, which originated from mammalian source. These data showed that similarly to many other cytokines, N-glycosylation of IL-24 is dispensable for receptor signaling. In contrast, whether fully deglycosylated IL-24 displays the same range of structural features and biological activities as its natural variant still remains unclear. For example, full glycosylation may enforce and stabilize a disulfide bond between two Cys residues of IL-24, the bond which was not unambiguously detected in present studies. Our initial results obtained with refolded preparations of IL-24 suggested that the presence of carbohydrate may determine the stability (reflected by solubility and tendency for aggregation) of this cytokine. However, crystal structure of the IL-24 complex suggests an additional possibility. We found that the two Cys residues do not contribute much to stabilization of the cytokine through formation of a disulfide bond, which, if formed, is likely labile or transient. In turn, the tendency of IL-24 to aggregate is eliminated in the complex by stabilization provided by the receptor chains. In contrast, inclusion of receptors practically eliminates an ability to discriminate between contributions of N-glycosylation and/or the disulfide bond to the structural stabilization of IL-24.

The destabilization of IL-24 by the apparent lack of a disulfide bridge between two Cys residues may provide a molecular mechanism ensuring localized and temporary action of IL-24. For example, it was reported that expression of IL-24 in rats was transiently elevated at the skin wound edges and base with the gradual decline of its expression during wound healing process (40). Thus, the following scenario is feasible when IL-24 produced after skin injury would act only in the vicinity of the wound, because only keratinocytes expressing the IL-24R would bind and respond to this locally produced cytokine. Excess of the unbound IL-24 would be quickly degraded because of its low stability without receptor binding. Other IL-10–related cytokines that act through the same receptor but are more stable than IL-24 may provide a long-range action and similar biological activities, and the differential functions of IL-24 and other cytokines may be achieved through their distinct expression kinetics and cellular sources.

The inability to prepare sufficient quantities of pure IL-24 in free form prevented us from directly determining the receptor-binding properties of this cytokine. Although it was demonstrated before (41) and was also shown in this study that IL-24 induces receptor signaling at low nanomolar concentrations, these results originate from experiments in which intact transmembrane receptor chains are present. The results reported in this study provide, at best, an indirect evidence that the shared receptor, IL-20R2, binds to IL-24 with slightly higher affinity than either IL-20R1 or IL-22R1. Whereas our conclusion is indirect, when combined with the previously published data for IL-19 and IL-20 (13, 35), it indicates that IL-20R2 may play the role of the RH of IL-20SFCs. This observation is contrary to the other subgroup of the IL-10 family members, those utilizing the IL-10R2 receptor chain, in which this shared receptor is invariably the low-affinity component.

The overall topology of the IL-24 ternary complex described in this study is very similar to the one described earlier for IL-20 (Fig. 5A, 5B) (13). Local structural differences between both complexes are partly attributed to low amino acid sequence identity between IL-24 and IL-20 and, more importantly, to the differences in the disulfide networks. These structural differences seem too subtle to explain the differences between the biological properties of IL-24 and other IL-20SFCs, particularly IL-20, which shares identical receptor complexes. Most likely, differences in biological responses to these cytokines result from specific patterns of their expression, their different tissue and cellular distribution, and activation profiles, and are less related to their structures. However, it is tempting to speculate that unique physicochemical properties of IL-24 may play a role in its activities that are not dependent on interactions with its receptors. In contrast to other IL-20SFCs, antitumor properties of IL-24 were recognized early on, but the specific molecular mechanisms behind a direct involvement or as an adjuvant are poorly understood.

A major breakthrough in our efforts to determine the structure described in this report was introduction of the strategy utilizing receptor–ligand fusions. The use of multicomponent fusions in structural studies is not new; however, we are not aware of any earlier reports describing such a strategy for structural studies of cytokine complexes. The original motivation for utilizing this approach was to stabilize IL-24. That goal, in principle, could be achieved by linking IL-24 to its RH, if such a fusion could be expressed in quantities needed for structural studies. In our study, this proved to not be the case. Because we had no prior knowledge about the affinity of IL-24 toward individual receptor chains, many plasmids encoding different IL-24 fusions were prepared and tested for expression. During this process, we realized that additional factors that were not IL-24 specific could be incorporated in the protocol, of which the most significant was the concept of fusing the cytokine to its RL chain (in contrast to IL-24, for most of the IL-10 family members, data on receptor affinities is available). It is well known that a subgroup of IL-10–related cytokines, sharing IL-10R2, binds it with very low affinity, usually precluding formation of stable ternary complex by a simple mixing of individual components. In a somewhat fortuitous way, we found that the IL-22R1–IL-24 fusion (but not the IL-20R2–IL-24 fusion), particularly in the presence of IL-20R2, was expressed at high levels.

Based on our studies of other cytokines from the IL-10 family, we believe that the strategy outlined in this article may be generally useful in preparation of other ternary complexes (data not shown). If applicable, it accomplishes two goals: providing a stoichiometric ratio of a cytokine and its RL, and assuring mutual access of both molecules. However, a few additional considerations should be taken into account. The optimal arrangement of the components in a fusion (i.e., RL-cytokine instead of cytokine-RL) and the structure of the linker (length and composition) may require testing in each individual case. Whereas coexpression of the RL-cytokine fusion with RH was shown to be most successful for the IL-24 complex, in at least one other case we found that the RL-cytokine fusion was expressed at a very high level even in the absence of RH. In that particular case, a ternary complex was subsequently formed by mixing the purified RL-cytokine fusion with an excess of RH (data not shown). When the ternary complex is a result of coexpression, it may be necessary to supply additional amounts of RH during purification to assure stoichiometry. This would be the case when the expression level of the RL-cytokine fusion is lower than that of RH and separation of the ternary complex from an excess of the fusion might not be trivial. Inclusion of a specific cleavage site within the linker is, in our opinion, of secondary importance. In the cases when interactions between an isolated RL and cytokine or between RL and cytokine/RH binary complex are weak, cleavage of the linker will most likely result in dissociation of the ternary complex.

Data were collected at Southeast Regional Collaborative Access Team 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Supporting institutions may be found at www.ser-cat.org/members.html.

The authors have no financial conflicts of interest.