Non-core Region Modulates Interleukin-11 Signaling Activity

Abstract

Human interleukin-11 (hIL-11) is a pleiotropic cytokine administered to patients with low platelet counts. From a structural point of view hIL-11 belongs to the long-helix cytokine superfamily, which is characterized by a conserved core motif consisting of four α-helices. We have investigated the region of hIL-11 that does not belong to the α-helical bundle motif, and that for the purpose of brevity we have termed “non-core region.” The primary sequence of the interleukin was altered at various locations within the non-core region by introducing glycosylation sites. Functional consequences of these modifications were examined in cell-based as well as biophysical assays. Overall, the data indicated that the non-core region modulates the function of hIL-11 in two ways. First, the majority of muteins displayed enhanced cell-stimulatory properties (superagonist behavior) in a glycosylation-dependent manner, suggesting that the non-core region is biologically designed to limit the full potential of hIL-11. Second, specific modification of a predicted mini α-helix led to cytokine inactivation, demonstrating that this putative structural element belongs to site III engaging a second copy of cell-receptor gp130. These findings have unveiled new and unexpected elements modulating the biological activity of hIL-11, which may be exploited to develop more versatile medications based on this important cytokine.

Introduction

Human interleukin-11 (hIL-11)² is a secreted cytokine inducing a wide variety of biological effects in vivo (1–3). For example, hIL-11 is a potent hematopoietic growth factor (2), possesses anti-inflammatory activity against chronic inflammatory diseases (4), protects intestinal epithelial cells from tissue damage (5), and attenuates cardiac fibrosis after myocardial infarction (6). Moreover, hIL-11 is a medically relevant protein of clinical potential. In particular, recombinant hIL-11 (oprelvekin) has been authorized by the FDA (Food and Drug Administration) to treat adult patients following myelosuppresive chemotherapy to prevent severe thrombocytopenia (7–9).

Interleukin-11 belongs to the superfamily of long helical cytokines. Invariably, this class of ligands display a four-helix bundle motif consisting of two pairs of anti-parallel α-helices connected by three loops of variable length (10–17). Signaling is initiated by cytokine binding to cell-surface receptors, leading to oligomeric complexes that subsequently induce activation of cytoplasmic factors of the STAT (signal transducers and activators of transcription) pathway, or MAPK (mitogen-activated protein kinase) cascade (13, 18).

Specifically, binding of interleukin-11 to cell-surface receptors comprises three mechanistically separate events. First, secreted interleukin binds with low affinity to a membrane-anchored specific receptor hIL-11R (also described as α-chain receptor) (19, 20). In the second step, the hIL-11·hIL-11R heterodimer engages with high-affinity to the shared receptor gp130 forming a stable heterotrimer. In the third stage, the heterotrimer further homodimerizes yielding a hexameric complex that elicits the cell-stimulatory signals (supplemental Fig. S1)(21, 22). This general mechanism is analogous to that of other gp-130 cytokines (reviewed in Refs. 12, 13, 23, and 24).

Although hIL-11 was first described 20 years ago (1), no high-resolution structural data has been yet reported. The only available experimental model of hIL-11 corresponds to cryoelectron microscopy density maps at 30-Å resolution of the hexameric complex with cognate receptor hIL-11R and shared receptor gp130 (22). Importantly, this study indicates that the relative position of interleukin and receptors in the complex is roughly similar to that described in the high-resolution crystal structure of the hexameric complex of interleukin 6 (hIL-6) with receptors (17, 25). Together with previous mutational studies, these data unveiled up to three epitopes within the “core” α-helical region that engage cell-surface receptors: Site I (to cognate α-receptor), Site II (to shared receptor gp130), and Site III (to a second receptor gp130) (26).

Several hypotheses have been proposed to explain how the relatively unsophisticated four-helix bundle motif can lead to such a vast array of specific and overlapping biological responses. For example, it has been suggested that pleiotropy and redundancy are regulated by characteristic cell-dependent distribution of cognate receptors, or that cytokines have the ability to signal through more than one receptor (reviewed in Ref. 27). More recently, an elegant mechanism based on the novel concept of “thermodynamic plasticity” has been developed to explain cross-reactivity in shared receptor gp130 (28).

Herein, we have turned our attention to the region of hIL-11 that does not belong to the core α-helical motif, and for the purpose of brevity we have termed it the “non-core region.” Our mutational analysis demonstrates that modifications in the non-core region can also modulate the biological function of hIL-11. The majority of muteins generated in this study showed enhanced signaling activity compared with wild-type interleukin. In contrast, altering the sequence of a putative mini α-helix strategically located in the vicinity of site III can specifically inactivate the interleukin. Overall, our study has generated two interesting interleukin variants, a superagonist and a strong antagonist, which may expand the pharmacological repertoire of this medically relevant cytokine.

EXPERIMENTAL PROCEDURES

Expression and Purification of hIL-11

Expression vector pSRα-hIL-11 containing the sequence of hIL-11 (SwissProt Accession Number P20809) or its variants was co-transfected together with vector pAdD26SVA#3 at a 10:1 ratio into dihydrofolate reductase-deficient Chinese hamster ovary cells (CHOdhfr−) using Lipofectamine 2000 (Invitrogen) (29). To select the clones of interest, transfectants were initially cultured for 2 weeks in α-MEM(−) medium (Invitrogen, CA) supplemented with 10% fetal calf serum (FCS). Methotrexate (Sigma) was added to CHOdhfr(+) cells cultured in α-MEM(−) medium to increase the yield of protein expression. The concentration of methotrexate was gradually increased from 0.01 to 1 μm. Cells resistant to 1 μm methotrexate were initially cultured in 75-cm² flasks and subsequently transferred to 250-ml spinner flasks (Iwaki, Japan) containing 0.3% Cytodex 3 microcarrier particles (GE Healthcare) (30). Cultured CHOdhfr(+) cells containing hIL-11 vector (or the desired variant) were grown in α-MEM supplemented with adenosine, deoxyadenosine, thymidine, and 10% FCS. Culture supernatants containing secreted protein were filtered to eliminate unwanted cell debris and subjected to a first chromatography step using a hydroxyapatite Bio-Gel column (Bio-Rad). Protein fractions in the flow-through were collected and subjected to a second chromatographic step using a silica column. Bound protein was washed with 80 ml of each of the following solutions: 20 mm phosphate (pH 6.8); 20 mm phosphate, 40% ethylene glycol (pH 6.8); 50 mm acetate, 50% ethylene glycol (pH 4.0); 15 mm HCl, 20% ethylene glycol. Finally the interleukin was eluted with a solution containing 15 mm HCl and 50% ethylene glycol. Eluted fractions containing interleukin-11 were neutralized with 1 ml of a solution containing 1 m Tris-HCl (pH 8.0). Protein concentration was determined with an ELISA kit specific for hIL-11(R&D Systems), and spectrophotometrically at 280 nm.

Expression and Purification of hIL-11R

Cognate receptor hIL-11R was expressed in a baculovirus expression system as described previously (31). The receptor was fused to an Fc fragment to improve its solubility without affecting binding of the interleukin (32, 33). The supernatant containing the secreted receptor was diluted with Fab binding solution and loaded into a protein A chromatography column following the manufacturer's instructions (ThermoFisher Scientific). The column was washed with 10 ml of Fab binding solution to remove unwanted protein contaminants. The receptor was recovered from the column with 10 ml of Fab elution solution. Fractions containing hIL-11R-Fc were subjected to an additional size-exclusion chromatographic step to remove aggregated material. Fractions containing the receptor were concentrated and stored at −80 °C.

Cell Proliferation Assay

T1165 cells from a murine plasma cytoma line were cultured in RPM1640 medium supplemented with 10% FCS and 10 ng/ml of wild-type hIL-11 at 37 °C. Prior to each assay, cells were washed three times with medium containing no interleukin, applied to 24-well plates at a concentration of 1 × 10⁴ cells per well, and cultured in RPM1640 medium supplemented with 10% FCS at 37 °C. Wild-type hIL-11 or its variants were added at concentrations of 0.001, 0.01, 0.1, 1, 10, or 100 ng ml⁻¹. Cell proliferation was determined after 4 days of incubation using a Coulter particle counter (Beckman).

Surface Plasmon Resonance (SPR)

Binding of hIL-11 to cognate receptor hIL-11R was analyzed by SPR in a Biacore T-100 instrument (GE Healthcare) in an analogous manner to that described in Sakamoto et al. (34). Briefly, a CM5 Biacore chip decorated with hIL-11R was injected with solutions containing increasing amounts of wild-type hIL-11 or its variants. Association rate constants (k_on) and dissociation rate constants (k_off) were obtained by a global fitting procedure using the BIAevaluation software (GE Healthcare). The apparent equilibrium dissociation constant was calculated from the ratio of the kinetic rate constants: K_D = k_off/k_on.

Differential Scanning Calorimetry (DSC)

Thermal unfolding of interleukin-11 was examined in a VP-DSC instrument (MicroCal). Protein samples were heated from 10 to 120 °C at a rate of 1 °C min⁻¹. Thermograms were analyzed using the Origin software package supplied by the manufacturer of the calorimeter. Each thermogram was subtracted with the buffer baseline, and the resulting data normalized by protein concentration.

Serum Clearance Experiment

Wild-type hIL-11 and N_T-3N mutein were subjected to a in vivo serum clearance experiment. A total of 20 mice were each injected intravenously with a solution containing 5 μg of interleukin. Mice were sacrificed at several time points, their blood was collected, and the concentration of interleukin-11 in serum determined with a hIL-11 specific ELISA kit (R&D Systems).

Quantification of Monosaccharides

Neutral, basic, and sialic acid sugars covalently attached to interleukin variants were determined in an LC-9A HPLC instrument (Shimazu, Japan). In a typical assay, 50-μl samples containing glycoprotein were first evaporated, and subsequently hydrolyzed in acid solution at 100 °C for 6 h, except those samples employed to determine sialic acid content, where milder hydrolysis conditions were employed (80 °C for 1 h). Following hydrolysis, sugar content was determined in an HPLC instrument using a TSK-gel AXG column (neutral sugars), a TSK-gel SCX column (basic sugars), or a Gelpack GL-C-620–10 column (sialic acid). Each column was calibrated with commercial samples of each of the monosaccharides analyzed. Protein content in the samples was determined with a hIL-11-specific ELISA kit (R&D Systems).

Mass Spectrometry

Equal volumes of purified L_A/B-O at a concentration of 0.1 mg/ml and sinapinic acid mixture were mixed, deposited on a MALDI plate, briefly dried, and the mass spectrum collected in a MALDI-TOF-MASS Voyayer instrument (Shimadzu, Japan).

Preparation of Deglycosylated N_T-3N Interleukin

Variant N_T-3N was deglycosylated with N-glycosidase A enzyme (Roche Applied Science) following the manufacturer's instructions. Briefly, the appropriate amount of N-glycosylated interleukin was mixed with 5 milliunits of N-glycosidase A, diluted with 150 μl of PBS buffer (pH 5.0), and incubated for 24 h at 37 °C. Proliferation of T1165 cells in the presence of glycosylated or deglycosylated N_T-3N mutein was monitored as described above.

RESULTS

Homology Model of hIL-11

The homology model of hIL-11 depicted in Fig. 1 was obtained from the ModBase database (35). This model is based on the crystallographic coordinates of human granulocyte colony-stimulating factor, which also belongs to the long helix cytokine superfamily (16). Granulocyte colony-stimulating factor was selected over other suitable candidates such as hIL-6 because of a higher sequence identity to hIL-11 (24 and 19%, respectively; Fig. 1C) and a more complete coverage of the primary sequence in the final model. Only the first five and last three residues of the mature hIL-11 protein are absent in the model depicted in Fig. 1A. Overall, the quality of this model is very acceptable as demonstrated during the analysis of its geometry with the program MolProbity (36), which indicated that only 3% of residues fall outside the allowed regions in the Ramachandran plot (not shown). As a comparison, 4.9% of residues of hIL-6 in the crystal structure in complex with gp130 and α-receptor fall in non-allowed regions of the Ramachandran plot (PDB accession code 1P9M).

FIGURE 1.

Modeled three-dimensional structure of hIL-11. A, schematic representation; B, topology model; and C, sequence alignment. The α-helices are named in alphabetical order as they appear in the primary sequence of the interleukin: helix A (magenta), helix B (yellow), helix C (blue), and helix D (green). Non-core regions are shown in blue, except for a predicted mini α-helix in Loop_A/B, which is highlighted in orange. Residues belonging to non-core regions (depicted in red) indicate the positions at which glycosylation sites were introduced. In panel B, the approximate location of these residues is indicated with red stars. The three-dimensional structure of hIL-11 was modeled from the crystal structure of granulocyte colony-stimulating factor in complex with its cognate receptor (16). Atomic coordinates were obtained from the ModBase and are depicted without further modification (35). Panel C shows the sequence alignment of human interleukin-11, human granulocyte-colony stimulating factor, and human interleukin-6. The colored boxes indicate the location of α-helices. The predicted mini helix in Loop_A/B of hIL-11 (orange), and other small helical elements (gray) are also highlighted. Arrows indicate the positions at which modifications were introduced. Molecular graphics and sequence alignment figures were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (54).

The homology model shown in Fig. 1 exhibits the main signature element of this family of cytokines, i.e. a four-helix bundle arranged in a top-top/down-down topology (12, 23). Residues not belonging to the core α-helical motif, for the purpose of brevity we termed non-core region, are arranged in loops of variable length connecting the α-helical elements. Interestingly, the model predicts the presence of a mini α-helix comprising residues 48–51 of Loop_A/B (Fig. 1) that we have modified in the course of this study (see below).

Cell Proliferation Activity of Wild-type hIL-11 and Muteins

To examine the role of non-core regions in the signaling activity of hIL-11, we prepared engineered variants of the interleukin containing N- or O-glycosylation consensus sequences at different locations of the non-core region: N terminus; C terminus; Loop_A/B connecting helix A to helix B; Loop_B/C connecting helix B to helix C; and Loop_C/D connecting helix C to helix D (Fig. 1 and supplemental Table S1). Thus each mutein contained between three and nine more residues than wild-type interleukin, and a carbohydrate moiety (see below). This mutational approach modifies significantly the primary structure of each targeted position, which may reveal specific roles of non-core elements in the activity of hIL-11.

Wild-type interleukin and muteins were expressed in CHO cells, purified to homogeneity, and their cell proliferation activity assessed in cultured T1165 cells. Fig. 2 shows that wild-type hIL-11 and all the muteins tested, except L_A/B-O, induced robust proliferation of T1165 cells at interleukin concentrations ranging from 0.1 to 100 ng ml⁻¹. Remarkably, most muteins exhibited higher cell proliferation activities than wild-type interleukin, i.e. superagonist behavior. For example, C_T-O, N_T-O, and N_T-3N are ∼10-fold more active than wild-type interleukin in a protein concentration basis. We could not observe significant differences between N- and O-glycosylated variants. The highest proliferation rates occurred in the presence of muteins carrying modifications at the terminal ends.

FIGURE 2.

Cell proliferation assay. T1165 cells derived from a murine plasmacytoma line were incubated with wild-type interleukin (black), or the following variants: N_T-O (light green), N_T-2N (light gray), N_T-3N (light pink), L_A/B-O (purple), L_B/C-O (light blue), L_B/C-N (light brown), L_C/D-O (orange), L_C/D-N (blue), and C_T-O (brown). Cell growth was monitored after 4 days. Data correspond to the average of four assays.

Incorporation of Glycan Moieties

N_T-3N and L_A/B-O variants, showing opposite effects in the proliferation activity assay above, were chosen to examine the extent of glycan incorporation in the muteins. Monosaccharide fragments obtained after hydrolysis of purified protein samples were quantified by HPLC (Table 1). The data indicated that both muteins, N_T-3N and L_A/B-O, contained several types of saccharides attached to them. In N_T-3N the relative molar ratios among the major sugars found fucose, GlcNAc, mannose, galactose, and Neu5Ac were 1:5.6:2.5:2.6:3.1. These values compared favorably with the composition of triantennary N-glycans 1:5:3:3:3 (supplemental Fig. S2) (37, 38). Moreover, the molar ratio between N-glycan and N_T-3N interleukin determined in the assay was 0.9 (taking fucose as reference), which demonstrated that in average each molecule of interleukin had incorporated, about one N-glycan moiety.

TABLE 1.

Sugar content of hIL-11 variants

Concentration of monosaccharides rhamnose, ribose, xylose, glucose, N-acetylmannose, and N-glycolylneuraminic acid was below 5% that of other sugars, or not detected in the assay.

Monosaccharide	NT-3Na		LA/B-Oa, sugar molar ratio (GalNAc = 1)
	Sugar molar ratio (fucose = 1)	Sugar/protein molar ratio
Mannose	2.5	2.3	NAb
Fucose	1.0	0.9	NA
Galactose	2.6	2.3	0.86
GlcNAcc	5.6	4.9	NA
GalNAcd	NA	NA	1.00
Neu5Ace	3.1	2.8	1.24

^a Although GalNAc is not incorporated in standard N-glycans, we found minor quantities of this sugar during analysis of N_T-3N (GalNAc to GlcNAc ratio = 0.15).

^b NA indicates that the value is not applicable.

^c GlcNAc is an abbreviation for N-acetylglucosamine.

^d GalNAc is an abbreviation for N-acetylgalactosamine.

^e Neu5Ac is an abbreviation for N-acetylneuraminic acid.

In L_A/B-O, the molar ratios among the major sugars GalNAc, galactose, and Neu5Ac were 1:0.9:1.3. These values contrasted with relative molar ratios of standard O-glycans (molar ratios = 1:1:2; supplemental Fig. S2) (37, 38). In particular, the relative concentration of sialic acid Neu5Ac was lower than expected. We collected MALDI-TOF mass spectrum of L_A/B-O to examine the origin of these differences (Fig. 3). The mass spectrum showed two major peaks appearing at m/z values of 20,840 and 20,546 Da that we assigned to interleukin modified with full size O-glycan (calculated mass = 20,799 Da), and O-glycan lacking one molecule of sialic acid (calculated mass = 20,508 Da), respectively. The presence of two forms of L_A/B-O differing in one molecule of sialic acid qualitatively explained the sugar ratios observed in the analytical HPLC assay. Moreover, the complete absence of m/z peaks around 19,850 Da (which corresponds to the theoretical mass of non-glycosylated interleukin) demonstrated that the L_A/B-O variant was quantitatively glycosylated. In summary, these data indicated that N- and O-glycans were incorporated in N_T-3N and L_A/B-O variants, respectively.

FIGURE 3.

MALDI-TOF spectrum of variant L_A/B-O. The position of the two major m/z peaks are indicated by arrows. The putative structure of O-glycans attached to L_A/B-O corresponding to each major m/z peak are shown in the panel. The symbols used for the sugar moieties are: GalNAc (open square), galactose (open circle), and sialic acid Neu5Ac (filled diamond). The symbols used to represent the sugar moieties are based on the recommendations of Varki et al. (55).

Binding to α-Receptor

SPR was employed to evaluate the affinity of wild-type interleukin, variant N_T-3N, and variant L_A/B-O, for cognate α-receptor (hIL-11R). Samples containing the cytokine were injected into a sensor chip in which hIL-11R had been previously immobilized. Representative real time response curves (sensorgrams) are shown in Fig. 4. Association rate constants (k_on) and dissociation rate constants (k_off) were determined from the sensorgrams by applying a global fitting regression assuming a stoichiometry of 1:1 (Table 2). Binding of wild-type hIL-11 to the receptor is characterized by a rapid association phase, followed by an also fast dissociation phase. In contrast, binding kinetics of mutein N_T-3N displayed slower association and dissociation steps, whereas the kinetic profile of L_A/B-O was intermediate between those of wild-type and N-glycosylated interleukins. The dissociation constant (K_D) calculated from the kinetic constants conveys the overall affinity of the ligand for the receptor, where higher values of K_D denote lower affinity. K_D values determined for wild-type, N_T-3N, and L_A/B-O interleukins were 8.1, 0.91, and 0.76 μm, respectively (Table 2). Variant L_A/B-O, which presented the lowest signaling activity in Fig. 2, was nonetheless the variant showing the strongest affinity for hIL-11R.

FIGURE 4.

Association kinetics of interleukin-11 to cognate receptor hIL-11R. SPR sensorgrams are shown as red traces. Black curves correspond to the global fitting of the sensorgrams with the BIAevaluation software package (GE Healthcare). An arrow pointing downwards indicates injection of running buffer containing wild-type hIL-11 (A), N_T-3N (B), or L_A/B-O (C) variants. An arrow pointing upwards indicates injection of running buffer containing no interleukin. The response signal is proportional to the amount of interleukin that binds to a chip surface decorated with immobilized receptor.

TABLE 2.

Kinetic parameters of the association of interleukin-11 to receptor hIL-11R

Protein	kon	koff	KD
	m−1s−1 × 10−3	s−1 × 103	μm
Wild type	4.7 ± 0.1	38 ± 0.5	8.1 ± 0.5
NT-3N	1.1 ± 0.05	1.0 ± 0.01	0.91 ± 0.05
LA/B-O	5.3 ± 0.05	4.0 ± 0.03	0.76 ± 0.06

We note that the fast binding kinetics of wild-type interleukin to α-receptor resulted in poorer fitting profiles than those of the muteins. To corroborate the robustness of the analysis above, the binding response in equilibrium was used to determine K_D (Scatchard plot, supplemental Fig. S3). The K_D value obtained by this alternative methodology (9.0 ± 1 μm) was essentially identical, within experimental error, to that estimated from Fig. 4A (8.1 ± 0.5 μm).

Variant L_A/B-O Is a Strong Antagonist of Wild-type hIL-11

SPR data above suggested that L_A/B-O is an antagonist of wild-type hIL-11, i.e. competes with wild-type interleukin for α-receptor without eliciting cell proliferation (Figs. 2 and 4). To test this hypothesis, T1165 cells were simultaneously incubated with a fixed amount of wild-type interleukin (100 ng ml⁻¹) and varying concentrations of L_A/B-O (Fig. 5A). In the absence of L_A/B-O mutein, wild-type interleukin promoted robust cell proliferation comparable with that shown in Fig. 2. However, addition of L_A/B-O in the cell culture rapidly reduced cell growth, even in the presence of a large excess of wild-type interleukin. When the concentration of L_A/B-O was equal or higher than 10 ng ml⁻¹ we ceased to observe cell stimulation. Interestingly, the concentration ratio at which the glycosylated variant inactivates wild-type interleukin ([wild-type]/[L_A/B-O] = 10) matches the relative affinities of the interleukins for α-receptor shown above (K_D(wild-type)/K_D(L_A/B-O) = 10).

FIGURE 5.

Effect of modifications introduced in Loop_A/B in cell proliferation. A, cell proliferation of T1165 cells in the presence of both, wild-type hIL-11 at a fixed concentration of 100 ng ml⁻¹, and variable concentrations of L_A/B-O. Data correspond to the average of four assays. B, cell proliferation assay of T1165 cells in the presence of wild-type hIL-11 (solid), variant L_A/B-O (dashed and dotted), variant L_A/B-N (dashed), or variant L_A/B-8Gly (dotted). Data correspond to the average of four assays.

An additional experiment was performed to investigate the specificity of the modification in Loop_A/B. Two new muteins engineered at the same positions as L_A/B-O were prepared: the first one contained a N-glycosylation site (L_A/B-N), whereas the second one displayed an octaglycine sequence not bearing glycosylation signal (L_A/B-8G). Two key observations were made. First, these two new variants showed a reduced ability to induce cell proliferation, supporting the idea that the amino acidic sequence of this region is important for the proper activity of the interleukin (Fig. 5B). Second, the use of variant L_A/B-8G demonstrated that N- or O-glycans are not necessarily involved in the mechanism of inactivation.

Effect of Modifications in the Stability of Interleukin

DSC was employed to assess the thermal stability of the interleukins (Fig. 6). On the one hand, wild-type and N_T-3N proteins showed a complex thermal unfolding behavior characterized by three transitions centered around 28, 62, and 76 °C (Table 3). The enthalpies of each transition were very similar, ranging between 53 and 68 kcal mol⁻¹. On the other hand, L_A/B-O exhibited a rather different thermal profile compared with that of the other two proteins. First, L_A/B-O showed a single and broad unfolding transition centered at 68 °C. Second, the transition enthalpy of L_A/B-O was considerably higher than that of any single transition of the other two variants.

FIGURE 6.

Thermal unfolding of interleukin-11 monitored by DSC. Samples of wild-type hIL-11 at 0.79 mg ml⁻¹ (solid line), N_T-3N variant at 0.97 mg ml⁻¹ (dashed line), and L_A/B-O variant at 0.43 mg ml⁻¹ (dashed and dotted line) in PBS buffer at pH 7.0 were heated from 10 to 120 °C at a rate of 1 °C min⁻¹. Heat capacity values (C_P) were obtaining after subtraction of the buffer baseline and normalized by protein concentration using the software package Origin.

TABLE 3.

Thermal Stability of wild-type interleukin-11 and muteins determined by DSC

T_M denotes the transition midpoint temperature, and ΔH indicates the enthalpy of the transition.

Protein	TM1	TM2	TM3	ΔH1	ΔH2	ΔH3	ΔHTotal
	°C			kcal mol−1
Wild type	27.5 ± 0.05	61.5 ± 0.1	76.3 ± 0.1	68.0 ± 0.3	53.6 ± 0.4	58.0 ± 0.4	179 ± 1
NT-3N	28.7 ± 0.05	63.6 ± 0.1	75.5 ± 0.1	64.8 ± 0.3	53.0 ± 0.4	65.5 ± .04	183 ± 1
LA/B-O	68 ± 0.5	NAa	NA	245 ± 2	NA	NA	245 ± 2

^a NA, not applicable.

To determine whether the dissimilar thermal behavior between L_A/B-O and wild-type interleukin (and N_T-3N variant) was caused by major changes of secondary structure, we acquired their circular dichroism (CD) spectra in the far UV region (200–250 nm) (supplemental Fig. S4). We could not observe significant differences among the spectra, indicating that the secondary structure of the cytokines remained relatively unchanged upon mutation. We should point out that CD data presented in supplemental Fig. S4 could not be used to estimate minor structural perturbations caused by the introduction of the glycosylation sites.

We further explored whether differences in stability (other than thermal stability) between wild-type interleukin and superagonist N_T-3N could explain the enhanced cell proliferation activity of the glycosylated variant. Two types of experiments were conducted. In the first one, cell proliferation activity of freshly purified protein was compared with that of samples stored 4 months in stabilizing buffer at 26 °C (Fig. 7A). Under both conditions, mutein N_T-3N showed higher cell growth stimulation than wild-type interleukin. More importantly, wild-type interleukin activity decayed faster than that of N_T-3N. The EC₅₀^NT-3N/EC₅₀^WT ratio, which indicates the relative proliferation strength of each protein, declined from a value of 0.2 when samples were prepared fresh, to about 0.03 after their long-term storage period. In other words, cell proliferation activity induced by wild-type interleukin decayed ∼7-fold faster than that of mutein N_T-3N.

FIGURE 7.

Stability and serum clearance of wild-type and N_T-3N variant. A, cell proliferation induced by the N_T-3N variant or wild-type hIL-11 were compared when fresh samples (black) or samples stored for 4 months at 26 °C (light gray) were employed. EC₅₀ indicates the protein concentration at which the interleukin induces 50% maximum proliferative effect. Bars correspond to the average of two measurements. B, time course of interleukin clearance from mouse serum after intravenous injection of 5 μg of either wild-type (solid line) or N_T-3N variant (dashed line). Protein concentration was determined with an hIL-11-specific ELISA kit.

In a second experiment, we monitored the in vivo clearance rate of wild-type and N_T-3N interleukins in serum of mice injected with a fixed amount of interleukin (5 μg). Differences between wild-type hIL-11 and N_T-3N were dramatic (Fig. 7B). Five minutes after injection the concentration of wild-type hIL-11 in serum had dropped below 1 ng ml⁻¹, whereas that of N_T-3N remained much higher (19 ng ml⁻¹). The levels of N_T-3N decreased steadily to 2 ng ml⁻¹ during the first 40 min after the injection, and after that more slowly up to 75 min where this variant was no longer detected under our assay conditions.

The role of N-glycan moiety in the stimulatory properties of N_T-3N was also investigated. Cell proliferation activity of the N_T-3N treated with N-glycosidase A (deglycosylated) was compared with that of untreated samples. Fig. 8 shows that both types of samples led to similar levels of cell growth for a wide range of concentrations, except at the highest concentrations of interleukin tested where the glycosylated interleukin N_T-N3 displayed significantly higher activity. These data suggests a critical role (but not an exclusive one) of the N-glycan in the superagonistic behavior of the muteins.

FIGURE 8.

Treatment of N_T-3N with N-glycosidase A limits cell stimulation. Solid squares and solid line correspond to the untreated N_T-3N protein, and empty squares and dotted line correspond to N_T-3N treated with N-glycosidase A. Cell growth was monitored after 4 days. Data correspond to the average of four assays.

Overall, this set of experiments demonstrated that engineering a N-glycosylation site at the N terminus of hIL-11 improved the cell stimulatory potential of the cytokine, and strengthened its long-term and in vivo stabilities. These enhanced properties occurred in a glycosylation-dependent manner without disrupting the secondary structure of the protein or its thermal denaturation profile.

DISCUSSION

Non-core Region Modulates hIL-11 Signaling Activity

Several mutational studies had established that certain regions within the four-helix core motif of hIL-11 (termed site I, site II, and site III) are essential for binding of the cytokine to cell-surface receptors (27, 32, 39–41) (supplemental Fig. S1). This core motif is well conserved among all helical cytokines, raising the interesting question of how this family of regulators can exert their specific stimulatory effect based on such a simple and widespread molecular design. This question is particularly relevant for those cytokines that share common cell-receptor gp130 (24, 26, 28). In this study we focused our efforts to the less structurally conserved region of hIL-11, termed the non-core region, to determine whether it could also play a significant role in the normal functioning of interleukin.

A previous site-directed mutagenesis study had identified Met-59 (located in Loop_A/B of the non-core region) as a key residue for the normal activity of hIL-11 (41). Our strategy consisted in altering, more substantially, selected locations within the non-core region of hIL-11 by introducing between three and nine residues containing glycosylation sites. The initial screening revealed that Loop_A/B (modified between His-49 and Asn-50) is critical for the stimulatory properties of hIL-11. In contrast, all other muteins exhibited comparable or even higher cell proliferation activities (up to 10-fold higher) than that of wild-type protein (Fig. 2). This observation indicated that the non-core region is very versatile at adapting to modifications in its primary structure and also to the attachment of carbohydrates. Interestingly, cytokine GSGF (whose three-dimensional structure was used to build the hIL-11 model depicted in Fig. 1) also tolerates large rearrangements of its primary sequence while keeping a biologically active conformation (42).

The high propensity (>50%) of hIL-11 muteins exhibiting superagonist behavior suggests that biological evolution has not optimized this interleukin to achieve maximum cell stimulatory strength. Perhaps an “over-optimized” interleukin with, for example, much higher affinity for receptor hIL-11R (as observed in N_T-3N variant) could limit the ability of other cytokines to signal through shared receptor gp130. An attractive hypothesis resulting from these observations is that the non-core region of hIL-11 has been biologically designed to restrain the full stimulatory potential of hIL-11.

Modification of a Putative Mini α-Helix at Loop_A/B Generates a Potent Antagonist

L_A/B-O is a powerful antagonist of hIL-11 that exhibits very low cell proliferation activity (Figs. 2 and 5). In particular, Fig. 5A demonstrates that a large excess of wild-type interleukin (∼10-fold) is required to restore cell growth when the L_A/B-O variant is also present in the medium.

Superposition of the three-dimensional model of hIL-11 onto the crystal structure of the hexameric complex of hIL-6 with receptors (17, 22), showed that the putative mini α-helix modified in L_A/B-O is located in the immediate vicinity of the second copy of receptor gp130 (supplemental Fig. S1). We have demonstrated that this predicted element is important for the stimulatory properties of hIL-11. First, modification of the mini helix led to glycan-independent inactivation (Figs. 2 and 5B). Second, inactivation rates were sensitive to the number and size of residues incorporated in the mutein (Fig. 5). And third, changes in the putative helix did not diminish the ability of L_A/B-O to bind to the cognate receptor hIL-11R (Fig. 4). Well on the contrary, the affinity of L_A/B-O for α-receptor was higher than that of wild-type interleukin, which explained why this variant displays antagonistic properties.

In addition, previous studies had shown that Trp-147 (located ∼12 Å from the mini helix modified in our study) is a key component of site III (27,32). Replacement of this key Trp residue with Ala (W147A mutein) generates a strong antagonist with little proliferation activity because of a defective binding to receptor gp130 (32). Together with these previous reports, our data supports the idea that the mini-helical element modified in L_A/B-O forms part of an expanded site III.

In summary, we have established that L_A/B-O is a new and strong antagonist of hIL-11, and thus a compound of potentially interesting pharmacological properties. This variant of hIL-11 (and other analogous variants with L_A/B-O-like properties) could be employed, for example, to ameliorate pathologies that involve endogenously hyperactive interleukin-11 such as in certain classes of inflammation (43), or in conditions where receptor hIL-11R is overexpressed such as in bone metastasis (44). However, further experimentation will be needed before these novel medications can be put into practice.

Generation of Superagonist Variant N_T-3N

N_T-N3 is a superagonist variant exhibiting ∼10-fold higher cell proliferation activity than wild-type interleukin (Fig. 2). Our data revealed several overlapping mechanisms contributing to the enhanced activity of N_T-3N. First, SPR binding data indicated that N_T-N3 forms a tight complex with α-receptor (10-fold stronger) that is characterized by slower dissociation rates (k_off) than those observed in wild-type interleukin. In other words, N_T-N3 remained attached to α-receptor for a longer period of time than wild-type interleukin, which in the biological environment of the cell could increase the likelihood of forming functional complexes through subsequent binding of receptor gp130. Second, although the modifications introduced in N_T-3N did not alter thermal stability or secondary structure of the mutein, inactivation rates upon long-term storage and serum clearance rates were greatly slowed down in this variant (Fig. 7). And third, the presence of carbohydrate favored the biological activity as illustrated in Fig. 8. Interleukin 11, contrary to other cytokines such as hIL-6 or granulocyte colony-stimulating factor (45, 46), is not glycosylated in its native form (47). Dissimilar glycosylation patterns among interleukins could contribute to the fascinating diversity of signaling events elicited by this structurally homologous family of cytokines.

Overall, our data are consistent with previous studies showing that protein glycosylation may result in enhanced stability but not necessarily higher thermostability (37, 48–50). Indeed, a similar mechanism has been proposed for a PEGylated form of hIL-11, although the unspecific nature of this chemical modification resulted in a cytokine with diminished ability to bind to receptors (51). The specificity of our mutational scheme (as demonstrated in N_T-3N) has overcome this limitation, leading to a modified interleukin with enhanced biological activity. In principle, the favorable properties of this new variant would seem suitable to replace the currently administered recombinant (non-glycosylated) hIL-11 in the treatment of severe thrombocytopenia (7–9).

Several alternative routes to produce the hIL-11 superagonists have been described in the literature, such as the design of soluble hIL-11·hIL-11R complexes (52), or the preparation of single-point mutations in the core α-helical region of the cytokine (40, 53). Herein we have shown that modification of the non-core region of hIL-11 is a plausible route to generate more potent bioactive variants of this pharmacologically valuable cytokine.

Conclusions

In this study we have shown that modifications of the primary sequence within the non-core region may modulate the function of hIL-11 in two opposite directions depending on the nature of the mutation. First, in the majority of muteins tested the signaling activity of the interleukin was augmented in a relatively nonspecific manner. And second, modification of a predicted mini helix in the vicinity of site III disrupted the cell stimulatory properties of the cytokine. The generation of these novel superagonist and antagonist variants of hIL-11 opens new avenues for the development of more versatile medications of this clinically relevant cytokine.