Abstract
Chemokines like CCL2 mediate leukocyte migration to inflammatory sites by binding to G-protein coupled receptors on the target cell as well as to glycosaminoglycans (GAGs) on the endothelium of the inflamed tissue. We have recently shown that the dominant-negative Met-CCL2 mutant Y13A/S21K/Q23R with improved GAG binding affinity is highly bio-active in several animal models of inflammatory diseases. For chronic indications, we have performed here a fusion to human serum albumin (HSA) in order to extend the serum half-life of the chemokine mutant. To compensate a potential drop in GAG-binding affinity due to steric hindrance by HSA, a series of novel CCL2 mutants was generated with additional basic amino acids which were genetically introduced at sites oriented towards the GAG ligand. From this set of mutants, the Met-CCL2 variant Y13A/N17K/S21K/Q23K/S34K exhibited high GAG-binding affinity and a similar selectivity as wild type (wt) CCL2. From a set of different HSA-chemokine chimeric constructs, the linked HSA(C34A)(Gly)4Ser-Met-CCL2(Y13A/N17K/S21K/Q23K/S34K) fusion protein was found to show the best overall GAG-binding characteristics. Molecular modeling demonstrated an energetically beneficial fold of this novel protein chimera. This was experimentally supported by GdmCl-induced unfolding studies, in which the fusion construct exhibited a well-defined secondary structure and a transition point significantly higher than both the wt and the unfused CCL2 mutant protein. Unlike the wt chemokine, the quaternary structure of the HSA-fusion protein is monomeric according to size-exclusion chromatography experiments. In competition experiments, the HSA-fusion construct displaced only two of seven unrelated chemokines from heparan sulfate, whereas the unfused CCL2 mutant protein displaced five other chemokines. The most effective concentration of the HSA-fusion protein in inhibiting CCL2-mediated monocyte attachment to endothelial cells, as detected in the flow chamber, was 8.6 µg/ml. This novel HSA-fusion protein exhibits not only high affinity but also selective displacement of chemokines from GAGs binding. HSA is therefore proposed to be a highly promising scaffold candidate for therapeutic, GAG-targeting chemokine mutants.
Introduction
Chemokines are small secreted proteins that function as intercellular messengers to orchestrate activation and migration of specific types of leukocytes from the lumen of blood vessels into tissues (Baggiolini, 2001). The chemokine-induced migration of leukocytes is mediated on the one hand by the interaction of chemokines with seven transmembrane G-protein coupled receptors (GPCRs) on the surface of the respective target cells. On the other hand, chemokines are interacting with cell surface glycosaminoglycans (GAGs). This interaction establishes a local concentration gradient and leads to immobilization of chemokines which ensures an optimal presentation of the proteins at the site of inflammation. GAGs therefore mediate the directional migration of leukocytes by providing the basis for a solid phase chemokine gradient (Esko and Lindahl, 2001; Nourshargh et al., 2010). GAGs are acidic linear polysaccharides consisting of repeating disaccharide units. These disaccharide units contain a hexosamine (N-acetylglucosamine or N-acetylgalactosamine or its derivate) and a hexuronic acid (glucuronic or iduronic acid) (Raman et al., 2005; Gesslbauer et al., 2007). Owing to their acidic composition, GAGs are able to interact with a variety of proteins such as proteases, cytokines, adhesive molecules and growth factors (Vlodavsky et al., 2002; Kolset et al., 2004).Three consensus amino acid sequences have been identified in typical GAG-binding proteins: –XBBXBX–, –XBBBXXBX– and –XBBXXBBBXXBBX–, where B is a basic amino acid and X stands for any residue (Cardin and Weintraub, 1989; Sobel et al., 1992).
Chemokines can be divided into CXC, CC, CX3C and X3C families according to their structure and the position of the typical cysteine residues. All chemokines (with the exception of lymphotactin/XCL1 and fraktaline/neurotactin/CX3CL1 that are members of the C and CX3C chemokine subfamily) have four cysteins in conserved positions and can be divided into the CXC and the CC subfamilies on the basis of the presence or absence of one amino acid between the two cysteines within the N-terminus. Monocyte chemoattractant protein-1 (MCP-1/CCL2), a member of the CC chemokine family, is a monocyte and lymphocyte-specific chemoattractant which activates CCR2 and CCR4 on monocytes, macrophages, dendritic cells and memory T cells (Ingersoll et al., 2011). It's up-regulation occurs in a variety of diseases that feature a monocyte-rich inflammatory component, such as atherosclerosis (Nelken et al., 1991; Ylä-Herttuala et al., 1991), rheumatoid arthritis (Koch et al., 1992; Hosaka et al., 1994; Robinson et al., 1995), inflammatory bowel disease (MacDermott, 1999) and congestive heart failure (Aukrust et al., 1998; Hohensinner et al., 2006). CCL2 is also involved in metastasis formation, by promoting tumor cell extravasation (Wolf et al., 2012). Chemokines have two major sites of interaction with their GPCRs, one in the N-terminal domain which functions as a triggering domain, and the other within the exposed loop after the second cysteine, which functions as a docking domain (Gupta et al., 1995). The GAG-binding sites of chemokines comprise clusters of basic amino acids spatially distinct (Ali et al., 2001). Some chemokines, such as CCL5/RANTES, have the –BBXB– motif in the 40s loop as major GAG-binding site; CXCL8/IL-8 interacts with GAGs through the C-terminal a-helix and Lys20 in the proximal N-loop. Other chemokines, such as CCL2/MCP-1, show a significant overlap between the residues that comprise the GPCR-binding site and the GAG-binding site (Lau et al., 2004).
To target the interaction of chemokines and GAGs, we have generated dominant-negative decoy proteins with increased GAG-binding affinity and knocked-out or reduced GPCR-binding affinity (Falsone et al., 2013). For a higher GAG-binding affinity, the wild type (wt) CCL2 was mutated in our earlier studies at positions 21 and 23 resulting in a mutant protein with significantly increased GAG binding compared with wt CCL2. For knocked-out GPCR binding, the mutation the Y13A was necessary to lower the activity of the wt (Paavola et al., 1998). A methionine residue at position −1 at the N-terminus—which was the result of Escherichia coli expression—further decreased the CCR2 activation of this mutant and appeared to have an additional positive impact on the GAG-binding affinity (Piccinini et al., 2010).
Here, we report the engineering and in vitro characterization of a novel HSA-CCL2 mutant fusion protein, which may have superior characteristics compared with unfused CCL2 mutants and which may therefore lead to a novel treatment of pathologies in which CCL2 levels are up-regulated such as oncological indications.
Materials and methods
Materials
GAG-binding plates, LMW heparin, HMW heparin, heparan sulfate (HS) and dermatan sulfate (DS) were purchased from Iduron (Manchester, UK), all chemicals, unless stated otherwise, from Sigma-Aldrich (St Louis, MO, USA). CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 were generated in house (see below). Phosphate-buffered saline (PBS) pH 7.2 contains 10 mM phosphate buffer and 137 mM NaCl.
Molecular design
For the HSA structure, the PDB entry 4K2C was used (DOI:10.2210/pdb4k2c/pdb) and opened in the molecular modeling program YASARA Version 13.9.8 (www.yasara.org) (Krieger et al., 2002; Wang et al., 2013). As the asymmetric unit contains two chains, chain B was deleted and missing atoms were completed using the ‘Clean’ command. Furthermore, the missing amino acids at the N-terminus (D1, A2 and H3) and at the C-terminus (L583, G584 and L585) were added with dihedral angles phi = X and psi = Y. For the subsequent linkage to dnCCL2, a glycine was introduced C-terminal at position 586. Then, a 1 ns-molecular dynamics simulation in aqueous solution was run at 298 K with YASARA using the AMBER03 force field (Duan et al., 2003) and PME (Essmann et al., 1995) to relax the resulting HSA structure. The standard macro md_run assigns not only the force field parameters, but also places counterions, predicts pKa values and conducts an energy minimization before starting the simulation. Snapshots were saved every 25 ps.
The dnCCL2 structure was designed by mutating the wt structure of CCL2. Therefore, one monomer of the PDB file 1DOM (DOI:10.2210/pdb1dom/pdb) was modified by introducing the mutations T13A, N17K, S21K, Q23K, S34K as well as an N-terminally leading methionine residue (Handel and Domaille, 1996). Finally, the linker GGGGS was added at the N-terminus of the CCL2 mutant and rigorous energy minimization was performed.
Expression and purification
Expression and purification of dnCCL2
Mutant genes were synthesized from DNA2.0 and were obtained in the pJExpress411 expression vector for further transformation into BL21 (DE3) Star E.coli cells (Invitrogen, Carlsbad, CA, USA).
Starting cultures were prepared and used for protein expression. Cultures were grown in 3-l Erlenmeyer flasks under 200 rpm shaking at 37°C in LB broth containing 30 μg/ml Kanamycin to an A600 of 0.8. Protein production was induced by the addition of 0.5 mM isopropyl β-d-thiogalactopyranoside. Cells were incubated with shaking for additional 3 h and harvested by centrifugation for 15 min at 6000g. Further expression and purification was performed as described earlier (Piccinini et al., 2010).
Protein expression and purification of the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant
Expression of HSA(C34A)-(Gly)4Ser-dnCCL2 was carried out using Pichia pastoris as expression host in a 1 l Multifors bioreactor (Infors AG, Bottmingen, Switzerland). The two-step fermentation process comprised a growth phase on glycerol followed by a production phase on methanol as sole carbon source. Pichia pastoris CBS7435 muts-PDI strain was inoculated in a starting volume of 0.4 l of fermentation medium (1/2 BSM containing 40 g/l glycerol) at a temperature of 28°C and pH 5.0. The temperature was phased-down from 28 to 24°C and pH was increased to 6.0 during the last 2 h of glycerol fed-batch and maintained at that level throughout the production time. The cultivation resulted in 6.5 g/l of the target protein with 62% purity.
To further purify the full length HSA(C34A)-(Gly)4Ser-dnCCL2, a two-step purification was performed. In the first step, a strong cation exchange resin—Fractogel EMD SO3– (Merck, Darmstadt, Germany)—which was specifically interacting with the chemokine part of the construct, was used. The second step was an affinity chromatography resin, Blue Sepharose 6 Fast Flow (GE Healthcare, Chalfont St Giles, UK), which is Cibacron™ Blue 3G covalently attached to the Sepharose 6 Fast Flow matrix by the triazine coupling method. The blue dye binds many proteins, such as albumin, interferon, lipoproteins and blood coagulation factors. Furthermore, it binds several enzymes including kinases, dehydrogenases and most enzymes requiring adenyl-containing cofactors, e.g. NAD+ (from GE Instruction manual for Bluespharose 6 Fast Flow).
For the initial purification step, the supernatant obtained from P.pastoris fermentation was diluted 1:2 using a 50 mM Tris pH 8 buffer (low salt buffer) and subsequently loaded on Fractogel EMD SO3– pre-equilibrated in low salt buffer. The elution was performed by applying a linear gradient from 0 to 2 M NaCl in 50 mM Tris pH 8 over 10 CV. The protein containing fractions were pooled and diluted 1:12 in low salt buffer for the second purification step. The diluted protein solution was loaded on Bluesepharose 6 Fast Flow pre-equilibrated in 50 mM Tris pH 8. The elution was carried out as described for the first step. Concentrating of the protein was performed by ultrafiltration using Amicon Ultra-15 (Ultracel-3k, Millipore, Billerica, MA, USA). Buffer exchange was performed by dialysis against PBS. To enhance the binding affinity of the HSA(C34A)-(Gly)4Ser-dnCCL2, a third purification step was added. For this step, the cation exchange resin SP Sepharose Fast Flow (GE Healthcare) was used with the same buffers as mentioned earlier. Elution, buffer exchange to PBS and concentrating were carried out as described earlier. The protein concentration was determined by UV280 measurement.
SDS–PAGE and Western blotting
The purity of the protein was analyzed by SDS–PAGE using 4–12% XT Criterion Precast gels (Biorad, Hemel Hempstead, UK) followed by Silver-staining, according to European Molecular Biology Laboratory. For western blot analysis, proteins were transferred via semi-dry blot (Biorad) onto Polyvinylidene fluoride membranes, blocked with 5% (w/v) non-fat dried skimmed milk powder in PBS for 1 h at room temperature (RT, 20°C). Incubation with primary and secondary antibodies was carried out at RT for 1 h. All proteins were detected using α-MCP1 antibody sc-1304 (SantaCruz Biotechnology, Dallas, TX, USA) diluted 1:200 and 1:10 000 anti-goat IgG/HRP antibody (Sigma-Aldrich) both diluted in dry milk. All proteins were visualized with the Immun star WesternC Kit (Biorad) and documented using the molecular imager Chemidoc XRS+ (Biorad) (Goger et al., 2002).
Circular dichroism
Circular dichroism (CD) measurements of 10 µM of the HSA(C34A)-(Gly)4Ser-dnCCL2, CCL2 and dnCCL2 in PBS were performed on a Jasco J-710 Spectropolarimeter (Easton, MD, USA). The CD spectra were collected between 190 and 250 nm with a response time of 1 s and a data point resolution of 0.2 nm using cuvettes with 0.1 cm path length. Five scans were averaged and background corrected in order to obtain smooth spectra. Mean residue ellipticities of the corrected spectra were calculated and plotted against the wavelength.
Size-exclusion chromatography
The size-exclusion chromatography (SEC) experiments were carried out on a Hitachi HPLC L-2100 system (Tokyo, Japan) equipped with an autosampler. A GE Superdex 75 PC 3.2/30 column (GE Healthcare) was used for separation. The flow rate of the separation buffer was 0.05 ml/min and it was composed of 136 mM sodium chloride, 8 mM sodium phosphate dibasic and 1.9 mM sodium phosphate monobasic in dH2O. The temperature of the column oven was set to 25°C and the detection was conducted at 214 nm. Prior to the measurement, each sample was diluted with separation buffer to a final concentration of 1 mg/ml and equilibrated for at least 30 min at 4°C before injection.
Guanidine hydrochloride-induced protein unfolding
Protein unfolding experiments were performed on a Jasco FP-Fluorometer FP6500 (Easton) coupled to an external water bath to ensure constant temperature during the measurements. Seven hundred nanomolar protein solutions in PBS containing different concentrations of guanidine hydrochloride (GdmCl) ultra-pure (MP Biomedicals, Solon, OH, USA) in the range of 0–6 M were prepared and equilibrated for 5 min at 20°C. All measurements were performed in triplicates and protein fluorescence emission spectra were recorded over the range of 300–400 nm following excitation at 280 nm. Slit widths were set to 5 nm for excitation and emission, the scan speed was 500 nm/min and the temperature was set to 20°C. After background (buffer) correction, the maximum emission intensity at a given GdmCl concentration was related to the intensity of the emission maximum of the native state (i.e. to the intensity at 336 or 334 nm, which corresponds to the emission maximum of each protein in 0 M GdmCl). This became necessary since quenching and subsequent de-quenching of the fluorescence emission was observed during GdmCl-induced unfolding experiments (see ‘Results and discussion’ section). The emission intensity quotient was then plotted against the GdmCl concentration and the sigmoid transition curves were fitted to the Boltzmann equation y = A2 + (A1 − A2)/1 + exp((x − x0)/dx)) of Origin 8.0 (OriginLab Corporation, Microcal Inc., Northampton, MA, USA), in which x0 represents the GdmCl transition midpoint concentration of unfolding, and dx represents a measure of cooperativity of unfolding (i.e. the GdmCl concentration range of unfolding).
Isothermal fluorescence titration
Isothermal fluorescence titration (IFT) measurements were carried out as described earlier (Gerlza et al., 2014) with the exception that GAGbody measurements were recorded with slit widths set at 3 nm for excitation and emission, and sensitivity was manually adjusted to 550 V. Titrations were performed with heparin and HS from Iduron, with additions between 50 and 1000 nM of ligand.
ELISA-like competition
Biotinylation of chemokines was performed as described recently (Gerlza et al., 2014).
ELISA-like competition protocol
A 2.5 µg GAG/250 nM biotinylated chemokine was diluted in PBS and coated onto Iduron GAG-binding plates over night at RT. Against the instructions of the manufacturer, we have coated the plates with GAG–chemokine complexes (instead of pre-coating with GAG only) in order to achieve maximum signal intensity. We assume that pre-coating with GAGs only would mask chemokine binding sites on the GAG due to interactions with the solid state surface thereby reducing the number of potentially binding biotinylated chemokines which are responsible for signal intensity. Similarly, the blocking step with HSA was left out. A washing step was performed to remove unbound biotinylated chemokine and GAG, followed by a 2 h incubation with different competitor concentrations diluted in PBS starting from 100 µM to 6 nM for decoy chemokines and 200 µM to 12 nM for wt chemokines, measuring each concentration thrice. To detect the remaining biotinylated chemokine we used an ELISA-like setup, therefore after another washing step we incubated the plates with high sensitivity Streptavidin HRP (Thermo Scientific, Waltham, MA, USA) diluted in 0.2% dry milk that binds to the non-displaced biotinylated chemokine on the plate. After another hour incubation at RT and removal of unbound Streptavidin by a washing step, we analyzed the plate by adding the substrate tetramethylbenzidine, resulting in a blue color change. After stopping the reaction with sulfuric acid, the absorbance at 450 nm was read in a Beckman Coulter DTX 800 Multimode Detector (Beckman Coulter, Austria). The reference (OD620) values were subtracted from the sample values (OD450), and the mean and standard deviation were calculated. Data analysis was performed using specialized statistical software Origin® (GE Healthcare).
Flow chamber
Laboratory reagents for neutrophil leukocyte isolation (BSA, Dextran, Histopaque) were purchased from Sigma-Aldrich. Human Tumor necrosis factor (TNF-α) was obtained from Peprotech (London, UK).
Culture of endothelial cells
Human lung microvascular endothelial cells (HMVEC-L) were purchased from Lonza (Basel, Switzerland) and were maintained in EGM-2 MV Bullet medium (Lonza, Basel, Switzerland) with 5% FCS. All culture surfaces were pre-coated with 1% gelatin for 1 h at 37°C to promote endothelial cell attachment and growth. The medium was changed every second day and cells were passaged upon reaching 90% confluence (5–6 days); the cultures were used between passages 5 and 9 in which the dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2-responsive proteoglycan co-receptors showed high expression in real-time PCR analyses (data not shown) (Konya et al., 2010, 2013).
Preparation of human peripheral blood monocytes
Blood was taken after informed consent from healthy non-atopic volunteers who were not taking any medication, according to a protocol approved by the Institutional Review Board of the Medical University of Graz. Peripheral blood leukocytes were obtained by dextran sedimentation of citrated whole blood. Preparations of peripheral blood mononuclear cells were prepared by Histopaque gradients, as described earlier (Sturm et al., 2005). Monocytes were further purified by negative magnetic selection using MACS Monocyte Isolation Kit II (Miltenyi Biotec, Bergisch Gladbach, Germany). Resulting purities and viabilities were >95%.
Monocyte capture to endothelial cells under flow conditions
Endothelial cells (4 × 105/substrate) were seeded on VenaEC biochips (Cellix, Dublin Ireland). Forty-eight hours later, the endothelial layers were stimulated with TNF-α (50 pM) or vehicle for 4 h. In some cases, endothelial cells were co-incubated with the indicated substances, i.e. various concentrations of dnCCL2 (0.01, 0.1, 1, 10 and 100 µg/ml) or HSA(C34A)-(Gly)4Ser-dnCCL2 (0.086, 0.86, 8.6, 86 and 860 µg/ml) in the last 1 h of the 4-h treatment with TNF-α. Monocytes were pre-conditioned in EBM-2 MV medium for 10 min at 37°C. If required, monocytes were incubated with dnCCL2 or HSA(C34A)-(Gly)4Ser-dnCCL2 in the same concentrations as used for endothelial cells for 10 min at 37°C. Endothelial monolayers were superfused with suspensions of 3 × 106/ml monocytes in EBM-2 MV medium at 0.5 dyne/cm2 for 5 min at 37°C in a OKOLAB H201-T1 heated cage. Monocyte capture to the endothelial layer was monitored by phase contrast on a Zeiss Axiovert 40 CFL microscope and Zeiss A-Plan 10×/0.25 Ph1 lens, using a Hamamatsu ORCA-03G digital camera and Cellix VenaFlux software. Computerized image analysis was performed by DucoCell analysis software (Cellix, Dublin, Ireland), where captured (i.e. slow rolling or in stationary adhesion) monocytes were quantified on each single image (Demyanets et al., 2011; Konya et al., 2011).
Pharmacokinetic profile
Animal experiments
Animal care and handling procedures including providing of food and water ad libitum, 12-h light and dark cycle and keeping in air-conditioned cages were performed in accordance with the European guidelines, and all the experiments were conducted under conditions previously approved by the local animal ethics committee.
Six to eight weeks old male C57BL/6 mice (Harlan, Indianapolis, IN, USA) were treated with an i.v. injection of dnCCL2 (200 µg/kg body weight) or HSA(C34A)-(Gly)4Ser-dnCCL2 (200 µg/kg body weight dnCCL2 equivalent) in the lateral tail vein. Group size was determined as n = 3. At defined time points blood was collected by heart puncture in deeply anesthetized mice. The serum concentration of dnCCL2 or HSA(C34A)-(Gly)4Ser-dnCCL2 was analyzed using human MCAF ELISA kit (Hölzel, Köln, Germany). ELISA setup was performed according to the manufacturer's protocol.
Statistical analysis
For the flow chamber investigation, all data are shown as mean + SEM for n observations. Statistical analyses were performed by SigmaPlot 12 software using one-way repeated measurement ANOVA with Bonferroni post hoc test. *P < 0.05 and ***P < 0.001 were considered as statistically significant. All treatments were performed in duplicates and experiments were repeated five to seven times. For all other investigations, data are shown as mean + SD. Statistical analyses were performed by GraphPad v5.04 using Student's t-test.
Results and discussion
CCL2 mutant design
We have established over the recent years a protein engineering approach that is based on generating dominant-negative GAG-binding therapeutic proteins. In the case of chemokines, we achieved this by genetically knocking out (or inactivating) the GPC receptor binding domain and by concomitantly increasing the GAG-binding affinity by replacing amino acids in the GAG-binding site by additional basic amino acids (Potzinger et al., 2006). In order to inhibit the GPC receptor activation of CCL2, we have replaced in earlier reported mutants Tyr13 by an alanine residue, which was further augmented by an N-terminal Met residue (process-related, due to E.coli expression). To increase the GAG-binding affinity, we have in addition replaced in these previous CCL2 mutants a serine by a lysine residue at position 21 (S21K) and a glutamine residue at position 23 by an arginine (Q23R) (Adage et al., 2012). Owing to their small molecular size, chemokines and their respective mutants commonly show a short serum half-life leading to low bio-availability, which is detrimental if such proteins are considered as biopharmaceuticals for chronic indications. This could be circumvented chemically by PEGylation or genetically by fusion to larger protein scaffolds like Fc fragments or human serum albumin (HSA). We have chosen HSA because of its natural affinity to chemokines which should give a stable fold of such fusion constructs.
Since we expected a possible drop in GAG affinity by a fusion to HSA—due to steric hindrance—we explored here further sites of CCL2 by exchanging wt residues against basic amino acids in order to maintain high GAG-binding affinity against the steric influence of HSA (see below). To do so, position 17—in close proximity to 21 and 23—as well as the distal sites 34 and 47 were genetically engineered to replace the natural amino acids by basic residues. The choice of sites for amino acid replacement to increase GAG-binding affinity followed two main rules: (i) to extend given and well-known GAG-binding site(s) and (ii) to impact as little as possible on the secondary structural core of the chemokine. Both engineering rules aimed at keeping the GAG selectivity of the wt protein within the mutant isoforms. The following CCL2 mutants were generated by site-directed mutagenesis and were investigated independently before fusing them to HSA: Met-CCL2(Y13A/N17K/S34K), Met-CCL2(Y13A/N17K/S21K/Q23K/V47K), Met-CCL2(Y13A/N17K/S21K/S34K), Met-CCL2(Y13A/N17K/S21K/Q23K/S34K) and Met-CCL2(Y13A/S21K/Q23K/S34K/V47K).
In order to assess class-selective GAG interactions, we investigated the affinities of the new mutants for HS, DS and heparin. In addition, we compared the new mutants with our previous Met-CCL2(Y13A/S21K/Q23R) mutant and wt CCL2. As can be seen from Fig. 1, all mutants except Met-CCL2(Y13A/N17K/S34K) showed higher GAG-binding affinities compared with wt CCL2. However, only Met-CCL2(Y13A/N17K/S21K/Q23K/S34K) exhibited the same selective GAG-binding profile as wt CCL2, namely highest affinity for HS followed by DS and heparin. HS and heparin are structurally closely related, but unlike heparin HS exhibits large un-sulfated regions (Rek et al., 2002). Since HS binds to the Y13A/N17K/S21K/Q23K/S34K mutant, like the wt CCL2, with higher affinity compared with the much higher overall charged heparin, this protein–GAG interaction is unlikely to be dominated by unspecific, electrostatic forces. In addition to charge, the conformation of the GAG-binding site might play a crucial role for the selectivity/affinity ratio of our mutants. We therefore assumed that the Y13A/N17K/S21K/Q23K/S34K mutant resembles closely the GAG-ligand selectivity of the wt chemokine. Consequently Met-CCL2(Y13A/N17K/S21K/Q23K/S34K) was selected for fusion to HSA.
IFT-derived dissociation constants (Kd values) of CCL2, CCL2(Y13A/S21K/Q23R), CCL2(Y13A/N17K/S34K), CCL2(Y13A/N17K/S21K/Q23K/V47K), CCL2(Y13A/N17K/S21K/S34K), CCL2(Y13A/N17K/S21K/Q23K/S34K) and CCL2(Y13A/S21K/Q23K/S34K/V47K) with respect to the three major GAG classes HS, DS and heparin. The data represent means ± SD of three independent measurements.
In the following sections, the term dnCCL2 (i.e. dominant-negative CCL2) will be used to abbreviate the full name of the mutant Met-CCL2(Y13A/N17K/S21K/Q23K/S34K).
Designing HSA-dnCCL2 mutant fusion proteins
Six HSA-dnCCL2 mutant fusion constructs were cloned and expressed in P.pastoris: HSA-dnCCL2, HSA-(Gly)4Ser-dnCCL2, HSA-[dnCCL2]2, dnCCL2-(Gly)4Ser-HSA, HSA(C34A)-(Gly)4Ser-dnCCL2 and HSA-[(Gly)4Ser]3-dnCCL2. The (Gly)4Ser peptide sequence is a highly suitable linker in protein fusion constructs as it provides sufficient spacing and conformational flexibility between the two protein moieties. In addition, this linker is highly unlikely to raise an immune response if considered to be used in human clinical trials. The (C34A) mutation in HSA was shown to prevent dimerization of the HSA moiety (McCurdy et al., 2004). The protein expression levels as well as the overall yields after purification differed for the six fusion constructs (9–200 mg/L and 27–98%, respectively; data not shown). In all cases, sufficient protein amounts were obtained with acceptable purity (∼95% according to SDS–PAGE and silver stain; data not shown) to be used for further in vitro characterization. Surface plasmon resonance and IFT affinity measurements revealed that constructs with a C-terminal HSA fused to the CCL2 mutant via a single (Gly)4Ser spacer gave highest GAG-binding affinity (data not shown). Therefore, the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant was chosen for a detailed structural and biological characterization. Interestingly, the affinity of the fusion constructs was not very different compared with the unfused CCL2 mutants and equally higher than for the wt chemokine. This means that the influence of the HSA-fusion tag on GAG-binding affinity was not as big as expected which could be attributed to the fact that the required exposure of the GAG-binding motif in the chimeric constructs is intact. We therefore investigated the potential structural consequences of chemokine–HSA fusion by molecular modeling and molecular dynamics computer simulations.
Modeling and molecular dynamics
As initial control of our fusion experiment, we wanted to investigate whether the artificially generated structure of the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant was able to exhibit an energetically stable fold. For this purpose, in silico molecular modeling studies were performed. First, the mutant (Gly)4Ser-dnCCL2 molecule was subjected to an extended 2 ns molecular dynamics simulation, and structures were collected after 0.13, 0.38, 0.75, 1.13, 1.5, 1.75 and after 2 ns. They were then individually and independently linked to the energetically and structurally relaxed HSA molecule via an overlapping Gly residue. Owing to the structural flexibility of the linker, elongated fusion constructs or more compact/tight structures were generated by this means. Subsequently, the loss of accessible surface area in the fusion constructs was calculated for each construct which was indicative for the increase in contact surface of the two protein sub-moieties. In combination with a lowest energy criterion, the maximal surface area was used to select a structurally most reasonable and stable 3D structure of the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion construct. In Table I, the structural and energetic parameters of this mutant protein are displayed. A surface model of the fusion construct indicating the accessible solvent area is shown in Fig. 2A.
Structural and energetic parameters of HSA(C34A)-(Gly)4Ser-dnCCL2
| HSA(C34A)-(Gly)4Ser-dnCCL2 | |
|---|---|
| Cα–Cα atoms that are closer than 5 Å | 14 contacts |
| Bumps | 62 contacts |
| Van der Waals contacts (atoms closer than the sum of their VdW radii + 0.5 Å) | 433 contacts (5417 kJ/mol) |
| Electrostatic interactions (pairs of Lys/His/Arg and Glu/Asp-residues whose interaction energy is lower than −5 kJ/mol) | 23 contacts (−240.32 kJ/mol) |
| Check hydrogen bondings between HSA and dnCCL2 | 838.85 kJ/mol |
| Atoms that are involved in π–π interactions | 4 interactions |
| Hydrophobic contacts (residues closer than 6 Å to dnCCL2) | 153 interactions (181.7 kJ/mol) |
| Total energy | −9431.10 kJ/mol |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | |
|---|---|
| Cα–Cα atoms that are closer than 5 Å | 14 contacts |
| Bumps | 62 contacts |
| Van der Waals contacts (atoms closer than the sum of their VdW radii + 0.5 Å) | 433 contacts (5417 kJ/mol) |
| Electrostatic interactions (pairs of Lys/His/Arg and Glu/Asp-residues whose interaction energy is lower than −5 kJ/mol) | 23 contacts (−240.32 kJ/mol) |
| Check hydrogen bondings between HSA and dnCCL2 | 838.85 kJ/mol |
| Atoms that are involved in π–π interactions | 4 interactions |
| Hydrophobic contacts (residues closer than 6 Å to dnCCL2) | 153 interactions (181.7 kJ/mol) |
| Total energy | −9431.10 kJ/mol |
Structural and energetic parameters of HSA(C34A)-(Gly)4Ser-dnCCL2
| HSA(C34A)-(Gly)4Ser-dnCCL2 | |
|---|---|
| Cα–Cα atoms that are closer than 5 Å | 14 contacts |
| Bumps | 62 contacts |
| Van der Waals contacts (atoms closer than the sum of their VdW radii + 0.5 Å) | 433 contacts (5417 kJ/mol) |
| Electrostatic interactions (pairs of Lys/His/Arg and Glu/Asp-residues whose interaction energy is lower than −5 kJ/mol) | 23 contacts (−240.32 kJ/mol) |
| Check hydrogen bondings between HSA and dnCCL2 | 838.85 kJ/mol |
| Atoms that are involved in π–π interactions | 4 interactions |
| Hydrophobic contacts (residues closer than 6 Å to dnCCL2) | 153 interactions (181.7 kJ/mol) |
| Total energy | −9431.10 kJ/mol |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | |
|---|---|
| Cα–Cα atoms that are closer than 5 Å | 14 contacts |
| Bumps | 62 contacts |
| Van der Waals contacts (atoms closer than the sum of their VdW radii + 0.5 Å) | 433 contacts (5417 kJ/mol) |
| Electrostatic interactions (pairs of Lys/His/Arg and Glu/Asp-residues whose interaction energy is lower than −5 kJ/mol) | 23 contacts (−240.32 kJ/mol) |
| Check hydrogen bondings between HSA and dnCCL2 | 838.85 kJ/mol |
| Atoms that are involved in π–π interactions | 4 interactions |
| Hydrophobic contacts (residues closer than 6 Å to dnCCL2) | 153 interactions (181.7 kJ/mol) |
| Total energy | −9431.10 kJ/mol |
(A) HSA(C34A)-(Gly)4Ser-dnCCL2 model showing the solvent accessible surface area of the HSA moiety in green. The surface of the dnCCL2 residues is colored dependent upon their charge: basic residues are shown in blue, neutral in cyan and acidic in red. (B) Model of the unfused dnCCL2 mutant.
In Fig. 2B, a surface representation of the unfused dnCCL2 mutant is shown. As can be seen, the GAG interaction surface, consisting of clustered basic amino acids (colored in blue, pointing out of the plane in Fig. 2B), exhibits a similar solvent-exposed orientation in the HSA-fusion construct as in the unfused dnCCL2 mutant. Therefore, a GAG-binding affinity of the chimeric protein similarly to the unfused chemokine mutant was expected.
GAG-binding affinity
In order to assess whether the fusion to HSA has changed the GAG-binding affinity of the dnCCL2 mutant, various GAG-binding affinity experiments were performed. Kd values of CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 to heparin and to HS were determined using IFT. By this method, similar Kd values in the range of 200 nM were detected for dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2, whereas wt CCL2 exhibited the expected significantly lower affinity for both HS (Kd = 1138 nM) and heparin (Kd = 2232 nM).
The bimolecular affinity between dnCCL2 and GAGs is apparently not significantly influenced by the HSA fusion. This is in accordance with the structural prediction of our molecular model (see Fig. 2A). However, it seems that although dnCCL2 is able to discriminate between heparin and HS, the fusion mutant is not able to differentiate between these two GAG ligands. This loss of differentiation capability may be due to protection of amino acids in the fusion construct which are responsible for specific hydrogen bonding and/or hydrophobic interactions with the glycan in the unfused protein (Fig. 3).
IFT-derived dissociation constants (Kd values) of CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 with respect to HS and heparin. Data are shown as mean ± SD, ***P < 0.001 was considered as statistically significant.
Selective chemokine displacement: ELISA-like competition
So far we have considered direct, i.e. bimolecular binding of GAGs to chemokine and chemokine mutants. We have recently established a method that enables a quick and reliable determination of IC50 values for a given GAG-binding protein in relation to other pre-bound proteins from surface-immobilized GAGs (Gerlza et al., 2014). We have called this method ELISA-like competition (ELICO) since the remaining protein in the reaction wells is quantified in an ELISA-like setup. For this purpose, the pre-bound protein needs to be biotinylated for which we have developed efficient and reproducible methods which guarantee that the biotinylated protein is still able to bind to GAGs with the same (or very similar) affinity as the unlabeled protein (Gerlza et al., 2014). In the current experiments, we have evaluated how efficiently our chemokine (fusion) mutants not only displace their corresponding wt protein but also other chemokines from HS. The rationale behind these experiments is that displacement of unrelated chemokines from a GAG ligand by a given chemokine (mutant) refers to unspecific GAG interactions of the labeled chemokines as well as of the chemokine mutants (see Fig. 4). As a potential drug, this could cause unwanted off-target effects. To evaluate the displacement pattern of our mutants, we have biotinylated several chemokines. From these, CCL2, CCL3, CCL5, CCL11, CXCL8, CXCL11 and CXCL12 were selected based mainly on conserved bio-equivalence (i.e. GAG binding and chemotaxis) after biotinylation compared with the unlabeled chemokines (data not shown). Since many chemokines changed their bioactivity significantly after biotinylation, the assessment of a larger chemokine panel was not possible at this stage.
ELICO-derived displacement profiles of seven biotinylated chemokines pre-incubated on HS by CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 (for further details see ‘Materials and methods’ section).
Regarding the displacement of wt CCL2 (Fig. 5), the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion construct gave an IC50 value of 2.3 µM. This is significantly better (2-fold) than CCL2 competing against CCL2, but not as good as the displacement capacity of dnCCL2 which gave an IC50 value of 82 nM (see Fig. 5). When specificity in the displacement pattern was considered by monitoring the displacement of chemokines other than CCL2, we observed that dnCCL2 displaced five more chemokines from HS (in addition to CCL2), whereas HSA(C34A)-(Gly)4Ser-dnCCL2 displaced only two more (namely CCL5 and CXCL8, see Fig. 4). This means that the HSA-fusion mutant is a much more selective competitor than the dnCCL2 mutant, and resembles more closely the displacement profile of CCL2. An explanation as to why dnCCL2 is a less selective competitor may be attributed to its larger accessible surface which enables more unspecific chemokine–GAG interactions than in the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant, where large parts of the chemokine moiety are protected by the HSA scaffold thus preventing unspecific GAG interactions and therefore unspecific displacements. In a non-competitive setup, we would assume wt CCL2 to exhibit highest selectivity in GAG recognition which cannot be evaluated in the current assay relating to the CCL2 isoforms as the mutants exhibit a much higher affinity and thus intrinsically lower IC50 values.
Enlarged ELICO-derived displacement profile of biotinylated CCL2 from HS by CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2. Data are shown as mean + SD, *P < 0.05 was considered as statistically significant.
Interestingly, both the dnCCL2 and the HSA(C34A)-(Gly)4Ser-dnCCL2 mutants displaced CCL5 and CXCL8 better (i.e. with lower IC50 values) than they displaced the CCL2 wt. This could be related either to the inability of both CCL2 mutants to primarily recognize the CCL2-specific GAG-binding motif, or to the unavailability of the cognate CCL2-binding motif in the GAG preparations used in the underlying study. Since the natural CCL2-specific GAG motif is unknown, we are currently not able to say whether our mutants are not targeting primarily the CCL2-specific GAG motif because this specific ligand is not available for any binding studies. We therefore plan for the future to isolate sufficient CCL2-specific GAG ligand(s) in order to tune our experiments towards chemokine-specific displacement.
Far-UV CD and unfolding
Next we investigated the (secondary) structure of the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion construct. For this purpose, we performed far-UV CD measurements of the fusion mutant in comparison with wt CCL2 and with the unfused mutant dnCCL2 (see Fig. 6). wt CCL2 was found to contain largely unstructured regions with reasonable low ß-sheet contributions. Interestingly, the amino acid replacements when mutating wt CCL2 to dnCCL2 led to a small but significant induction of secondary structure. The CD spectrum of HSA(C34A)-(Gly)4Ser-dnCCL2 clearly shows a well-folded mainly helical protein structure. The fusion to HSA apparently stabilizes the entire chimeric protein and can therefore be regarded as suitable scaffold to improve the structural characteristics of chemokines and chemokine mutants.
Far-UV CD spectrum (mean residue ellipticity, MRE) of CCL2, HSA(C34A)-(Gly)4Ser-dnCCL2 and dnCCL2.
To further demonstrate the stabilizing effect of the HSA fusion on the unfused mutant, GdmCl-induced fluorescence unfolding studies were performed (see Fig. 7). Compared with the wt chemokine, the unfused dnCCL2 mutant exhibited a significantly lower unfolding transition (UT) point. On the other hand, the HSA-fusion mutant HSA(C34A)-(Gly)4Ser-dnCCL2 showed the same UT point as the wt protein, i.e. at 4.5 mol/l GdmCl. In addition, the much stronger cooperativity (i.e. the narrower range of GdmCl concentration needed for unfolding) of the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant's unfolding curve compared with the wt and to the unfused dnCCL2 mutant refers to a structurally much more homogeneous protein. This is indicative of fewer structural isoforms of the HSA(C34A)-(Gly)4Ser-dnCCL2 mutant in solution. HSA itself showed a transition point at 3.2 M GdmCl with very strong cooperativity (i.e. very small dx; data not shown). Overall, this confirms the energetically favorable fold of the HSA-fusion construct as predicted the by molecular dynamics simulation. These data therefore suggest that the structure of the fusion mutant protein resembles a globular conformation as depicted in Fig. 2.
Guanidine-induced unfolding curves of CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2. Denaturation was followed by the (intrinsic) tryptophan fluorescence emission intensity change at a given GdmCl concentration relative to the intensity emission of the native state (native state). Data were analyzed using the Boltzmann equation (see ‘Materials and methods’ section), calculating the UT GdmCl midpoint concentration (x0) and the relative cooperativity, i.e. GdmCl concentration range of unfolding (dx).
Size-exclusion chromatography
The quaternary structure of chemokines and chemokine mutants plays an important role in chemokine function (Fernandez and Lolis, 2002). Many chemokines were found to exist in solution as dimers or in larger oligomeric structures. CCL2 for example was found mainly as a tetramer in solution (Lau et al., 2004). GAG binding induces further aggregation of chemokines which efficiently increases the local concentration of chemokines at the site of their secretion, thereby marking the hot spot of the chemotactic gradient (i.e. the site of highest concentration). We have found earlier that chemokine oligomerization negatively impacts GAG binding (Goger et al., 2002): in IFT binding studies oligomeric chemokines (i.e. chemokines at high concentrations) exhibited significantly lower affinities towards GAGs compared with monomeric or dimeric forms. We hence put forward a negative feedback model in which chemokines detach from GAG chains once they have oligomerized beyond a certain grade which correlates with high chemokine concentrations.
It was therefore important to see here, whether the introduced modifications had an impact on quaternary structure formation of the CCL2 mutant proteins. As can be seen in Fig. 8, wt CCL2 displayed a dominant peak which corresponds to the CCL2 tetramer (apparent molecular weight: 21.2 kDa) and a shoulder at longer retention times representing the dimeric form of the chemokine (apparent molecular weight: 15.1 kDa). In contrast to this, the unfused chemokine mutant dnCCL2 eluted from the size-exclusion column mainly as a dimer, no larger aggregates were observed. Finally, the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion mutant migrated on the SEC as monomer (apparent molecular weight: 75.6 kDa). This means that scaffolding the CCL2 mutant onto HSA led to an obligate monomer of the target chemokine. It is therefore not expected that the fusion mutant protein exhibits a concentration-dependent GAG-binding behavior which reflects the concentration-dependent oligomerization state of the protein. However, hetero-oligomerization between the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion mutant and wt CCL2 cannot be ruled out at this stage and will be the task of future studies.
SEC of CCL2, dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 (for experimental parameters, see ‘Materials and methods’ section).
Flow chamber
Stimulation of endothelial cells with TNF-α
To determine monocyte capture to pulmonary microvascular endothelial cells under physiological flow conditions, the endothelial cells were stimulated with vehicle or TNF-α. In the vehicle setup, four monocytes on average per high power field got captured spontaneously. In contrast, ∼130 monocytes adhered to endothelial monolayers stimulated with 50 pM TNF-α (see Fig. 9), indicating that TNF-α-induced monocyte capturing by endothelial cells very efficiently.
The effect of dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2 on monocyte capture to TNF-α-stimulated endothelial cells. Data are shown as means ± SEM of three independent measurements, *P < 0.05 and **P < 0.01 were considered as statistically significant.
Effect of dnCCL2 on TNF-α-induced monocyte capture to endothelial cells
In order to determine the effect of dnCCL2 on the capture of monocytes by TNF-α-stimulated endothelial cells, various concentrations of dnCCL2 (0.01, 0.1, 1.0, 10 and 100 µg/ml) were added to the endothelial cells during the last hour of the total 4-h TNF-α incubation time. Additionally, the freshly isolated human peripheral blood monocytes were also briefly pre-incubated with the same concentrations for dnCCL2 (see above) to avoid dilutions during the experiment. At 0.01, 0.1, 1.0 and 10 µg/ml concentrations, dnCCL2 induced a concentration-dependent, however, statistically not significant, reduction in monocyte capture to endothelial cells. At 100 µg/ml, dnCCL2 showed its maximal, statistically significant inhibitory effect (−44%, see Fig. 9A).
Effect of HSA(C34A)-(Gly)4Ser-dnCCL2 on TNF-α-induced monocyte capture to endothelial cells
The same setup was used for HSA(C34A)-(Gly)4Ser-dnCCL2 using equivalent concentrations (0.086, 0.86, 8.6, 86 and 860 µg/ml). Already at 0.856 µg/ml, HSA(C34A)-(Gly)4Ser-dnCCL2 reduced monocyte capture to endothelial cells significantly. Increasing the concentration of the HSA-fusion construct led to a further inhibition of monocyte adhesion which peaked at 86 µg/ml. Overall a plateau effect seems to be reach at concentration >8.6 µg/ml (see Fig. 9B). In comparison with dnCCL2, the HSA-fusion mutant was found to be bio-active at much lower concentrations in this TNF-α-induced monocyte capture set-up.
Pharmacokinetic profile
Finally, we assessed the pharmacokinetic profile of HSA(C34A)-(Gly)4Ser-dnCCL2 in C57BL/6 mice following single-dose intra venous application of the chimeric protein (see Table II). Analyses by ELISA allowed a comparison of detectable unfused dnCCL2 with the HSA(C34A)-(Gly)4Ser-dnCCL2 fusion construct in the murine serum. Both proteins reached similar maximal concentration in circulations (Cmax), however, dnCCL2 concentrations rapidly dropped following the intra venous administration. On the other hand, HSA(C34A)-(Gly)4Ser-dnCCL2 was detectable even after 72 h at biological relevant serum levels (data not shown). These results show that the fusion of the CCL2 mutant with HSA had the desired effect on prolonging exposure of the novel construct. In combination with the high and selective GAG-binding affinity, this novel construct could prove to be an efficient inhibitor of monocyte migration in clinical applications.
Pharmacokinetic parameters of dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2
| Compound | Cmax (ng/ml) | t1/2 (min) |
|---|---|---|
| dnCCL2 | 88.48 | 16.35 |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | 764.50 | 54 |
| Compound | Cmax (ng/ml) | t1/2 (min) |
|---|---|---|
| dnCCL2 | 88.48 | 16.35 |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | 764.50 | 54 |
Pharmacokinetic parameters of dnCCL2 and HSA(C34A)-(Gly)4Ser-dnCCL2
| Compound | Cmax (ng/ml) | t1/2 (min) |
|---|---|---|
| dnCCL2 | 88.48 | 16.35 |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | 764.50 | 54 |
| Compound | Cmax (ng/ml) | t1/2 (min) |
|---|---|---|
| dnCCL2 | 88.48 | 16.35 |
| HSA(C34A)-(Gly)4Ser-dnCCL2 | 764.50 | 54 |
Conclusion
Chemokines are small proteins which tend to oligomerize at higher concentrations and in the presence of GAGs. We have shown here that fusing an engineered chemokine, with higher GAG-binding affinity than the corresponding wt, to HSA does not necessarily lead to a loss in GAG-binding affinity but certainly improves its stability as well as its bio-availability in serum after i.v. injection. The HSA(C34A)-(Gly)4Ser-dnCCL2 fusion protein is therefore highly suited to be applied in indications where the CCL2–GAG interaction plays a major pathological role like in chronic inflammation and cancer. In the future, however, the selective displacement of the fusion mutant needs to be further improved since not only wt CCL2 is released from GAGs by the competition with HSA(C34A)-(Gly)4Ser-dnCCL2. For this purpose, the cognate GAG motif of CCL2 needs to be isolated and prepared in sufficient amounts for competition and binding studies.
Funding
This work was supported by the Austria Wirtschaftsservice (AWS) preseed project P1403518-PSL01.
References
Author notes
Edited by Thomas Kiefhaber
