Therapeutic Targeting of the G-CSF Receptor Reduces Neutrophil Trafficking and Joint Inflammation in Antibody-Mediated Inflammatory Arthritis

Neutrophils are essential cells in the innate immune system and are responsible for rapid responses to bacterial and fungal infections. Mature neutrophils contain an arsenal of antimicrobial agents within cytoplasmic granules, including proteinases, and carry the machinery to produce reactive oxygen and nitrogen species to facilitate destruction of pathogens for phagocytosis. However, these beneficial antimicrobial responses can be activated inappropriately in sterile inflammation, resulting in neutrophil-driven host tissue damage and the perpetuation of inflammation and amplification of autoimmunity. Neutrophils are therefore considered major contributors to numerous chronic inflammatory diseases, such as rheumatoid arthritis (RA), vasculitis, cystic fibrosis, chronic lung disease, and inflammatory bowel disease (1, 2).

RA is a chronic autoimmune disease that primarily targets synovial joints. The etiology of RA is multifactorial, and various cells contribute to the initiation and propagation of disease, including infiltrating inflammatory cells (neutrophils, T and B lymphocytes, dendritic cells, and macrophages) as well as activated stromal cells (synoviocytes, chondrocytes, bone cells, endothelial cells). Although MHC-mediated presentation of peptides to Ag-specific T cells and the generation of self-reactive autoantibodies are highly likely to be the initiating factors in RA (3), amplification steps may be critical in determining the expression and severity of disease. Neutrophils could be involved in each of these processes, both as initiators in very early disease (e.g., generation of autoantigens) (4), and/or in actively influencing the adaptive immune response (1, 5). Neutrophils comprise 70% of peripheral blood leukocytes in humans and up to 25% in mice (6). They are typically the most abundant cell type in inflamed RA synovial fluid (7, 8), and apheresis of leukocytes can provide clinical benefit in RA (9).

The majority (>98%) of neutrophils reside within the bone marrow (BM) (10). G-CSF binds specifically and exclusively to the G-CSF receptor (G-CSFR), which signals through the JAK1/STAT3 pathway (11). G-CSFR is predominantly expressed on neutrophils throughout all stages of maturation, but is also present on myeloid progenitors and endothelial cells (12, 13). G-CSF helps maintain homeostatic neutrophil levels and is widely used clinically for the treatment of neutropenia; however, mature neutrophil function (survival, adhesion, phagocytosis) is also regulated by G-CSF (2). The mobilization and trafficking of neutrophils are dependent on chemokines and homing molecules (integrins, adhesion molecules). Four subfamilies of chemokines can act on chemokine receptors that are differentially expressed on leukocytes, to direct cell trafficking from the bloodstream to extravascular tissues and inflammatory sites (14). The ELR⁺, but not the ELR⁻, CXC chemokines activate and attract neutrophils. G-CSF promotes neutrophil release from the BM into the circulation by enhancing the expression of CXCR2 ligands (KC, MIP-2) and/or by inhibiting the opposing SDF-1 and CXCR4 pathway (15, 16).

The specific nature of the G-CSF/G-CSFR interaction, and the limited number of cell types expressing the G-CSFR, makes this an appealing target for selective intervention in diseases in which neutrophils might play a crucial role in pathogenesis. Therapies aimed at blocking G-CSF–driven inflammatory diseases can target either the ligand or the receptor. G-CSF–deficient (G-CSF^−/−) mice were protected from collagen-induced arthritis (CIA) and acute monoarticular arthritis, and a neutralizing mAb to G-CSF alleviated established CIA in wild-type (WT) mice (17). G-CSF receptor–deficient (G-CSFR^−/−) mice were also protected from CIA (18); however, the effect of G-CSFR blockade on established disease has not been evaluated due to the lack of suitable Abs. Therefore, to further explore the potential value of therapeutic targeting of G-CSF in human diseases, we developed a mAb to mouse G-CSFR and evaluated its efficacy in a murine model of inflammatory polyarthritis—namely, type II collagen Ab-induced arthritis (CAbIA).

In this study, to our knowledge, we show for the first time the effects of therapeutic blockade of G-CSFR using this newly developed mAb. Anti–G-CSFR mAb blocked G-CSFR signaling and potently reduced established disease in a model of Ab-induced arthritis, with only a transient reduction in peripheral blood (PB) neutrophil levels. G-CSFR antagonism reduced neutrophil trafficking and the local expression of multiple inflammatory mediators/pathways without adversely impacting on viral clearance in a model of influenza infection. Our data suggest that G-CSFR signaling is essential for neutrophil homing to inflammatory sites and demonstrate the contribution of neutrophils to driving local inflammation, providing further support for the rationale of blocking G-CSF/G-CSFR in the treatment of inflammatory diseases, such as RA.

WT C57BL/6 (B6) mice and G-CSF^−/− mice (19) derived on the 129Sv × B6 background and backcrossed >20 generations onto B6 mice were used. All mice were males aged >7 wk. The CSL/Pfizer and Walter and Eliza Hall Institute of Medical Research Animal Ethics Committees approved all procedures and protocols with the exception of the influenza virus challenge experiments, which were approved by the University of Melbourne Animal Ethics Committee.

Mouse anti-type II collagen 5 clone mAb mixture kit was purchased from Chondrex (Redmond, WA) and used as described by the manufacturer. On day 0, mice were injected i.p. with 2 mg mAb mixture, followed by 50 μg LPS i.p. on day 3. Mice were monitored for up to 14 d for clinical signs of arthritis. A clinical score was assigned as follows: 0, normal; 0.5, swelling confined to digits; 1, mild paw swelling; 2, marked paw swelling; 3, severe paw swelling and/or ankylosis.

Mice were euthanized at various time points, and the paws were fixed in 10% (w/v) neutral buffered formalin, decalcified, and embedded in paraffin. Frontal tissue sections were stained with H&E and scored blinded to the treatment groups. The multiple joints of the front (carpal) and rear (tarsal) paws were scored collectively for three global features (exudate, presence of inflammatory cells within the joint space; synovitis, the degree of synovial membrane thickening and inflammatory cell infiltration; and tissue destruction, cartilage and bone erosion and invasion), each out of three (0, normal; 1, mild; 2, moderate; and 3, severe), and these were tallied for a total score out of nine.

Tissues, pooled from both knees of each individual mouse, were collected into DMEM plus 5% FCS on ice, washed, and digested (30 min, 37°C with 100 rpm agitation) in medium containing 1 mg/ml collagenase (CLS-1, 250 U/mg; Worthington Biochemical, Lakewood, NJ) and 0.10 mg/ml DNase I (type IV from bovine pancreas, 2100 kU/mg; Sigma-Aldrich). The digests were strained (70-μm cutoff), washed, and resuspended in PBS plus 2% FCS for cell counts. The cells were again pelleted, and the entire pellet was stained for flow cytometry, as detailed below.

Approximately 5 × 10⁴ cells in 0.2 ml 5–10% FCS were centrifuged (1000 rpm, 5 min) in a Cytospin 4 cytocentrifuge (Thermo Fisher Scientific), set on medium acceleration. The slides were dried, fixed in cold (−20°C) methanol for 6 min, dried again, and then stained with Giemsa (1:19 dilution with water; Sigma-Aldrich) for 30 min. Dried slides were mounted, and the relative abundance of neutrophils was evaluated under light microscope.

Abs directed against the recombinant extracellular domain of murine G-CSFR were isolated from a Dyax phagemid FAB-310 library (Dyax, Cambridge, MA). Candidate Fabs were screened by competitive phage ELISA, and high-affinity clones were reformatted as full-length IgG4 Abs with their respective κ or λ L chains (20). One clone, 5E2, was selected for affinity maturation, with the resultant binding affinities to murine G-CSFR assessed by surface plasmon resonance on a Biacore 4000 biosensor.

mAbs used for cytokine blockade were injected i.p. and included the following: anti-mouse G-CSFR (Ch5E2-VR81-mIgG1κ; CSL) or isotype control mAb (BM4, mouse IgG1κ; CSL); anti–G-CSF mAb (67604; R&D Systems, Minneapolis, MN); anti-TNF mAb (XT-22); or isotype control mAb (GL113, rat IgG1).

PB leukocyte numbers were determined on an ADVIA 120 automatic cell analyzer (Siemens, Deerfield, IL).

NFS-60 cells were maintained in RPMI 1640, 10% FCS, penicillin/streptomycin, GlutaMAX supplement (Life Technologies), and 2 ng/ml mIL-3 (R&D Systems) at 37°C under 5% CO₂. Cells were grown without cytokine for 18 h and then plated at 1 × 10⁴ cells/well in 96-well flat-bottom plates. Test Abs were added 30 min prior to addition of murine G-CSF (R&D Systems), and cells were incubated for 72 h at 37°C under 5% CO₂. ³[H]Thymidine (Perkin Elmer) was added for the final 6 h of incubation, and cells were harvested onto glass filter mats and air dried for 18 h before counting on a beta counter.

NFS-60 cells were grown without cytokine for 18 h and then plated at 2 × 10⁵ cells/well in 96-well round-bottom plates. Test Abs were added 30 min prior to addition of murine G-CSF, and cells were incubated for 20 min at 37°C, 5% CO₂. Cells were fixed in 500 μl 2% formaldehyde/PBS for 10 min at 37°C, washed (PBS), and then permeabilized in 500 μl 90% methanol for 30 min on ice. Cells were washed and then stained with PE-conjugated anti-pSTAT3 (pY705) Abs (612569; BD Biosciences) for 1 h at 22°C. Cells were analyzed on a BD Fortessa (BD Biosciences) using FlowJo software (Tree Star, Ashland, OR).

G-CSF levels in mouse sera were determined by sandwich ELISA using rat anti–G-CSF mAb (67604) as the capture Ab and biotinylated rabbit anti–G-CSF polyclonal Ab (R&D Systems) as the detection Ab, followed by addition of streptavidin-HRP using tetramethylbenzidine as substrate. Murine rG-CSF (PeproTech, Rocky Hill, NJ) was used as standard.

Single-cell suspensions of PB, BM, and joint digests were resuspended in PBS containing 2% FCS. Cells were stained with the following mAbs: FITC-conjugated anti-mouse Ly6G (1A8; BioLegend), eFluor450-conjugated anti-mouse CD11b (M1/70; eBioscience), allophycocyanin. Cy7-conjugated anti-mouse CD45 (30-F11; BD Pharmingen), PE-conjugated anti-mouse CXCR2 (242216; R&D Systems), allophycocyanin-conjugated anti-mouse CXCR4 (2B11; eBioscience), and PE.Cy7-conjugated anti-mouse CD62L (MEL-14; eBioscience). The 7-aminoactinomycin D (BD Pharmingen) was used to exclude dead cells, and live cells were analyzed on a BD Fortessa, using FlowJo software.

Joint washes and sera were evaluated for cytokine and chemokine levels using the Proteome Profiler Array kit (R&D Systems), which assays IL-1α, IL-1β, IL-6, IL-16, IL-23, IL-27, IL-1ra, G-CSF, M-CSF, KC, MCP-1, MCP-5, MIP-2, monokine induced by IFN-γ, RANTES, IFN-γ–inducible protein-10, BLC, TCA-2, SDF-1, triggering receptor expressed on myeloid cells 1 (TREM-1), sICAM-1, C5a, and TIMP. Chemiluminescent spots on the array membrane were visualized using ChemiDoc MP imaging system (Bio-Rad). Spot coordinates were then allocated individual zones, and the mean pixel intensity of each zone was determined using Image Lab software (Bio-Rad).

Arthritic mice were injected i.p. with luminol (200 mg/kg) and anesthetized (isoflurane inhalation) prior to bioluminescence imaging on an IVIS spectrum instrument (Caliper; exposure time 180 s; binning 4, field of view 12.5 cm) on specified days, based on previous studies (21, 22). Selected regions, termed regions of interest, were manually selected over front and rear paws, and identical regions of interest were used for time course analysis. Naive mice injected with luminol were used to control for background luminescence (∼400 average radiance). Living Image 3.1 software was used to quantitate bioluminescence as photons per second per square centimeter per selected region.

B6 mice with CAbIA were i.p. injected at day 5 with 50 μg anti–G-CSFR (n = 20 mice) or control mAb (n = 10 mice). At day 7, neutrophils (CD45⁺ CD11b^hi Ly6G⁺) were sorted (BD FACSAria) from pooled digests of the knee tissues, or PB, and the RNA was extracted from three such experiments, giving a total of 12 samples. RNA was analyzed using Illumina BeadArray microarray technology. Background correction, normalization, log transformation, and differential gene expression analysis were carried out in R using the limma (v3.18.11) Bioconductor package (23, 24). Manufacturer identifiers were mapped to accession numbers as per illuminaMousev2.db (v1.20.0) Bioconductor database, which contains Illumina MouseWG6v2 annotation, assembled from public repositories (25). All probes that were not at least 10% brighter than negative controls in at least three arrays were discarded. Probes for which no gene symbol was available were removed. Only the most highly expressed probe was kept when multiple gene expression measurements were present. Pathway network analyses were performed using MetaCore software (Thomson Reuters). Genes exclusively regulated by anti–G-CSFR mAb were visualized via a heat map from the gplots (v2.16.0) Bioconductor package. Data are provided in Gene Expression Omnibus (accession number GSE76966; http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE76966).

B6 mice were treated with four 100 μg doses of anti–G-CSFR mAb, or isotype control BM4, injected i.p. every 2 d starting 4 d before infection, as detailed below. The positive control for neutrophil depletion, anti-Ly6G (clone 1A8), required 500 μg i.p. and 200 μg administered intranasally (26), and rat IgG1 (each from the Walter and Eliza Hall Institute of Medical Research Antibody Facility) was used as its corresponding isotype control. The mice were infected with 100 PFU HK×31 (H3N2) influenza virus in 50 μl PBS delivered intranasally while under anesthesia and then monitored for 7 d before being bled for neutrophil determination and killed for lung tissue collection. Lungs were assessed for infectious virus by plaque formation in Madin–Darby canine kidney cells (27).

Statistical tests were performed using GraphPad Prism 6.0g for Mac OSX software. Analyses of differences between sample groups used either Student t test (two tailed), the Mann–Whitney U test (two tailed), or ANOVA with multiple comparisons tests, as indicated. Data shown are means ± SEM, unless otherwise stated. A p value <0.05 was considered statistically significant.

CAbIA could be induced with >80% incidence in B6 mice by i.p. injection of 2 mg anti-collagen mAb mixture, followed 3 d later with i.p. injection of 50 μg LPS. Clinical signs of disease began at day 4, peaked in severity between days 7 and 9, and then resolved (Fig. 1A, Supplemental Fig. 1A for schema). PB G-CSF levels rose during the course of the model and were maximal at the peak of disease (Fig. 1B), rapidly reducing as disease resolved. PB neutrophil levels increased concurrently with G-CSF and remained elevated (Fig. 1C). Histological assessment of joints at day 7 showed abundant neutrophils within the joint space exudate, often in close apposition with the articular cartilage surface (Fig. 1D, 1E, arrows). Cytospins of joint tissue digests confirmed the predominance of polymorphonuclear leukocytes (Fig. 1F). These data suggest a role for neutrophils and possibly G-CSF in the pathogenesis of CAbIA.

We previously showed that G-CSF was required for CIA (17). To test the requirement for G-CSF in CAbIA, we first used G-CSF^−/− mice. G-CSF^−/− mice were protected from the induction of CAbIA (incidence: WT, 5/5; G-CSF^−/−, 1/5), whereas disease susceptibility could be restored by i.p. injection of human rG-CSF (2.5 μg twice daily) on days 3–6 (incidence: 5/5)—corresponding to the rise in PB G-CSF levels and appearance of clinical disease in WT mice (Fig. 1G). G-CSF therefore has a critical role during the induction phase of CAbIA. To examine the role of G-CSF in established disease (day 5), arthritic WT mice were injected i.p. with varying doses of anti–G-CSF mAb on days 5, 7, 9 and 11. As little as 20 μg anti–G-CSF potently halted disease progression and reduced disease severity (Fig. 1H). TNF antagonists are currently used in the treatment of RA, but have proven ineffective in ∼40–50% of patients (28). Interestingly, anti–G-CSF was more potent than an optimum dose of anti-TNF mAb (200 μg) in reducing disease severity (Fig. 1I).

G-CSF binds specifically and exclusively to the G-CSFR (11), and the respective gene knockout mice have similar phenotypes (19, 29). There are no commercially available anti–G-CSFR mAbs that block the interaction of G-CSF with its receptor in mice. Therefore, to further examine the role of G-CSF in inflammatory arthritis, we developed mAbs to G-CSFR that block the binding of G-CSF. We screened a Fab phage-display library for Fabs targeting the murine G-CSFR. Following three rounds of selection, >3000 clones were screened for binding to murine G-CSFR using a phage ELISA. Screens identified 43 unique clones, which were then reformatted as whole IgGs, and the resulting Abs were screened for potency in a cell-based assay. Clone 5E2 was selected as the most potent anti–G-CSFR mAb, based on the dose-dependent inhibition of G-CSF–induced proliferation of NFS-60 cells (Fig. 2A, αGR [Par]).

To improve the potency of mAb 5E2, affinity maturation was undertaken using phage-display libraries covering all CDR regions except CDR L2. Following several rounds of panning with decreasing Ag concentration, enriched unique clones were screened for their G-CSFR–binding affinities (Supplemental Table I). Clone VR-81 showed the greatest improvement (40-fold) in binding affinity compared with the parental 5E2 Ab. The increased affinity of VR-81 translated to an improvement in potency, as shown by the inhibition of G-CSF–mediated cell proliferation (Fig. 2A, αGR [AM]), with a corresponding ability to block STAT3 phosphorylation (Fig. 2B). Clone VR-81 was subsequently reformatted as a chimeric Ab with a mouse IgG1 Fc domain (Ch5E2-VR81-mIgG1κ; hereafter referred to as anti–G-CSFR) and was used in all subsequent studies.

To examine the therapeutic effect of G-CSFR blockade, WT mice with established CAbIA were injected with different doses of anti–G-CSFR mAb on days 5, 7, and 10 (see Supplemental Fig. 1B, 1C for schemas). There was rapid and potent inhibition of joint inflammation, comparable in magnitude to that seen with anti–G-CSF mAb (compare Figs. 1H, 3A). Remarkably, only 10 μg anti–G-CSFR mAb was required for maximum protection from disease. PB neutrophilia was reduced by all anti–G-CSFR mAb doses (i.e., 10, 50, and 500 μg) at day 7, but returned to control levels by day 10 for the two lowest doses (Fig. 3D). PB neutrophil levels were equal to or greater than those of naive nondiseased littermate mice for all anti–G-CSFR mAb doses and time points examined during CAbIA. Anti–G-CSFR mAb (50 μg) had no effect on the percentage of BM neutrophils at day 7—the time of maximum effect on PB neutrophil levels (Fig. 3E).

In vivo bioluminescence imaging of luminol-injected mice with CAbIA was used to visualize neutrophil myeloperoxidase (MPO) activity, as previously described (21). MPO-positive neutrophils were present in the joints of control arthritic mice, and these were reduced in mice treated with anti–G-CSFR mAb (see Fig. 3F, 3G for day 10). Histological analysis of paws confirmed the clinical and in vivo imaging assessments (Fig. 3B, 3C). The joints of anti–G-CSFR mAb-treated mice showed significantly reduced histopathological features (exudate, synovitis, and tissue loss) compared with control arthritic mouse joints.

The MPO bioluminescence study and joint histopathology indicated fewer infiltrating neutrophils in the joints of mice treated with anti–G-CSFR mAb. To further evaluate this effect, arthritic joint tissues were removed at different times (days 7, 9, and 12) from mice treated with mAbs (sequentially at days 5, 7, and 10) and treated with mild enzymic digestion, and then single cells were examined by flow cytometry. Cells from joint digests of mAb control-treated arthritic mice were predominantly CD45⁺ leukocytes, and the majority of these were neutrophils (CD11b^hi, Ly6G⁺), with some monocyte/macrophage cells (CD11b^hi, Ly6G⁻) also present (Fig. 4A, upper panels, day 7 data shown). Within 48 h of i.p. injection of anti–G-CSFR (at day 7), there was a relative reduction of neutrophils and corresponding increase in monocyte/macrophage-like cells in joint cell digests, as determined both by flow cytometry (Fig. 4A, lower panels) and cytospins (Fig. 4B). Indeed at day 7, the joints of anti–G-CSFR–treated mice had markedly fewer total leukocytes and neutrophils than control joints (Fig. 4C, lower panels). By day 12, the number of joint leukocytes extracted from both treatment groups was reduced to background (i.e., naive control) levels.

To examine the effect of anti–G-CSFR treatment on neutrophil phenotype in CAbIA, flow cytometry was performed on BM, PB, and joint cell digestions. Markers of interest included the chemokine receptors CXCR2 and CXCR4, the integrin CD11b, and the adhesion molecule l-selectin (CD62L) (see Fig. 4D, Supplemental Fig. 2). Compared with control mice, neutrophils from all three sources had reduced CXCR2 and increased CD62L expression, consistent with a less activated phenotype. Joint neutrophils were unique in having elevated CXCR4 and CD11b expression. BM neutrophils differed from those of PB in having slightly reduced CXCR4 expression.

To further explore the effect of anti–G-CSFR treatment on joint inflammation, cytokine/chemokine protein arrays were performed on serum and joint tissue washes. KC and MCP-1 were significantly reduced in serum, and there were trends for reduced IL-1α, IL-16, TREM-1, sICAM-1, and complement component C5a (Fig. 5A). Interestingly, serum G-CSF levels rose with anti–G-CSFR treatment, most likely due to reduced receptor-mediated internalization and clearance.

In joint tissue washes, anti–G-CSFR mAb treatment resulted in a significant reduction in the proinflammatory cytokines IL-1β and IL-6, the chemokines KC (CXCL1, a ligand for CXCR2) and MCP-1 (CCL2), the monocyte/macrophage growth factor M-CSF, and the myeloid cell activation receptor TREM-1 (Fig. 5B). Trends were also seen for reduced SDF-1 (CXCL12, the ligand for CXCR4) and MIP-2 (CXCL2, another ligand for CXCR2).

Collectively, these data suggest that blockade of the G-CSFR results in reduced numbers of neutrophils at inflammatory sites and also within the circulation through antagonism of neutrophil chemotaxis, most notably, those involving CXCR2 and KC.

The effect of G-CSFR blockade on both circulating and local inflammatory neutrophils was further examined by differential gene expression. RNA was prepared from PB and joint neutrophils isolated from CAbIA mice treated therapeutically with either control or anti–G-CSFR mAbs. Genes were deemed to be differentially expressed if they had a log₂ fold change >1 and a false discovery rate <0.05. Differential gene expression analysis was performed on joint versus PB neutrophils, and this comparison was made for both control and anti–G-CSFR–treated mice (Fig. 6A). There were 227 genes that were exclusively upregulated in the joint neutrophils of control mAb-treated mice and 270 genes that were downregulated. Following anti–G-CSFR treatment, 572 genes were exclusively upregulated and 404 were downregulated in the joint neutrophils. The heat map of exclusively differentially expressed genes in these neutrophil populations suggests two distinct transcriptional profiles (Fig. 6A).

MetaCore software was used to identify the top (based on false discovery rate) 10 process networks in each data set (joint versus PB neutrophils of control and joint versus PB neutrophils of anti–G-CSFR–treated mice; Fig. 6B, 6C, respectively). Eight of the top 10 networks associated with the differentially expressed genes in joint neutrophils isolated from mice treated with control mAb were related to inflammation or the immune response (maroon bars), whereas only three networks relating to inflammation or the immune response were upregulated in joint neutrophils following anti–G-CSFR treatment. Notably, TREM-1 signaling, IL-6 signaling, neutrophil activation, and chemotaxis pathways featured prominently in the neutrophils of control mAb-treated joints, but these did not rank among the top 10 pathways in mice treated with anti–G-CSFR. This suggests that not only are there fewer neutrophils in the joints of anti–G-CSFR–treated mice, but those that do enter express a less inflammatory/immune-related profile.

As well as being key mediators in inflammatory processes, neutrophils are critically required for clearing infections, which is an important consideration in translating G-CSF antagonist therapies to the clinic. To determine whether G-CSFR blockade could have a detrimental effect on viral clearance, B6 mice were treated with four i.p. injections of 100 μg anti–G-CSFR or isotype control mAb BM4 every 2 d, starting 1 d before intranasal infection with HK×31 influenza virus (Fig. 7A). As a positive control for neutrophil depletion, separate mice were instead given anti-Ly6G (or rat IgG1 as isotype control). Anti-Ly6G reduced PB neutrophil counts to undetectable levels (Fig. 7D) and significantly delayed the clearance of HK×31, as seen by increased loss of body weight (Fig. 7B) and elevated lung viral load at day 7 (Fig. 7C). In contrast, anti–G-CSFR mAb treatment had no adverse effect on the day 7 pulmonary viral titres (Fig. 7C), and the mouse weights, like those of the isotype control mice, were starting to rise on day 7 (Fig. 7B). These data suggest that mice treated with anti–G-CSFR, at a dose that can reduce disease symptoms in the CAbIA model, could still mount a normal response to influenza infection without any signs of gross immunosuppression.

In this study, to our knowledge, we show for the first time the impact of therapeutic blockade of the G-CSFR on an inflammatory disease. Such an investigation has not previously been possible due to the lack of suitable Abs, and previous studies using G-CSFR^−/− mice do not provide the same information as these animals lack G-CSFR throughout the entire course of development and disease. The development of a mAb that binds to the G-CSFR and antagonizes ligand binding and signal transduction will enable more accurate modeling for subsequent clinical evaluation of this target in a range of human inflammatory diseases.

Our previous studies have identified a role for G-CSF in CIA (17, 18, 30), a model of experimental autoimmunity that is dependent on innate as well as adaptive immune responses (31). We showed that G-CSF signaling was essential for the innate immune response, and so this could also, at least partly, account for the effect of G-CSF blockade in CIA. We also showed normal CD4⁺ T cell and dendritic cell responses in G-CSF^−/− mice, but reduced anti-CII IgG production during CIA (17). The effect of G-CSF in that model is therefore likely to involve both the innate and adaptive immune responses. In contrast, CAbIA bypasses the need to develop anti-collagen Abs through the adaptive immune response, and thereby provides information on the clinically relevant effector phase of disease. CAbIA, like K/B×N serum-transfer arthritis, is independent of recipient T and B lymphocytes, but is dependent on innate immune cells, including neutrophils (32–35).

G-CSF^−/− mice were protected from CAbIA induction, and susceptibility could be restored by injection of G-CSF during the period when serum G-CSF levels were elevated in WT mice with CAbIA. G-CSF can also exacerbate suboptimal CIA when injected into WT mice in place of an Ag boost (30). The restoration of disease responsiveness in G-CSF^−/− mice shows that progenitor cells that have developed in the absence of G-CSF can generate a neutrophil-driven inflammatory response when supplied with G-CSF. This observation is important because there may be clinical settings (e.g., serious infection) in which rapid elevation of neutrophil levels is required in a patient receiving anti–G-CSF therapy. We found that therapeutic blockade of G-CSFR was effective in suppressing established disease and that G-CSFR signaling was essential for joint inflammation to persist and become chronic. The effect in halting the progression of disease was rapid and involved only a mild and transient reduction in circulating neutrophils, which remained above the levels of naive mice. Importantly, mice treated with anti–G-CSFR mAb could still mount a normal response to challenge with influenza virus, which is dependent on neutrophil involvement.

Approximately 40–50% RA patients do not respond to anti-TNF drugs (28), providing an opportunity for the introduction of new therapies. Inhibition of TNF only partially reduced disease in CAbIA and was inferior to G-CSF or G-CSFR blockade, as previously reported for CIA (17, 36). We also found that TNF-independent disease, revealed by using anti-TNF mAbs or TNF^−/− mice, was inhibited by anti–G-CSF mAb in CAbIA, as well as K/B×N serum-transfer arthritis (I.K. Campbell, unpublished observations). G-CSF and GM-CSF were originally identified as hemopoietic cell growth factors, but these molecules may have equally important roles as proinflammatory cytokines (reviewed in Refs. 37, 38). From this perspective, G-CSF and GM-CSF signaling pathways could provide appropriate targets for therapeutic intervention. The historical view of the role of the CSFs and the perception that interference could be highly detrimental are now being challenged. Indeed, recent phase 2 clinical trials targeting either GM-CSF (39) or GM-CSFR (40) in RA patients have proven to be remarkably effective, with minimal side effects. GM-CSF inhibition was also well tolerated in a phase 1b clinical trial in multiple sclerosis patients (41). The present study supports G-CSF blockade as another promising therapy for RA and other inflammatory disorders in which neutrophils play a prominent role. For example, recent studies have shown a requirement for G-CSF/G-CSFR in disease models of multiple sclerosis (experimental autoimmune encephalomyelitis) (42) and uveoretinitis (experimental autoimmune uveitis) (43).

Blockade of G-CSFR had a relatively specific effect on circulating neutrophil numbers with minimal effect on monocyte/macrophage numbers in the BM, blood, or inflammatory site. It would appear there was no downstream effect on other cell types, nor was there a general reduction in all myeloid cells that might be expected if mobilization of precursor cells was affected. BM neutrophil numbers were unaltered by G-CSFR blockade, and so a large reservoir of cells remains to be mobilized if required. Induction of CAbIA was associated with PB neutrophilia, which was reduced to naive levels by 500 μg anti–G-CSFR mAb—the highest dose tested. Surprisingly, even at this dose, the mice never became neutropenic, suggesting that other granulopoietic mediators (e.g., GM-CSF, IL-6, IL-3) were sufficient to maintain steady state neutrophil levels. We obtained similar findings with anti–G-CSFR blockade in nonhuman primates (K. Scalzo-Inguanti, K. Monaghan, K. Edwards, E. Herzog, D. Mirosa, M. Hardy, V. Sorto, H. Huynh, S. Rakar, H. Braley, N. Wilson, S. Busfield, A. Nash, and A. Andrews, manuscript in preparation). Importantly, lower mAb doses (10–50 μg) had only a transient and minor effect on PB neutrophil levels, while maintaining potent inhibition of disease throughout. In contrast, neutrophils were reduced in the joints of anti–G-CSFR (50 μg dose) mAb-treated mice compared with the untreated arthritic controls at all time points. These data imply that absolute PB neutrophil numbers may not directly relate to disease activity and that other G-CSF–dependent effects, such as neutrophil functional responses within local inflammatory microenvironments, may be more critical for sustaining disease.

A reproducible and persistent effect observed following G-CSFR blockade was the altered expression of chemokines and receptors involved in neutrophil homing. The chemoattractants, KC (CXCL1) and MCP-1 (CCL2), were both reduced in sera and in inflamed joints. KC is a ligand for CXCR2, which is homologous with the IL-8R type B in humans, and is rapidly (<45 min) increased in serum following administration of G-CSF (44). KC is produced by keratinocytes, monocytes/macrophages, endothelial cells, megakaryocytes, and osteoblasts, although not in direct response to G-CSF (15, 44). G-CSF stimulates CXCR2 expression on neutrophils in a STAT3-dependent manner (45). The absence of a local KC gradient and the reduced expression of CXCR2 on neutrophils together could account for the reduction of neutrophil trafficking into joints. Indeed, a small-molecule CXCR2/CXCR1 antagonist (SCH563705) ameliorated disease in mice with CAbIA and selectively reduced PB neutrophil levels in naive mice (46). MCP-1, in contrast, binds to CCR2 and is regarded as a chemokine for macrophages, although a recent study suggests neutrophils as well (47). In contrast to CXCR2/CXCR1 antagonism, a selective CCR2 antagonist (MK0812), which blocks the binding of MCP-1, had no effect on CAbIA (46). In a comprehensive study that used another model of Ab-induced arthritis (K/B×N serum-transfer arthritis), disease was attenuated in gene knockout mice lacking CXCR2, but not those deficient in other chemokine receptors (48). Using mixed BM chimeric mice, it was shown that CXCR2-expressing neutrophils were preferentially recruited to the arthritic joints over CXCR2-deficient cells. CXCR2 blockade was also effective in reducing disease in Ag-induced arthritis in rabbits and adjuvant arthritis in rats (49, 50). In our studies, MIP-2 (CXCL2), another ligand for CXCR2, also showed a trend for reduced joint expression, but serum levels were undetectable. Taken together, these observations suggest reduced levels of CXCR2 ligands and the lower expression of CXCR2 on neutrophils most likely account for reduced neutrophil trafficking into joints following anti–G-CSFR treatment.

The inflammatory cytokine profile of involved joints was also altered following G-CSFR blockade. Most notably, there were reduced levels of IL-1β and IL-6, which are both important mediators in RA. However, in the context of CAbIA, IL-1β but not IL-6 has been shown to be essential for disease (32). The reduction in local IL-6 production is therefore likely to be secondary and may simply reflect the general reduction in joint inflammation following anti–G-CSFR treatment. Interestingly, although per cell production is less, IL-1β is produced by neutrophils as well as macrophages. A large population of infiltrating neutrophils is likely to provide a major source of local IL-1β. In the K/B×N Ab-induced arthritis model, neutrophil-derived IL-1β was shown to be essential for disease (51). In that experimental model, immune complexes, through interaction with the FcγR on neutrophils, were thought to stimulate IL-1β production, which in turn enhanced chemokine secretion (including KC) by endothelial cells and synovial fibroblasts, thereby attracting more neutrophils (51). In our study, reduced IL-1β levels may therefore be a reflection of the lower numbers of neutrophils gaining access to joints (52). In a model of intraepidermal inflammation, local G-CSF production by keratinocytes mobilized neutrophils from the BM to the PB, whereas the secretion of CXCR2 ligands by these same cells was required to recruit the neutrophils from the PB to the inflammatory site (53). In the context of the inflamed joint, IL-1β can stimulate G-CSF production by resident joint cells (54, 55), which could result in mobilization of more neutrophils from the BM compartment. Anti–G-CSFR mAb may therefore inhibit this systemic feedback loop.

Whether or not neutrophils egress from the BM is dependent on their relative expression of CXCR2 and CXCR4 and the production of relevant ligands (15). A high CXCR4 to CXCR2 ratio serves to retain neutrophils within the BM. G-CSF downregulates CXCR4 expression on myeloid cells, promoting mobilization from the BM (16). Interestingly, we have observed elevated CXCR4 on neutrophils of the BM and PB as early as 3 h after anti–G-CSFR treatment of arthritic mice (I.K. Campbell, unpublished observations), whereas other markers remained unchanged. There was also a reduction in the percentage of neutrophils in the PB, but no change in the BM. This is consistent with previous reports (16) showing downregulation of CXCR4 by G-CSF, with G-CSFR blockade resulting in the retention of neutrophils within the BM. This effect appears to be relatively short-lived, as CXCR4 levels on neutrophils from the BM and PB returned to normal by 48 h; however, by this time, neutrophil CXCR2 expression was reduced in all three compartments (BM, PB, and joint), providing a second mechanism for reduced neutrophil trafficking from the PB.

Following extravasation into inflamed tissues, neutrophils do not appear to return to the circulation, but rather undergo apoptosis and are then phagocytosed by local tissue macrophages (56). This carefully controlled process limits any tissue damage that might arise from the lysis of necrotic neutrophils and the release of their destructive cargo (i.e., proteinases, reactive oxygen, and nitrogen species). The neutrophils remaining within the joints 48 h after G-CSFR blockade (i.e., day 7) had a phenotype distinct from those in the circulation or BM, having reduced CXCR2, but increased CD62L, CD11b, and CXCR4 expression compared with those from control mAb-treated mice. This surface marker expression is similar to one previously described as senescent and considered to be preapoptotic (57) and so might favor clearance by tissue macrophages. In further support of this, differential gene transcription profiling and MetaCore process pathway analysis showed that joint neutrophils from control mice displayed a distinct inflammatory phenotype, which was reduced in mice treated with anti–G-CSFR mAb. Pathways affected by anti–G-CSFR treatment included TREM-1 and IL-6 signaling, as well as neutrophil activation and chemotaxis, which are in general agreement with the cell surface marker expression and cytokine/chemokine protein array data. TREM-1 is a potent neutrophil and monocyte activator that stimulates cytokine/chemokine production, neutrophil degranulation, and phagocytosis (58). Soluble TREM-1 is secreted by neutrophils following their stimulation with LPS (59), which could account for its elevated levels in the serum and joint washes of mice in the CAbIA model, in which mice are injected with LPS. Soluble TREM-1 lacks a transmembrane domain and does not signal; consequently, it is thought to act as a competitive inhibitor for ligand binding to membrane-bound TREM-1, thereby providing a break on the inflammatory response. However, we found reduced levels of soluble TREM-1 in the serum and joint washes of anti–G-CSFR–treated mice, arguing against straightforward correlations.

In summary, this study has shown that blockade of G-CSFR can have prolonged effects on established inflammatory joint disease through modulation of neutrophil infiltration of the inflammatory site, and that this could be achieved without grossly affecting the response to infection. Our findings provide further support for utilizing this approach in the treatment of inflammatory diseases, such as RA. Individual chemokines and chemokine receptors are less appealing as targets for therapeutic intervention due to the considerable redundancy between receptors and ligands in this family. Blockade of a key regulatory cytokine, such as G-CSF, and its major cellular target, the neutrophil, may provide therapeutic benefit in a range of human inflammatory diseases. This study provides further rationale for such a strategy and reassurance that significant neutropenia is unlikely to be a limiting factor.

We thank Sandra Koernig (CSL Ltd.) for helpful discussion, and Laura McMillan (CSL Ltd.), Anne Walter (CSL Ltd.), Catherine Tarlinton (CSL Ltd.), Claire Morgan (Walter and Eliza Hall Institute of Medical Research), and Jane Murphy (Walter and Eliza Hall Institute of Medical Research) for technical assistance.

I.K.C., K.E.L., and I.P.W. are coinventors on a patent covering G-CSF antagonism in inflammation. I.K.C., D.L., K.M.E., V.R., M.N., N.J.W., K.S.-I., A.B.M., C.P., M.J.W., A.D.N., B.S.M., and A.E.A. are employed by CSL Ltd.; L.E.B. and C.M.-K. are contracted to CSL Ltd.; and I.P.W. is an advisor for CSL Ltd., which is currently developing G-CSF antagonists. The other author has no financial conflicts of interest.