Collagen-induced arthritis: severity and immune response attenuation using multivalent N-acetyl glucosamine



Rheumatoid arthritis is an autoimmunity leading to considerable impairment of quality of life. N-acetyl glucosamine (GlcNAc) has been described previously as a potent modulator of experimental arthritis in animal models and is used for osteoarthritis treatment in humans, praised for its lack of adverse effects. In this study we present a comprehensive immunological analysis of multivalent GlcNAc-terminated glycoconjugate (GC) application in the treatment of collagen-induced arthritis (CIA) and its clinical outcome. We used immunohistochemistry and FACS to describe conditions on the inflammation site. Systemic and clinical effects were evaluated by FACS, cytotoxicity assay, ELISA, cytometric bead array (CBA), RT–PCR and clinical scoring. We found reduced inflammatory infiltration, NKG2D expression on NK and suppression of T, B and antigen-presenting cells (APC) in the synovia. On the systemic level, GCs prevented the activation of monocyte- and B cell-derived APCs, the rise of TNF-α and IFN-γ levels, and subsequent type II collagen (CII)-specific IgG2a formation. Moreover, we detected an increase of anti-inflammatory IL-4 mRNA in the spleen. Similar to the synovia, the GCs caused a significant reduction of NKG2D-expressing NK cells in the spleen without influencing their lytic function. GCs effectively postponed the onset of arthritic symptoms, reduced their severity and in 18% (GN8P) and 31% (GN4C) of the cases completely prevented their appearance. Our data prove that GlcNAc glycoconjugates prevent the inflammatory response, involving proinflammatory cytokine rise, APC activation and NKG2D expression, leading to the attenuation of clinical symptoms. These results support the glycobiological approach to the treatment of collagen-induced arthritis/rheumatoid arthritis (CIA/RA) as a way of bringing new prospects for more effective therapeutic interventions.


Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by chronic joint inflammation resulting in subsequent cartilage and bone destruction [1]. Conventional therapeutic modalities used for the treatment of RA have significant side effects [2, 3], so there is a need to seek novel disease modulators. As the autoreactivity of immune cells is elicited in part by the dysregulated glycosylation, research during the past two decades has focused upon studying the molecular mechanisms underlying these changes and their role in autoimmune rheumatic diseases [4, 5].

N-acetyl-glucosamine (GlcNAc) is widely used as a supplement for treating osteoarthritis (its beneficial effects were proved in a placebo-controlled clinical trial [6]) and was reported to possess no adverse effects, even at high doses [7]. For RA, trials were performed with pure glucosamine only [8]. Suppressive effects of GlcNAc on experimental adjuvant arthritis in rats were reported by Hua et al. [9], and recently, Azuma et al. have described similar results in a laminarin-induced model of human RA in SKG/Jcl mice [3]. All the studies mentioned were performed with GlcNAc monosaccharide known to have low binding affinity to corresponding receptors. Multivalence thus plays an important functional role in carbohydrate–protein interactions [10]. Glycodendrimers with four (GN4C) and eight (GN8P) GlcNAc moieties on calix[4]arene and polyamidoamine scaffold, respectively, showing different three-dimensional (3D) structure and rigidity, were reported previously to modulate natural killer (NK) and B cell function [11-13], down-regulate the MGAT5 (mannosyl (alpha-1,6-)-glycoprotein beta-1,6-N-acetyl-glucosaminyltransferase) gene expression in synoviocytes of RA patients (GN8P) [14] and in fresh NK cells and the NK92 cell line (GN4C) [15]. We also found that GN8P influenced the NK cells isolated from RA-affected joints of the patients and that GlcNAc-carrying synthetic dendrimers have the ability to modulate the cytotoxic activity of NK cells [14] in a receptor repertoire-dependent manner; nevertheless, the exact mechanism remains unclear.

Collagen-induced arthritis (CIA) is an established experimental model of human RA induced by immunization of mice with bovine or chicken type II collagen (CII) emulsified in complete Freund's adjuvant (CFA) that results in rheumatic symptoms such as joint inflammation and swelling [16]. Unlike other models, such as collagen antibody-induced arthritis, that only elicits the effector functions leading to the rheumatic symptoms without the preceding immunopathology [17], the CIA model is suitable for studying the alterations of immune functions during disease development and progression.

The role of NK cells in RA and CIA is not fully understood, with evidence that they can both contribute to or protect against inflammation [18]. De Matos et al. showed a strong up-regulation of cytokine-producing NK cells (CD56bright) in the synovial fluid of RA patients supporting the ongoing inflammation [19]. Corresponding results were also presented by Dalbeth and Callan, who suggested that NK cells in the inflamed synovium interact with the macrophage/monocyte population, thus amplifying the production of inflammatory cytokines [20]. Conversely, Aramaki et al. reported low, impaired NK activity in human RA [21]. Lo et al. even reported that NK cells play a protective role in the development of CIA, which is mediated by interferon (IFN)-γ production suppressing the generation of Th17 cells [22]. Moreover, NK cells play different roles during different stages of the disease [23]. We were thus interested in the function of NK cells during the development of CIA and whether it can be modulated by the in-vivo administration of glycoconjugates that had previously been effective in alteration of NK cell function in rheumatoid arthritis in vitro [14]. The aforementioned effects of simple GlcNAc prompted us to examine the multivalent GlcNAc glycoconjugate efficacy, assuming their higher binding affinity to respective ligands on rheumatoid arthritis and its animal models.

The aim of our study was to evaluate the effect of GN8P and GN4C on CIA-bearing mice and their potential use as disease-modulating agents. First, we tested the GN8P treatment in two different administration schemes and focused on local reaction in arthritic joints by immunohistochemistry and FACS analysis of the synovial fluid with particular respect to NK, B and antigen-presenting cells. GN4C, a glycoconjugate with a more rigid structure, was added during the course of the study and compared with GN8P in terms of its systemic effect on antigen-presenting cells, NK and B cells and the clinical outcome. To evaluate the effect of the treatment, we employed clinical scoring [24], cytotoxicity assays, FACS, ELISA, real-time RT–PCR and cytometric bead array (CBA).

Materials and methods

Experimental animals

Ten-week-old male and female DBA/1 mice (Taconic, Lille Skensved, Denmark) were housed under specific pathogen-free conditions (22°C, 55% relative humidity, 12-h day/night cycle) and fed on NIH#31 rodent diet (Taconic) ad libitum. All procedures were conducted in accordance with the European Convention for the Care and Use of Laboratory Animals, as approved by the Czech Animal Care and Use Committee.

N-acetyl-D-glucosamine-coated dendrimers

Two structurally different glycodendrimers bearing four or eight GlcNAc moieties (Fig. 1a, GlcNAc monosaccharide) were used in this study. GN8P is a glycodendrimer based on a polyamidoamine scaffold bearing eight GlcNAc molecules (Fig. 1b). GN4C carries four GlcNAc molecules on calix[4]arene scaffold and is thus more rigid in its 3D structure (Fig. 1c). Both dendrimers were synthesized as described previously and kindly provided by Professor Lindhorst (CAU, Kiel, Germany) and Professor Kren (MBU, Prague, Czech Republic). The structure and purity was confirmed by electrospray ionization mass spectrometry and nuclear magnetic resonance [13, 25].

Figure 1.

Chemical structures of N-acetyl glucosamine (GlcNAc) and glycodendrimers. 2-(acetylamino)-2-deoxy-D-glucose or GlcNAc (a) is the simple monosaccharide derivative of glucose. GN8P (b) is a glycodendrimer bearing eight GlcNAc moieties on polyamidoamine (PAMAM) scaffold. GN4C (c) is a glycodendrimer bearing four GlcNAc moieties on calix[4]arene core providing a more rigid structure than PAMAM.

Induction of CIA

CIA was induced as described by Brand et al. [16]. The animals, fixed in a restrainer, were immunized by intradermal (i.d.) injection in the proximal third of the tail with 200 μg of bovine type II collagen (CII; Chondrex, Redmond, WA, USA) freshly dissolved in 0·1 M acetic acid and emulsified in equal volumes of complete Freund's adjuvant (CFA; Sigma-Aldrich, St Louis, MO, USA) fortified with 4 mg of Mycobacterium tuberculosis strain H27Ra per ml. The emulsion was prepared immediately before immunization by thorough mixing in a 0·5 ml syringe.

Glycoconjugate administration

Glycoconjugates (0·15 mg/kg) were administered intraperitoneally either in presymptomatic (days 10, 13, 16, 23 and 30) or symptomatic (days 23, 26, 29, 36 and 43) treatment schedules (Supporting information, Fig. S1). The concentration used was in accordance with the standard treatment dosage established and proved effective for immune modulation in our previous in-vivo studies [11-13]. Animals in the healthy control (HC) and untreated CIA (CIA) groups were injected with equal volumes of sterile phosphate-buffered saline (PBS) at the same time-points. The study comprised experimental groups of five to 10 animals in three to five independent experiments.

Histology and immunohistochemistry

Limbs from control and CIA mice were removed at euthanasia; joint tissue was micro-dissected, embedded in Jung tissue-freezing medium (Leica, Wetzlar, Germany) and frozen immediately in liquid nitrogen. Four-μm sections were fixed on lysine precoated slides in acetone and stained. Anti-CD11b-biotin and anti-NKG2D-biotin-conjugated primary antibodies (eBioscience, San Diego, CA, USA) were used. After 2-h incubation at 4°C the sections were washed and developed using a Vectastain ABC kit and diaminobenzidine (Vector, Burlingame, CA, USA), according to the manufacturer's protocol. Control slides for background and unspecific staining were prepared using normal rat serum instead of primary antibody, following the same procedures. Haematoxylin was used for counterstaining. Diagnosis was performed at ×10 , ×20 and ×40 magnifications (Zeiss transmission light microscope) by an expert pathologist. Microphotographs were obtained at ×40 magnification.

Isolation of spleen mononuclear cells (SMCs), lymph node cells and synovial fluid cells

Spleens were squeezed through nylon mesh and separated on Ficoll-Hypaque (Sigma Chemicals/Sigma Aldrich, St Louis, MO, USA) density gradient (1·086 g/ml: optimal density for murine leucocyte isolation). SMCs were washed three times in HEPES minimum essential medium (H-MEMd) (IMG, Prague, Czech Republic), resuspended in RPMI-1640 medium supplemented with 2 mM L-glutamine, 0·05 mg/ml gentamycin (IMG) and 5% heat-inactivated fetal calf serum (Biochrom, Berlin, Germany), and used immediately for assays. Lymph node cells were isolated in the same manner, omitting density gradient separation. Synovial fluid leucocytes from the arthritic paws were obtained after repeated intensive lavage of joints using H-MEMd, further washed and used for FACS analysis.

Flow cytometry (FACS)

Heparinized blood samples were seeded into U-bottomed 96-well microtitre plates (Nunc, Roskilde, Denmark) and erythrocytes were lysed using 0·15 M ammonium chloride buffer (22°C, 12 min). The cells were centrifuged (400 g for 2 min) and washed three times in ice-cold PBS containing 0·02% cold-water fish-skin gelatine and 0·01% sodium azide (Sigma-Aldrich). Cell suspensions prepared from spleens, lymph nodes and synovial fluid of individual mice were resuspended in the same PBS. The following monoclonal antibodies were used according to the manufacturer's protocol: CD45R/B220-Pacific Orange, CD3-phycoerythrin (PE), Nkp46-fluorescein isothiocyanate (FITC), CD11b-AlexaFluor 700, NKG2D-biotin, CD86-allophycocyanin (APC), CD11c-FITC, IA-IE-PE (BD Biosciences, San Jose, CA, USA or eBioscience, San Diego, CA, USA). We detected biotin by streptavidin-Qdot605 (Invitrogen, Grand Island, NY, USA). Samples were measured by BD LSRII flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) in a four-laser set-up (405 nm, 488 nm, 561 nm and 633 nm) and data were evaluated and offline-compensated based on single-stain controls in FlowJo version 9 (Tree Star, Ashland, OR, USA). Morphology and doublet exclusion was based on forward-scatter (FSC) area, FSC height and side-scatter (SSC) area; we used propidium iodide (BD Biosciences) for exclusion of non-viable cells.

Cytometric bead array (CBA)

Cytokine levels were measured by mouse Th1/Th2/Th17 CBA kit (BD Biosciences). Fluorescent beads of various sizes and fluorescence intensities coated with anti-cytokine antibodies were incubated with the serum samples, and a detection antibody conjugated with a different fluorochrome was used to detect the amount of the bound analyte. Serum samples were obtained by centrifugation of collected blood, processed according to the manufacturer's protocol, and measured on a BD LSRII. Data were evaluated in FlowJo version 9 and the median of fluorescence intensity was used to determine cytokine levels. Concentrations were calculated based on a standard curve.

Real-time RT–PCR

RNA was isolated from splenocytes with the RNeasy Mini Kit (Qiagen, Hilden, Germany) and transcribed into cDNA using a cDNA Archive Kit (Applied Biosystems, Foster City, CA, USA). Real-time RT–PCR was carried out with FastStart SYBR Green Master mix (Roche, Mannheim, Germany) in iCycler5 (Bio-Rad, Hercules, CA, USA). Interleukin (IL)-4 primer was designed in Primer3 software (forward primer: ACAGGAGAAGGGACGCCAT, reverse primer: TGAGCTCGTCTGTAGGGCTTC). Gene expression was normalized to β2-microglobulin (GeneriBio, Hradec Kralove, Czech Republic). Real-time RT–PCR data were evaluated in iQ5 softare (Bio-Rad).

ELISA for specific anti-collagen II (CII) IgG2a

Flat-bottomed, 96-well microtitre Maxisorp plates (Nunc) were coated with bovine CII (used for the induction of CIA) at 10 μg/ml in 0·1 M NaHCO3 (100 μl/well), incubated overnight at 4°C and blocked with 5% BSA–PBS (22°C, 2 h). Three dilutions (×10, ×100 and ×1000) of each sample were added in duplicate (100 μl/well). After overnight incubation (4°C) and extensive washing, anti-CII IgG2a was detected with horseradish peroxidase-conjugated goat anti-mouse IgG2a (Jackson, West Grove, PA, USA). Plates were developed using 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (KPL, Gaithersburg, MD, USA). Optical density was read at 450 nm by a Rainbow Thermo ELISA reader (Tecan, Salzburg, Austria). Sera from non-immunized mice (HC group) were used as controls.

Cytotoxicity assay

Cell-mediated cytotoxicity of SMC was estimated by the 51Cr-release assay, as described previously [26]. YAC-1 target cell line (mouse NK-sensitive T lymphoma) was maintained in RPMI-1640 medium supplemented with 2 mM L-glutamine, 1 mM sodium pyruvate, 0·05 mM 2-mercaptoethanol antibiotics (0·05 mg/ml gentamycin, 25 mg/ml amphotericin B) (IMG) and 10% heat-inactivated fetal calf serum (Biochrom). SMCs isolated from mice (effector cells, E) were pipetted into U-bottomed 96-well microtitre plates (Nunc). Then YAC-1 target cells (T) labelled with Na251CrO4 were added at an effector : target (E : T) ratio of 32:1 and incubated for 18 h (37°C, humidified atmosphere, 5% CO2) (Jouan, St Herblain, France). Cell-free supernatants were harvested (25 μl/sample) and mixed with 75 μl/sample of scintillation cocktail (SuperMix; Wallac, Turku, Finland). Radioactivity was measured using a Microbeta Trilux scintillation counter (Wallac). Effector cell lytic activity against target cells was calculated as follows: %cytotoxicity = 100 × (cpm) (specific lysis) − cpm (spontaneous lysis)]/[cpm (maximum lysis) − counts per minute (cpm (spontaneous lysis)].

Arthritis scores

Paws of individual mice were scored for clinical signs of arthritis three times a week based on standard scoring protocol: 0 = healthy/normal, 1 = erythema and swelling in digits, 2 = erythema and swelling in digits and metatarsal joints, 3 = erythema and swelling in digits, metatarsal and tarsal joints and 4 = erythema and severe swelling affecting the whole paw. Data are expressed as mean number of arthritic limbs per group, mean score per arthritic limb in the group, average symptom onset time and percentage of symptom-bearing animals [24].

Statistical analysis

The comparison of subgroups was based on various models of analysis of variance. Mouse was regarded as a random factor nested within the subgroup. A repeated-measure design was used to account for within-subject changes of score in time. Time to clinical onset was modelled by Cox's proportional hazards regression. Categorical data were described using absolute and relative frequencies and compared by Fisher's exact test. All statistical tests were treated as two-sided and evaluated at a significance level of 0·05. Results represent averages of multiple experiments and are expressed as mean ± standard deviation (s.d.). Statistical analysis was performed in Stata version 9·2 (StataCorp, College Station, TX, USA); graphs were produced by Prism 5 (GraphPad, San Diego, CA, USA) and Excel 2010 (Microsoft, Redmond, WA, USA).


Our previous research concentrated on the role of NK cells from different stages of human RA and their modulation by in-vitro-added GN8P glycoconjugate [14]. In the study presented herein, we were interested in the local reaction in rheumatic joints after in-vivo administration of GN8P in the animal model of autoimmune arthritis: CIA. We tested presymptomatic and symptomatic administration schemes and found that in the symptomatic schedule, the GC treatment had no substantial effect on either the clinical outcome or the followed immune parameters. We thus concentrated on the presymptomatic treatment starting at day 10, when the inflammatory response peaks, showing promising results presented thereafter. All data were first evaluated statistically with regard to the gender of the experimental animals. After we found no significant differences between males and females (represented 1:1) within the parameters of our interest, we processed the data obtained from both genders combined.

Attenuation of histological and immunological characteristics in CIA-affected joints by GN8P

Histological samples of the joint soft tissue demonstrated intensive inflammatory cellular infiltration under CIA conditions in comparison with healthy tissue, and more moderated infiltrate after GN8P treatment (Fig. 2a–c). The immunohistochemical staining of NKG2D and CD11b for the evaluation of specific cell types showed no cells expressing NKG2D receptor in the joints of healthy mice, whereas they were scarcely present in the CIA infiltrate (Fig. 3a,b). Cells expressing CD11b were rare in the healthy joint tissue, but abundant and confluent within the inflammatory infiltrate in untreated CIA mice (Fig. 4a,b). GN8P treatment did not cause marked changes in the NKG2D+ infiltration (Fig. 3c), while the CD11b-positive cells were still present in clusters but less massively than in the untreated CIA mice (Fig. 4c).

Figure 2.

Immune cell infiltration in the synovium. Representative images of the synovial infiltrate in hind paw joints of healthy controls (HC) (a), untreated collagen-induced arthritis (CIA) (b), and GN8P-treated CIA (c). Inflammatory infiltrate of the synovial and perisynovial tissue was stained by haematoxylin. Magnification ×40.

Figure 3.

Immunohistochemistry of NKG2D+ cell infiltration in the synovium. Representative images of the synovial infiltrate in hind paw joints of healthy controls (HC) (a), untreated collagen-induced arthritis (CIA) (b), and GN8P-treated CIA (c). NKG2D-positive cells in the synovial and perisynovial tissue were stained by anti-NKG2D antibody. NKG2D-positive cells are marked with arrows. Counterstained with haematoxylin. Magnification ×40.

Figure 4.

Immunohistochemistry of CD11b+ infiltration in the synovium. Representative images of the synovial infiltrate in hind paw joints of healthy controls (HC) (a), untreated collagen-induced arthritis (CIA) (b) and GN8P-treated CIA (c). CD11b-positive cells in the synovial and perisynovial tissue were stained by anti-CD11b antibody. Counterstained with haematoxylin. Magnification ×40.

In the synovial fluid, we detected a non-significantly lower proportion of NKG2D-positive cells after the GN8P treatment; however, NKG2D expression (MFI) on NK cells (CD3NKp46+) was decreased significantly (Fig. 5a,b). We found a slight reduction of CD11b+ cell infiltration, while the CD11b+CD11c+ antigen-presenting cell subset was markedly down-modulated (Fig. 5c). Following the adaptive immune response, GN8P suppressed the percentage of both total and CD86-positive B lymphocytes (Fig. 5d,e). Moreover, we detected a substantial decrease of total T cells and both cytotoxic (Tc) and helper (Th) subsets (Table 1).

Figure 5.

Cytometric (FACS) analysis of synovial fluid. NKG2D-expressing cells in the synovial fluid (a), NKG2D expression [mean fluorescence intensity (MFI)] on NK cells (mean ± standard deviation) (b), relative distribution of CD11b+CD11c+ antigen-presenting cells (APCs) (c), CD45R/B220+ B cells (d) and activated CD86+ antigen-presenting B cells (e) in the synovia were measured by flow cytometry (n = 5–10 mice per group, *P < 0·05; ***P < 0·001). Box-plots represent median, 25% and 75% percentiles, minimum and maximum.

Table 1. T cells distribution in the joints
  1. The table shows the relative levels of total T, cytotoxic (Tc) and helper (Th) cells in the synovia as measured by flow cytometry. Data are represented as mean ± standard deviation. CIA = collagen-induced arthritis.
T (CD3+NKp46) (%)3·6 ± 1·62·5 ± 0·7<0·05
Tc (CD3+NKp46/CD8+) (%)11·6 ± 2·16·1 ± 3·3<0·01
Th (CD3+NKp46/CD4+) (%)61·9 ± 5·952·9 ± 9·3<0·05

Cytokine serum levels, mRNA expression and anti-CII antibodies

The marked increase of tumour necrosis factor (TNF)-α caused by CIA was effectively prevented by the GN8P treatment (Fig. 6a). A similar observation was made with IFN-γ, where GN8P treatment kept the cytokine at the HC level (Fig. 6b). No significant changes were observed in the measured serum levels of IL-2, IL-4, IL-6 and IL-10. IL-17A showed a decreasing, although non-significant, trend after GN8P treatment (data not shown). As no significant changes of anti-inflammatory cytokines (IL-4, IL-10) were found in the sera, the mRNA expression was measured by real-time RT–PCR in SMC. While IL-10 expression was not altered by the GN8P treatment, IL-4 expression was markedly up-regulated (Fig. 6c).

Figure 6.

Serum levels of TNF-α and IFN-γ, mRNA expression of IL-4 in spleen mononuclear cells (SMC) and type II collagen (CII)-specific IgG2a level. The inflammatory cytokines TNF-α (a) and IFN-γ (b) were measured in sera of healthy and collagen-induced arthritis (CIA) mice by cytometric bead array. Relative fold expression of IL-4 mRNA (c) was measured in SMC by real-time RT–PCR and normalized to β2-microglobulin [healthy controls (HC) = 1)]. Levels of CII-specific IgG2a (d) were determined by ELISA (n = 5–10 mice per group, sample dilution ×100). Data are expressed as mean ± standard deviation. *P < 0·05; **P < 0·01; ***P < 0·001.

The Th1 (IFN-γ)-driven IgG2a level was measured by ELISA. The results depicted in Fig. 6d demonstrate a 14-fold increase of CII-specific antibody titre in CIA group relative to healthy controls. GN8P treatment prevented the rise of IgG2a by 7%.

Innate and adaptive immune responses to GCs

Due to the promising results obtained by GN8P administration to CIA mice, we further compared the effects of GN8P with another GlcNAc terminated conjugate on calix[4]arene scaffold – GN4C – providing better 3D structural control. We examined their effect on B cells, NK cells and antigen-presenting cells in the peripheral blood, spleen and lymph nodes and the effect on CIA clinical outcome.

We found a robust (fivefold) increase in total B cell percentage in the lymph nodes of CIA mice that was not changed by the GC treatment (Fig. 7a). Both GN8P and GN4C, however, prevented the activation of antigen-presenting (CD86+IA-IE+) B cells (Fig. 7b), as well as the generation of monocyte-derived (CD11b+CD11c+) antigen-presenting cells (Fig. 7c).

Figure 7.

Glycoconjugate-induced changes of relevant cell populations in the lymph nodes. Cells were isolated from the lymph nodes and analysed by flow cytometry (FACS). The relative distribution of CD45R/B220+ B cells (a), activated CD86+ antigen-presenting B cells (b) and myeloid CD11b+CD11c+ antigen-presenting cells (APCs) (c) are shown. Data represent results of three experiments performed (n = 5–10 mice per group) *P < 0·05; **P < 0·01; ***P < 0·001. Box -plots represent median, 25% and 75% percentiles, minimum and maximum.

NK cell-mediated cytotoxicity was measured at three time-points during CIA development and compared to HC. SMCs from CIA mice on day 10 (the peak of CII+CFA-induced inflammation) displayed significantly higher cytotoxicity. On day 37 it was comparable to controls, whereas on day 50 it was decreased by 40%. GC administration did not influence the NK cell function in either of the administration schedules (Fig. 8a).

Figure 8.

Function, distribution and phenotype of splenic NK cells. Gradual impairment of splenic NK cytotoxicity during collagen-induced arthritis (CIA) development expressed as a percentage of NK cell-mediated cytotoxicity of healthy mice stated as 100% (specific cytotoxicity = 46·8 ± 5·1%) and the effect of presymptomatic and symptomatic glycoconjugate (GC) treatment (a). Effect of CIA and GC treatment on the relative distribution of NK cells in the spleen (b) and the subset of NKG2D+ NK cells on day 37 (c). Results of three experiments performed (n = 5–10 mice per group) **P < 0·01; ***P < 0·001. Box-plots represent median, 25% and 75% percentiles, minimum and maximum.

With regard to the GCs' protective effect on NKG2D+ infiltration into the synovia, we also evaluated the relative distribution of NK cells along with the expression of NKG2D in the spleen, blood and lymph nodes. CIA caused a drop of splenic NK cell proportion on day 37 that was not modulated by the GC treatment (Fig. 8b); however, NKG2D-expressing NK cell levels, not altered by CIA itself, were decreased significantly by both GN8P and GN4C treatments by 11 and 15%, respectively (Fig. 8c).

In the peripheral blood, the NK cell levels were not altered; however, CIA caused a decrease of NKG2D-expressing subset (HC = 75·8 ± 3·0%; CIA = 65·2 ± 5·3%; P < 0·01), further slightly down-modulated by the GC treatment (CIA+GN8P: = 64·9 ± 5·9%; CIA+GN4C = 59·2 ± 7·6%). No changes in NK cell numbers and phenotype were found in the draining lymph nodes.

Amelioration of CIA clinical signs by the GC treatment

In our study, the CIA model evoked arthritic symptoms in 100% of untreated CII-immunized animals with the clinical onset on day 24·4 ± 2·5. The mean clinical onset of CIA was postponed significantly in GN4C-treated and non-significantly in the GN8P-treated mice (Fig. 9a). Concurrently, the GN4C and GN8P treatment caused a drop of incidence to 69 and 82%, respectively (Fig. 9b). The clinical symptoms and their severity, determined as mean score per affected limb, were attenuated significantly in GN4C-treated CIA mice (Fig. 9c). GN8P treatment resulted in a disease-moderating trend, reaching statistical significance in the final evaluated interval (day 32–37; P < 0·05). Examination of the percentage of affected limbs per group revealed that GN8P caused a decrease of 17%, while GN4C caused a drop of 35% to half the level of untreated CIA controls (65%) (Fig. 9d).

Figure 9.

Effects of glycoconjugate treatment on disease onset, incidence and clinical symptoms. Average time of onset of the collagen-induced arthritis (CIA) symptoms in untreated (CIA), GN4C and GN8P-treated groups (a). The percentage of ill animals in each group at the end of the experiment (day 37) (b). The progression of the disease severity was evaluated by clinical scoring (mean score per CIA-affected limb – c) and by the percentage of affected limbs per experimental group (d). Data represent results of three experiments performed (n = 5–10 mice per group), ***P < 0·001.


In this study we showed that the GlcNAc conjugates ameliorate the rheumatic symptoms of CIA, corresponding with the observations of Azuma et al. describing a similar effect of GlcNAc treatment in the SKG mouse model of RA [3]. GN8P prevented an inflammatory response in the arthritic joints by moderation of the immune cell infiltration and a significant reduction of both monocyte- and B cell-derived antigen-presenting T cells, and NKG2D-expressing NK cells in the synovial fluid. We found a GC-induced decrease of NKG2D+ NK subpopulation in the spleen and both antigen-presenting cell types in the lymph nodes. GN8P administration prevented the rise of anti-CII IgG2a, IFN-γ and TNF-α level and up-regulated IL-4 mRNA expression in SMC, supporting the anti-inflammatory action of GlcNAc-conjugates. The beneficial effect of glucosamine or N-acetyl-glucosamine on the disease progression and severity described previously in rodent arthritic models [3, 9] was thus supported by our results using multivalent GlcNAc-terminated conjugates.

The histological evaluation of the CIA-affected joints showed severe cell infiltration (comparable to the results of Joosten et al. [27]) and particularly the presence of NKG2D and CD11b-positive cells, reported previously to participate in bone erosion [28, 29]. The presymptomatic glycoconjugate treatment protected from the CD11b+ (notably the CD11b+CD11c+ antigen-presenting cells), B and T cell infiltration. Although the mechanism is unknown, Ma et al. demonstrated analogous suppression of dendritic cells (DC) and T lymphoblast activation in vitro by glucosamine [30]. These effects, together with the down-modulation of NKG2D expression by NK cells in the joints, could participate in the beneficial clinical outcome of the GC treatment, comparable to blocking of NKG2D receptor by antibodies demonstrated by others [31, 32].

In our presymptomatic treatment schedule, GN8P lowered the NKG2D+ NK cell percentage; however, GN4C was more effective, which corresponded with the attenuation of clinical symptoms. Unlike Andersson et al., who affected cells at the site of ongoing inflammation by anti-NKG2D treatment [32], we also managed to down-modulate NKG2D-positive cells at the systemic level. We consider these findings important with regard to the negative effect of NKG2D on autoimmunity development reported previously [31, 33]. Evaluating NK cytotoxicity, however, we also detected the suppression of NK effector function appearing during the disease progression, as reported by Aramaki et al. in human RA [21] or Lo et al. in CIA [22]. During the early arthritis phase the cytotoxicity was at the HC level, while in the late arthritis phase it showed a highly significant suppression. The lower relative distribution of NK cells in the spleen on day 37 in CIA-suffering mice also corresponded with this effect. A robust increase before the onset of clinical symptoms can be attributed to the stimulation by CFA [34]. With regard to the effect of our treatment, we incline to the concept that NK cells rather play a proinflammatory role in RA, as presented by de Matos et al., [19], Dalbeth et al. [20] or Pridgeon et al. [19, 20, 35], and their suppression is beneficial for the clinical outcome. The GC treatment leading to down-modulation of NKG2D on NK cells, which results in amelioration of the disease outcome, although in a manner yet unknown, merits further investigation.

As multi-competent lymphocytes with the ability to regulate innate and adaptive immune responses, NK cells have been shown to determine the outcome of B cell-mediated autoimmunity [36] and antigen-presenting cell activation [37]. B cells were also proved to be the key players in CIA development [38]. We saw a massive (fivefold) increase of B cells in the draining lymph nodes of CIA mice, confirming their activation and proliferation, corresponding with the subsequent CII-specific IgG2a antibody formation, partially prevented by the GN8P treatment. IgG2a is a Th1-driven isotype induced by IFN-γ [39], which is a prominent product of NK cells. Williams et al. [40] also described a correlation between antibody titres and disease severity in CIA, which corresponds with our results demonstrating the attenuation of the clinical symptoms, although the clinical improvement cannot be attributed to this phenomenon only.

In the synovia we observed a GN8P-caused decrease of total B cells and CD86-expressing B cell subset. The CD86-positive B cells are critical for B–T cell communication, and have an essential role in the activation of autoreactive T cells in RA [41]. Thus, the GC treatment, beside the reduction of antibody formation, inhibits the infiltration of autoreactive T cells into the synovia. This fact corresponds with the decreased numbers of T cells (particularly Tc) found in the joints.

Other important players in both human RA and animal models are the proinflammatory cytokines [42, 43]. TNF-α participates in the pathogenesis of RA as a paracrine stimulator of IL-1 and GM-CSF [44, 45] and drives the synovial inflammation leading to joint destruction [46]. GN8P prevented the CIA-evoked TNF-α production together with a drop of CD11b+CD11c+ synovia infiltration, resembling the favourable clinical effect caused by direct blockage of TNF-α [47]. Similar results were achieved by anti-TNF therapy of RA patients decreasing the CD86-expressing B cells [48]. IFN-γ contributes to arthritic inflammation by the activation of macrophages and IgG2a isotype switch [39, 49]. GC application, starting at the time of peak inflammatory reaction, when it can affect the processes leading to the disease establishment, can successfully prevent the IFN-γ production leading to partial suppression of the anti-CII IgG2a production and antigen-presenting cell activation.

It has been shown that IL-4 is a crucial anti-inflammatory mediator of CIA, correlating with tolerance induction. CIA can be suppressed by intravenous injection of CII; however, in DBA/1 IL4−/− mice, such treatment is completely ineffective [50]. IL-4 also potentiates CIA treatment by dexamethasone [51]. We thus examined the mRNA expression of IL-4 in splenocytes, although its serum level was not altered significantly. GN8P treatment caused a major increase in IL-4 mRNA expression, which supports the notion that it functions as an important suppressor of CIA development.

Grigorian et al. demonstrated a relationship between the inhibition of Th1 and Th17 cytokines and disease progression in experimental autoimmune encephalomyelitis (EAE) after oral administration of GlcNAc [52]. We assume a similar, although yet not fully unravelled, effect after the GlcNAc glycoconjugate treatment of CIA, orchestrating the complex interaction of both the innate and adaptive immune responses. We suggest that GCs influences directly, or via the cytokine network, antigen recognition and intercellular communication between antigen-presenting cells, NK and activated CD86+ B cells, resulting in down-modulation of Th1 and autoaggressive CD8+ T cells. Moreover, we assume a possible influence of IL-4 that would be effective in suppressing the autoimmune reaction, by induction of CD4+CD25+ regulatory T cells, as reported by Skapenko et al. [53, 54].

Taken together, the prophylactic GC treatment is capable of effective reduction of the incidence and postponement of the CIA onset as well as moderation of its severity. This is caused particularly by the prevention of inflammatory infiltration, inhibition of antigen presentation by both B cells and professional antigen-presenting cells and suppression of proinflammatory cytokine production. We suggest that inhibition of NKG2D-positive NK cells also plays an important role here, leading to the moderation of autoimmune processes. Our results prove that the glycobiological aspect should not be overlooked, as it may bring exciting new information and provide new prospects for more effective therapeutic interventions.


We thank Pavel Rossmann and Klara Klimesova for preparation of the histological samples and Katerina Fiserova for excellent technical assistance. This work was supported by Czech Science Foundation grants 310/06/0477 and 310/08/H077 and the Grant Agency of the Czech Ministry of Health grant NR/9106-3.


The authors declare no conflicts of interest.

Author contributions

J. R., K. C., V. H., I. D., L. V. and A. F. performed the experiments; J. R. and A. F. designed the study; M. M. performed the statistical evaluation; J. R., M. M. and A. F. wrote the manuscript.