Glycosylation of immunoglobulin G determines osteoclast differentiation and bone loss

Abstract

Immunglobulin G (IgG) sialylation represents a key checkpoint that determines the engagement of pro- or anti-inflammatory Fcγ receptors (FcγR) and the direction of the immune response. Whether IgG sialylation influences osteoclast differentiation and subsequently bone architecture has not been determined yet, but may represent an important link between immune activation and bone loss. Here we demonstrate that desialylated, but not sialylated, immune complexes enhance osteoclastogenesis in vitro and in vivo. Furthermore, we find that the Fc sialylation state of random IgG and specific IgG autoantibodies determines bone architecture in patients with rheumatoid arthritis. In accordance with these findings, mice treated with the sialic acid precursor N-acetylmannosamine (ManNAc), which results in increased IgG sialylation, are less susceptible to inflammatory bone loss. Taken together, our findings provide a novel mechanism by which immune responses influence the human skeleton and an innovative treatment approach to inhibit immune-mediated bone loss.

ARTICLE

Received 23 Dec 2014 | Accepted 10 Feb 2015 | Published 31 Mar 2015

DOI: 10.1038/ncomms7651 OPEN

Glycosylation of immunoglobulin G determines osteoclast differentiation and bone loss

Ulrike Harre1, Stefanie C. Lang1, Ren Pfeie1,2, Yoann Rombouts3,4, Sabine Frhbeir5, Khaled Amara6, Holger Bang7, Anja Lux8, Carolien A. Koeleman3,4, Wolfgang Baum1, Katharina Dietel1, Franziska Grhn5, Vivianne Malmstrm6, Lars Klareskog6, Gerhard Krnke1,2, Roland Kocijan1, Falk Nimmerjahn8, Ren EM Toes3, Martin Herrmann1, Hans Ulrich Scherer3 & Georg Schett1

Immunglobulin G (IgG) sialylation represents a key checkpoint that determines the engagement of pro- or anti-inammatory Fcg receptors (FcgR) and the direction of the immune response. Whether IgG sialylation inuences osteoclast differentiation and subsequently bone architecture has not been determined yet, but may represent an important link between immune activation and bone loss. Here we demonstrate that desialylated, but not sialylated, immune complexes enhance osteoclastogenesis in vitro and in vivo. Furthermore, we nd that the Fc sialylation state of random IgG and specic IgG autoantibodies determines bone architecture in patients with rheumatoid arthritis. In accordance with these ndings, mice treated with the sialic acid precursor N-acetylmannosamine (ManNAc), which results in increased IgG sialylation, are less susceptible to inammatory bone loss. Taken together, our ndings provide a novel mechanism by which immune responses inuence the human skeleton and an innovative treatment approach to inhibit immune-mediated bone loss.

1 Department of Internal Medicine 3, University of Erlangen-Nuremberg, Erlangen 91054, Germany. 2 Nikolaus Fiebiger Center of Molecular Medicine, University of Erlangen- Nuremberg, Erlangen 91054, Germany. 3 Department of Rheumatology, Leiden University Medical Center, Leiden 2300 RC, The Netherlands. 4 Center of Proteomics and Metabolomics, Leiden University Medical Center, Leiden 2300 RC, The Netherlands. 5 Department of Chemistry and Pharmacy, Interdisciplinary Center for Molecular Materials (ICMM), University of Erlangen-Nuremberg, Erlangen 91054, Germany. 6 Rheumatology Unit, Department of Medicine, Karolinska Institute and Karolinska University Hospital, Solna 17177, Sweden. 7 Orgentec Diagnostika, Mainz 55129, Germany.

8 Department of Genetics, University of Erlangen-Nuremberg, Erlangen 91054, Germany. Correspondence and requests for materials should be addressed to G.S. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651

IgG is a potent effector protein of the humoral immune response. Besides its role in antigen binding, IgG has been well established to regulate the activity of immune cells

by binding to activating or inhibitory FcgR, mediating key effector functions of IgG such as the clearance of pathogens or antibody-dependent cellular cytotoxicity. The binding of IgG to FcgR is inuenced by the glycan attached to asparagine-297 in the Fc part of IgG1. This glycan is composed of a conserved heptamer backbone that consists of N-acetylglucosamine and mannose residues and is modied by the addition of several sugar moieties (Supplementary Fig. 1). Changes in IgG glycosylation are considered to result in an altered Fc conformation and affect the binding afnity of IgG to FcgR2,3.

Hence, attachment of terminal sialic acid residues on IgG appears to mediate the anti-inammatory effects of intravenous IgG4,5. Conversely, the absence of terminal sialic acid residues increases the afnity to activating FcgR6,7. The exact mechanism, by which these sugar moieties affect biological functions, is, however, not fully claried8. Some911, but not all12, studies have suggested that anti-inammatory effects of sialylated IgG are attributed to an enhanced binding of IgG to anti-inammatory lectin receptors, such as the dendritic cell-specic intercellular adhesion molecule-3-grabbing non-integrin9 or the dendritic cell immunoreceptor10 according to changes in the Fc conformation.

Bone homeostasis is assured by a concerted action of bone forming osteoblasts and bone resorbing osteoclasts. This balance can be impaired by the immune system as it is observed in bone destructive autoimmune diseases such as rheumatoid arthritis. However, while cytokine-mediated effects have been extensively studied in the past13,14, there has been limited research regarding the impact of IgG or immune complexes on bone. Osteoclasts originate from the monocyte/macrophage lineage. Apart from specic growth factors and cytokines, such as the macrophage colony stimulating factor (M-CSF) and the receptor activator of NF-kB ligand (RANKL), osteoclastogenesis requires costimulatory signals from the immunoreceptor tyrosine-based activation motif-harbouring proteins, FcR common g chain (FcRg) and DNAX-activating protein (DAP)12 (refs 15,16). In addition, osteoclasts express FcgR in comparable amounts to macrophages and dendritic cells17,18, suggesting that they may be affected by immune complexes. This notion is supported by a study showing that crosslinking of FcgR on murine preosteoclasts leads to higher osteoclast numbers19 and our previous nding that antibodies against citrullinated proteins (ACPA) directly promote osteoclastogenesis20. As pathogenic antibodies such as ACPA have been reported to contain less sialic acid in their Fc glycan than whole-blood IgG21, we further hypothesized that IgG sialylation may have an impact on immune complexosteoclast interactions.

In this study, we used heat-aggregated IgG as an immune complex mimetic without antigen restrictions22 to investigate the effects of IgG sialylation on immune complexosteoclast interactions. We found IgG sialylation to be a main regulator for the pro-osteoclastogenic potential of immune complexes as only non-sialylated immune complexes stimulated osteoclastogenesis in vitro as well as in vivo. In addition, we found higher Fc sialylation levels of both random IgG as well as specic autoantibodies to be correlated with higher bone volume in patients with RA. Administration of the sialic acid precursor ManNAc resulted not only in elevated sialylation levels of IgG, but also in a decreased susceptibility to collagen-induced arthritis (CIA)-mediated bone erosion in DBA/1J mice, suggesting a protective role of sialylated IgG against autoimmune-mediated bone loss.

ResultsDesialylated immune complexes stimulate osteoclastogenesis. To investigate the effects of immune complexes on human osteoclast formation, we differentiated human monocytes in the presence of M-CSF, RANKL and TGF-b to the stage of mono- and bi-nucleated preosteoclasts and challenged them with 100 mg ml 1 of immune complex mimetics derived by heat aggregation of pooled human IgG. One fraction of the

IgG had been enzymatically desialylated before aggregation (Supplementary Fig. 2) to determine the importance of IgG sialylation for immune complex-mediated effects on preosteoclasts. In addition, we used immune complexes that had been deglycosylated, and therefore lost their FcgR binding capacity, as a control for unspecic effects by the amount of protein added to the osteoclast culture. After 24 h of stimulation, staining for tartrate-resistant acidic phosphatase (TRAP) was performed and osteoclasts were counted.

The stimulation with desialylated immune complexes (IC-ds) resulted in an accelerated osteoclastogenesis with an increase in osteoclast number, nuclei per osteoclast and bone resorption (Fig. 1). In contrast, after stimulation with untreated (sialylated) immune complexes (IC), we observed only a very weak tendency to higher osteoclast numbers and no difference in bone resorption. Neither the deglycosylated immune complexes (IC-dg) nor monomeric IgGs showed an effect, indicating a specic interaction between non-sialylated immune complexes and FcgR on preosteoclasts. This pro-osteoclastogenic effect of low-sialylated IgG complexes was conrmed with natural immune complexes consisting of low- or high-sialylated anti-TNP IgG and TNP-26-BSA (Supplementary Fig. 3).

Analyses by dynamic light scattering revealed similar particle sizes of native and desialylated aggregated IgG (Supplementary Fig. 4) excluding size differences as a reason for the strong osteoclastogenic effect observed with the desialylated immune complexes. In addition, all IgG samples had been veried to be free of endotoxin to avoid endotoxin-mediated side effects.

Immune complexes bind to FccRII and III on preosteoclasts. Next, we investigated how the distinct FcgR are regulated during the process of human osteoclastogenesis. Therefore, we isolated mRNA from osteoclast cultures every second day from the stage of monocytes (baseline, day 0) to the stage of mature osteoclasts (day 10). We observed an early upregulation of FcgRI and III as well as of the two immunoreceptor tyrosine-based activation motif-domain-harbouring accessory proteins FcRy and DAP12 (Fig. 2a). In contrast, the inhibitory FcgRIIB seemed to be mainly expressed by mature osteoclasts. The expression of FcgRIIA halved during the rst 2 days of culture but remained stable for the rest of the culture period. To further test which FcgRs are expressed on the protein level at the preosteoclast stage (day 68), we performed immunouorescence stainings on preosteoclasts. Conrming the mRNA data, all FcgRs could be detected (Fig. 2b), though we could not discriminate between FcgRIIA and FcgRIIB as the antibody against FcgRII recognizes both. Interestingly, smaller preosteoclasts exhibited the brightest expression signal, whereas larger, more mature cells that started to form podosomes seemed to have a less pronounced FcgR expression.

To determine which FcgR is responsible for the observed proosteoclastogenic effect of desialylated immune complexes, we added blocking antibodies against FcgRI, FcgRII and FcgRIII during the stimulation of preosteoclasts with desialylated immune complexes. Blocking of either FcgRII or FcgRIII resulted in a virtually complete abrogation of the pro-osteoclastogenic activity of desialylated immune complexes (Fig. 2c), indicating that a concerted interaction of those FcgRs is needed to provide the

2 NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651 ARTICLE

6 **

** P=0.007 P=0.09

P=0.007

P=0.059

Fold change

osteoclast number

5 4 3 2 1 0

Untreated

Untreated

Untreated

Fold change nuclei per osteoclast

2.01.51.00.5 0

2.5 2.0 **

P=0.0075

1.5

1.0

0.5

0

Fold change

resorbed area

IC

IC-ds

IC-dg

IgG

IgG-ds

IgG-dg Untreated Untreated

Untreated

IC IC-ds IC

IC

IC-dg IgG IgG-ds IgG-dg

IC-ds

IC

IC-ds

IC-dg

IC-ds

IgG

IgG-ds

IgG-dg

Figure 1 | Fc-glycosylation determines the osteoclastogenic effects of human immune complexes. (a) Fold change of osteoclast number and (b) nuclei per osteoclast after 24 h treatment of human preosteoclasts with 100 mg ml 1 of native, desialylated (ds) or deglycosylated (dg) monomeric (IgG) or complexed (IC) pooled human IgG. TRAP-positive cells with Z3 nuclei were considered as osteoclasts. Bars show means.e.m. of ve independent experiments. (c) Representative micro images. Scale bar, 100 mm. (d) Fold change of the resorbed area after 24 h treatment of human preosteoclasts with 100 mg ml 1 of native or desialylated (ds) IgG complexes (IC) in calcium phosphate-coated wells. Bars show means.e.m. of ve independent experiments (e) Representative micro images. Scale bar, 500 mm. Statistical analysis was performed with MannWhitney U-test. **Po0.01.

signal for enhanced osteoclast development. In contrast, blocking of FcgRI had no effect, conrming previous reports about the dispensability of the high-afnity FcgR in osteoclast development and arthritis19,23.

Desialylated immune complexes affect osteoclasts in vivo. We next wanted to investigate the implications of desialylated immune complexes on osteoclastogenesis in vivo. In patients with rheumatoid arthritis, immune complexes have been found in the synovial uid of the joint space2426. To simulate this situation we decided to inject murine-aggregated IgG into the knee joints of C57BL/6 mice. Again, a part of the IgG had been enzymatically desialylated before heat aggregation. The control group was injected with PBS. The injections were repeated after 4 days. After additional 3 days, bones were dissected and analysed for the number and size of osteoclasts in the trabecular bone adjacent to the injection site.

Mice that were injected with 5 mg of desialylated immune complexes (IC-ds) displayed a signicantly higher osteoclast number and surface compared with PBS-injected mice (Fig. 3ac). The injection of 5 mg of untreated (sialylated)

immune complexes (IC) resulted in a mild increase in osteoclast number and surface, but the effect was much less pronounced compared with desialylated immune complexes, conrming our in vitro data. Notably, this difference vanished after co-injection of 100 ng of TNFa. In this case, both aggregates were able to induce osteoclastogenesis already at an amount of 1 mg.

Haematoxylin-eosin staining revealed no differences in the degree of inammation between the groups with only localized, if

any, inammation around the small injury of the injection site (Fig. 3d,e). Hence, the observed pro-osteoclastogenic effect of desialylated immune complexes was independent of inamma-tory tissue inltration.

IgG glycosylation and bone loss in arthritis patients. To address to which degree IgG Fc sialylation affects human bone architecture, we measured the amount of Fc sialylation, galactosylation and fucosylation of total IgG and disease-specic autoantibodies (ACPA) in 30 patients with rheumatoid arthritis and related these results to bone microstructures of the distal radius measured by micro-computed tomography (mCT). We grouped the patients into tertiles with a low, medium or high degree of Fc sialylation, galactosylation and fucosylation of IgG, and compared bone volume, trabecular numbers and trabecular thickness between these groups.

In accordance with our previous experiments, patients exhibiting IgG with low-sialylated or low-galactosylated Fc portions displayed a signicantly decreased bone volume compared with patients with high levels of IgG Fc sialylation or galactosylation (Fig. 4a,b,d). This effect seems mainly be related to lower trabecular numbers, as no difference was observed in trabecular thickness. Similar results were obtained for disease-specic ACPA. The analysis of Fc fucosylation of IgG and ACPA did not show any relation to the bone structure (Fig. 4c). In addition, there was no difference between the groups regarding other parameters potentially impacting bone architecture or IgG glycosylation status27, such as age, sex, disease duration, disease activity or treatment status (Supplementary Table 1), suggesting

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651

FcRI FcRI

FcRIIA

FcRIII

FcRII

FcRIII

Isotype

Rel. mRNA expression

Rel. mRNA expression

Rel. mRNA expression

0.03

0.015

0.8

***

***

***

***

0.02

0.01

0.0080.0060.0040.002 0

0.250.200.150.100.05 0

0.010

Merge Receptor Phalloidin+DRAQ5

***

0.005

* *

0

0 0d 2d 4d 6d 8d 10d 0d 2d 4d 6d 8d 10d

0 0d 2d 4d 6d 8d 10d

0d 2d 4d 6d 8d 10d

1.5

FcRIIB

FcR DAP12

0.06

Rel. mRNA expression

Rel. mRNA expression

**

***

0.04

***

**

***

***

***

**

**

0.02

0d 2d 4d 6d 8d 10d

Rel. mRNA expression

1.0

0.60.40.2 0

*

*

P=0.014 P=0.014

1.0

0.5

IC-ds

*

Fold change

osteoclast number

0d 2d 4d 6d 8d 10d

0

+ + + +

+

FcRI FcRII FcRIII Isotype

Figure 2 | FccRII and III mediate the signal of non-sialylated immune complexes on preosteoclasts. (a) Quantitative RT-PCR for relative mRNA levels of FcgR and adaptor proteins normalized on b2-microglobulin during osteoclastogenesis. Day 0 represents the stage of monocytes, day 68 the stage of late preosteoclasts and day 10 the stage of mature osteoclasts. Bars show means.e.m. of three independent experiments. (b) Fluorescence microscopy images of FcgR expression on human preosteoclasts. FcgR are depicted in green, phalloidin staining for actin is depicted in red and DRAQ5 staining for the nuclei is depicted in blue. Scale bar, 25 mm (representative images of three independent experiments). (c) Fold change of osteoclast number after stimulation of preosteoclasts with 100 mg ml 1 of desialylated immune complexes (IC-ds) in the presence of 10 mg ml 1 of blocking antibodies against indicated FcgR. TRAP-positive cells with Z3 nuclei were considered as osteoclasts. Bars show means.e.m. of four independent experiments. Statistical analysis was performed with MannWhitney U-test. *Po0.5, **Po0.01, ***Po0.001.

that the observed differences in bone structure were conned to IgG sialylation and galactosylation.

To further dene the role of IgG sialylation in human bone structure, we addressed whether Fc sialylation inuences the bone phenotype even before the clinical onset of disease. We thereby made use of the observation that autoantibodies emerge several years before the onset of rheumatoid arthritis and that in some individuals bone changes precede the clinical onset of the disease28,29. When investigating IgG and ACPA Fc sialylation as well as bone structure in autoantibody-positive healthy individuals before the onset of rheumatoid arthritis, we again found that bone volume and trabecular numbers are signicantly lower in individuals with low IgG and low ACPA sialylation compared with those with a high-sialylation status (Supplementary Fig. 5).

Sialylated ACPA lose their pro-osteoclastogenic activity. To determine whether the observed association between IgG Fc galactosylation and bone volume is due to a direct effect of IgG Fc galactosylation or simply the fact that galactose is needed for the attachment of sialic acid to the IgG glycan, we enzymatically galactosylated and sialylated two human monoclonal ACPA (clones 109 and C7) (Supplementary Fig. 6) that have been generated from B cells from the joints of

rheumatoid arthritis patients30. In consistence with previously reported data20, both clones promoted osteoclast differentiation (Fig. 4e). Sialylation of these antibodies resulted in a loss of their pro-osteoclastogenic activity to the point of even a slight inhibition of osteoclastogenesis. In contrast, galactosylation alone, without sialylation, did not seem to have an effect on the osteoclastogenic properties of the antibodies. In accordance with these ndings, analysis of the sialic acid content per galactose residue on IgG in rheumatoid arthritis patients showed a signicant association of higher sialylation rates with increased bone volume and trabecular numbers (Supplementary Fig. 7), suggesting that IgG galactosylation is of much less importance for osteoclastogenesis and bone loss than IgG sialylation.

N-acetylmannosamine treatment protects mice from CIA. On the basis of these ndings, we hypothesized that an increase of IgG sialylation protects from autoimmune-mediated bone loss. To test this hypothesis, we supplemented the drinking water of DBA/J1 mice exposed to CIA with the sialic acid precursor N-acetylmannosamine (ManNAc). ManNAc is rapidly metabolized into 5-N-acetylneuraminic acid, which is the most abundant mammalian sialic acid in the body31. Supplementation of ManNAc is effective in treating certain myopathies caused by

4 NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651 ARTICLE

15 ***

***

P=0.06

PBS

IC 5 g IC-ds 5 g

PBS IC 5 g IC-ds 5 g

P=0.0002

*

P=0.033

P=0.057

P=0.073

10

N.Oc/B.Pm

5

0

PBS

IC

IC-ds

PBS

IC

IC-ds

IC IC-ds1 g 5 g 1 g

+ TNF

40

30

**

P=0.019

*

P=0.0002

P=0.035

*

P=0.024

P=0.083

PBS + TNFa IC 1 g + TNFa IC-ds 1 g + TNFa

PBS + TNFa IC 1 g + TNFa IC-ds 1 g + TNFa

Oc.S/BS

20

10

0

4

3

2

1

PBS IC IC-ds

1 g 5 g

IC-ds

IC

PBS

IC IC-ds

1 g

+ TNF

Inflammation score

IC

PBS

IC IC-ds

0 PBS IC IC-ds

1 g 5 g

IC-ds

1 g

+ TNF

Figure 3 | Non-sialylated immune complexes stimulate osteoclastogenesis in vivo. (a) Number of osteoclasts per bone perimeter (N.Oc/B.Pm) and (b) osteoclast surface per bone surface (Oc.S/BS) in tibial bone of C57BL/6 mice with intra-articular injection of the indicated amounts of native or desialylated (ds) murine immune complexes. Mice were injected at day 0 and day 4 with bone dissection at day 7. When indicated, 100 ng of TNFa was added to the rst injection. Bars show means.e.m. of Z12 knee joints of three independent experiments. (c) Representative images of TRAP-stained tibial sections. Scale bar, 50 mm. (d) Inammation score of haematoxylin-eosin stained joint sections with a score of 0 representing a healthy joint and a score of 4 representing a completely destroyed joint. Bars show means.e.m. of Z12 knee joints of three independent experiments. (e) Representative images of haematoxylin-eosin stained tibial sections. Scale bar, 500 mm. Statistical analysis was performed with MannWhitney U-test. *Po0.5, **Po0.01, ***Po0.001.

enzymatic defects associated with protein hyposialyation, which is currently evaluated in clinical studies32.

When using a similar protocol, mice fed with 10 g l 1 ManNAc displayed a signicantly higher IgG1 Fc sialylation compared with control mice that received either water or 10 g l 1 mannose (Fig. 5a). There was also a slightly higher IgG1 Fc galactosylation in ManNAc-treated mice, but this was not signicant. ManNAc treatment led to lower incidence, delayed onset and lower arthritis scores compared with controls challenged with either water or mannose (Fig. 5bd). No signicant difference in the amount of total IgG and collagen-specic antibodies between the groups was observed, suggesting that ManNAc treatment did not generally impair the IgG-based immune response (Fig. 5e). Furthermore, IgG Fc sialylation, but not galactosylation, was signicantly higher in mice that developed no or mild arthritis than in mice that developed severe arthritis (Fig. 5f), suggesting that higher IgG Fc sialylation levels reduce the susceptibility to experimental arthritis.

Most importantly, ManNAc treatment was very effective to protect from local and systemic bone loss. mCT measurements on the hind paws and the tibial bones revealed that ManNAc-treated mice did virtually not develop bone erosions and experienced less severe systemic bone loss during CIA compared with controls exposed to either water or mannose (Fig. 6a,b). Quantitative analysis showed that particularly trabecular bone loss was signicantly less severe in mice treated with ManNAc than in controls (Fig. 6c). Moreover, protection from bone erosions in ManNAc-treated mice was accompanied by signicant reduction in osteoclast numbers in the paws conrming the concept that higher IgG Fc sialylation impairs osteoclast differentiation and mitigates inammatory bone loss (Fig. 6d,e). The protective effect of high-sialylated IgG against bone loss was further conrmed in an IgG transfer model, in which untreated or in vitro sialylated IgG from mice, previously immunized against methylated bovine serum albumin (mBSA), was transferred to naive mice (Supplementary Fig. 8). Challenge with mBSA into the knee

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651

Sialylation (IgG) Sialylation (ACPA)

Galactosylation (ACPA)

BV/TV

Tb.N Tb.Th

BV/TV Tb.N Tb.Th

BV/TV Tb.N Tb.Th BV/TV Tb.N Tb.Th

BV/TV Tb.N

Tb.Th

0.25

** ** * *

*

3 0.15

0.10

0.05

0.00

mm

0.15

0.10

0.05

0.00

0.20

0.15

(%) (%) (%)

0.25

0.20

0.15

P=0.003 P=0.007

P=0.309 P=0.026 P=0.012 P=0.554

2

1

0

1/mm

1/mm

(%)

0.10

0.05

0.00

3

2

1

0

mm

0.10

0.05

0.00

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Galactosylation (IgG)

Fucosylation (IgG) Fucosylation (ACPA)

BV/TV Tb.N Tb.Th

* *

*

0.25

0.20

0.15

0.10

0.05

0.00

0.15

0.10

0.05

0.00

P=0.014

P=0.907 P=0.205 P=0.215 P=0.650 P=0.754 P=0.144

P=0.039

P=0.203

P=0.037

3 P=0.015

P=0.344

1/mm

mm

(%)

0.25

0.20

0.15

0.10

0.05

0.00

1/mm

mm

Low

High Sialylation

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

*

2.0

Fold change

osteoclast number

*

P=0.0495

P=0.0495

0.25 3

2

1

0

1/mm

3

2

1

0

0.20

0.15

0.10

0.05

0.00

mm

0.15

0.10

0.05

0.00

(%)

0.25

0.20

0.15

0.10

0.05

0.00

1/mm

3

2

1

0

1.5

1.0

0.5

0.0

Untreated

mm

0.15

0.10

0.05

0.00

0.15

0.10

0.05

0.00

109 109 Sial

C7

C7 Gal

C7 Sial

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Low

Medium

High

Figure 4 | Sialylation status of IgG and ACPA control bone mass in rheumatoid arthritis patients. (ac) Bone morphometric parameters including bone volume per tissue volume (BV/TV), trabecular number (Tb.N) and trabecular thickness (Tb.Th) of RA patients with low, medium or high levels of Fc (a) sialylation, (b) galactosylation or (c) fucosylation of total IgG and ACPA. Bars show means.e.m. of10 patients in each tertile. Cutoffs for tertiles are as follows: IgG Fc sialylation (low: o14%, middle: 1517.5%, high: 417.5%), ACPA Fc sialylation (o12%; 1216%; 416%), IgG Fc galactosylation (o43.5%; 43.551%; 451%); ACPA Fc galactosylation (o44%; 4454.5%; 454.5%); IgG Fc fucosylation (o85%; 8587%; 487%); ACPA Fc fucosylation (o93%; 9396%; 496%). (d) Two-dimensional (upper panel) and 3-dimensional (lower panels) reconstruction of the bone architecture of patients with low and high degree of Fc sialylation; (e) Fold change of osteoclast number after treatment of preosteoclasts for 72 h with 10 mg ml 1 of native, galactosylated (gal) or galactosylated and sialylated (sial) monoclonal ACPA (clone 109 and C7). TRAP-positive cells with Z3 nuclei were considered as osteoclasts. Bars show means.e.m. of three independent experiments. Statistical analysis was performed with KruskalWallis test with

Dunns correction (ac) and MannWhitney U-test (e). *Po0.5, **Po0.01.

joint of the recipient mice led to increased osteoclast numbers only in mice that received untreated IgG. In contrast, mice receiving sialylated IgG were protected from enhancement of osteoclastogenesis, although they developed joint swelling to a similar extent as mice receiving untreated IgG.

DiscussionIn the last years, there have been several studies with controversial ndings regarding the effects of immune complexes on osteoclastogenesis17,19,33. While the crosslinking of FcgR was found to be favourable for osteoclastogenesis19, it was unclear if the binding of immune complexes has the same effect. Here we show that the degree of IgG sialylation is of major importance for immune complexosteoclast interactions and that only low- or non-sialylated immune complexes drive osteoclastogenesis in vitro as well as in vivo. Using random IgG, we further demonstrated that the pro-osteoclastogenic effect of nonsialylated immune complexes is not restricted to distinct antigen specicities but a common feature of all IgG antibodies.

In human serum only a minority of IgG of less than 20% is sialylated5. This low amount of sialylation suggests that the

observed difference in preosteoclast stimulation between sialylated and non-sialylated immune complexes is not merely due to a reduced binding of sialylated IgG to the classical FcgR, but more likely due to an active suppression by sialylated IgG for example by binding to C-type lectins, such as dendritic cell immunoreceptor and dendritic cell-specic intercellular adhesion molecule-3-grabbing non-integrin, which recognize sialylated IgG and are expressed on myeloid cells9,10.

Interestingly, the putative suppressive function of sialylated IgG seems to be abrogated by strong pro-inammatory stimuli, as, in our mouse model, after co-injection of 100 ng of TNF-a also the sialylated immune complexes promoted osteoclastogenesis.

When investigating the IgG Fc glycans of rheumatoid arthritis patients, we found a signicant correlation between the Fc sialylation levels of random IgG as well as disease-specic autoantibodies (ACPA) and bone architecture. Patients with low levels of IgG Fc sialylation displayed lower bone volumes and lower trabecular numbers, while other parameters, such as age, sex or disease duration, did not differ among the groups. ACPA are well established as a major risk factor for the development of rheumatoid arthritis and are associated with a stronger disease course and enhanced bone erosion34,35. Of note, ACPA have been

6 NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651 ARTICLE

IgG1 galactosylation

50 40 30 20 10

0

IgG1 sialylation

*

P=0.0103

25

100

Control

P=0.058

% Of total glycans

20 15 10

5 0

Mannose

% Of total glycans

Incidence (%)

80 60 40 20

0

15 20 25 30 35

Days after immunization

ManNAc

Control

Water

Mannose

ManNAc

Control

Water

Mannose

ManNAc

Control

Water

Water

15 **

Score/affected mice

12

9

6

3

0

Control Water Mannose ManNAc

P=0.0024

Score day 32

10

5

0

Control

Water

Mannose

ManNAc

Mannose

ManNAc

15 20 25 30 35

Days after immunization

Total IgGP=0.058 P=0.065 P=0.0031 P=0.1481

Anti-collagen IgG

3

2

1

0

**

1.5

1.0

0.5

0.0

% IgG1 sialylation

10

5

0

20 50 40 30 20 10

0

% IgG1 galactosylation

15

mg ml1

OD 450nm

Control

Water

Mannose

ManNAc

Control

Water

Mannose

ManNAc

04 >4

Score

04 >4

Score

Figure 5 | Treatment with ManNAc increases sialylation of IgG1 and reduces susceptibility to CIA. (a) Quantication of total serum IgG1 Fc sialylation and galactosylation at day 32 in non-induced control mice and mice induced for CIA receiving treatment with water, 10 g l 1 mannose or 10 g l 1 ManNAc.

(b) Incidence of CIA. (c) Arthritis scores. (d) Representative images of hind paws at day 32. Scale bar, 5 mm. (e) Quantication of total serum IgG and collagen- specic IgG at day 32. Bars show means.e.m. of combined data from two independent experiments (non-induced control group: n 9 mice; all

other groups: n 13 mice). (f) Quantication of total serum IgG1 Fc sialylation and galactosylation at day 32 from mice that stayed healthy or developed

mild arthritis (score 04) and from mice that developed severe arthritis (score 44). Bars show means.e.m. of combined data from all mice subjected to CIA (score 04: n 13; score 44: n 25). Statistical analysis was performed with MannWhitney U-test. *Po0.5, **Po0.01.

reported to be less sialylated than random IgG21, which may explain their strong pathogenicity. We now provided evidence that this feature may also be a reason for the direct proosteoclastogenic effect of ACPA that has been published earlier20. Indeed, ACPA that were sialylated in vitro completely lost their capacity to drive osteoclastogenesis.

We have also found a signicant correlation between IgG Fc galactosylation and bone architecture in rheumatoid arthritis patients. But, as in vitro galactosylation of ACPA did not alter their pro-osteoclastogenic activity, IgG galactosylation does not seem to play a major role for direct immune complex preosteoclast interactions. However, we cannot exclude indirect effects of IgG galactosylation on osteoclastogenesis for example by inuencing the overall joint inammation.

On the basis of our data we hypothesized that an enhancement of IgG sialylation may be a treatment strategy to inhibit autoimmune-mediated bone loss. To test this hypothesis, we used the sialic acid precursor ManNAc, which is an interesting approach as it can be taken up orally, has been shown to affect myopathies related to hyposialylation32 and enhances overall protein sialylation36. Indeed, mice induced for CIA that were fed with ManNAc displayed a signicant increase in IgG sialylation compared with mice that received water or a mannose solution. Treatment with ManNAc not only resulted in a mitigated course of arthritis, but also blocked inammatory

osteoclastogenesis and bone erosion. Whether such an approach is effective in the treatment of rheumatoid arthritis and emerges as a more feasible treatment approach than costly intravenous immunoglobulin infusions, which have shown efcacy in the treatment of rheumatoid arthritis37,38, however, remains to be determined.

In summary, our data show that, apart from the regulation of immune effector functions, IgG sialylation controls osteoclast differentiation and bone mass in mice and humans pointing to a new link between the adaptive immune system and bone with direct relevancy for human disease.

Methods

Desialylation and deglycosylation of IgG and IC generation. Human IgG was taken from Beriglobin (Behring). Murine IgG was obtained from pooled serum of healthy C57BL/6 mice (Charles River) of different age and sex by purication over a protein G column (GE Healthcare) according to the manufacturers instructions. For desialylation, 1 mg of human or murine IgG was incubated with 5U or 10,000U neuraminidase (NEB) for 24 h or 48 h, respectively, at 37 C. For deglycosylation, human IgG was incubated with 500 U mg 1 PNGase F (NEB) for 24 h at 37 C.

The efciency of the enzymatic digestion was tested with a lectin blot. The digested IgG was puried over a protein G column (GE Healthcare) according to the manufacturers instructions and tested for endotoxin contamination using aLAL chromogenic endotoxin quantitation kit (Thermo scientic). Protein concentration was determined with the DC protein assay (Bio-Rad) and adjusted to 10 mg ml 1. Immune complexes were obtained by heat aggregation of the IgG at 63 C for 30 min.

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651

Control

Control

Water

Water

Mannose

Mannose

ManNAc

ManNAc

tBV/TV tBMD cBMD

cBV/TV

P=0.004 P=0.471 P=0.408

150 0.9 800

750

700

650

600

0.9

0.8

0.8

0.7

0.7

0.15

** **

P=0.007

0.10

0.05

0.00

Control

Control

%

mg HA cm3

100

50

0

%

mg HA cm3

Water

Mannose

ManNAc

Control

Water

Mannose

ManNAc

Control

Water

Mannose

ManNAc

Control

Water

Mannose

ManNAc

Water Mannose ManNAc

E.Ar/T. Ar N.Oc/T.Ar 5 100

80

60

40

20

4

3

2

1

**

**

P=0.0068 P=0.0055

%

0 Control

Water

Mannose

ManNAc

0 Control

Water

Mannose

ManNAc

Figure 6 | Treatment with N-acetylmannosamine (ManNAc) inhibits arthritis-mediated bone loss. (a,b) Representative 3-dimensional reconstructions of the bone architecture of hind paws (a) and tibial bones (b) of non-induced control mice and mice induced for collagen-induced arthritis (CIA) treated with water, 10 g l 1 mannose or 10 g l 1 ManNAc. (c) Bone morphometric parameters of the tibiae including trabecular and cortical bone volume per tissue volume (tBV/TV; cBV/TV) and trabecular and cortical bone mineral density (tBMD; cBMD). Bars show means.e.m. from one representative experiment (non-induced control group: n 5 mice; water treated group: n 7 mice; all other groups n 6 mice). (d) Representative images of TRAP-stained paw

sections. Scale bar, 200 mm. (e) Histological parameters of the hind paws including eroded area per tissue area (E.Ar/T.Ar) and osteoclast number per

tissue area (N.Oc/T.Ar). Bars show means.e.m. of combined data from two independent experiments (non-induced control group: n 9 mice; all other

groups: n 13 mice). Statistical analysis was performed with MannWhitney U-test. **Po0.01.

8 NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651 ARTICLE

Galactosylation and Sialylation of monoclonal antibodies. Monoclonal ACPA from the clones 109 and C7 and anti-TNP antibodies were generated as described elsewhere30,39. For galactosylation, 1 mg of IgG was incubated with 10 mM UDP-galactose (Calbiochem) and 2,5 mU of b1-4 galactosyltransferase (Sigma) in 50 mM MOPS, pH7.2 with 20 mM MnCl2 for 48 h at 37 C. For sialylation, 1 mg of IgG was incubated with 10 mM CMP-sialic acid (Calbiochem) and 10 mU of a2-6 sialyltransferase (Sigma) in 50 mM MES, pH 6,0 with 10 mM MnCl2 for 48 h at 37 C. The reactions were conrmed with a lectin blot.

Lectin blotting. IgG was resolved on a sodium dodecyl sulfatePAGE (SDS PAGE) gel under reducing conditions, transferred to PVDF membranes and blocked with 3% deglycosylated gelatine (Sigma). Blots were incubated with biotinylated sumbuccus nigra lectin (2 mg ml 1) for sialic acid, erythrina cristagalli lectin (5 mg ml 1) for galactose or lens culinaris agglutinin (5 mg ml 1, all vector laboratories) for the detection of the core glycan, followed by incubation with an alkaline phosphatase conjugated mouse-anti-biotin antibody (Sigma) and detection with 4-nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indolyl phosphate (Roche).

Dynamic light scattering measurements. Angular-dependent dynamic light scattering was performed using a 6,328 nm-Laser and a CGS3-Goniometer (ALV) at a temperature of 20 C with two avalanche diode detectors and an ALV-5000 correlator with 320 channels under cross-correlation. A range of scattering angels of 30oyo150 was covered in 10 steps. Intensity autocorrelation functions were transferred into electric eld autocorrelation functions via Siegert relation and then transformed into the distribution of relaxation times A(t) via regularized Laplace transformation. Resulting mean relaxation times for each angle were transferred into the apparent diffusion coefcient Dapp, which was then extrapolated to zero scattering angle (scattering vector square); the extrapolated result subsequently gave the hydrodynamic radius RH through StokesEinstein relationship.

Generation and stimulation of preosteoclasts. Human monocytes were puried by plastic adhesion of peripheral blood mononuclear cells that had been isolated from EDTA-blood of normal healthy donors using a Ficoll gradient (Lymphoot, BioRad). Preosteoclasts were generated in a-Mem (Invitrogen) containing 10%

fetal bovine serum (Biochrome) and 1% penicillin/streptomycin (Invitrogen) with 30 ng ml 1 M-CSF, 10 ng ml 1 RANKL and 1 ng ml 1 TGF-b (all Peprotech).

After 69 days (depending on the donor), rst binucleated cells appeared and preosteoclasts were incubated as indicated with the different IgG preparations. Osteoclast differentiation was evaluated by staining for TRAP using a Leukocyte Acid Phosphatase Kit (Sigma) according to the manufacturers instructions.

For the resorption assay, human preosteoclasts were generated and stimulated with IgG complexes on plates coated with calcium phosphate (Corning) under the same conditions as described above. Resorption was visualized with a von Kossa staining. In brief, we lysed the cells with water and incubated the wells for 30 min with 5% silver nitrate. After extensive washing, the silver stain was developed for 1 min with 5% sodium carbonate in 25% formaldehyde and unreacted silver was removed with 5% sodium thiosulfate for 5 min. For evaluation, photos were taken and the percentage of the resorbed area was calculated with Adobe Photoshop CS5 extended.

Blocking of Fcc receptors. Preosteoclasts were generated as described above and incubated for 1 h with 10 mg ml 1 of blocking antibodies against FcgRI (10.1, biolegend), FcgRII (AT10, abcam), FcgRIII (3G8, provided from the group of Falk Nimmerjahn) or isotype control (mouse IgG1, k, biolegend), respectively. Subsequently, 100 mg ml 1 of desialylated aggregated IgG was added for 6 h, followed by a complete medium exchange. After additional 18 h, osteoclast differentiation was evaluated by staining cells for TRAP using a Leukocyte Acid Phosphatase Kit (Sigma) according to the manufacturers instructions.

Immunouorescence staining. Preosteoclasts were xed with 4% PFA (pH 7.2) in PBS for 10 min at 37 C. Fixed samples were then permeabilized and blocked for 1 h with 3% BSA in 0.1% Triton X-100 in PBS (all Sigma). Next, samples were incubated for 90 min with primary antibodies against FcgRI (12.7 mg ml 1, 10.1.

Abcam), FcgRII (1:200, Abcam), FcgRIII (2 mg ml 1, J5511, Abcam) followed by 1 h incubation with the alexa-uor-488 conjugated secondary antibodies (goat-anti-mouse for FcgRI and III and goat-anti-rabbit for FcgRII, 10 mg ml 1) together with alexa-uor-568-labelled phalloidin (1:200). Samples were then incubated with

DRAQ5 (all Life technologies) and mounted with FluorSave reagent (Calbiochem). Fluorescence images were acquired with a Zeiss confocal microscope.

RNA isolation and quantitative RT-PCR analysis. Osteoclasts were differentiated from human monocytes as described above. Directly after monocyte isolation or after various times of culture, total RNA was extracted with peqGOLD TriFast (Peqlab) according to the manufacturers instructions. RNA was transcribed into cDNA using oligo(dT) primers and MuLV reverse transcriptase (Roche). Quantitative real-time PCR was performed with SYBR Green I-dTTP (Eurogentec) and

the following primer pairs: b2-microglobulin (b2-MG) forward 50-GATGAGTAT GCCTGCCGTGTG-30 and b2-MG reverse 50-CAATCCAAATGCGGCATCT-30, fcgr1 forward 50-GTGTCATGCGTGGAAGGATA-30 and fcgr1 reverse 50-GCAC TGGAGCTGGAAATAGC-30, fcgr2a forward 50-CCAGCATGGGCAGCTCTTC ACC-30 and fcgr2a reverse 50-TGGGCAGCCTTCACAGGATCA-30, fcgr2b forward 50-GCGGCCATTGTTGCTGCTGTA-30 and fcgr2b reverse 50-AGAGCATCC GGGTGCATGAGA-30, fcgr3 forward 50-ACAGGTGCCAGACAAACCTC-30 fcgr3 reverse 50-TTCCAGCTGTGACACCTCAG-30, fcrg forward 50-TGATTCCAGCAGTGGTCTTGCTCT-30 and fcrg reverse 50-ATGCAGGCATATGTGATGC CAACC-30, dap12 forward 50-CAGCGACCCGGAAACAGCGT-30 and dap12 reverse 50-CGGCCTCTGTGTGTTGAGGTCG-30.

Intraarticular injection of immune complexes. All animal experiments were approved by the government of Mittelfranken, Germany. Male 8week- old C57BL/6-mice (Charles River) were randomly allocated into groups and administered with an intraarticular injection through the patellar tendon of 5 ml of PBS containing the indicated amounts of murine immune complexes. The injection was repeated after 4 days. After additional 3 days, mice were killed and the bones were dissected for histological analysis.

Induction of CIA and treatment with N-acetylmannosamine. Chicken collagen type II (4 mg ml 1; Sigma) was emulsied in equal amounts with complete

Freunds adjuvant (Sigma) containing 5 mg ml 1 heat-inactivated Mycobacterium tuberculosis (H37Ra; Difco). For the induction of CIA, male 7-week-old DBA/1J mice (Janvier) were injected s.c. at the base of the tail with 100 ml of this emulsion.

Mice were rechallenged after 21 days. N-acetylmannosamine was constantly administered to the drinking water at a concentration of 10 g l 1, beginning with the primary immunization. Control mice received water or water containing10 g l 1 mannose (all Sigma). The allocation of the mice into the different groups was performed randomly. Clinical arthritis was evaluated every second day using a scoring system from 0 (no swelling) to 3 (severe swelling and erythema) for every limb, resulting in a maximally possible score of 12 per animal.

Antigen-induced arthritis transfer model. mBSA (2 mg ml 1; Sigma) was emulsied in equal amounts with complete Freunds adjuvant (Sigma) containing 5 mg ml 1 heat-inactivated Mycobacterium tuberculosis (H37Ra; Difco). Male 6-week-old Balb/c mice (Janvier) were injected s.c. with 100 ml of this emulsion and simultaneously injected i.p. with 5 108 heat-inactivated Bordtella pertussis

(National Institute for Biological Standards and Control (NIBSC)) at day 0 and 7. After 21 days, the mice were challenged with 100 mg mBSA into one knee joint. At day 28, blood was taken and IgG was isolated with a protein G column (GE Healthcare) according to the manufacturers instructions. For sialylation, 1 mg of IgG was incubated with 750 mM CMP-sialic acid (Calbiochem) and 30 mU of a2-6 sialyltransferase (Sigma) in 100 mM Tris/HCl, pH 8,0 with 1 mM MnCl2 for 4 days at 37 C. The reaction was conrmed with a lectin blot. For the IgG transfer, 2 mg of untreated or sialylated AIA-IgG was injected i.v. into naive male 9-week-old Balb/c (Janvier) mice. After 1 and 5 days, the right knee joints were injected with 100 mg mBSA. The left knee joints served as internal controls. Knee joint swelling was determined using a dial thickness gauge (Peacock) and expressed relative to the knee diameter at day 0. At day 8, mice were killed and the bones were dissected for histological analysis.

Measurement of IgG concentrations in mouse serum. The serum concentrations of IgG were measured with an enzyme-linked immunosorbent assay (ELISA) quantitation kit (Bethyl Laboratories). For the assessment of collagen-specic IgG, plates were coated with 10 mg ml 1 chicken collagen type II (Sigma) instead of the capture antibody.

Histology. Bones were xed for 6 h in 4% formalin and decalcied in EDTA (Sigma-Aldrich). Serial parafn sections (2 mm) were stained for TRAP using a Leukocyte Acid Phosphatase Kit (Sigma) according to the manufacturers instructions or with haematoxylin-eosin. All analyses were performed using a microscope (Nikon) equipped with a digital camera and an image analysis system for performing histomorphometry (Osteomeasure; OsteoMetrics).

Micro-computed tomography of mice. mCT analyses were performed with a SCANCO Medical mCT 40 scanner using the following parameters: voltage, 40 kV; X-ray current, 250 mA; exposure time, 5,000 ms/projection for 720 projections; matrix, 1,024 1,024; voxel size in reconstructed image, 9 mm. Images were ana

lysed with SCANCO evaluation software for parameters at the metaphyses of the proximal tibiae: ratio of bone volume to total volume, trabecular number and trabecular thickness.

Patients. Thirty rheumatoid arthritis patients were analysed in detail for (i) bone structure by micro-computed tomography, (ii) IgG and ACPA glycosylation status and (iii) their demographic and disease-specic characteristics. The methodology of bone structure and antibody glycosylation analysis is described below. Age and

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651

sex were collected as the main demographic parameters showing a standard population of rheumatoid arthritis patients with a mean (s.e.m.) age of 52.82.8 years and a dominance of female patients (20 out of 30). As disease-specic parameters we analysed disease duration, disease activity and anti-rheumatic treatment (conventional versus biological disease modifying anti-rheumatic drug therapy (DMARD)) in all the patients. Patients had established rheumatoid arthritis with a mean disease duration of 4.60.9 years. Their disease activity was3.50.2 units according to the disease activity score (DAS) 28, resembling moderate disease activity. Furthermore, both, patients with conventional DMARD (methotrexate) therapy (16 out of 30) and biological DMARD (tumour necrosis factor inhibitor) therapy (14 out of 30) were analysed. In addition to the 30 patients with rheumatoid arthritis, a small group of 12 healthy individuals with autoanti-body positivity, but no arthritis, was analysed for bone architecture and antibody glycosylation status as well as to exclude potential effects related to inammation. All analyses of human material were performed in full agreement with institutional guidelines and with the approval of the Ethical committee of the University Hospital Erlangen (permit # 248_13B). Informed consent and permission to use the obtained data for research were obtained from all subjects enroled in the study.

Micro-computed tomography in patients. HR-pQCT measurements were performed by an XtremeCT scanner (Scanco, Switzerland) at the ultra-distal radius of the right arm using the manufacturers standard in vivo protocol. The reference line was set manually. The rst CT-slice was 9.5 mm proximal to the reference line. For scanning, the hand was immobilized in a carbon bre cast. An antero-posterior scout view was then used to determine the region of interest. A total of 111 slices (voxel size 82 mm) were carried out. Details of the HR-pQCT measurements of the distal radius have been previously described by Boutroy et al.40 All measurements were performed with the same software (Version 6.0) by two well-trained physicians. Daily cross-calibrations with standardized control phantoms (Moehrendorf, Germany) were conducted to standardize measurements. Bone microstructure including trabecular bone volume fraction (BV/TV), trabecular number (Tb.N, 1/mm) and trabecular thickness (Tb.Th, mm) were analysed.

IgG isolation. Human ACPA-IgG and ACPA-depleted IgG were puried from 30 ml of serum. ACPA were isolated by antigen afnity chromatography using cyclic citrullinated peptide (CCP2)-coated beads prepared by mixing biotinylated CCP2 dissolved in PBS at 1 mg ml 1 with Neutravidin Plus UltraLink resin (Pierce, Thermo Scientic) for 1 h at room temperature (RT). Following extensive washing, the CCP2 beads were dissolved in PBS and loaded into an Of1100 96-well lter plate (Orochem). Prediluted serum samples were applied and binding of ACPA to the beads was allowed by shaking the plate at 600 r.p.m. for 2 h at RT. Following incubation, the plate was centrifuged for 2 min at 1,500 g and the ow

through was collected in a 96-deepwell storage plate (Thermo). Beads were washed with 200 ml PBS and 25 mM ammonium bicarbonate followed by ACPA elution using 100 mM formic acid at pH 2.5 (pro analysis for mass spectrometry; Merck). The elution fraction was directly neutralized with 2 M TRIS. IgG (ow-through) and eluted ACPA-IgG were further puried using CaptureSelect IgG-Fc (Hu) Afnity Matrix (life technologies). To this end, 20 ml of CaptureSelect afnity resin was loaded into a 96-well lter plate (Of1100, 0.7 ml per well, PE frit, Orochem) and washed three times with 200 ml PBS using a vacuum manifold (Millipore), before transferring the CCP2 ow-through/elution samples to the plate. The plate was then incubated on a multiwell plate shaker (1.5 mm orbit, VWR) for 1 h at 450 r.p.m. to facilitate the binding of antibodies to the beads. The latter were then washed three times with 200 ml PBS using vacuum before the elution of (ACPA)-

IgG by adding 100 ml of 100 mM formic acid at pH 2.4 (pro analysis for mass spectrometry; Merck). Again, the plate was shaken for 5 min on a shaking platform at 450 r.p.m. and elution of the antibodies was performed by centrifuging 1 min at 500 g. IgG of DBA/1J mice were isolated from 2 ml of serum (pre-diluted in 200 ml

PBS) using 20 ml of Protein G-Sepharose Fast Flow resin (GE Healthcare) according to the protocol described above for IgG capturing using CaptureSelect IgG-Fc (Hu) Afnity Matrix.

Fc-glycosylation analysis by nano-LC-ESI-MS. IgG and ACPA-IgG eluates were dried in a vacuum centrifuge and subjected to tryptic digest by adding 200 ng trypsin (sequencing grade, Promega) in 40 ml ammonium bicarbonate buffer followed by overnight incubation at 37 C. Digested ACPA and IgG antibodies were separated and analysed on an Ultimate 3000 UPLC system (Dionex Corporation, USA) coupled to a quadrupole-TOF mass spectrometer (MS) (micrOTOF-Q or maXis ultra-high resolution QTOF; Bruker Daltonics)41. The samples were injected and concentrated on a C18 solid phase extraction trap column (Dionex Acclaim PepMap100, 5 mm 300 mm i.d.) conditioned with 0.1% TFA (mobile phase A) for

1 min at 25 ml min 1. Sample separation was achieved on an Ascentis Express C18 nano-liquid chromatography column (50 mm 75 mm i.d., 2.7 mm HALO fused

core particles, Supelco) conditioned at 900 nl min 1 with mobile phase A after which the following gradients of mobile phase A and 95% acetonitrile (mobile phase B) were applied: 0 min 3% B, 2 min 6% B, 4.5 min 18% B, 5 min 30% B, 7 min 30% B, 8 min 1% B and 11 min 1% B for human IgG; 0 min 3% B, 2 min 6% B,4.5 min 18% B, 6 min 30% B, 8 min 30% B, 9 min 1% B and 12 min 1% B for mouse IgG. The UPLC was interfaced to the MS with a standard ESI source (Bruker

Daltonics) and a sheath-ow ESI sprayer (capillary electrophoresis ESI-MS sprayer; Agilent Technologies)41. Mass spectra were recorded from m/z 600 to 2,000 with two averages at a frequency of 0.5 hz. To reduce glycan decay during ion transfer, the quadrupole ion energy and collision energy of the MS were set at 2 and 4 eV, respectively. The total analysis time per sample was 13 and 14 min for human and mouse IgG samples, respectively. IgG Fc-glycopeptides were identied based on their retention times and their accurate monoisotopic masses of double protonated and triple-protonated charged ions (human IgG1: EEQYN*STYR, mouse IgG1: EEQFN*STFR and mouse IgG2A: EDYN*STLR; N* stands for an N-glycosylated asparagine). Internal calibration of LC-MS spectra was performed in Bruker DataAnalysis 4.0 using a list of known glycopeptides, before exportation of the data to the mzXML format. A list of glycopeptide features, dened as a retention time window of 1520 s and a peak maximum within mass window of m/z 0.07, was extracted from each data set using the in-house developed Xtractor2D or 3D software41. The extracted data, merged on a sample-data matrix, were nally evaluated using Microsoft Excel. Quality of mass spectra was determined based on intensities of total IgG1 glycoforms in both human and mice samples. The level of galactosylation, sialylation and fucosylation of human IgG1 were calculated using the following formulas: galactosylation (G1F G1FN G1FS G1FNS G1

G1N G1S)*0.5 G2F G2FN G2FS G2FNS G2 G2N G2S;

fucosylation G0F G1F G2F G0FN G1FN G2FN G1FS G2FS;

sialylation G1FS G2FS G1FNS G2FNS G1S G2S (G: galactose,

F: Fucose; N: bisecting N-Acetylglucosamine; S: sialic acid). The degree of galactosylation and sialylation of mouse IgG1and IgG2a was calculated using the following formulae: galactosylation (G1 G1F G1FN G1FNeuAc

G1FNeuGc)*0.5 (G2F G3F G2FNeuGc G3FNeuGc G2FNeuGc2)*1);

sialylation (G1FNeuAc G1FNeuGc G2FNeuGc G3FNeuGc)*0.5

G2FNeuGc2. (G: galactose, F: Fucose; N: bisecting N-Acetylglucosamine; NeuAc: N-Acetylneuraminic Acid; NeuGc: N-Glycolylneuraminic Acid).

Data procession and statistical analysis. We performed computations and charts with GraphPad Prism 5.03 software. If not indicated differently, two-sided MannWhitney U test was used for statistical analysis (*Po0.05; **Po0.01;

***Po0.001). For comparison of glycosylation tertiles with respect to bone data in humans, the KruskalWallis test with Dunns correction was used. We also performed a post-hoc power calculation for bone data in humans by taking into account the size of the observed effect and a type I error probability of 2.5% (adjusted for two tests, that is, low versus medium, low versus high), resulting in a power of 77.2% for the corresponding signicant nding. Data are presented as means.e.m. as well as medians and inter-quartile ranges. All analysis was performed in a blinded manner.

References

1. Arnold, J. N., Wormald, M. R., Sim, R. B., Rudd, P. M. & Dwek, R. A. The impact of glycosylation on the biological function and structure of human immunoglobulins. Annu. Rev. Immunol. 25, 2150 (2007).

2. Anthony, R. M. & Nimmerjahn, F. The role of differential IgG glycosylation in the interaction of antibodies with FcgammaRs in vivo. Curr. Opin. Organ Transplant. 16, 714 (2011).

3. Yamaguchi, Y. et al. Glycoform-dependent conformational alteration of the Fc region of human immunoglobulin G1 as revealed by NMR spectroscopy. Biochim. Biophys. Acta. 1760, 693700 (2006).

4. Anthony, R. M. et al. Recapitulation of IVIG anti-inammatory activity with a recombinant IgG Fc. Science 320, 373376 (2008).

5. Bohm, S., Schwab, I., Lux, A. & Nimmerjahn, F. The role of sialic acid as a modulator of the anti-inammatory activity of IgG. Semin. Immunopathol. 34, 443453 (2012).

6. Scallon, B. J., Tam, S. H., McCarthy, S. G., Cai, A. N. & Raju, T. S. Higher levels of sialylated Fc glycans in immunoglobulin G molecules can adversely impact functionality. Mol. Immunol. 44, 15241534 (2007).

7. Kaneko, Y., Nimmerjahn, F. & Ravetch, J. V. Anti-inammatory activity of immunoglobulin G resulting from Fc sialylation. Science 313, 670673 (2006).

8. Campbell, I. K. et al. Therapeutic effect of IVIG on inammatory arthritis in mice is dependent on the Fc portion and independent of sialylation or basophils. J. Immunol. 192, 50315038 (2014).

9. Anthony, R. M., Wermeling, F., Karlsson, M. C. & Ravetch, J. V. Identication of a receptor required for the anti-inammatory activity of IVIG. Proc. Natl Acad. Sci. USA 105, 1957119578 (2008).

10. Massoud, A. H. et al. Dendritic cell immunoreceptor: A novel receptor for intravenous immunoglobulin mediates induction of regulatory T cells. J. Allergy Clin. Immunol. 133, 853863 e855 (2014).

11. Sondermann, P., Pincetic, A., Maamary, J., Lammens, K. & Ravetch, J. V. General mechanism for modulating immunoglobulin effector function. Proc. Natl Acad. Sci. USA 110, 98689872 (2013).

12. Yu, X., Vasiljevic, S., Mitchell, D. A., Crispin, M. & Scanlan, C. N. Dissecting the molecular mechanism of IVIg therapy: the interaction between serum IgG and DC-SIGN is independent of antibody glycoform or Fc domain. J. Mol. Biol. 425, 12531258 (2013).

10 NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7651 ARTICLE

13. Braun, T. & Zwerina, J. Positive regulators of osteoclastogenesis and bone resorption in rheumatoid arthritis. Arthritis Res. Ther. 13, 235 (2011).

14. Zhao, B. & Ivashkiv, L. B. Negative regulation of osteoclastogenesis and bone resorption by cytokines and transcriptional repressors. Arthritis Res. Ther. 13, 234 (2011).

15. Koga, T. et al. Costimulatory signals mediated by the ITAM motif cooperate with RANKL for bone homeostasis. Nature 428, 758763 (2004).

16. Hayashi, M. et al. Osteoprotection by semaphorin 3A. Nature 485, 6974 (2012).

17. Grevers, L. C. et al. Immune complex-induced inhibition of osteoclastogenesis is mediated via activating but not inhibitory Fcgamma receptors on myeloid precursor cells. Ann. Rheum. Dis. 72, 278285 (2013).

18. Harre, U. et al. Moonlighting osteoclasts as undertakers of apoptotic cells. Autoimmunity 45, 612619 (2012).

19. Seeling, M. et al. Inammatory monocytes and Fcgamma receptor IV on osteoclasts are critical for bone destruction during inammatory arthritis in mice. Proc. Natl Acad. Sci. USA. 110, 1072910734 (2013).

20. Harre, U. et al. Induction of osteoclastogenesis and bone loss by human autoantibodies against citrullinated vimentin. J. Clin. Invest. 122, 17911802 (2012).

21. Scherer, H. U. et al. Glycan proling of anti-citrullinated protein antibodies isolated from human serum and synovial uid. Arthritis Rheum. 62, 16201629 (2010).

22. Henson, P. M. & Spiegelberg, H. L. Release of serotonin from human platelets induced by aggregated immunoglobulins of different classes and subclasses.J. Clin. Invest. 52, 12821288 (1973).23. Ji, H. et al. Arthritis critically dependent on innate immune system players. Immunity 16, 157168 (2002).

24. Zhao, X. et al. Circulating immune complexes contain citrullinated brinogen in rheumatoid arthritis. Arthritis Res. Ther. 10, R94 (2008).

25. Van Steendam, K. et al. Citrullinated vimentin as an important antigen in immune complexes from synovial uid of rheumatoid arthritis patients with antibodies against citrullinated proteins. Arthritis Res. Ther. 12, R132 (2010).

26. Mathsson, L., Lampa, J., Mullazehi, M. & Ronnelid, J. Immune complexes from rheumatoid arthritis synovial uid induce FcgammaRIIa dependent and rheumatoid factor correlated production of tumour necrosis factor-alpha by peripheral blood mononuclear cells. Arthritis Res. Ther. 8, R64 (2006).

27. Collins, E. S. et al. Glycosylation status of serum in inammatory arthritis in response to anti-TNF treatment. Rheumatology 52, 15721582 (2013).

28. Rantapaa-Dahlqvist, S. et al. Antibodies against cyclic citrullinated peptide and IgA rheumatoid factor predict the development of rheumatoid arthritis. Arthritis Rheum. 48, 27412749 (2003).

29. Kleyer, A. et al. Bone loss before the clinical onset of rheumatoid arthritis in subjects with anticitrullinated protein antibodies. Ann. Rheum. Dis. 73, 854860 (2014).

30. Amara, K. et al. Monoclonal IgG antibodies generated from joint-derived B cells of RA patients have a strong bias toward citrullinated autoantigen recognition. J. Exp. Med. 210, 445455 (2013).

31. Keppler, O. T. et al. UDP-GlcNAc 2-epimerase: a regulator of cell surface sialylation. Science 284, 13721376 (1999).

32. Galeano, B. et al. Mutation in the key enzyme of sialic acid biosynthesis causes severe glomerular proteinuria and is rescued by N-acetylmannosamine. J. Clin. Invest. 117, 15851594 (2007).

33. MacLellan, L. M. et al. Co-opting endogenous immunoglobulin for the regulation of inammation and osteoclastogenesis in humans and mice. Arthritis Rheum. 63, 38973907 (2011).

34. van Gaalen, F. A. et al. Autoantibodies to cyclic citrullinated peptides predict progression to rheumatoid arthritis in patients with undifferentiated arthritis: a prospective cohort study. Arthritis Rheum. 50, 709715 (2004).

35. Saeki, Y. et al. Baseline anti-citrullinated peptide antibody (ACPA) titers and serum interleukin-6 (IL-6) levels possibly predict progression of bone destruction in early stages of rheumatoid arthritis (ERA). Rheumatol. Int. 33, 451456 (2013).

36. Gagiannis, D., Gossrau, R., Reutter, W., Zimmermann-Kordmann, M. & Horstkorte, R. Engineering the sialic acid in organs of mice using N-propanoylmannosamine. Biochim. Biophys. Acta 1770, 297306 (2007).

37. Tumiati, B., Casoli, P., Veneziani, M. & Rinaldi, G. High-dose immunoglobulin therapy as an immunomodulatory treatment of rheumatoid arthritis. Arthritis Rheum. 35, 11261133 (1992).

38. Muscat, C. et al. Long term treatment of rheumatoid arthritis with high doses of intravenous immunoglobulins: effects on disease activity and serum cytokines. Ann. Rheum. Dis. 54, 382385 (1995).

39. Lux, A., Yu, X., Scanlan, C. N. & Nimmerjahn, F. Impact of immune complex size and glycosylation on IgG binding to human FcgammaRs. J. Immunol. 190, 43154323 (2013).

40. Boutroy, S., Bouxsein, M. L., Munoz, F. & Delmas, P. D. In vivo assessment of trabecular bone microarchitecture by high-resolution peripheral quantitative computed tomography. J. Clin. Endocrinol. Metab. 90, 65086515 (2005).

41. Selman, M. H. et al. Fc specic IgG glycosylation proling by robust nano-reverse phase HPLC-MS using a sheath-ow ESI sprayer interface.J. Proteomics 75, 13181329 (2012).

Acknowledgements

We thank Barbara Happich, Isabell Schmidt, Hedwig Symowski, Maurice H.J. Selman and Lise Hafkenscheid for technical assistance. We also want to thank Pierre Jurdic and his team for advice and technical assistance with microscopy. This study was supported by the Deutsche Forschungsgemeinschaft (Focus program SPP1468 IMMUNOBONE), the Bundesministerium fr Bildung und Forschung (BMBF; METARTHROS), the Marie- Curie project Osteoimmune, the TEAM and Masterswitch project of the European Union, the IMI funded project BTCure, and the Interdisciplinary Centre for Clinical Research of the University of Erlangen-Nuremberg. U.H. was supported by the doctoral training program GK1660 from the Deutsche Forschungsgemeinschaft. R.E.M.T. was supported by the Dutch Arthritis Foundation. H.U.S. is a recipient of a Clinical fellowship of The Netherlands Organization for Health Research and Development.

Author contributions

U.H., M.H. and G.S. designed and performed experiments and analysed data. U.H., S.C.L., W.B. and K.D. performed animal experiments. R.P. and A.L. contributed to the enzymatic modication of IgG. S.F. and F.G. performed and analysed the dynamic light scattering measurements. K.A., V.M. and L.K. generated and provided monoclonal ACPA. Y.R., C.A.K., R.E.M.T. and H.U.S. isolated and analysed IgG from sera. H.B., G.K., F.N. and M.H. provided valuable material and intellectual input. U.H. and G.S. wrote the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: There are no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Harre, U. et al. Glycosylation of immunoglobulin G determines osteoclast differentiation and bone loss. Nat. Commun. 6:6651 doi: 10.1038/ncomms7651 (2015).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 6:6651 | DOI: 10.1038/ncomms7651 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

& 2015 Macmillan Publishers Limited. All rights reserved.