Retraction and Correction

RETRACTION

MEDICAL SCIENCESRetraction for “Dominant suppression of inflammation by glycan-hydrolyzed IgG,” by Kutty Selva Nandakumar, Mattias Collin,Kaisa E. Happonen, Allyson M. Croxford, Susanna L. Lundström,Roman A. Zubarev, Merrill J. Rowley, Anna M. Blom, andRikard Holmdahl, which appeared in issue 25, June 18, 2013, ofProc Natl Acad Sci USA (110:10252–10257; first published May13, 2013; 10.1073/pnas.1301480110).The authors wish to note the following: “Using studies of IgG

hydrolyzed by the streptococcal glycan hydrolyzing enzyme EndoS,we found that treatment of mice with hydrolyzed IgG blockedantibody mediated arthritis. As an explanation for this observa-tion, we suggested that EndoS-hydrolyzed IgG per se dominantlyblocks local immune complex formation.“With new data from our own follow up experiments, we have

now found that this conclusion was incorrect.“Our new data shows that injection of EndoS is much more

potent in vivo than we could logically anticipate, as i.v. injectionof doses containing less than 0.1 μg EndoS mixed with IgGsuppressed arthritis using the same model as the one reported inthe initial paper (collagen antibody-induced arthritis). We pre-viously excluded the possibility that contaminating EndoS couldplay a role, as this contaminating amount was not detected usingstandard methods in the hydrolyzed IgG fraction we used in theexperiments. Furthermore, much higher doses of EndoS injectedin the same mouse strain as a control experiment did not affectcollagen induced arthritis in earlier experiments. The correctinterpretation of our collective data is that EndoS operates verypotently in vivo on an immune complex-mediated disease, pos-sibly by accumulating within immune complexes. Because thisinterpretation is different from our major conclusion of the pub-lished paper, the authors have unanimously decided to retract thispaper to be able to publish the data connected with a correct in-terpretation. We sincerely apologize to readers of this paper, whomight have been misled by our earlier interpretation.”

Kutty Selva NandakumarMattias CollinKaisa E. HapponenAllyson M. CroxfordSusanna L. LundströmRoman A. ZubarevMerrill J. RowleyAnna M. BlomRikard Holmdahl

www.pnas.org/cgi/doi/10.1073/pnas.1419043111

CORRECTION

MEDICAL SCIENCESCorrection for “Increased muscle PGC-1α expression protects fromsarcopenia and metabolic disease during aging,” by Tina Wenz,Susana G. Rossi, Richard L. Rotundo, Bruce M. Spiegelman, andCarlos T. Moraes, which appeared in issue 48, December 1, 2009,of Proc Natl Acad Sci USA (106:20405–20410; first publishedNovember 16, 2009; 10.1073/pnas.0911570106).The authors note that the α-tubulin loading control blot in

Fig. 4D appeared incorrectly. The corrected figure and its legendappear below.

www.pnas.org/cgi/doi/10.1073/pnas.1419095111

Fig. 4. Increased PGC-1α levels in aging muscle prevent degradative pro-cesses. (A) Immunohistochemistry of biceps femoris using anti-active caspase3 antibody to detect apoptosis. (B) Apoptotic index in skeletal muscle ho-mogenates of wild-type and MCK-PGC-1α of different age-groups based onnucleosome fragmentation (n = 6 for each group). *, P < 0.05, **, P < 0.01,***, P < 0.001. (C) Western blot of Bax and Bcl-2 in skeletal muscle homo-genates. (D) Western blot of the 20S subunit of the proteasome and tubulinin skeletal muscle homogenates. (E) Western blot of LC3-I and LC3-II inskeletal muscle homogenates.

www.pnas.org PNAS | November 4, 2014 | vol. 111 | no. 44 | 15851

RETR

ACT

IONAND

CORR

ECTION

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0 D

ownl

oade

d by

gue

st o

n N

ovem

ber

16, 2

020

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0 D

ownl

oade

d by

gue

st o

n N

ovem

ber

16, 2

020

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0 D

ownl

oade

d by

gue

st o

n N

ovem

ber

16, 2

020

Dow

nloa

ded

by g

uest

on

Nov

embe

r 16

, 202

0 D

ownl

oade

d by

gue

st o

n N

ovem

ber

16, 2

020

Dominant suppression of inflammationby glycan-hydrolyzed IgGKutty Selva Nandakumara,1, Mattias Collinb, Kaisa E. Happonenc, Allyson M. Croxfordd, Susanna L. Lundströme,Roman A. Zubareve, Merrill J. Rowleyd, Anna M. Blomc, and Rikard Holmdahla

aMedical Inflammation Research and eChemistry I Division, Department of Medical Biochemistry and Biophysics, Karolinska Institute, 171 77 Stockholm,Sweden; bDivision of Infection Medicine, Department of Clinical Sciences, Lund University, 22184 Lund, Sweden; cDepartment of Laboratory Medicine Malmö,Lund University, 205 02 Malmö, Sweden; and dDepartment of Biochemistry and Molecular Biology, Monash University, Clayton, VIC 3800, Australia

Edited by Raymond A. Dwek, University of Oxford, Oxford, United Kingdom, and accepted by the Editorial Board April 16, 2013 (received for reviewJanuary 24, 2013)

A unique anti-inflammatory property of IgG, independent of anti-gen specificity, is described. IgGwith modification of the heavy-chainglycan on asparagine 297 by the streptococcal enzyme endo-β-N-acetylglucosaminidase (EndoS) induced a dominant suppressionof immune complex (IC)-mediated inflammation, such as arthritis,through destabilization of local ICs by fragment crystallizable–fragment crystallizable (Fc-Fc) interactions. Small amounts (250μg) of EndoS-hydrolyzed IgG were sufficient to inhibit arthritisin mice and most effective during the formation of ICs in the targettissue. The presence of EndoS-hydrolyzed IgG disrupted largerIC lattice formation both in vitro and in vivo, as visualized withanti-C3b staining. Neither complement binding in vitro nor antigen–antibody binding per se was affected.

collagen | endoglycosidase | glycosylation | monoclonal antibody |rheumatoid arthritis

Glycosylation is an important posttranslational modificationaffecting the structure and biological properties of glyco-

proteins. One of the most intensively studied is the asparagine 297(Asn-297) glycan on the heavy chain (γ-chain) of IgG, which is se-questered within the internal space enclosed by the CH2 domains.Subtle changes in IgG N-glycome significantly change fragmentcrystallizable (Fc) conformation with dramatic consequences forIgG effector functions (1). Extensive noncovalent interactionsbetween the carbohydrate and the protein moiety in the IgG–Fcregion result in reciprocal influences on conformation (2). NMRstudies suggested a significant role for Fc–glycan dynamics inFc receptor (FcR) interactions (3). The minimal oligosaccharidestructure in IgG is a hexasaccharide (GlcNAc2Man3GlcNAc) withvariable sugar residues attached, resulting in the generation ofmany different glycoforms. Such altered IgG glycoforms lackingterminal sialic acid and galactose residues were identified in rheu-matoid arthritis (RA) patients (4). Recently, differential sialylationwas reported to regulate the inflammatory property of IgG (5).Endoglycosidases form a group (EC 3.2 subclass) of enzymes

that hydrolyze nonterminal glycosidic bonds in oligosaccharidesor polysaccharides. Endo-β-N-acetylglucosaminidase (EndoS)is a member of the GlcNAc polymer hydrolyzing glycosyl hydrolasesof family 18 (FGH18) secreted by group A β-hemolytic Strepto-coccus pyogenes. It exclusively hydrolyzes the β-1, 4-di-N-acetylchi-tobiose core of the asparagine-linked complex-type glycan onAsn-297 of the γ-chains of IgG (6). EndoS has similarities toendo-β-N-acetylglucosaminidases, which cleave the β1–4 link-ages between N-acetylglucosamines found in the core of theN-linked glycan of IgG.Antibodies [anti-citrullinated protein antibodies (ACPA), rheu-

matoid factors (RF), and anti-type II collagen antibodies] andimmune complexes (ICs) are prevalent in RA. ACPA and RFalso precede disease development (7). IC-mediated pathology isevident in several autoimmune diseases. Importantly, the path-ogenic effect of circulating ICs was shown to be dependent ontheir size and composition (8). Both antigen-driven (soluble and

target tissue-bound) and RF containing ICs present in RA patientsare of intermediate (6S–19S) to large (22S–30S) size, and larger(>22S) ICs containing RF were implicated in extraarticularmanifestations in RA (9). Furthermore, collagen type II (CII)-containing ICs from RA synovial fluid were shown to induceproduction of inflammatory cytokines (TNF-α, IL-1β, and IL-8)from peripheral blood mononuclear cells via FcγRIIA (10).Antibodies from patients with RA upon passive transfer in-

duced arthritis in mice (11). The effector phase of arthritis isoptimally studied using the collagen antibody-induced arthritismodel induced by anti-CII IgG mAbs (12). This model exhibitsfeatures of bone and cartilage erosions, major infiltrations ofgranulocytes, and deposition of IgG and complement factors onthe cartilage surface, characteristic of RA, and is dependent oncomplement, FcγRs, TNF-α, IL-1β, and neutrophils and mac-rophages (12). Interestingly, removal of the N-linked glycan ab-rogated pathogenic potential of antibodies (13) and abolished allof the proinflammatory properties of IC from systemic lupuserythematosus patients (14). In addition, EndoS-hydrolyzed IgGameliorated several antibody-mediated diseases in mice, includingarthritis (15). Attenuation of inflammation was reported to bedependent on IgG1 and IgG2b subclasses, but the mechanismswere not clarified. Here we demonstrate suppression of in-flammatory arthritis by EndoS-hydrolyzed IgG, which is domi-nant and mediated through disturbances in the formation ofICs within the target tissue. This dominant suppression of in-flammation by EndoS-hydrolyzed IgG is a different and uniquetherapeutic effect of EndoS modification.

Results and DiscussionDominant Inhibition of Inflammation by EndoS-Hydrolyzed IgG. EndoStreatment specifically cleaved the Asn-297 glycan on IgG (Fig.1A), which removed almost all (99%) of the variable glycanchains attached to the first N-acetylglucosamine (GlcNAc) res-idue of the Fc region (Fig.1 C and D and Tables S1 and S2).Upon anti-CII mAb (EndoS-unhydrolyzed IgG) transfer, ar-thritis developed in mice as early as 48 h with 100% incidence atday 10.Massive infiltration of immune cells, pannus formation, anddistinct bone and cartilage erosions were observed in this group ofmice (Fig. 1B). Interestingly, injection of EndoS-hydrolyzed IgG,irrespective of CII epitope specificity, potently inhibited such

Author contributions: K.S.N., M.J.R., and R.H. designed research; K.S.N., M.C., K.E.H., A.M.C.,S.L.L., and A.M.B. performed research; R.A.Z. contributed new reagents/analytic tools; K.S.N.,M.C., K.E.H., A.M.C., S.L.L., M.J.R., and A.M.B. analyzed data; and K.S.N. wrote the paper.

Conflict of interest statement: Hansa Medical AB has filed patent applications for usingEndoS-modified IgG for treatment of arthritis and K.S.N., M.C., and R.H. are listed asinventors. The authors have no additional financial interests. The funders had no rolein the preparation of the manuscript or the decision to publish.

This article is a PNAS Direct Submission. R.A.D. is a guest editor invited by the Editorial Board.

See Commentary on page 10059.1To whom correspondence should be addressed. E-mail: Nandakumar.Kutty-Selva@ki.se.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1301480110/-/DCSupplemental.

10252–10257 | PNAS | June 18, 2013 | vol. 110 | no. 25 www.pnas.org/cgi/doi/10.1073/pnas.1301480110

See Retraction Published October 20, 2014

inflammatory arthritis (Figs. 1 B and E–J and 2A and B), resultingin normal joint architecture. Similarly, mice treated with a mixtureof unhydrolyzed and EndoS-hydrolyzed IgG showed undisturbedjoints. Surprisingly, mAbs to trinitrophenol (TNP) (Hy2.15) andto human HLA-DR (L243), which are commonly used as controlantibodies in the mouse, also completely inhibited arthritis whenEndoS-hydrolyzed (Fig. 1 F and G). Inhibition by EndoS-hydro-lyzed antibodies occurred irrespective of the IgG subclass or theantigen specificity (Fig. 1 G and H).When we analyzed the time and dose dependence of the ar-

thritis suppressive effect, we found that 1 mg of EndoS-hydro-lyzed IgG within the mixture of 9 mg of antibodies completelyinhibited arthritis (Fig. 1I) and 3 mg of anti-CII mAb was suffi-cient to induce arthritis (12), and when we titrated the EndoS-hydrolyzed IgG we showed the dose needed for complete in-hibition of arthritis was as little as 250 μg (Fig. 1J).To find the effective therapeutic time point, EndoS-hydrolyzed

IgG was administered to groups of mice at different time points

and one group ofmicewas left untreated.Antibodymixture and thenLPS were injected at 0 and 3 h, respectively. In the 0-h treatmentgroup, EndoS-hydrolyzed IgG was injected initially, followed by aninjection of mAb mixture. Unlike before or after 48h treatment,complete blocking of arthritis was observed when the treatment withEndoS-hydrolyzed IgGwas done 3 h before or after the arthritogenicmixture injection (Fig. 2 A and B). Because antibodies are bound tocartilage within 30 min after injection (16), we concluded thatEndoS-hydrolyzed IgG could block the disease most effectively ifinjected during the time when antibodies start binding to thecartilage surface.

EndoS Hydrolysis of IgG Does Not Affect Antigen Binding. Surfaceplasmon resonance (SPR) analysis of the antigen–antibody bindingin the presence or absence of EndoS-hydrolyzed IgG clearlydemonstrated that removal of carbohydrate moieties from mAbsdid not affect their high-affinity binding to CII epitopes (Fig. 2C).We have earlier shown antibody-mediated damage in cartilage

Fig. 1. EndoS-hydrolyzed IgG dominantly inhibitinflammation. (A) SDS/PAGE and lectin blot analysisof mAbs incubated with (+) or without (−) EndoShydrolysis and separated by 10% SDS/PAGE. Theproteins were detected by PageBlue stain (Stain) orby blotting onto a PVDF membrane probed withLens culinaris agglutinin (LCA). (B) Representativefigures of H&E-stained ankle joints of mice (n = 3–4per group) injected with anti-CII mAbs; un-hydrolyzed (Left), EndoS-hydrolyzed (Center), ormixed IgG (Right). Magnification ×10. (C) Hy2.15and (D) EndoS-treated Hy2.15. Shown spectra wereacquired during the time period for which the ma-jority of glycosylated peptides from EEQFNSTFR(21.5–23.0 min) elute. Doubly and triply chargedions as well as predicted glycan structures areshown. All numbers given are for the monoisotopicmass charge. In all of the animal experiments, male(BALB/c × B10.Q) F1 mice were used. Unless other-wise stated all of the mice received antibodies i.v.(d 0) and 25 μg of LPS i.p. (d 5). For arthritis in-duction in experiments shown in E, F, I and J, 9 mgof two anti-CII mAb mixtures (M2139 + CIIC1) wereused, whereas for experiments in G and H 4 mgof four anti-CII mAbmixture (M2139 + CIIC1 + CIIC2 +UL1) was used. Antigen specificity is not requiredfor inhibition. Mice (n = 42) were injected with 4 mgof EndoS-hydrolyzed IgG (E) M2139H + CIIC1H orUL1H + CIIC2H or (F) Hy2.15H + L243H followed byanti-CII mAb. Dose and subclass dependency. (G)Mice (n = 39) were injected with EndoS-hydrolyzedor unhydrolyzed IgG1 (Hy2.15) or IgG2a (L243) mAbbinding to joint unrelated antigens at two differentconcentrations (1 mg and 0.25 mg), followed byanti-CII mAb. (H) Mice (n = 65) were injected withdifferent subclasses of EndoS-hydrolyzed anti-CII(M284H, M2139H, CIIC1H, CIIC2H, and UL1H) or anti-citrullinated CII peptide IgG (ACC4H) at two dif-ferent concentrations (1 mg and 0.25 mg), followedby anti-CII mAb. (I) Mice (n = 25) were injected witha mixture of EndoS-hydrolyzed and/or unhydrolyzedanti-CII IgG at different combinations. In mixed IgGgroups, group 1 received 4.5 mg of unhydrolyzedand 4.5 mg of EndoS-hydrolyzed IgG, group 2 had6.75 mg of unhydrolyzed and 2.25 mg of EndoS-hy-drolyzed IgG, and group 3 received 7.8775 mgof unhydrolyzed and 1.125 mg of EndoS-hydro-lyzed IgG. (J) Mice (n = 25) were injected withdifferent concentrations (50–4,000 μg) of EndoS-hydrolyzed single anti-CII IgG (M2139H), followedby anti-CII mAb. Three hours after the antibodytransfer, LPS was injected. H denotes EndoS-hydrolyzed IgG. Hy2.15 and L243 represent mAbs binding to TNP hapten and human HLA-DR antigen, respectively. Error bars indicate ± SEM.

Nandakumar et al. PNAS | June 18, 2013 | vol. 110 | no. 25 | 10253

MED

ICALSC

IENCE

SSE

ECO

MMEN

TARYSee Retraction Published

October 20, 2014

explants cultured in vitro with the mAb to CII used in this study:This preinflammatory effect does not require living cells, is medi-ated by fragment antigen-binding (Fab), and is epitope-dependent(17). To test whether EndoS hydrolysis could change this effect,FTIR microspectroscopy (FTIRM) was used for chemical analysisof cartilage (17) cultured in the presence of 100 μg/mL EndoS-

hydrolyzed or unhydrolyzed mAb (Fig. 2 D–G). The height andlocation of the amide 1 peak, representing protein, predominantlycollagen, and the height of the proteoglycan peak at 1,076 cm–1

were examined at the cartilage surface, where the mAb penetrates,and in the interior of cartilage explants. After 14 d, the amide1 peak at the surface and in the interior of the control cartilage

Fig. 2. Inhibition of inflammation and SPR and FTIRM analysis. (A) Mice (n = 30) were injected with 1 mg of EndoS-hydrolyzed anti-CII IgG (M2139H + CIIC1H +CIIC2H+UL1H) at different time points (−48, −3, 0,+3, or+48 h). At 0 h and 3 h, anti-CII mAb (M2139+ CIIC1 + CIIC2+UL1) and then LPSwere injected. One groupof mice received no treatment. (B) Effect of splenectomy. Mice (n = 21) were either splenectomized (Splx) or sham-operated (Sham). Three weeks later, they wereinjected with 4mg of EndoS-hydrolyzed IgG (M2139H + CIICH) or left untreated, followed by anti-CII mAb (M2139+ CIIC1). H denotes EndoS-hydrolyzed IgG. Errorbars indicate ±SEM. (C) SPR (Biacore) analysis of antibody binding capacity of EndoS-hydrolyzed and unhydrolyzed IgG was performed using CII immobilized onCM5 sensor chip. MAbswere injected at different concentrations throughflow cells at a flow rate of 30 μL/min. Antibodies were injected for 3min and dissociationof bound molecules was observed for 7 min. There was no difference in antibody binding when EndoS-hydrolyzed or unhydrolyzed IgGs were added at differentratios to anti-CII mAbmixture. (D and E) Changes in the chemical composition of the cartilage were assessed using FTIRM analysis. Representativemean spectra areshown fromcartilage cultureswithout antibody (D), and from cartilage cultured for 14 dwith 100 μg/mLof unhydrolyzedmAbM2139 (E). The results shownare themean of 10 measurements taken from the central areas (red line) and near the surface of the tissue (black line). The mean spectra for surface and interior werecalculated to assess the effects of antibody penetration on the peaks characteristic of CII and of proteoglycans. (F andG) Themean peaks from the surface cartilagewere compared with those from antibody-exposed surface of cartilage exposed to the EndoS-hydrolyzed or unhydrolyzed IgG. Cartilage exposed to either EndoS-hydrolyzed or unhydrolyzed IgG (CIIC1, M2139, and UL1) showed similar changes. (F) The height and location of the amide 1 peak, which represents the totalprotein content of the tissue, in the region 1,600–1,700 cm–1. (G) The height of the peak at 1,076 cm−1 represents proteoglycans.

10254 | www.pnas.org/cgi/doi/10.1073/pnas.1301480110 Nandakumar et al.

See Retraction Published October 20, 2014

cultured without mAb was located at 1,659 cm−1 (range 1,655–1,666 cm–1), but there were striking changes in spectra from carti-lage cultured in the presence of either EndoS-hydrolyzed or un-hydrolyzed anti-CII IgG. Beyond the region of penetration bymAb,the spectra were generally similar to those of controls culturedwithout mAb, but spectra from the surface of the cartilage showedsubstantial changes (Fig. 2D). For both unhydrolyzed and EndoS-hydrolyzed anti-CII, there was a shift in the location of the amide1 peak from 1,659 cm–1 to as low as 1,643 cm–1, indicative of de-naturation of the CII, accompanied by substantial decreasesin the height of the amide 1 peak, and the proteoglycan peak at1,076 cm−1, indicating a total loss of matrix (Fig. 2E). These dataconfirm that the EndoS-hydrolyzed IgG retained its antibody re-activity and that the suppressive effect in fact resides in the Fc butnot in the Fab part of the IgG molecules.

Spleen Is Not Required for Arthritis Inhibition. Recent studies dem-onstrated the anti-inflammatory property of terminal sialic acidspresent on IgG–Fc (5). C-type lectin receptor SIGN-R1 (CD209)expressed on macrophages in the splenic marginal zone is requiredfor recognition of such sialic acids (18), which results in theproduction of IL-33 and expansion of IL-4–producing basophilspromoting increased expression of the inhibitory FcγRIIb on ef-fector macrophages leading to attenuation of inflammation (19).However, in the present study splenectomy did not alter the in-hibitory capacity of EndoS-hydrolyzed IgG (Fig. 2B), suggestinginvolvement of mechanisms other than the SIGN-R1 pathway.

EndoS-Hydrolyzed IgG Disrupts Larger IC Formation. Specific IgGglycan hydrolysis alters both murine and human IgG–FcγR inter-actions (13, 15) and removal of outer-arm sugar residues affects thethermal stability and functionality of the CH2 domains of IgGs (20).However, the length and nature of residual carbohydrate structurescould also affect Fc–Fc interactions and thereby IC formation,complement binding and FcR binding.Binding of the arthritogenic antibodies to the cartilage matrix

and the subsequent formation of ICs on the joint surface is likely tobe the major factor leading to the clinically apparent inflammationand arthritis. In the process of local IC formation, Fc–Fc inter-actions play an important role (21), and the specific glycanspresent in the CH2 domain of IgG might have a vital function inthis process. Because the dominant suppression seems to be me-diated through more acute effects during the binding of antibodiesto the cartilage, we hypothesized this might be due to the in-stability of ICs formed within the target tissue, the articular joints.Because the Fab part did not influence the suppressive effect,

the next step was therefore to analyze whether the EndoS-hydrolyzed antibodies can disrupt the growth of Fc-dependent ICs.For this we used the dynamic light scattering (DLS) techniqueand analyzed the formation of IC by CII and anti-CII mAb in thepresence of EndoS-hydrolyzed or unhydrolyzed IgG. As shownin Fig. 3 A and B, larger IC formation was clearly disturbed bythe presence of EndoS-hydrolyzed IgG. Interestingly, distur-bance of large IC was more prominent in the presence of low-affinity (Fig. 3A) than high-affinity IgG (Fig. 3B). It has earlierbeen shown that the binding to low-affinity Fc receptors is de-creased by EndoS modification (13). However, the binding isdirectly related to IC formation. Previously, we reported RF-likeactivity of one of the mAb (CIIC1) present in the arthritogenicmixture (22); however, we did not observe any difference in itsbinding activity to EndoS-hydrolyzed and unhydrolyzed IgG (Fig.S1). Different data have been reported regarding complementbinding after EndoS hydrolysis (13, 15). To directly investigatethis we analyzed complement binding using either immobilizedEndoS-hydrolyzed or unhydrolyzed IgG, or CII–anti-CII ICs.The presence of EndoS-hydrolyzed IgG did not interfere with theability of antibodies to stimulate complement deposition either

when antibodies were immobilized directly on the plate (Fig. S2)or when bound to immobilized CII (Fig. 3C).The presence of EndoS-hydrolyzed IgG did not affect com-

plement activation but IC stability in vitro; hence, we furtheranalyzed deposition of C3b, the activated product of comple-ment factor C3, on the cartilage of mice as a measure of ICdeposition and complement activation in vivo. Twenty-fourhours after the injection of EndoS-hydrolyzed, unhydrolyzed, ormixed anti-CII IgG, mouse paws were analyzed for the binding ofmAbs to cartilage using anti-kappa antibodies as well as fordeposited C3b. The mAbs readily bound to the cartilage surface,but the pattern of C3b deposition between the groups was en-tirely different (Fig. 3 D and E). Minimal C3b staining was ob-served, only on the subchondral bone junction area of the jointsfrom mice injected with EndoS-hydrolyzed IgG compared witha significant level of deposition throughout the cartilage surfacein the unhydrolyzed IgG-injected group. In mice injected withmixed IgG, staining on the subchondral bone junction area wasmore intense and staining on the cartilage surface was weak. Weconclude from these results that the presence of EndoS-hydro-lyzed IgG decreased the formation of larger ICs in situ, mostlikely through the disturbance of Fc–Fc interactions, and thisleads to decreased complement activation in the tissue. Thesignificance of Fc–Fc interactions in precipitation (23) and for-mation of insoluble ICs (21) are early wisdoms. In addition, di-nitrophenol-specific nonprecipitating antibodies were shown toinhibit as well as solubilize IC between antigen and precipitatingantibodies (24) that might involve Fc–Fc interactions. Later,structural evidence for such interactions involving the glycosyl-ation loop of one Fc-fragment dimer binding to the CH2–CH3interface of another Fc fragment has been demonstrated (25).Although oligosaccharides have been reported not to be involvedin direct contacts with symmetry-related molecules (25), theirinteractions with the protein moiety in the IgG–Fc region couldvery well affect the reciprocal influences on conformation (2).S. pyogenes secretes several enzymes and proteins that bind

and modulate the functions of Igs as a part of its strategy forevading the immune system. Disruption of the development oflarger IC lattices by EndoS-cleaved IgG could very well be onesuch strategy. Conversely, antibodies as a constituent of ICs playan important role in triggering various inflammatory processesleading to the development of a number of autoimmune dis-eases. Neutrophils play a vital part during this process, andsequential complement fixation generating C5a and direct en-gagement of Fcγ receptors are needed to initiate and sustain suchneutrophil recruitment in vivo and subsequent inflammation (26).Recent studies demonstrated bidirectional regulation of C5aR andFcγRs, which could significantly influence effector functions (27).Here we demonstrate that a specific modification of the N-linkedglycan of IgG by EndoS leads to a profound anti-inflammatoryeffect and it does not require injection of a bacterial protein forprotection. Disruption of larger IC formation at the target organsusing host antibodies with a customized glycan profile (Fig. 3F)could be a unique therapeutic possibility for patients with IgG-mediated inflammatory diseases.

Materials and MethodsEndoS Hydrolysis of IgG. IgGs (CIIC1, M2139, M284, CIIC2, UL1, CB20, ACC4,Hy2.15, and L243)were hydrolyzedwith recombinant EndoS fused toGST (GST–EndoS) as previously described (6). Five micrograms of GST–EndoS in PBS wasadded per milligram of mAb followed by incubation for 16 h at 37 °C. GST–EndoS was completely removed by three serial passages over Glutathione-Sepharose 4B columns with a 1,000-fold overcapacity of GST binding (GEHealthcare). SDS/PAGE and Lens culinaris agglutinin (LCA) lectin blotting wereused to assess the purity and efficacy of EndoS cleavage. Briefly, 2 μg of EndoS-hydrolyzed and unhydrolyzed IgG were separated on 10% SDS/PAGE followedby staining with PageBlue protein stain (ThermoFisher Scientific), or blotted toPVDF using TransBlot Turbo transfer packs and apparatus (Bio-Rad). Mem-branes were blocked with 10 mM Hepes (pH 7.5) with 0.15 M NaCl, 0.01 mM

Nandakumar et al. PNAS | June 18, 2013 | vol. 110 | no. 25 | 10255

MED

ICALSC

IENCE

SSE

ECO

MMEN

TARYSee Retraction Published

October 20, 2014

MnCl2, 0.1 mM CaCl2, and 0.1% Tween-20 (HBST) and incubated with 1 μg/mLof biotinylated LCA lectin (Vector Laboratories). After washing in HBST,membranes were incubated with 50 ng/mL of peroxidase-labeled streptavidin(Vector Laboratories) and developed using Super Signal West Pico Chemilu-minescent Substrate (ThermoFisher Scientific) and a ChemiDoc XRS imagingsystem (Bio-Rad).

Glycopeptide Identification. EndoS-hydrolyzed or unhydrolyzed antibody (15μg) were trypsin-digested using Protease MAX Surfactant and trypsin

enhancer (Promega). Samples were analyzed using a reversed-phase liquidchromatography system (Easy-nLC; Proxeon) connected to a Velos Orbitrapmass spectrometer (MS) (ThermoFisher Scientific). The MS was operated inpositive mode and the survey MS scan in the range of m/z 300–2,000 wasobtained at a resolution of 60,000. Following each MS scan, the top four mostabundant precursor ions were selected for MS/MS using collision-induceddissociation and electron-transfer dissociation fragmentation. IgG Fc glyco-peptides were identified in liquid chromatography-MS/MS datasets by theircharacteristic retention times and accurate monoisotopic masses (within <10

Fig. 3. Disturbance of stable ICs and complement activation. CII (1 mg/mL) and anti-CII mAb CB20 (low affinity, A) or (M2139 (high affinity, B) at 1 mg/mL weremixed together at 1:1 ratio and incubated for 30 min at 37 °C, followed by addition of anti-hapten IgG, either unhydrolyzed (Hy2.15) or EndoS-hydrolyzed IgG(Hy2.15H) at the ratio 1:1:1. Twentymicroliters of thismixturewere loadedonto the capillary tube in theDLS instrument. Relative sizes of ICs present in the solutionare indicated in the x-axis and the y-axis denotes percentage of ICs present in the solution. Each sample was measured five to seven times and the bars representmean values from two experiments. Error bars indicate ± SEM. (C) Complement activation on CII bound anti-CII antibodies were monitored by measuring C3bdeposition. Each bar represents mean values from three experiments ± SD. (D) Deposition of C3b on the cartilage of mice was used as a measure of IC depositionand complement activation after the injection of EndoS-hydrolyzed, unhydrolyzed, or mixed anti-CII IgG. Mouse pups (three or fourmice per group) were injectedwith 1mgeachof unhydrolyzed IgG (M2139+CIIC2+UL1), EndoS-hydrolyzed IgG (M2139H+CIIC2H+UL1H), or amixture of IgGs at 1:1 ratio. In each group, 26–44joints were scored in total. **P < 0.01; ***P < 0.005. Error bars indicate± SEM. (E) Paw samples collected 24 h later were stainedwith biotinylated anti-kappa (Left)or goat anti-mouse anti-C3c antibodies (Right). Joint sections frommice injected with PBS (first row), unhydrolyzed (second row), EndoS-hydrolyzed (third row), ora mixture (1:1) of IgG (fourth row) are shown. Magnification ×20. Arrows indicate C3b deposition within ICs formed on the joint cartilage surface. (F) Diagramillustrating possible binding mechanisms involved in the suppression of arthritis by EndoS-hydrolyzed antibodies.

10256 | www.pnas.org/cgi/doi/10.1073/pnas.1301480110 Nandakumar et al.

See Retraction Published October 20, 2014

ppm from the theoretical values) of doubly and triply charged ions fromM2139: EDYNSTIR, CIIC1, and L243: EDYNSTLR as well as Hy2.15: EEQFNSTFR,respectively. Protein identity was confirmed using MASCOT search engine(version 2.3.2) using International Protein Index mouse concatenated data-base. Search parameters were as follows: MS mass error tolerance at 10 ppm,MS/MS mass accuracy at 0.5 Da, tryptic digestion with a maximum of twomissed cleavages, carbamidomethylation of cysteine as a fixed modification,asparagine and glutamine deamidation, and methionine oxidation as well asN-glycosylation (HexNAc[n]dHex[n]Hex[n]) as variable modifications.

Particle Size Measurement Using DLS. DLS is a useful tool to detect and confirmthe formation of a protein complex. Disturbance of stable IC formation by EndoS-hydrolyzed IgGswas carriedoutusing theDLS technique. Briefly, CII purified fromrat chondrosarcoma dissolved in 0.1M acetic acid at 5mg/mLwas diluted furtherin PBS (1 mg/mL) and anti-CII mAb (high-affinity M2139 or low-affinity CB20;1mg/mL) in PBS were mixed together at 1:1 ratio and incubated for 30 min at37 °C, followed by addition of either unhydrolyzed (Hy2.15) or EndoS-hydro-lyzed (Hy2.15H) anti-hapten IgG at the ratio 1:1:1. Twenty microliters of thismixture were loaded onto the capillary tube in the DLS instrument (PrecisionDetectors Inc.). The particle sizes expressed as the apparent Z-average (or in-tensity-weighted) hydrodynamic diameter and polydispersity index, which pro-vides information on the deviation from monodispersity, were measured (28)and compared between the groups.

SPR Analysis. SPR (Biacore 2000; Biacore) analysis was performed using thestandard procedure (29). Briefly, CII was immobilized on the surface of CM5sensor chips. EndoS-hydrolyzed and unhydrolyzed IgGs were injected at dif-ferent concentrations through flow cells in the running buffer [10 mM Hepes(pH 7.4), 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20] at a flowrate of 30 μL/min. Antibodies were injected for 3 min and dissociation ofbound molecules was observed for 7 min. Background binding to control flowcells was subtracted automatically. The chips were regenerated using pulseinjection of 100% ethylene glycol followed by 2 M NaCl and 100 mM HCl.

Analysis of the in Vitro Effects of mAbs Using Bovine Cartilage Explants. Ar-ticular cartilage samples extracted from adult bovine metacarpophalangealjoints and cartilage shavings (5 × 5 × 1 mm) were cultured for up to 14 d in

DMEM with 20% (vol/vol) FCS and 25 μg/mL ascorbic acid, with 100 μg/mL ofmAb or in medium alone. Medium was changed every 2 d, and fresh ascorbicacid and mAb were added at each change. Cartilage samples were tested induplicate, and all experiments were performed at least twice. On day 14, car-tilage explants were fixed in 4% (wt/vol) paraformaldehyde and embedded inparaffin for FTIRM. Sections (5 μm) were placed onto MirrIR low-e microscopeslides (Kevley Technologies), and adjacent sections were stained with toluidineblue. FTIR images were recorded with a Stingray Digilab FTS 7000 series spec-trometer coupled to a UMA 600 microscope equipped with a 64 × 64 focalplane array detector. For each spectrum, 16 scans were coadded at a resolutionof 6 cm–1. The spectra were analyzed using CytoSpec imaging software (17).Raw chemical maps were generated from the integrated intensities of specificfunctional groups identified in the spectra, and 10 spectra from the surface ofthe explant and 10 from the interior were extracted from the raw chemicalmaps. The mean spectra for “surface” and “interior” were calculated to assessthe effects of antibody penetration on the peaks characteristic of CII and pro-teoglycans. Analysis was performed on the location of the amide 1 peak (1,640–1,670 cm–1), which represents total protein, which for cartilage is primarily CII.For proteoglycans, analysis was based on the height of the peak at 1,076 cm–1,within the region of 1,175–960 cm–1 derived from carbohydrate moieties.

Statistical Analyses.All of themice with arthritis were included for calculationof severity. The severity of arthritis was analyzed by Mann–Whitney U testand the incidence by chi square test using Statview (version 5.0.1). Signifi-cance was considered when P < 0.05, for a 95% confidence interval.

ACKNOWLEDGMENTS. We thank Dr. Rajesh Ponnusamy for his help inmeasuring the sizes of ICs by DLS, Emma Mondoc for performing histology,and Carlos Palestro for taking care of animals. This study was supported bygrants from King Gustaf V:s 80 Years Foundation, Swedish RheumatismAssociation, Åke Wiberg, Alfred Österlund, Petrus and Augusta Hedlund,Clas Groschinsky, Torsten och Ragnar Söderberg, the Swedish Society forMedicine, the Royal Physiografic Society in Lund, the Medical Faculty at LundUniversity, Swedish governmental funding for clinical research (A.L.F.),Hansa Medical AB, Swedish Research Council Grants 2009-2338, 2010-57X-20240, and K2012-66X-14928-09-5, European Union MasterSwitch projectGrant HEALTH-F2-2008-223404, and the National Health and Medical Re-search Council of Australia, an Arthritis Australia Project grant.

1. Ferrara C, et al. (2011) Unique carbohydrate-carbohydrate interactions are requiredfor high affinity binding between FcgammaRIII and antibodies lacking core fucose.Proc Natl Acad Sci USA 108(31):12669–12674.

2. Deisenhofer J (1981) Crystallographic refinement and atomic models of a human Fcfragment and its complex with fragment B of protein A from Staphylococcus aureusat 2.9- and 2.8-A resolution. Biochemistry 20(9):2361–2370.

3. Barb AW, Prestegard JH (2011) NMR analysis demonstrates immunoglobulin G N-glycansare accessible and dynamic. Nat Chem Biol 7(3):147–153.

4. Parekh RB, et al. (1985) Association of rheumatoid arthritis and primary osteoarthritiswith changes in the glycosylation pattern of total serum IgG.Nature 316(6027):452–457.

5. Kaneko Y, Nimmerjahn F, Ravetch JV (2006) Anti-inflammatory activity of immuno-globulin G resulting from Fc sialylation. Science 313(5787):670–673.

6. Collin M, Olsén A (2001) EndoS, a novel secreted protein from Streptococcus pyogeneswith endoglycosidase activity on human IgG. EMBO J 20(12):3046–3055.

7. Rantapää-Dahlqvist S, et al. (2003) Antibodies against cyclic citrullinated peptide andIgA rheumatoid factor predict the development of rheumatoid arthritis. ArthritisRheum 48(10):2741–2749.

8. Cochrane CG, Hawkins D (1968) Studies on circulating immune complexes. 3. Factorsgoverning the ability of circulating complexes to localize in blood vessels. J Exp Med127(1):137–154.

9. Mageed RA, Kirwan JR, Thompson PW, McCarthy DA, Holborow EJ (1991) Charac-terisation of the size and composition of circulating immune complexes in patientswith rheumatoid arthritis. Ann Rheum Dis 50(4):231–236.

10. Mullazehi M, Mathsson L, Lampa J, Rönnelid J (2006) Surface-bound anti-type II col-lagen-containing immune complexes induce production of tumor necrosis factor al-pha, interleukin-1beta, and interleukin-8 from peripheral blood monocytes via Fcgamma receptor IIA: A potential pathophysiologic mechanism for humoral anti-typeII collagen immunity in arthritis. Arthritis Rheum 54(6):1759–1771.

11. Wooley PH, et al. (1984) Passive transfer of arthritis to mice by injection of humananti-type II collagen antibody. Mayo Clin Proc 59(11):737–743.

12. Nandakumar KS, Svensson L, Holmdahl R (2003) Collagen type II-specific monoclonalantibody-induced arthritis in mice: description of the disease and the influence ofage, sex, and genes. Am J Pathol 163(5):1827–1837.

13. Nandakumar KS, et al. (2007) Endoglycosidase treatment abrogates IgG arthritoge-nicity: importance of IgG glycosylation in arthritis. Eur J Immunol 37(10):2973–2982.

14. Lood C, et al. (2012) IgG glycan hydrolysis by endoglycosidase S diminishes theproinflammatory properties of immune complexes from patients with systemic lupuserythematosus: A possible new treatment? Arthritis Rheum 64(8):2698–2706.

15. Albert H, Collin M, Dudziak D, Ravetch JV, Nimmerjahn F (2008) In vivo enzymaticmodulation of IgG glycosylation inhibits autoimmune disease in an IgG subclass-dependent manner. Proc Natl Acad Sci USA 105(39):15005–15009.

16. Holmdahl R, Mo JA, Jonsson R, Karlstrom K, Scheynius A (1991) Multiple epitopes on

cartilage type II collagen are accessible for antibody binding in vivo. Autoimmunity

10(1):27–34.17. Amirahmadi SF, et al. (2004) An arthritogenic monoclonal antibody to type II colla-

gen, CII-C1, impairs cartilage formation by cultured chondrocytes. Immunol Cell Biol

82(4):427–434.18. Anthony RM, Wermeling F, Karlsson MC, Ravetch JV (2008) Identification of a re-

ceptor required for the anti-inflammatory activity of IVIG. Proc Natl Acad Sci USA

105(50):19571–19578.19. Anthony RM, Kobayashi T,Wermeling F, Ravetch JV (2011) Intravenous gammaglobulin

suppresses inflammation through a novel T(H)2 pathway. Nature 475(7354):110–113.20. Mimura Y, et al. (2000) The influence of glycosylation on the thermal stability and

effector function expression of human IgG1-Fc: Properties of a series of truncated

glycoforms. Mol Immunol 37(12-13):697–706.21. Easterbrook-Smith SB, Vandenberg RJ, Alden JR (1988) The role of Fc:Fc interactions

in insoluble immune complex formation and complement activation. Mol Immunol

25(12):1331–1337.22. Uysal H, et al. (2008) The crystal structure of the pathogenic collagen type II-specific

mouse monoclonal antibody CIIC1 Fab: Structure to function analysis. Mol Immunol

45(8):2196–2204.23. Møller NP (1979) Fc-mediated immune precipitation. I. A new role of the Fc-portion of

IgG. Immunology 38(3):631–640.24. Cosio FG, Birmingham DJ, Sexton DJ, Hebert LA (1987) Interactions between pre-

cipitating and nonprecipitating antibodies in the formation of immune complexes.

J Immunol 138(8):2587–2592.25. Kolenko P, et al. (2009) New insights into intra- and intermolecular interactions of

immunoglobulins: Crystal structure of mouse IgG2b-Fc at 2.1-A resolution. Immu-

nology 126(3):378–385.26. Sadik CD, Kim ND, Iwakura Y, Luster AD (2012) Neutrophils orchestrate their own

recruitment in murine arthritis through C5aR and FcγR signaling. Proc Natl Acad Sci

USA 109(46):E3177–E3185.27. Karsten CM, Köhl J (2012) The immunoglobulin, IgG Fc receptor and complement

triangle in autoimmune diseases. Immunobiology 217(11):1067–1079.28. Long Y, Philip JY, Schillén K, Liu F, Ye L (2011) Insight into molecular imprinting in

precipitation polymerization systems using solution NMR and dynamic light scatter-

ing. J Mol Recognit 24(4):619–630.29. Sjöberg AP, et al. (2007) The factor H variant associated with age-related macular de-

generation (His-384) and the non-disease-associated formbind differentially to C-reactive

protein, fibromodulin, DNA, and necrotic cells. J Biol Chem 282(15):10894–10900.

Nandakumar et al. PNAS | June 18, 2013 | vol. 110 | no. 25 | 10257

MED

ICALSC

IENCE

SSE

ECO

MMEN

TARYSee Retraction Published

October 20, 2014