Engineering antigens for in situ erythrocyte bindinginduces T-cell deletionStephan Kontosa,1, Iraklis C. Kourtisa, Karen Y. Danea, and Jeffrey A. Hubbella,b,c,2

aInstitute of Bioengineering, School of Life Sciences and School of Engineering, and bInstitute of Chemical Sciences and Engineering, School of Basic Sciences,Ecole Polytechnique Fédérale Lausanne, CH-1015 Lausanne, Switzerland; and cDiabetes Research Institute, Miller School of Medicine, University of Miami,Miami, FL 33136

Edited by George Georgiou, University of Texas at Austin, Austin, TX, and accepted by the Editorial Board November 17, 2012 (received for reviewSeptember 26, 2012)

Antigens derived from apoptotic cell debris can drive clonal T-celldeletion or anergy, and antigens chemically coupled ex vivo toapoptotic cell surfaces have been shown correspondingly to inducetolerance on infusion. Reasoning that a large number of erythro-cytes become apoptotic (eryptotic) and are cleared each day, weengineered twodifferent antigen constructs to target the antigen toerythrocyte cell surfaces after i.v. injection, one using a conjugatewith an erythrocyte-binding peptide and another using a fusionwith an antibody fragment, both targeting the erythrocyte-specificcell surface marker glycophorin A. Here, we show that erythrocyte-binding antigen is collectedmuchmore efficiently than free antigenby splenic and hepatic immune cell populations and hepatocytes,and that it induces antigen-specific deletional responses in CD4+ andCD8+ T cells. We further validated T-cell deletion driven by erythro-cyte-binding antigens using a transgenic islet β cell-reactive CD4+

T-cell adoptive transfer model of autoimmune type 1 diabetes:Treatmentwith the peptide antigen fused to anerythrocyte-bindingantibody fragment completely prevented diabetes onset induced bythe activated, autoreactive CD4+ T cells. Thus, we report a translat-able modular biomolecular approach with which to engineer anti-gens for targeted binding to erythrocyte cell surfaces to induceantigen-specific CD4+ and CD8+ T-cell deletion toward exogenousantigens and autoantigens.

autoimmunity | immune tolerance | BDC2.5 | TER119 |immunoengineering

Strategies to induce antigen-specific tolerance without the needfor systemic immunosuppression have great potential in pre-

venting immune responses in protein replacement therapies, graft-versus-host disease (GvHD), and autoimmune disorders (1). Forexample, in the treatment of diseases requiring replacement pro-tein administration, such as hemophilia, some patients mountimmune responses to the therapeutic protein antigen, renderingtreatment ineffective and dangerous (2). In autoimmune diseases,where peripheral tolerance mechanisms fail to inactivate autor-eactive T cells that have escaped central tolerance induction in thethymus, deletion or functional inactivation to clonal anergy is con-sidered key to nonprogression of the disease (3). Immunoregulatorymodalities capable of prophylactically educating or therapeuticallyreeducating the immune system to induce antigen-specific toleranceare greatly needed for these and many other situations.Apoptotic cells are well known as a source of tolerogenic anti-

gens, although the exact mechanisms are not fully understood (4,5). Literature evidence maintains that antigen displayed in a phys-iological noninflammatory manner by apoptotic cells to antigen-presenting cells (APCs) fails to induce a productive immune re-sponse, because Toll-like receptor signaling is inhibited by uptakeof apoptotic cellular debris (6), leading to antigen-specific T-celldeletion or anergy (7).The concept of engineering apoptotic cell carriers to induce

immunological tolerance has been explored. Notably, antigenschemically coupled ex vivo to splenocyte cell surfaces in a mannerso as to induce apoptosis in the splenocytes has been shown toinduce antigen-specific tolerance on infusion in such contexts as

autoimmunity in type 1 diabetes (8) and myocarditis (9), trans-plant tolerance (10), and allergy (11).Antigen has also been loadedosmotically into splenocytes, with the osmotic shock inducingapoptosis, in the context of tolerance to the model antigen ov-albumin (OVA) (12). Although results obtained with these cel-lular approaches have been remarkable, the complexities oftranslating such a cellular therapeutic approach to widespreadclinical use are formidable; despite these challenges, a phase I/IIA trial has been carried out with peripheral bloodmononuclearcells pulsed with multiple peptide epitopes in the context of re-lapsing-remitting multiple sclerosis (European Union DrugRegulating Authorities Clinical Trials no. 2008-004408-29). Forease of translation to widespread clinical use, an acellular tol-erogenic therapeutic approach composed only of simpler definedbiomolecules would be advantageous. For this, we turned toengineering the antigen to enable in situ cell surface binding aftersimple i.v. injection.We reasoned that the large number of erythrocytes cleared

after apoptosis-like programmed cell death (eryptosis) each day(more than 1% per day in humans and 4% in mice) (13) couldbe exploited to induce antigen-specific T-cell deletion. Althoughthe exact triggers of erythrocyte clearance remain unclear, eryp-totic erythrocytes are characterized by phosphatidylserine asym-metry,membraneheterogeneity, and annexin-Vbinding, analogousto apoptotic nucleated cells (13, 14). Rather than chemically con-jugate erythrocytes to an antigen ex vivo, as has been mentioned(15), we sought to develop a targeted antigen formulation thatwould spontaneously bind strongly and specifically to erythrocytesafter simple i.v. injection for eventual display to APCs, along withapoptotic markers, as the eryptotic erythrocyte fraction is clearedover the ensuing days.

ResultsCreation and Characterization of Erythrocyte-Binding OVA. As oneapproach to obtain strong and specific biophysical targeting of anantigen to erythrocyte cell surfaces in situ, we used a synthetic12-aa peptide (ERY1) that we discovered by phage display tobind to mouse glycophorin-A specifically (16), which is presentuniquely on erythrocytes. We used the model antigen OVA, so

Author contributions: S.K., K.Y.D., and J.A.H. designed research; S.K., I.C.K., and K.Y.D.performed research; S.K. and I.C.K. analyzed data; and S.K. and J.A.H. wrote the paper.

Conflict of interest statement: The Ecole Polytechnique Fédérale Lausanne has filed forpatent protection on the technology described here, and S.K., K.Y.D., and J.A.H. arenamed as inventors on those patents; J.A.H. is a shareholder in a company that haslicensed those patents.

This article is a PNAS Direct Submission. G.G. is a guest editor invited by theEditorial Board.

Freely available online through the PNAS open access option.1Present address: Anokion SA, CH-1015 Lausanne, Switzerland.2To whom correspondence should be addressed. E-mail: jeffrey.hubbell@epfl.ch.

See Author Summary on page 17 (volume 110, number 1).

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1216353110/-/DCSupplemental.

E60–E68 | PNAS | Published online December 17, 2012 www.pnas.org/cgi/doi/10.1073/pnas.1216353110

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

as to enable initial analysis in a transgenic mouse strain (OTI),the CD8+ T-cell population of which expresses the T-cell re-ceptor specific for the MHC I immunodominant OVA epitope,SIINFEKL. We chemically conjugated approximately three cop-ies of the ERY1 peptide to OVA to create an OVA variant(ERY1-OVA) that binds mouse erythrocytes with high affinityand specificity (Fig. 1A). High-resolution confocal microscopyshowed localization to the cell membrane equatorial periphery(Fig. 1B), and flow cytometric analysis demonstrated binding to besequence-specific (Fig. 1C). ERY1-OVA bound to erythrocyteswith high affinity, exhibiting a dissociation constant (Kd) of 6.2 ±1.3 nM, as determined by equilibrium binding measurements (Fig.1D). Based on this approximate value, an entire 10 μg dose ofERY1-OVA would be bound to the surface of erythrocytes fol-lowing i.v. injection in a mouse.We validated ERY1-OVA binding in vivo to circulating eryth-

rocytes following i.v. administration inmice.Whole-blood samplestaken 30min following injection of 150 μg of eitherOVAorERY1-OVA confirmed the specific erythrocyte binding capability ofERY1-OVA even amid the complex heterogeneous milieu ofblood and the vasculature (Fig. 2A). Consistent with glycophorin-A association, ERY1-OVA bound to erythrocytes (CD45−) butnot to leukocytes (CD45+). ERY1-OVA binding was unbiased asto the eryptotic state of the erythrocytes, binding strongly toboth annexin-V+ and annexin-V− CD45− populations (Fig. 2B).Pharmacokinetic studies of the OVA conjugate demonstratedthat ERY1-OVA erythrocyte binding was long-lived in vivo,exhibiting a cell surface t1/2 of 17.2 h (Fig. 2C). ERY1-OVAremained bound to erythrocytes for as long as 72 h followingadministration; during this time frame, ∼13% of erythrocytes arecleared in the mouse (17). Quantification of erythrocyte-bound

ERY1-OVA in vivo showed a relatively high loading of 1.174 ±0.005 ng of OVA per 106 erythrocytes. Erythrocyte binding byERY1-OVA elicited no detectable differences in hematocrit,corpuscular volume, or hemoglobin content, compared with ad-ministration of OVA (Fig. S1), demonstrating that glycophorin-A–mediated erythrocyte binding with antigen did not alter theirhematological parameters.

Erythrocyte Bound Antigen Uptake and Biodistribution. To reveal thecellular targets of antigen as it is collected after binding to eryth-rocytes after administration, we i.v. injected mice with the highlyfluorescent allophycocyanin protein, conjugated to either ERY1(ERY1-allophycocyanin) or a mismatch (MIS) scrambled peptide

100 101 102 1030

20

40

60

80

100

OVA binding

% o

f max

ERY1-OVA

SMCC-OVA

MIS-OVA

A OVA

1. SMCC2. ERY1-Cys

ERY1

ERY1ERY1

ERY1-OVA

erythrocyte

glycophorin-AERY1

ERY1

ERY1

C

OVA

ERY1-OVA

GY

PAO

VA

0 50 100 150 2000

20

40

60

80

100

ERY1-OVA concentration (nM)

K = 6.2 ± 1.3 nMd

OVA

geo

met

ric m

ean

fluor

esce

nce

inte

nsity

GY

PAO

VA

B

D

Fig. 1. ERY1-OVA binds the equatorial periphery of mouse erythrocyteswith high affinity. (A) Schematic of conjugation of ERY1 peptide to OVA,resulting in binding to erythrocyte surface glycophorin-A. (B) High-resolu-tion confocal microscopy images of mouse erythrocytes labeled ex vivo with(green) anti-mouse glycophorin-A (GYPA) and (red) either OVA (Upper) orERY1-OVA (Lower). (Scale bar = 5 μm.) (C) Binding of each OVA conjugateand intermediate, characterized by flow cytometry. ERY1, erythrocyte-binding peptide WMVLPWLPGTLD; MIS, MIS peptide PLLTVGMDLWPW;SMCC, sulfosuccinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate,used to conjugate ERY1 to OVA. (D) Equilibrium binding of ERY1-OVA toerythrocytes demonstrates the low dissociation constant of ERY1-OVA (R2 =0.97, one-site binding), determined by flow cytometry.

A

0.11 3.75e-3

0.199.8

1.96e-3 0.25

990.73

OVA bindingA

nnex

in-V

ERY1-OVAOVA

OVA

con

cent

ratio

n(n

g/10

ery

thro

cyte

s)6

OVA binding

CD45+ CD45-

ERY1-OVA

OVA

naive naive

OVA

ERY1-OVA

0

50

100

150

Time post injection (h)

ERY1-OVAOVA

t1/2 = 17.2 h

OVA geom

etric mean

fluorescence intensity

20 40 60 80 170150

0.5

1.0

1.5

2.0

0

B

C

0 101 102 1030 101 102 103

0 101 102 103 0 101 102 1030

101

102

103

Fig. 2. ERY1-conjugated antigen biospecifically binds circulating healthy anderyptotic erythrocytes on i.v. administration, inducing uptake by specific APCsubsets. (A) OVA (gray-filled histogram) and ERY1-OVA (black-filled histo-gram) binding to erythrocyte (CD45−) and nonbinding to leukocyte (CD45+)populations in vivo compared with noninjected mice (empty histogram),determined by flow cytometry. (B) ERY1-OVA binding and OVA nonbindingto circulating eryptotic (annexin-V+) and healthy (annexin-V−) erythrocytes,determined by flow cytometry. (C) Cell surface t1/2 of bound ERY1-OVA tocirculating erythrocytes, determined by geometric mean fluorescence in-tensity of flow cytometry measurements (n = 2; R2 = 0.98, one-phase expo-nential decay), and time-dependent ERY1-OVA cell surface concentration,determined by ELISA, at an administered dose of 150 μg (n = 2).

Kontos et al. PNAS | Published online December 17, 2012 | E61

IMMUNOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

(MIS-allophycocyanin) (binding shown in Fig. S2A). Flow cyto-metric analysis of splenic dendritic cell (DC) populations 12 and36 h following administration showed 9.4-fold enhanced uptakeof ERY1-allophycocyanin by MHC II+ CD11b− CD11c+ DCscompared with MIS-allophycocyanin yet similar uptake of ERY1-allophycocyanin and MIS-allophycocyanin by MHC II+ CD11b+

CD11c+ DCs (Fig. 3A). Additionally, MHC II+ CD8α+ CD11c+

CD205+ splenicDCswere found to uptakeERY1-allophycocyaninto a threefold greater extent than MIS-allophycocyanin, althoughthe absolute magnitude was markedly lower than for other DCpopulations in the spleen. Such targeting of antigen toward non-activated and CD8α+ CD205+ splenic DCs could strengthen thetolerogenic potential of erythrocyte binding, because these pop-ulations have been extensively implicated in apoptotic cell-driventolerogenesis (7, 18). Such flow cytometry-based results were vali-dated using confocal microscopy. Spleen sections taken from mice24 h following i.v. injection of OVA or ERY1-OVA showed OVAstaining in areas distinct from F4/80+ macrophage regions, withERY1-OVA–injected mice displaying more overall OVA stainingthan OVA-injected mice (Fig. 3C). In the liver, ERY1-allophyco-cyanin also greatly enhanced uptake by hepatocytes (CD45−MHCIIlow CD1d−, by 78.4-fold) and hepatic stellate cells (CD45− MHCII+ CD1d+, by 60.6-fold) compared with MIS-allophycocyanin(Fig. 3C); both populations have been reported as APCs thattrigger CD8+ T-cell deletional tolerance (19–21). Interestingly,preferential uptake was not seen in liver DCs (CD45+ CD11c+) or

Kupffer cells (CD45+ MHC II+ F4/80+), which serve as membersof the reticuloendothelial system that aids in clearance of eryth-rocytes and complement-coated particles. Analysis of liver sec-tions confirmed the flow cytometry measurements. Although nointact-OVA staining was seen in the liver 24 h following injection(perhaps due to antigen processing), only ERY1-OVA–treatedmouse livers stained strongly with the 25.d1.16 mAb, which specif-ically recognizes the immunodominant OVA peptide SIINFEKLdisplayed on H-2kb MHC I (Fig. 3D). In addition to supportingobservations of hepatic cell uptake of ERY1-allophycocyanin, thepresence of MHC I–SIINFEKL complexes on hepatic cell surfacessuggests their role in cross-presentation of exogenous antigens.

T-Cell Phenotype in Response to ERY1- and Single-Chain Fv AntigenConstructs. We next determined whether erythrocyte-bound an-tigen could be efficiently cross-presented and could corre-spondingly cross-prime reactive CD8+ T cells. We adoptivelytransferred carboxyfluorescein succinimidyl ester (CFSE)-labeledOTI CD8+ T cells (CD45.2+) into CD45.1+ mice and measuredproliferation of the OTI CD8+ T cells 5 d following i.v. adminis-tration of 10 μg of OVA; 10 μg of ERY1-OVA; or 10 μg of an ir-relevant erythrocyte-binding antigen, ERY1-GST (binding shownin Fig. S2B). OTI CD8+ T-cell proliferation, determined by di-lution of the fluorophore CFSE as measured by flow cytometry(Fig. 4A), was markedly enhanced over threefold in mice adminis-tered ERY1-OVA compared with OVA (Fig. 4B), demonstrating

5

10

0

ERY1-allophycocyanin

MIS-allophycocyanin

spleen liver

12 h 36 h 12 h 36 h

A

MHCII+ CD11b-CD11c+

MHCII+ CD11b+CD11c+

MHCII+ CD8α+CD11c+ CD205+

12 h

5

30

40

50

60

70

CD45- MHCIIlow

CD1d-CD45- MHCII+

CD1d+

5

10

CD45+CD11c+

CD45+ MHCII+F4/80+

* *

*

***

**

B

ERY1-allophycocyanin

MIS-allophycocyanin

Allo

phyc

ocya

nin+

cel

ls(%

of e

ach

subs

et)

Allo

phyc

ocya

nin+

cel

ls(%

of e

ach

subs

et)

DA

PI

F4/8

0O

VA

DA

PI

CD

45H

-2K

b-S

IINFE

KL

OVA ERY1-OVA OVA ERY1-OVAC D liverspleen

Fig. 3. Erythrocyte-bound allophycocyanin uptake by splenic DC subsets and nonprofessional APCs in the liver. (A) Increased cellular uptake of ERY1-allo-phycocyanin by MHC II+ CD11b− CD11c+ and MHC II+ CD8α+ CD11c+ CD205+ splenic DCs at 12 and 36 h postinjection, compared with MIS-allophycocyanin. (B)Increased cellular uptake of ERY1-allophycocyanin in the liver by hepatocytes (CD45− MHCIIlow CD1d−) and hepatic stellate cells (CD45− MHC II+ CD1d+) but notby liver DCs (CD45+ CD11c+) or Kupffer cells (CD45+ MHC II+ F4/80+), compared with MIS-allophycocyanin, 36 h following i.v. administration (n = 2). *P ≤ 0.05;**P ≤ 0.01; ***P ≤ 0.001. Data represent mean ± SE. (C) Spleenmicroscopy images of mice 24 h following administration of 10 μg OVA (Left) or ERY1-OVA (Right),stained for OVA (green), F4/80 (red), and DAPI nuclear staining (blue). (Scale bar = 50 μm.) (D) Liver microscopy images of mice 24 h following administration of10 μg OVA (Left) or ERY1-OVA (Right), stained for MHC I H-2Kb-SIINFEKL (green), CD45 (red), and DAPI for nuclear staining (blue). (Scale bar = 50 μm.)

E62 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

that erythrocyte binding increased antigen-specific CD8+ T-cellcross-priming compared with the soluble antigen. Similar resultswere also obtained by administration of a 10-fold lower antigendose of 1 μg, demonstrating the wide dynamic range of efficacy ofOTI CD8+ T-cell proliferation induced by ERY1-conjugatedantigen.To address the technical hurdle of conjugating ERY1 to small

peptide antigens insufficiently large for conjugation of multiplecopies of the ERY1 peptide, we engineered a second molecularimplementation of our concept, namely, an erythrocyte-specificsingle-chain Fv (scFv) antibody fragment to display peptide anti-gens. The murine glycophorin A-binding TER119 mAb was se-quenced and converted into the scFv format with a C-terminalSIINFEKL peptide fusion. We characterized the resulting eryth-rocyte-binding peptide fusion, TER119-SIINFEKL (binding shownin Fig. S2C), in an analogous in vivo OTI proliferation study aswith ERY1-OVA, with even more striking cross-priming results(Fig. 4C). TER119-SIINFEKL induced proliferation of the entire

(95%)OTICD8+T-cell population 5 d following adoptive transfer,indicating strong proteosomal processing and cross-presentation ofthe peptide antigen when fused to the erythrocyte-binding TER119scFv (Fig. 4D). Although administration of soluble SIINFEKL atan equimolar dose induced similar proliferation rates, it is im-portant to note that in this format, SIINFEKL can directly bind toMHC I and present to OTI CD8+ T cells without first being pro-teolytically processed by APCs. Our results on cross-presentationand cross-priming are consistent with those of other studies con-cerning tolerogenic antigen presentation on MHC I by APCsengulfing antigen from apoptotic cell debris (4, 22).To distinguish T cells being expanded into a functional effector

phenotype from those being expanded and deleted, we analyzedthe proliferating OTI CD8+ T cells for programmed death-1 (PD-1), a marker of T-cell exhaustion, and annexin-V binding, a hall-mark of apoptosis, and thus deletion. ERY1-OVA induced muchhigher total numbers of PD-1+ (23.2 ± 4.9%) and annexin-V+

(20.7 ± 7.4%) proliferating OTI CD8+ T cells 4 d after treatment

CFSEC

D8α

A ERY1-GST OVA ERY1-OVA

saline SIINFEKL TER119-SIINFEKL

CFSE

CD

8α

0

101

102

103

0 101 102 1030 101 102 1030 101 102 103

0 101 102 1030 101 102 1030 101 102 103

0

101

102

103

99 972

44112

***23.2±4.9

7.6±1.1

###43.1±10

8.3±3.6

OVA

ERY1-OVA

SIINFEKL

TER119-SIINFEKL

012345678Generation

PD-1+ OTI T cells (% of OTI)

012345678Generation

Annexin-V+ OTI T cells (% of OTI) 7%

0%

***20.4±7.4

4.8±1.3

###35.7±9.1

7.5±2.9

OVA

ERY1-OVA

SIINFEKL

TER119-SIINFEKL

10%

0%

saline OVA ERY1-OVA

Pro

lifer

atin

g O

TI T

cel

ls(%

of O

TI)

B

ERY1-GST

20

10

40

30

**

##

C D

E

25

50

75

100

saline TER119-SIINFEKL

SIINFEKL

Pro

lifer

atin

g O

TI T

cel

ls(%

of O

TI)

1 μg10 μg

Total

TotalF

Fig. 4. Erythrocyte-binding antigen formulations enhance cross-priming and apoptotic fate deletional proliferation of antigen-specific OTI CD8+ T cells invivo. (A) Proliferation of CFSE-labeled splenic OTI CD8+ T cells (CD3e+ CD8α+ CD45.2+) 5 d following i.v. administration of 10 μg of ERY1-GST (Left), 10 μg ofOVA (Center), or 10 μg of ERY1-OVA (Right). (B) Dose-dependent quantified proliferative populations of OTI CD8+ T-cell proliferation from A, as well as anidentical 1-μg dosing study; data represent mean ± SD (n = 5). **P < 0.01; ##P < 0.01. (C) Proliferation of CFSE-labeled OTI CD8+ T cells 5 d following i.v.injection of 7 μg of TER119-SIINFEKL (Right), an equimolar dose of SIINFEKL (Center), or saline (Left). (D) Quantified proliferative populations of OTI CD8+

T cells from C. (E) Heat map representing OTI CD8+ T-cell proliferation generations exhibiting larger PD-1+ populations on ERY1-OVA or TER119-SIINFEKLadministration compared with OVA or SIINFEKL, respectively (Left); the cumulative PD-1+ population (as a percentage of all OTIs) for each group is denoted(Right). Data represent mean ± SD (n = 5). ***P < 0.0001; ###P < 0.0001. (F) Heat map representing OTI CD8+ T-cell proliferation generations exhibiting largerapoptotic (annexin-V+) populations on ERY1-OVA or TER119-SIINFEKL administration compared with OVA or SIINFEKL, respectively (Left); the cumulativeannexin-V+ population (as a percentage of all OTIs) for each group is denoted (Right). Data represent mean ± SD (n = 5). ***P < 0.0001; ###P < 0.0001. All datawere determined by multiparameter flow cytometry.

Kontos et al. PNAS | Published online December 17, 2012 | E63

IMMUNOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

than did OVA (7.6 ± 1.1% and 4.8 ± 1.3%, respectively) (Fig. 4 EandF, upper two rows), suggesting an exhausted and apoptotic fatethat would eventually lead to clonal deletion. Even more strikingresults were obtained with the TER119-SIINFEKL treatment(Fig. 4 E and F, lower two rows), with very high PD-1 (43.1± 10%)and annexin-V+ (35.7± 9.1%) levels; although the free SIINFEKL

peptide was able to induce CD8+ T-cell proliferation (above), itdid not induce high levels of these markers (8.3 ± 3.6% and 7.5 ±2.9%, respectively). This phenotype of proliferating CD8+ T cell isconsistent with other reported OTI adoptive transfer models inwhich regulated antigen-specific T-cell receptor engagement byAPCs fails to induce inflammatory responses (23).

CD45.2

8D

Cα

) evil f o%(

sll ecI TO

CBAOVA , ERY1-OVA

Challenge control

Naive

G

mm(

emul ovr o

muT3 )

OVA+LPS

sllecTI T

Ognit ar efil or p- no

NCf o

%(8

Dα+

)

IH

*

3 ×105 OTIadoptive transfer

antigen + LPSchallenge i.d.

i.v. antigen1 6 15 19

3 ×105 OTIadoptive transfer

antigen + LPSchallenge i.d.

15 19

3 ×105 OTIadoptive transfer

19

naive ERY1- OVA

OVA challengecontrol

i.v. antigen

antigen+LPS

+ +- -+ +- +

OVA , ERY1-OVA

control

1 ×106 OTIadoptive transfer

antigen + LPSchallenge i.d.

i.v. antigen1 6 15 199

1 ×106

E.G7-OVA

1 ×106

E.G7-OVA

1 ×106 OTIadoptive transfer

antigen + LPSchallenge i.d.

15 199

D

0.0

0.4

0.8

1.2

1.6

salineERY1-OVAOVA

Time post graft (d)naive ERY1-

OVAOVA

0.0

0.1

0.2

0.3

0.4

4 5 6 7 8 9 100

50

100

150

naive ERY1-OVA

OVA challengecontrol

OTI

NFIγ

r pxe-ssein

gcI T

O) evil f o

%(sll e

**

0.0

0.1

0.2

0.3

0.4

naive ERY1- OVA

OVA challengecontrol

i.v. antigen

antigen+LPS

+ +- -+ +- +

**

0.1

1

10E

NFIγ

) Lm/ gn(

noi sser pxe

**

retitGgI

01gol -

0

2

4

6

naive ERY1- OVA

OVA challengecontrol

+ +- -+ +- +

naive ERY1- OVA

OVA challengecontrol

+ +- -+ +- +

F*

***

Fig. 5. Erythrocyte binding induces resistance to antigen challenge. (A) OTI CD8+ T-cell adoptive transfer tolerance model displays experimental protocol forexperimental as well as challenge and naive control groups (n = 5). i.d., intradermal. (B) Flow cytometric detection of OTI CD8+ T-cell populations (CD3e+ CD8α+

CD45.2+). (C) OTI CD8+ T-cell population quantification in the draining lymph nodes (inguinal and popliteal) 4 d following antigen challenge in CD45.1+ mice.**P < 0.01. (D) IFN-γ–expressing OTI CD8+ T cells in the draining lymph nodes 4 d following antigen challenge and restimulation with SIINFEKL peptide. **P <0.01. (E) IFN-γ concentrations in lymph node cell culture media 4 d following restimulation with SIINFEKL peptide, determined by ELISA. **P < 0.01. (F) OVA-specific serum IgG titers at day 19. *P < 0.05. Data represent mean ± SE. (G) Combination OTI- and OVA-expressing EL4 thymoma (E.G7-OVA) tumor tolerancemodel displays experimental protocol for experimental as well as control groups (n = 4 and n = 3, respectively). (H) Quantification of nonproliferating(generation 0) OTI CD8+ T cells circulating in blood 5 d following adoptive transfer; data represent median ±minimum to maximum. ***P = 0.0002. (I) Growthprofile of E.G7-OVA tumors that were s.c. injected 9 d following OTI adoptive transfer; data represent mean ± SE. *P < 0.05.

E64 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

OTI Challenge Model. Using an established OTI challenge model(12) (Fig. 5A), we demonstrated the ability of ERY1-OVA toprevent postdeletion immune responses to vaccine-mediatedantigen challenge, even with a very strong bacterially derived ad-juvant. To delete, we i.v. administered 10 μg of either OVA orERY1-OVA 1 and 6 d following adoptive transfer of OTI CD8+

(CD45.2+) T cells to CD45.1+mice.After an additional 9 d to allowpotential deletion of the transferred T cells, we then challenged therecipient mice with OVA adjuvanted with LPS by intradermal in-jection.Characterization of draining lymphnode and spleen cells, aswell as their inflammatory responses, 4 d after challenge allowed usto determine if deletion actually took place.The i.v. administration of ERY1-OVA resulted in profound

reductions in OTI CD8+T-cell populations in the draining lymphnodes (Fig. 5C, gating shown in Fig. 5B) and spleens (Fig. S3,gating shown in Fig. S3A) compared with administration of un-modified OVA, demonstrating deletion. Draining lymph nodesfrom ERY1-OVA–treated mice contained over 11-fold fewerOTI CD8+ T cells compared with OVA-treated mice and 39-foldfewer than challenge control mice that did not receive i.v. injec-tions of antigen; responses in spleen cells were similar (Fig. S3B).This effective clonal deletion exhibited in mice administered withERY1-OVA supported our earlier observations of enhancedOTI CD8+ T-cell cross-priming (Fig. 4); furthermore, it showsthat splenic and/or liver-localized cross-priming occurred in theabsence of APC presentation of costimulatory molecules, leadingto T-cell deletion.To evaluate the immune response following antigen challenge

further, the inflammatory nature of resident lymph node andspleen cells was characterized by expression of IFN-γ by OTICD8+ T cells. Following challenge with OVA and LPS, the lymphnodes of mice previously treated with ERY1-OVA harbored 53-fold fewer IFN-γ–expressing OTI CD8+ T cells compared withthose of challenge control mice (previously receiving no antigen)and over 19-fold fewer IFN-γ–expressing cells compared withthose of mice previously treated with an equivalent dose of OVA(Fig. 5D), demonstrating the importance of erythrocyte binding indeletional protection to challenge; responses in spleen cells weresimilar (Fig. S3D, gating shown in Fig. S3C). In addition, of thesmall OTI CD8+ T-cell population present in the lymph nodesand spleens of mice previously treated with ERY1-OVA, a lowerpercentage expressed IFN-γ, suggesting clonal inactivation (Figs.S3E and S4). Furthermore, the magnitude of total IFN-γ levelsproduced by cells isolated from the draining lymph nodes onSIINFEKL restimulation was substantially reduced in mice pre-viously treated with ERY1-OVA (Fig. 5E), with erythrocytebinding reducing IFN-γ levels 16-fold compared with OVA ad-ministration and over 115-fold compared with challenge controls.Interestingly, the suppressive phenomenon was also correlatedwith down-regulated IL-10 expression by lymph node cells (Fig.S5). Typically considered a regulatory CD4+ T cell-expressedcytokine in the context of APC–T-cell communication to dampenT helper 1 cell responses (24, 25), IL-10 expression was dispens-able for desensitization to challenge. Similar IL-10 down-regula-tion has been implicated in CD8+ T cell-mediated tolerogenesis(8, 26, 27). Similar OTI responses were seen when using TER119-SIINFEKL as the deletional antigen (Fig. S6).Erythrocyte binding also substantially attenuated humoral im-

mune responses against antigen, becausemice treated with ERY1-OVA exhibited 100-fold lower antigen-specific serum IgG titerscompared with mice treated with soluble OVA (Fig. 5F). A similarreduction in OVA-specific IgG titer reduction as a result of eryth-rocyte binding was seen in nonadoptively transferred C57BL/6(CD45.2+) mice. Following two i.v. injections of 1 μg of OVA orERY1-OVA 7 d apart, ERY1-OVA–treated mice exhibited 39.8-fold lower OVA-specific serum IgG levels 19 d after the first an-tigen injection (Fig. S7). This apparent reduction in B-cell acti-vation, following erythrocyte binding by the antigen, corroborates

current hypotheses concerning noninflammatory antigen pre-sentation during tolerance induction (3, 4, 28).

E.G7-OVA Thymoma Tolerance Model. To validate the induction ofantigen-specific immune tolerance in response to the engineerederythrocyte-targeting antigen further, we combined the OTIchallenge model with an OVA-expressing tumor graft model (Fig.5G). Similar to the previous experimental design, mice weretolerized by means of two i.v. injections of 10 μg of ERY1-OVA or10 μg of OVA following adoptive transfer of OTI CD8+ T cells.Marked T-cell deletion was detected 5 d following the first antigenadministration, because ERY1-OVA–injectedmice harbored 2.9-fold fewer nonproliferating (generation 0) OTI CD8+ T cells inthe blood (Fig. 5H). To determine the functional responsiveness ofproliferating OTI CD8+ T cells in the absence of a strong exoge-nously administered adjuvant, OVA-expressing EL-4 thymomacells (E.G7-OVA) were intradermally injected into the back skin ofmice 9 d following adoptive transfer. To assess the protective effi-cacy of erythrocyte-bound antigen, tumor-bearing mice were chal-lenged with LPS-adjuvanted OVA 6 d following tumor grafting,analogous in dose and schedule to the challenge model. Robusttumor growth was continuously observed in ERY1-OVA–treatedmice compared with OVA-treated or nontreated control micethrough to 8 d following tumor grafting (Fig. 5I), confirming thatERY1-OVA–driven OTI CD8+ T-cell proliferation induced func-tional immune nonresponsiveness to OVA. That tumor size wasarrested to a steady state 8 d following grafting may be indicative ofresidual OTI CD8+ T cells that had yet to undergo ERY1-OVA–driven deletion.Taken together, these results demonstrate that erythrocyte

binding by ERY1-OVA induces antigen-specific T-cell deletionto a strongly adjuvanted challenge, as well as implanted cellulargrafts expressing a xenoantigen. Moreover, this response wasachieved by functional inactivation and deletion of reactiveCD8+ T cells through interaction with antigen present oncirculating erythrocytes.

BDC2.5 Diabetes Induction Model. To explore erythrocyte-bindingantigen-based deletion of pathogenic T cells in an autoimmunemodel, we used the accelerated onset model of autoimmune type 1diabetes (29, 30). We demonstrated that a peptide islet β-cell auto-antigen fused to the erythrocyte-binding antibody fragment TER119protected mice from diabetes onset by deleting the aggressiveadoptively transferred activated diabetogenic T cells (Fig. 6). Weadoptively transferred transgenic diabetogenic CD4+ T cells fromthe nonobese diabetic (NOD) BDC2.5 mouse model previously ac-tivated ex vivo with the p31 mimetope peptide (YVRPLWVRME)to induce rapid diabetes onset; the BDC2.5 clonal epitope has re-cently been identified as the islet antigen chromogranin-A processedand displayed on I-Ag7 MHC II (31, 32).To determine if TER119-p31 induced analogous antigen-

specific CD4+ T-cell proliferation and deletion as TER119-SIIN-FEKL did in the OTI CD8+ T-cell context, activated BDC2.5 Tcells were labeled with CFSE and adoptively transferred into NODhosts. Mice were i.v. administered 10 μg of TER119-p31 or anequimolar dose of p31 8 h following adoptive transfer and wereeuthanized 4 d later. Splenocytes and pancreatic lymph node cellswere harvested, and BDC2.5 CD4+ T-cell proliferation wasassessed by dilution of the CFSE fluorophore via flow cytometry.Consistent with results obtained with ERY1-OVA and TER119-SIINFEKL, erythrocyte-binding p31 induced potent BDC2.5T-cell proliferation and deletion, because TER119-p31–treatedmice harbored 20-fold and 28-fold fewer nonproliferated BDC2.5CD4+ T cells compared with mice treated with p31 in the spleenand pancreatic lymph nodes, respectively (Fig. 6A).To assess the effects of TER119-p31–driven proliferation in

a pathological autoimmune setting, we performed an analogousyet longer term adoptive transfer study to allow the diabetogenic

Kontos et al. PNAS | Published online December 17, 2012 | E65

IMMUNOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

BDC2.5 CD4+ T cells to induce autoimmune diabetes. Follow-ing adoptive transfer of activated BDC2.4 CD4+ T cells, NODmice were i.v. administered 10 μg of TER119-p31 or an equi-molar dose of p31 every 3 d, for a total of three doses. Micetreated with p31 peptide or saline exhibited profound hyperglyce-mia as early as 3 d following adoptive transfer (Fig. 6C), with 100%of mice developing diabetes by days 5 and 6, respectively (Fig. 6D).Conversely, mice treated with TER119-p31 remained normogly-cemic throughout the experimental time course (Fig. 6C), with 0%incidence of diabetes 62 d following adoptive transfer (Fig. 6D).Histological confirmation of pathological findings showed low in-sulin expression and extensive T-cell infiltration (CD3e+) intopancreatic islets of mice treated with saline or p31 peptide, com-pared with robust insulin expression and no evident T-cell presencein mice administered TER119-p31 (Fig. 6B). These data demon-strate the efficacy of erythrocyte-binding antigens in driving clonaldeletional protection even in the context of preactivated, patho-genic CD4+ T cells in an autoimmune setting.

DiscussionRecent studies in the field of immune tolerance, motivated by thesearch for mechanistic understanding, have expanded the classicdogma of central tolerance induction to a far more interconnectedand dynamic process involving a cohort of cellular interactionsresponding amid a complex extracellular milieu. Aside from small-molecule–based immune suppressants, recent clinical trials usingbiologics-based broad-spectrum T cell- and B cell-depleting mol-ecules for the treatment of autoimmune disorders andGvHDhaveshown promising, yet mixed, results (1). Their varied extents ofsuccess and numerous side effects stem from their mode of action,namely, reducing all T-cell or B-cell populations, as opposed toselected removal of only the pathogenic subpopulation. For such

reasons, a long-standing goal in the induction of immunologicaltolerance is achieving lasting antigen-specific tolerance and at-tenuation of autoreactive T andB cells to within limits controllableby peripheral tolerance mechanisms (33). Tolerogenic regimensaimed at inducing regulatory T-cell (Treg) populations by systemicsoluble antigen administration have been attempted in the pre-vention of type 1 diabetes but have demonstrated little efficacy inthe clinic (34). Tregs have been extensively implicated as keyplayers in peripheral tolerance maintenance, and as such, severaltherapeutic strategies aimed at their induction in vivo and evendirect administration are currently being explored (35). Antigentherapy may further require administration of multiple antigenicepitopes carefully selected by in-depth clinical studies (36) orbiochemically modified for enhanced MHC binding (37). Withregard to targeted antigen delivery to promote tolerogenic pre-sentation, antibody-based antigen targeting to steady-state DCshas shown promise in preclinical studies as a biomolecular ap-proach to exploit the ability of APCs to induce clonal CD8+ T-cellanergy or deletion (38, 39).Here, we explore immunotherapy by antigen-specific T-cell

deletion, reasoning that systemic, quantitative removal of quies-cent or activated T cells from circulation would be beneficial ina number of applications. In our approach, we engineer an antigento bind circulating erythrocytes strongly and specifically, exploitingnatural regulatory mechanisms induced by apoptotic (here, eryp-totic) cell-derived antigen (3, 4, 40), yet without any ex vivo cellularmanipulation. Several molecular determinants of apoptotic cell-driven tolerance have been elucidated, and many more have beenproposed (5, 41). The most characterized source of such tolero-genic signals presented on apoptotic cells to APCs is phosphati-dylserine, which is recognized by phagocytic cells through eitherdirect ligation to receptors or indirect bridging by proteins, such

*

*

spleen pLN

0.1

0.2

0.3

0.4

0.5

0

Non

-pro

lifer

atin

g B

DC

2.5

cells

(% o

f CD

4+)

CD

4

CFSE

p31

TER119-p31

***

Blo

od g

luco

se (m

g/dL

)

Days following adoptive transfer

TER119-p31p31saline

TER119-p31p31

naive saline

p31 TER119-p31

DA

PI

CD

3εin

sulin

DA

PI

CD

3εin

sulin

A B

C

1 2 3 4 5 6 7 8 9 121110 49 55 62

150

300

450

600

Dia

bete

s in

cide

nce

(%)

0

60

20

40

80

100

1 2 3 4 5 6 7 8 9 121110

*** TER119-p31p31saline

i.v. antigen

62

Days following adoptive transfer

D

Fig. 6. Erythrocyte-binding autoantigen protects mice from T cell-induced autoimmune type 1 diabetes. (A) Increased proliferation and deletion ofadoptively transferred diabetogenic BDC2.5 CD4+ T cells in the spleen and pancreatic lymph nodes 4 d following administration of TER119-p31 compared withp31, as determined by dilution of CFSE fluorescence via flow cytometry (n = 4). *P < 0.01. (B) Microscopy images of pancreatic islets stained for insulin (red),CD3e (green), and DAPI nuclear staining (blue), demonstrating marked T-cell infiltration and islet destruction of saline- and p31-treated mice but not ofTER119-p31-treated mice. (Scale bar = 100 μm.) (C) Glycemia monitoring as measured by daily blood glucose measurements following adoptive transfer ofdiabetogenic BDC2.5 CD4+ T cells and a tolerogenic treatment regimen of either saline, p31, or TER119-p31 (n = 8, n = 9, and n = 9, respectively). ***P <0.0001. (D) Diabetes incidence rate quantified by measurements in C; arrows indicate antigen administration time points. ***P < 0.0001.

E66 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

as milk-fat globule EGF factor 8, growth arrest-specific 6 protein,and protein S (41, 42). Our results demonstrate that glycophorin-A targeted antigen can strongly and specifically bind both eryp-totic and normal (which will continue to age) erythrocytes in vivo(Fig. 2), and thus exploit the eryptotic erythrocyte as a toler-ogenic vehicle.Given the plasticity of T-cell subsets and their dynamic re-

sponse to antigens in vivo, phenotypic changes are key to the in-duction of tolerance vs. immunity. Blockade of the costimulatoryCD28/B7 signaling axis between T cells and APCs by cytotoxic T-lymphocyte antigen-4 serves as a constitutive regulatory controlswitch, and has been exploited as a tolerogenic regimen for auto-immunity treatment (43). PD-1 signaling on T cells by its ligandsPD-L1 and PD-L2 represents another regulatory mechanism ofimmune responses, and it has been extensively studied in thecontext of autoimmunity and T-cell exhaustion and deletion (44).Selective blockade of the PD-1/PD-L1 signaling axis was shown todisrupt insulin-coupled splenocyte tolerance induction in theBDC2.5 model by blocking T-cell receptor stop signals, thus de-creasing T-cell motility and increasing physical interactions withAPCs (45). Our results with erythrocyte-binding antigens are con-sistent with the implication of PD-1 in tolerogenesis, becausemarked expression of the regulatory marker was seen in pro-liferative generations of antigen-specific T cells following ERY1-OVA and TER119-SIINFEKL administration (Fig. 4E). Here, wedemonstrate that erythrocyte-binding antigens induce PD-1 sig-naling to drive deletional proliferation of clonal T-cell populations,both in the context of xenoantigens (Figs. 4 and 5I) and auto-antigens (Fig. 6A).Motivated by the success seen in mouse models in induction

of antigen-specific tolerance by antigen conjugated to or loadedwithin apoptotic cells, we aimed to engineer a simple, more readilytranslatable molecular approach that might achieve analogousends. We demonstrated two such biomolecular designs, one withprotein antigen conjugated to multiple copies of a glycophorin A-binding peptide and another with a peptide or protein antigenfused to a scFv antibody fragment specific for the same erythro-cyte-specific target. These demonstrated deletions of both CD8+

and CD4+ T cells, both antigen-experienced (in the case of theCD4+ T cells in the BDC2.5 model) and antigen-naive (in thecase of the CD8+ T cells in the OTI model). Although much workremains, such as exploring reversal of ongoing autoimmunity ina spontaneous onset model and determining the longevity of de-letion (which will likely depend on the age of the host), the abilityto delete activated pathogenic islet-specific CD4+ T cells, andthereby completely protect islets from destruction, leads us tobelieve this approach has the potential to serve as a simpleyet effective biomolecular modification to existing antigens forantigen-specific peripheral deletion of T cells. Because of themodular nature of the molecular designs presented above, thisapproach may be relevant to treatment of a number of autoim-mune diseases dependent on pathogenic T cells.

Materials and MethodsAnimals. The Swiss Veterinary Authority and the EPFL Centre d’Application duVivant previously approved all animal procedures. Female C57BL/6 mice(Charles River Laboratories) aged 8–12 wk were used for in vivo bindingstudies and as E.G7-OVA tumor hosts. C57BL/6-Tg(TcraTcrb) 1100Mjb (OTI)mice (Jackson Laboratories) were bred in specific pathogen-free (SPF) con-ditions at the Ecole Polytechnique Fédérale Lausanne Animal Facility, andfemales were used for splenocyte isolation at 6–12 wk of age. Female B6.SJL-PtprcaPepcb/Boy (CD45.1) mice (Charles River Laboratories) aged 8–12 wkwere used as recipient hosts for OTI CD8+ T-cell adoptive transfer and tol-erance induction studies. NOD.Cg-Tg(TcraBDC2.5, TcrbBDC2.5)1Doi/DoiJmice (BDC2.5 mice), which express the T-cell receptor transgene for specificrecognition of the chromogranin-A islet antigen, were bred in SPF con-ditions at the Ecole Polytechnique Fédérale Lausanne Animal Facility, andfemales were used for cell isolations. NOD/ShiLtJ mice (Charles River Labo-ratories) were used as BDC2.5 adoptive transfer hosts.

ERY1 Peptide Design. The ERY1 peptide was H2N-WMVLPWLPGTLDGGSGCRG-CONH2, and the MIS peptide was H2N-PLLTVGMDLWPWGGSGCRG-CONH2.The underlined sequence is the ERY1 12-mer sequence that we previouslydiscovered by phage display as a mouse glycophorin-A binder (16). TheGGSG region served as a linker to the cysteine residue used for conjugation;the flanking arginine residue served to lower the pKa, and thus to increasethe reactivity of the cysteine residue (46). SIINFEKL and p31 (YVRPLWVRME)peptides were purchased from Genscript. Detailed methods are provided inSI Materials and Methods.

ERY1-Antigen Design. The ERY1 or MIS peptide was conjugated via its freecysteine thiol to lysine side chains on OVA using the sulfosuccinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate linker (Thermo Scientific). Asa control, the ERY1 peptide was conjugated to GST derived from Escherichiacoli. For biodistribution studies, the ERY1 or MIS peptide was conjugated toallophycocyanin (Innova Biosciences). Methods are shown in SI Materialsand Methods.

TER119-Antigen Development. The TER119 scFv antibody fragment was cre-ated as described elsewhere (45). In brief, forward and reverse primer mix-tures were used to amplify the variable domains of the heavy and lightchains of cDNA of the TER119 hybridoma clone (provided by Shozo Izui,University of Geneva, Geneva, Switzerland). The products were digested andligated into the pMAZ360 subcloning vector, and the resultant sequence wasqueried in the ImMunoGeneTics database (www.imgt.org) to determine therelevant VH and VL domains. Using assembly PCR, the VH and VL genes werethen fused with a (G4S)3 linker and cloned into the SfiI and XbaI sites onpSecTagA (Invitrogen). From this parent vector, SIINFEKL and/or p31 wasthen additionally cloned into the 3′ end of the TER119 scFv gene using as-sembly PCR. The TER119-antigen constructs were expressed in suspensionculture of HEK293E cells under serum-free conditions with 3.75 mM valproicacid (Sigma–Aldrich) (47) for 5 d and purified from supernatant usingimmobilized metal ion affinity chromatography on an Akta FPLC system (GEHealthcare). Purified proteins were analyzed for purity using SDS/PAGE, forendotoxin level using THP-1 × Blue cells (InvivoGen), and for concentrationusing bicinchoninic acid assays (Thermo Scientific). The final product wassterile-filtered and stored at −80 °C in working aliquots.

OTI T-Cell Adoptive Transfer. CD8+ T cells from OTI (CD45.2+) mouse spleenswere isolated using a CD8 magnetic bead negative selection kit (MiltenyiBiotec) as per the manufacturer’s instructions. Freshly isolated CD8+ OTI cellswere resuspended in PBS and labeled with 1 μM CFSE (Invitrogen) for 6 minat room temperature, and the reaction was quenched for 1 min with anequal volume of Iscove’s modified Dulbecco’s medium (IMDM) with 10%(vol/vol) FBS (Gibco). Cells were washed, counted, and resuspended in pureIMDM before injection. A total of 3 × 106 CFSE-labeled CD8+ OTI cells wereinjected i.v. into the tail vein of recipient CD45.1+ mice. For short-termproliferation studies, 10 μg of ERY1-OVA or OVA or an equimolar dose ofSIINFEKL or TER119-SIINFEKL in a 100-μL volume was injected 24 h followingadoptive transfer. Splenocytes were harvested 5 d following antigen ad-ministration and stained for analysis by flow cytometry.

OTI Challenge Model. A total of 3 × 105 CFSE-labeled OTI CD8+ T cells wereinjected into CD45.1+ recipient mice as described above. At 1 and 6 d fol-lowing adoptive transfer, mice were i.v. administered 10 μg of ERY1-OVA orOVA or an equimolar dose of SIINFEKL or TER119-SIINFEKL in 100 μL of salineinto the tail vein. At 15 d following adoptive transfer, mice were challengedwith 5 μg of OVA and 25 ng of ultrapure E. coli LPS (InvivoGen) in 25 μL ofsaline injected intradermally into each rear leg pad (Hock method: total doseof 10 μg of OVA and 50 ng of LPS). Mice were killed 4 d following challenge,and spleen and draining lymph node cells were isolated for restimulation.For flow cytometry analysis of intracellular cytokines, cells were restimulatedin the presence of 1 mg/mL OVA or 1 μg/mL SIINFEKL peptide (Genscript) for3 h. Brefeldin-A (5 μg/mL; Sigma) was added, and restimulation was resumedfor an additional 3 h before staining and flow cytometry analysis. For ELISAmeasurements of secreted factors, cells were restimulated in the presence of100 μg/mL OVA or 1 μg/mL SIINFEKL peptide for 4 d. Cells were spun, and themedia were collected for ELISA analysis using IFN-γ and IL-10 Ready-Set-Gokits (eBioscience) as per the manufacturer’s instructions. OVA-specific serumIgG was detected by incubating mouse serum at varying dilutions on OVA-coated plates, followed by a final incubation with goat anti-mouse IgG-HRP(Southern Biotech).

OTI E.G7-OVA Tolerance Model. A total of 1 × 106 CFSE-labeled OTI CD8+ T cellswere injected into 8- to 12-wk-old female C57BL/6 mice as described above.

Kontos et al. PNAS | Published online December 17, 2012 | E67

IMMUNOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0

At 1 and 6 d following adoptive transfer, mice were i.v. administered 10 μgof ERY1-OVA or 10 μg of OVA in 100 μL of saline into the tail vein. Blood wasdrawn 5 d following adoptive transfer for characterization of OTI CD8+ T-cellproliferation by flow cytometry. OVA-expressing EL-4 thymoma cells (E.G7-OVA, CRL-2113; American Type Culture Collection) were cultured as perAmerican Type Culture Collection guidelines. In brief, cells were cultured inRPMI 1640 medium supplemented with 10% (vol/vol) FBS, 10 mM Hepes,1 mM sodium pyruvate, 0.05 mM β-mercaptoethanol, 1% puromycin/strep-tomycin (Invitrogen Gibco), and 0.4 mg/mL G418 (PAA Laboratories). Justbefore injection, cells were expanded in media without G418 and resus-pended on harvest in HBSS (Gibco). At 9 d following adoptive transfer, micewere anesthetized with isoflurane, the back area was shaved, and 1 × 106 E.G7-OVA cells were injected intradermally between the shoulder blades. At4 d following E.G7-OVA graft, tumor dimensions were measured every 24 hwith a digital caliper and tumor volume was calculated as an ellipsoid(V ¼ π=6

 l ·w · h; where V is volume, l is length,w is width, and h is the height ofthe tumor). At 15 d following adoptive transfer, mice were challenged with 5 μgof OVA and 25 ng of ultrapure E. coli LPS in 25 μL of saline injected intradermallyinto each rear leg pad (total dose of 10 μg of OVA and 50 ng of LPS).

BDC2.5 Diabetes Induction Model. Freshly isolated splenocytes from femaleBDC2.5 transgenic mice were stimulated for 4 d in DMEM supplemented with10% (vol/vol) FBS, 0.05 mM β-mercaptoethanol, 1% puromycin/streptomycin,and 0.5 μM p31 peptide. Following stimulation, cells were washed with basalDMEM and analyzed for purity by flow cytometry, and 5 × 106 T cells werei.v. injected into normoglycemic NOD/ShiLtJ mice. At 8 h following adoptive

transfer, mice were i.v. administered saline, 10 μg TER119-p31, or an equi-molar dose of p31 peptide every 3 d for three doses. Diabetes onset wasmonitored by measuring nonfasting blood glucose levels using an Accu-Check Aviva glucometer (Roche). Mice were considered diabetic at bloodglucose readings ≥250 mg/dL. After three hyperglycemic readings, micewere euthanized and organs were harvested for histology.

Data Analysis. All flow cytometry data were analyzed using FlowJo (v8.8.6 andv9.2; TreeStar). Graphing and statistical analyses of worked-up data wereperformed using Prism (v5; GraphPad). Mann–Whitney U tests were used forinterpreting flow cytometry population data, two-tailed Student t tests wereused for interpreting ELISA data, and ANOVA was used for interpreting OTIgeneration data (***P < 0.001; **P < 0.01; *P < 0.05). Log rank (Mantel–Cox)tests were used to analyze Kaplan–Meier plots of diabetes onset data.

Additional detailed materials and methods are available in SI Materialsand Methods.

ACKNOWLEDGMENTS. We thank C. Nembrini and J. K. Eby for helpfuldiscussions; Prof. M. A. Swartz (Ecole Polytechnique Fédérale Lausanne) forhelpful discussions and critically reading the manuscript; and P. C. Henrioud,C. Dessibourg, S. Thurnheer, V. Borel, and M. Pasquier for technical assis-tance. We also thank Prof. S. Izui (University of Geneva, Geneva, Switzer-land) for providing the TER119 hybridoma mRNA and Prof. K. Kretschmer(Center for Regenerative Therapies, Dresden, Germany) for helpful adviceconcerning the NOD.BDC2.5 model. Funding was provided by the Ecole Poly-technique Fédérale Lausanne School of Life Sciences.

1. Bluestone JA, et al. (2010) The Immune Tolerance Network at 10 years: Toleranceresearch at the bedside. Nat Rev Immunol 10(11):797–803.

2. De Groot AS, Scott DW (2007) Immunogenicity of protein therapeutics. TrendsImmunol 28(11):482–490.

3. Mueller DL (2010) Mechanisms maintaining peripheral tolerance. Nat Immunol 11(1):21–27.

4. Green DR, Ferguson T, Zitvogel L, Kroemer G (2009) Immunogenic and tolerogenic celldeath. Nat Rev Immunol 9(5):353–363.

5. Griffith TS, Ferguson TA (2011) Cell death in the maintenance and abrogation oftolerance: The five Ws of dying cells. Immunity 35(4):456–466.

6. Rothlin CV, Ghosh S, Zuniga EI, Oldstone MBA, Lemke G (2007) TAM receptors arepleiotropic inhibitors of the innate immune response. Cell 131(6):1124–1136.

7. Ferguson TA, Choi J, Green DR (2011) Armed response: How dying cells influence T-cell functions. Immunol Rev 241(1):77–88.

8. Fife BT, et al. (2006) Insulin-induced remission in new-onset NOD mice is maintainedby the PD-1-PD-L1 pathway. J Exp Med 203(12):2737–2747.

9. Godsel LM, et al. (2001) Prevention of autoimmune myocarditis through the inductionof antigen-specific peripheral immune tolerance. Circulation 103(12):1709–1714.

10. Martin AJ, et al. (2010) Ethylenecarbodiimide-treated splenocytes carrying male CD4epitopes confer histocompatibility Y chromosome antigen transplant protection byinhibiting CD154 upregulation. J Immunol 185(6):3326–3336.

11. Smarr CB, Hsu CL, Byrne AJ, Miller SD, Bryce PJ (2011) Antigen-fixed leukocytestolerize Th2 responses in mouse models of allergy. J Immunol 187(10):5090–5098.

12. Liu K, et al. (2002) Immune tolerance after delivery of dying cells to dendritic cells insitu. J Exp Med 196(8):1091–1097.

13. Berg CP, et al. (2001) Human mature red blood cells express caspase-3 and caspase-8,but are devoid of mitochondrial regulators of apoptosis. Cell Death Differ 8(12):1197–1206.

14. Lang KS, et al. (2003) Cation channels trigger apoptotic death of erythrocytes. CellDeath Differ 10(2):249–256.

15. Turley DM, Miller SD (2010) Prospects for antigen-specific tolerance based therapiesfor the treatment of multiple sclerosis. Results Probl Cell Differ 51:217–235.

16. Kontos S, Hubbell JA (2010) Improving protein pharmacokinetics by engineeringerythrocyte affinity. Mol Pharm 7(6):2141–2147.

17. Khandelwal S, Saxena RK (2006) Assessment of survival of aging erythrocyte incirculation and attendant changes in size and CD147 expression by a novel two stepbiotinylation method. Exp Gerontol 41(9):855–861.

18. Yamazaki S, et al. (2008) CD8+ CD205+ splenic dendritic cells are specialized to induceFoxp3+ regulatory T cells. J Immunol 181(10):6923–6933.

19. Holz LE, Warren A, Le Couteur DG, Bowen DG, Bertolino P (2010) CD8+ T celltolerance following antigen recognition on hepatocytes. J Autoimmun 34(1):15–22.

20. Ichikawa S, Mucida D, Tyznik AJ, Kronenberg M, Cheroutre H (2011) Hepatic stellatecells function as regulatory bystanders. J Immunol 186(10):5549–5555.

21. Thomson AW, Knolle PA (2010) Antigen-presenting cell function in the tolerogenicliver environment. Nat Rev Immunol 10(11):753–766.

22. Albert ML, et al. (1998) Immature dendritic cells phagocytose apoptotic cells viaalphavbeta5 and CD36, and cross-present antigens to cytotoxic T lymphocytes. J ExpMed 188(7):1359–1368.

23. Bursch LS, Rich BE, Hogquist KA (2009) Langerhans cells are not required for the CD8T cell response to epidermal self-antigens. J Immunol 182(8):4657–4664.

24. Darrah PA, et al. (2010) IL-10 production differentially influences the magnitude,quality, and protective capacity of Th1 responses depending on the vaccine platform.J Exp Med 207(7):1421–1433.

25. Lee MS, Kim Y-J (2007) Signaling pathways downstream of pattern-recognition

receptors and their cross talk. Annu Rev Biochem 76:447–480.26. Arnaboldi PM, Roth-Walter F, Mayer L (2009) Suppression of Th1 and Th17, but not

Th2, responses in a CD8(+) T cell-mediated model of oral tolerance. Mucosal Immunol

2(5):427–438.27. Saint-Lu N, et al. (2009) Targeting the allergen to oral dendritic cells with mucoadhesive

chitosan particles enhances tolerance induction. Allergy 64(7):1003–1013.28. Miller SD, Turley DM, Podojil JR (2007) Antigen-specific tolerance strategies for the

prevention and treatment of autoimmune disease. Nat Rev Immunol 7(9):665–677.29. Haskins K, McDuffie M (1990) Acceleration of diabetes in young NOD mice with

a CD4+ islet-specific T cell clone. Science 249(4975):1433–1436.30. Judkowski V, et al. (2004) Peptide specific amelioration of T cell mediated

pathogenesis in murine type 1 diabetes. Clin Immunol 113(1):29–37.31. Stadinski BD, et al. (2010) Chromogranin A is an autoantigen in type 1 diabetes. Nat

Immunol 11(3):225–231.32. Nikoopour E, et al. (2011) Vasostatin-1-derived peptide ChgA29-42 is an antigenic

epitope of diabetogenic BDC2.5 T cells in nonobese diabetic mice. J Immunol.33. Sakaguchi S, Powrie F, Ransohoff RM (2012) Re-establishing immunological self-

tolerance in autoimmune disease. Nat Med 18(1):54–58.34. Diabetes Prevention Trial–Type 1 Diabetes Study Group (2002) Effects of insulin in

relatives of patients with type 1 diabetes mellitus. N Engl J Med 346(22):1685–1691.35. Wood KJ, Bushell A, Jones ND (2011) Immunologic unresponsiveness to alloantigen in

vivo: A role for regulatory T cells. Immunol Rev 241(1):119–132.36. Bielekova B, et al. (2004) Expansion and functional relevance of high-avidity myelin-

specific CD4+ T cells in multiple sclerosis. J Immunol 172(6):3893–3904.37. Gabrysová L, Wraith DC (2010) Antigenic strength controls the generation of antigen-

specific IL-10-secreting T regulatory cells. Eur J Immunol 40(5):1386–1395.38. Mukhopadhaya A, et al. (2008) Selective delivery of beta cell antigen to dendritic cells

in vivo leads to deletion and tolerance of autoreactive CD8+ T cells in NOD mice. Proc

Natl Acad Sci USA 105(17):6374–6379.39. Stern JNH, et al. (2010) Promoting tolerance to proteolipid protein-induced

experimental autoimmune encephalomyelitis through targeting dendritic cells. Proc

Natl Acad Sci USA 107(40):17280–17285.40. Griffith TS, et al. (2007) Apoptotic cells induce tolerance by generating helpless CD8+

T cells that produce TRAIL. J Immunol 178(5):2679–2687.41. Nagata S, Hanayama R, Kawane K (2010) Autoimmunity and the clearance of dead

cells. Cell 140(5):619–630.42. Ravichandran KS (2011) Beginnings of a good apoptotic meal: The find-me and eat-

me signaling pathways. Immunity 35(4):445–455.43. Bour-Jordan H, et al. (2011) Intrinsic and extrinsic control of peripheral T-cell

tolerance by costimulatory molecules of the CD28/B7 family. Immunol Rev 241(1):

180–205.44. Francisco LM, Sage PT, Sharpe AH (2010) The PD-1 pathway in tolerance and

autoimmunity. Immunol Rev 236:219–242.45. Fife BT, et al. (2009) Interactions between PD-1 and PD-L1 promote tolerance by

blocking the TCR-induced stop signal. Nat Immunol 10(11):1185–1192.46. Lutolf MP, Tirelli N, Cerritelli S, Cavalli L, Hubbell JA (2001) Systematic modulation of

Michael-type reactivity of thiols through the use of charged amino acids. BioconjugChem 12(6):1051–1056.

47. Baldi L, Hacker DL, Meerschman C, Wurm FM (2012) Large-scale transfection of

mammalian cells. Methods Mol Biol 801:13–26.

E68 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 8,

202

0