Engineering for in situ erythrocyte binding induces T-cell deletion

Stephan 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 review September 26, 2012) Antigens derived from apoptotic cell debris can drive clonal T-cell in type 1 diabetes (8) and myocarditis (9), trans- deletion or anergy, and antigens chemically coupled ex vivo to plant tolerance (10), and (11). has also been loaded apoptotic cell surfaces have been shown correspondingly to induce osmotically into splenocytes, with the osmotic shock inducing tolerance on infusion. Reasoning that a large number of erythro- , in the context of tolerance to the model antigen ov- cytes become apoptotic (eryptotic) and are cleared each day, we albumin (OVA) (12). Although results obtained with these cel- engineered two different antigen constructs to target the antigen to lular approaches have been remarkable, the complexities of erythrocyte cell surfaces after i.v. injection, one using a conjugate translating such a cellular therapeutic approach to widespread with an erythrocyte-binding peptide and another using a fusion clinical use are formidable; despite these challenges, a phase I/ with an fragment, both targeting the erythrocyte-specific IIA trial has been carried out with peripheral blood mononuclear cell surface marker glycophorin A. Here, we show that erythrocyte- cells pulsed with multiple peptide in the context of re- binding antigen is collected much more efficiently than free antigen lapsing-remitting multiple sclerosis (European Union Drug by splenic and hepatic immune cell populations and hepatocytes, Regulating Authorities Clinical Trials no. 2008-004408-29). For + and that it induces antigen-specific deletional responses in CD4 and ease of translation to widespread clinical use, an acellular tol- + CD8 T cells. We further validated T-cell deletion driven by erythro- erogenic therapeutic approach composed only of simpler defined + cyte-binding antigens using a transgenic islet β cell-reactive CD4 biomolecules would be advantageous. For this, we turned to T-cell adoptive transfer model of autoimmune type 1 diabetes: engineering the antigen to enable in situ cell surface binding after Treatment with the peptide antigen fused to an erythrocyte-binding simple i.v. injection. antibody fragment completely prevented diabetes onset induced by We reasoned that the large number of erythrocytes cleared + the activated, autoreactive CD4 T cells. Thus, we report a translat- after apoptosis-like programmed cell death (eryptosis) each day able modular biomolecular approach with which to engineer anti- (more than 1% per day in humans and 4% in mice) (13) could gens for targeted binding to erythrocyte cell surfaces to induce be exploited to induce antigen-specific T-cell deletion. Although fi + + antigen-speci c CD4 and CD8 T-cell deletion toward exogenous the exact triggers of erythrocyte clearance remain unclear, eryp- antigens and autoantigens. totic erythrocytes are characterized by phosphatidylserine asym- metry, membrane heterogeneity, and annexin-V binding, analogous autoimmunity | | BDC2.5 | TER119 | to apoptotic nucleated cells (13, 14). Rather than chemically con- immunoengineering jugate erythrocytes to an antigen ex vivo, as has been mentioned (15), we sought to develop a targeted antigen formulation that trategies to induce antigen-specific tolerance without the need would spontaneously bind strongly and specifically to erythrocytes Sfor systemic immunosuppression have great potential in pre- after simple i.v. injection for eventual display to APCs, along with venting immune responses in protein replacement therapies, graft- apoptotic markers, as the eryptotic erythrocyte fraction is cleared versus-host disease (GvHD), and autoimmune disorders (1). For over the ensuing days. example, in the treatment of diseases requiring replacement pro- tein administration, such as hemophilia, some patients mount Results immune responses to the therapeutic protein antigen, rendering Creation and Characterization of Erythrocyte-Binding OVA. As one treatment ineffective and dangerous (2). In autoimmune diseases, approach to obtain strong and specific biophysical targeting of an where mechanisms fail to inactivate autor- antigen to erythrocyte cell surfaces in situ, we used a synthetic eactive T cells that have escaped central tolerance induction in the 12-aa peptide (ERY1) that we discovered by phage display to , deletion or functional inactivation to is con- bind to mouse glycophorin-A specifically (16), which is present sidered key to nonprogression of the disease (3). Immunoregulatory uniquely on erythrocytes. We used the model antigen OVA, so modalities capable of prophylactically educating or therapeutically reeducating the to induce antigen-specifictolerance are greatly needed for these and many other situations. Author contributions: S.K., K.Y.D., and J.A.H. designed research; S.K., I.C.K., and K.Y.D. Apoptotic cells are well known as a source of tolerogenic anti- performed research; S.K. and I.C.K. analyzed data; and S.K. and J.A.H. wrote the paper. gens, although the exact mechanisms are not fully understood (4, Conflict of interest statement: The Ecole Polytechnique Fédérale Lausanne has filed for 5). Literature evidence maintains that antigen displayed in a phys- patent protection on the technology described here, and S.K., K.Y.D., and J.A.H. are iological noninflammatory manner by apoptotic cells to antigen- named as inventors on those patents; J.A.H. is a shareholder in a company that has presenting cells (APCs) fails to induce a productive immune re- licensed those patents. sponse, because Toll-like receptor signaling is inhibited by uptake This article is a PNAS Direct Submission. G.G. is a guest editor invited by the of apoptotic cellular debris (6), leading to antigen-specific T-cell Editorial Board. deletion or anergy (7). Freely available online through the PNAS open access option. The concept of engineering apoptotic cell carriers to induce 1Present address: Anokion SA, CH-1015 Lausanne, Switzerland. immunological tolerance has been explored. Notably, antigens 2To whom correspondence should be addressed. E-mail: jeffrey.hubbell@epfl.ch. chemically coupled ex vivo to splenocyte cell surfaces in a manner See Author Summary on page 17 (volume 110, number 1). so as to induce apoptosis in the splenocytes has been shown to This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. induce antigen-specific tolerance on infusion in such contexts as 1073/pnas.1216353110/-/DCSupplemental.

E60–E68 | PNAS | Published online December 17, 2012 www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Downloaded by guest on September 27, 2021 as to enable initial analysis in a transgenic mouse strain (OTI), A CD45+ CD45- PNAS PLUS the CD8+ T-cell population of which expresses the T-cell re- ceptor specific for the MHC I immunodominant OVA , 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 affinity and specificity (Fig. 1A). High-resolution confocal microscopy naive naive showed localization to the cell membrane equatorial periphery B fl (Fig. 1 ), and ow cytometric analysis demonstrated binding to be OVA OVA sequence-specific (Fig. 1C). ERY1-OVA bound to erythrocytes with high affinity, exhibiting a dissociation constant (K ) of 6.2 ± d ERY1-OVA 1.3 nM, as determined by equilibrium binding measurements (Fig. ERY1-OVA 1D). Based on this approximate value, an entire 10 μg dose of 0101 102 103 0101 102 103 ERY1-OVA would be bound to the surface of erythrocytes fol- OVA binding lowing i.v. injection in a mouse. We validated ERY1-OVA binding in vivo to circulating eryth- B OVA ERY1-OVA rocytes following i.v. administration in mice. Whole-blood samples 0.11 3.75e-3 1.96e-3 0.25 taken 30 min following injection of 150 μg of either OVA or ERY1- 3 OVA confirmed the specific erythrocyte binding capability of 10 ERY1-OVA even amid the complex heterogeneous milieu of blood and the vasculature (Fig. 2A). Consistent with glycophorin- 102 − A association, ERY1-OVA bound to erythrocytes (CD45 ) but + not to leukocytes (CD45 ). ERY1-OVA binding was unbiased as 101 to the eryptotic state of the erythrocytes, binding strongly to + − − B both annexin-V and annexin-V CD45 populations (Fig. 2 ). Annexin-V 0 99.8 0.1 0.73 99 Pharmacokinetic studies of the OVA conjugate demonstrated 0101 102 103 0101 102 103 that ERY1-OVA erythrocyte binding was long-lived in vivo, OVA binding exhibiting a cell surface t1/2 of 17.2 h (Fig. 2C). ERY1-OVA remained bound to erythrocytes for as long as 72 h following 2.0 150

∼ C fluorescence intensity fluorescence administration; during this time frame, 13% of erythrocytes are ERY1-OVA OVAmean geometric cleared in the mouse (17). Quantification of erythrocyte-bound 1.5 OVA 100

OVA 1.0 A OVA ERY1-OVA B t1/2 = 17.2 h 6 50

1. SMCC OVA ERY1 OVA concentration OVA 0.5

(ng/10 erythrocytes) 2. ERY1-Cys ERY1 GYPA ERY1 ERY1 ERY1 0 0 ERY1-OVA 20 40 60 80 170150 ERY1 glycophorin-A Time post injection (h) OVA Fig. 2. ERY1-conjugated antigen biospecifically binds circulating healthy and erythrocyte eryptotic erythrocytes on i.v. administration, inducing uptake by specific APC GYPA subsets. (A) OVA (gray-filled histogram) and ERY1-OVA (black-filled histo- − gram) binding to erythrocyte (CD45 ) and nonbinding to leukocyte (CD45+) C D populations in vivo compared with noninjected mice (empty histogram), 100 100 ERY1-OVA determined by flow cytometry. (B) ERY1-OVA binding and OVA nonbinding + − 80 MIS-OVA 80 to circulating eryptotic (annexin-V ) and healthy (annexin-V ) erythrocytes, determined by flow cytometry. (C)Cellsurfacet of bound ERY1-OVA to SMCC-OVA 1/2 60 60 circulating erythrocytes, determined by geometric mean fluorescence in- K = 6.2 ± 1.3 nM fl 2

d tensity of ow cytometry measurements (n =2;R = 0.98, one-phase expo- % of max 40 40 nential decay), and time-dependent ERY1-OVA cell surface concentration, determined by ELISA, at an administered dose of 150 μg(n =2). 20 20 OVA geometric mean OVA fluorescence intensity

0 0 100 101 102 103 050100150200ERY1-OVA in vivo showed a relatively high loading of 1.174 ± OVA binding ERY1-OVA concentration (nM) 0.005 ng of OVA per 106 erythrocytes. Erythrocyte binding by Fig. 1. ERY1-OVA binds the equatorial periphery of mouse erythrocytes ERY1-OVA elicited no detectable differences in hematocrit, with high affinity. (A) Schematic of conjugation of ERY1 peptide to OVA, corpuscular volume, or hemoglobin content, compared with ad- resulting in binding to erythrocyte surface glycophorin-A. (B) High-resolu- ministration of OVA (Fig. S1), demonstrating that glycophorin- tion confocal microscopy images of mouse erythrocytes labeled ex vivo with A–mediated erythrocyte binding with antigen did not alter their (green) anti-mouse glycophorin-A (GYPA) and (red) either OVA (Upper)or hematological parameters. ERY1-OVA (Lower). (Scale bar = 5 μm.) (C) Binding of each OVA conjugate fl and intermediate, characterized by ow cytometry. ERY1, erythrocyte- Erythrocyte Bound Antigen Uptake and Biodistribution. To reveal the binding peptide WMVLPWLPGTLD; MIS, MIS peptide PLLTVGMDLWPW; cellular targets of antigen as it is collected after binding to eryth- SMCC, sulfosuccinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate, used to conjugate ERY1 to OVA. (D) Equilibrium binding of ERY1-OVA to rocytes after administration, we i.v. injected mice with the highly erythrocytes demonstrates the low dissociation constant of ERY1-OVA (R2 = fluorescent allophycocyanin protein, conjugated to either ERY1 0.97, one-site binding), determined by flow cytometry. (ERY1-allophycocyanin) or a mismatch (MIS) scrambled peptide

Kontos et al. PNAS | Published online December 17, 2012 | E61 Downloaded by guest on September 27, 2021 (MIS-allophycocyanin) (binding shown in Fig. S2A). Flow cyto- Kupffer cells (CD45+ MHC II+ F4/80+), which serve as members metric analysis of splenic (DC) populations 12 and of the reticuloendothelial system that aids in clearance of eryth- 36 h following administration showed 9.4-fold enhanced uptake rocytes and complement-coated particles. Analysis of liver sec- − of ERY1-allophycocyanin by MHC II+ CD11b CD11c+ DCs tions confirmed the flow cytometry measurements. Although no compared with MIS-allophycocyanin yet similar uptake of ERY1- intact-OVA staining was seen in the liver 24 h following injection allophycocyanin and MIS-allophycocyanin by MHC II+ CD11b+ (perhaps due to antigen processing), only ERY1-OVA–treated CD11c+ DCs (Fig. 3A). Additionally, MHC II+ CD8α+ CD11c+ mouse livers stained strongly with the 25.d1.16 mAb, which specif- CD205+ splenic DCs were found to uptake ERY1-allophycocyanin ically recognizes the immunodominant OVA peptide SIINFEKL to a threefold greater extent than MIS-allophycocyanin, although displayed on H-2kb MHC I (Fig. 3D). In addition to supporting the absolute magnitude was markedly lower than for other DC observations of hepatic cell uptake of ERY1-allophycocyanin, the populations in the spleen. Such targeting of antigen toward non- presence of MHC I–SIINFEKL complexes on hepatic cell surfaces activated and CD8α+ CD205+ splenic DCs could strengthen the suggests their role in cross-presentation of exogenous antigens. tolerogenic potential of erythrocyte binding, because these pop- ulations have been extensively implicated in apoptotic cell-driven T-Cell Phenotype in Response to ERY1- and Single-Chain Fv Antigen tolerogenesis (7, 18). Such flow cytometry-based results were vali- Constructs. We next determined whether erythrocyte-bound an- dated using confocal microscopy. Spleen sections taken from mice tigen could be efficiently cross-presented and could corre- 24 h following i.v. injection of OVA or ERY1-OVA showed OVA spondingly cross-prime reactive CD8+ T cells. We adoptively staining in areas distinct from F4/80+ regions, with transferred carboxyfluorescein succinimidyl ester (CFSE)-labeled ERY1-OVA–injected mice displaying more overall OVA staining OTI CD8+ T cells (CD45.2+) into CD45.1+ mice and measured than OVA-injected mice (Fig. 3C). In the liver, ERY1-allophyco- proliferation of the OTI CD8+ T cells 5 d following i.v. adminis- − cyanin also greatly enhanced uptake by hepatocytes (CD45 MHC tration of 10 μgofOVA;10μg of ERY1-OVA; or 10 μgofanir- − − IIlow CD1d , by 78.4-fold) and hepatic stellate cells (CD45 MHC relevant erythrocyte-binding antigen, ERY1-GST (binding shown II+ CD1d+, by 60.6-fold) compared with MIS-allophycocyanin in Fig. S2B). OTI CD8+ T-cell proliferation, determined by di- (Fig. 3C); both populations have been reported as APCs that lution of the fluorophore CFSE as measured by flow cytometry trigger CD8+ T-cell deletional tolerance (19–21). Interestingly, (Fig. 4A), was markedly enhanced over threefold in mice adminis- preferential uptake was not seen in liver DCs (CD45+ CD11c+)or tered ERY1-OVA compared with OVA (Fig. 4B), demonstrating

spleen liver A B *** 70 * * ERY1-allophycocyanin ERY1-allophycocyanin 10 60 10 MIS-allophycocyanin MIS-allophycocyanin

50 **

40 5 5

(% of each subset) 30 * (% of each subset)

Allophycocyanin+ cells

Allophycocyanin+ cells

5

0 12 h 36 h 12 h 36 h 12 h CD45- MHCIIlow CD45- MHCII+ CD45+ CD45+ MHCII+ MHCII+ CD11b- MHCII+ CD11b+ MHCII+ CD8α+ CD1d- CD1d+ CD11c+ F4/80+ CD11c+ CD11c+ CD11c+ CD205+ spleen liver CDOVA ERY1-OVA OVA ERY1-OVA CD45 F4/80 OVA H-2Kb-SIINFEKL DAPI DAPI

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 not by 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) Spleen microscopy images of mice 24 h following administration of 10 μgOVA(Left)orERY1-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 of 10 μgOVA(Left)orERY1-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. Downloaded by guest on September 27, 2021 PNAS PLUS ** A ERY1-GST OVA ERY1-OVA B 10 μg 40 1 μg 103 30 102 2 11 44 20 101 (% of OTI) 10

0 ##

CD8 α

Proliferating OTI T cells T Proliferating OTI 0101 102 103 0101 102 103 0101 102 103 saline ERY1- OVA ERY1- CFSE GST OVA

CDsaline SIINFEKL TER119-SIINFEKL 100

103 75

102 2 99 97 50

101 (% of OTI) 25

0

CD8 α

Proliferating OTI T cells T Proliferating OTI 0101 102 103 1 2 3 1 2 3 010 10 10 010 10 10 saline SIINFEKL TER119- CFSE SIINFEKL

PD-1+ OTI T cells (% of OTI) Total E 10% OVA 7.6±1.1 *** ERY1-OVA 23.2±4.9

SIINFEKL 8.3±3.6 TER119- ### SIINFEKL 43.1±10 012345678 0% Generation

Annexin-V+ OTI T cells (% of OTI) Total F 7% OVA 4.8±1.3 *** ERY1-OVA 20.4±7.4

SIINFEKL 7.5±2.9 TER119- ### SIINFEKL 35.7±9.1 012345678 0% Generation

Fig. 4. Erythrocyte-binding antigen formulations enhance cross-priming and apoptotic fate deletional proliferation of antigen-specific OTI CD8+ T cells in vivo. (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 μgof OVA (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 an identical 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-SIINFEKL administration 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 larger apoptotic (annexin-V+) populations on ERY1-OVA or TER119-SIINFEKL administration compared with OVA or SIINFEKL, respectively (Left); the cumulative annexin-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 data were determined by multiparameter flow cytometry.

that erythrocyte binding increased antigen-specificCD8+ T-cell (95%) OTI CD8+ T-cell population 5 d following adoptive transfer,

cross-priming compared with the soluble antigen. Similar results indicating strong proteosomal processing and cross-presentation of IMMUNOLOGY were also obtained by administration of a 10-fold lower antigen the peptide antigen when fused to the erythrocyte-binding TER119 dose of 1 μg, demonstrating the wide dynamic range of efficacy of scFv (Fig. 4D). Although administration of soluble SIINFEKL at OTI CD8+ T-cell proliferation induced by ERY1-conjugated an equimolar dose induced similar proliferation rates, it is im- antigen. portant to note that in this format, SIINFEKL can directly bind to To address the technical hurdle of conjugating ERY1 to small MHC I and present to OTI CD8+ T cells without first being pro- peptide antigens insufficiently large for conjugation of multiple teolytically processed by APCs. Our results on cross-presentation copies of the ERY1 peptide, we engineered a second molecular and cross-priming are consistent with those of other studies con- implementation of our concept, namely, an erythrocyte-specific cerning tolerogenic on MHC I by APCs single-chain Fv (scFv) antibody fragment to display peptide anti- engulfing antigen from apoptotic cell debris (4, 22). gens. The murine glycophorin A-binding TER119 mAb was se- To distinguish T cells being expanded into a functional effector quenced and converted into the scFv format with a C-terminal phenotype from those being expanded and deleted, we analyzed SIINFEKL peptide fusion. We characterized the resulting eryth- the proliferating OTI CD8+ T cells for programmed death-1 (PD- rocyte-binding peptide fusion, TER119-SIINFEKL (binding shown 1), a marker of T-cell exhaustion, and annexin-V binding, a hall- in Fig. S2C), in an analogous in vivo OTI proliferation study as mark of apoptosis, and thus deletion. ERY1-OVA induced much with ERY1-OVA, with even more striking cross-priming results higher total numbers of PD-1+ (23.2 ± 4.9%) and annexin-V+ (Fig. 4C). TER119-SIINFEKL induced proliferation of the entire (20.7 ± 7.4%) proliferating OTI CD8+ T cells 4 d after treatment

Kontos et al. PNAS | Published online December 17, 2012 | E63 Downloaded by guest on September 27, 2021 than did OVA (7.6 ± 1.1% and 4.8 ± 1.3%, respectively) (Fig. 4 E peptide was able to induce CD8+ T-cell proliferation (above), it and F, upper two rows), suggesting an exhausted and apoptotic fate did not induce high levels of these markers (8.3 ± 3.6% and 7.5 ± that would eventually lead to . Even more striking 2.9%, respectively). This phenotype of proliferating CD8+ is results were obtained with the TER119-SIINFEKL treatment consistent with other reported OTI adoptive transfer models in (Fig. 4 E and F, lower two rows), with very high PD-1 (43.1 ± 10%) which regulated antigen-specific T-cell receptor engagement by and annexin-V+ (35.7 ± 9.1%) levels; although the free SIINFEKL APCs fails to induce inflammatory responses (23).

A B C OVA , ERY1-OVA naive ERY1-OVA 1.6

3 ×105 OTI antigen + LPS OTI )evilfo%(sllecITO adoptive transfer challenge i.d. 1.2

1 6 15 19 ** i.v. antigen 0.8 Challenge control 5 0.4 3 ×10 OTI antigen + LPS OVA challenge adoptive transfer challenge i.d. control 0.0 15 19 naive ERY1- OVA challenge 5 Naive

3 ×10 OTI OVA control α

adoptive transfer 8DC i.v. antigen --++ antigen+LPS - ++ + 19 CD45.2 D E F 0.4 10 6

) * Lm/

)evilf **

g re

0.3 n(noi

ing sse ** ti 4

o%(sll 1

tG

rpx

s gI

0.2 s

erpx

e-

01

e γ

cITO 2

g NFI

e 0.1 ol-

0.1 γ

NF I 0.0 0 naive ERY1- OVA challenge naive ERY1- OVA challenge naive ERY1- OVA challenge OVA control OVA control OVA control i.v. antigen --++ --++ --++ antigen+LPS - ++ + - ++ + - ++ + G H I OVA , ERY1-OVA

6 sllecT 1 ×10 OTI antigen + LPS 0.4 *** 150 saline

adoptive transfer challenge i.d. )

3 ERY1-OVA

m ITOgn

m * OVA 0.3 (emul

1 6 9 15 19 +) 6 α 100

i.v. antigen 1 ×10 8D OVA+LPS

E.G7-OVA ita

Cfo% 0.2 ovromu control ref 6 i 50

1 ×10 OTI antigen + LPS lorp-no

adoptive transfer challenge i.d. ( 0.1 T

15 19 9 N 0.0 0 1 ×106 naive ERY1- OVA 45678910 E.G7-OVA OVA Time post graft (d)

Fig. 5. Erythrocyte binding induces resistance to antigen challenge. (A) OTI CD8+ T-cell adoptive transfer tolerance model displays experimental protocol for experimental 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 tolerance model 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) Growth profile 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. Downloaded by guest on September 27, 2021 OTI Challenge Model. Using an established OTI challenge model current hypotheses concerning noninflammatory antigen pre- PNAS PLUS (12) (Fig. 5A), we demonstrated the ability of ERY1-OVA to sentation during tolerance induction (3, 4, 28). prevent postdeletion immune responses to vaccine-mediated antigen challenge, even with a very strong bacterially derived ad- E.G7-OVA Thymoma Tolerance Model. To validate the induction of juvant. To delete, we i.v. administered 10 μg of either OVA or antigen-specific immune tolerance in response to the engineered ERY1-OVA 1 and 6 d following adoptive transfer of OTI CD8+ erythrocyte-targeting antigen further, we combined the OTI (CD45.2+) T cells to CD45.1+ mice. After an additional 9 d to allow challenge model with an OVA-expressing tumor graft model (Fig. potential deletion of the transferred T cells, we then challenged the 5G). Similar to the previous experimental design, mice were recipient mice with OVA adjuvanted with LPS by intradermal in- tolerized by means of two i.v. injections of 10 μg of ERY1-OVA or jection. Characterization of draining lymph node and spleen cells, as 10 μg of OVA following adoptive transfer of OTI CD8+ T cells. well as their inflammatory responses, 4 d after challenge allowed us Marked T-cell deletion was detected 5 d following the first antigen to determine if deletion actually took place. administration, because ERY1-OVA–injected mice harbored 2.9- The i.v. administration of ERY1-OVA resulted in profound fold fewer nonproliferating (generation 0) OTI CD8+ Tcellsin reductions in OTI CD8+ T-cell populations in the draining lymph the blood (Fig. 5H). To determine the functional responsiveness of nodes (Fig. 5C, gating shown in Fig. 5B) and spleens (Fig. S3, proliferating OTI CD8+ T cells in the absence of a strong exoge- gating shown in Fig. S3A)comparedwithadministrationofun- nously administered adjuvant, OVA-expressing EL-4 thymoma modified OVA, demonstrating deletion. Draining lymph nodes cells (E.G7-OVA) were intradermally injected into the back skin of from ERY1-OVA–treated mice contained over 11-fold fewer mice 9 d following adoptive transfer. To assess the protective effi- OTI CD8+ T cells compared with OVA-treated mice and 39-fold cacy of erythrocyte-bound antigen, tumor-bearing mice were chal- fewer than challenge control mice that did not receive i.v. injec- lenged with LPS-adjuvanted OVA 6 d following tumor grafting, tions of antigen; responses in spleen cells were similar (Fig. S3B). analogous in dose and schedule to the challenge model. Robust This effective clonal deletion exhibited in mice administered with tumor growth was continuously observed in ERY1-OVA–treated ERY1-OVA supported our earlier observations of enhanced mice compared with OVA-treated or nontreated control mice OTI CD8+ T-cell cross-priming (Fig. 4); furthermore, it shows through to 8 d following tumor grafting (Fig. 5I), confirming that that splenic and/or liver-localized cross-priming occurred in the ERY1-OVA–driven OTI CD8+ T-cell proliferation induced func- absence of APC presentation of costimulatory molecules, leading tional immune nonresponsiveness to OVA. That tumor size was to T-cell deletion. arrested to a steady state 8 d following grafting may be indicative of To evaluate the immune response following antigen challenge residual OTI CD8+ T cells that had yet to undergo ERY1-OVA– further, the inflammatory nature of resident lymph node and driven deletion. spleen cells was characterized by expression of IFN-γ by OTI Taken together, these results demonstrate that erythrocyte CD8+ T cells. Following challenge with OVA and LPS, the lymph binding by ERY1-OVA induces antigen-specific T-cell deletion nodes of mice previously treated with ERY1-OVA harbored 53- to a strongly adjuvanted challenge, as well as implanted cellular fold fewer IFN-γ–expressing OTI CD8+ T cells compared with grafts expressing a xenoantigen. Moreover, this response was those of challenge control mice (previously receiving no antigen) achieved by functional inactivation and deletion of reactive and over 19-fold fewer IFN-γ–expressing cells compared with CD8+ T cells through interaction with antigen present on those of mice previously treated with an equivalent dose of OVA circulating erythrocytes. (Fig. 5D), demonstrating the importance of erythrocyte binding in deletional protection to challenge; responses in spleen cells were BDC2.5 Diabetes Induction Model. To explore erythrocyte-binding similar (Fig. S3D, gating shown in Fig. S3C). In addition, of the antigen-baseddeletionofpathogenicTcellsinanautoimmune small OTI CD8+ T-cell population present in the lymph nodes model, we used the accelerated onset model of autoimmune type 1 and spleens of mice previously treated with ERY1-OVA, a lower diabetes (29, 30). We demonstrated that a peptide islet β-cell auto- percentage expressed IFN-γ, suggesting clonal inactivation (Figs. antigen fused to the erythrocyte-binding antibody fragment TER119 S3E and S4). Furthermore, the magnitude of total IFN-γ levels protected mice from diabetes onset by deleting the aggressive produced by cells isolated from the draining lymph nodes on adoptively transferred activated diabetogenic T cells (Fig. 6). We SIINFEKL restimulation was substantially reduced in mice pre- adoptively transferred transgenic diabetogenic CD4+ T cells from viously treated with ERY1-OVA (Fig. 5E), with erythrocyte the nonobese diabetic (NOD) BDC2.5 mouse model previously ac- binding reducing IFN-γ levels 16-fold compared with OVA ad- tivated ex vivo with the p31 mimetope peptide (YVRPLWVRME) ministration and over 115-fold compared with challenge controls. to induce rapid diabetes onset; the BDC2.5 clonal epitope has re- Interestingly, the suppressive phenomenon was also correlated cently been identified as the islet antigen chromogranin-A processed with down-regulated IL-10 expression by lymph node cells (Fig. and displayed on I-Ag7 MHC II (31, 32). + To determine if TER119-p31 induced analogous antigen- S5). Typically considered a regulatory CD4 T cell-expressed IMMUNOLOGY in the context of APC–T-cell communication to dampen specificCD4+ T-cell proliferation and deletion as TER119-SIIN- T helper 1 cell responses (24, 25), IL-10 expression was dispens- FEKL did in the OTI CD8+ T-cell context, activated BDC2.5 T able for desensitization to challenge. Similar IL-10 down-regula- cells were labeled with CFSE and adoptively transferred into NOD tion has been implicated in CD8+ T cell-mediated tolerogenesis hosts. Mice were i.v. administered 10 μg of TER119-p31 or an (8, 26, 27). Similar OTI responses were seen when using TER119- equimolar dose of p31 8 h following adoptive transfer and were SIINFEKL as the deletional antigen (Fig. S6). euthanized 4 d later. Splenocytes and pancreatic lymph node cells Erythrocyte binding also substantially attenuated humoral im- were harvested, and BDC2.5 CD4+ T-cell proliferation was mune responses against antigen, because mice treated with ERY1- assessed by dilution of the CFSE fluorophore via flow cytometry. OVA exhibited 100-fold lower antigen-specific serum IgG titers Consistent with results obtained with ERY1-OVA and TER119- compared with mice treated with soluble OVA (Fig. 5F). A similar SIINFEKL, erythrocyte-binding p31 induced potent BDC2.5 reduction in OVA-specific IgG titer reduction as a result of eryth- T-cell proliferation and deletion, because TER119-p31–treated rocyte binding was seen in nonadoptively transferred C57BL/6 mice harbored 20-fold and 28-fold fewer nonproliferated BDC2.5 (CD45.2+) mice. Following two i.v. injections of 1 μg of OVA or CD4+ T cells compared with mice treated with p31 in the spleen ERY1-OVA 7 d apart, ERY1-OVA–treated mice exhibited 39.8- and pancreatic lymph nodes, respectively (Fig. 6A). fold lower OVA-specific serum IgG levels 19 d after the first an- To assess the effects of TER119-p31–driven proliferation in tigen injection (Fig. S7). This apparent reduction in B-cell acti- a pathological autoimmune setting, we performed an analogous vation, following erythrocyte binding by the antigen, corroborates yet longer term adoptive transfer study to allow the diabetogenic

Kontos et al. PNAS | Published online December 17, 2012 | E65 Downloaded by guest on September 27, 2021 A B naive saline

0.5 p31 * insulin 0.4 TER119-p31 CD3 ε

0.3 * TER119-p31 DAPI 0.2 (% of CD4+)

0.1 insulin

Non-proliferating BDC2.5 cells p31 CD4

0 CD3 ε spleen pLN CFSE DAPI p31 TER119-p31 C D 600 100

saline 80 450 saline p31 p31 *** TER119-p31 60 *** 300 TER119-p31 40 150

Blood glucose (mg/dL) 20 Diabetes incidence (%) i.v. antigen 0 1234567891110 12 49 55 62 1234567891110 1262 Days following adoptive transfer Days following adoptive transfer

Fig. 6. Erythrocyte-binding autoantigen protects mice from T cell-induced autoimmune type 1 diabetes. (A) Increased proliferation and deletion of adoptively transferred diabetogenic BDC2.5 CD4+ T cells in the spleen and pancreatic lymph nodes 4 d following administration of TER119-p31 compared with p31, 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 of TER119-p31-treated mice. (Scale bar = 100 μm.) (C) Glycemia monitoring as measured by daily blood glucose measurements following adoptive transfer of diabetogenic 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.

BDC2.5 CD4+ T cells to induce autoimmune diabetes. Follow- reasons, a long-standing goal in the induction of immunological ing adoptive transfer of activated BDC2.4 CD4+ T cells, NOD tolerance is achieving lasting antigen-specific tolerance and at- mice were i.v. administered 10 μg of TER119-p31 or an equi- tenuation of autoreactive T and B cells to within limits controllable molar dose of p31 every 3 d, for a total of three doses. Mice by peripheral tolerance mechanisms (33). Tolerogenic regimens treated with p31 peptide or saline exhibited profound hyperglyce- aimed at inducing regulatory T-cell (Treg) populations by systemic mia as early as 3 d following adoptive transfer (Fig. 6C), with 100% soluble antigen administration have been attempted in the pre- of mice developing diabetes by days 5 and 6, respectively (Fig. 6D). vention of type 1 diabetes but have demonstrated little efficacy in Conversely, mice treated with TER119-p31 remained normogly- the clinic (34). Tregs have been extensively implicated as key cemic throughout the experimental time course (Fig. 6C), with 0% players in peripheral tolerance maintenance, and as such, several incidence of diabetes 62 d following adoptive transfer (Fig. 6D). therapeutic strategies aimed at their induction in vivo and even Histological confirmation of pathological findings showed low in- direct administration are currently being explored (35). Antigen sulin expression and extensive T-cell infiltration (CD3e+)into therapy may further require administration of multiple antigenic pancreatic islets of mice treated with saline or p31 peptide, com- epitopes carefully selected by in-depth clinical studies (36) or pared with robust insulin expression and no evident T-cell presence biochemically modified for enhanced MHC binding (37). With in mice administered TER119-p31 (Fig. 6B). These data demon- regard to targeted antigen delivery to promote tolerogenic pre- strate the efficacy of erythrocyte-binding antigens in driving clonal sentation, antibody-based antigen targeting to steady-state DCs deletional protection even in the context of preactivated, patho- has shown promise in preclinical studies as a biomolecular ap- genic CD4+ T cells in an autoimmune setting. proach to exploit the ability of APCs to induce clonal CD8+ T-cell anergy or deletion (38, 39). Discussion Here, we explore immunotherapy by antigen-specific T-cell Recent studies in the field of immune tolerance, motivated by the deletion, reasoning that systemic, quantitative removal of quies- search for mechanistic understanding, have expanded the classic cent or activated T cells from circulation would be beneficial in dogma of central tolerance induction to a far more interconnected a number of applications. In our approach, we engineer an antigen and dynamic process involving a cohort of cellular interactions to bind circulating erythrocytes strongly and specifically, exploiting responding amid a complex extracellular milieu. Aside from small- natural regulatory mechanisms induced by apoptotic (here, eryp- molecule–based immune suppressants, recent clinical trials using totic) cell-derived antigen (3, 4, 40), yet without any ex vivo cellular biologics-based broad-spectrum T cell- and -depleting mol- manipulation. Several molecular determinants of apoptotic cell- ecules for the treatment of autoimmune disorders and GvHD have driven tolerance have been elucidated, and many more have been shown promising, yet mixed, results (1). Their varied extents of proposed (5, 41). The most characterized source of such tolero- success and numerous side effects stem from their mode of action, genic signals presented on apoptotic cells to APCs is phosphati- namely, reducing all T-cell or B-cell populations, as opposed to dylserine, which is recognized by phagocytic cells through either selected removal of only the pathogenic subpopulation. For such direct ligation to receptors or indirect bridging by proteins, such

E66 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al. Downloaded by guest on September 27, 2021 PNAS PLUS as milk-fat globule EGF factor 8, growth arrest-specific 6 protein, ERY1 Peptide Design. The ERY1 peptide was H2N-WMVLPWLPGTLDGGSGCRG- and protein S (41, 42). Our results demonstrate that glycophorin- CONH2, and the MIS peptide was H2N-PLLTVGMDLWPWGGSGCRG-CONH2. A targeted antigen can strongly and specifically bind both eryp- The underlined sequence is the ERY1 12-mer sequence that we previously totic and normal (which will continue to age) erythrocytes in vivo discovered by phage display as a mouse glycophorin-A binder (16). The GGSG region served as a linker to the cysteine residue used for conjugation; (Fig. 2), and thus exploit the eryptotic erythrocyte as a toler- the flanking arginine residue served to lower the pKa, and thus to increase ogenic vehicle. the reactivity of the cysteine residue (46). SIINFEKL and p31 (YVRPLWVRME) Given the plasticity of T-cell subsets and their dynamic re- peptides were purchased from Genscript. Detailed methods are provided in sponse to antigens in vivo, phenotypic changes are key to the in- SI Materials and Methods. duction of tolerance vs. . Blockade of the costimulatory CD28/B7 signaling axis between T cells and APCs by cytotoxic T- ERY1-Antigen Design. The ERY1 or MIS peptide was conjugated via its free antigen-4 serves as a constitutive regulatory control cysteine thiol to lysine side chains on OVA using the sulfosuccinimidyl-4-(N- switch, and has been exploited as a tolerogenic regimen for auto- maleimidomethyl) cyclohexane-1-carboxylate linker (Thermo Scientific). As immunity treatment (43). PD-1 signaling on T cells by its ligands a control, the ERY1 peptide was conjugated to GST derived from Escherichia coli. For biodistribution studies, the ERY1 or MIS peptide was conjugated to PD-L1 and PD-L2 represents another regulatory mechanism of allophycocyanin (Innova Biosciences). Methods are shown in SI Materials immune responses, and it has been extensively studied in the and Methods. context of autoimmunity and T-cell exhaustion and deletion (44). Selective blockade of the PD-1/PD-L1 signaling axis was shown to TER119-Antigen Development. The TER119 scFv antibody fragment was cre- disrupt insulin-coupled splenocyte tolerance induction in the ated as described elsewhere (45). In brief, forward and reverse primer mix- BDC2.5 model by blocking T-cell receptor stop signals, thus de- tures were used to amplify the variable domains of the heavy and light creasing T-cell motility and increasing physical interactions with chains of cDNA of the TER119 hybridoma clone (provided by Shozo Izui, APCs (45). Our results with erythrocyte-binding antigens are con- University of Geneva, Geneva, Switzerland). The products were digested and sistent with the implication of PD-1 in tolerogenesis, because ligated into the pMAZ360 subcloning vector, and the resultant sequence was marked expression of the regulatory marker was seen in pro- queried in the ImMunoGeneTics database (www.imgt.org) to determine the relevant VH and VL domains. Using assembly PCR, the VH and VL genes were liferative generations of antigen-specific T cells following ERY1- then fused with a (G4S)3 linker and cloned into the SfiI and XbaI sites on OVA and TER119-SIINFEKL administration (Fig. 4E). Here, we pSecTagA (Invitrogen). From this parent vector, SIINFEKL and/or p31 was demonstrate that erythrocyte-binding antigens induce PD-1 sig- then additionally cloned into the 3′ end of the TER119 scFv gene using as- naling to drive deletional proliferation of clonal T-cell populations, sembly PCR. The TER119-antigen constructs were expressed in suspension both in the context of xenoantigens (Figs. 4 and 5I) and auto- culture of HEK293E cells under serum-free conditions with 3.75 mM valproic antigens (Fig. 6A). acid (Sigma–Aldrich) (47) for 5 d and purified from supernatant using Motivated by the success seen in mouse models in induction immobilized metal ion affinity chromatography on an Akta FPLC system (GE fi of antigen-specific tolerance by antigen conjugated to or loaded Healthcare). Puri ed proteins were analyzed for purity using SDS/PAGE, for endotoxin level using THP-1 × Blue cells (InvivoGen), and for concentration within apoptotic cells, we aimed to engineer a simple, more readily using bicinchoninic acid assays (Thermo Scientific). The final product was translatable molecular approach that might achieve analogous sterile-filtered and stored at −80 °C in working aliquots. ends. We demonstrated two such biomolecular designs, one with protein antigen conjugated to multiple copies of a glycophorin A- OTI T-Cell Adoptive Transfer. CD8+ T cells from OTI (CD45.2+) mouse spleens binding peptide and another with a peptide or protein antigen were isolated using a CD8 magnetic bead negative selection kit (Miltenyi fused to a scFv antibody fragment specific for the same erythro- Biotec) as per the manufacturer’s instructions. Freshly isolated CD8+ OTI cells cyte-specific target. These demonstrated deletions of both CD8+ were resuspended in PBS and labeled with 1 μM CFSE (Invitrogen) for 6 min and CD4+ T cells, both antigen-experienced (in the case of the at room temperature, and the reaction was quenched for 1 min with an ’ fi ’ CD4+ T cells in the BDC2.5 model) and antigen-naive (in the equal volume of Iscove s modi ed Dulbecco s medium (IMDM) with 10% + (vol/vol) FBS (Gibco). Cells were washed, counted, and resuspended in pure case of the CD8 T cells in the OTI model). Although much work IMDM before injection. A total of 3 × 106 CFSE-labeled CD8+ OTI cells were remains, such as exploring reversal of ongoing autoimmunity in injected i.v. into the tail vein of recipient CD45.1+ mice. For short-term a spontaneous onset model and determining the longevity of de- proliferation studies, 10 μg of ERY1-OVA or OVA or an equimolar dose of letion (which will likely depend on the age of the host), the ability SIINFEKL or TER119-SIINFEKL in a 100-μL volume was injected 24 h following to delete activated pathogenic islet-specificCD4+ Tcells,and adoptive transfer. Splenocytes were harvested 5 d following antigen ad- thereby completely protect islets from destruction, leads us to ministration and stained for analysis by flow cytometry. believe this approach has the potential to serve as a simple 5 + yet effective biomolecular modification to existing antigens for OTI Challenge Model. A total of 3 × 10 CFSE-labeled OTI CD8 T cells were fi injected into CD45.1+ recipient mice as described above. At 1 and 6 d fol- antigen-speci c peripheral deletion of T cells. Because of the μ modular nature of the molecular designs presented above, this lowing adoptive transfer, mice were i.v. administered 10 g of ERY1-OVA or OVA or an equimolar dose of SIINFEKL or TER119-SIINFEKL in 100 μL of saline approach may be relevant to treatment of a number of autoim- into the tail vein. At 15 d following adoptive transfer, mice were challenged IMMUNOLOGY mune diseases dependent on pathogenic T cells. with 5 μg of OVA and 25 ng of ultrapure E. coli LPS (InvivoGen) in 25 μLof saline injected intradermally into each rear leg pad (Hock method: total dose Materials and Methods of 10 μg of OVA and 50 ng of LPS). Mice were killed 4 d following challenge, Animals. The Swiss Veterinary Authority and the EPFL Centre d’Application du and spleen and draining lymph node cells were isolated for restimulation. Vivant previously approved all animal procedures. Female C57BL/6 mice For flow cytometry analysis of intracellular , cells were restimulated (Charles River Laboratories) aged 8–12 wk were used for in vivo binding in the presence of 1 mg/mL OVA or 1 μg/mL SIINFEKL peptide (Genscript) for studies and as E.G7-OVA tumor hosts. C57BL/6-Tg(TcraTcrb) 1100Mjb (OTI) 3 h. Brefeldin-A (5 μg/mL; Sigma) was added, and restimulation was resumed mice (Jackson Laboratories) were bred in specific pathogen-free (SPF) con- for an additional 3 h before staining and flow cytometry analysis. For ELISA ditions at the Ecole Polytechnique Fédérale Lausanne Animal Facility, and measurements of secreted factors, cells were restimulated in the presence of females were used for splenocyte isolation at 6–12 wk of age. Female B6.SJL- 100 μg/mL OVA or 1 μg/mL SIINFEKL peptide for 4 d. Cells were spun, and the PtprcaPepcb/Boy (CD45.1) mice (Charles River Laboratories) aged 8–12 wk media were collected for ELISA analysis using IFN-γ and IL-10 Ready-Set-Go were used as recipient hosts for OTI CD8+ T-cell adoptive transfer and tol- kits (eBioscience) as per the manufacturer’s instructions. OVA-specific serum erance induction studies. NOD.Cg-Tg(TcraBDC2.5, TcrbBDC2.5)1Doi/DoiJ IgG was detected by incubating mouse serum at varying dilutions on OVA- mice (BDC2.5 mice), which express the T-cell receptor transgene for specific coated plates, followed by a final incubation with goat anti-mouse IgG-HRP recognition of the chromogranin-A islet antigen, were bred in SPF con- (Southern Biotech). ditions at the Ecole Polytechnique Fédérale Lausanne Animal Facility, and females were used for cell isolations. NOD/ShiLtJ mice (Charles River Labo- OTI E.G7-OVA Tolerance Model. A total of 1 × 106 CFSE-labeled OTI CD8+ T cells ratories) were used as BDC2.5 adoptive transfer hosts. were injected into 8- to 12-wk-old female C57BL/6 mice as described above.

Kontos et al. PNAS | Published online December 17, 2012 | E67 Downloaded by guest on September 27, 2021 At 1 and 6 d following adoptive transfer, mice were i.v. administered 10 μg transfer, mice were i.v. administered saline, 10 μg TER119-p31, or an equi- of ERY1-OVA or 10 μg of OVA in 100 μL of saline into the tail vein. Blood was molar dose of p31 peptide every 3 d for three doses. Diabetes onset was drawn 5 d following adoptive transfer for characterization of OTI CD8+ T-cell monitored by measuring nonfasting blood glucose levels using an Accu- proliferation by flow cytometry. OVA-expressing EL-4 thymoma cells (E.G7- Check Aviva glucometer (Roche). Mice were considered diabetic at blood OVA, CRL-2113; American Type Culture Collection) were cultured as per glucose readings ≥250 mg/dL. After three hyperglycemic readings, mice American Type Culture Collection guidelines. In brief, cells were cultured in were euthanized and organs were harvested for histology. RPMI 1640 medium supplemented with 10% (vol/vol) FBS, 10 mM Hepes, 1 mM sodium pyruvate, 0.05 mM β-mercaptoethanol, 1% puromycin/strep- Data Analysis. All flow cytometry data were analyzed using FlowJo (v8.8.6 and tomycin (Invitrogen Gibco), and 0.4 mg/mL G418 (PAA Laboratories). Just v9.2; TreeStar). Graphing and statistical analyses of worked-up data were before injection, cells were expanded in media without G418 and resus- performed using Prism (v5; GraphPad). Mann–Whitney U tests were used for pended on harvest in HBSS (Gibco). At 9 d following adoptive transfer, mice interpreting flow cytometry population data, two-tailed Student t tests were were anesthetized with isoflurane, the back area was shaved, and 1 × 106 E. used for interpreting ELISA data, and ANOVA was used for interpreting OTI G7-OVA cells were injected intradermally between the shoulder blades. At < < < – 4 d following E.G7-OVA graft, tumor dimensions were measured every 24 h generation data (***P 0.001; **P 0.01; *P 0.05). Log rank (Mantel Cox) – with a digital caliper and tumor volume was calculated as an ellipsoid tests were used to analyze Kaplan Meier plots of diabetes onset data. (V ¼ π= l ·w · h; where V is volume, l is length, w is width, and h is the height of Additional detailed materials and methods are available in SI Materials 6 the tumor). At 15 d following adoptive transfer, mice were challenged with 5 μg and Methods. of OVA and 25 ng of ultrapure E. coli LPS in 25 μL of saline injected intradermally into each rear leg pad (total dose of 10 μgofOVAand50ngofLPS). ACKNOWLEDGMENTS. We thank C. Nembrini and J. K. Eby for helpful discussions; Prof. M. A. Swartz (Ecole Polytechnique Fédérale Lausanne) for BDC2.5 Diabetes Induction Model. Freshly isolated splenocytes from female helpful discussions and critically reading the manuscript; and P. C. Henrioud, C. Dessibourg, S. Thurnheer, V. Borel, and M. Pasquier for technical assis- BDC2.5 transgenic mice were stimulated for 4 d in DMEM supplemented with β tance. We also thank Prof. S. Izui (University of Geneva, Geneva, Switzer- 10% (vol/vol) FBS, 0.05 mM -mercaptoethanol, 1% puromycin/streptomycin, land) for providing the TER119 hybridoma mRNA and Prof. K. Kretschmer μ and 0.5 M p31 peptide. Following stimulation, cells were washed with basal (Center for Regenerative Therapies, Dresden, Germany) for helpful advice DMEM and analyzed for purity by flow cytometry, and 5 × 106 T cells were concerning the NOD.BDC2.5 model. Funding was provided by the Ecole Poly- i.v. injected into normoglycemic NOD/ShiLtJ mice. At 8 h following adoptive technique Fédérale Lausanne School of Life Sciences.

1. Bluestone JA, et al. (2010) The Immune Tolerance Network at 10 years: Tolerance 25. Lee MS, Kim Y-J (2007) Signaling pathways downstream of pattern-recognition research at the bedside. Nat Rev Immunol 10(11):797–803. receptors and their cross talk. Annu Rev Biochem 76:447–480. 2. De Groot AS, Scott DW (2007) Immunogenicity of protein therapeutics. Trends 26. Arnaboldi PM, Roth-Walter F, Mayer L (2009) Suppression of Th1 and Th17, but not Immunol 28(11):482–490. Th2, responses in a CD8(+) T cell-mediated model of oral tolerance. Mucosal Immunol 3. Mueller DL (2010) Mechanisms maintaining peripheral tolerance. Nat Immunol 11(1): 2(5):427–438. – 21 27. 27. Saint-Lu N, et al. (2009) Targeting the to oral dendritic cells with mucoadhesive 4. Green DR, Ferguson T, Zitvogel L, Kroemer G (2009) Immunogenic and tolerogenic cell chitosan particles enhances tolerance induction. Allergy 64(7):1003–1013. – death. Nat Rev Immunol 9(5):353 363. 28. Miller SD, Turley DM, Podojil JR (2007) Antigen-specific tolerance strategies for the fi 5. Grif th TS, Ferguson TA (2011) Cell death in the maintenance and abrogation of prevention and treatment of autoimmune disease. Nat Rev Immunol 7(9):665–677. fi – tolerance: The ve Ws of dying cells. Immunity 35(4):456 466. 29. Haskins K, McDuffie M (1990) Acceleration of diabetes in young NOD mice with 6. Rothlin CV, Ghosh S, Zuniga EI, Oldstone MBA, Lemke G (2007) TAM receptors are a CD4+ islet-specificTcellclone.Science 249(4975):1433–1436. pleiotropic inhibitors of the innate immune response. Cell 131(6):1124–1136. 30. Judkowski V, et al. (2004) Peptide specific amelioration of T cell mediated 7. Ferguson TA, Choi J, Green DR (2011) Armed response: How dying cells influence T- pathogenesis in murine type 1 diabetes. Clin Immunol 113(1):29–37. cell functions. Immunol Rev 241(1):77–88. 31. Stadinski BD, et al. (2010) Chromogranin A is an autoantigen in type 1 diabetes. Nat 8. Fife BT, et al. (2006) Insulin-induced remission in new-onset NOD mice is maintained Immunol 11(3):225–231. by the PD-1-PD-L1 pathway. J Exp Med 203(12):2737–2747. 32. Nikoopour E, et al. (2011) Vasostatin-1-derived peptide ChgA29-42 is an antigenic 9. Godsel LM, et al. (2001) Prevention of autoimmune myocarditis through the induction of antigen-specific peripheral immune tolerance. Circulation 103(12):1709–1714. epitope of diabetogenic BDC2.5 T cells in nonobese diabetic mice. J Immunol. 10. Martin AJ, et al. (2010) Ethylenecarbodiimide-treated splenocytes carrying male CD4 33. Sakaguchi S, Powrie F, Ransohoff RM (2012) Re-establishing immunological self- – epitopes confer histocompatibility Y chromosome antigen transplant protection by tolerance in autoimmune disease. Nat Med 18(1):54 58. – inhibiting CD154 upregulation. J Immunol 185(6):3326–3336. 34. Diabetes Prevention Trial Type 1 Diabetes Study Group (2002) Effects of insulin in – 11. Smarr CB, Hsu CL, Byrne AJ, Miller SD, Bryce PJ (2011) Antigen-fixed leukocytes relatives of patients with type 1 diabetes mellitus. N Engl J Med 346(22):1685 1691. tolerize Th2 responses in mouse models of allergy. J Immunol 187(10):5090–5098. 35. Wood KJ, Bushell A, Jones ND (2011) Immunologic unresponsiveness to alloantigen in 12. Liu K, et al. (2002) Immune tolerance after delivery of dying cells to dendritic cells in vivo: A role for regulatory T cells. Immunol Rev 241(1):119–132. situ. J Exp Med 196(8):1091–1097. 36. Bielekova B, et al. (2004) Expansion and functional relevance of high-avidity myelin- 13. Berg CP, et al. (2001) Human mature red blood cells express caspase-3 and caspase-8, specific CD4+ T cells in multiple sclerosis. J Immunol 172(6):3893–3904. but are devoid of mitochondrial regulators of apoptosis. Cell Death Differ 8(12): 37. Gabrysová L, Wraith DC (2010) Antigenic strength controls the generation of antigen- 1197–1206. specific IL-10-secreting T regulatory cells. Eur J Immunol 40(5):1386–1395. 14. Lang KS, et al. (2003) Cation channels trigger apoptotic death of erythrocytes. Cell 38. Mukhopadhaya A, et al. (2008) Selective delivery of beta cell antigen to dendritic cells Death Differ 10(2):249–256. in vivo leads to deletion and tolerance of autoreactive CD8+ T cells in NOD mice. Proc 15. Turley DM, Miller SD (2010) Prospects for antigen-specific tolerance based therapies Natl Acad Sci USA 105(17):6374–6379. for the treatment of multiple sclerosis. Results Probl Cell Differ 51:217–235. 39. Stern JNH, et al. (2010) Promoting tolerance to proteolipid protein-induced 16. Kontos S, Hubbell JA (2010) Improving protein pharmacokinetics by engineering experimental autoimmune encephalomyelitis through targeting dendritic cells. Proc fi – erythrocyte af nity. Mol Pharm 7(6):2141 2147. Natl Acad Sci USA 107(40):17280–17285. 17. Khandelwal S, Saxena RK (2006) Assessment of survival of aging erythrocyte in 40. Griffith TS, et al. (2007) Apoptotic cells induce tolerance by generating helpless CD8+ circulation and attendant changes in size and CD147 expression by a novel two step T cells that produce TRAIL. J Immunol 178(5):2679–2687. – biotinylation method. Exp Gerontol 41(9):855 861. 41. Nagata S, Hanayama R, Kawane K (2010) Autoimmunity and the clearance of dead 18. Yamazaki S, et al. (2008) CD8+ CD205+ splenic dendritic cells are specialized to induce cells. Cell 140(5):619–630. Foxp3+ regulatory T cells. J Immunol 181(10):6923–6933. 42. Ravichandran KS (2011) Beginnings of a good apoptotic meal: The find-me and eat- 19. Holz LE, Warren A, Le Couteur DG, Bowen DG, Bertolino P (2010) CD8+ T cell me signaling pathways. Immunity 35(4):445–455. tolerance following antigen recognition on hepatocytes. J Autoimmun 34(1):15–22. 43. Bour-Jordan H, et al. (2011) Intrinsic and extrinsic control of peripheral T-cell 20. Ichikawa S, Mucida D, Tyznik AJ, Kronenberg M, Cheroutre H (2011) Hepatic stellate tolerance by costimulatory molecules of the CD28/B7 family. Immunol Rev 241(1): cells function as regulatory bystanders. J Immunol 186(10):5549–5555. – 21. Thomson AW, Knolle PA (2010) Antigen-presenting cell function in the tolerogenic 180 205. liver environment. Nat Rev Immunol 10(11):753–766. 44. Francisco LM, Sage PT, Sharpe AH (2010) The PD-1 pathway in tolerance and – 22. Albert ML, et al. (1998) Immature dendritic cells phagocytose apoptotic cells via autoimmunity. Immunol Rev 236:219 242. alphavbeta5 and CD36, and cross-present antigens to cytotoxic T . J Exp 45. Fife BT, et al. (2009) Interactions between PD-1 and PD-L1 promote tolerance by – Med 188(7):1359–1368. blocking the TCR-induced stop signal. Nat Immunol 10(11):1185 1192. 23. Bursch LS, Rich BE, Hogquist KA (2009) Langerhans cells are not required for the CD8 46. Lutolf MP, Tirelli N, Cerritelli S, Cavalli L, Hubbell JA (2001) Systematic modulation of T cell response to epidermal self-antigens. J Immunol 182(8):4657–4664. Michael-type reactivity of thiols through the use of charged amino acids. Bioconjug 24. Darrah PA, et al. (2010) IL-10 production differentially influences the magnitude, Chem 12(6):1051–1056. quality, and protective capacity of Th1 responses depending on the vaccine platform. 47. Baldi L, Hacker DL, Meerschman C, Wurm FM (2012) Large-scale transfection of J Exp Med 207(7):1421–1433. mammalian cells. Methods Mol Biol 801:13–26.

E68 | www.pnas.org/cgi/doi/10.1073/pnas.1216353110 Kontos et al. Downloaded by guest on September 27, 2021