Sensitization to self (virus) antigen by in situ expression of murine interferon gamma

(1)

Sensitization to self (virus) antigen by in situ

expression of murine interferon-gamma.

M S Lee, … , M B Oldstone, N Sarvetnick

J Clin Invest.

1995;

95(2)

:486-492.

https://doi.org/10.1172/JCI117689

.

Autoimmune disease results from inflammatory destruction of tissues by aberrant

self-reactive lymphocytes. We studied the autoimmune potential of T lymphocytes

immunologically ignorant of viral antigens acting as self antigens and whether the host

defense molecule IFN-gamma could stimulate these cells to cytotoxic competency. For this

purpose, we produced double transgenic mice expressing pancreatic IFN-gamma as well

as lymphocytic choriomeningitis virus (LCMV) nucleoprotein (NP) or glycoprotein (GP)

antigen. 100% of the NP+/IFN-gamma+ mice became diabetic before 2 mo of age, while

none of the NP single transgenic littermates and only 10% of IFN-gamma single transgenic

littermates did. Strikingly, NP+/IFN-gamma+ mice spontaneously developed cytotoxic T

lymphocyte activity on LCMV-infected targets and vaccinia virus-NP-infected ones without

prior LCMV infection but NP+/IFN-gamma- mice did not, which indicates specific

sensitization to the viral antigen by IFN-gamma. These results suggest that lymphocytes

ignorant of self antigens can be activated by IFN-gamma released after immunologic

stimulation such as viral infection. This mechanism may account for the loss of apparent

tolerance to self antigens in autoimmune diseases such as insulin-dependent diabetes

mellitus.

Research Article

(2)

Sensitization

to

Self

(Virus)

Antigen by In

Situ Expression

of

Murine Interferon-y

Myung-ShikLee, MatthiasvonHerrath, Hans Reiser,* Michael B. A. Oldstone,andNora Sarvetnick

Department ofNeuropharmacology, The Scripps Research Institute, La Jolla, California 92037;and*DepartmentofPathology, Harvard Medical School, Boston, Massachusetts 02115

Abstract

Autoimmunedisease results from inflammatory destruction oftissuesby aberrantself-reactivelymphocytes. We studied the autoimmune potential of T lymphocytes immunologi-cally ignorantof viral antigensacting as self antigens and whether the host defense molecule IFN-y could stimulate these cells to cytotoxic competency. For this purpose, we

produced double transgenic mice expressing pancreatic

IFN-y

as well as lymphocytic

choriomeningitis

virus

(LCMV) nucleoprotein (NP)or glycoprotein (GP) antigen.

100% ofthe

NP+/IFN-y'

mice became diabetic before 2 mo of age, while none ofthe NP single transgenic littermates and only 10% of

IFN-y

single transgenic littermates did.

Strikingly,

NP'/IFN-y'

micespontaneously developed

cyto-toxicTlymphocyte

activity

onLCMV-infected targets and vacciniavirus-NP-infectedones withoutprior LCMV infec-tionbut

NPf/IFN-y-

mice didnot,which indicates specific sensitization to the viral antigen by

IFN-y.

These results suggest that lymphocytes ignorantofselfantigens can be activated by

IFN-y

released afterimmunologic stimulation

suchasviral infection.Thismechanismmay account for the loss ofapparent tolerance to selfantigens in autoimmune diseases such as insulin-dependent diabetes mellitus. (J. Clin. Invest. 1995.95:486-492.) Key words: autoimmunity

*transgenic mice*pancreatic islets * diabetes *B7-1

Introduction

Thequestionof howapathogenicautoimmune response devel-ops has been of interest to immunologists and clinicians for many years. In many autoimmunediseases,sensitized and acti-vatedautoreactiveTlymphocytesdestroy target cellsharboring

thecorresponding tissue-specific antigens.Ininsulin-dependent

diabetes mellitus patients, circulating autoreactive T

lympho-cytes thatrespondtotheislet/brainantigen glutamicacid

decar-boxylase have been demonstrated (1-3). If an understanding

of theeventsthatcausetheselymphocytestobecome sensitized

to antigens such as glutamic acid decarboxylase could be

at-AddresscorrespondencetoNoraSarvetnick, Departmentof Neurophar-macology,CVN-10,TheScripps Research Institute, 10666NorthTorrey

PinesRoad,LaJolla,CA92037.Phone:619-554-7066;FAX:

619-554-6477.

Receivedfor publication19July1994 andinrevisedform20

Sep-tember1994.

tained, then development ofpreventativetreatment for autoim-mune diseases might follow.

Tostudy this process of lymphocyte sensitization in insulin-dependent diabetes mellitus, weinitially addressed the conse-quences of targeted expression of the host defense molecule

IFN-y

in the pancreatic islets (Ins-IFN-y mice) (4, 5). These studies demonstrated that islet expression of IFN-y does lead

toimmunological sensitization to wholepancreaticislets, how-ever, the target antigenwas notidentified,and thusthe mecha-nism ofautoimmunity couldnotbe studied. Specifically, it is not known whether sequestered antigens within islets are

en-countered by lymphocytes for the first time after islet damage

orwhether circulating cells within the normal repertoire become

activated todestroy their target when theyareincontact with the

IFN

in the pancreas.

To address these mechanistic issues, we used transgenic mice that expresslymphocytic choriomeningitisvirus

(LCMV)'

antigenontheirpancreatic

,1

cells(RIP-LCMV mice)toprovide

aunique tissue-specific antigen towardwhich thecell-mediated

and humoral immune responses can be monitored. The RIP-LCMVtransgenicmicearenormoglycemicand free ofinsulitis;

however, they developautoimmunediabetes when infected with LCMV (6, 7). These results suggested that lymphocytes that could reacttoperipheral self antigensare notdeleted but have

neverencountered these antigens because of theirlocationon

non-antigen-presentingcells

(APC).

Itispossiblethatthe acti-vation of such T cells specific for self antigensmaybe mediated via inflammatory cytokines elicited by viral infection.

To determine whether T cells that hadnever encountered

selfantigens could be activated by inflammatorycytokines,we

crossed Ins-IFN-y mice withRIP-LCMVmice. LCMVantigen

wasin the form of either the viralglycoprotein (GP)or

nucleo-protein (NP). The incidence of spontaneous diabetes,

histopa-thology, andanti-viralcytotoxicTlymphocyte (CTL)responses

werestudied. Strikingly, weobserved the spontaneous genera-tionof CTLspecificforself(viral) antigenindoubletransgenic

mice but not in LCMV single transgenic littermates. Since

LCMV-specificTcells would be present in both the double and single transgenic mice (6, 7),ourdata suggest that IFN-y can

induce thedevelopmentof CTL fromcirculatingT

lymphocytes

within the repertoireofanormal individual.

Methods

Doubletransgenicmice.TheIns-IFN-ymurine linewasmaintainedby repeated crossingswithBALB/c mice.RIP-LCMV-GPmice(line

34-1.Abbreviationsused in thispaper:APC,antigen-presentingcell; CTL, cytotoxicT lymphocyte; LCMV, lymphocytic choriomeningitis virus; NP, nucleoprotein; VV,vaccinia virus.

J.Clin.Invest.

C TheAmericanSocietyforClinicalInvestigation,Inc.

0021-9738/95/02/0486/07 $2.00

(3)

Table I. Incidence of Spontaneous Diabetes* in LCMV/IFN-y Double Transgenic Mice

NP+ orGP' IFN-y+

Both+ only only

Both-LCMV-NP/IFN-y 13/13t 0/6 1/10 0/10 LCMV-GP/IFN-y 2/9 0/7 4/9 0/9

*_{Diabetes was confirmed by a glucose level of >300 mg/dl in}_monthly

determinations of ocular blood samples over the 4-mo observation pe-riod. *_The_incidence_{of diabetes in}_NP+fIFN-y'_mice_was_{significantly} higher than that in all other groups of mice (P < 0.005).

20) were selected from those maintained by repeated crossings to

C57BL/6. Ins-IFN-y mice were intercrossed to RIP-LCMV-NP (line

25-3) or RIP-LCMV-GP mice to derive lines positive for both transgenes. Fl mice were screened byPCRamplification ofgenomic

DNA.NP-specific primerswereCAGTTATAGGTGCTCTTCCGCand

AGATCTGGGAGCCTTGCTLTG. GP-specific primers were

CGC-CGGTCTFTlGCATGTTCTAG and

GCACATTCACCTGGACTT-TGTC. Primers for human insulin promoter sequences were CCTGGT-CTAATGTGGAAAGTG and TGCAATI7CCGGACCATTTCC. 20

Mtg/ml

tail DNA, 1 tM each primer, 200 1M each deoxynucleoside

triphosphate, 2 mMMgCl2, and 40 U/ml Taqpolymerase weremixed for PCR. Annealing and polymerization were done at 60 and720C,

respectively, for a total of 25 cycles. The incidence of spontaneous diabetes inFlmicewasdeterminedby monthlydeterminationofblood glucose level over the 4-mo observation period. In some mice, weekly measurementof blood glucose levelwasdone. Diabeteswasconfirmed

byaglucoselevelof >300mg/dlin ocular bloodsamplestestedwith

aGlucometer3 (Miles Inc.,Elkhart,IN).

Immunohistochemistry. Fresh-frozen sectionsofthepancreatic tissue

were incubated with anti-CD4(PharMingen,San Diego, CA),anti-CD8

(PharMingen), anti-B220 (PharMingen), anti-Mac-1 (Boehringer

Mannheim,Indianapolis, IN), F4/80 (Serotec Ltd., Kidlington,Oxford,

UnitedKingdom),oranti-B7-1antibody (8) for 30 minat room

temper-ature.Theconcentration ofprimary mAbs was between 5 and 10

Mg/

ml. Incubations with biotinylated secondary antibody and then with

avidin-biotin-peroxidase complex followed for 30 min each. After the color reactionwithdiaminobenzidine,thesections were counterstained inhematoxylin solutionormethylgreen forhistopathologic examina-tion. We evaluated 15-30 islets from 2-3 histological sectionsofeach mouse.

Doubleimmunofluorescence.Fresh-frozensections ofpancreatic

tis-suewereincubatedwith either one of anti-B220,anti-Mac-l,orF4/80

antibodyfollowedbyincubation withtetramethylrhodamine

isothiocya-nate (TRITC)-labeled anti-rat IgG for 30 min at room temperature. They were sequentially incubated with anti-B7-1 antibody and then FITC-labeled anti-hamsterIgGfor30 min,respectively.Control

experi-mentsusing one of the primary antibodies and an irrelevant one of the

twosecondaryantibodieswereperformedandconfirmed thespecificity

ofourresults.

CTLresponse. SecondaryCTLactivitywasmeasured aspreviously

described(9).Splenocytes and PBLwereculturedfor> 1 wkon macro-phage-coated plates that were infected with LCMV Armstrong at a moi of 1 and then irradiated2,000 rad. Thesemacrophageshad been har-vested 3 dafter intraperitoneal injection of 3.85% thioglycolate. BALB C1.7 (H-2d) orMC57 (H-2b)targets wereinfected with LCMV Arm-strongat amoi of 1 orwith vaccinia virus (VV)-NP (or VV-GP)at a

moi of 3.Uninfectedtarget cells wereusedascontrols. All cells were labeledwith5"Crand incubated with stimulatedlymphocytesatvarious ET ratios.Lymphocytesfrom BALB/c orC57BL/6miceinfected with 2 x 105 plaque forming unitsof LCMV Armstrong servedaspositive

controls.

600-500

-la

a 400

-0 ₃₀₀

-200 -100 0o 0 NP+/AFN+ -0--<- NP+IAFN--0-i3- NP-AFN+

3 4 5 6 7 8 wk

Figure 1. Weekly bloodglucosedeterminationinfour

NP'/IFN-y'

mice. All

NP+/JFN-y'

mice became diabetic(>300mg/dl)between5 and 6 wk withwasting.OneNP-/IFN-y' mouseandone

NP+/IFN-y-mouse werenotdiabeticalthoughtheformer showed mild

hyperglyce-mia.Bloodglucoselevel of500mg/dlat8 wkofagemeans >500 mg/dl.

ELISAofanti-LCMVantibody.Humoralimmune responsetoLCMV

protein was assayed using an ELISA accordingto a modification of

previously described methods (9). Aftercoating eachwell of 96-well plates with 1 pg of LCMV antigen and blocking with 10% bovine serum/0.2% Tween20, serial dilutions ofserumfrom 1:10to _1:1,000

wereadded,and incubationwasdone for 1 hat roomtemperature. After

washing, peroxidase-conjugated antimurine IgG was added.

Subse-quently,substrate solutioncontainingo-phenylenediamine(Sigma

Im-munochemicals,St. Louis, MO)was added,andabsorptionat 492nm

wasmeasured.

Statisticalanalysis. The Z test wasused tocompare the incidence of spontaneous diabetes between groups.

Results

Development ofdiabetes. To characterize the islet function in double and single transgenic mice, we initially measured the bloodglucose levels. Allof 13

NP'/IFN-y'

double

transgenic

mice became diabetic, butnone of 6

NP'/IFN-y-

and only 1

of 10

NP-/IFN-y'

mice became diabeticoverthe4-mo observa-tion period (Table I). In contrast, two of nine

GP'/IFN-y'

doubletransgenicmice becamediabetic,whereasnoneofseven

GP+IIFN-y- and four of nine

GP-/IFN-y'

littermates became diabetic (Table I). The incidence of diabetes in

NP+/IFN-y'

micewassignificantly higherthanthat in

NP-/IFN-y'

orGP'/

IFN-y'

mice

(both

P<0.005); however,theincidence of

diabe-tesin

GP+/IFN-y'

wasnot

significantly

different from that in

GP-JIFN-y'

mice(P>0.1).All

NPT/IFN-y'

doubletransgenic

mice became diabetic before 2moof ageonthebasis ofmonthly

bloodglucose determination. Weeklyblood glucoselevels on

micefromtypicallitters showed that all fourNPA+IFN-y+ dou-bletransgenicmicebecamediabetic between5and6wkof age

(Fig. 1).Inaddition,all

NP'fIFN-y'

miceshowed emaciationat

the age of2 mo. These suggestedthatprobably all

NP+/IFN-Y'

double

transgenic

mice became diabetic before 6 wk of age. Theother threekinds of mice

(NP-/IFN-y',

GP'AIFN-y',

and

GP-/IFN-y')

developed

diabetes after2 moof age.

Histology

ofthe pancreas.

Histopathological

studiesofthe NPorGP

single transgenic

littermates (n = 4

each)

revealed

no

pancreatic inflammation; however,

in both

NP+/IFN-y+ (n

=

₅₎

_and

_GP'/IFN-y'

_(n = ₄₎_double

_transgenic

miceas well

(4)

Figure 2. Histopathology of the doubletransgenicmice. (A) The pancreas from a 4-mo-old NP+/

IFN-y'mouse showed infiltration of mononuclear cells into the islets that were almost completely de-stroyed(x400).(B) The pancreas from a 4-mo-oldGP'/IFN-y+

mouse showed mononuclear cell infiltration of similar pattern with remaining islet tissue(x400).

mononuclear cell infiltration into the islets was documented with islet destruction ofvariabledegree(Fig. 2, A and B).

Immu-nohistochemical studies showed that infiltrating cells comprised macrophages stained with anti-Mac-IorF4/80,CD4'orCD8'

Tlymphocytes, andalso Blymphocytes stainedwithanti-B220

in both NP'/IFN-y' and GP'/IFN-y' double transgenic mice at the age of2-4 mo (n = 3 each) and in a 3-mo-old IFN-y

single transgeniclittermate (n= 1)(Fig.3, A-D). Weestimated

thatCD4+ Tlymphocytes and CD8+ Tlymphocytesaccounted

j.

for 65 and 15% oftotal infiltrating cells, respectively, whereas

macrophages and B lymphocytes each represented 10% ofthe infiltrating cells.

Spontaneous generation ofCTL againstself(viral)antigens. Todetermine if lymphocytesfrom

NP'/IFN--y'

orGP'/IFN-'y+

mice could be specifically sensitizedtoLCMV antigen without LCMV infection, secondaryCTL assays wereperformed.These

studies revealedthatlymphocytes from three

NP+/IFN-y'

mice testedkilled20, 30, and 41%ofLCMV-infectedtarget cells at

C

whA

B

_D

*: S

.i..

.*

Figure3. Immunohistochemical stainingof the double transgenic pancreas. (A)CD4+

lympho-cytes in the pancreas ofa3-mo-old

NP+/IFN-y+ mouse; (B) CD8+lymphocytesin the

pan-creasof the

NPJ/IFN-y'

mouse;(C) CD4+ lym-phocytesinthe pancreas ofa3-mo-old GP+/

IFN-y+mouse; and (D) CD8+lymphocytesin thepancreas of theGP'/IFN-y+mouse.B

lym-phocytes stained with anti-B220 andMac-I

-positive macrophages werealsoobservedin bothNP+/IFN-y' andGP+/IFN-y+double

transgenic pancreata.

:*

.73

j

(5)

-- InfectedControl

-A-

NP+/IFN+ U

NP+/IFN---- _Uninfected_Target

B

%

Release

75

1

50

25

1:1 20:1

-*- InfectedControl

A

NP+AIFN+

U-

*NP+flFN-- *NP+flFN-- _{ininfadm* Tarmt}

10:1 5:1

a0- InfectedControl

I _GP+IIFN+

U

GP+IIFN-- GP+IIFN-- GP+IIFN-- _Uninfected_Target

m

_ _

5:1 2.5:1 1:1

Figure 4. Secondary CTLresponsetoLCMV antigens in double transgenic mice and single transgenic littermates usedascontrols. Lymphocytes

fromNP'AIFN-y'mice killed both LCMV-infected (A) and VV-NP-infected targets(B)toamoderate degree, whereas lymphocytes from NP+/ IFN-y-miceexerted only negligible killing of thesetargets(A and B). Lymphocytes fromGP+/IFN-y'mice killedasmall portion of

LCMV-infectedtargets(C). Lymphocytes from LCMV-infected mice usedaspositive controls showedstrongCTLresponsetoLCMV antigens (A-C).

an E/T ratio of5 ormore (Fig. 4A). To further characterize

theCTLresponse,weusedVV-NP-infectedtargetcellsinstead

of LCMV-infected targets. Lymphocytes from NP/+IFN-y+

mice killed17, 19,and 31% ofVV-NP-infectedtargetsbutno

VV-GP-infectedtargetsoruninfected cells (Fig. 4B).

Lympho-cytes from three

GP'/IFN-y'

nondiabetic mice killed 12, 14,

and22% of theLCMV-infectedtargets atanE/T ratio of5but

notuninfected targets (Fig. 4 C). CTLresponse of

GP'fIFN-y' micewasnotcharacterized further.

IncontrasttoCTLresponse,thehumoralresponsetoLCMV antigenswasnotdetected in doubletransgenic mice. Sera from

five

NP'/IFN-y',

two

GP'/IFN-y',

one

NP'/IPN-y-,

andone

GP+/IFN-y-

mice did not show any antibody activity in an

ELISA test up to 1:10 dilution, whereas sera from LCMV-infected miceshowed strongantibody activity evenat 1:1,000

dilution.

Expression of costimulatory molecules. Activationor

sensi-tization of naive T cells depends critically onthe recognition

ofcostimulatory signals. We histochemically examined the

ex-pression of B7-1, a membrane-bound costimulatory molecule expressed byAPC(10). We foundthat B7-1wasexpressedon

infiltratingcells in thepancreasof

NP'AIFN-y'

and

GP'/IFN-y'_{double transgenic mice} _at_theageof 2-4mo(n = 3each)

aswellasina3-mo-old IFN-y single transgenicmousebutnot

on pancreatic parenchymal cells (Fig. 5, A-C). Additionally,

doubleimmunofluorescence studies disclosed thatB7-1 mole-cules were expressed on macrophages and B lymphocytes in the pancreata of NP+/IFN-y+ and

GP+/IFN-y'

double transgenic mice(Fig. 6, A-D).

Discussion

We observedaveryhighincidence of diabetes in

NPJ/IFN-y'

mice with accelerated onset, while that in GP/IrFN-y+ mice

was not different from

IFN-y'

single transgenic littermates. The incidence ofearly onsetdiabetes was high in the initial generations ofIns-EFN-y mice (4)but gradually decreased to 30-40% in males and 0-10% in females after repeatedcrossing

withBALB/c(Lee, M.-S.,etal., unpublished data).Thischange appearedtobe causedbyashift in the balance between

destruc-tion and regenerationof islet cells (11, 12). The incidence of diabetes in

NP-/IFN-y',

GP-/IFN-y',

andGP+/IFN-y+ mice

is consistent with those recent observations. However, that

100% of

NP'/IFN-y'

mice spontaneouslydevelopeddiabetes

A # Reiea

75

50

25

>5:1 2.5:1

C % Release

75

50

25

---- - = =

Iwow*to fuax

%PI III I

Aftftftm.._

(6)

Figure 5. ExpressionofB7-1 mol-ecules. Immunohistochemical

staining disclosed that B7-1 mole-cules wereexpressedon infiltrat-ingcellsbutnot on islets or other

parenchymalcells in the pancreata

ofa3-mo-oldNP'/IFN-y'mouse (A)(x400)and a 3-mo-oldGP'/ IFN-y+ mouse(B)(X 1,000). In-filtrating cells in the pancreata of a3-mo-oldIFN-ysingle transgenic mousealsoexpressed

B7-1 molecules (C)(X 400).

represents a significantly increased incidence compared with

other mice.

We demonstrated the spontaneous generation of CTL against self antigens (LCMV) in our LCMV/IFN-y double

transgenicmicebutnotinLCMVsingle transgeniclittermates. TheseCTL wereobserved both inNP+/IFN-y' mice andto a

lesser degree in GP+f1FN-y' mice. The less robust CTL re-sponseof

GP'/IFN-y'

micetoLCMV-infectedtargetsand their lower incidence of diabetes compared withNP'/IFN--y' mice

arealmostcertainly explainedby MHC restriction of

lympho-cyte-mediated cytotoxicityandgene dosageeffects.

GP+/IFN-y' mice have a b x d MHC haplotype. The GP antigen of LCMVcanbepresentedtoCTLonH-2b butnotonH-2d

mole-cules, whereas the NP antigenis recognized by such cells on

both H-2b and H-2d molecules (13).Theincidence of diabetes

evenafterLCMV infection of GP single transgenicmiceof b

xd MHC haplotypewasonly 33% (3/9)overthe4-mo

observa-tion period, althoughall mice developed positive antibody

re-sponsestotheLCMVantigen indicatingsuccessful LCMV in-fection(14).Thedifference in the incidence of diabetes between

NP/IFN-y' and

GP+/IFN-y'

mice underscores the differing

abilities of individualantigenstoelicit immuneresponses,and theveryhighincidence of diabetes in

NPf/IFN-y'

micecould be attributed to the presence of a single antigen that can be

effectively recognized by CTL.Besides thegenedosage effect,

differentexpressionlevels between NP and GPtargetantigens

Figure6.Double

immunofluores-cencestudiesshowed that

macro-phages stainedwithF4/80 antibody (A) expressed B7-1 molecules (B)in

thepancreasofa3-mo-old

NP'/IFN-Y+mouse.SomeBlymphocytes

stainedwithanti-B220antibody (C) alsoexpressed B7-1 molecules(D) (x 1,000). In additiontoscattered double-positive cells,many single-positivecellswerepresent.

Double-positivecellswereobserved inthe

pancreasofGP+/IFN-y+miceas well.

P...: w.,

.1 ::

..

44

...d,

A

'1; :' ..A.Ibe

*.

P.

do4. I

f'p

9..:

Ve. An

4;:!

C

.il

ace:.

(7)

may play a role in thephenotypic differences between NP+/ IFN-y+ and

GP+/IFN-y'

mice. However,possible thymic

ex-pression of GP antigen wouldnotcontributetothe lower level of CTLactivity and low incidence of diabetes in GP+/IFN-y+

micebecause GP was not expressed in the thymus of the

RIP-LCMV-GP line 34-20 (15).

The development of CTL and spontaneous diabetes in LCMVJIFN-y double transgenic micecontrastswith results of

TNF-a transgenic mice, which did not develop spontaneous diabetes themselves or when crossed with mice expressing

LCMVantigens(16). The induction of CTL responsesto viral antigens noted in LCMV/IEFN-y mice also contrasts with the observation that IL-2orIL-6transgenicmicecrossed with RIP-LCMV mice did notdevelop a spontaneous CTL response to

viral antigens (17). Similarly, no spontaneous autoimmunity arose in RIP-IL-2 transgenicmice orRIP-IL-2/RIP-Kb double transgenic mice (18). Sensitization to islet antigens (Kb) was notdirectly demonstratedintransgenic mice thatexpressed

H-2Kb and IL-2 in pancreatic 6 cells and also had the T cell receptor specific forKb, although the development of diabetes

was accelerated (19). These discrepancies indicate that cyto-kines differ in their abilityto causespecific immune sensitiza-tion andemphasizetheunique capacity of IFN-ytoelicitTcell responses.

Theexpression of B7-1 ontheAPC butnot on pancreatic islet cellsis consistent with the notion that the killing of targets by effector CTL does not require costimulatory signals (20). Transgenic expression ofB7-1 moleculesonpancreatic

63

islet cells has been reported to conferantigenpresenting ability to

islet cells (21, 22); however, it isnotclear ifparenchymal cells

canexpress costimulatory moleculesphysiologicallyinvivoor

if such cells expressing costimulatory molecules are able to presentantigen in natural autoimmune disorders or models. In contrastto thosemodels, specific sensitization to selfantigens byIEN-y inoursystem mightbe related to theexpression of costimulatory molecules suchasB7-1 on, and theirpresentation by, APC infiltrating the pancreas of transgenic mice. In this regard, the pathway for the activation of ignorant lymphocytes

in this model is different from that postulated by Bottazzo et al. (23) in which epithelial cells express aberrant class II MHC molecules and present self antigens without the help of APC leadingto autoimmunity.

The pathway to immunological sensitization to pancreatic LCMVantigens by local production of IFN-y is clearly distinct from thesystemicsensitizationtoLCMVantigensafter LCMV infection with diabetes resulting from crossreactivity to the

sameviral (islet) antigen (6, 7). Our system probably involves sensitization of lymphocytes to islet LCMV antigens within nonlymphoid peripheral organs where IFN-y is expressed. However, in related experiments, systemic administration of IFN-y touninfected RIP-LCMV mice failed to elicit diabetes (7). Thediscrepancy betweenthese results and our data could reflect adifference in the concentration or duration of IFN-y experienced locallyor ahierarchy effect mediated by a general suppression of lymphoid tissues, rendering peripheral inflam-matory responses unlikely. Recent studies using double transgenic mice expressing both IFN-y and human CD4 in the pancreatic /3 cells failedtodemonstrate humoralautoimmunity directedtotheCD4 molecule(24). As a Thl-promoting, Th2-inhibiting cytokine, the effect of production of IFN-y would favoracellularimmune response rather thanahumoralresponse

(25), which could

explain

the absence of humoral responseto

CD4. This would also

explain

theabsence ofanti-LCMV

anti-bodyin

NPJ/IFN-y'

and

GP+/IFN-y'

mice inour_system.

Our

results,

the

striking

demonstrationofspontaneous

CTL

generationin mice witha

normal,

unmanipulated

Tcell

reper-toire, suggest thatonemechanism of

autoimmunity

may result from local activation of

lymphocytes

specific

for self

antigens

but

ignorant

of those

antigens

because of their locationon

non-APC. These

lymphocytes

becomeactivated

by

alteration of the

pancreatic

environment suchaselicitation of

cytokines.

Those

lymphocytes

corresponding

toself constituents would then be ableto

respond

totheir

respective antigens

andinitiatean

auto-immune response which could be

perpetuated by

their

prolifera-tion withinthe

tissue, subsequent

cellular

damage,

and libera-tion of other

sequestered antigens allowing

the

amplification

of the anti-self response. The acceleration of diabetes observed in

ourdouble

transgenic

mice could reflect the

ability

ofa

_single,

well-represented antigen

to

intensify

this process.

Acknowledgments

Wearegratefulto S. Zhu and J. Dockterfortheir excellenttechnical

assistance. We thank H. Fox for criticalreadingand P.Minikfor edito-rialassistance.

M. Leewassupported byafellowshipfrom theMultipleSclerosis

Society (FA 1007-A-1),and M.vonHerrathwassupportedbya

fellow-shipfrom the Juvenile DiabetesFoundation(393357).N.Sarvetnickwas

supported by acareerdevelopmentaward from the Juvenile Diabetes

Foundation.This workwassupported bygrants from the National

Insti-tutes of Health (HD-29764 and HL-48728 to N. Sarvetnick and

AGO4342toM. B. A.Oldstone).Thisismanuscriptno.8590-NP from TheScrippsResearch Institute.

References

1.Kaufman,D.L.,M.Clare-Salzler,J.Tian,T.Forsthuber,G. S. P.Ting,P.

Robinson,M. A.Atkinson,E. E.Sercarz,A.J.Tobin,and P. V. Lehmann. 1993.

Spontaneous lossof T-cell toleranceto_glutamicaciddecarboxylasein murine insulin-dependentdiabetes. Nature(Lond.). 366:69-75.

2.Tisch, R.,X.-D. Yang,S. M.Singer, R.S.Libiau,L.Fugger,and H.0.

McDevitt.1993. Immune responsetoglutamicaciddecarboxylasecorrelates with insulitis in non-obese diabetic mice. Nature(Lond.). 366:72-75.

3.Harrison, L.C.,M.Honeyman,H. J. Deaizpurua, R.S.Schmidli, P. G.

Colman,B. D.Tait,and D. S. Cram. 1993. Inverse relation between humoral and cellularimmunitytoglutamicaciddecarboxylaseinsubjectsatrisk of

insulin-dependentdiabetes. Lancet. 341:1365-1369.

4.Sarvetnick, N.,D.Liggitt, S. L.Pitts, S. E.Hansen, and T. A.Stewart. 1988.Insulin-dependentdiabetes mellitus induced intransgenicmicebyectopic expressionof class H MHC andinterferon-gamma.Cell. 52:773-782.

5.Sarvetnick, N.,J.Shizuru,D.Liggitt,L.Martin,B.McIntyre,A.Gregory,

T.Parslow,and T.Stewart. 1990. Loss ofpancreaticislettolerance inducedby

/3-cellexpressionofinterferon-y.Nature(Lond.).346:844-847.

6.Oldstone,M. B.A.,M.Nerenberg,P.Southern,J.Price,and H.Lewicki. 1991. Virus infectiontriggers insulin-dependentdiabetes mellitus inatransgenic

model: role of anti-self(virus)immune response. Cell.65:319-331.

7.Ohashi,P.S.,S.Oehen,K.Buerki,H.Pircher,C. T.Ohashi,B.Odermatt,

B. Malissen,R. M. Zinkernagel,and H.Hengartner. 1991. Ablation of

"toler-ance" and induction of diabetesby virus infection in viral antigen transgenic

mice.Cell. 65:305-317.

8. Razi-Wolf, Z.,G. J.Freeman, F. Galvin,B. Benacerraf, L.Nadler, and H.Reiser. 1992.Expressionand function of the murine B7antigen, themajor costimulatorymoleculeexpressed by peritoneal exudatecells. Proc.Natl.Acad. Sci. USA. 89:4210-4214.

9. Klavinskis,L. A., L. Whitton,and M. B. A. Oldstone. 1989. Molecular

engineeredvaccine which expressesanimmunodominant T-cellepitopeinduces

cytotoxic Tlymphocytes that conferprotection from lethal virus infection. J. Virol. 63:4311-4316.

(8)

foractivation of murine T lymphocytes via the T-cell receptor/CD3 complex. Proc.Natl.Acad. Sci. USA. 89:271-275.

11.Gu, D., and N. Sarvetnick. 1993. Epithelial cell proliferation and islet neogenesis in IFN-y transgenic mice. Development (Camb.). 11:33-46.

12.Gu, D., M.-S. Lee, T. Krahl, and N. Sarvetnick. 1994. Transitional cells in the regenerating pancreas. Development (Camb.). 120:1873-1881.

13.Whitton,J. L.,P. J. Southern, and M. B. A. Oldstone. 1988. Analysis of thecytotoxic T lymphocyte responses to glycoprotein and nucleoprotein compo-nentsoflymphocyticchoriomeningitisvirus. Virology. 162:321-327.

14.Lee, M.-S., L. Wogensen, J.Shizuru,M. B.A.Oldstone,and N. Sarvetnick. 1994. Pancreatic islet production of murineinterleukin-10doesnot inhibit im-mune-mediated tissue destruction. J. Clin.Invest. 93:1332-1338.

15.von Herrath, M., J. Dockter, and M. B. A.Oldstone. 1994. Role of the

thymusin virus-inducedinsulin-dependentdiabetes mellitus in a transgenic model. Immunity. 1:231-242.

16.Higuchi,Y., P. Herrera, P.Muniesa, P. Huarte, D. Belin, P. Ohashi, P.

Aichele,L.Orci, J.-D. Vassalli, and P. Vassalli. 1992. Expression of a tumor necrosis factor a transgene in murine pancreatic (3 cells results in severe and permanentinsulitis without evolution towards diabetes. J. Exp. Med.

176:1719-1731.

17. vonHerrath, M.G.,J.Allison, J. F. A. P. Miller, and M. B. A. Oldstone. 1995. Focal expression ofinterleukin-2does not break unresponsiveness to "self" (viral) antigen expressed in /3 cells but enhances development ofautoimmune

disease (diabetes) after initiation of anti-self immune response. J. Clin. Invest. 95:477-485.

18.Allison,J., L.Malcolm, N.Chosich,and J. F. A.P. Miller. 1992.

Inflam-mationbutnotautoimmunityoccursintransgenicmiceexpressing constitutive levels ofinterleukin-2 in islet/scells. Eur. J. Immunol. 22:1115-1121.

19.Heath, W. R., J. Allison, M. W. Hoffmann, G. Schonrich, G.Hammerling,

B.Arnold,and J. F. A. P. Miller.1992. Autoimmune diabetes as a consequence oflocally producedinterleukin-2. Nature(Lond.). 359:547-549.

20. Harding, F. A., and J. P. Allison. 1993. CD-28-B7 interactions allow the induction of CD8+ cytotoxic lymphocytes in the absence of exogenous help. J. Exp. Med. 177:1791-1796.

21. Harlan, D. M., H. Hengartner, M. L. Huang, Y.-H. Kang, R. Abe, R. W.

Moreadith,H.Pircher,G. S.Gray,P.S.Ohashi,G.J.Freeman,etal. 1994. Mice expressing both B7-1 and viral glycoprotein on pancreatic beta cells along with glycoprotein-specific transgenic T cells develop diabetes due to a breakdown of

T-lymphocyte unresponsiveness.Proc. Natl. Acad.Sci. USA. 91:3137-3141. 22. Guerder, S., D. E. Picarella, P. S. Linsley, and R. A. Flavell. 1994. Costimulator B7-1 confersantigen-presenting-cellfunction to parenchymal tissue and in conjunction with tumor necrosis factor-a leads to autoimmunity in transgenic mice. Proc. Natl. Acad. Sci. USA. 91:5138-5142.

23. Bottazzo, G. F., R.Pujoll-Borrell,and T. Hanafusa. 1983. HLA-DR expres-sion and antigen presentation in induction of endocrine autoimmunity. Lancet. ii:ll 5-1118.

24.Martin, L., M. Mora-Worms, C. Lucas, C. Reynolds, and T. A. Stewart. 1993. An evaluation of mechanismsbywhichtolerance toorgan-specificantigens islostusingatransgenicmousemodel. J. Immunol. 150:1234-1243.