Pancreatic gene expression in rare cells of thymic medulla: evidence for functional contribution to T cell tolerance

International Immunology, Vol. 9, No. 9, pp. 1355–1365 © 1997 Oxford University Press

Pancreatic gene expression in rare cells ofthymic medulla: evidence for functionalcontribution to T cell tolerance

Kathleen M. Smith 1, David C. Olson, Ryutaro Hirose and Douglas Hanahan

Department of Biochemistry and Biophysics, and Hormone Research Institute, University of California,San Francisco, CA 94143-0534, USA1Present address: DNAX Research Institute, 901 California Avenue, Palo Alto, CA 94304, USA

Keywords: pancreatic gene, T cell tolerance, thymic medulla

Abstract

We report the initial characterization of rare cells within the thymus that express ‘peripheral’ self-antigens and are capable of inducing partial tolerance to a model protein. Mice from two transgenicfamilies that express SV40 T antigen (Tag) in pancreatic islet β cells under control of a rat insulinpromoter (RIP) develop T cell tolerance toward this neo-self antigen. These mice express low levelsof Tag mRNA in the thymus. Transplantation of thymus from tolerant RIP-Tag mice into athymichosts is sufficient to confer tolerance by CD4 F Th cells and elicits variable tolerance by CD8 F

cytotoxic T cells. Thymic medulla is shown to contain rare cells that express the endogenousinsulin and somatostatin genes, and in the transgenic animals, Tag. These cells are referred to as‘peripheral antigen-expressing’ (PAE) cells. Thymic cell fractionation reveals the PAE cellsexpressing insulin and Tag to be present in a fraction enriched for non-lymphoid, MHC class II F

cells. Notably, absence of thymic expression of the RIP-Tag gene in another transgenic familycorrelates with failure to establish self-tolerance and susceptibility to autoimmunity. Thus,expression of tissue-restricted genes such as insulin in PAE cells of thymic medulla may serve tolimit development of potentially autoimmune T cells.

Introduction

There is now a compelling body of experimental evidence and Oldstone (5) first suggested that the unexpected resultof thymic expression of insulin promoter transgenes wasthat self-tolerance toward widely abundant self-antigens is

established at the T cell level during development in the functionally significant; the data indicated the thymic expres-sion could elicit some degree of non-responsiveness towardthymus, whereby T cells are first positively selected in the

cortex for functional TCR that bind self-MHC and then nega- the transgene product. Initially these results, while intriguing,were interpreted to be a consequence of promiscuous expres-tively selected in the medulla for death or paralysis when their

TCR recognizes self-peptides presented by MHC molecules sion of the hybrid insulin–promoter transgenes. Now, however,the evidence points to an alternative explanation, i.e. thaton epithelial and dendritic cells (1,2). In contradistinction,

conventional wisdom has had it that tolerance of rare cells thymic expression of pancreatic genes is a normal propertyof these ‘tissue-restricted genes’. Thus, this laboratory hasand non-abundant antigens, as exemplified by the insulin-

producing pancreatic β cell, involves non-thymic (or peri- shown that a number of endogenous pancreas-specificgenes, including the insulin genes, as well as a hybrid insulinpheral) mechanisms which act on T cells following their

emigration from the thymus (3). However, the view that transgene, can be detected by RT-PCR at very low levels inthe thymus (6). In turn, Flavell and co-workers reported thatperipheral mechanisms are the key to β cell tolerance has

begun to change, in part because of reports demonstrating the exocrine pancreatic gene elastase and a hybrid elastasetransgene were both expressed in thymus, and demonstratedthat transgenic mice utilizing insulin gene regulatory regions

can express the transgenes both in islet β cells and in that transplantation of transgenic thymus could convey toler-ance of the transgene product, SV40 T antigen (Tag) (7).thymus. Notably, studies from the laboratories of Miller (4)

Correspondence to: D. Hanahan

Transmitting editor: S. Korsmeyer Received 26 March 1997, accepted 2 June 1997

1356 Rare thymic cells express tissue-restricted genes

We have used a transgenic mouse model of alternative self observed in mice receiving fresh or cultured thymi. Kidneyswere removed from transplanted nu/nu mice at the time oftolerance or autoimmunity conferred by hybrid insulin genes

to investigate the thymic cells expressing insulin, the hybrid sacrifice and the transplanted thymi inspected to confirmthat they had become populated with T cell progenitors.insulin transgene and other endogenous pancreatic genes,

and have further assessed the functional contributions of Additionally, several transplanted thymi were analyzed byhistology and found to be normal.thymic expression in regard to these alternative immunological

phenotypes. In this model, the rat insulin promoter (RIP) hasRNA-PCR analysisbeen used to direct expression of the SV40 Tag to the insulin-

producing β cell, a rare cell type located in the pancreatic RNA was prepared from tissues using Trizol (Gibco/BRL,Gaithersburg, MD) according to the manufacturer’s instruc-islets which is the target of autoimmune destruction in type I

diabetes. Families of RIP-Tag mice develop one of two tions. Aliquots of RNA, 3.0 or 15 µg, were DNase treated andsplit in half. cDNA was synthesized under standard conditionsalternative immunological phenotypes (8). RIP1-Tag2 and

RIP3-Tag2 mice develop systemic T cell tolerance toward this using random primers. Half the sample received MMLVreverse transcriptase (Gibco/BRL) while the other half did notnovel self antigen, manifested as severely impaired humoral

and undetectable cytotoxic T cell responses toward the (–RT). Use of more RNA in the reverse transcription reactionincreases the sensitivity of detection of specific messages.Tag antigen, and an absence of inflammation of the Tag-

expressing islets (6). In contrast, RIP1-Tag5 mice develop an The PCR analysis has been described (6).autoimmune response directed at their Tag-expressing islets,

Proliferation assayas evidenced by circulating autoantibodies (8), leukocyteinfiltration of the islets (9) and the development of Tag-specific This assay has been described (6). Briefly, mice were immun-

ized s.c. at the base of the tail with 50 µg Tag emulsified inT cells (10).In this report we demonstrate that thymus from tolerant RIP- complete Freund’s adjuvant (CFA). Ten days later, the dis-

persed draining lymph node cells were cultured at 4–63105Tag transgenic mice is sufficient to induce tolerance of Thcells in recipient non-transgenic mice. In addition, we provide cells/well in HL-1 medium (Hycor, Irvine, CA), supplemented

with L-glutamine, with either dilutions of Tag protein or heatevidence that insulin and Tag, as well as several otherendogenous pancreatic genes, are expressed by rare cells killed Mycobacterium tuberculosis (Difco, Detroit, MI) in triplic-

ate for 3 days and pulsed with 1 µCi [3H]thymidine. Plates werelocalized in the thymic medulla; we refer to these cells as‘peripheral antigen-expressing’ (PAE) cells. A family of RIP- harvested and incorporation of [3H]thymidine determined.Tag mice prone to autoimmune disease does not express Tag

Antibody assayin thymus, implicating failure to express as a predisposition toautoimmunity. The results suggest a role for the thymic PAE This assay has been described (6). Briefly, mice were immun-

ized i.p. with a mixture of 10 µg Tag protein and 5 µg β-cells in inducing tolerance to proteins which are normallyconsidered tissue restricted and imply a new form of cell type galactosidase (Sigma, St Louis, MO) emulsified in CFA on

day 0, and primed with the same dose of antigen in incompletespecific regulation for their genes.Freund’s adjuvant on days 14 and 27. Mice were bled on day32 and the sera tested in an ELISA for the presence of IgG

Methods recognizing either Tag or β-galactosidase.

Mice Cytotoxic assayRIP1-Tag2 (11) and RIP3-Tag2 (6) have been described. The This assay has been described (6). Briefly, mice were immun-RIP1-Tag2 and RIP3-Tag2 mice used in these studies were ized i.p. with 23107 SV40 transformed fibroblasts, C57SVbackcrossed to C57BL/6 mice for at least 15 generations. (12). Ten days later, 23107 splenocytes were cultured withC57BL/6, C57BL/6 nu/nu, C57BL/6 nu/1 and BALB/c mice 2.53105 irradiated C57SV (3500 rad) or 107 BALB/c spleenwere purchased from Jackson Laboratories (Bar Harbor, cells (2500 rad) in six-well plates in DMEM supplementedME) or bred in the UCSF transgenic facility under specific with 10% FCS and 50 µm 2-mercaptoethanol. After 5 days,pathogen-free conditions. the re-stimulated spleen cells were collected and tested for

their ability to lyse Tag-expressing fibroblasts (C57SV), aThymic transplantation syngeneic fibroblastic cell line kindly provided by Barbara

Knowles [MC57G (13)], or the allogeneic mastocytoma P815Male and female C57BL/6 nu/nu mice, between 6 and 26weeks of age, were transplanted under the kidney capsule (ATCC, Rockville, MD) in a standard 4h 51Cr-release assay.

To test for helper-independent cytotoxic T lymphocytes (CTL)with four intact newborn thymic lobes from either RIP1-Tag2,RIP3-Tag2 or non-transgenic littermates. Thymi were removed (14), splenocytes were cultured with irradiated C57SV either

with or without added murine IL-2 at 10 U/ml (Boehringeron p0, the day of birth, or p1 and either transplanted thesame day or after overnight culture. The genotype of the Mannheim, Indianapolis, IN) and assayed as described

above.newborn mice was determined by PCR of a tail biopsyprior to transplantation. The majority of nu/nu mice received

Statistical analysistransplants before 12 weeks of age and donor thymi wereremoved on p1 and transplanted the same day. These nu/nu A Student’s unpaired t-test was performed at the 1:100 effector

to target ratio to evaluate the differences between the CTLmice were allowed to reconstitute their T cells for 10–21weeks before analysis. No difference in phenotype was response of non-transgenic and RIP1-Tag2 mice and of nu/

Rare thymic cells express tissue-restricted genes 1357

nu mice transplanted with non-transgenic thymus and nu/nu antibodies both to Tag and to a control antigen, β-galactosid-ase (Fig. 1A). For comparison, transgenic, normal non-trans-mice transplanted with transgenic thymus.genic and untransplanted nu/nu mice were immunized. As

Immunohistochemistry expected, nu/nu mice transplanted with non-transgenicthymus, much like normal mice, made strong antibodyThymi were harvested and immersion fixed in 3.5% paraform-responses to both antigens. Remarkably, nu/nu mice trans-aldehyde at 4°C overnight. Fixed tissues were imbeddedplanted with RIP1-Tag2 or RIP3-Tag2 thymus responded inin paraffin and 5 or 10 µm sections were processed forstrikingly similar fashion to intact transgenic mice, generatingimmunohistochemistry using standard techniques. Endogen-strong antibody responses to β-galactosidase, but little anti-ous peroxidase was quenched in 0.7% H2O2/MeOH. Antigenbody to Tag. Untransplanted nu/nu mice failed to makeretrieval was performed using Citra solution (Biogenex, Sanantibody to either antigen. Thus, Th cells developing in theRamon, CA) according to the manufacturers’ instructions.context of thymic Tag expression are tolerant to Tag, as shownSections were incubated overnight at 4°C with primary antiseraby their inability to provide help for antibody production. This[Tag S, insulin and somatostatin (Dako, Carpenteria, CA)]result has been confirmed in an additional cohort of nu/nuin 5% goat serum/PBS, 30 min at room temperature withmice transplanted with RIP1-Tag2 thymus.biotinylated goat anti-rabbit antiserum (5 µg/ml; Vector

To assess Th cell tolerance more directly, lymph node cellsLaboratories, Burlingame, CA) and then with streptavidin–from primed reconstituted nu/nu mice were tested for theiraminohexanol–biotin–horseradish peroxidase complex (Vect-capacity to proliferate to Tag in vitro. Once again, the nu/nuastain-Elite; Vector). Protein was visualized with the peroxidasemice transplanted with non-transgenic thymus respondedsubstrate 3,39-diamino-benzidine tetrahydrochloride (Sigma).similarly to normal mice, while the nu/nu mice transplanted withFor the double immunohistochemical analysis, staining withtransgenic thymus demonstrated impaired responsiveness toprimary rabbit polyclonal antisera for rel B (C-19; Santa CruzTag, similar to unmanipulated transgenic mice (Fig. 1B andBiotechnology, Santa Cruz, CA) or the lectin ulex agglutinin ID). Lymph node cells from all animals proliferated well to the(UEA I; Vector) was performed immediately after the firstcontrol antigen, M. tuberculosis (Fig. 1C and E). This resultreaction by reblocking peroxidase and proceeding throughhas been confirmed in three additional sets of nu/nu micethe horseradish peroxidase reaction, using the SG chromogentransplanted with either RIP1-Tag2 or RIP3-Tag2 thymus.(Vector). Single antigen assays were followed by brief counterThus, using two assays which measure the tolerance inductionstain in either Gills #1 hematoxylin or eosin (Sigma) prior toof CD41 Th cells, the low level of Tag expressed in transgenicmounting. Negative controls included staining of sectionsthymus is sufficient to confer tolerance, similar to that observedfrom non-transgenic mice and omission of primary antiserum.in intact transgenic mice.

Cell fractionation studiesTransgenic thymus is not sufficient to ensure development ofCells were prepared according to the protocol of Steinmancomplete cytotoxic T cell toleranceand colleagues (15,16). Briefly, individual thymic lobes wereUsing a third lymphocyte function assay, transplanted nu/nucollected from 3.5- to 6-week-old mice and perfused withmice were tested for the ability to generate CTL recognizingcollagenase, 100 U/ml (Type III; Worthington Biochemicals,Tag. As expected, lymphocytes from non-transgenic miceFreehold, NJ) in HBSS. The perfusate, fraction 1, was collectedand nu/nu mice transplanted with non-transgenic thymusseparately. Thymic lobes were torn into small pieces withlysed Tag-expressing targets efficiently, while splenocytesforceps. The released cells were collected, while the largerfrom RIP1-Tag2 (Fig. 2A) and RIP3-Tag2 (Fig. 2E) micepieces of tissue were further digested in 400 U/ml collagenasefailed to recognize them. Remarkably, individual nu/nu miceat 37°C for 15 min. The cells released by mechanical disrup-transplanted with either RIP1-Tag2 (Fig. 2C) or RIP3-Tag2tion and those released by collagenase digestion were pooled(Fig. 2G) thymus gave variable killing of the Tag-expressingand passed through a wire mesh, fraction 2. The cells werecells. This variability is reflected by the large error bars shownpelleted at 1000 r.p.m. in a Sorvall RT6000B and resuspendedfor the mean response of nu/nu mice transplanted within dense BSA (Cohn fraction V, 3220-75; Intergen, Newtransgenic thymus. This variable CTL response is unlikely toYork, NY), refractive index 1.384–1.385 at 25°C. For detailedresult from unusual NK cell activity in the nu/nu geneticpreparation of dense BSA see (16). Aliquots of resuspendedbackground, since nu/nu mice which had not received acells were overlaid with RPMI and spun for 15 min at 9000 gtransplant did not lyse either target (Fig. 2C and D). Addition-(7000 r.p.m.) in a Sorvall HS-4 rotor, without a brake. Theally, every nu/nu mouse which received a thymus generatedpellet, fraction 3, and the cells at the interface, fraction 4,an allogeneic response (Fig. 2D and H), indicating that thiswere collected separately.variability in a specific CTL response is not a reflection ofpoor reconstitution of the immune system. Finally, variable

Results CTL activity was not seen in nu/nu mice transplanted withnormal thymus nor in mice transplanted with transgenic

Transgenic thymus confers tolerance of Th cells thymus but analyzed in one of the other two assays. The CTLassays have been repeated several times with additional nu/To investigate the functional significance of thymic Tag expres-

sion in tolerant RIP-Tag mice, we transplanted thymi from nu mice transplanted with either RIP1-Tag2 or RIP3-Tag2thymus with similar results (Fig. 2I). In summary, thymusnewborn RIP1-Tag2, RIP3-Tag2 or non-transgenic mice under

the kidney capsule of syngeneic nu/nu mice. First, reconstit- from a tolerant RIP-Tag mouse can by itself induce non-responsiveness of CD41 Th lymphocytes, while in most casesuted nu/nu mice were tested for their ability to produce

1358 Rare thymic cells express tissue-restricted genes

Fig. 1. Thymus from a tolerant RIP-Tag mouse is sufficient to induce tolerance of Th cells. Antibody response of nu/nu mice reconstituted withRIP1-Tag2 and RIP3-Tag2 thymus (A). Non-transgenic, RIP1-Tag2, RIP3-Tag2, nu/nu mice transplanted with non-transgenic thymus, nu/numice transplanted with RIP1-Tag2 thymus, nu/nu mice transplanted with RIP3-Tag2 thymus, and unmanipulated nu/nu mice were immunizedand boosted with Tag and β-galactosidase in CFA. Sera were tested for the presence of IgG recognizing Tag (d) and β-galactosidase (s).The mean response of all animals in each group is shown by a bar. The tertiary antibody response, as well as the level of antibody detectedin the preimmune serum from non-transgenic mice, is shown. In vitro proliferation of lymphocytes from nu/nu mice transplanted with RIP1-Tag2 and RIP3-Tag2 thymus (B–E). Lymph node cells from immunized non-transgenic mice (d, solid line, n 5 4), RIP1-Tag2 mice (s, solidline, n 5 3), RIP3-Tag2 mice (n, solid line, n 5 3) (B and C), nu/nu mice transplanted with non-transgenic thymus (d, dashed line, n 5 4),nu/nu mice transplanted with RIP1-Tag2 thymus (s, dashed line, n 5 3) and nu/nu mice transplanted with RIP3-Tag2 thymus (n, dashed line,n 5 3) (D and E) were cultured with dilutions of Tag (B and D) and the control antigen M. tuberculosis (C and E). The antigen-dependentproliferation was measured after 72 h by adding [3H]thymidine for the last 6 h. The mean 6 SEM of all animals in each group is shown.

it is insufficient to induce complete tolerance of CD81 cytotoxic added during the in vitro boost to simulate Th cell function(Fig. 3). No Tag-specific cytotoxic activity was induced in theT cells.

Only minimal Tag-specific T cell help is presumably avail- presence of IL-2, indicating that the CTL in intact RIP1-Tag2mice have been directly tolerized, and that tolerance thereforeable to the developing CTL both in intact transgenic and in

nu/nu mice reconstituted with transgenic thymus, since the results from lack of both Th and CTL specific for the neo-self antigen.helper cells in each case are similarly tolerant (at least at

the humoral level). Assuming no differential Th1/Th2 biasesImmunohistochemical analysis of Tag-expressing cells in thebetween intact and transplant mice, the data suggests thatthymusthe CTL response observed in the reconstituted nu/nu mice

must be largely Th independent. In contrast, therefore, one Next, we sought to begin characterization of the Tag-expressing cells in the thymus. Thymi were collected from micewould predict that CTL as well as Th cells in intact RIP-Tag

mice are tolerized. To address that possibility, a CTL assay between the day of birth (newborn) and 1 week of age, the timeframe in which Tag was initially detected by RNA-PCR in RIP1-was performed using RIP1-Tag2 mice in which IL-2 was

Rare thymic cells express tissue-restricted genes 1359

Fig. 2. Transgenic thymus elicits variable tolerance by cytotoxic T cells. CTL response of nu/nu mice transplanted with thymus from newbornRIP1-Tag2 mice (A–D). Splenocytes from immunized non-transgenic mice (j, dashed line, n 5 5), RIP1-Tag2 mice (s, dashed line, n 5 8)(A and B), nu/nu mice transplanted with non-transgenic thymus (m, solid line, n 5 8), nu/nu mice transplanted with RIP1-Tag2 thymus (.,solid line, n 5 8) and unmanipulated nu/nu mice (,, dotted line, n 5 2) (C and D) were tested for their ability to kill Tag-expressing fibroblasts,C57SV (A and C), syngeneic Tag negative fibroblasts, MC57G (data not shown) or P815 (B and D), allogeneic. CTL response of nu/nu micetransplanted with thymus from the second transgenic line, RIP3-Tag2 (E–H). Splenocytes from immunized non-transgenic mice (j, dashedline, n 5 4), RIP3-Tag2 mice (s, dashed line, n 5 4), (E and F) nu/nu mice transplanted with non-transgenic thymus (m, solid line, n 5 3) andnu/nu mice transplanted with RIP3-Tag2 thymus (., solid line, n 5 5) (G and H) were cultured and tested for their ability to kill C57SV (E andG), MC57G (data not shown) or P815 (F and H). The mean 6 SEM of all animals in each group is shown. Pooled data at a 1:100 effector totarget ratio from five separate experiments using thymus chimeric mice is shown in panel (I) ( C57SV targets, d; P815 targets, s).

Tag2 and RIP3-Tag2 mice (6). The tissues were processed for nu mice, indicating that Tag expression persists well into adult-hood (Fig. 4I). In thymus from 1-week-old RIP3-Tag2 mice weimmunohistochemistry and immunostained with Tag-specific

antiserum (9). Tag-staining cells were observed in RIP1-Tag2 detected ~100 Tag-expressing cells (Table 1).The anatomical location of the Tag-staining cells was alsoand RIP3-Tag2 thymic sections (Fig. 4 and Table 1), but not in

non-transgenic thymus (data not shown). In RIP1-Tag2 mice, investigated. Thymic sections were double stained with anti-serum recognizing Tag, and either antibody to rel B, which isroughly 5 times more Tag-expressing cells were identified in

newborn thymic sections than in 3-day-old thymi (Table 1), expressed by thymic dendritic and medullary epithelial cells, orUEA I, which only binds medullary epithelium (17). The doublewhich correlates well with RNA-PCR analysis of thymic Tag

expression. Tag-staining cells were also detected in RIP1-Tag2 staining revealed that all Tag1 cells (n 5 96) were present inregions populated by rel B1 (Fig. 4A and B) and UEA I1 (Fig.thymi between 15 and 52 weeks after transplantation into nu/

1360 Rare thymic cells express tissue-restricted genes

thymic cell population, a hypothesis could be made for theexistence of a specialized thymic function capable of inducingtolerance to ‘peripheral proteins’. To investigate this possibility,we fractionated thymic cells (15) and performed RNA-PCRanalyses for a set of pancreatic transcripts. Briefly, thymiwere perfused with collagenase, producing fraction 1, theperfusate. The perfused thymi were then mechanically dis-rupted and further digested with collagenase, to generatefraction 2, the total released cells. In turn, fraction 2 wascentrifuged through a BSA step gradient, resulting in a cellpellet, fraction 3, and a low-density fraction of cells, fraction4, which remained at the interface. The fractions were charac-terized by FACS as well as by light microscopy (data notshown). The predominant cell population, comprising 98–99% of all cells contained in fractions 1–3, consisted of small,round cells which stained with Thy-1, identifying them asthymocytes. The low-density fraction, 4, which contained only0.9–1.5% of the total cells, consisted of a mixed population,including most of the large granular cells released by thecollagenase treatment. The cells in this fraction were identifiedby FACS analysis as dendritic cells, macrophages, thymo-cytes and an unidentified population of small class II MHC1

cells, which are most likely monocytes, B cells or restingmacrophages. In one typical experiment these cell typescomprised 26.4, 14.5, 49.2 and 7.9% of the total fraction 4Fig. 3. RIP1-Tag2 mice fail to exhibit Th independent CTL activity.

Splenocytes from RIP1-Tag2 (s, dashed line, n 5 3: ,, solid line, population respectively.n 5 3) and non-transgenic (j, dashed line, n 5 3; m, solid line, n 5 RNA-PCR analysis of the four thymic cell fractions revealed3) mice immunized with C57SV were cultured in the presence (C and

Tag mRNA only in the low-density fraction, 4 (Fig. 5a).D) or absence (A and B) of 10 U/ml mIL-2 during the in vitro re-Messages for five neuroendocrine markers, insulin, glucagon,stimulation. Five days later cells were tested for killing of Tag

expression targets (A and C) or syngeneic normal fibroblasts (B somatostatin, pancreatic polypeptide and GAD67, made byand D). four different cell types in pancreatic islets, were all detected

in the low-density fraction 4. In addition, these mRNAs weredetected in the precursor fraction 2, depleted from fraction 3and increased in fraction 4, indicating that the major cell type4C and D) cells, indicating that they reside in the medulla.

Interestingly, cells expressing Tag were frequently observed in expressing these genes is enriched in the low-density fraction.Low levels of glucagon, somatostatin, pancreatic polypeptidepairs (Fig. 4J). At the age of peak expression, ~25% of all

Tag-staining cells were found as doublets. Collectively, these and GAD67 transcripts were also found in the perfusate,fraction 1. It has been shown that a proportion of the thymiccharacteristics are consistent with interdigitating dendritic or

medullary epithelial cells (18). However, the diffuse staining for dendritic cells can be released by thymic perfusion (19,20),which may explain the presence of some of these mRNAs inboth rel B and UEA I, and the inherent cell density of the thymus

made it difficult to unambiguously identify the Tag1 cells using fraction 1.The expression of four exocrine pancreatic markers wasa variety of immunohistochemical techniques.

Our previous study indicated that a number of pancreatic also investigated. Two of these, trypsin and elastase, werepreviously detected in newborn and adult thymus, whilegenes, normally thought of as being ‘peripheral’ or ‘tissue

restricted’, were expressed as mRNAs in newborn and adult mRNA for amylase and carboxypeptidase A1 were not (6).Messages for trypsin and elastase were clearly identified inthymus (6). Therefore we asked whether thymic cells

expressing their gene products could be detected. Double fraction 4, the low-density cells, while trypsin mRNA was alsodetected in fractions 1 and 2. In contrast to our earlier studies,immunohistochemical staining with antibodies recognizing rel

B and either insulin or somatostatin revealed that both islet cell we have now detected mRNA for the exocrine gene productamylase both in newborn thymus, as well as in thymic fractionsproteins could also be detected in a small number of cells in

the thymic medulla (Fig. 4E–H). No immunoreactive cells were from adult mice, most likely because the RNA-PCR techniqueemployed in the present study is more sensitive. Amylaseobserved without primary antiserum. Although mRNAs for glu-

cagon, amylase and pancreatic polypeptide were detectable message was most readily observed in the low-density fraction4, as well as in fraction 1, and, at low levels, in the other twoin thymic cDNA, antisera recognizing these proteins failed to

identify cells in the thymus (data not shown). fractions. We were, however, still unable to detect messagefor carboxypeptidase A1 in any of the thymic samples.

mRNA for insulin, Tag and other pancreatic genes is enriched In sum, mRNAs for nine pancreatic genes preferentiallyin fractionated thymic cells segregated into the low-density fraction of cells, which is

highly enriched in antigen-presenting cells, including dendriticIf insulin and somatostatin and indeed the other pancreaticgenes detectable as RNAs were all expressed by a distinct cells and macrophages.

Rare thymic cells express tissue-restricted genes 1361

Table 1. Summary of Tag protein detection in the thymus

Transgene Agea No. of No. of No. of Tag1 Range per No. of Tag1 peranimals sections cells section thymusb

RIP1-Tag2 P0 3 56 60c 0–6 [275]p3 3 59 11 0–4 [50]

RIP3-Tag2 p0 4 535d 16 0–2 5p7 2 423d 169 0–7 100

The thymi from animals in each group were collectively fixed and serially sectioned. Tag protein was detected by immunohistochemistryand the number of Tag-expressing cells was scored by microscopic examination.

aAge in days, where p0 is the day of birth.bApproximate number, based on extrapolation. The confidence level is greater for RIP3-Tag2 samples because all of the recovered sections

were stained, whereas only representative sections of RIP1-Tag2 samples were examined (in brackets).cNo Tag staining was observed in 38 sections from new born non-transgenic littermates of RIP1-Tag2.dRepresents ~85% of the total volume of the thymi in each set.

Tag and insulin expressing cells are MHC class II1 Discussion

In the current study we have shown that the intrathymicThe transplantation experiments revealed that transgenicexpression of an insulin-promoted transgene can inducethymus specifically conferred a tolerant phenotype on Thtolerance to the novel self protein it encodes. Several othercells. Given that developing CD41 Th cells recognize antigenstudies have previously reported that thymic expression ofpresented by MHC class II, we asked whether the thymic celltransgenes promoted by pancreatic regulatory elements cantype producing insulin and Tag expressed MHC class II. Theinduce tolerance (4,5,7). We have extended these findingslow-density fraction of thymic cells was treated with antibodiesby demonstrating that in RIP-Tag mice, the thymic contributionto MHC class II, and as controls Thy-1 and the B cell markerof tolerance to Tag, a protein which can be processed forB220, and antibody-reactive cells were then depleted usingpresentation by either MHC class I or class II molecules,magnetic beads. Figure 5(b) shows the result of one suchdiffers for CD41 helper and CD81 cytotoxic T cells. Weexperiment. Antibody to MHC class II depleted both messagesthereby provide support for the proposition that peripheralfrom this fraction of cells, whereas antibodies to Thy-1 hadexpression of Tag is also important for complete tolerancelittle effect and antibody recognizing B220 had none. FACSinduction, in particular of CD81 T cells, consistent withanalysis of the low-density cell fraction had revealed severalresults from other model systems (3,14). Another recentpopulations that express MHC class II, including dendriticstudy implicated thymic stroma in the induction of transgene-cells and macrophages, among others. Of these, we notespecific tolerance (7), but did not investigate the responsiblethat dendritic cells express very low levels of Thy-1, no B220cells. We show herein that Tag expressed under control ofand very high levels of MHC class II (15), and our data notthe insulin gene regulatory region can be localized to discreteshown, consistent with the pattern of antibody depletions.cells of the thymic medulla (by immunohistochemistry), thatthese Tag-expressing cells co-express MHC class II (byAutoimmune RIP1-Tag5 mice fail to express Tag mRNA infractionation and depletion analyses) and are therefore cap-thymusable of functioning as antigen-presenting cells to directly

Another measure of the significance of thymic expression has induce tolerance in developing T cells.come from the analysis of a line of autoimmune RIP-Tag mice,

Roles for both thymus and periphery in the induction ofRIP1-Tag5. Mice in this and several other families begintolerance to tissue-restricted self proteinsexpressing Tag in scattered β cells between 9 and 11 weeks

of age (data not shown), and develop hallmarks of autoimmune We have shown that low-level thymic Tag expression in twodisease, including the appearance of serum autoantibodies families of transgenic mice, RIP1-Tag2 and RIP3-Tag2, isand infiltration of the pancreatic islets by T and B lymphocytes sufficient to confer tolerance of the CD41 Th cells in the(8–10). RNA was prepared from thymus and pancreas of absence of pancreatic Tag expression. The toleranceperinatal, newborn and 1-week-old RIP1-Tag5 mice, the ages observed in these nu/nu mice reconstituted with transgenicat which thymic Tag expression was initially detected in thymus is similar to that observed in unmanipulated RIP1-tolerant RIP-Tag mice (6). We also collected RNA from adult Tag2 and RIP3-Tag2 mice. In contrast to the CD41 Th cellmice, ranging from 9 to 26 weeks of age, wherein Tag is tolerance elicited by thymic transplantation, thymic expressionknown to be expressed in pancreas. The analysis revealed of Tag was insufficient to ensure tolerization of the CD81

that Tag transcripts were not detectable in thymus of RIP1- cytotoxic T cells; CTL phenotypes ranged from completelyTag5 mice at any age (Fig. 6), even when a more sensitive tolerant to fully responsive, and thus non-tolerant. This charac-RNA-PCR assay was used than that employed for the previ- teristic differs from the phenotype of intact tolerant RIP-Tagously reported detection of Tag in the thymus of tolerant RIP- mice, which are profoundly non-responsive when challengedTag mice (6). In contrast, Tag transcripts were easily detected to generate Tag-specific CTL (Fig. 2) (6). Significantly, non-

responsiveness is not overcome by the in vitro addition of thein adult RIP1-Tag5 pancreas.

1362 Rare thymic cells express tissue-restricted genes

Th cell product IL-2 during CTL generation (Fig. 3), implyingthat the CTL in RIP1-Tag2 mice have received an additionaltolerization signal outside of the thymus, a signal which islacking in the transplanted nu/nu mice. The source of antigenthat elicits this signal in intact RIP1-Tag2 and RIP3-Tag2mice is almost certainly the main site of Tag synthesis, thepancreatic β cell. However, an alternative mechanism couldin principle contribute to the failure of CTL to become tolerantto Tag. It has recently been proposed that antigen-specificregulatory T cells are generated in the thymus to activelycontrol autoreactive T cells (21). If such cells only arise duringthe prenatal phase of T cell development, they could havefailed to develop recognizing Tag in our neonatal thymictransplant chimeras. When the characteristics of these putat-ive regulatory cells are better defined it will be of interest toassess their possible role in tolerance toward our modelself-antigen.

Kosaka and Sprent (14) have shown that tolerance ofhelper-independent CD81 T cells in parent → F1 chimeras isinduced by a two-step mechanism. An initial tolerization signalis delivered in the thymus and a second signal in the periphery,which drives the T cells into a further state of non-respons-iveness. Similar findings have been reported by Hammerling’sgroup for transgenic T cells recognizing the class I moleculeKb (3). Recently, Forster and Lieberam have shown usingTCR transgenic mice that CD41 T cells can be tolerized toTag in the draining lymph nodes of the pancreas (22). Thus,tolerant RIP-Tag mice have revealed mechanistic roles forboth thymic and peripheral expression in the induction oftolerance to a tissue-specific antigen. The present studyfurther suggests that the extent of tolerance induction bythese rare PAE cells in the thymus differs for helper andcytotoxic T cells recognizing the same protein.

Specialized thymic cells express peripheral antigens?

Our previous study demonstrated that Tag mRNA wasexpressed in the thymus of perinatal mice and that otherpancreatic mRNAs could be identified in both newborn andadult thymus (6). Herein we show that Tag, insulin andsomatostatin are detected by immunostaining in discrete cellsin transgenic thymic medulla, within fields of dendritic cellsand medullary epithelium. The frequency of Tag-producingcells correlates well with the level of mRNA detected by PCRin the thymus during perinatal life. In addition, Tag proteinwas identified in cells in transplanted thymi, up to 1 year post-transplant, indicating that thymic Tag expression persists intoadulthood. Fractionation of thymic cells demonstrated that acell fraction highly enriched for dendritic and other antigen-presenting cells is preferentially abundant in the mRNAs forFig. 4. Immunohistochemical detection of peripheral antigens in theTag, insulin, somatostatin, as well as the other pancreaticthymus. Sections of transgenic thymus (RIP3-Tag2, A–I; RIP1-Tag2,

J) were stained for Tag (A–D, I and J), insulin (E and F) and markers found to be expressed in the thymus. Depletion ofsomatostatin (G and H). Protein was visualized with an horseradish the fractionated thymic cells with antibody to MHC class IIperoxidase reaction and DAB chromogen (brown). Double demonstrated that the Tag and insulin transcripts are pro-immunohistochemistry for rel B (A, B and E–H) or UEA I (C and D)

duced by cells expressing MHC class II, consistent withwas then performed using SG chromogen (gray-blue). Panel (I) showsthe ability of these antigen-expressing cells to preferentiallyTag staining in a transgenic thymus 1 year after transplantation under

the kidney capsule of a nu/nu mouse. Single antigen assays were induce tolerance of CD41 Th cells. We note, however, a recentfollowed by brief counterstaining in either hematoxylin (I) or eosin (J) report (23) in which expression of another model antigenprior to mounting. Magnification: 3196 (A, C, E, G and I) or 3588

(nuclear β-galactosidase) by medullary thymic epithelium, but(B, D, F, H and I).not bone marrow-derived cells, resulted in tolerance of CD41

Rare thymic cells express tissue-restricted genes 1363

Fig. 5. Fractionation of thymic cell types reveals the PAE cell in a rare class enriched for non-lymphoid cells. (A) mRNA for Tag as well as forendogenous pancreatic genes is present in the low-density fraction of cells. cDNA from the four thymic cell fractions, prepared from RIP1-Tag2 mice, was tested for enrichment of sequences for Tag and several endogenous pancreatic genes by RNA-PCR. Presence or absenceof reverse transcriptase in the cDNA synthesis reaction is indicated by a ‘1’ or ‘–’. The amount of RNA, in µg, reverse transcribed is indicatedover the lane. Included as controls are pancreas (P) from a 6-week-old RIP1-Tag2 mouse and thymus (T) from a newborn (nb) RIP1-Tag2mouse. (B) Messages or Tag and insulin are present in low-density cells expressing MHC class II. Low-density cells, prepared from RIP1-Tag2 mice, were immunomagnetically depleted of thymocytes (Thy-1), B cells (B220) or MHC class II. cDNA was prepared from 7.5 µg ofRNA from cells which either bound (Bound) to the beads or which remained in the supernatant (Free). Samples were analyzed for the presenceof message for Tag, insulin and, as a control, β2-microglobulin by PCR. The antibodies used in the immunodepletions are indicated over thelanes. The lane marked with C is a newborn RIP1-Tag2 thymus cDNA. Non-depleted samples from fractions 1, 2, 3 and 4, from the samepreparation, are also included as controls. This experiment has been repeated with similar results.

cells via MHC class II presentation of the endogenouslyexpressed β-galactosidase.

Our data localizes the expression of certain peripheralantigens to a small number of MHC class II1 cells within themedulla; those cells may prove to be dendritic, epithelial ornovel. The cells we detect, collectively termed PAE cells, areimplicated as the source of a specific functional capacity ofthe thymus—induction of tolerance in CD41 T cells to a modelperipheral self antigen. Future experiments, including bonemarrow transfers and transplantation of fetal thymi, will allowthe lineage of PAE cells to be clarified.

We estimate that only a few hundred PAE cells express Tagin the thymic medulla. How can these few hundred cellspossibly tolerize the millions of thymocytes which transitthe medulla? Several studies have estimated that maturingthymocytes reside in the medulla between 12 and 16 days

Fig. 6. Failure to express Tag in thymus correlates with autoimmunity. prior to exit to the periphery (24). This long residence mayTag RNA is not detected in the thymus of RIP1-Tag5 mice. RNA, provide the necessary time for developing thymocytes toprepared from thymus or pancreas of RIP1-Tag5 mice of various

encounter their cognate PAE cell and become tolerized.ages, was assayed for the presence of Tag and β2-microglobulinHowever, it is possible to overwhelm the thymic tolerizationmessages by RNA-PCR. A ‘1’ and ‘–’ indicate whether reverse

transcriptase was added to the sample. The age of each mouse is machinery. When RIP1-Tag2 mice were bred to transgenicindicated over the lane: newborn (nb), 1, 9, 11, 15, 23 and 26 weeks. mice expressing a CD41 Tag-specific TCR on a majority ofThe standard (STD) is cDNA prepared from 1.5 µg of a 1/100 dilution developing T cells, tolerance of CD41 T cells of Tag failed toof newborn RIP1-Tag2 pancreatic RNA diluted into newborn non-

be induced (10). In light of our finding herein that there aretransgenic thymic RNA. Two to four animals were examined ateach age. only a few hundred Tag-expressing cells per thymus, it is not

1364 Rare thymic cells express tissue-restricted genes

surprising that greatly increasing the abundance of antigen- Summary and new directionsspecific T cells can overwhelm the tolerizing capacity of

We have identified rare MHC class II1 cells, termed PAE cells,the thymus.

which reside in the thymic medulla, express tissue-specificgenes and, in the case of a model antigen, are capable of

Does thymic expression of tissue-specific proteins play a eliciting CD41 Th cell tolerance. Notably, the absence ofgeneral role in limiting the development of potentially auto- thymic expression of the insulin–Tag transgene correlatesimmune T cells? with susceptibility to autoimmunity in an otherwise similar

transgenic family. These results raise a number of generalWe present evidence that tissue-specific genes normallyquestions for the future. First, are thymic PAE cells dendritic,produced by five different pancreatic cell types are preferen-epithelial, novel or a combination thereof? Does a single PAEtially expressed in a rare thymic cell population. The genescell express many different peripheral antigens or restrictedinclude three selectively expressed in the β cells, i.e. Tag,subsets of these tissue-restricted antigens? Do PAE cellsinsulin and GAD67, as well as glucagon, somatostatin andexpress antigens from other tissues besides the pancreas?pancreatic polypeptide, produced by the α, δ and PP cellsHow is gene regulation controlled in these cells; are thereof the endocrine islets respectively; in addition three exocrinecryptic thymic transcriptional enhancer elements associatedgene products, elastase, trypsin and amylase, are expressedwith tissue-specific genes? The results of Vafiadis and Pug-with a similar pattern. While this low level expression mightliese and their coworkers (25,26) suggest that the VNTRbe considered spurious or inconsequential, we have pro-element upstream of the human insulin genes might representduced mice whose only source of Tag is their thymus toa component of a thymic PAE gene regulatory element, sincedemonstrate that thymic Tag expression is sufficient to conferallelic variation in this element affects thymic expression levelsnon-responsiveness of CD41 Th cells. Moreover, we infer thatof insulin mRNA. Thus, the results from our study in mice,failure to express the Tag neo-self antigen in these medullaryexploiting a hybrid insulin transgene, and from the study ofcells is profound, since RIP1-Tag5 mice which develop aninsulin gene expression in human thymus, collectively makeautoimmune response directed against their pancreatic βa provocative case that variability in expression levels of βcells, the target of autoimmune destruction in type I diabetes,cell autoantigens such as insulin (33) or GAD65 (34) in PAEfail to express Tag in the thymus (Fig. 6). We suggest,cells may predispose individuals for autoimmune diabetes.therefore, that expression of these endogenous pancreatic

genes in medullary cells also serves a similar immunologicalfunction. In support of this hypothesis, Vafiadis and Pugliese

Acknowledgementsand their coworkers (25,26) have recently shown that allelesWe thank Raquel Nagal for animal husbandry and analysis of trans-of a genetic susceptibility locus for type I diabetes in humansgenic mice, Phil Galante for early immunohistochemical studies, Maryare associated with differing levels of insulin gene expressionCrowley for technical advice in the preparation of the thymic cell

in the thymus. The IDDM2 susceptibility locus maps to the 59 populations, Irmgard Forster, Nigel Killeen and Lewis Lanier forregion of the human insulin gene; humans carrying ‘protective’ critical reading of the manuscript, and Patti Keefe for editorial

assistance. We gratefully acknowledge postdoctoral support from thealleles of IDDM2 express relatively high levels of insulin mRNAAmgen Corp. and the American Cancer Society (D. O.) and thein thymus, in comparison to individuals carrying alleles thatUCSF Molecular Medicine Program (R. H.). This work was supportedare associated with increased risk of IDDM, who express by grants to D. H. from the National Cancer Institute and the Human

markedly lower levels. The data, while strictly correlative, Frontiers of Science Program, and benefited from an equipment grantto the UCSF transgenic mouse facility by the Markey Charitable Trust.suggest that allelic variations in thymic expression of ‘peri-

pheral’ autoantigens could have a profound effect on thebalance between self-tolerance and autoimmunity.

AbbreviationsOther investigators have reported expression of tissue-CFA complete Freund’s adjuvantrestricted genes in the thymus, principally of neuroendocrineCTL cytotoxic T lymphocytegenes (7,27–31). Beyond that, a comprehensive survey ofPAE peripheral antigen expressingorgan-specific genes for possible expression in the thymus RIP rat insulin promoter

has not yet been performed; it should be noted that extremely Tag SV40 large tumor antigensensitive methods will be required to detect both protein and UEA I ulex agglutinin ImRNA in these rare cells of the thymic medulla. Salient tothis general issue is a hypothesis, presented by Boone and

Referencesvan Pel (32), suggesting that the thymus expresses many orall tissue-specific genes as subgenomic fragments. They 1 Fowlkes, B. J. and Ramsdell, F. 1993. T-cell tolerance. Curr. Opin.

Immunol. 5:873.proposed that promiscuous expression of peripheral proteins2 Sprent, J. and Webb, S. R. 1995. Intrathymic and extrathymicin the thymus would allow immature thymocytes to become

clonal deletion of T cells. Curr. Opin. Immunol. 7:196.tolerant of self proteins which they were destined to encounter 3 Hammerling, G. J., Schonrich, G., Ferber, I. and Arnold, B.in the periphery, thus limiting the development of potentially 1993. Peripheral tolerance as a multistep mechanism. Immunol.

Rev. 133:93.autoreactive T cells. Consistent with that postulate, we have4 Heath, W. R., Allison, J., Hoffman, M. W., Schonrich, G.,presented evidence that rare thymic medullary cells express

Hammerling, G., Arnold, B. and Miller, J. F. A. P. 1992. Autoimmunea set of genes which are considered tissue restricted and diabetes as a consequence of locally produced interleukin-2.have further shown that this expression can have functional Nature 359:547.

5 von Herrath, M. G., Dockter, J. and Oldstone, M. B. 1994. Howconsequences for immune self-tolerance.

Rare thymic cells express tissue-restricted genes 1365

virus induces a rapid or slow onset insulin-dependent diabetes The physiological role of regulatory T cells in the prevention ofautoimmunity: the role of the thymus in the generation of themellitus in a transgenic model. Immunity 1:231.

6 Jolicoeur, C., Hanahan, D. and Smith, K. M. 1994. T cell tolerance regulatory T cell subset. Immunol. Rev. 149:195.22 Forster, I. and Lieberam, I. 1996. Peripheral tolerance of CD4toward a transgenic β cell antigen and transcription of

endogenous pancreatic genes in the thymus. Proc. Natl Acad. T cells following local activation in adolescent mice. Eur. J.Immunol. 26:3194.Sci. USA 91:6707.

7 Antonia, S. J., Geiger, T., Miller, J. and Flavell, R. A. 1995. 23 Oukka, M., Colucci-Guyon, E., Tran, P., Cohen-Tanoudji, M.,Babinet, C., Lotteau, V. and Kosmatopoulos, K. 1996. CD4 TMechanisms of immune tolerance induction through the thymic

expression of a peripheral tissue-specific protein. Int. Immunol. cell tolerance to nuclear proteins induced by medullary thymicepithelium. Immunity 4:545.7:715.

8 Adams, T., Alpert, S. and Hanahan, D. 1987. Non-tolerance 24 Scollay, R. and Godfrey, D. I. 1995. Thymic emigration: conveyerbelts or lucky dips? Immunol. Today 16:268.and autoantibodies to a transgenic self antigen expressed in

pancreatic β-cells. Nature 325:223. 25 Vafiadis, P., Bennett, S. T., Todd, J. A., Nadeau, J., Grabs, R.,Goodyer, C. G., Wickramasinghe, S., Colle, E. and Polychronakos,9 Skowronski, J., Jolicoeur, C., Alpert, S. and Hanahan, D. 1990.

Determinants of the B cell response against a transgenic C. 1997. Insulin expression in human thymus is modulated byINS VNTR alleles at the IDDM2 locus. Nat. Genet. 15:289.autoantigen. Proc. Natl Acad. Sci. USA 87:7487.

10 Forster, I., Hirose, R., Arbeit, J. M., Clausen, B. and Hanahan, D. 26 Pugliese, A., Zeller, M., Fernandez, A., Zalcberg, L. J., Bartlett,R. J., Ricordi, C., Pietropaolo, M., Eisenbarth, G. S., Bennett, S. T.1995. Limited capacity for tolerization of CD41 T cells specific

for a pancreatic β cell neoantigen. Immunity 2:573. and Patel, D. D. 1997. The insulin gene is transcribed in thehuman thymus and transcription levels correlate with the allelic11 Hanahan, D. 1985. Heritable formation of pancreatic β-cell tumors

in transgenic mice harboring recombinant insulin/simian virus 40 variation at the INS VNTR-IDDM2 susceptibility locus for type 1diabetes. Nat. Genet. 15:293.oncogenes. Nature 315:115.

12 Knowles, B. B., Koncar, M., Pfizenmaier, K., Solter, D., Aden, D. 27 Geenen, V., Legros, J.-J., Franchimont, P., Baudrihaye, M.,Defresne, M.-P. and Bonviver, J. 1986. The neuroendocrineand Trinchieri, G. 1979. Genetic control of the cytotoxic T cell

response to SV40 tumor-associated specific antigen. J. thymus: coexistence of oxytocin and neurophysin in the humanthymus. Science 232:508.Immunol. 122:1798.

13 Trinchieri, G., Aden, D. P. and Knowles, B. B. 1976. Cell-mediated 28 Fuller, P. J. and Verity, K. 1989. Somatostatin gene expression inthe thymus gland. J. Immunol. 143:1015.cytotoxicity to SV40-specific tumour-associated antigen. Nature

261:312. 29 Bantanero, E., De Leeuw, F.-E., Jansen, G. H., van Wichen, D. F.,Huber, J. and Schuurman, H.-J. 1992. The neural and neuro-14 Kosaka, H. and Sprent, J. 1993. Tolerance of CD81 T cells

developing in parent → F1 chimeras prepared with supralethaal endocrine component of the human thymus. Brain Behav.Immun. 6:249.irradiation; stepwise induction of tolerance in the intrathymic and

extrathymic environments. J. Exp. Med. 177:367. 30 Mathisen, P. M., Pease, S., Garvey, J., Hood, L. and Readhead,C. 1993. Identification of an embryonic isoform of myelin basic15 Crowley, M., Inaba, K., Witmer-Pack, M. and Steinman, R. M.

1989. The cell surface of mouse dendritic cells: FACS analysis protein that is expressed widely in the mouse embryo. Proc. NatlAcad. Sci. USA 90:10125.of dendritic cells from different tissues including thymus. Cell.

Immunol. 19:108. 31 Pribyl, T. M., Campagnoni, C. W., Kampf, K., Kshima, T., Handley,V. W., McMahon, J. and Campagnoni, A. T. 1993. The human16 Swiggard, W., Nonacs, R. M., Witmer-Pack, M. D., Steinman, R. M.

and Inaba, K. 1992. Enrichment of dendritic cells by plastic myelin basic protein gene is included within a 179-kilobasetranscription unit: expression in the immune and central nervousadherence and EA rosetting. Curr. Prot. Immunol. 3:7.1.

17 Burkly, L., Hession, C., Ogata, L., Marconi, L. A., Olson, D., systems. Proc. Natl Acad. Sci. USA 90:10695.32 Boone, T. and van Pel, A. 1989. T cell-recognized antigenicTizard, R., Cate, R. and Lo. D. 1995. Expression of rel B is

required for the development of thymic medulla and dendritic peptides derived from the cellular genome are not proteindegradation products but can be generated directly bycells. Nature 373:531.

18 Fairchild, P. J. and Austyn, J. M. 1990. Thymic dendritic cells: transcription and translation of short subgenic regions. Ahypothesis. Immunogenetics 29:75.phenotype and function. Int. Rev. Immunol. 6:187.

19 Kyewski, B. A., Fathman, C. G. and Rouse, R. V. 1986. Intrathymic 33 Eisenbarth, G. S., Jackson, R. A. and Pugliese, A. 1992. Insulinautoimmunity: the rate limiting factor in pre-type I diabetes.presentation of circulating non-MHC antigens by medullary

dendritic cells. J. Exp. Med. 163:231. J. Autoimmun. 5 (Suppl. A): 241.34 Ellis, T. M. and Atkinson, M. A. 1996. The clinical significance of20 Lorenz, R. G. and Allen, P. M. 1989. Thymic cortical epithelial

cells can present self-antigens in vivo. Nature 337:560. an autoimmune response to glutamic acid decarboxylase. Nat.Med. 2:148.21 Saoudi, A., Seddon, B., Heath, V., Fowell, D. and Mason, D. 1996.