Autoimmune islet destruction in spontaneous type 1 diabetes ... - Nature

5 downloads 0 Views 2MB Size Report
Jan 21, 2003 - tis') and eventually inside it ('invasive insulitis'). The patho- genicity of intra-islet infiltrates is curiously low, with months of T-cell accumulation ...
ARTICLES

Autoimmune islet destruction in spontaneous type 1 diabetes is not β-cell exclusive SHAWN WINER1, HUBERT TSUI1, AMBROSE LAU1, AIHUA SONG , XIAOMAO LI2, ROY K. CHEUNG1, ANASTAZIA SAMPSON1, FATEMEH AFIFIYAN1, ALISHA ELFORD3, GEORGE JACKOWSKI2, DOROTHY J. BECKER4, PERE SANTAMARIA5, PAMELA OHASHI3 & H.-MICHAEL DOSCH1

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

1

1

Hospital For Sick Children, Research Institute and Departments of Pediatrics and Immunology, University of Toronto, Toronto, Ontario, Canada, 2 Syn*X Pharma Inc., Toronto, Ontario, Canada 3 UHN, Ontario Cancer Institute, Departments of Medical Biophysics and Immunology, Toronto, Ontario, Canada 4 Department of Pediatrics, Children’s Hospital and University of Pittsburgh, Pittsburgh, Pennsylvania, USA 5 Department of Microbiology and Infectious Diseases and Julia MacFarlane Diabetes Research Centre, Faculty of Medicine, Health Sciences Centre, University of Calgary, Calgary, Canada S.W. and H.T. contributed equally to this study. Correspondence should be addressed to H.-M.D.; email: [email protected]

Published online 21 January 2003; doi:10.1038/nm818

Pancreatic islets of Langerhans are enveloped by peri-islet Schwann cells (pSC), which express glial fibrillary acidic protein (GFAP) and S100β. pSC-autoreactive T- and B-cell responses arise in 3- to 4-week-old diabetes-prone non-obese diabetic (NOD) mice, followed by progressive pSC destruction before detectable β-cell death. Humans with probable prediabetes generate similar autoreactivities, and autoantibodies in islet-cell autoantibody (lCA) –positive sera co-localize to pSC. Moreover, GFAP-specific NOD T-cell lines transferred pathogenic peri-insulitis to NOD/severe combined immunodeficient (NOD/SCID) mice, and immunotherapy with GFAP or S100β prevented diabetes. pSC survived in rat insulin promoter Iymphocytic choriomeningitis virus (rip–LCMV) glycoprotein/CD8+ T-cell receptorgp double-transgenic mice with virus-induced diabetes, suggesting that pSC death is not an obligate consequence of local inflammation and βcell destruction. However, pSC were deleted in spontaneously diabetic NOD mice carrying the CD8+/8.3 T-cell receptor transgene, a T cell receptor commonly expressed in earliest islet infiltrates. Autoimmune targeting of pancreatic nervous system tissue elements seems to be an integral, early part of natural type 1 diabetes.

Type 1 diabetes (T1D) in humans and its premier animal model, the non-obese diabetic (NOD) mouse, are polygenic autoimmune diseases whose penetrance is controlled by environmental factors1–3. Insulin deficiency is the end result of a slowly progressive process, pre-diabetes, characterized by the accumulation of T-cell infiltrates around the islet (‘peri-insulitis’) and eventually inside it (‘invasive insulitis’). The pathogenicity of intra-islet infiltrates is curiously low, with months of T-cell accumulation before pronounced β-cell death late in prediabetes, around the development of insulin deficiency4. This slow progression of pre-diabetes is not well understood, but is typical for human and NOD disease. Early prediabetes in NOD mice has been targeted by a variety of immunotherapies that successfully ameliorate its progression, but interventions later in prediabetes, when T cells have infiltrated the islet interior, are generally ineffective and may precipitate disease5–7. The development of invasive insulitis thus marks a critical checkpoint in prediabetes progression whose underlying mechanisms remain uncertain8. More comprehensive staging of prediabetes, identification of the factors that control its progression and greater insight into the mecha198

nisms of effective immunotherapies in diabetes-prone rodents are all important steps in the development of intervention strategies that can be considered for human T1D. Although pathogenic tissue infiltrations in T1D are largely islet restricted, few at best of the target autoantigens are exclusively expressed in islets, and many are constituents of central and peripheral nervous system tissue (for example, glutamic acid decarboxylases (GAD65, GAD67), IA-2 and ICA69)9–11. Diabetic autoimmunity in humans and NOD mice includes spontaneous development of clinically latent T-cell autoreactivities that target nervous system autoantigens. T1D is more common in patients and families with multiple sclerosis (MS)12. An MS-like autoimmune encephalitis (AENOD) develops following treatment of NOD mice with pertussis toxin13 and a high-incidence, peripheral autoimmune neuropathy develops in NOD mice lacking the co-stimulatory CD28 ligand, B7-2 (ref. 14). In the search for possible links between islet and nervous system autoimmunity, we observed spontaneous autoimmune targeting of pancreatic nervous system tissue elements, peri-islet Schwann cells (pSC), in both diabetes-prone humans and NOD mice. NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

ARTICLES

a

b

c

d

e

f

g

h

i

k

j

l

m

Fig. 1 GFAP+ pSC are the first target tissue in NOD prediabetes. a, Absence of leukocytic infiltrations from an H&E stained islet from a 5week-old NOD female. b, Serial section of the same islet, double stained with antibodies against insulin (blue) and GFAP (green). c, A transmission electron micrograph of a pSC (black arrow, cell body) separating exocrine (right) from endocrine (left) tissue. White bar, 10 µm. d, A transmission electron micrograph of a pSC, separating exocrine (top) from endocrine tissue (bottom). The black triangle depicts a thin cytoplasmic extrusion originating from the pSC body in close contact with the islet. White bar, 4 µm. e, The neuro-insular complex (arrow) from a 6-week-old NOD female stains heavily

for GFAP. f, GFAP+ pSC (green) in 10-week-old NOD/SCID mice. g, Purified human islet stained for GFAP (green) and insulin (blue). h, Double-stained image of peri-insulitis in a 7-week-old NOD female with accumulation of CD3+ T cells (red) at the pSC border (green). i and j, Localized pSC breaches in two 9-week-old females. k, T cells encircle pSC (white arrows), but spare islet interior regions free of pSC, which contain an apparently intact β-cell mass (inset: insulin stained blue). l, Destroyed pSC (green) and invasive insulitis (red) in a 14-week-old NOD female. m, Destroyed islet at diabetes onset with absent GFAP (green) and insulin staining (blue). Magnification: a, b, e–j, l and m, ×200; k, ×400.

Peri-islet Schwann cells The pancreatic islets of Langerhans are surrounded by Schwann cells15,16, marked by the expression of S100β and glial fibrillary acidic protein (GFAP)17,18. These cells form a tight, pSC mantle that envelops the endocrine islet tissue (Fig. 1a and b). Electron microscopy reveals thin, bi-layered membrane extrusions (Fig. 1c and d) from the Schwann cell body, separating endocrine

from exocrine pancreas. pSC converge at the neuro-insular complex (Fig. 1e, arrow) with nude axons and sympathetic nerve fibers17, and they show a similar configuration in NOD/SCID (Fig. 1f) mice, as well as human islets (Fig. 1g) and SJL/J, BALB/c and C57BL/6 mice (data not shown). NOD mice in our colony develop overt T1D in a staggered fashion between the ages of 17 and 35 weeks in 85% of females

NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

199

ARTICLES

a

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

d

s

i

i

Stimulation index

b

c

Fig. 2 B- and T-cell autoimmunity to pSC constituents in NOD females. a, NOD mice, but not C57BL/6, BALB/c or SJL/J mice (4–6/group), show significant T-cell responses to GAD65 (10 µg/ml, ), GFAP (10 µg/ml, ) and S100β (50 µg/ml, ) starting at 4–5 weeks of age and persisting to diabetes (background 692–1,120 cpm, mean 954 ± 153). OVA, control, . b, T-cell responses in various regional lymph nodes from 10-week NOD females. Lymph node cells were pooled from 5–7 mice, experiments were repeated three times and data were averaged (background 421–776 cpm, mean 608 ± 241). Pancreatic lymph node cells, ; axillary lymph node cells, ; inguinal lymph node cells, . c, GFAP transcripts are detectable by template-calibrated RT-PCR in brain but not the NIT β-cell line, which expresses traces of S100β and large amounts of ICA69. β-glucuronidase (GUS) was used for calibration. d, Covalently GFAP-coupled proteomic chip arrays (top) were incubated with sera from 3- to 10-week-old NOD females or males as indicated. Chips were washed and read in a SELDI time of flight

mass spectrograph. Control chip surfaces (middle) were identical except for the absence of GFAP. The differential peak signal (∆p) is shown below (bottom). Large peaks at 150 kDa (IgG) were observed only in female sera; 4 of 31 similar profiles are shown. e, Western blot for the detection of antibodies against GFAP in mouse sera (0.5 µg of GFAP in each lane). Lane 1, 8-week C57BL/6; lanes 2–7, NOD female sera (1:150), 3.5, 5, 6, 8, 10 and 20 weeks old, respectively; lane 8, 1:1,000 IgG GFAP antibody.

and 15% of males19. In 1-month-old NOD females, T cells immigrate to the pSC-defined endo-exocrine tissue junction, and they accumulate over several weeks in tight contact with the pSC layer (Fig. 1h and i). In 2-month-old females, many weeks before diabetes, islets can be found that show circumscript breaches in the pSC mantle (Fig. 1i and j), with T lymphocytes encircling Schwann cells from the islet exterior and interior (Fig. 1k, arrows). Even after T cells penetrate the pSC mantle, they usually accumulate contiguous to pSC, rather than moving to the islet interior, which remains largely free of T cells and has apparently intact β-cell mass (Fig. 1k, inset).

As T cells accumulate, pSC lose their regular organization and are progressively deleted, suggesting that this tissue is an early target of the autoimmune response in prediabetic mice (Fig. 1l). At best, few pSC remnants are observed in late prediabetic and overtly diabetic animals (Fig. 1m). The absence of pSC destruction in non-autoimmune BALB/c, C57BL/6 and NOD/SCID mice (Fig. 1f) documents the autoimmune nature of early pSC demise in prediabetic NOD females. pSC destruction is not synchronous, but in animals aged 2 months or less, pSC seem to be largely intact except for localized breaches of mantle integrity in a growing number of islets. All stages of pSC breach and T-cell invasion can

a

charge

e

1

2

3

4

5

6

7

8

47.5

c

b

d Fig. 3 pSC autoimmunity in human T1D. a, Proliferative T-cell responses to GFAP (green) and S100β (yellow) in peripheral blood mononuclear cells from newly diabetic children, ICA autoantibody-positive (ICA+) FDRs with probable prediabetes, or FDRs with no sign of diabetic autoimmunity. Data are presented as stimulation index (SI, cpm antigen stimulated/medium control; background 900–1800 cpm, mean 1,265 ± 215). Positive responses were ≥3 s.d. above mean OVA responses, P values < 0.002 versus healthy controls, Mann Whitney test. Red triangles, see text. b, Anti-GFAP autoantibody detection on SELDI proteomic chip arrays in sera of diabetic or healthy donors (see Fig. 2d). Top, GFAP surface; middle, control surface; bottom, ∆p. c–d, Sera from healthy, ICA negative (c) or prediabetic, ICA positive (d) FDR were used for indirect immunofluorescence of islet cryo-sections (red, human IgG), counterstained with anti-GFAP for pSC detection (green). Overlapping fluorescence emission (yellow) co-localizes autoantibodies to the pSC. 200

NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

be observed in different islets from the same NOD females at 10–12 weeks of age. Littermate NOD males, which resist T1D, show essentially intact pSC barriers. These observations confirm and extend a previous structural study of pSC, with our data suggesting that pSC death precedes invasive insulitis and β-cell death, rather than reflecting a consequence of β-cell injury18. pSC autoimmunity in NOD mice Earlier studies identified S100β, GFAP and glutamic acid decarboxylase (GAD) as pSC components17,18,20. We found that autoreactive T cells specific for these pSC proteins appeared spontaneously in 4-week-old NOD females (Fig. 2a). This extends previous reports of the early appearance of GAD65-specific T cells21,22 to other pSC proteins. These T-cell pools were detected by T-cell proliferation assays of spleen cell cultures and thus had considerable systemic distribution. They likely arise earlier, as occasional animals had positive spleen cell responses at weaning (Fig. 2a). pSC-specific T cells were also detected and may arise in pancreatic lymph nodes (Fig. 2b). This localized spontaneous pSC autoimmunity to the pancreas, as pSC autoreactivity was absent from other regional lymphatic tissues. T cells of 2-month-old SJL/J, C57BL/6 and

a

e

b

f

Fig. 4 Autoimmunity to pSC antigens induces insulitis and modifies adoptively transferred diabetes development. a, T-cell proliferative response of the GFAP-specific T cell line. b, A representative islet from a female NOD/SCID mouse, 6 d after transfer of a GFAP-specific T cell line. The arrow points to a neuro-insular complex. Green, GFAP; red, CD3; blue, insulin. Magnification, ×200. c, Peri-insulitis in an islet from a female NOD/SCID mouse 6 days after transfer of a GFAP-specific T cell line. Green, GFAP; red, CD3; blue, insulin. Magnification, ×200. d, Immunofluorescence of progressive pSC destruction in NOD/SCID 3 weeks after T-cell line transfer (GFAP, green; CD3, blue; CD4, red; CD3+/CD4+/CD8– cells, purple; CD3+/CD4–, likely CD8+ T cells, blue; original magnification ×200). e, GFAP (, 5 µg, n = 13) and S100β (, 50 µg, n = 21), but not OVA (, 50 µg, n = 21) or PBS (, n = 29), emulsified in IFA prevented the adoptive transfer of T1D with 107 diabetic spleen cells. f, Proliferative GFAP and S100β T-cell responses are boosted in disease-protected adoptive transfer recipients from (e), 35 d post transfer (n = 3–5/group). , GFAP responses; , S100β responses; , OVA NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

BALB/c mice failed to respond to pSC antigens, associating pSC autoimmunity with the diabetes-prone (NOD) phenotype. β-Cells express trace amounts of GAD65 (ref. 9) as well as S100β, but lack GFAP expression detectable by RT-PCR (Fig. 2c). We therefore focused on GFAP as a local pSC marker in studies of autoantibodies, using two different approaches. IgG autoantibodies to GFAP were measured using covalently GFAP-coupled proteomic chip arrays in a SELDI time-of-flight mass spectrometry system23, calibrated with a monoclonal IgG antibody against GFAP. Three of five sera from 23- to 26-day-old females, seven of nine sera from 1-month-old females and nine of ten sera from older females contained a GFAP-binding protein of the expected mass for IgG (149,805.7 ± 1,200 Da; Fig. 2d). Sera from young males were negative (0/8; Fig. 2d, top right), but occasionally, low-level signals were observed in 10- to 16-week-old males (2/9 sera; Fig. 2d, bottom right). IgG autoantibodies against GFAP were also detected in western blots, albeit with less sensitivity, in sera from NOD females, and they were maintained in overtly diabetic mice at 5 months of age (Fig. 2e, lanes 2–7). Sera from 7- to 8-week-old non-autoimmune C57BL/6 (Fig. 2e, lane 1), BALB/c or SJL mice were negative in both assays (data not shown).

c

d

g

h

responses. GFAP- and S100β responsive T cells secrete copious amounts of IFNγ (n = 3–5/group, mean values >200 pg/ml are shown as purple circles). g, Equal distribution of CD4+ (red) and CD8+ (green) cells in a representative islet from a non-diabetic PBS treated animal, 35 d post transfer (>30 islets were stained, ×200). h, Near absence of CD8+ cells (green) and abundance of CD4+ (red) cells in a representative islet from a GFAP-treated, T1D-protected adoptive transfer recipient, 35 d post transfer (>30 islets were stained, ×200). 201

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

ARTICLES

a

b

c

d

e

f

g

h

i

j

Fig. 5 Differential effects of two transgenic TCR diabetes models. a, Islet from LCMV RIP-gp, TCRgp-double transgenic C57BL/6 mouse not infected with LCMV. b–e, Progressive destruction of insulin-producing β-cells, 2 (b, c) to 4 (d, e) days

after LCMV infection. f, pSC damage during peri-insulitis in wild-type NOD female. g–j, Progressive pSC damage in NOD mice carrying the NY8.3 TCR transgene. (GFAP, green; CD3, red; insulin, blue; magnification, ×200).

These observations suggested that autoimmunity to the pSC proteins GFAP, S100β and GAD65 is associated with the progressive disease course in NOD females. Autoantibodies against GFAP may represent a new and early serum marker of high diabetes risk in NOD mice.

trols contained autoantibodies against GFAP detected by SELDI (Fig. 3b, P = 0.03, Fisher; ref. 23). Sera of ICA+ and ICA-negative FDRs were also compared by two-color immunofluorescence in fresh-frozen pancreas (Fig. 3c and d). When pSC were counterstained for GFAP, autoantibodies present in ICA+ FDR sera, but not ICA-negative control sera, distinctly co-localized to the pSC mantle, providing direct visualization of pSC autoimmune targeting by human diabetes-associated autoantibodies. Identification and prospective follow-up of GFAP-autoantibody positive FDRs is under way to determine if anti-GFAP autoimmunity has a value for predicting long-term disease.

pSC autoimmunity in humans Based on these conclusions, we examined members of T1D kindreds from the Pittsburgh diabetes registry24,25. Patients with recent onset T1D and first-degree relatives (FDRs) of index cases were recruited through informed consent, and heparinized blood samples were shipped to Toronto by overnight courier and analyzed by in vitro proliferation assays for responses to a panel of 14–22 disease-relevant islet and control self and foreign proteins/peptides as described25. Patients (n = 56) had positive responses to multiple diabetes autoantigens in 89% of cases (505/667 responses measured; data not shown), and all but three of these had GFAP- and/or S100β-specific T cells (Fig. 3a). FDRs were stratified according to the presence (n = 53) or absence (n = 27) of islet cell cytoplasmic autoantibodies (ICA) and T1D-risk associated HLA DQ8 genotype24,25. In this registry, ICA+ high-risk FDRs have a cumulative T1D risk of 70–80% over 15 years24. T-cell responses in ICA+ FDRs varied widely, from responses to multiple test antigens (60% of subjects) to absent autoreactivities (29%). GFAP-specific T cells were detected in 73% of subjects with multiple autoreactivities, almost always including GAD65, IA-2 and ICA69. In the ICA-negative cohort, there were only occasional positive responses to test autoantigens (18/400 different autoantigen responses measured). However, 13 of these 18 responses clustered in four subjects (red triangles, Fig. 3a) who, by this token, may carry a higher T1D risk; these four subjects all had GFAP responses (Fig. 3a). The differences between GFAP/S100β responses in ICA-negative, low-risk FDRs and patients or ICA+, high-risk FDRs were significant (P values < 0.0001, chi square). Serum samples from 6 of 8 new onset patients, 12 of 15 ICA+ FDRs with probable prediabetes, and only 2 of 12 healthy con202

Relevance of pSC autoimmunity in prediabetes Several experiments were carried out to further characterize pSCspecific autoimmunity. We generated T-cell lines, specific for GFAP, from female NOD splenocytes (Fig. 4a). These were infused into NOD/SCID mice that lack endogenous lymphocytes. Within 1 week, infused lymphocytes emigrated from exocrine vessels to the peri-islet regions of the pancreas, accumulating at the neuro-insular complex (Fig. 4b, arrow), and the cells rapidly progressed to typical peri-insulitis profiles with notable disruption and damage of the pSC envelope (Fig. 4c). The infiltrating T cells stained for CD3 and CD4, and they progressively expanded pSC damage over the ensuing weeks (Fig. 4d). These animals developed neither T1D nor obvious neurological abnormalities. pSC-autoreactive T cells selectively home to the islet in NOD mice and initiate pSC damage, but progression to diabetes seems to require autoimmunity to other (β-cell) antigens. To determine the possible role of pSC autoimmunity within prediabetes progression, we analyzed the effectiveness of pSCbased immunotherapy. The adoptive transfer model of T1D was used as a robust, accelerated disease model7,26 involving spleen cell transfer from newly diabetic mice into young, irradiated NOD recipients. GFAP, S100β or controls (phosphate buffered saline (PBS) or ovalbumin (OVA)) were emulsified in incomplete Freund’s adjuvant (IFA) and injected intraperitoneally (i.p.) 48 NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

ARTICLES hours after spleen cell transfer7. Acute diabetes developed in control animals within 2–3 weeks (Fig. 4e). Animals injected with GFAP or S100β showed significantly delayed and reduced diabetes development (P values < 0.001, log-rank tests). Specific GFAP or S100β T-cell autoreactivity in GFAP- or S100β-treated mice was boosted (Fig. 4f, P < 0.0001 versus PBS, Mann-Whitney) and showed a strong bias for interferon (IFN)-γ (Fig. 4f; values of 200 islets scored/group; data not shown, but see Fig. 4h). To understand how immunotherapy with a non-β-cell protein might mediate diabetes protection, we stained islets for T-cell markers. CD4+ (Fig. 4g and h) and CD8+ T cells showed an approximately equal distribution in PBS- or OVA-treated animals (Fig. 4g), whereas CD8+ T cells were absent from the ring-like peri-islet infiltrates of GFAP-treated, protected mice (Fig. 4h). The distribution of CD4+ and CD8+ T cells was normal in spleens from all mice (data not shown). Progression of prediabetes critically depends on CD8+ T cells27,28, and the immunotherapy-induced deviation towards CD4+ bias may identify a new T1D protective mechanism. Our observation of CD4+ biased islet infiltration may be related to high IFN-γ levels, as preliminary observations in IFN-γ–mutant NOD congenics failed to show disease protection and islet CD4+ bias after GFAP immunotherapy (data not shown). There is precedence for IFN-γ–dependent diabetes protection in NOD mice treated with complete Freund’s adjuvant29. pSC destruction is not a bystander effect Loss of pSC during prediabetes may reflect pSC-specific autoimmunity or collateral damage (that is, bystander effects) mediated by β-cell–specific T cells. To test this possibility, we compared two disease models in which diabetes is mediated by CD8+ T cells carrying a transgenic T-cell receptor (TCR). In the first model, the rat insulin promoter (RIP) was used to direct the expression of the lymphocytic choriomeningitis virus (LCMV) glycoprotein (gp) in β-cells30. These transgenic C57BL/6 mice were bred with P14 TCR transgenics that express a rearranged TCR specific for LCMV-gp and H-2Db on most CD8+ T cells31. The second model involves NOD mice carrying the diabetogenic NY8.3 TCR transgene28,32. This TCR was derived from and is commonly found in the earliest islet infiltrates of NOD females33,34. Untreated RIP-gp/P14 double-transgenic mice have normal, pSC-enveloped islet structures and absent T cell infiltrates (Fig. 5a). Tissue attack in LCMV-induced diabetes is narrowly targeted to one β-cell (auto-) antigen (gp) and overt diabetes develops rapidly after infection with live virus. T-cell infiltrates were observed as early as 2 days after LCMV infection, with depletion of insulin-positive β-cells and some alteration, but no major pSC elimination (Fig. 5b and c). In animals with overt diabetes, insulin-expressing β-cells were severely depleted, but the pSC islet envelope remained largely intact (Fig. 5d and e). In this experimentally induced diabetes model, pSC survive despite local inflammation and massive β-cell death, suggesting that pSC are not merely a physical barrier that succumbs to collateral or bystander damage during β-cell attack. In contrast, spontaneous diabetes in 8.3 TCR transgenic NOD mice mimicked the NOD disease course, although at a faster pace. Wild-type NOD mice show peri-insulitis with typical pSC depletion and apparently intact β-cell mass (Fig. 5f). Similar but more progressive pSC destruction was observed in 8.3 TCR-transNATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

genic NOD mice, with peri-insulitis development (Fig. 5g) and complete pSC deletion (Fig. 5h–j). pSC demise in this model maps to a distinct TCR, naturally generated in the course of early NOD prediabetes33,34 and pathogenic for pSC as well as β-cells28. Discussion The data presented here document that autoimmunity in T1D targets pSC, a nervous system component, early in prediabetes. Similar pSC autoantibodies and T-cell autoreactivities were found in humans with probable prediabetes and young NOD females, suggesting that pSC autoimmune targeting may represent a natural element of the prediabetes progression program in spontaneous T1D. Although diabetes occurred without pSC destruction in the LCMV model, pSC and β-cells were destroyed in NOD and 8.3 TCR-transgenic mice. Autoimmune diabetes can thus develop through two distinct pathways, differentiated by the presence or absence of pSC demise. Collectively, the data lead to the hypothesis that early autoimmunity in spontaneous T1D targets nervous system tissue elements in the pancreas, and is thus not β-cell exclusive. The evidence is compelling that NOD pSC are the first pancreatic tissue elements to be destroyed en masse. While β-cell damage may occur at that time, it would be subtle. However, it will be difficult to formally rule out bystander effects, whereby β-cell–specific T cells ‘inadvertently’ damage neighboring pSC or whereby pSC-directed autoimmunity inadvertently damages adjacent β-cells. Data from the LCMV model demonstrate that inflammation and autoimmune killing of β-cells are not sufficient for pSC destruction. We observed that pSC-specific T-cell lines initiated insulitis but not diabetes in NOD/SCID hosts, recapitulating early phases of NOD prediabetes. The early appearance of pSC-specific T cells and autoantibodies is most consistent with direct autoimmune targeting of pSC as an early component of progressive prediabetes. pSC-derived GAD65 (ref. 20) may contribute to the autoimmune targeting of GAD65 early in prediabetes21,22, which has been difficult to understand in the face of low β-cell GAD65 levels9. Schwann cells can present antigen through MHC classes I and II (refs. 35,36), and early autoimmune targeting of β-cell antigens such as insulin37 may involve cross-presentation, facilitated by the close proximity of pSC, β-cells and high local insulin concentrations. Antigen spreading is a well-established process in T1D (ref. 21), including the human disease, in which progressively more islet proteins are targeted by expanding autoimmune repertoires in mid-late prediabetes25,38. It will be interesting to determine whether pSC autoimmune targeting evolves through antigen spreading from initially β-cell–specific autoreactivity, whether the opposite is true, or whether both events can occur. pSC autoreactivity can be detected as early as 3 weeks of age, before detectable pancreas pathology, and a single 8.3+ T-cell clone arising early in NOD prediabetes33,34 killed both pSC and β-cells. The 8.3 target autoantigen is unknown, but may be expressed in both cell types (analogous to GAD65 and S100β), and killing would be cognate, initiated by either βcell- or pSC-derived antigen initially presented in draining pancreatic lymph nodes39. Immunotherapy with GFAP and S100β protected wild-type NODs from diabetes. Protected animals developed strong, TH1biased peri-insulitis with pSC death, but lack of progression to islet invasion and β-cell kill. Because some S100β is expressed in β-cells, the protection following S100β immunization is not unexpected, as the presence of autoantigen in target tissue is re203

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

ARTICLES quired and may be sufficient for effective immunotherapy19. The effect of GFAP immunotherapy is less easily explained, as GFAP expression was below RT-PCR detection limits in the NIT-1 β-cell line and negative by histochemistry in pancreatic islets. In a new transgenic model, we observed selective pSC ablation in 6-week-old NOD males carrying a herpesvirus thymidine kinase transgene under GFAP promoter control (F.A. et al., unpublished data). The absence of detectable β-cell damage from these and comparable40,41 mice suggests that if the GFAP promoter is active in β-cells, it would be at very low levels. GFAP immunotherapy effects were reminiscent of the similar behavior of GFAP-specific T-cell lines in NOD/SCID mice, with pathogenic peri-insulitis but not T1D. Immunotherapymediated disease protection may be related to the essential absence of CD8+ T cells from peri-insulitis lesions. We were initially surprised by the TH1 bias, but our immunotherapy protocol used low GFAP doses (5 µg) and the ensuing T-cell responses in protected mice were of high affinity, which can contribute to TH1 preference42. Protective immunotherapy can depend on IFN-γ (ref. 29), and our preliminary observation that GFAP did not protect IFN-γ–deficient NOD congenics is consistent with this previous report. While mechanisms require further studies, these observations imply that modification of T-cell pools that recognize pSC constituents can modify the course of prediabetes, again associating pSC-autoimmunity with prediabetes progression. Our findings pose several questions. For example, we do not know the function of the pSC mantle or of the non-myelinated, nociceptor-like axons associated with it17,43. NOD mice with pSC genetically modified to resist or invite death should eventually allow further distinction between pSC and β-cell autoimmunity. The presence of pSC autoantibody and T-cell autoreactivities in humans with probable prediabetes may imply that pSC autoimmune targeting is a common feature of spontaneous autoimmune diabetes. Prospective studies of diabetes kindreds are under way to measure the disease predictive power, and thus the possible linkage of pSC autoimmunity and progression to overt T1D in humans. Methods Mice. NOD, NOD/SCID, C57BL/6, SJL/J and BALB/c mice were purchased (The Jackson Laboratory, Bar Harbor, Maine) and maintained under approved protocols in our conventional vivarium (female diabetes incidence: 85%). NY8.3 TCR-transgenic NOD mice28 and LCMV-gp/TCRgp doubletransgenic C57BL/6 mice30 were maintained as described. In adoptive transfer experiments, splenocytes from 4–6 diabetic NOD females were pooled and 107 cells were injected intravenously (i.v.) in 100 µl of PBS into irradiated (650 rad) 6- to 8-week-old male NOD recipients. Forty-eight hours after transfer, GFAP (5 µg), S100β (50 µg), OVA (50 µg) or PBS were emulsified in IFA and injected i.p. Glucosuria was used to screen for diabetes (TesTape, Lily, Toronto, Canada). Diabetes was confirmed by blood glucose measurements on 2 consecutive days (>13.8 mM/l; SureStep, Life Technologies Inc., Burnaby, British Columbia, Canada). T-cell studies. Splenocytes from 3- to 24-week-old NOD females were cultured (4 × 105 cells/well) in AIM V serum-free medium (Life Technologies, Burnaby, Canada) containing antigen (.002–50 µg/ml). After 72 h, cultures were pulsed (1 µCi, [3H]thymidine/18 h) and counted by liquid scintillation. Supernatants were collected after 72 h of culture for cytokine measurements by ELISA (BD Pharmingen, Mississauga, Ontario, Canada). Lymph-node assays were similar but used 2 × 105 lymph node cells plus 2 × 105 irradiated (1,100 rad) syngeneic splenocytes/well. T-cell lines were generated through several rounds of antigen stimulation in the presence of 10 U/ml IL-2 and irradiated splenic APC, separated by 7–10 d of rest. Cell lines responsive to GFAP, but not other test antigens, were injected (107, i.v.) 204

into 8-week-old female NOD/SCID mice. PBMC from human subjects were purified on Ficoll-Hypaque gradients and cultured for 1 week in protein-free Hybrimax 2897 medium (Sigma, St. Louis, Missouri) with 10 U of human IL-15 and 0.01–10 µg test antigen. To normalize pooled results from human and mouse experiments, some data are presented as stimulation index (SI, cpm antigen stimulated/medium control). Positive human responses were defined as previously reported (SItestantigen > 3 s.d. above SIOvalbumin)25. Immunofluorescence and histology. Frozen murine pancreas sections were fixed in 1% paraformaldehyde, blocked with 5% normal donkey serum (The Jackson Laboratory), and stained with polyclonal rabbit antibodies against GFAP (Signet Pathology Systems, Dedham, Massachusetts), goat antibody against CD3 (1:50; Santa Cruz Biotechnology Inc., Santa Cruz, California), and/or guinea-pig antibody against insulin (DAKO, Carpinteria, California). Bound antibodies were detected with biotinylated donkey anti-guinea pig IgG (1:200, Jackson), Streptavidin AlexaFluor 546 or 633 (1:300, Molecular Probes, Eugene, Oregon), and FITC conjugated donkey anti-rabbit IgG (1:25, Jackson). When the biotin-streptavidin system was used, sections were also blocked with an avidin/biotin blocking kit (Vector, Burlingame, California). ICA staining was performed on snap-frozen sections of 4-week-old NOD female pancreas with an overnight incubation of human test sera (1:160) at room temperature and detection with anti-human AlexaFluor 546 (1:400, Molecular Probes). Purified human islets were provided by the islet core of the Pittsburgh JDRF Diabetes center. To score insulitis severity, pancreata were fixed in 10% buffered formalin for a minimum of 24 h. Histological sections were stained with hematoxylin and eosin and two blinded observers scored at the following scale: 0, normal islet; 1, peri-insulitis or infiltration of 50% of the islet surface area. Antibody detection. Surface-enhanced laser desorption/ionization (SELDI) time-of-flight mass spectrometry used the Ciphergen instrument (Ciphergen Biosystems, Fremont, California)23. GFAP was covalently coupled to the chip array surface, followed by saturating free binding sites according to the manufacturer’s protocols44. Replicate dilutions of test sera were incubated on array spots (1 µl, 1:1–1:20, 15 min/20 °C), washed (1–2 min, PBS-Tween 20, then double distilled water to remove free ions) and read in the instrument. Each array spot was laser-sampled about 100 times to generate signal averages. Autoantibodies against GFAP were detected at 150 ± 3 kDa and the average peak height was noted. IgG anti-GFAP antibody (Signet) provided a positive control for calibration. Reagents. OVA (Sigma), baculovirus-derived human GAD65 (Diamyd Diagnostics, Stockholm, Sweden), human GFAP (> 90% pure) and bovine S100β (>98% pure; Calbiochem, San Diego, California) were purchased. Standard template-calibrated RT-PCR of brain and the NIT β-cell line used the following primers: GFAP, 5′-CCATGCCACGTTTCTCCTTG-3′ and 5′-CTGCAGTTGGCGGCGATAG-3′; S100β, 5′-TACTCGGACACTGAAGCCAGAG3′ and 5′-GTCTCACTCATGTTCAAAGAAC-3′; GUS, 5′-GTGATGTGGTCTG TGGCCAA-3′ and 5′−TCTGCTCCATACTCGCTCTG-3′; ICA69, 5′-GGCTGTGGCACCAGAGCCGAGG-3′ and 5′-TCATGCATTGAGCAATTCGTG-3′. Statistics. Numeric values were compared by Mann-Whitney tests, incidence data were analyzed by life tables and endpoints were compared by Tables. Significance was set at 5% and all tests were two-tailed. Acknowledgments We thank M. Trucco and R. Boti for provision of purified human islets from the Pittsburgh JDRF Diabetes Center; D. Homeard for transmission electron microscopy; D. Midha, C. McKerlie and D. Winer for helpful discussions; and A. Darnley and K. Reilly for collection of human blood samples and ICA assays. This work was supported by grants from the Canadian Institutes of Health Research, the Juvenile Diabetes Research Foundation, the National Institutes of Health and the Renziehausen Fund. Competing interests statement The authors declare competing financial interests: see the website (http://www.nature.com/naturemedicine) for details. NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturemedicine

RECEIVED 7 AUGUST; ACCEPTED 19 DECEMBER 2002 1. Serreze, D.V. & Leiter, E.H. Genetic and pathogenic basis of autoimmune diabetes in NOD mice. Curr. Opin. Immunol. 6, 900–906 (1994). 2. Wicker, L.S., Todd, J.A. & Peterson, L.B. Genetic control of autoimmune diabetes in the NOD mouse. Annu. Rev. Immunol. 13, 179–200 (1995). 3. Karges, W., Ilonen, J., Robinson, B.H. & Dosch, H.-M. Self and non-self antigen in diabetic autoimmunity: Molecules and mechanisms. Mol. Aspects Med. 16, 79–213 (1995). 4. Shimada, A., Charlton, B., Taylor, E.C. & Fathman, C.G. β-cell destruction may be a late consequence of the autoimmune process in nonobese diabetic mice. Diabetes 45, 1063–1067 (1996). 5. Atkinson, M.A. & Leiter, E.H. The NOD mouse model of type 1 diabetes: As good as it gets? Nature Med. 5, 601–604 (1999). 6. Tisch, R., Wang, B. & Serreze, D.V. Induction of glutamic acid decarboxylase 65specific Th2 cells and suppression of autoimmune diabetes at late stages of disease is epitope dependent. J. Immunol. 163, 1178–1187 (1999). 7. Winer, S. et al. Peptide dose, MHC-affinity and target self-antigen expression are critical for effective immunotherapy of NOD mouse prediabetes. J. Immunol. 165, 4086–4094 (2000). 8. Andre, I. et al. Checkpoints in the progression of autoimmune disease: lessons from diabetes models. Proc. Natl. Acad. Sci. USA 93, 2260–2263 (1996). 9. Kim, J. et al. Differential expression of GAD65 and GAD67 in human, rat, and mouse pancreatic islets. Diabetes 42, 1799–1808 (1993). 10. Lu, J. et al. Identification of a second transmembrane protein tyrosine phosphatase, IA-2β, as an autoantigen in insulin-dependent diabetes mellitus: Precursor of the 37-kDa tryptic fragment. Proc. Natl. Acad. Sci. USA 93, 2307–2311 (1996). 11. Karges, W., Pietropaolo, M., Ackerley, C. & Dosch, H.M. Gene expression of islet cell antigen p69 (ICAp69) in man, mouse and rat. Diabetes 45, 513–521 (1996). 12. Marrosu, M.G. et al. Patients with multiple sclerosis and risk of type 1 diabetes mellitus in Sardinia, Italy: A cohort study. Lancet 359, 1461–1465 (2002). 13. Winer, S. et al. Type I diabetes and MS patients target islet plus CNS-autoantigens, non-immunized NOD mice can develop autoimmune encephalitis. J. Immunol. 166, 2831–2841 (2001). 14. Salomon, B. et al. Development of spontaneous autoimmune peripheral polyneuropathy in b7-2-deficient nod mice. J. Exp. Med. 194, 677–684 (2001). 15. Donev, S.R. Ultrastructural evidence for the presence of a glial sheath investing the islets of Langerhans in the pancreas of mammals. Cell Tissue Res. 237, 343–348 (1984). 16. Smith, P.H. Structural modification of Schwann cells in the pancreatic islets of the dog. Am. J. Anat. 144, 513–517 (1975). 17. Sunami, E. et al. Morphological characteristics of Schwann cells in the islets of Langerhans of the murine pancreas. Arch. Histol. Cytol. 64, 191–201 (2001). 18. Teitelman, G., Guz, Y., Ivkovic, S. & Ehrlich, M. Islet injury induces neurotrophin expression in pancreatic cells and reactive gliosis of peri-islet Schwann cells. J. Neurobiol. 34, 304–318 (1998). 19. Winer, S. et al. ICA69null NOD mice develop diabetes but resist disease acceleration by cyclophosphamide. J. Immunol. 168, 475–482 (2002). 20. Saravia-Fernandez, F. et al. Localization of γ-aminobutyric acid and glutamic acid decarboxylase in the pancreas of the nonobese diabetic mouse. Endocrinology 137, 3497–3506 (1996). 21. Tisch, R. et al. Immune response to glutamic acid decarboxylase correlates with insulitis in non-obese diabetic mice. Nature 366, 72–75 (1993). 22. Kaufman, D.L. et al. Spontaneous loss of T cell tolerance to glutamic acid decarboxylase in murine insulin dependent diabetes. Nature 366, 69–72 (1993). 23. Merchant, M. & Weinberger, S.R. Recent advancements in surface-enhanced laser

NATURE MEDICINE • VOLUME 9 • NUMBER 2 • FEBRUARY 2003

desorption/ionization-time of flight-mass spectrometry. Electrophoresis 21, 1164–1177 (2000). 24. Lipton, R.B. et al. Autoimmunity and genetics contribute to the risk of insulin-dependent diabetes mellitus in families: Islet cell antibodies and HLA DQ heterodimers. Am. J. Epidemiol. 136, 503–512 (1992). 25. Dosch, H.-M., Cheung, R.K., Karges, W., Pietropaolo, M. & Becker, D.J. Persistent T cell anergy in human type 1 diabetes. J. Immunol. 163, 6933–6940 (1999). 26. Wicker, L.S., Miller, B.J. & Mullen, Y. Transfer of autoimmune diabetes mellitus with splenocytes from nonobese diabetic (NOD) mice. Diabetes 35, 855–860 (1986). 27. DiLorenzo, T.P. et al. Major histocompatibility complex class I-restricted T cells are required for all but the end stages of diabetes development in nonobese diabetic mice and use a prevalent T cell receptor α chain gene rearrangement. Proc. Natl. Acad. Sci. USA 95, 12538–12543 (1998). 28. Amrani, A. et al. Progression of autoimmune diabetes driven by avidity maturation of a T-cell population. Nature 406, 739–742 (2000). 29. Serreze, D.V. et al. Th1 to Th2 cytokine shifts in nonobese diabetic mice: sometimes an outcome, rather than the cause, of diabetes resistance elicited by immunostimulation. J. Immunol. 166, 1352–1359 (2001). 30. Ohashi, P.S. et al. Ablation of “tolerance” and induction of diabetes by virus infection in viral antigen transgenic mice. Cell 65, 305–317 (1991). 31. Pircher, H., Burki, K., Lang, R., Hengartner, H. & Zinkernagel, R.M. Tolerance induction in double specific T-cell receptor transgenic mice varies with antigen. Nature 342, 559–561 (1989). 32. Verdaguer, J. et al. Spontaneous autoimmune diabetes in monoclonal T cell nonobese diabetic mice. J. Exp. Med. 186, 1663–1676 (1997). 33. Santamaria, P. et al. β-cell-cytotoxic CD8+ T cells from nonobese diabetic mice use highly homologous T cell receptor α-chain CDR3 sequences. J. Immunol. 154, 2494–2503 (1995). 34. Verdaguer, J. et al. Acceleration of spontaneous diabetes in TCR-β-transgenic nonobese diabetic mice by β-cell cytotoxic CD8+ T cells expressing identical endogenous TCR-α chains. J. Immunol. 157, 4726–4735 (1996). 35. Spierings, E. et al. Mycobacterium leprae-specific, HLA class II-restricted killing of human Schwann cells by CD4(+) Th1 cells: a novel immunopathogenic mechanism of nerve damage in leprosy. J. Immunol. 166, 5883–5888 (2001). 36. Wekerle, H., Schwab, M., Linington, C. & Meyermann, R. Antigen presentation in the peripheral nervous system: Schwann cells present endogenous myelin autoantigens to lymphocytes. Eur. J. Immunol. 16, 1551–1557 (1986). 37. Wong, F.S. et al. Identification of an MHC class I-restricted autoantigen in type 1 diabetes by screening an organ-specific cDNA library. Nature Med. 5, 1026–1031 (1999). 38. Brooks-Worrell, B., Gersuk, V.H., Greenbaum, C. & Palmer, J.P. Intermolecular antigen spreading during the preclinical period of human type I diabetes. J. Immunol. 166, 5265–5370 (2001). 39. Mathis, D., Vence, L. & Benoist, C. β-Cell death during progression to diabetes. Nature 414, 792–798 (2001). 40. Delaney, C.L., Brenner, M. & Messing, A. Conditional ablation of cerebellar astrocytes in postnatal transgenic mice. J. Neurosci. 16, 6908–6918 (1996). 41. Bush, T.G. et al. Fulminant jejuno-ileitis following ablation of enteric glia in adult transgenic mice. Cell 93, 189–201 (1998). 42. Leitenberg, D. & Bottomly, K. Regulation of naive T cell differentiation by varying the potency of TCR signal transduction. Semin. Immunol. 11, 283–292 (1999). 43. Hahn von Dorsche, H., Schafer, H. & Titlbach, M. Histophysiology of the obesitydiabetes syndrome in sand rats. Adv. Anat. Embryol. Cell Biol. 130, 1–95 (1994). 44. Austen, B.M., Frears, E.R. & Davies, H. The use of SELDI proteinchip arrays to monitor production of Alzheimer’s βamyloid in transfected cells. J. Pept. Sci. 6, 459–469 (2000).

205