1999 Nature America Inc. http://medicine.nature.

com

ARTICLES

Blocking both signal 1 and signal 2 of T-cell activation prevents apoptosis of alloreactive T cells and induction of peripheral allograft tolerance
YONGSHENG LI1, XIAN CHANG LI1, XIN XIAO ZHENG1, ANDREW D. WELLS2, LAURENCE A. TURKA2 & TERRY B. STROM1
1

Department of Medicine, Harvard Medical School, Division of Immunology, Beth Israel Deaconess Medical Center, Boston, Massachusetts 02215, USA 2 Department of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA X.C.L. and Y.L. contributed equally to this study. Correspondence should be addressed to T.B.S.; email: tstrom@caregroup.harvard.edu

The alloimmune response against fully MHC-mismatched allografts, compared with immune responses to nominal antigens, entails an unusually large clonal size of alloreactive T cells1. Thus, induction of peripheral allograft tolerance established in the absence of immune system ablation and reconstitution is a challenging task in transplantation. Here, we determined whether a reduction in the mass of alloreactive T cells due to apoptosis is an essential initial step for induction of stable allograft tolerance with non-lymphoablative therapy. Blocking both CD28B7 and CD40CD40 ligand interactions (co-stimulation blockade) inhibited proliferation of alloreactive T cells in vivo while allowing cell cycle-dependent T-cell apoptosis of proliferating T cells, with permanent engraftment of cardiac allografts but not skin allografts. Treatment with rapamycin plus co-stimulation blockade resulted in massive apoptosis of alloreactive T cells and produced stable skin allograft tolerance, a very stringent test of allograft tolerance. In contrast, treatment with cyclosporine A and co-stimulation blockade abolished Tcell proliferation and apoptosis, as well as the induction of stable allograft tolerance. Our data indicate that induction of T-cell apoptosis and peripheral allograft tolerance is prevented by blocking both signal 1 and signal 2 of T-cell activation. Bcl-XL transgenic mice are resistant to induction of cardiac allograft tolerance by co-stimulation blockade. The failure of tolerance induction in Bcl-XL transgenic mice is linked to the resistance of alloreactive T cells to apoptotic cell death after deprivation of T-cell growth factors (A.D.W. and X.C.L.). Moreover, IL-2 knockout mice with profound defects in activation-induced T-cell apoptosis are also resistant to the induction of tolerance to islet and cardiac allografts by co-stimulation blockade2 or by rapamycin3. Rapamycin, in contrast to the calcineurin inhibitor cyclosporin A (CsA), blocks growth factor-imparted proliferative signals4 but does not block antigen priming for activation-induced cell death5 (AICD). Thus, treatment with rapamycin produces islet and cardiac allograft tolerance in wildtype control mice, but it fails to do so in IL-2 knockout recipients3. The failure to induce allograft tolerance in both Bcl-XL transgenic and IL-2 knockout mice may be due to a defect of activated alloreactive T cells in undergoing apoptosis. Here, we determined whether a reduction in the clonal size of alloreactive T cells through T-cell apoptosis, a process essential in central tolerance6, is a prerequisite for induction of peripheral
1298

allograft tolerance using non-lymphoablative therapy. We transplanted cardiac allografts from BALB/c mice (H-2d) into C3H/He (H-2k) recipients and treated them with cytotoxic T-lymphocyte antigen-4 immunoglobulin (CTLA4Ig) plus the antibody against CD40 ligand (CD40L) MR1, with or without treatment with rapamycin or CsA. In keeping with previous reports79, treatment with CTLA4Ig and antibody against CD40L induced long-term cardiac allograft survival with a mean survival time of more than 120 days (n = 7), whereas untreated control mice rejected cardiac allografts within 10 days (n = 6). Addition of CsA to the co-stimulation blockade protocol resulted in rejection of BALB/c cardiac allografts with a mean survival time of 30 days (n = 6). The ability of calcineurin inhibitors to block co-stimulation blockade-induced tolerance has been noted in this mouse model8. A similar observation has been made in a monkey renal transplant model10. In contrast, rapamycin fully preserved the tolerogenic effect of co-stimulation blockade here. All recipients treated with rapamycin and co-stimulation blockade experienced permanent engraftment (mean survival time, more than 120 days; n = 5) (Table 1). Rapamycin as a monotherapy was also very effective in this model, and all cardiac allografts survived for >120 days (n = 5). The combination of rapamycin and CsA was also very detrimental to long-term cardiac allograft survival (mean survival time, 33 days; n = 8). Thus, CsA antagonized the tolerizing effects of two non-lymphoablative protocols (rapamycin and costimulation blockade) to produce allograft tolerance. Rapamycin, in contrast to CsA, is very compatible with costimulation blockade in tolerance induction. To determine whether rapamycin may synergize with co-stimulation blockade to produce stable allograft tolerance, we used the stringent skin allograft model. We transplanted full-thickness tail skin grafts from BALB/c (H-2d) mice onto the thoracic walls of C3H/He (H-2k) recipient mice. Untreated mice rejected the skin allografts with a mean survival time of 9 days (n = 5)(Table 1). In contrast to the cardiac allograft model, all recipient mice treated with co-stimulation blockade (mean survival time, 15 days; n = 6) or rapamycin (mean survival time, 34 days; n = 5) or CsA (mean survival time, 15 days; n = 6) rejected BALB/c skin allografts. Treatment with combined co-stimulation blockade and CsA resulted in an uniform rejection of all skin allografts (mean survival time, 22 days; n = 6). In contrast, all C3H/He recipients treated with combined co-stimulation blockade and rapamycin experienced permanent engraftment of BALB/c skin
NATURE MEDICINE VOLUME 5 NUMBER 11 NOVEMBER 1999

1999 Nature America Inc. http://medicine.nature.com

 1999 Nature America Inc. http://medicine.nature.com

ARTICLES
priming for AICD and may encourage other apoptotic events through inhibition of Bcl-2/Bcl-XL expression15. Rapamycin and CsA might produce different effects on T-cell activation and T-cell apoptosis, thereby accounting for the considerable difference in tolerance induction by co-stimulation blockade. To quantitatively analyze the proliferation and apoptosis of alloreactive T cells in vivo, we labeled splenic lymphocytes from C3H/He mice with CFSE, a fluorochrome whose per cell fluorescent intensity halves with each round of cell proliferation16. We injected the dye-labeled C3H/He lymphocytes into BALB/c mice irradiated with a dose sufficient to ablate their immune systems. At 25 days after adoptive cell transfer, we recovered cells from the hosts and stained them with Cy-chrome-conjugated antibody against CD4 or antibody against CD8 and PEconjugated annexin V. Thus, proliferation of alloreactive T cells in vivo can be monitored precisely through analysis of their CFSE profiles, and apoptotic cell death in each round of dividing cells can be quantitated (Fig. 1). Day 3 after passive cell transfer of naive lymphocytes represents the time at which maximum waves of proliferating T cells can be clearly identified. In untreated control mice, about 21% of CFSE-labeled CD4+ T cells proliferated in the host spleen by this time, and we detected at least seven discrete generations of dividing cells (Fig. 2a). As shown by annexin V staining, T cells became increasingly susceptible to apoptosis in vivo with each cycle of cell division (Fig. 2b and c). Treatment with co-stimulation blockade substantially inhibited proliferation of T cells in vivo in response to alloantigen; only 9.4% of CFSE-labeled CD4+ T cells proliferated in vivo. However, after T cells in hosts treated with co-stimulation blockade proliferated, they divided for many generations (seven to eight times). The proportion of apoptotic cells present in each generation of dividing cells was increased in recipient mice treated with costimulation blockade. About 16% of undivided cells were annexin V-positive; the annexin V-positive T cells increased to 32% after three cell divisions, and to 57% after six cell divisions. Overall, co-stimulation blockade reduced the frequency of proliferating alloreactive T cells, while increasing the susceptibility of proliferating T cells to apoptosis. Compared with treatment with co-stimulation blockade alone, treatment with combined rapamycin and co-stimulation blockade did not lead to further inhibition of cellular proliferation. With either protocol, about 9% of CFSE-labeled CD4+ T cells proliferated in vivo. However, the combination of rapamycin and co-stimulation blockade resulted in a considerable increase in apoptotic cell death in every generation of dividing cells, especially in the first four cycles of cell division. All the dividing cells showed the phenotype of activated cells, as they expressed high levels of CD44 and low levels of CD45RB and CD62L (data not shown). Thus, rapamycin may be important in promoting apoptosis of activated T cells. Perhaps rapamycin plus co-stimulation blockade blocks expression of anti-apoptotic or survival molecules such as Bcl-2/BclXL in cycling T cells15,17 while at least permitting IL-2-mediated priming for AICD (ref. 3). In contrast, the combination of CsA and co-stimulation blockade completely blocked T-cell proliferation, thereby precluding the induction of cell cycle dependent T-cell apoptosis (Fig. 2). These non-dividing T cells had a naive phenotype, as they expressed high levels of CD45RB and CD62L, low levels of CD44, and lacked CD25. CFSE-labeled CD8+ T cells recovered from the host spleen had patterns of proliferation and apoptotic cell death (data not shown) similar to those of CD4+ T cells in mice treated with co-stimulation blockade and rapamycin or CsA.
1299

1999 Nature America Inc. http://medicine.nature.com

Fig. 1 Quantitative analysis of proliferation and apoptosis of alloreactive T cells in vivo. CFSE-labeled splenic lymphocytes from C3H/He mice were injected into irradiated BALB/c hosts. Splenic lymphocytes from the BALB/c hosts were collected 3 d later and stained with Cy-chrome-conjugated antibody against CD4 and PEannexin V. Proliferation and apoptosis of CFSElabeled CD4+ T cells in each generation (numbers beside graphs) of dividing cells were analyzed by flow cytometry. Data are representative of five experiments.

allografts (mean survival time, more than 120 days; n = 5). C3H/He recipient mice with long-term BALB/c skin allografts accepted second skin allografts from BALB/c mice (mean survival time, more than 40 days; n = 2) but promptly rejected third party C57BL/6 skin allografts (mean survival time, 17 days; n = 2). The protocol of rapamycin and co-stimulation blockade is indeed synergistic in tolerance induction. This tolerizing protocol is unique in that stable skin allograft tolerance across full MHC barriers has been achieved so far only with aggressive lymphoablative therapies and/or by creation of bone marrow chimerism11. Rapamycin and CsA have diverging effects on the induction of transplantation tolerance by co-stimulation blockade. In certain models, CsA, which inhibits calcineurin activation triggered by T-cell receptors12, blocks induction of T-cell anergy and T-cell apoptosis13,14. In contrast, rapamycin does not block
NATURE MEDICINE VOLUME 5 NUMBER 11 NOVEMBER 1999

 1999 Nature America Inc. http://medicine.nature.com

ARTICLES

1999 Nature America Inc. http://medicine.nature.com

Fig. 2 Proliferation frequency and apoptosis of alloreactive CD4+ T cells in vivo. CFSE-labeled C3H/He lymphocytes were injected into irradiated BALB/c hosts; mice were then treated with MR1 and CTLA4Ig for 3 d. Groups of mice were also treated with rapamycin or CsA for 3 d. Syngeneic controls, CFSE-

labeled BALB/c lymphocytes injected into irradiated BALB/c mice. a, Proliferation frequency (percent, right column) of CFSE-labeled CD4+ T cells. b, Dot plot of apoptosis in each generation of dividing cells. c, Apoptosis curve in each generation of dividing cells. Data are representative of four experiments.

T-cell apoptosis is an integral component of alloantigendriven cell responses, and this response can be profoundly modulated by co-stimulation blockade plus rapamycin or CsA. In this setting, apoptosis is cell cycle-dependent and linked to tolerance induction. Blockade of both CD28- and CD40L-costimulatory pathways reduced the proliferation frequency of alloreactive cells, and a substantial proportion of these proliferating cells are committed to apoptotic cell death. Costimulation blockade does not completely block cell proliferation, and a small population of cells that proliferate can divide for multiple generations. This may explain the basis for the acceptance of cardiac allografts in such recipients who nonetheless can reject more immunogenic skin allografts (Table 1). The combination of co-stimulation blockade and CsA totally abolished the proliferation capacity and apoptosis of alloreactive T cells. In global immunosuppression (blocking both signal 1 and signal 2), apoptosis and induction of allograft tolerance are precluded. The alloreactive T cells can probably mount a vigorous rejection response after the immunosuppression is stopped. The proliferation frequency of alloreactive T cells is similar in hosts treated with co-stimulation blockade and combined co-stimulation blockade and rapamycin. However, the addition of rapamycin, in contrast to CsA, considerably increased the cell cycle-dependent apoptosis of alloreactive T cells, permitting the induction of stable skin allograft tolerance.
1300

Apoptosis of alloreactive T cells seems essential for the induction of stable peripheral allograft tolerance with regimens that are not lymphoablative. Global immunosuppression completely blocks T-cell activation and apoptosis, thereby precluding the induction of true allograft tolerance. Depletion of certain cytopathic T cells in the inductive phase of tolerance to MHC-mismatched allografts is probably necessary to reduce the mass of alloreactive clones. A reduced mass of alloreactive T cells is likely to be permissive to achieve tolerance. Regulated AICD may also encourage the development or selective survival of immunoregulatory cells that may prove essential in maintaining the tolerant status over time18. For example, the oligoclonal Tcell response to minor histocompatibility antigens is far more sensitive to immune deviation from T helper-cell type 1 to T helper-cell type 2 than the polyclonal response to MHC antigens is19. It has troubled the clinical community that tolerizing effects of co-stimulation blockade are blocked by calcineurin inhibitors or corticosteroids10. Thus, our data have important clinical implications; an essential component of non-lymphoablative therapy in organ transplantation should avoid global immunosuppression and should selectively promote apoptotic cell death of activated T cells. Methods
Animals. BALB/c (H-2d) and C3H/He (H-2k) mice, 810 weeks old, were obtained from the Jackson Laboratory (Bar Harbor, Maine).
NATURE MEDICINE VOLUME 5 NUMBER 11 NOVEMBER 1999

 1999 Nature America Inc. http://medicine.nature.com

ARTICLES
Table 1
Donor BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d) BALB/c (H-2d)
RPM, rapamycin.

Survival of BALB/c cardiac and skin allografts in C3H/He recipients treated with co-stimulation blockade, rapamycin and CsA.
Treatment Untreated CTLA4Ig+MR1 CsA RPM CTLA4Ig+MR1+RPM CTLA4Ig+MR1+CsA CsA+RPM Untreated CTLA4Ig+MR1 CsA RPM CTLA4Ig+MR1+RPM CTLA4Ig+MR1+CsA Graft survival (days) 8, 9, 10, 10, 10, 11 >120 x 7 13, 19, 19, 20, 27, 27 >120 x 5 >120 x 5 25, 26, 29, 30, 35, 41 28, 31, 31, 33, 33, 34, >120, >120 8, 8, 9, 10, 10 8, 13, 13, 17, 18, 19 14, 14, 15, 15, 15, 18 7, 19, 34, 103, 110 >120 x 5 19, 20, 22, 22, 23, 24 Mean survival time (days) 10 >120 20 >120 >120 30 33 9 15 15 34 >120 22

Recipient C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k) C3H/He (H-2k)

Graft Cardiac Cardiac Cardiac Cardiac Cardiac Cardiac Cardiac Skin Skin Skin Skin Skin Skin

1999 Nature America Inc. http://medicine.nature.com

Reagents and tolerizing protocols. Rapamycin was provided by S. Sehgal (Wyeth-Ayerst, Princeton, New Jersey). Cyclosporine A (Sandimmune) was obtained from the Beth Israel Deaconess Medical Center pharmacy. Murine CTLA4Ig was constructed and expressed in our laboratory as described20. A B-cell hybridoma producing a hamster monoclonal antibody against mouse CD40L (MR1, IgG2a) was obtained from American Type Culture Collection (ATCC, Rockville, Maryland). The hybridoma cells were grown in serum-free UltraCulture medium (BioWhittaker, Walkersville, Maryland). The MR1 monoclonal antibody was purified using protein G columns. Treatment of cardiac allograft recipients with rapamycin consisted of 0.2 mg/kg per day intraperitoneally for the first 3 days after transplantation, followed by 0.2 mg/kg every other day for 14 days. Skin allograft recipients received 3 mg/kg rapamycin per day for 14 days. Treatment with CsA consisted of 20 mg/kg per day subcutaneously for 14 days. Costimulation blockade treatment consisted of 0.2 mg CTLA4Ig intraperitoneally on days 0, 2, 4 and 6 after transplant and 0.25 mg monoclonal antibody against CD40L (MR1) intraperitoneally on days 0, 2 and 4 after transplant. Cardiac transplantation. Cardiac grafts from BALB/c donors were collected by dividing and excising the aorta and the pulmonary artery. The cardiac grafts were then transplanted into C3H/He recipients by suturing donor aorta and donor pulmonary artery end-to-side to the recipients abdominal aorta and vena cava21, respectively. Graft function was monitored every other day by trans-abdominal palpation and scored on a scale of 14 based on the strength and the rate of impulses. Rejection was defined as a complete cessation of palpable beat and was confirmed by direct visualization after laparotomy. Skin transplantation. Full-thickness tail skin was collected from donor BALB/c mice. The skin grafts (1 1 cm in size) were grafted onto the thoracic walls of recipient C3H/He mice. The skin graft survival was monitored daily and rejection was defined as a complete necrosis of the skin grafts. CFSE labeling and in vivo quantitation. A single-cell suspension of spleens and lymph nodes from C3H/He mice was prepared in Hanks balanced salt solution (HBSS). Red blood cells were lysed by hypotonic shock. Lymphocytes were washed and resuspended in HBSS at a concentration of 1 107 cells/ml for labeling with a tracking fluorochrome 5-carboxyfluorescein diacetate succinimidyl ester (CFSE; Molecular Probes, Eugene, Oregon) as described16. Cells were incubated with CFSE at a final concentration of 5 M in HBSS for 6 min. The labeling was then terminated by the addition of FCS (10% of the total volume). Cells were washed twice in complete RPMI-1640 medium and resuspended in HBSS for intravenous injection. BALB/c mice were irradiated with 1,000 rad, a dose sufficient to ablate their immune systems, with a GammaCell irradiator (Ontario, Canada).
NATURE MEDICINE VOLUME 5 NUMBER 11 NOVEMBER 1999

Each mouse then received 4 107 to 6 107 CFSE-labeled cells through the penile vein. Mice were killed 25 d later, spleens and peripheral lymph nodes were collected from host mice, and single-cell suspensions were prepared. Cells were stained on ice for 20 min with a biotinylated antibody against mouse CD4 (GK1.5) or mouse CD8 (53-6.7) (PharMingen, San Diego, California), then stained on ice for 15 min with streptavidinCy-Chrome and PE-conjugated annexin V (PharMingen, San Diego, California). Cells were washed twice in annexin V labeling buffer. The proliferation and apoptotic cell death of CFSE-labeled CD4+ or CD8+ T cells in each distinct generation of dividing cells were analyzed by flow cytometry. Data were collected and analyzed by gating onto CD4+CFSE+ cells or CD8+CFSE+ cells. Calculation of frequency of proliferating T cells in vivo. The frequency of T cells proliferating in response to alloantigen in vivo was calculated as reported16,22. Distinct cycles of dividing cells were identified by their CFSE profiles. The absolute number of cells in each cycle was counted using the FACS acquisition software (CellQuestTM; Becton Dickinson, Mountain View, California (XCL). The number of precursors that proliferated and gave rise to the absolute number of daughter cells was extrapolated using the formula: y/2n (y =absolute number of cells in each cell cycle, n = number of cell divisions). For example, 16 daughter cells in the third cell division is the progeny of 2 precursors, each of which have divided 3 times (16/23=2). The frequency of proliferating T cells in the responder population was then calculated by dividing the total number of precursors by the sum of total CFSE-labeled cells collected. Acknowledgments Grant support for this work was provided by JDF international 1-1999-16 (X.C.L.) and 1-1999-317 (X.X.Z.), National Institutes of Health RO1 AI42298 (T.B.S.), and National Institutes of Health PO AI/GF 41521 (T.B.S. and L.A.T.).

RECEIVED 18 JUNE; ACCEPTED 14 SEPTEMBER 1999

1. Warrens, A.N., Lombardi, G. & Lechler, R.I. Presentation and recognition of major and minor histocompatibility antigens. Transpl. Immunol. 2, 103107 (1994). 2. Dai, Z., Konieczny, B.T., Baddoura, F.K. & Lakkis, F.G. Impaired alloantigen mediated T-cell apoptosis and failure to induce long term allograft survival in IL-2 deficient mice. J. Immunol. 161, 16591663 (1998). 3. Li, X.C., Zheng, X.X., Zand, M.S., Li, Y. & Strom, T.B. IL-2 dependent rapamycin resistent signals are required for tolerance induction and T-cell apoptosis. Transplantation 65, S169665 (1998). 4. Abraham, R.T. & Wiederrecht, G.J. Immunopharmacology of rapamycin. Annu. Rev. Immunol. 14, 483510 (1996). 5. Bierer, B.E. et al. Two distinct signal transmission pathways in T lymphocytes are inhibited by complexes formed between an immunophilin and either FK506 or rapamycin. Proc. Natl. Acad. Sci. USA 87, 92319235 (1990). 6. Ikuta, K., Uchida, N., Friedman, J. & Weissman, I.L. Lymphocyte development from stem cells. Annu. Rev. Immunol. 10, 759783 (1992). 7. Lenschow, D.J. et al. Long term survival of xenogeneic pancreatic islet grafts induced by CTLA-4Ig. Science 257, 789792 (1992).

1301

 1999 Nature America Inc. http://medicine.nature.com

ARTICLES
8. Larsen, C.P. et al. Long term acceptance of skin and cardiac allografts after blocking CD40 and CD28 pathways. Nature 381, 434438 (1996). 9. Kirk, A.D. et al. CTLA-4Ig and anti-CD40L prevent renal allograft rejection in primates. Proc. Natl. Acad. Sci. USA 94, 87898794 (1997). 10. Kirk, A.D. et al. Treatment with humanized monoclonal antibody against CD154 prevents acute renal allograft rejection in nonhuman primates. Nature Med. 5, 686693 (1999). 11. Sykes, M. Chimerism and central tolerance. Curr. Opin. Immunol. 8, 694703 (1996). 12. Schreiber, S.L. & Crabtree, G.R. The mechanism of action of cyclosporine A and FK 506. Immunol. Today 13, 136142 (1992). 13. Shi, Y.F., Sahai, B.M. & Green, D.R. Cyclosporin A inhibits activation-induced cell death in T-cell hybridomas and thymocytes. Science 339, 625626 (1989). 14. Yazdanbakhsh, K., Choi, J. W., Li, Y., Lau, L.F. & Choi, Y. Cyclosporin A blocks apoptosis by inhibiting the DNA binding activity of the transcriptional factor Nur77. Proc. Natl. Acad. Sci. USA 92, 437441 (1995). 15. Adachi, M. et al. Interleukin-2 upregulates BAG-1 gene expression through serine-rich region within IL-2 receptor beta chain. Blood 88, 41184123 (1996). 16. Wells, A. D., Gudmundsdottir, H. & Turka, L. A. Following the fate of individual T cells throughout activation and clonal expansion. J. Clin. Invest. 100, 31733183 (1997). 17. Boise, L. H. et al. CD28 co-stimulation can promote T-cell survival by enhancing the expression of Bcl-XL. Immunity 3, 8798 (1995). 18. Cobbold, S. & Waldmann, H. Infectious tolerance. Curr. Opin. Immunol. 10, 518524 (1998). 19. Li, X. C., Zand, M. S., Li, Y., Zheng, X. X. & Strom, T. B. On histocompatibility barriers, Th1 to Th2 immune deviation, and the nature of the allograft responses. J. Immunol. 161, 22412247 (1998). 20. Tivol, E.A. et al. CTLA-4Ig prevents lymphoproliferation and fatal multiorgan tissue destruction in CTLA-4 deficient mice. J. Immunol. 158, 50915094 (1997). 21. Larsen, C.P., Morris, P.J. & Austyn, J.M. Migration of dendritic leukocytes from cardiac allografts into host spleens. J. Exp. Med. 171, 307314 (1990). 22. Song, H. K. et al. Alloimmune responses against major and minor histocompatibility antigens: distinct division kinetics and requirement for CD28 co-stimulation. J. Immunol. 162, 24672471 (1999).

1999 Nature America Inc. http://medicine.nature.com

1302

NATURE MEDICINE VOLUME 5 NUMBER 11 NOVEMBER 1999