Nucleosome-Specific Regulatory T Cells Engineered by Triple Gene Transfer Suppress a Systemic Autoimmune Disease¹

Systemic autoimmune diseases have traditionally been treated using nonspecific immunosuppressive agents, but these agents often lead to opportunistic infections and an increased rate of malignancy. There remains the need to develop selective or specific therapies that target individual autoantigens. Several strategies have been developed as potential Ag-specific immunotherapies, such as using Ag-pulsed dendritic cells (DCs),³ but the majority of these approaches will require further investigation (1, 2). A more detailed understanding of autoimmune diseases, including autoantigen presentation, is required for the development of reasonable immunotherapies.

Systemic lupus erythematosus (SLE) is a life-threatening autoimmune disease characterized by the production of a variety of autoantibodies (3). Anti-DNA Abs are thought to be one of the major pathogenic products of the autoimmune response (4, 5, 6). Datta and colleagues (7, 8, 9), as well as other groups (10, 11), have noted that nucleosomes could be a major immunogen for pathogenic autoantibody-inducing T cells in lupus-prone mice. Datta and coworkers (7, 8, 9) showed that the majority of pathogenic T_H clones specific for nucleosomes were capable of rapidly inducing anti-DNA autoantibody production, and that these clones were also able to induce nephritis when injected into young lupus-prone mice. Moreover, anti-nucleosome ELISAs have demonstrated better sensitivity, specificity, and diagnostic confidence with regard to human SLE than anti-DNA ELISAs. Anti-nucleosome ELISAs are also correlated with disease activity, as determined by the SLE Disease Activity Index (12, 13).

Although evidences have accumulated demonstrating the importance of nucleosomes as major pathogenic autoantigens, the cellular mechanisms for the immunological recognition of nucleosomes are poorly understood. Generalized hyperresponsiveness of B cells has been reported in both mice and human lupus (14, 15). However, these nonspecific immune disorders cannot provide a sufficient model of nuclear autoantigen-specific autoimmunity encountered in patients with lupus.

To better understand the mechanisms of autoantigen recognition, we first reconstituted nucleosome specificity by TCR gene transfer in CD4⁺ T cells from (NZB × NZW)F₁ (NZB/W F₁) lupus model mice (3, 16). Using this model, we demonstrated an abnormal autoantigen presentation of splenic DCs. Among the lymphoid organs, this elevated autoantigen presentation of DCs was relatively localized in the spleen. We then developed a triple gene transfer system to generate autoantigen-specific regulatory cells. These regulatory cells preferentially accumulated in the spleen and suppressed the progression of the disease without obvious systemic immunosuppression.

Line 3A is a cell line from lupus-prone (SWR × NZB)F₁ (SNF1; I-A^d/q) that can recognize the immunodominant nucleosomal epitope (histone H4; aa 71–94) in the context of I-A^d (7, 8) and many other I-A molecules (8). Both TCR-α and -β cDNA fragments were synthesized using PCR based on the published sequences of line 3A (7, 8) and designated as AN3 TCR-α and -β. Vα13 and Vβ4 fragments identical to CDR1 and two sequences of line 3A were obtained from NZB splenic cDNA and an added CDR3 sequence by PCR. Jα41-Cα fragment and Jβ2.6-Cβ fragment were also obtained from NZB splenic cDNA and an added CDR3 sequence by PCR. Vα13-CDR3 fragment and CDR3-Jα41-Cα fragment were combined in a subsequent “fusion” reaction in which the overlapping ends anneal, allowing the 3′ overlap of each strand to serve as a primer for the 3′ extension of the complementary strand. The resulting fusion product is amplified further by PCR. Vβ4-CDR3 fragment and CDR3-Jβ2.6-Cβ fragment were combined similarly. The full-length fragments were cloned into a pMXW retroviral vector to obtain pMXW-AN3α and pMXW-AN3β (Fig. 1 A). pMXW is an improved vector generated by insertion of the woodchuck hepatitis virus posttranscriptional regulatory element (WPRE) (17, 18) between the NotI and SalI sites of pMX (19). WPRE enhances expression of transgenes delivered by retroviral vectors (18), and the expression efficiency of the pMXW vector is improved 1.5 times when WPRE is inserted, compared with the efficiency of the pMX vector. Murine CTLA4Ig cDNA was synthesized by PCR as described previously (20) and was then cloned into the pMX-IRES-GFP (21). Complementary DNAs for the TCR α- and β-chains were isolated from a cDNA library of DO11.10 TCR-transgenic splenocytes and were inserted into the retroviral vector pMX to generate pMX-DOTAE and pMX-DOTBE, respectively (22).

NZB/W F₁ and BALB/c mice were obtained from Japan SLC (Shizuoka, Japan). All animal experiments were conducted in accordance with the institutional and national guidelines.

Total splenocytes were isolated and cultured for 48 h in RPMI 1640 medium supplemented with 10% FCS, 2 mM l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 5 × 10⁻⁵ M 2-ME in the presence of Con A (10 μg/ml) and IL-2 (50 ng/ml) before the transduction. Retroviral supernatants were obtained by transfection of pMXW, pMXW-AN3α, pMXW-AN3β, pMX-IRES-GFP, pMX-CTLA4Ig-IRES-GFP, pMX-DOTAE, or pMX-DOTBE DNA into PLAT-E packaging cell lines (22, 23) with the use of the FuGENE 6 transfection reagent (Roche Diagnostic Systems, Somerville, NJ).

Falcon 24-well plates (BD Biosciences, San Jose, CA) were coated with the recombinant human fibronectin fragment CH296 (Retronectin; Takara, Otsu, Japan) according to the manufacturer’s instructions. Before infection, virus-bound plates were prepared. The viral supernatant (500 μl) was preloaded onto each well of the CH296-coated plate, and the plate was spun at 2000 × g for 3 h at 32°C. The virus-coating procedure was repeated three times. Before infection, the viral supernatant was washed away and splenocytes prestimulated for 48 h were added to each well (1 × 10⁶ cells/well). Cells were cultured for 36 h to allow infection to occur. To control the viral expression efficiency, we produced a viral supernatant (pMXW, pMXW-AN3α, pMXW-AN3β, pMX-IRES-GFP, and pMX-CTLA4Ig-IRES-GFP, simultaneously) and prechecked the uniformity of the infection efficiency before in vitro and in vivo experiments.

A CD4⁺ T cell population was prepared by negative selection with MACS using anti-CD19 mAb, anti-CD11c, mAb, and anti-CD8a mAb. CD11c⁺ DCs were prepared as previously described (24, 25). Briefly, spleen cells or lymph node cells were digested with collagenase type IV (Sigma-Aldrich, St. Louis, MO) and DNase I, and the CD11c⁺ cells were selected twice by positive selection using MACS CD11c microbeads and magnetic separation columns. The purity (85% in average) was determined by visualization with anti-CD11c-biotin followed by streptavidin-PE. A CD19⁺ B cell population was prepared by positive selection with MACS using anti-CD19 mAb. For CFSE labeling (Molecular Probes, Eugene, OR), cells were resuspended in PBS at 1 × 10⁷/ml and incubated with CFSE at a final concentration of 5 mM for 30 min at 37°C, followed by two washes in PBS.

Pure nucleosomes were prepared as previously described (26). Briefly, frozen pure chicken erythrocytes were thawed and suspended in lysis buffer on ice (10 mM Tris-HCl, 10 mM NaCl, 5 mM MgCl₂, 0.5% Nonidet P-40, and 0.25 mM PhMeSO₂F, pH 7.5). The nuclei were recovered by centrifugation and the nuclear pellet was washed and digested with micrococcal nuclease. The nuclear pellet was lysed into 0.2 mM Na₂EDTA, and nuclear debris was removed by centrifugation. The soluble chromatin at A₂₆₀ ≈100 was dialyzed against 5 mM triethanolamine HCl, 60 mM NaCl, 1 mM Na₂EDTA (pH 7.5), and subsequently fractionated in the same buffer, usually in sucrose gradients. Gradients were fractionated and monitored at 280 nm, and the appropriate fractions were pooled.

At 24 h postinfection, purified CD4⁺ T cells were cultured at 2 × 10⁴cells/well, with 1 × 10⁵ cells/well of irradiated CD19⁺ B cells or 1 × 10⁴ cells/well of irradiated CD11c⁺ DCs in 96-well flat-bottom microtiter plates in volumes of 100 μl of complete medium with or without 1 μg/ml nucleosome or 0.3 μM chicken OVA_323–339 peptide. After 24 h of culture, the cells were pulse labeled with 1 μCi of [³H]thymidine/well (NEN Life Science Products, Boston, MA) for 15 h and the [³H]thymidine incorporation was determined. In some experiments, we calculated the ratio of (group A − cpm)/(group B − cpm) in each experiment and showed the average ratio of three to five experiments as “average ratio of (group A − cpm)/(group B − cpm) to clarify the reproducibility of the data.

The indicated number of cells suspended in PBS were i.v. injected into mice. For the transfer of gene-transduced cells, cell viability was always >97%, as detected by trypan blue exclusion.

IgG anti-DNA Abs were quantified using ELISA plates coated with calf thymus DNA (Sigma-Aldrich, and the DNA-binding activities were expressed in units, referring to a standard curve obtained by serial dilutions of a standard serum pool from 7- to 9-mo-old NZB/W F₁ mice, containing 1000 U/ml. IgG antinucleosome Abs or IgG anti-histone Abs were quantified using ELISA plates coated with purified nucleosome or purified histone. Methods for detection of CTLA4Ig protein were described previously (27). Briefly, ELISA plates were coated with anti-mouse CTLA4 (BD PharMingen, San Diego, CA) overnight at 4°C, blocked with blocking solution, and then incubated sequentially for 1 h at 37°°C with serial dilutions of serum or culture supernatants followed by peroxidase-conjugated F(ab′)₂ goat anti-mouse IgG2a (Accurate Antibodies, Westbury, NY) and ABTS substrate (Kirkegaard & Perry, Gaithersburg, MD). Serial dilutions of a known concentration of purified CTLA4Ig were used in each plate to establish a standard curve.

Organs were fixed in 4% paraformaldehyde, embedded in paraffin, and stained with periodic acid-Schiff solution. For three-color immunofluorescence staining, sections were incubated with biotinylated peanut agglutinin (Vector Laboratories, Burlingame, CA) and then with Cy5.5-conjugated streptavidin (Cortex Biochemicals, Irvine, CA). The sections were then stained with a rat Alexa488-labeled mAb to B220 and tetramethylrhodamine-conjugated mAbs to CD4 and CD8 (Vector Laboratories). To detect the deposition of immune complexes at glomeruli, we incubated sections with FITC-labeled goat Abs to mouse IgG or to C3 (ICN Pharmaceuticals, Costa Mesa, CA).

Statistical significance was determined using the unpaired Student’s t test or the Mann-Whitney U test.

We previously reported successful TCR gene transfer and reconstitution of the Ag specificity to OVA in BALB/c CD4⁺ T cells (27). In the present study, we transferred nucleosome-specific TCR genes (AN3α and β) into NZB/W F₁ splenocytes. To improve the expression of the introduced genes, we generated a Moloney-based retroviral vector, pMXW, by insertion of the woodchuck fragment (17, 18) into pMX (19). We selected the TCR of line 3A from lupus-prone SNF1 mice. A hybridoma transfected with this TCR did not exhibit any significant response to either H-2^d or H-2^z APCs (28). Each TCR gene was inserted into pMXW, and the resulting retrovirus vectors (pMXW-AN3α and pMXW-AN3β) were used for the gene transfer (Fig. 1 A).

Retroviral infection of the AN3 TCR genes into NZB/W F₁ splenocytes induced a 40–45% increase in the Vβ4⁺ population in CD4⁺ T cells compared with mock-infected splenocytes (Fig. 1 B). The calculated efficiency of the Vβ4 introduction into the CD4⁺Vβ4⁻ population was 50–60%. The lack of anti-Vα13 or anti-clonotypic Abs prevented direct visualization of AN3 TCR surface expression. However, based on the transduction of other TCR pairs (i.e., OVA-specific DO11.10 TCR detected by a clonotypic Ab KJ1-26; and AV8/BV7, detectable by anti-Vα8 and anti-Vβ7 Abs; data not shown), we speculate that Vα chain expression is approximately equal to that of the Vβ chain. Thus, the proportion of clonotypic AN3 TCR-expressing cells was estimated to be 25–36% in CD4⁺ T cells. These cells were referred to as BWF.AN3, and the mock-infected CD4⁺ T cells were referred to as BWF.mock.

We investigated the specific reactivity to the nucleosomes of BWF.AN3 in the presence of NZB/W F₁ B cells and DCs. Although BWF.mock cells showed minimal proliferation in the presence of B cells and the nucleosomes, BWF.AN3 showed strong proliferation in the presence of B cells and the nucleosomes, but not in the presence of B cells alone (Fig. 1,C). The average ratio of (BWF.AN3 - cpm)/(BWF.mock − cpm) was 1.13 ± 0.12 and that of (BWF.AN3 with nucleosome (nuc) − cpm)/(BWF.mock with nuc − cpm) was 3.12 ± 0.51 in three experiments (p < 0.005). These results demonstrate that the introduction of AN3 TCR reconstitutes the specificity for the nucleosome on CD4⁺ T cells. Furthermore, BWF.AN3 showed proliferation in the presence of splenic DCs without nucleosome (Fig. 1 D). The average ratio of (BWF.AN3 − cpm)/(BWF.mock − cpm) was 6.97 ± 1.63 in five experiments (p < 0.001). Consistent with previous report that CD4⁺ T cells of lupus-prone mice responded to nucleosome ex vivo (7), BWF.mock showed relatively weak proliferation in the presence of splenic DCs and nucleosome. The average ratio of (BWF.mock with nuc − cpm)/(BWF.mock − cpm) was 2.21 ± 0.73 in five experiments (p < 0.05). Despite these endogenous responses of BWF.mock to nucleosome, BWF.AN3 showed stronger proliferation compared with BWF.mock in the presence of splenic DCs with the nucleosomes. The average ratio of (BWF.AN3 with nuc − cpm)/(BWF.mock with nuc − cpm) was 4.01 ± 2.18 in five experiments (p < 0.05).

AN3α-infected or β-infected cells failed to respond to the DCs (data not shown), and the autoreactive response was blocked by anti-I-A^d Ab (Fig. 1 E). Thus, this autoreactivity of BWF.AN3 to splenic DCs suggests that NZB/W F₁ splenic DCs spontaneously present a considerable amount of nucleosomal epitopes.

To investigate the relative contribution of either T cells or splenic DCs to the autoreactive response, we also transduced the AN3 TCR into BALB/c CD4⁺ T cells and these cells were referred to as BAL.AN3. Although BAL.AN3 showed no proliferative response to BALB/c splenic DCs, BWF.AN3 showed a moderate proliferative response to BALB/c splenic DCs (Fig. 1 F). These results suggest that the BWF1 T cell hyperreactivitiy enables BWF.AN3 to recognize small amounts of nucleosomal epitope presented on BALB/c splenic DCs, but these small amounts are ignored by BAL.AN3. As expected, BWF.AN3 strongly responded to BWF1 splenic DCs. Proliferative response of BWF.AN3 in the presence of BALB/c splenic DCs amounted to ∼14–18% of that to BWF1 splenic DCs, indicating that the abnormal presentation of splenic DCs may contribute more to the autoreactive response than does T cell hyperreactivity.

To determine the general Ag recognition and reactivity of NZB/W F₁ mice, we examined the proliferation of T cells transduced with OVA-specific TCR (DO11.10). Fifty to 60% of the total CD4⁺ T cells expressed the introduced DO11.10 TCR, as determined by the anti-clonotypic Ab KJ1-26. DO11.10-transduced BWF1 T cells cultured with DCs plus OVA_323–339 peptide exhibited stronger proliferation than BALB/c T cells, again suggesting that BWF1 T cells possess general hyperreactivity. In contrast, the OVA peptide-presentation (Fig. 1 G) and the whole OVA presentation (data not shown) of NZB/W F₁ splenic DCs appeared to be quite similar to that of BALB/c splenic DCs. Thus, the hyperpresentation of DCs seems to be restricted to a certain Ag.

DCs in every type of lymphoid tissue may present nucleosomal epitopes, because nucleosomal Ags are available in every organ. To investigate this possibility, we fluorescently labeled either BWF.mock or BWF.AN3 T cells in vitro with CFSE and injected them into NZB/W F₁ mice. Two days after the transfer, T cells from the spleen and those from the peripheral lymph nodes (LNs) were harvested and analyzed. BWF.mock isolated from the spleen exhibited a convergent strong fluorescence peak, indicating that these cells had not proliferated extensively. In contrast, BWF.AN3 isolated from the spleen exhibited several weaker fluorescence peaks. Moreover, AN3 CD4⁺ T cells underwent multiple divisions over a 5-day period of the experiment, and mock CD4⁺ T cells underwent a very slight progression of cell division (Fig. 2 A). These findings suggested that BWF.AN3 encountered the nucleosomal epitope in the spleen. It was of note that both CFSE-labeled BWF.mock and BWF.AN3 isolated from the peripheral LNs exhibited a strong convergent fluorescence peak, suggesting that BWF.AN3 encountered the nucleosomal epitope less frequently in the LNs.

A comparison of the stimulative capacity for BWF.AN3 also suggested that splenic DCs presented more nucleosomal epitope than DCs from the peripheral LNs (Fig. 2 B). The average ratio of (BWsplDC − cpm)/(BWLNDC − cpm) was 2.79 ± 0.44 in three experiments (p < 0.005). These results showed that nucleosome-specific T cells are stimulated predominantly in the spleen.

We next tried to generate nucleosome-specific regulatory cells by introducing an immunosuppressive molecule, CTLA4Ig, as the third gene in BWF.AN3 T cells. Long-term administration of CTLA4Ig to NZB/W F₁ mice has been shown to prevent disease onset for a period of months (29).

We constructed a pMX-CTLA4Ig-IRES-GFP vector (Fig. 1,A). We then performed a triple gene transfer of the AN3αβ and CTLA4Ig genes to investigate the effect on CTLA4Ig expression. The experimental groups consisted of CD4⁺ T cells transduced with either AN3 + CTLA4Ig-IRES-GFP(CTLA4Ig), AN3 + IRES-GFP(IG), pMXW(mock) + CTLA4Ig, or mock + IG. The average expression efficiency from several different sets of infection was 45.2% for Vβ4 and 47.3% for GFP in CD4⁺ cells (Fig. 3,A). The average expression efficiency is expected to be 45% for the AN3α gene, and the average percentage of GFP⁺AN3⁺ cells expressing all three gene products in CD4⁺ T cells was estimated to be ∼10% (0.45 × 0.45 × 0.45). As shown in Fig. 3,B, the CTLA4Ig secreted from T cells blocked the proliferation of the endogenous T cell population to the nucleosome to a moderate degree. The average ratio of (mock + CTLA4Ig with nuc − cpm)/(mock + IG with nuc − cpm) was 0.40 ± 0.07 in three experiments (p < 0.005). But the T cell stimulation mediated by AN3 TCR was not blocked by CTLA4Ig. The average ratio of (AN3 + IG − cpm)/(mock + IG − cpm) was 7.85 ± 1.07 and that of (AN3 + CTLA4Ig − cpm)/(mock + IG − cpm) was 7.18 ± 0.96 in three experiments. The AN3 + CTLA4Ig transduced cells showed the increase of CTLA4Ig secretion on T cell activation in the presence of DCs (Fig. 3 C).

We transferred cell suspensions containing 1 × 10⁶ cells of CD4⁺ T cells, calculatedly expressing either AN3 + CTLA4Ig, AN3 + IG, mock + CTLA4Ig, or mock + IG into 10-wk-old NZB/W F₁ mice.

The autoantibodies usually found in NZB/W F₁ mice were measured in the sera from the different groups. The elevations of anti-dsDNA and anti-histone Abs were suppressed in AN3 + CTLA4Ig-injected mice at 22 wk of age (Fig. 4 A). AN3 + CTLA4Ig-injected mice showed the lowest average titer of anti-nucleosome Ab, but the titer in this group was not significantly different from those in the controls. This inefficient suppression may be due to the fact that autoimmunity to the nucleosome is the driving reaction and that this reaction is stronger than the subsequent response.

The mice were monitored biweekly for proteinuria. By week 22, control mice that had received PBS, mock + IG, AN3 + IG, or mock + CTLA4Ig started developing severe nephritis, as diagnosed by persistent proteinuria of >300 mg/dl. By 30 wk of age, 89% of the PBS control group, 88% of the mock + IG group, 63% of the AN3 + IG group, and 75% of the mock + CTLA4Ig group of mice had developed severe proteinuria, whereas none of the AN3 + CTLA4Ig mice showed excess proteinuria (Fig. 4 B). However, the AN3 + CTLA4Ig-transferred mice started to develop severe proteinuria at 32 wk of age. Splenomegaly and an increase in the CD4:CD8 ratio, usually observed in aged NZB/W F₁ mice, were suppressed in AN3 + CTLA4Ig-injected mice (data not shown).

The kidneys from the controls and AN3 + CTLA4Ig-injected mice were examined at 30 wk of age (Fig. 5, A–F). Control mice had severe glomerulonephritis with mesangial proliferation and thickening of the capillary walls with marked deposition of IgG and complement. AN3 + CTLA4Ig-injected mice had mild glomerular lesions and deposition of IgG and complement was only restricted to the mesangial area. Although mock + CTLA4Ig-transferred mice showed formation of a number of large follicles with T cell invasion in the spleen, AN3 + CTLA4Ig-transferred mice showed only a limited number of small follicles (Fig. 5, G and H).

We next examined the T cell-dependent humoral immune response to active immunization of OVA. Mice transferred with the engineered T cells at 10 wk of age were immunized with OVA (100 μg) with CFA at 14 wk of age and boosted with OVA with IFA at 16 wk of age. The level of anti-OVA IgG Ab titer from 17-wk-old mice treated with AN3 + CTLA4Ig was not significantly different from those of the control mice (Fig. 4,C). AN3 + CTLA4Ig transferred mice, but not other experimental groups, had low but detectable levels of serum CTLA4Ig (13.4 ± 10.1 μg/ml) (Fig. 4,D), findings consistent with in vitro data shown in Fig. 3 C. These results suggest that the engineered regulatory cells are sufficient to suppress autoimmune disease. However, they are not enough to induce general immunosuppression, because of the low serum level of CTLA4Ig in AN3 + CTLA4Ig-transferred mice.

In this study, we demonstrated T cell hyperresponsiveness and the possibility of nucleosomal hyperpresentation of splenic DCs in NZB/W F₁ mice. In addition to the involvement of T cell hyperresponsiveness in Ab-mediated autoimmune disease (30), our results strongly suggest that the autoantigen hyperpresentation of DCs could contribute to the initiation and propagation of the response to the autoantigen, thereby resulting in florid autoimmune disease. This observation is consistent with those from previous reports indicating that mice with T cell hyperresponsiveness develop only a mild form of lupus-like symptoms (31, 32). Since hyperpresentation was not observed in the case of an exogenous Ag, OVA (peptides and whole protein), it is possible that the autoantigen hyperpresentation of splenic DCs was not due to the general hyperpresentation, e.g., excessive costimulatory signals, but rather to some Ag-restricted phenomenon. These features may be nucleosome specific, as reported in a previous study demonstrating that lupus-prone B6.NZMc1 mice showed nucleosome reactivity of T cells without generalized immunological deficits of B cells and T cells (33).

Although disease-related increases in the number of splenic DCs and chemokine production by myeloid DCs have been reported (34), these abnormalities have been observed in aged lupus-prone mice. Our finding of autoantigen hyperpresentation in the splenic DCs of young mice (10 wk) suggests the significance of the autoantigen hyperpresentation of splenic DCs in the pathogenesis of lupus.

Autoreactive response of nucleosome-specific T cells was much more prominent in the spleen than in the LNs. Although the mixed I-A haplotype of Aβz/Aαd molecules in NZB/W F₁ mice (35) may be associated with autoreactive response of AN3 infectant, the absence of the autoreactivity to B cells and DCs from peripheral LNs strongly suggests the requirement of an autoantigen for the autoreactivity. Differences between the splenic DCs and DCs from other peripheral lymphoid organs have been reported, including differences in the expression of chemokines (36) and chemokine receptors (37). Otherwise, localization of tissue-specific autoantigen among secondary lymphoid organs may be one explanation. For example, although DCs in the gastric LNs are known to exhibit constitutive presentation of gastric parietal cell-specific H⁺/K⁺-ATPase, peripheral or mesenteric DCs do not (38). Thus, the spleen could be one of the main sources of nucleosomes. Increased frequency of splenic apoptosis in SNF1 lupus mice has also been reported (23). Moreover, an insufficient complement system may allow apoptotic waste material to accumulate in the spleen (i.e., the “waste disposal” hypothesis) (39).

In our study, the therapeutic effect with minimal systemic immunosuppression was archived by the use of nucleosome-specific T cells secreting CTLA4Ig. Although elevation of CTLA4Ig protein was detected in the serum of AN3 + CTLA4Ig mice, the average concentration of CTLA4Ig in AN3 + CTLA4Ig mice is less than one-tenth of the level of previous systemic CTLA4Ig treatment with 5 × 10⁸ PFU of adenovirus (27). Although the systemic adenoviral-CTLA4Ig (5 × 10⁸ PFU) treatment exhibited a therapeutic effect equivalent to that of our experiment, the systemic treatment was accompanied with generalized immunosuppression. Since autoantigen-specific CTLA4Ig-secreting T cells showed normal Ab production on active immunization, this treatment may be superior to systemic CTLA4Ig administration. However, a systemic effect of a very low level of CTLA4Ig cannot be excluded and should be investigated further.

It is not surprising that 10⁶ AN3 + mock cells did not aggravate the disease, since as many as 4 × 10⁷ original L3A clone cells were needed to accelerate lupus nephritis in young lupus-prone mice (40). Thus, a relatively small amount of Ag-specific and potentially pathogenic T cells could be used for the immunotherapy. Foxp3, a member of the transcription factor family, has been identified as a key molecule for the development of CD4⁺CD25⁺ regulatory T cells (41). Retroviral transfer of Foxp3 confers regulatory function on CD4⁺CD25⁻ T cells. The introduction of such regulatory molecules with TCR could possibly generate Ag-specific regulatory T cells.

In a preliminary analysis of the persistence of the transferred genes in the spleen and LNs from 30-wk-old mice with RT-PCR, expression of AN3α gene was detected in the spleens from two of two AN3 + IG⁻ and AN3 + CTLA4Ig⁻-injected mice (data not shown). These results may suggest the persistence of introduced genes at 20 wk after the transfer in the spleen.

Although several models of adoptive cell gene therapy have been reported using T cell hybridomas or lines (42, 43), our method has the advantage of using autologous lymphocytes for gene recipients. However, TCR-transduced recipient T cells could gain heterodimeric TCR consisting of endogenous and exogenous chains. If such an unexpected TCR recognizes a certain unrelated self-derived molecule, the transduced T cells may be harmful. We did not observe evident autoreactivity in single AN3α or AN3β genes transferred into CD4⁺ T cells (data not shown), and the renal disease of AN3 TCR-transferred mice was not accelerated (Fig. 5 B). There was a recent report of tumor rejection mediated by retrovirally reconstituted Ag-specific T cells without any significant autoimmune pathology (44, 45). However, the possibility of developing autoimmunity should be carefully investigated further in application of TCR gene transfer.

In the present study, the efficacy of triple gene transfer in peripheral T cells was demonstrated for the first time. Although several improvements of the present method are still necessary, these findings suggest that the direct engineering of Ag-specific functional cells with multiple gene transfer is a powerful technique for the development of future Ag-specific therapies.

We are grateful to Kazumi Abe and Shiho Ohta for their excellent technical assistance.