Profound autoimmunity in CD3-mutant mice

We first compared the spectrum of T cell development in recombination-activation gene 1–deficient (Rag1^{- /-} ) mice reconstituted with retroviral vector–transduced donor bone marrow from mice lacking genes encoding CD3 and CD3 (CD3-KO mice) that also lacked CD3 and CD3 mRNA and protein because of gene silencing caused by the insertion of a phosphoglycerate kinase–neomycin-resistance cassette into the Cd3e locus^{9, 10, 11} (Supplementary Fig. 1a online). The retroviral vectors used encoded wild-type or ITAM-mutant CD3 chains (Supplementary Fig. 1b–d). Rag1^{- /-} recipient mice reconstituted with CD3-KO bone marrow transduced with a vector encoding all four wild-type CD3 subunits (CD3^WTCD3^WTCD3^WTCD3^WT; called 'CD3^WT' here) had CD4 and CD8 expression patterns on thymocytes and splenocytes similar to those of Rag1^{- /-} recipients of wild-type bone marrow transduced with empty vector (Supplementary Fig. 1e). The TCR surface expression on thymocytes and splenocytes from these mice was also similar to that of cells from C57BL/6 mice (Supplementary Fig. 1f). In contrast, Rag1^{- /-} recipient mice reconstituted with CD3-KO bone marrow transduced with empty vector or with a vector encoding ITAM-mutant CD3 subunits (CD3^MCD3^MCD3^MCD3^M; called 'CD3^M' here) showed severely impaired T cell development, even though B cell reconstitution was normal; thus, these mice had a phenotype similar to that of unmanipulated CD3-KO mice (Supplementary Fig. 1e). We then generated 23 additional retroviral vectors encoding various combinations of wild-type and mutant CD3 ITAMs for subsequent analysis (Table 1). Some combinations included mutant CD3 subunits that had only one wild-type ITAM (for example, in CD3a^WTbc^M, 'a' is the membrane-proximal ITAM, 'b' is the middle ITAM and 'c' is the membrane-distal ITAM).

Table 1: CD3 mutants

Full table

Unexpectedly, mice with two wild-type ITAMs (CD3^WTCD3^M; called 'CD3^WTM' here) or four wild-type ITAMs (CD3^WTCD3^M; called 'CD3^WTM' here) became sick 5–6 weeks after bone marrow transfer and died within 1 week of the onset of signs that included lethargy, hunched posture, ruffled coat, dehydration and squinting but limited weight loss (Fig. 1a). This early and rapid disease morbidity and mortality was particularly notable, as T cells were not detectable in the periphery until approximately 3 weeks after bone marrow transfer, and complete T cell reconstitution required approximately 8 weeks. Rag1^{- /-} recipients of CD3-KO bone marrow transduced with CD3^M retrovirus had an autoimmune disease similar to that noted when CD3-KO donor bone marrow was used, but recipients of CD3-KO bone marrow transduced with CD3^WT retrovirus did not (Supplementary Fig. 2 online). This finding indicated that ITAM number rather than the retroviral vector transduced (single versus 2A-linked multicistronic) and/or bone marrow donor (CD3-KO versus CD3-KO) was responsible for the disease phenotype (data not shown). This disease could also be adoptively transferred with splenocytes from sick CD3^WTM and CD3^WTM bone marrow chimeras but not those from healthy CD3^WT bone marrow chimeras (Fig. 1b).

Figure 1: Profound autoimmunity in CD3-mutant mice.

(a) Time of onset of sickness in Rag1^{- /-} mice after transfer of bone marrow from CD3^WT mice (n = 60 recipient mice), CD3^WTM mice (n = 45 recipient mice) or CD3^WTM mice (n = 43 recipient mice). (b) Time of onset of sickness in Rag1^{- /-} recipients of 8 10⁶ whole splenocytes from spleens collected from CD3^WT mice (n = 21 recipient mice), CD3^WTM mice (n = 21 recipient mice) or CD3^WTM mice (n = 10 recipient mice) after the mutant mice had become sick. (c) Hematoxylin and eosin staining of formalin-fixed, paraffin-embedded sections of organs from the mice in a. (d) Time to onset of sickness (mean and s.e.m.) after bone marrow transfer (designations below graph as in Table 1). (e) Immunofluorescence staining and DAPI counterstaining of formalin-fixed cryostat sections to detect CD4⁺ T cells and Iba-1⁺ activated macrophages. (f) Reactivity profiles of serum from CD3-wild-type and CD3-mutant mice compared by autoantigen array. Analysis of microarrays¹⁷ identified antigens with statistically significant differences in array reactivity for CD3-wild-type and CD3-mutant mice (Supplementary Table 3). Serum from mice with pristane-induced systemic lupus erythematosus serves as a positive control. ssDNA, single-stranded DNA; dsDNA, double-stranded DNA; Flax_f1, flax flagellen; C-anca, antibody to neutrophil cytoplasmic antigen. Data are the mean of eight to ten separate experiments (a) or two to three separate experiments (b) or are representative of five to ten mice per group (c), two to four experiments with 9–20 mice (d), two experiments with two mice per group (e) or three independent experiments with 33 serum samples from six groups (f).

Full size image (154 KB)

The development of autoimmunity in the CD3^WTM bone marrow chimeras was also notable because CD3-KO mice expressing a 'Tg1-6^M' mutant transgene in which sequence encoding each ITAM tyrosine was replaced with sequence encoding phenylalanine (and thus having the same number of wild-type ITAMs as the CD3^WTM bone marrow chimeras) develop only very mild indications of inflammatory disease in old age^{12, 13, 14}. This difference was not due to our experimental approach, as Rag1^{- /-} recipients of CD3-KO Tg1-6^M-transgenic bone marrow transduced with empty vector also failed to develop disease (Supplementary Fig. 2a). Notably, thymocytes and splenocytes from CD3-KO Tg1-6^M-transgenic mice had much higher TCR surface expression than that of C57BL/6 mice or mice with retroviral transgenic ('retrogenic') expression of CD3^WT or CD3^M, which may have masked and/or compensated for any signaling deficiencies during negative selection (Supplementary Fig. 2b–d). It is noteworthy that T cells in CD3-KO mice, which have lower expression of surface TCR, overtly react with self major histocompatibility complex (MHC) once normal TCR expression is restored¹⁵.

Histological examination of the sick mice showed a broad spectrum of tissues affected (Supplementary Results and Tables 1, 2 online). The most severe pathology was in the lungs, liver and gastrointestinal tract (Fig. 1c). Death was probably due to multiorgan inflammatory processes characterized by colitis, and vasculitis and perivasculitis in the lung and liver.

Subsequent analysis of all 25 groups of mice expressing wild-type or mutant CD3 showed that 13 groups developed very similar signs and had similar lesions by histological examination (Table 1 and Supplementary Table 1). These groups had two to six wild-type ITAMs. Notably, only two of the four groups with six wild-type ITAMs became sick, which suggested that the type of ITAM present can also affect disease outcome. In general, the kinetics of disease onset slowed with fewer wild-type ITAMs, which suggested a direct correlation between CD3 signal strength (proportional to functional wild-type ITAM number) and disease severity and/or time to disease onset (Fig. 1d and Supplementary Fig. 3c online). Disease penetrance was 100% in these 13 groups of mice. Immunofluorescence analysis of lung, liver, intestine and kidney showed substantial CD4⁺ T cell infiltration and macrophage activation but more limited infiltration of CD8⁺ T cells and CD19⁺ B cells (Fig. 1e and Supplementary Fig. 4 online). We confirmed an absolute requirement for CD4⁺ T cells in the initiation and progression of autoimmunity by adoptive transfer of disease with purified CD3^WTM CD4⁺ T cells but not CD8⁺ T cells or CD3^WT CD4⁺ wild-type control cells (Supplementary Fig. 5 online).

Many autoimmune diseases are characterized by the development of autoantibodies to a variety of host antigens. We used autoantigen microarrays to compare the reactivity profiles of serum from CD3-wild-type and CD3-mutant retrogenic mice¹⁶. Hierarchical clustering and significance analysis of microarrays¹⁷ showed that a small proportion of serum from CD3-mutant mice had statistically significant autoantibody reactivity patterns. Whereas serum from most of the CD3^WTM mice did not show an antigen-binding pattern that was significantly different from that of CD3-wild-type controls, except for some borderline positive reactivities, serum from CD3^WTM and CD3^WTM mice clustered together with positive control serum from mice with pristane-induced systemic lupus erythematosus (Fig. 1f and Supplementary Fig. 6 and Table 3 online). Notably, there were both overlapping reactivities (for example, to single-stranded and double-stranded DNA) and unique reactivities (for example, to flax flagellen, a peptide of bacterial flagella). In addition, serum from one of the two CD3a^WTbc^M mice developed vasculitis-associated autoantibodies directed against neutrophil antigens (Supplementary Fig. 6). It is unclear why only certain CD3-mutant mice developed autoantibodies. The detection of autoantibodies in CD3^WTM mice was notable, given that these mice failed to develop symptomatic disease in the time frame of this study (50 d), and may suggest the existence of more chronic forms of autoimmunity in mice with six to nine functional CD3 ITAMs.

High ITAM number required for T cell development

To identify the defects that led to the severe autoimmune disease of the ITAM-mutant mice, we first assessed the efficiency of T cell development and establishment of a peripheral T cell pool in mice expressing each of the 25 CD3 ITAM combinations. Another goal of this analysis was to calculate the number and type of ITAMs required for various T cell–developmental parameters (such as thymocyte number, percentage of CD4 single-positive thymocytes, splenic T cell number and so on). We initially generated 446 mice having at least 40% reconstitution with green fluorescent protein–positive (GFP⁺) splenocytes in 15 experiments (Table 1, Fig. 2 and Supplementary Fig. 7 online). Each parameter examined required a specific threshold number of wild-type ITAMs (such as seven ITAMs for thymocyte number, four for splenic T cell number, two for CD4⁺ cell/CD8⁺ cell ratio), rather than a linear relationship correlating with ITAM numbers.

Figure 2: A specific number of ITAMs is required for each T cell developmental parameter.

Analysis of Rag1^{- /-} recipient (retrogenic) mice 5–8 weeks after transfer of transduced CD3-KO bone marrow. (a) Total number of thymocytes in each group. (b) Flow cytometry of thymocytes stained for CD4, CD8, CD25 and CD44 and gated on live GFP⁺CD4^- CD8^- cells, with a quadrant gate set to obtain percent cells at each double-negative stage (DN1–DN4). (c) Flow cytometry of thymocytes stained for CD4 and CD8 and gated on live GFP⁺ cells, with a quadrant gate set to obtain percent of each population. (d) Flow cytometry of percent live GFP⁺TCR⁺ cells to assess splenic T cell numbers. (e) Ratio of CD4⁺ splenocytes to CD8⁺ splenocytes for cells analyzed as described in d. Data are the mean and s.e.m. of 9–40 mice from two to eight separate experiments per group.

Full size image (98 KB)

In general, mice expressing seven to nine wild-type CD3 ITAMs had a pattern of T cell development and peripheral T cell seeding similar to that of wild-type control mice (Fig. 2). There were differences in the total number of splenic T cells in each group, and the low error and large number of mice examined indicated that these findings were not due to experimental variation but instead were due to subtle differences imposed by the individual ITAMs. The most notable difference among the groups with seven to nine wild-type CD3 ITAMs was that CD3^WTM mice had 2.5-fold higher thymocyte numbers than did CD3^WT mice (Fig. 2a). CD3^WTM mice did not have this larger number of thymocytes, which suggested that small differences in ITAM sequence and/or location can influence TCR signaling and thymocyte development.

In contrast, we noted substantially lower total thymocyte numbers in mice with six or fewer wild-type ITAMs, which suggested alterations in selection, survival and/or egress (Fig. 2a). Pre-TCR signaling, which mediates the transition from the double-negative 3 stage to the double-negative 4 stage, seemed essentially normal in mice with two or more wild-type ITAMs (Fig. 2b and Supplementary Fig. 8a online). These data also show that effective pre-TCR signaling can occur with only two functional ITAMs, which supports the idea that nominal tonic signaling induced by expression of rearranged TCR–pre-TCR complexes is sufficient to mediate transition from the double-negative 3 stage to the double-negative 4 stage¹⁸. Upregulation of CD5 on single-positive thymocytes, a marker of positive selection¹⁹, for the most part correlated with thymic output, such that mice with six or more functional ITAMs that did not develop autoimmunity had normal CD5 expression, whereas mice with six or fewer functional ITAMs that did develop disease had low CD5 upregulation (Supplementary Fig. 8b,c). For unknown reasons, mice with very few functional ITAMs seemed to have higher CD5 expression than that of other groups. Many of the groups that developed autoimmunity (three to six wild-type ITAMs) also had a much lower percentage of double-positive thymocytes and an associated increase in the ratio of single-positive cells to double-positive cells (Fig. 2c and Supplementary Fig. 7c). Although some of the decrease in double-positive populations could have been due to disease-induced release of steroid hormones, much of this analysis was done before disease onset. Furthermore, there were two groups of mice with six wild-type ITAMs (CD3^WTM and CD3b^WTac^M) that had fewer total thymocyte numbers but did not develop the autoimmune disease or pathology.

Thymocyte cellularity did not always correlate with the size of the peripheral T cell pool, as six groups had a low number of thymocytes but essentially normal splenic T cell numbers (Fig. 2d). Discrepancies between thymocyte and splenic T cell numbers did not correlate with the ratio of single-positive cells to double-positive cells or disease state and thus are more likely a consequence of homeostatic proliferation that resulted in 'filling up' of the peripheral T cell space. Notably, these six groups were the only groups with an intact CD3 ITAM and at least one intact CD3 ITAM, which suggested that these ITAMs may be key in T cell homeostasis. Whereas the proportion of T cells in the spleen varied considerably among the CD3-mutant groups, the percentage of T cells in the spleen and Peyer's patch of CD3-mutant groups was the same or higher than that in mice with wild-type CD3 over a broad ITAM range, which suggested that relatively low TCR signal strength may be sufficient for normal development and homeostasis of T cells (Supplementary Fig. 9 online). CD3^WTM mice, which developed rapid autoimmunity, had many more T cells, especially in the Peyer's patches. Whether this is a consequence of the autoimmune status of these mice remains to be determined.

TCR surface expression on peripheral T cells and TCR variable -chain (V) use were unaltered between wild-type mice and mutant mice that developed disease (Supplementary Fig. 7d, 8d). The CD4⁺ T cell/CD8⁺ T cell ratio was normal (2:1) in the spleens of mice with three or more wild-type ITAMs but was much higher (approximately 4:1) in mice with fewer than three wild-type ITAMs (Fig. 2e). The one exception was CD3^WTM mice, which had a normal CD4⁺ cell/CD8⁺ cell ratio; this suggested that a functional CD3 ITAM alone was uniquely sufficient to control this balance. Finally, the two groups with only one functional ITAM had almost no double-negative 4, double-positive or single-positive thymocytes. Of these two groups, mice with only a wild-type CD3 ITAM had no splenic T cells, but 40% of mice with only a wild-type CD3 ITAM nevertheless had detectable splenic CD4⁺ T cell populations (Fig. 2d).

The data reported above collectively demonstrate that each parameter examined required a specific threshold of wild-type ITAMs. It is evident that the number (quantitative parameter) rather than the type (qualitative parameter) of wild-type ITAMs was the main defining factor. However, we noted a secondary qualitative influence of certain CD3 ITAMs, most notable for CD3 verses CD3. Thus, differential recruitment of 'downstream' signaling molecules and/or location in the TCR-CD3 complex may also influence T cell development.

Low ITAM number and T cell function

We next assessed the number of CD3 ITAMs required for peripheral T cell proliferation and cytokine production in response to the 'superantigen' staphylococcal enterotoxin B (there were no differences in the percentage of V7⁺ or V8⁺ T cells reactive to staphylococcal enterotoxin B among any of the groups examined; Supplementary Fig. 8d and data not shown). In contrast to the discrete number of wild-type ITAMs required for each of the developmental parameters examined above, there was a precise linear relationship between the number of CD3-wild-type ITAMs and T cell–proliferative capacity (Fig. 3a,b and Supplementary Fig. 10 online). The groups that were most displaced from the linear regression line all had a single CD3 or CD3 wild-type ITAM (Supplementary Fig. 10a), which suggested a unique modulatory influence for these ITAMs on T cell proliferation.

Figure 3: Cells from CD3-mutant mice proliferate less but have normal cytokine production.

(a–g) Analysis of cells from Rag1^{- /-} recipient (retrogenic) mice 5–8 weeks after transfer of CD3-KO bone marrow; splenic T cells (2 10⁵) were purified by magnetic-activated cell sorting and stimulated for 48 h with staphylococcal enterotoxin B (30 g/ml) plus irradiated C57BL/6 splenocytes (5 10⁵). (a,b) Cells pulsed for a further 24 h with [³H]thymidine. Data in b are a linear regression of the data in a. (c–g) Analysis of cytokines in 50 l of the supernatant immediately after stimulation; far right bars, limit of detection or media control. Data are the mean and s.e.m. of four to ten separate experiments. (h–k) Analysis of cells from Rag1^{- /-} recipient (retrogenic) mice 8 weeks after transfer of transduced CD3-KO 3A9 TCR–transgenic bone marrow: splenic T cells (2 10⁵) purified by magnetic-activated cell sorting were stimulated for 48 h with various concentrations of HEL (48–62)plus irradiated B10.BR splenocytes (5 10⁵). (h) Cells pulsed for a further 24 h with [³H]thymidine. (i–k) Analysis of cytokines in 50 l of the supernatant immediately after stimulation. Data are the mean s.e.m. of two separate experiments.

Full size image (96 KB)

In contrast, the presence of fewer wild-type CD3 ITAMs had no significant effect on cytokine production (Fig. 3c–g). However, the type of CD3 ITAM present did seem to have some modulatory effect on cytokine production. T cells from mice that developed rapid autoimmunity produced more of the proinflammatory cytokine interferon- (IFN-) than did those from wild-type mice. Notably, T cells with only a single wild-type CD3 ITAM produced almost the same amount of interleukin 2 (IL-2), IL-4 and IFN- as wild-type T cells did (Fig. 3c–e) and produced nearly twice as much IL-5 and IL-10 as wild-type T cells did (Fig. 3f,g). These data indicate that a low threshold number of CD3 ITAMs is required for cytokine production.

Although V use was unaltered by the number of wild-type CD3 ITAMs, it is possible that there may have been alterations in the TCR repertoire as a consequence of selection of peptide-MHC recognition with higher affinity to compensate for fewer wild-type ITAMs²⁰. To address that possibility, we transduced bone marrow from CD3-KO mice expressing a transgene encoding the 3A9 TCR (specific for the peptide hen egg lysozyme amino acids 48–62 (HEL (48–62)) and H-2A^k restricted) with either CD3^WT or CD3^M retroviral vectors. Consistent with their non–TCR-transgenic counterparts, they had fewer thymic and splenic CD3^M 3A9 T cells than the wild-type controls had (Fig. 2a,d, and Supplementary Fig. 11a,b online). However, we found no disease in these mice, which may have been due in part to the high CD4⁺ splenic T cell/CD8⁺ splenic T cell ratio induced by the MHC class II–restricted TCR transgene (data not shown and Supplementary Fig. 11c). CD3^WT 3A9 T cells but not CD3^M 3A9 T cells proliferated in response to peptide stimulation (Fig. 3h). In contrast, CD3^M 3A9 T cells secreted much more IL-2 and IFN- than CD3^WT 3A9 T cells did, although some secretion of the former was due to a lack of proliferation-driven autocrine consumption (Fig. 3i,j). Production of IL-10 by CD3^WT and CD3^M 3A9 T cells was similar, which indicated that having 60% fewer functional ITAMs had little effect on cytokine production (Fig. 3k). Thus, whereas lower CD3 signal strength limited T cell–proliferative capacity, the ability to produce a wide array of cytokines remained mostly intact.

Lymphopenia is not required for autoimmune disease

To determine what caused disease in the CD3-mutant mice, we first assessed the contribution of T cell lymphopenia. Although many of the groups that developed this disease were lymphopenic, several factors suggested an etiology distinct from lymphopenia-driven diseases (Supplementary Results and Supplementary Fig. 12 online). Despite those differences, we addressed this issue directly by generating mixed retrogenic chimeras by reconstituting Rag1^{- /-} mice with various combination of CD3-KO bone marrow transduced with CD3^WTM, CD3^WT or empty vector (Supplementary Fig. 13a online). The reconstitution, T cell development and disease onset of wild-type CD3^WT and mutant CD3^WTM control groups was similar to that in previous experiments. CD3-KO bone marrow transduced with empty vector and injected together affected the percentage of 'marked' bone marrow and B cells but had no effect on T cell reconstitution or disease score (Supplementary Fig. 13a–c). Mice injected with equal proportions of bone marrow transduced with CD3^WT and bone marrow transduced with CD3^WTM had T cell populations similar in size to those of mice injected with bone marrow transduced with CD3^WT alone or with a 50:50 mixture of bone marrow transduced with CD3^WT and bone marrow transduced with empty vector (Supplementary Fig. 13a,b). Notably, disease nevertheless developed in CD3^WT-CD3^WTM mixed bone marrow chimeras, albeit to a lesser extent, which suggested that lymphopenia was not required for disease development (Supplementary Fig. 13c,d). Disease in these mice seemed to resolve after about 6 weeks after transfer, which may have been because of the presence of regulatory T cells (T_reg cells) derived from wild-type CD3^WT bone marrow. These data suggest that the autoimmune disease of CD3-mutant mice is exacerbated but not caused by lymphopenia.

T_reg cell number and function

We next sought to determine whether autoimmune disease was due to a breakdown of peripheral or central tolerance. T_reg cells are a critical subpopulation of CD4⁺ T cells that are essential for maintaining self tolerance and preventing autoimmunity^{21, 22, 23, 24, 25} and for limiting chronic inflammatory diseases, such as asthma and inflammatory bowel disease^{26, 27}. T_reg cells are key in enforcing peripheral tolerance, and their absence can cause profound autoimmunity^{28, 29}. The percentage of splenic Foxp3⁺ T_reg cells was essentially normal in four groups of CD3-mutant mice with a range of wild-type ITAM numbers (Fig. 4a). Immunofluorescence analysis suggested that CD3^WTM Foxp3⁺ T_reg cells were able to migrate into the lung and liver lesions of autoimmune mice (data not shown). Notably, despite their TCR signaling deficiencies, CD3^WTM T_reg cells had nearly normal suppressive activity relative to that of their wild-type counterparts in controlling the proliferation of fully functional naive effector T cells in vitro (Fig. 4b). Thus, the autoimmune disease that developed in CD3-mutant mice seemed to do so in the presence of a normal proportion of functionally competent Foxp3⁺ T_reg cells. However, it is possible that this population had some functional deficiency in vivo.

Figure 4: Normal proportion and function of T_reg cells in CD3-mutant mice.

(a) Staining of CD4 and Foxp3 in splenocytes from retrogenic mice generated by retrovirus-mediated stem cell gene transfer, assessed 5–6 weeks after transplant with gating on CD4⁺ lymphocytes. Data are the mean and s.e.m. of two separate experiments. (b) Suppressive activity of spleen and lymph node cells from the mice in a 6 weeks after transplant, sorted based on a GFP⁺CD4⁺CD45RB^loCD25⁺ profile for T_reg cells (suppressor) and a GFP⁺CD4⁺CD45RB^hiCD25^- profile for naive T cells (responder (Resp)), mixed at various ratios (horizontal axis) and stimulated with antibody to CD3 plus irradiated C57BL/6J splenocytes. Data are the mean s.e.m. of two separate experiments.

Full size image (34 KB)

Lower CD3 signaling converts negative to positive selection

It was possible that the autoimmune disease of CD3-mutant mice could have been due to a breakdown in central tolerance^{30, 31}. For example, self-reactive thymocytes that are normally deleted by negative selection could instead have been positively selected because of lower TCR signaling capacity. To test that hypothesis directly, we expressed wild-type or mutant CD3 molecules together in cells bearing a transgene encoding the MataHari TCR (V8.3, specific for the male antigen Uty^{32, 33}). We transduced bone marrow from male or female CD3-KO Rag1^{- /-} mice with MataHari TCR-2A-TCR retrovirus plus either wild-type or mutant CD3 retrovirus and then used that bone marrow to reconstitute irradiated Rag1^{- /-} recipients. In CD3^WT MataHari female mice, a uniform population of CD8⁺V8.3⁺ T cells was present in the periphery (Fig. 5a–c). However, in the presence of the male self antigen, T cells were deleted, as indicated by the lack of GFP⁺ cells in the blood (Fig. 5a,b). Some T cells accumulated in the spleen but had low expression of CD8, lacked V8.3 TCR expression and failed to produce IFN- after antigen stimulation (Fig. 5c and data not shown). Such features are indicative of effective negative selection and peripheral tolerance induction^{32, 34}. In contrast, male mice expressing three of four mutant CD3 ITAM combinations had a high percentage of GFP⁺ cells in the blood, and all groups had a substantial percentage of CD8⁺V8.3⁺ T cells in the spleen (Fig. 5a–c). However, these T cells were unresponsive to stimulation with cognate antigen, consistent with previous observations³⁵ (data not shown). These data suggest that a lower wild-type ITAM number resulted in positive rather than negative selection of self-reactive T cells.

Figure 5: Lower signaling converts negative selection into positive selection and generates IFN--producing autoreactive T cells.

(a–c) Flow cytometry of cells from blood (a,b) and spleens (c) of Rag1^{- /-} mice 8 weeks after they were reconstituted with CD3-KO Rag1^{- /-} bone marrow transduced with retrovirus encoding MataHari TCR and retrovirus encoding wild-type or mutant versions of CD3. Blood (a), peripheral blood lymphocytes gated on live lymphocytes (with a forward scatter–side scatter gate); blood GFP⁺ (b), peripheral blood lymphocytes gated on GFP⁺ live lymphocytes (with FL1 and a forward scatter–side scatter gate). Data are representative of 5–15 mice per group in one to three separate experiments. (d) Proliferation of splenocytes and lymph node cells from CD3^WT and CD3^WTM retrogenic mice given drinking water containing BrdU for 8 d, assessed 4 weeks after transplant by staining for CD4 and BrdU and gating on CD4⁺ lymphocytes. Data are representative of two experiments with four to five mice per group. (e–h) Proliferation (e) and cytokine secretion (f–h) of cells from the brachial, cervical and mesenteric lymph nodes of CD3^WT and CD3^WTM retrogenic mice, assessed 6 weeks after transplant by being plated for 48 h as single-cell suspensions in 96-well round-bottomed plates. (e) Cells pulsed for a further 24 h with [³H]thymidine. (f–h) Analysis of cytokines in 50 l of the supernatant immediately after 48 h of culture. Data are the mean and s.e.m. of two to five mice per group in two separate experiments.

Full size image (123 KB)

If there were a failure of negative selection in CD3-mutant mice that developed autoimmunity, autoreactive T cells would be expected in the periphery of sick mice. We first assessed whether T cell proliferation was higher in CD3-mutant mice by in vivo labeling with the thymidine analog BrdU (5-bromodeoxyuridine). Whereas the percentage of CD3^WT and CD3^WTM BrdU⁺CD4⁺ T cells in the spleen was identical, CD3^WTM mice had 43% more BrdU⁺CD4⁺ T cells than CD3^WT mice had in lymph nodes draining the sites of inflammatory lesions (Fig. 5d). In addition, CD3^WTM T cells showed more ex vivo proliferation in response to lymph node–resident, self antigen–expressing antigen-presenting cells (Fig. 5e). These findings are notable, given that CD3-mutant T cells had a distinct proliferative defect (Fig. 3a,b). Finally, although CD3^WTM T cells produced minimal IL-10 and only twofold more IL-2 than CD3^WT T cells did, they secreted 100-fold more IFN- (Fig. 5f–h). These data collectively suggest that autoreactive T cells with lower CD3 signaling capacity escape negative selection, show enhanced proliferation in the periphery and produce substantial IFN- in response to antigen-presenting cells expressing self antigen.

Mice.

Rag1^{- /-} , CD3-KO and C57BL/6J mice were from Jackson Laboratories. CD3-KO mice, generated by the crossing of Cd3e^P/P mice (provided by C. Terhorst) with CD3-KO mice, have been described^{9, 10}. CD3-KO Tg1-6^M mice were from N. van Oers⁴⁴. Mice transgenic for the 3A9 TCR were from M. Davis⁴⁵; this HEL (48–62)-specific TCR is H-2A^k restricted but will mediate T cell development on the H-2^b background of the mice used in these experiments. Rag1^{- /-} , CD3-KO, CD3-KO Rag1^{- /-} , CD3-KO, 3A9.CD3-KO and CD3-KO Tg1-6^M mice were re-derived, bred and maintained as helicobacter-free, citrobacter-free, specific pathogen–free colonies at St. Jude. All animal experiments were done according to national, state and institutional guidelines in a helicobacter-free and specific pathogen–free facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. Animal protocols were approved by the Institutional Animal Care and Use Committee of St. Jude.

Generation of CD3 multicistronic vectors.

Constructs linked to the 2A peptide were generated by recombinant PCR and were cloned into pMIG, a mouse stem cell virus–based retroviral vector containing an internal ribosomal entry site–GFP cassette, as described^{9, 46, 47}. Additional restriction sites were introduced, and an additional construct was designed with substitution of each tyrosine with phenylalanine in all CD3-chain ITAMs. Subsequent CD3-mutant constructs were then generated by the appropriate subcloning.

Retrovirus-producing cells.

Cell lines producing retrovirus were generated as described^{48, 49, 50} with some modifications⁹. Human embryonic kidney 293T cells were transiently transfected with 4 g pMIG vector with or without sequences encoding TCR, together with packaging and envelope vectors, with the TransIt 293T transfection reagent (Mirus). GP+E86 cells, a mouse embryonic fibroblast 3T3–based packaging cell line, were transduced every 12 h for 3–4 d with virus and polybrene (6 g/ml) until a viral titer above 1 10⁵ viruses per ml after 24 h was obtained.

Retrovirus-mediated stem cell gene transfer.

Retroviral transduction of mouse bone marrow cells was done as described^{9, 11}. Bone marrow was collected from 8- to 10-week-old donor mice 48 h after treatment with 150 mg 5-fluoruracil (Pharmacia & UpJohn) per kg body weight. Bone marrow cells were cultured for 48 h in complete DMEM with 20% (vol/vol) FBS, mouse IL-3 (20 ng/ml), human IL-6 (50 ng/ml) and mouse stem cell factor (50 ng/ml; Biosource-Invitrogen). Cells were then cultured together for a further 48 h with irradiated (1,200 rads) retrovirus-producing cell lines plus polybrene (6 g/ml) and cytokines as described above. Nonadherent transduced bone marrow cells were collected and washed and were resuspended in PBS containing 2% (vol/vol) FBS plus heparin (20 U/ml). Bone marrow cells (4 10⁶ cells per mouse) were injected through the tail vein into irradiated (450 rads) Rag1^{- /-} recipient mice. After initial mice were noted to become sick and die within 5–7 weeks of transfer, all further experiments were completed when the mice were first noted to be sick (lethargy, hunched posture, ruffled coat, dehydration and squinting). Necropsy showed that the few sick wild-type mice had no inflammation and were sick because of an unrelated cause.

Additional methods.

Information on histology and immunofluorescence, flow cytometry and cell sorting, proliferation and cytokine measurements, T_reg cell assays, autoantigen array production and probing, and microarray data analysis is available in the Supplementary Methods online.

Accession codes.

UCSD-Nature Signaling Gateway (http://www.signaling-gateway.org): A000552, A000550, A000549 and A000553.

Note: Supplementary information is available on the Nature Immunology website.

Author contributions

All authors contributed to discussions of experimental design, data analysis and/or editing of the manuscript; J.H. did all experiments unless stated otherwise; H.W., K.D.E. and C.J.W. generated mice for the immunofluorescence microscopy and autoantibody analysis, did the T_reg cell experiments, chimera-lymphopenia experiments, some of the flow cytometry analysis and the BrdU and lymph node assays; K.L.B. did all the histological analysis and scoring while 'blinded' to sample identity; K.F. assisted in cellular isolation and analysis and managed the breeding colonies; Z.B. and R.S. did the immunofluorescence microscopy, H.S., A.C. and P.J.U. did the autoantibody analysis; N.S.C.v.O. provided the CD3-KO Tg1-6^M-transgenic mice and discussions about their use; J.H. and D.A.A.V. wrote the manuscript; and D.A.A.V. conceptualized and directed the project.