Self-Organized Criticality Theory of Autoimmunity

Background The cause of autoimmunity, which is unknown, is investigated from a different angle, i.e., the defect in immune ‘system’, to explain the cause of autoimmunity. Methodology/Principal Findings Repeated immunization with antigen causes systemic autoimmunity in mice otherwise not prone to spontaneous autoimmune diseases. Overstimulation of CD4+ T cells led to the development of autoantibody-inducing CD4+ T (aiCD4+ T) cell which had undergone T cell receptor (TCR) revision and was capable of inducing autoantibodies. The aiCD4+ T cell was induced by de novo TCR revision but not by cross-reaction, and subsequently overstimulated CD8+ T cells, driving them to become antigen-specific cytotoxic T lymphocytes (CTL). These CTLs could be further matured by antigen cross-presentation, after which they caused autoimmune tissue injury akin to systemic lupus erythematosus (SLE). Conclusions/Significance Systemic autoimmunity appears to be the inevitable consequence of over-stimulating the host's immune ‘system’ by repeated immunization with antigen, to the levels that surpass system's self-organized criticality.


Introduction
Since 'clonal selection theory of immunity' of F. Macfarlane Burnet and subsequent molecular biological discoveries on V(D)J recombination and the diversity and individuality of immune response, how autoimmunity arises remains unclear. Apart from the term 'autoimmunity' which is now ready-made, in the present study, we tried to see the pathogenesis of autoimmunity from different angle and test the integrity of immune 'system'. The method we have chosen was to stimulate the system maximally by antigen to the levels far beyond its steady-state just like testing the capability of automobile. In a perfectly reproducible experiments in which the mice not prone to autoimmune diseases were immunized repeatedly with antigen, we have unexpectedly and surprisingly discovered that overstimulation of immune system beyond its self-organized criticality inevitably leads to systemic autoimmunity. Subsequent detailed molecular analyses revealed in the first that autoantibodies are induced not by cross reaction to antigen but by de novo T cell receptor (TCR) revision. Second, final maturation of effector cytotoxic T lymphocyte (CTL) via antigen cross-presentation is sine qua non for generating autoimmune tissue injury. Most importantly, we now show that autoimmunity arises not from 'autoimmunity', but as a natural consequence of normal immune response when stimulated maximally beyond system's self-organized criticality.

Induction of Autoantibodies
Consistent with the common observation that T cells become anergic after strong stimulation with antigen [1], we observed that 26 immunization with staphylococcus enterotoxin B (SEB) caused SEB-reactive Vb8 + CD4 + T cells from BALB/c mice to become anergized. However, these cells recovered from anergy to divide and produce IL-2 after further immunization 86 with SEB ( Figure S1A). This was accompanied by the induction of autoantibodies, including IgG-and IgM-rheumatoid factor (RF), anti-Sm antibody, and in particular, RF reactive against galactose-deficient IgG, typically found in human autoimmunity [2] ( Figure 1A). Autoantibodies can also be induced by other conventional antigens, including ovalbumin (OVA) or keyhole limpet hemocyanin (KLH) ( Figure S2) as long as immunizing antigen is correctly presented to T cells ( Figure  S1B). CD4 + T cells of repeatedly-immunized mice become fully matured, expressing CD45RB lo , CD27 lo and CD122 hi (data not shown), and these primed CD4 + T cells can confer RF generation in naïve recipients following adoptive transfer ( Figure 1B). The induction of autoantibodies is independent of CD8 + T cells or MHC class I-restricted antigen presentation for the following reasons. First, both RF and anti-dsDNA antibody can be consistently induced upon repeated immunization of b 2 -microglobulin (b 2 m)-deficient BALB/c mice with OVA. b 2 m-deficient mice are deficient in CD8 + T cells, which are reduced to ,0.8% of splenic T cells [3] ( Figure S3). Second, the ability to induce autoantibodies was transferable from OVA-immunized BALB/c mice to b 2 m-deficient mice solely via CD4 + T cells ( Figure 1C). Thus, CD4 + T cells from repeatedly-immunized mice acquire the ability to induce autoantibodies. We refer to these as autoantibody-inducing CD4 + T (aiCD4 + T) cells in this communication.

Mechanism of Autoantibody Induction
To further clarify the characteristics of aiCD4 + T cells, we examined their TCR repertoire by spectratyping of their complementarity determining region 3 (CDR3) [4]. Combinatorial assessment of Vb and Jb showed that the CDR3 length profiles of CD4 + splenocytes in mice immunized either 86with PBS or 26 with SEB fit a normal Gaussian curve, typical of a diverse and unbiased TCR repertoire ( Figure 2A). However, splenocytes, but not thymocytes, from mice immunized 86 with SEB showed skewed length profiles, suggesting that TCR revision was in progress at periphery of the spleen. Genes encoding components of the V(D)J recombinase complex were specifically re-expressed in mice immunized 86 with SEB, including the recombinationactivating genes 1 and 2 (RAG1/2), terminal deoxynucleotidyl transferase (TdT) and surrogate TCRa chain (pTa) [5] ( Figure 2B). The RAG1 gene is expressed in vivo after immunization 86 with SEB in rag1/gfp knock-in mice [6] ( Figure 2C). In these mice, serum RF was increased in conjunction with an increase of GFPexpressing Vb8 + CD4 + T cells in the spleen. To directly prove that V(D)J recombination took place at the periphery in spleen, we used ligation-mediated PCR (LM-PCR) to detect blunt-end DNA fragments harboring a rearranged coding V region flanked by recombination signal sequences (RSS) [7,8]. We identified rearranged intermediates corresponding to the TCRa variable region 2 (TCRAV2) in the splenocytes of mice immunized 86 with SEB ( Figure 2D). These findings indicate that repeated immunization with conventional antigen can induce the generation of aiCD4 + T cells which have undergone TCR revision and are capable of stimulating B cells [9]. This observation is in line with previous findings showing that such somatic mutations occur often in lymphocytes, a process which is considered to be a major stochastic element in the pathogenesis of autoimmunity [10,11]. Thus, overstimulation of CD4 + T cells by repeated immunization with antigen and induction of full maturation inevitably leads to the generation of aiCD4 + T cells which have undergone TCR revision and are capable of inducing autoantibodies. Importantly, the present study shows that such aiCD4 + T cells are induced by de novo TCR revision but not by cross-reaction to antigen.

Induction of Autoimmune Tissue Injury
Repeated immunization with OVA can also lead to autoimmune tissue injury and the production of autoantibodies reactive against IgG, Sm and dsDNA ( Figure 3 and Figure S2A). Serum immune complex (IC), proteinuria, and the deposition of IC and OVA in the kidney were noted in mice immunized 126with OVA ( Figure 3A). Typical diffuse proliferative glomerular lesions were seen in the kidney, and these glomeruli were infiltrated with CD8 + T cells. These observations resemble the clinical features observed in lupus patients, who typically exhibit an increase in CD8 + T cells in the peripheral blood and infiltration of CD8 + T cells in kidney [12,13]. Immunization of mice 126 with OVA led to reexpression of the V(D)J recombinase complex and enlargement of the spleen ( Figure S4A), and an increase in anti-dsDNA antibody, which is uniquely linked to autoimmune tissue injury in lupus nephritis [14] ( Figure S2A). Pathological findings included diffuse membranous (wire-loop) and/or proliferative glomerulonephritis in the kidney ( Figure 3A), infiltration of plasma cells around hepatic bile ducts ( Figure S4B), enlarged lymphoid follicles with marked germinal center in spleen ( Figure S4B), occasional lymphocyte infiltration into the salivary glands (data not shown), lymphoid cell infiltration into the thyroid, and perivascular infiltration of neutrophils and macrophages into the skin dermis of the auricle ( Figure S4B). The lupus band test, diagnostic of SLE, was positive in the skin at the epidermal-dermal junction ( Figure 3B).

Mechanism of Autoimmune Tissue Injury
It has been shown previously that IFNc is increased in association with autoimmune tissue injury [15][16][17]. Consistent with this, we found that the number of IFNc + CD8 + T cells, but not regulatory T or T helper 17 cells, was increased following immunization 126 with OVA ( Figure 4A and data not shown). We also observed an expansion of IFNc-producing effector/ memory CD8 + T cells, which are necessary for adaptive immunity [18] ( Figure 4A). These IFNc-producing CD8 + T cells were observed to have infiltrated into OVA-deposited glomeruli of OVA-immunized mice ( Figure 3A). CD8 + T cells are required for tissue injury based on the following observations. First, the transfer of CD8 + T cells can induce renal lesions in mice ( Figure 4B), as well as the generation of new IFNc + CD8 + T cells in the spleens of recipient mice following cell transfer ( Figure S5). Second, autoimmune tissue injury is not induced by the transfer of CD8 + T cells from OVA-immunized wild-type mice into b 2 m-deficient mice ( Figure 1C). And finally, CD8 + T cell transfer must be accompanied by at least a 16 booster immunization with OVA to induce autoimmune tissue injury in the recipient mice ( Figure S6). The findings indicate that full-matured, IFNc-producing effector CD8 + T cells are required for the induction of autoimmune tissue injury, provided that the relevant antigen is correctly presented on the target organs. These are well-established characteristic of CTL and not novel. We show, however, that (i) CTL is induced through an immune, but not 'autoimmune', process, and that (ii) autoimmune tissue injury inevitably occurs when CD8 + T cells are overstimulated to become matured effector CTLs. The latter means that regardless of how CTL is induced, the consequence of CTL over-induction is immune tissue injury.

Antigen Cross-Presentation
We next show that antigen cross-presentation is required for the induction of CTL and tissue injury. To test this, we co-cultured OVA-pulsed dendritic cells (DC) from mice immunized 126 with OVA together with T cells from OVA-TCR transgenic DO11.10 mice exclusively expressing OVA-reactive TCR [19]. We show that OVA-reactive DO11.10 CD8 + T cells are activated upon coculture with OVA-pulsed DCs ( Figure 4C and Figure S7). Further, autoimmune tissue injury and the increase in IFNc + CD8 + T cells, but not of autoantibody generation, were both abrogated by adding chloroquine (CQ), an inhibitor of antigen cross-presentation ( Figure 4C). This indicates that antigen cross-presentation is required for the expansion of IFNc-producing CD8 + T cells and autoimmune tissue injury.
aiCD4 + T Cell Helps CD8 + T Cell to Induce Tissue Injury Since CTL appear to play a rather passive role in autoimmunity, we next studied whether or not aiCD4 + T cell help is required for the induction of autoimmune tissue injury. Since anti-CD4 treatment almost abrogates generation of IFNc-producing CD8 + T cell and autoimmune tissue injury in OVA-immunized BALB/c mice ( Figure S8), to test whether this CD4 + T cell-mediated help is mediated by aiT cells or antigen-specific T cells, we have transferred CD4 + T cells from mice immunized 126 with KLH into CD4 + T-depleted BALB/c mice immunized 86 with OVA ( Figure 4D). Because full-matured IFNc + CTLs do not develop with less than 86 immunization with OVA ( Figure S9), this experiment can test the ability of aiCD4 + T cells that have undergone TCR revision to promote the maturation of OVAspecific CTL. The result showed that both autoimmune tissue injury and OVA-specific IFNc + CD8 + T cells arose in these mice after transfer, indicating that aiCD4 + T cells with de novo TCR revision are required for the maturation of CD8 + T cell and autoimmune tissue injury ( Figure 4D).

Discussion
The present findings are consistent with the current consensus that CD4 + T cells normally die via activation-induced cell death (AICD) after repeated exposure to a single antigen, while naïve CD4 + T cells having a 'cross-reactive' TCR with lower affinity can be activated through repeated exposure to the same antigen and survive due to weak TCR signaling, ultimately acquiring autoreactivity [20]. We show here, however, that aiCD4 + T cells are induced not by cross-reaction, but by de novo TCR revision. The aiCD4 + T cells thus generated induce not only autoantibodies but also full-maturation of CD8 + T cells leading to autoimmune  . Subsets of CD8 + T cells categorized into naïve (CD44 low CD62L high ), effector (CD44 high CD62L low ), and memory (CD44 high CD62L high ) fractions (middle). Flow cytometry of IFNc + cells within naïve or effector/memory CD8 + T cell populations. Spleen cells were separated into naïve (CD44 low ) and effector/memory (CD44 high ) cells using CD44 MACS beads, and IFNc + cells within the CD8 + T population was evaluated (lower). (B) Adoptive transfer of splenocytes of OVA-immunized BALB/c mice into naïve recipients. The recipients were injected with 500 mg OVA 24 h after cell transfer, and proteinuria examined 2 weeks later. (C) Cross-presentation of OVA to CD8 + T cells. Splenic CD11c + DC from OVA-immunized or control mice were incubated in the presence (OVA(+)) or absence (OVA(2)) of 1 mg/ml OVA with or without chloroquine (CQ) (20 mg/ml) for 3 h, followed by a co-culture with KJ1-26 + CD8 + T cells of DO11.10 transgenic mice for 24 h to examine surface expression of CD69 (upper). Inhibition of cross-presentation in vivo by administration of 250 mg CQ per mouse 3 h prior to immunization with OVA or PBS. IFNc + CD8 + T cells (middle), autoantibodies and proteinuria (lower) after 126 immunization. (D) Requirement of autoantibody-inducing CD4 + T cells for CD8 + T cell-mediated autoimmune tissue injury. BALB/c mice were immunized 126with KLH, and CD4 + T cells were isolated using MACS beads. Cells were transferred into the anti-CD4 antibody-treated recipient mice immunized 86with OVA. Percent matured CTL, i.e., IFNc + CD8 + T cells, and proteinuria were measured 2 weeks after booster immunization 16 with KLH. doi:10.1371/journal.pone.0008382.g004 tissue injury akin to human SLE. Thus, induction of aiCD4 + T cells is a critical step, and subsequent induction of effector CTL is a critical next step in the development of autoimmunity [21,22]. The question of how autoimmunity is triggered can therefore be deduced to the quantitative response of host against immunizing antigen, i.e., the ability of host to present and/or cross-present antigen. It then follows that the ability of certain antigens such as measles virus to cause autoimmunity may be due to their ability, in conjunction with its ability to present antigen, to overstimulate CD4 + and/or CD8 + T cells of certain hosts beyond integrity of their immune system. Living organisms are constantly exposed to a broad range of environmental antigens, as exemplified by the recent re-emergence of measles virus infection among a subpopulation of Japanese young adults who were not vaccinated against the virus. We therefore conclude that systemic autoimmunity necessarily takes place when host's immune 'system' is overstimulated by external disturbance, i.e., repeated exposure to antigen, to the levels that surpass system's self-organized criticality, and propose here 'self-organized criticality theory' explaining the cause of autoimmunity.
For adoptive cell transfer, B, T, CD4 + T and CD8 + T cells were isolated from spleens to .90% purity using MACS beads (Miltenyi Biotec, Germany). The cells were transferred into naïve BALB/c or b 2 m-deficient mice via i.p. (5610 6 /mouse) or i.v. (2.5610 7 /mouse) injection. The recipients received a single i.p. injection of 25 mg SEB or 500 mg OVA 24 h after cell transfer, and sera, urine and organ of recipients were studied 2 weeks afterwards.
BALB/c mice were injected i.p. with 200 mg anti-CD4 antibody (GK1.5; BioLegend) to deplete CD4 + T cell 24 h after immunization 86 with OVA. Four days later, CD4 + T cells from mice immunized 126 with KLH were transferred to the CD4 + T-depleted mice. The recipient mice received a single i.p. injection of 100 mg KLH 24 h after the cell transfer.
To inhibit cross-presentation, mice were immunized in vivo with 250 mg of chloroquine (Sigma) 3 h prior to immunization with 500 mg OVA or PBS every 5 d. Presence of autoantibodies was analyzed 2 d after each immunization, and proteinuria and IFNc + CD8 + T cells were examined 9 d after the final immunization.

Statistical Analysis
Statistical analyses were performed using Student's t test, and the data are expressed as the mean 6 SD. 3.4261.02% in peripheral blood mononuclear cells (PBMC) 9 d after 3rd treatment with anti-CD4 Ab. (B) Mice were immunized 126 with OVA with or without adding anti-CD4 antibodies, and the number of IFNc + cells within the CD8 + T population (upper and lower left) (mean 6 SD, 5 mice/group) and proteinuria (lower right) were evaluated. Found at: doi:10.1371/journal.pone.0008382.s008 (2.02 MB TIF) Figure S9 Study on the requirement of autoantibody-inducing CD4 + T cells for autoimmune tissue injury. Neither OVA-specific matured IFNc + CD8 + T cells or autoimmune tissue injury were observed until BALB/c mice were immunized at least 106 with OVA. The percent splenic IFNc + CD8 + T cells (left, mean 6 SD, 4 or 5 mice/group) and proteinuria (right) were examined after immunization 66, 86, 106 and 126 with OVA. Found at: doi:10.1371/journal.pone.0008382.s009 (0.67 MB TIF)