A Proposed Classification of the Immunological Diseases

The formal recognition and genetic understanding of the autoinflammatory diseases has defined mechanisms of self-directed inflammation that are independent of adaptive immunity.


Research in Translation
August 2006 | Volume 3 | Issue 8 | e297 A century ago Paul Ehrlich proposed that immune reactivity against self, which he called "horror autotoxicus" and which is now called autoimmunity, would be incompatible with life because of potentially devastating consequences for the host. But Ehrlich was proven wrong after the demonstration of autoantibodies and the emergence of a theoretical basis for autoreactivity [1]. Conceptually, autoimmunity is viewed as a defect of either B or T lymphocyte selection, with aberrant lymphocytic responses to autoantigens [2]. In recent years, an improved genetic understanding of both common and rare diseases, collectively associated with mutations refl ecting immune system perturbations-ranging from the thymus, to B and T cells, to T regulatory cells-has vindicated the autoimmunity paradigm [3] (Table 1).

Problems with the Concept of Autoimmunity
Nevertheless, there are several diffi culties with the autoimmunity concept when considering selfdirected tissue infl ammation. These diffi culties include a lack of major histocompatibility complex (MHC) and autoantibody associations in many diseases, tentatively labelled as autoimmune. A gradual appreciation of these diffi culties has led to revised defi nitions of autoimmunity, but this approach fails to defi ne when selfdirected tissue infl ammation is not autoimmune in origin [4].
And there is yet another weakness in the concept of autoimmunity: the idea that the immune system functions by making a distinction between self and nonself has come under scrutiny for failing to explain a number of fi ndings. For example, "Why do we fail to reject tumors, even when many clearly express new or mutated proteins? Why do most of us harbor autoreactive lymphocytes without any sign of autoimmune disease, while a few individuals succumb?" [5].
To answer these questions, Polly Matzinger proposed the "danger signal theory," which proposes that the immune system is not so much concerned with self/nonself discrimination but with mounting responses to danger signals, including exogenous pathogenic bacteria and endogenous damaged tissues [5]. However, the danger model does not account adequately for the exquisite specifi city of the adaptive immune responses in autoimmune diseases. This article draws on recent advances from genetic and molecular studies and improved clinical insights into disease in order to propose a unifi ed classifi cation and theoretical framework for all immunological diseases.

Box 1. Defi nitions of Autoimmunity and Autoinfl ammation Generic Defi nition of Autoimmunity
Self-directed infl ammation, whereby aberrant dendritic cell, B and T cell, responses in primary and secondary lymphoid organs lead to breaking of tolerance, with development of immune reactivity towards native antigens. The adaptive immune response plays the predominant role in the eventual clinical expression of disease. Organ-specifi c autoantibodies may predate clinical disease expression by years and manifest before target organ damage is discernible.

Proposal for a Defi nition of Autoinfl ammation
Self-directed infl ammation, whereby local factors at sites predisposed to disease lead to activation of innate immune cells, including macrophages and neutrophils, with resultant target tissue damage. For example, disturbed homeostasis of canonical cytokine cascades (as in the periodic fevers), aberrant bacterial sensing (as in Crohn disease), and tissue microdamage predispose one to sitespecifi c infl ammation that is independent of adaptive immune responses.
Paradoxically, the background to these discoveries is over a century old, with Eli Metchnikoff's seminal observations that described how phagocytic cells, rather than serum factors (or antibodies), were responsible for infl ammatory tissue reactions against foreign antigens.
Fifty years later came recognition of the clinical entities subsequently known as hereditary periodic fevers (HPFs) [6], which are now known to include tumour necrosis factor (TNF) receptor-associated periodic fever syndrome (TRAPS) [7], familial Mediterranean fever (FMF), hyperimmunoglobulinaemia D with periodic fever syndrome (HIDS), and several others ( Table 1). The key breakthrough came in Daniel Kastner's laboratory by using a candidate gene approach in families with a rare autosomal dominant HPF termed familial Hibernian fever, initially in the prototypic familial Hibernian fever family from Nottingham, as well as in a series of families drawn from both Europe and the United States. Mutations in the TNF1 receptor, which is widely distributed on both immune and nonimmune cells, were shown in six families. This led the authors to propose the term TNF receptor-associated periodic syndrome (TRAPS) and to coin the term autoinfl ammation, in recognition of an immunopathogenesis that was distinct from autoimmunity [7].
It now appears that TRAPS and other monogenic periodic fever disorders share a common thread. They all show disturbances in pathways associated with innate immune cell function, encompassing abnormal signalling in key cytokine pathways that include TNF and interleukin-1 (IL-1β) (via adaptor molecules collectively termed the infl ammasome [8]), as well as through mutations in proteins associated with bacterial sensing [9,10] (Table 2). Jérôme Galon and colleagues proposed that polygenic diseases sharing clinical features in common with the HPFs and lacking autoantibody or MHC associations could, by default, be termed autoinfl ammatory in nature [10]. Indeed, the recognition of innate immune-related factors at target sites of disease, rather than adaptive immunity, has led to the idea of classifying some conditions (such as Crohn disease and Behçet syndrome) as being autoinfl ammatory [10,11]. However, this classifi cation remains highly controversial, given that evidence for autoantibodies and autoimmune-like reactions is also a feature of these diseases [12]. Also, the logical consequence of this approach is a resulting two-tiered classifi cation for some, but not all, immunological diseases.

Autoimmunity versus Autoinfl ammation
The issues pertaining to immunological disease classifi cation are compounded by the absence This table lists some genes known to be associated with self-directed infl ammation, as well as their distribution and putative functions, where known. All autoimmune disease-associated mutations described so far have been in proteins associated with adaptive immune responses. These proteins are expressed in the thymus or in B and T cells and T regulatory cells. MHC polymorphisms have also been reported in the diseases commonly regarded as autoimmune in nature. Conversely, all mutations described thus far in the autoinfl ammatory diseases are in proteins that do not specifi cally play a role in adaptive immunity but are expressed on innate immune cells or with a more widespread nonimmune cell distribution. Collectively, these mutations are typically associated with episodic bouts of uncontrolled infl ammation. of a precise defi nition of what constitutes autoinfl ammation in the common polygenic diseases. This is in contrast to polygenic autoimmune disease, where a broad consensus on a generic defi nition exists (Box 1). Autoinfl ammation may simply be defi ned as self-directed tissue infl ammation, where local factors at disease-prone sites determine activation of the innate immune system. Such a defi c boundaries for what constitutes self-directed infl ammation. Furthermore, all immunological disease can then be conceptualised as being purely autoinfl ammatory or autoimmune, or being a combination of autoinfl ammatory-autoimmune mechanisms that variably interact in the phenotypic expression of disease ( Figure 1). Thus, the boundaries for autoimmunity are set by mutations associated with the monogenic autoimmune diseases, which show an increased propensity towards adaptive immune responses and which are recognisable by the presence of autoantibodies. On the other hand, the boundaries of autoinfl ammation are defi ned by mutations in cells or molecules involved in innate immune responses at disease-prone sites, where disease expression cannot be explained by autoimmune mechanisms.
For example, Crohn disease is the fi rst polygenic disease with a genetically defi ned autoinfl ammatory component, which was defi ned simultaneously by two groups who showed that the disease-associated mutation occurred in a protein involved in innate immune responses [13,14]. Specifi cally, the NOD2-associated mutations are thought to be linked to aberrant intracellular innate immune responses to bacterial peptidoglycan [15]. In addition to its expression on cells of the monocyte lineage, the NOD2 protein is also expressed on gut epithelial cells. Moreover, carriage of two copies of the NOD2 mutation is associated with sitespecifi c involvement of the ileum and severe stricturing disease [16].
Gout is the fi rst common polygenic condition with a molecular basis that is reminiscent of the monogenic autoinfl ammatory diseases. The causative urate crystals have a tendency for site-specifi c deposition in the joints, which may only periodically lead to infl ammation, despite the continuous presence of crystals [17]. At a molecular level, attacks of gout are associated with activation of the IL-1β signalling cascade, via the NALP3 infl ammasone, in a manner similar to some of the HPFs [18,19].

Clinical Studies That Helped Defi ne Autoinfl ammatory Diseases
Crohn disease is closely associated with the seronegative spondyloarthropathies, which include ankylosing spondylitis, reactive arthritis, and psoriatic arthritis. Indeed, most patients with ankylosing spondylitis have subclinical Crohn disease. Of course some of these disorders show striking human leukocyte antigen (HLA)-B27 MHC associations, unlike Crohn disease, and immune reactivity against self has long been suspected as an underlying immunopathogenetic mechanism [20]. However, recent magnetic resonance imaging studies have shown that early disease localisation in ankylosing spondylitis, reactive arthritis, and psoriatic arthritis is maximal at, and adjacent to, sites of relatively high shear and tensile forces • Psoriasis: anti-TNF therapy and efalizumab or alefacept, both of which are thought to act primarily by blocking T cell migration and activation.
In Group C, there is considerable heterogeneity in the extent of response to biological therapies in diseases such as RA and psoriasis. In these two conditions, both anticytokine and antilymphocyte strategies may be effective, which is in keeping with a signifi cant interplay between autoimmune and autoinfl ammatory components, as proposed in Figure 1. Future therapy development and an improved understanding of the basis for drug resistance in immunological diseases will be enhanced by considering the relative role of adaptive immunity and innate immune factors at target sites of disease.

Box 2. The Autoinfl ammatory-Autoimmune Continuum and Targeted Therapy in the Immunological Diseases
August 2006 | Volume 3 | Issue 8 | e297 Based on this defi nition, many common diseases with strong infl ammatory components could be classed as predominantly autoinfl ammatory in nature, although most of these conditions also have evidence for autoimmunity in the clinical setting ( Figure 1). Importantly, this defi nition of autoinfl ammation allows for the establishment of specifi nition would encompass autoinfl ammatory mechanisms across the spectrum of immunological disease. Indeed, several tissue-specifi c factors that could contribute to infl ammation have been recognised (Text S1).
at tendon and ligament insertions [21]. Bone infl ammation adjacent to insertions may be seen in all of these conditions [22]. However, when bone infl ammation is extensive, it is often associated with carriage of the HLA-B27 gene, suggesting that local factors determine the degree of activation of the adaptive immune response at certain predisposed sites [23].
Collectively, Crohn disease and the seronegative arthropathies are associated with acneform lesions, skin pustulosis, and occasionally multifocal osteitis. All of these clinical features are variably shared with two recently identifi ed monogenic autoinfl ammatory conditions: (1) pyogenic arthritis, pyoderma gangrenosum, and severe cystic acne (PAPA) syndrome and (2) a type of chronic multifocal recurrent osteomyelitis (CMRO) [24,25]. The molecular basis of these monogenic equivalents of more common polygenic clinical counterparts relates to mutations in proteins associated with innate immune cell functioning rather than with adaptive immunity. This suggests that tissue-specifi c factors in the bones, joints, or skin may lead to clinical disease expression at certain sites (Table 1).
In common with ankylosing spondylitis and psoriatic arthropathy, Behçet disease also has MHC class I associations. However, in contrast to ankylosing spondylitis and psoriatic arthropathy, Behçet disease has clinical features that seem to be mostly autoinfl ammatory in nature (Table  S1). Furthermore, particular variants of both the FMF gene (MEFV) and TNFRSF1A are more common in people with Behçet disease. There appears to be overlap between Behçet disease, infl ammatory bowel disease, and MEFV mutations in general, and Ahmet Gül has postulated that poorly defi ned tissue-specifi c factors in Behçet disease may eventually lead to the development of secondary autoimmune responses [26].
In the case of autoimmune diseases such as rheumatoid arthritis (RA), studies by Ai Lyn Tan and colleagues [27] and by Laura Rhodes and colleagues [28] showed that the degree of joint infl ammation, joint erosions, and therapeutic responses are variably affected by tissue-specifi c factors, including the position of joint ligaments. For instance, RA erosive changes are more pronounced adjacent to the site of maximal stress, as exemplifi ed in the index fi nger compared with the ring fi nger of the dominant hand [27]. These studies show how secondary autoinfl ammatory mechanisms contribute to the clinical expression of RA. Based on magnetic resonance imaging observations in infl ammatory arthritis, McGonagle and colleagues proposed a classifi cation whereby RA is viewed as the archetypal autoimmune-mediated synovitis, and the seronegative arthropathies are considered from the perspective of tissue-specifi c factors related to joint insertions [29]. The implications of this dichotomous classifi cation of joint disease can be extended to all immunological diseases.

Implications for Autoimmunity
The autoimmunity paradigm has dominated immunology for so long that our concepts of many disorders, including Crohn disease, have been moulded to fi t the prevailing dogma. Placing immunological disease along this proposed continuum allows the relative contribution of different types of self-directed infl ammation to be considered without assuming that an adaptive immune response is central to disease pathogenesis.
The case of vasculitis (blood vessel infl ammation) illustrates the usefulness of a continuum view of immunological disease. The autoimmune-mediated vasculitides can be distinguished clinically by the presence of pathogenic autoantibodies, including  The monogenic "autoinfl ammatory" diseases may be exclusively determined by local tissuespecifi c factors. For rare monogenic "autoimmune" conditions, the disease localisation appears to be determined predominantly by the adaptive immune response. The clinical heterogeneity within the immunological diseases, both among patients and between populations, may refl ect the variable expression of autoinfl ammatory and autoimmune factors in disease causation. For example, in humans, there is considerable genetic and molecular evidence for uveitis falling into all of the disease categories, with the exception of the rare monogenic autoimmune diseases. There is also considerable overlap between polygenic autoinfl ammatory diseases and MHC class 1-associated diseases, but to simplify classifi cation, these are split up into different categories. This fi gure does not include all immunologically recognised diseases because of their large number.
antineutrophil cytoplasmic antibodies (ANCA). It is of note, therefore, that the nonautoantibody-associated vasculitides, including Takayasu arteritis and giant cell arteritis, affect particular vascular territories in a patchy manner, thereby illustrating the contribution of local factors to disease pathogenesis (Table S1). Thus far, evidence for tissue-specifi c factors infl uencing the expression of autoimmune diseases in humans, such as type 1 diabetes, is lacking, but the concept of secondary autoinfl ammation in autoimmunity offers an alternative perspective on how genetic or environmental factors affecting disease-prone sites could lead to, or alter, clinical disease expression.
The classifi cation of MHC class I-associated diseases as autoimmune has been contentious given that these conditions lack specifi c autoantibody associations. The present classifi cation places HLA-B27 -related conditions and other MHC class I-associated diseases as intermediates-or "at a half-way house"-between autoinfl ammation and autoimmunity. As outlined earlier, tissue-specifi c factors at disease-prone sites appear to be instrumental in localisation of these conditions. While this article deals exclusively with selfdirected tissue infl ammation, immunereactivity reactions against nonself (such as reactions to organ transplants) help illustrate the concept of MHC class I-associated diseases being "halfway houses." In the transplantation fi eld, renal rejection reactions are strongly associated with MHC class I antigens and with tissue-specifi c factors, especially the duration of organ ischemia. In fact, if organ ischemia is minimised, then MHC-mismatched grafts survive as well as matched grafts, thus showing how adaptive immunity and local tissue factors interact in disease expression [30] in the context of MHC class I associations.
In certain clinically defi ned autoimmune scenarios, including RA, it appears that some patients do in fact have a disease that is predominately autoinfl ammatory in nature. In a study of patients with benign polyarthritis of the elderly, which often meets diagnostic criteria for RA and has a good prognosis but lacks the autoantibody association, McGonagle and colleagues observed that the pattern of disease localisation was similar to the seronegative arthropathies. In other words, joint disease tended to involve the periarticular structures rather than the synovium [31]. The good prognosis in benign polyarthritis is similar to reactive arthritis, which is a type of seronegative arthritis. Thus, the variable prognosis in diseases such as RA and multiple sclerosis may be related to the predominance of either autoinfl ammation or autoimmunity, with the former generally equating to a better prognosis.
The formal recognition of autoinfl ammation also has implications for a better clinical understanding of the targeted therapy of the immunological diseases. For example, anticytokine therapy is especially effective in autoinfl ammatory disease; the interleukin-1 receptor antagonist anakinra shows good effi cacy in some monogenic autoinfl ammatory disorders [32]. On the other hand, strategies to target lymphocytes are especially effective in autoimmune diseases, such as lupus. In some cases, both anticytokine and antilymphocyte strategies are effective in disorders that lie somewhere along the autoimmuneautoinfl ammatory disease continuum Landmark Papers That Set the Scene for Proposing an Autoimmune-Autoinfl ammation Spectrum

Burnet et al. [1]
The seminal work of Burnet and others set the scene for understanding the nature of autoimmunity.

McDermott et al. [7]
This article showed that some infl ammation directed against self was due to mutations in the TNF receptor and introduced the concept of autoinfl ammation. Since TNF is pivotal in innate immune responses, the work confi rmed that this disease process was very different from autoimmunity at the molecular level.

Hugot et al. and Ogura Y et al. [13,14]
Until this work, Crohn disease was conceptualised in relationship to autoimmune mechanisms. These studies were published simultaneously, and showed that mutations in a protein associated with innate immune responses played a key role in a subgroup of patients with Crohn disease.

Martinon et al. [18]
This paper showed that the molecular pathways associated with immune activation in gout and pseudogout were very similar to those associated with immune activation in some of the monogenic autoinfl ammatory diseases. This mechanistic link shows how multifactorial common diseases, without a clearly defi ned genetic basis, are linked to autoinfl ammation.

Matzinger [34]
The author argued that the danger theory stood on the shoulders of the self/nonself discrimination theory, and thus explained autoimmunity. However, the danger theory nicely illustrates the role of innate immune responses, which are independent of self/nonself discrimination, as a mechanism for selfdirected tissue infl ammation.  In early RA, joint disease localisation is to the synovium-in keeping with the concept of the synovium being the primary target organ. However, in early psoriatic arthritis, the infl ammatory changes have a widespread distribution and appear to relate to patterns of joint stressing around ligaments, adjacent bone, and soft tissues, rather than a specifi c antigen territory. The fi gure shows a contrastenhanced high-resolution magnetic resonance image of a distal interphalangeal joint optimised for showing sites of infl ammation (pixel size, 100 ×100 microns). There are extensive infl ammatory changes in all tissues. Asterisk, site of diffuse osteitis; arrowhead, synovial enhancement; solid arrows, joint ligaments that show fl orid infl ammatory changes at insertions and within ligaments; open arrow, extracapsular soft-tissue enhancement.
(Box 2). Finally, cytokine blockade of diseases with a strong autoinfl ammatory basis may aggravate or precipitate autoimmune diseases such as lupus. Placing infl ammatory disease along a continuum, therefore, may have relevance for therapy development, since disorders with a prominent autoinfl ammatory component could be targeted via innate immune pathway blockade.

Refi ning our Understanding of Immunology
The concept of autoinfl ammation has implications for our theoretical understanding of immunology. The two theories of self-directed immunity-the self/nonself discrimination theory and the danger signal theory-place the emphasis on different aspects of how autoimmunity develops and argue for different theoretical rationales for immunological disease. In physics, the nature of matter cannot be readily understood in terms of either particles or waves, so the wave-particle duality arose. Likewise, many aspects of classically recognised autoimmune diseases are best viewed in terms of self/nonself discrimination and, conversely, the autoinfl ammatory diseases are best viewed from the perspective of tissue-specifi c danger signals. For example, uric acid-the causative molecule in gout-is a recognised danger signal [33]. From the standpoint of the danger theory, the ultimate susceptibility to disease lies not with the adaptive immune system but with the target tissue itself, from which danger signals emanate [34]; this hypothesis is closely allied with the proposed generic defi nition of autoinfl ammation. Looking at immunological disease from the autoimmunity perspective, several groups have drawn attention to the possible role of tissue-specifi c factors in autoimmunity [35].
A better clinical understanding of diseases may be achieved by purposely looking for specifi c autoinfl ammatory and autoimmune features. For example, the idea of coeliac disease being autoimmune in nature is questionable given the exogenous nature of the causative gluten antigen. However, when viewed from the perspective of tissue-specifi c components, such as altered gut permeability following inciting infections and gut tissue transglutaminase-mediated gluten peptide modifi cation, then a unifying basis can be conceptualised [36]. Conversely, the idea that Crohn disease and ulcerative colitis are autoinfl ammatory in nature is also questionable given their associations with p-antineutrophil cytoplasmic autoantibodies [37]. It remains to be determined whether the autoantibody association represents secondary autoimmunity or disease heterogeneity, with some cases being predominantly autoimmune and others autoinfl ammatory.

Conclusions
The formal recognition and genetic understanding of the autoinfl ammatory diseases has defi ned mechanisms of self-directed infl ammation that are independent of adaptive immunity. If we adopt a "continuum model" of immunology, in which diseases lie on a spectrum from autoimmune to autoinfl ammatory, we can begin to defi ne the relative contributions of both the innate and the adaptive immune responses to particular diseases. All noninfectious infl ammatory disease can be accommodated within this classifi cation.
Animal models of autoimmune disease have been very instructive for elucidating molecular pathways of many conditions. However, in order to adequately assess the role of individual tissue-specifi c factors, studies will need to focus on the role of site-specifi c factors in humans. Future clinical studies are needed to develop imaging strategies, including molecular imaging, as well as to determine the basis for infl ammation at certain predisposed sites ( Figure  2). Studies of tissues that are subject to autoinfl ammatory reactions in diseases such as multiple sclerosis need to explore autoinfl ammatory mechanisms in disease expression that have been neglected to date. The autoinfl ammatory-autoimmune continuum offers an inclusive classifi cation of immunological disease and a better understanding of the pathogenesis and treatment of selfdirected infl ammation.  The danger signal theory of Matzinger, with tissue-specifi c factors determining disease localisation

Supporting Information
The major factor determining disease is aberrant SNS discrimination, with breakdown of immunological tolerance Immunological basis Genetically related to perturbations of innate immune function, including pro-infl ammatory cytokine signalling abnormalities/ bacterial sensing/local tissue abnormalities Acquired immune perturbation key-to-disease expression Cellular basis Expression determined by cells of innate immune system, including neutrophils and macrophages or nonimmune cells Expression mainly determined by factors affecting B and T cell activity Genetic mutations in HPFs, including TRAPS and FMF, affect these cells Genetic mutations in rare autoimmune diseases affect these cells or their selection in thymus DOI: 10.1371/journal.pmed.0030297.t002 This table represents some of the key features that allow differentiation of a "pure autoinfl ammatory disease" from a "pure autoimmune disease." The rare monogenic HPFs are the prototypic autoinfl ammatory diseases, whereas the prototypes for autoimmune diseases include the polygenic MHC and autoantibody-related diseases, as well as some rare monogenic diseases. SNS, self/nonself.