Abstract
Macrophage activation syndrome (MAS) is a potentially fatal complication of rheumatic diseases. The condition is considered part of secondary hemophagocytic lymphohistiocytoses (HLH). There are similarities in genetic background, pathogenesis, and clinical and laboratory features with primary HLH (p-HLH). We describe findings in mouse models of secondary HLH, comparing them with models of p-HLH and the cellular and molecular mechanisms involved, and relate them to recent findings in patients with secondary HLH. A multilayer model is presented in which background inflammation, infections, and genetics all contribute in different proportions and in several ways. Once the “threshold” has been reached, inflammatory cytokines are the final effectors, independent of the interplay between different upstream pathogenic factors.
- MACROPHAGE ACTIVATION SYNDROME
- JUVENILE IDIOPATHIC ARTHRITIS
- HEMOPHAGOCYTIC LYMPHOHISTIOCYTOSES
- ETIOLOGY
The term macrophage activation syndrome (MAS) is commonly used to identify a potentially life-threatening complication of rheumatic diseases1,2,3. It usually occurs in the context of systemic juvenile idiopathic arthritis (s-JIA), but also, albeit more rarely, in adult-onset Still disease, systemic lupus erythematosus, or Kawasaki disease.
MAS is characterized by high fever, lymphoadenomegaly, hepatosplenomegaly, pancytopenia, and coagulopathy, and may proceed to multiple organ failure. Biochemical abnormalities include increased hypofibrinogenemia, hypertriglyceridemia, and increased levels of soluble CD25, ferritin, and inflammatory cytokines.
Although the term MAS is commonly used, particularly by rheumatologists, the condition has recently been linked to secondary hemophagocytic lymphohistiocytosis (HLH). Indeed, similarities in clinical and laboratory features, genetic background, and pathogenesis with primary HLH are becoming more evident. The term rheumatologic-HLH (rheuma-HLH or r-HLH) has been recently proposed4.
Primary HLH (p-HLH) is caused by mutations of genes coding for proteins involved in granule exocytosis, including perforin, Munc13-4, syntaxin11, and Munc18-25. In the absence of a genetic cause or familial inheritance, secondary HLH occurs in response to infections, associated with malignancies (particularly lymphoma), or in the context of rheumatic disease (MAS/r-HLH). Studies in patients with p-HLH, as well as in murine models of p-HLH6,7,8,9,10,11, support the hypothesis that a defective cytotoxic CD8+ T cell response and an alteration of the antigen-presenting cell (APC)–CD8+ T cell crosstalk leads to defective silencing of the immune response and abnormal T cell activation2,3,6,9,12. The end result of this process is uncontrolled production of proinflammatory cytokines by macrophages and T lymphocytes, leading to organ damage and hematologic alterations13,14. Indeed, characteristic of HLH, although not necessary for the diagnosis, is the presence of macrophages phagocyting blood cells in liver or bone marrow15.
In this review, we describe findings in mouse models of secondary HLH, comparing them with models of p-HLH and the cellular and molecular mechanisms involved, and relate them to recent findings in patients with secondary HLH.
Animal models of primary and secondary HLH
Murine models of p-HLH have been generated by gene deletion or gene mutation of the murine orthologs of the genes involved in human p-HLH, including perforin, Munc13-4, and Rab27a6,7,8,10,11. Perforin deleted (prf−/−) mice are the most studied. These mice do not spontaneously develop HLH-like symptoms, but they show features characteristic of HLH after infection with a noncytopathic virus, lymphocytic choriomeningitis virus (LCMV)6. They develop fever, splenomegaly, pancytopenia, hypertriglyceridemia, and hypofibrinogenemia, and die within 2 weeks. Histological analysis reveals activated lymphohistiocytic infiltrates, bone marrow hypoplasia, and hemophagocytosis. Serum cytokine profiles are also very similar to those of humans with p-HLH. Mice harboring nonfunctional mutations of Unc13d (murine ortholog of the human Munc13-4) behave normally in pathogen-free conditions; upon infection with LCMV they develop an HLH-like phenotype with reduced platelet counts, splenomegaly, neutrophilia, and macrophage infiltration with bone marrow hemophagocytosis7. After LCMV infection, mice deficient for Rab 27a, a protein that colocalizes with Munc13-4, develop a mild decrease in hemoglobin and in platelet counts, neutrophilia (and not neutropenia), increased ferritin, splenomegaly, and liver hemophagocytosis, with negligible lethality8.
These models based on genetic alterations in the granule exocytic pathway are models of p-HLH and therefore represent tools to investigate cellular and molecular mechanisms of diseases caused by genetic mutations. However, the findings may not be directly translated to the setting of secondary HLH triggered by infections or presenting in the context of rheumatic diseases. Four models of secondary HLH have been published in the last year, of which 2 are triggered by living pathogens (Table 1).
A model of secondary HLH caused by Epstein-Barr virus (EBV) infection has been obtained using humanized mice. Human hematopoietic CD34+ cells are introduced into immunodeficient mice, thus rebuilding a “humanized” immune system. Infection of these mice with EBV induces a condition resembling EBV-associated HLH in humans, with thrombocytopenia and hemophagocytosis. By 10 weeks postinfection, two-thirds of the infected mice die after exhibiting high and persistent viremia, leukocytosis, anemia, and thrombocytopenia16. Chronic typhoid fever following infection with Salmonella enterica has been proposed as a model of secondary HLH17. In this experimental model, an acute response is followed by a chronic inflammatory disease characterized by anemia, thrombocytopenia, reduced lymphocytes, increased neutrophils and monocytes, hypofibrinogenemia, hyperferritinemia, fever, and hepatosplenomegaly. Hemophagocytic macrophages were found in bone marrow and spleen.
A third model of secondary HLH has been described, in which normal mice develop an HLH-like syndrome upon repeated stimulation with the dinucleotide CpG. CpG mimics bacterial and viral DNA and is a synthetic ligand of Toll-like receptor-9 (TLR-9)18. This model of secondary HLH caused by infections is based on repeated CpG administration in genetically normal mice in the absence of underlying chronic inflammation. It is characterized by leukopenia, anemia, thrombocytopenia, splenomegaly, hyperferritinemia, and liver necrosis due to microthrombosis. Neither mortality nor hemophagocytosis was observed. However, in the absence of interleukin 10 (IL-10), these mice show hemophagocytosis and die, supporting the antiinflammatory role for IL-10 in this setting.
Lastly, a model of secondary HLH has been generated in IL-6 transgenic mice (IL-6TG)19. IL6TG mice express high levels of IL-6 and do not present obvious signs of inflammatory disease. However, when challenged with TLR ligands, they show increased mortality, with increased ferritin and soluble CD25, as well as decreases in platelet, neutrophil counts, and hemoglobin levels; hemophagocytosis was not found. This model provides information about the development of HLH triggered by TLR ligands mimicking infections, in the presence of high levels of IL-6 (a cytokine that is markedly increased in patients with s-JIA20), in mice with normal cytotoxic pathways. Therefore, these mice represent a model of secondary HLH in the context of rheumatic disease (MAS/r-HLH).
Cellular agents
Cell types that have been thought to play a pathogenic role in HLH include CD8+ T lymphocytes, natural killer (NK) cells, and macrophages.
Murine models of p-HLH provide clear evidence for a role of CD8+ T lymphocytes in p-HLH. In prf−/− mice, depletion of CD8+ T lymphocytes stops disease activity6. In models of p-HLH, CD8+ T cells producing high levels of interferon-γ (IFN-γ) are found6,7,8. An increase in viral antigens by APC appears to be responsible for the abnormal CD8+ T lymphocyte activation9. To date, no definitive information is available on the role of CD8+ T cells in the salmonella and EBV humanized mouse models of secondary HLH or in the IL-6TG model of MAS/r-HLH17,19. However, a marked increase in CD8+ T cells is reported in the EBV humanized mouse model16. CD8+ T lymphocytes are increased only marginally, and more importantly, are dispensable in the repeated CpG model of secondary HLH18. This difference suggests that the involvement of CD8+ T cells might be different in p-HLH versus secondary HLH. However, IFN-γ-producing CD8+ T lymphocytes have been found in the livers of children with HLH secondary to infections or MAS/r-HLH, supporting the hypothesis of a pathogenic role of these cells in humans with secondary HLH21. Studies of CD8+ T cell activity from patients with MAS/r-HLH did not yield clear abnormalities. Therefore, the role of T cells in MAS/r-HLH is far from being defined and this may represent a difference compared to p-HLH.
The pathogenic role of NK cells in all forms of HLH remains unclear. NK cells are dispensable for HLH development in prf−/− mice and in the repeated CpG model of secondary HLH6,18. However, in a murine experimental setting of IFN-γ-increased serum levels, NK cell differentiation was impaired, suggesting that abnormal IFN-γ levels, which can be found in many HLH murine models, may block NK cell maturation22. Defective NK cytotoxicity is a hallmark of p-HLH, being the consequence of mutations of genes involved in cytotoxic pathways. An imbalance of NK cell subpopulations (decrease in the CD56+ subset) and a defective cytotoxic activity have been found in some s-JIA patients with MAS/r-HLH3,23,24. Reduced levels of perforin expression have been reported in NK cells from patients with s-JIA who are susceptible to MAS/r-HLH24. NK cell number and cytotoxic activity are reduced during MAS/r-HLH, and this was related to impaired response to IL-1825,26. Although IL-18 enhances NK functions, such as response to TLR and IFN-γ production, high and persistent levels of this cytokine were found to be correlated with increased NK cell apoptosis in systemic autoimmune diseases, including rheumatoid arthritis27. Taking these observations together, it is not clear whether abnormalities of NK cells during MAS/r-HLH play a primary role or whether they are secondary to high levels of cytokines and other abnormal stimuli. Moreover, it is unclear whether and how other functions of NK cells, such as cytokine production (i.e., IFN-γ), are affected during MAS/r-HLH.
Although macrophage hemophagocytosis is considered highly specific for MAS/r-HLH, the pathogenic role of macrophages is a matter of debate. Macrophages have been proposed to have a causal role, to be secondarily activated, or even to have antiinflammatory activity2. Although cells staining positive for CD68, a macrophage cell marker, are increased during the development of HLH in prf−/− mice, a causal role of macrophages in the pathogenesis of HLH was not established in this experimental model6. On the other hand, nonlymphoid cells are highly expanded and may play a role in the production of IFN-γ in the repeated CpG model of secondary HLH18. Macrophages have been demonstrated to produce IFN-γ under different kinds of stimuli28,29. Macrophages producing TNF-α and IL-6 have been found in livers of patients with HLH secondary to infections or MAS/r-HLH21. A characteristic of MAS/r-HLH is the presence of macrophage-expressing CD163, which is traditionally correlated with antiinflammatory scavenger activity of fully differentiated macrophages30. Liver biopsies of patients with HLH secondary to infections or MAS/r-HLH also show infiltrating CD68+ macrophages, which are often considered highly proinflammatory21. Immunohistochemistry of bone marrow aspirates from patients with MAS/r-HLH in s-JIA shows similar numbers of CD68+ and CD163+ macrophages (Bracaglia and Strippoli, unpublished data). However, the phagocytic activity and production of inflammatory cytokines by macrophages from patients with MAS/r-HLH in s-JIA have not been reported. The observation of a synergy in monocyte/macrophage activation between IL-6, a cytokine involved in s-JIA, and TLR ligands in inflammatory cytokine production19 (Strippoli, et al, unpublished data) supports the pathogenic role of macrophages in the generation of an exaggerated inflammatory response contributing to the development of overt r-HLH.
While CD8+ T lymphocytes appear to play a major role in p-HLH, murine models of secondary HLH, as well as the available data from patients with MAS/r-HLH, do not provide clear evidence of a role of NK cells or CD8 T lymphocytes. However, defective cytotoxicity, as shown by decreased NK activity in patients with r-HLH, may suggest a role for these 2 cell types, but more studies are required to clarify their contribution.
A multilayer model of MAS/r-HLH pathogenesis
Based on the data from animal models of primary and secondary HLH and from patients with MAS/r-HLH, it is reasonable to hypothesize that multiple factors, at different levels, contribute to the development of MAS/r-HLH in patients with s-JIA (Figure 1).
High disease activity may contribute to hyperresponse to an infectious trigger (as occurs in a significant proportion of patients with MAS/r-HLH). Evidence in vitro and in vivo points to a major role for high levels of IL-619,31, at least for TLR ligands. In general, genetic predisposition to high response to pathogen-associated molecular patterns by pattern recognition receptors might be involved. High disease activity may also contribute to functionally defective cytotoxic activity. Albeit with some interpatient variability, defective perforin expression, defective NK cytotoxic activity, and/or defective NK cell numbers have been shown in patients with active disease with or without active MAS/r-HLH23,26. A potential role of increased IL-18 level and abnormal IL-18 receptor response has been suggested26,32,33.
Concerning genetic background, recent evidence suggests a potential role of the same genes involved in p-HLH. One study showed heterozygosity for low-penetrance variants of perforin in some patients with MAS/r-HLH24; another reported the association between a specific Munc13-4 haplotype and development of MAS/r-HLH in patients with s-JIA34. To our knowledge, no information is available on other genes involved in p-HLH. One could hypothesize that hypomorphic variants in each of these genes, as well as in other unknown genes of the cytotoxic pathways, may contribute. The genetic background may also affect the intensity of the response to TLR ligands. A recent report demonstrated an association of an interferon regulatory factor-5 (IRF-5) polymorphism with susceptibility to MAS/r-HLH in Japanese patients with s-JIA35. IRF-5 is one of the transcription factors that mediate TLR signaling. Although this observation needs confirmation in white subjects, it points to an additional potential contribution of the genetic background.
When one considers the entire spectrum of HLH, the relative contribution of the factors described above may be variable in different forms (Figure 2). In p-HLH, loss of function mutations are sufficient to cause development of HLH even in the apparent absence of infectious triggers. Most deaths caused by the H5N1 avian influenza are associated with HLH features, suggesting that some infectious triggers might be sufficiently strong to cause HLH, with little, if any, contribution from genetics or from a preexisting inflammatory status4. Other infectious triggers are known to be associated with development of HLH (i.e., Leishmania, EBV); however, this occurs in a small percentage of infected patients, suggesting a relevant contribution of the genetic background. In MAS/r-HLH, subtle genetic defects and the inflammatory status, with variable contribution from each one of these factors, create a pathophysiological environment in which, as a consequence of an infectious stimulus of moderate/mild “strength,” the activation of macrophages and T cells may reach the threshold for clinically full-blown HLH. All forms of HLH have common clinical and laboratory features. Uncontrolled T cells and macrophage activation fuels a cytokine storm that is eventually responsible for the clinical features. Exaggerated levels of inflammatory cytokines are thus the final effectors, independent of the interplay between different upstream pathogenic factors.
Potential therapeutic targets
Identification of the key agents in the cytokine storm might provide the rationale for novel therapies. In s-JIA, 2 cytokines have emerged: IL-1 and IL-6. However, their role in the development and progression of MAS/r-HLH is poorly understood. Evidence in IL-6TG mice supports the hypothesis that the high levels of IL-6 in s-JIA might contribute to triggering MAS/r-HLH. In vitro studies have also shown that IL-6 can lead to TLR-mediated signaling in monocytes, inducing increased production of inflammatory cytokines, including IL-1β19. Even if neutralization of IL-6 has proven to be effective in a significant portion of patients with s-JIA, some MAS/r-HLH episodes were reported in the clinical trial with tocilizumab. Their relation with drug exposure and their severity remain to be clarified31. While IL-6 is increased both in prf−/− mice and in the repeated CpG model of secondary HLH, its role in progression and maintenance has never been tested using blocking antibodies6,18,36. With the exception of a recent report on a single patient with adult-onset Still disease with a favorable outcome, no information on the response to tocilizumab in MAS/r-HLH or in other HLH syndromes is available37.
In gene profiling of patients with s-JIA, increased expression of components of the IL-1 and TLR pathways has been identified38,39, and IL-1 blockade with anakinra is effective in a significant proportion of patients. Reports suggest also that administration of anakinra is effective in treatment of established MAS/r-HLH38,40,41,42. However, initiation of anakinra treatment has occasionally been associated with triggering of MAS/r-HLH43. Moreover, IL-1 blockade is not effective in mice harboring nonfunctional mutations of Unc1344. More information is needed to fully understand the role of IL-1 and the potential of IL-1 blockade in MAS/r-HLH42. An IL-1–IL-6 synergy may be hypothesized in patients with s-JIA exposed to infections. IL-1 induces IL-6 production from a variety of cell types45,46,47. IL-6 strengthens TLR ligand-induced IL-1β production by peripheral blood mononuclear cells (Strippoli, et al, unpublished data), and blockade of IL-1 may affect biological effects of IL-6 in multiple myeloma48.
Although its role in s-JIA has not been investigated, IFN-γ levels are increased in mouse models of both primary and secondary HLH, including the MAS/r-HLH in IL-6TG mice6,8,9,16,18 (De Benedetti, unpublished data). However, the increase is variable in the different models, with levels being generally higher in the models of p-HLH. In models of p-HLH, neutralization of IFN-γ reduces biochemical abnormalities, reduces tissue infiltrates, and prevents death6,8; in the nonlethal repeated CpG model of secondary HLH it reduces hematological abnormalities18. In addition, systemic chronic exposure to IFN-γ is sufficient to cause cytopenia and hemophagocytosis in mice49. Nevertheless, IFN-γ appears less effective, and another inflammatory cytokine, TNF-α, acquires pathogenic relevance when prf−/− mice are challenged with murine cytomegalovirus, a cytopathic virus29. In the repeated CpG model of secondary HLH, leukopenia and hyperferritinemia were IFN-γ-independent18. Information on the effect of IFN-γ neutralization is not available in other models of secondary HLH, including in IL-6TG mice. CD8+ T cells are the source of IFN-γ in p-HLH models. In the repeated CpG model of secondary HLH, a myeloid cell population may play a role in IFN-γ production18. The source of IFN-γ in the IL-6TG models of MAS/r-HLH is still to be determined. IFN-γ-producing CD8+ T lymphocytes are present in the liver of children with HLH secondary to infections or MAS/r-HLH21. Thus biological inhibition of IFN-γ as well as the IL-18/IL-18BP system, which plays a major role in IFN-γ production and is unbalanced in patients with HLH, may be a new frontier in the treatment of both primary and secondary HLH50.
There is some evidence suggesting efficacy of IL-1 inhibition in overt MAS/r-HLH, supporting a role of IL-1 in maintaining MAS/r-HLH. While there is still no solid information on the role of IL-6 in overt MAS/r-HLH, data from animals suggest that IL-6 may contribute to triggering it. Some data point to major involvement of IFN-γ in p-HLH; however, a comprehensive analysis of the levels and the signature of this cytokine in patients with MAS/s-HLH is still needed. Incidentally, increased levels of neopterin, a marker of IFN-γ activation, have been reported in 5 patients with MAS/r-HLH; this finding needs confirmation in a larger number33. A better understanding of the interplay between these cytokines might allow identification of the most suitable therapeutic target, to dampen as rapidly as possible the cytokine storm leading to multiple organ damage in patients with severe MAS/r-HLH.
- Accepted for publication January 31, 2013.