Disruption of MyD88 signaling suppresses hemophagocytic lymphohistiocytosis in mice

Hemophagocytic lymphohistiocytosis (HLH) is a rare disorder of the immune system. The familial form (FHL) arises in children and is caused by mutations that impair the cytotoxic activity of NK cells and CD8⁺ T cells. The second form, sporadic HLH, may arise in individuals with pathogen infections or with autoimmune diseases or malignancies.¹  Both HLH variants are believed to be initiated by viral infections.²  These pathogenic triggers have remained elusive, although Epstein-Barr virus has been proposed as one cause.^3,4 

After symptomatic onset of HLH, survival declines rapidly despite chemotherapy with cytotoxic and immunosuppressive drugs.⁵  To date, hematopoietic stem cell transplantation is the best long-term therapy for familial HLH.

We previously described an N-ethyl-N-nitrosourea (ENU)–germ line mutant called jinx in which a mutation in Unc13d prevents the release of cytolytic granules by NK cells and CD8⁺ T cells.⁶  We demonstrated that clinical features of the HLH-like syndrome observed in humans are recapitulated in Unc13d^jinx/jinx mice infected with lymphochoriomeningitis virus (LCMV). Unc13d^jinx/jinx remains the only animal model of human type 3 FHL, which is caused by mutations in MUNC13-4, the human ortholog of Unc13d.

On infection of Unc13d^jinx/jinx mutants, a positive feedback loop is initiated between CD8⁺ T cells and antigen-presenting cells (APCs) such as macrophages, leading to their overactivation and HLH-like symptoms.⁶  We hypothesized that interrupting this positive feedback loop could ameliorate the onset and/or extent of HLH-like disease and potentially reveal novel therapeutic targets for human HLH. We therefore examined the impact of disrupting cell adhesion molecules, effector cytokines, and innate immune sensors on the progression of HLH-like disease in Unc13d^jinx/jinx mutants, and identified myeloid differentiation primary response gene 88 (MyD88) as a key protein required for the development of the syndrome.

All mouse studies were performed in accordance with institutional regulations governing animal care and use. C57BL/6J mice were bred locally at The Scripps Research Institute. The following mutants have been previously reported and are described at http://mutagenetix.scripps.edu: Unc13d^jinx/jinx6  (MGI: 3628822; MMRRC: 016137-UCD), Myd88^poc/poc7  (MGI: 3641255; MMRRC:010475-UCD), and Tnf^PanR1/PanR18  (MGI:3616888; MMRRC:010462-UCD). Itgb2^Joker/Joker mice (MGI:3808883; MMRRC:016138-UCD) carry an A to T transversion in the acceptor splice site of intron 6 (TCCCAG→TCCCTG) in the Itgb2 gene, resulting in a complete deletion of exon 7. The mutated CD18 protein cannot associate with CD11a, CD11b, or CD11c.⁹  Interferon γ receptor (IFNγR)–deficient mice were kindly provided by Dr Charles D. Surh (The Scripps Research Institute) and B6.129S7-Il1r1^tm1Imx/J, mice were obtained from The Jackson Laboratory. Compound mutants were generated by intercrossing and genotyping the respective F1 progeny.

Blood samples were taken from the retro-orbital plexus of mice and analyzed using a Hemavet 950 veterinary hematology system (Fisher Scientific). Alternatively, IFNγ or TNF were assayed in the serum by ELISA (eBioscience).

The Armstrong strain of LCMV was injected intravenously at a dose of 2 × 10⁵ plaque-forming units per mouse. Viral titers were determined after organ homogenization by standard plaque assays on VERO cells.

The following antibodies were used for flow cytometry, to stain splenocytes: CD11b (M1/70), CD3ϵ (145-2C11), CD8α (53-6.7), CD80 (B7-1), CD86 (GL1), F4/80 (BM8), IFNγ (XMG1.2), TNF (MP6-XT22). Specific CD8⁺ T-cell responses were determined 12 days after infection, from splenocytes, by intracellular IFNγ and TNF staining after a 5-hour stimulation with 10⁻⁷ M LCMV GP33 (KAVYNFATM) peptide as described before.⁶  Specific cytokine secretion after stimulation with LCMV-derived peptide was measured as (% cytokine⁺ CD8⁺ T cells after peptide stimulation) − (% cytokine⁺ CD8⁺ T cells after stimulation without peptide). For spontaneous cytokine production, % cytokine⁺ CD8⁺ T cells after stimulation without peptide are indicated.

For histologic analysis, tissues were fixed in 10% buffered formalin. Paraffin-embedded sections were stained with H&E.

The statistical significance of differences was determined by unpaired Student 2-tailed t test. Differences with a P < .05 were considered statistically significant. For all figures: ns indicates not significant; *P < .05; **P ≤ .01; ***P ≤ .001.

Lymphocyte-function associated antigen (LFA-1) is a heterodimeric integrin that is involved in lymphocyte migration and intercellular adhesion. Because LFA-1 is found at the immunologic synapse between LCMV-infected target and CD8⁺ T cells,¹⁰  we hypothesized that impairing its function would restrain uncontrolled CD8⁺ T-cell proliferation in LCMV-infected Unc13d^jinx/jinx mice. To test this, we bred Unc13d^jinx/jinx mutants with Joker mice, in which a point mutation in the Itgb2 gene precludes surface expression of CD18, the β chain of the LFA-1 complex.⁹ 

Despite a decreased response on day 7 after LCMV infection, when T-cell expansion peaks (supplemental Figure 1A, available on Blood Web site; see the Supplemental Materials link at the top of the online article), the LCMV GP33-specific T-cell response in Unc13d^jinx/jinx;Itgb2^Joker/Joker mice was as severe as in Unc13d^jinx/jinx mice on day 12, when T cells usually contract (supplemental Figure 1B-C). The Joker mutation did not alter the spontaneous production of IFNγ in jinx CD8⁺ T cells 7 days after infection (supplemental Figure 1D). A similar differential behavior among CD8⁺ T-cell populations has been previously reported in LCMV-infected, CD18-deficient mice around the peak of the T-cell response.¹¹  On day 12 after infection, Unc13d^jinx/jinx;Itgb2^Joker/Joker CD8⁺ T cells accumulated, produced IFNγ spontaneously and displayed an activated phenotype (supplemental Figure 1C-E). Furthermore, both Unc13d^jinx/jinx;Itgb2^Joker/Joker and Unc13d^jinx/jinx mice displayed hematologic symptoms of HLH-like disease on day 12, namely low hematocrit, leukocytosis and neutrophilia (supplemental Figure 1F-H, respectively). Therefore, inhibiting integrin-dependent interaction is not sufficient to dampen the later progression of HLH-like syndrome in Unc13d^jinx/jinx mice.

HLH-affected individuals exhibit elevated concentrations of circulating cytokines including TNF and IFNγ, which are indicative of a poor clinical outcome in young patients.^12,13  We evaluated the role of these cytokines in the development of experimental HLH. To this end, Unc13d^jinx/jinx mutants were bred to Tnf^PanR1/PanR1 mice, in which TNF bioactivity is completely abrogated,⁸  and to IFNγ receptor (IFNγR)–deficient mice. On day 12 after LCMV infection, Unc13d^jinx/jinx and Unc13d^jinx/jinx;Tnf^PanR1/PanR1 mice had enlarged spleens (supplemental Figure 2A), with elevated proportions of CD8⁺ T cells and macrophages (supplemental Figure 2B-C). By contrast, Unc13d^jinx/jinx;Ifngr^−/− mice spleens did not develop splenomegaly and the proportions of splenic CD8⁺ T cells and macrophages were comparable with those observed in wild-type mice (supplemental Figure 2B-C). However, circulating levels of IFNγ and TNF in Unc13d^jinx/jinx;Ifngr^−/− mice were approximately 40-fold and 5-fold higher than in Unc13d^jinx/jinx controls on day 12 after LCMV infection (supplemental Figure 2D and E, respectively). A substantial elevation in the concentration of serum IFNγ was already observable in Unc13d^jinx/jinx;Ifngr^−/− mice relative to Unc13d^jinx/jinx mice on day 7 after LCMV infection (supplemental Figure 2F).

These data demonstrate that TNF signaling has no major role in the development of HLH-like syndrome in LCMV-infected Unc13d^jinx/jinx mice, whereas IFNγ/IFNγR signaling is required for uncontrolled APC and CD8⁺ T-cell proliferation. Unc13d^jinx/jinx;Ifngr^−/− mice became hunched and weak by day 12 after infection, with 100 000-fold higher viral loads in their kidneys than Unc13d^jinx/jinx mice (supplemental Figure 2G); LCMV infection led to a fatal outcome in some of these mice (data not shown). Unresolved viral proliferation in the compound mutants is consistent with previous studies showing that IFNγ-deficient mice are susceptible to LCMV¹⁴  and that elevated viral loads in Ifngr^−/− mice are further aggravated by the absence of perforin.^15,16  In addition, administration of an IFNγ-blocking antibody to perforin-deficient mice during the peak CD8⁺ T-cell response to LCMV resulted in remission of HLH-like symptoms.^17,18  Despite the use of different LCMV strains, these complementary observations suggest that the production of IFNγ at early times during viral infection may be critical to generate an initial antiviral state by priming immune responses in HLH prone individuals, before high quantities of IFNγ secreted by activated CD8⁺ T cells trigger the development of immunopathology. In our model of IFNγR-deficiency, this early antiviral state may be prevented, resulting in increased viral susceptibility. Early blockade of IFNγ/IFNγR signaling may therefore be deleterious in HLH-affected individuals. Alternatively, and as previously reported in another model of virus infection, deficiency of IFNγR may result in a different effect than deficiency of its ligand, IFNγ.¹⁹ 

TLRs are sensors of pathogens, and most TLRs recruit the adaptor MyD88 on stimulation. During LCMV infection, MyD88 expressed by APCs is implicated in proinflammatory cytokine production²⁰  and CD8⁺ T-cell expansion,^21,22  2 hallmarks of HLH. To examine the effects of a disruption of MyD88 signaling in the development of HLH-like disease in Unc13d^jinx/jinx mice, we introduced the pococurante (poc) allele of Myd88⁷  onto the Unc13d^jinx/jinx background. LCMV-infected Unc13d^jinx/jinx;Myd88^poc/poc mice and wild-type controls displayed comparable densities of red blood cells (RBCs) and white blood cells (WBCs) on day 12 (Figure 1A-B). LCMV infection did not affect hematocrit, hemoglobin, circulating lymphocytes, neutrophils, and monocytes in Unc13d^jinx/jinx;Myd88^poc/poc mice (supplemental Figure 3), and the spleen cellularity of these double mutants was half that observed in Unc13d^jinx/jinx mice (Figure 1C). Moreover, whereas the proportion of splenic macrophages in Unc13d^jinx/jinx; Myd88^poc/poc mice was comparable with the one in Unc13d^jinx/jinx mice after infection (Figure 2A), the absolute number of these cells was 6-fold lower than in Unc13d^jinx/jinx mice (Figure 2B). Unc13d^jinx/jinx;Myd88^poc/poc macrophages did not up-regulate CD86 (Figure 2C) and CD80 (Figure 2D) on their surface as opposed to Unc13d^jinx/jinx macrophages. Because MyD88 is not involved in the up-regulation of CD86 and CD80 on APCs²⁰  or in antigen presentation in LCMV-infected wild-type mice,²²  we hypothesize that the down-modulation of these molecules at the surface of Unc13d^jinx/jinx;Myd88^poc/poc macrophages is cell-extrinsic and possibly attributed to an indirect effect mediated by other cells.

As the ablation of MyD88 signaling in the Unc13d^jinx/jinx background affects the expression of costimulatory molecules on APCs, T-cell activation in Unc13d^jinx/jinx;Myd88^poc/poc mice may be dampened as well. The proportions and the absolute numbers of splenic CD8⁺ T cells in Unc13d^jinx/jinx;Myd88^poc/poc mice and wild-type mice were similar (Figure 3A-B). The percentage of CD8⁺ T cells that spontaneously produced IFNγ in wild-type and Unc13d^jinx/jinx;Myd88^poc/poc mice were comparable and approximately one-tenth the percentage observed in Unc13d^jinx/jinx mice on day 12 after infection (Figure 3C). In vitro stimulation of splenocytes with the LCMV-derived peptide GP33 showed that the total percentage of IFNγ-producing CD8⁺ T cells (Figure 3D) and their absolute numbers (Figure 3E) were high in Unc13d^jinx/jinx mice, but significantly lower and equivalent to wild-type in Unc13d^jinx/jinx;Myd88^poc/poc mice. The percentage of LCMV GP33-specific cytokine-producing CD8⁺ T cells was comparable in all strains (Figure 3F).

Taken together, CD8⁺ T-cell responses in Unc13d^jinx/jinx;Myd88^poc/poc mice were lower than in Unc13d^jinx/jinx mutants, consistent with the reduced activation state of their macrophages. Therefore, inhibition of MyD88 signaling in the Unc13d^jinx/jinx background not only prevented overproliferation of immune cells, but also dampened antigen-presenting cell maturation and overactivation of CD8⁺ T cells.

Although they exhibited similar spleen size, complete blood counts, macrophage and CD8⁺ T-cell numbers as wild-type mice, Unc13d^jinx/jinx;Ifngr^−/− mice produced overwhelming amounts of systemic IFNγ and TNF (supplemental Figure 2D-E). Serum IFNγ concentration was higher in Unc13d^jinx/jinx;Myd88^poc/poc double mutants than in wild-type controls on day 7 at the peak of the CD8⁺ T-cell response (supplemental Figure 2F). However, 12 days after infection with LCMV, IFNγ was undetectable in Unc13d^jinx/jinx;Myd88^poc/poc mice (Figure 4) which additionally displayed only a moderate quantity of serum TNF (supplemental Figure 2E). These observations show that compared with deletion of the Ifngr gene, the Myd88^poc mutation mitigates systemic inflammation in a remarkably effective manner. Although IFNγ may play a major role in the activation of APCs by primed CD8⁺ T cells, it has nonetheless nonredundant functions in controlling viral proliferation, and is therefore necessary for host survival.¹⁴  LCMV tends to persist in kidneys,²³  and whereas Unc13d^jinx/jinx;Ifngr^−/− compound mutants failed to clear the virus from the kidney on day 12 after inoculation, no infectious particles were detected in Unc13d^jinx/jinx;Myd88^poc/poc mice (supplemental Figure 2G). Noteworthy, LCMV was controlled in blood and spleens of Unc13d^jinx/jinx;Myd88^poc/poc and Unc13d^jinx/jinx;Ifngr^−/− mice on day 12 (data not shown).

Finally, we performed histologic analysis to further investigate the dampening effects of MyD88 disruption on HLH-like symptoms in Unc13d^jinx/jinx mice. Sections from spleens revealed absent or disrupted B-cell follicles with infiltrations of activated lymphocytes and macrophages in LCMV-infected Unc13d^jinx/jinx mice. Unc13d^jinx/jinx;Myd88^poc/poc compound mutants, however, displayed distinct germinal centers, although their overall splenic architecture was to some extent less organized than in wild-type controls (Figure 5).

We have shown that mutational inactivation of MyD88 prevents HLH-like disease in Unc13d^jinx/jinx mice by abrogating the positive feedback loop between CD8⁺ T cells and APCs while permitting effective control of infection. Replicating virus was absent in Unc13d^jinx/jinx;Myd88^poc/poc mutants, in contrast with a large infectious reservoir in Unc13d^jinx/jinx;Ifngr^−/− kidneys (supplemental Figure 2G). This has important clinical implications toward a strategy to eliminate HLH immunopathology while maintaining antiviral immunity. Several mechanisms and effector cell types may account for the balance between immunopathology and protection observed in Unc13d^jinx/jinx;Myd88^poc/poc mice.

Besides TLR signaling, MyD88 mediates signaling from the IL-1 receptor family, which includes the IL-18 receptor expressed by CD8⁺ T cells. In a LCMV model of neuropathogenesis, disruption of the IL-18 or IL-1 receptors did not affect CD4⁺ T-cell response whereas disruption of MyD88 did.²⁴  IL-18 synergizes with IL-12 in maintaining IFNγ secretion by LCMV-specific CD8⁺ T cells.^25,26  Furthermore, administration of an IL-18–blocking antibody to LCMV-infected perforin-deficient mice, which also exhibit HLH-like clinical features, did not improve the survival of treated animals.¹⁷  Therefore abrogation of IL-18 signaling in CD8⁺ T cells by compound homozygosity for the Myd88^poc mutation is unlikely to explain rescue of HLH-like disease in Unc13d^jinx/jinx;Myd88^poc/poc mice.

High levels of IL-1 are found in the serum of HLH-affected patients,¹³  and the administration of IL-1 receptor antagonists seems beneficial for patients affected with the reactive form of HLH.²⁷  To determine the degree of involvement of IL-1 in the development of the HLH-like syndrome in response to LCMV infection, Unc13d^jinx/jinx mice were bred to IL-1R1–deficient mice. Disrupting IL-1R1 signaling neither decreased the maturation state of Unc13d^jinx/jinx macrophages (supplemental Figure 4A) nor restrained activation of Unc13d^jinx/jinx CD8⁺ T cells in response to LCMV (supplemental Figure 4B-C), suggesting that the rescue of LCMV-infected Unc13d^jinx/jinx mice by concomitant MyD88-deficiency does not result from impairment of IL-1 receptor signaling.

Our data exclude a role for TNF and IL-1 for the progression of HLH-like disease in Unc13d^jinx/jinx mutants (supplemental Figures 2 and 4, respectively). Yet IL-6–producing macrophages have been identified in liver biopsies and blood of patients suffering from hemophagocytic syndrome^28,29  and significant amount of IL-6 was found in sera of mice displaying symptoms of HLH.¹⁷  On LCMV infection, MyD88-deficiency results in reduced production of IL-6,^20,22  which has been suggested to support T-cell proliferation.³⁰  However, Unc13d^jinx/jinx;Myd88^poc/poc double mutants displayed serum IL-6 concentrations that were not significantly different from control C57BL/6J or Unc13d^jinx/jinx mice on day 7 after LCMV infection (supplemental Figure 4D). These data suggest a minor role of this cytokine for the outcome of the T-cell response in our model.

The observation that mice with deficiencies in both TLR7 and TLR9 mount normal adaptive immune responses to LCMV supports a TLR-independent role of MyD88 for containing the infection.²⁰  Alternatively, sensors independent of MyD88 may be sufficient to induce normal adaptive responses to LCMV in MyD88-deficient mice. Recently, the endoplasmic RNA helicases RIG-I and MDA-5 were shown to bind LCMV-derived RNA and trigger type I IFN production.³¹  Additional cell-intrinsic roles have also been ascribed to MyD88 as for instance its ability to regulate expression of effector immune responses in IFNγ-activated macrophages.³²  MyD88 has been shown to interact with IFNγR thereby transducing IFNγ-mediated signals that induce a pathway for the control of expression and stability of many short-lived mRNAs, including those encoding TNF and IFNγ-inducible protein 10 (IP-10).³³ 

During LCMV infection, the priming of the immune response largely depends on LCMV-infected APCs.^34,35  It has been suggested that impaired type I IFN production by plasmacytoid dendritic cells is an underlying cause of reduced priming of endogenous T cells in Myd88^−/− mice.²⁰  However, MyD88-deficient APCs were shown to be as efficient as wild-type APCs in priming infused, MyD88-competent T cells with a transgenic T-cell receptor specific for the immunodominant epitope of LCMV.²²  Furthermore, serum IFNβ remained undetectable in LCMV-infected mice on day 7, although substantial concentrations of IFNγ were measured (data not shown and supplemental Figure 2F), thereby further emphasizing the importance of IFNγ for the modulation of the T-cell response in Unc13d^jinx/jinx mice.

Thus, a need for MyD88 in costimulation and antigen presentation leading to T-cell activation is unlikely to be a critical driver of HLH-like disease per se in LCMV-infected Unc13d^jinx/jinx mice, at least during the acute stage of the infection. Yet, alteration of T-cell responses observed in the MyD88-deficient state might potentially result from a requirement for MyD88 within APCs.

MyD88 is expressed by T cells and required for their function. Therefore, Unc13d^jinx/jinx;Myd88^poc/poc CD8⁺ T cells might have an intrinsic defect imparted by the Myd88 mutation. In mice infected with LCMV, MyD88 signaling is necessary for sustained expansion of effector CD8⁺ T cells,^21,22  especially during systemic infection with the faster-replicating Traub strain of LCMV.³⁶  However MyD88 is not involved in early activation and proliferation of antigen-activated CD8⁺ T cells nor in the generation and maintenance of functional memory CD8⁺ T cells.³⁷  Considering the importance of IFNγ-producing CD8⁺ T cells for the development of LCMV-induced HLH disease,¹⁷  it is plausible that disruption of MyD88 rescues the disease in Unc13d^jinx/jinx;Myd88^poc/poc mutants because fewer CD8⁺ T cells secreting less IFNγ are induced. This may, in turn, result in fewer APCs being activated, thereby restricting the stimulation of CD8⁺ T cells. In addition to such a quantitative effect on APC activation, the transcription of many IFNγ-dependent genes in macrophages also relies on MyD88,³²  which may further affect those macrophages that become activated in Unc13d^jinx/jinx;Myd88^poc/poc mice.

In summary, our data demonstrate no role for CD18, TNF, and IL-1R1 in the progression of HLH in our model. Furthermore, we identified MyD88 as a candidate drug target superior to IFNγ, given the unrestrained viral load resulting from a disruption of IFNγ signaling.

We thank Christine Domingo, Sara Kalina, Diantha LaVine, Terrence Robinson, and Charles Ross for technical support. Dr Nissi Varki, University of California, San Diego helped with the evaluation of the histology. We are indebted to Carrie N. Arnold, Amanda L. Blasius, Eva Marie Y. Moresco, Sophie Ugolini, and Eric Vivier for critical review of the manuscript.

This work was supported by National Institutes of Health (NIH) contract HHSN2722007000 38C (to B.B.) and NIH grants 5P01AI070167 (to B.B.) and AI09484 (to M.B.O.), by fellowships from EMBO and the Swiss National Science Foundation (to P.K.), and by NIH training grant AI007244 (to D.P.).

National Institutes of Health

Contribution: P.K. and K.C. conducted experiments and performed data analyses; D.P. conducted experiments, performed data analyses, and provided critical discussions; M.B.O. provided critical discussions; and P.K., K.C., and B.B. designed experiments and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Bruce Beutler, MD, Professor and Chair, Department of Genetics, The Scripps Research Institute, 10550 N Torrey Pines Rd, SP-293, La Jolla, CA 92037; e-mail: bruce@scripps.edu.