IFN-γ and CD25 drive distinct pathological features during hemophagocytic lymphohistiocytosis

Abstract

Background

The inflammatory activation of CD8⁺ T cells can, when unchecked, drive severe immunopathology. Hyper-stimulation of CD8⁺ T cells, through a broad set of triggering signals, can precipitate hemophagocytic lymphohistiocytosis (HLH), a life-threatening systemic inflammatory disorder.

Objective

The mechanism linking CD8⁺ T cell hyper-activation to pathology is controversial, with excessive production of IFN-γ and, more recently, excessive consumption of IL-2, proposed as competing hypotheses. We formally tested the proximal mechanistic events of each pathway in a mouse model of HLH.

Methods

In addition to reporting a complete IFN-γR1 deficient patient with multiple aspects of HLH pathology, we used the mouse model of Prf1^KO mice infected with LCMV to genetically eliminate either IFN-γ production or CD25 expression and assess the immunological, hematological and physiological disease measurement.

Results

We found a striking dichotomy between mechanistic basis of the hematological and inflammatory components of CD8⁺ T cell-mediated pathology. The hematological features of HLH were completely dependent on IFN-γ production, with complete correction following loss of IFN-γ production, without any role for CD8⁺ T cell-mediated IL-2 consumption. The mechanistic contribution of the immunological features, by contrast, were reversed, with no role for IFN-γ production, but substantial correction following the reduction of IL-2 consumption by hyper-activated CD8⁺ T cells. These results were complemented by the characterization of an IFN-γR1-deficient HLH-like patient, where multiple aspects of HLH pathology was observed in the absence of IFN-γ signaling.

Conclusion

These results synthesize the competing mechanistic models of HLH pathology into a dichotomous pathogenesis driven through discrete pathways. A holistic model provides a new paradigm for understanding HLH, and, more broadly, the consequences of CD8⁺ T cell hyper-activation, thereby paving the way for clinical intervention based on the features of HLH in individual patients.

Capsule summary

Studying the pathogenesis of hemophagocytic lymphohistiocyosis, we found a striking dichotomy between the hematological and inflammatory components of HLH. Hematological features are strictly dependent of IFN-γ, while inflammation is driven by dysregulated IL-2 signaling.

Graphical Abstract

INTRODUCTION

Hemophagocytic lymphohistiocytosis (HLH) is a life-threatening systemic inflammatory disorder. The inflammation in this disease is characterized by an excessive activation of the immune system (mainly cytotoxic T cells and macrophages) and a cytokine storm (including IFN-γ, IL-6 and TNF-α). Clinical manifestations include prolonged fever, cytopenia, hepatosplenomegaly, hepatitis, and, in some patients, neurological symptoms. Classical presentation can lead to patient death by uncontrolled inflammation, multiple organ failure or secondary severe infection, if left untreated ^1–3. Successful treatment requires drastic clinical intervention through steroids, etoposide and immunosuppressive reagents, with only a 50 to 60% survival rate ^4–6.

HLH is a broad clinical term, encompassing a diverse set of inflammatory contexts mediated by CD8⁺ T cell hyper-activation. Primary HLH refers to patients with severe inherited defects in genes related to the cytotoxic pathways of T and NK cells, or in genes involved in Epstein Barr virus (EBV) clearance, with viral infection frequently serving as a trigger for clinical manifestation ⁷. When a genetic cause is established, allogeneic hematopoietic stem cell transplantation (allo-HSCT) is the only curative treatment. Secondary or acquired HLH encompasses patients who develop the syndrome as a consequence of escalating CD8⁺ T cell responses against infections (mainly after EBV and cytomegalovirus (CMV), but also after bacterial, fungal or parasitic infections), malignancies (mainly lymphomas) or autoimmune targets, such as occurs in systemic juvenile idiopathic arthritis or systemic lupus erythematosus ^1,3,5,8. HLH thus serves as the archetype for understanding the pathology mediated by hyper-activation of CD8⁺ T cells within a general clinical context. The poor prognosis of HLH patients, even when given the best current treatments, indicates that our understanding of the molecular basis of this disease is unsatisfactory. HLH is a complex disease comprising two main features with both immunological manifestations, including cytotoxic T cell activation and cytokine storm, and hematological manifestations, with severe anemia. In addition, hemophagocytosis describes the presence of activated macrophages able to engulf blood cells, and its presence has been reported to be a direct contributor to anemia ⁹. Whether the immunological and hematological aspects of the disease share a common mechanistic basis, or independently arise from distinct molecular mechanisms has not been explored to date.

One of the key developments in the understanding of HLH has been the development of a mouse model that recapitulates the key features of patients. The mouse model utilizes Perforin^KO mice (Prf1^KO), replicating the common PRF1 mutation in primary HLH. Prf1^KO mice are healthy in the baseline state, however when infected with lymphocytic choriomeningitis virus (LCMV), Prf1^KO mice develop many of the same hematological and immunological characteristics of HLH, with severe anemia and fatal inflammation ^10,11. Analysis of HLH patients and the HLH mouse model has been used to develop a working model for HLH pathogenesis. In this model, defects in cytotoxicity result in higher viral titers and an inability to clear antigen presentation cells, leading to extended immune stimulation and increased activation of CD8⁺ T cells ¹². These CD8⁺ T cells then have amplified expression of cytokines which further precipitate systemic inflammation including the activation of other immune cells, especially macrophages, creating a cytokine storm ¹¹. IFN-γ is thought to be the most critical cytokine in HLH pathology, with both the hematological and inflammatory aspects of the disease attributed to its excessive production ^9,11,13,14. The role of IFN-γ has, however, been recently challenged, with the findings of HLH in patients with genetic defects in the IFN-γ pathway ^15–17. Specifically, three patients have been reported with HLH disease and an IFN-γR defect, although among these only one patient received a definitive diagnostic of a complete absence of IFN-γ signaling through IFN-γR2, while the other two patients had either a partial defect or an implied defect of IFN-γR1^15,16. Supporting the IFN-γ-independence of disease, development of a systemic juvenile idiopathic arthritis-like inflammation has been described in IFN-γ knockout mice ^18,19. And finally, gene expression profiling of peripheral blood mononuclear cells (PBMC) from children with untreated HLH failed to identify differential expression of IFN-γ or IFN-γ regulated genes as an HLH signature ²⁰. We recently proposed an additional alternative model of HLH pathogenesis ²¹. CD8⁺ T cells from LCMV-infected Prf1^KO mice not only exhibit hyper-production of IFN-γ, but also manifest extreme expression level of the IL-2 receptor alpha (CD25). The enhanced consumption of IL-2 by activated CD8⁺ T cells is accompanied by a collapse in regulatory T cell (Treg) numbers ²¹, a feature which could account for the feed-forward inflammatory loop observed in HLH patients. While these models are not mutually exclusive, resolution of this controversy is of great importance as a clinical trial using anti-IFN-γ antibody in HLH patients is currently ongoing ²².

Here we sought to formally test the competing hypotheses for the pathological mechanisms of CD8⁺ T cell hyper-stimulation in HLH. In addition to reporting a complete IFN-γR1 deficient patient with multiple aspects of HLH pathology, we used the mouse model of Prf1^KO mice infected with LCMV to genetically eliminate either IFN-γ production or CD25 expression and assess the immunological, hematological and physiological disease measurement. We found a striking dichotomy between the hematological and inflammatory components of HLH pathology. IFN-γ production during HLH was essential for the hematological features, with the loss of IFN-γ completely correcting anemia, but having no effect on immune activation, wasting and survival. By contrast, restricting IL-2 consumption by CD8⁺ T cells had no impact on the hematological features of HLH, but restored Treg numbers, suppressed the hyper-inflammatory state and extended the life-span. These results provide a paradigm shift for the understanding of the consequences of CD8⁺ T cell hyper-activation, with the two sets of symptoms (inflammatory and hematological) driven through dichotomous pathways, which together drive the HLH clinical presentation.

MATERIALS AND METHODS

Patient

Clinical data were obtained from the patient developing HLH symptoms during clinical care in Casablanca, Morocco. Further diagnostic investigation was performed at the Laboratory of Human Genetics of Infectious Diseases, INSERM U1163, Imagine Institute, in accordance with local regulations, and with the approval of the IRB of Necker Hospital for Sick Children, Paris, France. Written informed consent was obtained by the patient family.

Mice

Perforin-deficient (C57BL/6-Prf1^tm1Sdz/J) ²³, IFN-γ-deficient (C57BL/6-Ifng^tm1Ts) ²⁴ and CD8-Cre transgenic mice (C57BL/6-Tg(Cd8a-cre)1Itan/J) ²⁵ were obtained from The Jackson Laboratory (Bar Harbor, ME) and were used on the C57BL/6 background. CD25^KO mice were generated from EUCOMM embryonic stem cells, with production through blastocyst injection. Transgenic founders were transiently intercrossed with Flpase mice (FLPe) ²⁶, to excise the neomycin resistance cassette and convert the allele to a floxed status. The resulting CD25^fl/fl mice were backcrossed to the C57BL/6 background >8 generations prior to intercross with CD8-Cre mice. All mice were maintained in specific pathogen-free facilities of University of Leuven. All experiments were approved by the University of Leuven Animal Ethics Committee. All mice were used at 8–15 weeks of age.

Virus and infection

LCMV-Armstrong was produced and titrated as previously described ²⁷. Briefly, The LCMV strain was propagated in baby hamster kidney cells and titrated on Vero African green monkey kidney cells as previously described ²⁸.At 8–12 weeks of age, mice were infected with 10⁵ PFU of LCMV-Armstrong intraperitoneally (day 0) and were monitored and analyzed at day 10 post-infection for immunological and hematological parameters.

Blood and plasma analysis

Mouse blood samples were analyzed on Cell-Dyn 3700 (Abbott) to determine red blood cell and platelet counts as well as hematocrit and hemoglobin contents. Plasma samples were analyzed determined by a mouse IL-2Rα DuoSet ELISA (R&D System) and the MSD inflammatory cytokine panel according the manufacturers’ protocol.

Genomic DNA and whole exome-sequencing (WES)

Genomic DNA was extracted from whole blood and 3μg was sent for WES. Briefly, an adapter-ligated library was prepared with the Paired-End Genomic DNA Sample Prep kit V1 (Illumina). The SureSelect Human All Exon kit (Agilent Technologies) was then used for exome capture. Exome capture was performed with the 71Mb SureSelect Human All Exon kit (Agilent Technologies).

In vitro patient cell line analysis

EBV-transformed B lymphocytes (EBV-B cells) were generated from freshly collected PBMCs. IFN-γR1 expression (phycoerythrin PE labeled mAbs specific human or isotype-matched controls from BD Biosciences) on the surface of EBV-B cells were performed by flow cytometry analysis as previously described ²⁹. EBV-B cells were stimulated for 30 minutes with IFN-γ (10⁵ IU/ml) or IFN-α (10⁵ IU/ml) and phospho-STAT1 (anti-Y701-pSTAT1-Alexa468, 612132, BD Transduction laboratory) was detected by flow cytometry analysis. Staining was assessed on a Galios cytometer (Beckman Coulter), and the results were analyzed with FlowJo v10.

Mutation testing

The wild-type IFNGR1 allele was inserted into V5-topo-pcDNA3.1 (Invitrogen). The p.W99R and c.131del mutants were generated by site-directed mutagenesis (Agilent Technologies, PfuUltra II Fusion HS DNA polymerase). IFN-γR1-deficient SV40-fibroblasts were transfected with an insert-less V5-tagged pcDNA3.1 plasmid (Empty Plasmid, EV) and one of the IFNGR1 V5-tagged pcDNA3.1 plasmids in the presence of X-treme GENE9 transfection reagent (Sigma Aldrich). Total proteins were solubilized in extraction buffer as previously described ³⁰. Samples were run on SDS-PAGE in 10% polyacrylamide gels (Bio-Rad), and the resulting bands were transferred onto fluorescence PVDF membranes (Bio-Rad) with the Trans-Blot Turbo Transfer System (Bio-Rad). Immunoblotting was performed with antibodies against GAPDH (14C10, Cell Signaling Technology) and the V5 tag (Thermo Fisher Scientific). Bound antibody was detected with ECL western blotting detection reagents (Bio-Rad) on Chemidoc. Electrophoretic mobility shift assay (EMSA) was carried out as previously described ³⁰. Briefly, cells were stimulated for 20 minutes with IFN-γ at 10⁴ IU/ml. We then obtained a nuclear extract from these cells, and incubated 5–10 μg of this nuclear extract with IRDye 700 gamma-activating sequence (GAS) probe, corresponding to the Fc-γR1 promoter, subjected the mixture to electrophoresis in a polyacrylamide gel and read with LI-COR Clx machine.

Degranulation assay

Degranulation of human NK cells and CD8⁺ T cells were performed as following; 3×10⁵ PBMC were incubated for 3 hours in vitro alone, in presence of 1:1 K562 cells, or with phorbol 12-myristate 13-acetate (50 ng/mL; Sigma-Aldrich), and Ionomycin (250 ng/mL; Sigma-Aldrich). Cells were then stained for surface staining including CD3 (SK7), CD8 (RPA-T8), CD56 (MEM188), CD16 (CB16), CD107a (LAMP-1) and CD107b (H4B4) (all from Invitrogen, Thermofisher) and CD14 (M5E2, from Biolegend) and analyzed by flow cytometer CANTO II (BD Bioscience).

Histology

Tissues were fixed in 4% PFA and continued to trimming and processing for hematoxylin and eosin staining. Images were evaluated on a Nikon Labophot-2 microscope and captured using a Nikon D5000 digital microscopic camera. Histopathological analysis was performed blinded and subjectively by BioGenetics Research Laboratories (WA, USA) as before ^31,32. Briefly, severity score 1 was used to include pathological findings that showed no significant lesions and normal morphology. Severity score 2 was assigned to all tissues showing minimal (minor and infrequent, <10% of defined area) and mild (noticeable microscopic change but no overt feature, 11–20% of defined area involved in case of focal, multifocal or diffuse effects). Moderate findings including prominent changes (21–40% of defined area for focal, multifocal or diffuse effects) were assigned severity score 3. Overwhelming microscopic features (41–100% for focal, multifocal or diffuse effects in defined area) were assigned severity grade 4.

Flow cytometry

Single-cell suspensions were prepared from mouse spleen and pooled lymph nodes (cervical, inguinal, mesenteric, axillary and brachial). For intracellular cytokine staining, lymphocytes were plated at 5×10⁵ cells/well in 96 well tissue culture plates in complete RPMI containing phorbol myristate acetate (50 ng/mL; Sigma-Aldrich), ionomycin (250 ng/mL; Sigma-Aldrich) and monensin (1/1500; BD, Bioscience) for 4 hours at 37°C. All cells were fixed with BD Cytofix™ (BD, Biosciences) or fixed and permeabilized with the eBioscience Foxp3 staining kit (eBioscience). Anti-murine antibodies included anti-CD4 (RM4–5), anti-CD8a (53–6.7), anti-FoxP3 (FJK-16s), anti-CD25 (PC61.5), anti-CD127 (A7R34), anti-CTLA4 (UC10–4B9), anti-CD69 (H1.2F3), anti-CD103 (2E7), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-IL-2 (JES6–5H4), anti-IFNγ (XMG1.2), anti-KLRG1 (2F1), anti-PD1 (J43 and anti-phospho-STAT5 (SRBCZX) and anti-Ki67 (SolA15) from eBioscience. Data were collected on BD CANTO II (BD Biosciences) and analyzed using FlowJo for Mac version 9.6 (Tree Star Inc.).

Statistics

Statistical analysis was performed with a one-way ANOVA with Tuckey’s multiple comparisons test. All statistical analyses were carried out with GraphPad Prism (version 7; GraphPad Software, La Jolla, Calif).

RESULTS

IFN-γ production drives the hematological, but not immunological, components of HLH pathology

LCMV infection of Prf1^KO mice constitutes a robust mouse model of HLH, with hematological abnormalities, exuberant inflammation and early mortality ¹¹. Previous studies using genetic deficiency or antibody-mediated neutralization have highlighted the role of IFN-γ in the anemia occurring in LCMV-infected Prf1^KO mice ^9,11,13,14, however the other aspects of HLH pathology have not been assessed. We backcrossed Prf1^KO mice with IFN-γ^KO, creating Prf1^KO IFN-γ^KO double knockout (DKO) mice. LCMV infection of wildtype, Prf1^KO mice and Prf1^KO IFN-γ^KO DKO mice produced a severe disruption of the erythroid cell compartment in Prf1^KO mice which was completely corrected in DKO mice (Figure 1A–C). IFN-γ^KO mice, with wildtype perforin, did not exhibit any hematological abnormalities (Figure 1A–C). Intriguingly, while the anemia was prevented by IFN-γ-deficiency, DKO mice still exhibited liver hemophagocytosis (Figure 1D, E). This data demonstrates that low red cell counts, hemoglobin and hematocrit are IFN-γ-dependent pathologies.

Figure 1. Hematological component of murine HLH is dependent on excessive IFN-γ production.

Wildtype mice, IFN-γ^KO, Prf1^KO, and Prf1^KOIFN-γ^KO mice were infected with LCMV Armstrong. On day 10, blood samples from each strain were assessed for (A) hemoglobin (B) red blood cells and (C) hematocrit (A-C: n=11,5,4, 6). Results pooled from 4 experiments. (C) Representative histology from the liver and (D) histological scoring for hemophagocyte presence in liver on H&E sections on day 10 (n=3,4,3,3). Arrows indicate hemophagocytes. Scale bar is 50 μm. Analyzed with one-way ANOVA with Tuckey’s multiple comparisons test, SEM and individual data points shown.

Having established the IFN-γ-dependency of the hematological signature of HLH, we next assessed the physiological and immunological components of the disease. Unlike wildtype mice, Prf1^KO mice present with distressed behavior during the late stages of LCMV infection, coupled with severe wasting necessitating euthanasia at day 10 (Figure 2A). Despite a complete recovery of hematological abnormalities, Prf1^KO IFN-γ^K DKO mice developed the same distressed behavior and wasting, also requiring euthanasia at day 10 (Figure 2A). Prf1^KO mice demonstrated an enlarged spleen and liver damage after LCMV infection, which was also unaffected IFN-γ genotype (Supplementary Figure 1). In HLH, one of the primary pathological events is the infiltration and destruction of tissues, leading to multiple organ failure ^3,5. While wildtype and IFNγ^KO recovered from LCMV infection without widespread liver damage, Prf1^KO and Prf1^KO IFN-γ^KO DKO both experienced extensive reticuloendothelial cell infiltration (Figure 2B). The physiological manifestations of HLH pathology were therefore not dependent on excessive IFN-γ production.

Figure 2. Inflammatory and physiological aspects of murine HLH arise independent of IFN-γ.

Wildtype mice, IFN-γ^KO, Prf1^KO, and Prf1^KOIFN-γ^KO mice were infected with LCMV Armstrong. (A) Mouse weights on days 0, 7 and 10 after LCMV infection (n=7,9,7,9). (B) Histological scoring for immune cell infiltrates in liver on H&E sections on day 10 (n=5,6,3,6). (C) Percentage of CD8+ T cells in the lymph nodes (n=7,9,7,9). (D) Percentage of CD8⁺ T cells in the CD8⁺CD62L⁻CD44⁺ effector memory (TEM) (n=7,9,7,9) and percentage of activated CD8⁺ T cells in the CD8⁺CD127⁺KLRG1^high (SLEC) subsets (n=4,5,7,6). (E) Percentage of CD8⁺ T cells expressing CD25⁺ (n=7,9,7,9). (F) Representative CD25 staining on CD8⁺ T cells. (G) CD25 geometric MFI on total CD8⁺ T cells (n=4,5,7,6). (H) Relative ratio of sCD25 in the serum (n=14,6,15,6). Analyzed with one-way ANOVA with Tuckey’s multiple comparisons test, SEM and individual data points shown. Results pooled from 4 experiments and 3 experiments for sCD25 measurements.

Previous publications from our group and others have previously demonstrated a clear correlation between serious weight loss and critical inflammation in mouse HLH model ^21,33. Therefore, we performed a detailed scrutiny of the immune compartment. First, CD8⁺ T cells, which are known to be responsible for the disease severity ¹¹, were elevated in the lymph nodes of Prf1^KO IFN-γ^KO DKO mice to an even greater degree than Prf1^KO mice (Figure 2C). In Prf1^KO mice infected with LCMV, there is not only an increase in the size of the CD8⁺ T cell compartment, but also a marked shift towards an activated phenotype, with high percentages of effector memory subsets (CD8⁺CD44⁺CD62L⁻) and short-lived effector cells (SLEC; CD8⁺CD44⁺KLRG1⁺CD127⁻) which are both hallmarks of HLH inflammatory condition (Figure 2D). Prf1^KO IFN-γ^KO DKO mice, despite the absence of IFN-γ, exhibited the same escalation of these inflammatory mediators as Prf1^KO mice, well above the levels normally associated with LCMV infection in wildtype or IFN-γ^KO mice (Figure 2D). We recently added exuberant CD25 expression on CD8⁺ T cells to the characteristic hallmarks of murine HLH, a condition associated with Treg collapse ²¹. This immunological manifestation was reproduced here in Prf1^KO mice, with an identical presentation in Prf1^KO IFN-γ^KO DKO mice. Specifically, CD25 expression as both a percentage (Figure 2E) and intensity (Figure 2F) were higher on CD8⁺ T cells from LCMV-infected Prf1^KO and Prf1^KO IFN-γ^KO DKO mice than the LCMV-infected wildtype or IFN-γ^KO controls. Moreover, high levels of soluble CD25 (sCD25) in the serum has been used as a diagnostic marker for HLH ^1,5, and when analyzed in our cohort, sCD25 was found at the highest level in Prf1^KO and Prf1^KO IFN-γ^KO DKO mice, clearly emphasizing the presence of systemic inflammation (Figure 2G). Similar results were seen with CD4⁺ T cells, with elevated activation in Prf1^KO mice which was not ameliorated by the additional loss of IFN-γ (Supplementary Figure 2). These data clearly demonstrated that the immunological features of HLH arise independent of the expression of IFN-γ, with hyper-activated CD8⁺ T cells still possessing deleterious effects on normal mouse physiology.

Development of HLH in a patient with an absence of IFN-γ signaling

We identified a patient affected by a complete IFN-γR1 deficiency that developed multiple HLH diagnostic criteria secondary to severe mycobacterial infection (Figure 3A). A 2 month old girl, from first degree consanguineous parents from Morocco, was diagnosed with BCG vaccine reactivation with presence of homolateral lymphadenopathy. There was no improvement despite antibiotherapy. Two months later, her clinical status further deteriorated and HLH diagnosis was made in presence of splenomegaly, high fever, pancytopenia (anemia and thrombocytopenia), hyperferritinemia and hypertriglyceridemia (Supplementary Table 1). While her blood showed normal level of NK cells (number and percentage according to the age of the patient), NK cell function was impaired (Supplementary Figure 3) and samples for sCD25 assessment were not available. Hemophagocytosis was not found in her bone marrow. The patient received antibiotics (ciproxin, amikacin, clarithromycin, rifampicin, isoniazid, ethambutol), steroids and cyclosporine and blood transfusions. The patient stabilized under this regime. Further clinical investigation of the patient was not possible. With a severe mycobacterial infection at a young age further investigation regarding the genetic origin of the disease was performed suspecting Mendelian susceptibility to mycobacterial disease (MSMD)³⁴. A rare homozygote missense mutation, p.W99R, within IFNGR1 gene was detected by WES. This mutation was confirmed by Sanger. Mother was heterozygous for this mutation. No DNA sample was available for father. No mutation in the genes involved in primary HLH was detected. (Figure 3B). While IFNGR1 protein was detected at a level comparable to healthy control (Figure 3C), further in vitro assay demonstrated a complete deficiency in the receptor signaling, with an absence of STAT1-Y701phosphorylation in the patient EBV-B cells after IFN-γ stimulation in vitro, while stimulation after IFN-α promoted normal STAT1 phosphorylation (Figure 3D). To formally characterize the IFNGR1 allele we used an exogenous expression system. IFN-γR1-deficient primary human fibroblasts immortalized with SV-40T antigen (SV40-fibroblasts) were transiently transfected with wild-type (WT), p.W99R, and c.131del (defined loss of expression allele, responsible of complete AR IFNγR1 deficiency) IFNGR1 cDNA. As with the analysis of primary patient cells, the p.W99R allele in this transfection system demonstrated normal levels of protein expression (Figure 3E). Cells were stimulated with IFN-γ and gamma-activated factor-DNA binding activity was evaluated by EMSA using a GAS probe. SV-40 fibroblasts transfected with p.W99R had an abolished response to IFN-γ, similar to a negative control (Figure 3F). These findings demonstrate that the p.W99R allele is a loss of signaling function allele with normal protein expression. In the patient, a high level of IFN-γ on plasma, absence of IL-12p70 production (after BCG+ IFN-γ activation), normal IFN-γ production (after BCG+ IL-12p40 activation) in whole blood assay were also detected (data not shown), as found in other patients with complete IFN-γR1 deficiency³⁴. Building upon previously reported HLH cases with impaired IFN-γ^15–17, this patient demonstrates that multiple aspects of HLH pathology can arise with a complete absence of IFN-γ signaling. Furthermore, the absence of hemophagocytosis in an IFN-γR1-deficient HLH-like patient complements the mouse data (above), in that hematological aspects of HLH may be IFN-γ-dependent, even though overall disease progression was not.

Figure 3. IFN-γR1 deficiency in a HLH-like patient.

(A) Pedigree of kindred, showing the index case and parents; each generation is designated by a Roman numeral (I-II). No gDNA was available for testing for father labeled “E?”. (B) Electropherogram (healthy control, patient (II,1) and her mother) showing the position of the mutation in the IFNGR1 gene. (C) Flow cytometry analysis of IFN-γR1 expression on the surface of EBV-B cells from healthy control, IFN-γR1 deficient patient and patient (II,1) with the recessive W99R mutation. The histograms represent the expression of IFN-γR1 (gray filled) and the isotype control (dashed line). (D) Phospho-STAT1 response of EBV-B cells to IFN-γ (black line) or IFN-α (dashed line) stimulation in healthy control, IFN-γR1 deficient patient and patient (II,1). (E) IFN-γR1^−/− SV40 fibroblasts were left untransfected (NT) or were transiently transfected with EV, WT, c.131del (negative control) o p.W99R mutant- IFNGR1 plasmids. Total lysis and western blots were performed with anti-V5 antibody, with GAPDH as the loading control. (F) IFN-γR1^−/− SV40 fibroblasts were left untransfected (NT) or were transiently transfected with empty plasmid (EV), wild type (WT), c.131del (negative control) and p.W99R mutant-IFNGR1 plasmids, and were either left non-stimulated (NS) or stimulated with 10⁵ IU/ml IFN-γ (IFN-γ). DNA-binding activity was then analyzed by EMSA with a GAS probe.

IL-2 consumption by hyper-activated CD8⁺ T cells contributes to the immunological and pathophysiological processes of HLH without altering the hematological features

In order to test the role of IL-2 consumption by CD8⁺ T cells during hyper-stimulation, we generated CD25^flox mice from EUCOMM ES cells (Supplementary Figure 4), deleted the lacZ and neo with a FLPe deleter strain ²⁶ and intercrossed them with mice bearing the CD8⁺ peripheral enhancer-driven Cre transgene ²⁵. The system preserves the expression of CD25 on Tregs, and allows CD8⁺ T cells access to IL-2 consumption via the constitutively expressed low-affinity dimeric IL-2 receptor, but prevents the activation-induced expression of the high-affinity trimeric IL-2 receptor. The CD8^ΔCD25 mice were then backcrossed to the Prf1^KO background, allowing the comparison of wildtype, Prf1^KO and Prf1^KOCD8^ΔCD25 mice during LCMV-infection. The IL-2 consumption model is based on the observation that CD8⁺ T cells from LCMV-infected Prf1^KO mice express CD25 levels greater than those observed on Tregs, resulting in a rewiring of IL-2 consumption from Tregs to CD8⁺ T cells and a collapse in Treg number ²¹. We therefore first checked whether the CD8^ΔCD25 strategy prevented rewiring of IL-2 consumption. Few CD8⁺ T cells express CD25 in wildtype mice, even 10 days after LCMV infection (Figure 4A). Prf1^KO mice, by contrast, show a dramatic increase, with CD25 expressed by >50% of CD8+ T cells. This increase was absent in Prf1^KOCD8^ΔCD25 mice, which, akin to wildtype mice, showed no prominent CD25⁺ population (Figure 4A). The effect of CD25 expression on IL-2 consumption (as measured through phosphorylation of STAT5) was dramatic: wildtype mice demonstrated no notable phosphoSTAT5 in CD8⁺ T cells (Figure 4C) and a high level in Tregs (Figure 4B), while Prf1^KO mice exhibited a dramatic increase in phosphoSTAT5 in CD8⁺ T cells and a concomitant decrease in Tregs (Figure 4B, C). Prf1^KOCD8^ΔCD25 mice showed an almost complete rescue of this IL-2 consumption rewiring, with phospho-STAT5 patterns recapitulating the wildtype situation (Figure 4C, D). Notably, this system did not completely eliminate CD8⁺ T cell access to IL-2 signaling, with phospho-STAT5 levels marginally higher than wildtype mice (Figure 4C), rather it just corrected the exuberant rewiring occurring in Prf1^KO mice. We previously hypothesis that this “IL-2 theft” was responsible for the observed reduction in Treg numbers ²¹, our results here formally demonstrate this, with an increase of IL-2 availability in the serum (Figure 4E) and a rescue of Treg numbers in Prf1^KOCD8^ΔCD25 mice (Figure 4B). In addition, while evaluating the presence of soluble CD25 (sCD25), a hallmark diagnostic marker of HLH, we found that Prf1^KOCD8^ΔCD25 mice presented with very low amounts of sCD25 in their serum, definitively identifying CD8⁺ T cells as the source of sCD25 during the severe HLH inflammation (Figure 4F). Overall, these data validate the Prf1^KOCD8^ΔCD25 system for functional testing of the role of CD8⁺ T cell-mediated IL-2 consumption on the pathogenesis of HLH.

Figure 4. Removal of CD25 expression on CD8+ T cells prevents rewiring of IL-2 consumption and the collapse of Tregs during HLH.

Wildtype mice, Prf1^KO, Prf1^KOCD8^ΔCD25 mice were infected with LCMV Armstrong, and assessed on day 10. (A) Percentage of CD8⁺ T cells expressing CD25 (n=13,9,16). (B) Percentage of Treg within the CD4⁺ T cell population (n=13,9,16). (C) (left) Percentage of cells positive for phospho-STAT5 staining in CD8⁺ T cells, (middle) geometric MFI for phospho-STAT5 staining on CD8⁺, (right) representative phospho-STAT5 staining on CD8⁺ T cells (n=5,7,7). (D) (left) Percentage cells positive for phospho-STAT5 staining in CD4⁺CD25⁺ Treg cells within the CD4⁺ T cell population, (middle) geometric MFI of phospho-STAT5 staining on CD4⁺CD25⁺ Treg cells and (right) representative phospho-STAT5 staining on CD4⁺Foxp3⁺ Treg cells (n=6,7,7). (E) IL-2 concentration in the serum (n=10,9,12). (F) Soluble CD25 in the serum (n=10,9,12). Analyzed with one-way ANOVA with Tuckey’s multiple comparisons test, SEM and individual data points shown. Results pooled from 4 experiments and 3 experiments for phospho-STAT5 measurements.

To determine the role of IL-2 consumption on the hematological features of HLH, we infected wildtype, Prf1^KO and Prf1^KOCD8^ΔCD25 mice with LCMV and measured hemoglobin, red cells, hematocrit, reticulocytes and platelets at day 10. Unlike IFN-γ-deficiency, which completely rescued the anemic features of Prf1^KO mice (Figure 1), loss of CD25 expression on CD8⁺ T cells had no effect on any parameter, with hematological deficiencies equivalent to LCMV-infected Prf1^KO mice in each case (Figure 5). These results clearly establish the hematological manifestations of HLH as IFN-γ-dependent, CD25-independent.

Figure 5. Correction of the IL-2 consumption phenotype of hyper-activated CD8+T cells does not impact the hematological features of HLH.

Wildtype mice, Prf1^KO, Prf1^KOCD8^ΔCD25 mice were infected with LCMV Armstrong. On day 10, blood samples from each strain were assessed for (A) hemoglobin (n=11, 9, 14), (B) red blood cells (n=10, 9, 14), (C) hematocrit (n=11, 9, 14) (D) reticulocytes (n=10, 8, 14), (E) immature reticulocyte fraction (IRF) (n=8, 6, 11) and (F) platelets (n=11, 9, 14). Analyzed with one-way ANOVA with Tuckey’s multiple comparisons test, SEM and individual data points shown. Results pooled from 4 experiments.

Despite the severe anemia of Prf1^KOCD8^ΔCD25 mice following LCMV-infection, the physiological response to infection was markedly improved over Prf1^KO mice. Prf1^KOCD8^ΔCD25 mice did not develop the distressed behavior observed in Prf1^KO mice at day 10 (data not shown), however they experienced the same reduction in weight (Figure 6A). More critically with regards to pathology, the Prf1^KOCD8^ΔCD25 mice experienced a substantial extension in survival (Figure 6B). This correction of inflammatory pathologies was accompanied by a (at least partial) correction of most of the critical immunological parameters. The CD8⁺ T cell expansion observed in Prf1^KO mice was largely abrogated (Figure 6C), with the shift towards a TEM and SLEC phenotype intermediate in Prf1^KOCD8^ΔCD25 mice compared to Prf1^KO and wildtype mice (Figure 6D, E, F). With the exception of CD69, partial corrections were also observed in PD-1, Ki67 and importantly IFN-γ expression by CD8⁺ T cells (Figure 6G–J). A near complete correction was also observed in the CD4⁺ T cell population, with the shift from naïve to TEM in Prf1^KO mice effectively quenched by negation of IL-2 consumption by CD8⁺ T cells (Figure 6K). By contrast, neither the liver damage observed in LCMV-infected Prf1^KO mice (Supplementary Figure 5) nor the cytokine storm (Supplementary Figure 6) were corrected. Together these data demonstrate that escalated CD25 expression, and the resulting “IL-2 theft”, is one of the major mechanistic drivers of the immunological and pathophysiological features of HLH.

Figure 6. Immunological and pathophysiological processes of HLH are dependent on CD25 expression by CD8⁺ T cells.

Wildtype mice, Prf1^KO, Prf1^KOCD8^ΔCD25 mice were infected with LCMV Armstrong, and assessed for immunological features on day 10. (A) Mouse weights on days 0, 7 and 10 after LCMV infection (n=10,7,13). (B) Survival following LCMV infection (n=8,5,8). (C) Percentage of CD8⁺ T cells in the lymph nodes(n=12,9,16). (D) Percentage and (E) representative flow cytometry for CD8⁺ T cells in the naïve (CD62L⁺CD44⁻), central memory (TCM; CD62L⁺CD44⁺), and effector memory (TEM; CD62L⁻CD44⁺) class (n=12,9,16). (F) Percentage of activated CD8⁺ T cells in the CD8⁺CD127⁺KLRG1^low (MPEC) (n=12,9,16) and CD8⁺CD127⁻KLRG1^high (SLEC) subsets (n=11,9,15). (G) Percentage of CD8⁺ T cells expressing CD69 (n=12,9,16), (H) PD-1 (n=12,9,16), (I) Ki67 (n=13,9,16) and (J) IFN-γ (n=9,9,12). (K) Percentage of CD4⁺ T cells in the naïve (CD62L⁺CD44⁻), central memory (TCM; CD62L⁺CD44⁺), and effector memory (TEM; CD62L⁻CD44⁺) class (n=12,9,16). Analyzed with one-way ANOVA with Tuckey’s multiple comparisons test, SEM and individual data points shown. Results pooled from 3 to 4 experiments.

DISCUSSION

Together these data provide a novel, comprehensive model for the pathogenesis of HLH. A disparate group of disease triggers are possible, each of which initiate the upstream events of CD8⁺ T cell hyper-stimulation. In the case of primary HLH, the initiation event is typically an infectious trigger, with defective elimination of infected cells due to impaired cytotoxicity by CD8⁺ T cells and NK cells, and an induced numerical loss of NK cells ²¹. This in turn drives excessive antigen-presentation to CD8⁺ T cells ⁷, likely due to a combination of elevated viral titers and prolonged antigen stimulation (defective elimination of antigen presentation cells) ¹². In secondary HLH, the initiation trigger varies, but regardless results in hyper-stimulation of CD8⁺ T cells against viral, tumor or self-antigens ^1,5,8. The CD8⁺ T cell hyper-stimulation leads to excessive activation and proliferation of CD8⁺ T cells, creating a hyper-activated state ¹¹. The downstream events directly proximal to pathology here diverge: excessive production of IFN-γ leads to the hematological features, while excessive consumption of IL-2 contributes to the immunological features and ultimately to death due to severe inflammation. The IFN-γ and CD25 pathogenic pathways act independent of each other, with correction of one pathway leaving the other intact. In the mouse model, CD25-deficiency in CD8⁺ T cells gave a partial recovery of the inflammatory components of HLH, demonstrating that CD25 expression, while central, is complemented by additional inflammatory pathways. The remaining mortality in CD25-deficient HLH mice may be due to these residual inflammatory pathways, or to the pathological consequences of the IFN-γ-dependent hematological components. In a translational context, this incomplete recovery is consistent with the incomplete response to cyclopsorin A treatment.

This synthesized model paints a more complex picture for the role of IFN-γ in HLH. While studies focusing on the hematological features have found IFN-γ inhibition to be an effective approach ^9,13,14, our study reveals that it does not prevent the lethal inflammatory signature of HLH in both patients and mouse models. Indeed, IFN-γ, if anything, exhibited regulatory properties over the inflammatory aspects of disease, with enhanced inflammation following genetic removal of IFN-γ in HLH mice. Rather than acting as an inflammatory mediator, it is therefore likely that the key effects of IFN-γ are hematological. The mechanistic pathway for the anaemia is thought to be macrophage-mediated destruction of red cells in the spleen ³⁵, as observed in HLH models ^9,11,13,14. As we also observed a low reticulocyte count (Figure 5), IFN-γ may also function via disrupting the bone-marrow support of erythroid lineage differentiation ³⁶. Thus, while the ongoing clinical trials of anti-IFN-γ antibody in HLH patients are likely to reverse anemia, they may not be effective as a monotherapy, with the inflammatory aspects requiring a distinct approach. This adjunct therapy will need to target the consumption of IL-2 by CD8⁺ T cells, the collapse of Tregs, or the downstream (IFN-γ-independent) inflammatory cascade that drives the immune aspects of the disease. Interestingly, inhibition of universal cytokine receptor signaling JAK pathway has been very recently used with success in the same preclinical model of HLH used here ^33,37. The use of ruxolitinib, a JAK1/JAK2 inhibitor, has been proven to be efficient at controlling CD8⁺ T cells expansion and dampening the cytokine storm in HLH ³³. While this activity has been attributed to IFN-γ inhibition (and, indeed, hematological correction should be attributed to this function), the inhibitor also strongly impacts on other cytokines, including IL-2 which potentially explains the correction of the inflammatory aspects ^33,37. TNF-α in particular is an interesting candidate responsible for the inflammatory cascade downstream of Treg loss, as early experiments have previously demonstrated a long term survival after infection with LCMV in mice doubly knockout for Perforin and TNFR1 ³⁸, and TNF-α levels are reduced in ruxolitinib-treated mice ^33,37. However, as the TNF-α pathway does not depend on JAK½, it is likely that a complex milieu of other cytokines participate upstream.

Finally, we would note that these results have implications beyond HLH. While primary HLH is caused by a rare and specific set of genetic lesions, the diverse causes of secondary HLH demonstrate these principles to be more generic outcomes of excessive hyper-stimulation of CD8⁺ T cells. Indeed, the same pathways (IFN-γ production and IL-2 consumption) that drive clinical pathology in the extreme context during HLH appear to operate at a sub-clinical level during “normal” CD8⁺ activation events. Thus even wildtype mice infected with LCMV exhibit a transient anemia and limited temporal drop in Treg number ³⁹, a feature observed in other strong infection responses ⁴⁰. These deleterious outcomes can thus perhaps be considered the tolerable costs of mounting an effective CD8⁺ T cell response, which only rarely escalate into frank immunopathology.

Key messages.

• In HLH, hematological features are strictly dependent of IFN-γ. However, in both patient and mouse model, absence of IFN-γ signaling does not prevent inflammatory aspect of the disease.

• IL-2 signaling plays a key role in the excessive CD8⁺ T cell activation during HLH inflammation, providing potential therapeutic target for HLH treatment.

Abbreviations used

IFN-γ

Interferon-gamma

HLH

Hemophagocytic lymphohistiocytosis

IL-2

Interleukin 2

IL-6

Interleukin 6

TNF-α

Tumor necrosis factor alpha

NK cells

Natural killer cells

EBV

Epstein Barr virus

CMV

Cytomegalovirus

allo-HSCT

allogeneic hematopoietic stem cell transplantation

PRF1

Perforin

IFNγR1

Interferon-gamma receptor 1

PBMC

peripheral blood mononuclear cells

Treg

regulatory T cell

LCMV

lymphocytic choriomeningitis virus

KO

knockout

SLEC

short-lived effector cells

MPEC

memory precursor effector cells

BCG

Bacille Calmette and Guérin

MSMD

Mendelian susceptibility to mycobacterial disease

STAT

Signal transducer and activator of transcription

EMSA

Electrophoretic mobility shift assay

ES cells

embryonic stem cells

JAK

Janus kinase