Oncogene-induced maladaptive activation of trained immunity in the pathogenesis and treatment of Erdheim-Chester disease

Trained immunity (TI) is a functional program physiologically elicited in monocyte/macrophages upon sensing of specific pathogens, which results in enhanced production of proinflammatory cytokines.¹ Mechanistically, TI is characterized by epigenetic changes promoting transcription of proinflammatory genes,^2,3 as well as immunometabolic changes, such as increased glycolysis, glutaminolysis, and cholesterol synthesis, which sustain enhanced proinflammatory activation.^4,5 TI evolved as a protective mechanism against repeated infection, yet maladaptive activation of TI (ie, in the absence of infection) is emerging as a contributing factor in the pathogenesis of diseases characterized by excessive cytokine production by myeloid cells.^6,7 However, the exact role and extent of inappropriate activation of TI in the pathogenesis of human inflammatory diseases remain undetermined.

Rare diseases characterized by extreme phenotypes often provide invaluable information about the relevance of specific molecular mechanisms to human pathophysiology. Erdheim-Chester disease (ECD) is a rare, severe inflammatory myeloid cancer characterized by infiltration of tissues by foamy macrophages (or histiocytes).^8,9 Because virtually any tissue can be affected, the disease has pleomorphic clinical manifestations, with cardiovascular, pulmonary, and central nervous system involvement conferring a poor prognosis.¹⁰ The pathogenesis of ECD hinges upon 2 different, interconnected mechanisms occurring in monocytes: (1) deregulated activation of the MAPK pathway, most commonly caused by a mutation in the proto-oncogene BRAF (BRAFV600E)^11-13; (2) rampant proinflammatory activation and cytokine production, resembling a systemic “cytokine storm.”^14-16 Of note, mutated myeloid cells infiltrating affected tissues in ECD do not proliferate; however, cytokine and chemokine production results in progressive lesion growth through the recruitment of both mutated and nonmutated myeloid cells.¹⁷ Although this dual clonal-inflammatory nature of ECD has long fascinated scientists,^17,18 the mechanistic link between deregulated activation of the MAPK pathway and downstream production of proinflammatory cytokines is still undetermined. Of note, monocytes that activate TI programs and ECD macrophages exhibit striking functional and phenotypic similarities: besides abundant cytokine production, both exhibit increased cholesterol synthesis that leads to a large size and a foamy appearance^5,6 and rely on activation of the MAPK pathway for inflammatory activation.^2,9

In this study, we hypothesized that induction of a maladaptive TI phenotype represents the missing link between clonal transformation and proinflammatory activation in ECD. The extreme phenotype of ECD illustrates how maladaptive activation of TI can result in the development of human inflammatory myeloid neoplasms, while also revealing the therapeutic potential of TI inhibition in conditions characterized by deregulated production of cytokines by myeloid cells.

This study was approved by the Institutional Review Board of San Raffaele Hospital (approval 100/INT/2020) and was conducted in accordance with the Declaration of Helsinki.

Peripheral blood mononuclear cells (PBMCs) were purified from buffy coats of healthy donors, provided by the Institutional Blood Bank. Monocytes were isolated by negative selection with magnetic beads (details are in the supplemental Material, available on the Blood Web site).

Lentiviral constructs and production are described in the supplemental Material.

Purified monocytes were transduced by adding noninfectious viruslike particles carrying Vpx (Vpx-VLPs) together with lentiviral constructs encoding for either BRAFWT or BRAFV600E at multiplicity of infection 1 (details are in the supplemental Material).¹⁹ 

For macrophage differentiation, human monocytes isolated from PBMCs were cultured at 37°C for 6 to 7 days in complete RPMI 1640 medium supplemented with 20 ng/mL mouse macrophage colony-stimulating factor (Miltenyi Biotec). Medium was replaced every 2 days.

Flow cytometric analysis is described in the supplemental Material.

Oil Red-O staining was performed on ECD histiocytes, freshly isolated from a skin lesion upon liberase TM (12.5 μg/mL; Roche) treatment, and on myelomonocytic cells obtained and cultured for 7 days as just described (see details in the supplemental Material).

Cell morphology, BRAFWT/BRAFV600E expression, and Oil Red-O staining were examined by wide-field microscopy. Appropriate fluorescence filters were used to detect GFP-fused proteins and 4′,6-diamidino-2-phenylindole (DAPI)-stained nuclei. Images were taken with a 40× objective on a Zeiss AxioImagerM2m (AxioCam MRc5).

Immunohistochemistry of ECD tissues was performed as previously reported,^15,20 using the antibodies anti-human interleukin-6 (IL-6; R&D System, Minneapolis, MN) and anti-phospho-p44/42 MAPK ERK1/2 (Thr202/Tyr204; Cell Signaling Technology, Danvers, MA).

Methods and reagents for western blot analysis are described in the supplemental Material.

Cells were cultured for 6 days as described in “Monocyte differentiation into macrophages.” RNA was extracted from 1 × 10⁶ cultured cells or PBMCs isolated from the blood of patients with ECD by using ReliaPrep RNA Miniprep Systems (Promega Corporation) according to the manufacturer’s instructions. RNA from formalin-fixed, paraffin-embedded biopsy tissue was extracted upon deparaffinization using the AllPrep DNA/RNA FFPE Kit (Qiagen) according to the manufacturer’s instructions. RNA was quantified with The Qubit 2.0 Fluorometer (ThermoFisher), and its quality was assessed by the 4200 TapeStation (Agilent Technologies). Minimum quality was defined as RNA integrity number >7. One nanogram of total RNA was used for library preparation with SMART-Seq v4 Ultra Low Input RNA Kit (Takara Bio USA) and sequenced on a NextSeq 500 (Illumina). Total RNA sequencing and analysis are detailed in the supplemental Material.

Oxygen consumption rate and extracellular acidification rate (ECAR) were measured with a Seahorse XFe96 Analyzer (Agilent Technologies, Santa Clara, CA), using the SeaHorse XF Cell Mito Stress Test and Glycolysis Stress Test, according to the manufacturer’s instructions. Detailed procedures are reported in supplemental Material.

Metabolomics studies were performed as previously reported.^21,22 Detailed procedures are reported in supplemental Material.

Cytokine concentrations in supernatants and sera were determined with commercial enzyme-linked immunosorbent assay kits for human tumor necrosis factor α and IL-6 (both from R&D Systems, Minneapolis, MN), according to the manufacturer’s instructions (see details in supplemental Material).

In situ hybridization was performed according to the smiFISH (single-molecule inexpensive FISH) approach, where unlabeled primary probes are prehybridized to a secondary common fluorescently labeled probe.²³ Detailed procedures are reported in the supplemental Material.

Cells were cultured for 6 days, as described in “Monocyte differentiation into macrophages,” and fixed with 1% formaldehyde (Sigma) at a concentration of 10⁶cells per milliliter. Fixed cell preparations were sonicated, and immunoprecipitation was performed with antibodies against histone 3 lysine 4 trimethylation (H3K4me3) or histone 3 lysine 27 acetylation (H3K27Ac) (Diagenode). After chromatin immunoprecipitation, DNA was processed for quantitative polymerase chain reaction analysis (see details in supplemental Material).

Treatment with the selective BRAFV600E inhibitor vemurafenib was administered at a dose of 480 mg daily, as per clinical recommendations.⁸ 

For preparation of patients and image acquisition strategy, see the supplemental Material.

ECD skin biopsy samples were obtained for diagnostic purposes and cultured in the RCCS bioreactor (Synthecon, Houston, TX), as previously described.²⁴ In brief, biopsy tissue was cut into 10 mm³ fragments and cultured (3 fragments/vessel) for 4 days in the presence or absence of the MAPK pathway inhibitor trametinib (GSK1120212; 1 nM). At the end of culturing, supernatants were collected for metabolite determination.

Statistical significance was determined as stated in the figure legends. P < .05 indicated statistically significant differences. Data were analyzed with GraphPad Prism 5.0. Data are shown as the mean ± standard error of the mean (SEM).

ECD is a rare condition, and only a negligible subfraction of circulating monocytes of patients with ECD bear the causative BRAFV600E mutation, thus making it impractical to perform studies on primary cells. We thereby developed an innovative model to recapitulate the pathogenesis of ECD in vitro, by ectopically expressing BRAFV600E in primary human monocytes. Specifically, we developed a lentiviral vector expressing BRAFV600E and GFP under the ubiquitous human phosphoglycerate kinase promoter (Figure 1A). A lentiviral vector encoding wild-type BRAF provided an essential control, ensuring that observed biological effects are not caused by BRAF overexpression. This strategy, schematically represented in Figure 1B, yielded efficient delivery of the BRAFV600E transgene into primary human monocytes, whereas the reporter gene GFP allowed for the identification of cells expressing the transgene. GFP tracking by flow cytometry demonstrated effective transduction of BRAFV600E into healthy human monocytes (Figure 1C). Of note, forward and side scatter rates were significantly increased in monocytes transduced with BRAFV600E (Figure 1D), consistent with an increase in cell size and granularity. Bright-field microscopy studies confirmed that monocytes transduced with BRAFV600E underwent morphologic changes leading to a foamy appearance (Figure 1E). Notably, phenotypic changes also included an increase in cytoplasmic lipid content (Figure 1E; Oil Red O staining) and were overall highly reminiscent of foamy macrophages isolated from lesion tissue of patients affected by ECD (Figure 1F). The BRAFV600E mutation in ECD histiocytes results in constitutive activation of the downstream MAPK pathway intermediates ERK1 and ERK2^25,26 (Figure 1G). Western blot analysis confirmed that phosphorylation of ERKs occurs in myeloid cells transduced with BRAFV600E (Figure 1H). Constitutive activation was specific and restricted to the ERK1/2 MAPKs, as p38 MAPK phosphorylation was not increased in cells expressing BRAFV600E.

Macrophages in ECD lesions (Figure 2A) or isolated from the blood of patients with ECD (Figure 2B) exhibited deregulated spontaneous production of proinflammatory cytokines such as IL-6. These findings were recapitulated in macrophages expressing BRAFV600E, which exhibited spontaneous production of IL-6 (Figure 2C) and TNFα (Figure 2D). Consistent with this evidence, advanced microscopy studies using RNA smFISH confirmed that macrophages transduced with BRAFV600E exhibit spontaneous transcription of IL6 messenger RNA (mRNA), comparable to levels observed in primary macrophages stimulated with TNFα (Figure 2E-F). These findings indicate that macrophages expressing ectopic BRAFV600E exhibit constitutive activation of the MAPK pathway, transformation into foamy macrophages with a large lipid-laden cytoplasm, and production of proinflammatory cytokines, thus recapitulating genetic, phenotypic, and functional features of ECD macrophages.

To study the transcriptional effects induced by BRAFV600E in macrophages, we performed whole-transcriptome analyses on transduced cells. Principal component and hierarchical clustering analysis of gene expression datasets revealed that the overall transcriptome of nontransduced and wild-type BRAF (BRAFWT)–expressing cells was similar and clustered together; conversely, the transcriptome of cells transduced with BRAFV600E clustered independently from controls (Figure 3A-B). Specifically, segregation of BRAFV600E-expressing cells displayed 5906 significantly differentially expressed genes (DEGs; 2906 upregulated and 3000 downregulated; false discovery rate [FDR] <0.01) compared with the controls (Figure 3C). We next confirmed that transcriptional effects induced by BRAFV600E in macrophages in vitro effectively recapitulate those spontaneously occurring in ECD lesions in vivo. Because myelomonocytic cells expressing BRAFV600E constitute up to 10% to 50% of cells in ECD tissue specimens, but only a marginal fraction (<1%) of circulating PBMCs,²⁰ we performed whole-transcriptome analyses on matched tissue specimens and PBMCs from 3 patients with ECD to reveal the transcriptional effects induced by BRAFV600E in vivo. As expected, ECD tissue samples and PBMCs exhibited profound transcriptional differences (Figure 3C-D), with 4461 DEGs (3070 upregulated and 1391 downregulated (FDR <0.01; Figure 2E). Gene set enrichment analysis (GSEA) against Hallmark Gene Sets from Molecular Signatures Database (v7.1) revealed that DEGs identified in macrophages transduced with BRAFV600E and in ECD lesions were enriched for multiple shared categories (Figure 3G), notably including inflammatory activation (ie, inflammatory response) and immunometabolic activation (ie, hypoxia, glycolysis). Pathways associated with DNA damage responses (UV response) and mitogenic signaling (KRAS signaling) were also consistently activated, in line with expression of the BRAFV600E oncogene. A comparison between DEGs in macrophages transduced with BRAFV600E and ECD tissue biopsies confirmed tight consistency, again indicative of effective recapitulation of the disease phenotype (Figure 3H). Some differences in gene expression signatures were also observed (most notably related to interferon signaling) that likely reflect the cellular homogeneity in vitro vs the architectural and functional complexity of the lesion tissue microenvironment.

TI is characterized by distinctive metabolic changes that sustain enhanced proinflammatory activation of monocyte/macrophages. Typical immunometabolic changes of TI induced by β-glucan include increased glycolysis, increased glutaminolysis through the tricarboxylic acid (TCA) cycle and increased cholesterol synthesis.^4-627 We thereby investigated whether expression of BRAFV600E induces such immunometabolic changes in macrophages.

Induction of aerobic glycolysis through an AKT/mammalian target of rapamycin (mTOR)–dependent pathway is the metabolic basis of TI.⁴ Western blot analysis of myelomonocytic cells transduced with BRAVF600E revealed constitutive AKT phosphorylation (Figure 4A), and Seahorse Flux analyzer confirmed downstream induction of glycolysis (Figure 4B). Specifically, macrophages expressing BRAFV600E showed increased glycolysis compared with cells transduced with BRAFWT, as assessed by ECAR (Figure 4B-C; supplemental Figure 1). In addition, macrophages expressing BRAFV600E exhibited increased glycolytic capacity after glucose supplementation (supplemental Figure 1).

To investigate whether BRAFV600E induces immunometabolic rewiring, transduced cells were cultured in either plain culture medium or culture medium enriched with ¹³C₆-glucose or ¹³C₅-glutamine and were subjected to metabolomics studies (Figure 4D). Similar to gene expression studies, principal component analysis revealed that the metabolome of control and BRAFWT-expressing cells was comparable and clustered together (Figure 4E). Conversely, the metabolome of cells expressing BRAFV600E clustered independently from controls, indicative of profound metabolic rewiring (Figure 4E). Hierarchical clustering analysis of the top 50 significant metabolites grouped by pathways relevant to TI confirmed that BRAFV600E induced compatible and extensive immunometabolic rewiring (Figure 4F). Tracing experiments using either ¹³C₆-glucose or ¹³C₅-glutamine allow for determining incorporation of these metabolites into specific intermediates and their fueling of metabolic pathways. BRAFV600E induced marked activation of the glycolysis pathway, as revealed by increased generation of glycolysis intermediates as well as the end products pyruvate and lactate (Figure 4G). An additional immunometabolic hallmark of TI (that is, increased glutaminolysis fueling the TCA cycle) was also observed in macrophages expressing BRAFV600E, as determined by incorporation of ¹³C₅-glutamine–derived carbons into the TCA intermediates α-ketoglutarate, citrate, malate, fumarate, and succinate (Figure 4H). Increased cholesterol synthesis is an additional immunometabolic hallmark of trained myeloid cells. Tracking incorporation of carbons derived from ¹³C₆-glucose or ¹³C₅-glutamine revealed increased de novo cholesterol synthesis in macrophages transduced with BRAFV600E, mostly fueled by increased glycolysis (Figure 4I). Of note, increased cholesterol synthesis was also observed in macrophages from patients with ECD, and resulted in a foamy, lipid-laden appearance (Figure 4J; compare with Oil Red-O staining in Figure 1F). Accumulation of intracellular lipids in macrophages expressing BRAFV600E was paralleled by upregulation of genes involved in lipid uptake, synthesis, and storage (ie, PPARγ, SREBF1, SREBF2, and ACSL4) and downregulation of genes related to lipid export (ie, CD36, ABCA1; supplemental Figure 2). Overall, these findings indicate that BRAFV600E in macrophages induces hallmark immunometabolic features of TI by exposure to β-glucan, such as glycolysis, increased glutaminolysis through the TCA cycle, and increased cholesterol synthesis.^4-627 

In TI, metabolites generated by immunometabolic changes activate chromatin-modifying enzymes, leading to gene-specific chromatin modifications associated with enhanced transcription of genes encoding inflammatory cytokines.²⁸ Two key epigenetic marks characterize TI: H3K27Ac at distal enhancers and H3K4me3 at promoters of genes encoding cytokines.^1,29 We therefore investigated epigenetic changes and enhanced cytokine production in myelomonocytic cells expressing BRAFV600E.

Macrophages expressing ectopic BRAFV600E exhibited the hallmark epigenetic features of TI on the promoters of genes encoding IL-6, and TNFα (Figure 5A-D). These epigenetic modifications altered the chromatin configuration that maintained these genes transcriptionally repressed in quiescent cells, thus increasing chromatin accessibility to transcription and enabling more robust cytokine production in response to stimulation with pathogens. Indeed, functional validation to the activation of TI in macrophages expressing BRAFV600E came from assessment of cytokine production after secondary stimulation with LPS, which simulates a subsequent encounter with a pathogen. Compared with native homologues, macrophages expressing the BRAFV600E mutation exhibited strikingly enhanced production of IL-6 (Figure 5E) and TNFα (Figure 5F) after stimulation with LPS. These findings indicate that the BRAFV600E oncogene in macrophages induces the typical epigenetic changes of TI, resulting in the hyperresponsiveness to inflammatory triggers, which represents the functional hallmark of TI.

Evaluation of patients with ECD with fluorodeoxyglucose-positron emission tomography (FDG-PET), an imaging technique that measures glucose uptake by cells, revealed that macrophages infiltrating ECD lesions were characterized by marked metabolic activation and avid glucose uptake, a finding consistent with increased glycolysis induced by BRAFV600E (Figure 6A-B left). The selective BRAFV600E inhibitor vemurafenib is used clinically for the treatment of ECD. Treatment with vemurafenib resulted in metabolic shutdown of lesions (Figure 6A-B right), thus confirming that BRAFV600E is causative of the induction of glycolysis in ECD macrophages. Of note, previous studies from our group showed that vemurafenib treatment does not affect the viability of ECD macrophages, thus indicating that metabolic changes at FDG-PET are not attributable to the death of macrophages bearing the BRAFV600E mutation.²⁴ Increased glucose metabolism at FDG-PET in patients with ECD is paralleled by high serum levels of the proinflammatory cytokines TNFα and IL-6, which were reduced upon treatment with vemurafenib (Figure 6C), again substantiating the causative link between BRAFV600E and macrophage proinflammatory activation in vivo. To explore the immunometabolic mechanisms linking BRAFV600E and proinflammatory activation of ECD macrophages, we used the bioreactor technology,^17,24 which allows for long-term 3D culture of tissue biopsies obtained for diagnostic purposes (supplemental Figure 1). With this technique, we cultured ECD biopsy specimens for 4 days, with or without MAPK inhibition, and changes in the metabolome were assessed in supernatants. Metobolomics studies revealed that the hallmark immunometabolic changes of TI (ie, increased glycolysis and accumulation of fumarate and malate) occurred in ECD lesions in vivo; again, MAPK pathway inhibition dampened these metabolic effects (Figure 6D).

Finally, we evaluated the therapeutic potential of inhibiting TI for suppressing inflammation in ECD. The glycolysis inhibitor 2-deoxyglucose (2DG) inhibits activation of TI.⁴ Macrophages expressing BRAFV600E were cultured in the presence or absence of vemurafenib or 2DG for 48 hours, and supernatants were assessed for cytokine content. Treatment with 2DG effectively reduced production of IL-6 (Figure 6E). Of note, treatment with vemurafenib also reduced this cytokine, thus recapitulating the effects observed in patients with ECD. Treatment with the mTOR inhibitor rapamycin did not suppress cytokine production (Figure 6E), despite at least partial inhibition of glycolysis (supplemental Figure 4). Treatment with the glutaminase inhibitor BPTES, which inhibits glutaminolysis, resulted in a slight but significant decrease in cytokine production (Figure 6F); however, cell toxicity limited exploration of BPTES treatment at higher doses.

Because disease almost always recurs in patients with ECD upon discontinuation of MAPK inhibitors, we also investigated whether inhibition of TI would have more sustained effects in lowering proinflammatory activation. Macrophages expressing BRAFV600E were treated with vemurafenib or 2DG, and cytokine production was assessed after 48 hours; then, treatment-free culture medium was replaced and cytokine production was reassessed at 72 hours. As expected, the effects of vemurafenib on cytokine production waned upon treatment withdrawal, whereas suppression of cytokine production obtained with 2DG was maintained at 72 hours (supplemental Figure 5).

These findings indicate that effective therapeutic strategies for ECD combat the TI phenotype and that pharmacologic inhibition of glycolysis with 2DG effectively dampens cytokine production by macrophages expressing BRAFV600E.

This study indicates that an oncogenic mutation along the MAPK pathway, specifically BRAFV600E, can induce a TI phenotype in myeloid cells in a cell-intrinsic, pathogen-independent fashion. This maladaptive, oncogene-induced activation of TI is relevant to the development of human disease, as illustrated by ECD, a severe inflammatory myeloid neoplasm characterized by excessive production of cytokines by macrophages. Therapeutic modulation of TI results in suppression of excessive cytokine production and in potential therapeutic applications for this and other conditions characterized by deregulated production of cytokines by myeloid cells.

TI is a proinflammatory program of myeloid cells that evolved to provide protection against repeated infections. Recognition of several pathogens can induce TI, but induction by β-glucan, a component of the cell wall of Candida albicans, was first and best characterized.^2,28 β-Glucan binds the pattern recognition receptor dectin-1, which activates the AKT/mTOR signaling axis and induces changes in cell energy metabolism, such as increased glycolysis, glutaminolysis through the TCA cycle, and cholesterol synthesis.^4-627 Metabolic intermediates derived from these processes, such as fumarate and acetyl-CoA, modulate the activity of enzymes involved in remodeling the epigenetic landscape of myeloid cells.²⁸ Two key epigenetic changes ensue: an increase in H3K27Ac at distal enhancers and in H3K4me3 at promoters of genes encoding cytokines. Thus, induction of TI programs enables myeloid cells to deploy stronger transcription of proinflammatory genes, while also providing the energy that is necessary to sustain enhanced inflammatory activation. Functionally, this results in hyperresponsiveness to inflammatory triggers.

In this study, all these mechanisms and their functional consequences were recapitulated in macrophages expressing BRAFV600E. Although, strictly speaking, the BRAFV600E genetic mutation in macrophages does not induce “innate immune memory,” as the mutation is obviously persistent, the functional and molecular changes undergone by the ECD macrophages are all indicative of a maladaptive TI phenotype. Specifically, BRAFV600E induced activation of the AKT/mTOR signaling axis; increased glycolysis, glutaminolysis, and cholesterol synthesis; caused typical epigenetic changes in genes encoding IL-6 and TNFα; and enhanced constitutive cytokine production and hyperresponsiveness to stimulation with LPS, which is the functional hallmark of TI. This line of evidence indicates that TI in ECD is a nodal pathogenic mechanism, a link between oncogenic mutation and immunometabolic activation of macrophages, and a novel therapeutic target. These findings hold translational relevance, as the clinical effectiveness of vemurafenib and other kinase inhibitors in ECD is not fully satisfactory because of their remarkable toxicity, incomplete responses, or disease recurrence upon discontinuation.^26,30,31 Combination therapy with kinase inhibitors and cytokine-blocking agents has been proposed in ECD, to target oncogenic as well as inflammatory activation.^33-38 By targeting a novel, causal pathway in the pathogenesis and inflammatory activation of ECD, inhibition of TI has a strong clinical rationale and may attain results comparable to this dual pharmacologic approach.

Because immunometabolic circuits in TI are typically redundant, we screened multiple inhibitors as potential therapeutic agents for ECD. The glycolysis inhibitor 2DG effectively and persistently suppressed cytokine production, whereas the glutaminolysis inhibitor BPTES yielded more limited responses. A concept emerges that although metabolites generated during glutaminolysis and TCA cycle activation (ie, fumarate) induce the epigenetic modifications of TI, glycolysis represents the main energy supply mechanism of macrophages during hyperinflammatory activation. Indeed, although glycolysis is a relatively inefficient pathway for the generation of cellular adenosine triphosphate (ATP), netting only 2 molecules of ATP for every unit of glucose, it allows for rapid generation of ATP and biosynthetic precursors, both of which are immediately needed to enhance cytokine production.³⁹ Of note, the mTOR inhibitor rapamycin did not suppress cytokine production, despite at least partial inhibition of glycolysis. This finding is consistent with previous evidence of pleiotropic and mixed anti-inflammatory and proinflammatory effects of rapamycin on myeloid cells,⁴⁰ as well as with sparse clinical responses of patients with ECD treated with rapamycin.⁴¹ Overall, our findings identified the glycolysis inhibitor 2DG as a promising treatment for further development, either as a monotherapy or in combination with current strategies.

Our study has some limitations. Given the rarity of ECD, we developed a model to recapitulate the disease pathogenesis and phenotype in vitro, which was accomplished through lentiviral transduction of BRAFV600E. For practical reasons, we used primary human monocytes for our experiments; however, oncogenic mutations in patients with ECD affect hematopoietic precursors in the bone marrow.^17,42-44 It is possible that transduction of differentiated cells would yield some differences in phenotypic output.⁴² Of note, this limitation marks an additional analogy with TI: although it is typically induced in vitro by stimulating monocytes obtained from the peripheral circulation, immunometabolic and epigenetic changes characteristic of TI in vivo occur in myeloid progenitors.⁴⁵ Strengths of this study include the development of an innovative, clear-cut experimental model to recapitulate the pathogenesis of an otherwise difficult-to-study disease and the use of a broad array of relevant methodologies, to capture the intricacy of biological mechanisms underlying TI and the pathogenesis of ECD. Despite the complexity and the adoption of different experimental angles, our findings point at the same conclusion: oncogenic activation of the MAPK pathway by BRAFV600E induces a maladaptive TI phenotype in myeloid cells.

This study provides proof of concept for the hypothesis that maladaptive activation of TI leads to hyperinflammation and pathological tissue damage in human diseases.⁴⁶ The role and extent of inappropriate activation of TI in the pathogenesis of ECD implicate investigational opportunities in other, related human inflammatory myeloid neoplasms (ie, Langerhans cell histiocytosis).^28,42,47 Equally important, therapeutic opportunities may arise from pharmacologic targeting of TI in these and other conditions characterized by maladaptive, myeloid-driven inflammation.

The authors thank the Advanced Light and Electron Microscopy BioImaging Center (ALEMBIC) for support with image acquisition and analysis, the Erdheim-Chester Disease Global Alliance (ECDGA) for support of patients and physicians fighting ECD, and the Eureka Institute for Translational Medicine.

G.C. is supported by Italian Association for Cancer Research - AIRC under MFAG 2018 (ID 22136 project; principal investigator [PI], G.C.), FOREUM (Career Award 2020; PI G.C.), and the Interleukin Foundation. M.G.N. is supported by The Netherlands Organization for Scientific Research (Spinoza grant) and by the European Research Council (Consolidator Grant).

Contribution: G.C. conceived the study; R.B., R.M., D.S., T.N., J.D.-A., R.J.A., I.M., D.M., S.Z., M.P., E.C., I.W.T., F.P., D.C., A.L., D.G., L.C., G.D., C.D., L.A.B.J., A.K.-R., C.A.D., M.F., and E.F. performed the experiments and analyzed data; P.M., G.D., J.H., G.D.L., A.T., C.C., and L.D. took clinical care of the patients; L.A.B.J., A.K.-R., C.A.D., S. Cardaci, A.B., S. Cenci, A.D., E.M., M.G.N., and G.C. supervised the study; and R.B., R.M., M.G.N., and G.C. drafted the manuscript.

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

Correspondence: Giulio Cavalli, IRCCS San Raffaele Scientific Institute, Via Olgettina 60, 20132 Milan, Italy; e-mail: cavalli.giulio@hsr.it.