Circulating CD1c⁺ myeloid dendritic cells are potential precursors to LCH lesion CD1a⁺CD207⁺ cells

Langerhans cell (LC) histiocytosis (LCH) is a disorder characterized by the accumulation of pathogenic dendritic cells (DCs; LCH DCs) in inflammatory lesions. Clinical presentations are highly variable, ranging from single lesions to potentially lethal disseminated disease.¹  Clinically, LCH is defined by high-risk (lesions involving liver, spleen, and/or bone marrow involvement) and low-risk disease (all other sites) based on risk of death.²  LCH lesion biopsies are characterized by CD1a⁺CD207⁺ “histiocytes” among extensive inflammatory infiltrate with common histology regardless of extent of disease or clinical risk.^3,4  Mutually exclusive activating somatic mutations in MAPK pathway genes localizing to LCH lesion CD1a⁺CD207⁺ DCs have been identified in ∼85% cases, with BRAFV600E mutation as the most common.^3,5-10 

Despite recent advances in identifying mutations in LCH, little is known about the differentiation pathway(s) and cell(s) of origin of LCH CD1a⁺CD207⁺ DCs. Historically, pathologic LCH lesion CD1a⁺CD207⁺ DCs were proposed to arise from CD207⁺ epidermal LCs or along a parallel pathway.^11,12  However, transcriptome analysis of LCH lesion CD1a⁺CD207⁺ DCs identified increased expression of genes associated with early myeloid differentiation compared with epidermal LCs.^13-15  Furthermore, despite early reports of langerin/CD207 expression being unique to epidermal LCs, multiple alternative CD207-expressing DC populations have been discovered,¹⁶  highlighting the possibility of multiple potential differentiation pathways for the LCH lesion CD1a⁺CD207⁺ DCs. Lineage-mapping experiments identified the BRAFV600E allele in CD34⁺ hematopoietic stem cells and peripheral blood of patients with high-risk LCH. The BRAFV600E allele in peripheral blood mononuclear cells (PBMCs) localized primarily to myeloid populations (eg, CD11c⁺ cells and CD14⁺ cells), though in some cases, it was also detected in other lineages, including lymphoid cells.^3,17  In a mouse model, enforced expression of BRAFV600E mutation in CD11c⁺ cells recapitulated a phenotype with features of high-risk LCH, including development of lesions with CD1a⁺CD207⁺ DCs.³  These findings suggest that pathogenic LCH lesion CD1a⁺CD207⁺ DCs may arise from myeloid cell precursors with an activating MAPK gene mutation.^1,3 

The pathway from CD34⁺ hematopoietic stem cell or early myeloid precursor to LCH lesion CD1a⁺CD207⁺ DCs is not known. Monocytes are the most abundant myeloid mononuclear cells, comprising ∼10% of PBMCs. Blood monocyte populations include CD14⁺CD16⁻ “classical” monocytes and less abundant CD16⁺ monocytes.^18,19  Classical monocytes have the ability to differentiate into tissue DCs and macrophages in vitro (reviewed in Collin and Bigley¹⁹ ). CD16⁺ monocytes have been further classified into 2 subpopulations based on CD16 and CD14 expression, with CD16⁺⁺CD14⁺ “nonclassical monocytes” more abundant than CD16⁺CD14⁺ “intermediate” monocytes.^18,20  Intermediate monocytes, which constitute a very small percentage in blood, appear to have both phagocytic and inflammatory functions, whereas nonclassical monocytes are more limited to inflammation with ability to produce tumor necrosis factor α (TNF-α) and interleukin 1β (IL-1β) upon activation.²¹  Peripheral blood DCs are rarer than monocytes (1% to 2% of PBMCs), including plasmacytoid DCs (pDCs) and myeloid DCs (mDCs).^19,22,23  mDCs express antigens including CD11c, CD13, CD33, and CD11b, but lack monocyte antigens CD14 and CD16. This population may be further split into CD141⁺ mDCs (cDC1s) and CD1c⁺ mDCs (cDC2s). CD1c⁺ mDCs (cDC2s) are the major DCs in peripheral blood, tissues, and lymphoid cells (reviewed in Collin and Bigley,¹⁹  Collin and Bigley,²²  and Collin et al²³ ). CD1c⁺ mDCs from peripheral blood have less “activated” phenotype compared with the tissues CD1c⁺ mDCs (reviewed in Collin and Bigley,¹⁹  Collin and Bigley,²²  and Collin et al²³ ). In general, CD1c⁺ mDCs have low intrinsic capacity to cross-present antigen to CD8⁺ T cells, but they have enhanced capacity to stimulate naive CD4⁺ T cells.^24-28  As for CD141⁺ mDCs, they have high capacity to cross-present antigens to CD8⁺ T cells, mediate efficient recognition of viral antigens, and engage necrotic cells via CLEC9A.^24-27,29,30 

Both CD14⁺ monocytes and CD1c⁺ mDCs may acquire a CD1a⁺CD207⁺ DC phenotype in vitro.^14,15,17,31  This is in line with ontogeny of the epidermal LCs where epidermal LC precursors are formed during embryonic and fetal hematopoiesis and self-renew in the steady state, but may be replaced by phenotypically and functionally indistinguishable bone marrow–derived myeloid cells.^32-35  Similarly, dermal DCs that express CD207 may arise from common DC precursors (CDPs), macrophage DC precursors (MPPs), as well as circulating hematopoietic stem cells (HSCs) and monocytes.^16,36-38  Given the plasticity of mononuclear cells and the broad range of tissue distribution, there are many candidates for cell of origin and differentiation pathways leading to LCH CD1a⁺CD207⁺ DCs and associated lesion formation (Figure 1).

In this study, we sought to define potential origins of LCH lesion CD1a⁺CD207⁺ DCs as an opportunity to identify potential therapeutic targets for LCH and also to define the impact of MAPK pathway activation in myeloid differentiation.

This study was performed according to protocols approved by the institutional review boards of Baylor University College of Medicine, Singapore Singhealth, and National Health Care Group research ethics committees. Peripheral blood and biopsy specimens were obtained from LCH patients (supplemental Table 1a-b). LCH biopsy specimens were identified through clinical diagnosis, including CD207⁺ immunohistochemistry. Control skin specimens were obtained from discarded tissues from elective procedures, and control peripheral blood specimens were collected from healthy pediatric donors without blood or cancer disorders or inflammatory conditions.

Peripheral blood specimens from healthy donors and LCH patients as well as LCH biopsy samples were collected and processed as described previously.¹³  Antibodies used for LCH biopsy samples and peripheral blood specimens are listed in supplemental Table 2. Fluorescence-activated cell sorting (FACS) was performed using BD FACS Aria Fusion and BD FACSCanto and data analyzed with FCS Express 6 (De Novo Software, Glendale, CA) and FlowJo (TreeStar, Ashland, OR).

For genomic DNA extraction, FACS-sorted populations were collected in ice-cold phosphate-buffered saline supplemented with 1 mM EDTA and 1% bovine serum albumin. Genomic DNA was isolated from sorted subpopulations from peripheral blood and LCH lesion specimens using QIAamp DNA micro with RNAse A treatment (Qiagen, Valencia, CA) and then amplified with the REPLI-g Midi kit per the manufacturer’s instructions. The BRAF-V600E mutation assay (Somatic Mutation Assay for BRAF_476; Qiagen) was performed as described previously³  with 30 ng of amplified genomic DNA. Triplicate reactions were performed for each sample. All experiments were performed on a CFX96 real-time system (Bio-Rad Laboratories, Hercules, CA). All samples that were used in this analysis are listed in supplemental Table 3b.

Total RNA was purified from sorted subpopulations from peripheral blood and lesion specimens according to the Qiagen RNeasy Micro kit (Qiagen). RNA integrity was determined using Agilent Bioanalyzer, and the RNA integrity numbers were calculated. Biotinylated complementary RNA was prepared according to the protocol by Epicentre TargetAmp 2-Round Biotin-aRNA Amplification kit 3.0 using 500 pg of total RNA. Hybridization of complementary DNA (cDNA) was performed on Illumina Human-HT12 version 4 chips (Illumina, San Diego, CA).

Array data were extracted at the probe set level with no background subtraction using Illumina’s BeadStudio software. These raw data were then normalized by the quantile method using the lumi package in R/Bioconductor v2.13.1. A part of this data was previously reported in Haniffa et al²⁴  and McGovern et al³⁹  and the data set can be found in the Gene Expression Omnibus data repository (GSE35457 and GSE85305).

For generation of human myeloid subpopulation gene signatures for connectivity map (CMAP) analysis⁴⁰  as previously described in Haniffa et al,²⁴  1 cell subset was compared with other cell subsets pooled using the Student t test in R statistical software. Differentially expressed genes (DEGs) were selected with a Benjamini-Hochberg (BH) multiple testing⁴⁰  corrected P < .05. CMAP analysis⁴⁰  was performed comparing myeloid cell signature gene subsets with the LCH lesion CD1a⁺CD207⁺ DC gene-expression data after removal of the tissue-specific probes. The samples used in this analysis are listed in supplemental Table 3a.

Hierarchal clustering was performed by comparing the expression profiles across the set of samples using 1 − (centered) correlation for the distance metric with average linkage clustering. All samples used in this analysis are listed in supplemental Table 3a.

BubbleGUM software as described in Spinelli et al⁴¹  was used to perform multiple gene set enrichment analysis (GSEA) on all possible pairwise comparisons. A GCT file containing the preprocessed and normalized expression data were input into the BubbleGum module alongside a CLS class file, defining cell-type–specific phenotype labels associating each sample in the expression data. A GMT file containing the predefined gene signatures for CD1c⁺ mDCs, CD141⁺ mDCs, LCs, CD14⁺ monocyte-derived macrophages (also referred as CD14⁺ DCs), macrophages, CD14⁺ monocytes, and CD16⁺ monocytes, to be tested for enrichment and a CHIP file, corresponding to the CHIP platform were also included. The gene signature for each myeloid subpopulation is listed in supplemental Table 4. A weighted enrichment statistic (described in Subramanian et al⁴² ) was used to calculate the degree of the enrichment of each gene signature. The data were displayed as an array of circles, or a BubbleMap in which the color of the circle denotes in which of the classes the enrichment occurs and the area of circle denotes the normalized enrichment score. The intensity of the colors shows the limit of significance of the enrichment or false discovery rate. Samples used in this analysis are listed in supplemental Table 3a.

Total RNA was purified from sorted subpopulations from peripheral blood and lesion specimens according to the Arcturus PicoPure RNA Isolation kit protocol (Applied Biosystems). RNA quality was verified using the Pico Chip at the Baylor University College of Medicine Microarray Facility. cDNA amplification was performed using the Ovation Pico WTA V2 system according to the manufacturer’s protocol (Nugen, San Carlos, CA). Fragmented and biotinylated cDNA was hybridized to GeneChip Human Transcriptome Array 2.0 according to the manufacturer’s procedures (Affymetrix, Waltham, MA). Raw data from all samples were normalized using the SST-RMA algorithm implemented in the Affymetrix Expression Console.

A 1-way analysis of variance was used to compare LCH CD1a⁺CD207⁺ DCs to healthy control blood CD1c⁺ mDCs. DEGs were identified using the Transcriptome Analysis Console 4.0 with false discovery rate controlled at 0.05 using the BH method and a fold change >2. All samples that were used in the analysis are listed in supplemental Table 3a. Among the DEGs, 2190 of them were differentially expressed between the 2 populations. A heatmap was generated showing the 50 genes with highest significant relative expression in LCH CD1a⁺CD207⁺ DCs (supplemental Figure 2).

A paired 1-way repeated measure analysis of variance was used to compare paired samples of CD1a⁺CD207^high, CD1a⁺CD207^low, CD1a⁺CD207⁻ cells. Significance was determined using the false discovery rate controlled at 0.05 using the BH method. The samples used in analysis are listed in supplemental Table 3a.

A principal component analysis (PCA) was performed using the samples listed in supplemental Table 3a with R statistical software. The ggbiplot package was used to draw a normal data ellipse for each group.

Total RNA was isolated, and cDNA was generated as described in “Affymetrix gene chip processing and analysis.” Quantitative real-time polymerase chain reactions (PCRs) were performed with TaqMan gene-expression assays (Applied Biosystems) and the PrimePCR probe assay (Bio-Rad Laboratories, Hercules, CA) on a CFX96 real-time system (Bio-Rad Laboratories) according to the manufacturer’s protocol. TaqMan probe sets used were CD1a (HS00381754_g1), CD1c (HS00233509_m1), CD207 (Hs00210451_m1), CD141 (HS00264920_s1), CLEC9A (HS001651638_m1), HLA-DQA2 (Hs00607448_gH), HLA-DQB2 (Hs00794552_m1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH; HS02758991_g1). Prime PCR probe sets were GAPDH (qHsaCEP0041396) and HLA-DQA2 (qHsaCIP0040206). All data were normalized to GAPDH expression. Statistical analyses were performed with Prism 7.0 (GraphPad Software, La Jolla, CA). All samples used in this analysis are listed in supplemental Table 3b.

To evaluate the relatedness of transcriptome of LCH lesion pathogenic DCs to potential precursor populations, we first compared full gene-expression profiles of the lesion CD1a⁺CD207⁺ DCs to gene signatures generated from human myeloid cells (supplemental Table 3a). All of these myeloid cells including CD14⁺ classical monocytes, CD16⁺ nonclassical monocytes, CD14⁺ DCs (also referred as CD14⁺ monocyte-derived macrophages), resident macrophages, epidermal LCs, CD1c⁺ mDCs, and CD141⁺ mDCs have been previously used in human and mouse DC ontogeny studies.^24,39  According to the CMAP analysis, the CD1c⁺ mDC gene signature had the highest enrichment in the gene-expression profile of LCH lesion CD1a⁺CD207⁺ DCs, followed by CD141⁺ mDCs and LCs, then CD14⁺ DCs and CD14⁺ monocytes, with least enrichment for the gene signatures of macrophages and CD16⁺ monocytes (Figure 2A). Analyzing these data with a complementary approach, we also compared gene-expression profiles of LCH lesion CD1a⁺CD207⁺ DCs to gene-expression profiles from other cell types, including human peripheral blood and skin DCs, monocytes, and macrophages obtained from microarray data sets using the BubbleGUM platform (supplemental Table 3a).⁴¹  Consistent with CMAP analysis, the CD1c⁺ mDC gene signature had the highest degree of enrichment in LCH lesion CD1a⁺CD207⁺ DC gene set compared with gene-expression profiles from other DCs and macrophage and monocyte subsets (Figure 2B). Collectively, these data revealed that the transcriptome of lesion CD1a⁺CD207⁺ DCs in this series was most closely related to that of CD1c⁺ mDCs.

To further investigate whether LCH lesion CD1a⁺CD207⁺ cells are biologically related to less mature CD1c⁺ mDCs (blood CD1c⁺ mDCs) or more differentiated CD1c⁺ mDCs (tissue CD1c⁺ mDCs), we compared transcriptomes from LCH CD1a⁺CD207⁺ DCs to blood CD1c⁺ mDCs and skin CD1c⁺ mDCs along with CD14⁺ monocytes and CD14⁺ DCs through unsupervised hierarchical clustering. Interestingly, LCH lesion CD1a⁺CD207⁺ DCs clustered more closely to the blood CD1c⁺ mDCs than to the skin CD1c⁺ mDCs (Figure 2C).

Although LCH lesion pathogenic cells are generally described as CD1a⁺CD207⁺, flow cytometry analysis of LCH lesions demonstrate some heterogeneity of the CD207 expression level variable relative abundance of CD1a⁺ subpopulations (Figure 3A; supplemental Table 5). LCH lesion CD1a⁺ subpopulations including CD1a⁺CD207^high cells, CD1a⁺CD207^low cells, and CD1a⁺CD207⁻ cells were analyzed along with CD3⁺ lymphocytes from LCH lesions with respect to global gene expression (Figure 3A-B; supplemental Figure 4). PCA of gene-expression profiles from all subpopulations analyzed showed that CD1a⁺CD207⁻, CD1a⁺CD207^high, and CD1a⁺CD207^low subpopulations were tightly clustered relative to the CD3⁺ population (Figure 3C). Comparison of global gene-expression profiles of these 3 LCH lesion CD1a⁺ subpopulations did not identify any significantly differentially expressed genes after controlling for multiple comparisons. However, specific analysis of CD207 expression identified significantly increased expression from CD1a⁺CD207⁻ through CD1a⁺CD207^high cells with consistent levels of genes expressing CD1a, CD1c, and TNFα (supplemental Figure 1).

Somatic BRAFV600E mutation alleles have been previously demonstrated to be enriched in LCH lesion CD1a⁺CD207⁺ DCs.^3,43  Assuming BRAFV600E to be a driver mutation and cells to be clonal, candidate precursor and pathogenic cells would be expected to share this mutation. We therefore analyzed BRAFV600E allele burden in LCH lesion subtypes and found that, in addition to CD1a⁺CD207^high, CD1a⁺CD207^low, and CD1a⁺CD207⁻ cells, the BRAFV600E allele could also be identified in CD1c⁺ mDCs, but was not detected in other LCH lesion subpopulations, including CD11c⁺CD1c⁻ cells that contain myeloid precursors, monocytes, and CD141⁺ mDCs (Figure 3D).

These gene-expression and mutation-mapping studies identified CD1c⁺ mDCs as transcriptionally and mutationally similar to LCH lesion CD1a⁺CD207⁺ cells. To identify potential biomarkers specific to differentiating DCs, we compared the gene-expression profiles between healthy donor blood CD1c⁺ mDCs and LCH lesion CD1a⁺CD207⁺ DCs. This analysis identified differential expression of known LCH-related genes including CD1a, CD1c, CD207, and TNFα, which were much more highly expressed in LCH lesion CD1a⁺CD207⁺ DCs relative to blood CD1c⁺ mDCs from healthy donors (Figure 4A; supplemental Figure 2). An unexpected finding from this experiment was significantly increased expression of HLA-DQA2 and HLA-DQB2 in LCH lesion CD1a⁺CD207⁺ cells (Figure 4A). These genes were not previously associated with LCH, but have been previously reported to be specific to human epidermal LCs.⁴⁴ HLA-DQA2 and HLA-DQB2 expression was further assessed across LCH lesion subpopulations (Figure 4B; supplemental Figure 5), with purity of all sorted LCH subpopulations confirmed by real-time PCR (supplemental Figure 3A-B). Where HLA-DQA2 was expressed in many subpopulations in our study, HLA-DQB2 expression was specific to the LCH lesion subpopulations that harbored BRAFV600E mutations (Figure 4B; supplemental Figure 5). Further flow cytometry analysis of LCH lesions identified surface expression of HLA-DQB2 on CD1c⁺ mDCs, CD1a⁺CD207⁻ cells, CD1a⁺CD207^low cells, and CD1a⁺CD207^high cells, but not in other lesion myeloid and lymphoid subpopulations (Figure 4C; supplemental Figure 6). Gene-expression analysis found HLA-DQB2 in CD1a⁺CD207⁺ cells from low-risk and high-risk lesions, with no statistical difference in expression between risk groups (supplemental Figure 7).

The BRAFV600E mutation may be detected at low levels (generally <0.1%) in PBMCs) of patients with active BRAFV600E⁺ lesions characterized as high risk (liver, spleen, and/or bone marrow involvement).³  Using real-time PCR to characterize HLA-DQB2 expression in PBMCs of patients with active BRAFV600E⁺ LCH lesions, HLA-DQB2 expression was identified in PBMCs of high-risk LCH, but not in PBMCs from both low-risk multisystem and low-risk single system patients, correlating with identification of detectable BRAFV600E allele (Figure 5A-B). PBMCs were then analyzed by flow cytometry for HLA-DQB2 surface expression. Interestingly, HLA-DQB2 expression was only identified in high-risk LCH patients (but not in low-risk patients or healthy controls) (Figure 5C; supplemental Figure 8), and it was localized to CD1c⁺ mDCs (3.3% to 23.7% of total peripheral blood CD1c⁺ mDCs) (Figure 5C). Although the BRAFV600E allele was identified across multiple peripheral blood cell subpopulations, it was highly enriched in CD1c⁺HLA-DQB2⁺ mDCs (Figure 5D-E).^3,17 

According to the Misguided Myeloid DC Model, LCH DCs arise from hematopoietic stem cells or myeloid precursors with pathologically activated MAPK signaling, and the extent of disease depends on the stage of differentiation of the myeloid precursor in which the activating mutation arises.^1,12  Analyses of peripheral blood from children and adults with LCH demonstrate that the BRAFV600E mutation may be carried by a range of myeloid and lymphoid cells.^3,17  We hypothesize that LCH lesion CD1a⁺CD207⁺ DCs share a common pathway of terminal differentiation, though the pathway from precursor to CD1a⁺CD207⁺ DC remains undefined. In this study, we aimed to identify potential origin(s) of LCH lesion CD1a⁺CD207⁺ DCs.

Given the ability to differentiate into langerin⁺ (CD207⁺) cells in vitro, both CD1c⁺ mDCs and CD14⁺ monocytes are potential candidates to give rise to LCH CD1a⁺CD207⁺ DCs.^{14,15,17,31,45}  To determine the differentiation pathway(s) of LCH CD1a⁺CD207⁺ DCs, we first compared the gene-expression profiles of CD1a⁺CD207⁺ DCs to gene signatures from human myeloid cell lineages. LCH CD1a⁺CD207⁺ cells were most highly enriched with CD1c⁺ mDC signature, with relatively lower enrichment with the CD14⁺monocyte signature using both CMAP and BubbleGum analysis platforms (Figure 2A-B). Furthermore, unsupervised clustering of transcriptomic data demonstrated that LCH CD1a⁺CD207⁺ DCs were more related to blood CD1c⁺ mDCs than to tissue CD1c⁺ mDCs, blood CD14⁺ monocytes, or tissue CD14⁺ DCs (Figure 2C). Analysis of BRAFV600E allele burden in LCH lesion cells identified the mutation in CD1a⁺CD207⁻ cells, CD1a⁺CD207^low cells, and CD1a⁺CD207^high cells as well as in CD1c⁺ mDCs, but not in lesion CD14⁺ DCs (Figure 3D). These transcriptome analysis and mutation-tracing experiments are consistent with blood CD1c⁺ mDC as a potential precursor to the LCH lesion CD1a⁺CD207⁺ DC.

It is therefore plausible that blood CD1c⁺ mDC could migrate to the lesion, become tissue CD1c⁺ mDC, and then acquire CD1a and CD207 expression (Figure 6). To test changes in gene expression along such a pathway of differentiation, we compared gene-expression profiles of LCH lesion CD1a⁺CD207⁺ cells and healthy control blood CD1c⁺ mDCs. In addition to some expected differences in gene expression (CD207, CD1a, TNF, MMP9), an unexpected finding was remarkably increased expression of HLA-DQA2 and HLA-DQB2 in LCH lesion CD1a⁺CD207⁺ DCs (Figure 4A; supplemental Figure 2). These HLA class II (HLA-II) genes were previously reported with physiologic expression specific to epidermal LCs.⁴⁴  In our study, we found HLA-DQA2 expression in multiple myeloid populations in LCH lesions (supplemental Figure 5). However, HLA-DQB2 expression was specific to CD1c⁺ mDCs, CD1a⁺CD207⁻ cells, and CD1a⁺CD207⁺ cells (Figure 4B-C), cells with the common feature of also harboring the BRAFV600E allele (Figure 3D).

Given the expression pattern of HLA-DQB2 on potential LCH lesion CD1a⁺CD207⁺ DC precursors, we investigated peripheral blood populations for HLA-DQB2 expression. As in the study by Lenormand et al,⁴⁴ HLA-DQB2 expression was absent from any peripheral blood populations from healthy donors. However, we did identify HLA-DQB2 expression in CD1c⁺ mDCs in peripheral blood of patients with high-risk LCH (Figure 5C). Despite the presence of BRAFV600E among multiple PBMC populations in patients with high-risk LCH (and BRAFV600E⁺ lesions), HLA-DQB2 localized to CD1c⁺ mDCs (Figure 5D-C). Furthermore, BRAFV600E alleles were enriched in CD1c⁺HLA-DQB2⁺ mDCs (Figure 5E). HLA-DQB2 mRNA and cell-surface expression may therefore represent novel biomarkers for circulating LCH precursors and pathologic LCH lesion DCs. HLA-DQB2 was demonstrated to stimulate T cells in vitro.⁴⁴  It is plausible that, in addition to representing a clinically relevant biomarker, it may play a role in mediating inflammation in LCH lesions.

Although this study supports a model of CD1c⁺ mDCs as precursors for LCH CD1a⁺CD207⁺ DCs, it is possible that >1 cell of origin may exist. It has been shown previously that CD14⁺ monocytes have the capacity to become CD1a⁺CD207⁺ cells in response to notch ligands δ-like protein 1 (DLL1), DLL4, GMCSF, and TGFβ in vitro.^14,15,17,31  Furthermore, a recent study showed that CD14⁺ monocytes could acquire gene signature with similarities to the LCH CD1a⁺CD207⁺ cell signature in the presence of Notch ligands JAG2 in vitro.⁴⁵  It remains possible that CD14⁺monocytes could differentiate rapidly into LCH CD1a⁺CD207⁺ DCs once entering the lesion, or that a small population of CD14⁺ DCs that present in LCH lesions may not be captured by the experiments in this study. Notably, unlike a previous report,⁴⁶  we did not detect any CD207⁺ cells in peripheral blood of either LCH patients (high or low risk) or healthy controls (supplemental Figure 9).

The findings from this study support an updated Misguided Myeloid DC Model of LCH pathogenesis in which blood CD1c⁺ mDCs with hyperactive ERK may have the capacity to migrate to the LCH lesion sites and differentiate into LCH CD1a⁺CD207⁺ DCs (Figure 6). If differentiation from CD1c⁺ mDCs to CD1a⁺CD207⁺ DCs is critical to LCH pathogenesis, then blocking chemokines or integrins that allow CD1c⁺ mDCs to enter tissues may represent a novel therapeutic strategy. Additionally, these findings suggest potential roles of MAPK activation in physiologic CD1c⁺ mDC differentiation.

The authors thank Linna Zhang and Elizabeth Pacheco (College of Medicine, Baylor University; Texas Children’s Cancer Center) for data management support, and Tatiana Goltsova and Amos S. Gaikwad from the Flow Cytometry Core Laboratory (Texas Children’s Hospital) for assistance with flow cytometry.

The TXCH Histiocytosis Program is supported by a research grant from the HistioCure Foundation. Additional support was received from National Institutes of Health, National Cancer Institute grants R01 CA154489 (M.M. and C.E.A.), CA154947 (M.M. and C.E.A.), and P50CA126752 (Specialized Program of Research Excellence in Lymphoma) (C.E.A.); the St. Baldrick’s Foundation Innovation grant (C.E.A.); the St. Baldrick’s Foundation Consortium grant supporting the North American Consortium for Histiocytosis (NACHO) (C.E.A.); the Leukemia & Lymphoma Society Translational Research Program (C.E.A.); Singapore Immunology Network core funding (F.G.); a Histiocytosis Association grant (F.G.); and Cancer Research UK grant C30484/A21025 (M.C.). F.G. is an EMBO Young Investigator Programme awardee. K.P.H.L. was supported by the Howard Hughes Medical Institute to the Baylor University College of Medicine Med into Grad Initiative; National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases T32 training grants in Hematology (5T32DK60445-13 and 5T32DK60445-14); and a National Institutes of Health, National Institute of General Medical Sciences T32 training grant in Translational Biology and Molecular Medicine (2T32GM88129-6). T.M.B. was supported by National Cancer Institute Ruth L. Kirschstein National Research Service Award, Individual Predoctoral Fellowship to Promote Diversity in Health-Related Research grant 5F31CA228360-02. O.S.E. was supported by National Institutes of Health, National Cancer Institute K12 grant 2K12CA090433-16.

Contribution: K.P.H.L. and P.M. designed research, performed experiments, analyzed data, and wrote, reviewed, and edited the manuscript; M.P., K.D., H.L., and N.M. designed research, analyzed data, and reviewed and edited the manuscript; H.A., D.Z., T.M.B., O.S.E., R.C., A.S., and B.S. provided technical advice, performed experiments, reviewed clinical data, and reviewed and edited the manuscript; E.N., M.M., K.L.M., and T.-K.M. contributed to the study concept and experimental design and reviewed and edited the manuscript; and F.G., M.C., and C.E.A. conceived the study, supervised the project, and wrote, reviewed, and edited the manuscript.

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

Correspondence: Carl E. Allen, Texas Children’s Hospital, Feigin Center, Suite 730.06, 1102 Bates St, Houston, TX 77030; e-mail: ceallen@txch.org; Matthew Collin, Human Dendritic Cell Laboratory, Institute of Cellular Medicine, Newcastle University, Framlington Pl, Newcastle upon Tyne NE2 4HH, United Kingdom; e-mail: matthew.collin@ncl.ac.uk; or Florent Ginhoux, Singapore Immunology Network (SIgN), Agency for Science, Technology and Research (A*STAR), Singapore 138648; e-mail: florent_ginhoux@immunol.a-star.edu.sg.