BRAFV600E-induced senescence drives Langerhans Cell Histiocytosis pathophysiology

Langerhans cell histiocytosis (LCH) is a potentially fatal condition characterized by granulomatous lesions with characteristic clonal mononuclear phagocytes (MNP) harboring activating somatic mutations in MAPK pathway genes, most notably BRAFV600E. We recently discovered that the BRAFV600E mutation can also affect multipotent hematopoietic progenitor cells (HPC) in multisystem LCH disease. How BRAFV600E mutation in HPC leads to LCH is not known. Here we show that enforced expression of the BRAFV600E mutation in early mouse and human multipotent HPC induced a senescence program that led to HPC growth arrest, apoptosis resistance and senescence-associated secretory phenotype (SASP). SASP, in turn, promoted HPC skewing towards the MNP lineage leading to the accumulation of senescent MNP in tissue and the formation of LCH lesions. Accordingly, elimination of senescent cells using INK-ATTAC transgenic mice as well as pharmacologic blockade of SASP improved LCH disease in mice. These results identify senescent cells as a novel target for the treatment of LCH. These new results revealed a new conundrum in LCH pathophysiology: how does the somatic BRAFV600E mutation in multipotent progenitor cells lead to LCH lesions? In accordance with these studies, we found that mTOR inhibition reduced HPC skewing toward the MNP lineage and improved disease phenotype in vivo. Altogether, our results showing that SASP contributes to enhanced myeloid cell differentiation in the BM, could explain the myeloid-biased hematopoiesis observed in aged patients (reviewed in (Montecino-Rodriguez et al., 2013)) and if conrmed could suggest that mTOR inhibition in HPC could help restore immune balance and homeostasis in older patients. sacriced (J) The percentage of senescent-p16INK4a overexpressing cells was assessed by ow cytometry in the bone marrow from untreated and treated BRAFV600EScl+ ATTAC+ mice (n= 5). Graph shows the percentage of GFP+ cells among CD45+ cells. (K) Liver weight in each group (n=8). percentage of tissue inltration in liver and lung are with representative image of hematoxylin and eosin staining of tissues isolated from untreated and treated BRAFV600EScl+ ATTAC+ mice Data representative of 3 experiments and are represented as mean ± s.e.m, statistical signicance analyzed by an unpaired and paired (for 5A and B) two-sided t-test is indicated by *p < 0.05; **p < 0.01; 0.001.


Introduction
Histiocytoses are a group of conditions characterized by the accumulation of histiocytes in various tissues. Histiocytes, now more commonly known as mononuclear phagocytes (MNP), consist of heterogeneous cells of myeloid origin that include dendritic cells and macrophages. Speci cally, Langerhans Cell Histiocytosis (LCH), the most common histiocytic disorder, results from the accumulation of CD207 + CD1a + cells that resemble epidermal MNP, called Langerhans cells (Merad et al., 2008). LCH occurs mostly in children with an incidence of 8.9/10 6 children per year, similar to pediatric Hodgkin's lymphoma (Stålemark et al., 2008). LCH can arise in a variety of settings from single bone lesions to life-threatening disseminated disease that affects the spleen, liver, and bone marrow (BM).
Currently, front-line chemotherapy fails in more than 50% of patients with multisystem disease, highlighting the need for novel therapies for this group of patients (Allen et al., 2018). However, uncertainty with regard to disease pathophysiology as well as the paucity of a reliable LCH preclinical model have challenged the development of optimal clinical strategies for patients with aggressive LCH.
Activating somatic mutations in MAPK pathway genes, most notably BRAFV600E, have been identi ed in 85% cases of LCH, speci cally in the pathogenic CD207 + cells within LCH lesions (Badalian-Very et al., 2010;Chakraborty et al., 2014 and2016;Lim et al., 2020). CD207 + LCH cells express many features of MNP, and in particular dendritic cells, such as expression of MHC II, CD11c, and CD207. Normal dendritic cells have a short half-life in peripheral tissues, which they leave upon expression of the chemokine receptor CCR7, following a CCL21 gradient expressed by lymphatic vessels and lymph node stromal cells that guide them to the T zone of tissue-draining lymph nodes (Merad et al., 2013). We recently showed that the BRAFV600E mutation prolonged LCH cell survival through the induction of the anti-apoptotic molecule BCL-xL and abolished CCR7 expression, thus trapping LCH cells in tissues and leading to LCH lesions (Hogstad et al., 2018). Importantly, we and others (Berres et al., 2015;Durham et al., 2017;Milne et al., 2017) recently discovered that the BRAFV600E mutation can also be identi ed in CD34 + BM hematopoietic progenitor cells (HPC) in LCH patients with systemic disease, which has led to the reclassi cation of multi-focal LCH as a myeloid neoplasia (Allen et al., 2018;Berres et al., 2015;Lim et al., 2020). These new results revealed a new conundrum in LCH pathophysiology: how does the somatic BRAFV600E mutation in multipotent progenitor cells lead to LCH lesions? HPC differentiate into lymphoid and myeloid lineages in response to cell intrinsic and extrinsic cues in the BM. Once committed to the myeloid lineage, HPC can either differentiate into the granulocytic or the MNP lineage, the latter of which gives rise to dendritic cells, monocytes, monocyte-derived dendritic cells, and monocyte-derived macrophages (Merad et al., 2013).
Here, we explored the mechanisms by which, the BRAFV600E mutation could potentially drive a multipotent hematopoietic progenitor to give rise to LCH lesions. Strikingly, we found that enforced expression of the BRAFV600E mutation in early mouse and human multipotent HPC induced a senescence program that led to HPC growth arrest, apoptosis resistance, and senescence-associated secretory phenotype (SASP) induction. This SASP, in turn, promoted multipotent HPC skewing towards the MNP lineage away from other hematopoietic lineages. BRAFV600E-induced senescence persisted in differentiated MNP that accumulated in tissues, where SASP induction and sustained MNP survival contributed to the formation of LCH lesions. Accordingly, we found that genetic elimination of senescent cells using INK-ATTAC transgenic mice, as well as pharmacologic blockade of SASP, improved LCH disease burden in mice. These results transform our understanding of LCH pathophysiology, and identify senescent cells as a novel target for the treatment of LCH.

Results
Expression of the BRAFV600E mutation in mouse multipotent hematopoietic progenitor cells (HPC) is su cient to drive to LCH lesion formation We previously identi ed that the BRAFV600E mutation is found not only in LCH lesions, but also in CD34 + BM HPC and blood circulating CD11c + and CD14 + monocytes in patients with multi-system LCH (Berres et al., 2015). These results prompted us to explore how expression of the BRAFV600E mutation in multipotent HPC could lead to LCH lesions. To examine whether BRAFV600E + HPC can drive LCH lesions, we enforced expression of the BRAFV600E mutation in cells expressing Scl (also known as Tal1), a molecule expressed speci cally in long-term and short-term hematopoietic stem cells (HSC) in mice (Göthert et al., 2005). We genetically engineered somatic mosaicism for a BRAFV600E allele linked to a yellow uorescent protein (YFP) in HSC using tamoxifen (Tam)-inducible targeting in Scl cre-ER mice, which we named BRAFV600E Scl+ , and used BRAFwt Scl+ as control littermates ( Figure 1A). Thus, Rosa YFP+ cells carried the BRAFV600E mutation in BRAFV600E Scl+ mice, whereas Rosa YFP+ cells in BRAFwt Scl+ mice did not carry the mutation but underwent Rosa26-locus driven cre recombination. Tamoxifen (Tam) pulsedadministration led to YFP expression in 30% of HSC providing the opportunity to compare BRAFV600E + (YFP + ) Scl+ cells to BRAFwt(YFP -) Scl+ cells within the same mouse ( Fig 1B). Strikingly, BRAFV600E Scl+ mice developed substantial organomegaly, granulomatous bone lesions and BM aplasia within 4 weeks of Tam administration, while BRAFwt Scl+ control littermates did not (Fig 1C, Supplementary Fig 1A, B and C). Histological analysis of hematoxylin & eosin (H&E) stained formalin-xed para n-embedded (FFPE) tissue sections revealed an accumulation of histiocytes within granulomatous lesions, multinucleated giant cells and an in ammatory in ltrate in the liver, lungs and dermis of BRAFV600E Scl+ mice (Fig 1D,  Sup Fig 1D). Immunohistochemistry analysis revealed that many histiocytes express CD207 (Fig 1D).
Lesions were diagnosed as LCH by a clinical pathology expert blinded to the experiment. Flow cytometry analysis con rmed the signi cantly increased immune cell in ltrate and the expansion of the MNP compartment within the immune cell in ltrate (Fig 1F, Sup Fig 1E). Importantly, in the skin, MNP mostly accumulated in the dermis and were absent from the epidermis (Fig 1F), mimicking the dermotropism of the human disease. Flow cytometry analysis also revealed the preferential expansion of MNP within the BRAFV600E YFP+ immune cell compartment in tissues known as targets of LCH lesions (Fig 1E and Sup Fig 1F).
Humanized mice reconstituted with human CD34 + HPC expressing the BRAFV600E mutation develop LCH-like disease To determine whether expression of the BRAFV600E mutation in human HPC could also lead to LCH disease, we generated a lentiviral vector expressing the green uorescent reporter (GFP) together with the BRAFV600E mutation (BRAFV600E hu ) or the NGFR truncated gene (NGFR hu ) under the EF-1 alpha 1 (EF1A1) promoter (Sup Fig 1G). HPC transduction e ciency, around 40%, led to sustained ERK phosphorylation in BRAFV600E hu but not in NGFR hu HPC (Sup Fig 1H and I). We reconstituted sublethally irradiated 8-week old NOD-SCID gamma (NSG) mice with human CD34 + cord blood HPC transduced with this lentiviral vector ( Fig 1G). NSG mice deteriorated within two months following reconstitution with BRAFV600E hu CD34 + HPC, but not NGFR hu CD34 + HPC, and developed an expansion of blood circulating CD11c + CD14 + MNP cells (Sup Fig 1J) accompanied with an accumulation of CD1a + CD207 + cells in the liver, lung and spleen pathognomonic of LCH disease (Fig 1H). These results, together with our data showing that BRAFV600E Scl+ mice develop LCH-like lesions, establish that expression of the BRAFV600E mutation in mouse and human early HPC is su cient to drive LCH lesion formation.
BRAFV600E HPC differentiate predominantly toward the mononuclear phagocyte (MNP) lineage To explore how the BRAFV600E mutation in early HPC, which should retain their multipotent lineage capabilities, drives the speci c accumulation of MNP in the periphery, we measured the differentiation potential of BRAFV600E+ YFP+ HPC in BRAFV600E Scl+ mice. Strikingly, we found that BRAFV600E + HSC and multipotent progenitors (MPP) were reduced, while BRAFV600E + granulocyte-macrophage progenitor (GMP) were signi cantly expanded in the BM compared to BRAFwt progenitors (Fig 2A) suggesting that the BRAFV600E + HPC failed to expand and were instead biased towards the myeloid lineage. The expansion of BRAFV600E + GMP was associated with a remarkable expansion of BM BRAFV600E + monocytes, macrophages and dendritic cells, whereas BM BRAFV600E + neutrophils did not expand ( Fig  2B). To directly examine whether the BRAFV600E mutation skewed HPC commitment to the MNP lineage, we cultured puri ed HPC from BRAFV600E Scl+ or BRAFwt Scl+ mice in methylcellulose and measured HPC progeny using a colony forming unit (CFU) assay that re ects the differentiation into committed progenitors. Consistent with the lineage bias observed in vivo, we found that BRAFV600E HPC preferentially differentiated into CFU consisting of macrophage progenitors (CFU-M), while the numbers of granulocytic progenitors (CFU-G) and erythroid progenitors also called burst forming unit-erythroid (BFU-E) were reduced ( Fig 2C).
We also examined whether the BRAFV600E mutation affected human HPC lineage commitment.
Consistent with the results obtained from murine BRAFV600E Scl+ HPC, we found that human BRAFV600E hu HPC preferentially differentiated into the MNP lineage ( Fig 2D). Remarkably, using the same CFU assay applied in mice, we found that human BRAFV600E + HPC differentiated preferentially into CFU-GM and CFU-M colonies but not into CFU-G colonies, whereas NGFR + HPC retained a pluripotent differentiation potential (Fig 2E and Sup Fig 2). Gene expression pro ling of human BRAFV600E + GFP + and control NGFR + GFP + HPC seven days after culture in stem-cell media revealed enrichment for an MNP signature in BRAFV600E + HPC that included the expression of the classical dendritic cells genes (BATF3, IRF4, CLEC10A) as well as the macrophage genes (CSF1R, CD68, SPP1), while HPC differentiation into granulocytes was strongly reduced compared to control NGFR + HPC ( Fig 2F).
To further con rm the MNP skewing of BRAFV600E + HPC in LCH patients, we performed RNA sequencing of puri ed CD34 + HPC isolated from the BM of LCH patients (Supplementary Table 1). Similar to the signature of BRAFV600E transduced HPC, CD34 + HPC isolated from LCH patients showed reduced expression of genes associated with granulopoiesis (ELANE, MPO, PRTN3, CSF3R), and an increased expression of genes involved in dendritic cells/macrophages commitment (BATF3, IRF4, CSF1R, CLEC10A) compared to CD34 + HPC isolated from an age-matched individual ( Fig 2G). Taken together, these data establish that the BRAFV600E mutation directs mouse and human HPC to differentiate into the MNP lineage away from the lymphoid and the granulocytic lineage.
The skewing of BRAFV600E + HPC into the MNP lineage is driven by both cell intrinsic and extrinsic cues To examine whether the enhanced MNP differentiation of BRAFV600E + HPC was cell intrinsic or a result of external cues, we measured the expansion of BRAFV600E + and BRAFwt HPC within the same animal.
Strikingly, while we failed to observe an expansion of GMP in BRAFwt Scl+ mice, there was an expansion of BRAFV600E + and BRAFwt GMP within the same BRAFV600E Scl+ mice (Fig 2H), associated with an increase of BM BRAFV600E + and BRAFwt MNP but not neutrophils (Fig 2I). The expansion of BM BRAFwt MNP in BRAFV600E Scl+ mice but not in BRAFwt Scl+ mice suggest that MNP skewing may not only be driven by cell intrinsic cues.
To examine whether similar cell extrinsic cues can also drive human HPC differentiation into MNP, we took advantage of the fact that HPC transduction e ciency with BRAFV600E or NGFR lentiviral vectors did not exceed 40% (Sup Fig 1G), to compare within the same culture the differentiation of BRAFV600E + GFP + and BRAFV600E -GFPas well as NGFR + GFP + and NGFR -GFP -HPC. Similar to our mouse results, BRAFV600E -GFPcells co-cultured with BRAFV600E + GFP + cells had an enhanced MNP differentiation potential, although at a reduced scale compared to BRAFV600E + GFP + HPC counterparts ( Fig 2J). This contrasts with our earlier results showing that puri ed human BRAFV600E -GFP -HPC cultured alone in CFU assays did not have an MNP differentiation advantage ( Fig 2E, Sup Fig 2). This suggests that secreted molecules produced by BRAFV600E + GFP + HPC contributed to driving MNP differentiation of healthy HPC. To directly measure whether the BRAFV600E + HPC secretome skewed the differentiation of BRAFwt HPC toward the MNP lineage, we cultured human CD34 + HPC in stem-cell media in the presence or absence of supernatant isolated from human BRAFV600E + HPC or NGFR + HPC. Consistent with our hypothesis, we observed that healthy HPC cultured in the presence of supernatant obtained from human BRAFV600E + HPC cultures were more prone to differentiate into MNP compared to healthy HPC cultured in the presence of NGFR + HPC supernatant, and that the enhanced MNP differentiation potential was dependent on the dose of BRAFV600E + HPC supernatant added ( Fig 2K).

The BRAFV600E mutation drives mouse and human hematopoietic progenitors into oncogene-induced senescence
The realization that the BRAFV600E mutation limited the expansion of HSC (Fig 2A) prompted us to further assess the proliferation potential of BRAFV600E mutant HPC using a competitive reconstitution assay in which, puri ed BRAFV600E + HPC or BRAFwt HPC (isolated from CD45.2 + animals) were injected together with host HPC (CD45.1 + ) at a 2:1 ratio into sub-lethally irradiated CD45.1 + recipients. Strikingly, we found that at 4 and 8 weeks after transplantation, the number of engrafted BRAFV600E + HPC was strongly reduced compared to BRAFwt HPC ( Fig 3A, Sup Fig 3A), thus revealing the proliferative disadvantage of BRAFV600E HPC compared to BRAFwt HPC.
The BRAFV600E mutation has been previously described to drive cellular senescence in human naevi lesions (Michaloglou et al., 2005). Cellular senescence classically occurs secondary to DNA damage or because of oncogene-activation and results in cell cycle arrest driven by cell cycle regulators such as p16 INK4a , prolonged survival driven by the anti-apoptotic proteins Bcl-xL and Bcl-2. Senescent cells also form heterochromatin foci, and express senescence-associated b galactosidase (SAbGal), as well as a senescence-associated secretory phenotype (SASP) (reviewed in (Hernandez-Segura et al., 2018;Muñoz-Espín and Serrano, 2014)), which leads to the production of a wide range of in ammatory molecules such as IL-1, IL-6, IL-8 and several matrix metalloproteinases. Furthermore, senescence results in morphological alterations with senescent cells being enlarged with an irregular shape.
Strikingly, we found that BRAFV600E Scl+ HPC were unable to cycle, as shown by the low BrdU incorporation potential which, likely accounts for their poor ability to engraft compared to BRAFwt Scl+ mice in a competitive reconstitution assay ( Fig 3B). HPC isolated from BRAFV600E Scl+ mice were also enlarged compared to HPC from BRAFwt Scl+ mice, a common feature of senescent cells (Sup Fig 3B); expressed high levels of CDKN2A transcripts which codes for cell cycle regulator p16 INK4a (Fig 3C), were positive for SAbGal ( Fig 3D); and produced SASP-associated proteins (Fig 3E), all features consistent with a senescent state. LCH cells isolated from peripheral tissues of BRAFV600E Scl+ mice were also in a senescent state, as shown by reduced Ki-67 expression (Fig 3F), expression of CDKN2A (Fig 3G), SAbGal activity ( Fig 3H) and the production of SASP-associated proteins ( Fig 3I). These results suggest that the BRAFV600E mutation induced a senescence program in HPC which was maintained in differentiated LCH cells.
Similar to mice, the BRAFV600E mutation in human HPC led to an initial phase of proliferation (Sup Fig  3C) followed by a marked cell-cycle arrest in vitro ( Fig 3J, Sup Fig 3D). Human BRAFV600E transduced HPC expressed canonical markers of cellular senescence including enlarged cellular size (Sup Fig 3E), positivity for SAbGal staining (Fig 3L), senescence-associated heterochromatin foci ( Fig 3M) and increased production of SASP cytokines (Fig 3N, Sup Fig 3F). Gene expression pro ling of transduced human BRAFV600E + GFP + HPC con rmed the senescence signature (Kuilman et al., 2008) ( Fig 3K) and LCH lesions that formed in humanized mice reconstituted with human BRAFV600E CD34 + cells also expressed SAbGal activity and high p16 INK4a levels (Sup Fig 3G and H).
Importantly, CD34 + BM cells isolated from LCH patients also expressed a senescence signature including increased expression of CDKN2A, CDKN2C, CDKN2D, CD9, MDM2, and matrix-metalloproteinases, which was absent from CD34 + BM cells isolated from age-matched healthy individuals (Fig 4A and Sup Fig 4A).
Puri ed CD207 + cells from human LCH lesions expressed high CDKN2A, CDKN2B, CDKN2C transcripts and high SASP transcript levels (MMP1, MMP3, MMP9, MMP13) ( Fig 4B). LCH cells from LCH lesions expressed low level or no Ki-67 ( Fig 4C) while they expressed high levels of p16 INK4a protein expression and high SAbGal activity (Fig 4D,E and Sup Fig 4B,C). In addition, we found high IL-6 and IL-8 cytokine levels in the plasma of LCH patients with multisystem disease (Fig 4F).
A neurodegenerative syndrome (LCH-ND) characterized by progressive ataxia, learning and behavior di culties arises in some patients with LCH and remains one of the most di cult to treat LCH complication. We previously demonstrated the presence of BRAFV600E + perivascular microglial-like cells along with blood circulating BRAFV600E + mononuclear cells in patients with LCH-ND. To determine whether senescence could also contribute to LCH-ND pathophysiology, we examined brain autopsy sections of a patient with fatally progressive ND-LCH . Strikingly, we found multifocal aggregates of enlarged BRAFV600E + cells expressing monocyte-macrophage markers (CD14 + , CD163 + , CD33 + ) overexpressing the senescent marker p16 INK4a aggregating in areas of the white matter ( Fig 4G). These cells also lacked the microglial marker (P2RY12, Sup Fig 4E) and accumulated in perivascular regions suggesting that senescent circulating MNP that are recruited from the blood to the brain parenchyma rather than brain resident microglial cells contribute to ND in this patient.
Taken together, these data suggest that the BRAFV600E mutation induces a senescence program in HPC that persisted in differentiated LCH cells that accumulate in the peripheral tissue sand brain parenchyma.

Pharmacological blockade of mTOR pathway inhibits SASP induction and limits MNP differentiation of BRAFV600E + HPC
Our results above suggesting that the secretome produced by BRAFV600E + HPC contributed to the skewing of HPC toward the MNP lineage, prompted us to examine whether SASP inhibition could help rescue HPC multi-lineage differentiation potential. SASP includes a wide range of cytokines that can contribute to MNP skewing and therefore targeting a speci c cytokine might not be su cient to obtain a therapeutic bene t. SASP induction in senescent cells is thought to be driven by the sustained activation of the mammalian target of rapamycin (mTOR) pathway (Herranz et al., 2015;Laberge et al., 2015). Thus, we asked whether rapamycin, a potent mTOR inhibitor could, could reduce MNP differentiation skewing of BRAFV600E + senescent HPC by its ability to blunt the pro-in ammatory phenotype of senescent cells.
We rst con rmed that inhibition of the mTOR pathway was su cient to reduce the production of in ammatory cytokines by BRAFV600E + HPC senescent cells in vitro ( Fig 5A). Importantly, and in line with our hypothesis, SASP inhibition signi cantly reduced BRAFV600E + HPC skewing into MNP in vitro ( Fig 5B). Rapamycin treatment also partially reduced excess MNP differentiation from BRAFV600E -GFP -HPC co-cultured with senescent BRAFV600E + GFP + HPC, further establishing that mTOR inhibition reduced the release of SASP-associated molecules driving HPC sustained MNP differentiation potential ( Fig 5B).
Importantly, rapamycin administration to BRAFV600E Scl+ mice reduced the accumulation of BM GMP and MNP, while BM neutrophils were not affected by the treatment (Fig 5C, D and Sup Fig 5A). Rapamycin administration reduced organomegaly in BRAFV600E Scl+ animals ( Fig 5E) and the in ammatory in ltrate in tissues (Fig 5F), improving LCH disease. Of note, rapamycin did not induce apoptosis of BRAFV600E + cells (Sup Fig 5B). While we cannot exclude that improved LCH disease in treated BRAFV600E Scl+ animals may also be linked to a direct effect of rapamycin on T cells-in ltrating LCH lesions, the ability of rapamycin to inhibit the release of in ammatory cytokines from BRAFV600E HPC and the MNP skewing induced by secreted factors of BRAFV600E + HPC, in addition to rapamycin inhibition of GMP and MNP accumulation in the BM of BRAFV600E Scl+ animals, emphasizes the strong contribution of mTOR-driven SASP to the increased MNP differentiation observed in mice and human BRAFV600E HPC.

Elimination of senescent cells improves the clinical outcome of LCH bearing mice
To measure the exact contribution of senescence to LCH pathogenesis, we took advantage of mice expressing the INK-ATTAC transgene (inducible elimination of p16 INK4a -positive senescent cells upon administration of AP20187) generated as described (Baker et al., 2011). The INK-ATTAC transgene includes an open reading frame (ORF) coding for the enhanced green uorescence protein (eGFP) and the FK506-binding-protein-caspase 8 (FKBP-Casp8) fusion protein, expressed under the promoter of the CDKN2A gene, which encodes for p16 INK4a protein. These mice enable CDKN2A/p16 INK4a+ senescent cells to be visualized based on GFP expression and conditionally deleted upon administration of AP20187 (AP), a synthetic drug that induces the dimerization of a membrane-bound myristoylated FKBP-Casp8 leading to the apoptosis of Caspase 8 expressing cells.
INK-ATTAC mice were crossed to BRAFV600E ca/ca ;Scl cre-ER (BRAFV600E Scl+ ATTAC + ) or to BRAFwt ca/ca ;Scl cre-ER mice (BRAFwt Scl+ ATTAC + ) as a control group. Both groups received Tamoxifen for 5 days to induce LCH lesions and we measured the accumulation and composition of senescent cells in the BM ( Fig 5G). In line with our previous results, we observed a signi cant accumulation of GFP + senescent cells in the bone marrow of BRAFV600E Scl+ ATTAC + animals compared to BRAFwt Scl+ ATTAC + control littermates ( Fig 5H). The majority of bone marrow GFP + senescent cells were skewed towards the MNP lineage and led to enhanced accumulation of senescent macrophages and dendritic cells but not neutrophils in BRAFV600E Scl+ ATTAC + compared to BRAFwt Scl+ ATTAC + mice ( Fig 5I). BRAFV600E Scl+ ATTAC + were then treated with AP to delete senescent cells or with the vehicle control for 3 weeks. AP treatment cleared GFP + cells from the bone marrow of BRAFV600E Scl+ ATTAC + mice ( Fig 5J). Importantly, clearance of GFP + senescent cells from BRAFV600E Scl+ ATTAC + reduced organomegaly of LCH -bearing mice ( Fig 5K) and reduced the immune in ltrate in liver and lung tissues (Fig 5L), showing that senescent cells are responsible of LCH pathophysiology.
Finally, to further examine whether pharmacological elimination of senescent cells can improve LCH outcome, we used ABT-263 a speci c inhibitor of the anti-apoptotic proteins BCL-2 and BCL-xL, which are both highly expressed in senescent cells. While not speci c to senescent cells, ABT-263, was shown to improve the clinical outcome of several senescence -associated diseases (Bussian et al., 2018;Chang et al., 2016;Jeon et al., 2017). Similar to most senescent cells, BCL-2 and BCL-xL are highly expressed in BRAFV600E + HPC and LCH cells in peripheral tissues. We had previously shown that inhibition of BCL-xL can e ciently eliminate LCH cells (Hogstad et al., 2018). Here we asked whether ABT-263 administration can improve outcome in LCH bearing mice. We administered ABT-263 to BRAFV600E Scl+ animals or BRAFwt Scl+ control littermates for 3 weeks (Sup Fig 5C). ABT-263 diluent prevents its systemic administration and both gavage of ABT-263 or diluent control are quite toxic to mice making it di cult to measure the bene cial outcome of ABT-263 in treated mice. Nonetheless and in accordance with ex vivo experiment, we found in BRAFV600E Scl+ mice that were able to survive, speci c clearance of BRAFV600E + cells (Sup Fig 5D), reduced organomegaly (Sup Fig 5E) and reduced immune lung in ltrate (Sup Fig 5F) in mice treated with ABT-263. These results suggest that ABT-263 could provide a novel clinical strategy to treat LCH patients either alone or in combination with BRAF inhibitors.

Discussion
Using genetically engineered mouse models and humanized mouse models, we demonstrate that expression of the BRAFV600E mutation in mouse and human HPC is su cient to drive the formation of LCH lesions with a tissue distribution comparable to human systemic LCH. We also showed that enforced BRAFV600E mutation in mice and human HPC as well as primary human CD34 + hematopoietic progenitors isolated from LCH patients are in a senescent state and that BRAFV600E-induced senescence in hematopoietic progenitors contribute to LCH pathophysiology. Finally, using the INK-ATTAC transgene, we con rmed the accumulation of senescent cells in the bone marrow and demonstrate that depletion of senescent cells in LCH bearing mice improves disease outcome.
Our ndings showing that BRAFV600E-induced HPC senescence contributes to LCH lesions shed light into key features of LCH disease, such as (1) the accumulation of poorly proliferative MNP driven by the expression of the senescent-associated cell cycle inhibitor p16 INK4a , and (2) the large immune in ltrate, with subsequent brotic injuries that accumulates in LCH lesions, which we show is driven by SASP + senescent LCH cells.
We have previously generated a mouse model in which, one allele of BRAFV600E was induced under the CD11c promoter expressed by MNP progenitors and mature dendritic cells (Berres et al., 2015). In that model, mice developed LCH-like disease in the spleen, lung and liver but only minimal disease in the skin and bones, which was surprising as human LCH lesions frequently involve skin and bone tissues (Berres et al., 2015). In the current study, we showed that when BRAFV600E mutation was driven by the Scl promoter, expressed in pluripotent hematopoietic progenitors upstream of the MNP progenitors (BRAFV600E Scl+ mice), mice developed -in addition to LCH-like lesions in the spleen, lung and liver tissues -skin lesions in the dermis that mimicked human LCH lesions, as well as granulomatous lesions, indicating that expression of the BRAFV600E mutation at the HSC level drives a more clinically relevant LCH-phenotype than the disease that forms upon induction of the BRAFV600E mutation in MNP progenitor stage.
The BRAFV600E oncogene-induced senescence is best known for its ability to protect human naevi from malignant transformation through the induction of cell cycle arrest (Michaloglou et al., 2005). In this study, we showed that expression of the BRAFV600E oncogene in early hematopoietic progenitors also promotes the induction of a senescence program and that senescence contributes to shaping the LCH phenotype. Cellular senescence refers to a cellular state of irreversible cell cycle arrest induced by various cell stressors such as oncogenes. The molecular effectors of senescence include the activation of canonical tumor suppressors like p16 INK4a and p53. This event leads to an arrest in proliferation that prevents senescent cells from re-entering the cell cycle, while also undergoing anabolic processes driven by high levels of stress-induced nutrient-dependent mTOR activation. The senescent cell signature includes positivity for SAbGal staining, expression of the cell cycle inhibitor CDKN2A (p16 INK4a ), and the production of mTOR-driven in ammatory cytokines including IL-1, IL-6, IL-8, as part of a SASP program. We show here that BRAFV600E HPC have a much lower proliferative capacity in vitro and in vivo which likely explains their decreased ability to engraft in competitive transplant assays when compared to wild type HPC (Fig 3A). Instead, they are more prone to differentiate into pathogenic BRAFV600E senescent MNP, which persist in tissues for prolonged periods of time where they continue to release SASP-driven cytokines. The senescence program explains poorly understood features of LCH cells including their reduced proliferation rate and the production of in ammatory cytokines that have been widely reported in patients (Allen et al., 2010(Allen et al., , 2018. The SASP-driven cytokine production also contributes to subsequent brotic injury, which at critical sites (i.e. CNS including pituitary, hepatic bile ducts) leads to LCH morbidities. In addition to inducing senescence, our previous ndings that BRAFV600E expression downregulates CCR7 expression in LCH cells and traps them in peripheral tissues may further contribute to their pathogenic accumulation locally (Hogstad et al., 2018).
While CDKN2A/B deletion has never been described in LCH lesions, clonal deletion of CDKN2A/B have been described in Langerhans Cell Sarcoma (Xerri et al., 2018), a rare and highly aggressive disease de ned by potent histiocyte proliferation with cytologic atypia and increased mitotic index. These results suggest that BRAFV600E-induced senescence likely protects LCH cells from malignant transformation, similar to what has been observed in melanocytic naevi.
The nature of the in ammatory in ltrate in LCH lesions, while well-known and widely described (Picarsic and Jaffe, 2015), has remained poorly understood, prompting several groups to search for a microbial origin of LCH (McClain et al., 1994;Jenson et al., 2000). Here we showed that it is instead the SASP program that drives the release of in ammatory cytokines in BM HPC and differentiated cells in the periphery. We also showed that the senescence-driven cytokine release by BM HPC drives the induction of a MNP-differentiation program. SASP is thought to be induced upon mTOR activation, which is activated in senescent cells upon stress-induced cell cycle arrest. The release of in ammatory cytokines not only promotes the differentiation of HPC into MNP cells, but also likely contributes to the development of granulomatous lesions in peripheral tissues, common in LCH lesions, and subsequent brotic injury. mTOR inhibition by rapamycin has been shown to reduce SASP expression (Herranz et al., 2015;Laberge et al., 2015). In accordance with these studies, we found that mTOR inhibition reduced HPC skewing toward the MNP lineage and improved disease phenotype in vivo. Altogether, our results showing that SASP contributes to enhanced myeloid cell differentiation in the BM, could explain the myeloid-biased hematopoiesis observed in aged patients (reviewed in (Montecino-Rodriguez et al., 2013)) and if con rmed could suggest that mTOR inhibition in HPC could help restore immune balance and homeostasis in older patients.
Systemic LCH can lead to severe morbidity in patients including an increased risk of neurodegenerative syndrome (LCH-ND) that can arise years after patients are presumably cured (Allen et al., 2018). The pathogenesis of the ND syndrome in LCH patients remains poorly understood and a clinical unmet need in LCH patients. Recent studies in mice suggest that BRAFV600E expression in murine embryonic erythromyeloid-progenitor (EMP) can lead to a neurodegenerative in ammatory disorder without systemic involvement (Mass et al., 2017), although this is rarely the case in patients where LCH-ND arises more frequently in patients with longer periods of uncontrolled disease (reviewed in (Haroche et al., 2017;Yeh et al., 2018)).
We recently found that circulating BRAFV600E + monocytes are present in higher frequency at all stages of LCH-ND suggesting that LCH-ND patients likely harbor BRAFV600E + HPC in the BM . In addition, we also identi ed diffuse perivascular in ltration of BRAFV600E + cells with monocyte phenotype lacking microglial markers in the brain tissue of patient with LCH-ND , indicating that the accumulation of circulating MNP in the brain of LCH patients may contribute to subsequent neurodegeneration in these patients. Prior results have shown that senescence can lead to cognition-associated neuronal loss in mice (Bussian et al., 2018). Here we show that circulating MNP that accumulate in LCH-ND lesions also express the senescent marker p16 INK4a suggesting that senescent cells may drive LCH-ND in patients with LCH. These results suggest senolytics may help improve this grueling LCH complication.
Systemic LCH remains a di cult to treat condition that can require the use of high-dose chemotherapy followed by allogenic BM transplantation (reviewed in (Allen et al., 2015;Donadieu et al., 2015)). BRAF or MEK inhibition therapies can improve clinical outcome, although this approach is not curative in most cases as most patients recur upon treatment interruption (Cohen Aubart et  Yeh, E.A., Greenberg, J., Abla, O., Longoni, G., Diamond, E., Hermiston, M., Tran, B., Rodriguez-Galindo, C., Allen, C.E., McClain, K.L., et al. (2018). Evaluation and treatment of Langerhans cell histiocytosis patients with central nervous system abnormalities: Current views and new vistas. Pediatr. Blood Cancer 65, e26784.

Animal experiments
All animal experiments performed in this study were approved by the Institutional Animal Care  2005)). BRAFV600E ca/-X Scl cre-ER were then crossed with Rosa26 YFP/YFP mice (C57BL/6 background; The Jackson Laboratory). To induce the cre recombinase, BRAFV600E Scl+ , BRAFwt Scl+ , or BRAFV600E Sclcontrol littermates were injected intra-peritoneally with tamoxifen for 5 consecutive days (5 mg, day 1; 2 mg, days 2, and 3; 1 mg, days 4 and 5). All animals were housed under specific pathogen-free conditions and sacrificed at the indicated time points. All experiments were controlled using cre-positive littermates negative for the BRAFV600E construct or littermates negative for the cre recombinase transgene construct. BRAFV600E or BRAFwt chimeras were generated by transplantation of 1 to 3 million whole BM cells flushed from the femurs of 8 weeks old BRAFV600E Scl+ or BRAFwt Scl+ mice following the tamoxifen injections, into lethally irradiated (650 cGy X 2) CD45.1 mouse aged 8-12 weeks (C57BL/6 background; Charles River Laboratory). Mice were kept on Sulfamethoxazole / Trimethoprim (STI Pharma) for 3 weeks.
Mice were allowed to recover for 2 weeks after transplantation before initiation of drug treatments. BRAFV600E Scl+ ATTAC + and BRAFwt Scl+ ATTAC + mice were generated by crossing the litters from BRAFV600E ca/ca mice and Scl cre-ER with ATTAC +/mice (kindly provided by Lorenzo Galuzzi, Weill Cornell Medical College University, New York, NY). The cre recombinase was induced as described earlier.
NOD scid gamma (NSG) mice were purchased from the Jackson Laboratory (BALB/c background) and were housed under specific pathogen-free conditions. Humanized mice were generated by transplanting 8 weeks old NSG mice sub-lethally irradiated with 250 cGy with 200,000 human CD34 + cells previously transduced for 16 to 24 hours with the BRAFV600E or NGFR lentiviral vector. Poulikakos) (plasmid) using High Fidelity Taq polymerase. PCR products were purified using the QIAquick PCR purification Kit (Qiagen), and restriction enzymes SalI and XbaI were used to digest the vector and PCR products. PCR products were ligated with the expression vector using NEB Quick Ligase and the plasmid was verified with Sanger Sequencing.

Lentiviral vector production and titration
Lentiviral vectors were produced as previously described (Baccarini et al., 2011). Briefly, 293T cells were seeded approximately 24 hours before transfection in a 15 cm plate and incubated at 37°C with 5% CO . The following day, the cells were transfected with a third generation packaging system (pVSV, pMDLg/pRRE, pRSV-REV) and the appropriate transfer plasmid using calcium phosphate. The media was changed 14 hours later. Another 30 hours later, the supernatant containing the vector was harvested, passed through a 0.22μm filter and ultracentrifuged (20,000 rpm for 2 hours) to concentrate the vector 300 -500-fold. The vector particles were resuspended in sterile PBS, aliquoted and stored at -80°C. Vector titration was performed on 293T cells using the limiting dilution method.

Cord blood processing and transduction
Mononuclear cells from cord blood were isolated via a Ficoll-Paque PLUS (GE Healthcare) gradient. Mononuclear cells were washed and resuspended in FACS buffer (PBS without Ca 2+ 2 and Mg 2+ supplemented with 2% heat inactivated FBS and 5mM EDTA). CD34 + cells were purified using a CD34 MicroBead kit Human from Miltenyi Biotec (130-100-453). CD34 + human hematopoietic progenitor cells were resuspended at a concentration of 1 million cells/mL in Stemspan media SFEM II (StemCell Technologies, 09605) supplemented with 100ng/mL SCF, 100ng/mL FLT3-L and 50ng/mL TPO (so called stem-cell media). 200,000 cells were plated in a 48 well plate and 5μL of lentiviral construct (either BRAFV600E or NGFR) was added to the media for a MOI=50. The cells were incubated at 37°C and 5% CO .
Supplemented Stemspan media was added 24 hours after the initiation of transduction and cells were kept in culture for up to 14 days. GFP-transduced cells were analyzed for myeloid and lymphoid surface marker expression (see antibody list) starting two days after the transduction.

Flow cytometry and FACS
Adult skin (from the ears), lung, liver, spleen, and femurs were dissected from mice at the age of 12-16 weeks. Adult skin was first incubated overnight at 4° in dispase II solution (Roche) (2.5mg/mL in PBS) with the dermal side facing down. The dermis was separated mechanically from the epidermis and both tissues were digested in a solution of collagenase D (Roche) (1 mg/mL) and DNase I (Roche) (1 mg/mL) in RPMI (Corning) +10% FBS for one hour at 37°C.
After digestion, the dermis and epidermis were homogenized using a 18G needle. The cell suspension was filtered through a 70μm cell strainer into a flow cytometry tube.
For the spleen and the lung, tissues were enzymatically digested in a solution of collagenase D (Roche) (1 mg/mL) and DNase I (Roche) (1 mg/mL) in HBSS (Corning) for 30 minutes at 37°C, followed by mechanical trituration with an 18G blunt tipped syringe and filtration through a 70μm filter.
BM single cell suspensions were obtained after flushing out the BM from the femur using a 27G needle and then were incubated with ACK lysis buffer (420301, BioLegend) for 3 minutes at room temperature. Blood was drawn from the liver sinus and treated twice with ACK lysis buffer for 5 minutes at room temperature.
For flow cytometry, cells were stained in FACS buffer (PBS without Ca 2+ and Mg 2+ supplemented with 2% heat-inactivated FBS and 5mM EDTA) with flow cytometry monoclonal antibodies for 20 minutes at 4°C. Murine hematopoietic progenitor cells were sorted as singlets of DAPIlineagec-kit + Sca1 -, and murine LSK + cells were sorted as singlets DAPIlineagec-kit + Sca1 + . Human transduced HPC were sorted as DAPI -, GFP + , or GFP-singlets.
Flow cytometry was performed using a Fortessa analyzer (BD) and FACS sorting was performed using a FACS Aria II (BD) or LSRII (BD). Flow cytometry data analysis was performed using FlowJo (TreeStar) software.

Markers
Reactivity Multiplexed Immunohistochemical Consecutive Staining on a Single Slide Tissues were fixed in 4% formaldehyde for 24 hours and embedded in paraffin. Five mm-thick formalin-fixed paraffin-embedded (FFPE) tissue sections on glass slides were baked at 37°C overnight, deparaffinized in xylene, and rehydrated in decreasing concentrations of ethanol.
Then, tissue sections were incubated in retrieval solution (pH6 or 9) for antigen retrieval at 95°C for 30 minutes. Tissue sections were incubated in 3% hydrogen peroxide and in serumfree protein block solution (Dako, X0909) before adding the primary antibody for 1 hour at room temperature. After signal amplification using EnVison + System-HRP labelled Polymer anti-mouse (Dako, K4001) or anti-rabbit (Dako, K4003), chromogenic revelation was performed using 3-amino-9-ethylcarbazole (AEC, Vector Laboratories SK4200). Slides were counterstained with hematoxylin, mounted with a glycerol-based mounting medium (Dako, C0563) and scanned for digital imaging (Hamamatsu NanoZoomer S60 Whole Slide Scanner). Then, the same slides were successively stained as described (Remark et al., 2016).   -7). (G) Cartoon shows the scheme used to generate the humanized LCH mouse.
(H) Liver, lung and spleen tissues sections isolated from humanized mice reconstituted with BRAFV600Ehu HPC were stained with anti-CD207 and anti-CD1a antibodies. Graph represents the number of CD207+ or CD1a+ cells per mm2 in tissue sections isolated from animals (n= 2-3). Data are represented as mean ± s.e.m; statistical signi cance analyzed by an unpaired two-sided t-test is indicated by *p < 0.05; **p < 0.01; ***p < 0.001. statistical signi cance analyzed by an unpaired two-sided t-test is indicated by *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. transduced HPC (BRAFV600Ehu) that were treated with rapamycin or DMSO diluent, was quanti ed by ow cytometry after 7 days of culture. (C-E) BRAFV600EScl+ mice were treated with rapamycin (0.5 mg/ kg body weight/ day) or vehicle control for 10 days (C) Percentage of GMP (de ned as DAPI-lineage-ckit+ Sca-CD16/32+ CD34high/int) among lineage negative BM cells and (D) The percentage of dendritic cells, macrophages and neutrophils among total CD45+ BM cells were measured using ow cytometry, of 3 experiments and are represented as mean ± s.e.m, statistical signi cance analyzed by an unpaired and paired (for 5A and B) two-sided t-test is indicated by *p < 0.05; **p < 0.01; ***p < 0.001.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.