key: cord-0426698-nfxn7l27
authors: Anselmi, Giorgio; Vaivode, Kristine; Dutertre, Charles-Antoine; Bourdely, Pierre; Missolo-Koussou, Yoann; Newell, Evan; Hickman, Oliver; Wood, Kristie; Saxena, Alka; Helft, Julie; Ginhoux, Florent; Guermonprez, Pierre
title: Engineered niches support the development of human dendritic cells in humanized mice
date: 2019-11-08
journal: bioRxiv
DOI: 10.1101/835223
sha: e8014deacf2ebb6d85682fbe76f8215b1b5b15ec
doc_id: 426698
cord_uid: nfxn7l27

Classical dendritic cells (cDCs) are rare sentinel cells specialized in the regulation of adaptive immunity. Modeling cDC development is both crucial to study cDCs and harness their potential in immunotherapy. Here we addressed whether cDCs could differentiate in response to trophic cues delivered by mesenchymal components of the hematopoietic niche where they physiologically develop and maintain. We found that expression of the membrane bound form of human FLT3L and SCF together with CXCL12 in a bone marrow mesenchymal stromal cell line is sufficient to induce the contact-dependent specification of both type 1 and type 2 cDCs from CD34+ hematopoietic stem and progenitor cells (HSPCs). Engraftment of these engineered mesenchymal stromal cells (eMSCs) together with CD34+ HSPCs creates an in vivo synthetic niche in the dermis of immunodeficient mice. Cell-to-cell contact between HSPCs and stromal cells within these organoids drive the local specification of cDCs and CD123+AXL+CD327+ pre/AS-DCs. cDCs generated in vivo display higher levels of resemblance with human blood cDCs unattained by in vitro generated subsets. Altogether, eMSCs provide a novel and unique platform recapitulating the full spectrum of cDC subsets enabling their functional characterization in vivo.

Classical human dendritic cells (cDCs) are sentinels of the immune system with a unique 44 ability to regulate the function of T lymphocytes 1 . Dendritic cells can induce immune 45 tolerance 2 or drive the development of immunity 3 . 46 The analysis of blood circulating subsets has revealed that cDCs consist in two major 47 subtypes: CD141 + XCR1 + Clec9A + DCs (cDC1) and CD1c + CD11c + CD172a (SIRPa) + DCs 48 (cDC2s) 4-6 . Both cDC1s and cDC2s arise from a bone marrow committed progenitor 7 or 49 from early IRF8 + multipotent progenitors 8, 9 , which generate a common circulating precursor 50 10 that progressively diverge in pre-cDC1s and pre-cDC2s 10-12 . Type 1 DCs are conserved 51 between mouse and human, and they share the expression of specific surface markers such as 52 Clec9A 13 and XCR1 5 as well as the transcription factor (TF) IRF8, which is essential for the 53 development of murine cDC1s 4-6, 13-15 . Conversely, human CD1c + type 2 DCs have been 54 shown to express the IRF4 TF 16 , which controls the development of their phenotypic 55 equivalent in the mouse model 16, 17 . This rather simple picture is complicated by the diversity 56 of CD1c + cells, which encompass migratory DCs as well as CD14 int inflammatory DCs 57 recruited during inflammation [18] [19] [20] . More recently, un-biased approaches have uncovered a 58 deeper complexity in the DC network with the identification of 2 types of cDC2s with 59 distinct transcriptional profiles and the identification of AXL + CD11c + CD1c + cells which 60 have been proposed to act as a precursor for cDCs 12, 21 . other niche factors 28 . At steady state, Flt3-ligand (FLT3L) is delivered as a membrane-bound 67 precursor expressed on radio-resistant stromal cells [29] [30] [31] . After egressing from the bone 68 marrow, DC precursors circulate in the blood and seed the peripheral tissues 32 . In the lymph 69 node, stromal fibroblastic reticular cells provide FLT3L 33 , and FLT3-dependent proliferation 70 of cDC in periphery is required for their maintenance 34, 35 . 71 Modelling the development of cDCs in culture systems is instrumental to better understand 72 their ontogeny and define their immunological function. Pioneer work from Banchereau et al. 73 have identified that GMCSF and TNF-a cooperate to produce CD1a + cells with features of Here we aimed at modeling human cDC development by providing physiological factors 92 associated to hematopoietic niches. We found that engineered mesenchymal stromal cells 93 (eMSCs) expressing a combination of membrane-bound FLT3L and SCF/KITL together with 94 CXCL12 provide a scaffold for human cDC differentiation. Engraftment of eMSCs along 95 with CD34 + HSPCs leads to the local development of cDCs in immunodeficient mice. This in 96 vivo system recapitulates the differentiation of pre/AS-DCs, cDC1s and cDC2s with an 97 unreached level of similarity with the phenotype of human blood cDCs. 

Stromal membrane-bound FLT3L is sufficient to support human cDC differentiation 100 from CD34 + HSPCs 101 We hypothesized that the interaction of hematopoietic progenitors with membrane bound 102 factors expressed by stromal cells of the niche would promote the specification of the cDC 103 lineage.

To test this, we engineered a bone marrow-derived murine mesenchymal cell line (MS5) 7 53 105 to stably and homogeneously express the transmembrane form of human FLT3L (MS5_F) as 106 probed by flow cytometry (Fig. 1a) . Co-culture of MS5_F with CD34 + HSPCs drives the 107 appearance of cDC1-like Clec9A + CD141 + and cDC2-like CD14 -CD1c + cells. Importantly, 108 MS5_F is more efficient than recombinant soluble FLT3L (MS5+recFL) in generating cDC-109 like cells (Fig. 1b) .

In contrast, OP9 54 hematogenic stromal cells stably expressing membrane bound FLT3L 111 (OP9_F) were less efficient than MS5_F in driving cDC differentiation (Fig. 1c) . Besides, 112 MS5_F also promoted the appearance of CD123 + CD303/4 + cells resembling either pDCs or 113 pre/AS-DCs 12, 21 (Supplementary Fig.1a and b) . 114 Next, we wanted to test whether cell-to-cell interactions mediate the differentiation of cDCs 115 driven by FLT3L-expressing MS5 stromal cells. Using transwell permeable to soluble factors 116 but preventing cognate interactions, we found that direct contact is required to support 117 efficiently cDC differentiation (Fig. 1d) . 118 Collectively, these data show that membrane FLT3L expression in stromal cells provide an 119 improved platform to trigger the differentiation of cDC-like cells from CD34 + HSPCs in vitro 120 via cell-to-cell contact. 124 Next, we sought to improve the efficiency of cDC production in MS5_F by co-expressing 125 additional niche factors. We focused on SCF, CXCL12 and TPO because of their essential 126 role in supporting HSPCs maintenance in the bone marrow niche 22, 24-27, 55 . SCF had also 127 been extensively used in previously published DC culture protocols 7, 39, 40, 56 .

To this end, we generated a collection of MS5 stromal cells stably expressing either one, two, 129 three or four human factors by combining CXCL12, TPO and membrane-bound SCF/KITL, 130 with or without membrane-bound FLT3L ( Supplementary Fig. 2a) . 131 We screened this collection of engineered mesenchymal stromal cell (eMSC) lines based on 132 their ability to support human cDC differentiation from cord blood-derived CD34 + HSPCs.

At day 15, only FLT3L-expressing eMSCs successfully supported the differentiation of 134 CD141 + Clec9A + and CD14 -CD1c + cells ( Fig. 2a and Supplementary Fig. 3a) . We conclude 135 that FLT3L is necessary for the differentiation of cDCs using eMSCs. Besides, optimal cDC 136 production was obtained in cultures containing eMSC co-expressing membrane bound SCF 137 and CXCL12 together with FLT3L (MS5_FS12) (Fig. 2a) , whereas no difference was 138 observed for CD14 + CD16monocytes and CD14 + CD16 + macrophages as compared to 139 MS5_CTRL ( Supplementary Fig. 3b ).

Furthermore, we noticed that in vitro differentiated CD14 -CD1c + cDC2-like cells were 141 heterogeneous for the expression of the mannose receptor CD206 (Fig. 2a) . Circulating blood 142 cDC2s do not generally express CD206 ( Supplementary Fig. 3c ) whereas CD206 is a marker 143 of skin and migratory cDC2 19, 57 .

Most of the previously described protocols to generate human DC-like cells in vitro from 145 8 both CD14 + monocytes and CD34 + HSPCs made an extensive use of the cytokine GM-CSF 7, 146 8, 37, 39, 40, 56 , with one exception 49 . Since we did not include GM-CSF in our protocol, we 147 wanted to assess whether human GM-CSF was spontaneously produced in CD34 + cultures. 148 We could not detect any GM-CSF from co-culture supernatant (Fig. 2b) . Accordingly, GM-149 CSF blocking antibody did not impact the generation of cDCs driven by MS5_FS12 (Fig.   150 2c). We conclude that GM-CSF is dispensable for the generation of cDCs in vitro, as 151 previously reported both in mouse 58, 59 and human 49 . 152 We also observed that MS5_FS12 stromal cells significantly improve the differentiation of 153 CD123 + CD303/4 + cells (Fig. 3a) , a phenotype shared by both plasmacytoid DC and pre/AS- Human DCs generated in vitro using MS5_FS12 align with circulating blood DCs 168 In order to validate the identity of the cDCs generated using the MS5_FS12 stromal niche, 169 9 we compared the transcriptome (RNA-seq) and phenotype (CyTOF) of in vitro differentiated 170 subsets to circulating blood cDC1s and cDC2s ( Fig. 4a-f ).

Hierarchical clustering of RNA-seq data revealed that subsets generated in culture maintain a 172 strong "culture imprinting" (Supplementary Fig. 4a ). Indeed, we could identify a 2000 genes 173 signature (1000 genes up and 1000 genes down-regulated), which clearly separates in vitro 174 generated cells from circulating blood subsets regardless of their subset identity 175 ( Supplementary Fig. 4b ). The majority of these genes were associated to cell cycle and 176 metabolism as shown by pathways enrichment analysis ( Supplementary Fig. 4c ).

Nonetheless, once this "in vitro culture signature" was subtracted from the total protein 178 coding genes, CD141 + Clec9A + and CD1c + CD206 +/cells generated in culture 179 transcriptionally align to blood cDC1 and cDC2, respectively (Fig. 4a) . CD1c + cells have recently been shown to contain two distinct subsets termed as cDC2 and 186 cDC3 21 . Alignment of cultured cells was probed towards total CD1c + cells (CD1c>cDC1), 187 cDC2 (cDC2>ALL and cDC2>DC3) and DC3 signatures (DC3>cDC2 and DC3>ALL). We 188 found that in vitro generated CD141 + Clec9A + and CD1c + CD206 +/cells are enriched in genes 189 defining circulating blood cDC1 and cDC2, respectively (Fig. 4b) . The expression of the top 190 50 genes for each signature in the differentiated subsets further supports this conclusion (Fig.   191 4c). Importantly, both CD206 + and CD206subsets aligned preferentially to cDC2 as 192 compared to DC3 and cDC1 ( Fig. 4b and Supplementary Fig. 4d ). CD163 was recently 193 described as a marker selectively expressed in blood cDC3 as compared to cDC2 21 .

Supporting our previous conclusion, CD163 was neither expressed in CD1c + CD206nor in 195 CD1c + CD206 + cells generated in vitro, while CD163 + cells were detected among CD14 + 196 monocytes and CD14 + CD16 + macrophages ( Supplementary Fig. 4e ).

To obtain a more exhaustive characterization of the phenotype of in vitro generated subsets 198 we performed CyTOF analysis using a panel of 38 metal-conjugated monoclonal antibodies.

Dimension reduction of the CyTOF data was performed using the Uniform Manifold However, the yield of in vitro generated cDC was significantly higher at day14, when most of 220 the cultures were therefore analyzed ( Supplementary Fig. 4i ).

Collectively, our data demonstrate that: i) in vitro generated CD141 + Clec9A + recapitulate the 222 phenotype of bona fide blood cDC1; ii) CD14 -CD1c + cells align to cDC2 regardless of their 223 CD206 expression; iii) CD123 + CD303 + cells contain some recently described pre-DC/AS-224 DC phenotypically and functionally distinct from pDCs. However, we identified two major 232 We next wanted to assess whether we could use MS5_FS12 to recapitulate a more 233 physiological niche microenvironment supporting human HSPCs maintenance in vivo. 234 To this end, we designed an experimental strategy based on the subcutaneous injection of 235 cord blood-derived CD34 + HSPCs together with MS5_FS12 in a basement membrane matrix 236 (Matrigel) in NSG mice (Fig. 5a ).

Clusters of cells embedded in Matrigel can be identified as early as day 12 by tissue histology 238 (Fig. 5b) . Flow cytometry analysis demonstrated that MS5_FS12 but not MS5_CTRL 239 induced the expansion of human leukocytes within the Matrigel plugs (Fig. 5c ). We then 240 tested whether cell-to-cell interactions of eMSC with human progenitors play a role in this 241 process. We injected two independent plugs of CD34 + HSPCs with either MS5_CTRL or 242 MS5_FS12 in the same recipient mouse (contralateral plugs) (Fig. 5d ). We found a relative The MS5_FS12 niche efficiently supports human cDC1, cDC2, pre/AS-DC and pDC 263 development from CD34 + HSPCs in vivo 264 We investigated whether the engraftment of CD34 + HSPCs together with MS5_FS12 could 265 lead to the local differentiation of human DC subsets.

Flow cytometry analysis of Matrigel organoids demonstrates that the MS5_FS12 but not the 267 13 MS5_CTRL niche specifically supports the differentiation of CD141 + Clec9A + cDC1-like 268 cells and CD14 -CD1c + cDC2-like cells ( Fig. 6a and Supplementary Fig. 6a ). This finding was 269 supported by immunofluorescence staining on plug sections highlighting the occurrence of 270 human CD45 cells expressing either Clec9A or CD1c (Fig. 6b) .

Further analysis revealed the expansion of CD123 + CD303/4 + cells in MS5_FS12 when 272 compared to MS5_CTRL plugs ( Fig. 6c and Supplementary Fig. 6a ). All these cells also 273 expressed CD45RA and heterogeneous levels of AXL and CD327, as previously described 274 for their in vitro counterparts (Fig. 6c) . However, only MS5_FS12 induced a strong 275 accumulation of AXL + CD327 + pre/AS-DC expressing various levels of CD1c (Fig. 6c) . In 276 addition, bona fide CD123 + CD45RA + AXL -CD327 lo/-pDCs could also be detected (Fig. 6c) .

RNA-seq analysis of in vivo generated CD123 + AXL -CD327 lo/and CD123 + AXL + CD327 + 278 cells further support this conclusion and unequivocally align them to blood circulating pDC 279 and AS-DC, respectively (Fig. 6d) . Supplementary Fig. 6c ). On the contrary, pDC were not significantly increased in the same 291 comparison ( Fig. 6f and Supplementary Fig. 6c ). We conclude that local cues associated to 292 14 the MS5_FS12 niche control cDC lineage commitment. In support of this view, we could not 293 detect a systemic increase in the levels of serum FLT3L in mice carrying engineered stromal 294 cell plugs (Fig. 6g) . Accordingly, spleen resident murine cDCs did not expand upon 295 MS5_FS12 engraftment while they massively do so upon administration of recombinant 296 soluble human FLT3L (Fig. 6g) . Together with the 2-plugs experiments (Fig. 6f) , these 297 observations suggest that most of the FLT3L aegis relies on its membrane bound form Table 6 ).

Moreover, we found that: i) AXL + Siglec6 + pre/AS-DCs generated in vivo (or in vitro) align 331 to their primary counterparts and selectively express a signature that distinguish them from 332 bona fide pDCs (DAB2, CD22, ADLH2 e.g.) (Fig. 7c) . ii) conversely, AXL -Siglec6bona fide 333 pDCs generated in vivo (or in vitro) align to their primary counterparts and express high 334 levels of markers distinguishing them from pre/AS-DCs (IRF7, GZMB, TCF4, BCL11A, e.g.) 335 (Fig. 7c) . iii) cDC2s generated in vivo (in NSG mice organoids carrying MS5_FS12) had 336 higher levels of similarity with blood cDC2s (including higher expression of BTLA, FCER1A, 337 e.g.) (Fig. 7b, 7d and 7e) . Recently, both CD5 + and CD5 -cDC2s subsets have been reported 338 in human blood 65, 66 and we found that in vivo generated cDC2s aligned particularly well 339 with blood CD5 + cDC2s (with the expression of CD5, CD2, e.g.) (Fig. 7b and 7e) . By 340 contrast, in vitro generated cDC2s expressed high levels of activation genes such MHC 341 molecules (HLA-DR, DQ, e.g.); co-stimulatory molecules (CD80, CD40, e.g.), activation 342 markers (ETS2, CCR6, CCR7, CXCL13, CCL22, e.g.) (Fig. 7e and 7f) and genes associated 343 with type I and type II interferon pathways (STAT1, IRF9, IGS15, GBP1, e.g.) 344 (Supplementary Fig. 7d and Supplementary Table 6 ).

All together, we conclude that MS5_FS12-containing organoids provide a unique scaffold for (Fig. 8a) . We then performed a mixed leukocyte reaction (MLR) by co-culturing CD4 + naïve T cells proliferation (Fig. 8b) , as expected and reported for circulating blood 367 cDC2 12, 21 (Supplementary Fig. 8b) . Conversely, pDC were significantly less effective on 368 triggering T cells activation, as shown by the consistent reduction in the frequency of 369 dividing CD4 + T cells when compared to cDC2 and pre/AS-DC (Fig. 8b) . Frozen samples were shipped to GENEWIZ ® where they were processed. RNA was extracted 546 and libraries were prepared using an ultra-low input RNA library preparation kit (Illumina).

Libraries were sequenced on HiSeq2500 (Illumina).

The raw sequencing data was initially aligned on the human reference genome hg38 using 549 STAR aligner (v2.5.3a) 76 . Raw read counts matrix was made by STAR (with the option - Table 4 ). Briefly, the transciptome of blood cDC1s and blood CD1c + cells was taken from 567 previously published data sets 63 . cDC1>CD1c + and CD1c + >cDC1 signatures were defined 

Dendritic cells and the control of immunity

Tolerogenic dendritic cells

Dendritic-cell-based therapeutic cancer vaccines

Superior antigen cross-presentation and XCR1 expression define 721 human CD11c+CD141+ cells as homologues of mouse CD8+ dendritic cells

The XC chemokine receptor 1 is a conserved selective marker of 724 mammalian cells homologous to mouse CD8alpha+ dendritic cells

Human CD141+ (BDCA-3)+ dendritic cells (DCs) represent a 727 unique myeloid DC subset that cross-presents necrotic cell antigens

Restricted dendritic cell and monocyte progenitors in human cord blood 730 and bone marrow

Dendritic Cell Lineage Potential in Human Early Hematopoietic 732 Progenitors

Lineage specification of human dendritic cells is marked by IRF8 734 expression in hematopoietic stem cells and multipotent progenitors

Circulating precursors of human CD1c+ and CD141+ dendritic cells

Human dendritic cells (DCs) are derived from distinct circulating 739 precursors that are precommitted to become CD1c+ or CD141+ DCs

Mapping the human DC lineage through the integration of high-742 dimensional techniques

DNGR-1 is a specific and universal marker of mouse and human 744

Batf3-dependent dendritic cells in lymphoid and nonlymphoid tissues

Type I interferons produced by dendritic cells promote their 747 phenotypic and functional activation

Human tissues contain CD141hi cross-presenting dendritic cells 749 with functional homology to mouse CD103+ nonlymphoid dendritic cells

IRF4 transcription factor-dependent CD11b+ dendritic cells in 752 human and mouse control mucosal IL-17 cytokine responses

IFN regulatory factor-4 and -8 govern dendritic cell subset 755 development and their functional diversity

Human inflammatory dendritic cells induce Th17 cell differentiation

Characterization of resident and migratory dendritic cells in human 759 lymph nodes

A 761 common pronociceptive pain modulation profile typifying subgroups of chronic 762 pelvic pain syndromes is interrelated with enhanced clinical pain

Single-cell RNA-seq reveals new types of human blood dendritic 765 cells, monocytes, and progenitors

Sl/Sld hematopoietic progenitors are deficient in situ

Early transplantation to a normal microenvironment prevents the 769 development of Steel hematopoietic stem cell defects

CXCL12 in early mesenchymal progenitors is required for 772 haematopoietic stem-cell maintenance

Chemokine receptor CXCR4-dependent internalization and resecretion 774 of functional chemokine SDF-1 by bone marrow endothelial and stromal cells

Critical role of thrombopoietin in maintaining adult quiescent 777 hematopoietic stem cells

Thrombopoietin/MPL signaling regulates hematopoietic stem cell 779 quiescence and interaction with the osteoblastic niche

Hematopoietic stem cell niche maintenance during 782 homeostasis and regeneration

Inflammatory Flt3l is essential to mobilize dendritic cells and 784 for T cell responses during Plasmodium infection

Dendritic cell homeostasis is 786 maintained by nonhematopoietic and T-cell-produced Flt3-ligand in steady state and 787 during immune responses

Flt3 permits survival during infection by rendering dendritic 789 cells competent to activate NK cells

The origin and development of nonlymphoid tissue CD103+ DCs

Transcriptional profiling of stroma from inflamed and resting 794 lymph nodes defines immunological hallmarks

The receptor tyrosine kinase Flt3 is required for dendritic cell 796 development in peripheral lymphoid tissues

Origin of dendritic cells in peripheral lymphoid organs of mice

GM-CSF and TNF-800 alpha cooperate in the generation of dendritic Langerhans cells

Efficient presentation of soluble antigen by cultured

 1953-1962 (1996 independent cord blood donors. ** p<0.01, ***p<0.001, one-way ANOVA test.