Distinct ontogenetic lineages dictate cDC2 heterogeneity

Conventional dendritic cells (cDCs) include functionally and phenotypically diverse populations, such as cDC1s and cDC2s. The latter population has been variously subdivided into Notch-dependent cDC2s, KLF4-dependent cDC2s, T-bet+ cDC2As and T-bet− cDC2Bs, but it is unclear how all these subtypes are interrelated and to what degree they represent cell states or cell subsets. All cDCs are derived from bone marrow progenitors called pre-cDCs, which circulate through the blood to colonize peripheral tissues. Here, we identified distinct mouse pre-cDC2 subsets biased to give rise to cDC2As or cDC2Bs. We showed that a Siglec-H+ pre-cDC2A population in the bone marrow preferentially gave rise to Siglec-H− CD8α+ pre-cDC2As in tissues, which differentiated into T-bet+ cDC2As. In contrast, a Siglec-H− fraction of pre-cDCs in the bone marrow and periphery mostly generated T-bet− cDC2Bs, a lineage marked by the expression of LysM. Our results showed that cDC2A versus cDC2B fate specification starts in the bone marrow and suggest that cDC2 subsets are ontogenetically determined lineages, rather than cell states imposed by the peripheral tissue environment.

To refine cDC2A and cDC2B identification, we used Clec9a Cre Rosa26 LSL-tdTomato Rbpj loxP/loxP mice (C9a tdTomatoΔRBPJ ) that lack Notch signaling in the cDC lineage and compared them to Clec9a Cre Rosa26 LSL-tdTomato controls (C9a tdTomato ).The number of cDC2As, but not cDC2Bs (as defined by the UMAP clusters), was reduced in C9a tdTomatoΔRBPJ mice in all organs analyzed (Fig. 1c).C9a tdTomatoΔRBPJ mice also displayed an increase in cluster 3 (CD8α − CD117 + Esam − ) across all tissues (Fig. 1c and Extended Data Fig. 2b), suggesting that these cells were immediate precursors of cDC2As whose terminal differentiation was arrested in the absence of Notch signals (hereafter early cDC2As) 4,5 .CD8α + tDCs were only found in spleen and MLN but were not decreased in C9a tdTomatoΔRBPJ mice (Fig. 1c and Extended Data Fig. 2c).Together with reports showing that cDC2Bs, but not cDC2As, are KLF4-dependent 17 , our data suggested that the overall heterogeneity of cDC2s can be distilled down to two main Notch-dependent T-bet + cDC2A and Notch-independent T-bet − cDC2B branches and states of differentiation along them.

Single-cell RNA-seq defines cDC2 heterogeneity at the pre-cDC2 level
We next identified pre-cDCs in tissues using a protocol developed for isolating lung pre-cDCs 18 .We gated on Lin − CD11c + MHC-II −/lo CD11b −/lo SIRPα − CD135 + CD43 + cells while excluding Ly6D + cells (precursors of both plasmacytoid cells 37,38 and tDCs 25 ) and CD11b hi SIRPα + CD16/32 + cells (monocyte-like cells and DC3 progenitors 39 ) (Extended Data Fig. 3a).Using in vitro differentiation assays (Extended Data Fig. 3b), fate mapping (Extended Data Fig. 3c) and in vivo Fms-like tyrosine kinase 3 ligand (Flt3L) dependence (Extended Data Fig. 3d), we confirmed that the gating strategy identified bona fide pre-cDCs in the bone marrow and spleen, as previously shown for the lung 18 .We used the gating strategy (Extended Data Fig. 3e) to sort pre-cDCs from the bone marrow, spleen and lung of C57BL/6J wild-type (WT) mice.We performed single-cell RNA-seq (scRNA-seq) analysis on 2,649 bone marrow, 4,371 spleen and 358 lung-sorted pre-cDCs after excluding a small number of dying cells and contaminants (identified using immune cell transcriptome profiles; https://www.immgen.org/)(Fig. 2a).We integrated the three tissues (bone marrow, spleen and lung) and generated a UMAP that identified nine clusters that, although varying in proportion, overlapped across all tissues (Fig. 2a).Therefore, we concatenated the cells from all tissues and used published gene signatures 15,30 to annotate the UMAP clusters.This approach indicated that clusters 4, 5 and 6 corresponded to proliferative early pre-cDCs (Fig. 2b).They were enriched in bone marrow (Fig. 2a,b), which is consistent with the fact that they originate in that tissue.Clusters 0 and 1 probably represented more differentiated pre-cDCs about to leave the bone marrow 40 or pre-cDCs that recently colonized peripheral tissues (Fig. 2b).Clusters 3, 2, 7 and 8 (late pre-cDCs) were overrepresented in peripheral tissues (Fig. 2a,b), where pre-cDCs complete differentiation into cDCs 1 .Overall, pre-cDCs segregated into two groups: one consisting of clusters 3 and 6 with a gene expression signature of pre-cDC1s/cDC1s; and one consisting of clusters 0, 1, 2, 4, 5, 7 and 8 and similar in gene expression to pre-cDC2s/cDC2s (Fig. 2c) 29,30 .We did not identify any RORγt fate mapping 15 , later shown to constitute a distinct lymphoid cell type rather than bona fide cDCs 16 .In a further study, KLF4-dependent cDC2s were suggested to correspond to cDC2Bs 17 .Finally, infection or cancer can drive the appearance of cells termed 'inflammatory cDC2s' and 'mature dendritic cells enriched in regulatory molecules', respectively 12,18,19 .Thus, at present, mouse cDC2s variably include cDC2As, cDC2Bs, Notch-dependent cDC2s, KLF4-dependent cDC2s, inflammatory cDC2s and mature dendritic cells enriched in regulatory molecules.Some of these subpopulations might overlap or correspond to different developmental or activation states of the same DC lineage, while others might represent distinct cDC2 subsets.Adding to the complexity, another population, variably termed transitional DCs (tDCs), AXL + DCs, AS DCs or plasmacytoid-like DCs has been identified in humans and mice 17,[20][21][22][23][24][25] .tDCs are proposed to have a lymphoid origin and recent work suggests that they are part of the plasmacytoid DC lineage, although they can differentiate into cells resembling cDC2As 20,25,26 .
One approach to disentangle this complexity is to study cDC ontogeny.The lifespan of cDCs in tissues is short (3-6 days 27 ) such that the cDC tissue network needs to be constantly replenished from bone marrow precursors.The conventional or common DC progenitor (CDP) is the earliest bone marrow cell with DC-restricted potential 1,28 .These CDPs give rise to pre-cDCs, which leave the bone marrow through the blood to seed all tissues and generate terminally differentiated cDC1s and cDC2s 1 .Specification toward the cDC1 or cDC2 lineage starts already at the CDP stage and generates pre-cDC1s and pre-cDC2s 29,30 .The prevailing view is that the latter then diversify by acquiring distinct phenotypic or functional traits in different tissue niches or under different inflammatory conditions 15,31 .In line with this notion, retinoic acid supports the differentiation of Notch2-dependent cDC2s in the intestine and spleen 32,33 ; type 3 innate lymphoid cells (ILC3s) in the spleen promote the differentiation of cDC2As through the production of lymphotoxin 34 .However, it is possible that cDC2 diversity specification might occur at the pre-cDC level in the bone marrow and that signals in tissue are permissive rather than instructive.
In this study, we used a binary definition of cDC2s, splitting them, as proposed 15 , into T-bet + cDC2As and T-bet − cDC2Bs.We showed that cDC2As and cDC2Bs in mice at steady state phenotypically encompass the previously described Notch-dependent and KLF4-dependent cDC subsets.Notably, we found that pre-cDC2s in the bone marrow could already be divided into two subtypes that preferentially gave rise to cDC2As or cDC2Bs.The identification of biased pre-cDC2A and pre-cDC2B populations in mouse and human bone marrow supports the notion that cDC2As and cDC2Bs represent distinct ontogenetic lineages.

Article
https://doi.org/10.1038/s41590-024-01745-9cluster that appeared uncommitted at the level of the gene expression signature (Fig. 2c), as expected 29,30 .Pre-cDC2s were relatively more heterogenous than pre-cDC1s (seven compared to two clusters) (Fig. 2c).Within the late pre-cDC2s clusters, there were two broad groups: clusters 0, 2 and 8 showed increased similarity in gene expression profile to cDC2A; clusters 1 and 7 expressed more genes in common with cDC2B (Fig. 2d).These data suggested that subdivision of cDC2s into cDC2As and cDC2Bs could be recapitulated at the level of their pre-cDC precursors using gene expression profiling.

Pre-cDC2s are biased toward the cDC2A or cDC2B fate
We used Comet, a tool for predicting cell population surface markers from scRNA-seq data 41 , to design a strategy to identify putative pre-cDC subsets using flow cytometry.Comet identified markers previously used to distinguish pre-cDC1s (CD117 and CD24) from pre-cDC2s (Ly6C and CD115, among others) 29,30 (Supplementary Table 2), the accuracy of which we confirmed using in vitro differentiation assays (Extended Data Fig. 4a,b).Comet further identified CD8α as a marker for the putative pre-cDC2As, in addition to marking cDC1s and tDCs    9) of pre-cDC1s and pre-cDC2s on the concatenated UMAP, and violin plots for the scores within the 3 and 6, and 0, 1, 2, 4, 5, 7 and 8, cluster groups.d, Feature plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table 9) of cDC2As and cDC2Bs on the concatenated UMAP, and violin plots for the scores within the 0, 2 and 8, and 1 and 7, cluster groups.Expression levels are shown as a gradient from low (light gray) to high (teal).In c,d, a two-tailed Mann-Whitney U-test was used for comparison (median ± the interquartile range (IQR)).P values are indicated above the graphs.in some tissues (Fig. 3a, Extended Data Fig. 4c-e and Supplementary Table 2).Using flow cytometry, we confirmed that Ly6C + pre-cDC2s encompassed CD8α − and CD8α + cells (Fig. 3b and Extended Data Fig. 4a,c-e).UMAP analysis of Lin − spleen cells stained for multiple cDC and pre-cDC markers positioned CD8α − pre-cDC2s on a branch leading to cDC2B, and CD8α + pre-cDC2s on a distinct one leading to cDC2A (Fig. 3b and Extended Data Fig. 4c,d).We sorted spleen CD8α + pre-cDC2s and CD8α − pre-cDC2s (Extended Data Fig. 4a) and performed bulk RNA-seq analysis (Extended Data Fig. 5a).Differentially expressed genes (DEGs) from either population (Supplementary Table 3) were used as a gene signature, which when overlaid on the earlier scRNA-seq UMAP analysis (Extended Data Fig. 5b), indicated that CD8α was indeed able to segregate putative precursors of cDC2As (CD8α + pre-cDC2s) and cDC2Bs (CD8α − pre-cDC2s) in mouse spleen (Extended Data Fig. 5b).This analysis also indicated that although tDCs express CD8α, their gene expression profile was distinct from that of CD8α + pre-cDC2s (Extended Data Fig. 5a).

Two bone marrow pre-cDC2 subsets are related to cDC2As and cDC2Bs
Next, we investigated whether the lineage bias of pre-cDC2As and pre-cDC2Bs occurred as they entered the tissue or, as for pre-cDC1s and pre-cDC2s, before leaving the bone marrow.Pseudotime analysis of scRNA-seq data from bone marrow pre-cDCs suggested two mutually exclusive cDC2A and cDC2B differentiation trajectories (Fig. 4a).We compared the gene expression profiles of the cell clusters that defined the two trajectories (Fig. 4b).Among the transcripts that segregated clusters 0 and 1 in the bone marrow, we found 87 that overlapped with some of the transcripts that segregated late pre-cDC2As (clusters 2 and 8) and late pre-cDC2Bs (cluster 7) in the periphery, as well as those that segregated cDC2As and cDC2Bs (Fig. 4b and Supplementary Table 4).This overlap was statistically significant Fig. 3 | Peripheral pre-cDC2s are biased toward the cDC2A or cDC2B fate.a, Feature plot (left) and violin plot (right) showing Cd8a expression on the concatenated UMAP or in cluster groups 0, 2 and 8, or 1 and 7, as in Fig. 2b-d.b, Representative UMAP of flow cytometry analysis of splenic pre-cDC and cDC populations generated on CD11c + Lin − cells using CD11c, MHC-II, CD26, CD64, CD88, XCR1, SIRPα, Esam, CLEC12A, CD11b, CD43, CD135, CD117, Ly6C and CD8α (left), and CD8α + cells overlaid onto the UMAP (right).c, CD45.2 + cDC2s (derived from CD8α − or CD8α + pre-cDC2s) recovered from the spleen of CD45.1 recipient mice overlaid onto a UMAP representing the cDC lineage of the host (left) and flow cytometry analysis showing the number and percentage of WT CD45.2 Esam + cDC2As and CLEC12A + cDC2Bs recovered from the spleen of WT CD45.1 recipient mice 3 days after transfer of the CD8α − and CD8α + CD45.2 pre-cDC2s populations (right).Populations are annotated in b. d, ZsGreen MFI (after subtracting the autofluorescence background) in cDC2As and cDC2Bs or CD8α − or CD8α + pre-cDC2s from T-bet-ZsGreen mice and representative flow cytometry plots with overlaid CD8α + pre-cDC2s and CD8α − pre-cDC2s in the spleen, MLN, lung and liver depicting T-bet-ZsGreen expression (fluorescence intensity) in each pre-cDC2 population.e, Percentage of cDC2As and cDC2Bs or CD8α − or CD8α + pre-cDC2s in the spleen, MLN, lung and liver and representative UMAP for the spleen, MLN, lung and liver showing the clusters containing cDC2As and cDC2Bs or CD8α − or CD8α + pre-cDC2s.In c,d, each dot represents one mouse (n = 4 in c and n = 8 in d,e); data were pooled from two experiments (mean ± s.e.m.; median ± IQR for the violin plot).In c-e, quantifications come from the UMAPs (as shown in b and Extended Data Fig. 4c-e).A two-tailed Mann-Whitney U-test was used for comparison.P values are indicated above the graphs.

Fig. 4 | The bone marrow contains two populations of pre-cDC2s that can be segregated according to Siglec-H expression and are related to cDC2As
and cDC2Bs.a, Pseudotime analysis of scRNA-seq data (Fig. 2b-d) from cluster 4 to clusters 7 and 8 concatenated from the bone marrow, spleen and lung.b, Heatmap of 87 DEGs between early pre-cDC2s (clusters 0 and 1) in the bone marrow (left), late pre-cDC2 clusters (clusters 2 and 8, and cluster 7) from the bone marrow, spleen and lung (middle) and comparison of our pre-cDC scRNAseq data to those of splenic cDC2As and cDC2Bs from Brown et al. 15 (right).Expression levels ranged from low (blue) to high (orange).c, Expression of CD8α on pre-cDC2s from the bone marrow, spleen, MLN, lung and liver, gated as in Extended Data Fig. 4c-e Flagelin (ng ml −1 ) R848 (µg ml −1 ) CpG (µg ml −1 ) Zymosan (µg ml −1 )
In contrast to peripheral tissues, we did not detect expression of CD8α in any pre-cDC2s in the bone marrow (Fig. 4c).However, scRNA-seq and quantitative PCR with reverse transcription (RT-qPCR) analysis identified Siglec-H as a potential marker for the putative bone marrow pre-cDC2As in cluster 0 (Fig. 4d,e).Flow cytometry analysis confirmed that bone marrow pre-cDC2s could be segregated into Siglec-H + and Siglec-H − populations 30 (Fig. 4f and Extended Data Fig. 7a-d).Siglec-H expression was very low in pre-cDC2s or cDC2s from peripheral tissues, such as the spleen (Extended Data Fig. 8a), suggesting that Siglec-H expression was lost as early pre-cDCs differentiated into late pre-cDCs that leave the bone marrow, which is consistent with previous reports 30 .Accordingly, scRNA-seq data analysis showed that Siglech expression was higher in cells in cluster 0 and lower in more differentiated pre-cDC2As in clusters 2 and 8 (Extended Data Fig. 8b).We sorted Siglec-H + and Siglec-H − pre-cDC2s from the bone marrow and performed bulk RNA-seq analysis to obtain a DEG signature for both populations (Extended Data Fig. 8c-d and Supplementary Table 5).When mapped onto the scRNA-seq UMAP, the signature of the Siglec-H + pre-cDC2s highlighted cells in clusters 0, 2 and 8, whereas the signature of the Siglec-H − pre-cDC2s highlighted cells in clusters 1 and 7 (Extended Data Fig. 8d).We further used principal component analysis (PCA) to probe the relationship between bone marrow Siglec-H + pre-cDC2s and Siglec-H − pre-cDC2s and the CD8α + pre-cDC2As and CD8α − pre-cDC2Bs found in the spleen.Principal component 1 segregated cells according to tissue, while principal component 2 split the cells according to subset (Extended Data Fig. 8c), indicating similarity between Siglec-H + and CD8α + pre-cDC2s and Siglec-H − and CD8α − pre-cDC2s.
Siglec-H + pre-cDC2s displayed a greater proliferation index than Siglec-H − pre-cDC2s, which was similar to the difference between cDC2As and cDC2Bs (Fig. 4g).cDC2As and Siglec-H + pre-cDC2s responded more strongly to flagellin stimulation, whereas cDC2Bs and Siglec-H − pre-cDC2s were more responsive to R848, CpG and zymosan (Fig. 4h).Bone marrow Siglec-H + pre-cDC2As and Siglec-H − pre-cDC2Bs displayed comparable labeling to bone marrow pre-cDC1s in Clec9a Cre lineage-tracing mice (Extended Data Fig. 6a-b) and were Flt3L-dependent (Extended Data Fig. 6c), suggesting that they all descended from CDPs and not monocytes.These data showed that Siglec-H + pre-cDC2s and Siglec-H − pre-cDC2s in the bone marrow resemble peripheral cDC2As and cDC2Bs, respectively in terms of gene expression, proliferation capacity and pattern of responsiveness to innate immune stimuli 4,14,15 .

Lymphotoxin and Notch ligands sustain pre-cDC2A specification
We next sorted Siglec-H + and Siglec-H − pre-cDC2s from the bone marrow of T-bet-ZsGreen mice for in vitro differentiation assays.Both Siglec-H +   and Siglec-H − pre-cDC2s cultured with Flt3L alone differentiated into cDC2s, as measured by the upregulation of MHC-II and SIRPα (Fig. 5a).However, they did not give rise to T-bet-ZsGreen + cells unless cocultured with OP9-DL4 feeder cells, which provide Notch ligands (Extended Data Fig. 8e), in the presence of recombinant mouse lymphotoxin (Fig. 5a,b).In this setting, Siglec-H + pre-cDC2s, but not Siglec-H − pre-cDC2s, generated T-bet-ZsGreen + cDC2As (Fig. 5a,b).This reiterated the importance of Notch signaling in the cDC2A differentiation pathway and led us to assess its effect on pre-cDC2s.Although C9a tdTomato and C9a tdTomatoΔRBPJ mice had equivalent numbers of Siglec-H + and Siglec-H − pre-cDC2s in the bone marrow and CD8α + and CD8α − pre-cDC2s in the periphery (Extended Data Fig. 8f), bulk RNA-seq analysis showed that bone marrow pre-cDC2s from C9a tdTomatoΔRBPJ mice displayed an altered gene expression profile (Extended Data Fig. 8g and Supplementary Table 6).f, GSEA analysis showing significantly modified pathways in mouse bone marrow pre-cDC2As versus pre-cDC2Bs as in Fig. 2 (1), mouse peripheral pre-cDC2As versus pre-cDC2Bs as in Fig. 2 (2) and mouse splenic cDC2As versus cDC2Bs from Brown et al. 15 (3), with human cDC2A versus cDC2B lineages from the bone marrow (4).In c, a one-way ANOVA (with Tukey correction) was used for comparison (median ± IQR).Comparisons are from one group of clusters relative to all other groups and indicated when not significant.Reference groups are (from left to right): 1, 3, 5, 9; 10; 11; 0, 7, 12; and 2, 4, 6, 8, 13.P values are indicated above the graphs.

Bone marrow specification of cDC2s is conserved across species
We reanalyzed a published dataset that reported cDC2As and cDC2Bs among HLA-DR isotype (HLA-DR) + cells from human spleen 15 .

Discussion
Distinct cell types or different cell states can contribute to the heterogeneity of cDC2s.In this study, we identified pre-cDC2s in mouse bone marrow and peripheral tissues that displayed differential propensity to generate cDC2As versus cDC2Bs and could account for previously described cDC2 types.Much like the separation between cDC1s and cDC2s, the specification of cDC2As and cDC2Bs started in the bone marrow.These data argue for a model in which cDC subsets (cDC1, cDC2A and cDC2B) and related lineages (DC3s, plasmacytoid cells, tDCs) are prespecified in the bone marrow and constitute bona fide DC subsets rather than tissue-determined cell states.We could not ascertain whether pre-cDC2As and pre-cDC2Bs are unipotential as we noted residual capacity of bone marrow Siglec-H + or spleen CD8α + pre-cDC2 to generate cDC2Bs.This might reflect plasticity but could equally represent technical limitations in cell sorting or in the penetrance of Cre-mediated recombination in lineage tracing.In addition, some of the output cells in our lineage-tracing experiments, and in vivo transfer and in vitro differentiation assays, did not express markers that allowed us to assign them to either the cDC2A or cDC2B lineages.Clonal analysis, as well as more extensive phenotyping, will be important in the future to distinguish precursor bias from absolute commitment.Siglec-H + and Siglec-H − pre-cDC2s are proposed to represent distinct developmental stages of cDC2s 30 .We further found a population of bone marrow pre-cDC2s that never expressed Siglec-H and generates cDC2Bs.We also showed that Siglec-H + pre-cDC2As lost the expression of Siglec-H as they left the bone marrow, concomitant with the acquisition of CD8α expression and before final differentiation into cDC2As in tissues.This is consistent with a previous report that Siglec-H + pre-cDC2s can give rise to cDC2s 15,17,30 but argues that it is the case only for cDC2As and not cDC2Bs.
Specific organ niches can drive adult monocytes to become resident macrophages akin to those that colonized the organs during embryonic life 44 .In this setting, tissue signals override ontogeny to specify myeloid cell fate.However, unlike tissue macrophages that can live up to 18 months in mice and 11 years in humans 45 , the lifespan of cDCs in mouse tissues is estimated to be 3-6 days in most organs 27,46 .This might explain why cDC2 subsets are prespecified in the bone marrow, as they may not have enough time to be 'instructed' by their niche.

Article
https://doi.org/10.1038/s41590-024-01745-9 However, this does not negate the importance of the tissue microenvironment 15,31,34,47 as we showed that pre-cDC2s required a permissive setting to complete their differentiation.Different environmental cues in lymphoid versus nonlymphoid organs could modulate the proliferation and lifespan of pre-cDC2 types or their progeny, explaining the contrasting proportion of cDC2As and cDC2Bs in these organs.In line with this notion, Esam + cDC2As proliferate more than Esam − cDC2s in response to lymphotoxin expressed by splenic ILC3s 4,48,49 .Differential expression of chemokine receptors in pre-cDC2As versus pre-cDC2Bs (for example, Ccr1, Ccr2 and Ccr9, as noted in our scRNA-seq analysis) could additionally affect the tropism of pre-cDC subsets toward different organs.
We focused on ontogeny and gene expression as the primary tool for cDC definition, as done by others 17,25 .It has been suggested that progenitors that express Siglec-H + and share other markers with plasmacytoid cells (most likely corresponding to the pre-cDC2As described in this study) act as cDC2 precursors 17 .tDCs can generate Esam + cells that show phenotypic overlap with, yet are distinct from, cDC2As 25 .Our data suggest that pre-cDC2As display phenotypic similarities to tDCs, but arise from Ly6D − precursors, display distinct gene expression signatures from tDCs, can be distinguished by higher expression of SIRPα, MHC-II, CLEC12A and CD43 and lower expression of CD24, and display lower labeling than tDCs in SigH RFP mice.As such, our data are consistent with the notion that tDCs and pre-cDC2As represent distinct populations, although we note that both can give rise to Esam + DCs (this work and Sulczewski et al. 25 ).Based on the expression of CD11b and CD24, tDC-derived Esam + DCs may not be canonical cDC2As, although expression of T-bet remains to be assessed.Finer delineation of the cDC2A and the tDC lineages will require a genetic approach, such as hCD2 or CD300c lineage-tracing mice.
DC3s have recently been shown to be distinct from cDCs and monocytes and arise from Ly6C + monocyte-DC progenitors that do not go through a pre-cDC stage 39 .Similarly, tDCs originate from Ly6D + bone marrow progenitors shared with plasmacytoid DCs 25 .The discovery of ontogenetically distinct DC3s, tDCs, together with our observations, supports a model in which the bone marrow is the original site of DC precursor bias toward the cDC1, cDC2A, cDC2B, DC3 and tDC fate.Additional studies will be necessary to establish the degree of plasticity in pre-cDC commitment during inflammation and assess the functional properties of progeny cDC2As, cDC2Bs, DC3s and tDCs.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41590-024-01745-9.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.© The Author(s) 2024 https://doi.org/10.1038/s41590-024-01745-9

Ethics
The research in this manuscript complies with all relevant ethical regulations.Mouse experiments were planned in accordance with the principles of the three Rs (replacement, reduction, refinement).All experiments were performed in accordance with the United Kingdom Animals (Scientific Procedures) Act of 1986.The UK Home Office accredited all researchers for animal handling and experimentation.Dispensation to carry out animal research at the Francis Crick Institute was approved by the institutional ethical review body and granted by the UK Home Office under PPL PF40C0C67.

Human bone marrow
Human bone marrow was purchased from STEM CELL Technologies and processed as described previously 52 .Briefly, cells from three independent donors (female aged 31 years, and males aged 29 and 24 years) were thawed in prewarmed FCS containing DNase I (10 μg ml −1 ), washed and stained for fluorescence-activated cell sorting (FACS) as described below (the antibodies used for staining are listed in Supplementary Table 10).After sorting, human pre-DCs and DCs from the three individuals were pooled to minimize individual variability before submission for scRNA-seq.

Preparation of single-cell suspensions
Mice were perfused intracardially through the left ventricle using cold PBS before tissue collection.Livers were further perfused in situ via the portal vein.This procedure efficiently removed circulating cells as assessed by injection of CD45 antibody (intravenously) 2 min before tissue collection and processing 40 .Spleens, MLNs, lungs and livers were cut into small pieces and digested with collagenase VIII (1 mg ml −1 , Sigma-Aldrich) and DNase I (0.4 mg ml −1 , Roche) in Roswell Park Memorial Institute (RPMI) 1640 medium for 15 min (spleen and MLN) or 25 min (lung and liver) at 37 °C.Digested tissues were passed through a 70-μm cell strainer (BD Biosciences) and washed with FACS buffer (3% FCS and 5 mM EDTA in PBS).For lung and liver, leukocytes were enriched using Percoll gradient centrifugation (GE Healthcare) as described previously 18 .For bone marrow, the femur, tibia and hip extremities were cut and spun for 30 s at 10,000 r.p.m.Cells were resuspended in FACS buffer after centrifugation.For the transfer assays, the spine and humerus were also collected and crushed with a mortar before collecting a cell suspension with a micropipette and filtering using a 100-μm cell strainer.

Pre-cDC enrichment and isolation
Single-cell suspensions from the bone marrow, spleen and lung were enriched for pre-cDCs by staining for lineage-restricted markers with biotin-conjugated or fluorescein isothiocyanate (FITC)-conjugated antibodies (CD3, Ly6G, Siglec-F, B220, CD19, Ly6D, NK1.1 and Ter119) and depleting T, B and plasmacytoid cells, as well as red blood cells, neutrophils, eosinophils and their precursors, using the EasySep Mouse Biotin Positive Selection Kit II (STEMCELL Technologies).Cells were stained as described below.Pre-cDC and cDC subsets were FACS-sorted on an Aria Fusion (BD Biosciences) with a 100-μm nozzle using the gating strategy shown in Extended Data Figs.1b, 3a, 4a and 7a as indicated.

Flow cytometry analysis
Cells were preincubated with blocking anti-CD16/32 in FACS buffer for 10 min at 4 °C and then stained for 40 min at 4 °C with an antibody cocktail and LIVE/DEAD Fixable Dead Cell Stain Kit (Thermo Fisher Scientific) in FACS buffer.Lineage (Lin) markers included CD3, Ly6G, Siglec-F, B220, CD19, Ly6D, NK1.1 and Ter119, unless otherwise specified.The antibodies used for flow cytometry are listed in Supplementary Table 10.Samples were acquired using a BD FACSymphony A5 (BD Biosciences) or in an ID7000 (Sony Biotechnology) or SpectroFlo Aurora (Cytek) spectral analyzers.Data were analyzed using FlowJo (v.10.8.2) as shown in Extended Data Figs. 1, 4 and 8. UMAP analysis 53 of the flow cytometry data was carried out on the basis of CD11b, CD11c, CD26, CD43, CD64, CD88, CD135, SIRPα, MHC-II, CD117, Ly6C, Siglec-H, CD8α, XCR1, CLEC12A and Esam expression.Annotation of clusters on the UMAP plots was done by using defining markers for each immune population.Validation of the accuracy of the UMAP analysis versus manual gating was confirmed by overlaying different immune populations identified by either strategy.Monocytes and MDCs were identified as in Cabeza-Cabrerizo et al. 18 .Earlier bone marrow progenitors were identified as in Cardoso et al. 54 .

scRNA-seq
Mouse and human pre-cDCs (viability >95%) were processed according to the manufacturer's instructions on a 10X Genomics Chromium platform.Library generation was performed using the Chromium Single Cell 3′ Reagents Kits (10X Genomics) and sequenced on an HiSeq 4000 (Illumina) to achieve an average of approximately 63,000 reads per cell and approximately 4,000 cells per sample.Raw reads were initially processed using the Cell Ranger v.3.0.2 pipeline 55 , which deconvolved reads to their cell of origin using the unique molecular identifier tags, aligned these to the mm10 transcriptome (to which we added the eGFP sequence (https://www.addgene.org/browse/sequence/305137/)to detect GFP-expressing cells) using STAR (v.2.5.1b) 56 and reported cell-specific gene expression count estimates.All subsequent analyses were performed in R v.3.6.1 using the Seurat (v.3) package 57 .Genes were considered to be 'expressed' if the estimated (log 10 ) count was at least 0.1.Primary filtering was then performed by removing from consideration cells expressing fewer than 50 genes and cells for which mitochondrial genes made up greater than three standard deviations from the mean of mitochondria-expressed genes.PCA decomposition was performed and, after consideration of the eigenvalue 'elbow-plots', the first 30 components were used to construct the UMAP plots per sample.Multiple samples were integrated using 2,000 variable genes and Seurat's canonical correlation analysis.Cluster-specific gene markers were identified using a Wilcoxon rank-sum test; the top 10 or 20 genes ranked according to log fold change per cluster were used to generate a heatmap.Clusters were annotated using known marker genes and gene signatures (refs.15,30 and Supplementary Table 9).Contamination with plasmacytoid cells and MDCs was ruled out by integrating our data with previous scRNA-seq analysis that included these cells 22,58 and checking for cluster segregation.GSEA was used to identify pathways enriched in a cluster or a group of them against others.CytoTRACE 59 was used to determine the differentiation states of cells.Trajectories were identified using the package Slingshot (v.1.4.0) 60 , using the undifferentiated cluster as a starting point and the dimensionality reduction UMAP coordinates.Lineages were identified showing different trajectories ending in specific differentiated cells (Supplementary Table 11).Comet analysis 41 was used to identify putative flow cytometry markers for populations defined using scRNA-seq.The analysis was performed by loading the scRNA-seq data, the UMAP and the clustering from Seurat on the Comet portal 41 ).

Bulk RNA-seq
Pre-cDCs and cDCs (gating strategy shown in Extended Data Figs.1b, 4a and 7a) were FACS-sorted from the bone marrow and spleen either from https://doi.org/10.1038/s41590-024-01745-9 WT or C9a tdTomato and C9a tdTomatoΔRBPJ mice.Cells (0.6 × 10 4 to 3.2 × 10 4 ) were sorted directly into lysis buffer to avoid loss of material.RNA was extracted using the RNeasy Mini Kit (QIAGEN).The NuGEN Ovation RNA-Seq System (V2) was used for complementary DNA (cDNA) synthesis followed by the NuGEN Ultralow Library System (V2) for library preparation.Samples were normalized to 1 ng of RNA for input; the preparation was performed according to the manufacturer's guidelines.Sequencing was performed on an Illumina HiSeq 4000, with 100-base pair single-end reads.After sequencing, samples were normalized and analyzed.GSEA was used to identify pathways enriched in cells from different genotypes.

RNA extraction and RT-qPCR
Cells were collected in RLT buffer and RNA extraction was performed using the RNeasy Micro Kit (QIAGEN).cDNA synthesis was carried out using SuperScript II Reverse Transcriptase (Invitrogen).RT-qPCR was performed using the TaqMan Universal PCR Master Mix (Thermo Fisher Scientific) and primers (Supplementary Table 12).Analysis was performed on a QuantStudio PCR system (Thermo Fisher Scientific) using Δ Ct quantification.

Proliferation assessment
Mice were injected intraperitoneally with 1 mg EdU (Lumiprobe) 2 h before tissue collection for assessment of cell proliferation.EdU detection was carried out using the Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry Kit (Thermo Fisher Scientific) after surface staining and fixation and permeabilization.Intranuclear staining of Ki-67 was performed in parallel to EdU detection.Cells were analyzed using flow cytometry as described above.

Statistical analysis and reproducibility
No statistical methods were used to predetermine sample sizes but our sample sizes were similar to those reported in previous publications 18 .Mice were not randomized in cages, but each cage was randomly assigned to a treatment group.Investigators were not blinded to mouse identity during necropsy and sample analysis.Male and female mice were used to perform the experiments.However, we did not observe differences between sexes.In all cases measurements were taken from distinct samples and no individual data points were excluded under any circumstances.Statistical analyses were performed using Prism 9 (GraphPad Software).Results are depicted as the mean ± s.e.m. and median ± IQR in violin plots.The statistical test used is specified in each figure legend.For pair comparisons, a nonparametric two-tailed Mann-Whitney U-test was used.When ANOVA was used, Tukey correction was performed.Data distribution was assumed to be normal, but this was not formally tested.For Tables 1, 3 and 5, a two-sided Wald test with Benjamini-Hochberg correction was used.For Supplementary Tables 2, 4, 6 (DEGs) and 8, a one-sided Wilcoxon signed-rank test with Benjamini-Hochberg correction was used.For Supplementary Table 6 (enrichment), a one-sided hypergeometric test with Benjamini-Hochberg correction was used.For Supplementary Table 11, a one-sided Wald test not corrected for multiple testing was used.These comparisons were made using the DESeq2.Genes with P adj < 0.05 were taken forward and used to draw a heatmap using the ComplexHeatmap R package or to generate a gene signature.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
After excluding CD8α + tDCs (dark blue), cDC2As (teal) and cDC2Bs (orange) are identified using ESAM and CLEC12A, respectively.Arrows denote gate hierarchy.c, Manual gates from B are overlaid onto a UMAP (same as 1b) of the Lin -CD11c + cells (from the first gate of the manual strategy in a).The UMAP was generated on the basis of CD11c, MHC-II, CD26, CD64, CD88, XCR1, SIRPα, ESAM, CLEC12A, CD11b, CD43, CD135, CD117, Ly6C, and CD8α.cDC2s in the leftmost UMAP are gated and zoomed in the following panels, where cDC2 subsets gated manually are overlaid.See also Fig. 1b.d, The expression of key markers used to define different cDC and tDC subpopulations in the UMAPs is shown in the form of heatmaps.Expression levels are represented as a colour gradient from low (blue) to high (orange).
Extended Data Fig. 2 | Validation of spleen cDC2 gating strategy.a, (Left) Heatmap representation of the top differentially expressed genes (an adjusted p value of < 0.05) from a new bulk RNAseq analysis of the two cDC2 populations (ESAM + cDC2s and CLEC12A + cDC2s) sorted using the gating strategy shown in Extended Data Fig. 1b (PCA is shown later in Extended Data Fig. 5a).Expression levels are represented as a colour gradient from low (blue) to high (orange).Each column represents a sample coming from a pool of 5 mice.Note that the expression of Esam, Clec12a and Tbx21 was either not detected or not significant in the statistical analysis.(Right) Feature plots representing the score of DEGs from a (used as signatures) of ESAM + cDC2s and CLEC12A + cDC2s overlaid onto a UMAP of cDC2As and cDC2Bs generated from the Brown et al scRNAseq dataset 15 .
Expression levels are shown as a gradient from low (light grey) to high (teal).The quantification of the scores is shown on the bottom in the form of violin plots.b, (from left to right and top to bottom) FACS analysis showing CD43, MHC-II, CD8α, CD117, ESAM, CD11b, CLEC12A, CD24, MGL-2 and PD-L2 expression on spleen cDC2 and tDC populations (identified from UMAP gates as shown in 1b and Extended Data Fig. 1c, d).c, FACS analysis showing the percentage of different populations (identified as shown in 1b and Extended Data Fig. 1c, d) in the indicated tissues.Data in (c-d) are a pool of two experiments (n = 8) (means ± SEM, median ± IQR for violin plot).Each dot in b represents one mouse (n = 8).Mann-Whitney test (two-tailed) was used to compare cDC2As and cDC2Bs in A. P values are indicated on top of the graphs.
Extended Data Fig. 3 | Validation of gating strategy for sorting total pre-cDC populations from tissues.a, Sorting strategy for pre-cDCs (and other precursor cells to ascertain which ones are bona fide pre-cDCs).Live single cells from spleen or bone marrow cell suspensions negative for lineage markers (CD3, Ly6G, SiglecF, B220, CD19, NK1.1, and Ter119) and positive for CD45.2 were analysed as follows: CD11c + MHC-II −/lo were selected, from this gate, the CD135 + CD43 + cells contained the pre-cDCs and other contaminants.CD135 + CD43 + cells contained two populations: Ly6D + and Ly6D − cells.The Ly6D + cells were directly sorted as one population (grey gate).The Ly6D -cells were further split into three subpopulations that were sorted as shown on the fourth panel: CD11b -(light blue gate), CD11b lo (dark blue gate) and CD11b hi (orange gate).Arrows denote gate hierarchy.b, The populations highlighted in panels 3 and 4 were sorted from the bone marrow (top) or spleen (bottom) and cultured for 3 days with OP9-DL1 stromal cells in the presence of Flt3L.Data are FACS analysis showing the % recovery after differentiation and frequency of cDC subsets and plasmacytoid cells among the progeny.These populations were analysed using manual gating and were defined as: single, live, CD45.2 + , CD11c + MHC-II + cells.cDC1s are defined as XCR1 + while cDC2s express SIRPα.The right panel shows the cDC1/cDC2 subset distribution of progeny from the sorted cells after differentiation.c, FACS analysis showing TdTomato labelling of the indicated cell populations from the bone marrow or spleen of C9a TdTOM mice gated as shown in a. d, FACS analysis showing the abundance of the indicated cell populations (gated as shown in a) in the bone marrow and spleens of WT and Flt3L-deficient mice.e, Refined gating strategy used to sort total pre-cDCs from tissues taking into account the results from a-d.In this sorting strategy, pre-cDCs are identified as leukocytes that are negative for many lineage-restricted markers (CD3, Ly6G, SiglecF, B220, CD19, Ly6D, NK1.1, and Ter119), as well as negative/low for surface expression of MHC-II, CD11b and SIRPα, but positive for CD11c, CD135, and CD43.Each dot represents one mouse (n = 3 in b and d and 8 in c).Data are from one out of two experiments (b, d) or a pool of two (c) (means ± SEM).Mann-Whitney test (two-tailed) was used to compare WT and Flt3l −/− mice in (d).P values are indicated on top of the graphs.
Extended Data Fig. 4 | Pre-cDC subset identification in the spleen, MLN, lung and liver.a, Sorting strategy for spleen pre-cDC subsets.Leftmost panel has been pre-gated on single, live, CD45 + , and lineage − spleen cells.The lineage cocktail includes antibodies against CD3, Ly6G, SiglecF, B220, CD19, NK1.1, Ly6D, and Ter119.CD117 and Ly6C are used to identify pre-cDC1s (dark grey) and pre-cDC2s, respectively.CD8α labels the putative pre-cDC2As (light green) whereas the putative pre-cDC2Bs are CD8α − (yellow).Arrows denote gate hierarchy.b, (left) Violin plots showing the expression of Kit and Ly6c2 in pre-cDC1s (clusters 3 and 6) or pre-cDC2s (clusters 0, 1, 2, 4, 5, 7, 8) from scRNAseq analysis (UMAP of data concatenated from all tissues).(right) Total pre-cDCs or the indicated subsets were sorted from spleen (sorting strategy as in Extended Data Fig. 4a) and cultured for 3 days with OP9-DL1 stromal cells in the presence of Flt3L.The progeny after differentiation was analysed by FACS for cDC subset distribution.Cells were analysed using manual gating and defined as: single, live, CD45.2 + , CD11c + MHC-II + .cDC1s are defined as XCR1 + , whereas cDC2s express SIRPα.c, Manual gates as in Extended Data Fig. 4a for pre-cDCs and as in Extended Data Fig. 1b for cDC were overlaid onto a UMAP analysis of the spleen (same as 3b).Colours for pre-cDCs correspond to the gates in a.The UMAP was generated using the Lin -CD11c + cells from the first gate of the manual strategy in a, and using the following markers: CD11c, MHC-II, CD26, CD64, CD88, XCR1, SIRPα, ESAM, CLEC12A, CD11b, CD43, CD135, CD117, Ly6C, and CD8α.d, The expression of key markers used to define different pre-cDC subpopulations in the UMAPs (from spleen in 3b) is shown in the form of heatmaps.Expression levels are represented as a colour gradient from low (blue) to high (orange).e, Representative UMAP analysis from the spleen, MLN, lung and liver.UMAP was generated as in b.Ungated cells are migratory cDC1s and cDC2s, and probably DC3s and were not analysed in this study.In b (right) data are a pool of two experiments (n = 4) (means ± SEM and median ± IQR for violin plot).Mann-Whitney test (two-tailed) was used for comparisons.P values are indicated on top of the graphs.
Extended Data Fig. 5 | Validation of the strategy to identify splenic pre-cDC2 subsets.a, PCA of all expressed genes from a bulk RNAseq (same as Extended Data Fig. 2a) of the indicated populations sorted from spleen as shown in Extended Data Figs.1b and 4a.b, (left) Heatmap representation of the top DEGs (an adjusted p value of <0.05) defining CD8α − pre-cDC2 and CD8α + pre-cDC2 analysed by bulk RNAseq (same analysis as a).Expression levels are represented as a colour gradient from low (blue) to high (orange).Each column represents a sample coming from a pool of 5 mice.(right) Feature plots representing the score of the CD8α − and CD8α + pre-cDC2 signatures (signatures are the list of DEGs from the heatmap on the left) projected on the concatenated UMAP.Expression levels are shown as a gradient from low (light grey) to high (teal).The quantification of the scores is shown on top of the plots.c, FACS analysis showing (left) recovery (number of cells), (middle) differentiation (upregulation of MHC-II) and (right) cDC2 specification (upregulation of SIRPα) of WT CD45.2 cells recovered from spleens of WT CD45.1 recipient mice 3 days after transfer of the indicated CD45.2 pre-cDC2s populations (1-4x10 4 cells sorted as shown in Extended Data Fig. 4a).d, qRT-PCR analysis showing expression of Cd8a (top left) and Tbx21 (bottom) in the indicated spleen cell populations (FACS-sorted as shown in Extended Data Figs.1b and 4a).(top right) Flow cytometric quantification of CD8α expression in the indicated populations (gated as in Extended Data Fig. 4c-e).e, FACS analysis of CD45.2 cells recovered from spleens of CD45.1 mice 3 days after receiving the indicated CD45.2 pre-cDC2s populations from T-bet-ZsGreen mice (1-4 × 10 4 cells sorted as shown on top -negative gate was set using a WT counterpart).Data are: (top left) recovery (number of cells), (top middle) differentiation (upregulation of MHC-II), (top right) cDC2 specification (upregulation of SIRPα), (bottom left) % ZsGreen + , (bottom middle) % ESAM + and (bottom right) % CLEC12A + cells.Each dot represents one mouse, and data are a pool of two experiments (n = 4 in c and e and 6 in d) (means ± SEM, median ± IQR for violin plot).Mann-Whitney test (two-tailed) was used for comparisons.P values are indicated on top of the graphs.In d CD8α + pre-cDC2 were compared against CD8α − pre-cDC2, and cDC2A (and early cDC2A) against cDC2B.
Extended Data Fig. 6 | cDC2As and cDC2Bs are bona fide cDC subsets.a, (left) schematic depicting strategy for labelling of cDC lineages in DNGR-1 lineage tracer mice (C9a tdTOM ). Figure was generated with BioRender.(right) FACS analysis showing % Tomato + bone marrow progenitors identified as in reference 54 .b, FACS analysis showing % Tomato + cells in the indicated cDC and pre-cDC subtypes and MDCs as reference for a poorly-labelled lineage.c, FACS analysis showing relative number of the indicated cDC and pre-cDC subtypes in WT versus Flt3L-deficient mice.Number of monocytes and MDCs from different tissues is also shown as reference for a Flt3L-independent lineage.Tissues analysed are indicated at the left of the graphs.Each dot represents one mouse (n = 8), and data were pooled from two experiments, in c data are expressed as fold-difference from WT (means ± SEM).Gating and quantifications come from UMAPs as shown in Extended Data Fig. 7b-d (see later) for the bone marrow and Extended Data Fig. 4c-e for the spleen, MLN, lung and liver.Monocytes and MDCs were identified as in ref. 18.Each dot represents one biological replicate (n = 8), and data are a pool of two experiments (means ± SEM).For panels (a, c) one-way ANOVA (with Tukey correction) was used for comparison of the groups against the labelling of MDPs or against the WT control.P values are indicated on top of the graphs.
Extended Data Fig. 7 | Pre-cDC subset identification in the bone marrow.a, Sorting strategy for bone marrow pre-cDC subsets.Leftmost panel has been pre-gated on single, live, CD45 + , and lineage − spleen cells.The lineage cocktail includes antibodies against CD3, Ly6G, SiglecF, B220, CD19, NK1.1, Ly6D, and Ter119.CD117 and Ly6C are used to identify pre-cDC1s (dark grey) and pre-cDC2s, respectively.SiglecH labels the putative pre-cDC2As (light green) whereas the putative pre-cDC2Bs are SiglecH − (yellow).Arrows denote gate hierarchy.b, Manual gates used in a overlaid onto a UMAP analysis.The UMAP was generated using the Lin − CD11c + cells from the first gate of the manual strategy in a and used the following markers: CD11c, MHC-II, CD26, CD64, CD88, XCR1, SIRPα, ESAM, CLEC12A, CD11b, CD43, CD135, CD117, Ly6C, and SiglecH.c, The expression of key markers used to define different pre-cDC subpopulations in the UMAPs is shown in the form of heatmaps.Expression levels are represented as a colour gradient from low (blue) to high (orange).d, Analysis strategy for pre-cDC subsets in the bone marrow.The plot has been zoomed in the population of pre-cDCs shown in the second panel of b (highlighted in blue).Dark grey gate are pre-cDC1s, green gate are SiglecH + pre-cDC2s and yellow gate are SiglecH − pre-cDC2s.
Extended Data Fig. 8 | cDC2A differentiation trajectory post bone marrow egress.a, FACS analysis of SiglecH expression by the indicated pre-cDC2 or cDC2 populations isolated from the tissues indicated on top of the graphs.Gating is shown in Extended Data Fig. 7b-d for the bone marrow and Extended Data Fig. 4c-e for peripheral organs.b, Violin plot depicting the expression of Siglech in the clusters from the concatenated UMAP of the scRNAseq analysis (see Fig. 2a).c, PCA of all expressed genes from a bulk RNAseq of the indicated pre-DC2 populations from spleen (Sorted as shown in Extended Data Fig. 4a) and bone marrow (sorted as shown in Extended Data Fig. 7a).d, (left) Heatmap representation of the top DEGs (an adjusted p value of < 0.05) defining SiglecH − pre-cDC2 and SiglecH + pre-cDC2 analysed by bulk RNAseq (same analysis as c).Expression levels are represented as a colour gradient from low (blue) to high (orange).Each column represents a sample coming from a pool of 8 mice.(right) Feature plots representing the score of the DEGs (shown in the heatmap, used as signatures) of SiglecH − and SiglecH + pre-cDC2 on the concatenated UMAP.
Expression levels are shown as a gradient from low (light grey) to high (teal).On the right is a violin plot depicting the expression of the DEG-derived signatures by the indicated clusters.e, FACS analysis of transduced OP9 cells showing overexpression of DL4.Sorted DL4 hi cells (bottom right panel) were used as feeder cells for Fig. 5a, b. f, FACS analysis showing the number of cells in the indicated pre-cDC2 populations from C9a tdTOM (dark grey) or C9a TdTOMΔRBPJ (light grey) mice.Gating is shown in Extended Data Fig. 7b-d for the bone marrow and Extended Data Fig. 4c-e for peripheral organs.g, PCA of all expressed genes from a new bulk RNAseq of pre-DC2 populations (same as 5c) sorted (as shown in Extended Data Fig. 7a) from the bone marrow of C9a TdTOM versus C9a TdTOMΔRBPJ mice.Each dot represents a sample coming from a pool of 3 mice.In panel a and f, each dot represents one mouse (n = 7 in a 9 in f), and data were a pool from two experiments (means ± SEM, median ± IQR for violin plot).Two-way ANOVA (with Tukey correction, a,b and f) or Mann-Whitney test (two-tailed, d) was used to compare the different groups.P values are indicated on top of the graphs.
Extended Data Fig. 9 | Model for cDC2A and cDC2B ontogeny.a, qRT-PCR analysis showing the expression of Lyz2 in cDC, tDC and pre-cDC populations from the spleen (sorted as shown in Extended Data Figs.1b and 4a).Data are normalised to housekeeping gene Hprt.b, FACS analysis showing the percentage of RFP + in splenic plasmacytoid cells (defined as CD45.2 + , Lin + , CD11c + , MHC-II + , SiglecH + , CD26 + CD64 − cells) from SiglecH lineage tracing (SigH RFP ) mice crossed to Lyz2 eGFP reporter mice.c, FACS analysis showing the percentage of RFP + (top) or eGFP + (bottom) among early cDC2As or tDCs across the indicated organs.Gating is shown in 1b and Extended Data Fig. 1c-d.Dotted line is the reference value for RFP + pre-cDC2A (top) or eGFP + cDC2B (bottom) percentage in each tissue.Each dot represents one mouse (n = 5 in b and c and 6 in a), and data from one of two experiments (b-c) or pooled from two experiments (a) (means ± SEM).One-way ANOVA (with Tukey correction) was used to compare: in a, CD8α − pre-cDC2 were compared against CD8α + pre-cDC2 and cDC2B against cDC2A and in c, the tDCs and the early cDC2As (separately) with the pre-cDC2As (top) or the cDC2Bs (bottom).P values are indicated on top of the graphs.d, Schematic representation of a model for cDC2A and cDC2B ontogeny: In cDC2A differentiation, SiglecH-positive pre-cDC2As downregulate SiglecH as they leave the bone marrow and acquire the expression of CD8α as they colonise the tissues.Subsequent differentiation of these pre-cDC2As into tissue cDC2As involves downregulation of CD8α and upregulation of CD117 and MHC-II.T-bet expression is progressively upregulated throughout the entire cDC2A differentiation trajectory.cDC2A development is RBP-Jκ-dependent.In cDC2B differentiation, the bone marrow generates pre-cDC2Bs that express LYSM but lack SiglecH and CD8α.This population differentiates into cDC2Bs marked by increased LYSM tracing and upregulation of MHC-II and CLEC12A.cDC2B development is KLF4-dependent.The question marks denote the gaps that remail to be addresses in our model: Clonal analysis, as well as the use of better or additional markers will be necessary to assess the level of plasticity within bone marrow cDC2 progenitors (top question mark).Similarly, the split between the cDC2A and the tDC lineage remains to be confirmed by a genetic approach (bottom question mark).Figure was generated with BioRender.

Fig. 2 |
Fig. 2 | cDC heterogeneity can be recapitulated at the pre-cDC level.a, UMAPs displaying scRNA-seq analysis of pre-cDCs sorted as shown in Extended Data Fig. 3e from the bone marrow (2,649 cells), spleen (4,371 cells) and lung (358 cells) with unsupervised clustering (each sample is a pool of six mice).The proportion of the nine clusters identified in the UMAPs for each organ is shown on the right.b, Representative plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table9) of proliferation (middle) and early (left) or late (right) pre-cDC projected onto the concatenated UMAP space.Expression levels are shown as a gradient from low (light gray) to high (teal).c, Feature plots

9
Fig. 2 | cDC heterogeneity can be recapitulated at the pre-cDC level.a, UMAPs displaying scRNA-seq analysis of pre-cDCs sorted as shown in Extended Data Fig. 3e from the bone marrow (2,649 cells), spleen (4,371 cells) and lung (358 cells) with unsupervised clustering (each sample is a pool of six mice).The proportion of the nine clusters identified in the UMAPs for each organ is shown on the right.b, Representative plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table9) of proliferation (middle) and early (left) or late (right) pre-cDC projected onto the concatenated UMAP space.Expression levels are shown as a gradient from low (light gray) to high (teal).c, Feature plots Fig. 2 | cDC heterogeneity can be recapitulated at the pre-cDC level.a, UMAPs displaying scRNA-seq analysis of pre-cDCs sorted as shown in Extended Data Fig. 3e from the bone marrow (2,649 cells), spleen (4,371 cells) and lung (358 cells) with unsupervised clustering (each sample is a pool of six mice).The proportion of the nine clusters identified in the UMAPs for each organ is shown on the right.b, Representative plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table9) of proliferation (middle) and early (left) or late (right) pre-cDC projected onto the concatenated UMAP space.Expression levels are shown as a gradient from low (light gray) to high (teal).c, Feature plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table9) of pre-cDC1s and pre-cDC2s on the concatenated UMAP, and violin plots for the scores within the 3 and 6, and 0, 1, 2, 4, 5, 7 and 8, cluster groups.d, Feature plots depicting the score for the gene signatures (refs.15,30 and Supplementary Table9) of cDC2As and cDC2Bs on the concatenated UMAP, and violin plots for the scores within the 0, 2 and 8, and 1 and 7, cluster groups.Expression levels are shown as a gradient from low (light gray) to high (teal).In c,d, a two-tailed Mann-Whitney U-test was used for comparison (median ± the interquartile range (IQR)).P values are indicated above the graphs.

9 CD45
Articlehttps://doi.org/10.1038/s41590-024-01745-/doi.org/10.1038/s41590-024-01745-9 Fig.4| The bone marrow contains two populations of pre-cDC2s that can be segregated according to Siglec-H expression and are related to cDC2As and cDC2Bs.a, Pseudotime analysis of scRNA-seq data (Fig.2b-d) from cluster 4 to clusters 7 and 8 concatenated from the bone marrow, spleen and lung.b, Heatmap of 87 DEGs between early pre-cDC2s (clusters 0 and 1) in the bone marrow (left), late pre-cDC2 clusters (clusters 2 and 8, and cluster 7) from the bone marrow, spleen and lung (middle) and comparison of our pre-cDC scRNAseq data to those of splenic cDC2As and cDC2Bs from Brown et al.15 (right).Expression levels ranged from low (blue) to high (orange).c, Expression of CD8α on pre-cDC2s from the bone marrow, spleen, MLN, lung and liver, gated as in Extended Data Fig.4c-e.d, Siglech expression projected on the scRNA-seq UMAP of bone marrow pre-cDCs as in Fig.2a(left) and expression of Siglech in cluster 0 or 1 from bone marrow pre-cDCs (right).e, RT-qPCR for Siglech normalized to Hprt in spleen cDCs sorted as in Extended Data Fig.1band bone marrow pre-cDCs sorted as in Extended Data Fig.7a.f, Representative flow cytometry plot showing Siglec-H and CD26 on pre-cDC2s from the bone marrow gated as single live Lin − CD11c + MHC-II −/lo CD11b −/lo SIRPα − CD135 + CD43 + Ly6C + cells as in Extended Data Fig.4b-d.g, 5-Ethynyl-2′-deoxyuridine (EdU) incorporation and Ki-67 staining on CD8α − or CD8α + (or Siglec-H − or Siglec-H + in the bone marrow) pre-cDC2s identified from the UMAP gates as in Extended Data Figs.4e and 7d(top) and cDC2As and cDC2Bs identified from the UMAP gates as in Extended Data Fig.4efrom bone marrow, spleen, MLN, lung and liver.h, OX40L MFI and Il12b mRNA normalized to Hprt (RT-qPCR) in splenic cDC2As and cDC2Bs sorted as in Extended Data Fig.1band bone marrow Siglec-H − or Siglec-H + pre-cDC2s sorted as in Extended Data Fig.8aafter overnight culture with flagellin, R848, CpG or zymosan.In c,e,g, each dot represents one mouse (n = 3 in h, n = 6 in e, n = 7 in g, n = 8 in c).Data are from one of two experiments (h) or a pool of two (c,e,g) (mean ± s.e.m.; median ± IQR for the violin plot).A two-tailed Mann-Whitney U-test (d,g) or two-way analysis of variance (ANOVA) (with Tukey correction, e,h) was used to compare groups (in e, the comparison is relative to Siglec-H lo pre-cDC2s).P values are indicated above the graphs.

Fig. 5 |Fig. 6 |
Fig. 5 | Bone marrow Siglec-H + and Siglec-H − pre-cDC2 populations respond differentially to lymphotoxin and Notch ligands to become cDC2s.a, Cell number, expression of MHC-II, expression of SIRPα and expression of T-bet-ZsGreen on bone marrow Siglec-H lo pre-cDC2s and Siglec-H hi pre-cDC2s after the culture of Siglec-H + and Siglec-H − pre-cDC2s sorted from the bone marrow of T-bet-ZsGreen mice (as in Extended Data Fig. 7a) with OP9 or OP9-DL4 stromal cells for 3 days in the presence of Flt3L with or without recombinant mouse lymphotoxin.b, Representative flow cytometry plots showing the expression of MHC-II, SIRPα and T-bet-ZsGreen on Siglec-H + pre-cDC2s and Siglec-H − pre-cDC2s on day 3 of coculture with OP9-DL4 stromal cells, Flt3L