Novel mouse models based on intersectional genetics to identify and characterize plasmacytoid dendritic cells

Plasmacytoid dendritic cells (pDCs) are the main source of type I interferon (IFN-I) during viral infections. Their other functions are debated, due to a lack of tools to identify and target them in vivo without affecting pDC-like cells and transitional DCs (tDCs), which harbor overlapping phenotypes and transcriptomes but a higher efficacy for T cell activation. In the present report, we present a reporter mouse, pDC-Tom, designed through intersectional genetics based on unique Siglech and Pacsin1 coexpression in pDCs. The pDC-Tom mice specifically tagged pDCs and, on breeding with Zbtb46GFP mice, enabled transcriptomic profiling of all splenic DC types, unraveling diverging activation of pDC-like cells versus tDCs during a viral infection. The pDC-Tom mice also revealed initially similar but later divergent microanatomical relocation of splenic IFN+ versus IFN− pDCs during infection. The mouse models and specific gene modules we report here will be useful to delineate the physiological functions of pDCs versus other DC types.

Plasmacytoid dendritic cells (pDCs) are the main source of type I interferon (IFN-I) during viral infections. Their other functions are debated, due to a lack of tools to identify and target them in vivo without affecting pDC-like cells and transitional DCs (tDCs), which harbor overlapping phenotypes and transcriptomes but a higher efficacy for T cell activation. In the present report, we present a reporter mouse, pDC-Tom, designed through intersectional genetics based on unique Siglech and Pacsin1 coexpression in pDCs. The pDC-Tom mice specifically tagged pDCs and, on breeding with Zbtb46 GFP mice, enabled transcriptomic profiling of all splenic DC types, unraveling diverging activation of pDC-like cells versus tDCs during a viral infection. The pDC-Tom mice also revealed initially similar but later divergent microanatomical relocation of splenic IFN + versus IFN − pDCs during infection. The mouse models and specific gene modules we report here will be useful to delineate the physiological functions of pDCs versus other DC types.
Host survival from viral infections depends on IFN-I, exerting both antiviral and immunoregulatory functions 1 . However, dysregulated IFN-I production fuels immunopathology in autoimmune diseases and certain viral infections 1,2 . Hence, identifying the cellular sources of IFN-I and their molecular regulation is important to design treatments to boost or dampen IFN-I responses depending on the pathophysiological context.
The pDCs are specialized in rapid and high-level production of IFN-I in response to viruses 1,3 . They engulf virus-derived material into endosomes equipped with toll-like receptor 9 (TLR9) for sensing unmethylated CpG DNA and TLR7 for single-stranded RNA. TLR7/9 activates an MyD88-to-IRF7 signaling cascade, leading to IFN-I production. Recently, new DC types sharing surface markers and gene expression with pDCs were identified, including pDC-like cells and transitional DCs (tDCs), which can contaminate pDC populations and confound their characterization [4][5][6] .
Only a small fraction of pDCs produces IFN-I during viral infections 7 . How this process is regulated remains enigmatic. We do not know precisely when, where and how pDCs sense and sample virus-derived material, and how this shapes host antiviral defense 3 . Answering these questions has been hampered by the lack of mutant mouse models enabling specific and penetrant targeting of pDCs 8 . This bottleneck was caused by lack of a gene expressed in pDCs with high enough specificity to target them by classic knock-out or knock-in approaches 9 . In the Siglech-based deleter/reporter mice, other DC populations and macrophage subsets are targeted 10 . Moreover, activated pDCs downregulate SiglecH 7 . Hence, Siglech-green fluorescent protein (GFP) mice are not suitable for reliably identifying pDCs in vivo 10 , even if this was attempted by selecting cells with high GFP intensity and plasmacytoid morphology 11 . Transgenic Siglech-Cre mice additionally suffer from a low pDC-targeting efficacy 12 . In Itgax Cre ;Tcf4 flox/− mice, tDCs and pDC-like cell development is compromised 6 ; macrophage and Technical Report https://doi.org/10.1038/s41590-023-01454-9 intensity (MFI), in pDCs' immediate precursors, the CD11c + pre-pDCs ( Fig. 2c and Extended Data Fig. 2a).
Upstream along the lymphoid path, low tdT levels were detected in CD127 + SiglecH + Ly6D + progenitors ( Fig. 2g and Extended Data Fig. 2b), consistent with these cells giving rise to pDCs 4 . Very low tdT levels were detected in the Ly6D + SiglecH − progenitor and none upstream ( Fig. 2f-g and Extended Data Fig. 2b).
Thus, in pDC-Tom mice, tdT expression is exclusively induced in late bone marrow precursors committed to the pDC lineage, with a strong increase in CD11c + pre-pDCs and maximal level in differentiated pDCs.

ZeST mice distinguish pDCs, pDC-like cells and tDCs
Refined, flow cytometry phenotypic keys can discriminate pDCs from pDC-like cells and tDCs at steady state 6 . This remains challenging in inflammation and in tissues by immunohistofluorescence. Within hematopoietic cells, Zbtb46 is specifically expressed in the cDC lineage including in pre-cDCs 18 , in pDC-like cells, as confirmed in Zbtb46 GFP mice 4,26 , and in tDCs 6 . Therefore, to rigorously identify pDC-like cells and tDCs, we generated Zbtb46 GFP ;Siglech iCre ;Pacsin1 LSL-tdT (ZeST) mice (Fig. 3a).
To validate the identity of the DC types gated manually, we characterized their phenotype further (Fig. 3c). As expected, beyond being Ly6D + SiglecH + , both tdT − and tdT + pDCs were CCR9 high CD11c int . A fraction of Zbtb46 + Ly6D + cells expressed lower levels of tdT and CCR9, and higher levels of CD11c, compared with tdT + pDCs. CD11c high tDCs were GFP high-tdT − . The pDC-like cells were GFP + tdT − CCR9 −/low . The surface phenotype of these populations was not modified during MCMV infection ( Fig. 3c and Extended Data Fig. 3). Hence, tdT was expressed only in cells harboring a phenotype of bona fide pDCs, whereas GFP was mainly expressed in cDC1s, cDC2s, pDC-like cells and CD11c high tDCs.
Next, we performed an unsupervised analysis of the phenotypic relationships for all the Lin − , CD11c high or SiglecH + cells. Onto a t-distributed stochastic neighbor embedding (t-SNE) representation of the data, we projected the populations identified through manual B cell subsets coexpressing Tcf4 and Itgax 6,13 might also be affected. BDCA2-DTR mice 14 are the most trusted model for pDC depletion; however, they should be used with caution because their serial injection with diphtheria toxin causes artefactual chronic IFN-I production and severe immunopathology, with one dose sufficient to induce IFN-I (ref. 15 ). Thus, there is an unmet scientific need for mutant mouse models allowing specific and penetrant targeting of pDCs without technical artefacts 8 .

Results
The pDC-Tom mice allow specific pDC detection by flow cytometry We generated mice knocked in for Cre expression from Siglech, which is highly expressed by mouse pDCs 3 . We crossed Siglech iCre and Rosa26 LoxP-STOP-LoxP(LSL)-RFP mice 16 to generate S-RFP mice for fate-mapping Siglech-targeted cells (Fig. 1a). Over 95% of splenic pDCs were red fluorescent protein positive (RFP + ) ( Fig. 1b and Extended Data Fig. 1a,b). Variable proportions of myeloid and lymphoid lineages expressed RFP ( Fig. 1b and Extended Data Fig. 1a,b), consistent with Siglech expression 10,17 . We reasoned that enhanced specificity could be achieved by harnessing intersectional genetics, driving expression of a reporter under the control of two genes coexpressed only in pDCs. We aimed for activation by Siglech-driven Cre of a conditional fluorescent reporter cassette knocked in a gene exclusively expressed by pDCs within Siglech fate-mapped cells. We selected Pacsin1, expressed exclusively in pDCs within hematopoietic cells 18 and promoting their IFN-I production 19 . We generated Pacsin1 LoxP-STOP-LoxP-tdTomato (Pacsin1 LSL-tdT ) mice, knocked in with a floxed cassette for tdTomato (tdT) conditional expression. We crossed them with Siglech iCre mice, to generate pDC-Tom mice (Fig. 1c). In splenocytes from pDC-Tom mice, tdT was exclusively expressed in pDCs ( Fig. 1d and Extended Data Fig. 1a,c). The tdT + cells expressed neither lineage markers nor CD11b (Extended Data Fig. 1d). The CD45 + tdT + cells isolated from different organs were CD11c int and BST2 high (Fig. 1e), as expected for pDCs 20 . CD45 + tdT + cells coexpressed Ly6D, B220, SiglecH and CCR9 (Fig. 1f), a combination specific to pDCs. Thus, tdT expression in pDC-Tom mice is sufficient to specifically and unambiguously identify most pDCs.

The pDC-Tom mice allow refining pDC gating strategies
Defining pDCs as coexpressing CD11c, BST2 and SiglecH can lead to contamination by conventional DCs (cDCs), pDC-like cells 4 or tDCs 5,6,21 . Moreover, on inflammation, such as during mouse cytomegalovirus (MCMV) infection, the expression of these markers is altered, causing a phenotypic convergence of pDCs and cDCs 7 . Hence, we harnessed pDC-Tom mice to define a gating strategy allowing unequivocal pDC identification both at steady state and during infection. We defined pDCs as tdT + cells and used HyperFinder for unsupervised computational generation of a gating strategy to identify them, based on surface markers without using tdT. We included Ly6D, which is selectively expressed on pDCs and B cells, discriminating them from cDCs and tDCs 4,5 . At steady state, splenic pDCs were identified as Bst2 high L y6D + B220 + CD19 − CCR9 + SiglecH + cells (Fig. 1g). During MCMV infection, they were identified as Ly6D + CX 3 CR1 low/int CD19 − CCR9 high B220 high BST2 high cells (Fig. 1h). Hence, current identification of pDCs as lin − CD11 b − CD11c low-to-int BST2 high cells 20 can be improved by addition of positivity for Ly6D or CCR9 and eventual exclusion of CX 3 CR1 high cells. SiglecH is not a good marker postinfection (p.i.), because it is downregulated (Fig. 1i), especially on IFN-I-producing pDCs 7 . We thus propose identifying pDCs as Ly6D high BST2 high CD19 − B220 + CD11b − CD11c + cells.

Technical Report
https://doi.org/10.1038/s41590-023-01454-9 We sorted cDC2s, CD11c high tDCs, pDC-like cells and pDCs from steady-state mouse spleens to examine their morphology (Fig. 3e). The cDC2s harbored many pseudopods or dendrites, translating into a low circularity index. Most pDC-like cells and pDCs harbored a round morphology, translating into a high circularity index. The CD11c high tDC population was morphologically heterogeneous, with a bimodal distribution of circularity indices, half of the cells being dendritic, like cDC2s, and the other half round, like pDCs. Overall, quantitative and unbiased analysis of cellular morphology supported success in high-degree purification of the DC types.
Finally, to better discriminate the tdT signal from autofluorescence, and analyze more cell-surface markers in ZeST mice, we harnessed spectral flow cytometry (Extended Data Figs. 4 and 5). Unsupervised cell clustering based on all surface markers (Extended Data Fig. 4a,b), without considering tdT and GFP, showed that most GFP + cells were cDC2s or cDC1s (Extended Data Fig. 4c). They also encompassed lymphoid cells and a cluster of myeloid cells, but with a low MFI (Extended Data Fig. 4d). The pDCs represented 81.5 ± 6% of tdT + splenocytes (Extended Data Fig. 4e). The individual contribution of other phenotypic cell clusters to the tdT + gate was very small and their tdT MFI below that of pDCs, barely above background (Extended Data Fig. 4f).
Complementary analysis by supervised identification of cell types via manual gating allowed study of both tDC populations: the Ly6C − versus Ly6C + fractions of Lin − CD11b − XCR1 − , CD11c + or BST2 + , CX3CR1 + CD26 + cells (Extended Data Fig. 5a-d). The populations phenotypically defined as pDC-like cells 4 or Ly6C + tDCs 6 largely overlapped, as was the case for CD11c high tDCs and Ly6C − tDCs (Extended Data Fig. 5b-e). Most Lin − CD11b − XCR1 − , CD11c + or BST2 + , Ly6D − CX3CR1 − cells were CD11c + CD26 + GFP + CD64 − (Extended Data Fig. 5f), indicating DC lineage. They encompassed major histocompatibility complex II (MHC-II) −/low and MHC-II high cells putatively corresponding to pre-DCs versus differentiated DCs, respectively. A decrease in the absolute numbers of most of DC types was observed in the spleen at 48 h p.i. (Supplementary Table). This analysis confirmed the high specificity and penetrance of GFP expression in cDC1s and cDC2s, and both tDC populations, as well as the high specificity and penetrance of tdT expression in pDCs (Extended Data Fig. 5g).

ScRNA-seq confirms DC-type identification in ZeST mice
We harnessed ZeST mice to perform single-cell RNA sequencing (scRNA-seq) for the five DC types, on index sorting from the spleen of animals either infected or not infected with MCMV, using the gating strategy shown in Fig. 3b and the FB5P-seq (FACS-based 5′-end scRNA-seq) protocol 27,28 .
We first analyzed 343 splenocytes isolated from noninfected (NI) mice. They were clustered and annotated for cell types (Supplementary Table) by Seurat, based on gene expression profiles (Extended Data Fig. 6a), and by Rphenograph, based on surface marker expression (Extended Data Fig. 6b,c). DC-type assignment to individual cells was rather consistent between these two strategies, but suggested heterogeneity of Seurat clusters II and III (Extended Data Fig. 6d). Therefore, to unambiguously and reliably assign a DC type to individual cells in an unbiased manner, we used a combinatorial strategy: to be selected, a cell had to belong to the expected intersection between the Rphenograph and Seurat clusters (colored cells in Extended Data  Table).
Our next objective was to assign a DC type to the 951 cells from the whole dataset, from both NI and MCMV-infected mice (Extended Data Fig. 7a and Supplementary Table). We aimed at achieving a robust DC-type assignment, irrespective of cell states and infection conditions, by combining transcriptomic and phenotypic analyses. Based on the intersection of the clustering with Seurat (Extended Data Fig. 7b) versus Rphenograph (Extended Data Fig. 7c,d), a final DC type was assigned to 851 cells (colored cells in Extended Data Fig. 7e): 310 pDCs, 170 pDC-like cells, 146 CD11c high tDCs, 103 cDC2s and 122 cDC1s, with 100 cells left unannotated (Fig. 4a). We confirmed DC-type assignment by a complementary method: we generated cell type-specific transcriptomic signatures from the dataset focused on cells from NI mice (Supplementary Table) and used them for sgCMap analysis of the whole dataset (Extended Data Fig. 7f). All the cells sorted as pDCs, and most of the tdT − pDCs, were computationally assigned to pDCs (Extended Data Fig. 7e,f). The assignment to cDC1s was also consistent between cell sorting and deductive re-annotation. Cell clustering on sgCMap scores tended to confirm the distinction between cDC2s and CD11c high tDCs, although many cells had a null score for the 'tDC_vs_cDC2' sgCMap signature, emphasizing the proximity between these two DC types (Extended Data Fig. 7f). As expected, Zbtb46 + Ly6D + cells were mostly assigned to pDC-like cells. Some cells sorted as pDC-like cells were in time assigned to CD11c high tDCs (Extended Data Fig. 7e,f). Not only CD11c high tDCs but also pDC-like cells were CX3CR1 +/high (Fig. 4b). Akin to cDCs, CD11c high tDCs were GFP high , whereas pDC-like cells expressed SiglecH, BST2, Ly6D and CCR9 to levels intermediate between those of pDCs (high) and cDCs (low). The pDCs and pDC-like cells shared high expression of Siglech (Fig. 4c,d), Tcf4 and Runx2 (Fig. 4c). CD11c high tDCs and pDC-like cells shared high expression of Crip1, Lgals3 and Vim (Fig. 4c), previously reported to discriminate pDC-like cells from pDCs 4 . The pDC-like cells and CD11c high tDCs shared with cDCs higher expression of Zbtb46, Spi1, Slamf7 and S100a11, compared with pDCs (Fig. 4c). Contrary to cDC2 cells, a fraction of pDC-like cells and CD11c high tDCs expressed Cd8a (Fig. 4c,d), as reported for tDCs 6,21,29 . The pDC-like cells, CD11c high tDCs and cDC2s specifically expressed Ms4a4c (Fig. 4c,d) and Ms4a6c (Fig. 4c). CD11c high tDCs and cDC2s selectively shared expression of Ms4a6b (Fig. 4d) and S100a4 (Fig. 4c). CD11c high tDCs expressed higher levels of certain cDC genes than pDCs and pDC-like cells, including Batf3, Rogdi and Cyria (Fig. 4c). The pDC-like cells expressed very high levels of Ly6c2 (Fig. 4c,d). Hence, the pDC-like cells characterized in the present report were confirmed to align with both the originally described pDC-like cells 4 and the CD11c low Ly6C + tDCs 6 .
All cDC1s expressed the XCR1 protein (Fig. 4b). However, a fraction was low/negative for Xcr1 and other genes specific of steady-state cDC1s (Gpr141b, Tlr3, Cadm1 and Naaa; Fig. 4c), consistent with DC-type activation decreasing the expression of many of the genes used to identify them at steady state [30][31][32] , preventing use of individual genes for reliable DC-type identification in scRNA-seq datasets 33 .
As expected, only pDCs expressed rearranged immunoglobulin (Ig) genes, Ccr9, Klk1 and Cox6a2 (Fig. 4c) as well as the tdT fluorescent reporter above autofluorescence levels (Fig. 4e). The tdT − pDCs were largely overrepresented in our scRNA-seq dataset because we enriched them for characterization. As we did not use the tdT signal in our analysis, these results confirm the specificity of pDC-Tom mice for pDC identification; they also confirmed proper identification of pDC-like cells and CD11c high tDCs in ZeST mice, and allowed refining of their characterization through side-by-side pangenomic transcriptomic profiling.

The pDC-Tom mice allow mapping of pDC microanatomical localization
We harnessed the pDC-Tom mice to determine the microanatomical localization of pDCs in various organs. In the spleen, a relatively high density of tdT + cells was observed in the T cell zone (TCZ) within the white pulp (WP) (Fig. 5a). Scattered tdT + cells were detected in the red pulp (RP), delimited by CD169 + metalophilic marginal zone (MZ) macrophages and densely populated by F4/80 + RP macrophages

Technical Report
https://doi.org/10.1038/s41590-023-01454-9 ( Fig. 5b). Most TCZ or RP tdT + splenocytes expressed BST2 (Fig. 5a,b). Conversely, whereas most TCZ BST2 + cells were tdT + (Fig. 5a), this was not the case in the RP (Fig. 5b), consistent with BST2 expression on plasma cells, macrophage subsets, pDC-like cells and tDCs 4,6,34 . We quantified the density of pDCs (tdT + cells) in the whole spleen (Fig. 5c,d) and in its microanatomical compartments: the RP, the MZ, the TCZ and the B cell zone (BCZ) (Fig. 5c,e,f). The pDC density in the whole spleen was around 400 cells mm −2 (Fig. 5d). Steady-state splenic pDCs were primarily located in the TCZ and RP (Fig. 5e). The pDC density was much higher in the TCZ (~1,500 cells mm −2 ) than in other splenic compartments (≤400 cells mm −2 ) (Fig. 5c,f). In lymph nodes, pDCs were frequent in paracortical T cell (CD3 + ) areas (Extended Data Fig. 8a). In the small intestine, pDCs had been observed in both Peyer's patches 35 and the lamina propria (LP), primarily based on BST2 signal 36 . However, interpretation may have been confounded by BST2 expression on macrophage and DC subsets in the gut 37,38 . Hence, we re-examined pDC microanatomical location in the small intestine. The tdT + cells were detectable in the LP, generally close to EpCAM + epithelial cells (Extended Data Fig. 8b). Very few tdT + cells were observed in Peyer's patches (Extended Data Fig. 8c). The pDCs were not detectable in the LP of the large intestine (Extended Data Fig. 8d). Thus, at steady state, in the gut, pDCs are detectable quasi-exclusively in the small intestine LP.

SCRIPT mice allow tracking IFN-I + and IFN-I − pDCs in situ
During MCMV infection, the Ifnb1 EYFP reporter mouse model 39 does not only allow detection, at 36 h p.i., of the pDCs actively producing IFN-β 7,27 , but, in addition, at 48 h, of the pDCs that have already produced and secreted IFN-β but maintained enhanced YFP expression for >12 h 27 . Hence, the Ifnb1 EYFP mice are reliable and well suited to fate-map the pDCs that have previously produced IFN-I, enabling determination of their activation trajectory in vivo during infection 27 . The pDCs clustered in the MZ near to infected cells at 36 h p.i., at the time of peak IFN-I production, then migrated to the TCZ after termination of their IFN-I production between 40 h and 48 h p.i., while acquiring transcriptomic, phenotypic and functional features of mature cDCs 27   One-way analysis of variance (ANOVA) was used for the statistical analysis, * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

Technical Report
https://doi.org/10.1038/s41590-023-01454-9 of adequate markers prevented us from assessing the spatiotemporal repositioning of the pDCs that were not producing IFN-I. To answer this question, we generated Siglech iCre ;Ifnb1 EYFP ;Pacsin1 LSL-tdT (SCRIPT) mice (Fig. 6a) for unambiguous identification of both the pDCs that are producing (or have produced) IFN-I (tdT + YFP + ) and those that do not (tdT + YFP − ) (Fig. 6a,b). By flow cytometry, both tdT + YFP − and tdT + YFP + cells were detectable at 36 h and 48 h p.i., expressing Ly6D and BST2 consistent with their pDC identity (Fig. 6b). We detected only very rare tdT − YFP + Ly6D − BST2 low cells, consistent with pDCs being the main source of IFN-I during infection 7,20,27,40 . We examined the localization of both tdT + YFP − and tdT + YFP + pDCs in spleens from MCMV-infected mice (Fig. 6c-f). At 36 h p.i., both pDC populations formed clusters in the MZ (Fig. 6c,d), leading to an increase in the proportion of pDCs at this microanatomical site (Fig. 6e). Although pDCs were still detectable in the TCZ of infected animals, their local density was reduced compared with uninfected mice (Fig. 6c,e).

Only IFN-I fate_mapped pDCs entered the TCZ during infection
At 48 h p.i., we detected large aggregates of tdT + YFP − cells in the MZ, whereas tdT + YFP + cells were mostly detected in the TCZ (Fig. 6c,d).
Hence, we observed an opposite trend in the spatiotemporal repositioning of the pDCs that either produced or did not produce IFN-I within the spleen during infection: whereas most of the pDCs fate-mapped for IFN-I production (YFP + ) were located in the MZ at 36 h and then in the TCZ at 48 h, the reverse distribution was observed for the pDCs that did not produce IFN-I (YFP − ) (Fig. 6f). This analysis showed that, at the peak of IFN production, the high recruitment of pDCs to the MZ occurs independent of their ability to produce IFN-I, whereas, at later time points, only the pDCs that have produced IFN-I are licensed to migrate to the TCZ; the other pDCs are retained in the MZ. Next, we sorted DC types from the spleens of infected mice and examined their morphology, also comparing YFP + versus YFP − pDCs at 60 h p.i. (Fig. 6g,h and Extended Data Fig. 9). A fraction of YFP + pDCs acquired an irregular morphology with pseudopods or dendrites (Fig. 6g), harboring significantly lower circularity indices compared both with their YFP − counterparts and with steady-state pDCs (Fig. 6h). Hence, the pDCs that had produced IFN-I during infection selectively acquired a dendritic morphology, consistent with their known transcriptomic and functional convergence toward cDCs 27 .

The pDCs clustered around infected cells but few produced IFN-I
Due to the lack of specific markers to track pDCs in situ before their peak IFN-I production, determination of the early kinetics of their microanatomical redistribution in the spleen during infection has not been possible previously. The pDC-Tom mice allowed us to address this issue (Fig. 7). Recruitment of pDCs and their clustering in the MZ were detectable as early as 12 h p.i. (Fig. 7a,b), with a clear increase in the proportion of MZ pDCs approaching the plateau values observed between 18 h and 48 h (Figs. 6e and 7c). Cells replicating MCMV (expressing the viral immediate early gene 1, IE1) were already detectable at 12 h p.i., mainly in the MZ (Fig. 7d-f). The pDC clusters were already localized in close proximity to MCMV-infected cells at 12 h (Fig. 7d), consistent with the proximity between IFN-I-producing pDCs and infected cells observed at later time points 20 . Thus, the recruitment of pDCs to the vicinity of infected cells in the MZ occurred early, already at 12 h p.i., 24 h before their peak IFN-I production.
In the MZ of the spleen of 40-h infected mice, three-dimensional (3D) reconstructions from confocal microscopy images showed that IFN-I-producing pDCs established tight interactions with virus-infected cells ( Supplementary Videos 1-3), consistent with in vivo establishment of interferogenic synapses as previously observed only in vitro 41,42 .

The pDC-like cells and CD11c high tDCs diverge on activation
We analyzed our FB5P-seq data to compare the responses of the five DC types to MCMV infection in vivo. Four Seurat clusters were identified for pDCs, corresponding to distinct activation states ( Fig. 8a and Supplementary Table): quiescent pDCs, intermediate pDCs harboring induction of IFN-stimulated genes (ISGs) but lacking expression of cytokines genes, activated pDCs expressing moderate levels of cytokine genes and IFN-I-producing pDCs expressing high levels of Ifnl2, and all the genes encoding IFN-I. CD11c high tDCs, cDC1s and cDC2s each split into three clusters, corresponding to quiescent, intermediate and activated states. The pDC-like cells split into two clusters only: quiescent and activated.
The cell types and activation states identified by scRNA-seq were shared across individual mice ( Supplementary Fig. 1a), showing similar proportions between mice at the same time p.i. (Supplementary Fig. 1b).
Expression of tdT remained highly selective in pDCs, irrespective of their activation states (Fig. 8b), despite SiglecH downregulation in IFN-I-producing pDCs 7 (Extended Data Fig. 10a). Activation increased autofluorescence but did not induce tdT in other DC types (Fig. 8b). The pDCs remained Ly6D high CCR9 high across activation states. BST2 expression increased on all activated DC types (Fig. 8b), as reported previously 7,43,44 . Activation decreased CX3CR1 expression in CD11c high tDCs (Extended Data Fig. 10). Together with their increased BST2 expression (Fig. 8b), this contributed to shifting a fraction of the CD11c high tDC population from infected mice into the phenotypic gate used for sorting pDC-like cells, which was corrected by deductive cell-type reassignment on scRNA-seq computational analysis (Extended Data Fig. 7f and Supplementary Table). The cells assigned to CD11c high tDCs expressed higher levels of CD11c and GFP than pDC-like cells (Extended Data Fig. 10a). Although the expression of XCR1 on cDC1s and CD11b on cDC2s decreased with activation, these surface markers remained clearly detectable, allowing phenotypic identification of these DC types (Extended Data Fig. 10a), contrasting with near extinction of Xcr1 expression in activated cDC1s (Extended Data Fig. 10b).
As observed in previous reports based on flow cytometry or bulk RNA-seq data 6 , quiescent CD11c high tDCs and pDC-like cells were close together in the Uniform Manifold Approximation and Projection (UMAP) space, between pDCs and cDC2s (Fig. 8a). On activation, they moved in opposite directions, with activated pDC-like cells close to intermediate pDCs, but intermediate or activated CD11c high tDCs and cDC2s close together (Fig. 8a) Fig. 7a). DC-type assignment is the same as in Fig. 4a. Seurat clusters are indicated on the UMAP (Extended Data Fig. 7b). Activation states were assigned based on mining of the marker genes of Seurat clusters (see Supplementary  Table). b, Violin plots showing the expression of selected phenotypic markers across DC types and activation states. c, Heatmap showing mRNA expression levels of selected genes (rows) across all 951 individual cells (columns), with hierarchical clustering of genes using Euclidean distance, and ordering of individual cells (column) according to their assignment into cell types and activation states using the same color code (top) as in a. The color scale for gene expression levels is the same as in Fig. 4c.

Technical Report
https://doi.org/10.1038/s41590-023-01454-9 whereas only 84 genes were differentially expressed between CD11c high tDCs and pDC-like cells at the steady state, this number increased to 527 for activated cells (Supplementary Table). Whereas all DC types induced ISGs (Fig. 8c, cluster 1), only CD11c high tDCs, cDC2s and cDC1s induced high levels of genes linked to DC maturation/migration and interactions with T cells, encompassing Fscn1, Ccr7, Il15, Il15ra, Cd80, Cd200 and Cd274 (Fig. 8c, cluster 3). Another set of ISGs and genes associated with DC interactions with T cells gradually increased in expression from intermediate to activated to IFN-I-producing pDCs (Fig. 8c, cluster 2). These genes were induced to similar levels in activated pDC-like cells, but to even higher levels in activated CD11c high tDCs, cDC2s and cDC1s. Only a few genes were differentially expressed between CD11c high tDCs and cDC2s: 16 at steady state, 1 at the intermediate activation state and 11 for activated cells (Supplementary Table). The number of genes differentially expressed between pDCs and pDC-like cells remained stable over activation: 113 at steady state versus 138 in activated cells (Supplementary Table). Hence, on activation, pDC-like cells behaved more like pDCs, although not producing IFN-I (Fig. 8c, cluster 5), whereas CD11c high tDCs converged further toward cDC2s. Activated pDC-like cells maintained a higher Ly6c2 expression (Extended Data Fig. 10c) and a lower CD11c MFI (Extended Data Fig. 10a) than CD11c high tDCs, confirming the reliability of these markers even in inflammation. Genes encoding MHC-II molecules, such as H2-DMb2 and H2-Eb1, were higher in CD11c high tDCs than in pDC-like cells, reaching the same levels as in cDC2s (Extended Data Fig. 10c). Tmem176a and Tmem176b, encoding cell-surface markers, were higher in CD11c high tDCs than in pDC-like cells, pDCs, cDC1s and to a lesser extent cDC2s, across activation conditions (Extended Data Fig. 10c,d).
Cd8a expression remained lower in cDC2s across activation conditions (Extended Data Fig. 10c). Ms4a4c was higher in activated pDC-like cells, intermediate and activated CD11c high tDCs and cDC2s than in pDCs and cDC1s (Extended Data Fig. 10c). Apod was expressed to higher levels in pDC-like cells, although it was induced in all other DC types except cDC2s on activation (Extended Data Fig. 10c,d).
Finally, beyond confirming the unique ability of pDCs to produce IFN-I/III during MCMV infection 7,20,40 , our analysis unraveled selective induction in cDC1s of genes encoding costimulation molecules or cytokines involved in the crosstalk with natural killer (NK) or T cells, encompassing Cd70, Pdcd1lg2, Pvr and Ccl22 (Fig. 8c, cluster 4), Il27 (Fig. 8c, cluster 3), Il18 and Il12b (Extended Data Fig. 10e). This is consistent with a division of labor between DC types, with cDC1s promoting NK and CD8 T cell antiviral responses 45,46 .

Discussion
We generated the first reporter mouse model, to our knowledge, specifically and efficiently tagging pDCs, across organs and activation conditions: the pDC-Tom (Siglech iCre ;Pacsin1 LSL-tdT ) mice.
We did not provide data to solve the current debate on pDC ontogeny 4,5,47 , because tdT expression in our model starts only from the late CD11c + pre-pDC differentiation state that is common to the 'lymphoid' and 'myeloid' paths.
Using pDC-Tom mice, we could study the choreography of the relocation of all pDCs in the spleen during MCMV infection. All pDCs were attracted to the microanatomical sites of viral replication early p.i. Thus, the failure of most pDCs to produce IFN-I is unlikely to result from restricted access to infected cells. Alternatively, pDC tight clustering raises the hypothesis that the first to produce IFN-I may have repressed this function in their neighbors, through a quorum-sensing mechanism, to prevent excessive inflammation and consequent immunopathology. Future studies using pDC-reporter mice and pDC-specific genetic manipulations should help to better understand this regulation.
We generated the first side-by-side transcriptomic comparison, to our knowledge, of pDCs, pDC-like cells and CD11c high tDCs, through scRNA-seq. We uncovered a divergent activation between CD11c high tDCs and pDC-like cells. Only CD11c high tDCs underwent a maturation closely resembling that of cDCs.
CD11c high tDCs and activated cDC2s selectively expressed high levels of H2-DMb2 and Tmem176a/b, new candidate markers for these DC types. Tmem176a/b colocalizes with human leukocyte antigen (HLA)-DM in the late endolysosomal system, promotes antigen presentation to naive T cells and contributes to tolerance/immunosuppression [48][49][50] . This raises the question of the division of labor between CD11c high tDCs and cDC2s, including for CD4 T cell tolerance. Our scRNA-seq dataset will be a valuable resource to mine the gene expression profiles of pDC-like cells and CD11c high tDCs, compared with pDCs, cDC1s and cDC2s, to infer and experimentally test hypotheses on their functional specialization and molecular regulation.

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/s41590-023-01454-9.

Technical Report
https://doi.org/10.1038/s41590-023-01454-9 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/.

Mice
All animal experiments were performed in accordance with national and international laws for laboratory animal welfare and experimentation (EEC Council Directive 2010/63/EU, September 2010). Protocols were approved by the Marseille Ethical Committee for Animal Experimentation (registered by the Comité National de Réflexion Ethique sur l'Expérimentation Animale under no. 14; APAFIS no. 1212-2015072117438525 v.5 and APAFIS no. 21626-2019072606014177 v.4). C57BL/6 mice were purchased from Janvier Labs. All other mouse strains were bred at the Centre d'ImmunoPhénomique (CIPHE) or the Centre d'Immunologie de Marseille-Luminy (CIML), under specific pathogen free-conditions and in accordance with animal care and use regulations. Mice were housed under a 12 h dark:12 h light cycle, with a temperature range of 20-22 °C and a humidity range of 40-70%. Siglech iCre mice (B6-Siglech tm1(iCre)Ciphe ) 20 and Pacsin1 LoxP-STOP-LoxP-tdTomato(LSL-tdT) (B6-Pacsin1 tm1(tdT)Ciphe ) mice were generated by CIPHE. Siglech iCre ;Pacsin1 LSL-tdT mice (pDC-Tom) were generated by crossing Siglech iCre mice with Pacsin1 LSL-tdT mice, and then maintained and used in a double homozygous state. SCRIPT mice were generated by crossing pDC-Tom mice with Ifnb1 Eyfp mice (B6.129-Ifnb1 tm1Lky ) 39 , and then maintained and used in a triple homozygous state. ZeST mice were generated by crossing Zbtb46 GFP mice (B6.129S6(C)-Zbtb46 tm1.1Kmm/J ) 26 with pDC-Tom mice and were used in a heterozygous state. All animals used were sex and age matched (used between 8 and 16 weeks of age).

Virus and viral infection
Virus stocks were prepared from salivary gland extracts of 3-week-old, MCMV-infected BALB/c mice. All mice used in the experiments were infected intraperitoneally with 10 5 plaque-forming units of Smith MCMV and sacrificed at the indicated time points.

Cell preparation for flow cytometry analysis or cell sorting
Spleens or lymph nodes were harvested and submitted to enzymatic digestion for 25 min at 37 °C with Collagenase IV (Worthington biochemical) and DNase I (Roche Diagnostics). Organs were then mechanically digested and passed over 100-μm cell strainers (Corning). Bone marrow cells were flushed from mouse femurs. Red blood cells (RBCs) were then lysed by using RBC lysis buffer (Life Technologies) for spleen and bone marrow cell preparation. Livers were harvested, minced and submitted to enzymatic digestion, as for the spleen. Liver pieces were then crushed and the cell suspension obtained was washed 2× with phosphate-buffered saline (PBS) 1×, before performing a 80:40 Percoll gradient. Small intestines were harvested, opened longitudinally and then cut into 1-cm pieces. Pieces were washed extensively with PBS 1×, then incubated 3× at 37 °C on shaking (200 r.p.m.) with PBS 1× containing 2% fetal calf serum (FCS) and 5 mM EDTA. At the end of each incubation, supernatants were collected and centrifuged. Pelleted cells, mainly intraepithelial lymphocytes, from the three incubations were pooled together and submitted to a 67:44 Percoll gradient. Cells isolated from the middle ring of Percoll gradients were washed once with PBS 1×, then used for flow cytometry.

Spectral flow cytometry analysis
All antibodies were purchased from Becton Dickinson Biosciences, Bio-Legend or eBioscience. Dead cells were discriminated in all experiments using Live/dead (ZombieNIR) fixable dead stain (Life Technologies). All staining was carried out with Fc-block (2.4G2 and 9E9) in staining buffer (PBS, 2 mM EDTA, 0.5% bovine serum albumin (BSA) and 20% Brilliant Violet stain buffer (BD Biosciences)). Cell suspensions were stained first for 15 min at 37 °C by monoclonal antibodies targeting chemokine receptors and certain antigens (CCR9, CD26, CD64, CX3CR1, SiglecH and XCR1) and for 25 min more by monoclonal antibodies targeting other antigens (B220, Bst2, CD45, CD3, CD8a, CD11b, CD11c, CD19, CD88, F4/80, IgD, IgM, Ly6C, Ly6D, Ly6G, NK1.1 and MHC-II). Cells were washed 3× in FACS buffer (PBS, 2 mM EDTA and 0.5% BSA) and resuspended in PBS. In all flow cytometric plots, doublets, aggregates, dead cells and autofluorescence were excluded. Acquisition was performed on an Aurora five lasers (Cytek Biosciences) using SpectroFlo v.3.0.1 software (Cytek Biosciences); the quality control (QC) indicates a similarity and complexity index of 11.7. Data were analyzed using OMIQ (app.OMIQ.ai). First, the data were run through flowCut 51 to check for aberrant signal patterns or events. Second, the data were cleaned by manual gating to remove doublets, debris and dead cells. Third, the autofluorescence of the tissue was subtracted. UMAP 52 was run to reduce the dimensions to a two-dimensional space and thus group phenotypically similar events into 'islands' to illustrate differences both between and inside each population. PARC (phenotyping by accelerated refined community) 53 was subsequently used to cluster the events based on UMAP parameters.

ScRNA-seq data generation
For the generation of the scRNA-seq data, we followed the FB5P method previously published 28 . Briefly, single cells were FACS sorted into ice-cold 96-well PCR plates (Thermo Fisher Scientific) containing 2 μl of lysis mix per well. Immediately after cell sorting, each plate was covered with an adhesive film (Thermo Fisher Scientific), briefly spun down in a benchtop plate centrifuge and frozen on dry ice. The reverse transcription (RT) reaction was performed with SuperScript II (Thermo Fisher Scientific) in the presence of RNaseOUT (Thermo Fisher Scientific), dithiothreitol (Thermo Fisher Scientific), betaine (Sigma-Aldrich), MgCl 2 (Sigma-Aldrich) and well-specific template-switching oligonucleotide. For complementary DNA amplification, KAPA HiFi HotStart ReadyMix (Roche Diagnostics) was used with adapted primers. For library preparation, amplified cDNA from each well of a 96-well plate was pooled, purified with two rounds of 0.6× solid-phase reversible immobilization beads (AmpureXP, Beckman or CleanNGS, Proteigene) and finally eluted in nuclease-free water. After tagmentation and neutralization, tagmented cDNA was amplified with Nextera PCR Mastermix containing Nextera i5 primer (Illumina) and customized i7 primer mix. Libraries generated from multiple 96-well plates of single cells and carrying distinct i7 barcodes were pooled for sequencing on an Illumina NextSeq2000 platform, with 100 cycles of P2 flow cells, targeting 5 × 10 5 reads per cell in paired-end, single-index mode with the following cycles: Read1 (Read1_SP, 67 cycles), Read i7 (i7_SP, 8 cycles) and Read2 (Read2_SP, 16 cycles). Two to three individual mice were used as a source for the single cells for each time point, with three https://doi.org/10.1038/s41590-023-01454-9 independent sorts performed with two or three animals each time (sorts for mice nos. 56 and 58 on 3 November 2020, for nos. 52, 53 and 61 on 17 December 2020 and for nos. 81, 84 and 86 on 11 February 2021); sorting plates were frozen until all samples had been collected and all libraries were generated and sequenced simultaneously to avoid eventual batch effects.

Bioinformatics analyses of scRNA-seq data
A mark-up file of how the scRNA-seq analysis was performed is provided ( Supplementary Fig. 2). FB5P-seq data were aligned and mapped to a reference genome using STAR (v.2.5.3a) and HTSeqCount (v.0.9.1) and processed to generate a single-cell, unique molecular identifier (UMI) counts matrix as described 28 . The counts matrix was loaded to R (v.4.0.3) and Seurat (v.3.2.0) 54 was used for downstream analyses as described 27 . Gene expression is shown as log(normalized values) and protein expression as inverse hyperbolic arcsine (asinh) of fluorescence intensity-scaled values (asinh(fluorescence)/100). For dimensionality reduction, we performed UMAP, using the RunUMAP function. The differentially expressed genes (DEGs) were determined using the FindMarkers function.
A convergent transcriptional reprogramming occurs in all DC types during their maturation, which can lead to their clustering primarily according to their activation states rather than to their cell types 31 . In particular, many of the genes that are specific for a given DC type at steady state are strongly downregulated on activation [30][31][32] , such that they cannot be used individually to identify DC types 33 . These phenomena recurrently caused issues for identifying DC types in the analyses of certain scRNA-seq data, which led to an inadequate inflation of DC subset nomenclature, with new DC subset names coined for clusters that actually corresponded to activation states of already identified DC types. This problem is exemplified by the current use of the name 'DC3' for mature DC clusters in scRNA-seq studies, whereas 'DC3' is also being used for a human DC type that is ontogenetically and functionally distinct from human pDCs, cDC1s, cDC2s and monocyte-derived DCs, as discussed in a recent commentary 55 . Similar to the authors of this commentary, we emphasize, in scRNA-seq data analysis, the necessity for clearly distinguishing distinct DC types from different activation states of the same DC type. To achieve this aim, proper strategies are required for robust assignment of a cell-type identity to each individual cell, irrespective of its activation state, in a manner enabling resolution of the heterogeneity of mixed-cell clusters. This can be achieved by integrating phenotypic and transcriptomic data (as enabled with index sorting or CITE-seq 33 ) or by performing a gene set enrichment analysis at the level of individual cells, by using the single-cell CMap algorithm 56 with specific composite up/down gene modules carefully defined from an independent dataset. We used both strategies in the present study as a technical guide to help readers in their future analyses of scRNA-seq data.
A first Seurat analysis was performed only on cells from uninfected mice (345 cells after QC, 2 of which were removed due to lack of index sorting data). Clusters were identified based on either gene expression using Seurat (k-nearest neighbor = 5; resolution = 0.2) or phenotypic marker expression using Rphenograph 57 (v.0.99.1) (number of nearest neighbors k = 20), taking into account the protein expression for GFP, B220, Ly6D, XCR1, CX3CR1, BST2, SiglecH, CCR9, CD11b and CD11c (but not for tdTomato). A CMap analysis 56 was performed on this dataset using cell-type-specific signatures (tDCs, cDC2s, cDC1s and pDCs) established on reanalysis with the BubbleGUM GeneSign module 58 of a published 29 , independent, bulk RNA-seq dataset (Gene Expression Omnibus (GEO) accession no. GSE76132), and a relative pDC_vs_pDClike signature retrieved from a previously published study 4 . The GSE76132 dataset encompassed four cell types, tDCs, cDC2s, cDC1s and pDCs, analyzed through bulk RNA-seq. We retrieved and reanalyzed these data to generate signatures for each DC type compared with the three others. Hence, we ended up with four composite signatures, one for each DC type, with its 'UP' and 'DN' genes when compared with the three other DC types. The study by Rodrigues et al. 4 encompassed only two cell types, pDCs and pDC-like cells, analyzed through scRNA-seq. Hence, from this dataset, we retrieved one composite signature encompassing the genes expressed higher in pDCs above pDC-like cells ('UP') and reciprocally the genes less expressed in pDCs than in pDC-like cells ('DN'). The CMap algorithm has been made available on Github (https:// github.com/SIgN-Bioinformatics/sgCMAP_R_Scripts), as well as an example and recommendations on how to use it (https://github.com/ DalodLab/MDlab_cDC1_differentiation/blob/main/scRNAseq_pipeline. md#cell-annotation-using-cmap).
Integration of Seurat and Rphenograph cluster information together with the CMap scores allowed identification of, and focus on, bona fide steady-state pDCs, pDC-like cells, CD11c high tDCs, cDC1s and cDC2s (205 total cells), from which we computed relative signatures based on all pairwise comparisons between cell types, using the FindMarkers function (default parameters), for consecutive single-cell CMap analyses.
A second Seurat analysis was then performed on cells from both uninfected and MCMV-infected mice (1,132 cells after QC). Clusters of contaminating cell types (macrophages, NK cells and a small cluster of proliferative cells of mixed types) were identified by their top markers using the FindMarkers function (test.use = 'bimod') and removed (181 cells). A third Seurat analysis was performed on the remaining cells (951 cells).
Clusters were identified based on gene expression using Seurat (k-nearest neighbor = 9; resolution = 0.7) or based on phenotypic marker expression using Rphenograph (number of nearest neighbors k = 50). Integration of Seurat and Rphenograph clusters allowed identification of unambiguous cell types and activation states (851 total cells), which was corroborated on performing a single-cell CMap analysis using the relative signatures identified in the previous step. DEGs between cell types at equivalent activation states or between activation states for a given cell type were then extracted, using the FindMarkers function (default parameters, threshold for adjusted P < 0.05) (see Supplementary Table).

Immunohistofluorescence, microscopy and image analysis
a B220-rich area within the WP. The RP was defined as an F4/80-rich region and the MZ as the space between the CD169 staining and the F4/80 staining. For the calculation of pDC counts mm −2 , the function 'analyze particles' was used and a threshold for tdT intensity and the size (>8 μm 2 ) was applied. For the quantification of the repartition of pDCs on infection, the intensity of tdT was calculated in the different zones after a threshold for the intensity had been applied. A ratio between the intensity in each zone and the total intensity of the whole section was then calculated.

Cell immunofluorescence and cell morphology analysis
The different cell subsets were FACS sorted in a cold 5-ml FACS tube containing 1 ml of RPMI (Roswell Park Memorial Institute) medium supplemented with 10% FCS, 1% l-glutamine (Gibco), 100 U ml −1 of penicillin-streptomycin, 1% nonessential amino acids, 1% sodium pyruvate and 0.05 mM 2-mercaptoethanol. After washing in PBS, cells were resuspended in RPMI and 1% FCS. Then, 1,000-50,000 cells were allowed to adhere for 1 h at 37 °C to coverslips coated with 5 μg cm −2 of poly(d-lysine) (Sigma-Aldrich). Cells were then washed in PBS, fixed in 4% paraformaldehyde for 10 min and permeabilized/blocked in PBS, 0.2% Triton X-100, 2% FCS, 2% rat serum, 2% goat serum and 2% donkey serum. F-actin staining was performed at 4 °C overnight in PBS, 0.1% Triton X-100 and 2% FCS with phalloidin AF405plus (Life Technologies). Samples were washed in PBS and mounted with Prolong Antifade Gold mounting medium. Images were acquired by spectral confocal microscope (Zeiss LSM 880) with a ×63 objective and analyzed with ImageJ v.1.52p software. For the cell morphology, a binary image was created for each individual cell based on the F-actin staining. The circularity index ((4π × area)/(perimeter × 2)) was then calculated with the adequate function of the software.

Statistical analysis
No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications 20,27 . Data distribution was assumed to be normal but this was not formally tested. All quantifications were performed with awareness of experimental groups, meaning not in a blinded fashion. Animals were matched in age and gender between experimental groups, without randomization. No animals or data were excluded. Statistical parameters including the definitions and exact value of n (number of biological replicates and total number of experiments) and the types of statistical tests are reported in the figures and corresponding legends. Statistical analyses were performed using Prism v.8.1.2 (GraphPad Software) or R v.4.0.3 statistical programming language. Statistical analysis was conducted on data with at least three biological replicates. Comparisons between groups were planned before statistical testing and target effect sizes were not predetermined. Error bars displayed on graphs represent the mean ± s.e.m. Statistical significance was defined as: * P < 0.05, ** P < 0.01, *** P < 0.001 and **** P < 0.0001.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The scRNA-seq data have been deposited in the GEO repository under accession no. GSE196720. All other data generated or analyzed during the present study are included in this report (and its Supplementary  Information files). Source data are provided with this paper.
Extended Data Fig. 4 | Spectral flow cytometry-based unsupervised characterization of the expression pattern of tdT and GFP in splenocytes of ZeST mice. a, Unsupervised dimensional reduction, and cell clustering, for the analysis by spectral flow cytometry of the expression of surface markers on CD45 + splenocytes from ZeST mice and one control C57BL/6 mouse, without considering tdT and GFP signals. The UMAP was calculated for all samples together, with downsampling to 200,000 CD45 + cells/sample, and is shown for each infection time point with individual mice pooled together (ZeST not infected, n = 2; 36 h, n = 4; 48 h, n = 3; control C57BL/6 mouse, not infected, n = 1). The cluster color code represents related cell types in similar colors (lower right, see also panel b). b, Cell cluster annotation for cell type identities, based on cell surface marker expression. The mean fluorescence intensity (MFI) of each cluster for each marker was calculated by averaging cluster MFI values across the 10 mice analyzed, and used for hierarchical clustering (heatmap). The corresponding expression patterns were then used to assign cell cluster to indicated cell types. c, Percent of each cluster within GFP + cells, shown first as mean percent of each cluster within GFP + cells across all ZeST mice, and then for each mouse. The color code of the clusters is the same as in (a, b) and their ordering in each pie chart shown in the color key (lower right). d, GFP MFI (mean on scaled data, Y-axis) for each ZeST mouse (one dot per mouse) and each cluster (X-axis, with same color code and ordering as in (a, c, d). For each cluster, the mean ± SEM of the MFI across all ZeST mice is shown as the black lines. For comparison, the autofluorescent signal in the GFP channel in each cell cluster in the C57BL/6 mouse is shown as a black square. e, Percent of each cluster within tdT + cells, designed as in (c). f, tdT MFI for each cluster and each ZeST mouse, designed as in (d). The data shown are from one experiment representative of two independent ones.