The spleen contains a myriad of conventional dendritic cell (cDC) subsets that protect against systemic pathogen dissemination by bridging antigen detection to the induction of adaptive immunity. How cDC subsets differentiate in the splenic environment is poorly understood. Here, we report that LTα1β2-expressing Rorgt+ ILC3s, together with B cells, control the splenic cDC niche size and the terminal differentiation of Sirpα+CD4+Esam+ cDC2s, independently of the microbiota and of bone marrow pre-cDC output. Whereas the size of the splenic cDC niche depended on lymphotoxin signaling only during a restricted time frame, the homeostasis of Sirpα+CD4+Esam+ cDC2s required continuous lymphotoxin input. This latter property made Sirpα+CD4+Esam+ cDC2s uniquely susceptible to pharmacological interventions with LTβR agonists and antagonists and to ILC reconstitution strategies. Together, our findings demonstrate that LTα1β2-expressing Rorgt+ ILC3s drive splenic cDC differentiation and highlight the critical role of ILC3s as perpetual regulators of lymphoid tissue homeostasis.
Host protection requires continuous detection and response to an overwhelming myriad of pathogenic insults. By their widespread tissue distribution and unrivaled capacity to recognize danger-associated signals and process and present pathogen-derived antigens, conventional dendritic cells (cDCs) are essential components against disease. Tissue cDCs derive from bone marrow pre-cDC progenitors that circulate in the bloodstream and continuously seed tissues (Liu et al., 2009; Naik et al., 2007). Whereas pre-cDC commitment to cDC1 or cDC2 fate seems to begin in the bone marrow (Grajales-Reyes et al., 2015; Schlitzer et al., 2015), terminal differentiation requires the integration of tissue-specific cues that lead to the emergence of unique tissue-specific cDC features (Bosteels et al., 2020; Sichien et al., 2017).
In the spleen, XCR1+ cDC1s are a relatively homogenous population that excels at cross-presentation (den Haan et al., 2000; Lehmann et al., 2017; Schnorrer et al., 2006). In contrast, Sirpα+ cDC2s, which preferentially prime CD4+ T cells, are phenotypically, transcriptionally, and functionally heterogeneous (Dudziak et al., 2007; Lehmann et al., 2017; Vander Lugt et al., 2014). Two main cDC2 subsets can be distinguished, Sirpα+CD4+Esam+ cDC2s, which play pivotal roles in T helper type 17 cell (Th17 cell) polarization (Lewis et al., 2011; Satpathy et al., 2013), and Sirpα+CD4−Esam− cDC2s, which appear to be specifically involved in Th2 cell fate decisions (Tussiwand et al., 2015). While the transcription factors involved in the commitment to these disparate cDC fates are partially known (Murphy et al., 2016), the cellular and molecular signals that instruct their expression remain largely unidentified.
To date, two main signals have been identified that control cDC development and differentiation in the spleen. Delta-like 1 (DLL1)–Notch2 and lymphotoxin (LTα1β2)–LTβ receptor (LTβR) interactions are required for the development and/or homeostasis of Sirpα+CD4+Esam+ cDC2s (Abe et al., 2003; Briseño et al., 2018; Fasnacht et al., 2014; Kabashima et al., 2005; Lewis et al., 2011; Satpathy et al., 2013; Wang et al., 2005; Wu et al., 1999). Regarding the latter ligand-receptor pair, cDC intrinsic LTβR expression and signaling regulates local CD4+ cDC2 proliferation (De Trez et al., 2008; Kabashima et al., 2005; Satpathy et al., 2013). While the origin of membrane-bound heterotrimeric LTα1β2 seems to be hematopoietic (Wu et al., 1999), its precise source remains controversial. Using mixed chimeric systems, evidence was obtained that LTα1β2-expressing B cells were critical in splenic cDC2 homeostasis (Kabashima et al., 2005). Other reports questioned whether the B cell requirement was direct since gross abnormalities in lymphoid tissue architecture are common to B cell– and LTα1β2-deficient states, potentially leading to secondary defects in cDC homeostasis (Crowley et al., 1999; Moseman et al., 2012; Phan et al., 2009; Wu et al., 1999; Zindl et al., 2009).
LTα1β2 is expressed by multiple hematopoietic cells. Among these, innate lymphoid cells (ILCs) might constitute an alternative source of ligand for cDC homeostasis (Vivier et al., 2018), as it is now well appreciated that ILCs and cDC communicate extensively. For example, cDC1-derived IL-12 activates ILC1s for early control of toxoplasma and viral infections (Klose et al., 2014; Weizman et al., 2017); cDC2-derived IL-23 activates ILC3s coordinating mucosal immunity against bacterial infections (Cella et al., 2009; Kinnebrew et al., 2012).
While the data highlighted above indicate unidirectional communication between cDCs and ILCs, in which the former instructs the function of the latter, they fostered the hypothesis of reverse communication events whereby ILCs could directly influence cDC development and/or homeostasis. In agreement with this hypothesis, we show here that Rorgt+ ILC3s are a critical, nonredundant source of LTα1β2 that regulates the size of the splenic cDC niche and the terminal differentiation of Sirpα+CD4+Esam+ cDC2s.
Results and discussion
Splenic cDC development is temporally asynchronous
The splenic cDC compartment is highly heterogeneous, comprising multiple subsets that are not only transcriptionally and phenotypically diverse (Fig. 1 A) but also functionally distinct (den Haan et al., 2000; Dudziak et al., 2007; Sathe and Shortman, 2008; Schnorrer et al., 2006; Vander Lugt et al., 2014). To unravel the origin of such heterogeneity, we examined early developmental stages of the murine spleen. We noticed that the cDC compartment did not emerge fixed but matured over time. While cDC numbers steadily rose as mice aged and the spleen increased in size (Fig. 1 B, left), XCR1+ cDC1 and Sirpα+ cDC2 subsets did it at different rates (Fig. 1 B, middle). Indeed, the rate at which Sirpα+CD4+Esam+ cDC2 numbers increased was so pronounced that the early over-representation of cDC1s quickly disappeared, giving place to a splenic cDC compartment dominated by the former cells, which seemed to arrive at its mature representation around post-natal day 21 (Fig. 1 B, right).
The perinatal period represents a critical window in development, in which the first microbial encounters set the stage for subsequent immune homeostasis and host–pathogen interactions (Gensollen et al., 2016). Hence, we hypothesized that microbial colonization could be responsible for the observed alterations in splenic cDC ratios. To test this hypothesis, we analyzed mice raised in germ-free conditions or under broad antibiotic treatment. Neither approach modified the numbers of cDCs isolated from spleens (Fig. 1 C), suggesting that microbial exposure has no/minimal impact on splenic cDC development.
Tissue cDCs develop in situ from bone marrow–derived pre-cDC precursors (Cabeza-Cabrerizo et al., 2019; Grajales-Reyes et al., 2015; Liu et al., 2009; Naik et al., 2007; Schlitzer et al., 2015). Given that cDC1 and cDC2 have different immediate precursors, Ly6G−CD11cintMHC-II−Flt3L+SirpαintSiglecH−Ly6C− pre-cDC1s and Ly6G−CD11cintMHC-II−Flt3L+SirpαintSiglecH−Ly6C+ pre-cDC2s, respectively (Fig. 1 D; Schlitzer et al., 2015), we considered the hypothesis that the temporal evolution we observed in mature splenic cDCs resulted from changes in their upstream precursors. Analysis of pre-cDCs by flow cytometry in either bone marrow or spleen revealed only mild changes throughout our observation period (Fig. 1 D); as these did not coincide with the kinetics of mature cDCs, we suggest that changes neither in pre-cDC bone marrow output nor in splenic “infiltration” account for the final composition of the splenic cDC compartment.
Taken together, these data suggested that splenic cDC development was programmed within the spleen independently of inputs from the bone marrow or the microbiota. More importantly, it seemed that the intrasplenic capacity to support cDC development varied in time.
ILCs regulate splenic Sirpα+CD4+Esam+ cDC2 development
cDCs display diverse phenotypes across and within tissues, suggesting significant tissue adaptation (Sichien et al., 2017). In this regard, it is noteworthy that cDC subsets are not randomly distributed but seem to localize to distinct niches in several tissues (Baptista et al., 2019; Calabro et al., 2016; Dudziak et al., 2007; Gatto et al., 2013; Gerner et al., 2012; Yi and Cyster, 2013). In the spleen, XCR1+ cDC1s were distributed between the T cell area and the red pulp; Sirpα+CD4−Mgl2+ cDC2s seemed to localize mostly to the red pulp; and Sirpα+CD4+Mgl2− cDC2s localized predominantly to the bridging channels entering the T cell area (Fig. 2 A). These observations raised the hypothesis that splenic cDC development could be regulated by intercellular communications events taking place in dedicated niches. To test this hypothesis, we looked into animals lacking diverse immune cell subsets. Analysis of Rag2−/− mice, which lack B and T cells, revealed a decrease in splenic cDC numbers that affected all subsets equally (Fig. 2 B); this phenotype was the result of B cell deficiency (Fig. S1 A). Surprisingly, analysis of Rag2−/−x γc−/− spleens, which in addition to B and T cell deficiency are also devoid of innate lymphocytes, revealed a further reduction in cDC numbers (Fig. 2 B). This reduction affected Sirpα+CD4+Esam+ cDC2 specifically, leading to an unbalanced cDC compartment (Fig. 2 B). Overall, these data suggested that B cells provide a general niche and/or signal(s) that promote broad splenic cDC development, whereas innate lymphocytes may be specifically required for the development and/or homeostasis of Sirpα+CD4+Esam+ cDC2s.
To establish which type of innate lymphocytes, cytotoxic natural killer (NK) cells versus helper ILCs, promoted splenic Sirpα+CD4+Esam+ cDC2 development, we took advantage of depleting antibodies. Whereas administration of anti-NK1.1 antibodies effectively depleted NK cells, leaving the ILC compartment intact, anti-CD90 antibodies had the opposite effect, eliminating Tbet+ ILC1s, Gata3+ ILC2s, and Rorgt+ ILC3s efficiently while leaving NK cells untouched (Fig. S1, B and C). Depletion of NK cells had no effect on splenic cDCs (Fig. 2 C), a finding further confirmed in NKp46-Cre-DTA transgenic mice that in addition to NK cell deficiency also exhibited partial deficiencies in Tbet+NKp46+ ILC1s and Rorgt+CD4−NKp46+ ILC3s (Fig. S1 D). In contrast, depletion of ILCs led to a specific reduction in the number of Sirpα+CD4+Esam+ cDC2s (Fig. 2 C). These data suggested that ILCs, rather than NK cells, were required for proper development/homeostasis of Sirpα+CD4+Esam+ cDC2s. To determine whether ILCs were also sufficient, we proceeded to reconstitute CD45.2+Rag2−/−x γc−/− mice with ILCs sorted from CD45.1+Rag2−/− congenic mice. 2 wk after adoptive transfer, low numbers of donor-derived ILCs could be readily recovered from the spleen of acceptor mice (Fig. 2 D), correlating with increased retrieval of Sirpα+CD4−Esam− and Sirpα+CD4+Esam+ cDC2s (Fig. 2 E). cDC1s remained unaffected (Fig. 2 E). Of note, in these experiments, all ILC subsets increased concomitantly, precluding an accurate inference as to which ILC subset could be implicated in the homeostasis of Sirpα+CD4+Esam+ cDC2s as all seemed to dose-dependently influence the relatively small but consistent restoration of this cDC2 compartment (Fig. S1 E). Taken together, our data suggest that ILCs (possibly in an ILC subset–specific manner) are necessary and partly sufficient for the homeostasis of splenic Sirpα+CD4+Esam+ cDC2s.
Sirpα+CD4+Esam+ cDC2 development requires ILC3s
To identify which ILC subset(s) was involved in Sirpα+CD4+Esam+ cDC2 development and/or homeostasis, we combined immunostaining, confocal microscopy, and histocytometry (Gerner et al., 2012) to simultaneously assess ILC and cDC spatial distribution (Fig. 3 A). Early analysis showed that ILC subsets were unevenly distributed throughout the spleen. In particular, NK1.1+Eomes− ILC1s and Gata3+ ILC2s localized to the red pulp, whereas Rorgt+ ILC3s were present mostly in bridging channels (Fig. S2 A). These observations suggested that Rorgt+ ILC3s and Sirpα+CD4+ cDC2s were enriched in identical locations and, indeed, intimate contacts between the two cell types could be observed (Fig. 3 A; Hoorweg et al., 2015; Kim et al., 2007; Magri et al., 2014). To formally quantify these observations, we treated the records of intrasplenic ILC and cDC subset localization (Fig. 3 B) as multitype spatial point patterns for the purpose of statistical analysis. First, we calculated the nearest neighbor distance between all ILC and cDC subsets. These analyses revealed that, on average, Rorgt+ ILC3s and Sirpα+CD4+ cDC2s resided in closer proximity to each other than any other ILC–cDC pair (Fig. 3 C and Fig. S2 B). Considering, however, that the proportions of the different cellular subsets were very different and this could bias our results, we took an in silico approach to “regress out” the influence of an overabundant Sirpα+CD4+ cDC2 compartment by sampling at random a number of Sirpα+CD4+ cDC2s equivalent to the least represented cDC subset in our datasets. Remeasuring the nearest neighbor distances showed that, after controlling for differences in cDC abundance, Rorgt+ ILC3s still resided closer to Sirpα+CD4+ cDC2s than to any other cDC subset (Fig. S2 C). To further validate these findings, we compared our results to a random permutation null hypothesis where ILC labels, at their observed representation, were shuffled across their positional coordinates. As illustrated in Fig. S2 D, this strategy eliminated the preferential enrichment between Rorgt+ ILC3s and Sirpα+CD4+ cDC2s in two out of three samples analyzed. Taken together, these analyses suggested that the spatial codistribution pattern of Rorgt+ ILC3s and Sirpα+CD4+ cDC2s could not have emerged by chance alone. Hence, to gain further insight into the spatial cDC organization potentially promoted by Rorgt+ ILC3s, we took Rorgt+ ILC3s as points of origin from which the probabilities of observing cDCs of a given subset were calculated as a function of spatial two-dimensional distance. As before, these probabilities were compared with a random permutation null hypothesis by relabeling of ILCs. While Sirpα− cDC1s and Sirpα+CD4− cDC2s exhibited a tendency to be segregated away from Rorgt+ ILC3s, statistically this seemed to occur by chance. In contrast, Sirpα+CD4+ cDC2s exhibited a higher probability of being near Rorgt+ ILC3s than would be expected (Fig. 3 D and Fig. S2 E). Taken together, these data suggested that ILC3s and Sirpα+CD4+Esam+ cDC2s may engage in bi-directional communication in vivo.
To probe whether the differentiation of Sirpα+CD4+Esam+ cDC2s depended on ILC3s, as the prior data suggested, we analyzed spleens from RorcGFP/GFP mice. In these mice, the presence of two copies of the Rorc-GFP transgene results in constitutive Rorgt deficiency, impairing ILC3 development (Eberl et al., 2004). As compared with Rorc+/+ and RorcGFP/+ littermate controls, RorcGFP/GFP mice exhibited reduced numbers of Sirpα+CD4+Esam+ cDC2s (Fig. 3 E). In this experimental setup, in which B cells are present in normal numbers, XCR1+ cDC1 and Sirpα+CD4−Esam− cDC2 homeostasis remained unaffected (Fig. 3 E). Identical results were obtained in Rag2−/− x RorcGFP/GFP mice, in which deletion of ILC3s incremented Sirpα+CD4+Esam+ cDC2 deficiency without affecting XCR1+ cDC1 and Sirpα+CD4−Esam− cDC2 numbers (Fig. 3 F). Combined, these results suggested that ILC3s specifically regulate Sirpα+CD4+Esam+ cDC2 homeostasis in both lymphoreplete and lymphopenic environments. However, given that the reduction in Sirpα+CD4+Esam+ cDC2 numbers was much more pronounced in Rag2-deficient as compared with Rag2-sufficient mice, i.e., in the absence versus the presence of B cells, it seems that some degree of compensation for ILC3 deficiency may be exerted by Rag2-dependent B cells.
Sirpα+CD4+Esam+ cDC2 development requires ILC3-derived LTα1β2
The preceding data revealed a specific role for ILC3s in the regulation of splenic Sirpα+CD4+Esam+ cDC2 homeostasis. To identify candidates mediating the intercellular communication events between ILC3s and Sirpα+CD4+Esam+ cDC2s responsible for the maintenance of the latter cells, we took a computational approach (Fig. 4 A). First, we screened publicly available transcriptional profiles of splenic cDCs (Brown et al., 2019) to identify differentially expressed genes between the three cDC subsets. As depicted in Fig. 4 B, in which individual genes are plotted as points in a hexagonal diagram containing three axes (one per cDC subset) placed at 120° angles, and where the distances to the center and angle represent the log2 fold induction and the directionality of such induction, respectively, differentially expressed genes overpopulated the horizontal axis representing the global comparison XCR1+ cDC1s versus CD11b+ (Sirpα+) cDC2s. These data, which are further highlighted in the adjacent rose plots depicting the percentage of differentially expressed genes, are consistent with a major bifurcation between cDC1 and cDC2 development as previously demonstrated (Ma et al., 2019; Schlitzer et al., 2015). Comparatively, Tbet− (Sirpα+CD4−Esam−) cDC2s and Tbet+ (Sirpα+CD4+Esam+) cDC2s were closer to each other. Inspection of the genes specifically associated with Tbet+ (Sirpα+CD4+Esam+) cDC2s revealed the presence of several modules related to leukocyte differentiation and intercellular communication (signaling and cell–cell adhesion; Fig. 4 C). This result prompted the use of NicheNet (Browaeys et al., 2020), an algorithm capable of inferring the ligand–receptor interactions occurring between cellular pairs by combining transcriptome data of the interacting cells with a priori knowledge on signaling and gene regulatory networks. Specifically, we applied NicheNet to infer which ligand–receptor pairs expressed by Rorgt+ ILC3s and Tbet+ (Sirpα+CD4+Esam+) cDC2s could regulate the expression of the Tbet+ cDC2–associated genes found. Among the top predicted ligands expressed by ILC3s that could induce the Tbet+ cDC2 gene signature (Fig. 4 D), two signals known to affect splenic cDC homeostasis were present, Dll1 (Fasnacht et al., 2014; Lewis et al., 2011) and Lta (Abe et al., 2003; Kabashima et al., 2005; Wang et al., 2005; Wu et al., 1999). Of note, while Ltb did not appear in the top 20 predicted ligands, it was also present in the list of predicted ligands produced by NicheNet, and moreover, it was highly expressed by ILC3s (Björklund et al., 2016; Gury-BenAri et al., 2016; Magri et al., 2014; Reboldi et al., 2016). Hence, on the basis of absolute ligand and receptor expression on ILC3s and Tbet+ cDC2s, respectively (Fig. 4 D), and of the known biology of ILC3s (van de Pavert and Mebius, 2010), we hypothesized that ILC3-derived LTα1β2 was responsible for Sirpα+CD4+Esam+ cDC2 maintenance.
In support of our hypothesis, we found that cDC2s from Rag2−/− and Rag2−/−x γc−/− mice expressed higher levels of LTβR as compared with WT cDC2s (Fig. 4 E); comparatively, LTβR levels on cDC1s were minorly affected (Fig. 4 E). This increase in receptor expression, which was more pronounced as the degree of Sirpα+CD4+Esam+ cDC2 deficits increased, may represent lack of ligand-induced receptor down-regulation (Kabashima et al., 2005; Lewis et al., 2011; Yi and Cyster, 2013). Most crucially, this observation simultaneously suggested that ILC3s and cDC2s communicate via LTα1β2 and that LTα1β2 signaling regulates Sirpα+CD4+Esam+ cDC2 homeostasis. Consistently, we observed that treatment of mice with LTβR Fc fusion antagonist (LTβR-Fc), a decoy receptor that blocks the biological activity of LTα1β2, specifically reduced Sirpα+CD4+Esam+ cDC2 numbers (Fig. 4 F and Fig. S3 A); conversely, treatment of Rag2−/−x γc−/− mice with an agonistic LTβR antibody distinctively restored this cDC compartment (Figs. 4 G and Fig. S3 B). Given that LTβR-Fc treatment reduced Sirpα+CD4+Esam+ cDC2 numbers in WT and Rag2−/− but not Rag2−/−x γc−/− mice (Fig. 4 F), combined these experiments implicated LTα1β2 signaling as a crucial regulator of Sirpα+CD4+Esam+ cDC2 homeostasis and revealed adaptive and innate lymphocytes as the critical sources of such signals.
By studying mice with B cell–specific transgenic overexpression of LTα1β2 and mice with μMT:Ltα−/− mixed bone marrows, we have previously shown a critical role for B cell–derived LTα1β2 in splenic cDC2 homeostasis (Kabashima et al., 2005); these effects might have been indirect and caused by a defect in lymph node architecture (Crowley et al., 1999; Moseman et al., 2012; Zindl et al., 2009). Hence, to conclusively examine whether ILC3s mediated Sirpα+CD4+Esam+ cDC2 homeostasis via LTα1β2 and distinguish between the function of B cell– and ILC3-derived LTα1β2 in a cell-intrinsic manner, we combined Ltbflox alleles with Cd19-Cre and/or Rorc-Cre transgenes. The resulting mice lacked LTβ expression on B cells (Cd19-Cre; Tumanov et al., 2002), ILC3s and T cells (Rorc-Cre; Kruglov et al., 2013) or both (Cd19-Cre x Rorc-Cre). Analysis of their spleens revealed that LTβ expression on B cells and ILC3s independently regulated the size of the splenic cDC compartment. Both B cell and ILC3 LTβ deficiency induced significant reductions in XCR1+ cDC1, Sirpα+CD4−Esam− cDC2, and Sirpα+CD4+Esam+ cDC2 numbers (Fig. 4 H and Fig. S3 C). Combined B cell and ILC3 LTβ deficiency caused a further contraction specifically in Sirpα+CD4+Esam+ cDC2 numbers (Fig. 4 H and Fig. S3 C). Taken together, these results suggest that LTα1β2 signals with origin in B cells and ILC3s additively (or synergistically) regulate the splenic cDC niche size. Albeit smaller, the resulting niche in Ltbflox/flox x Cd19-Cre and Ltbflox/flox x Rorc-Cre mice seems to adequately support Sirpα+CD4+Esam+ cDC2 differentiation. This process becomes adversely affected only when neither B cells nor ILC3s (Ltbflox/flox x Cd19-Cre x Rorc-Cre mice) can provide LTα1β2. Consistent with this latter finding and thus multiple sources of LTα1β2, treatment of uMT−/− mice with LTβR-Fc provoked a specific reduction in splenic Sirpα+CD4+Esam+ cDC2 numbers (Fig. S3 D). Although Ltbflox/floxx Rorc-Cre mice lack LTβ in both ILC3s and T cells (Kruglov et al., 2013), the role of T cell–derived LTβ in cDC2s homeostasis seems not essential compared with ILC3-derived LTβ, as Rag2−/− and uMT−/− mice had similar phenotypes and LTβR-Fc treatment effectively reduced splenic Sirpα+CD4+Esam+ cDC2 numbers in Rag2−/−, but not Rag2−/−x γc−/−, mice (Fig. 4 F).
Our data show a synergistic role of LTα1β2 expressed on ILC3s and B cells in splenic cDC2 homeostasis. Whereas LTα1β2-expressing B cells are numerous in the spleen, LTα1β2-expressing ILC3s are relatively scant (Schaeuble et al., 2017). While this disparity probably explains why B cells seem to have a bigger cumulative influence in splenic cDC homeostasis as compared with ILC3s, it also suggests that, at the single-cell level, LTα1β2-expressing ILC3s might be more potent regulators of the splenic cDC compartment. Consistently, in various tissues, ILC3s are known to express significantly higher levels of LTα1β2 as compared with B cells (Björklund et al., 2016; Gury-BenAri et al., 2016; Magri et al., 2014; Reboldi et al., 2016). Alternatively, it may be that by acting during embryogenesis in the promotion of lymphoid tissue development (van de Pavert and Mebius, 2010), ILC3s, which include lymphoid tissue inducer cells, affect the ability of B cells to communicate with developing cDC and thus secondarily limit the provision of B cell–derived LTα1β2. In favor of this hypothesis, combined LTβ deficiency in B cells and ILC3s indeed does not grossly increase the deficits in cDC1s and Sirpα+CD4+Esam− cDC2s observed in either B cell– or ILC3-specific LTβ-deficient hosts, and neither LTβR blockade nor agonism in adulthood is able to influence the numbers of these two cDC subsets. These observations imply that the splenic cDC niche is formed early on and requires LTα1β2 signals provided by both ILC3s and B cells that act on the same developmental pathway without built-in redundancy. Once formed, however, the cDC niche becomes LTα1β2-insensitive.
In contrast to this temporally restricted prerequisite, Sirpα+CD4+Esam+ cDC2s required LTα1β2 continuously as numbers of these cells were influenced by both LTβR blockade and LTβR agonism well into adulthood. Whereas also here B cells and ILC3s seemed to be the major providers of LTα1β2, no single source seemed essential, and both cells were capable of compensating for the lack of the other.
In Rag2−/− x γc−/−, Rag2−/− x RorcGFP/GFP, and Ltbflox/flox x Cd19-Cre x Rorc-Cre mice, residual numbers of Sirpα+CD4+Esam+ cDC2s persist. Whether these cells are identical to those found in WT conditions or represent precursors arrested in their differentiation process due to lack of sufficient input from B cells and ILC3s remains to be determined. Importantly, however, these cells seemed to develop independently of LTα1β2 signaling as they were not only insensitive to LTβR-Fc treatment but also present in Ltb−/− mice (data not shown). Although we clearly found ILC3s and Sirpα+CD4+Esam+ cDC2s to occupy the same spatial niche, the bridging channel, and illustrated how several ILC3-expressed molecules have the potential to modulate gene expression in cDC2s, an additional factor provided by nonlymphocytes hence seems to quick-start cDC2 differentiation. This factor is likely to be stromal cell–derived DLL1 (Fasnacht et al., 2014), as Notch2 and LTβR deletions seem to target the same cDC populations in spleen and small intestine (Briseño et al., 2018; Fasnacht et al., 2014; Lewis et al., 2011; Satpathy et al., 2013).
Splenic white pulp development requires LTα1β2 (Fu et al., 1998; Gonzalez et al., 1998; Kim et al., 2007; Ngo et al., 2001; Schaeuble et al., 2017; Tumanov et al., 2002). We have previously shown that LTβR expression in hematopoietic cells, rather than in radioresistant stromal cells, is essential for proper development of splenic CD4+ cDC2s (De Trez et al., 2008). In those experiments, reconstitution of the CD4+ cDC2 compartment occurred in spite of abrogated Ccl19, Ccl21, and Cxcl13 levels in LTβR-deficient recipients (De Trez et al., 2008). These results, which suggest at least partial independence between LTα1β2-mediated white pulp and cDC development, indicate that the traditional dichotomy between T cell area and B cell follicle stromal cells may not be at play in cDC development and thus raise the question as to how ILC3s and Sirpα+CD4+Esam+ cDC2s come to localize in the same niche. Given our previous work and that of others, we speculate that the chemoattractant receptor Ebi2 might be implicated in this process. Ebi2 was shown to be involved both in the localization of ILC3s in intestinal cryptopatches and isolated lymphoid tissues (Chu et al., 2018; Emgård et al., 2018) and in cDC2 distribution in splenic bridging channels and the B cell follicle/T cell zone interface (Gatto et al., 2013; Lu et al., 2017; Yi and Cyster, 2013). Similarly, CD4+ T cells, which are predominantly activated by cDC2s (Baptista et al., 2019; Dudziak et al., 2007; Eickhoff et al., 2015; Vander Lugt et al., 2014), congregate near the latter cells in an Ebi2-dependent manner (Baptista et al., 2019; Li et al., 2016). Consistent with this hypothesis, loss of CD4+ cDC2s in Ebi2-deficient mice could be rescued by LTβR agonism (Yi and Cyster, 2013), suggesting that Ebi2 expression on cDCs controls access to LTα1β2. Alternatively, CCL20–CCR6 interactions might be involved. Both ILC3s and cDC2s express CCR6, and LTα1β2 leads to the induction of CCL20 (Kucharzik et al., 2002; Rumbo et al., 2004).
In summary, we showed here that splenic cDC homeostasis depends on both LTα1β2-expressing ILC3s and B cells. Synergism governs the overall size of the splenic cDC compartment and redundantly determines the terminal differentiation of Sirpα+CD4+Esam+ cDC2s. Given the strong bias exhibited by the different cDC subsets in the induction of distinct T cell responses (den Haan et al., 2000; Dudziak et al., 2007; Lehmann et al., 2017; Schnorrer et al., 2006; Vander Lugt et al., 2014), our observations regarding the ability to manipulate the splenic Sirpα+CD4+Esam+ cDC2 compartment in adulthood have relevance for the design of therapeutic approaches that aim to target CD4+ T cell activity.
Materials and methods
Specific pathogen–free (SPF) WT C57BL/6 mice were obtained from Janvier Laboratories. Germ-free WT mice were bred and maintained at the University of Ghent gnotobiotic facility. B6.129S2-Ighmtm1Cgn/J (uMT−/−), B6(Cg)-Rag2tm1.1Cgn/J (Rag2−/−), B6.129P2-Gt(ROSA)26Sortm1(DTA)Lky/J (DTA), B6.129P2(Cg)-Rorctm2Litt/J (RorgtGFP), B6.129P2(C)-Cd19tm1(cre)Cgn/J (CD19-Cre), and B6.FVB-Tg(Rorc-Cre)1Litt/J (Rorc-Cre) mice were obtained from The Jackson Laboratory. B6.B10(Cg)-Rag2tm1FwaIL2rgtm1Wjl (Rag2−/− x γc−/−) mice were purchased from Taconic Farms. B6(Cg)-Ncr1tm1.1(iCre)Viv (NKp46-Cre; Narni-Mancinelli et al., 2011) and B6(Cg)-Ltbtm1Avt (Ltbflox; Tumanov et al., 2002) mice were described previously. All animals were maintained in SPF conditions at accredited animal facilities either at the VIB-UGhent or at the University of Texas at San Antonio. Housing conditions entailed individually ventilated cages in a controlled day–night cycle and food and water ad libitum. Unless stated otherwise, animals were enrolled in experiments between 8 and 12 wk of age. Both genders were used with no gender-specific effects being noticed. Animals were randomized into the different experimental groups blindly. Experiments were approved by either the VIB-UGhent ethical review board or the University of Texas at San Antonio Institutional Animal Care and Use Committee in accordance with the specific local legislation.
In vivo treatments
Mice were treated with broad-spectrum antibiotics (1 mg/ml ampicillin, 1 mg/ml neomycin, 1 mg/ml gentamycin, 1 mg/ml metronidazole, and 0.5 mg/ml vancomycin) targeting Gram-negative and -positive bacteria and protozoans in the drinking water (sugar supplemented to increase palatability) for 14 d.
Antibody-mediated cell depletion
To deplete NK cells or ILCs, 300 μg of rat anti-mouse NK1.1 (clone PK136; Bioceros) or 300 μg of rat anti-mouse CD90 (clone YTS154; Bioceros) was administered, respectively. Littermate controls were treated with an isotype control antibody (rat IgG2a, clone RG7/1.30; BioXCell).
To block LTα1β2 activity in vivo, mice were treated with an LTβR-Fc decoy receptor (Browning et al., 1997). Mice were injected intraperitoneally with 100 μg of this fusion protein on days 0, 3, 7, and 10 and analyzed on day 12 or 13. Control mice received identical injections of PBS.
LTβR agonism was induced by intraperitoneal injection of 100 μg of an agonistic rat anti-mouse LTβR antibody (clone 4H8; Bioceros), on days 0, 3, 7, and 10. Mice were sacrificed on day 12 or 13. Control mice received 100 μg of an isotype control antibody (rat IgG2a, clone RG7/1.30; BioXCell).
Adoptive ILC transfer
To expand ILCs in vivo, Rag2−/− mice were treated intraperitoneally with a mixture of 2.5 μg recombinant human IL-7 (StemCell Technologies) and 15 μg mouse anti-human IL-7 antibody (clone M25; BioXCell) in PBS every other day for 1 wk.
ILC isolation and transfer
On day 6 of rhIL-7:anti–IL-7 treatment, the spleen, mesenteric lymph nodes, and small intestine were processed into single-cell suspensions, which were stained for FACS sorting. ILCs were sorted as live CD45+Lineage(CD3e, CD19, NK1.1)− Ly6G−CD11c−CD90hiCD127+ cells on a FACSAria II (BD Biosciences). Approximately 4 × 105 ILCs were adoptively transferred via intrasplenic injection. Briefly, acceptor mice were anesthetized with isoflurane, the flank skin and peritoneum above the spleen were opened, and 100 μl of ILCs in PBS was injected directly into the spleen with a 29-G needle. The peritoneum was closed with surgical thread and the skin stapled. Buprenorphine analgesia (0.1 mg/kg) was administered subcutaneously. Littermate control mice underwent the same surgical procedure but received only PBS. Acceptor mice were analyzed 14 d after adoptive transfer.
Spleens were enzymatically digested with collagenase IV (3 mg/ml; Worthington) and DNase1 (40 μg/ml; Sigma-Aldrich) for 30 min at 37°C. Digestion was stopped by adding ice-cold PBS, and cell suspensions were filtered through a 70-µm nylon mesh. Bone marrow single-cell suspensions were obtained by flushing femurs with PBS over 70-µm nylon mesh filters. Erythrocytes were lysed in a solution of ammonium chloride (10 mM KHCO3, 155 mM NH4Cl, and 0.1 mM EDTA). To obtain cell counts, samples were spiked with counting beads (123count eBeads; Thermo Fisher Scientific).
Staining, acquisition, and analysis
Single-cell suspensions were first incubated with fixable viability dyes (eFluor506 or eFluor780; eBioscience) to identify dead cells and with an FcγRII/III antibody (clone 2.4G2) to limit nonspecific antibody binding. After washing, cells were incubated with mixtures of fluorescently and/or biotin-labeled antibodies for 30 min at 4°C. The antibodies used are listed in Table S1. When biotin-labeled antibodies were used, a second surface-staining step with fluorescently labeled streptavidin was included. For intracellular staining of transcription factors, cells were fixed using the Foxp3 fixation/permeabilization kit (eBioscience) per the manufacturer’s instructions after surface staining. Samples were acquired on an LSRFortesssa cytometer (BD Biosciences) and analyzed using FlowJo software (BD Biosciences).
Immunofluorescence and image analysis
Spleens were fixed in 1% paraformaldehyde, dehydrated in 30% sucrose, and frozen in optimal cutting temperature medium. 10-µm sections were prepared and stained overnight at 4°C. Antibodies are listed in Table S2.
Image acquisition, processing, and histocytometry
Multicolor images were acquired on a Zeiss 880 tiling confocal microscope equipped with 25×/0.8 numerical aperture and 40×/1.4 numerical aperture immersion oil objectives. The resulting images were corrected for fluorochrome spillover in ImageJ and imported into Imaris for cellular visualization and segmentation. cDCs were visualized by selecting voxels with CD11c signal—the Imaris’s surface creation wizard provided automatic thresholds around positively stained cells expediting this operation—and masking all other parameters of interest within using channel arithmetic operations in the ImarisXT extension. DC surfaces were created on the resulting CD11c channel using Imaris’s surface creation module, which employs a watershed algorithm to define individual cells. ILC1s surfaces were segmented on NK1.1 signal and filtered for the absence of nuclear Eomes. ILC2s and ILC3s were segmented directly on Gata3 and Rorgt signals, respectively.
Computational analysis of RNA sequencing data
Publicly available transcriptomic data regarding splenic cDCs and intestinal ILC3s were retrieved from GEO under accession no. GSE130201 (Brown et al., 2019) and accession no. GSE109125 (Immunological Genome Project Consortium), respectively. Here, we reanalyzed only bulk RNA sequencing profiles using the published count matrixes using R/Bioconductor.
Differential gene expression
To identify genes specifically associated with Tbet+ (Sirpα+CD4+Esam+) cDC2s, data were processed with edgeR (Robinson et al., 2010) for normalization and limma-voom (Ritchie et al., 2015) for pairwise differential expression testing. Genes were considered differentially expressed if log2 fold change > 1 or less than −1 (adjusted P value < 0.05, Benjamini–Hochberg correction for multiple testing). Of note, one Tbet− (Sirpα+CD4−Esam−) cDC2 sample was removed from the analysis due to aberrant clustering as also seen in the original publication (Brown et al., 2019). For visualization, we used the package Triwise (van de Laar et al., 2016) reducing the three-dimensional dataset into a two-dimensional barycentric coordinate system displaying the log2 mean gene expression values of each cell type.
Gene ontology (GO) analysis
To obtain higher-order insights into the potential physiology of the Tbet+ (Sirpα+CD4+Esam+) cDC2 gene signature, we used clusterProfiler (Yu et al., 2012) to identify overrepresented GO terms. GO term redundancy was eliminated by calculating the semantic similarity between GO terms (Yu et al., 2010).
To infer how ILC3s communicate with Tbet+ (Sirpα+CD4+Esam+) cDC2s, we used NicheNet (Browaeys et al., 2020). We assumed that the Tbet+ cDC2 gene signature identified during differential gene expression testing partly represented the influence of ILC3s on Tbet+ cDC2 gene expression; and we defined the sets of ligands and receptors expressed by ILC3s and Tbet+ cDC2s, respectively, using edgeR. Subsequently, these data were used in NicheNet, using default parameters, to prioritize the ILC3-expressed ligands that explain the target gene signature. Potential Tbet+ cDC2 receptors were identified by querying the NicheNet’s built-in ligand-receptors prior models considering only bona fide interactions. To select hits for experimental validation, we further considered the absolute mRNA levels detected for each of the ligand-receptor pairs identified—log2(cpm).
Statistical analyses were performed with GraphPad Prism (GraphPad Software). The statistical methods used in each analysis are mentioned in the corresponding figure legends. Data are presented as bars representing the mean ± standard deviation; dots represent individual measurements. n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.
Online supplemental material
Fig. S1 depicts the role of B cells, NK cells, and ILCs in splenic cDC homeostasis; Fig. S2 highlights the spatial localization of ILC and cDC subsets; and Fig. S3 shows the dependency of cDCs on LTα1β2 signaling. Table S1 and Table S2 contain information regarding the antibodies used in this study.
We thank all members of the Laboratory of Immunoregulation and Mucosal Immunology (VIB-UGhent Center for Inflammation Research) for intellectual input during the course of these studies and the VIB Center for Inflammation Research Flow Cytometry and Microscopy Core Facilities for continuous assistance.
M. Vanderkerken is supported by a fellowship from Fonds Wetenschappelijk Onderzoek Vlaanderen; A.P. Baptista is supported by a Marie-Sklodowska Curie Action fellowship as part of Horizon 2020; C.L. Scott is supported by the Fonds Wetenschappelijk Onderzoek Vlaanderen and a European Research Council starting grant; Y. Saeys is supported by the Fonds Wetenschappelijk Onderzoek Vlaanderen and the Marylou Ingram Scholar program; H. Hammad is supported by a research initiative grant from Ghent University; A.V. Tumanov is supported by grants from the National Institutes of Health (AI135574) and the Max and Minnie Tomerlin Voelcker Fund; and B.N. Lambrecht is supported by a European Research Council advanced grant, a research initiative grant from Ghent University, and an Excellence of Science research grant.
Author contributions: M. Vanderkerken designed, performed, and analyzed experiments and wrote the paper; A.P. Baptista conceptualized the research, designed, performed, and analyzed experiments, and wrote the paper; M. De Giovanni designed, performed, and analyzed experiments; S. Fukuyama, R. Browaeys, and C.L. Scott performed experiments; P.S. Norris, G. Eberl, J.P. Di Santo, E. Vivier, and Y. Saeys provided key reagents/materials; H. Hammad and J.G. Cyster designed and analyzed experiments; C.F. Ware and A.V. Tumanov designed experiments and provided key reagents; and C. De Trez and B.N. Lambrecht conceptualized the research, designed experiments, analyzed the data, and wrote the paper. All authors reviewed the paper.
Disclosures: E. Vivier is an employee of Innate Pharma. C.F. Ware reported grants from Capella Biosciences, grants from Eli Lilly, and grants from Boehringer Ingelheim Pharmaceuticals outside the submitted work; in addition, C.F. Ware had a patent to USP 8,974,787 issued and a patent to USP 8,349,320 issued. No other disclosures were reported.
M. Vanderkerken, A.P. Baptista, and M. De Giovanni contributed equally to this paper.
A.V. Tumanov, C. De Trez, and B.N. Lambrecht contributed equally to this paper.