Stromal cells have for a long time been viewed as structural cells that support distinct compartments within lymphoid tissues and little more. Instead, an active cross-talk between endothelial and fibroblastic stromal cells drives the maturation of lymphoid niches, a relationship that is recapitulated during lymph node organogenesis, steady-state conditions, and following inflammation. In this review, we go over recent advances in genetic models and high-resolution transcriptomic analyses that have propelled the finer resolution of the stromal cell infrastructure of lymph nodes, revealing that the distinct subsets are strategically positioned to deliver a catered mixture of niche factors to interacting immune cell populations. Moreover, we discuss how changes in the activation state of poised stromal cell–underpinned niches rather than on-demand differentiation of new stromal cell subsets govern the efficient interaction of Ag, APC, and cognate B and T lymphocytes during adaptive immune responses.
Lymph nodes (LN) are situated at convergence points of afferent lymphatic vessels to monitor lymph-borne Ags draining from local tissues. LN stromal cells (LNSCs) comprise distinct cell types: blood endothelial cells (BECs), lymphatic endothelial cells (LECs), and fibroblastic reticular cells (FRCs), which in turn embody distinct subsets. Spatially distributed stromal cell subsets secrete a unique array of niche factors, creating regional microenvironments that compartmentalize the distribution and interaction of Ag, APC, and leukocytes in the LN. Advances in genetic models targeting LNSCs combined with high-dimensional molecular analyses have propelled the resolution of cell types involved in LN development as well as the molecular identity and immunological function of distinct LNSC types, which are reviewed in this work.
The organogenesis of LNs relies on the interaction of nonhematopoietic lymphoid tissue organizer (LTo) cells and lymphoid tissue inducer (LTi) cells in the embryonic LN anlagen (1). Recent studies have unveiled that the sequential activation of different LTo cell subtypes is mandatory for the development of the LN anlage and the formation of the characteristic LN infrastructure (reviewed in Refs. 2, 3). Redefining the identity of LTo cells has helped not only to shed light on the sequence of events in LN formation but also suggested potential progenitor cell populations for the different subsets of LNSCs in adult LNs. Prior to LN anlage initiation, LTi cells migrate through lymphatic vessels of the embryo and are attracted to and retained at vascular intersections by specialized RANK+ CCL21+ lymphatic endothelial LTo cells, an initial and decisive step for LN organogenesis (4). The retention of sufficient LTi cell numbers in the lymphatic bed promotes their prolonged interaction with local mesenchymal and blood endothelial LTo cells, allowing for the lymphotoxin-β receptor (LTβR)–dependent activation of adhesion molecules (5) and CCL19 and CCL21 chemokine expression (4, 6) and, ultimately, LN development. FRC subset differentiation from fibroblast activation protein (FAP)–expressing LTo cells in LN anlagen has been documented using cell fate mapping (7). Several mechanisms that trigger the activation of mesenchymal LTo cells have been described, such as retinoic acid–mediated and interstitial lymph fluid flow–induced CXCL13 expression (8, 9). The conditional deletion of Ltbr or abrogation of the alternative NF-κB pathway in Ccl19-Cre– or Cxcl13-Cre–expressing mesenchymal LTo cells abrogated FRC differentiation and niche maturation (4, 10), as did the genetic abrogation of the canonical NF-κB pathway in Ccl19-Cre–expressing cells (11). Moreover, effectors of Hippo signaling, YAP and TAZ, have recently been shown to govern FRC maturation in concert with LTβR signaling (12). Importantly, FRC differentiation (13) and the maturation of high endothelial venules (HEVs) (10, 14) require constant LTβR signaling throughout postnatal life. Overall, these findings indicate that three LTo cell subtypes (lymphatic, blood endothelial, and mesenchymal LTo cells) determine LN organogenesis (Fig. 1A) and form the progenitor populations of the different LNSC types and subsets (Fig. 1B). Future studies are needed for the further resolution of lineage trajectories and differentiation programs of LTo cells and their progeny in adult LNSC subsets.
LNSC subset heterogeneity
LNSC subsets create distinct microenvironments that cater to the efficient interaction and activation of immune responses (15). Although heterogeneity within FRCs is well recognized, all LNSC types comprise distinct subsets (Fig. 1B). In recent years, single-cell transcriptomic analyses paired with high-resolution imaging and cell-specific genetic targeting have helped to reveal heterogeneity in all LNSC types, which correlate to their anatomical location and immunological function.
LECs are the first LNSC type to interact with lymph-borne material, forming a physical barrier between the lymph and the LN parenchyma. Peripheral Ags, as well as effector and memory lymphocytes, activated dendritic cells (DCs), and monocytes drain via the afferent lymphatics into the subcapsular sinus (SCS), which envelopes the LN cortex and meets with the medullary sinuses collecting at the hilum of the LN. Medullary sinuses eventually converge into a single efferent lymphatic vessel that further transports Ag and leukocytes to the next LN. Additionally, the LN cortex is penetrated by blind-ended cortical sinuses that converge with medullary sinuses and help to direct egressing lymphocytes (16). Given the spatial localization of these sinuses, the interaction with entering or egressing lymph-borne material and parenchymal cells, it comes as no surprise that heterogeneous LEC subsets form LN sinuses (Fig. 1C). The SCS, the first lymph Ag-sampling zone, is lined by two distinct LEC subsets: floor LECs (fLECs) and ceiling LECs (cLECs). fLECs line the SCS and support the transport of Ags into the LN parenchyma. Expression of plasmalemma vesicle–associated protein (PLVAP) and caveolin-1 (CAV1) by fLECs forms size-exclusion diaphragms that control the entry of small molecular mass Ags into the conduit system descending from the SCS floor (17). Alternatively, larger Ags are shuttled directly to parenchymal B cells by CD169+ SCS macrophages (18) that are maintained via the secretion of CSF-1 by fLECs (19). cLECs line the LN capsule and express the atypical chemokine receptor ACKR3 to scavenge CCL21 and direct lymph-draining APCs into the LN parenchyma (20). In addition to differential expression of chemokines and niche factors, fLECs and cLECs are also distinguished by their expression of the lymphatic vessel endothelial hyaluronan receptor-1 (LYVE-1), which is expressed by fLECs but not cLECs (20).
Phenotypic and functional studies suggest that at least two main processes occur in medullary sinuses: lymphocyte egress and myeloid cell accumulation. Indeed, LECs are a major source of the sphingosine-1–phosphate (S1P) gradient that mediate the egress of activated lymphocytes from LNs via the medullary sinuses (21, 22). LEC-produced S1P and CSF-1 gradients also promote NK cell and medullary sinus macrophage retention, respectively (19, 23), and the expression of CD209 by medullary LECs may support neutrophil accumulation in human LNs (24). Consistent with these divergent functions, single-cell RNA–sequencing (scRNA-seq) studies reveal the presence of at least two subsets of medullary LECs (25, 26). In murine LNs, pentraxin 3 (Ptx3)–expressing LECs occupy the central and paracortical areas of medullary sinuses and express genes involved in lymphocyte egress. In contrast, macrophage receptor with collagenous structure (MARCO)–expressing LECs line the perifollicular medullary sinuses and express genes associated with innate immunity, complement cascades, and coagulation cascades. Transcriptomic comparisons of human and mouse LN LECs show a high degree of conservation with two medullary LEC subsets, as well as ceiling and floor SCS LECs (24, 25). Cross-species conservation of LN LEC subsets reflects the immunological relevance of spatially distributed, specialized LEC subsets for the coordinated trafficking of Ag and leukocytes.
The vascular tree transverses LNs as arteries entering the hilum branching into arterioles and then capillaries that penetrate the LN cortex and converge as HEVs before connecting to veins in the medulla (Fig. 1C). Recent transcriptional analyses have revealed regional heterogeneity of LN BECs according to their location along the vascular tree (27). For example, arteriole and prearteriole endothelial cells (ECs) express genes induced by laminar shear flow as well as the gene Ly6c, the expression of which is also shared by capillaries. Five clusters of capillary cells were identified based on their expression of previously identified genes: Cdh13, Emcn, Gja1, and Gpihbp1 (28). The most abundant cluster, type I capillary ECs, expresses a classical gene signature related to vascular endothelial growth factor receptor activity but not angiogenesis (27). The molecular identity of type II and transitional capillary ECs reflects an intermediate phenotype between capillary and arteriole ECs or capillary and high ECs (HECs), respectively. A small cluster was found to highly express IFN signaling genes, which may reflect a distinct activation state. Additionally, a population of capillary resident precursors was identified, which express progenitor-like genes and functions (27).
Capillaries converge to form HEVs, a specialized subset of venous BECs that have a cuboidal shape, express high levels of peripheral node addressin (PNAd) and CCL21 to mediate entry of l-selectin–, CCR7-expressing naive lymphocytes into the LN parenchyma (29). LTβR signaling in HECs is required to maintain luminal PNAd expression (30) as well as the expression of CCL21 and glycosylation-dependent cell adhesion molecule 1 (GLYCAM1) (10, 31). LTα1β2-expressing DCs interacting with HEVs are proposed to be the cellular source of lymphotoxin ligands maintaining HEC functionality (32). Recently, Girard and colleagues (31) have also described regional heterogeneity in HECs, with those situated near medullary sinuses expressing PNAd but not GLYCAM1 in contrast to the overlapping expression of these prototypical HEV addressins throughout the cortex. High-resolution histological analyses suggest that HEVs converge with veins traversing the medulla that lack PNAd expression, consistent with the transcriptomic identification of venous BECs lacking HEV-associated genes (Chst4, Fut7, Ccl21) (27). Non-HEC, venous ECs instead express genes encoding E- and P-selectins as well as ICAM1 and VCAM1 and may preferentially mediate inflammation-associated myeloid cell influx. Further studies are warranted to define the topology of heterogeneous addressin expression on the venous endothelium and its functional relevance as portals to lymphocyte and myeloid cell entry in the steady-state and following inflammation.
LN FRC subsets can be broadly grouped according to the immune cells they interact with and their anatomical location (Fig. 1C): B cell–interacting reticular cells (BRCs), T cell zone reticular cells (TRCs), medullary reticular cells (MedRCs), and perivascular reticular cells, including a subset of CD34+ reticular cells. TRCs populate the LN cortex and secrete the T cell chemoattractants and survival factors CCL19, CCL21, and IL-7 (33, 34). Single-cell transcriptomic analyses have resolved at least three distinct TRC subsets: TRCs expressing high or low levels of CCL19 and CXCL9-expressing TRCs (35). CCL19-high TRCs appear to be canonical, conduit-ensheathing TRCs, whereas CCL19-low TRCs and CXCL9-expressing TRCs express higher transcript levels of Cxcl13 and may be situated at the T-B border and interfollicular regions, respectively (35). TRCs bordering the B cell follicle and interfollicular regions transcriptionally resemble and have also been referred to as T-B border reticular cells (TBRCs) and interfollicular reticular cells (36, 37). In addition to the expression of B and T cell chemoattractants, TBRCs also support the accumulation of Ab-secreting cells via the production of the plasma cell niche factors a proliferation-inducing ligand (APRIL), CXCL12, and IL-6 (38, 39). CXCL9+ TRCs express a strong IFN-inducible gene signature and were postulated to direct the movement of CX3CR1-expressing myeloid cells following inflammation (35). Genetic abrogation of type I IFN sensing in Ccl19-Cre–targeted FRCs indeed led to changes in LN myeloid cell composition already in the steady-state, revealing a role for tonic antiviral-sensing programs in myeloid cell–interacting FRC subsets such as medullary and interfollicular reticular cells (37).
The B cell follicle is underpinned by CXCL13-expressing FRCs, collectively termed as BRCs. CXCL13 promotes B cell clustering and follicle formation (6), and concentrated, immobilized gradients of the chemokine have recently been shown to enhance B cell trafficking and Ag encounters within the follicle (40). BRCs include marginal reticular cells lining the SCS follicular DCs (FDCs) that capture and display immune complexes to B cells and TBRC. Using the Cxcl13-Cre/tdTomato mouse model and scRNA-sequencing, Pikor and colleagues (36) recently elucidated that the B cell follicle is underpinned by two subsets of FDCs: light zone (LZ)– and dark zone (DZ)–FDCs. LZ-FDCs recapitulate the classical profile of FDCs, expressing complement and Fc-receptors, whereas DZ-FDCs express lower levels of these receptors compared with LZ-FDCs but share expression of canonical genes involved in dendrite morphology (36). Notably, specification of BRC subsets, including LZ- and DZ-FDCs, was found to occur already in the steady-state (36).
High-resolution histological analysis using different gene reporter mouse models has identified distinct reticular cell subsets in the deep cortex periphery and medullary sinuses (41). Both subsets express high levels of CXCL12 and embed ERTR7-rich conduits; however, FRCs in the deep cortex periphery express higher levels of CCL21 and underpin B cells accumulated in the region adjacent to the medulla. In contrast, MedRCs express high levels of the leptin receptor and interact with Ab-secreting cells, NK cells, and macrophages (41), consistent with previous studies that medullary FRCs express high levels of CXCL12, IL-6, APRIL, and BAFF (38, 42) as well as CCL2 (43) to recruit and retain these cell populations. scRNA-seq identifies at least two populations of MedRCs (35, 36), although further studies are needed to confirm the spatial distribution and function of molecularly distinct MedRC subsets.
Inflammation-induced LNSC remodeling ensures immunological function
The presence of foreign Ags (soluble or taken up by DCs) in draining LNs triggers an influx of immune cell entry, activation, and proliferation accompanied by LNSC remodeling. FRC, LEC, and BEC all proliferate in response to bacterial and viral pathogens, adjuvants, and immunogenic tumors (44–47) to accommodate LN swelling, lymphangiogenesis, vascular sprouting, and stretched FRC networks. Collectively, the cross-talk of poised stromal cell niches with activated immune cells drives the topological and transcriptional remodeling of LNSCs, which in turn coordinates the efficient interaction of Ag, APCs, and cognate T and B cells.
Lymphatic remodeling leads to increased lymph flow, and an influx of DCs and lymphocytes into LNs draining the site of inflammation (16). In the context of acute infections, B cells (48, 49), macrophages (50), and DCs (51) have all been identified as key producers of vascular endothelial growth factor (VEGF) and drivers of lymphangiogenesis. B cell–dependent mechanisms of lymphangiogenesis require LTβR-dependent interactions with FRCs (49). Transcriptional analysis of LEC subsets in draining LNs 48 h after topical oxazolone application demonstrates that although both fLECs and PTX3+ medullary LECs upregulate IFN response genes, they also maintain site-specific upregulation of inflammatory mediators (25). fLECs selectively upregulate Ccl20 encoding the ligand for CCR6 expressed on DC, effector/memory T cells and B cells, and Ccl5 encoding the ligand for CCR5 expressed on monocytes while downregulating Csf1. In contrast, PTX3+ LECs upregulate Ccl2 encoding the ligand for the monocyte chemoattractant CCR2. Transcriptional upregulation of Ccl20, Ccl2, and Ccl5 has also been reported in bulk RNA–sequencing analyses of LECs following viral infection (45). The transient downregulation of Csf1 may facilitate the migration of SCS macrophages into B cell follicles of inflamed LNs (19, 52), whereas high CCL2 expression in medullary sinuses may help support a negative feedback–limiting humoral immunity (43). Collectively, inflammation-induced lymphangiogenesis and inflammatory LEC states promote leukocyte influx and regional distribution in draining LNs.
Inflammation-induced angiogenesis in LNs.
Neovascularization helps to meet the increased nutrient and oxygen demand by proliferating cells and promotes lymphocyte accumulation in inflamed LNs. Within the LN parenchyma, sprouting angiogenesis of HEVs (53) is triggered by IL-1β–producing DCs that induce the production of VEGF by TRCs (51). However, B cells are required to reach maximal EC numbers in draining LNs (45). Using an inducible, multicolored tracking of single Cdh5-Cre–targeted ECs, Bajénoff and colleagues (54) have shown that endothelial progenitor cells situated along LN HEVs undergo clonal expansion and sequentially assemble to create capillaries and HEV neovessels in inflamed LNs. Recent fate-mapping experiments further demonstrate that within the Cdh5-expressing population, progenitor potential is recapitulated by Apln-CreER–targeted capillary resident precursors that contribute to the inflammation-induced neovascularization of capillaries and HEVs in inflamed LNs (27).
LN BECs appear to undergo modest transcriptional remodeling over the course of an antiviral immune response. Bulk RNA–sequencing identified Cxcl11, C3, and Vegfc among a handful of genes upregulated by BECs in HSV-1–draining LNs (45). scRNA-seq analysis has further distinguished inflammation-induced transcriptional changes in HECs compared with capillary BECs. Girard and colleagues (31) observed that HECs in inflamed LNs transiently downregulate mature HEV genes such as Glycam1 and Gnct1 and upregulate inflammatory genes such as Tnfrsf9 encoding the costimulatory marker CD137 and Ch25h. The transcriptional dedifferentiation of mature HECs in inflamed LNs may reflect a temporary loosening of the selectivity of lymphocyte infiltration into the LN via HEVs (55) or may reflect an immature transcriptional signature of newly assembled HEVs (54). Combining single-cell sequencing with fate-mapping models would clarify the transcriptional remodeling of HECs during inflammation.
FRC remodeling steers optimal adaptive immunity.
Inflammation-induced FRC proliferation requires B cells and LTβR signaling to support the maturation of FRC progeny (34, 45, 56) as well as additional cues and interacting cell partners. In the T cell zone, the interaction of FRCs and DCs promotes CLEC2-dependent reticular cell relaxation and the physical stretching of the network (57). Additionally, IL-17–dependent cross-talk between locally activated Th cells and TRC may metabolically support fibroblast proliferation and survival (58). In the B cell follicle, chemokine gradients localize proliferating B cells to the T-B border, causing a stretching of the BRC network to create the germinal center (GC) DZ (36). In addition to topological remodeling, inflammation induces transcriptional FRC remodeling. Bulk sequencing of PDPN+ reticular cells reveals that genes involved in Ag presentation, extracellular matrix, and chemokine and cytokine signaling are among the functional gene groups differentially regulated in LN FRCs draining the site of viral infection (45) or tumors (47). Single-cell transcriptomic analyses revealed that inflammation does not induce the maturation of new reticular cell subsets (35–37) but rather reflects changes in the activation state of poised subsets. Indeed, Ccl19-Cre–targeted FRCs demonstrate profound changes in transcriptional programs related to type I IFN responses, Ag presentation, chemokine-mediated immune cell recruitment, and immune regulation at early time points following lymphocytic choriomeningitis virus infection (37). Such transcriptional remodeling may be transient, as FRC gene expression profiles at the peak of the adaptive immune response were relatively unaltered in lymphocytic choriomeningitis virus–immunized mice (35). Moreover, relatively modest transcriptional changes in FDCs were shown to accompany vesicular stomatitis virus–induced GC formation; LZ-FDCs further upregulated genes involved in Ag presentation and capture, whereas DZ-FDCs upregulated a number of genes encoding for cytokines and chemokines, cell adhesion, and extracellular matrix remodeling (36). Thus, the degree of inflammation-induced transcriptional remodeling may vary temporally and according to the pathogen. Because these transcriptional changes occur in neighboring, phenotypically similar subsets (59), the subsequent challenge will be using and developing appropriate genetic models to selectively dissect the immunological function of distinct FRC subsets.
Transcriptional and topological FRC remodeling not only alleviates the spatial constraints of inflamed LNs but also promotes efficient immune cell interaction and responses. For instance, the topological remodeling of LZ- and DZ-FDCs driven by the CXCL12-dependent cross-talk with B cells is required for optimal encounter of GC B cells with follicular helper T cells, class-switch recombination, and affinity maturation (36). Moreover, the physical stretching of the FRC network during inflammation may promote temporary conduit permeability (60, 61), which in turn has been shown to promote the passive entry of IgM into conduits for their expedited distribution to the periphery (62). Although size exclusion to large molecular mass Ags (500 kDa dextran) was maintained in inflamed LNs (60), Shields and colleagues (47) demonstrate an increased permeability of conduits to 70 kDa dextran in tumor-draining LNs, suggesting a functional, but more relaxed, size exclusion of conduits in swollen LNs. It is notable that the passive conduit uptake of IgM, a critical first wave of protection against cytopathic viruses (63, 64), spatially overlaps with the T-B border, an area of poised chemokine gradients supporting T and B cell migration (36), B cell class-switching (65), and Ab-secreting cell survival factors (38, 39), suggesting that the choreographed fine-tuning of poised BRC subsets orchestrates efficient humoral immunity.
In terms of T cell responses, FRCs may modulate T cell activation, differentiation, and peripheral tolerance. Examination of murine LN FRCs demonstrates that a cross-talk between FRCs and T cells limits the T cell proliferation and effector functions in an NO-dependent manner (66–69). Notably, human FRCs have recently been shown to attenuate T cell proliferation in an NO- and T cell–independent manner (70). T cell anergy was mediated by FRC secretion of IDO, cyclooxygenase-1 and -2 (COX1, COX2), adenosine 2A receptor (A2AR), and TGF-β and was reversible either through pharmacological blockade or removing T cells from FRC cultures (70). Importantly, effector functions of preactivated or chimeric AgR T cells were unperturbed in FRC cocultures, in contrast to findings in the murine setting. This disconnect may reflect differences related to the strength of T cell stimulation in murine experiments or an altered milieu in chronically activated human lymphoid tissues. In line with the notion that FRCs do not strictly suppress T cell activation, FRCs were shown to synergize with IL-6 to condition antiviral CD8 T cells to adopt a tissue resident phenotype compared with T cells activated in FRC cultures in the absence of IL-6 (71). Although IL-6 was upregulated in FRCs cocultured with splenocytes (71), the source of IL-6 in inflamed LNs remains unclear. IL-6 was previously found to be upregulated by BECs and LECs, but not FRCs in bulk RNA–sequencing (45) and microarray analyses (72). Furthermore, type I IFN responsiveness in Ccl19-Cre–targeted FRC has recently been shown to enhance Ag presentation capacity, although dampening expression of immune regulatory factors to favor effector T cell activation rather than exhaustion (37). Future studies will need to examine to what extent FRC-produced T cell–activating and –attenuating cues are spatially constrained, varied over the course of inflammation, and whether species-specific mechanisms guide FRC–T cell interaction.
Recent advances in stromal cell–targeted genetic models and high-resolution transcriptomic analyses have elaborated with an unprecedented resolution the cellular and molecular processes involved in LN development as well as the spatial and molecular identity of LNSC subsets. These studies highlight that subset specification occurs in the steady-state, and poised FRC subsets act as a road map to direct the optimal interaction of activated immune cells, Ags, and LNSCs. Local cross-talk with activated leukocytes, and possibly with soluble inflammatory mediators draining inflamed tissues, mediates transcriptional changes in state of LNSCs and topological remodeling that further optimize the influx of immune cells into the LN parenchyma, the interaction of cognate T and B cells, T cell activation, and the rapid delivery of neutralizing Abs to peripheral tissues. The inflammatory microenvironment as a whole contributes to the efficiency of adaptive immune responses.
Recent advances in transcriptomic analyses at the single-cell level have helped to resolve previously unknown LNSC subsets. As with any new technology, it is important to understand the technical limitations and sources of variation in these analyses, such as sequencing depth, total cell number, and clustering strategy. In particular, the molecular identity of scarce subsets is the most susceptible to variation depending on the total yield and clustering resolution. Repeated experiments from different groups over time will reveal the most robust gene signatures that identify distinct LNSC subsets and activation states. Moreover, the quality and interpretation of the data rely on competent cell isolation protocols and validation using independent methods. Although in vitro and in situ culture conditions are the only available option for human studies, the next challenge to resolve the identity and differentiation of closely related FRC subsets during LN development or inflammation-induced remodeling in the murine setting will be the selection or generation of appropriate genetic models. Ultimately, a thorough understanding of the role of LNSCs in orchestrating immunity will require a dissection of the spatial and molecular topology of stromal cell–defined niches in relation to leukocyte migration, activation, and differentiation.
This work was supported by the Swiss National Science Foundation Grants 166500 and 159188 (to B.L.) and 180011 (to N.B.P.). The funder had no role in design, decision to publish, or preparation of the manuscript.
Abbreviations used in this article:
blood endothelial cell
B cell–interacting reticular cell
fibroblastic reticular cell
high endothelial venule
lymphatic endothelial cell
LN stromal cell
lymphoid tissue inducer
lymphoid tissue organizer
medullary reticular cell
peripheral node addressin
T-B border reticular cell
T cell zone reticular cell.
The authors have no financial conflicts of interest.