Summary
The lymph node (LN) is home to resident macrophage populations that are essential for immune function and homeostasis. The T cell paracortical zone is a major site of macrophage efferocytosis of apoptotic cells, but key factors controlling this niche are undefined. Here we show that fibroblastic reticular cells (FRCs) are an essential component of the LN macrophage niche. Macrophages co-localised with FRCs in human LNs, and murine single-cell RNA-sequencing revealed that most reticular cells expressed master macrophage regulator CSF1. Functional assays showed that CSF1R signalling was sufficient to support macrophage development. In the presence of LPS, FRCs underwent a mechanistic switch and maintained support through CSF1R-independent mechanisms. These effects were conserved between mouse and human systems. Rapid loss of macrophages and monocytes from LNs was observed upon genetic ablation of FRCs. These data reveal a critically important role for FRCs in the creation of the parenchymal macrophage niche within LNs.
Introduction
In lymph nodes, stromal cell communication with leukocytes is key to the initiation of a healthy immune response and eventual pathogen control (Fletcher et al., 2020; Lütge, Pikor and Ludewig, 2021). Fibroblastic reticular cells (FRCs) are the most prevalent non-hematopoietic cell type in lymph nodes. Together with sinusoidal vascular elements, FRCs create the structural scaffolding on which leukocytes migrate and interact, including the T cell and dendritic cell-rich paracortex, the cortical B cell follicles, and the medullary plasma cell niche (Cremasco et al., 2014; Huang et al., 2018; Fletcher et al., 2020). From this unique position at the coalface of the immune response, FRCs have evolved an immune-specialised role in regulating the survival, interaction, migration and function of T cells, B cells, dendritic cells, plasma cells, and innate lymphoid cells (ILCs) (Acton et al., 2012; Cremasco et al., 2014; Huang et al., 2018; Knoblich et al., 2018; Fletcher et al., 2020; Pikor et al., 2020; Kapoor et al., 2021; Lütge, Pikor and Ludewig, 2021).
The importance of FRCs in adaptive immunity is well established, and emerging evidence suggests that they also play a critical role in innate immunity. Specialised Gremlin1+ subsets of fibroblasts engage in multifaceted crosstalk with dendritic cells in lymphoid tissues, and in some systems fibroblasts and macrophages are capable of co-operative support through mutual provision of growth factors (Zhou et al., 2018; Bellomo et al., 2021; Franklin, 2021; Kapoor et al., 2021). Mouse FRCs also respond to viral and bacterial ligands (Fletcher et al., 2010; Malhotra et al., 2012; Yang et al., 2014), regulate activation of group 1 ILCs and viral clearance through expression of the innate immunological sensing adaptor MyD88 (Gil-Cruz et al., 2016), and in nonclassical secondary lymphoid organs, FRC-dependent MyD88 signaling steers B cell responses via TNF-dependent interactions with inflammatory monocytes (Perez-Shibayama et al., 2018).
Despite the pivotal role of macrophages as managers of immune homeostasis and drivers of humoral and anti-viral immunity, our understanding of macrophage biology in lymph nodes is still evolving. Macrophages are found in every FRC niche (Baratin et al., 2017; Bellomo et al., 2018), but key factors controlling their development, function and localisation in lymph nodes remain unclear. Lymph node macrophages are categorized by the niches they occupy. Sinus-lining macrophages are relatively well-studied as highly phagocytic “flypaper” macrophages bathed in lymph and capable of engulfing incoming antigen for efficient presentation to other cell types (Bellomo et al., 2018). Subcapsular sinus macrophages capture particles or immune complexes for direct transfer to follicular B cells, and are important for limiting viral spread (Carrasco and Batista, 2007; Junt et al., 2007; Phan et al., 2009; Gonzalez et al., 2010; Iannacone et al., 2010). They can present antigen to naïve T cells, but their ability to internalise and process antigen is lower than medullary sinus macrophages, which are highly efficient at pathogen and apoptotic cell clearance (Phan et al., 2009). Depletion of both subsets in mice reduced anti-tumor immunity through a reduction in CD8+ T cell activation (Asano et al., 2011). Mesenchymal lymphoid tissue organizer (LTO) cells and marginal reticular cells (MRC) utilize RANKL to support the development of subcapsular and medullary CD169+ sinusoidal macrophages, but not other macrophage subsets (Camara et al., 2019).
Less well studied are the macrophages found in parenchymal and FRC-rich regions of lymph nodes. Medullary cord macrophages regulate medullary plasma cell maturation and survival (Mohr et al., 2009; Huang et al., 2018), while in the T cell paracortical zone, a rich, previously misidentified network of immunosuppressive macrophages play a unique role in immune homeostasis through potent efferocytosis of apoptotic T cells (Baratin et al., 2017; Bellomo et al., 2018). Both macrophage populations strongly co-localise with their local FRC network (Bellomo et al., 2018; Huang et al., 2018).
Here, we identified an essential support system provided by FRCs to macrophages. We demonstrate that expression of CSF1R ligands by FRCs is capable of regulating macrophage differentiation and survival. We further show in two genetic models of FRC ablation that depletion of FRCs drives a rapid loss of myeloid cells in lymph nodes. These data show that FRCs are a critically important component of the lymph node macrophage niche.
Results
T zone macrophages colocalize with FRCs in secondary lymphoid organs
Examination of mouse lymph nodes confirmed that macrophages identified by expression of MERTK co-localized with the FRC network in the T cell zone (Figure 1A) (Baratin et al., 2017). We quantified these interactions by calculating macrophage proximity and alignment with extracellular matrix fibers secreted by FRCs, based on co-localization of >10% of macrophage perimeter with a visible fiber, for cells with a clear cross-section (Figure 1B). Macrophages were significantly more likely to be associated with an FRC fiber than not associated (Figure 1C), and FRC-associated macrophages showed significantly greater elongation (Figure 1D), supporting a physical association between these two cell types.
Next, we examined whether the same held true for human FRCs and macrophages in tonsils. CD163+ macrophages co-localized with the FRC network in the paracortical T cell zone (Figure 1E), and also preferentially associated with FRC fibers, with significant elongation when associated (Figure 1F, G).
These data provide evidence for a close relationship between macrophages and the FRC network that is conserved across human and mouse lymph nodes.
Conserved expression of genes crucial for myeloid cell maturation, migration and function
To determine whether FRCs express factors relevant to myeloid cell maintenance, and whether these were broadly concordant across species, we performed transcriptomic analysis and comparison of a range of FRC datasets obtained from cultured human tonsil-derived FRC samples (Figure 2A), cultured versus freshly isolated human tonsil-derived FRCs (Figure 2B), freshly isolated FRCs from mouse skin-draining (SLN) or mesenteric lymph nodes (MLN) (first published in 31) (Figure 2C) and cultured mouse SLN FRCs (Figure 2D). All datasets showed robust expression of macrophage colony stimulating factor CSF1 and myeloid chemoattractants CCL2, CXCL1, and CXCL12. CXCL8 has no mouse equivalent but was strongly expressed in human cells. Tonsil and SLN, but not MLN FRCs, showed high expression of monocyte differentiation and growth factor IL-6. These data showed that expression of myeloid regulatory factors is a property that is conserved across human and mouse FRCs, and suggested a potential role for the FRC-macrophage dyad in regulating the myeloid response to infection.
FRCs exhibit TLR4 signalling and expression of myeloid cell immunoregulatory factors
FRCs express pattern recognition receptors TLR3 and TLR4, and can respond to cognate bacterial and viral stimuli or analogues (Fletcher et al., 2010; Malhotra et al., 2012; Gil-Cruz et al., 2016; Severino et al., 2017). However, the kinetics, signalling mechanisms and effects on downstream transcription are not well defined, particularly for human FRCs. We therefore examined the effect of pathogen sensing on FRC-derived factors relevant to myeloid cell migration and function.
RNA sequencing (RNA-Seq) analysis of cultured human FRCs showed that TLR4 expression was the highest and least variably expressed TLR between the 3 donors (Figure 3A).
To confirm functional MyD88-dependent TLR4 signalling within human FRCs, we assessed phosphorylation of p65 (RelA, NF-kB pathway) and p38 (MAPK pathway) in response to LPS (Figure 3B) (Akira and Takeda, 2004). TLR4 inhibitor CLI-095 (TAK-242), which specifically binds the intracellular domain of the TLR4 receptor to block signalling (Kawamoto et al., 2008), was used to confirm TLR4 dependence. Flow cytometric analysis showed rapid phosphorylation of both p38 and p65 in FRCs after 5 minutes of LPS exposure, peaking at 10 minutes, confirming that FRCs quickly sense LPS and respond via functional TLR4 signalling. Accordingly, LPS stimulation increased the secretion of downstream targets IL-6 and IL-8 (CXCL8) by cultured human FRCs within 24h (Figure 3C, D).
Next, we investigated global transcriptional outcomes of TLR signalling to FRCs in vivo. CCL19-Cre x R26R-EYFP mice were injected with LPS to induce an inflammatory response. After 3 days, EYFP+ reticular cells from the lymph nodes of LPS treated and control mice were sorted for single-cell RNA-Seq analysis. Eight fibroblastic clusters were identified and validated via expression of expected markers (Figure 3E; Supp. figure 1A-C) (Rodda et al., 2018; Pikor et al., 2020). The subsets identified comprised T zone reticular cells (TRC); follicular dendritic cells (FDC); pericytes; perivascular reticular cells (PVRC); marginal reticular cells (MRC); T-B zone reticular cells (TBRC); and two subsets of medullary reticular cells (medRC1 and medRC2). (Figure 3E).) All clusters were represented in both treatment groups (Figure 3E; Supp. figure 1D).
Next, we compared LPS-treated versus treatment-naïve mice and identified the 30 most differentially expressed genes (15 upregulated with treatment, 15 downregulated) across all clusters. LPS treatment was associated with upregulated expression of myeloid cell-attracting chemokines including Cxcl1, Cxcl2 and Ccl2. Conversely, expression of homeostatic chemokines Ccl19 and Ccl21a was downregulated, as well as genes encoding ribosomal proteins. KEGG pathway analysis (P<0.05, FDR and Benjamini-Hochberg <0.05) showed that LPS treatment drove significant over-representation of genes related to innate immune function, including TNF, chemokine, cytokine and NOD-like receptor signalling (Figure 3G).
Based on their prominent roles in myeloid cell regulation, we chose to further validate expression of Ccl2 and Csf1, which were robustly expressed across all human and mouse RNA screening data (Figure 2, 3). Ccl2 was significantly upregulated with LPS treatment in mice, while Csf1 showed unchanged robust expression across fibroblast subsets (Figure 3F, H). To test if this held true in human cultures, FRCs from 3 donors were treated with LPS for 24h and secreted proteins analysed. LPS treatment drove a significant upregulation of CCL2 secreted protein, which was abrogated in the presence of TLR4 inhibitor TAK-242 (Fig. 3I). As the PI3K/AKT signaling pathway can be activated in MyD88-dependant TLR4 signalling (Laird et al., 2009), we also examined the effect of PI3K inhibition of LPS-induced protein secretion by FRCs. In the presence of PI3K inhibitor LY294002, the increase in CCL2 was similarly abrogated (Figure 3I). These data confirmed that CCL2 upregulation by human FRCs was driven by LPS signaling through TLR4. Conversely, Csf1 expression did not appear as a differentially expressed gene in the mouse scRNA-Seq analysis, a finding borne out by flow cytometry staining for CSF1 protein expression in human FRCs, which was unchanged after LPS treatment (Figure 3J).
Mouse and human FRCs can regulate the survival and differentiation of monocytes via signalling to CSF1R
A functional role for CSF1 expression by lymph node FRCs has not previously been reported. Robust expression of CSF1 across our mouse and human datasets, together with the close connection between FRCs and myeloid cells observed in vivo, led us to hypothesise that FRCs may have the capacity to support myeloid cell development or differentiation.
In the lymph nodes, T cell zone resident macrophages are long-lived, with slow replacement by blood-borne monocytes (Baratin et al., 2017). Monocytes can be defined as classical (CD11b+Ly6Chi in mouse, CD14+CD16− in human) or non-classical (CD11b+Ly6Clo in mouse, CD14−/loCD16+ in human) based on their ability to perform a pro- or anti-inflammatory functions (Ziegler-Heitbrock et al., 2010).
The proliferation, differentiation and survival of macrophages and monocytes is regulated by CSF1, which acts via autocrine or paracrine signaling through its receptor, CSF1R (Chitu and Stanley, 2006; Lenzo et al., 2012). To investigate the effects of FRC-derived CSF1 signaling on myeloid cells, we performed co-culture assays with mouse bone marrow cells as a source of myeloid precursors. In the presence of recombinant CSF1, as expected, bone marrow cells differentiated to CD11b+F4/80+ macrophages (Figure 4A). This macrophage differentiation was inhibited by the addition of a CSF1R blocking antibody (Figure 4A, Supp. Fig 2A). Notably, the addition of FRCs to bone marrow cells, without exogenous CSF1, was sufficient to yield an expansion of (CD11B+ F4/80+) macrophages (Figure 4A).
Next, bone marrow cells were co-cultured for 3 days with or without mouse FRCs or CSF1R blocking antibody (Figure 4B). The addition of FRCs significantly increased the number of macrophages (CD11b+F4/80+) and classical monocytes (GR-1− CD11b+ LY6Chi) over the culture period, dependent on CSF1R signalling (Figure 4B). Absolute numbers of macrophages and classical monocytes significantly increased after plating, suggesting that FRCs were able to actively foster the proliferation and/or differentiation of these cells. FRCs were also able to maintain numbers of non-classical monocytes (GR-1− CD11b+ LY6Clo), which otherwise underwent attrition over the culture period (Figure 4B).
LPS treatment, which did not alter CSF1 transcription, nonetheless had synergistic effect when administered with FRCs, increased macrophage number two-fold above the FRC-mediated increase, and this additional boost was not CSF1 dependent (Suppl. Fig. 2B),.
Based on these results, we explored protein-level expression of CSF1R ligands by human FRCs. In addition to expressing CSF1, human FRCs also expressed IL-34 protein (Figure 4C, D, confirmed by RNAseq data), which also binds CSF1R and induces the maturation of monocytes into macrophages (Foucher et al., 2013).
To examine the effects of human FRCs and FRC-derived CSF1R ligands on human monocyte and macrophage phenotypes, we used a co-culture system with PBMCs from healthy donors as a monocyte source and examined numbers and phenotypes of major monocyte and macrophage subsets after 3 days. Macrophages exhibit effects on a continuum from strongly pro-inflammatory (often denoted M1; CD64+HLA-DR+) through to strongly suppressive (M2; CD206+CD64−) (Buchacher et al., 2015; Tarique et al., 2015). Their differentiation depends on the cues given by the local tissue micro-environment (Lenzo et al., 2012). Human FRCs did not alter M1-phenotype macrophages (Figure 4E), but provided a strong differentiation stimulus for M2-phenotype macrophages (Figure 4F). This effect of FRCs was significantly reduced in the presence of a CSF1R blocking antibody, capable of neutralizing the effects of both CSF1 and IL-34 in culture. FRCs also drove an average 8-fold increase in CD16−CD14+ classical monocytes, however this was not mediated through CSF1R signalling, since CSFR1 blockade did not prevent the increase (Figure 4G). As with mouse monocytes, this suggested the presence of undescribed CSF1R-independent mechanisms of FRC-support. Human FRCs did not alter numbers of CD16+CD14− non-classical monocytes (Figure 4H). These data were concordant with mouse results.
We also tested the effect of LPS on this system, having observed that LPS did not alter CSF1 transcription. In the presence of LPS, FRCs still provided significant support for M2 and classical monocytes, but this occurred via CSF1R-independent mechanisms (Figure 4F, G, Supplementary figure 2B). Fold-change increases in M2 macrophages and non-classical monocytes with FRC co-culture represented an absolute increase in cell numbers over 72h of culture, showing that FRCs support an active increase in these cells rather than simply fostering survival (Supplementary figure 2C, D).
Together, these data show that mouse and human FRCs are able to promote the differentiation or expansion of M2 macrophages and classical monocytes, at least in vitro. FRCs expressed CSF1R ligands CSF1 and IL-34, and signaling to the CSF1R played a major role in the support observed under steady-state conditions. Additional undefined FRC-derived signals were involved in driving the FRC-mediated increase in M2 macrophages and classical monocytes that occurred in the presence of LPS.
FRC ablation diminishes myeloid cell lineages within lymph nodes
Our data showing accumulative effects of FRCs on M2 macrophages and classical monocytes drove us to test the hypothesis that the removal of FRCs would lead to a loss of major myeloid cell types in the lymph node.
To test this, we used two mouse models of FRC depletion. In vivo depletion of FRCs was achieved, by crossing a CCL19-Cre strain with a Rosa26-diphtheria toxin receptor strain (CCL19-DTR), as described (Cremasco et al., 2014) . In this mouse model, diphtheria toxin (DTx) treatment specifically depletes FRCs within lymph nodes. 48h after DTx treatment, lymph nodes underwent significant and prolonged involution, involving overall reduced cellularity that did not recover for the duration of the study (3 weeks) (Figure 5A). Accordingly, FRCs were fully ablated at day 2 post-treatment, and did not recover within the study duration (Figure 5B). Monocytes underwent an initial influx into the lymph node prior to their loss (Figure 5C), and macrophages similarly showed a delayed loss, with no effect at day 2 and a significant depletion at day 8 that did not recover by day 22 (Figure 5D).
Monocyte and macrophage populations were therefore not sustainable in the absence of FRCs. FRCs were dispensable for a monocyte influx at Day 2, but this increase was short-lived, and did not translate to an observable recovery in macrophage numbers at timepoints studied.
In a separately developed model where FRCs express DTR under the control of Fibroblast Activation Protein (DM2 mice (Denton et al., 2014)), we observed a similar effect of FRC depletion on monocyte and macrophage numbers (Supp. Figure 3). Lymph nodes harvested 2 days after the cessation of DTx treatment showed that FRC ablation led to a significant reduction in monocytes (classical and non-classical), and macrophages (subcapsular and medullary).
Taken together, these findings demonstrate the provision of a supportive niche by FRCs to myeloid lineage cells within secondary lymphoid organs.
Discussion
FRCs form the structural highway on which leukocytes interact. FRCs have been shown to facilitate deletional tolerance (Fletcher et al., 2010), antigen presentation (Baptista et al., 2014; Dubrot et al., 2014), and lymphocyte and dendritic cell homing (Bajenoff et al., 2006; Link et al., 2007) . FRCs promote T cell, B cell, plasma cell and ILC survival (Link et al., 2007; Cremasco et al., 2014; Gil-Cruz et al., 2016; Huang et al., 2018); they regulate T cell activation in mice and humans (Fletcher et al., 2020), and are known to respond to LPS (Malhotra et al., 2012; Fletcher et al., 2014; Gil-Cruz et al., 2016; Perez-Shibayama et al., 2018). Recently, targeted deletion of type I interferon receptor (Ifnar) from FRCs revealed a role in infection-driven monocyte and neutrophil accumulation and recruitment, revealing an important biological imperative to understand FRC-innate immune cell interactions within lymph nodes and secondary lymphoid organs (Perez-Shibayama et al., 2020).
While the presence of resident medullary macrophages within lymph nodes has been long reported (Bellomo et al., 2018) and macrophages are confirmed in FRC-rich zones (Zhang et al., 2012; Baratin et al., 2017; Huang et al., 2018), the cells and factors supporting their residence remain undefined. Here we sought to discover the immunoregulatory factors involved in the intimate relationship between FRCs and myeloid cells under steady-state and inflammatory conditions.
Immunofluorescence microscopy of the human tonsil and mouse lymph nodes revealed that macrophages colocalize with FRC fibers and elongate when aligned with FRC fibers. Our transcriptomic analysis of mouse and human FRCs built on previous findings and showed a strong conservation of expression of genes relevant to innate immunity, particularly the expression of CSF1, CCL2, IL-6, IL-8/CXCL8 and CXCL12, which all have well-established roles in the maturation, function, adhesion and/or chemoattraction of myeloid cells (Witmer-Pack et al., 1993; Gerszten et al., 1999; Pixley and Stanley, 2004; MacDonald et al., 2010; Mauer et al., 2014).
TLR4 (CD284), in complex with CD14 and Ly96/MD2, forms the major LPS receptor on mammalian cells, and it also binds endogenous proteins such as oxidized low-density lipoprotein and beta-defensins, as well as polysaccharides including hyaluronic acid and heparin sulfate proteoglycan (Brubaker et al., 2015). TLR4 complex ligation drives multiple intracellular signalling cascades. The myeloid differentiation primary response 88 (MyD88)-dependent pathway activates the NF-κB pathway, driving transcription of pro-inflammatory cytokines, such as IL-1B and IL-6. MyD88 also activates mitogen-activated protein kinase (MAPK), which leads to activation of transcription factor AP-1, controlling gene transcription as well as mRNA stability, notably for IL-6. Some genes, such as CXCL8 which encodes IL-8, contain binding regions for both NF-κB and MAPKs, and are synergistically influenced by both (Akira and Takeda, 2004; Cronin et al., 2012).
Our work showed that lymph node FRCs are poised to respond quickly to LPS-driven inflammation by upregulating innate cytokines IL-6, IL-8/CXCL8 and CCL2, through rapid TLR4 signalling, which began within 5 mins and occurred via both NF-κB and MAPK pathways. The relevance of upregulation of these cytokines in vivo requires further exploration, but it is reasonable to assume that production of these factors in lymph nodes following TLR signalling is very likely to stimulate resident macrophages to respond.
Single-cell RNA-seq data identified and validated eight reticular cell clusters from LPS-treated and treatment-naïve mice, correlating with previous single-cell RNA-seq analyses, and demonstrating the niche specific heterogeneity of these cell types (Rodda et al., 2018; Kapoor et al., 2021). From this, myeloid cell-associated chemokines, Cxcl1, Cxcl2 and Ccl2, were expressed in LPS-treated mice, suggesting an as-yet untested role for FRCs in promoting macrophage and monocyte migration. This correlates with recent findings that the recruitment of monocytes into the lymph node following immunisation with OVA-Alum is dependent on CCL2 derived from FRCs (Dasoveanu et al., 2020), though these findings may be model-dependent as neither Ccl2 nor Cxcl1 expression by FRCs was upregulated in an Herpes Simplex Virus infection model, despite observed myeloid infiltration (Gregory et al., 2017). Murine 3T3 fibroblasts have been shown to facilitate the migration of macrophages via cellular contractions and tunnel formation (Ford, Orbach and Rajagopalan, 2019), however the chemotactic mechanisms responsible have not been validated.
Co-culture experiments suggested that FRCs support the differentiation and survival of M2-like macrophages via signalling through CSF1R in the absence of TLR4 signalling, and through other, yet to be defined, mechanisms in the presence of LPS. It is attractive to speculate that FRCs promote macrophage polarisation towards a regenerative and repair state, and away from an inflammatory state, as FRCs dampen innate immune-driven inflammation in murine sepsis (Fletcher et al., 2014; Xu et al., 2019).
The importance of an intact FRC network for macrophage maintenance was demonstrated using in vivo depletion of FRCs in two mouse models, CCL19-DTR and FAP-DTR. Both models showed a rapid loss of resident macrophages. The lack of recovery was not unexpected; alterations in FRC networks can take weeks-to-months to resolve (Novkovic et al., 2016; Gregory et al., 2017). The molecular mechanisms are unknown; certainly both models induce a loss of various cell types including T cells, B cells and dendritic cells that are dependent on FRCs for survival (Cremasco et al., 2014; Denton et al., 2014), but there is nonetheless a clear dependence on FRCs, direct or indirect, for maintenance of monocyte and macrophage numbers within lymph nodes.
While mouse and human FRCs exhibit some clear molecular differences in regulation of T cell activation (Knoblich et al., 2018), our results showed that both the effects of FRCs on macrophages and monocytes, and a CSF1R-signalling mechanism are strongly conserved between mice and humans. There was one notable difference: in mice, LPS stimulation further enhanced the capacity of FRCs to promote M2 macrophage differentiation, while in human FRC:monocyte co-cultures, it did not. The basis for this remains unclear, but differences in cell source (human PBMC vs mouse bone marrow) and markers may play a role (Murray et al., 2014).
The biology of human FRCs is still largely unexplored, and it is still unclear which subset/s of FRCs are best represented through in vitro culture, highlighting the importance of in vivo follow up. Recent findings (Huang et al., 2018; Rodda et al., 2018; Zhang et al., 2018) have shown that mouse FRC subsets include a distinct medullary population that regulates plasma cell function via production of APRIL and IL-6. Medullary macrophage populations are reportedly CSF1-independent in mice (Witmer-Pack et al., 1993; Cecchini et al., 1994; MacDonald et al., 2010). These sinusoidal and medullary myeloid populations have recently been shown to respond to RANKL signalling from marginal reticular fibroblasts and LECs (Camara et al., 2019), with LEC-derived CSF1 shown to regulate sinus-lining macrophage populations (Mondor et al., 2019). T zone macrophages were relatively recently defined (Baratin et al., 2017), and whether they specifically rely on CSF1 is not yet known. CSF1 and IL-34 were expressed in scRNA sequencing analysis of freshly isolated human Gremlin1+ fibroblasts, further supporting our in vitro microarray analysis (Kapoor et al., 2021)All fibroblast subsets, including T zone subsets, expressed either CSF1 or IL-34 in mice and we demonstrated protein expression of these ligands on the whole FRC population when grown in culture. IL-34 and CSF1 both bind the CSF1R and possess similar ability to promote macrophage differentiation, though their roles diverge beyond this point, driving differential cytokine secretion by macrophages (Boulakirba et al., 2018).
Using complementary mouse and human studies, our data shows that macrophages interact intimately with FRCs in vivo and ultimately rely on FRCs for survival. Provision of CSF1R ligands, resulting in increased survival and M2 differentiation was observed in vitro. FRCs are poised to swiftly respond to inflammation through TLR4 signalling, upregulating factors relevant to myeloid cell function and localization. CSF1 expression was not altered with LPS treatment, and both mouse and human FRCs switched to a CSF1R-independent mechanism of myeloid cell support in the presence of LPS. Importantly, in vivo depletion experiments revealed that an intact stromal network is critically important to the maintenance of macrophages and monocytes. These findings show that FRCs provide microenvironmental support to macrophages and monocytes.
Contact for Reagent and Resource Sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Anne Fletcher (Anne.L.Fletcher{at}monash.edu).
Experimental Model and Subject Details
Experimental Animals
C57BL6J mice were obtained from Monash Animal Services and housed at the Animal Research Laboratories at Monash University, Clayton. CCL19-Cre mice (Chai et al., 2013) were crossed with Rosa26-iDTR mice and maintained at the Peter Doherty Institute, The University of Melbourne. DM2 mice expressing the diphtheria toxin receptor (DTR) under the regulatory elements of the murine Fap gene (Roberts et al., 2013; Denton et al., 2014) were housed at the University of Birmingham Biomedical Services Unit. For single cell RNA-Seq, BAC-transgenic C57BL/6N-Tg (Ccl19-Cre)489Biat (Ccl19-Cre) mice (Chai et al., 2013) have been previously described. These mice were on the C57BL/6 genetic background, were maintained in individually ventilated cages and were used between 8 and 10 weeks of age. Experiments were performed in accordance with federal and cantonal guidelines (Tierschutzgesetz) under permission numbers SG07/19 following review and approval by the Cantonal Veterinary Office (St. Gallen, Switzerland). All mice were specific pathogen– free and cared for in accordance with institutional guidelines. All experiments received approval from relevant institutional ethics committees.
Human tissues
Palatine tonsils were obtained from consenting donors from the National Disease Research Interchange (NDRI) resource centre or Human Biomaterials Resource Centre (HBRC), Birmingham (HTA licence 12358, 15/NW/0079), under project approval number REC/RG/HBRC/12-071. All tissues were obtained and utilised in accordance with institutional guidelines and according to the principles expressed in the Declaration of Helsinki.
Method Details
Lymph node digestion and FRC purification
Axillary, mesenteric and brachial lymph nodes were harvested from 5 – 10 euthanized C57BL6J mice and digested according to published protocols (Fletcher et al., 2011). Human tonsils were cut into 5mm segments and enzymatically digested according to previous protocols (Fletcher et al., 2011). Human or mouse cell suspensions were seeded at approximately 2900 cells/cm2 in αMEM complete media (Invitrogen) and cultured in 10% batch selected FBS (Sigma Aldrich) and 1% penicillin/streptomycin (Invitrogen) for 1 passage before weaning to antibiotic free conditions. Cells were harvested with 0.2% Trypsin with 5mM EDTA (Invitrogen, USA). Mouse FRCs required sorting using MACS (Miltenyi Biotec), with anti-mouse CD45 and CD31 magnetic beads, according to manufacturer’s instructions. Cells were quantified using a hemocytometer, and assessed for viability using trypan blue. Viability and purity of samples were routinely >95%.
RNA-Seq and microarray analysis
Data from freshly isolated and cultured mouse FRCs is available at NCBI GEO accession number GSE15907 (Malhotra et al., 2012) and GSE60111 (Fletcher et al., 2014) and processed as described. Data from human FRCs is accessible at monash.figshare.com doi: 10.4225/03/ 5a2dae0c9b455 and was prepared and processed as described (Knoblich et al., 2018). All heatmaps were generated using Morpheus matrix visualization software (Broad Institute, https://software.broadinstitute.org/morpheus).
Mice and LPS administration
Ccl19-Cre x R26R-EYFP mice were subcutaneously injected in one flank with 5 mg of LPS from E.coli (Sigma) together with 100 mg Ovalbumin grade VI (Sigma). Mice were sacrificed at day 3 post administration and brachial lymph nodes were removed for further lymph node stromal cell isolation and scRNA sequencing analysis.
Preparation of stromal cells
Brachial lymph nodes were transferred into a 24-well dish filled with RPMI 1640 medium containing 2% FCS, 20 mM Hepes (all from Lonza), 1 mg/ml Collagenase D (Sigma), 25 μg/ml DNAse I (Applichem) and Dispase (Roche). Dissociated tissues were incubated at 37°C for 30 minutes. After enzymatic digestion, cell suspensions were washed with PBS containing 0.5% FCS and 10 mM EDTA. Stromal cells were enriched by depleting CD45+ hematopoietic cells and TER119+ erythrocytes using MACS microbeads (Miltenyi, Germany) as described previously (Chai et al., 2013). Cells were sorted for EYFP+CD31−CD45− reticular cells and processed using the 10x Chromium (10X Genomics) system. Two biological replicates of LPS-treated and two biological replicates of naïve controls were used for the analysis.
Droplet-based single cell RNA-seq analysis
Isolated cells were sorted for EYFP+ reticular cells and processed on a 10x Chromium (10X Genomics) (62) to generate cDNA libraries. The processing followed the recommended protocol for the Chromium Single Cell 3’ Reagent Kit (v3 Chemistry) and the generated libraries were sequenced on an Illumina NovaSeq 6000 sequencing system at the Functional Genomic Center Zurich. Sequencing files were pre-processed using CellRanger (v3.0.2) (Zheng et al., 2017) with Ensembl GRCm38.94 release as reference and damaged cells or duplicates were removed in a subsequent quality control using scater R/Bioconductor package (v1.11.2) (McCarthy et al., 2017) running in R v3.6.0. In detail, cells were excluded if they had very high or low UMI counts or total number of detected genes (more than 2.5 median absolute deviations from the median across all cells) or high mitochondrial gene content (more 2.5 median absolute deviations above the median across all cells). In addition, cells expressing one of the genes Top2a, Mki67, Pclaf or Cenpf were excluded as proliferating cells resulting in a total of 10 019 cells from naïve lymph nodes and 14 427 cells from LPS treated mice.
Further downstream analysis was performed using functions from the Seurat package (v3.1.1) (Stuart et al., 2019) running in R v3.6.1 including data normalization, scaling, dimensional reduction with PCA and UMAP and graph-based clustering. Clusters were characterized based on marker genes and conditions were compared based on differentially expressed genes inferred from Wilcoxon test as implemented in the FindMarkers function (Seurat v3.1.1) (Stuart et al., 2019). Finally, functional differences between conditions were summarized based on a gene set enrichment analysis. For this all genes were ranked based on a signal-to-noise ratio statistic calculated on normalized expression values. Resulting ranked gene lists were used as input for GSEA-Preranked (v7.0.3) in a GenePattern notebook (Reich et al., 2017) with gene sets from the mSigDB (v7.0) C2 collection. Accession number: E-MTAB-10908.
In vivo FRC ablation
CCL19-Cre x Rosa26-iDTR (CCL19-DTR) mice and Cre-negative Rosa26-iDTR+ control mice received two injections of Diphtheria toxin (DTx) at 10ng/g of body weight, 24h apart. Skin draining LN were harvested at different days following DTx treatment and digested in RPMI containing 2% FCS, 1mg/mL collagenase D and 0.1mg/mL DNAse for 25 min at 37C. LN were further incubated for 15 min with the addition of 0.8mg/mL dispase (67). Single cell suspensions were filtered (70μm) before staining with antibodies for flow cytometry .
Diphtheria toxin receptor (DTR) FAP+ DM2 mice received 25ng/g diphtheria toxin (List Biological Laboratories) i.p. on days 0, 2 and 4, and were euthanised on day 6. Lymph nodes (axillary, brachial, inguinal, mesenteric) were dissected and enzymatically digested for flow cytometric analysis. Lymph node cell suspensions were labelled with cocktail containing anti-mouse conjugated antibodies. Cells were then analysed on a BD LSR Fortessa using Flowlogic software version 1.7 (Inivai Technologies).
Luminex bead assay
Human FRCs were plated overnight at a density of 4×104 cells/ 0.32cm2 and left to adhere to surface. Cells were then stimulated with either 1ug/ml LPS (Serotype O111:B4; Sigma Aldrich) over a 24 hour time course. Specific wells were pre-cultured for 2 hours with either 10mg/ml of TLR4 inhibitor (CLI-095, Invivogen) or 10mg/ml PI3 Kinase inhibitor (LY294002, Invivogen) as controls. Supernatants were removed at stated time points and secreted cytokine quantities were measured using Luminex Bead technology according to Bioplex protocols and cytokine Kit (Biorad).
Phosphorylation of MAPK p38 and NFκB p65 in TLR4 stimulation
Human FRCs were plated at 4 × 104 cells/ 0.32cm2 in complete αMEM media and left to adhere. After 4 hours, media was removed and replaced with serum free media overnight to induce basal phosphorylation levels. The following day, cells were stimulated with 1 μg/ml LPS (Serotype O111:B4; Sigma Aldrich) and cells were removed from wells at stated time points with 0.2% Trypsin with 5mM EDTA (Invitrogen), then fixed with 4% PFA and methanol, and labelled immediately according to protocols from Cell Signalling Technologies (CST) with unconjugated rabbit anti-human phosphorylated NF-κB p65 (Clone: Ser536 93H1) followed by goat anti-rabbit IgG (H+L) Alexa Fluor 488 (Thermo Fisher). For p38, the CST staining protocol was followed with mouse anti-human phosphorylated p38 MAPK (Clone: Thr180/Thr182) used with goat anti-mouse IgG (H+L) conjugated to Alexa Fluor 488 (Thermo Fisher) as a secondary. Cells were then labelled with specific stromal markers podoplanin and CD31 and were analysed using FACS Canto (BD Biosciences) with analysis undertaken using Flowlogic software version 1.7 (Inivai Technologies).
Immunofluorescent microscopy
Murine axillary, mesenteric and brachial lymph nodes, or human tonsils, were snap-frozen in OCT (Sakura) and stored at −80C. 7μm sections were cut, air dried for 30 minutes, then fixed in chilled acetone for 20 minutes, followed by 2x PBS washes. Sections were stained with primary antibodies (Table 1) for 20-30 minutes in a dark humidified box at room temperature, followed by 2x 5 minute washes in PBS. Secondary antibodies were added for 20-30 minutes at room temperature. Slides were washed twice in PBS, with DAPI added to sections for 2 minutes at room temperature prior to mounting (ProLong Gold anti-fade mountant, Thermo Fisher). Slides were imaged on a Zeiss LSM 800 confocal scanning microscope.
Cell morphology analysis
The perimeter of cells was manually drawn and the perimeter and area calculated using ImageJ software. To ensure cells were seen in cross-section, only those cells showing a DAPI+ nucleus were chosen for measurement. The morphology index is an inverse roundness metric calculated as perimeter2/4πArea, where the minimum value of 1 would be returned for a perfectly circular cell, while the further a cell deviates from circularity, the larger the values (Pinner and Sahai, 2008) .
FRC and monocyte co-cultures
Human FRCs at passage 3 were plated overnight at 2×104 cells per 0.32cm2 well. The following day, 4×105 human PBMCs were isolated from healthy donors using Ficoll-Paque or Lymphoprep according to the manufacturer’s instructions, and added to appropriate wells. Where indicated, 10ng/ml of LPS (Cell Signalling Technology, USA), 0.1μg/ml human anti-CSF1R (Clone: 61701 – R&D systems) or anti-human isotype control (Clone: 11711 – R&D systems) were added. Cells were incubated for 72 hours at 37°C, then harvested, quantified using a Countess (Thermo Fisher), labelled with anti-human myeloid and stromal markers (Table 1) and analysed by flow cytometry on a LSR Fortessa (BD Biosciences).
For mouse assays, 2×105 mouse FRCs were left to adhere to 6-well plates overnight. The next day 1×106 mouse bone marrow cells were added to each well, with 100U/ml recombinant M-CSF (Peprotech), 10μg/ml purified anti-mouse MCSFR/CD115 (Clone: AFS98 – eBioscience) or 1μg/ml LPS (Cell Signalling Technology, USA) added as required. Some wells were pre-treated for 30 minutes with 10μg/ml TLR4 inhibitor TAK-242 (Invivogen). Cells were harvested after 4 days, quantified using a Z2 Coulter Counter (Beckman Coulter, USA) and labelled for flow cytometry then analysed on a FACS Canto (BD Biosciences) using Flowlogic software version 1.7 (Inivai Technologies).
Monocyte populations were described as classical (CD11b+Ly6Chi in mouse, CD14+CD16− in human) and non-classical (CD11b+Ly6Clow in mouse, CD14−/loCD16+ in human). Macrophage populations were defined in human experiments as M1 macrophages (CD64+HLA-DR+CD206−) and M2 macrophages (CD64−CD206+). Mouse macrophages were defined as CD11b+F480+.
Statistical analysis
A one-way ANOVA with Tukey’s post-test was used to compare 2 or more groups of parametric data. Mann-Whitney U test was used for non-parametric data. Normality was assessed using D’Agostino-Pearson. P< 0.05 was considered significant. To assess fold change data, a Wilcoxon rank test with a ratio paired T test was used. For pathway analysis, the top 15 upregulated genes with LPS treatment were used for KEGG analysis. Pathways shown yielded P values <0.05, with FDR and Benjamini-Hochberg values <0.05).
Author contributions
JD and ALF performed experiments, designed the study, analysed data and wrote the paper; TH designed the study, analysed data and wrote the paper; KK, HWC, YA, JDDC, JA, ED, performed experiments; ML, CPS and DR performed and analysed experiments; AD and SJT provided essential reagents; RB contributed essential guidance; MS designed experiments, SM and BL provided essential reagents, designed and analysed experiments.
Declaration of interests
The authors declare no competing interests.
Supplementary Figure legends
Acknowledgements
This work was supported by a Wellcome Trust Seed Award and a Birmingham Fellowship to AF, and by Monash University. TH was supported by a NHMRC Career Development Fellowship. The authors thank D. Fearon for usage of the DM2 mice model in this study. We gratefully acknowledge the contribution made by the University of Birmingham’s Human Biomaterials Resource Centre, which has been supported through a Birmingham Science City – Experimental Medicine Network of Excellence project.