Notch and TLR signaling coordinate monocyte cell fate and inflammation

Conventional Ly6Chi monocytes have developmental plasticity for a spectrum of differentiated phagocytes. Here we show, using conditional deletion strategies in a mouse model of Toll-like receptor (TLR) 7-induced inflammation, that the spectrum of developmental cell fates of Ly6Chi monocytes, and the resultant inflammation, is coordinately regulated by TLR and Notch signaling. Cell-intrinsic Notch2 and TLR7-Myd88 pathways independently and synergistically promote Ly6Clo patrolling monocyte development from Ly6Chi monocytes under inflammatory conditions, while impairment in either signaling axis impairs Ly6Clo monocyte development. At the same time, TLR7 stimulation in the absence of functional Notch2 signaling promotes resident tissue macrophage gene expression signatures in monocytes in the blood and ectopic differentiation of Ly6Chi monocytes into macrophages and dendritic cells, which infiltrate the spleen and major blood vessels and are accompanied by aberrant systemic inflammation. Thus, Notch2 is a master regulator of Ly6Chi monocyte cell fate and inflammation in response to TLR signaling.


Introduction
Infectious agents or tissue injury trigger an inflammatory response that aims to eliminate the inciting stressor and restore internal homeostasis (J. Bonnardel & Guilliams, 2018). The mononuclear phagocyte system (MPS) is an integral part of the inflammatory response and consists of the lineage of monocytes and macrophages (MF) and related tissue-resident cells. A key constituent of this system are monocytes of the major (classic) monocyte subtype, in mice called Ly6C hi monocytes. They originate from progenitor cells in the bone marrow (BM), circulate in peripheral blood (PB) and respond dynamically to changing conditions by differentiation into a spectrum of at least three distinct MPS effector phagocytes: MF, dendritic cells (DC), and monocytes with patrolling behavior (Arazi et al., 2019;J. Bonnardel & Guilliams, 2018;Chakarov et al., 2019;Gamrekelashvili et al., 2016;Hettinger et al., 2013). The diversity of monocyte differentiation responses is thought to be influenced by environmental signals as well as tissue-derived and cell-autonomous signaling mechanisms to ensure context-specific response patterns of the MPS (Okabe & Medzhitov, 2016). However, the precise mechanisms underlying monocyte cell fate decisions under inflammatory conditions are still not fully understood.
When recruited to inflamed or injured tissues, Ly6C hi monocytes differentiate into MF or DC with a variety of phenotypes and function in a context-dependent-manner and regulate the inflammatory response (Krishnasamy et al., 2017;Xue et al., 2014).
Toll-like receptor 7 (TLR7) is a member of the family of pathogen sensors expressed on myeloid cells. Originally identified as recognizing imidazoquinoline derivatives such as Imiquimod (R837) and Resiquimod (R848), TLR7 senses ssRNA, and immunecomplexes containing nucleic acids, in a Myd88-dependent manner during virus defense, but is also implicated in tissue-damage recognition and autoimmune disorders (Kawai & Akira, 2010). TLR7-stimulation induces cytokine-production in both mouse and human patrolling monocytes and mediates sensing and disposal of damaged endothelial cells by Ly6C lo monocytes (Carlin et al., 2013;Cros et al., 2010), while chronic TLR7-stimulation drives differentiation of Ly6C hi monocytes into specialized macrophages and anemia development (Akilesh et al., 2019

TLR and Notch signaling promote monocyte conversion
We first studied the effects of TLR and/or Notch stimulation on monocyte conversion in a defined in vitro system (Gamrekelashvili et al., 2016). Ly6C hi monocytes isolated from the bone marrow of Cx3cr1 gfp/+ reporter mice (GFP + ) were cultured with recombinant Notch ligand DLL1 in the presence or absence of the TLR7/8 agonist R848 and analyzed after 24 hours for the acquisition of key features of Ly6C lo monocytes (Gamrekelashvili et al., 2016;Hettinger et al., 2013). In contrast to control conditions, cells cultured with DLL1 showed an upregulation of CD11c and CD43, remained mostly MHC-II negative, and expressed transcription factors Nr4a1 and Pou2f2, leading to a significant, five-fold increase of Ly6C lo cells, consistent with enhanced monocyte conversion. Cells cultured with R848 alone showed a comparable phenotype response, both qualitatively and quantitatively ( Figure 1A-C). Interestingly, on a molecular level, R848 stimulation primarily acted on Pou2f2 induction, while Notch stimulation primarily induced Nr4a1. Furthermore, the combination of DLL1 and R848 strongly and significantly increased the number of CD11c + CD43 + Ly6C lo cells above the level of individual stimulation and significantly enhanced expression levels of both transcriptional regulators Nr4a1 and Pou2f2 ( Figure 1A-C), suggesting synergistic regulation of monocyte conversion by TLR7/8 and Notch signaling. By comparison, the TLR4 ligand LPS also increased Ly6C lo cell numbers and expression levels of Nr4a1 and Pou2f2. However, the absolute conversion rate was lower under LPS and there was no synergy with DLL1 ( Figure 1D and E).
Since monocyte conversion is regulated by Notch2 in vitro and in vivo (Gamrekelashvili et al., 2016), we next tested TLR-induced conversion in Ly6C hi monocytes with conditional deletion of myeloid Notch2 (N2 ΔMy ). Both, wt and N2 ΔMy monocytes showed comparable response to R848, but conversion in the presence of DLL1, and importantly, also DLL1-R848 co-stimulation was significantly impaired in mutant cells ( Figure 1F). This suggests independent contributions of TLR and Notch signaling to monocyte conversion.
To study whether the TLR stimulation requires Myd88 we next tested Ly6C hi monocytes with Myd88 loss-of-function (Myd88 -/-). Compared to wt cells, Myd88 -/monocytes showed strongly impaired conversion in response to R848 but a conserved response to DLL1. The response to DLL1-R848 co-stimulation, however, was significantly impaired ( Figure 1G). Furthermore, expression of Nr4a1 and Pou2f2 by R848 was strongly reduced in Myd88 -/monocytes with or without DLL1 co-stimulation, while DLL1dependent induction was preserved ( Figure 1H). Thus, Notch and TLR signaling act independently and synergistically to promote monocyte conversion.
To address the role of TLR stimulation for monocyte conversion in vivo we adoptively transferred sorted Ly6C hi monocytes from CD45.2 + GFP + mice into CD45.1 + congenic recipients, injected a single dose of R848 and analyzed transferred CD45.2 + GFP + cells in BM and Spl after two days ( Figure 2A). Stimulation with R848 significantly promoted conversion into Ly6C lo monocytes displaying the proto-typical Ly6C lo CD43 + CD11c + MHC-II lo/phenotype ( Figure 2B and C). In contrast, transfer of Myd88 -/-Ly6C hi monocytes resulted in impaired conversion in response to R848 challenge ( Figure 2D and E). Together, these data indicate that TLR and Notch cooperate in the regulation of monocyte conversion.

Notch2-deficient mice show altered myeloid inflammatory response
To characterize the response to TLR stimulation in vivo we applied the synthetic TLR7 agonist Imiquimod (IMQ, R837) in a commercially available crème formulation (Aldara) daily to the skin of mice (El Malki et al., 2013;van der Fits et al., 2009) and analyzed the systemic inflammatory response in control or N2 ΔMy mice (Gamrekelashvili et al., 2016) ( Figure 3A). While treatment with IMQ induced comparable transient weight loss and ear swelling in both genotypes ( Figure S1A), splenomegaly in response to treatment was significantly more pronounced in N2 ΔMy mice ( Figure S1B).
To characterize the spectrum of myeloid cells in more detail we next performed flow cytometry of PB cells with a dedicated myeloid panel (Gamrekelashvili et al., 2016) and subjected live Lin -CD11b + GFP + subsets to unsupervised t-SNE analysis ( Figure 3B). This analysis strategy defined 5 different populations, based on single surface markers: Ly6C + , CD43 + , MHC-II + , F4/80 hi and CD11c hi ( Figure 3C). Applying these 5 gates to samples from separate experimental conditions identified dynamic alterations in blood myeloid subsets in response to IMQ, but also alterations in N2 ΔMy mice ( Figure 3D).
Specifically, abundance and distribution of Ly6C + cells, containing classical monocytes, in response to IMQ were changed to the same extend in both genotypes. In contrast, the MHC-II + and F4/80 hi subsets were more abundant in N2 ΔMy mice, but also showed more robust changes in response to IMQ. On the other hand, the CD43 + subset, containing the patrolling monocyte subset, showed prominent enrichment in wt mice, but was less abundant and showed diminished distribution changes after IMQ treatment in N2 ΔMy mice ( Figure 3D).
To analyze the initially defined subsets more precisely we applied a multi-parameter gating strategy to define conventional cell subsets ( Figure S1C and Table S1) (Gamrekelashvili et al., 2016).
In response to IMQ, Ly6C hi monocytes in wt mice increased transiently in blood, and this response was not altered in mice with conditional Notch2 loss-of function ( Figure   3E). In contrast, while Ly6C lo monocytes robustly increased over time with IMQ treatment in wt mice, their levels in N2 ΔMy mice were lower at baseline (Gamrekelashvili et al., 2016) and remained significantly reduced throughout the whole observation period ( Figure 3E and F and S2A and B). At the same time, untreated N2 ΔMy mice showed increased levels of MHC-II + atypical monocytes ( Figure 3E and F and S2A and B) (Gamrekelashvili et al., 2016), accompanied by a significant wave of macrophages appearing in blood and spleen at d5 ( Figure 3E and F and S2A and B). This was followed by a peak in the DC population at d7 ( Figure 3E and F and S2A and B). These latter changes did not occur in bone marrow but were only observed in the periphery ( Figure S2C and D

Global gene expression analysis identifies expansion of macrophages in blood of Notch2 deficient mice during acute inflammation
To characterize more broadly the gene expression changes involved in monocyte differentiation during inflammation, we next subjected monocyte subsets from PB of wt and N2 ΔMy mice after Sham or IMQ treatment ( Figure S3A) to RNA-sequencing and gene expression analysis. After variance filtering and hierarchical clustering, 600 genes were differentially expressed between 6 experimental groups ( Figure 4A). Comparative gene expression analysis of Ly6C lo cell subsets during IMQ treatment identified 373 genes significantly up-or down-regulated with Notch2 loss-of-function (Pvalue <0.01, Figure 4C-F), which were enriched for phagosome formation, complement system components, Th1 and Th2 activation pathways and dendritic cell maturation by ingenuity canonical pathway analysis ( Figure S3B). Notably, signatures for autoimmune disease processes were also enriched (Table S2) (Table S3). Overall, these data suggest regulation of Ly6C hi monocyte cell fate and inflammatory responses by Notch2.
Furthermore, changes in cell populations resulted in altered systemic inflammatory response patterns. Levels of TLR-induced cytokines and chemokines, such as TNF-α, CXCL1, IL-1β, IFN-α, were elevated to the same extend in wt and N2 ΔMy mutant mice in response to IMQ treatment, suggesting normal primary TLR-activation ( Figure S3C).
However, circulating levels of chemokines produced by Ly6C lo monocytes (Carlin et al., 2013), such as CCL2, CCL3, CXCL10, and IL-10 were higher in wt mice compared to N2 ΔMy mice, while the levels of pro-inflammatory cytokines IL-17A, IL-6 and GMCSF were strongly enhanced in N2 ΔMy mice, confirming systemic pro-inflammatory alterations in addition to cellular changes in Notch2 loss-of-function mice in response to IMQ ( Figure 4G, and S3C).

Notch2-deficiency promotes macrophage differentiation of Ly6C hi monocytes
To match the observed gene expression pattern of inflammatory Ly6C lo cells from wt and N2 ΔMy mice under IMQ treatment with previously described cells of the monocyte- to Lyve-1 hi MHC-II lo -like macrophages, typically associated with blood vessels ( Figure   5B and C). Together, these data demonstrate a cell fate switch from Ly6C lo monocytes to macrophages in the absence of Notch2.

Notch2 regulates cell fate decisions of Ly6C hi monocytes during inflammation
In the steady-state, Ly6C hi monocytes differentiate into Ly6C lo monocytes and this process is regulated by Notch2 (Gamrekelashvili et al., 2016). In order to confirm that Notch2 controls differentiation potential of Ly6C hi monocytes in response to TLR stimulation we performed adoptive transfer of CD45.2 + wt or N2 ΔMy BM Ly6C hi monocytes into IMQ-treated CD45.1 + congenic recipients and analyzed the fate of donor cells after three days ( Figure 5D). Unsupervised t-SNE analysis of flow cytometry data showed an expanded spectrum of expression patterns in cells from N2 ΔMy donors compared to wt controls ( Figure 5E). More precisely, Ly6C hi monocytes from wt mice converted preferentially to Ly6C lo monocytes (Ly6C lo F4/80 lo/-CD11c + CD43 + MHC-II lo/phenotype) during IMQ treatment ( Figure 5F and G). In contrast, conversion of Notch2deficient Ly6C hi to Ly6C lo monocytes was strongly impaired, but the development of donor-derived F4/80 hi macrophages was strongly enhanced ( Figure 5H). Furthermore, this expansion of macrophages was also observed in aortas of Notch2-mutant mice in vivo, which was followed by granulocyte (GC) infiltration in aortas of IMQ-treated N2 ΔMy mice ( Figure 5J and K). Adoptive transfer studies confirmed that MF in IMQ-treated aortas originated from N2 ΔMy Ly6C hi monocytes ( Figure 5L and M). These data confirm that Notch2 is a master regulator of Ly6C hi monocyte differentiation potential, regulating a switch between Ly6C lo monocyte or macrophage cell fate during inflammation. These data also demonstrate that in the context of inactive myeloid Notch2 signaling, TLRstimulation results in systemic pro-inflammatory changes and vascular inflammation.  (Guilliams, Mildner, & Yona, 2018). In the steady-state, a subset of Ly6C hi monocytes converts to Ly6C lo monocytes in mice and humans, which is regulated by Notch2 and the endothelial Notch ligand Delta-like 1 (Dll1) (Gamrekelashvili et al., 2016;Patel et al., 2017;Yona et al., 2013). However, when recruited into tissues, Ly6C hi monocytes can give rise to two types of monocyte-derived resident tissue macrophages ( (Gamrekelashvili et al., 2016) and Ly6C hi monocyte numbers in Notch2-mutant mice are not altered at steady-state or during TLR stimulation, which suggests cell fate decision at or below the level of Ly6C hi monocytes. However, whether Notch2 loss-of-function differentially impacts monocyte subset composition requires further study.

Discussion
While our current data clearly demonstrate that Notch2 loss-of-function promotes macrophage development from Ly6C hi monocytes and a pro-inflammatory milieu during TLR stimulation, we have previously shown that Dll1-Notch signaling promotes maturation of anti-inflammatory macrophages from Ly6C hi monocytes in ischemic muscle (Krishnasamy et al., 2017). Furthermore, Dll4-Notch signaling initiated in the liver niche was recently shown to promote Kupffer cell development after injury (Johnny Bonnardel et al., 2019;Sakai et al., 2019) or to promote pro-inflammatory macrophage development (Xu et al., 2012). This suggests that the role of Notch is ligand-, cell-and context-specific, which emphasizes the differential effects of specific ligand-receptor combinations (Benedito et al., 2009). Our data demonstrate that Notch2 is a master regulator of Ly6C hi monocyte cell fate during inflammation, which contributes to the nature of the inflammatory response.
Lastly, our data also reveal in important function of myeloid Notch2 for regulation of systemic and vascular inflammation with potential implications for autoimmune disease.

Tissue and cell preparation
For single cell suspension mice were sacrificed and spleen, bone marrow, blood and aortas were collected. Erythrocytes were removed by red blood cell lysis buffer (BioLegend) or by density gradient centrifugation using Histopaque 1083 (Sigma-Aldrich). Aortas were digested in DMEM medium supplemented with 500U/ml Collagenase II (Worthington). After extensive washing cells were resuspended in PBS containing 10%FCS and 2mM EDTA kept on ice, stained and used for flow cytometry or for sorting.

Flow cytometry and cell sorting
Non-specific binding of antibodies to Fc-receptors was blocked with anti-mouse CD16/CD32 (TruStain fcX from BioLegend) in single cell suspensions prepared from Spl, PB or BM. After subsequent washing step cells were labelled with primary and secondary antibodies or streptavidin-fluorochrome conjugates and were used for flow cytometry analysis (LSR-II, BD Biosciences) or sorting (FACSAria; BD Biosciences or MoFlo XDP; Beckman Coulter). Antibodies and fluorochromes used for flow cytometry are described in Table S5. Flow cytometry data were analyzed using FlowJo software (FlowJo LLC). Initially cells were identified based on FSC and SSC characteristics.

RNA isolation, library construction, sequencing and analysis
Peripheral blood monocyte subpopulations were sorted from Aldara treated mice or untreated controls ( Figure S3A)  Gene set enrichment analysis (GSEA) (Mootha, Lindgren, et al., 2003; was performed on 373 DEGs using GSEA software (Broad institute) and C5 GO biological process gene sets (Liberzon et al., 2015) from MsigDB with 1000 gene set permutations for computing P-values and FDR.

In vitro conversion studies
96 well flat bottom plates were coated at room temperature for 3 hours with IgG-Fc or DLL1-Fc ligands (all from R&D) reconstituted in PBS. Sorted BM Ly6C hi monocytes were cultured in coated plates and were stimulated with Resiquimod (R848, 0.2µg ml -1 , Cayman Chemicals) or LPS (0.2µg ml -1 , E. coli O55:B5 Sigma-Aldrich) in the presence of M-CSF (10ng ml -1 , Peprotech) at 37 C for 24 hours. One day after culture, cells were harvested, stained and subjected to flow cytometry. Frequency of Ly6C lo monocyte-like cells (CD11b + GFP + Ly6C lo/-CD11c lo MHC-II lo/-CD43 + ) in total live CD11b + GFP + cells served as an indicator of conversion efficiency and is shown in the graphs. Alternatively, cultured cells were harvested and isolated RNA was used for gene expression analysis.

Induction of acute systemic inflammation using Aldara
Mice were anesthetized and back skin was shaved and depilated using depilating crème. Two days after 50mg/mouse/day Aldara (containing 5% Imiquimod, from Meda) or Sham crème were applied on depilated skin and right ear (where indicated) for 4-5 consecutive days (El Malki et al., 2013;van der Fits et al., 2009). Mouse weight and ear thickness were monitored daily. Mice were euthanized on the indicated time points after start of treatment (day 0, 5, 7 and 12) PB, Spl, BM and aortas were collected for further analysis.

Quantitative real-time PCR analysis
Total RNA was purified from cell lysates using Nucleospin RNA II kit (Macherey Nagel).

Statistical analysis
Results are expressed as mean ± standard error of mean (SEM). N numbers are biological replicates of experiments performed at least three times unless otherwise indicated. Significance of differences was calculated using unpaired, two-tailed Student's t-test with confidence interval of 95%. For comparison of multiple experimental groups one-way or two-way ANOVA and Bonferroni's multiplecomparison test was performed.

Data availability
Data from RNA sequencing have been deposited to NCBI's Gene Expression Omnibus and are available under the accession number GSE147492.
All the relevant data are available upon request.           Tables   Table S1. Surface phenotype signatures for identification of distinct myeloid populations in vivo.