MyD88-dependent TLR signaling oppositely regulates hematopoietic progenitor and stem cell formation in the embryo

Hemogenic endothelial (HE) cells in the dorsal aorta undergo an endothelial to hematopoietic transition (EHT) to form lympho-myeloid biased progenitors (LMPs), pre-hematopoietic stem cells (pre-HSCs) and adult-repopulating HSCs. These briefly accumulate in intra-arterial hematopoietic clusters (IAHCs) before being released into the circulation. It is generally assumed that the number of IAHC cells correlates with the number of HSCs. Here we show that changes in the number of IAHC cells, LMPs, and HSCs can be uncoupled. Mutations impairing MyD88-dependent toll-like receptor (TLR) signaling decreased the number of IAHC cells and LMPs but increased the number of HSCs in the aorta-gonad-mesonephros region of mouse embryos. TLR4 -deficient embryos generated normal numbers of HE cells but the proliferation of IAHC cells was decreased. Loss of MyD88-dependent TLR signaling in innate immune myeloid cells had no effect on IAHC cell numbers. Instead, TLR4 deletion in endothelial cells recapitulated the phenotype observed with germline deletion, demonstrating that MyD88-dependent TLR signaling in endothelial cells and/or in IAHCs regulates the balance between generating LMPs and HSCs. Summary Statement Toll-like receptor signaling in endothelial cells restricts the number of hematopoietic stem cells but increases the number of committed progenitors and intra-arterial hematopoietic cluster cells by promoting their proliferation.


Introduction
Hematopoietic ontogeny occurs in multiple waves in which hematopoietic stem and progenitor cells (HSPCs) with different potentials differentiate from hemogenic endothelial (HE) cells via an endothelial to hematopoietic transition (EHT). HE cells in the extra-embryonic yolk sac undergo EHT to become committed erythro-myeloid (EMP) and lymphoid progenitors, while HE cells in the caudal arteries (dorsal aorta in zebrafish; dorsal aorta, vitelline and umbilical in mouse) differentiate into multipotent, lympho-myeloid biased progenitors (LMPs), HSCs, and HSC precursors called pre-HSCs (Auerbach et al., 1996;Hadland and Yoshimoto, 2018). Following EHT, the newly formed HSPCs in the major arteries briefly accumulate in intra-arterial hematopoietic clusters (IAHCs) consisting of cells that uniformly express endothelial markers, hematopoietic transcription factors such as RUNX1, and high levels of Kit (Bollerot et al., 2005).
Innate immune and inflammatory signaling pathways have been shown to regulate the number of HSPCs generated in the arteries of zebrafish and mouse embryos. Mutations in or morpholino knockdowns of genes encoding type I and II interferons (Li et al., 2014;Sawamiphak et al., 2014), tumor necrosis factor (TNF) (Espin-Palazon et al., 2014), inflammasomes (Frame et al., 2020), interleukin 1b (IL-1b) (Frame et al., 2020), granulocyte colony stimulating factor (G-CSF) (He et al., 2015), toll-like receptor 4 (TLR4) (He et al., 2015), nuclear factor kappa B (NF-kB) (Espin-Palazon et al., 2014;He et al., 2015) and RIG-1-like receptors (Lefkopoulos et al., 2020) reduced the number of myb:GFP + kdrl:mCherry + HSPCs in transgenic reporter zebrafish embryos and decreased the levels of runx1 and myb transcripts in the dorsal aorta (DA). Innate immune signaling is initiated by binding of ligands to pattern recognition receptors (PRRs). PRRs recognize pathogen-associated molecular patterns (PAMPs), which are small molecules common to a particular class of infectious agents, and host molecules referred to as danger-associated molecular patterns (DAMPs) which are endogenous proteins or small molecules that are released from the cell or the extracellular matrix in response to cell damage, stress, or death. Although, in theory, PAMPs from infectious agents might be responsible for activating innate immune signaling in mammalian or pre-hatched zebrafish embryos, the most plausible ligands for PRRs in the relatively sterile environment of the mammalian embryo are DAMPs.
Activation of toll-like receptors (TLRs) drives the production of type I interferons, TNF, IL-1b and G-CSF, and therefore directly or indirectly stimulates multiple inflammatory pathways (Monlish et al., 2016). TLRs signal through two intracellular pathways that involve either the MyD88 (myeloid differentiation primary response protein) and TIRAP (toll-interleukin 1 receptor domain containing adaptor protein) or the TRIF (toll IL-1 receptor (TIR)-domain-containing adaptor inducing interferon b) and TRAM (translocating chain-associated membrane protein) adaptor proteins.
Signaling from TLR4 on the plasma membrane through MyD88/TIRAP results in the nuclear localization of the CREB, AP1, and NF-kB transcription factors and expression of inflammatory cytokine genes such as TNF, IL-1b and G-CSF. Signaling through endosomal TLR4, TRIF/TRAM and the interferon regulatory factors (IRFs) primarily activates the expression of interferon genes.
Morpholino knockdown of the genes encoding TLR4 (tlr4bb) and MyD88 (myd88) in zebrafish embryos reduced the expression of runx1, the number of myb:GFP + kdrl:mCherry + HSPCs in the DA, and HSPCs in secondary sites of colonization including the caudal hematopoietic tissue and thymus (He et al., 2015). However, markers that discriminate between committed progenitors (LMPs) and pre-HSCs/HSCs are not available for zebrafish embryos, therefore how TLR4 signaling affects the balance between them is difficult. Understanding how this balance is regulated is important for specifically optimizing generation of the desired hematopoietic stem or progenitor cells from arterial HE cells ex vivo.
Here we examine TLR signaling to determine in which cells it is required, through which intracellular adaptors it signals to regulate arterial HSPC ontogeny, and the types of HSPCs it regulates. We show that signaling through the MyD88 arm of the TLR pathway increases the number of IAHC cells and LMPs, while restricting the number of multi-lineage adult-repopulating HSCs.

TLR4 signaling regulates RUNX1 levels and the number of IAHC cells.
It was reported that loss of TLR4 reduced expression of runx1 in the DA of zebrafish embryos and the number of RUNX1 + cells in the DA of mouse embryos (He et al., 2015). To examine whether the reduction in runx1 expression could be due to fewer RUNX1 + HE cells, lower RUNX1 protein levels in HE cells, or a decrease in IAHC cells we analyzed Tlr4 -/mouse embryos and their wildtype littermates by whole mount confocal microscopy ( Fig. 1A-F). IAHC cells were identified by their rounded appearance and expression of RUNX1, CD31, and high levels of c-Kit, while HE cells were flatter, located within the endothelial layer, and expressed RUNX1, CD31, and very low or undetectable levels of c-Kit. We found that the median fluorescence intensity of RUNX1 protein was lower in HE cells in the DA of E10.5 Tlr4 -/embryos compared to Tlr4 +/+ littermates but the numbers of RUNX1 + HE cells were equivalent ( Fig. 1A-C). The number of IAHC cells, on the other hand, was reduced by 42% in E10.5 Tlr4 -/embryos, and IAHC cells expressed lower levels of RUNX1 protein (Fig. 1D,E). The reduction in IAHC cell number is consistent with observations in tlr4bb morphant zebrafish embryos, which had fewer myb:GFP + kdrl:mCherry + HSPCs in the DA (He et al., 2015).
The DA contains approximately 0-5 IAHC cells at E9.5 and ~600 IAHC cells at E10.5 (Yokomizo and Dzierzak, 2010). Since the number of HE cells at a single snapshot in time is a function of the number that were originally specified minus the number that have undergone EHT, the optimal time for determining how many HE cells are specified is at E9.5, prior to the major egress of HE cells via EHT. The numbers of RUNX1 + HE cells were equivalent in E9.5 Tlr4 +/+ and Tlr4 -/embryos (Fig. 1F), and levels of RUNX1 in HE cells were slightly increased (Fig. 1G).
Therefore, TLR4 deficiency does not affect the number of HE cells that are specified at E9.5, but it decreases RUNX1 levels in HE cells at E10.5.
The decreased number of IAHC cells in E10.5 Tlr4 -/embryos could be caused by inefficient EHT, decreased proliferation, or increased apoptosis of IAHC cells. Inefficient EHT is unlikely as it would have resulted in an accumulation of HE cells in the DA in Tlr4 -/embryos at E10.5, which we did not observe. Tlr4 -/-IAHC cells were, however, less proliferative, as reflected by a smaller fraction that were positive for the proliferation marker Ki-67 ( Fig. 1H,I). Flow cytometric analysis also showed loss of TLR4 decreased proliferation, determined by EdU labeling, specifically of CD45 + IAHC cells, which at E10.5 lack pre-HSCs but are enriched for LMPs (Zhu et al., 2020) (Fig. 1K,L and Fig. S1). TLR4 deficiency did not affect the proliferation of endothelial cells, HE cells, or CD45 -IAHC cells, which at E10.5 include pre-HSCs (Fig. 1J,L). In summary, loss of TLR4 decreased RUNX1 protein levels in E10.5 HE and IAHC cells and also decreased the number and proliferation of CD45 + IAHC cells, but had no effect on the number of HE cells.
Both TLR4 and TLR2 are expressed in and regulate inflammatory signaling in endothelial cells (Salvador et al., 2016), therefore we determined whether TLR2 also regulates the number of IAHC cells. Loss of TLR2 reduced the number of IAHC cells in the DA to a similar extent as TLR4 deficiency (Fig. 2B). TLR2 signals exclusively through the MyD88 adaptor, while TLR4 signals through both the MyD88 and TRIF adaptors ( Fig. 2A). To determine which signaling arm of TLR4 was important for regulating the number of IAHC cells we enumerated IAHC cells in both Myd88 -/and TRIF-deficient (Ticam1 -/-) embryos. Loss of MyD88, but not TRIF, decreased the number of IAHC cells (Fig. 2B). Therefore, signaling via cell surface TLR2 and TLR4 through the MyD88 adaptor, but not by endosomal TLR4 through the TRIF adaptor, regulates the number of IAHC cells.

MyD88-dependent TLR signaling positively regulates the number of committed lymphoid progenitors but restrains the number of adult repopulating HSCs.
To determine whether loss of TLR signaling also decreased the number of adult-repopulating HSCs capable of multilineage engraftment, we transplanted 1 embryo equivalent (ee) of aortagonad-mesonephros (AGM) regions from E11.5 wild type, Tlr4 -/-, Tlr2 -/-, Myd88 -/-, and Ticam1 -/embryos into irradiated adult recipient mice, and measured contribution to peripheral blood and bone marrow 16 weeks post-transplant (Fig. S2). One E11.5 wild type AGM region contains ~1 adult repopulating HSC, and therefore, 1ee is a limiting dose (Kumaravelu et al., 2002). Despite the decrease in the number of IAHC cells, loss of TLR4, TLR2, and MyD88 increased the proportion of mice with multi-lineage engraftment (Fig. 2C), indicating that adult-repopulating HSCs were not depleted in the AGM regions of TLR4, TLR2, or MyD88-deficient mouse embryos but instead appeared to be increased. Loss of Ticam1, on the other hand, which did not significantly change IAHC cell numbers, did not affect multi-lineage engraftment (Fig. 2B,C).
To quantify the numbers of multi-lineage adult repopulating HSCs in embryos with mutations affecting MyD88-dependent TLR signaling, we transplanted a cohort of mice with limiting doses (0.3 ee and 1.0 ee) of dissociated cells from the AGM regions of wild type, Tlr4 -/and Myd88 -/embryos. There was a trend towards increased numbers of HSCs in the AGM regions of E11.5 Tlr4 -/embryos and a significant 4.2-fold increase in the number of multi-lineage repopulating HSCs in Myd88 -/embryos (Fig. 2D).
LMPs with the potential to produce B and T cells in culture are >10-fold more abundant in AGM regions of E10.5 embryos compared to HSCs and pre-HSCs, and are enriched in the CD45 + population (Li et al., 2014;Rybtsov et al., 2016;Zhu et al., 2020). To determine whether the number of LMPs was altered by mutations affecting MyD88-dependent signaling we performed a limiting dilution analysis of cells from the AGM regions of wild type and Myd88 -/embryos by plating them into wells containing OP9 stromal cells to enumerate progenitors with B cell potential, and on OP9 cells expressing the Notch ligand delta-like 1 (OP9-DLL1) to measure the number of T cell progenitors (Li et al., 2014;Schmitt and Zuniga-Pflucker, 2002). Myd88 -/embryos contained significantly fewer progenitors capable of producing B or T cells compared to wild type embryos (Fig. 2E). Therefore, MyD88-dependent TLR signaling increases the generation of LMPs at the expense of long-term multilineage-repopulating HSCs.

TLR4 signaling in endothelial cells and/or IAHCs, but not in primitive innate immune myeloid cells, regulates IAHC cell numbers.
Professional innate immune myeloid cells are the primary producers of inflammatory cytokines and chemokines and are proposed to be an important cellular source of the inflammatory molecules promoting the generation of HSPCs in the DA. Depletion of macrophages and granulocytes in zebrafish embryos using morpholinos targeting spi1a +spi1b, of macrophages alone with morpholinos against irf8, or of neutrophils using morpholinos to cebp1 decreased runx1 expression in the DA and the number of myb:GFP + kdrl:mCherry + HSPCs (Espin-Palazon et al., 2014;He et al., 2015;Li et al., 2014). Depletion of macrophages in explanted AGM regions from mouse embryos by clodronate also decreased HSPCs (Mariani et al., 2019). Vital imaging documented macrophages intimately contacting IAHC cells, and macrophages were occasionally observed to extravasate through endothelial cells to reach the IAHCs (Mariani et al., 2019;Travnickova et al., 2015).
We determined whether TLR4 signaling in yolk sac-derived myeloid cells regulates the number of IAHC cells in the DA by deleting Tlr4 floxed alleles (Tlr4 f/f ) in EMPs using Csf1r-Cre (Gomez Perdiguero et al., 2015;Schulz et al., 2012). Csf1r-Cre should result in TLR4 loss in both yolk sac-derived macrophages and granulocytes, although since granulocytes are a minor population in the mouse embryo, macrophages are the more likely source of inflammatory cytokines (Mariani et al., 2019). We confirmed that cell surface TLR4 was depleted on macrophages following Csf1r-Cre deletion (Fig. 3A, Fig. S3A), and that macrophages had an impaired response to LPS stimulation, as evidenced by lower TNF production (Fig. 3B). Deletion of TLR4 by Csf1r-Cre had no impact on the number of IAHC cells (Fig. 3C, Fig. S3B), indicating that the cytokines produced by primitive myeloid cells in response to TLR4 signaling do not regulate the number of IAHCs. RUNX1 levels were significantly lower in IAHCs, but whether this was due to Csf1r-Cre mediated deletion in EMPs, or in differentiating IAHC cells could not be determined (Fig. 3D). These results do not, however, exclude a role for primitive myeloid cells in regulating IAHC cell numbers, as other signaling pathways that regulate inflammatory cytokine production or other processes mediated by primitive myeloid cells may be responsible for their effect on hematopoietic ontogeny in the DA.
We next assessed whether TLR4 signaling in endothelial or IAHC cells was necessary for generating normal numbers of IAHC cells by deleting Tlr4 f/f alleles with an endothelial cell-specific, tamoxifen-regulated Cre (Cdh5-Cre ERT ) (Sorensen et al., 2009a). We timed the tamoxifen injections in pregnant dams to preferentially delete TLR4 in arterial endothelial cells and IAHCs but not in EMPs. EMPs differentiate from HE cells in the yolk sac vascular plexus between E8.5-E9.5 (Frame et al., 2015), just prior to the major wave of EHT in arterial endothelium that produces IAHC cells (Tober et al., 2013;Yokomizo and Dzierzak, 2010). We previously performed a temporal analysis of Cdh5-Cre ERT activity following tamoxifen injection and demonstrated that deletion in EMPs could be mitigated by injecting tamoxifen at E9.0 or later in development (Tober et al., 2013). Therefore, to delete TLR4 in arterial endothelial cells while sparing EMPs we injected tamoxifen into pregnant dams at E9.5 and analyzed HE and IAHC cells one day later in Tlr4 f/f and Tlr4 f/f ; Cdh5-Cre ERT embryos. We confirmed that TNF production was unchanged in macrophages from Tlr4 f/f ; Cdh5-Cre ERT embryos following stimulation with LPS at E10.5, confirming that deletion of Tlr4 using a carefully timed activation of Cdh5-Cre ERT has minimal impact on EMPs or their progeny (Fig. 3E). The number of IAHC cells at E10.5 was significantly decreased following deletion of Tlr4 f/f alleles at E9.5 by Cdh5-Cre ERT , as were RUNX1 levels in HE and IAHC cells (Fig. 3F,G). We conclude that TLR4 is required in arterial endothelial cells and/or IAHC cells to generate normal numbers of IAHC cells.
To confirm that TLR4 signaling can be directly activated in embryonic endothelial cells we purified CD31 hi VEC + ESAM + CD44 -CD45 -Ter119 -Mac1 -CD41endothelial cells, stimulated them ex vivo with LPS and monitored expression of the integrin CD62E (E-selectin, encoded by Sele) which is a direct NF-kB target gene (Collins et al., 1995). CD62E levels increased in purified endothelial cells in response to LPS compared to unstimulated endothelial cells (Fig. 3H), indicating that embryonic endothelial cells respond to TLR4 signals. We were unable to accurately detect TLR4 by flow cytometry on the surface of endothelial cells. However, since signaling through TLR4 on the cell surface requires MyD88, and the genetic data indicate that TLR signaling through MyD88 but not TRIF adaptors regulates IAHC numbers, we conclude that the relevant signaling is occurring through extracellular TLR4 signaling.
In summary, the number of cells in the IAHCs, and the number of LMPs are positively regulated by MyD88-dependent TLR signaling (Fig. 4). The number of specified HE cells, on the other hand, is not. MyD88-dependent TLR signaling is required in endothelial cells or in the IAHC cells that differentiate from HE cells to regulate IAHC cell numbers. The selective decrease in LMPs in the absence of MyD88-dependent signaling inversely correlates with an increase in adult-repopulating HSCs, suggesting that MyD88-dependent TLR signaling regulates the balance between stem cells and progenitors with different potentials (Fig. 4). Previous single cell RNA sequencing demonstrated that LMPs express higher levels of proliferation genes, such as Myc, compared to pre-HSCs/HSCs (Zhu et al., 2020). Loss of TLR4 signaling and decreased proliferation may differentially affect both committed progenitors and pre-HSCs/HSCs. TLR4 signaling in adult HSCs increases proliferation but reduces fitness (Takizawa et al., 2017), therefore decreased proliferation of pre-HSCs and HSCs in MyD88-deficient embryos may contribute to the selective increase in functional HSCs.
Haploinsufficiency for RUNX1 decreases the number of IAHC cells and results in the precocious emergence of HSCs in the embryo (Cai et al., 2000). However, Runx1 haploinsufficiency also decreases the number of HE cells (Zhu et al., 2020), whereas loss of MyD88-dependent TLR signaling does not. HE cells differentiate from a precursor population called pre-HE, and the efficiency at which pre-HE differentiates into HE is regulated by RUNX1 dosage (Zhu et al., 2020). We believe the reason loss of MyD88-dependent TLR signaling does not reduce the number of HE cells is that RUNX1 levels are normal in mutant embryos at E9.5 when specification occurs. The subsequent decrease in RUNX1 levels in HE and IAHC cells at E10.5 may contribute to the lower number and proliferation of IAHC cells, and reflect a later requirement for RUNX1. RUNX1 promotes the G1 to S phase transition in several contexts (Cai et al., 2011;Nimmo et al., 2005;Osorio et al., 2008), and loss of RUNX1 decreases proliferation, consistent with these observations. Our data are consistent with those from zebrafish embryos which demonstrated lower levels of runx1 expression and decreased numbers of HSPCs upon loss of MyD88-dependent TLR signaling (He et al., 2015), but they add an extra layer of complexity. They also provide a cautionary note for investigators studying HSC ontogeny--that the level of RUNX1 expression, and the numbers of IAHC cells and progenitors do not always correlate with the number of HSCs.

Animal husbandry
The morning post mating is considered embryonic day (E) 0.5. E9.5-E11.5 embryos were staged at the time of harvest by counting somites. Embryos that showed abnormal development were discarded. Mice were handled according to protocols approved by the University of Pennsylvania's Institutional Animal Care and Use Committee and housed in a specific-pathogenfree facility.

Cell sorting and flow cytometry
Antibodies used can be found in Supplemental Table 1. All antibodies were used at a concentration of 1:200. DAPI, LIVE/DEAD Fixable Aqua (ThermoFisher L34957), or Fixable Viability Stain 780 (BD Biosciences 565388) were used to exclude dead cells. We used the Cytofix/Cytoperm Kit (BD Biosciences) for intracellular flow cytometry to detect TNF in cells from embryos that were treated ex vivo with 100ng/mL LPS or vehicle for 4 hours.
Yolk sacs were removed from embryos, and vitelline vessels were retained with the embryonic portion. The head, cardiac and pulmonary regions, liver, digestive tube, tail and limb buds were removed. The remaining portion containing the aorta-gonad-mesonephros (AGM) region, portions of somite, umbilical and vitelline vessels were collected. Tissues were dissected in phosphate buffered saline (PBS)/10% Fetal Bovine Serum (FBS) and Penicillin/Streptomycin (Sigma), followed by dissociation in 0.125% Collagenase (Sigma) for 1 hour. Tissues were washed and filtered through a 40-micron filter and resuspended in antibody solution. Cells were sorted on a BD Influx equipped with a 100-micron nozzle and run at a pressure of 17 psi with flow rates less than 4000 events/second and collected in PBS/20% FBS/25mM HEPES. Flow cytometry was performed on an LSR II (BD) and data were analyzed using FlowJo software (v10.7, BD).

T and B cell progenitor assays
OP9 and OP9-DL1 stromal cells were maintained in αMEM containing 20% FBS and antibiotics and plated into 96-well plates at a concentration of 5500 cells/well 1 day before dilutions of the dissociated embryonic tissues were added. Medium was supplemented with purified recombinant 5ng/ml Flt3L and 10ng/ml IL-7 (OP9) or 5ng/ml Flt3L and 1ng/ml IL-7 (OP9-DL1) (Peprotech). Cells were harvested and analyzed by flow cytometry 12 days after culture. B cells were identified as CD19 + B220 + , T cells as CD25 + Thy1.1 + . The progenitor frequency was calculated using the method of maximum likelihood applied to Poisson distribution with L-Calc software (StemCell Technologies). OP9s were purchased from ATCC, and OP9-DL1s were obtained from Juan Carlos Zún̋iga-Pflücker.

EdU labeling
Yolk sacs were removed and intact embryos were placed in αMEM containing 20%FBS and antibiotics. Embryos were incubated in a 24-well plate at 37°C for 2 hours in EdU (5-ethynyl-2´deoxyuridine, 100uM). Immediately following incubation with EdU, embryos were dissociated using 0.125% collagenase (Sigma) for 30 minutes and stained for flow cytometry. EdU detection was performed using the Click-iT Plus EdU Flow Cytometry Assay kit (Thermo Fisher, C10632).

Statistics
Significance for multiple comparisons was determined using ordinary one-way ANOVA with Tukey's test for multiple comparisons or two-way ANOVA with Sidak's correction for multiple comparisons. Two-tailed student's t-test was used for pairwise comparisons. All statistics were   for Tlr4 +/+ and n=157 for Tlr4 -/with approximately 50-60 cells measured from 3-4 embryos per genotype.
F) Quantification of HE cells/mm at E9.5 in the DA of Tlr4 +/+ and Tlr4 -/embryos. Cells were counted in the region spanning the caudal boundary of the heart to the intersection with the umbilical artery. Mean±SD; Student's t-test, two-tailed; n=3 and 6 for Tlr4 +/+ and Tlr4 -/-, respectively.