Previously, we have demonstrated that a subpopulation of microglia, known as Hoxb8 microglia, is derived from the Hoxb8 lineage during the second wave (E8.5) of yolk sac hematopoiesis, whereas canonical non-Hoxb8 microglia arise from the first wave (E7.5). Hoxb8 microglia have an ontogeny distinct from non-Hoxb8 microglia. Dysfunctional Hoxb8 microglia cause the acquisition of chronic anxiety and an obsessive-compulsive spectrum-like behavior, trichotillomania, in mice. The nature and fate of the progenitors generated during E8.5 yolk sac hematopoiesis have been controversial. Herein, we use the Hoxb8 cell lineage reporter to define the ontogeny of hematopoietic cells arising during the definitive waves of hematopoiesis initiated in the E8.5 yolk sac and aorta-gonad-mesonephros (AGM) region. Our murine cell lineage analysis shows that the Hoxb8 cell lineage reporter robustly marks erythromyeloid progenitors, hematopoietic stem cells and their progeny, particularly monocytes. Hoxb8 progenitors and microglia require Myb function, a hallmark transcription factor for definitive hematopoiesis, for propagation and maturation. During adulthood, all immune lineages and, interestingly, resident macrophages in only hematopoietic/lymphoid tissues are derived from Hoxb8 precursors. These results illustrate that the Hoxb8 lineage exclusively mirrors murine definitive hematopoiesis.
There are at least three successive waves of hematopoiesis in the developing mouse embryo (Ginhoux and Guilliams, 2016; Godin et al., 1995; Hoeffel and Ginhoux, 2015). The activity of these hematopoietic waves temporally overlaps. The first wave initiates in the blood islands of the yolk sac. Starting at embryonic day (E) 7.5, this primitive wave generates erythromyeloid progenitors (EMPs), which includes hematopoietic progenitors for canonical non-Hoxb8 microglia that directly migrate to and colonize the developing brain at E9.5 (Ginhoux et al., 2010). The second wave initiates in the hemogenic endothelium of the yolk sac. This definitive wave generates a second population of EMPs at E8.5, including the Hoxb8 microglia progenitors that proceed through the aorta-gonad-mesonephros (AGM) region and fetal liver before they colonize the developing brain starting at E12.5 (De et al., 2018). Together, non-Hoxb8 microglia and Hoxb8 microglia serve as the immune cells of the brain. It is this latter subpopulation of microglia that, when dysfunctional, results in the acquisition of chronic anxiety and obsessive-compulsive spectrum-like behavior (i.e. either genetic disruption of Hoxb8 or specific ablation of this microglial subpopulation in mice results in chronic anxiety and trichotillomania-like pathology) (Nagarajan et al., 2018; Trankner et al., 2019). A third, definitive wave initiates in the hemogenic endothelium of the para-aortic splanchnopleura (P-Sp), which then becomes the AGM region and is presumed to generate pre-hematopoietic stem cells (pre-HSCs) starting at E8.5 (Cumano and Godin, 2007).
The ontogeny of tissue-resident macrophages, including microglia, is under debate. For example, although it is believed that yolk sac EMPs generate microglia through yolk sac macrophage intermediates, it is still being debated whether these yolk sac EMPs originate from the first wave, second wave or both waves of yolk sac hematopoiesis. Recently, several models have been put forward that propose different origins and ontogenies of tissue-resident macrophages and microglia. The first model, proposed by Hoeffel et al., states that two waves of EMPs exist in the yolk sac during yolk sac hematopoiesis. The first wave of EMPs (E7.5) gives rise to microglia via yolk sac macrophage intermediates, whereas the second wave (E8.5) gives rise to tissue-resident macrophages via fetal monocytes. EMPs born in the E8.5 yolk sac do not require colony-stimulating factor 1 receptor (Csf1r) expression for survival but do require Myb expression (Hoeffel et al., 2015; Kierdorf et al., 2013). Csf1r is essential for regulating the development of myeloid cells (Stanley and Chitu, 2014) and is necessary for the viability of microglia (Elmore et al., 2014). Myb is required to establish hematopoiesis in the fetal liver and the renewal of adult HSCs (Mucenski et al., 1991; Schulz et al., 2012; Sumner et al., 2000). Most E8.5 EMPs vacate the yolk sac and seed the primordial fetal liver by E9.5, where they differentiate into fetal liver monocytes and are proposed, in this model, to be the precursors of most tissue-resident macrophages (Ginhoux and Guilliams, 2016; Hoeffel et al., 2015).
The second model, proposed by Gomez Perdiguero et al., argues that all EMPs are born in the yolk sac at E8.5 and differentiate into yolk sac macrophages that, in turn, populate all tissues of the developing embryo, including the brain, to become tissue-resident macrophages and microglia, respectively (Gomez Perdiguero et al., 2015; Kierdorf et al., 2015; Perdiguero and Geissmann, 2016). In this model, fetal monocytes in the fetal liver, whether derived from yolk sac EMPs or AGM-derived fetal-HSCs (fHSC), minimally contribute to tissue-resident macrophage populations, including microglia. Interestingly, Myb is expressed in E8.5 EMPs and has been suggested to be required to differentiate erythroid, but not myeloid, progenitor cells (Perdiguero and Geissmann, 2016). In the fetal livers of Myb-null embryos, there is an absence of fetal liver monocytes, but fetal macrophages remain present (Schulz et al., 2012), suggesting that fetal macrophages may arise from a Myb-independent lineage originating in the yolk sac (i.e. E8.5 EMPs). These results may not exclude the possibility that fetal liver monocytes can give rise to fetal macrophages (Hoeffel et al., 2015; Sheng et al., 2015).
Another hematopoietic site within the embryo proper has been suggested to generate pre-HSCs at E8.5 in the P-Sp region (third wave of hematopoiesis). The third model, proposed by Sheng et al., states that all tissue-resident macrophages arise from fHSCs. Nascent pre-HSCs that mature into fHSCs begin seeding the primordial fetal liver by E10.5. Like E8.5 yolk sac EMPs, fHSCs also contribute to the generation of fetal liver monocytes in the developing liver (Ginhoux and Guilliams, 2016; Hoeffel et al., 2015; Sheng et al., 2015). By E14.5, fetal liver monocytes exit the fetal liver to populate all tissues of the developing embryo except the brain (Hoeffel et al., 2015).
The models proposed above agree that EMPs generate microglia. One debate centers around whether there are two distinct waves or one continuous wave of EMPs in the yolk sac. The ontogeny of tissue-resident macrophages, whether these cells differentiate from yolk sac macrophages or fetal liver monocytes, also remains unclear. Adding to this complexity, the precise origin of fetal liver monocytes is uncertain because of the contribution from AGM-derived fHSCs (Sheng et al., 2015) and yolk sac-derived EMPs (Hoeffel et al., 2015).
Are there additional Cre drivers that could help clarify the proposed models of the ontogeny of tissue-resident macrophages? The Hoxb8-IRES-Cre driver has been shown to label immune cell populations in various tissues, including blood, bone marrow and brain (Chen et al., 2010; De et al., 2018; Trankner et al., 2019). However, the extent to which this Cre driver labels immune cell populations, including tissue-resident macrophages and their hematopoietic progenitors, has not been described.
In this comprehensive study, we define the ontogeny of the immune cells of the Hoxb8 lineage using the Hoxb8-tdTomato lineage reporter mouse described by De et al. (2018) . During embryonic development, the Hoxb8 cell lineage marker exclusively marks the definitive hematopoietic waves of the yolk sac (E8.5) and AGM (E10.5), both of which are fundamentally distinct from the primitive hematopoietic wave of the yolk sac (E7.5). In support of the Hoeffel model, we show that fetal monocytes are exclusively derived from the Hoxb8 lineage, whereas yolk sac macrophages are not, implying that Hoxb8 fetal monocytes contribute to the generation of tissue-resident macrophages. We show that the Hoxb8 cell lineage marker labels adult tissue-resident macrophages preferentially located in hematopoietic/lymphoid tissues. In contrast to the Myb independence of canonical non-Hoxb8 microglia, we further describe a requirement of Myb for the propagation and maturation of Hoxb8 hematopoietic progenitors and their intermediates, including Hoxb8 microglia, during embryonic development. We propose that the Hoxb8 cell lineage reporter robustly defines the ontogeny of definitive hematopoiesis.
Hoxb8 lineage CD16/32+ EMPs become the dominant EMP population in the fetal liver
Previously, we identified the earliest developmental time point at which Hoxb8 is expressed during embryonic hematopoiesis. We detected Hoxb8-tdTomato+ cells in the CD41+ Kit+ hematopoietic lineage in the yolk sac as early as E8.5 (De et al., 2018), coincident with the emergence of E8.5 definitive EMPs (CD41+ Kit+ CD16/32+) (McGrath et al., 2015). CD16/32, proposed by McGrath et al., distinguishes between primitive E7.5 yolk sac EMPs (CD16/32− CD41+ Kit+) and definitive E8.5 yolk sac EMPs (CD16/32+ CD41+ Kit+). CD16/32+ EMPs also lack Sca-1 expression, which renders them distinct from HSCs. Combining these markers with the Hoxb8-tdTomato lineage reporter (Hoxb8IRES-Cre/+; Rosa26CAG-LSL-tdTomato/+), we were able to follow the consumption and expansion of EMP subpopulations. The amount of tdTomato+ EMPs were similar in the E8.5 yolk sac (CD16/32−, 3.42%; CD16/32+, 7.07%) and E8.5 embryo proper (CD16/32−, 0.42%; CD16/32+, 0.00%); however, the percentage of Hoxb8 CD16/32+ EMPs significantly increased 4.1-fold compared with Hoxb8 CD16/32− EMPs in the E9.5 yolk sac (CD16/32−, 3.12%; CD16/32+, 12.76) (Fig. S1A,B). E8.5 CD16/32+ EMPs, which include a subset that is tdTomato+ (7.07%) (Fig. 1A,C), have been shown to exit the yolk sac to seed and proliferate in the fetal liver by E9.5 (Hoeffel et al., 2015; McGrath et al., 2015). At E11.5, the fetal liver already contains Kit+ cells that are primarily tdTomato+ (De et al., 2018) (Fig. 1D) and the percentage of tdTomato+ CD16/32+ EMPs (35.3%) was significantly higher compared with all time points examined in the yolk sac (E8.5-E11.5, 7.07-24.5%) (Fig. 1B,C,E). Interestingly, there was an inflection point in the ratio of non-Hoxb8 lineage (nB8L) and Hoxb8 lineage (B8L) CD16/32+ EMP populations prior to establishing fetal liver hematopoiesis between E11.5 and E12.5 (Fig. 1C). By E12.5 and beyond, tdTomato+ CD16/32+ EMPs become the dominant EMP population in the fetal liver (E12.5, 72.8%; E14.5, 86.6%) (Fig. 1C), with a significant increase in number compared with B8L CD16/32− and nB8L CD16/32− EMP populations between E12.5 and E14.5 (Fig. S1C,D), which is likely to accommodate the production of fetal monocytes. Collectively, these data suggest that the second wave of yolk sac CD16/32+ EMPs, marked by the Hoxb8 lineage, selectively expands in the fetal liver to become the main EMP population.
Hematopoietic stem cells are derived from the Hoxb8 lineage
As stated in the Introduction, almost all LSK+ (Lineage− Sca-1+ Kithi) HSCs obtained from bone marrow of adult mice are labeled by the Hoxb8 cell lineage marker system (De et al., 2018). As adult HSCs do not express Hoxb8, the lineage marker must have labeled the precursors of HSCs. Previously, we have shown that CD41+ and CD45+ Kit+ hematopoietic progenitors, and CD45+ Sca-1+ Kithi fHSCs can be identified in the fetal liver as early as E10.0 using the Hoxb8-tdTomato lineage reporter (De et al., 2018). However, the Sca-1+ Kithi fHSC compartment is a heterogeneous population containing subsets of short-term (ST) CD48−/CD150− and long-term (LT) CD48−/CD150+ fHSCs, and CD48+ CD150− multipotent progenitors (MPPs) (Pietras et al., 2015; Gudmundsson et al., 2020; Kowalczyk et al., 2015; Upadhaya et al., 2018). In the E12.5, E14.5 and E17.5 fetal liver (FL) and E17.5 spleen (Sp), the majority of ST-fHSCs (E12.5 FL: nB8L, 34.3%; B8L, 64.8%; E14.5 FL: nB8L, 16.4%; B8L, 82.6%; E17.5 FL: nB8L, 9.2%; B8L, 87.8%; E17.5 Sp: nB8L, 3.1%; B8L, 96.4%), LT-fHSCs (E12.5 FL: nB8L, 26.4%; B8L, 72.4%; E14.5 FL: nB8L, 16.4%; B8L, 91.7%; E17.5 FL: nB8L, 9.5%; B8L, 83.5%; E17.5 Sp: nB8L, 1.6%; B8L, 96.1%) and MPPs (E12.5 FL: nB8L, 3.07%; B8L, 96.9%; E14.5 FL: nB8L, 1.42%; B8L, 98.5%; E17.5 FL: nB8L, 0.21%; B8L, 99.7%; E17.5 Sp: nB8L, 1.6%; B8L, 98.5%) were tdTomato+ (Fig. 2A,B,D-F). In a similar fashion, the Hoxb8-tdTomato reporter efficiently labeled nearly all ST-HSCs (nB8L, 0.092%; B8L, 99.8%), LT-HSCs (nB8L, 7.67%; B8L, 92.1%) and MPPs (nB8L, 0.02%; B8L, 99.8%) in adult bone marrow (Fig. 2C-F), which unequivocally shows that virtually all fetal and adult HSCs and MPPs arise from Hoxb8-expressing precursors.
As HSCs are defined functionally by their ability to engraft and reconstitute the immune system of the recipient, we tested the reconstitution properties of Hoxb8 lineage-derived adult LSK+ HSCs in immune-deficient NBSGW (KitW-41J/W-41J; Prkdcscid/scid; Il2rgtm1Wjl/tm1Wjl; Tyr+/+) recipient mice. NBSGW mice allow engraftment of hematopoietic progenitors from non-autologous sources such as human without the need for pre-conditioning (i.e. radiation) (McIntosh et al., 2015). Two to three months after intra-hepatic injections of a low cell concentration of ∼2.0×103 donor-derived tdTomato+ LSK+ HSCs into recipient newborn mice (Fig. S2A), engraftment of total tdTomato+ cells was detected in bone marrow (6.45%), spleen (6.94%), thymus (5.02%) and liver (0.28%) (Fig. S2B,C), indicating that donor cells successfully homed to their appropriate hematopoietic niches. Multi-lineage engraftment was also detected in the granulocyte (CD11bhi Gr-1hi), monocyte (CD11blo Gr-1int), T cell (CD11b− B220−) and B cell (CD11b− B220+) compartments in spleen and bone marrow (Fig. S2D,E), suggesting that tdTomato+ LSK+ HSCs, even at a low concentration, are functionally capable of giving rise to multi-lineage hematopoietic cells.
Hoxb8 hematopoietic progenitors in the AGM region are selectively affected in the absence of Myb function
Previously, we have shown that Hoxb8 CD41+ and CD45+ Kit+ hematopoietic progenitors can be identified in both the yolk sac and in the AGM region. To determine in which hematopoietic tissue Hoxb8 hematopoietic progenitors propagate and mature, we used embryos that lacked Myb function. Primitive hematopoiesis in the yolk sac is unaffected by the disruption of Myb function (Mucenski et al., 1991; Schulz et al., 2012; Sumner et al., 2000). We also did not observe a significant reduction in the number and percentage of total Ter119− Kit+ hematopoietic progenitors in the yolk sac (Fig. 3A-C), including CD41+ Kit+ EMPs and CD45+ Kit+ progenitors (Fig. S3A,B,D,E), of either the Hoxb8 or non-Hoxb8 lineages in E10.5 Myb−/− embryos relative to Myb+/+ embryos. These data agree with previous reports (Mucenski et al., 1991; Schulz et al., 2012; Sumner et al., 2000) that the loss of Myb function should not impact canonical primitive (non-Hoxb8 lineage) hematopoietic progenitors.
We next examined the effect of disruption of Myb function on Hoxb8 hematopoietic progenitors detected in the AGM region during active AGM hematopoiesis. A previous study showed that Myb transcripts are found predominantly in the CD45lo/− Kit+ population in the E10.5 AGM and that, in the absence of Myb function in E10.5 embryos, there appears to be a reduction of the Kit-expressing population (Mizuochi et al., 2012). It has also been suggested that pre-HSCs (CD45lo/− Kit+) are generated and mature in the P-Sp/AGM, migrate to the fetal liver (Ferkowicz et al., 2003; Rybtsov et al., 2011; Sheng et al., 2015; Taoudi et al., 2008), and acquire Myb function for their propagation (Mukouyama et al., 1999). We have previously shown that various Kit-expressing hematopoietic progenitor populations (>60% of CD41+ Kithi, >80% of CD45+ Kithi and >90% of CD45+ Sca-1+ Kithi cells) in the E12.5 fetal liver are labeled by the Hoxb8 cell lineage reporter (De et al., 2018). Therefore, it is reasonable to hypothesize that prior to the establishment of fetal liver hematopoiesis at E12.5, Hoxb8 lineage Kit-expressing cells in the AGM region are dependent on Myb function. Indeed, Fig. 3D,E and Fig. S3C,F show that >50% of the total Ter119− Kit+ tdTomato+ and Ter119− CD45lo/− Kit+ tdTomato+ hematopoietic progenitors (a population that includes EMPs and pre-HSCs) in the AGM region are dependent on Myb function. This dependence results in a ∼2-fold reduction in the number of Hoxb8 hematopoietic progenitors in Myb−/− embryos, whereas the number of non-Hoxb8 hematopoietic progenitors in the AGM is essentially unchanged. Notably, as yolk sac-derived EMPs migrate through the AGM region via the aorta to reach the fetal liver, the Kit+ population might represent migrating EMPs and emerging pre-HSCs. As a result, our data likely reflect the impact of a loss-of-function of Myb on both yolk sac-derived EMPs and AGM-derived pre-HSCs: hematopoietic progenitor populations that arise during definitive hematopoiesis.
Myb function is required for both Hoxb8 and non-Hoxb8 hematopoietic progenitors during fetal liver hematopoiesis
We have previously reported that the Hoxb8-tdTomato lineage reporter labeled ∼5% of CD41+ Kithi hematopoietic progenitors in the yolk sac from E8.5 to E10.0, whereas >60% of CD41+ Kithi and 82% of CD45+ Kithi hematopoietic progenitors in the fetal liver were labeled from E12.5 onwards (De et al., 2018). These data suggest that the Hoxb8-tdTomato lineage reporter efficiently marks hematopoietic progenitor populations, including EMPs and fHSCs, during hematopoiesis in the fetal liver. As several studies have shown a loss of hematopoietic Kit+ progenitors in fetal livers of Myb−/− embryos (Schulz et al., 2012; Sumner et al., 2000), we anticipated that the Hoxb8 hematopoietic progenitor populations expressing Kit would also be strongly affected by the loss of Myb function. Indeed, total Hoxb8 and non-Hoxb8 Ter119− Kit+ progenitors (Fig. 4A-C,F-H) and CD41+ Kithi, CD45lo Kithi, and CD45hi Kithi hematopoietic progenitors (Fig. S4A-C) are virtually absent in the fetal livers of E11.5, E12.5 and E14.5 Myb−/− embryos. In the fetal livers of E12.5 Myb−/− embryos, both the Hoxb8 and non-Hoxb8 CD16/32+ EMP populations are significantly reduced by the loss of Myb function (Fig. 4D,E,I). The complete loss of non-Hoxb8 progenitors could represent the rapid consumption of primitive hematopoietic progenitors, as already observed in previous reports (Schulz et al., 2012). Collectively, these data suggest that EMPs from the second wave of yolk sac hematopoiesis and fHSCs from AGM hematopoiesis are compromised by the loss of function of Myb.
Most fetal monocytes and tissue-resident macrophages derive from the Hoxb8 lineage after the establishment of hematopoiesis in the fetal liver and are maintained throughout adulthood
Hoxb8 lineage cells are detected in postnatal tissues, including peripheral blood, bone marrow and brain (Chen et al., 2010; De et al., 2018; Trankner et al., 2019). To gain a better understanding of the development of the Hoxb8 lineage, we broadened our tracking of Hoxb8-tdTomato presence in the CD45+ pan-hematopoietic population in non-hematopoietic/lymphoid tissues (heart, liver, kidney and lung) and hematopoietic/lymphoid tissues (bone marrow, spleen, thymus and blood). CD45+Hoxb8-tdTomato+ hematopoietic cells were mainly detected in hematopoietic/lymphoid tissues (BM, 75.2%; spleen, 96.9%; thymus, 98.4%) versus non-hematopoietic/lymphoid tissues (heart, 0.17%; liver, 1.35%; kidney, 7.48%; lung, 21.16%) (Fig. S5). This prompted us to survey the development of Hoxb8 lineage committed hematopoietic cells, specifically monocytes and tissue-resident macrophages.
As EMPs and fHSCs have been shown to differentiate into fetal monocytes in the fetal liver (Hoeffel et al., 2015, 2012; Hoeffel and Ginhoux, 2018; McGrath et al., 2015; Sheng et al., 2015), and the Hoxb8-tdTomato cell lineage reporter labels a significant fraction of these populations (De et al., 2018) (Figs 1 and 2), we looked at the development of fetal monocytes (CD45+ F4/80lo CD11bhi CD64+Cx3cr1-GFP+ Ly6Chi/−) in embryonic tissues (fetal liver, fetal lung, fetal heart, thymus, spleen and blood) and adult hematopoietic/lymphoid tissues (bone marrow, spleen, thymus and blood) for the appearance of the Hoxb8 lineage marker. Fetal monocytes can be separated into subsets consisting of classical inflammatory (Ly6Chi) and non-classical patrolling (Ly6C−) cells and have differential expression of Cx3cr1 (Geissmann et al., 2003; Guilliams et al., 2018; Hoeffel et al., 2015). In both the Ly6C− and Ly6Chi subsets, the percentage of fetal monocytes that was tdTomato+ (B8L) substantially increased between E12.5 and E14.5 in the fetal liver (Ly6C−, 84.7 to 98.4%; Ly6Chi, 52.6 to 97.1%), fetal lung (Ly6C−, 80.2 to 95.2%; Ly6Chi, 39.8 to 95.4%), fetal heart (Ly6C−, 30.1 to 64.3%; Ly6Chi, 28.5 to 94.9%), and circulating fetal blood (Ly6C−, 76.9 to 82.5%; Ly6Chi, 10.4 to 87.6%) (Fig. 5A-D). Notably, tdTomato− (nB8L) fetal monocytes are rapidly reduced by E14.5. By E17.5, nearly all fetal monocytes are tdTomato+ (fetal liver: Ly6C−, 99.8%; Ly6Chi, 99.1%; fetal lung: Ly6C−, 93.6%; Ly6Chi, 98.9%; fetal heart: Ly6C−, 85.1%; Ly6Chi, 98.9%), including spleen (Ly6C−, 97.3%; Ly6Chi, 99.5%) and thymus (Ly6C−: 97.5%, Ly6Chi: 99.5%) (Fig. 5E), showing that non-Hoxb8 lineage fetal monocytes are rapidly consumed by Hoxb8 lineage fetal monocytes. In the adult, virtually all Ly6C− and Ly6Chi monocytes, and Ly6Cint granulocytes in each examined tissue (bone marrow, thymus, spleen) were tdTomato+ (Fig. S6), a result that is also detected in adult blood (Fig. S7), indicating that Hoxb8 lineage-derived monocytes and granulocytes are maintained throughout adulthood.
As embryonic and adult spleen, thymus, blood and adult bone marrow are heavily populated with Hoxb8 monocytes (Fig. 5, Figs S6 and S7) and Hoxb8 HSCs (Fig. 2), we wondered whether these hematopoietic/lymphoid tissues would also contain Hoxb8 tissue-resident macrophages. We tracked the development of Hoxb8 tissue-resident macrophages (TRMs) in different tissues during embryonic development and during adulthood, in non-hematopoietic/lymphoid tissues (heart, liver, kidney and lung) and hematopoietic/lymphoid tissues (bone marrow, spleen and thymus). In fetal liver, the percentage of Hoxb8 TRMs reached as high as 25.9% at E14.5 but remained significantly less than non-Hoxb8 TRMs from E12.5 to E18.5 (Fig. 5F,K, Fig. S8). In fetal lung, whereas the percentage of nB8L TRMs decreased, the percentage of Hoxb8 TRMs increased between E12.5 (nB8L, 62.7%; B8L, 27.5%) and E17.5 (nB8L, 16.5%; B8L, 72.7%) (Fig. 5G,K, Fig. S8). Likewise, in circulating fetal blood, an increase in percentage of Hoxb8 TRMs was present between E12.5 (nB8L, 52.0%; B8L, 37.7%) and E14.5 (nB8L, 25.8%; B8L, 62.4%) (Fig. 5I). In fetal heart, low levels of Hoxb8 TRMs were detected throughout development from E12.5 (nB8L, 63.3%,; B8L, 8.12%) to E17.5 (nB8L, 76.8%; B8L, 3.15%) (Fig. 5H,K, Fig. S8). By E17.5, nearly all fetal TRMs were tdTomato+ in thymus (nB8L, 2.72%; B8L, 96.9%) and spleen (nB8L, 2.12%; B8L, 96.1%) (Fig. 5J). Whereas other embryonic tissues had detectable levels of Hoxb8 TRMs, such as skin (E17.5) (Fig. S8), negligible levels of Hoxb8 TRMs were present in the yolk sac (E11.5) (Fig. 5K), midgut (E12.5), stomach (E12.5) and kidney (E14.5, E17.5) (Fig. S8).
In the adult, nearly all resident macrophage populations (CD45+ F4/80+ CD11blo/− CD64+ Ly6C−) were tdTomato+ (thymus: nB8L, 1.18%; B8L, 96.6%; spleen: nB8L, 0.83%; B8L, 94.8%; bone marrow: nB8L, 0.17%; B8L, 99.6%) (Fig. S9A,B), with the majority being tdTomato+ GFP− (thymus, 84.6%; spleen, 83.0%; bone marrow, 91.1%) (Fig. S9A,C). Using the myeloid/macrophage markers Iba-1 and Cx3cr1-GFP, most tdTomato+ cells were GFP− Iba1− in the heart and liver, whereas lungs contained tdTomato+ cells that were largely Iba1+ GFP+/− (Fig. S9D). Adult lungs were particularly interesting as ∼50% of alveolar macrophages (CD45+ F4/80+ Ly6C− CD11b−Cx3cr1-GFP−) (Liegeois et al., 2018) and nearly all interstitial macrophages (CD45+ F4/80+ Ly6C− CD11b+Cx3cr1-GFP+) (Liegeois et al., 2018) were tdTomato+ (Fig. S9E). The higher percentage in the lung can be explained by myeloid populations derived from the fetal liver (alveolar macrophages) and bone marrow (interstitial macrophages) (Tan and Krasnow, 2016), tissues where Hoxb8-tdTomato signal is widespread (De et al., 2018).
As hematopoietic/lymphoid tissues also contain other immune cells, we examined what percentage of these cells the Hoxb8-tdTomato lineage reporter would label. Over 93% of eosinophils, neutrophils, T cells and B cells were tdTomato+ (Fig. S10). Overall, these data suggest that the great majority of innate and adaptive immune cell populations in hematopoietic/lymphoid tissues are predominantly derived from Hoxb8-expressing precursors.
The Hoxb8-tdTomato lineage reporter efficiently and selectively labels fetal macrophages derived from the second wave of yolk sac hematopoiesis
To show that the Hoxb8-tdTomato lineage reporter does not efficiently label primitive fetal macrophages arising during the first wave of hematopoiesis in the yolk sac (E7.5), we detected Hoxb8 macrophages (tdTomato+ GFP−, tdTomato+ GFP+) as early as E9.5 in the yolk sac, but at significantly lower levels than non-Hoxb8 macrophages (tdTomato− GFP+) (tdTomato− GFP+, 41.63%; tdTomato+ GFP+, 2.93%; tdTomato+ GFP−, 3.96%) (Fig. 6A). Similarly, we detected negligible levels in the embryo proper (tdTomato+ GFP+, 2.64%; tdTomato+ GFP−, 0.99%) and no tdTomato+ cells in the E9.5 head (Fig. 6A). Notably, the low level of tdTomato+ cells in the yolk sac could represent circulating macrophages at this time. In the E10.5 embryo, the percentage of Hoxb8 macrophages increases in the yolk sac (tdTomato+ GFP+, 17.1%; tdTomato+ GFP−, 0.98%) and head (tdTomato+ GFP+, 10.5%; tdTomato+ GFP−, 0.22%) but remains significantly lower than non-Hoxb8 macrophages (Fig. 6B). Similarly, Hoxb8 macrophages remain significantly lower compared with non-Hoxb8 macrophages in yolk sac (tdTomato− GFP+, 61.0%; tdTomato+ GFP+, 8.49%; tdTomato+ GFP−, 1.47%), fetal liver (tdTomato− GFP+, 27.86%; tdTomato+ GFP+, 8.82%; tdTomato+ GFP−, 4.59%) and head (tdTomato− GFP+, 81.57%; tdTomato+ GFP+, 11.63%; tdTomato+ GFP−, 0.00%) in E12.5 embryos (Fig. 6C). These results show that a negligible percentage of primitive yolk sac-derived macrophages (first wave of hematopoiesis) arise from the Hoxb8 lineage.
The presence of Hoxb8 microglia in the brain is selectively sensitive to loss of Myb function compared with non-Hoxb8 microglia
As the Hoxb8-tdTomato lineage reporter does not efficiently label primitive macrophages, including canonical non-Hoxb8 microglia that are independent of Myb function during development, we asked whether non-canonical Hoxb8 microglia could be dependent on Myb function. Hoxb8 microglial progenitors, unlike canonical non-Hoxb8 microglial progenitors, transit through the fetal liver before their migration into the developing brain at E12.5 (De et al., 2018). Because of this alternative migration route, the number of Hoxb8 microglia present in the developing brain would be predicted to be selectively dependent on Myb function. The density of Hoxb8 microglia was significantly reduced in the brains of E14.5 embryos (prior to embryonic lethality) by the absence of Myb function (Fig. 7A,B). In contrast, the density of non-Hoxb8 microglia was not significantly affected in the brains of Myb−/− embryos relative to Myb+/+ embryos (Fig. 7C). These results further support the unique ontogeny of Hoxb8 microglia relative to canonical non-Hoxb8 microglia.
Border-associated macrophages are mainly derived from the Hoxb8 lineage
Recent reports have identified border-associated macrophages (BAMs), which include subdural meninges macrophages (mMΦ), perivascular spaces macrophages (pvMΦ) and choroid plexus macrophages (cpMΦ), in the brain that are distinct from canonical microglia (Goldmann et al., 2016; Mrdjen et al., 2018). BAMs are first detected in the subdural meninges at E9.5 and therefore, have a prenatal origin, either arising from yolk sac or fetal liver (Goldmann et al., 2016). Using the markers CD206, Cx3cr1-GFP and Hoxb8-tdTomato, we were able to identify B8L (GFP+ tdTomato+ CD206+) and nB8L (GFP+ tdTomato− CD206+) mMΦ BAMs and B8L (GFP+ tdTomato+ CD206−) and nB8L (GFP+ tdTomato− CD206−) parenchymal microglia (Fig. 7D). While the percentages of B8L microglia and nB8L microglia are in agreement with prior studies (Chen et al., 2010; De et al., 2018; Nagarajan et al., 2018), the percentage of B8L BAMs (F4/80+ CD11b+ CD206+ GFP+ tdTomato+) (68.72%) was higher compared with nB8L BAMs (F4/80+ CD11b+ CD206+ GFP+ tdTomato−) (20.44%) (Fig. 7E,F), suggesting that most BAMs are of the Hoxb8 lineage.
The Hoxb8 cell lineage reporter Hoxb8IRES-Cre; Rosa26CAG-LSL-tdTomato robustly labels hematopoietic progenitor and fHSC populations initiated during the definitive waves of yolk sac hematopoiesis at E8.5 and AGM hematopoiesis, respectively. Approximately 9% of the existing CD41+ Kit+ hematopoietic progenitor pool in the yolk sac is labeled with the Hoxb8-tdTomato reporter. We next observed Hoxb8 CD41+ Kit+ hematopoietic progenitors in the AGM (E10.0-11.5), where they are expanded ∼8-fold relative to their presence in the yolk sac and represent ∼80% of the total progenitor pool (De et al., 2018) (and herein). By E12.5, Hoxb8 CD41+ and CD45+ Kit+ hematopoietic progenitors are present in the fetal liver where they are further amplified (∼280-fold, again relative to their presence in the yolk sac) and comprise greater than 90% of the hematopoietic progenitor pool in the fetal liver.
Multiple aspects of the current ontological models for tissue-resident macrophages remain unresolved. For example, are tissue-resident macrophages generated from fetal monocytes (Hoeffel model) or yolk sac macrophages (Gomez-Perdiguero model)? In the current study, the yolk sac macrophages that were Hoxb8-tdTomato+ in the E9.5 yolk sac (<4%) were depleted by E10.5, whereas 98% of Ly6C− and 97% of Ly6Chi fetal monocytes were Hoxb8-tdTomato+ by E14.5 in the fetal liver. By comparison, our data using the Hoxb8 lineage reporter supports the Hoeffel model. However, we were unable to completely resolve whether the contribution of fetal monocytes arises from AGM-derived fHSCs (Sheng Model) or yolk sac-derived EMPs (Hoeffel Model). We found that the great majority of LSK+ ST-fHSCs (83%), LT-fHSCs (91%), MPPs (98%) and E8.5 yolk sac-derived CD16/32+ EMPs (87%) are derived from the Hoxb8 lineage, suggesting that Ly6C− and Ly6Chi fetal monocytes are derived from both hematopoietic sources. According to the Hoeffel model, E8.5 yolk sac-derived EMPs acquire Myb expression by E9.5, when they seed the fetal liver (Hoeffel et al., 2015; Kierdorf et al., 2013). We show that Hoxb8 and non-Hoxb8 CD16/32+ EMPs in the fetal liver are very sensitive to the presence of Myb, suggesting that Myb function is required for the great majority of the hematopoietic progenitor populations upon seeding the fetal liver. Therefore, currently, our data support both the Hoeffel and Sheng models.
We recognized that Hoxb8 lineage CD41+ and CD45+ hematopoietic progenitors, including CD45+ Sca-1+ Kit+ fHSCs, can be identified in both the yolk sac, as early as E8.5, and in the AGM region, and express detectable levels of Hoxb8 transcripts (De et al., 2018). Therefore, we hypothesized that a disruption of Myb function would affect the Hoxb8 hematopoietic progenitor populations arising from definitive hematopoiesis. Whereas Hoxb8 and non-Hoxb8 hematopoietic progenitor cells are unaffected in the yolk sac, Myb function appears to specifically affect the majority of Hoxb8 hematopoietic progenitors in the AGM region. These results provide evidence that hematopoiesis in both Hoxb8 and non-Hoxb8 lineages in the yolk sac is not affected by the absence of Myb, whereas a subset of hematopoietic progenitors derived from the Hoxb8 lineage are selectively sensitive to Myb function in the AGM region during AGM hematopoiesis. In contrast, both non-Hoxb8 and Hoxb8 hematopoietic progenitors require Myb function in the fetal liver. From these experiments, we propose that the Hoxb8-IRES-Cre lineage mouse model firmly marks definitive hematopoiesis and efficiently labels all definitive-derived hematopoietic lineages starting in the E8.5 yolk sac.
Hoxb8 and non-Hoxb8 microglia have different migratory routes into the developing brain. Canonical non-Hoxb8 microglia directly migrate from the yolk sac into the embryonic brain, starting at E9.5 as the vascular system forms (Ginhoux et al., 2010). Hoxb8 microglia have an alternative ontogeny and passage from the yolk sac to the fetal liver before entering the developing brain starting at E12.5 (De et al., 2018). As primitive hematopoiesis is independent of Myb function and the production of yolk sac macrophages is still active in Myb-null embryos, we would anticipate that their number in the brain would not be affected by disruption of Myb function. On the other hand, as Hoxb8 microglial progenitors passage through the fetal liver, their number should be significantly affected by the loss of Myb function. We observed a significant effect on the number of Hoxb8 microglia, further supporting the unique ontogeny of Hoxb8 microglia and its contribution to their unique roles in regulating homeostasis in the brain.
At E11.5, 1 day before the expansion of fetal monocytes in the fetal liver, the percentages of yolk sac-derived EMPs and AGM-derived fHSCs in the fetal liver labeled with the Hoxb8-tdTomato reporter are ∼41% and ∼51%, respectively, and significantly increase by E12.5 (Hoxb8 EMPs, ∼82%; Hoxb8 fHSCs, ∼88%). At E12.5 in the fetal liver, ∼85% of Ly6C− and ∼53% of Ly6Chi fetal monocytes are derived from the Hoxb8 lineage; by E14.5, it is ∼98% and 97%, respectively. Our data, combined with the current models, suggest that yolk sac-derived Hoxb8 EMPs and AGM-derived Hoxb8 fHSCs generate Hoxb8 fetal monocytes in the fetal liver, which, in turn, become Hoxb8 tissue-resident macrophages.
Shared with our previous observations regarding adult bone marrow HSCs (Chen et al., 2010; De et al., 2018), it appears that the Hoxb8 lineage predominantly labels hematopoietic cells and tissue-resident macrophages in tissues of active adult hematopoiesis/lymphopoiesis. These findings are congruent with our previous results that, during embryonic development, hematopoietic progenitors, AGM-derived fHSCs, yolk sac-derived EMPs and fetal monocytes are derived mainly from the Hoxb8 lineage (De et al., 2018). Nearly all granulocytes, monocytes and lymphocytes (i.e. T cells and B cells) are of the Hoxb8 lineage in adult blood. This makes sense because most adult innate immune blood cells are generated from LSK+ HSCs in the bone marrow and are labeled with the Hoxb8-tdTomato reporter.
Functional heterogeneity is apparent in the hematopoietic lineages. Immune cell populations contain sub-lineages with an extensive diversity of expression profiles and functions (Satpathy et al., 2012; Yona and Jung, 2010). Tissue-resident macrophages represent a diverse population of cells among tissues and within a single tissue. Their developmental origins contribute to their functional diversity and distribution in these tissues (den Haan and Kraal, 2012; Epelman et al., 2014; Hoeffel et al., 2012; Tan and Krasnow, 2016). Debate on the precise ontogeny of this diverse population of cells is ongoing (Gomez Perdiguero et al., 2015; Hoeffel et al., 2015; Sheng et al., 2015). It will be essential to determine the extent to which the developmental origin(s) and ontogeny of immune cells contribute to their diverse functions.
MATERIALS AND METHODS
All experiments in this study have been performed on mice. Experimental procedures were approved by the Institutional Animal Care and Use Committee at the University of Utah. Gt(ROSA)26Sortm14(CAT-tdTomato)Hze (Ai14, 007908) and NOD.Cg-KitW-41J; Tyr+; Prkdcscid; Il2rgtm1Wjl/ThomJ (NBSGW, #026622) mice were obtained from the Jackson Laboratory. Cx3cr1GFP mice (005582, Jackson Laboratory) were a kind gift from Dr Monica Vetter (Department of Neurobiology and Anatomy, University of Utah, Salt Lake City, UT, USA). Myb+/− mice (Mucenski et al., 1991) were a kind gift from Dr James Palis (Department of Pediatrics, University of Rochester Medical Center, Rochester, NY, USA). Hoxb8IRES-Cre mice are described by Chen et al. (2010).
Flow cytometry analysis of embryonic and postnatal tissues
Embryo isolation and dissection were performed as previously described (De et al., 2018). Briefly, dissected embryos and adult tissue (2-3-months-old) were placed on ice in 10% fetal bovine serum (FBS, Atlanta Biologicals) in 1× Hanks' balanced salt solution (HBSS, Gibco). Embryonic and adult blood was collected via cardiac puncture and decapitation (embryo only). Blood samples were collected in 0.5 M EDTA-AC/10% FBS to prevent coagulation. Embryonic tissue was gently mechanically dissociated into single cells. Adult tissue was minced into small pieces. Embryonic and adult tissue was then incubated in 10% FBS/0.05% collagenase/0.025 U/ml DNaseI for 30 min at 37°C followed by a gentle mechanical dissociation into single cells. Cells were passed through a 80 µm mesh to obtain a single-cell suspension.
Anti-mouse antibodies used were as follows: TER-119 PerCP-Cy5.5 (1:50, BioLegend, 116228), B220 APC-Cy7 (1:100, BioLegend, 103224), CD11b BV510 (1:100, BioLegend, 101263), CD11b APC (1:100, BioLegend, 101212), CD115 PE-Cy7 (1:100, BioLegend, 135524), CD150 BV510 (1:100, Biolegend, 115929), CD206 Pe-Cy7 (1:100, Biolegend, 141720), CD48 PerCP-Cy5.5 (1:50, BioLegend, 103422), F4/80 PE-Cy7 (1:100, BioLegend, 123114), F4/80 PerCP-Cy5.5 (1:50, BioLegend, 123128), Kit PE-Cy7 (1:100, BioLegend, 105814), CD45 APC (1:100, BioLegend, 103112), CD16/32 APC (1:100, BioLegend, 101326), Ly6C APC-Cy7 (1:100, BioLegend, 126026), CD41 APC-eFluor 780 (1:100, eBioscience, 47-0411-80) and CD41 APC-Cy7 (1:100, BioLegend, 133928) in 10% FBS/1% BSA/1×HBSS. Cells were incubated with their respective antibody cocktail for 30 min on ice. Cells were counterstained with 3 µM DAPI in 10% FBS/1×HBSS. Flow cytometry data were obtained using the BD Bioscience FACSCanto II flow cytometry sorter or BD Bioscience FACS ARIA flow cytometry sorter. All FACS data were analyzed with FlowJo 10.0.7 (Celeza).
Cryosectioning and immunohistochemistry
Embryonic and postnatal brain tissue were processed and sectioned as previously described (De et al., 2018). Briefly, for immunohistochemistry, brain samples were permeabilized with 0.2% Triton X-100, 1% sodium deoxycholate solution, then incubated overnight with primary antibody mixture at 4°C. The sections were incubated with secondary antibodies for 2 h at room temperature the following day. Sections were counterstained with DAPI (D1306, Molecular Probes) and mounted with ProLong Gold antifade reagent (P36934, Molecular Probes) and microscope cover glass (Fisherbrand, 22-266882). Images were acquired on the Leica TCS SP5 confocal microscope, and processed and analyzed using Imaris 7.7 (Bitplane), as described below.
Primary antibodies used were as follows: chicken anti-GFP (1:250, Aves Labs, GFP-1020), guinea pig anti-tdTomato-GP-Af430 (1:250, Frontier Institute, AB_2631185), rat anti-mouse Kit (1:250, BioLegend, 105801), rat anti-mouse CD41 (1:250, BioLegend, 133901) and rat anti-mouse CD206 Alexa Fluor 647 (1:200, Biolegend, 141712). Secondary antibodies used were: goat anti-chicken Alexa Fluor 488 (1:500, Thermo Fisher Scientific, A-11039,), goat anti-guinea pig Alexa Fluor 555 (1:500, Thermo Fisher Scientific, A-21428) and goat anti-rat Alexa Fluor 647 (1:500, Thermo Fisher Scientific, A-21247). Primary and secondary antibodies used have been tested for specificity and cross-reactivity.
Confocal imaging parameters and Imaris image analysis
Images were acquired on a Leica TCS SP5 confocal system. For image acquisition, the tissue sections were imaged with a 10× (0.4 NA, Leica) or 20× objective (0.4 NA, Leica), and 1.5 or 2.0× digital zoom, and images were acquired at 1024×1024 resolution and 400-600 Hz scan speed, using a 5.0 µm z-depth through the tissue. For microglial counting, the brain sections were imaged with a 10× objective and 1.5× digital zoom, and images were acquired at 512×512 resolution and 400-600 Hz scan speed, using a 5.0 µm z-depth through the tissue. Images were processed using Imaris Image Analysis Software x64 (v 7.7.2, Bitplane). The ‘Spots’ function was used to count the number of microglia per unit area (mm2). To further identify the subsets of cells that co-label with Cx3cr1-GFP, the spots were filtered using ‘mean intensities’ of the fluorescence of the marker. The ‘Surface’ function was used to quantify the area of the analysis region.
Neonatal intra-hepatic transplantations
LSK+ HSCs were isolated from postnatal bone marrow (P15) of Hoxb8IRES-Cre/+; Rosa26CAG-LSL-tdTomato/+ mice and sorted by flow cytometry. Sorted Hoxb8 LSK+ HSCs (∼2.0×103 DAPI− Lin− Sca-1+ Kithi tdTomato+ cells) were transplanted into the livers of P1-P4 NBSGW mice.
Data from all experiments were analyzed with GraphPad Prism 8 software (San Diego, CA). Unpaired t-test was used for direct comparison between two data groups. Standard two-way ANOVA followed by post-hoc analysis using Tukey's or Šidák's multiple comparisons test compared multiple data groups. Data are graphically reported as mean±s.e.m. (pooled and individual biological replicates). P<0.05 was considered significant.
We wish acknowledge the extensive help we have received from our animal husbandry support team (Karl Lustig, Kerry Prettyman and June Wangerin), from members of the Capecchi laboratory, especially Anne Boulet PhD and Naveen Nagarajan PhD for their critique of the manuscript, and from Kay Higgins for the preparation of the manuscript. We also thank Florent Ginhoux PhD (Singapore Immunology Network, A*STAR) for his careful reading and editing of our manuscript.
Conceptualization: M.R.C., D.A.V.D.; Methodology: D.A.V.D., S.D., S.Z., K.M.E., B.X.; Validation: D.A.V.D., S.Z., B.X.; Formal analysis: D.A.V.D., S.D., B.X.; Investigation: D.A.V.D., S.D.; Resources: D.A.V.D.; Data curation: D.A.V.D.; Writing - original draft: D.A.V.D.; Writing - review & editing: M.R.C., D.A.V.D.; Visualization: D.A.V.D.; Supervision: M.R.C.; Funding acquisition: M.R.C.
This work was supported by the National Institutes of Health (R01 MH093595) (M.R.C.) and the University of Utah Flow Cytometry Facility. Open access funding provided by the University of Utah. Deposited in PMC for immediate release.
The authors declare no competing or financial interests.