Tissue-resident macrophages originate from yolk sac-derived erythro-myeloid progenitors

Introductory paragraph

Most haematopoietic cells renew from adult hematopoietic stem cells (HSC)^1–3, however, macrophages in adult tissues can self-maintain independently of HSC^4–7. Progenitors with macrophage potential in vitro have been described in the yolk sac before emergence of HSC^8–13, and fetal macrophages^13–15 can develop independently of Myb⁴, a transcription factor required for HSC¹⁶, and can persist in adult tissues^4,17,18. Nevertheless, the origin of adult macrophages and the qualitative and quantitative contributions of HSC and putative non-HSC progenitors are still unclear¹⁹. Here we show that the vast majority of adult tissue-resident macrophages in liver (Kupffer cells), brain (microglia), epidermis (Langerhans cells), and lung (alveolar macrophages) originate from a Tie2⁺ cellular pathway generating Csf1r⁺ erythro-myeloid progenitors (EMP) distinct from HSC. EMP develop in the yolk sac at embryonic day (E) 8.5, migrate and colonise the nascent fetal liver before E10.5, and give rise to fetal erythrocytes, macrophages, granulocytes and monocytes until at least E16.5. Subsequently, HSC-derived cells replace erythrocytes, granulocytes and monocytes, whereas Kupffer cells, microglia, and Langerhans cells are not replaced in 1-year old animals, while alveolar macrophages may be progressively replaced in aging mice. Our fate mapping experiments identify, in the fetal liver, a sequence of yolk sac EMP-derived, followed by HSC-derived haematopoiesis, and identify yolk sac EMP as a common origin for tissue macrophages.

Csf1r-expressing cells in the mouse embryo give rise to tissue-resident macrophages in adult tissues⁴. To identify in the developing embryo the site of origin of Csf1r-expressing cells, we performed time course analyses by constitutive (Csf1r^iCre) and inducible (Csf1r^MeriCreMer) fate mapping of cells in yolk sac (YS), head, limbs, caudal region and fetal liver (FL) (Fig. 1a). Progenitors, defined as Kit⁺ CD45^{low 12} (gate R1 in Fig. 1b), were first detected in Csf1r^iCre Rosa26^YFP embryos in the YS from 16-18 somite pairs (sp) stage onwards (E8.5, Fig. 1b, and Extended Data Fig. 1a-c). Csf1r^iCre YFP⁺ Kit^- CD45⁺ cells (gate R2 in Fig. 1b), characterised in Fig. 2 as myeloid cells, were detected in the YS at 20-25 sp (E9, Fig. 1b), and subsequently in the caudal and head regions of the embryo from E9.5, and the fetal liver from E10.5 onwards (Extended Data Fig. 1a-d). To discriminate migration of YFP⁺ cells from de novo labelling, we induced YFP expression in Csf1r^MeriCreMer Rosa26^YFP embryos at E6.5 or E8.5. In embryos pulsed at E6.5, YFP⁺ cells were not detected (Extended Data Fig. 2a, b). When pulsed at E8.5, Csf1r^MeriCreMer YFP⁺ Kit⁺ CD45^low progenitors were detected between E9.5-11.5 in the YS, and in the fetal liver from E10.5 (Fig. 1 c, d). In the fetal liver, numbers of YFP⁺ Kit⁺ CD45^low progenitors increased 3-fold from E10.5 to E11.5, at which time they were 25-fold more numerous in the fetal liver than in the YS (Fig. 1d). At E8.5 all YS Csf1r^iCre YFP⁺ Kit⁺ CD45^low progenitors expressed AA4.1, an antigen expressed on early hematopoietic progenitors¹² (Extended Data Fig. 1e). Csf1r^MeriCreMer YFP⁺ AA4.1⁺ Kit⁺ CD45^low cells were also present in the YS from E9.5 to E10.5, and in the fetal liver from E10.5 (Fig. 1e). These progenitors were undetectable at E10.5 in the aorta-gonad-mesonephros (AGM) region (Fig. 1e), indicating they do not originate within the embryo proper.

Figure 1. E8.5 Csf1r⁺ progenitors originate in the yolk sac and expand in the fetal liver.

a, Fate-mapping analysis of Csf1r-expressing cells. Arrows indicate time points for analysis, and green shades the genetic labelling period. b, YFP expression on live cells from Csf1r^iCre Rosa26^YFP yolk sac (YS), separated by somite pairs between E8.25 and E9 (0sp, n=3; 3/6sp, n=3; 7-9sp, n=5; 11-13sp, n=3; 16-18sp, n=4; 20-25sp, n=4) (upper panels), and Kit and CD45 expression on YFP⁺ cells (lower panels). R1 indicates Kit⁺ CD45^low, and R2 indicates Kit^- CD45⁺ cells. c, Schematic representation of sites analysed in mouse embryos: YS, AGM region, fetal liver and head. Kit and CD45 phenotype of YFP⁺ cells from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (E9.5, n=3; E10.25, n=3; E10.5, n=4; E11.25, n=4; E12.5, n=9). d, Number of YFP⁺ Kit⁺ CD45^low cells (R1 in panel b) per organ or region (mean ± s.e.m.) in Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5. e, Number of YFP⁺ AA4.1⁺ Kit⁺ CD45^low cells per embryonic region and time points (mean ± s.e.m.) in Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5. See also Supplementary Table 1 and Extended Data Figs. 1 and 2.

Figure 2. E8.5 Csf1r⁺ progenitors differentiate into myeloid cells and red blood cells in the fetal liver.

a, Distribution of mixed (Mix), G/M, and E/Mk CFU-C from unsorted, AA4.1⁺ Kit⁺ CD45^low and YFP⁺ AA4.1⁺ Kit⁺ CD45^low cells from E9 YS from Csf1r^iCre Rosa26^YFP and E12.5 fetal liver from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (three independent experiments each). CFU-erythroid and/or megakaryocyte (E/Mk); CFU-granulocyte and/or monocyte/macrophage (G/M); CFU-mix, at least three of the following: G, E, M and Mk. See also Extended Data Fig. 3. b, F4/80 and CD11b expression on YFP⁺ CD45⁺ from YS, head (brain for E11.5), limbs and liver of Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 and analysed on E9.5 (n=3), E10.5 (n=4), E11.5 (n=4), and E12.5 (n=9); Dashed lines represent FMO (fluorescence minus one) control. See also Extended Data Fig. 4. c, F4/80, CD11b, Gr1, Ly-6G and Ly-6C expression in fetal liver CD45⁺ YFP⁺ cells from Csf1r^MeriCreMer Rosa26^YFP embryos pulse-labelled at E8.5, and analysed on E12.5 (n=9) and E16.5 (n=14). May-Grünwald-Giemsa stained cytospin preparations of fetal liver YFP⁺ F4/80^bright and YFP⁺ CD11b^high cells sorted from E16.5 Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5. See also Extended Data Fig. 6a for sorted cells from E14.5 Csf1r^MeriCreMer embryos. Scale bar, 10μm. d, YFP labelling efficiency (%) among red blood cells in fetal liver from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5, mean ± s.e.m. (E11.5, n=4; E12.5, n=9; E14.5, n=5; E16.5, n=11; E18.5, n=4).

Together, these fate-mapping experiments demonstrate that YS-derived progenitors colonise the liver primordium, as proposed earlier^8,20,21, and their expression of AA4.1 suggests that they represent erythro-myeloid progenitors (EMP)¹². In in vitro colony forming assays, the AA4.1⁺ population contained most of the total E9 YS colony forming units-culture (CFU-C, 266±137 vs 296±75, mean ± standard deviation). Frequencies and distributions of different CFU-C, i.e. erythroid (E) / megakaryocyte (Mk) (E/Mk), granulocyte / macrophage (G/M), and G, M, E, and/or Mk (Mix) potential, were comparable between overall AA4.1⁺ and Csf1r^iCre YFP⁺ AA4.1⁺ progenitors (Fig. 2a, Extended Data Fig. 3). Moreover, in the E12.5 fetal liver, the CFU potential of overall AA4.1⁺ and Csf1r^MeriCreMer YFP⁺ AA4.1⁺ cells was comparable to the yolk sac progenitors (Fig. 2a).

These results indicated that yolk-sac-derived, E8.5-labelled YFP⁺ AA4.1⁺ Kit⁺ CD45^lo progenitors have erythroid and myeloid potential in yolk sac and fetal liver. Next, we investigated by fate-mapping their contribution to fetal liver haematopoiesis in vivo. Csf1r^iCre YFP⁺ and Csf1r^MeriCreMer YFP⁺ F4/80^bright fetal macrophages were first detected among Kit^- CD45⁺ (R2 in Fig 1b) at E10.5 in the yolk sac, liver, head and forelimbs (Fig. 2b, Extended Data Fig. 4a-c). In addition, the fetal liver from E12.5 to E16.5 contained Csf1r^MeriCreMer YFP⁺ monocytes and granulocytes (Fig. 2c). The fetal liver also contained Csf1r^MeriCreMer YFP⁺ red blood cells from E11.5 until at least E14.5 (Fig. 2d, Extended Data Fig. 4d). Red blood cells were not labelled before E11.5, indicating that, in contrast to yolk-sac-derived erythrocytes in the fetal liver, primitive erythrocytes in the yolk sac did not arise from Csf1r-expressing cells. Collectively, yolk-sac -derived Csf1r⁺ progenitors contribute to fetal liver haematopoiesis by giving rise to F4/80^bright macrophages, monocytes, granulocytes and red blood cells.

We next investigated the transition from yolk-sac-derived to HSC-derived haematopoiesis. To trace the latter, we used Flt3^Cre which labels fetal and adult HSC-derived multipotent hematopoietic progenitors ²², and their progeny (Extended Data Fig. 5). We compared progeny of yolk-sac-derived progenitors in Csf1r^MeriCreMer mice to progeny of HSC in Flt3^Cre mice. In the fetal liver from E14.5 to E18.5, the progenies of Csf1r⁺ and Flt3⁺ precursors were distinct but complemented each other (Fig. 3a, b). At E14.5, YS-derived CD45⁺ populations included Kit⁺ progenitors, F4/80^bright macrophages, and CD11b^high Gr1⁺ monocytes/granulocytes (Fig. 3b, Extended Data Fig. 6a). Of note, monocytes/granulocytes were present in Myb-deficient fetal liver (Fig. 3a). Csf1r^MeriCreMer YFP⁺ macrophages remained detectable throughout fetal development, and were not replaced by Flt3^Cre YFP⁺ cells. However, yolk-sac-derived Kit⁺ cells and myeloid cells were no longer detectable by E16.5 and E18.5, respectively (Fig. 3a). In contrast, Flt3^Cre YFP⁺ Kit⁺ cells, and CD11b^high Gr1⁺ granulocytes/ monocytes increased in numbers between E14.5 and E18.5 (Fig. 3a-c). The progenies of Csf1r⁺ progenitors and Flt3⁺ progenitors also complemented each other during development in the lung and skin (Extended Data Fig. 6 b, c). Quantitative analyses in fetal and adult tissues indicated that Flt3^Cre YFP labelling of Kit⁺ progenitors preceded that of monocytes/granulocytes, with 80% of progenitors labelled at E18.5, and 80% of monocytes at postnatal day 8 (P8) (Fig. 3b). In contrast, Flt3^Cre YFP labelling plateaued at 14% for adult liver F4/80^bright Kupffer cells, at 2% for CD45^low brain microglia, and 30% for epidermal Langerhans cells up to 1 year of life (Fig. 3b). In contrast, Flt3^Cre YFP labelling of CD45⁺ F4/80⁺ brain macrophages, and lung alveolar macrophages was 16 % in 12-week-old adults but increased progressively over time to reach 40% in 1 year-old mice (Fig. 3b).

Figure 3. Fetal liver HSC-derived Flt3⁺ progenitors give rise to monocytes and granulocytes in late embryos and adults but do not replace YS-derived macrophages.

a, F4/80, Kit, CD11b and Gr1 expression on total CD45⁺ cells (black) and YFP⁺ CD45⁺ cells from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5 (green) in the FL at the indicated days of embryonic development (E14.5, n=5; E16.5, n= 10; E18.5, n=9). F4/80, Kit, CD11b and Gr1 expression on YFP⁺ CD45⁺ cells from Flt3^Cre Rosa26^YFP embryos (orange) (E14.5, n=7; E16.5, n=6; E18.5, n=6). F4/80 and CD11b expression on CD45⁺ cells in Myb^-/- embryos (E14.5, n=4; E16.5, n= 7). b, YFP labelling efficiency in Kit⁺ lin^- cells, CD11b^high F4/80^low cells (characterised in Extended Data Fig. 5) and F4/80^bright macrophages (Kupffer cells in adults) in fetal and adult Flt3^Cre Rosa26^YFP liver (first panel on the left). YFP labelling efficiency in blood monocytes, brain microglia (CD45^low F4/80⁺) and CD45⁺ F4/80⁺ brain macrophages in Flt3^Cre Rosa26^YFP pups and mice (second panel). YFP labelling efficiency in alveolar macrophages (F4/80^bright Siglec-F⁺ CD11b^-) and F4/80^low CD11b^high myeloid cells in Flt3^Cre Rosa26^YFP lungs (third panel). YFP labelling efficiency in epidermal Langerhans cells (LCs) and dermal CD11b^high (MHC II⁺ EpCAM^-) myeloid cells in Flt3^Cre Rosa26^YFP skin (fourth panel). See Extended Data Fig. 6b and c. Mean ± s.e.m.; P8, n=3; 4-week-old, n=6; 12-week-old, n=11-14; 40-week-old, n=7; 1-year-old, n=3).w, week; y, year. c, Representative images of May-Grünwald-Giemsa stained cytospin preparations of YFP⁺ CD11b^high F4/80^low cells sorted from E18.5 Flt3^Cre Rosa26^YFP fetal liver. Scale bar, 10μm.

Altogether, these data indicate that Flt3^Cre YFP⁺ Kit⁺ progenitors and monocytes account for only a minor fraction of microglia, Kupffer cells, alveolar macrophages and Langerhans cells in young adults. To investigate whether the presence of these adult Flt3^Cre YFP⁺ F4/80^bright macrophages corresponds to their HSC origin, we performed non-myeloablative transplantations of YFP⁺ long term-HSC (LT-HSC) from adult bone marrow into Rag2^-/-γ_c^-/-Kit^W/Wv recipients²³ (Extended Data Fig. 7). Eight weeks after transplantation, the vast majority of HSC, myeloid progenitors, monocytes, and F4/80^low tissue myeloid cells in the recipients were of donor HSC origin. In contrast, only 7% of F4/80^bright macrophages in spleen, 2% in liver, 5% in lung, 13% in pancreas, 2% in epidermis, and 0% in the brain were donor-derived. Thus, recruitment of HSC-derived precursors is not a major mechanism for the maintenance of F4/80^bright macrophages in these tissues.

Collectively, these findings reveal that the transition from yolk-sac- to HSC-derived haematopoiesis occurs late in fetal development for monocytes (E14.5) and granulocytes (E16.5), and suggest that HSC-derived progenitors only marginally replace yolk-sac-derived microglia in the brain, Kupffer cells in the liver, Langerhans cells in the epidermis, although alveolar macrophages and brain CD45⁺ F4/80⁺ macrophages may undergo progressive replacement with age.

Labelling efficiency of most tissue-resident macrophage populations in adult Csf1r^MeriCreMer Rosa26^YFP mice pulse-labelled with 4-hydroxytamoxifen (OH-TAM) at E8.5 was low^4,24. The strength of most genetic pulse-labelling systems is that they allow fate-mapping of cells during a specific time window, however, a weakness is the commonly incomplete labelling which could explain why a large fraction of tissue-resident macrophages remained unlabelled. Hence, based on these data we cannot formally exclude a fetal HSC origin of the unlabelled cells as suggested by others based on transfer of fetal precursors ^24–26.

We thus made use of a newly generated inducible Cre knock-in mouse (Tie2^MeriCreMer) to track hematopoietic output from hematopoietic progenitors and HSC in situ (Busch K. et al., submitted). Tie2/Tek is expressed in endothelial cells, yolk-sac progenitors, AGM, fetal liver and adult HSCs³. We assessed the time window at which Tie2⁺ cells contributed to emerging HSCs and macrophages by injecting tamoxifen at different time points (Fig. 4a, Extended Data Figs. 8-10). Fetal liver E12.5 and E15.5 LT-HSC were labelled efficiently in Tie2^MeriCreMer embryos pulsed at E6.5, E7.5 or E10.5 (Fig. 4b, Extended Data Fig. 8, 9). Yolk sac E9.5 Kit⁺ CD45^low progenitors were also labelled in Tie2^MeriCreMer Rosa26^YFP embryos pulsed at E7.5 (Extended Data Fig. 10). Interestingly, fetal liver cells with a megakaryocyte-erythrocyte progenitor (MEP) phenotype, and F4/80^bright macrophages in yolk sac, brain, and fetal liver were labelled with high efficiency (60%) in embryos pulsed at E6.5 and E7.5, but not in embryos pulsed at E10.5 (Fig. 4b, Extended Data Fig. 8). These fate mapping experiments directly demonstrate that E12.5 and E15.5 fetal macrophages originate from cells that express Tie2⁺ as early as E6.5 and, importantly, before E9.5, and strongly support the notion that fetal liver erythro-myeloid progenitors, and all fetal tissue macrophages up to E15.5 are of yolk sac origin.

Figure 4. Fetal macrophages and adult tissue-resident macrophages originate from Tie2-expressing progenitors prior to E10.5.

a, Fate mapping analysis of Tie2-expressing cells after tamoxifen (TAM) administration at E7.5, or E8.5, or E9.5 or E10.5. Arrows indicate time points for analysis. b, Flow cytometric analysis of fetal liver long-term or short-term hematopoietic stem cells (LT-HSCs, ST-HSCs), multipotent progenitors (MPPs), common myeloid progenitors (CMPs), granulocyte-monocyte progenitors (GMPs), megakaryocyte-erythrocyte progenitors (MEPs) (left panel) and of fetal macrophages (right panel) in the yolk sac, brain, and fetal liver. Time points of labelling (E7.5 (n=7); E10.5 (n=7)) and analysis are indicated, and for each experiment one representative analysis is shown. See Extended Data Fig. 8 for quantitative analysis. c, Frequencies of labelled HSCs and progenitor cells, splenocytes, and F4/80^low CD11b^high myeloid cells and F4/80^bright resident macrophages in spleen, liver lung, epidermis and brain were analysed (mean ± s.d.) from 6-8-week-old Tie2^MeriCreMer animals pulse-labelled at E7.5 (n=4), E8.5 (n=4), E9.5 (n=4) or E10.5 (n=6).

In adult mice pulsed at embryonic stages (E7.5, or E8.5, or E9.5 or E10.5), bone marrow HSC-derived progenitors, peripheral cells (T and B cells, and granulocytes) in the spleen, and CD11b^high F4/80^low myeloid cells in peripheral tissues (spleen, liver and lung) were homogenously labelled at frequencies comparable to HSC labelling, consistent with their adult HSC origin (Fig. 4c). In contrast, YFP labelling frequencies of adult tissue-resident macrophages were maximal in animals pulse-labelled at E7.5, declined at later time points and were minimal when labelled at E10.5 (Fig. 4c). The fact that adult HSC are disconnected from resident macrophages is further underscored by the finding that resident macrophages in mice pulsed at E7.5 were labelled at higher frequencies than adult HSC, i.e. labelling efficiency did not equilibrate with mouse development. In summary, these inducible temporal analyses demonstrate that while both macrophages and HSC originate from progenitors expressing Tie2 as early as E6.5, adult tissue-resident macrophages in the brain (microglia), liver (Kupffer cells), lung (alveolar macrophages), skin (Langerhans cells), and (to some extent) spleen (F4/80^bright macrophages) develop almost exclusively from an Tie2-expressing progenitor pathway distinct from HSCs. These data are consistent with results from Csf1r^MeriCreMer pulse-labelling experiments (see Figs. 1 and 2), with our earlier observation that resident macrophages are independent of the transcription factor Myb⁴, and, finally, complement our data obtained in Flt3^Cre Rosa26^YFP mice (see Fig. 3).

This study demonstrates that Myb-independent tissue resident macrophages⁴ originate from yolk-sac-derived EMP, characterised by expression of Csf1r from E8.5 (16-18 somites). The data do not distinguish whether resident macrophages originate from erythro-myeloid, granulocyte-macrophage, or macrophage only-progenitors because these potentials coexist within the yolk-sac-derived EMP population.

We also provide strong in vivo evidence for engraftment of yolk-sac-derived EMP in the early fetal liver. These cells substantially contribute to the first wave of fetal liver hematopoiesis, followed later by bona fide fetal liver HSC-derived hematopoiesis^8,19,21. Conclusions from recent studies that Langerhans cells and alveolar macrophages are not of yolk sac origin based on transfer of fetal precursors^25,26 should be interpreted in light of our findings that yolk sac EMPs expand in the fetal liver, and are the main source for tissue-resident macrophages.

Under steady-state conditions, yolk-sac-derived macrophages are only marginally replaced by HSC-derived cells in the brain, liver and epidermis. It is remarkable that macrophages of yolk sac origin persist in functionally very distinct tissues, suggesting that the origin is more deterministic of the life span than the tissue location. However, some yolk-sac-derived macrophages can undergo replacement in older mice, as for lung alveolar macrophages. In a third group, exemplified by gut-associated macrophages²⁷, yolk-sac-derived macrophages are replaced by HSC-derived macrophages in the first weeks of post-natal life. The mechanisms responsible for the maintenance of yolk-sac-derived macrophages in certain adult tissues require further investigations. Although yolk-sac- and HSC-derived macrophages can co-exist in the same environment, and their balance be perturbed by pathology, the contributions of these developmentally distinct macrophage populations to homeostasis and inflammation remain to be characterised.

Methods

Animals

Myb^{−/− 28}, Csf1r^{MeriCreMer 29}, Csf1r^{iCre 30}, Flt3^{Cre 31}, Rag2^−/− γ_c^−/−Kit^{W/Wv 23} and Rosa26^YFP reporter ³² mice have been previously described. Rag2^−/−γ_c^−/−Kit^W/Wv were on a mixed genetic background²³, Csf1r^MeriCreMer and Csf1r^iCre mice were on FVB background, other mice were on C57BL/6 (CD45^.2/.2; Ly5.2) background.

Csf1r^MeriCreMer and Csf1r^iCre mice were generated and provided by Jeffrey W. Pollard. Myb^−/− mice were generated and provided by Jon Frampton. Flt3^Cre mice were generated by Conrad Bleul and provided by Thomas Boehm and Sten Eirik Jacobsen. Rosa26^YFP (B6.129X1-Gt(ROSA)26Sor^tm1(EYFP)Cos/J) reporter mice were purchased from The Jackson Laboratory. We generated a new inducible Cre knock-in mouse (Tie2^MeriCreMer) crossed to Rosa26^YFP mice to track the hematopoietic output from HSC in situ (Busch K. et al., submitted). No randomization method was used and the investigators were blinded to the genotype of the embryos and animals during the experimental procedure. Results are displayed as mean ± s.e.m (Fig.1, 2, 3) or s.d. (Fig. 4, Supplementary Table 1). All experiments included littermate controls and the minimum sample size used was 3. Embryonic development was estimated considering the day of vaginal plug formation as 0.5 days post-coitum (dpc), and staged by developmental criteria⁸. In Figs. 1 and 2 and Extended Data Figs. 1-4, embryos were included based on their somite number for embryonic days <E11.5, as described in⁸. No statistical method was used to predetermine sample size.

All animal procedures were performed in adherence to our project licence issued by the United Kingdom Home Office under the Animals (Scientific Procedures) Act 1986, or by the German regional council at the Regierungspräsidium Karlsruhe, Germany, respectively.

Genotyping

PCR genotyping of Myb²⁸, Csf1r^{iCre 30}, Rag2^−/− γ_c^−/−Kit^{W/Wv 23}, Csf1r^MeriCreMer and Flt3^Cre mice⁴ was performed according to protocols described previously. PCR genotyping of Tie2^MeriCreMer will be described elsewhere (Busch K. et al., submitted).

Processing of tissues for flow cytometry

Pregnant females were killed by cervical dislocation or by exposure to CO₂. Embryos ranging from embryonic day (E) 8.25 to E18.5 were removed from the uterus and washed in 4°C phosphate-buffered saline (PBS, Invitrogen). The yolk sac (YS) was harvested from embryos between E8.25 and E12.5. Embryos were exsanguinated through decapitation in 10mM EDTA. To obtain single cell suspensions, organs were incubated in PBS containing 1mg/ml Collagenase D (Roche), 100U/ml DNAse I (Sigma) and 3% fetal calf serum (FCS, Invitrogen) at 37°C for 30min.

Adult tissues (P8 to 1 year) were prepared as follows. Blood was collected by cardiac puncture from anesthetized (isoflurane inhalation) mice. Under terminal anesthesia, mice were perfused by gentle intracardiac injection of 10ml prewarmed (37°C) PBS 1x. The spleen, right liver lobe and right lung lobes were harvested and processed for flow cytometry. To obtain single cell suspensions, organs were incubated 30min in PBS containing 1mg/ml Collagenase D (Roche), 100U/ml DNAse I (Sigma), 2.4mg/ml of Dispase (Invitrogen) and 3% FCS (Invitrogen) at 37°C. Brains from Tie2^MeriCreMer were dissociated and incubated for 60min at 37°C in HBSS with 0.2mg/ml Collagenase D, 20μg/ml Dispase I (Roche), and 50U/ml DNAse I (Sigma). Brain cells were resuspended in isotonic Percoll (Pharmacia) at a final density of 1.072g/ml in HBBS containing 3% FCS. The suspension was underlayered with Percoll solution at 1.088g/ml and overlayered with additional layers of Percoll (1.06, 1.05 and 1.03g/ml). After centrifugation cells were collected from 1.06 and 1.072g/ml layers. Brains from the other strains were processed like described for the spleen. For collection of Langerhans cells from Tie2^MeriCreMer, epidermal sheets were prepared using an epidermis dissociation kit (Miltenyi Biotec). In the other strains, epidermal sheets were separated from the dermis after incubation for 45min at 37°C in 2.4 mg/ml of Dispase (Invitrogen) and 3% FCS (Invitrogen) and the epidermis was further digested for 30min in PBS containing 1mg/ml Collagenase D (Roche), 100U/ml DNAse I (Sigma), 2.4mg/ml of Dispase (Invitrogen) and 3% FCS (Invitrogen) at 37°C.

Flow cytometric analysis of embryonic and adult tissues and cell sorting

Tissues were mechanically dissociated and passed through a 100μm cell strainer (BD). Red blood cell lysis of fetal liver and adult lung and spleen was performed as described³³. Cells were centrifuged at 320 g for 7min, resuspended in 4°C PBS, plated in multi-well round-bottom plates and immunolabelled for FACS analysis. After 15min incubation with purified anti-CD16/32 (FcγRIII/II) diluted 1/50, or ChromPure mouse IgG whole molecule (Dianova) diluted 1/20 in staining buffer (PBS 1X; 0.5% BSA; 2mM ETDA), antibody mixes were added and incubated for 30min. Where appropriate, cells were further incubated with streptavidin conjugates for 20min. The full list of antibodies used can be found in Supplementary Table 2.

Flow cytometry was performed using a BD Biosciences FACSCanto II flow cytometer or a BD Biosciences LSR Fortessa cell analyzer. All data were analysed using FlowJo 9.5 (Tree Star Inc.) or FACS Diva software (BD Bioscience).

Fetal liver, skin and lung YFP⁺ F4/80^bright and YFP⁺ CD11b^high cells from E18.5 Flt3^Cre Rosa26^YFP embryos and from E14.5 and E16.5 Csf1r^MeriCreMer Rosa26^YFP pulsed at E8.5 were sorted into FCS-coated tubes using FACSAria II for cytospin preparations.

Pulse labelling of Csf1r⁺ and Tie2⁺ progenitors

For genetic cell labelling we crossed tamoxifen-inducible Csf1r^MeriCreMer and Tie2^MeriCreMer transgenic mouse strains with Rosa26^YFP reporter mice. In Csf1r^MeriCreMer Rosa26^YFP embryos recombination was induced by single injection at E8.5 of 75μg per g (body weight) of 4-hydroxytamoxifen (Sigma) into pregnant females. OH-TAM was supplemented with 37.5μg per g Progesterone (Sigma). In Tie2^MeriCreMer Rosa26^YFP embryos recombination was induced by treatment of pregnant females by gavage at different time points (between E7.5 and E10.5) with a single dose of 2.5mg tamoxifen (Sigma) and 1.75mg progesterone (Sigma) to counteract the mixed estrogen agonist effects of tamoxifen, which can result in fetal abortions.

Continuous labelling of Csf1r⁺ progenitors

For fate mapping analysis of Csf1r⁺ precursors, Csf1r^iCre females were crossed with homozygous Rosa26^YFP reporter males. Indicated tissues from embryos and adult F1 mice were analysed by flow cytometry.

Fate mapping of Flt3⁺ haematopoietic progenitors

For fate mapping analysis of Flt3⁺ precursors, Flt3^Cre males (the transgene is located on the Y chromosome) were crossed to homozygous Rosa26^YFP reporter females. For adult experiments, Flt3^Cre males were blood phenotyped. Animals with YFP labelling efficiency above 60% in the lymphocytes, monocytes and granulocytes were used for experiments and female littermates were used a Cre negative controls.

Colony Forming assays

Colony-forming unit-culture (CFU-C) assays were performed using Methocult M3434 (Stem Cell Technologies) as described in³⁴. Embryos were collected and dissected in PBS (Gibco, Invitrogen) supplemented with 10% FCS (batch tested and obtained from Gibco), 50 U/ml penicillin, and 50μg/ml streptomycin (Cambrex Corporation). E9 embryos were staged by somite counting. E9 yolk sac and E12 fetal livers were each pooled and incubated for 30 min at 37 °C in PBS supplemented with 10% FCS, 50U/ml penicillin, 50μg/ml streptomycin, 1mg/ml Collagenase D (Roche) and 100U/ml DNase I (Sigma), and dissociated by pipetting. Suspensions were washed, and viable cells were counted on the basis of trypan blue (Sigma) exclusion using a Kova hemocytometer slide.

AA4.1⁺ progenitors were isolated by flow cytometry using using FACSAria II or FACSAria III. Labelling of cells was performed as described above using the following antibodies: CD45-APC-Cy7, Kit-PE and AA4.1-APC, and live cells were gated on the basis of Hoechst 33258 exclusion. Cells were collected into FCS-coated tubes and recounted prior to plating where possible. Gates were defined using unstained, single stained and fluorescence minus one (FMO) stained cells.

Cells were plated in duplicate in 35mm culture dishes according to manufacturer’s instructions. Cultures were grown at 37 °C with 5% CO2 with colonies scored after 10 days.

Colonies were picked and washed once with phosphate-buffered saline (PBS; Gibco, Invitrogen) supplemented with 10% fetal calf serum (FCS; batch tested and obtained from Gibco). Cytospin preparations were stained with May-Grünwald-Giemsa method for morphological inspection of colonies (see below).

Morphological Analysis of sorted cells and colonies

Cytospin preparations were performed using a Cytospin 3 (Thermo Shandon) by centrifuging (i) cells from colonies at 400 rpm for 4min (medium acceleration) or (ii) sorted cells at 500 rpm for 10min (low acceleration). Slides were air-dried for at least 30min, and fixed for 5min in methanol. Methanol-fixed cytospin preparations were manually stained in 50% May-Grünwald solution for 5min, 14% Giemsa for 15min, washed with Sorensons buffered distilled water (pH 6.8) for 5 min and rinsed with Sorensons buffered distilled water (pH 6.8). After air-drying, slides were mounted with Entellan New (Merck) and representative pictures were taken using a Nikon eclipse E6000 microscope with a Nikon Plan Fluor 60X/1.40 NA oil DIC H objective and NIS-elements BR2.30 software (Nikon).

Transplantation of HSC without irradiation

HSC transplantation in non-irradiated Rag2^−/−γ_c^−/−Kit^W/Wv mice was performed as described previously²³. In brief, approximately 1000 LT-HSC (lin^-Sca-1⁺Kit⁺ CD150⁺CD48^-) isolated from the bone marrow of panRosa^YFP mice, which carry a constitutively active YFP reporter allele, were injected into Rag2^−/−γ_c^−/−Kit^W/Wv mice. Recipients were analysed 2 months after transplantation for donor/host chimerism in Blood, spleen, lung, liver, pancreas, brain and epidermis. To test the functionality of E12.5 phenotypic LT-HSC, 10 YFP⁺ LSK CD150⁺CD48^- (phenotypic LT-HSC) from Tie2^MeriCreMerRosa26^YFP pulsed at E7.5 were transplanted into Rag2^-/-γ_c^-/-Kit^W/Wv mice and blood lineages were analysed 16 weeks after.

Extended Data

Extended Data Figure 1. Analysis of Csf1r reporter expression in fetal progenitor cells in Csf1r^iCre Rosa26^YFP.

a, Schematic representation of the different hematopoietic and non-hematopoietic sites dissected in the mouse embryos: yolk sac (YS), aorta-gonado-mesonephros (AGM) region, fetal liver and head. b, Experimental design for fate mapping analysis of Csf1r-expressing cells. Arrows indicate analysed time points c, Kit and CD45 expression on YFP⁺ cells from Csf1r^iCre Rosa26^YFP embryos (E8.25, n=7; E8.5, n=4; E9.25-E9.5, n=16; E10.25, n=9; E10.5, n=5; E11.5, n=8; E12.5, n=5) d Number of YFP⁺ Kit⁺ CD45^low cells per organ/region and developmental time points (mean±s.e.m.) in Csf1r^iCre Rosa26^YFP embryos (upper panel). Number of YFP⁺ AA4.1⁺ Kit⁺ CD45^low cells per embryonic region and developmental time points (mean±s.e.m.) in Csf1r^iCre Rosa26^YFP embryos (lower panel). e, AA4.1 and Kit expression on YFP⁺ cells from Csf1r^iCre Rosa26^YFP embryos (upper panel) and from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (lower panel).

Extended Data Figure 2. Fate mapping analysis of Csf1r-expressing cells.

a, Experimental design for fate mapping analysis of Csf1r-expressing cells. Arrows indicate analysed time points b, YFP expression on live cells from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E6.5 with OH-TAM and analysed at E10.5 (n=2) and E12.5 (n=4). c, Percentage of YFP⁺ cells among Kit⁺ CD45^low cells (YFP labelling efficiency) per organ/region (mean±s.e.m.). Upper panel, Csf1r^iCre Rosa26^YFP embryos (E8.25, n=7; E8.5, n=4; E9.5, n=16; E10.25, n=9; E10.5, n=5; E11.5, n=8; E12.5, n=5); lower panel Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5 (E9.5, n=3; E10.25, n=3; E10.5, n=4; E11.5, n=4; E12.5, n=9). e, Percentage of YFP⁺ cells among AA4.1⁺ Kit⁺ CD45^low cells (YFP labelling efficiency) per embryonic organ/region and developmental time points (mean±s.e.m.). Upper panel, Csf1r^iCre Rosa26^YFP embryos; lower panel: Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5.

Extended Data Figure 3. Csf1r⁺ progenitors have erythro-myeloid potential ex vivo.

a, Sorting strategy for CFU-C (colony forming unit-culture) assays for E9 Csf1r^iCre Rosa26^YFP YS (upper panel) and E12.5 fetal liver from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (lower panel). Dead cells were excluded based on Hoechst 33258 incorporation and, after doublet exclusion, cells were gated based on CD45 and Kit expression. AA4.1⁺ Kit⁺ CD45^low and YFP⁺ AA4.1⁺ Kit⁺ CD45^low cells were isolated from Cre^- and Cre⁺ embryos respectively. b, Mean CFU-C (colony forming unit-culture) frequency from three independent experiments each of E9 Csf1r^iCre Rosa26^YFP YS and E12.5 fetal liver from Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5. CFU-erythroid and/or megakaryocyte (E/Mk); CFU-granulocyte and/or monocyte/macrophage (G/M); CFU-mix, at least three of the following: G, E, M and Mk. c, Morphological validation of colony types obtained from E9 YS Csf1r^iCre YFP⁺ AA4.1⁺ Kit⁺ CD45^low CFU-C assays. Representative images from May-Grünwald-Giemsa stained cytospin preparations of Mixed, E/Mk and G/M colonies. Black arrowhead: macrophages; Granulocyte pathway: blue arrows: Erythroid and Megakaryocyte pathway: red arrows. Scale bar, 10μm.

Extended Data Figure 4. Analysis of Csf1r reporter expression in fetal macrophages and red blood cells in Csf1r^iCre Rosa26^YFP.

a, F4/80 and CD11b expression on YFP⁺ CD45⁺ from YS, head (brain for E11.5), limbs and liver of Csf1r^iCre Rosa26^YFP embryos (E8.5, n=4; E9.5, n=16; E10.5, n=5; E11.5, n=9; E12.5, n=5). Dashed line represent FMO (fluorescence minus one) control. b, Percentage of macrophages (F4/80^bright) among YFP⁺ cells, mean±s.e.m., in Csf1r^iCre Rosa26^YFP embryos (left) and Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (right). See also Extended Data Table 1. c, Percentage of YFP⁺ cells among F4/80^bright cells (YFP labelling efficiency) per embryonic organ/region and developmental time points (mean±s.e.m.). Left panel, Csf1r^iCre Rosa26^YFP embryos (E8.25, n=7; E8.5, n=5; E9.5, n=15; E10.25, n=9; E10.5, n=5; E11.5, n=9; E12.5, n=5); right panel: Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5 (E9.5, n=3; E10.25, n=3; E10.5, n=4; E11.25, n=4; E12.5, n=9). d, YFP expression in erythrocytes (CD45^- Ter119⁺) from YS and fetal liver of Csf1r^iCre Rosa26^YFP embryos (left) and Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (right).

Extended Data Figure 5. Analysis of Flt3 reporter expression in blood leucocytes, stem/progenitor cells, fetal red blood cells, and adult liver, lung and spleen in Flt3^Cre Rosa26^YFP mice.

a. YFP labelling efficiency in blood lineages at different embryonic and adult time points (E14.5, n=9; E16.5, n=9; E18.5, n=7; P8, n=7; 4-week-old, n=6; 12 week-old, n=9; 40 week-old, n=7) in Flt3^Cre Rosa26^YFP mice are shown. Lymphocytes were gated as CD3⁺/CD19⁺, Granulocytes (CD11b⁺ Gr1⁺ CD115^-), Gr1⁺ monocytes (CD11b⁺ Gr1⁺ CD115⁺), Gr1^- monocytes (CD11b⁺ Gr1^- CD115⁺) and Red Blood Cells (RBCs, CD45^- Ter119⁺). b, YFP labelling efficiency in bone marrow LT-HSC, ST-HSC, MPP and Lin^- Sca1^- Kit⁺ progenitors in 4 (n=3) and 12 week-old (n=6) Flt3^Cre Rosa26^YFP mice. c, YFP labelling efficiency in fetal liver RBC progenitors (CD45⁺ Ter119⁺) and RBCs (CD45^- Ter119⁺) in Flt3^Cre Rosa26^YFP mice (E14.5, n=5; E16.5, n=5; E18.5, n=7) and comparison of YFP labelling efficiency in fetal liver and blood RBCs in Flt3^Cre Rosa26^YFP mice at E14.5 (n=5), E16.5 (n=5) and E18.5 (n=7). d, Expression of Gr1 and MHC II, Ly-6G and Siglec-F, CD11c and CD64, and Nkp46 and CD19 among F4/80^low CD11b^high myeloid cells in the liver. Histograms represent Flt3^Cre YFP labelling efficiency in the following defined populations: Granulocytes (Gr1⁺ MHC II^- or Ly-6G⁺), Eosinophils (Siglec-F⁺), Dendritic cells (CD11c⁺), B cells (CD19⁺) and NK cells (Nkp46⁺) (n=3). e, Analysis of F4/80^low CD11b^high myeloid cells in the lung as in (b). f, Analysis of F4/80^low CD11b^high myeloid cells in the spleen as in (b). g, Expression of CD64 in F4/80^bright macrophages and F4/80^low myeloid cells in the liver, lung and spleen (FMO, Fluorescence minus one).

Extended Data Figure 6. Characterisation of fetal F4/80^low CD11b^high myeloid cells in liver, lung and skin.

a. F4/80, Kit, CD11b and Gr1 expression on YFP⁺ CD45⁺ cells in the fetal liver at E14.5 in Csf1r^MeriCreMer Rosa26^YFP embryos pulsed at E8.5 (left panel). Representative images of May-Grünwald-Giemsa stained cytospin preparations of fetal liver YFP⁺ F4/80^bright and YFP⁺ CD11b^high cells sorted from E14.5 Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (right panel). Scale bar,10μm. b, c, F4/80, CD11b, Gr1 and Siglec F expression on CD45⁺ cells in the embryonic and post natal lung (b) and skin (c) in Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5 (green) and Flt3^Cre Rosa26^YFP embryos (orange). Representative images of May-Grünwald-Giemsa stained cytospin preparations of lung YFP⁺ F4/80^bright and YFP⁺ CD11b^high F4/80^low (b) and skin YFP⁺ F4/80^bright and YFP⁺ Kit⁺ F4/80^- CD11b^- mast cells (c) sorted from E18.5 Flt3^Cre Rosa26^YFP embryos and E16.5 Csf1r^MeriCreMer Rosa26^YFP embryos pulsed with OH-TAM at E8.5. Scale bar 10μm.

Extended Data Figure 7. Adult BM transplantation reconstitutes the hematopoietic system but does not replace tissue resident F4/80^bright macrophages.

a, Schematic representation of transplantation experiments. LT-HSC isolated from bone marrow of panRosa26^YFP donor mice were injected into Rag2^-/-γ_c^-/- Kit^W/Wv recipients (approx. 1000 cells/recipient). Eight weeks after transplantation stem cells, myeloid progenitors, monocytes and macrophages of recipient mice were analysed for donor chimerism. b, Long-term or short-term hematopoietic stem cells (LT-HSC, ST-HSC), multipotent progenitors (MPP), common myeloid progenitors (CMP), granulocyte-monocyte progenitors (GMP), megakaryocyte-erythrocyte progenitors (MEP), and circulating Ly6C^hi and Ly6C^lo monocytes were isolated from transplanted Rag2^-/-γ_c^-/-Kit^W/Wv mice and analysed for YFP expression. c, F4/80^bright macrophages and F4/80^low myeloid cells in spleen, liver, lung, pancreas, epidermis and brain were analysed for YFP expression.

Extended Data Figure 8. Analysis of fetal stem/progenitor cells and fetal macrophages in Tie2^MeriCreMer Rosa26^YFP embryos pulse-labelled from E6.5 to E10.5.

a, Experimental design for fate mapping analysis of Tie2^MeriCreMerRosa26^YFP embryos pulse-labelled at E6.5, or E7.5, or E8.5, or E9.5 or E10.5. b, c, Representative flow cytometry of fetal liver stem/progenitor cells (b) and of fetal macrophages (c) in the yolk sac, head region, and embryo body at E12.5, injected at E6.5 or at E10.5. d, Representative images of May-Grünwald-Giemsa stained cytospin preparations of sorted YFP⁺ and YFP^- CD45⁺ F4/80^bright macrophages of the embryo proper or the head region of E13.5 Tie2^MeriCreMerRosa26^YFP embryos pulsed at E7.5. Scale bar, 10μm e, Quantification of the percentage of YFP⁺ stem/progenitor cells in the fetal liver and macrophages in yolk sac, brain (head) and embryo body at E12.5. Embryos were labelled at E6.5 (n=5), or E7.5 (n=7), or E8.5 (n=4), E9.5 (n=5) or E10.5 (n=7) and analysed at E12.5 (mean±s.d.).

Extended Data Figure 9. Transplantation of YFP⁺ fetal liver LT-HSC from Tie2^MeriCreMerRosa26^YFP into Rag2^-/- γ_c^-/-Kit^W/Wv mice.

a, Experimental design for pulse-labelling and LT-HSC sort from Tie2^MeriCreMerRosa26^YFP embryos pulsed at E7.5. b, Fetal livers of E12.5 embryos pulsed at E7.5 were harvested, YFP⁺ LSK CD150⁺CD48^- LT-HSC were sorted by flow cytometry and 10 LT-HSC were injected into Rag2^-/-γ_c^-/-Kit^W/Wv recipients. c, Blood analysis of recipients 16 week after LT-HSC transplantation. Donor-derived (YFP⁺) and recipient-derived (YFP^-) blood cells were analysed for expression of CD19, CD3, CD11b, and Gr-1. One representative example is shown.

Extended Data Figure 10. Analysis of E9.5 YS progenitor cells in Tie2^MeriCreMer Rosa26^YFP embryos pulse-labelled at E7.5.

a, Experimental design for fate mapping analysis of Tie2-expressing cells in Tie2^MeriCreMerRosa26^YFP embryos. Embryonic cells were pulse-labelled by Tamoxifen (TAM) administration into pregnant Tie2^MeriCreMer mice at E7.5. Yolk sac (YS) and embryo proper (EP) of E9.5 Tie2^MeriCreMerRosa26^YFP embryos were analysed by flow cytometry. b, Quantification of total living cells and YFP⁺ living cells in YS and EP of analysed embryos (mean±s.d., n=5). c, Flow cytometry analysis of Kit⁺ CD45^low cells among total living cells (black) or living YFP⁺ cells (blue) in YS and EP of a representative E9.5 Tie2^MeriCreMerRosa26^YFP embryo (left) and quantification of all analysed embryos (right; mean±s.d., n=5). d, Analysis of F4/80⁺ fetal macrophages among CD45⁺ cells (black) or YFP⁺ CD45⁺ cells (blue) in YS and EP (mean±s.d., n=5) and quantification of all analysed embryos (right; mean±s.d., n=5). e, Percentage of YFP⁺ cells (YFP labelling efficiency) among live cells, Kit⁺ CD45^low, CD45⁺ Kit^- cells and F4/80⁺ cells from the YS and EP of E9.5 Tie2^MeriCreMerRosa26^YFP embryos pulsed at E7.5.