Quantitative developmental anatomy of definitive haematopoietic stem cells/long-term repopulating units (HSC/RUs): role of the aorta-gonad-mesonephros (AGM) region and the yolk sac in colonisation of the mouse embryonic liver

Embryonic development of the mammalian haematopoietic system is complex and, in many aspects, poorly understood. Two major sources of haematopoietic activity have been identified in vertebrate embryos. For a long time the yolk sac (YS), where haematopoietic activity is first observed, was assumed to be the primary site of formation of the haematopoietic stem cells (HSCs) that migrate to and colonise the foetal liver and subsequently the bone marrow(Moore and Metcalf, 1970). However, identification of a powerful intraembryonic HSC activity in avian,amphibian and murine embryos cast doubts on this assumption. In experimentally engineered chick-quail chimeras, haematopoiesis in the adult organism was found to originate from the body but not from the YS of the chimera,suggesting that the YS/embryonic haematopoietic hierarchy is transitory(Dieterlen-Lievre, 1975;Martin et al., 1979). In amphibian embryos the dorsal lateral plate (DLP) mesoderm next to the dorsal aorta contributes mainly to definitive haematopoiesis, whereas the ventral blood island (VBI), which is equivalent to the YS, contributes to both primitive and to some extent definitive haematopoiesis(Chen and Turpen, 1995). Recent experiments on amphibian embryos, using individual blastomere labelling techniques, have demonstrated that although at the gastrulation stage both DLP and VBI originate from a common ventral bipotential mesoderm layer(Turpen et al., 1997) their precursors are spatially separated at the blastula stage(Ciau-Uitz et al., 2000). Relatively recently an intra body site of definitive haematopoietic activity has been identified in the mouse embryo prior to definitive haematopoiesis in the liver (Godin et al., 1993;Medvinsky et al., 1993;Muller et al., 1994;Cumano et al., 1996;Medvinsky and Dzierzak, 1996;Medvinsky et al., 1996). Before the onset of organogenesis and the establishment of the circulatory system in the mouse embryo, this site, termed the visceral para-aortic splanchnopleura (P-Sp) region, but not the YS, contains multipotent lymphomyeloid progenitors, although at this stage they are incapable of reconstituting adult recipients (Cumano et al., 1996; Cumano et al.,2001). During organogenesis, part of the splanchnopleura transforms into a morphologically distinct composite axial structure consisting of the dorsal aorta, genital ridges and mesonephroi (AGM region). High numbers of spleen-colony forming units (CFU-S) are concentrated in the AGM region prior to the presence of CFU-S activity in the liver(Medvinsky et al., 1993;Medvinsky and Dzierzak, 1996). The first definitive (long-term reconstituting) HSCs appear in the AGM region and umbilical vessels at late 10/early 11 d.p.c.(Muller et al., 1994;Medvinsky and Dzierzak, 1996;de Bruijn et al., 2000). The AGM region is the only tissue in the 10 d.p.c. mouse embryo capable of the autonomous initiation/expansion of definitive HSCs, as demonstrated using an organ culture approach. Slightly later, by 11 d.p.c., HSCs appear both in the liver rudiment and the YS but again only the AGM region is capable of expanding the number of HSCs in organ culture conditions(Medvinsky et al., 1996). These and other features of AGM biology, including the kinetics of CFU-S and HSC development, led to the hypothesis that the AGM region is the primary site of formation of definitive HSCs which then colonise secondary haematopoietic organs, primarily the foetal liver(Dzierzak and Medvinsky, 1995;Dzierzak and Medvinsky, 1998). Recently, more compelling evidence has emerged that indicates HSCs originate in the body of the embryo independently of YS activity. Indeed, as early as day 7 p.c., before circulation is established in the mouse embryo, the P-Sp but not the YS contains ancestors of cells that are capable of long-term,multi-potential repopulation of adult irradiated recipients devoid of NK cells(Cumano et al., 2001). Experiments on human embryos also revealed dramatic differences in the lymphohaematopoietic potential of the dorsal aorta and the YS(Tavian et al., 2001). Upon transplantation into NOD-SCID mice, cultured dorsal aorta cells showed lymphomyeloid reconstitution whereas YS cells were only capable of contributing to the myeloid lineage. Cumulatively, these data from several different research groups point to the embryo body as the site of origin of definitive HSCs.

The above data does not, however, rule out a possibility that at later stages the YS is involved in the independent production or expansion of definitive HSCs. Although early YS cells are unable to repopulate adult irradiated recipients upon direct transplantation, when transplanted into the embryo they can contribute to adult haematopoiesis(Weissman et al., 1978;Toles et al., 1989). Analogous results have been achieved by transplantation of YS cells into newborn recipients (Yoder and Hiatt,1997; Yoder et al.,1997). In addition, by day 8 p.c. both the P-Sp and the YS contain cells that can mature into definitive HSCs by co-culture with an AGM-derived stromal cell line (Matsuoka et al.,2001). Thus at least from day 8 p.c. the YS contains cells(pre-definitive HSCs), which are capable of development into definitive HSCs upon maturation in an embryonic or newborn microenvironment(Medvinsky and Dzierzak,1999). However, it remains unclear if and when during normal embryo development these cells mature into definitive HSCs and whether they have to migrate to an inductive AGM microenvironment in order to do so.

In our previous papers we focused on the initiation of definitive HSC production in the mouse embryo (Muller et al., 1994; Medvinsky and Dzierzak, 1996). Here we explore the subsequent stage of HSC development from day 11 until day 13 when the number of definitive HSCs increases in the embryo. From day 11 p.c. onwards the number of HSC/RUs increases dramatically in the liver(Morrison et al., 1995;Ema and Nakauchi, 2000) but little is known about their distribution in the rest of the embryo, the routes of their migration and the mechanisms underlying their expansion in the liver. Various types of multipotent, pluripotent and bi-potent myeloid and lymphoid progenitors have been identified during embryogenesis in the YS, AGM region,liver, thymic and splenic rudiments (Moore and Metcalf, 1970; Velardi and Cooper, 1984; Johnson and Barker, 1985; Wong et al.,1986; Eren et al.,1987; Liu and Auerbach,1991; Cumano et al.,1992; Morrison et al.,1995; Ema et al.,1998; Kawamoto et al.,1998; Nishikawa et al.,1998; Liu et al.,1999; Ohmura et al.,2001; Palis et al.,2001; Traver et al.,2001; Douagi et al.,2002) and different types of progenitor cells are disseminated via the circulation (Moore and Metcalf,1970; Johnson and Barker,1985; Rodewald et al.,1994; Delassus and Cumano,1996). However, as the links within the haematopoietic hierarchy and between tissues are unclear, it is difficult to resolve an entire anatomical picture of the development of definitive HSCs. Although data from some publications has revealed fragments of it(Moore and Metcalf, 1970;Ikuta et al., 1990;Morrison et al., 1995;Berger and Sturm, 1996;Sanchez et al., 1996;de Bruijn et al., 2000;Ema and Nakauchi, 2000;Hsu et al., 2000;de Bruijn et al., 2002;North et al., 2002), a comprehensive quantitative anatomical map of HSC/RUs development during embryogenesis has not been produced. Here, we have attempted to create a temporal and spatial map of development of definitive HSC/RUs in the 11-13 d.p.c. mouse embryo. To this end, the number of HSC/RUs in different embryonic tissues has been estimated using a limiting dilution method(Szilvassy et al., 1990). In addition, the potential of the AGM region and the YS to produce and/or maintain definitive HSC/RUs has been assessed using an organ culture approach(Medvinsky and Dzierzak,1996).

We have found that expansion of the pool of HSC/RUs within the liver occurs concurrently with increasing numbers of HSC/RUs in the circulation. In addition to the previously reported activity of the AGM region during day 10-11 p.c. (Medvinsky and Dzierzak,1996; de Bruijn et al.,2000), we report here that a day later, at 12 d.p.c., the YS becomes competent to generate (and/or expand) definitive HSC/RUs. This finding suggests that both the AGM region and the YS produce HSCs that colonise the developing liver in two subsequent waves, the peaks of which fall on days 10-11 p.c. and day 12 p.c. respectively.

CBA/Ca, C57BL/6 and (CBA×C57BL6) F₁ mice were bred in the animal breeding unit of the University of Edinburgh or purchased from Harlan. Cell suspensions were prepared from non-cultured or cultured 11-13 d.p.c. embryonic tissues after incubation with 0.1% collagenase-dispase (Sigma) in PBS at 37°C. Embryonic blood was collected within 2-3 minutes after separation of the YS from the embryo body and kept on ice before transplantation to preserve HSCs. Special care was taken to remove umbilical and vitelline arteries from preparations of the YS and embryonic circulation.(CBA×C57BL6) F₁ mouse male embryos were used as donors for transplantation into (CBA×C57BL6) F₁ female recipients. In some experiments (CBA×C57BL6) F₁ Ly5.1/2 embryos were used for transplantation into (CBA×C57BL6) F₁ Ly5.2 female recipients. Recipient mice were irradiated at 9.5 Gr split into two doses separated by a 3 hour interval in the Cs source at a rate 21.6 rad/minute. The mice received neomycin (0.16 g/100 ml) in acidified drinking water for the first 4 weeks after transplantation.

Tissues were cultured in myelo-cult medium (Stem Cell Technology)supplemented with 10^-6 M hydrocortisone hemi-succinate (Sigma) on Durapore 0.65 μm filters (Millipore) supported by stainless steel stands(5% CO₂ in air) at the gas-liquid interface as previously described(Medvinsky and Dzierzak,1996). 12 d.p.c. foetal liver was explanted in small pieces comparable in size to the AGM region. Cultures were set up for 3-5 days and 7-9 days.

The contribution of donor cells was assessed 6-8 weeks and 3.5-5 months after transplantation as described previously(Medvinsky and Dzierzak,1996). Briefly, the percentage of male donor cells in the haematopoietic system of female recipients was assessed by comparison with standards of serially diluted male in female DNA (0.1%, 1%, 10%, 100%). Both test DNA and standards were amplified by PCR using primers specific for male Y2B and mouse myogenin sequences. As previously, we have restricted our analysis to those cells (definitive HSCs), which upon transplantation,contributed at a level of 10% or higher in the haematopoietic system of irradiated recipients (Muller et al.,1994; Medvinsky and Dzierzak,1996).

Transplantations of day 12 and 13 tissues in some cases were carried out using Ly5.1 embryos and multilineage contribution was assessed in recipient mice by assessing co-expression of the Ly5.1 marker and lineage specific markers by antibody staining and subsequent analysis on FACSCalibur(Beckton-Dickinson). For this purpose biotinylated anti-Mac-1, anti-B220(secondary stained with Strepdavidin-PE; Sigma), PE-conjugated anti-CD3ϵand FITC-conjugated anti-Ly5.1 antibodies (Pharmingen) were used. In some cases the bone marrow of reconstituted mice was transplanted into secondary recipients.

Upon transplantation, one definitive HSC is sufficient to differentiate into all lymphoid and myeloid cell types and to contribute over several months at a high level to the haematopoietic system of an irradiated recipient(Lemischka, 1992;Morrison et al., 1997). In order to estimate the number of HSCs/RU in various organs we have adopted a limiting dilution method by transplanting low numbers of HSCs (several dilutions) into irradiated recipients(Szilvassy et al., 1990). Test cells were co-transplanted intravenously with 2×10⁴ bone marrow cells to ensure short-term survival of the recipient. The number of HSCs in tested tissues was estimated by Poisson statistics based on the proportion of non-repopulated recipients in long-term (longer than 3.5 months)repopulating experiments. Serial dilutions are expressed in embryo equivalents(e.e.). A minimum of 2 and maximum of 12 different dilutions were used for each tissue. For each dilution between 4 and 21 recipients were transplanted in a minimum of two independent replicate experiments. The final numbers of HSCs were estimated by the maximum likelihood method using Genstat 5 package and expressed as the most probable numbers (MPN)(GenStat 5 Release 3 Reference Manual,1993). The asymmetric error range in parentheses next to MPN and also showing on the graphs is typical for this kind of analysis which involves confidence interval estimation of the Poisson mean.

Our calculations are based on the assumption that HSCs from different tissues and at different stages of development have an equal seeding efficiency. However, this may not be the case and therefore we, as others(Ema and Nakauchi, 2000),introduce an operational term `repopulating unit' (RU), which is not necessarily related to a single cell and appears next to HSC abbreviation in the text.

In separate experiments we have found that the 2×10⁴carrier bone marrow cells injected per recipient contain on average 2 (1.6,2.6) long-term repopulating HSC/RUs, which is similar to numbers reported by some other groups (Abkowitz et al.,2000). This may explain why when other researchers transplanted 10 times more bone marrow carrier cells (2×10⁵) (about 20 HSC/RUs) along with day 11 embryonic liver no liver contribution was detected in recipient mice (Ema and Nakauchi,2000), as the donor HSCs may have been out competed by an excess of HSCs in the carrier bone marrow.

According to Poisson statistics the number of HSC/RUs within the AGM region, YS and the liver of the 11 d.p.c. embryo is close to 1 per tissue(Fig. 1). In the present experiments in contrast to our previous report(Muller et al., 1994) we are now able to detect some HSC/RUs in 11 d.p.c. embryonic blood, possibly because of improved methods of preservation of embryonic blood cells before transplantation (see Materials and Methods). Bone marrow from primary recipients reconstituted with HSC/RUs from the 11 d.p.c. circulation could be successfully transferred to secondary recipients (data not shown). This finding has revealed a route for the dissemination of HSC/RUs from the AGM region as early as day 11 p.c., which has previously been suggested but never confirmed experimentally. In these present experiments we were not able to reconstitute recipient mice with the body remnants of day 11 embryos (data not shown) and therefore attribute the previously reported rare cases of reconstitution with body remnants (Muller et al., 1994) to the presence of circulating HSC/RUs and/or the occasional inclusion of umbilical and vitelline vessels into the transplant(de Bruijn et al., 2000).

On day 12 p.c. both the AGM region and the YS contain approximately two to three HSC/RUs each, which is higher than on day 11 p.c.(Fig. 1A). At this time the embryonic circulation also contains about three HSC/RUs. The significant numbers of HSC/RUs in the circulation suggests intensive trafficking of HSC/RUs within the embryo. By day 12 p.c. the number of HSC/RUs in the embryonic liver reaches 53 (43,69) thus increasing approximately 50-fold from day 11 p.c. Multilineage contribution to recipient mice was confirmed by analysis of selected mice(Fig. 2).

Estimates of possible HSC/RUs numbers derived from the circulation in these organs showed that HSC/RUs numbers within both the AGM region and the YS are significantly above the numbers of HSC/RUs attributable to circulating blood present in these tissues (Table 1). To directly test the number of circulating HSC/RUs contained within the AGM region we flushed out 12 d.p.c. dorsal aorta(Table 2). As expected the flushed out samples of embryonic blood contained fewer HSC/RUs than remained in the dorsal aorta [0.3 (0, 1.7) and 1.6 (1.2, 2.4)respectively]. Therefore, either both the AGM region and the YS on day 12 p.c. are involved in the specific production of HSC/RUs, or circulating HSC/RUs have been selectively retained in these tissues. This issue has been more closely examined in organ culture experiments described below.

In contrast to day 11 p.c., on day 12 p.c. HSC/RUs were consistently detected in the body of the embryo. The number of HSC/RUs in the body of the 12 d.p.c. embryo without the liver was estimated to be 12.1 (8.1,19.3). When in addition to the liver, the AGM region and blood were also removed the total number of HSC/RUs in all remaining tissues was 5.8 (4.4,8.2) (Table 1). From the amount of blood in body transplants (Table 1) we estimate that about 1.6 HSC/RUs in the embryo body belong to the pool of circulating HSC/RUs. Therefore, it may be that apart from the AGM region and the YS a few HSC/RUs are harboured in other tissues of the body. Amongst individually tested tissues (thymus, spleen, lung,forelimbs, heart and head) transplanted separately, the lungs consistently reconstituted irradiated recipients. They contained 0.4 (0.2, 0.6)HSC/RUs, which is above the expected 0.06 HSC/RUs that would be brought there by the circulation (Table 1). Forelimbs also contained about 0.5 (0.3, 0.9) HSCs. Some untested tissues may contain solitary HSC/RUs as well.

By day 13 p.c. of development the number of circulating HSC/RUs in the embryonic vasculature remains high 5.9 (4.7, 7.7) and the number of HSC/RUs in the liver continues to grow, reaching 260 (212, 320)(Fig. 1). The number of HSC/RUs decreases in both the AGM region and the YS to 0.8 (0.6, 1.2) and0.8 (0.6, 1.2) HSC/RU per tissue respectively(Fig. 1,Table 1). The total number of HSC/RUs in the body outside the liver remains stable compared to 12 d.p.c. body (Table 1) and HSC/RUs are no longer detectable in the lungs. Forelimb transplants reconstituted three out of five recipient mice in two dilutions. Since the amount of blood in distal limbs was extremely low, freely circulating HSC/RUs are unlikely to account for this (Table 1). HSC/RUs found in 12-13 d.p.c. limbs may reflect early colonisation of developing long bones with HSC/RUs. Further analysis of the limbs of 14 d.p.c. embryos will be required to test the reliability of this conclusion.

In the day 11 p.c. embryo five sites contain HSC/Rus: the AGM region, the YS, the liver, blood and umbilical vessels(Muller et al., 1994;Medvinsky and Dzierzak, 1996;de Bruijn et al., 2000). It has been shown using an organ culture approach that the AGM region is the only one of these tissues capable of expanding or generating HSC/RUs at this age(Medvinsky and Dzierzak, 1996;de Bruijn et al., 2000). Here we have quantitatively reassessed the potential of 11 d.p.c. embryonic tissues to produce HSC/RUs, by quantifying the number of HSC/RUs that are produced and maintained in organ culture. After 3-5 days in culture, the number of HSC/RUs in 11 d.p.c. AGM region increased from 0.9 (0.7, 1.1) to 12 (10.0,17.6) (P<0.05) (Fig. 3A) followed by a drop in the numbers of HSC/RUs after 7-9 days in culture.

As was shown previously, 11 d.p.c. YS is incapable of expanding the initial numbers of explanted HSC/RUs. Each explant of 11 d.p.c. YS contains approximately one HSC before and after 3 days in culture(Fig. 3A). However, in contrast to the AGM region, 11 d.p.c. YS explants are not able to maintain HSC/RUs for a longer time in culture. Similarly, in 11 d.p.c. liver about 1 HSC can be detected after 3 but not after 7 days in culture.

When 12 d.p.c. AGM region was tested, we found that after 3-5 days in vitro it contained the same number of HSC/RUs as it initially contained in the embryo (Fig. 3B). A slight increase in the numbers of HSC/RUs was observed after 7-9 days in culture. 12 d.p.c. AGMs are larger than 11 d.p.c. AGMs and therefore culture conditions could be suboptimal. To optimise the culture we reduced the size of the explants by subdissection of 12 d.p.c. AGMs and found no signs of HSC expansion in these cultures either (data not shown). In addition, we found that the microenvironment of the 12 d.p.c. AGM is highly supportive of long-term (up to 4 weeks) maintenance of HSC/RUs (Kumaravelu et al.,unpublished observation). Thus, we infer that the ability of the AGM region to expand (and/or generate) definitive HSC/RUs is significantly attenuated on day 12 p.c., as compared to 11 d.p.c., concurrent with progressive specification of the AGM region into gonads and mesonephric derivatives.

In contrast to 11 d.p.c. YS, 12 d.p.c. YS explants acquire the capacity to increase the numbers of HSC/RUs during culture. Before culture 12 d.p.c. YS contains 1.8 (1.4, 2.4) HSC/RUs but after 3 days in culture it contains6.8 (5.0, 9.8) HSC/RUs per explant (P<0.05)(Fig. 3B). However, like the 11 d.p.c. YS and in contrast to the AGM region, 12 d.p.c. YS was not able to maintain HSC/RUs in long-term cultures; after 7 days in culture the number of HSC/RUs dropped to less than 1, being 0.6 (0.4, 1.2) HSC/RUs per YS. This may possibly reflect the transitory nature of haematopoietic activity in the YS in vivo.

Explants of 12 d.p.c., foetal liver were not able to maintain the initial number of HSC/RUs in culture, which could be explained either by suboptimal culture conditions for this tissue or by immaturity of day 12 liver microenvironment (Fig. 3C).

Shortly before the onset of organogenesis the embryo starts to generate a transitory population of embryonic haematopoietic cells that serve its immediate needs. These first haemopoietic cells, consisting mainly of committed myeloid progenitors, CFU-S and primitive erythroid cells, appear in the embryonic circulation in growing numbers and then colonise the initially haematopoietically inactive embryonic liver(Moore and Metcalf, 1970;Johnson and Barker, 1985;Medvinsky, 1993;Medvinsky et al., 1996). Definitive HSCs, which give rise to the adult haematopoietic hierarchy,develop slightly later and gradually form a massive pool in the foetal liver,which becomes the main source of HSCs which subsequently colonise the bone marrow (Dzierzak and Medvinsky,1995; Morrison et al.,1997; Dzierzak et al.,1998). Owing to the absence of unique markers for definitive HSCs at present, direct monitoring of their origin and movements during development, as well as definition of the developmental boundaries of the haematopoietic system is not possible and can only be inferred indirectly from HSC functional assays. Some markers have been successfully used to narrow down the anatomical location of definitive HSCs but their expression is not restricted purely to HSCs (North et al.,1999; Manaia et al.,2000; Ma et al.,2002). In addition, HSC migration is not necessarily restricted to the vascular network and their location in tissues may not be accompanied, as has been shown for the AGM region, by active haematopoiesis(Medvinsky et al., 1996;Godin et al., 1999). Although some progress has been achieved, in contrast to the development of solid organs, an anatomical and histological description of development of the definitive haematopoietic system is lacking. Here, using a quantitative approach we analyse the early development of definitive HSC/RUs, the `germinal layer' of the definitive haematopoietic system.

Early development of definitive HSC/RUs in the mouse embryo involves at least two key stages; (i) initiation of definitive HSC/RUs (late day 10-early day 11 p.c.) and (ii) expansion of the pool of definitive HSC/RUs. In addition, some tissues may be involved in the maintenance of HSC/RUs. Since HSCs may rapidly change their location during embryogenesis, detection of HSCs in tissues does not identify whether these tissues are capable of generating,expanding, or maintaining them or whether these tissues merely transiently contain them. In order to try and distinguish between these possibilities, we and others previously developed an organ culture approach which enables isolated tissues to be individually tested to reveal their HSC activity(Medvinsky and Dzierzak, 1996;Cumano et al., 2001). For example, the number of HSC/RUs within the AGM region on days 10-11 p.c. is no higher than within the YS and the liver. However, in contrast to other embryonic tissues at this age, the AGM region is capable of autonomously initiating and expanding HSC/RUs in vitro suggesting that the initial pool of definitive HSC/RUs is generated within the AGM region and these then colonise the liver (Morrison et al.,1995; Medvinsky and Dzierzak,1996; Ema and Nakauchi,2000).

Here, using an improved protocol we have been able to detect rare HSC/RUs in day 11 circulation (compare with our previous report)(Muller et al., 1994). This reveals a previously proposed, but never experimentally shown, route by which 11 d.p.c. AGM derived HSCs can colonise the liver. During days 12-13 p.c. the number of HSC/RUs in the circulation rises (approximately 6 HSC/RUs were detected on day 13 p.c.) consistent with the more intense colonisation of the foetal liver with HSCs from extra-hepatic sources during this period. It is worth bearing in mind that HSC/RUs present in the circulation at the moment of embryo dissection are likely to represent only a small proportion of total HSC/RUs present in the circulation over the entire day of gestation.

Our present calculations indicate that the total number of HSC/RUs within the developing embryo increases dramatically from about 3 (day 11 p.c.) to 66(day 12 p.c.) mainly due to accumulation of HSC/RUs in the liver(Table 3). This is in accordance with previously published numbers of HSC/RUs in day 12 liver(Ema and Nakauchi, 2000). Owing to the length of the cell cycle it is unlikely that such an increase in HSC number occurs entirely from amplification of a few HSC/RUs that initially colonized the liver. If the possibility of de novo/primary formation of HSC/RUs in the liver is ruled out then this increase must be the result of a massive immigration of HSC/RUs from extra-liver source(s). Our quantitative data are in accord with this hypothesis as explained below.

In the 11 d.p.c. embryo the number of HSC/RUs within the liver is low and can be easily explained by colonisation from the AGM region. In fact, one 11 d.p.c. AGM region can produce as many as twelve HSC/RUs after 3 days in culture; and we assume that in vivo this process may be significantly more efficient. By 12 d.p.c. the HSC productivity of the AGM region decreases as assessed by the organ culture test. However measurement of HSC/RUs numbers in uncultured AGMs showed that the number of HSC/RUs in the AGM region is still higher than would be expected in a non-haematopoietic tissue. Conversely, by 12 d.p.c. the YS showed noticeable HSC productivity in vitro. It was capable of expanding the number of HSC/RUs from about 1.8 to 6.8 during a 3-day culture period.

Thus, at early stages of liver colonisation the high cumulative HSC productivity of the AGM region and the YS may provide the liver with a large part of the `ready-to-use' pool of HSCs(Fig. 4). It is interesting that in the same culture conditions the liver itself was unable to increase or even maintain the initial number of HSCs. This suggests either that by day 12 p.c. the liver is not yet competent to expand HSCs or that the culture conditions used are not fully adequate. It is important to note in relation to this that a cell line derived from day 14 foetal liver is capable of maintaining HSCs over a period of 1 month(Moore et al., 1997).

The generation/expansion of definitive HSCs/RUs in 12 p.c. YS culture is likely related to previous observations that from 8-10 d.p.c. the YS contains immature cells that when placed into an embryonic or newborn environment become capable of contributing into adult haematopoiesis(Weissman et al., 1978;Toles et al., 1989;Yoder and Hiatt, 1997;Yoder et al., 1997;Matsuoka et al., 2001). Our data suggest that these early YS cells may not necessarily need processing in the AGM region to become functional definitive HSCs but can mature in situ on day 12 p.c. The controversy over the origin of HSCs has always centred on the issue of whether the P-Sp/AGM region or the YS is the initial site of their generation. However, these data indicate that definitive HSCs may develop independently and asynchronously in these two different sites of the mouse embryo suggesting that the argument over the first source of HSCs is not relevant. However, at present, the possibility of cross-seeding of the YS and the AGM region with definitive HSCs and/or their ancestor cells cannot be excluded. Lineage tracing of YS and AGM haematopoiesis from early stages of development is required to finally resolve this issue.

In summary, we have carried out a comprehensive anatomical mapping of the development of definitive HSC/RUs in the mouse embryo from 11-13 d.p.c. during which time the embryonic liver becomes colonised. We have shown that increasing numbers of HSC/RUs in the liver is accompanied by the appearance of growing numbers of HSC/RUs in the embryonic blood. The data presented here suggests that in addition to early waves of colonisation with committed and multipotent haematopoietic progenitors(Moore and Metcalf, 1970;Johnson and Barker, 1985;Dzierzak and Medvinsky, 1995)the liver is colonised by two consecutive waves of definitive HSC/RUs. The initial wave of HSC/RUs arrives from the AGM region on day 10 p.c., reaches a maximum by day 11 p.c. and disappears by day 13 p.c. On day 12 p.c. when AGM activity is decreasing, the second wave of colonisation arrives from the YS. This wave marks the embryonic stage when early YS cells mature into definitive HSC/RUs.

A visual demonstration of development of HSC/RUs in 10-12 d.p.c. mouse embryo as it is viewed by the authors is presented as a movie and accompanies the Web version of the article(Supplementary Information).

We are grateful to Prof. John Bishop (Edinburgh) and Prof. Josef Chertkov(Moscow) for helpful discussions; Dr Tony Hunter (Edinburgh) for help with statistical analysis. We thank Dave Kwant (Edinburgh) for preparing the movie. We thank Noemi Cambray for technical help and the staff of the animal house,John Verth, John Tweedie, Carol Manson for taking care of our experimental animals. This work was supported by the Leukaemia Research Fund grant 9656 to A. M. and J. A. P. K. was a fellow of the International Journal of Experimental Pathology. A. M. is an MRC Senior Fellow.