Nature Editor's Overview

A mouse embryo (day 8.5) with labeled cells from the yolk sac in its blood, after the labeled blood cells enter the embryo proper. Credit: Igor Samakhalov

For the better part of a century, hematologists have wondered where blood comes from. Understanding the origin of blood has implications not only for blood diseases but also for growing blood stem cells (otherwise known as hematopoetic stem cells or HSCs) in culture, a difficulty that stymies therapies and currently requires HSCs to be collected through painful bone marrow extractions.

Scientists have long known that the yolk sac supplies HSCs to the embryo, and many once assumed this to be the source of adult HSCs. This view was supported by experiments showing that yolk sac cells could form mature blood cells both in culture and when injected into the spleens of irradiated mice. Others thought adult HSCs originated from a region of the fetus called the aorta–gonad–mesonephros (AGM). Yolk sac blood, they reasoned, was just for embryos and fetuses, which have unique forms of hemoglobin. That camp won, and their view has stood in textbooks for the past 30 years. However, in the April 26 issue of Nature, Igor Samokhvalov at the RIKEN Center for Developmental Biology and colleagues reverse current orthodoxy12.

Using an elegant lineage tracing technique, the researchers showed that stem cells from the yolk sac contribute to blood-forming cells in adult mice. The team labeled cells in week-old mouse embryos that expressed a gene called Runx1, which is predominantly expressed in yolk sac at that stage (but see the Experts' Corner). At this point, the yolk sac and embryo are distinct and do not share vasculature (indeed, blood vessels do not yet exist). Therefore, the team reasoned, labeled cells would likely derive from the yolk sac.

Surprisingly, they found labeled cells in the blood of adult mice, and the cells persisted for 15 months, the course of the study. That means at least some of the adult blood supply originates from the yolk sac. Further studies will be needed to determine just how much the yolk sac contributes to the adult blood supply, and whether the AGM provides a separate supply, or whether the yolk sac progenitors pass through the AGM. Either way, textbook accounts are due for a rewrite.

Read below to see a panel of experts' comments (in black) on this article, and responses from the authors (in italics). Table and reference numbers refer to those in the research article. Note that we have not included comments from all the reviewers.

Care to comment? Blog here on The Niche

The Experts' Corner: Experts in developmental hematology weigh in, followed by author's responses in italics.

Expert #1

The spatial and temporal origin of cells in the embryo that will contribute to hematopoiesis in the adult is a fundamental question of developmental hematopoiesis. This paper has two significant assertions: 1) that the precirculation yolk sac contributes to adult hematopoiesis, and 2) all hematopoietic stem cells are specified in the mouse embryo by E9.5. The experimental approach taken by the authors is a lineage marking system that is based on runx1 expression. This approach has the advantage of allowing the marked cells to follow their endogenous spatial and temporal development in situ within the embryo. The authors appropriately ensure that only the downstream promoter of Runx1 is used to mark cells. However, the experimental approach also has significant weaknesses.

It is known that changes in Runx1 dosage may well alter both the location and timing of hematopoietic stem cell specification in the mouse embryo.

The second difficulty with this marking approach is that it is based on a time of gestation and not a developmental stage. Variability between litters and within litters will add a range of starting stages to this experiment. As described by Downs (1993), there is a significant spread in developmental stages in albino, outbred mice at 7.5-8.0. The ranges in stages that were observed in litters particularly at 7.5 need to be provided and compared with the percentage of animals in comparably timed pregnancies that contained marked hematopoietic cells as adults. I think a statistical argument can then be made comparing the range of stages marked at E7.5 and the percentage of adult mice with contributions of marked cells in the peripheral blood.

Authors' Response

We observed the same degree of stage variation in our mixed (ICR×C57BL/6) background embryos as was reported for outbred PO mice (Downs K.M. and Davies T. 1993). Even though we did not perform systematic analysis in this respect, the degree of variations can be seen in Figure 3 and especially in Figure 3f. According to Downs and Davies report, at gestation period E7.5-E7.75 there is a certain slowdown of the embryo growth and development, the stage spread looks quite similar and provides a time window for less variable starting points. In principle, first quasi-intraembryonic Runx1 expression starts in omphalomesenteric artery at the embryo turning stage, which is about 24 hours after our standard E7.5 injection point. It is clear that our labeling stops within 24 hours postinjection (Fig. 2). Moreover the detailed labeling kinetics puts this interval well within 12 hours, and even 6 hours after tamoxifen injection, so we assumed that this time difference accommodates well the embryo stage variations in order for E7.5 injections to label only early Runx1+ cells, i. e. proximal yolk sac cells. Most importantly, it was directly confirmed by the labeling pattern analysis of the activated embryos (Fig. 3). No omphalomesenteric artery cell labeling was observed 12 hours after injection (not to mention other embryonic sites) despite noticeable embryo stage variation (Fig. 3f). At this time point the tamoxifen stimulation fails to label even distal Runx1+ yolk sac cells, which indicates that no more labeling occurs in the developing concepti.

Suggested statistical analysis of yolk sac's contribution to adult hematopoiesis is a relevant idea to pinpoint the start of pre-HSC generation. We did perform injections at 6 hours before standard time for E7.5 injections, which are effectively E7.25 injections, and now we include the data in the Figure 6c of the revised manuscript. Later E7.5-E8.5 injections would show only gradual statistical increase of adult blood contribution (Figure 6). Would it be an indication that the cell labeling happens at intraembryonic sites of Runx1 expression in “advanced” embryos? Probably not so, because in addition to quite modest induction of intraembryonic expression, there is a strong expansion of Runx1+ cells in post-primitive yolk sac clusters at around E8.5. We could not do the comparative expression studies, but the larger number of the strongly expressing yolk sac cell may explain the higher labeling of HSC lineage after E8.5 activation. This uncertainty underscores the limitations of our labeling system, but at least for the E7.5 activation points our tracing system provides exclusive labeling of targeted Runx1+ yolk sac population.

Expert

This inability to determine the exact stage of the embryo being marked makes it uncertain if other sites in the embryo, particularly the endothelium of the aorta, omphalomesenteric artery and umbilical vessels are also being marked. Some of these sites express Runx1 as early as day 8.5 (as stated by the authors and shown in North, et al.), but the expression pattern of Runx1 between day 7.5 and 8.5 has not been carefully delineated in either normal mice or in Runx1-lacZ heterozygotes. When does expression of Runx1 initiate in these various vascular beds? This analysis would also require examination of embryo sections to identify endothelial expression that may not be visible by examination of intact whole mount embryos.

Authors' Response

The concern about possible marking of intraembryonic cells at E7.5 injections is well grounded and it had prompted us before conducting any cell tracing studies to 1) carefully examine the Runx1 expression pattern between early neural plate stages and the start of Runx1 liver primordium expression (E7.5-E9.5). 2) analyze in detail the E7.5-marking process itself to assess the anatomical leakiness of our system. For both the Runx1 expression and ROSA26 marking analysis we have chosen whole-mount LacZ staining since even at the E8.0-E8.5 stages the embryo is essentially flat, transparent and its non-head region consists of 2×104 cells only (J. Palis et al., Development 1999). The whole mount staining gives an integral picture of the expression/marking and it allows scoring even a single LacZ+ cell in embryo (some 4-hour E7.5-activated embryos had few single faintly labeled cells in the proximal yolk sac region (Fig. 3a) and they were scored as LacZ+). We concluded that first non-circulatory Runx1+ cells appear in omphalomesenteric artery at around E8.5 confirming the data of T. North et al., (1999), followed by umbilical vessel expression. First bona fide (non-umbilical/vitelline vessels) embryonic Runx1 expression starts at around 20 somite pairs in the prospective AGM region. The E7.5 marking studies confirmed absolute lack of intraembryonic cell labeling at 12 hour after injections when some embryos reached the late head fold stages. It proved that our E7.5 injection regiment labels only yolk sac – derived cells. Nevertheless, we added modifications and revisions in our text to address the concern in more details.

Expert

Marking at E7.5 results in significant contribution to umbilical vessels at E11.5 (as shown in Fig. 2), (“interestingly, endothelium of the umbilical cord was labeled in the majority of cases (n=16 out of 25, Fig. 5)”). Interestingly Figure 5a shows spindle shaped, isolated mesenchyme cells near the umbilical vessels that are marked by stain. Are these cells, along with the extensive staining of umbilical endothelial cells, assumed to be derived from yolk sac blood islands or are they the result of marking of endogenous cells within these vessels?

Authors' Response

In relation to the remarks on Figure 5a, the origin of the cells mentioned by the reviewer is not clear. The section was performed through so called physiological embryonic hernia containing some small vessels and it is possible that come of Runx1+ cell outside of umbilical vessels might be the labeled primitive blood cells. Another alternative is that some cells (particularly those in upper left side of the picture) might be the intruding mesenchymal cells though it needs some further investigation. As mentioned in above paragraph the cells originate in proximal yolk sac. We modified the Figure legend according to this remark.

Expert

The authors put significant efforts into determining when cre recombinase is active after the injection of the dams at E7.5 with tamoxafin. Expression studies are consistent with tagging at 6 hours. However, the results shown in figure 3e (that the number of embryos in a litter that are tagged doubles between 6 and 12 hours and continues to rise even afterwards) are at odds with labeling being complete earlier as claimed.

Authors' Response

We agree that our statements in this regard need further clarification and revision. Recombination-competent Runx1 expression starts in the proximal yolk sac at around neural plate/early bud stage. Consequently, younger embryos in a litter may not express yet sufficient levels of Runx1 at the moment of E7.5 injection and the recombination in these embryos is postponed until the proper induction of the gene expression. This explains only 70% (in good agreement with Downs and Davies' embryo stage variation data) genetic recombination labeling at 6 hours postinjection. Indeed, we observed that all non-recombined after 6 hours embryos were lagging in the development. It is difficult to say when tamoxifen action (binding to cytoplasmic MER-Cre-MER) happens in the lagging embryos (it may well be within 6 hours postinjection provided that enough recombinase is synthesized) since some time is required for activated MER-Cre-MER to enter the nucleus and perform the recombination in ROSA26 locus. To our mind a good indication of short labeling interval (less than 6 hours) is that there was no distal expansion of cell labeling within Runx1 expression domain after 6 hours. Sharp-cut edges of labeled cellular rings argue against “lagging” LacZ synthesis in more distal yolk sac cells. We included in the revised text the above clarification.

Expert

Finally, the placenta has recently been shown as a site of HSC expansion (and possibly emergence). Was this organ examined for the presence of marked cells in any of these experiments?

Authors' Response

Accumulation of HSCs in placenta starting from E11.0 is the most interesting discovery in developmental hematology in recent years. However, we did not study the contribution of labeled cells to placenta since the hematopoietic role of placenta remains to be determined and any our published data may prove misleading in this respect. Main question is whether adult-type HSCs emerge in the placental vascular beds and/or mesenchyme, or placenta is a capacious niche for accumulating HSCs generated elsewhere. Therefore, lack of the labeled cells contribution to placenta may give unfounded support for the first possibility, and the opposite situation could be misleadingly interpreted as an argument against it. In our proposed model, adult-type HSCs do not emerge in yolk sac, rather the yolk sac pre-stem cells move to embryonic sites in order to mature. The AGM region and U/V vessels were shown as the sites of adult-type HSC maturation, but for placenta we simply need more data. This was the reason that we felt it is a bit premature to involve placenta in our studies.

Expert #2

This is an outstanding paper that utilizes a novel approach to addressing a question that has been of interest to developmental biologists, hematologists, and stem cell transplant clinicians for the past century. While it is recognized that hematopoietic stem cells (HSC) migrate to specific niches at specific times during murine development, the origin of the HSC has been elusive. In this paper, the authors use a an inducible lineage mapping study to trace the commitment of yolk sac derived cells into the hematopoietic lineage and observe that yolk sac derived cells contribute to lymphoid and myeloid lineages in the bone marrow of adult mice.

The paper is clearly written with excellent use of images that do not appear to be overinterpreted. However, I do have some questions that will require additional work to address.

It would be important, given the limited time between the administration of the tamoxifen and the known onset of the systemic circulation, to provide more information on the potential for the drug to be present in the embryo at a time when hematopoiesis emerges in the PSp region. Are there more detailed studies published that address this point (there is a statement to this effect but no factual data presented)? If not, the authors should determine these kinetics in embryos.

Authors' Response

This remark is based on the assumption that the hematopoiesis starts in paraaortic splanchnopleura (PSp) region at around the start of somitogenesis (based on studies of Ana Cumano and colleagues). However, the timing is hardly accurate since the hematopoietic progenitors were shown to emerge only after lengthy (in mouse development terms) tissue explant culture and it reflects merely the potential of cells from the caudal part of early embryo in the particular culture conditions. During the ex vivo culture the cell developmental fate can be seriously skewed. Therefore it is difficult to say whether hematopoiesis emerges in PSp at this early embryonic stage. The retrospective potential studies in developmental hematology may be misleading, for example, the definitive progenitors in yolk sac are detected beginning at E8.25, whereas first mature definitive cells are seen in circulation only after E11.5, and therefore one can not say that E8.25 yolk sac is the site of emerging definitive hematopoiesis. Also it is conceivable that other caudal elements like primordial germ cells may be responsible for apparent hematopoietic differentiation due to the ease of EG cells derivation at around E8.0. The 12 hour kinetics of cell labeling at E7.5 in our system showed no staining in PSp region, therefore we concluded that even if at E8.0 there are some cells in PSp with ex vivo hematopoietic potential, they are not expressing Runx1 at the level which is enough for the labeling and, consequently, we are not reading their contribution into adult hematopoiesis.

Expert

One key point of the paper is that the yolk sac cells marked at E7.5 contribute to adult marrow hematopoiesis. To be clear that the marked cells in the marrow cavity are indeed stem cells, the authors need to use standard transplantation techniques to prove this point (a HSC will give rise to both lymphoid and myeloid lineages for more than 4 months following transplantation into a conditioned recipient animal). Assuming that the animals are congenic, this should be straightforward, and if not, one may have to resort to the use of NOD/SCID mice.

Authors' Response

The suggestion is warranted from widely accepted point of view. The transplantation/ repopulation assays are extremely important from therapeutic standpoint. In fundamental biology the idea of the assays boils down to the simplicity of the cell labeling in a donor-recipient system, which even allows quantitative long-term engraftment studies. Essentially the repopulation assay is a deterministic long-term cell tracing in an artificial donor-recipient system. It is applied due to extreme technical difficulties associated with specific in vivo labeling of a single cell clone or even a cell population in question. Consequently, if we are able to pulse-label specifically a distinct cell population in vivo, and trace it long-term, we would not need to do cell transplantation into conditioned recipients. In our system, a very small group of labeled proximal yolk sac cells (a great majority of E7.5-labeled cells are primitive blood) survive ontogenesis and populate native adult sites of hematopoiesis for at least 1 year, giving rise to all major hematopoietic cell lines while competing with an excess of unlabeled progenitors. This conforms with the strictest definition of stem cells. Indeed, there were no differences between the adult hematopoiesis contribution of in vivo genetically labeled cells and corresponding transplanted cells (Göthert, J.R. et al., Blood (2005), 105:2724). Moreover, we showed a substantial contribution of E7.5-labeled cells to CD34− LSK bone marrow cells which are regarded as practically pure HSC population.

A repopulation assay compared to a long-term genetic cell tracing is significantly less strict: in our system the cells compete with the 10-20-fold excess of the unlabeled progenitors not just for BM niches but also for several embryonic, fetal and neonatal hematopoietic niches. In addition, in a repopulation assay we study only the potential of any given cell population for efficient self-renewal and differentiation which is assessed retrospectively by the presence of lingering labeled cells in recipient's blood. Conditioning of the recipient for severe compromising of endogenous hematopoietic system may drive even non-hematopoietic cells to repopulate. Even though the last phenomenon is pretty limited, the extent of overlapping between the populations of cells with long-term repopulation potential and the actual HSCs remains to be determined.

Expert

The main point that must be addressed in some detail in the discussion, is the fact that it has been reported by several groups that the heterozygous Runx1 mouse displays some abnormalities in HSC emergence (HSC appear earlier in the PSp and yolk sac than in wild-type mice). Since the marking strategy used in the present study causes the test animals to be heterozygous for Runx1 expression, the major question most readers will raise is whether the results obtained herein are relevant to the normal ontogeny of hematopoiesis (the authors state this point but do not discuss it).

Authors' Response

The main question of our study was whether yolks sac cells can give rise to adult HSC lineage. If this is the case it would manifest the existence of a special complex genetic program which controls the following chain of events: commitment and mobilization of a yolk sac mesodermal cell, support of blood-borne phenotype of the cell, recognizing distinct vascular beds in embryo, attachment and integration into endothelial layer or chemotaxis into subjacent mesenchyme, and initiation of maturation. What conditions are permissive for the program is another question. If it works only at lower dose of Runx1 the question is why the program is fixed in evolution. Mouse strains are very different in the number of HSCs and hematopoietic progenitors (Muller-Sieburg and Riblet, J Exp Med 1996, 183:1141). It is conceivable that the levels of Runx1 are also different and some strains develop apparently normal adult hematopoiesis at lower levels of the protein using the developmental pathway we uncovered in our studies. The question of relative contribution of different hemogenic sites in conceptus to adult hematopoiesis remains open and needs further addressing.

On the other hand, the embryonic effects observed in Runx1-heterozygotes may not be necessarily due to the gene haploinsufficiency, but rather because of known DNA hitchhiking effect. In general, haploinsufficiency effects found in knockout models should be taken with caution due to persistence of a substantial segment of differential DNA surrounding targeted locus even after multiple backcrosses. There are three genes of interferon receptors located within 1.1cM from the Runx1 locus (MGI 3.43, Mouse Chromosome 16 Linkage Map, 62.20 - 64.00 cM; http://www.informatics.jax.org/). The heterozygotes selected for the mutated Runx1 allele would carry strain 129 interferon receptor alleles on BL/6 background, whereas “wild type” homozygotes do not. The interferon receptors genes may be included in the gene repertoire controlling the HSC frequencies, because interferons strongly affect hematopoietic progenitors' proliferation. Since all HSCs arise during embryogenesis (Göethert, J. R. et al. Blood 2005, 105, 2724), their frequency is already variable early in development depending on specific genetic makeup, which in turn may shift the developmental pattern for hematopoietic progenitors in congenic rather than coisogenic Runx1 heterozygotes without inactivation of a Runx1 allele. It is of note in this regard, that recently IL-3 was shown to rescue unexpectedly the Runx1 “haploinsufficiency” (E. Dzierzak's Lab, Dev. Cell (2006) 11, 171), whereas interferon alpha is know to be an antagonist of IL-3 in hematopoietic progenitors (Jaster, R. publications).

In the revised manuscript we added the discussion about the limits of our conclusions and relevancy of our data to normal ontogeny.