In a recent paper in Development, Jin et al. (Jin et al., 2009) reported that, in zebrafish embryos, the definitive haematopoietic cells of the trunk and tail have different differentiation outputs; that is, whereas cells in the tail can give rise to erythrocytes, those remaining in the trunk cannot. This finding that definitive erythropoiesis is absent in the trunk depends entirely on the absence of a whole-mount in situ hybridisation (WISH) signal for a single erythroid marker (αe1-globin) in these cells. They then describe a series of elegant experiments to examine and to attempt to explain this differing fate. If correct, these findings would be interesting, although surprising.
The finding that no definitive erythropoiesis occurs in the trunk is puzzling, because it contradicts our previously published results (Zhang and Rodaway, 2007). In this paper, we clearly showed expression of the erythroid markers gata1 and βe1-globin, as well as histochemical staining for haemoglobin, in haematopoietic clusters in the trunk of embryos at 4 days post-fertilisation (dpf). In light of the Jin et al. (Jin et al., 2009) study, we repeated our experiments and confirmed that βe1-globin is expressed in both trunk and tail haematopoietic cells (see Fig. S1A-C in the supplementary material), implying that both regions are erythropoietic. Detection of βe1-globin expression in the trunk required a longer incubation in chromogenic substrate than that in the tail, but was clearly specific. It is well known in the field that it is more difficult to detect gene expression in the trunk of older zebrafish embryos than in other tissues.
To avoid reliance on the expression of a single globin mRNA, we re-examined the emergence of definitive erythropoiesis by histochemistry for haemoglobin (see Fig. S1D-P in the supplementary material). At 2 dpf, staining is strong in circulating primitive erythrocytes and clear, but weaker, staining is seen in cells lying ventral to the dorsal aorta (DA) in the tail (see Fig. S1E in the supplementary material). These might be erythroblasts derived from the transient tail erythro-myeloid progenitors, as described by Bertrand et al. (Bertrand et al., 2007). In the trunk at this stage, no staining above background can be detected in the mesenchyme between the DA and the posterior cardinal vein (PCV). By 3 dpf, however, scattered haemoglobin-expressing cells are located in the mesenchyme between the major vessels in both trunk and tail. By 4 dpf, this expression has coalesced into clusters in the trunk (typically adjacent to somite boundaries) and into a more continuous strip of cells in the tail. In both trunk and tail, the erythroblasts are on the dorsal aspect of the major vein [PCV in trunk, caudal vein (CV) in tail; see Fig. S1J,K in the supplementary material]. This expression pattern is maintained at least as late as 7 dpf. At this age, erythropoiesis is also clearly detectable in the pronephros. Both trunk and tail haematopoietic tissues are, therefore, erythropoietic.
The presence of histochemically detectable haemoglobin in the trunk, along with the expression of βe1-globin mRNA, implied that at least one α-globin was likely to be expressed in this region. Although Jin et al. (Jin et al., 2009) reported that αe1-globin was not expressed in the trunk, we re-tested the expression of this gene by WISH. We detected expression in mesenchyme overlying the cardinal vein both in the tail and (after longer staining) in the trunk (see Fig. S1Q-T in the supplementary material). There are two possible reasons for the discrepancy between our data regarding αe1-globin expression and that of Jin et al. (Jin et al., 2009). Firstly, WISH detection of gene expression in the trunk of older embryos requires careful optimisation of proteinase K treatment. However, the fact that Jin et al. (Jin et al., 2009) were able to detect other mRNAs [for example, L-plastin (lcp1 – Zebrafish Information Network) and cmyb] in this region makes this explanation less likely. The second possibility is that, when probing for αe1-globin, Jin et al. (Jin et al., 2009) stopped the chromogenic reaction once staining had appeared in the tail, and did not continue the reaction until trunk expression was detected.
The fact that it is difficult to detect mRNA in the trunk vessels and the haematopoietic clusters of older embryos by WISH is well known in the field. The difference in WISH sensitivity is likely to result from the fact that most of the posterior blood island (PBI) is covered only by a thin layer of epidermis, whereas the trunk clusters are surrounded by other tissues that might impede access of fixative and that, once fixed, can present a barrier to probe and/or antibody penetration. Our findings also show that although globin mRNA is much harder to detect by WISH in the trunk than in the tail, the amount of haemoglobin (assayed by rate of staining by peroxidase histochemistry) is similar in the two sites. Therefore, we believe that the difficulty of detecting mRNA in trunk haematopoietic clusters in older embryos is a limitation of the WISH technique, rather than a result of genuinely lower expression levels.
Finally, we tested whether the caudal haematopoietic tissue (CHT; also known as the posterior blood island, PBI) is necessary for definitive erythropoiesis. Primitive and definitive erythrocytes can be distinguished by their morphology: primitive erythrocytes are disc shaped, whereas definitive erythrocytes are rugby ball shaped (Belair et al., 2001). Primitive cells enter circulation early in the second day of development and form the majority of circulating erythrocytes as late as 5 dpf. Thereafter, they are lost from circulation (Weinstein et al., 1996) and by 7 dpf almost all the circulating erythrocytes have definitive morphology (Belair et al., 2001). In order to determine the requirement for the CHT/PBI, we surgically removed the tails of embryos either before (1 dpf, 30 embryos) or after (3.5 dpf, 30 embryos) the colonisation of the CHT/PBI by trunk-derived cells described previously by Jin et al. (Jin et al., 2007). We examined these embryos at 7 dpf and, although we cannot completely exclude an effect on total number, there were ample circulating erythrocytes. Significantly, the proportion of erythrocytes with definitive morphology (∼95%) was indistinguishable from unoperated controls (see Fig. S1U-Z and Appendix S1 in the supplementary material; 5 embryos per treatment). Thus, although erythropoiesis clearly does occur in the CHT/PBI, it is not the sole site of definitive erythropoiesis, nor is it indispensable.
In summary, our findings show that haematopoietic cells in the trunk of zebrafish embryos: (1) express both αe1- and βe1-globin mRNA; (2) contain histochemically detectable haemoglobin; and (3) can give rise to definitive erythrocytes in the absence of the PBI/CHT. Thus, in contrast to the findings of Jin et al. (Jin et al., 2009), definitive erythropoiesis does occur in the trunk of zebrafish embryos.