NACA deficiency reveals the crucial role of somite-derived stromal cells in haematopoietic niche formation

The ontogeny of haematopoietic niches in vertebrates is essentially unknown. Here we show that the stromal cells of the caudal haematopoietic tissue (CHT), the first niche where definitive haematopoietic stem/progenitor cells (HSPCs) home in zebrafish development, derive from the caudal somites through an epithelial–mesenchymal transition (EMT). The resulting stromal cell progenitors accompany the formation of the caudal vein sinusoids, the other main component of the CHT niche, and mature into reticular cells lining and interconnecting sinusoids. We characterize a zebrafish mutant defective in definitive haematopoiesis due to a deficiency in the nascent polypeptide-associated complex alpha subunit (NACA). We demonstrate that the defect resides not in HSPCs but in the CHT niche. NACA-deficient stromal cell progenitors initially develop normally together with the sinusoids, and HSPCs home to the resulting niche, but stromal cell maturation is compromised, leading to a niche that is unable to support HSPC maintenance, expansion and differentiation. How do stromal cells contribute to the haematopoietic stem cell (HSC) niche? In zebrafish, Murayama et al. identify somite-derived stromal cells in the first niche where HSCs home, and depletion of the nascent polypeptide-associated complex alpha subunit in these cells prevents HSC maturation.

I n vertebrate ontogeny, emerging definitive haematopoietic stem/progenitor cells (HSPCs) migrate and colonize successive specialized haematopoietic niches, which in turn regulate proliferation and differentiation of HSPCs, and release of their progeny into the blood. In the most studied haematopoietic niche, the bone marrow of mammals, the cellular constituents so far identified as important for haematopoietic niche function have been endothelial cells of venous sinusoids and arterioles, osteoblasts and associated cells at the endosteal surface, and various stromal cell subsets often closely associated with the blood vessels. Among these are CXCL12-abundant reticular cells 1 , Nestin þ perivascular cells 2 , and within the latter subset, Leptin receptor þ cells with reticular morphology around venous sinusoids 3 , and NG2 þ cells with pericyte morphology around arterioles 4 . The relationship and degree of overlap between these various stromal cell subsets is still to be understood, as well as their relative localization and interaction with HSPCs and with other stromal elements in vivo.
In zebrafish development, as in mammals, definitive HSPCs initially emerge by an endothelial-to-haematopoietic transition from the ventral wall ('floor') of the dorsal aorta (VDA) 5 , and then migrate through the blood to colonize the first niche where they will expand and differentiate into multiple blood lineages. In fish, this initial niche is the caudal haematopoietic tissue (CHT 6,7 ), while in mammals it is the fetal liver. The fetal liver haematopoietic niche is still poorly characterized 8 . We previously reported that the CHT of zebrafish is organized around a transient network of venous sinusoids, the caudal vein (CV) plexus, and that its ultrastructural organization is quite similar to the peri-sinusoidal haematopoietic microenvironment in the haematopoietic niche of adults-the bone marrow in mammals, or the kidney marrow in fish-with typical fibroblastic reticular cells (FRCs) in close contact with the abluminal side of venous sinusoids and extending a network between the latter, within which haematopoietic cells expand and differentiate 6 . In the CHT, these FRCs appear to be the main stromal cell type making up the niche together with the endothelium of the CV plexus/sinusoids. Thanks to the transparency of the zebrafish larva, the CHT is uniquely accessible to non invasive in vivo observation, offering a unique opportunity to explore in situ the development of this haematopoietic niche at a cellular level.
In the present study, we show that the stromal FRCs of the CHT niche derive from the ventral border of the caudal somites, through an epithelial-mesenchymal transition (EMT). We find that a deficiency in the a-subunit of the nascent polypeptideassociated complex (NACA, aNAC) prevents the maturation of the EMT-derived mesenchymal progenitors of the FRCs, and that this results in a CHT niche that can still trap circulating HSPCs, but is no longer able to retain them and support their expansion and differentiation. The NACA-deficient mutant thus highlights the essential role of these stromal FRCs in the making of a functional haematopoietic niche.

Results
The oloca mutant is defective in definitive haematopoiesis. We identified the zebrafish mutant oloca as a recessive mutant with fewer neutrophils by 3.5-4 days post fertilization (d.p.f.), using Sudan Black (SB) staining in a F3 mutant screen following ENU mutagenesis. SB staining of the oloca mutant showed fewer neutrophils in the head (Fig. 1b), and no neutrophils in the CHT (Fig. 1d), which normally produces numerous neutrophils by this stage 9 (Fig. 1a,c). The few SB þ cells found in the oloca tail were not localized within the CHT but at the body surface (Fig. 1d,  arrows). Oloca mutants develop apparently normally up to 2.5-3 d.p.f. Morphological defects become obvious after 3.5 d.p.f.: a small head with immature lower jaw, cardiac oedema, disorganized iridophores in eyes, reduced circulation from 90 h post fertilization (h.p.f.) onward, and finally the larvae die by B4.5 d.p.f.
As the SB phenotype suggested a lack of definitive granulopoiesis in the CHT of oloca larvae, we analysed the status of definitive haematopoiesis in the CHT by whole mount in situ hybridization (WISH). At 72 h.p.f., much fewer cmyb þ HSPCs were found in the oloca CHT compared with siblings (Fig. 1e,f), and none at all from 90 h.p.f. onwards (Fig. 1g,h). Accordingly, by 90 h.p.f., granulopoiesis, probed by lysC, cebp1 and nephrosin expression (Fig. 1i-n), erythropoiesis, probed by e1-globin expression (Fig. 1q,r), and lymphopoiesis probed by ikaros expression (Fig. 1s,t) were all strongly reduced. The final definitive haematopoietic tissues, the thymus and kidney, which normally become seeded by HSPCs from the CHT by 3 and 5 d.p.f. respectively 6 , similarly lacked cmyb expression in oloca embryos (Fig. 1u,v), and the lymphoid markers ikaros and lck were missing in the mutant thymus ( Fig. 1w-z). In contrast with the other haematopoietic lineages, csf1ra expression, which marks the macrophage lineage, was similar to siblings or even slightly increased in the CHT of oloca mutants (Fig. 1o,p). This is consistent with our observations that primitive macrophages, born in the yolk sac, colonize the nascent CHT through blood circulation before definitive haematopoiesis takes place there 6 .
Oloca is a mutant in the naca gene. To identify the mutation responsible for the oloca phenotype, we used simple sequence length polymorphism mapping. The mutation was mapped between flanking markers z7590 and w27205 (Fig. 2a) on chromosome 23. This interval is B450 kb and contains around 20 genes. One of them, naca, had been previously identified in an insertional mutagenesis screen as no knack 10 . no knack is an essential gene and the terminal phenotype was similar to oloca. In a cross with oloca, no knack failed to complement, strongly suggesting that oloca was a naca mutant. Sequencing of the zebrafish naca gene, in mutant and sibling embryos, revealed a premature stop codon in the mutant allele due to a C-T transition, resulting in a Q4stop alteration at codon 69 (Fig. 2b,c). The naca gene encodes the a subunit of the heterodimeric nascent polypeptide-associated complex (aNAC/ NACA), which binds to nascent polypeptides on the ribosome 11 . NACA appears to be part of the chaperon system that ensures the proper folding of nascent proteins, notably those to be translocated in the endoplasmic reticulum, ER 12,13 . The naca gene also encodes a muscle-specific, much larger isoform, skNAC, through differential RNA splicing 14,15 . NACA and skNAC have also been described as transcriptional co-activators in osteoblasts and myotubes, respectively 14,16,17 .
To provide independent evidence that the naca mutation causes the oloca phenotype, we tested whether morpholinos (MOs) targeted against naca RNA would phenocopy the oloca mutant, and whether wild-type naca RNA injection would rescue the oloca phenotype. Two naca splice-blocking MOs, MO1 and MO2, phenocopied the oloca mutant. In the CHT of naca MO1injected embryos, very few cmyb-positive HSPCs and almost no SB stained neutrophil were observed at 3.5 d.p.f. (Fig. 2d-g). Co-injection of 125 or 250 pg of naca wild-type RNA with 4 ng of naca MO1 resulted in rescue of cmyb þ HSPCs and neutrophil differentiation (SB staining) in the naca morphants (Fig. 2h,j). In contrast, co-injection of the same amounts of naca mutant RNA with naca MO1 did not rescue the morphant phenotype to any extent (Fig. 2i,k). WISH analysis through development revealed ubiquitous expression of naca mRNA at 25-30 h.p.f., with a stronger signal in the ventral half of somites (Fig. 2l, asterisks) and in mesenchymal cells ventral to the somites in the caudal area (Fig. 2n, arrowheads). By 3 and 5 d.p.f. (by which stage haematopoiesis peaks in the CHT), naca has become specifically expressed in the CHT (Fig. 2p,q, arrows), a trend that appears to initiate by 42 h.p.f. (Fig. 2o, arrows). In oloca mutant embryos, WISH detected no naca transcripts, suggesting nonsense mediated mRNA decay (Fig. 2m).
HSPCs initially home to the mutant CHT. We used naca MO1 to assess the early effects of NACA deficiency on haematopoiesis, at stages at which mutant embryos cannot yet be distinguished morphologically. naca MO1-injected embryos showed normal primitive haematopoiesis, including granulopoiesis of yolk sac origin ( Supplementary Fig. 1, arrows), and erythrocyte circulation. By 36 h.p.f., cmyb-positive definitive HSPCs were found along the aorta ventral wall (VDA) and nascent CHT of naca morphant embryos as in wild type (Fig. 3a,b). To follow their fate in live embryos, we turned to transgenic Tg[CD41:gfp] embryos, as definitive HSPCs are CD41:GFP low from as soon as they emerge from the VDA 18 (from here onwards, gfp will refer to the transgene, and, for example, CD41:GFP to the fluorescence of the GFP protein expressed from the CD41:gfp transgene). Consistently, naca MO1-injected Tg[CD41:gfp] embryos showed normal emergence of CD41:GFP low HSPCs from the VDA at 54 h.p.f. (Fig. 3c,d,k) and their settling in the CHT (Fig. 3e,f,k). Thus, definitive HSPCs were normally specified and produced from the morphant DA floor (VDA), and initially colonized the CHT as in wild type.
In contrast, 30 h later, by 84 h.p.f., the morphant CHT contained much less CD41:GFP low HSPCs than control embryos (Fig. 3g,h,k). This decline in HSPC number in the morphant CHT was accompanied by a strong reduction in the dorso-ventral width of the CHT. In addition, whereas CD41:GFP low lymphoid progenitors normally start colonizing the thymus rudiment from 54 h.p.f. onwards 18 , the naca morphant thymus rudiment showed no such colonization even at 84 h.p.f. (Fig. 3i,j), consistent with our previous WISH data with the oloca mutant ( Fig. 1u-z). Thus NACA deficiency did not affect HSPC emergence from the aorta, nor their initial homing in the CHT, but their subsequent maintenance, expansion and differentiation in the CHT. The same outcomes were observed with oloca:CD41gfp mutant embryos (Fig. 3l).    (Fig. 4b,c). In sharp contrast, in the mutant recipient larvae, CMTPX þ cells were no longer observed in the CHT and very few transplanted cells were detected in the kidney area by this stage (Fig. 4b, HSPCs are not intrinsically compromised in the oloca mutant. We next addressed whether the haematopoietic defect in the oloca mutant was solely due to its deficient CHT niche, or whether it also involved a cell-autonomous defect of mutant HSPCs. To address this point, we took advantage of a recent method for generating parabiotic embryo pairs 19   control MO-injected lysC:DsRed embryos whereas only few cells of either origin were found in the naca morphant CHT (Fig. 4g,i,j). These results confirm that naca deficiency leads to a deficient CHT niche and show that, in contrast, the homing capacity and granulocytic differentiation ability of NACA-deficient definitive HSPCs is intact.
CHT stromal cells originate from the caudal somites. In our initial description of the CHT niche 6 , our electron microscopy data suggested it to consist in two main cell types, the endothelial cells of the CV plexus sinusoids, and mesenchymal cells in between the sinusoids that appeared to evolve into what we called FRCs, owing to their ultrastructural resemblance to FRCs in mammalian bone marrow and adult fish kidney marrow. . We thus combined the time-lapse confocal imaging of the TCF:nls-mCherry þ nuclei of these emigrating cells with vital staining with the interstitial fluorescent marker Bodipy ceramide, which has been used to delineate somite and cell contours within the somitic epithelium 21,22 ( Supplementary Fig. 3b). The four-dimensional tracking of individual nuclei of the emigrating cells confirmed that they were initially all intra-somitic, and located within the entire medio-lateral extent of the ventral border of the somites (Supplementary Fig. 3c-e and Supplementary Movie 4). We studied this emigration process in more detail by combining fluorescence with video-enhanced Differential Interference Contrast (VE-DIC) microscopy 6 . From 19 h.p.f., epithelial cells from the ventral part of the caudal somites underwent an EMT. They started to show active blebbing through the somite basal lamina ( Fig. 5a and Supplementary Movie 5), then more extended protrusions in ventral direction (Fig. 5b), and emigrated as strings of mesenchymal cells into the more ventral caudal space, with frequent mitoses before and during their migration ( Fig. 5c-e, Supplementary Movie 6). In that space, at the same time, the primordial CV started sprouting in ventral direction to give rise to the CV plexus by 30 h.p.f. To observe the interaction between the two processes, we crossed Tg[TCF:nls-mCherry] with Tg[fli1:gfp] fish. At 25 h.p.f., the angiogenic sprouting of endothelial cells from the primordial CV was very dynamic, and TCF:nls-mcherry þ cells were found interspersed with the endothelial sprouts (Supplementary Movie 3b), interacting with them and with each other through filopodial extensions (Fig. 5f,g). By 35 h.p.f., the CV plexus was complete and functional 6    cytoplasm and long filopodia by which they connected to each other and to endothelial cells (Fig. 5i,j). By 48 h.p.f., their nuclei had become smaller, with a central, well-defined nucleolus, and while they were still making a dynamic network of interconnecting filopodial extensions connecting them to each other and other cells ( Fig. 5k and Supplementary Movie 7), many of them now had their cell body in close apposition to the abluminal side of endothelial cells of the CV plexus (Fig. 5h,l). This configuration is identical to that of the FRCs that we previously described by electron microscopy as the main stromal cell type making up the CHT niche together with the endothelial cells of the CV plexus 6 . We further evidenced the somitic origin of these cells by transiently expressing GFP from a skeletal muscle a-actin promoter in  among the developing stromal cells both between and in close apposition to the CV plexus endothelia ( Supplementary  Fig. 4g-m) as previously observed (Fig. 5h,l). Finally, in vivo observation of Tg[sox10:gfp] which highlights neural crest cells (NCCs) 23 , revealed no NCC contribution to the CHT other than pigment cells (Supplementary Fig. 5a,c). Moreover, sox10 mutants lacking most NCC derivatives revealed no sox10:GFP þ cells in the CHT, yet no morphological abnormality of the CHT, and the presence of typical stromal cells as in the wild type ( Supplementary Fig. 5a-h), indicating that NCCs do not contribute to stromal cell development during CHT niche formation. We therefore conclude that the FRCs/stromal cells initially arise from the caudal somites through a typical EMT, and that their progenitors accompany the formation of the CV plexus, leading altogether to the final haematopoietic niche. From now on, we will designate these cells as stromal cells or stromal reticular cells (SRCs).
Naca is required for stromal cell maturation. We next compared the development of caudal endothelial (Fli1:GFP þ ) and stromal (TCF:nls-mCherry þ ) cells between control and naca morphant embryos. Time-lapse confocal analysis on [Fli1:gfp; TCF:nls-mCherry] double transgenic embryos indicated that the initial sprouting of Fli1:GFP þ CV endothelial cells and parallel emigration of TCF:nls-mCherry þ stromal cell progenitors from the somites occurred normally in morphant embryos compared with controls (Fig. 6a,b).
By 48 h.p.f., about two times less TCF:nls-mCherry þ stromal cells were detected in the naca morphant CHT relative to controls (Fig. 6c,d,g). However this did not necessarily reflect a true decrease in stromal cell number at this stage, because the mean mCherry fluorescence per detected cell was also diminished over 2-fold in both the somites and CHT stromal cells ( Supplementary  Fig. 6a-d) but not in the fin mesenchymal cells ( Supplementary  Fig. 6e) in morphants relative to controls. Close inspection by combined VE-DIC and fluorescence microscopy did confirm the presence, only in the morphants, of stromal cells visible by VE-DIC that had virtually no more detectable mCherry signal ( Supplementary Fig. 6b, arrowheads). By 60-72 h.p.f., both fluorescence and VE-DIC microscopy evidenced a virtually complete lack of stromal cells in the naca morphant (Fig. 6e,f,  Supplementary Fig. 6f,g).  ARTICLE In parallel, the CV plexus was slightly thinner in the naca morphant at 48 h.p.f. and displayed fewer and often larger intervascular spaces in sagittal optical section (Fig. 6c,d,h,i). Between 48 and 60 h.p.f., these spaces flattened and the difference in CV plexus thickness between control and morphant larvae kept increasing (Fig. 6e,f, Supplementary Movie 8), as observed above in the Tg[CD41:gfp] background (Fig. 3e-h). In addition, in the thinner naca morphant CHT, blood kept circulating in the various branches of the CV plexus, unlike in controls in which blood eventually flows only through the ventral-most branch, the definitive CV, as the more dorsal branches become packed with haematopoietic cells 6 . Both features thus are likely consequent to the lack of haematopoietic cells accumulation in the naca morphant CHT.
NACA is cell-autonomously required for stromal cell survival. Our data suggest naca deficiency affects stromal cells survival  during CHT niche maturation. To further investigate whether the requirement for naca in stromal cell development is cellautonomous, gastrula transplantation was performed to generate chimeric embryos, in which naca function is partially recovered or affected within the stromal cell population. At early gastrula stage (40% epiboly), cells located at the outer margin of the blastoderm (where mesoderm fates arise) of wild-type (control MO-injected) Tg[H2A-gfp] (ref. 24) embryos (which express nuclear GFP ubiquitously) were transplanted into naca MO-injected Tg[TCF:nls-mCherry] at the same stage and place ( Fig. 7a-d). Reciprocally, cells from naca MO-injected Tg[H2A-gfp] embryos were transplanted in the same way into control MO-injected Tg[TCF:nls-mCherry] embryos (Fig. 7e-h). Recipient embryos were individually followed up at 42 and 60 h.p.f. At 42 h.p.f., SRCs originating from the wild-type H2A-GFP þ donor could be observed in the naca morphant recipient in 7/120 successful transplants (Fig. 7a,b). These transplanted H2A-GFP þ stromal cells had slightly expanded by 60 h.p.f., while the naca-deficient endogenous TCF:nls-mCherry þ stromal cells had mostly disappeared (Fig. 7c,d,i). In the reciprocal experiment, cells transplanted from naca MO-injected H2A-GFP þ donors into wt recipient Tg[TCF:nls-mCherry] embryos also provided stromal cells associated with the caudal plexus at 42 h.p.f., in 6/120 successful transplants (Fig. 7e,f); however their number had greatly decreased by 60 h.p.f., unlike the TCF:nls-mCherry þ stromal cells from the wild-type recipient which had slightly expanded (Fig. 7g,h,j). These results demonstrate that naca is cellautonomously required for SRC survival.
NACA deficiency triggers the UPR and stromal cell apoptosis.
Previous reports have shown that NACA depletion can cause ER stress and ER stress-induced apoptosis through activation of the unfolded protein response (UPR) pathway in mammalian cultured cells 25 . We therefore analysed the expression of seven UPR-related genes in sibling and mutant larvae at 3.5 d.p.f. Four genes (atf4, bip, chop and xbp-1) coding for UPR effectors were significantly up-regulated in the mutant (Fig. 8a)  oloca mutant and sibling larvae (Fig. 8b). To clarify whether NACA-deficient cells may die by apoptosis, we performed whole mount TUNEL staining in oloca mutants in different transgenic backgrounds. By 3 d.p.f., TUNEL þ signals were observed in the CHT of mutant but not sibling larvae. Using tyramide-amplified immunodetection, we could detect double positive, TCF:nls-mCherry þ /TUNEL þ apoptotic bodies, evidencing apoptotic death of TCF:nls-mCherry þ mutant stromal cells (Fig. 8c-e). In contrast we found virtually no TUNEL þ apoptotic bodies among gata1:DsRed þ erythroid cells, mpx:GFP þ neutrophils, CD41:GFP low HSPCs and ikaros:GFP þ lymphoid progenitors in the mutant CHT ( Supplementary Fig. 7a-h).
We then sought to block the ER stress/UPR pathway during CHT niche formation to determine whether activation of the UPR pathway causes stromal cell death and consequent defects in niche formation. Chemical chaperones such as 4-phenylbutyric acid (4-PBA) enhance protein folding and thus act as a suppressor of ER stress. When we treated oloca sibling and mutant embryos with 4-PBA, we found first that the fold enhancement of UPR genes (atf4, atf6, bip and chop) expression in mutant relative to sibling embryos was significantly decreased ( Supplementary Fig. 7i); second, the number of TUNEL þ cells was also decreased in the CHT of oloca mutants (Fig. 8f). Third, 4-PBA treatment significantly rescued the number of HSPCs marked by cmyb in situ hybridization at 3.5 d.p.f. (Fig. 8g-j).
Finally, injection of mRNA encoding the anti-apoptotic protein Bcl-XL in oloca mutant embryos totally suppressed the occurrence of TUNEL þ cells in the mutant CHT (Fig. 8k). It only partly rescued the number of TCF:nls-mCherry þ stromal cells at 60 h.p.f. (Fig. 8l,n-p), possibly reflecting the faster extinction of the transgene in the mutant, and it fully rescued the number of CD41-GFP þ HSPCs at 3 d.p.f., up to the wild-type level (Fig. 8m,q-s).

Discussion
We previously described that in zebrafish development, the CHT-the initial niche where definitive HSPCs home, expand and differentiate-consists of two main cell types: the venous endothelial cells of the CV plexus, and a cytomorphologically homogeneous stromal cell type which we named the FRCs, given their similarity in cytology and spatial organization to the FRCs of adult fish kidney marrow and mammalian bone marrow 6 . The origin of these FRCs was so far unknown. Here we have discovered that they originate from the ventral border of the caudal somites, through a typical EMT. Having now better described the development of these cells in vivo, we prefer to designate them as SRCs, as this name represents their two salient characteristics. Like NCCs after their EMT from the neural tube, the SRCs emigrate from the caudal somites as strings of cells that often divide on the way. They will constitute an excellent, readily accessible model for future in vivo studies of EMT. Although very little is known about the corresponding haematopoietic niche in mammals, in the fetal liver, Chagraoui et al. 26 reported that in vitro haematopoiesis-supportive stromal cells isolated from fetal liver consisted of 'EMT cells', as defined by co-immunostaining with a series of epithelial and mesenchymal markers.
Recently, Nguyen et al. 27 demonstrated that trunk somites give rise to cells that incorporate in the DA and facilitate the hemogenic specification of other endothelial cells in the DA floor. Thus, our present work and this study together reveal an amazingly pervasive, and so far entirely unsuspected, cellular contribution of somite-derived cells in nurturing HSC development and then function, in first the specification of HSCs within the aorta 27 , and later their expansion and differentiation in the first niche where they home (this work).
In the CHT, the ventral-wards emigration of SRC progenitors from the somite border closely accompanies the formation of the CV plexus in the same space, such that by 32-35 h.p.f., the very onset of HSPC budding from the aorta floor, the CV plexus is complete, with blood circulating through all its branches, and SRC progenitors interconnecting the CV branches. The niche is thus ready just in time to trap the first circulating aorta-derived HSPCs. Over the next 24 h, the SRC progenitors reorganize and mature into SRCs partly lining the abluminal face of the venous sinusoids.
In mammals, neural crests contribute to bone marrow niche formation via nestin þ mesenchymal stem cells 28 . In contrast, we observed no contribution of NCCs to the CHT other than pigment cells, and found normal SRCs and CHT haematopoiesis in a sox10 null mutant which lacks most neural crest derivatives.
Here we could observe the mesenchymal SRC progenitors in vivo due to their labelling through a Wnt signalling reporter transgene. This does not necessarily imply that these cells keep receiving and transducing Wnt signals. Indeed, the TCF:GFP or TCF:nls-mCherry reporter is highly expressed in the somites and consequently in the SRC progenitors as they emerge from them, but then the fluorescent signal progressively decays in these cells, such that by 3 d.p.f., they are no longer detectable in vivo by this means, suggesting that the reporter protein was mostly inherited from the somitic origin.
The second main result of our study is that a deficiency in NACA drastically compromises the sustainability and differentiation of HSPCs in the CHT, due to a defect not in the HSPCs, but in the CHT niche, making the oloca mutant the first of this kind in the zebrafish literature so far. In the oloca mutant, HSPCs, whether autologous or transplanted from wild-type adult kidney marrow home to the CHT niche just as in a wt larva, but then are unable to stay, expand and differentiate there. We traced this defect of the CHT niche to a defective maturation of the SRCs. Initially, SRC progenitors emerge normally from the somites, and accompany the formation of a functional CV plexus, which traps circulating HSPCs as well as in wild type. But then, over the next 2 days, the SRC progenitors progressively disappear instead of maturing into SRCs, and the resulting CHT is unable to support HSPC maintenance. We found that NACA is required cell autonomously for the maturation and survival of SRCs.
Why does NACA deficiency cause the disappearance of SRC progenitors in particular? In mammals, NACA deficiency can cause ER stress and consequently the UPR, which, depending on the context, can trigger apoptosis 25,29 . We found that NACA deficiency does trigger the UPR in zebrafish larvae, and that treatment of naca mutants with the chemical chaperon 4-PBA significantly suppressed SRC apoptosis, and restored CHT haematopoiesis. Interestingly, EMT-derived mesenchymal cells in general were recently shown to be exquisitely sensitive to ER stress-induced apoptosis 30 . This feature was shown to be linked to their high level of extracellular matrix secretory activity, which involves steady ER development. Here in the CHT, the mesenchymal SRC progenitors are most likely responsible for secreting the large amount of extracellular matrix that surrounds the CV plexus by 48 h.p.f. 6 , which would render them more sensitive to the ER stress caused by NACA deficiency.
Finally, the transplantation of WKM cells from adult fish into 2 d.p.f. larvae revealed an unexpected feature. The WKM cells readily homed not only to the CHT, but also to the kidney area. This was surprising because the developmental colonization of kidney by larval HSPCs only starts by 4 d.p.f. 6,31 . This unexpected observation suggests that developmentally, the kidney niche is potentially able to trap circulating HSPCs already by 2 d.p.f., but that larval HSPCs are not yet able to home to it. Now, whereas wild-type WKM cells similarly homed to the wt and oloca mutant kidney short after transplantation, 2 days later these cells had disappeared from the mutant kidney area, while they had expanded in the wild type. This observation suggests that NACA deficiency may affect not only the CHT niche, but also the kidney marrow niche, in a similar way. The early (pronephric) kidney niche has not been much studied yet, but similar SRCs are found in the CHT and adult kidney marrow niches 6 . Therefore we suggest that NACA deficiency may affect CHT and kidney marrow SRCs/FRCs in the same way, with similar consequences. The same may well apply to mesenchymal cells of mammalian haematopoietic niches such as the bone marrow.

Methods
Zebrafish. Transgenic and mutant stocks of zebrafish were raised and staged according to Westerfield 32  Generation and identification of the oloca mutant. The oloca mutant (allele t30953) was identified in a large-scale N-ethyl-N-nitrosourea (ENU) mutagenesis screen (Tübingen 2005 Screen Consortium) by whole-mount Sudan Black staining of neutrophils 9 in 4 d.p.f. larvae. Meiotic mapping of oloca t30953 was performed using standard simple sequence length polymorphisms. The naca gene was cloned from cDNA made from mutant and wild-type siblings (primers nacaF1 and nacaR1). DNA was sequenced on both strands (primers nacaF2 and nacaR2). Genomic sequencing was performed by PCR amplification of naca exon 3 from mutant and wild-type larvae (primers nacaF3, nacaR3, nacaF4 and nacaR4). Primers used are listed in Supplementary Table 1. The naca wild-type and mutant cDNAs were cloned into pCS2 þ for mRNA injections (primers nacaF5 and nacaR5). Morpholino and mRNA injections. One to two-cell-stage embryos were microinjected with 4 ng of a splice blocking antisense morpholino (MO) synthesized against naca; naca MO1, 5 0 -TTAAAAGCAGTACTAACCCGTCTCA-3 0 , Genetools). Efficacy of MO knockdown was assessed by typical morphological abnormalities of oloca mutant and following SB staining. Parallel studies using another splice blocking MO (naca MO2, 5 0 -AAAAGAAAGGCATCTTACCTGTGCC-3 0 ) produced similar results, so all data presented in this study were derived from using naca MO1. For control embryos, the same amount of a standard control morpholino (5 0 -CCTCT TACCTCAGTTACAATTTATA-3 0 ) was used. Wild-type and mutant mRNA used in the rescue assays were generated using the SP6 mMessage mMachine kit (Ambion) and co-injected with naca MO1 at a concentration of 125 ng ml À 1 .
Wide-field microscopy. VE-DIC and fluorescence microscopy were performed as described previously 6,9 through the 60x/1.00 NA water-immersion objective of a Nikon 90i microscope. Time-lapse recording of VE-DIC images was performed using BTVpro software (Bensoftware, London).
Confocal microscopy, image processing and quantification. Confocal fluorescence microscopy was performed on Leica SPE and SP8 setups as described previously 18 . For quantification of the mean nls-mCherry fluorescence intensity per cell type in double Tg [TCF:nls-mCherry; Fli1:gfp] embryos, a volume of interest (VOI) was defined in 10 cells of each population of interest (stromal, somite muscle and fin mesenchymal cells) per embryo, and for each VOI the mean fluorescence/ voxel was calculated using Imaris software (Bitplane). Eight control and eight naca MO-injected embryos were thus processed. The mean of 10 cells from each cell population was subjected to statistical analysis using unpaired two-tailed Student's t-test (GraphPad Prism software). For nuclear tracking of Bodipy ceramide-stained TCF:nls-mCherry embryo, Imaris software was used for semiautomated four-dimensional tracking of nuclei over time; the tracks were manually checked under stereoscopic viewing, using Nvidia 3D-Vision glasses coupled to a double-frequency (120 Hz) video display.
Whole kidney marrow transplantation. Six adult Tg[mpx:gfp] transgenic fish were deeply anesthetized and all parts of the kidney (head, trunk and tail) were isolated in ice-cold PBS. WKM cells were prepared as described previously 49 . The cell suspension (5-8 Â 10 5 ml À 1 in PBS) of WKM was labelled with 5 mM of CellTracker Red CMTPX (Invitrogen) in PBS at 32°C for 15 min, then washed several times in the zebrafish culture media previously reported by Stachura et al. 50 , then incubated at 32°C for another 40 min to release excess CMTPX. Stained cells were prepared in the zebrafish culture media at a concentration of 5 Â 10 7 ml À 1 for the transplantation. Cell sorting for mpx-GFP þ neutrophils was based on forward and side scatter and GFP fluorescence as described previously 37,51 , using a FACSAria III flow cytometer. Sorted mpx-GFP þ cells were spun down and resuspended in DPBS (Invitrogen) at a concentration of 3 Â 10 7 ml À 1 for transplantation. CMTPX-labelled WKM cells or sorted mpx-GFP þ cells were transplanted into 2 d.p.f. larvae (mutant or sibling) using the same setup as in Sarris et al. 52 . The cells were transplanted into the duct of Cuvier so as to enter circulation. Larvae with similar numbers of transplanted cells were kept overnight at 28°C in E3 medium (5 mM NaCl, 0.15 mM KCl, 0.33 mM CaCl 2 , 0.33 mM MgSO 4 , methylene blue and 1-phenyl-2-thiourea, PTU). Two to thirty hours post transplantation larvae were mounted in 1% low melting point agarose in E3 medium supplemented with 0.16 mg ml À 1 tricaine (Sigma) and used for imaging.
Generation of parabiotic zebrafish embryos. Four ng of naca MO or control MO were injected in 1-cell stage Tg[mpx:gfp] or Tg[lysC:DsRed] embryos, respectively. Injected morphants were raised until 1k-cell (late blastula) stage, then the two blastoderms were fused superficially as described 19 and let to develop into partially fused larvae. Non-injected Tg[mpx:gfp] and Tg[lysC:DsRed] embryos were fused for the control experiment.
Gastrula transplantation. naca or control MO-injected embryos were dechorionated in a glass petri dish filled with high-calcium Ringer solution 19 at late blastula stage. Two donor (naca or control MO-injected H2A-gfp) embryos and 4-5 host (control or naca MO-injected TCF:nls-mCherry) embryos were gently loaded using a glass pipet on a thin layer of 3% methylcellulose spread on a 30 mm petri dish filled with Ringer solution with antibiotics (penicillin/ streptomycin; 1% v/v). The transplant capillary was cut at 25 mm inner diameter and filled with Ringer solution before use. At 40% epiboly stage (5 h.p.f.), B50-100 cells from the donor embryos were aspirated, targeting the ventral margin of the blastoderm that will give rise to somitic mesoderm, using the CellTram micromanipulator (Eppendorf). Around 50 cells were transplanted into the exact same location of each recipient embryo. The transplanted embryos were kept in the same solution overnight at 28°C, then the medium was replaced with Volvic/PTU the next day. The number of GFP and nls-mCherry positive stromal cells in the CHT was counted at 42 and 60 h.p.f. for in each successfully transplanted recipient embryo. Statistical analysis was performed using paired t-test (GraphPad Prism software).
4-PBA treatment. oloca mutant and sibling embryos were treated with 50 mM 4-phenylbutyrate (4-PBA; Sigma) in 10 mM HEPES (pH 7.4)/E3 medium from 24 h.p.f. The medium was renewed every day during the treatment period. Control embryos were incubated with vehicle (DMSO) alone at the equivalent concentration (0.01%). The 4-PBA-or DMSO-treated embryos were separated into sibling and mutant groups then fixed at 3.5 d.p.f. for in situ hybridization or TUNEL staining. Total RNAs for quantitative PCR (qPCR) analysis were extracted from groups of 15 oloca sibling or mutant embryos at 3.5 d.p.f.
Bcl-XL mRNA injection. For synthesis of Bcl-XL mRNA, the plasmid pCS2þ flag-zBcl XL 54 was linearized with Not I then 1 mg of linearized plasmid was subjected to in vitro transcription using SP6 mMessage mMachine kit (Ambion). oloca mutant and sibling embryos in CD41:gfp or TCF:nls-mCherry transgenic background were injected with 200 pg of Bcl-XL mRNA per embryo at the one-cell stage. The injected embryos and non-injected controls were raised till 60 h.p.f. (oloca:TCF:nls-mCherry) or 3 d.p.f. (oloca:CD41:gfp), then subjected to confocal microscopy for cell counting. For the detection of oloca:TCF:nls-mCherry mutant embryos, confocal images of the CHT from 27 randomly chosen embryos were acquired at 60 h.p.f., and the embryos were then raised separately until 3 d.p.f. to ascertain their mutant or sibling phenotype. qPCR and TUNEL staining. oloca sibling and mutant larvae were collected separately at 3-3.5 d.p.f. Total RNA was extracted using TRIzol reagent (Sigma) from each pool of 30 sibling and mutant embryos and 3 mg of total RNA were reverse transcribed to cDNA using M-MLV reverse transcriptase (Promega). Gene-specific primers were designed to detect an amplicon of each target gene of B150 bp. Each PCR reaction was performed with 10 ng cDNA using Power SYBR Green Master Mix (Life Technologies) and 250 nM of each primer. Primer sets used in this study are listed in Supplementary Table 2. qPCR was carried out for three biological replicates with measurements taken from three technical replicates using a ABI7300 real-time PCR machine (Life Technologies). The relative expression of each gene was determined after being normalized to zebrafish elongation factor 1a (ef1a). Data in bars represent mean ± s.d. Statistical analyses were performed using unpaired two-tailed Student's t-test (GraphPad Prism software) and P values o0.001 were considered as significant.