CXCR1 remodels the vascular niche to promote hematopoietic stem and progenitor cell engraftment

The hematopoietic microenvironment is a critical regulator of hematopoietic stem and progenitor cell (HSPC) function under normal conditions, in the presence of malignancy, and under conditions of stress such as regeneration after cytotoxic chemotherapy and engraftment after hematopoietic stem cell transplantation (HSCT; Krause and Scadden, 2012; Morrison and Scadden, 2014). Within the hematopoietic microenvironment there are several niches, each of which hosts cell types with distinct functional roles in the biology of one or more subsets of HSPCs. The vascular niche and the osteoblastic niche have long been appreciated to be important for supporting hematopoietic stem cell (HSC) biology. Elegant work using tissue-specific knockout mice has shown that the vascular niche is indispensable for HSC maintenance and regeneration, whereas the osteoblastic niche is critical for supporting a subset of lymphoid progenitors (Zhu et al., 2007; Hooper et al., 2009; Ding et al., 2012; Ding and Morrison, 2013; Morrison and Scadden, 2014). Within the vascular niche, the arteriolar niche is thought to contain quiescent NG2⁺Nestin^brightα-smooth muscle actin⁺ perivascular stromal cells that express high levels of CXCL12/SDF-1 and maintain HSCs in a state of quiescence under steady-state conditions (Kunisaki et al., 2013). In contrast, the sinusoidal niche is composed of VEGFR2⁺VEGFR3⁺ sinusoidal endothelial cells and Nestin^dimLepr⁺ perivascular stromal cells that express a large number of secreted molecules including Notch ligands, CXCL12/SDF-1, BMP ligands, stem cell factor (SCF), and others (Fernandez et al., 2008; Butler et al., 2010; Ding et al., 2012; Ding and Morrison, 2013). The sinusoidal niche is a dynamic locale with angiogenesis and hematopoietic regeneration occurring in concert after myelotoxic stress. Without an intact sinusoidal vascular niche in the marrow, long term hematopoietic repopulation after myeloablation and HSCT is severely compromised (Hooper et al., 2009). Likewise, the sinusoidal niche is necessary for hematopoiesis in the spleen under conditions of stress such as recovery from chemotherapy, pregnancy, and blood loss (Inra et al., 2015). A complete understanding of the mechanisms by which the sinusoidal niche regulates hematopoiesis during stress will bring to light new therapies that can improve hematopoietic reconstitution.

Recently, we identified an enhancer element for the transcription factor Runx1 that specifically marks HSCs in the developing zebrafish that have long-term hematopoietic repopulating capacity (Tamplin et al., 2015). Using Runx1 transgenic reporter lines, we identified a novel interaction between HSCs and sinusoidal endothelial cells during a period of development characterized by dynamic changes in the niche and expansion of the HSC pool (Tamplin et al., 2015). Runx1⁺ HSCs arise from endothelial cells in the aorta-gonad-mesonephros (AGM) region and enter circulation; beginning around 36 h after fertilization (hpf), they colonize the caudal hematopoietic tissue (CHT), a vascular plexus in the tail of the zebrafish embryo (Murayama et al., 2006). HSCs adhere to the luminal surface of the sinusoidal endothelial cells, transmigrate to the extraluminal space and there they interact intimately with components of the niche including endothelial cells, stromal cells, and possibly other cells, in a process known as “cuddling” (Tamplin et al., 2015; Mahony et al., 2016). HSCs undergo rapid expansion within this temporary niche until ∼6 d post fertilization (dpf), when they migrate to the kidney marrow where they remain for the life of the animal (Chen and Zon, 2009).

Here, we sought to identify the molecular factors that mediate HSPC interactions with the sinusoidal endothelial cells of the CHT niche. Using gene expression studies, gain- and loss-of-function genetics and single-cell track analysis, we show that cxcl8/cxcr1 signaling increases endothelial cell cuddling and enhances expression of cxcl12a/sdf-1a leading to increased HSPC residency time within the niche. These effects allow for additional HSPC cell divisions to occur with a consequent increase in CHT colonization. Using digital reconstruction of the CHT and a parabiotic zebrafish system, we show that cxcl8/cxcr1 signaling positively regulates both CHT volume and HSPC engraftment in a stem cell nonautonomous manner. Collectively, these data identify a new role for cxcl8/cxcr1 in remodeling the sinusoidal vascular niche during the process of HSPC colonization in the developing zebrafish. Further, they suggest that modulation of CXCL8/CXCR1 signaling may be beneficial in recipients of HSCT.

To identify genes that might be involved in HSPC–endothelial cell interactions, endothelial cells and HSPCs were sorted from Tg(kdrl:mCherry;Runx1+23:GFP) transgenic zebrafish embryos at 72 h after fertilization (hpf) and from the kidney marrow of adult transgenic zebrafish. The kdrl (vegfr2) and murine Runx1 enhancer/promoter elements specifically mark zebrafish endothelial cells and HSPCs, respectively (Jin et al., 2005; Chi et al., 2008; Tamplin et al., 2015). We undertook gene expression profiling on all four populations (adult and embryonic mCherry⁺ endothelial cells and GFP⁺ HSPCs) by microarray analysis (Affymetrix Zebrafish 1.0 ST array). To identify sets of genes with differential expression between endothelial cells and HSPCs, we searched the NCBI Gene database for keywords representing 18 categories of extracellular and secreted molecules in the zebrafish genome (Fig. S1). The number of unique genes across all gene sets was 1,900. Some genes were present in more than one gene set; accounting for this overlap, 2462 gene set entries were created. Table S1 lists all 2462 gene set entries ordered by gene set with Entrez unique identifier and gene symbol, if annotated. Gene set enrichment analysis (GSEA; Subramanian et al., 2005) was then performed using these gene sets and the adult and embryonic gene expression data to identify gene sets enriched in the endothelial cell phenotype relative to the HSPC phenotype. In this analysis, the zebrafish chemokine gene set had the most significant enrichment in both adult and embryonic endothelial cells (P = 0.000 for both comparisons, Fig. 1 a and Tables S2 and S3). The intersection of the leading-edge chemokine genes in the embryonic and adult datasets contained 20 genes (Fig. 1 b and Tables S4–S6). 16 of these 20 chemokine genes were subsequently used as candidates in a gain-of-function screen.

Candidate chemokine genes were expressed at high levels in F₀ mosaic transgenic zebrafish by microinjecting single-cell embryos with DNA encoding the candidate gene under the control of the heat shock–inducible promoter for hsp70l (Adám et al., 2000; Halloran et al., 2000). Heat shock induction was performed at 36 and 48 hpf, coincident with the initial wave of HSPC colonization of the CHT. This approach resulted in >20-fold overexpression of the candidate genes compared with animals that did not undergo heat shock induction (Fig. S2). Whole-mount in situ hybridization (WISH) was performed using a mixture of probes marking HSPCs (runx1/c-myb), followed by blinded scoring for HSPC colonization of the CHT. Embryos injected with DNA encoding GFP under the same hsp70l promoter served as a control (Fig. 1 c). Mosaic overexpression of cxcr1 (loc797181) enhanced HSPC colonization in this assay (P = 0.033; n = 63 embryos; Fig. 1, d and e). The ligand for cxcr1 in humans, cxcl8, similarly enhanced HSPC colonization of the CHT (P = 0.0028; n = 41 embryos; Fig. 1, f and g).

WISH for zebrafish cxcr1 in wild-type embryos revealed a diffuse low-level staining pattern with specific enhancement in the CHT (Fig. 1, h and i). Expression of both cxcr1 and cxcl8 was detected by RT-PCR in kdrl:mCherry⁺ endothelial cells sorted from embryos at 72 hpf (Fig. 1 j). Neither cxcr1 nor cxcl8 were detectable in sorted HSPCs at the mRNA level (unpublished data). Sorted endothelial cells were treated for 30 min in vitro with 11,12-epoxyeicosatrienoic acid (EET), a small arachidonic acid derivative recently shown by our laboratory to enhance HSC engraftment in embryonic zebrafish and in adult zebrafish and murine HSCT assays (Li et al., 2015; Tamplin et al., 2015). This brief treatment enhanced endothelial cell cxcl8 expression by approximately sixfold and cxcr1 expression by approximately threefold (Fig. 1, k and l). Thus, cxcr1 and cxcl8 are present within the endothelial cell niche early during its development and both are inducible by a compound known to play a role in promoting hematopoiesis.

To quantitate the effects of cxcr1 and cxcl8 on HSPC colonization, we performed gain- and loss-of-function experiments in the Tg(Runx1+23:NLS-mCherry) HSPC transgenic reporter line (referred to hereafter as Runx1:mCherry; Tamplin et al., 2015). Mosaic, high-level gene expression was induced in F₀ transgenic animals injected with DNA encoding hsp70l:cxcr1 or hsp70l:GFP by heat shock at 36 and 48 hpf. Stable transgenic lines (F₁ generation) did not tolerate heat shock induction of cxcr1, likely because the transgene was expressed at toxic levels. The entire CHT was imaged at 72 hpf by fluorescence confocal microscopy (Fig. 2 a, black box). Digital image analysis (Imaris; Bitplane Scientific) was used for unbiased image segmentation and classification of spots by a composite measure of fluorescence intensity, size and roundness. The threshold for identifying HSPCs from background fluorescence was set such that ∼3–8 such spots were present in the control group, a number known to be present in this transgenic line based on limiting dilution experiments and imaging of the entire CHT (Tamplin et al., 2015). The same thresholds were then used to segment and classify spots as HSPCs in experimental groups. Fig. 2 (b_i and c_i) shows the raw fluorescence images from a portion of the CHT (denoted by the red box in Fig. 2 a). Fig. 2 (b_ii and c_ii) shows the HSPCs as identified by unbiased digital analysis, and Fig. 2 (b_iii and c_iii) shows an overlay of the fluorescence images and segmented spots. Mosaic overexpression of cxcr1 beginning at 36 hpf increased HSPC colonization of the CHT at 72 hpf (3.8 ± 1.0 vs. 9.1 ± 1.9 HSPCs per CHT; P = 0.020, Fig. 2 d). HSC specification in the AGM requires cxcl8 expression and so this experiment was repeated with heat shock induction at 48 and 60 hpf, a period of time after peak HSC specification but still during the initial wave of CHT colonization (Jing et al., 2015). HSPC colonization of the CHT was enhanced by cxcr1 under these conditions as well (4.3 ± 0.9 vs. 11.8 ± 3.1 HSPCs per CHT, P = 0.023, Fig. 2, e–g).

Next, to understand the requirement for cxcl8 in HSPC colonization of the CHT, we bred the Runx1:mCherry reporter transgene into animals carrying a loss-of-function missense mutation in cxcl8 (cxcl8^−/− (Jing et al., 2015)). Runx1:mCherry;cxcl8^+/− animals were incrossed and the progeny imaged at 72 hpf, followed by genotyping for cxcl8. Runx1:mCherry;cxcl8^+/+ animals had significantly higher HSPC colonization of the CHT compared with Runx1:mCherry;cxcl8^−/− clutchmates (7.2 ± 1.4 vs. 2.6 ± 0.7 HSPCs per CHT, P = 0.02, Fig. 2, h–k). There was no significant difference between Runx1:mCherry;cxcl8^+/− animals and either wild-type or mutant clutchmates.

Finally, we wished to understand if a related chemokine receptor, cxcr2 (loc796724), would similarly enhance HSPC colonization. Global overexpression of zebrafish cxcr2 beginning at 36 hpf did not increase HSPC colonization when tested in this assay (Fig. 2 l).

CXCL8 and other immune mediators such as IL-6, IFN-γ, TNF, prostaglandin E₂ (PGE₂), and EET may have a role in directly stimulating HSC emergence from the AGM (North et al., 2007; Kobayashi et al., 2010; Espín-Palazón et al., 2014; Sawamiphak et al., 2014; Jing et al., 2015; Li et al., 2015). Cxcl8 morphants and cxcl8^−/− mutants have reduced numbers of HSCs in the AGM (Jing et al., 2015). Although induction of cxcr1 in our gain-of-function experiments begins at the end of AGM hematopoiesis, it is possible that enhanced HSC specification and emergence might contribute to the increased numbers of HSPCs seen in the CHT. To rule out this possibility, Tg(sclβ:GFP;kdrl:mCherry) animals were generated and cxcr1 expression was induced at 36 hpf by DNA microinjection and heat shock. The AGM was imaged from 38–49 hpf by time-lapse video microscopy and emerging sclβ:GFP⁺ HSPCs were identified as they budded from the floor of the dorsal aorta into the AGM (Kissa and Herbomel, 2010; Zhen et al., 2013; Jing et al., 2015; Fig. 3, a–e). The number of HSPCs per somite over this period is plotted in Fig. 3 f. Induction of cxcr1 caused no significant increase in HSC emergence compared with animals injected with control DNA encoding an empty vector or uninjected animals treated with DMSO. In contrast, animals treated with PGE₂ as a positive control beginning at 16-somite stage had significantly higher numbers of HSPCs per somite compared with the other groups. Collectively, these experiments using gain- and loss-of-function genetic techniques demonstrate a new and independent role for cxcl8/cxcr1 signaling in supporting HSPC colonization of the CHT.

We next aimed to understand the cellular mechanisms by which cxcr1 supports HSPC colonization. Heat shock induction of cxcr1 gene expression was performed at 36 and 48 hpf, followed by time-lapse fluorescence microscopy of Runx1:mCherry embryos from 52 to 72 hpf. Digital image analysis was performed, as previously described, and HSPCs were enumerated over these 20 h. Mosaic overexpression of cxcr1 had little effect on HSPC numbers within the CHT until ∼55 hpf, after which there were consistently more HSPCs in animals expressing cxcr1 compared with GFP control (Fig. 4 a). As our previous experiments had suggested that AGM hematopoiesis was not enhanced by cxcr1 under these conditions, we hypothesized that the increased numbers of HSPCs in the CHT of these animals would be due in part to increased retention time within the niche itself. Single-cell track analysis was performed to determine the residency time of individual HSPCs within the CHT (n = 648 tracks analyzed, Fig. 4 b). Control animals expressing GFP showed a distribution of residency times with the large majority of cell tracks lasting <6 h, consistent with previously published results (Tamplin et al., 2015). In contrast, animals expressing cxcr1 had an expanded number of cell tracks lasting between 7 and 15 h, increasing the residency time of the top 10% of HSPCs by nearly 2 h (Fig. 4 b, dashed lines). These long-term resident HSPCs in animals overexpressing cxcr1 significantly increased the mean residency time of this group compared with control (3.0 ± 0.1 vs. 3.4 ± 0.2 h, P = 0.029).

After the observation of increased CHT residency time in animals overexpressing cxcr1, it next seemed likely that HSPCs in these animals would have the opportunity to undergo additional rounds of cell division while within the niche. Mitotic events were enumerated in the time lapse datasets, and it was found that HSPCs in animals overexpressing cxcr1 were 1.8-fold more likely to undergo mitosis relative to GFP control (0.31 ± 0.05 vs. 0.56 ± 0.08 mitoses per HSPC, P = 0.0041, Fig. 4, c and d). The increased number of HSPC cell divisions was not only a consequence of residency time, as overexpression of cxcr1 also increased the intrinsic rate of HSPC mitosis by 1.5-fold (0.07 ± 0.01 vs. 0.11 ± 0.02 mitoses/HSPC/hour within the CHT, P = 0.02, Fig. 4 e). Finally, we hypothesized that endothelial cell cuddling would be increased in association with the effects on CHT retention time and HSPC mitosis seen in these time-lapse experiments. To test this hypothesis, cxcr1 was overexpressed in Runx1:mCherry;kdrl:GFP double transgenic animals, and cuddling events were identified as shown in the still frames in Fig. 4 f (n = 468 total tracks evaluated; Video 1). In this example, the HSPC identified by the arrow is in transit from 51–52 hpf and remains cuddled in an endothelial cell pocket from 52–59 hpf before exiting. Fig. 4 f_vi shows mitosis of this cell and release of a daughter cell (arrowhead) from the pocket. The percent cuddling time relative to overall time each HSPC spent in the CHT was significantly increased in animals overexpressing cxcr1 compared with control animals (40.1 ± 2.3 vs. 66.2 ± 4.5% cuddling time; P = 6.28 × 10⁻⁷; Fig. 4 g). Collectively, these findings show that cxcl8/cxcr1 signaling acts as an overall positive regulator of CHT colonization in the developing zebrafish. This effect is likely a consequence of enhanced interactions with the sinusoidal endothelial cell niche, leading to prolonged retention time, increased rate of HSPC mitosis, and an overall expansion of the HSPC pool.

To better understand the mechanisms by which cxcr1 acts on the endothelial cell niche to enhance HSPC colonization, we generated a stable transgenic line expressing cxcr1 under the control of the endothelial cell-specific promoter, kdrl (Tg(kdrl:cxcr1), hereafter referred to as kdrl:cxcr1). Expression of cxcr1 in sorted endothelial cells from these transgenic embryos was roughly 2.4-fold higher than in endothelial cells from WT clutchmates by quantitative PCR (Fig. 5 a). There were no significant differences in expression of other angiogenesis-related genes in sorted endothelial cells by RNA sequencing. Runx1:mCherry;kdrl:cxcr1 double transgenic animals were imaged at 72 hpf and found to have significantly increased HSPC colonization of the CHT compared with Runx1:mCherry clutchmates without the cxcr1 transgene (6.1 ± 1.1 vs. 10.8 ± 1.2 HSPCs per CHT; P = 0.001, Fig. 5, b–d). Mpx:GFP;kdrl:cxcr1 transgenic animals did not have increased numbers of mature neutrophils within the CHT suggesting that overexpression of cxcr1 does not enhance colonization of all myeloid cell types (Fig. 5 e).

CXCL12 (SDF-1) is a chemokine ligand expressed by endothelial and perivascular stromal cells that is essential for HSC maintenance and engraftment after transplantation (Nagasawa et al., 1996; Ara et al., 2003; Sugiyama et al., 2006; Ding and Morrison, 2013; Inra et al., 2015). Enforced expression of CXCL12 by endothelial cells in CXCL12^−/− mice is sufficient to rescue a profound defect in HSC numbers in the bone marrow (Ara et al., 2003). We therefore hypothesized that the increased numbers of HSPCs in kdrl:cxcr1 transgenics might be associated with induction of cxcl12 in the CHT. Zebrafish cxcl12 exists in two isoforms, cxcl12a and cxcl12b, both of which have critical roles in embryonic hematopoiesis (Zhang et al., 2011; Nguyen et al., 2014; Tamplin et al., 2015). Overexpression of cxcr1 in kdrl:cxcr1 transgenic animals increased expression of cxcl12a within the CHT compared with WT clutchmates by WISH (P = 0.03; n = 60 embryos, Fig. 5, f–h) whereas expression of cxcl12b was unaffected (Fig. 5, i–k). This suggests that cxcr1 modifies the function of the niche to favor HSPC colonization.

To look specifically at the response of endothelial cells to cxcr1 stimulation, primary human umbilical vein endothelial cells (HUVECs) were treated for 6 h in vitro with purified recombinant human CXCL8 (rhCXCL8, RND systems). This brief treatment induced endothelial cell CXCL12 mRNA expression by 1.5-fold compared with control (Fig. 5 l). Expression of survivin and VEGFA, known targets of CXCL8/CXCR1 in endothelial cells, was induced by 2.4- and 1.5-fold, respectively (Fig. 5 l; Li et al., 2003). In addition, rhCXCL8 treatment induced expression of CXCL8 itself (Fig. 5 l). After 24 h in culture, VEGF protein was undetectable in the culture supernatants of control-treated cells but could be detected at low levels in cells treated with rhCXCL8 (4.8 ± 0.86 pg/ml). There was no difference in CXCL12 levels in culture supernatants (22.8 ± 1.1 vs. 24.4 ± 0.97 pg/ml). RNA sequencing and IPA analysis were performed to understand the global gene expression changes induced by CXCL8/CXCR1 signaling (Fig. 5 m). The induction of anti-apoptotic and angiogenic factors (survivin and VEGFA) as well as the global gene expression changes in favor of cellular maintenance and movement suggested the possibility that CXCL8/CXCR1 signaling might cause structural reorganization or expansion of the niche that could provide space for additional HSPC colonization.

To address this hypothesis, cxcr1 expression was induced at 36 and 48 hpf in wild-type zebrafish embryos using the heat shock-inducible system and the structure of endothelial cell niche was assessed by WISH for the endothelial cell marker, kdrl. Overexpression of cxcr1 increased the size and intensity of kdrl staining in a blinded scoring assay (P = 0.041; n = 40; Fig. 6, a–c). The larger vasculature was grossly unaffected. To measure the effect of cxcr1 on the size of the niche directly, cxcr1 expression was induced by DNA microinjection and heat shock induction in Tg(kdrl:GFP) or Tg(kdrl:hRAS-mCherry; hereafter referred to as kdrl:GFP and kdrl:mCherry) reporter zebrafish as before. Animals were imaged by confocal microscopy at 72 hpf (Fig. 6 d). The CHT was then digitally reconstructed in three dimensions and its volume measured using Imaris digital image analysis software (Fig. 6, e and f). Induction of cxcr1 expression increased the volume of the CHT by 26% compared with animals that did not undergo heat shock induction (1.06 ± 0.08 vs. 1.34 ± 0.07 × 10⁶ µm³; P = 0.02, Fig. 6 g). Treatment of kdrl:mCherry transgenics with the selective inhibitor of CXCR1/CXCR2, SB225002 (Cayman Chemical), from 48 to 72 hpf reduced CHT volume at 72 hpf (8.91 ± 0.49 vs. 6.4 ± 0.70 × 10⁵ µm³; P = 0.012; Fig. 6 h). These findings at static time points were confirmed by measuring CHT volume continuously from 52 to 72 hpf using time-lapse video microscopy. Overexpression of cxcr1 beginning at 36 hpf progressively increased CHT volume from 53 hpf onward (Fig. 6 i). To look specifically at the effect of cxcr1 expression in the endothelial cell compartment on the volume of the niche, we again used the kdrl:cxcr1 transgenic line. Double transgenic kdrl:cxcr1;kdrl:mCherry animals were generated and were found to have significantly greater CHT volume compared with kdrl:mCherry clutchmates (1.1 ± 0.05 vs. 1.3 ± 0.06 × 10⁶ µm³; P = 0.02; Fig. 6 j). Finally, we wished to confirm these findings in an independent reporter line. Lymphatic vessel endothelial hyaluronan receptor 1b (lyve1b) is expressed in lymphatic vessels and in the CHT at 72 hpf (Fig. 6, k–p; Flores et al., 2010; Okuda et al., 2012). The Tg(lyve1b:GFP; referred to as lyve1b:GFP) line was crossed to the kdrl:cxcr1 line to generate lyve1b:GFP;kdrl:cxcr1 and lyve1b:GFP (referred to as WT) clutchmates, followed by imaging of the CHT at 72 hpf. Embryos carrying the kdrl:cxcr1 transgene had a significantly larger CHT compared WT control embryos (5.3 ± 0.17 vs. 6.3 ± 0.18 × 10⁵ µm³; P = 0.0006; Fig. 6 p). These findings demonstrate that in concert with induction of cxcl12a expression, cxcr1 induces structural changes that support HSPC colonization within the sinusoidal vascular niche.

We aimed to better understand the cell autonomy of the effects of cxcr1 on HSPC engraftment. To do this, pairs of zebrafish embryos were fused at 4 hpf to create parabiotic organisms (Demy et al., 2013; Murayama et al., 2015; Hagedorn et al., 2016). All parabiotics used in these experiments developed with conjoined anterior structures, shared circulation, and morphologically normal bodies and tails. Approximately 70% of all viable parabiotics met these criteria. Kdrl:GFP embryos were modified by microinjection of DNA encoding hsp70l:cxcr1 or empty vector as a control, and were fused to uninjected clutchmates (Fig. 7, a and b). The injected side of each pair was marked by co-injecting fluorescent dextran (Fig. 7, c and d). Gene expression was induced by heat shock at 36 and 48 hpf and CHT volume was measured in both sides of each parabiotic at 72 hpf by fluorescence confocal microscopy followed by 3D digital reconstruction. The fold change in CHT volume (injected:uninjected) is plotted in Fig. 7 e. As expected, when parabiotics were injected with control DNA, the fold change in CHT volume was close to 1 (0.9 ± 0.05-fold change). However, when hsp70l:cxcr1 was injected, CHT volume was significantly increased on the injected side (1.27 ± 0.11-fold change; P = 0.012, compared with control parabiotics). This increase in the CXCR1 group is in line with what was observed in our heat shock induction experiments in single organisms. These data indicate that cxcr1 acts locally within the niche and not via circulating cells or other factors that would otherwise have transited to the uninjected side and nullified any difference in CHT volume.

Finally, we created a parabiotic zebrafish system that allowed us to quantify HSPC engraftment in a genetically distinct niche. Runx1:mCherry donor zebrafish were fused to casper recipient zebrafish that had been injected with DNA encoding either cxcr1 or GFP under control of the hsp70l promoter (Fig. 7 f). In this system, the donor niche is unmodified and donor-derived HSPCs can be detected in both donor and recipient CHTs (Fig. 7, g and h). Gene expression was induced by heat shock at 36 and 48 hpf, parabiotics were imaged at 72 hpf and donor-derived HSPCs colonizing the autologous donor niche and those engrafted in the recipient niche were quantified using unbiased digital image analysis. In control parabiotics expressing GFP within the recipient niche, there was no difference in donor-derived HSPC colonization of the donors and engraftment of the recipients (3.8 ± 1.7 vs. 5.0 ± 1.8 HSPCs per CHT, p = NS, Fig. 7 i). However, when cxcr1 was overexpressed within the recipient niche, donor-derived HSPC engraftment of the recipient niche was significantly greater than colonization of the autologous donor niche (11.4 ± 2.1 vs. 19.8 ± 3.1 per CHT; P = 0.019; Fig. 7 i). Moreover, this increase was not simply the result of preferential homing, as overexpression of cxcr1 increased HSPC numbers in both donors and recipients compared with control parabiotics (P = 0.045 for donors and P = 0.007 for recipients, Fig. 7 i). Rather, these data suggest that the action of cxcr1 within the recipient niche allowed expansion of donor-derived HSPCs, likely by increasing residency time and mitotic rate. This expanded pool of donor cells then freely circulated back to the donor niche. These parabiotic models show that the activity of cxcr1 is nonautonomous to HSPCs and other circulating cells. Together with the effects seen in the kdrl:cxcr1 transgenic model, they provide further data in support of a role for cxcr1 in acting autonomously within the vascular niche to promote HSPC engraftment.

HSPC–microenvironment interactions are critical for maintaining the hematopoietic system under steady-state conditions and to allow the organism to respond to systemic stress such as hemorrhage, infection, toxin exposure, chemotherapeutic treatment or radiation damage (Hooper et al., 2009; Ding et al., 2012; Inra et al., 2015). Vertebrate development is characterized by rapid expansion of the hematopoietic system, and thus embryonic hematopoiesis in the zebrafish CHT reflects the biology of hematopoietic regeneration after toxic, inflammatory, or infectious insult (Tamplin et al., 2015). A growing body of literature has made it clear that the microenvironment communicates with HSPCs via proinflammatory cytokine signals in the settings of stress, embryonic or fetal development, and hematologic malignancies (Mirantes et al., 2014; Luis et al., 2016). Here, we have used gain- and loss-of-function experiments in the developing zebrafish to show that signaling via the proinflammatory chemokine receptor, cxcr1, enhances HSPC colonization of the sinusoidal endothelial cell niche.

Work from our laboratory has recently shown that an intronic enhancer sequence from the murine Runx1 gene (Runx1+23) marks zebrafish hematopoietic stem cells with long-term repopulating capacity when linked to mCherry or GFP reporter transgenes (Tamplin et al., 2015). This work showed that there are between 4 and 5 phenotypically defined Runx1⁺ cells within the CHT at this developmental stage, and that ∼1 in 3 of these are long-term HSCs by limiting dilution transplant experiments (Tamplin et al., 2015). The Runx1 reporter lines were used to demonstrate endothelial cell cuddling of HSPCs within the CHT (Tamplin et al., 2015). We have now used these same lines to begin to uncover the molecular mechanisms by which cuddling and stem cell colonization of the CHT occurs. Visualization of CHT colonization by Runx1⁺ HSPCs in live embryos followed by single cell track analysis demonstrated that cxcl8/cxcr1 signaling enhances endothelial cell cuddling of HSPCs, prolonging HSPC residency time within the niche and allowing additional mitotic events to occur. Gene expression studies in zebrafish and primary human endothelial cells suggest that cxcl8/cxcr1 signals in a positive feedback loop, inducing expression of cxcl12a in the process and providing a molecular mechanism to promote HSPC retention.

The embryonic zebrafish is unique among vertebrates in that the entire hematopoietic niche can be imaged at fine spatiotemporal resolution. Digital image analysis allowed us to reconstruct the CHT in three dimensions and to quantitate changes in CHT volume in the presence of enhanced cxcr1 signaling. Though we cannot rule out the possibility that cxcr1 may act via induction of secondary paracrine factors that act locally within the niche, our experiments using kdrl:cxcr1 transgenic fish, HUVECs, and parabiotic zebrafish strongly support the notion that cxcr1 acts directly on the vascular niche, expanding its volume and enhancing HSPC engraftment and proliferation. Based on these data, we propose the following novel mechanism: before the onset of HSPC colonization, the primitive vascular niche of the CHT expresses low levels of cxcr1 and cxcl8. Cxcl8 is induced within the niche in the presence of local vasoactive lipid mediators such as EET. It is possible that myeloid cells or a subset of HSPCs produce cxcl8 under some circumstances as even stem and progenitor cells have been shown to be potent cytokine producers (Zhao et al., 2014; Li et al., 2015; Sinclair et al., 2016). Cxcl8 signals in an autocrine/paracrine fashion via cxcr1 on endothelial cells, inducing expression of vegfa and promoting survival and motility of the sinusoidal endothelial cells within the CHT. Cxcl12a is induced, allowing colonization and maintenance of HSPCs within the niche. Cxcr1 promotes endothelial cuddling, which allows HSPCs to expand and produce more mature myeloid cells which may also support the niche through cxcl8 expression. By 5–6 dpf, the kidney marrow has developed and outcompetes the CHT as a favorable HSPC niche. We speculate that the loss of cxcl8-producing progenitors and myeloid cells at this time interrupts the proangiogenic positive feedback loop in the CHT and allows the terminal differentiation of the sinusoids into intersegmental veins.

Though cxcl8/cxcr1 signaling has well defined roles in host defense, angiogenesis, and inflammation, this pathway is only beginning to be understood as an important factor in hematopoiesis (Cacalano et al., 1994; Terashima et al., 1998; Waugh and Wilson, 2008; Deng et al., 2013; Zhang et al., 2015). We have recently found that loss of cxcl8 signaling leads to a lower rate of endothelial–hematopoietic transformation (EHT) and reduced HSC numbers in the AGM (Bertrand et al., 2010; Kissa and Herbomel, 2010; Jing et al., 2015). This role for cxcl8 in the specification of HSCs from hemogenic endothelium is in accordance with other work demonstrating that TNF, IFN-γ, PGE₂, EET, TLR agonists, and other proinflammatory mediators are important for the earliest phase of definitive hematopoiesis (North et al., 2007; Espín-Palazón et al., 2014; Sawamiphak et al., 2014; He et al., 2015; Li et al., 2015).

Beyond HSC specification, cxcl8 and other chemokines have been implicated in the complex interplay between HSPCs and mature myeloid cells in the process of mobilization from the niche. In an early study, a single dose of recombinant human CXCL8 was found to rapidly mobilize HSPCs in mice (Laterveer et al., 1995). Extensive work by the Pelus laboratory has compared the mobilization characteristics of several members of the CXC family of chemokines and has highlighted the importance of CXCL2 (Gro-β) signaling via CXCR2 in this process (Pelus et al., 2002). In these studies, stem cell mobilization appears to be dependent on MMP-9 and CXCR2 expression by neutrophils (Cacalano et al., 1994; Pelus et al., 2004; Pelus and Fukuda, 2006). In addition, a recent study has shown that CXCR2^−/− mice have impaired vascular regeneration in the marrow after lethal irradiation (Hale et al., 2015). Nevertheless, placing CXCL8/CXCR1 precisely within this conceptual framework has been difficult because mice lack a close ortholog of CXCL8, its absence likely complemented by CXCL1 (Gro-α/KC) and CXCL2 (Gro-β/MIP-2; Lee et al., 1995; Wang et al., 1997).

In zebrafish, the CXC–chemokine system is complicated by an ancestral genome amplification event that produced three cxcl8 genes, located on chromosomes 1, 7, and 17 (van der Aa et al., 2010). Zebrafish cxcl8 is expressed in leukocytes, though the different contributions of each cxcl8 locus remain poorly understood (Oehlers et al., 2010). Cxcr1 and cxcr2 are broadly expressed in leukocytes, vasculature, the developing gut, and elsewhere (Oehlers et al., 2010; Deng et al., 2013). In a zebrafish model of pseudomonas infection, cxcr2 knockdown, but not cxcr1 knockdown, reduced neutrophil chemotaxis to cxcl8 and leukotriene B4 and impaired localization at the site of infection, indicating that cxcl8/cxcr2 is the critical pathway in response to infection. Intriguingly, the same study noted that whereas cxcr1 knockdown did not alter neutrophil responses to infection, it did substantially reduce the number of neutrophils present in the CHT at 72 hpf (whereas cxcr2 knockdown had no effect in the CHT), suggesting that cxcr1 may be involved in myeloid development (Deng et al., 2013). We now directly show that cxcr1 is an important mediator of HSPC engraftment within the vascular niche, whereas cxcr2 may have little or no role in this process.

Understanding the interactions between HSPCs and the vascular niche in the setting of vertebrate development has shed light on how the hematopoietic system responds to stress conditions. Additional mechanistic studies in animal models of normal and stress hematopoiesis will be necessary to fully understand the role of CXCL8, CXCR1, and CXCR2 in hematopoietic recovery after administration of chemotherapy, engraftment after HSCT, and in the development of hematologic malignancies and myeloproliferative disorders. However, one can speculate that modulating CXCL8/CXCR1 signaling in these patients might aid in favorably remodeling the niche to promote the recovery of normal hematopoiesis.

Wild-type AB, casper, and transgenic lines Runx1:GFP, Runx1:mCherry (Tamplin et al., 2015), kdrl:GFP (Jin et al., 2005), kdrl:mCherry (Chi et al., 2008), lyve1b:GFP (Okuda et al., 2012), sclβ:GFP (Zhen et al., 2013), and kdrl:cxcr1 were used in this study. Embryos were used up to 84 hpf. Cxcl8^−/− fish were generously provided by A. Huttenlocher. All animals were housed at Boston Children’s Hospital and were maintained according to institutional animal care and use committee protocols.

Zebrafish embryos were manually dissociated in PBS containing collagenase (Liberase; Sigma-Aldrich) at 72 hpf. Adult zebrafish were euthanized by tricaine overdose and ice water immersion, and kidney marrow was harvested. Single-cell suspensions were made using a 40-µm filter and indicated cell populations were sorted by FACS using a FACSAria machine (BD).

Total RNA was purified from flow-sorted cells, and gene expression profiling was performed by microarray analysis (Zebrafish 1.0 ST Array; Affymetrix) as previously described (Tamplin et al., 2015). Gene sets were created by searching the NCBI Gene database for the categories of extracellular and secreted molecules listed in Table S1. Gene set enrichment analysis was performed using adult and embryonic gene expression datasets with three biological replicates for endothelial cells and HSPCs. GSEA v2.2.0 software was used for the analysis. Entrez IDs were used to identify genes in the gene sets and gene expression data. Between 44 and 90% of the genes in each gene set were included in the expression datasets. Leading edge subsets of each gene set were identified for the adult and embryonic data and overlap was identified using Venny 2.0.

Microarray and RNaseq data are available from the Gene Expression Omnibus under accession nos. GSE56015 and GSE92543.

There is variability in the literature and informatics databases regarding the nomenclature for the zebrafish cxcr1 and cxcr2 genes. We have used the identifications suggested by Deng et al., 2013 for cxcr1 (Entrez ID: 797181; Ensembl gene: ENSDARG00000052088) and cxcr2 (Entrez ID: 796724; Ensembl gene: ENSDARG00000054975).

Candidate zebrafish coding sequences were amplified from kidney marrow total RNA using the Superscript III RT kit (Thermo Fisher Scientific) and gene-specific primers. PCR products were cloned into the pentr SD Topo middle entry vector (Invitrogen), and then into Tol2-based expression vectors via Gateway reaction (Kwan et al., 2007). Hsp70l and kdrl promoters were used as 5′ sequences (Adám et al., 2000; Jin et al., 2005). Tol2 expression vectors (20 pg per construct) were microinjected into the cell of zebrafish zygotes. Hsp70l:GFP (5 pg) was co-injected to mark successfully injected embryos with high mosaicism. Tol2 mRNA (200 pg) was co-injected to enhance genomic integration of the constructs. Viable embryos were selected at 24 hpf.

Viable embryos were transferred to 96-well PCR plates at indicated time points in 100 µl embryo water. Heat shock was performed in a standard PCR thermal cycler. Embryos were incubated at 40°C for 30 min before being returned to the standard incubator (28°C). A second heat shock was performed 12 h later, and animals were removed from the plates by gently flushing each well with embryo water.

Zebrafish embryos were treated by adding SB225002 (0.5 µM) or dimethyl PGE₂ (10 µM) directly to the fish water at the indicated time points (Cayman Chemicals). An equal volume of DMSO was added to the fish water for control groups to serve as a negative control.

Zebrafish embryos were euthanized by tricaine overdose and fixed in 4% PFA before in situ hybridization for runx1 was performed and c-myb, cxcl12a, cxcl12b, or kdrl mRNA using standard techniques (Thisse and Thisse, 2008). For scoring, groups of 10–20 embryos were deidentified and suspended in glycerol before performing a semiquantitative assessment of the intensity of specific staining in the CHT region. Zebrafish groups were unblinded and scores compared using the Wilcoxon rank sum test (R v3.2.2).

Parabiotic zebrafish were made according to published protocols (Demy et al., 2013; Hagedorn et al., 2016). In brief, zebrafish embryos were fused at the 1,000-cell stage by embedding in methylcellulose and covering in high-calcium Ringer’s solution containing penicillin and streptomycin. Approximately 70% of such animals developed with fused heads, morphologically normal tails, and shared circulation. Successfully joined parabiotic animals underwent heat shock induction of gene expression as described.

Zebrafish embryos and parabiotics were anesthetized with tricaine and embedded in 1.2% agarose on MatTek glass-bottom 6-well plates using standard techniques (Westerfield, 2007). Images were acquired using a Yokogawa spinning disk confocal and Nikon inverted Ti microscope. A Nikon 20× air Plan-Apo differential interference contrast NA 0.75 objective was used. Images were acquired with Andor iXon x3 EMCCD cameras and NIS Elements software. Confocal z-stacks were acquired at 2 µm intervals. Time lapse video frames were acquired every 10–30 min, depending on acquisition time, using an automated stage.

All image analysis was performed using Imaris software (Bitplane) as described in detail in the Results section.

Primary zebrafish endothelial cells were treated with EET (5 µM) or DMSO control for 30 min at room temperature before lysis and RNA purification (Li et al., 2015). Primary HUVECs were cultured in supplemented M200 medium containing hEGF, hydrocortisone, hFGF, VEGF, R3-IGF-1, ascorbic acid, heparin, and fetal bovine serum (Gibco). Cells were grown in a humidified 5^% CO₂ atmosphere at 37°C. Cells were starved in unsupplemented medium for 12 h before addition of recombinant human CXCL8 (R&D Systems) at a final concentration of 10 ng/ml or an equal volume of vehicle control (water). Cells were harvested after 6 h of treatment, and RNA was purified using the RNeasy Plus Mini kit (QIAGEN). For protein expression experiments, HUVEC culture supernatants were harvested and filtered before analysis by ELISA. Human VEGF and CXCL12 ELISA kits were purchased from Abcam and used according to the manufacturer’s instructions.

Total RNA was purified from zebrafish endothelial cells, HSPCs, and HUVECs using the RNeasy Plus Mini kit (QIAGEN). Complementary DNA was reverse transcribed using the SuperScript III RT kit (Thermo Fisher Scientific) and a mixture of poly-A and random hexamer primers. Quantitative PCR was performed using SSOFast Supermix with EvaGreen (Bio-Rad Laboratories) and a Bio-Rad thermal cycler. The following primer sequences were used in the study: zf-18s: 5′-TCGCTAGTTGGCATCGTTTATG-3′ (forward), 5′-CGGAGGTTCGAAGACGATCA-3′ (reverse); zf-cxcl8: 5′-TCATTGAAGGAATGAGCTTGAGAG-3′ (forward), 5′-CCAGTTGTCATCAAGGTGGC-3′ (reverse); zf-cxcr1: 5′-GTGATCGTACGCGCTATGGA-3′ (forward), 5′-ATTCGGGTTGCTAATCGCCA-3′ (reverse) ; hs-CXCL12: 5′-ACATGGCTTTCGAAGAATCG-3′ (forward), 5′-GCTGGTCCTCGTGCTGAC-3′ (reverse); hs-survivin: 5′-TTGGTGAATTTTTGAAACTGGA-3′ (forward), 5′-CTTTCTCCGCAGTTTCCTCA-3′ (reverse); hs-VEGFa: 5′-AGGGCAGAATCATCACGAAGT-3′ (forward), 5′-AGGGTCTCGATTGGATGGCA-3′ (reverse) ; hs-CXCL8 5′-AAATTTGGGGTGGAAAGGTT-3′ (forward), 5′-TCCTGATTTCTGCAGCTCTGT-3′ (reverse).

For the HUVEC RNA-seq, ribosomal RNA was depleted from total RNA using the Ribo-Zero kit (Epicentre) according to the manufacturer’s instructions. Sequencing libraries were constructed using the NEBNext Ultra kit for Illumina (New England Biolabs). For zebrafish endothelial cell RNA-seq, cells were lysed in TRIzol LS and total RNA was extracted by phenol-chloroform extraction and isopropanol precipitation. Messenger RNA selection and first-strand cDNA synthesis was performed using the SMARTer Universal Low Input RNA kit (Takara Bio Inc.). Libraries were synthesized using the Low Input Library Prep kit for Illumina (Takara Bio Inc.). All libraries were sequenced on an Hi-Seq 2500 instrument (Illumina).

For the GSEA experiments, enrichment scores and P-values were calculated by the GSEA software package (Subramanian et al., 2005). The remainder of the statistical analyses were performed and data were plotted using R v3.2.2. No statistical method was used to predetermine sample size. Embryos were randomly selected for experimental and control groups before DNA injection. All animals with normal morphology, expected levels of reporter transgene expression, and normal circulation were included in the analyses. Sample variations were compared using the Kolmogorov-Smirnov test. HSPC numbers and CHT volume were then compared using the two-sided Student’s t test or the Wilcoxon rank sum test, depending on the sample distributions. In situ hybridization data were compared using the Wilcoxon rank sum test. In all tests, P < 0.05 was considered to be statistically significant. Mean ± SEM is reported unless indicated otherwise.

Fig. S1 shows the bioinformatic workflow used to identify differentially regulated genes between endothelial cells and HSCs and to select candidate genes for the gain-of-function screen. Fig. S2 provides an example of the degree of global overexpression achieved in our gain-of-function experiments. Video 1 shows a representative HSC-endothelial cell cuddling event. Tables S1–S6, available as Excel files, identify the gene sets, component genes, and results of GSEA.

This work was supported by National Institutes of Health R01HL04880, P01HL032262, P30DK049216, R01DK53298, U01HL10001, R24DK092760 (L.I.Z.) and postdoctoral awards from Alex’s Lemonade Stand Foundation, the Cancer Research Institute, and the Conquer Cancer Foundation of ASCO (B.W.B.). L.I.Z. is a founder and stock holder of Fate Therapeutics, Marauder Therapeutics, and Scholar Rock. The authors declare no competing financial interests.

Author contributions: Conceptualization, B.W. Blaser and L.I. Zon; Methodology, B.W. Blaser and L.I. Zon; Investigation, B.W. Blaser, J.L. Moore, E.J. Hagedorn, B. Li, V.B., O.T., R. Riquelme, A. Lichtig, S. Yang, and Y. Zhou; Writing (original draft), B.W. Blaser; Writing (Review) & Editing, B.W. Blaser and L.I. Zon; Funding Acquisition, B.W. Blaser and L.I. Zon; Resources, L.I. Zon; Supervision, L.I. Zon.