1. Developmental Biology
  2. Stem Cells and Regenerative Medicine

Defining the ultrastructure of the hematopoietic stem cell niche by correlative light and electron microscopy

  1. Sobhika Agarwala
  2. Keun-Young Kim
  3. Sebastien Phan
  4. Saeyeon Ju
  5. Ye Eun Kong
  6. Guillaume A Castillon
  7. Eric A Bushong
  8. Mark H Ellisman  Is a corresponding author
  9. Owen J Tamplin  Is a corresponding author
  1. Center for Stem Cell and Regenerative Medicine, Department of Pharmacology, College of Medicine, University of Illinois at Chicago, United States
  2. Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, United States
  3. Department of Neurosciences, University of California at San Diego School of Medicine, United States
https://doi.org/10.7554/eLife.64835
Share this article
Cite this article
  1. Sobhika Agarwala
  2. Keun-Young Kim
  3. Sebastien Phan
  4. Saeyeon Ju
  5. Ye Eun Kong
  6. Guillaume A Castillon
  7. Eric A Bushong
  8. Mark H Ellisman
  9. Owen J Tamplin
(2022)
Defining the ultrastructure of the hematopoietic stem cell niche by correlative light and electron microscopy
eLife 11:e64835.
https://doi.org/10.7554/eLife.64835
  2. Download BibTeX
  3. Download .RIS

Abstract

The blood system is supported by hematopoietic stem and progenitor cells (HSPCs) found in a specialized microenvironment called the niche. Many different niche cell types support HSPCs, however how they interact and their ultrastructure has been difficult to define. Here, we show that single endogenous HSPCs can be tracked by light microscopy, then identified by serial block-face scanning electron microscopy (SBEM) at multiscale levels. Using the zebrafish larval kidney marrow (KM) niche as a model, we followed single fluorescently labeled HSPCs by light sheet microscopy, then confirmed their exact location in a 3D SBEM dataset. We found a variety of different configurations of HSPCs and surrounding niche cells, suggesting there could be functional heterogeneity in sites of HSPC lodgement. Our approach also allowed us to identify dopamine beta-hydroxylase (dbh) positive ganglion cells as a previously uncharacterized functional cell type in the HSPC niche. By integrating multiple imaging modalities, we could resolve the ultrastructure of single rare cells deep in live tissue and define all contacts between an HSPC and its surrounding niche cell types.

Editor's evaluation

The manuscript reports on an extensive body of work, achieving the still highly challenging identification of HSPCs within the ultrastructure of their niche. The study highlights the heterogeneous nature of HSC-niche interactions, which is consistent with heterogeneity identified through genomic and functional studies. The work presented is of high interest to the field.

https://doi.org/10.7554/eLife.64835.sa0

Introduction

Hematopoietic stem and progenitor cells (HSPCs) give rise to all blood cell types throughout the life of an organism (Orkin and Zon, 2008). HSPCs reside in a complex microenvironment called the niche that is made up of many different kinds of support cells, including various types of mesenchymal stromal cells (MSCs) and endothelial cells (ECs) (Pinho and Frenette, 2019). However, our understanding of HSPC interactions with niche cells has been limited to low-resolution light microscopy. Recently developed transgenic reporter lines have allowed identification of well-defined endogenous HSPCs in both mouse and zebrafish model organisms (Acar et al., 2015; Chen et al., 2016; Christodoulou et al., 2020; Tamplin et al., 2015). The dynamic behavior of HSPCs in the niche of mouse and zebrafish can be observed using live imaging (Bixel et al., 2017; Christodoulou et al., 2020; Itkin et al., 2016; Koechlein et al., 2016; Lo Celso et al., 2009; Spencer et al., 2014; Tamplin et al., 2015). Yet, it remains challenging to observe HSPC-niche interactions at high resolution. Certain events during hematopoietic ontogeny, such as the colonization of the fetal bone marrow, have so far only been studied in fixed tissues because they are difficult to access (Coşkun et al., 2014). Our goal is to better define the ultrastructure of single endogenous HSPCs deep in niche tissue.

We have taken advantage of the transparency and external development of the zebrafish larva to study the earliest migration events of HSPCs into the presumptive adult kidney marrow (KM) niche. We considered using two different correlative light and electron microscopy (CLEM) techniques (Karreman et al., 2016) to resolve the ultrastructure of HSPCs. For the first approach, to confirm cell identity we used a label-based approach similar to what was done using an APEX2-Venus-CAAX fusion protein (Hirabayashi et al., 2018) that could be visualized both with light microscopy and high contrast in EM imaging. A similar method was applied in zebrafish using an APEX-GBP (GFP Binding Protein) fusion to resolve GFP+ transgene expression on electron micrographs (Ariotti et al., 2015). Building on these previous studies, we used genetically encoded APEX2 engineered peroxidase (Lam et al., 2015), together with a fluorescent protein, to track single HSPCs as they migrated into and lodged in the larval KM during the earliest colonization stages. For the second approach, we used improved software alignments to merge datasets across different imaging platforms and could correlate single cells without the need for endogenous labels. We found clusters of HSPCs around the glomerulus as previously described, as well as single HSPCs lodged in a perivascular niche. The heterogeneity between sites of HSPC lodgement and their surrounding niche cells supports the model that functionally distinct sites exist within the hematopoietic microenvironment (Zhang et al., 2021). These large 3D CLEM datasets also allowed us to identify previously uncharacterized dopamine beta-hydroxylase (dbh) positive ganglion-like cells in the larval kidney niche that were in direct contact with HSPCs.

Results

Colonization of the larval kidney niche by circulating HSPCs

To characterize the larval kidney HSPC niche in the zebrafish, we first wanted to determine the location of all HSPCs in the anterior KM. We used HSPC-specific Runx:mCherry+ transgenic zebrafish larvae (Tamplin et al., 2015) that were fixed at 5 days post fertilization (dpf) for anti-mCherry immunofluorescence, then optically cleared using benzyl alcohol/benzyl benzoate (BABB). This technique allowed visualization and quantification of all mCherry+ HSPCs in the larvae. We observed bilateral clusters of ~50 HSPCs, for a total of ~100 HSPCs, in the region of the anterior KM at 5 dpf (Figure 1—figure supplement 1A-D). During early HSPC colonization of the KM at 4 dpf we observed ~50 total HSPCs, which increased by ~100% from 4 to 5–6 dpf, and then again by ~50% from 5–6 to 7–8 dpf (Figure 1—figure supplement 1E). We performed phospho-histone H3 (PH3) antibody labeling to determine if HSPCs were highly proliferative. On average, we only found one to two mitotic HSPCs in the KM at 5 dpf (Figure 1—figure supplement 1F-J). Previous studies of hematopoietic cells in the larval KM at 7 dpf also showed low levels of proliferation (van Rooijen et al., 2009). These data suggest that the increase in HSPC numbers during the early stages of KM colonization results from arrival via circulation, that is HSPCs that originated in the caudal hematopoietic tissue (CHT) and dorsal aorta (Murayama et al., 2006), and not extensive proliferation of resident HSPCs.

To follow the dynamics of HSPC colonization in the KM niche, we performed time-lapse live imaging at 4 and 5 dpf. Although the early zebrafish larva is transparent, imaging live KM using point scanning confocal microscopy is challenging due to the relatively slow acquisition time and depth of the tissue. To rapidly capture HSPC colonization events throughout the entire depth of the larval KM, we performed light sheet fluorescence microscopy (Huisken and Stainier, 2009). Using this technique, we could rapidly acquire a Z stack through the entire KM in less than 30 s (>200 slices with 1 μm spacing). Consistent with our observations of fixed embryos, a depth-coded projection of light sheet time-lapse images showed bilateral Runx:mCherry+ HSPC clusters (‘green’ cells on left and ‘blue’ cells on right; Figure 1—figure supplement 2; Figure 1—video 1). Together with the cdh17:GFP reporter line for pronephric tubules (Zhou et al., 2010), we localized Runx:mCherry+ HSPCs within the anterior kidney region, mediolateral to the proximal pronephric tubules (Figure 1—figure supplement 3). To determine the location of these HSPC clusters relative to the vasculature, we imaged HSPC-specific Runx:mCherry together with the flk:ZsGreen vascular-specific transgenic reporter (Cross et al., 2003). HSPC clusters were located between lateral dorsal aortae and cardinal veins (Figure 1—figure supplement 4). Light sheet live imaging of the KM niche at 5 dpf allowed us to directly visualize the locations and dynamics of early HSPC niche colonization.

To observe the interaction between single HSPCs as they arrive in the larval KM niche, we imaged the Runx:GFP transgenic line together with flk:mCherry to label vessels (Figure 1—video 2). As previously described, Runx:GFP is a more restricted marker of HSPCs and is expressed in the cytoplasm, allowing cellular morphology to be resolved, while Runx:mCherry more broadly labels the progenitor pool and is only localized to the nucleus, making it more appropriate for quantifying single cells (Tamplin et al., 2015; Figure 1—figure supplement 5A). Upon arrival within the anterior kidney region, rare circulating Runx:GFP+ HSPCs were seen interacting with and lodging in the perivascular niche (Figure 1A). We resolved single lodged HSPCs surrounded by ECs in a pocket-like structure, as we observed previously in the zebrafish CHT and mouse fetal liver (Tamplin et al., 2015; Figure 1B).

Figure 1 with 7 supplements see all
Download asset Open asset
Single hematopoietic stem and progenitor cells (HSPCs) lodge in a perivascular region of the larval kidney niche.

(A) Snapshot of single optical sections (XY, XZ, YZ planes) from light sheet live image of a Runx:GFP;flk:mCherry double transgenic zebrafish larva. A single Runx:GFP+ HSPC (white arrowhead) is lodged in a perivascular region lateral to the dorsal aorta (DA). (B) Detail of optical sections (1 µm steps) through the single lodged Runx:GFP+ HSPC in (A). mCherry+ endothelial cells contact the HSPC and form a surrounding pocket. The +5 µm section is also shown in XZ and YZ planes. Abbreviations: DA, dorsal aorta; SeA, intersegmental artery; SeV, intersegmental vein; PCV, posterior cardinal vein; CCV, common cardinal vein; ISVs, intersegmental vessels; D, dorsal; V, ventral; A, anterior; P, posterior.

We also examined expression of the cd41:GFP HSPC reporter line in the region of the anterior KM and found substantial overlap with Runx:mCherry+ HSPCs (Figure 1—figure supplement 5A). The widespread labeling of the 5 dpf HSPC pool by the cd41:GFP reporter, and its cytoplasmic expression that reveals cellular morphology, makes it a valuable tool for further analysis of HSPCs in the larval KM. Furthermore, cd41:GFP+ HSPCs were visible in EC pockets, confirming that they had lodged in the KM niche (Figure 3—figure supplement 1A). Together, the Runx:mCherry, Runx:GFP, and cd41:GFP reporter lines show similar lodgement in the larval KM niche, and represent an array of valuable tools to interrogate the interaction of HSPCs with various niche cell types.

HSPCs in the larval KM niche form direct contacts with an MSC and multiple ECs

After characterizing the specific location of HSPCs in the larval KM niche, and their lodgement in EC pocket structures that are also formed in the CHT (Tamplin et al., 2015), we sought to identify the specific types of contacts that form between HSPCs and niche cells. We considered that tight junctions may form between HSPCs and ECs and would be marked by tight junction protein 1 (tjp1), the scaffolding protein also known as zonula occludens-1 (ZO-1) that has two orthologues in zebrafish (tjp1a and tjp1b) (Anderson et al., 1988; Stevenson et al., 1986). We injected Oregon Green dextran into the circulation of Runx:mCherry transgenic larvae to label the vessel lumen, followed by fixation and immunofluorescence with anti-ZO-1 and anti-mCherry antibodies. We observed expression of ZO-1 broadly on ECs, as well as localization between Runx:mCherry HSPCs and surrounding niche cells (Figure 2A and B). These data suggest that tight junctions form at the contact points between HSPCs and the niche.

Figure 2
Download asset Open asset
Hematopoietic stem and progenitor cells (HSPCs) lodged in the larval kidney niche make direct contacts with endothelial cells (ECs) and mesenchymal stromal cells (MSCs).

(A) Single optical section from confocal image of larval kidney (fixed) shows Runx:mCherry+ HSPCs (magenta) lodged in the perivascular niche. Oregon Green dye labels the vessel lumen. Blue dotted lines surround the dorsal aorta (DA) and red dotted lines surround the glomerulus (G). Tight junction protein is marked by zonula occludens-1 (ZO-1) (white). (B) High-resolution optical sections (0.5 µm steps) through the boxed regions in (A) show ZO-1+ contact points between mCherry+ HSPCs and the niche (yellow arrowheads). (C) Orthogonal slices (XY and YZ planes) from live light sheet 3D volume of larval kidney niche. Single cd41:GFP+ HSPCs (green) is in contact or in close proximity (yellow arrowhead) to cxcl12:DsRed2+ MSCs (magenta). The white dotted line represents the DA. (D) Quantification of distances measured between GFP+ HSPC and DsRed2+ MSCs shows ~60% of HSPCs are in contact with MSCs, and the remaining are within 9 µm. Numbers above the columns indicate the cell numbers counted in each group (from n=8 embryos). Abbreviations: D, dorsal; V, ventral; A, anterior; P, posterior.

We then went on to characterize additional niche cell types that are present in the larval KM. We performed imaging using the cd41:GFP HSPC transgenic reporter line together with cxcl12:DsRed2 (Glass et al., 2011) to label MSCs (Figure 2C). We measured the distance between HSPCs and MSCs and observed 57% (n=16/28) of HSPCs were in direct contact with an MSC, 29% were <5 μm (n=8/28), and 14% (n=4/28) were <10 μm away (Figure 2D). These data from the larval KM niche are very similar to previous observations from the CHT showing that the majority of HSPCs are in contact with, or close proximity to, an MSC (Tamplin et al., 2015). Together, our results demonstrate that HSPC lodgement in the larval KM niche occurs close to or in contact with a single MSC, and that HSPCs are in contact with multiple ECs, similar to what we observed previously in the CHT.

A CLEM approach to characterize the ultrastructure of HSPCs in the larval KM niche

Next we wanted to explore the ultrastructure of HSPCs and their surrounding cells in the larval KM niche. One approach to analyze the ultrastructure of an entire region of tissue is serial section electron microscopy (EM), a technique that has been used to resolve the projectome of the complete zebrafish larval brain (Hildebrand et al., 2017). We previously used CLEM based on anatomical landmarks to match the position of fluorescently labeled HSPCs in the CHT niche between confocal and SBEM datasets (Tamplin et al., 2015). However, we found this approach was difficult to apply in the KM because the tissue is much larger and denser than the CHT (Cell Image Library [CIL] accession numbers CIL:54845 and CIL:54850). To confirm precise correlation of single cells between light and EM imaging modalities, we developed two distinct approaches depending on the goals of the experiment. In Workflow #1, we wanted to track lodgement of HSPCs in the larval KM niche using time-lapse live imaging, followed by high-contrast DAB (3,3'-diaminobenzidine) staining with genetically encoded APEX2 to label single endogenous HSPCs in EM sections. In Workflow #2, we wanted to correlate the position of fluorescently labeled cells in existing transgenic reporter lines across confocal and EM datasets. Using these workflows, we were able to observe HSPC lodgement in the KM, then resolve the ultrastructure of those same cells together with the surrounding niche.

The first step in Workflow #1 was to generate a transgenic construct that expressed mCherry for light microscopy, and APEX2 as a genetic tag that allowed electron-dense contrast on target subcellular structures (Figure 3A). Our rationale was that mCherry+ HSPCs are also APEX2+ and will be identifiable by both fluorescence imaging and EM, respectively. We fused APEX2 with H2B and mito tags for localization to the nucleus and the mitochondrial matrix, respectively. To drive expression of this construct in HSPCs, we cloned these elements under control of the draculin (drl) promoter, generating the drl:mito-APEX2_p2A_APEX2-H2B_p2A_mCherry transgene, hereafter referred to as drl:APEX2-mCherry. The drl promoter is a marker of vascular and hematopoietic lineages (Herbomel et al., 1999; Mosimann et al., 2015), and we chose it because of its high HSPC expression level compared to other available promoters, such as Runx1+23 (Tamplin et al., 2015; Nottingham et al., 2007) or cd41 (Ma et al., 2011). Furthermore, it was previously confirmed that drl:GFP+ HSPCs almost completely overlap with Runx:mCherry+ HSPCs from embryo to adult (Henninger et al., 2017; Mosimann et al., 2015).

Figure 3 with 4 supplements see all
Download asset Open asset
Correlative light and electron microscopy (CLEM) Workflow #1 to genetically encode a label in endogenous hematopoietic stem and progenitor cells (HSPCs) for live tracking by light microscopy and high-contrast resolution in serial block-face scanning electron microscopy (SBEM) sections.

(A) Fusion construct encoding p2A-linked proteins mito-APEX2, APEX2-H2B, and mCherry that localize to the mitochondria, nucleus, and cytoplasm, respectively. The draculin promoter was used to transiently drive strong mosaic expression in HSPCs. Random insertion in the genome was by Tol2-mediated transgenesis. (B) Tol2 draculin:mito-APEX2_p2A_APEX2-H2B_p2A_mCherry (drl:APEX2-mCherry) fusion construct was injected together with tol2 mRNA in one cell wild type zebrafish embryos. (C) At 5 days post fertilization (dpf), embryos with circulating mCherry+ HSPCs were visually screened and retro-orbitally injected with alpha bungarotoxin to paralyze the embryo, and Oregon Green dye to label the vasculature. (D) Dye-injected mCherry+ double positive embryos were visually screened and used for light sheet microscopy (example shows a 439 × 439 × 115 µm3 volume of the anterior kidney marrow (KM); ISVs, intersegmental vessels; yellow dotted line, DA, dorsal aorta; gut AF, gut autofluorescence). (E) Brightfield example of a single embryo after fixation and DAB (3,3'-diaminobenzidine) staining to label APEX2+ HSPCs that are located within the dotted box (dotted line marks DA, dorsal aorta; Y, yolk; D, dorsal; V, ventral; A, anterior; P, posterior). (F) After embedding, the sample was oriented and trimmed based on images acquired using micro-computed tomography (microCT) (example shows orthogonal sections in three planes, N; notochord, G; glomerulus, S; swim bladder). (G) Single plane from ~3000 sections of SBEM data (example shows a 233 × 331 × 213 µm3 volume; s1-s5, somites 1–5; G, glomerulus; PD, pneumatic duct).

To track and correlate single cells through multiple imaging modalities, we required sparse labeling of HSPCs. Therefore, we generated transient F0 transgenics with a mosaically labeled HSPC pool. Although a caveat of F0 mosaic transgenics in zebrafish is the inherent variability of labeling between embryos, this allowed us to select embryos that had similar numbers of lodged HSPCs in the KM niche compared to other HSPC reporter lines. We observed only one to two rare HSPCs surrounded by EC pockets in the KM niche of each cd41:GFP or Runx:GFP transgenic larvae, and selected F0 drl:APEX2-mCherry larvae with similar HSPC numbers (Figures 1 and 3, Figure 3—figure supplement 1). Both cd41:GFP+ and selected F0 drl:APEX2+ larvae had similar numbers of positive HSPCs in the mediolateral clusters of the anterior KM niche (Figure 3—figure supplement 1B). We gained further confirmation of HSPC-specific expression in F0 drl:APEX2-mCherry+ larvae by injection of the drl:APEX2-mCherry construct into cd41:GFP transgenic embryos, and observed ~15% of HSPCs were both mCherry+;APEX2+ and GFP+ (data not shown). These data demonstrate that transient expression of the drl:APEX2-mCherry construct in F0 larvae can generate sparse labeling of HSPCs in the KM niche with similar frequency to other stable HSPC-specific reporter lines.

To generate larvae for CLEM Workflow #1, we injected drl:APEX2-mCherry construct together with tol2 transposase into single-cell stage wild type zebrafish embryos (Figure 3B). At 5 dpf, larvae with circulating mCherry+ HSPCs were injected retro-orbitally with alpha-bungarotoxin and dextran-conjugated Oregon Green dye, to paralyze the larvae and label the vasculature, respectively (Figure 3C). Once immobilized and mounted for light sheet live imaging, optical sections were acquired through the entire depth of the KM. A short time-lapse was performed to confirm an HSPC was lodged in the KM and not circulating (~30 min; Figure 3—video 1). Single lodged HSPCs could be identified relative to the surrounding tissues (Figure 3D), and larvae were immediately fixed after imaging. Larvae were stained with DAB to label APEX2+ cells for identification by brightfield microscopy (Figure 3E). Larvae were treated with osmium tetroxide and embedded for micro-computed tomography (microCT; Figure 3F and Figure 3—figure supplement 2; Figure 3—video 2). This intermediate microCT step allowed the larvae to be oriented and trimmed to select a discrete region of interest (ROI) for SBEM (Figure 3G and Figure 3—figure supplement 2). Last, automated SBEM generated over 3000 high-resolution sections of the ROI (e.g., XY = 10 nm/pixel, Z=70 nm/pixel; Figure 3—figure supplement 2; Figure 4—video 1). Using focal charge compensation (Deerinck et al., 2018) which effectively eliminates specimen charging, we obtained a high-resolution SBEM dataset without the need for excessive post-processing alignment. Together, these experimental steps in Workflow #1 allowed us to label an endogenous HSPC that was tracked live, then stained for high-contrast detection in a large SBEM dataset.

To correlate the position of a single labeled HSPC across multiple imaging modalities, we performed 3D software alignment of both light sheet and SBEM datasets. First, we identified anatomical features as landmarks in the SBEM data, such as the somites, glomerulus, and pronephric tubules (Figure 4—figure supplement 1). We also observed clustered hematopoietic cells around the glomerulus in the same region as seen by light sheet microscopy (compare Figure 1—figure supplement 3 and Figure 4—figure supplement 2). We merged 3D rendered light sheet and SBEM datasets using image analysis software (Imaris) and aligned matching anatomical features in all three planes (Figure 4A–C and Figure 4—figure supplement 3). By performing these 3D alignments, we could locate a single APEX2+ cell in the SBEM dataset that was <5 μm from the corresponding mCherry+ HSPC imaged in the light sheet volume (Figure 4D). Furthermore, the APEX2+ cell had dark nuclear staining with much higher contrast than any of the surrounding cells, confirming we had identified the same APEX2+;mCherry+ HSPC across multiple imaging modalities (Figure 4D). By correlating 3D light sheet and SBEM data, we could confirm lodgement of a single HSPC in the larval KM niche, allowing us to further define the ultrastructure of this rare cell relative to its surrounding cells in the niche.

Figure 4 with 4 supplements see all
Download asset Open asset
3D alignment of light sheet and serial block-face scanning electron microscopy (SBEM) datasets localizes a single rare hematopoietic stem and progenitor cell (HSPC) across multiple imaging modalities.

(A) Single Z plane from light sheet imaging of drl:APEX2-mCherry+ transgenic larva showing the lodged mCherry+ HSPC (white arrowhead). (B) Global alignment of 3D rendered models generated from light sheet and SBEM datasets using Imaris software. (C) Orthogonal views of the white boxed region within B shows a 3D view of the alignment between light sheet and SBEM datasets. White arrowhead points to the single lodged HSPC in the aligned light sheet and SBEM datasets. (C’) APEX2+ HSPC in SBEM data. (C’’) mCherry+ HSPC in light sheet data. Green: Injected Oregon Green dextran dye marking vessels. Magenta: Runx:mCherry+ HSPCs and autofluorescence in gut. (D) Detail of the alignment shows mCherry+ HSPC and APEX2+ HSPC are <5 µm apart (dotted white line). Abbreviations: ISVs, intersegmental vessels; D, dorsal; V, ventral; A, anterior; P, posterior.

3D modeling of SBEM data reveals all cells in contact with an endogenous HSPC

To reconstruct the spatial relationships of the single APEX2+ HSPC relative to its surrounding niche cells, we performed extensive tracing of cell membranes within the SBEM data using 3D modeling software (IMOD) (Kremer et al., 1996). ECs are generally elongated cells with a large nucleus and little cytoplasm, while MSCs can be distinguished by their granular cytoplasm and a nucleus that occupies three-fourths of the cell volume (Tamplin et al., 2015). This morphological analysis revealed that the lodged APEX2+;mCherry+ HSPC (Figure 5A) was enclosed in a pocket of five ECs, and attached to a single MSC, that all directly contact the surface of the HSPC (Figure 5B; Figure 5—video 1). This same configuration of cells was seen previously in the CHT (Tamplin et al., 2015), demonstrating that this cellular structure is also conserved in the larval KM niche. Within other sections of the SBEM dataset, we observed the clusters of hematopoietic cells posterior to the glomerulus that were also seen with light sheet imaging (compare Figure 1—figure supplement 3 and Figure 5—figure supplement 1).

Figure 5 with 4 supplements see all
Download asset Open asset
Hematopoietic stem and progenitor cells (HSPCs) lodge in a multicellular niche in the perivascular kidney marrow (KM).

The ultrastructure of a single APEX2+ HSPC (white arrow) and its surrounding niche cells are modeled using 3D SBEM (00:15 from Figure 4—video 1). (A) The APEX2+ HSPC is lodged in the perivascular KM niche. (Ai) Surrounding tissues are labeled; the HSPC is anterior to the pneumatic duct, dorsal to intestine and pancreas, and ventral to the somites and pronephric tubule. (Aii) Higher magnification shows the APEX2+ HSPC is only two-cell diameters from the vessel lumen (white area). (Aiii) Full resolution detail of the APEX2+ HSPC showing high-contrast labeling of the nucleus (APEX2-H2B), mitochondria (mito-APEX2; white arrowhead), and extracellular space dorsal to the cell. (B) (i–iii) SBEM sections at different levels through the APEX2+ HSPC (white arrows) as shown in the schematic (iv). The HSPC is simultaneously in contact with multiple niche cells: five endothelial cells (EC1–5), 1 mesenchymal stromal cell (MSC), and a ganglion-like (GL) cell. Two unlabeled APEX2 negative putative HSPCs were lodged in the same niche (HSPC2 and HSPC3). HSPC2 is attached to HSPC3, and the APEX2+ HSPC (Biii; asterisk). (C) 3D rendered models of the APEX2+ HSPC (solid green) in contact with niche cells. 3D contours are in the same colors as outlines in (B). The APEX2+ HSPC is directly contacted by: (i) five ECs; (ii) one MSC; (iii) one HSPC, and a chain of GL-like cells. (iv) The GL-like cell is part of a long continuous chain of similar cells that extends through the niche. Scale bars: 5 µm unless otherwise labeled. Abbreviations: D, dorsal; V, ventral; A, anterior; P, posterior.

A significant advantage of SBEM datasets is that they provide a complete 3D picture of the cellular composition of a tissue that is not dependent on prior knowledge of transgenic or immunolabeled markers. Known transgenic markers allowed us to characterize HSPC-EC (Figure 1) and HSPC-MSC (Figure 2C and D) interactions, but not discover novel HSPC-niche cellular interactions. Careful analysis of our SBEM dataset and the cells in direct contact with the APEX2+;mCherry+ HSPC identified ganglion-like cells in the larval KM (Figure 5B and C; Figure 5—video 1). The ability to move through all adjacent sections of the 3D SBEM dataset allowed us to follow the length of these ganglion-like cells and discover that it was part of a chain of at least eight morphologically similar cells that extended throughout the larval KM niche.

Further tracing of the neighboring cells in contact with the APEX2+;mCherry+ HSPC revealed two unlabeled cells with the distinctive morphology of putative HSPCs (i.e., scant cytoplasm, large round nucleus, ruffled membrane; HSPC2 and HSPC3; Figure 5B and C). These other two putative HSPCs were APEX2 negative, suggesting they were not progeny derived by division from the APEX2+;mCherry+ HSPC, and were more likely independent HSPC clones that had lodged in the same niche. All three HSPCs were in direct contact with the chain of ganglion-like cells that we found extends through the larval KM niche (Figure 5C; Figure 5—video 1). This 3D SBEM dataset allowed identification of all surrounding cells in contact with the APEX2+;mCherry+ HSPC, and strikingly showed that a single endogenous HSPC can be in direct physical contact with as many as eight other cells.

Finally, given this multicellular HSPC niche structure we observed, we reasoned that an unlabeled HSPC in an independent dataset should be identifiable based on location and morphology alone. We generated a second SBEM dataset that also had a single APEX2+;mCherry+ HSPC, however it was found in a vessel lumen in the larval KM niche and attached to the vessel wall (Figure 5—figure supplement 2). We searched the perivascular regions of the larval KM niche within this second SBEM dataset and found two unlabeled putative HSPCs. Not only did both putative HSPCs share a distinct morphology, but following 3D modeling of all surrounding niche cells in contact with the HSPCs, we found each one was also in its own pocket of five ECs, and attached to a single MSC, exactly as we had observed previously in the CHT (Tamplin et al., 2015), and with the APEX2+;mCherry+ labeled HSPC (compare Figure 5C and Figure 5—figure supplement 3). Furthermore, one of the unlabeled putative HSPCs was also in contact with a chain of ganglion-like cells (Figure 5—figure supplement 3). In summary, using just the anatomical location and distinctive morphology of an HSPC, we could reconstruct the 3D niche configuration of two more putative HSPCs in the larval KM niche.

A second CLEM approach to characterize fluorescently labeled cells in the larval KM niche

While the APEX2+;mCherry+ labeling and CLEM approach developed in Workflow #1 was effective for characterizing individual cells lodged in the larval KM niche, we also wanted an approach to characterize all cells in the region that were labeled by an existing fluorescent transgenic reporter line. Toward that goal, we developed Workflow #2 for CLEM analysis of GFP+ cells in thick sections of the larval kidney niche (Figure 6). Briefly, GFP+ transgenic larvae were fixed, sectioned using a vibratome, incubated with DRAQ5 fluorescent nuclear dye, and imaged by confocal microscopy. After imaging, DAB staining solution was added to the sections, followed by illumination to photooxidize DRAQ5. This step locally generates reactive oxygen species that trigger DAB polymerization and darkening of the nuclei (Ou et al., 2017). As in Workflow #1, samples were embedded, microCT was used to orient and trim the block, followed by SBEM. Confocal microscopy, microCT, and SBEM datasets are aligned in 3D using Imaris software.

Figure 6 with 2 supplements see all
Download asset Open asset
Correlative light and electron microscopy (CLEM) Workflow #2 to align all cd41:GFP+ hematopoietic stem and progenitor cells (HSPCs) in the larval kidney marrow (KM) niche.

Five days post fertilization (dpf) cd41:GFP+ HSPCs (green) and DRAQ5 nuclear dye (blue). (A-D) The same region of the KM niche is marked by a white dotted rectangle. (A) Confocal and brightfield image of thick vibratome section. (B) Confocal Z projection of thick vibratome section. (C) Aligned overlay of micro-computed tomography (microCT) and confocal data. (D) Aligned overlay of serial block-face scanning electron microscopy (SBEM) and confocal data (XY plane only). (E) Aligned overlay of SBEM and confocal data (XY, XZ, YZ planes). (F) Detail of single SBEM section with aligned overlay of cd41:GFP+ HSPCs. Anatomical features are labeled and color-coded. (G) Summary of processing steps used in Workflow #2. Abbreviations: D, dorsal; V, ventral; A, anterior; P, posterior.

Having established Workflow #2 to perform CLEM using existing GFP+ transgenic lines, we generated a dataset for the KM niche of a cd41:GFP+ larva. We chose the cd41:GFP+ transgenic line because of its consistent labeling of HSPCs in mediolateral clusters (Figure 1—figure supplement 5) and lodgement of HSPCs in EC pockets (Figure 3—figure supplement 1). This allowed us to characterize the different configurations of many individual HSPCs and their surrounding niche cells, as well as compare these data from an established HSPC-specific reporter line with our other datasets. After brief fixation we could still see a small cluster of cd41:GFP+ HSPCs in the region of the anterior KM niche (Figure 6A). Confocal imaging was performed on a thick 100 μm vibratome section after DRAQ5 staining (Figure 6B). After embedding, microCT data was used to orient the sample for SBEM. Confocal and microCT datasets were aligned based on general nuclear staining of the tissue (Figure 6C). After SBEM, Imaris software was used to align confocal and SBEM datasets in 3D (Figure 6D and E). Multiple cd41:GFP+ HSPCs in the larval KM niche were precisely aligned across these different imaging modalities at the single-cell level (Figure 6F).

To further confirm the 3D alignment of all cd41:GFP+ HSPCs across imaging platforms, we compared the position of predicted cells by software analysis of the confocal Z stack (using Imaris ‘Spots’ function), with segmentation of putative HSPCs based on morphology in the SBEM dataset (using Imaris manual ‘Surfaces’ tracing function). The software predicted 10 cd41:GFP+ HSPCs in the confocal Z stack, and all of those cells overlapped with traced models of the nearest putative HSPC in the SBEM dataset (Figure 6—figure supplement 1). Interestingly, extensive tracing of putative HSPCs in the SBEM dataset revealed an additional seven cells that were not well defined by their GFP signal in the confocal dataset (Figure 6—figure supplement 2). These putative HSPCs were all deeper in the sample, so were either GFP- or not detected because of low GFP signal.

All 17 putative HSPCs in the larval KM niche (GFP+, GFP-, or GFPlow) were classified based on their location (i.e., within a hematopoietic cluster, vessel lumen, or perivascular niche), and by their cellular contacts (i.e., EC, MSC, HSPC, etc.). These data were summarized together with all putative HSPCs found by SBEM in this study (n=22; Table 1). Regardless of their location, all putative HSPCs were in contact with between 2 and 6 ECs. Strikingly, 50% of putative HSPCs were in contact with a single MSC (n=11/22; Table 1), similar to what we found by light microscopy in the larval KM niche (60% cd41:GFP+ HSPCs in contact with cxcl12:DsRed2+ MSC; Figure 2), and previously in the CHT (60% Runx:GFP+ HSPCs in contact with cxcl12:DsRed2+ MSC; Tamplin et al., 2015). Our second CLEM workflow enabled analysis of all cd41:GFP+ HSPCs within an SBEM dataset, and allowed identification of common features shared by many putative HSPCs. Considering all our SBEM data together (Table 1), we have also identified considerable heterogeneity between the different structural configurations of a single HSPC and its surrounding niche cells. This suggests that there could be functional heterogeneity as well between different sites of hematopoiesis in the microenvironment. Our current SBEM approaches are not high throughput, so acquisition of more replicates in future studies will better inform the extent of structural heterogeneity between sites of HSPC lodgement in the niche.

Table 1
Summary of all putative hematopoietic stem and progenitor cells (HSPCs) in serial block-face scanning electron microscopy (SBEM) datasets, their locations, and niche cell contacts.

‘Cell Image Library (CIL)’ refers to the public database accession numbers. ‘SBEM dataset’ refers to one of the four datasets used in the study: APEX2 #1, APEX2 #2, cd41:GFP, and dbh:GFP. ‘Figure’ refers to the figure where the putative HSPC is shown in the manuscript. ‘Putative HSPC’ refers to the number of the HSPC as it is annotated within its respective figure. ‘Label’ refers to the endogenous genetic label for HSPCs in each sample. Label is not applicable (N/A) in the dbh:GFP sample because GFP marks the niche cells and not the HSPCs. ‘Color code’ refers to the outline, 3D model, or color overlay used to distinguish between different putative HSPCs in the SBEM dataset. ‘Location’ describes where the putative HSPC is located within the niche: perivascular lodgement, in a vessel lumen, or in the hematopoietic clusters adjacent to the glomerulus. Cell contact columns indicate how many individual niche cells (based on morphology, except in the case of dbh:GFP) are in direct contact with a single putative HSPC.

Cell Image Library (CIL)SBEMdatasetFigurePutativeHSPCLabelColor codeLocationEC contactMSC contactRBC contactHSPC contactGL contact
CIL:54847APEX2 #151APEX2+GreenPerivascular51021
CIL:54846APEX2 #25-S21APEX2+NoneLumen20000
CIL:54846APEX2 #25-S31APEX2-GreenPerivascular51001
CIL:54846APEX2 #25-S32APEX2-GreenPerivascular51000
CIL:54849cd41:GFP61GFP+RedCluster50000
CIL:54849cd41:GFP62GFP+OrangeLumen30300
CIL:54849cd41:GFP63GFP+YellowCluster51220
CIL:54849cd41:GFP64GFP+LimeCluster30100
CIL:54849cd41:GFP65GFP+Bright greenCluster51000
CIL:54849cd41:GFP66GFP+TealCluster30300
CIL:54849cd41:GFP67GFP+Light blueCluster50100
CIL:54849cd41:GFP68GFP+Dark blueLumen30100
CIL:54849cd41:GFP69GFP+MagentaLumen30200
CIL:54849cd41:GFP610GFP+Dark redCluster51110
CIL:54849cd41:GFP611GFP lowWhiteCluster21130
CIL:54849cd41:GFP612GFP lowWhiteCluster31100
CIL:54849cd41:GFP613GFP lowWhiteCluster60100
CIL:54849cd41:GFP614GFP lowWhiteCluster41100
CIL:54849cd41:GFP615GFP-WhiteCluster50110
CIL:54849cd41:GFP616GFP-WhiteCluster50200
CIL:54849cd41:GFP617GFP lowWhiteCluster51200
CIL:54848dbh:GFP71NoneMagentaPerivascular51011

  1. EC, endothelial cell; MSC, mesenchymal stromal cell; RBC, red blood cell; HSPC, hematopoietic stem and progenitor cell; GL, ganglion-like cell.

Ganglion-like cells discovered by SBEM are dbh positive niche cells

Our SBEM datasets allowed us to identify all cells in contact with putative HSPCs and discover uncharacterized ganglion-like cells in the larval KM niche. To determine the identity of the ganglion-like niche cells that were in contact with HSPCs (Figure 5 and Figure 5—figure supplement 3), we searched for candidate neuronal cells present in the kidney region at 5 dpf, and identified dbh positive ganglion cells to be the most likely candidate (An et al., 2002; Guo et al., 1999; Stewart et al., 2006; Stewart et al., 2004; Zhu et al., 2012). Previous studies have shown that dopamine signaling regulates HSPC function in the mammalian bone marrow (Afan et al., 1997; Katayama et al., 2006; Liu et al., 2021; Méndez-Ferrer et al., 2008). We crossed the dbh:GFP (Zhu et al., 2012) and Runx:mCherry transgenic lines, and observed dbh:GFP+ cell projections in contact with or in close proximity to mCherry+ HSPCs (Figure 7A and B; Figure 7—figure supplement 1A).

Figure 7 with 3 supplements see all
Download asset Open asset
Dopamine beta-hydroxylase positive ganglion-like cells are present within the larval kidney marrow niche.

(A) 3D rendering generated using light sheet movies of Runx:mCherry;dbh:GFP double transgenic larva shows mCherry+ clusters (dotted ovals) in close proximity to GFP+ cells. DA, dorsal aorta. (B) Oblique slice through the 3D volume shows GFP+ extensions from the dbh:GFP+ cells into the mCherry+ hematopoietic stem and progenitor cell (HSPC) clusters. (B’) Detail of the boxed region in B shows contact formation between the GFP+ extensions and mCherry+ HSPCs (white arrowheads) in all three planes. (C) Quantification showing significantly reduced number of Runx:mCherry+ HSPCs in 6-hydroxydopamine (6-OHDA)-treated transgenic larvae compared to DMSO controls. Unpaired t test with Welch’s correction. DMSO vs. 250 μM 6-OHDA, p=0.0003; DMSO vs. 300 μM 6-OHDA, p=0.0002. Sample size (n), DMSO, n=13; 250 μM 6-OHDA, n=18; 300 μM 6-OHDA, n=15. Number of biological replicates = 3. (D) The ultrastructure of a dbh:GFP+ cell (labeled green) in proximity of a cell with HSPC-like morphology (labeled magenta) (white arrowheads). Surrounding tissues are labeled, and the dbh:GFP+ cell HSPC pair is posterior to the glomerulus, dorsal to pneumatic duct, pronephric tubule and intestine, and ventral to the notochord. (D’) Higher magnification shows the dbh:GFP+ cell-HSPC pair is only three-cell diameters from the vessel lumen (white area). (D’’) Full resolution detail of the dbh:GFP+ cell-HSPC pair showing contact formation. Abbreviations: D, dorsal; V, ventral; A, anterior; P, posterior.

To understand the functional significance of dopamine signaling during KM niche colonization, we treated Runx:mCherry+ larvae with 6-hydroxydopamine (6-OHDA), a neurotoxin that induces lesions in dopaminergic neurons (Jackson-Lewis et al., 2012; Matsui and Sugie, 2017; Vijayanathan et al., 2017). We confirmed the efficacy of the drug treatment by reduced locomotor activity (Feng et al., 2014) of the larvae (data not shown). A previous study indicated that disrupting dopamine signaling during early stages of development (five somite stage to 30 hours post fertilization (hpf)) resulted in reduced HSPC numbers within the CHT at 48 hpf (Kwan et al., 2016). Since our aim was to examine the effect of disrupting dopamine signaling on HSPCs during KM niche colonization, we treated larvae with 6-OHDA from 4 to 5 dpf during KM niche colonization. We observed a significant reduction in the number of Runx:mCherry+ HSPCs within the niche after treatment with this dopaminergic cell neurotoxin (Figure 7C; Figure 7—figure supplement 1C, D).

Last, we validated the presence of dbh+ cells within the KM niche by confocal imaging of dbh:GFP+ transgenic larva, followed by SBEM, and then correlation of the two datasets using our CLEM approach to analyze any GFP+ transgenic line (Figure 6G; Workflow #2). dbh:GFP+ cells aligned with ganglion-like cells in the KM niche. We identified one ganglion-like cell in contact with a lodged cell that morphologically resembled an HSPC, with the caveat that this cell lacks an independent marker of HSPC identity (Figure 7D; Figure 7—figure supplement 1B; Figure 7—videos 1; 2). We have confirmed that dbh+ ganglion-like cells are present in the larval KM niche and are in direct contact with putative HSPCs. Furthermore, we used a small molecule approach to functionally test the role of dopamine signaling in HSPC colonization of the larval KM niche.

Discussion

We developed two novel CLEM workflows to study the ultrastructure of HSPCs in the larval KM niche. In the first (Workflow #1), we used a genetically encoded fluorescent reporter (i.e., mCherry) together with APEX2 that produces a high-contrast label in SBEM. An advantage of this approach was that we could label specific organelles with a dark stain on EM sections (e.g., mitochondria and nuclei). In a second approach (Workflow #2), we could use any GFP+ transgenic line to match the position of a single labeled cell across confocal and EM platforms. In all of the SBEM datasets, we could also identify putative unlabeled HSPCs based on their distinct size and morphology (i.e., 6–7 μm diameter, round, ruffled membrane, large nucleus, scant cytoplasm). In all datasets, lodged putative HSPCs made similar contacts with niche cells: two to six ECs, zero or one MSC, and sometimes a red blood cell, another HSPC, and/or a ganglion-like cell. Strikingly, all of these niche cells could simultaneously contact a single HSPC. This highlights the complexity of signals that an HSPC could receive, as well as the challenges ahead to resolve the functional significance of each contact on stem cell regulation.

In the zebrafish larva, dbh+ sympathetic ganglion cells are found in the larval KM region at 5 dpf (An et al., 2002; Stewart et al., 2006; Stewart et al., 2004; Zhu et al., 2012). We considered that these cells could be the ganglion-like cells that we found in direct contact with HSPCs. We used additional rounds of CLEM to confirm that dbh:GFP+ cells were adjacent to HSPCs in the perivascular KM niche. Chemical inhibition of dbh during KM colonization significantly reduced the number of lodged HSPCs. Our findings are consistent with the established role for neuronal regulation of HSPCs in zebrafish and mammals (Agarwala and Tamplin, 2018). Intriguingly, these dbh+ cells originate from the same neural crest lineage as a subtype of fetal bone marrow MSCs (An et al., 2002; Isern et al., 2014). Together, our results have demonstrated that CLEM is a viable approach to identify single rare stem cells deep in live tissue, and that 3D models of a stem cell in its niche built from SBEM data provides a complete picture of stem cell-niche interactions.

The identification and characterization of endogenous stem cells in complex tissues remains extremely challenging. Taking advantage of the transparent zebrafish larva and light sheet imaging, we found it possible to track a single putative HSPC deep in live tissue. We directly observed the larval KM niche during the earliest stages of HSPC colonization (4–5 dpf). This is comparable to HSPC seeding of the fetal bone marrow in mammals. Multiple HSPC-specific reporter lines (e.g., cd41:GFP, Runx:GFP, Runx:mCherry), as well as transient labeling using the drl promoter, all marked putative HSPCs in clusters adjacent to the glomerulus, lodged in perivascular regions, or in the lumen of vessels. Many of these perivascular HSPCs were found in pockets of ECs and adjacent or in contact with MSCs, similar to what we previously observed in the CHT (Tamplin et al., 2015). HSPCs found in a vessel lumen were in contact with two to three ECS, while those in clusters or the perivascular niche were in contact with three to six ECs (Table 1). We found that these HSPC-EC contacts were positive for the tight junction marker ZO-1. Interestingly, Drosophila equivalent occludin junctions have been found to regulate niche interactions with hematopoietic cells and germline stem cells (Fairchild et al., 2016; Khadilkar et al., 2017). Approximately 50% of putative HSPCs identified in SBEM datasets were in contact with a single MSC, while contact of HSPCs with other putative HSPCs and/or red blood cells was more variable (Table 1). Taking all our data together, we only found three examples of putative HSPCs in contact with ganglion-like cells (Table 1), and none of these were in HSPC clusters adjacent to the glomerulus. This suggests that there is heterogeneity between sites of HSPC lodgement in the niche, although more definitive conclusions about the degree of heterogeneity will depend on more replicate samples acquired in future studies.

A major outstanding question is if our observed structural heterogeneity between HSPC locations in the niche also indicates functional heterogeneity. This would result in differential regulation of HSPCs depending on the combination of surrounding niche cells. For example, the 1:1 attachment of an HSPC to an MSC that we observe for 50–60% of HSPCs could regulate quiescence vs. expansion. In our previous studies of the CHT, the MSC appeared to provide an anchor for symmetric vs. asymmetric HSPC divisions (Tamplin et al., 2015). The CHT of the zebrafish differs from the fetal liver because there is not the massive exponential expansion of HSPCs required to sustain the mammalian embryo in utero (Kumaravelu et al., 2002; Tamplin et al., 2015). Unlike mammalian embryos, zebrafish embryos develop externally and can survive for many days with no definitive hematopoiesis (Sood et al., 2010; Zhang et al., 2011). We do not know if the HSPCs that transit through the CHT to the KM have equal potential to become long-term quiescent hematopoietic stem cells, or if there is already heterogeneity within the HSPC pool. Some HSPCs may be fated to produce the blood lineages required during development, while others may retain their quiescence as they transit through the CHT to the KM. Although we do not have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible. Recent studies showed that subtypes of ECs within the vasculature of the mouse bone marrow microenvironment provide different regulatory functions (Zhang et al., 2021). It will be fascinating to unravel how heterogeneity of the HSPC pool interacts with heterogeneity of different sites within the niche, and ultimately decides the functional output of HSPCs.

Materials and methods

Animal care

Request a detailed protocol

Adult zebrafish (Danio rerio) were maintained at 28°C on a 14 hr:10 hr light:dark cycle. The Tg(Runx1+23:GFP) (Tamplin et al., 2015) transgenic line used in this study labels single rare HSPCs, and Tg(Runx1+23:mCherry) (Tamplin et al., 2015) and Tg(cd41:GFP) (Lin et al., 2005) label broader HSPC populations. Tg(flk:ZsGreen) (Cross et al., 2003) and Tg(flk:mCherry) (Chi et al., 2008) mark endothelial cells. Tg(cdh17:EGFP) (Zhou et al., 2010) labels the pronephric tubules. Tg(cxcl12/sdf-1a:DsRed2) labels stromal cells (Glass et al., 2011). Tg(dbh:EGFP) (Zhu et al., 2012) marks sympathetic neurons of superior cervical ganglion. Larval zebrafish were raised in Petri dishes containing E3 solution (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, and 0.33 mM MgSO4) until 5 dpf. All experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committees at the University of Illinois at Chicago (Protocol ACC 19-051) and the University of Wisconsin-Madison (Protocol M006348).

Plasmid construction

Request a detailed protocol

The mito-APEX2_p2A_APEX2-H2B_p2A_mCherry cassette was synthesized on a kanamycin-resistant pUC19 backbone (GenScript) and used as a middle entry vector for assembly using LR Clonase II Plus for Multisite Gateway Cloning (Invitrogen). The four plasmids used in the reaction were: (1) pCM293 –6.3 kbdraculin (drl) 5’ entry promoter construct (Herbomel et al., 1999; Mosimann et al., 2015); (2) synthesized middle entry vector pME_mito-APEX2_p2A_APEX2-H2B_p2A_mCherry (a.k.a. pME-APEX2-mCherry; Addgene #188944); (3) Tol2kit #302 3’ entry vector p3E-polyA (Kwan et al., 2007); (4) Tol2kit #394 destination vector pDestTol2pA2 (Kwan et al., 2007). The Gateway assembled LR reaction product, draculin:mito-APEX2_p2A_APEX2-H2B_p2A_mCherry_polyA plasmid (a.k.a. Tol2pA2-drl:APEX2-mCherry; Addgene #188945), was sequence-verified before injection into one-cell zebrafish embryos.

Microinjections

Request a detailed protocol

Transient transgenesis was established by injecting Tol2pA2-drl:APEX2-mCherry plasmid together with tol2 transposase into wild type zebrafish embryos at the single-cell stage. For light sheet imaging, 1 mg/ml of alpha-bungarotoxin (Swinburne et al., 2015) and 2.5% of Oregon Green dye (Dextran, Oregon Green 488; 70,000 MW, Invitrogen) were injected retro-orbitally into 5 dpf transgenic larvae to paralyze them and label their vessels, respectively.

Whole mount immunofluorescence

Request a detailed protocol

The 5 dpf Runx1+23:mCherry+ zebrafish larvae were fixed overnight in AB buffer containing 4% PFA in ×2 fix buffer (234 mM sucrose, 0.3 mM CaCl2, and ×1 PBS). Post fixation, the larvae were permeabilized with 0.5% Triton in ×1 PBS and washed in deionized water for 2.5 hr. Blocking was performed with 2% BSA in 0.5% Triton/PBS and rabbit polyclonal anti-mCherry primary antibody (Abcam) was used in 1:500 dilution overnight at 4°C. Donkey anti-rabbit Alexa Fluor 568 (Invitrogen) was used as a secondary antibody at 1:1000 dilution. For PH3 labeling, mouse monoclonal anti-PH3 (Ser10, clone 3H10) antibody (Millipore Sigma) was used in 1:250 dilution overnight at 4°C. Donkey anti-mouse Alexa Fluor 647 (Invitrogen) was used as a secondary antibody at 1:1000 dilution.

For ZO-1 antibody staining, the larvae were fixed in 4% PFA overnight followed by 3 PBST (×1 PBS with 0.1% Tween 20) rinses. Larvae were dehydrated in increasing methanol/PBST series (25%, 50%, 75%) and stored in 100% methanol at –20°C overnight followed by rehydration in the reversed methanol/PBST series (75%, 50%, 25%). After PBST rinses, they were blocked in 2% serum solution in PBDT (×1 PBS with 1% BSA, 1% DMSO, and 0.5% Triton) and then incubated in mouse monoclonal ZO-1 (clone 1A12) primary antibody (Thermo Fisher Scientific) in 1:500 dilution overnight at 4°C. Donkey anti-mouse Alexa Fluor 647 (Invitrogen) was used as a secondary antibody at 1:1000 dilution.

Drug treatment

Request a detailed protocol

Runx1+23:mCherry+ zebrafish larvae were treated at 4 dpf with 250 and 300 μM 6-OHDA (Cayman Chemical) for 24 hr. The drug was washed off at 5 dpf followed by fixation, tissue clearing, and imaging.

Tissue clearing and confocal imaging of KM

Request a detailed protocol

To image the KM in fixed samples, following antibody staining, the larvae were embedded in 1% low melting agarose in glass-bottomed dishes (35 mm, MatTek Corporation) with the dorsal surface in contact with the glass bottom. To improve visibility through the larvae, optical tissue clearing was performed according to the protocol (Dodt et al., 2007) with some modifications. Embedded samples were washed six to seven times with 100% methanol, followed by overnight incubation at 4°C to dehydrate the tissues. Next, the samples were washed six to seven times with a 1:1 ratio of 100% methanol and BABB clearing reagent (1 part benzyl alcohol:2 parts benzyl benzoate) and incubated for 30 min at room temperature before transferring to 100% BABB. These organic solvents have high refractive indices which result in the solvation of lipids within tissues to align their refractive indices to prevent scattering of light through the tissues (Dodt et al., 2007). Zeiss LSM 880 confocal microscope was used to image the KM. Volume dimensions of 354.25 × 354.25 × 74 µm3 was acquired every 2 µm (0.35 × 0.35 × 2 µm3/pixel) with a line sequential scan mode and averaging of 2.

For ZO-1 antibody staining, the stained larvae were sectioned in the sagittal plane through vibratome sectioning. Sections of 100 µm were generated that were imaged using Olympus FluoView 1000 at ×60 magnification.

Light sheet imaging

Request a detailed protocol

For light sheet imaging, stage-matched transgenic zebrafish larvae were paralyzed by retro-orbital injection of alpha-bungarotoxin (Swinburne et al., 2015) and embedded in 1% low melting point agarose within thin capillary tubes. The KM was illuminated with a light sheet from one axis, and fluorescence was detected at a perpendicular axis. Larvae were imaged with a ×20 objective using the Zeiss Light sheet Z.1 for Single Plane Illumination Microscopy (SPIM). The capillary tube was inserted into the sample chamber to release the larva into the chamber containing E3 while remaining attached to the capillary tube through the agarose layer. The larva was rotated such that the lateral surface of the larva closer to the edge of the agarose layer was perpendicular to the detection objective to optimize fluorescence detection. The laser path was aligned to illuminate the KM from one axis. Z stacks were acquired through the entire KM with 1 µm spacing in less than 1 min and time-lapse videos were recorded to track the circulating HSPCs entering the KM. Through a full 360° rotation of the larva relative to the detection lens, only about 5° were optimal for light emission from the sample and image acquisition of the KM. Some additional light sheet images were acquired using a Mizar Tilt microscope.

Image analysis

Request a detailed protocol

Image processing and rendering were done using Imaris.v.9 (Bitplane), IMOD (Kremer et al., 1996), and ImageJ/Fiji (Schindelin et al., 2012). Imaris was used to count the number of cells within the KM niche. Imaris ‘Spot’ module was used first to specify the ROI around the KM in confocal images followed by automated counting of the spots generated corresponding to the single HSPCs. The accuracy of the spots generated was further confirmed by using the manual mode of ‘Spot’ function. ‘Co-localization’ feature on Imaris was used to identify PH3/mCherry double positive cells. Distances between HSPCs and niche cells were measured in 3D light sheet image volumes using Imaris. In microCT datasets, ImageJ/Fiji was used to orient and define the ROI that was trimmed and then scanned by SBEM. The ‘Add image to’ function in Imaris was used to add a light sheet or confocal dataset to an already open SBEM dataset, followed by alignment of both datasets in three dimensions, first using gross anatomical markers (e.g., somites, vessels), then single cells and/or nuclei as reference. Orthogonal views through the two datasets in the XY, YZ, and XZ planes were used to confirm the alignment. After alignment of light microscopy and SBEM datasets, single putative HSPCs and their surrounding niche cells were manually traced in high-resolution SBEM datasets using IMOD software. Separate objects were defined by drawing contours around the plasma membrane of the target cell (and sometimes nuclei) as it moved slice by slice in the Z plane. Individual contours were meshed with imodmesh to reveal 3D reconstructions of the cells of interest. In this way, HSPCs, ECs, MSCs, GLs, and their nuclei were identified and represented in the 3D model. Fluorescent image levels and background were adjusted, using the default ‘background subtraction’ (Imaris), threshold adjustment, and/or brightness/contrast adjustment (Fiji/ImageJ, Imaris, Zeiss Zen, Adobe Photoshop). Imaris was used to define ‘Spots’ in the cd41:GFP+ CLEM dataset using default settings, except ‘Spot Detection’ was set to: Estimated XY diameter: 3.00 μm; model PSF-elongation along Z axis ‘ON’, estimated Z diameter: 6.00 μm; background subtraction ‘ON’. SBEM brightness/contrast adjustment was performed using IMOD and/or Fiji, Imaris, Adobe Photoshop.

Larvae preparation for APEX2+;mCherry+ CLEM (Workflow #1)

Request a detailed protocol

Five dpf zebrafish larvae were prepared for microCT and SBEM as previously described (Deerinck et al., 2010). Briefly, immediately after light sheet imaging, larvae were fixed in 2.5% glutaraldehyde and 4% paraformaldehyde in 0.15 M cacodylate buffer (CB, pH 7.4) at 4°C overnight. After removing the fixative, larvae were treated for 15 min with 20 mM glycine in 0.15 M CB, on ice to quench the unreacted glutaraldehyde. Preincubation was done in 2.5 mM DAB solution (25.24 mM stock in 0.1 M HCl) in 0.15 M CB for 1 hr on ice. For staining, 0.03% H2O2 containing DAB solution was added to the larvae on ice. The reaction was monitored every 5–10 min and stopped when the desired intensity was achieved. The staining buffer was washed off with ×5 washes using 0.15 M CB. Then, larvae were washed with 0.15 M CB and then placed into 2% OsO4/1.5% potassium ferrocyanide in 0.15 M CB containing 2 mM CaCl2. The larvae were left for 30 min on ice and then 30 min at room temperature (RT). After thorough washing in double distilled water (ddH2O), larvae were placed into 0.05% thiocarbohydrazide for 30 min. Larvae were again washed and then stained with 2% aqueous OsO4 for 30 min. Larvae were washed and then placed into 2% aqueous uranyl acetate overnight at 4°C. Larvae were washed with ddH2O at RT and then stained with 0.05% en bloc lead aspartate for 30 min at 60°C. Larvae were washed with ddH2O and then dehydrated on ice in 50%, 70%, 90%, 100%, 100% ethanol solutions for 10 min at each step. Larvae were then washed twice with dry acetone and placed into 50:50 Durcupan ACM:acetone overnight. Larvae were transferred to 100% Durcupan resin overnight. Larvae were then flat embedded between glass slides coated with mould-release compound and left in an oven at 60°C for 72 hr.

Larvae preparation for GFP+ CLEM (Workflow #2)

Request a detailed protocol

To facilitate the 3D correlation between different imaging modalities when using any GFP+ transgenic line, we added DRAQ5, a nuclear DNA binding fluorescent dye. Briefly, the 5 dpf transgenic zebrafish larvae were fixed with 0.5% glutaraldehyde and 2% paraformaldehyde in 0.15 M cacodylate buffer (CB, pH 7.4) for 2.5 hr and 100 μm thick sagittal sections were collected and incubated in DRAQ5 (1:1000, Cell Signaling Technology) on ice for an hour. Confocal images of GFP and DRAQ5 signals were collected on a Leica SPE II confocal microscope with a 20x oil-immersion objective lens using 488 and 633 nm excitation. Following, the samples were preincubated in 2.5 mM DAB solution (25.24 mM stock in 0.1 M HCl) in 0.15 M CB for 30 min on ice. Next, the photo-oxidation of DAB by DRAQ5 was done in 2.5 mM DAB in 0.15 M CB using a solar simulator (Spectra-Physics 92191-1000 solar simulator with 1600 W mercury arc lamp and two Spectra-Physics SP66239-3767 dichroic mirrors to remove infrared and ultraviolet wavelengths), while bubbling oxygen in the solution. The light was filtered through a 10 cm square bandpass filters (Chroma Technology Corp.) for illumination at 615 nm (40 nm band pass). The reaction was monitored every 20 min and stopped when the desired darkening in the nuclei was achieved. Larvae were then washed five times with 0.15 M CB and incubated in 2% OsO4/1.5% potassium ferrocyanide in 0.15 M CB containing 2 mM CaCl2 to get an EM visible stain. The larvae were left for 30 min on ice and then 30 min at RT. After thorough washing in ddH2O, larvae were placed into 0.05% thiocarbohydrazide for 30 min. Larvae were again washed and then stained with 2% aqueous OsO4 for 30 min. Larvae were washed and then placed into 2% aqueous uranyl acetate overnight at 4°C. Larvae were washed with ddH2O at RT and then stained with 0.05% en bloc lead aspartate for 30 min at 60°C. Larvae were washed with ddH2O and then dehydrated on ice in 50%, 70%, 90%, 100%, 100% ethanol solutions for 10 min at each step. Larvae were then washed twice with dry acetone and placed into 50:50 Durcupan ACM:acetone overnight. Larvae were transferred to 100% Durcupan resin overnight. Larvae were then flat embedded between glass slides coated with mould-release compound and left in an oven at 60°C for 72 hr.

MicroCT and SBEM

Request a detailed protocol

The microCT tilt series were collected using a Zeiss Xradia 510 Versa (Zeiss X-Ray Microscopy) operated at 80 kV (88 µA current) with a 20x magnification and 0.872 µm pixel size. MicroCT volumes were generated from a tilt series of 2401 projections using XMReconstructor (Xradia). SBEM were accomplished using Merlin SEM (Zeiss, Oberkochen, Germany) equipped with a Gatan 3View system and a focal nitrogen gas injection setup (Deerinck et al., 2018). This system allowed the application of nitrogen gas precisely over the blockface of ROI during imaging with high vacuum to maximize the SEM image resolution. Even a minimal accumulation of charge on the specimen surface resulted in poor image quality and distortions in the non-conductive biological samples. This could happen through 3000 sequential images of the charge-prone zebrafish blockface due to image jitter. Deerinck et al. recently introduced a new approach to SBEM that uses focal nitrogen gas injection over the sample block surface to eliminate the charging, while allowing the high vacuum to be maintained in the specimen chamber, resulting in a tremendous improvement in image resolution, even with specimens that were not intensely heavy-metal stained (Deerinck et al., 2018). Images were acquired in 3 kV accelerating voltage and 0.5 µs dwell time; Z step size was 70 nm; raster size was 30k × 18k and Z dimension was ~3000 image samples. Volumes were collected using 40% nitrogen gas injection to samples under high vacuum. Once volumes were collected, the histograms for the slices throughout the volume stack were normalized to correct for drift in image intensity during acquisition. Digital micrograph files (.dm4) were normalized and then converted to MRC format. The stacks were converted to eight bit and volumes were manually traced for reconstruction using IMOD (Kremer et al., 1996).

Data availability

SBEM datasets have been deposited in the National Center for Microscopy and Imaging Research (NCMIR) publicly accessible resource database Cell Image Library (CIL). There are six SBEM datasets (accession numbers: CIL:54845, CIL:54846, CIL:54847, CIL:54848, CIL:54849, CIL:54850) that are accessible as group with the following link: http://cellimagelibrary.org/groups/54850. CIL accession numbers are referenced in Table 1. Newly generated plasmids have been deposited in Addgene (#188944 and #188945).

The following data sets were generated
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54845
    Transverse sections of 5 days post-fertilization wild-type zebrafish larva in the region of the anterior kidney.
    https://doi.org/10.7295/W9CIL54845
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54846
    Sagittal sections of 5 days post-fertilization drl:APEX2-mCherry+ zebrafish larva in the region of the anterior kidney (eLife Table 1: APEX2 #2 5 dpf).
    https://doi.org/10.7295/W9CIL54846
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54847
    Sagittal sections of 5 days post-fertilization drl:APEX2-mCherry+ zebrafish larva in the region of the anterior kidney (eLife Table 1: APEX2 #1 5 dpf).
    https://doi.org/10.7295/W9CIL54847
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54848
    Sagittal sections of 5 days post-fertilization dbh:gfp+ zebrafish larva in the region of the anterior kidney (eLife Table 1: dbh:GFP 5 dpf).
    https://doi.org/10.7295/W9CIL54848
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54849
    Sagittal sections of 5 days post-fertilization cd41:gfp+ zebrafish larva in the region of the anterior kidney (eLife Table 1: cd41:GFP 5 dpf).
    https://doi.org/10.7295/W9CIL54849
    1. Kim KY
    2. Agarwala S
    3. Phan S
    4. Ju S
    5. Kong YE
    6. Castillon GA
    7. Bushong EA
    8. Ellisman MH
    9. Tamplin OJ
    (2022) CIL:54850
    Transverse sections of 5 days post-fertilization wild-type zebrafish larva in the region of the anterior kidney.
    https://doi.org/10.7295/W9CIL54850

References

Decision letter

  1. Cristina Lo Celso
    Reviewing Editor; Imperial College London, United Kingdom
  2. Didier YR Stainier
    Senior Editor; Max Planck Institute for Heart and Lung Research, Germany
  3. Anne Schmidt
    Reviewer; CNRS UMR3738, Institut Pasteur, France

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Defining the ultrastructure of the hematopoietic stem cell niche by correlative light and electron microscopy" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Didier Stainier as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Anne Schmidt (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Three reviewers (including myself) have now carefully read and considered your manuscript, and we all agree that it presents a very interesting study and a technical tour de force establishing multimodal imaging of HSPC niches in the zebrafish kidney, nicely combining confocal microscopy, light sheet microscopy and serial block-face scanning electron microscopy. The results achieved are interesting but still overall preliminary and we agree that major revisions should be completed prior to final publication.

Essential revisions:

The main points of agreement are the following:

1) Multiple different lines of fish are used to highlight HSPCs, but since their signals do not overlap, it is hard to come to a definitive conclusion of what cells are analysed each time. It seems that Runx1:GFP would be the best line to use, but it is actually the least used. CD41:GFP is used in some analyses too and especially in the validation of the APEX transgenic line (though not shown).

2) Only two examples of HSCPs surrounded by the 7 cells niche are presented. Is this the case for all HSPCs, or what sort of variation is observed? Variation would be expected given the reporters used are marking HSPCs rather than primitive HSCs, and therefore a heterogeneous population. In general, more data should be presented and statistically analysed to support the conclusions presented.

3) APEX signal is not always shown, but should be used to unequivocally identify the cell presented.

4) In figure 2 ZO-1 does not look particularly enriched around Cherry positive cells, and the relative conclusions should be rephrased.

5) Scale bars are not always clear, and a number of figure panels are hard to read (see specific comments). The same applies to multiple videos.

6) One point to be better discussed is how the Authors reconcile the similar organisation of CHT and KM niches, when functionally they are very different, ie they support rapidly expanding and more quiescent HSPCs, respectively.

7) Is it the case that there are two anatomically and functionally distinct niches in the KM, the glomerular and perivascular one? The functional difference needs to be proven. One possibility would be to identify differential influence of dopamine on these two niches.Reviewer #1 (Recommendations for the authors):

ED1A. Could some contrast indicating the glomerulous and tubules be shown?

ED1I. Please show average and error bars rather than the summarised pie chart.

Video 1 is very hard to follow as cells and structures move significantly from one frame to the next.

Video 2. Perhaps the more novel finding is that most of the interactions are relatively short lived. This should be at least noted and commented on.

Figure 2C. Why CD41gfp and not Runx1:GFP? DsRed resolution is not sufficient to rule out that each HSPC is in touch with just one MSC.

Line 106. Most other HSPCs were within 1 cell diameter: 4 vs 8 are further.

The characterization of the new transgenic fish line needs to be more precise and better shown. How can there be 1 or 2 HSPC/EC pockets, but many more HSPC clusters? IS there a functional difference between the single cells in EC pockets and the cells in clusters?

It is explained that the APEX-cherry cells are identified as also GFP+ in CD41+ fish (why not Runx-GFP?), but there is no CD41-GFP in Figure 3. Are the APEX-cherry cells truly HSPCs?

Figure 5. How many HSPCs are observed in this kind of niche in how many fish? The reporters used are not very strict for primitive HSPCs and variability would be expected and should be shown. At the moment it cannot be claimed that 'HSPCs lodge in a highly ordered multicellular niche'.

Figure 7 and 8. APEX staining is missing but CLEM is equally completed. Is APEX necessary then?

Reviewer #2 (Recommendations for the authors):

The manuscript by Agarwala S et al., is a very nice and interesting study and a technical tour de force to establish CLEM as an approach to study the stem cell niche composition. The authors use a variety of genetic reporters in zebrafish and multimodal imaging, comprising confocal and light sheet microscopy, microCT and serial block-face scanning electron microscopy (SBEM) to study the initial stages of hematopoiesis in the developing kidney marrow. The results suggest that the approach is "doable" and can provide qualitative and descriptive information, therefore significantly adding to the field. Among the initial description provided by this technology is the identification of several cell types that might have niche function for hematopoietic stem cells (HSCs) in the kidney marrow, including endothelial cells, mesenchymal stromal cells (MSCs), sympathetic neurons and glial cells, resembling the results obtained by other groups in the mouse bone marrow. However, the data provided is essentially very nice images of individual HSCs or few cells that might well be representative examples, but a systematic study providing quantitative information and analysis to support the conclusions seems to be lacking. It is not clear how many zebrafish have been used in each experiment to obtain conclusions, as many images focus on a single HSPC. In fact, n numbers are only indicated in Figure 8, where 3 biological replicates have been studies. The conclusions would require additional experiments and statistical analysis of the results. Other comments are below:

1) The authors indicate that Runx:GFP is a better HSPC marker than Runx:mCherry. Yet the data with Runx:GFP are scarce. Runx:GFP is only used in Figure 1 and the less specific Runx:mCherry is used in the other main figures. Figure 1 also lacks quantification. How many HSPCs are surrounded by how many endothelial cells? Where are GFP+ cells located in relationship to other niche cells?

2) The authors provide two examples of HSPCs surrounded by 7 cell niche (5 ECs, 1 MSC, 1 Glial like cell), but we don't know if this the case for every HSPC or just a subset/what proportion.

3) Figure 7 importantly identifies putative HSPCs based on morphological criteria, but again only few examples are shown and confirmation that these are HSPCs using the reporter lines seems required.

4) In Figure 8D, HSPCs have been identified as 'morphologically resembling an HSPC'; some of the reporters available should be used to demonstrate that this is the case.

5) L. 101: "This suggests that ZO-1 is a potential candidate for mediating adhesion between HSPCs and the surrounding niche cells in the KM niche" and L. 244-245 of Discussion – Figure 2A does not show a particular enrichment of ZO-1 expression near mCherry+ cells.

6) L.226-7: "suggesting a role for dopamine signalling in colonization of the presumptive adult KM niche" should probably be rephrased. There could be different reasons and neurotransmitters/other molecules responsible for this effect.

7) Scale bars in some figures e.g 3D-G weren't always clear.Reviewer #3 (Recommendations for the authors):

The work presented by Owen Tamplin and collaborators is aimed at defining the different cellular components and chemical signals that provide structural and functional specificity to the niche into which hematopoietic stem cells and progenitors (HSPCs) are homing. These investigations are essential to determine precisely the niches that will favour the survival and maintenance of hematopoietic stem cells with the highest degree of stemness which, ultimately, should allow producing this type of cells in the context of regenerative medicine.

Importantly, this kind of work would have not been possible in any other vertebrates, which is why the developing zebrafish is a unique living model organism, owing to its transparency and relative easiness for genetic engineering.

Here, Owen Tamplin and collaborators succeed in providing impressive 3-dimensional information on the structural organization of pre-definitive niches in the developing kidney of the zebrafish larva. In analogy with what has been described in the mammalian bone marrow, they also provide functional evidence for a role of dopamine signalling via glial-like cells, which empowers the use of the zebrafish to comprehend the biology of immunity establishment, a process essential for life. Overall, this piece of work paves the way for further, more detailed analyses on the physical and fonctional interactions between hematopoietic stem cells and their 3D-niche environment.

Comments to the Authors

In this manuscript, the authors address, at high spatial 3-dimensional resolution that includes light sheet microscopy on the whole zebrafish larva and correlative light/electron microscopy, critical aspects of hematopoietic stem cells and progenitors (HSPCS) and their niches: their conserved structural and cellular identities as well as the regulatory function of contacting glial-like cells.

The work is overall quite impressive, ambitious technically, and provides unprecedented 3D reconstitution of the pre-definitive hematopoietic niche region, in the developmental zebrafish kidney (the future functional equivalent of the adult bone marrow in mammals). However, I think that the manuscript, in its present form, requires additional experimental work, in particular regarding physical contact between HSPCs (unambiguously identified) and glial-like cells (see comments on Figure 8 and text), which is the functional piece of work of the study. In addition, the manuscript is difficult to follow for the reader in particular for the figures that are, for many of them, lacking annotations (even with the support of Extended Data, the reader has to make serious efforts to be able to critically analyze the Figures (which is clearly a difficult task for authors that are publishing correlative images and EM analysis)).

Finally and conceptually, it would be good if the authors could propose a more precise vision in regard to the potential functional implications of the 2 geographical/structurally defined niches described here, i.e the one proximal to the glomerulus, and the more distant perivascular niche (this should be, on my opinion, included in the Discussion of the manuscript). Somehow, along the manuscript, we lose the track on the authors aim regarding the characterization of these 2 niche types/regions respective peculiarities (number of HSPCs versus other cell types, influence of neurotransmitters, etc …) and, in fine, the perivascular niche in which 'rare' cells are homing seems to be the centre of attention without any further and precise explanation.

Please find my major comments beneath on Figures and related text:

– Figure 3 and text

The information is not easily captured by the reader.

In general, panels D-G need annotations (I would suggest: DA and ISVs in D, the position of DAB stained regions in E, the main organs that we see in F (I guess we see the notochord and part of the swim bladder?), G: the position of somites (s1, s2, s3, s4; with somite 2 nearby the kidney marrow as shown in the very nice suppl. Video 4)).

More specific comments on specific panels:

D: I doubt that the large, pink, elongated structure is indeed HSPCs (could it be the auto-fluorescence (AF) from the gut as seen in Extended Data Figure 3? I guess this is the same for Figure 4, panel A)

E: the local DAB precipitate(s) is not clearly visible (could the authors make attempt to obtain an image before and after DAB staining, on the same larva (and show the comparison of the 2 of them?)).

For the accompanying Extended Data Figure 5, the legend of panel D says we see a reconstitution of HSPCs and multicellular niche (line 851), where are they? could the authors delineate the corresponding regions?

For the Supplementary Video 5, which is truly impressive, and beautiful in resolution, with plenty of information, what a pity not to have annotations to get a maximum benefit from it; the frustration is immense ! I propose that authors decompose it for the images that are the most in relation to HSPC clusters and the relevant surrounding organs (including vascular structures) and build an additional piece of Extended Data. I realize that one of the sections is composing Figure 5A (00:15 of the pile of images), which should appear in the legend to this Figure. In addition, are the panels Bi-iii of Figure 5 also extracted from Supplementary Video 5 ? In addition, for Video 5, should we expect to see the two different regions in which clusters and single (or 1-2) cells are supposed to establish 2 specific (and different) niches?

Finally, it appears that red blood cells (RBCs, which are nucleated in the zebrafish) have a very dark cytoplasm 00:11 for ex ? is it DAB staining (in the cytoplasm then and not in nuclei?) or is it endogenous cytosolic peroxydase activity which is revealed here?

In the text that relates to Figure 3, I find a potential inconsistency in the intention of having sparse labelling of HSPCs (line 131), which I understand is an obvious advantage if one wants to increase the chance to visualize single cells with no other labelled contacting hematopoietic cell (and hence using F0 larvae), but to have as an outcome the same result as when using stable transgenic line (see lines 139-140) is rather unexpected. Could the authors comment on that?

Also, do the 1-2 mCherry+ cells/EC pockets and 1-5 mCherry+/clusters were observed in each of the 11 larvae that were analyzed (owing to expected mosaicism, this may not be the case)?

– Figure 4 and text

Figure 4 is quite informative for comprehending the strategy followed technically but, somehow, is not illustrating the text lines 173-175 that makes the statement '… dark nuclear and mitochondrial staining … confirming that we have identified the same APEX2+/mCherry+ HSPC …'; thus, the authors should provide a magnification of that HSPC (from the EM image in panel Dii, with a clear nuclear and mitochondrial dark APEX2-derived staining).

– Figure 5 and text

I find an inconsistency saying that the posterior perivascular niche encloses a single HSPC as stated line 187 (or may be 2 maximum? as stated lines 137 and 140 if I understand correctly, which is also considered as a 'single rare' cell in Figure 6) (with the ratio 1xHSPC/5xEC/1xMSC) when the figure (see also the legend of Biii lines 709-710) shows two additional – non labelled – HSPCs (HSPCs 2 and 3); can the authors comment on that?

Supplementary Video 6 is impressive, but again (in particular for the planes 00:00 to 00:33), some annotations should be added on most relevant tissues/organs/cell types.

In addition, and on conceptual grounds, it is proposed that the CHT niche(s) is (are), in their vast majority, the site(s) of progenitors expansion, meaning most probably a very minor proportion would home cells bearing/maintaining full, long-term HSC potential (as opposed to the kidney marrow in which future adult HSCs are homing, thus with an expected higher proportion of HSC-specific niches). Based on this, how would the authors conceive that such conserved structures would accommodate such differentially fated cells (see the statement lines 187-188 'demonstrating this cellular structure is conserved between developmental and presumptive adult niche')?

– Figure 8 and text

The Figure with the light sheet information on close proximity between HSPCs and dbh cells is nice and the quantitative analysis panel (C) convincing.

However, the data should be substantiated using a double Tg line (draculin:2APEX2-mCherry) X (dbh:GFP) to confirm the identity of the HSPC(s) contacting the dbh cell(s) and not only a resemblance as state line 231 (hence labelled by the APEX2 nuclear + mitochondrial activities). This would very much substantiate the work and strengthen the manuscript on this important aspect of HSPC/niche establishment/regulation (and the authors have all the tools/approaches at hand). This is also required since the authors clearly state at the end of the Discussion (line 262) '… additional rounds of CLEM to confirm that dbh:GFP+ cells were adjacent to HSPCs …'; clearly, identifying unambiguously HSPCs is mandatory.

In the accompanying Extended Data Figure 9, panel (A) line 892, what do (n=10) refers to? 10 sections of 1 experiment? 1 section, each from 10 independent experiments? What cells does the blue colour in panel Bii underline?

In the text, the authors say that they wish to address a role of dopamine signalling in the KM niche colonization (hence they treat with 6-OHDA from 4 to 5dpf, during niche colonization); would a shorter treatment (ex: few hours) also affect HSPCs in the niche, which would indicate a function of dopamine in the niche per se (ex: survival, maintenance) rather than colonization?

Finally, the authors propose, if I understand correctly, that there may be 2 different types of functional niches (1) proximal/adjacent to the glomerulus, (2) more distant and referred to as posterior perivascular niche; do they have any information on a possible differential influence of dopamine and dbh-cells on these 2 niche types (or to their colonization)? Answering to that question would as well strengthen the functional impact of the work.

Comments on the Discussion:

The authors should, in my opinion, discuss the possible functional discrepancies between the 2 geographic positions of the niches that are visualized here, i.e the one adjacent to the glomerulus and the one that is perivascular (more distant from the glomerulus); they should do so also in the context of the current work on mammalian bone marrow niches. This is an important issue in the field because, ultimately, one wants to comprehend what are the functional differences between the niches that allow the maintenance of long-term HSCs and the ones that are more devoted to support specific lineage differentiation. Do the authors have a vision on that issue?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Defining the ultrastructure of the hematopoietic stem cell niche by correlative light and electron microscopy" for further consideration by eLife. Your revised article has been evaluated by Didier Stainier (Senior Editor) and a Reviewing Editor.

The manuscript has been improved and all reviewers commended the extra work and data added to the study. They also agree that it is important for the readers to include a few further edits to better reflect the limitations of the techniques used and the heterogeneity uncovered when a larger number of HSPC is analysed. However, no further experiments are needed. Please see below for details of their recommendations.

Reviewer #2 (Recommendations for the authors):

The manuscript describing this technically groundbreaking and very impressive approach to study the niche is largely improved and provides a more balanced interpretation of the results, considering the inherent limitation of the approach including datasets available.

I understand that the approach is technically very challenging and time-consuming, and would not allow to obtain larger datasets to perform quantitative analyses and obtain more definitive conclusions. Therefore, these should be rephrased to emphasise the qualitative (rather than quantitative) nature of the analysis and acknowledge some limitations (such as the lack of markers for putative HSCs in some cases). The contact with glial cells is emphasised but it should be noted that only 3 out of 22 HSPCs were found in contact with glial cells. Besides the numerical difference of HSPCs found in contact with different niche cells, whether HSPCs contacting these different niche cells might be functionally distinct could be incorporated into the discussion. The conclusions should acknowledge the qualitative nature of the study and emphasise the strengths and limitations of this novel approach to study the niche in the future, since it could potentially elicit significant interest in the field to use it in different experimental settings.

Reviewer #3 (Recommendations for the authors):

Many thanks to the authors to have provided a significant revision of their work; obviously, this correlative multimodal microscopy work is a tedious, difficult, time-consuming but essential task.

While many of my points have been taken into account, there still remains 2 critical aspects that I believe should be addressed and that are partly emphasized by the new dataset (dataset 4 using the CD41:gfp line, workflow 2).

Point 1 (that also relates to essential point 2 of the Evaluation Summary): The authors have provided a new dataset (see new Figure 6 and the 2 Supplemental Figures+ Table 1) that significantly increases the number of HSPCs that are analysed, including the identification of potential niche contacting cells (Table 1). However, the dataset raises the critical point of heterogeneity of the niche environment because the cluster of HSPCs that is analysed appears disconnected to any glial-like cells (GL, see outcome = 0 in Table 1 for the 17 HSPCs). Hence, this new dataset strengthens the idea of the heterogeneity which is not quantitatively appreciated in the work. Also, on qualitative aspects, it appears that the HSPC cluster is more distal to the glomeruli than the HSPCs illustrated in Figure 5 and Figure 5 – figure supplement 1 and 3 for example (which appear from the images to be more proximal to the glomerulus while for new Figure 6 the cluster appears to be connecting the pronephric tubule and the margin of the intestinal epithelium). Indeed, there may be different niches that may have been underappreciated in the work as it stands (niche peculiarities may depend on geo-localization, without considering the other, more posterior, perivascular niche).

Point 2 (that also relates to essential point 3 of the Evaluation Summary): The point on the APEX2 signal to unequivocally identify the HSPCs under study has not been addressed in the revision. This is important in particular for the data that illustrate the contact between dbh cells and putative HSPCs which is scarce (1 couple of cells in new Figure 7 (previous Figure 8, kept unchanged)); as it stands, this result is still preliminary (the HSPC is not identified unequivocally. The authors argue in their response that they did not manage to optimize the protocol so as to retain the GFP and mCherry signals but can't they take the peroxidase signal as reference (together with the gfp signal from dbh cells))?

[editor note: please reword the specific section to discuss other options and limitations]

Remark on essential Point 7: the answer to that point is well taken and understandable. In addition, since the strategy is to use mosaicism (hence reduced number of labelled HSPCs), this decreases the chance to capture the cells that immobilize in the perivascular, more distant niche (the HSPCS proximal to glomeruli appear to be more clusterized, increasing the chance to capture some of them). The authors have added in the discussion (lines 400-401 and not 392-394) that they are working on a labelling approach that would make it possible. I believe it is fair considering this point beyond the scope of this study.

https://doi.org/10.7554/eLife.64835.sa1

Author response

Essential Revisions:

The main points of agreement are the following:

1) Multiple different lines of fish are used to highlight HSPCs, but since their signals do not overlap, it is hard to come to a definitive conclusion of what cells are analysed each time. IT seems that Runx1:GFP would be the best line to use, but it is actually the least used. CD41:GFP is used in some analyses too and especially in the validation of the APEX transgenic line (though not shown).

We acknowledge we have used multiple different zebrafish HSPC reporter lines throughout this study. They are all well-established HSPC reporters that have varying degrees of overlap with each other. Different lines were selected for each experiment based on their color (GFP or mCherry) and compatibility with other reagents, and if they marked a large or small proportion of the HSPC pool. In fact, we are greatly encouraged that the different HSPC lines have all produced consistent results that confirm the main conclusions from our study.

Previously, we quantified and reported the overlap between these lines in the caudal hematopoietic tissue (CHT) at 72 hours post fertilization (hpf) (Owen J Tamplin et al., 2015). We have copied that information here: “Runx:GFP overlaps 92±11% with Runx:mCherry. Runx:mCherry overlaps 13±6% with Runx:GFP. (CHT of n=11 embryos scored)” and “cd41:GFP overlaps 60±12% with Runx:mCherry. Runx:mCherry overlaps 44±8% with cd41:GFP. (CHT of n=12 embryos scored).”

For mosaic HSPC expression we used the draculin (drl) promoter because it marks HSPCs and drives expression at a higher level than other available promoters (e.g., Runx+23 or cd41). Others have validated overlap between Runx:mCherry and drl:GFP in HSPCs (Henninger et al., 2017; Mosimann et al., 2015), and showed that virtually 100% of Runx:mCherry positive cells are also drl:GFP positive from embryo to adult. One caveat of this F0 mosaic transgenic approach is that there is inherent variability between embryos. This is what we wanted because it allowed us to select embryos that had positive cells lodged in the kidney that were also sparse for single cell correlation between imaging modalities. Therefore, we selected embryos that had similar cell numbers to those in cd41:GFP embryos (i.e., 1-2 in EC pockets and 1-5 in clusters; new Figure 3 —figure supplement 1B).

Overall, we observed excellent overlap between Runx:mCherry, Runx:GFP, cd41:GFP, and draculin promoter expression in the HSPC compartment that has also been validated in other studies. Together, these lines have demonstrated HSPCs in the larval kidney marrow niche of 5 dpf zebrafish are consistently found in three locations: (1) vessel lumen; (2) clusters adjacent to the glomerulus; (3) and lodged in perivascular sites (new Table 1). We have quantified all of the data from our light sheet experiments and consistently found there are only 1-2 HSPCs in pockets, regardless of which line is used (new Figure 3 —figure supplement 1B).

Regarding the suggestion that we use the Runx:GFP line for additional experiments, over the last 2 years our Runx:GFP line has undergone transgene silencing that is a common problem with tol2 transposase-generated zebrafish lines after multiple generations. We have been working to recover the line from frozen stocks, and regenerate the line using I-SceI meganuclease transgenesis to avoid this problem in the future, but we do not yet have a line available to perform new experiments.

2) Only two examples of HSCPs surrounded by the 7 cells niche are presented. Is this the case for all HSPCs, or what sort of variation is observed? Variation would be expected given the reporters used are marking HSPCs rather than primitive HSCs, and therefore a heterogeneous population. In general, more data should be presented and statistically analysed to support the conclusions presented.

In this revised manuscript we present more data to address the variation of HSPC lodgement in the niche of larval zebrafish at 5 dpf. We have generated one new serial section EM dataset (#4 cd41:GFP).

First, to provide context for our system, HSPC lodgement in the nascent kidney marrow of the larval zebrafish at 5 dpf is a very rare event. The predicted number of HSPCs produced from the dorsal aorta of the zebrafish embryo is considerably less (~20-fold) than mouse models. Instead of hundreds of cells produced from the dorsal aorta of the mouse embryo (Ganuza et al., 2017), in zebrafish it is less than 30 (Henninger et al., 2017). This makes tracking and quantification of many events very difficult, particularly at the early stages of larval kidney niche colonization when we are tracking the arrival of the very first cells. Our Runx:mCherry line has the broadest coverage of the HSPC pool and it shows there are only 50-100 positive cells in the entire kidney marrow niche between 4-5 dpf.

Another considerable challenge with this project is acquisition of serial section electron microscopy dataset (SBEM). From initial light imaging, to embedding, microCT scans, serial section scanning, post-acquisition data alignment and analysis, takes many months of work. This explains why some high impact studies using serial section EM only present a single dataset (Hildebrand et al., 2017). In the first submission of our study, we presented three serial section EM datasets: #1 drl:APEX2-mCherry; #2 drl:APEX2-mCherry; #3 dbh:GFP. For this resubmission we have produced a new serial section EM dataset: #4 cd41:GFP (new Figure 6). This new dataset using the cd41:GFP HSPC-specific reporter line has allowed us to analyze many lodged HSPCs in a single SBEM dataset. This has allowed us to make conclusions about common features of HSPC lodgement and confirm consistent observations with our previous datasets. Although each of the four SBEM datasets was generated to address slightly different questions, we did observe and quantify common features of HSPC lodgement in the niche of each sample. We have described niche interactions for a total of n=22 putative HSPCs collected from all SBEM datasets and summarized these results in new Table 1.

3) APEX signal is not always shown, but should be used to unequivocally identify the cell presented.

We acknowledge that the APEX2 staining approach was not always used to generate our serial section EM datasets. We did not clearly explain in the first version of the manuscript that we developed two distinct correlative light and electron microscopy (CLEM) workflows. Workflow #1 was dependent on having sparse but high levels of APEX2 expression in the HSPC compartment. For that purpose, we used the draculin promoter to express mCherry (for light microscopy) and APEX2 (for EM) from the same construct. This was injected transiently in F0 embryos because we wanted sparse mosaic labeling. APEX2 is a genetically encoded peroxidase that we tagged to nuclei and mitochondria (Lam et al., 2015), allowing us to match position across imaging modalities. The light and EM 3D datasets were merged and aligned using Imaris software.

Workflow #2 was a significant technical advance because it could be used with any GFP-expressing transgenic line. The embryos were treated with DRAQ5 nuclear stain before confocal imaging. Nuclear labeling provided the landmarks that were needed in both 3D light microscopy and EM datasets for software correlation and alignment. This second workflow was more rapid and robust in its alignments, and was used to confirm the location of cd41:GFP+ cells (new Figure 6) and dbh:GFP+ cells (Figure 7) in the niche. What we found so striking about the approach developed in Workflow #2 was the precision of alignments at the single cell level (new Figure 6 —figure supplements 1 and 2), and the ability to (as stated by the reviewer) unequivocally identify any cell between light microscopy and EM modalities.

Our results were consistent using the two different workflows and identified HSPCs in three locations (as mentioned above): (1) vessel lumen; (2) clusters adjacent to the glomerulus; (3) and lodged in perivascular sites (new Table 1).

4) In figure 2 ZO-1 does not look particularly enriched around Cherry positive cells, and the relative conclusions should be rephrased.

We have edited the text in Lines 135-138 to rephrase it as shown below. Since ZO-1 staining is present on all ECs, including those ECs that surround the HSPC, our observation suggests that it could be a potential candidate for mediating connections between ECs and HSPCs.

“We observed expression of ZO-1 broadly on ECs, as well as localization between Runx:mCherry HSPCs and surrounding niche cells (Figure 2A,B). These data suggest tight junctions form at the contact points between HSPCs and the niche.”

5) Scale bars are not always clear, and a number of figure panels are hard to read (see specific comments). The same applies to multiple videos.

We have made the requested changes to figure panels and the videos.

6) One point to be better discussed is how the Authors reconcile the similar organisation of CHT and KM niches, when functionally they are very different, ie they support rapidly expanding and more quiescent HSPCs, respectively.

We have addressed this excellent point in the Discussion section with the following paragraph:

“A major outstanding question is if there is functional heterogeneity within different niche sites of the CHT and KM. For example, the 1:1 attachment of an HSPC to an MSC that we observe for 50-60% of HSPCs could regulate quiescence versus expansion. There may also be subtypes of ECs within the vasculature of the CHT and KM that provide different regulatory functions, as is observed in the adult mouse bone marrow (J. Zhang et al., 2021). The CHT of the zebrafish differs from the fetal liver because there is not the massive exponential expansion of HSPCs required to sustain the mammalian embryo in utero (Kumaravelu et al., 2002; Owen J Tamplin et al., 2015). Unlike mammalian embryos, zebrafish embryos develop externally and can survive for many days with no definitive hematopoiesis (Sood et al., 2010; Y. Zhang, Jin, Li, Qin, and Wen, 2011). We do not know if the HSPCs that transit through the CHT to the KM have equal potential to become long term quiescent HSCs, or if there is already heterogeneity. Some HSPCs may be fated to produce the blood lineages required during development, while others may retain their quiescence as they transit through the CHT to the KM. Although we don’t have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible.”

7) Is it the case that there are two anatomically and functionally distinct niches in the KM, the glomerular and perivascular one? The functional difference needs to be proven. One possibility would be to identify differential influence of dopamine on these two niches.

This is a very interesting point however the genetic tools are not yet available to address this point. We do not have a single cell labeling strategy for HSPCs in the 5 dpf larval kidney that would allow us to mark and track cells in different locations. We have added the following lines to the Discussion section (392-394):

“Although we don’t have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible.”

Reviewer #1 (Recommendations for the authors):

ED1A. Could some contrast indicating the glomerulous and tubules be shown?

A new image has been added as Figure 1 —figure supplement 1A with a single z-section through the sample with better contrast for glomerulus and the tubules.

ED1I. Please show average and error bars rather than the summarised pie chart.

A plot showing average and error bars has been added and is the new Figure 1 —figure supplement 1J.

Video 1 is very hard to follow as cells and structures move significantly from one frame to the next.

We acknowledge there is lots of cell movement in this video because of circulating HSPCs, however the HSPCs in the mediolateral clusters are mostly static. We added this note to the Video 1 figure legend:

“We observed HSPCs that were rapidly moving in circulation, as well as those that were mostly static in the mediolateral clusters.”

Video 2. Perhaps the more novel finding is that most of the interactions are relatively short lived. This should be at least noted and commented on.

We added this note to the Video 2 legend:

“Occasionally, HSPCs slow and attached to vessels in the perivascular region of the larval KM niche.”

Figure 2C. Why CD41gfp and not Runx1:GFP? DsRed resolution is not sufficient to rule out that each HSPC is in touch with just one MSC.

A different image has been used in Figure 2C to show cxcl12:DsRed2+ stromal cell interaction with cd41:GFP+ HSPCs.

Line 106. Most other HSPCs were within 1 cell diameter: 4 vs 8 are further.

The text in Lines 140-142 has been edited to make this clearer:

“We measured the distance between HSPCs and MSCs and observed 57% (n=16/28) of HSPCs were in direct contact with an MSC, 29% were <5 µm (n=8/28), and 14% (n=4/28) were <10 µm away (Figure 2D).”

The characterization of the new transgenic fish line needs to be more precise and better shown. How can there be 1 or 2 HSPC/EC pockets, but many more HSPC clusters? IS there a functional difference between the single cells in EC pockets and the cells in clusters?

Regarding the F0 transient drl:APEX2-mCherry transgenics, we discussed this above (Essential Revision #1) and have copied our response here:

For mosaic HSPC expression we used the draculin (drl) promoter because it marks HSPCs and drives expression at a higher level than other available promoters (e.g., Runx+23 or cd41). Others have validated overlap between Runx:mCherry and drl:GFP in HSPCs (Henninger et al., 2017; Mosimann et al., 2015), and showed that virtually 100% of Runx:mCherry positive cells are also drl:GFP positive from embryo to adult. One caveat of this F0 mosaic transgenic approach is that there is inherent variability between embryos. This is what we wanted because it allowed us to select embryos that had positive cells lodged in the kidney that were also sparse for single cell correlation between imaging modalities. Therefore, we selected embryos that had similar cell numbers to those in cd41:GFP embryos (i.e., 1-2 in EC pockets and 1-5 in clusters; new Figure 3 —figure supplement 1B).

It is explained that the APEX-cherry cells are identified as also GFP+ in CD41+ fish (why not Runx-GFP?), but there is no CD41-GFP in Figure 3. Are the APEX-cherry cells truly HSPCs?

To verify that drl:APEX2-mCherry+ cells are HSPCs, we injected drl:APEX2-mCherry construct into cd41:GFP transgenics and we observed 15% overlap between mCherry+/GFP+ cells (Line 191).

Figure 5. How many HSPCs are observed in this kind of niche in how many fish? The reporters used are not very strict for primitive HSPCs and variability would be expected and should be shown. At the moment it cannot be claimed that 'HSPCs lodge in a highly ordered multicellular niche'.

We have described all the HSPC-niche interactions we have observed for n=22 putative HSPCs from 4 independent SBEM datasets (new Table 1). We have also reworded our conclusions as follows to describe a “multicellular niche”:

Line 264: Finally, given this multicellular HSPC niche structure we observed…

Figure 5 title: HSPCs lodge in a multicellular niche in the perivascular KM.

Figure 7 and 8. APEX staining is missing but CLEM is equally completed. Is APEX necessary then?

Our understanding is that the Reviewer is referring to Figures 6 and 7 in the previous version of the manuscript. These are now renamed as Figure 5 —figure supplement 2 and 3. These represent the same sample. We performed CLEM and aligned the APEX2+/mCherry+ HSPC. However, this HSPC was attached to a vessel wall but was not lodged in the posterior perivascular niche. Since HSPCs are generally rounded with a large nucleus, as shown previously (O. J. Tamplin et al., 2015). In Figure 5 —figure supplement 3, we looked at the SBEM data to identify other unlabeled putative HSPCs in the perivascular KM niche based on cellular morphology alone. The identification of two unlabeled putative HSPCs in the same anatomical location further confirmed that HSPCs lodge in the perivascular KM niche and interact with multiple niche cells.

In Essential Revision #3 above, we describe in detail our two CLEM workflows and how they are used to tackle different questions. Workflow #1 uses APEX2 and Workflow #2 uses any GFP+ transgenic line.

Reviewer #2 (Recommendations for the authors):

1) The authors indicate that Runx:GFP is a better HSPC marker than Runx:mCherry. Yet the data with Runx:GFP are scarce. Runx:GFP is only used in Figure 1 and the less specific Runx:mCherry is used in the other main figures. Figure 1 also lacks quantification. How many HSPCs are surrounded by how many endothelial cells? Where are GFP+ cells located in relationship to other niche cells?

We have now described all the HSPC-niche interactions we have observed for n=22 putative HSPCs from 4 independent SBEM datasets (new Table 1). In Essential Revision #1 above, we describe in detail the different applications for the various transgenic lines used in this study.

2) The authors provide two examples of HSPCs surrounded by 7 cell niche (5 ECs, 1 MSC, 1 Glial like cell), but we don't know if this the case for every HSPC or just a subset/what proportion.

See (1) above.

3) Figure 7 importantly identifies putative HSPCs based on morphological criteria, but again only few examples are shown and confirmation that these are HSPCs using the reporter lines seems required.

See (1) above.

4) In Figure 8D, HSPCs have been identified as 'morphologically resembling an HSPC'; some of the reporters available should be used to demonstrate that this is the case.

Using our Workflow #2, we validated the presence of dbH+ cells within the KM niche by confocal imaging of dbh:GFP+ transgenic larva, followed by SBEM, and then correlation of the two datasets using the DRAQ5 approach. We agree with the reviewer that it would have been ideal to perform this correlation using runx:mCherry+/dbh:GFP+ or APEX2:mCherry+/dbh:GFP+ double transgenics. However, we were not able to optimize the protocol in a way that we could retain both the GFP and mCherry signals and perform SBEM. Therefore, we proceeded with dbh:GFP+ transgenic alone. Instead, we present light microscopy data and quantification showing that most HSPCs are in contact or close proximity to dbh:GFP+ cells (Figure 7B and Figure 7 —figure supplement 1A).

5) L. 101: "This suggests that ZO-1 is a potential candidate for mediating adhesion between HSPCs and the surrounding niche cells in the KM niche" and L. 244-245 of Discussion – Figure 2A does not show a particular enrichment of ZO-1 expression near mCherry+ cells.

Copied from above (Essential Revision #4):

We have edited the text in Lines 135-138 to rephrase it as shown below. Since ZO-1 staining is present on all ECs, including those ECs that surround the HSPC, our observation suggests that it could be a potential candidate for mediating connections between ECs and HSPCs.

“We observed expression of ZO-1 broadly on ECs, as well as localization between Runx:mCherry HSPCs and surrounding niche cells (Figure 2A,B). These data suggest tight junctions form at the contact points between HSPCs and the niche.”

6) L.226-7: "suggesting a role for dopamine signalling in colonization of the presumptive adult KM niche" should probably be rephrased. There could be different reasons and neurotransmitters/other molecules responsible for this effect.

We have revised this statement as follows (Line 350-352):

“We observed a significant reduction in the number of Runx:mCherry+ HSPCs within the niche after treatment with this dopaminergic cell neurotoxin (Figure 7C; Figure 7 —figure supplement 1C,D).”

7) Scale bars in some figures e.g 3D-G weren't always clear.

The scale bars have been modified.

Reviewer #3 (Recommendations for the authors):

Please find my major comments beneath on Figures and related text:

– Figure 3 and text

The information is not easily captured by the reader.

In general, panels D-G need annotations (I would suggest: DA and ISVs in D, the position of DAB stained regions in E, the main organs that we see in F (I guess we see the notochord and part of the swim bladder?), G: the position of somites (s1, s2, s3, s4; with somite 2 nearby the kidney marrow as shown in the very nice suppl. Video 4)).

More specific comments on specific panels:

D: I doubt that the large, pink, elongated structure is indeed HSPCs (could it be the auto-fluorescence (AF) from the gut as seen in Extended Data Figure 3? I guess this is the same for Figure 4, panel A)

These are excellent suggestions and we have updated Figure 3D-G to include these details.

E: the local DAB precipitate(s) is not clearly visible (could the authors make attempt to obtain an image before and after DAB staining, on the same larva (and show the comparison of the 2 of them?)).

We do not have an example of a larva before DAB staining; however, we have selected an example at higher magnification with the region of the anterior KM niche labeled with a box (new Figure 3E).

For the accompanying Extended Data Figure 5, the legend of panel D says we see a reconstitution of HSPCs and multicellular niche (line 851), where are they? could the authors delineate the corresponding regions?

The data discussed previously was referring to new Figures 5 and 6. To address this comment, we have edited the figure legend for Figure 3 —figure supplement 2 as follows:

(D) Representative SBEM volume (dimensions: 31x22x30 µm).

For the Supplementary Video 5, which is truly impressive, and beautiful in resolution, with plenty of information, what a pity not to have annotations to get a maximum benefit from it; the frustration is immense! I propose that authors decompose it for the images that are the most in relation to HSPC clusters and the relevant surrounding organs (including vascular structures) and build an additional piece of Extended Data.

We have added a new Figure 5 —figure supplement 1 to show another region taken from a single frame of Video 5 (i.e., another section of the full SBEM dataset). We have also annotated the video to label landmark tissues. Upon publication the full resolution SBEM data will be made publicly available.

I realize that one of the sections is composing Figure 5A (00:15 of the pile of images), which should appear in the legend to this Figure.

The information has been added to the Figure 5 legend.

In addition, are the panels Bi-iii of Figure 5 also extracted from Supplementary Video 5 ? In addition, for Video 5, should we expect to see the two different regions in which clusters and single (or 1-2) cells are supposed to establish 2 specific (and different) niches?

Figure 5 panels Bi-iii are taken from the full resolution (9x9 nm/pixel) SBEM dataset. Video 5 was compiled from low resolution SBEM images (1/10 the original resolution) simply to provide the reader with an overview of the full dataset. New Figure 5 —figure supplement 1 shows a different section from the SBEM dataset that shows the hematopoietic cluster near the glomerulus.

Finally, it appears that red blood cells (RBCs, which are nucleated in the zebrafish) have a very dark cytoplasm 00:11 for ex ? is it DAB staining (in the cytoplasm then and not in nuclei?) or is it endogenous cytosolic peroxydase activity which is revealed here?

It is known that zebrafish erythrocytes have high endogenous peroxidase activity (Yamasaki and Nakayasu, 2003), and we believe our data shows these high cytosolic levels. To insure specificity of our genetically encoded APEX2 peroxidase, we tagged the nucleus and mitochondria, distinguishing our target cells from any other cells that may have dark contrast (Figure 5Aiii).

In the text that relates to Figure 3, I find a potential inconsistency in the intention of having sparse labelling of HSPCs (line 131), which I understand is an obvious advantage if one wants to increase the chance to visualize single cells with no other labelled contacting hematopoietic cell (and hence using F0 larvae), but to have as an outcome the same result as when using stable transgenic line (see lines 139-140) is rather unexpected. Could the authors comment on that?

This was addressed in Essential Revision #1 (copied below):

For mosaic HSPC expression we used the draculin (drl) promoter because it marks HSPCs and drives expression at a higher level than other available promoters (e.g., Runx+23 or cd41). Others have validated overlap between Runx:mCherry and drl:GFP in HSPCs (Henninger et al., 2017; Mosimann et al., 2015), and showed that virtually 100% of Runx:mCherry positive cells are also drl:GFP positive from embryo to adult. One caveat of this F0 mosaic transgenic approach is that there is inherent variability between embryos. This is what we wanted because it allowed us to select embryos that had positive cells lodged in the kidney that were also sparse for single cell correlation between imaging modalities. Therefore, we selected embryos that had similar cell numbers to those in cd41:gfp embryos (i.e., 1-2 in EC pockets and 1-5 in clusters; new Figure 3 —figure supplement 1B).

Also, do the 1-2 mCherry+ cells/EC pockets and 1-5 mCherry+/clusters were observed in each of the 11 larvae that were analyzed (owing to expected mosaicism, this may not be the case)?

As described above, there was inherent variability between embryos, but we intentionally selected F0 drl:APEX2-mCherry embryos that had HSPC labeling that was similar to cd41:GFP stable lines (see new Figure 3 —figure supplement 1B).

– Figure 4 and text

Figure 4 is quite informative for comprehending the strategy followed technically but, somehow, is not illustrating the text lines 173-175 that makes the statement '… dark nuclear and mitochondrial staining … confirming that we have identified the same APEX2+/mCherry+ HSPC …'; thus, the authors should provide a magnification of that HSPC (from the EM image in panel Dii, with a clear nuclear and mitochondrial dark APEX2-derived staining).

We revised the statement to make it clear that Figure 4D only the dark nuclei is visible (Lines 224-227):

“Furthermore, the APEX2+ cell had dark nuclear staining with much higher contrast than any of the surrounding cells, confirming we had identified the same APEX2+;mCherry+ HSPC across multiple imaging modalities (Figure 4D).”

The data requested is presented in Figure 5 Aii and Aiii.

– Figure 5 and text

I find an inconsistency saying that the posterior perivascular niche encloses a single HSPC as stated line 187 (or may be 2 maximum ? as stated lines 137 and 140 if I understand correctly, which is also considered as a 'single rare' cell in Figure 6 ) (with the ratio 1xHSPC/5xEC/1xMSC) when the figure (see also the legend of Biii lines 709-710) shows two additional – non labelled – HSPCs (HSPCs 2 and 3); can the authors comment on that ?

Thank you for the comment. We understand clarification is necessary. We believe this relates to Essential Revision #2 and have copied the response below:

We have described niche interactions for a total of n=22 putative HSPCs collected from all SBEM datasets and summarized these results in new Table 1.

Supplementary Video 6 is impressive, but again (in particular for the planes 00:00 to 00:33), some annotations should be added on most relevant tissues/organs/cell types.

We added annotations to Video 6 at 00:05 (the frame that corresponds to Figure 5) to outline relevant tissues.

In addition, and on conceptual grounds, it is proposed that the CHT niche(s) is (are), in their vast majority, the site(s) of progenitors expansion, meaning most probably a very minor proportion would home cells bearing/maintaining full, long-term HSC potential (as opposed to the kidney marrow in which future adult HSCs are homing, thus with an expected higher proportion of HSC-specific niches). Based on this, how would the authors conceive that such conserved structures would accommodate such differentially fated cells (see the statement lines 187-188 'demonstrating this cellular structure is conserved between developmental and presumptive adult niche')?

This is similar to Essential Revision #6 and we have copied our response below:

We have addressed this excellent point in the Discussion section with the following paragraph:

“A major outstanding question is if there is functional heterogeneity within different niche sites of the CHT and KM. For example, the 1:1 attachment of an HSPC to an MSC that we observe for 50-60% of HSPCs could regulate quiescence versus expansion. There may also be subtypes of ECs within the vasculature of the CHT and KM that provide different regulatory functions, as is observed in the adult mouse bone marrow (J. Zhang et al., 2021). The CHT of the zebrafish differs from the fetal liver because there is not the massive exponential expansion of HSPCs required to sustain the mammalian embryo in utero (Kumaravelu et al., 2002; Owen J Tamplin et al., 2015). Unlike mammalian embryos, zebrafish embryos develop externally and can survive for many days with no definitive hematopoiesis (Sood et al., 2010; Y. Zhang et al., 2011). We do not know if the HSPCs that transit through the CHT to the KM have equal potential to become long term quiescent HSCs, or if there is already heterogeneity. Some HSPCs may be fated to produce the blood lineages required during development, while others may retain their quiescence as they transit through the CHT to the KM. Although we don’t have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible.”

– Figure 8 and text

The Figure with the light sheet information on close proximity between HSPCs and dbh cells is nice and the quantitative analysis panel (C) convincing.

However, the data should be substantiated using a double Tg line (draculin:2APEX2-mCherry) X (dbh:GFP) to confirm the identity of the HSPC(s) contacting the dbh cell(s) and not only a resemblance as state line 231 (hence labelled by the APEX2 nuclear + mitochondrial activities). This would very much substantiate the work and strengthen the manuscript on this important aspect of HSPC/niche establishment/regulation (and the authors have all the tools/approaches at hand). This is also required since the authors clearly state at the end of the Discussion (line 262) '… additional rounds of CLEM to confirm that dbh:GFP+ cells were adjacent to HSPCs …'; clearly, identifying unambiguously HSPCs is mandatory.

This is similar to Reviewer #2 comment #4 above and we have copied our response below:

Using our Workflow #2, we validated the presence of dbH+ cells within the KM niche by confocal imaging of dbh:GFP+ transgenic larva, followed by SBEM, and then correlation of the two datasets using the DRAQ5 approach. We agree with the reviewer that it would have been ideal to perform this correlation using Runx:mCherry+/dbh:GFP+ or APEX2:mCherry+/dbh:GFP+ double transgenics. However, we were not able to optimize the protocol in a way that we could retain both the GFP and mCherry signals and perform SBEM. Therefore, we proceeded with dbh:GFP+ transgenic alone. Instead, we present light microscopy data and quantification showing that most HSPCs are in contact or close proximity to dbh:GFP+ cells (Figure 7B and Figure 7 —figure supplement 1A).

In the accompanying Extended Data Figure 9, panel (A) line 892, what do (n=10) refers to? 10 sections of 1 experiment? 1 section, each from 10 independent experiments? What cells does the blue colour in panel Bii underline?

n=10 indicates the number of larvae in which cells were counted. Blue cells are DRAQ5 labeled nuclei. This is clarified in the revised Figure 7 —figure supplement 1 legend (relevant section copied below):

“Numbers above the columns indicate the cell numbers counted in each group (combined data from n=10 larvae). (B) 3D CLEM alignment of confocal dataset with DRAQ5 labeled nuclei (blue in (i), (ii), and (iv)) and SBEM datasets identifies a dbh:GFP+ cell in the KM niche. (i) Alignment between confocal and SBEM datasets in the XY plane. (ii) White arrowhead points to the single dbh:GFP+ cell in the aligned confocal and SBEM datasets, and in (iii) the green outlined dbh:GFP+ cell in SBEM data alone. (iv) dbh:GFP+ cell in confocal data alone.”

In the text, the authors say that they wish to address a role of dopamine signalling in the KM niche colonization (hence they treat with 6-OHDA from 4 to 5dpf, during niche colonization); would a shorter treatment (ex: few hours) also affect HSPCs in the niche, which would indicate a function of dopamine in the niche per se (ex: survival, maintenance) rather than colonization?

We performed drug treatments for shorter durations (2 hours and 4 hours) starting at 120 hpf (5 dpf) and we did not see an effect on HSPCs in the niche (data not shown). We are pursuing the role of dopamine signaling in the niche in more detail in a future study that we believe is beyond the scope of this current manuscript.

Finally, the authors propose, if I understand correctly, that there may be 2 different types of functional niches (1) proximal/adjacent to the glomerulus, (2) more distant and referred to as posterior perivascular niche; do they have any information on a possible differential influence of dopamine and dbh-cells on these 2 niche types (or to their colonization)? Answering to that question would as well strengthen the functional impact of the work.

This is similar to Essential Revision #7 and we have copied the response below:

This is a very interesting point however the genetic tools are not yet available to address this point. We have added the following lines to the Discussion section (393-395):

“Although we don’t have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible.”

Comments on the Discussion:

The authors should, in my opinion, discuss the possible functional discrepancies between the 2 geographic positions of the niches that are visualized here, i.e the one adjacent to the glomerulus and the one that is perivascular (more distant from the glomerulus); they should do so also in the context of the current work on mammalian bone marrow niches. This is an important issue in the field because, ultimately, one wants to comprehend what are the functional differences between the niches that allow the maintenance of long-term HSCs and the ones that are more devoted to support specific lineage differentiation. Do the authors have a vision on that issue?

This is similar to Essential Revision #6 and again we have copied our response below:

We have addressed this excellent point in the Discussion section with the following paragraph:

“A major outstanding question is if there is functional heterogeneity within different niche sites of the CHT and KM. For example, the 1:1 attachment of an HSPC to an MSC that we observe for 50-60% of HSPCs could regulate quiescence versus expansion. There may also be subtypes of ECs within the vasculature of the CHT and KM that provide different regulatory functions, as is observed in the adult mouse bone marrow (J. Zhang et al., 2021). The CHT of the zebrafish differs from the fetal liver because there is not the massive exponential expansion of HSPCs required to sustain the mammalian embryo in utero (Kumaravelu et al., 2002; Owen J Tamplin et al., 2015). Unlike mammalian embryos, zebrafish embryos develop externally and can survive for many days with no definitive hematopoiesis (Sood et al., 2010; Y. Zhang et al., 2011). We do not know if the HSPCs that transit through the CHT to the KM have equal potential to become long term quiescent HSCs, or if there is already heterogeneity. Some HSPCs may be fated to produce the blood lineages required during development, while others may retain their quiescence as they transit through the CHT to the KM. Although we don’t have the genetic tools yet to functionally distinguish between different sites of HSPC lodgement within the larval KM niche, we are developing a labeling approach that would make this possible.”

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #2 (Recommendations for the authors):

The manuscript describing this technically groundbreaking and very impressive approach to study the niche is largely improved and provides a more balanced interpretation of the results, considering the inherent limitation of the approach including datasets available.

I understand that the approach is technically very challenging and time-consuming, and would not allow to obtain larger datasets to perform quantitative analyses and obtain more definitive conclusions. Therefore, these should be rephrased to emphasise the qualitative (rather than quantitative) nature of the analysis and acknowledge some limitations (such as the lack of markers for putative HSCs in some cases). The contact with glial cells is emphasised but it should be noted that only 3 out of 22 HSPCs were found in contact with glial cells. Besides the numerical difference of HSPCs found in contact with different niche cells, whether HSPCs contacting these different niche cells might be functionally distinct could be incorporated into the discussion. The conclusions should acknowledge the qualitative nature of the study and emphasise the strengths and limitations of this novel approach to study the niche in the future, since it could potentially elicit significant interest in the field to use it in different experimental settings.

We thank the reviewer for their comments. We have updated the text to emphasize the qualitive nature of the approach and acknowledge some limitations along with the strengths. We will discuss that only 3 out of 22 HSPCs were found in contact with glial cells. Also, we have discussed that HSPCs could be functionally distinct based on their contact with different niche cells.

Reviewer #3 (Recommendations for the authors):

Comments to the Authors

Many thanks to the authors to have provided a significant revision of their work; obviously, this correlative multimodal microscopy work is a tedious, difficult, time-consuming but essential task.

While many of my points have been taken into account, there still remains 2 critical aspects that I believe should be addressed and that are partly emphasized by the new dataset (dataset 4 using the CD41:gfp line, workflow 2).

Point 1 (that also relates to essential point 2 of the Evaluation Summary): The authors have provided a new dataset (see new Figure 6 and the 2 Supplemental Figures+ Table 1) that significantly increases the number of HSPCs that are analysed, including the identification of potential niche contacting cells (Table 1). However, the dataset raises the critical point of heterogeneity of the niche environment because the cluster of HSPCs that is analysed appears disconnected to any glial-like cells (GL, see outcome = 0 in Table 1 for the 17 HSPCs). Hence, this new dataset strengthens the idea of the heterogeneity which is not quantitatively appreciated in the work.

We thank the reviewer for their comments. This is a similar point as stated by Reviewer #2 and we have copied the answer here:

We have updated the text to emphasize the qualitive nature of the approach and acknowledge some limitations along with the strengths. We will discuss that only 3 out of 22 HSPCs were found in contact with glial cells. Also, we have discussed that HSPCs could be functionally distinct based on their contact with different niche cells.

Also, on qualitative aspects, it appears that the HSPC cluster is more distal to the glomeruli than the HSPCs illustrated in Figure 5 and Figure 5 – figure supplement 1 and 3 for example (which appear from the images to be more proximal to the glomerulus while for new Figure 6 the cluster appears to be connecting the pronephric tubule and the margin of the intestinal epithelium). Indeed, there may be different niches that may have been underappreciated in the work as it stands (niche peculiarities may depend on geo-localization, without considering the other, more posterior, perivascular niche).

We thank the reviewer for their thoughtful observations. Regarding the observation that the HSPC cluster in new Figure 6 is less proximal to the glomerulus compared to the HSPC cluster of Figure 5, we believe it is simply the slices we have chosen to present for these figures. Depending on the oblique angle chosen for visualization through the 3D serial section blockface electron microscopy (SBEM) dataset, the glomerulus and/or pronephric tubules may or may not be visible. In Figure 4 —figure supplement 2, it is easier to see that the HSPC clusters are always located between the glomerulus and pronephric tubules. Visualization of the data aside, it is possible that depending on where in the cluster a single HSPC resides, it may be closer to a particular cell type or signal that differentially regulates its function. We will make all datasets publicly available so readers can explore them further and perform new analyses.

Point 2 (that also relates to essential point 3 of the Evaluation Summary): The point on the APEX2 signal to unequivocally identify the HSPCs under study has not been addressed in the revision. This is important in particular for the data that illustrate the contact between dbh cells and putative HSPCs which is scarce (1 couple of cells in new Figure 7 (previous Figure 8, kept unchanged)); as it stands, this result is still preliminary (the HSPC is not identified unequivocally). The authors argue in their response that they did not manage to optimize the protocol so as to retain the GFP and mCherry signals but can't they take the peroxidase signal as reference (together with the gfp signal from dbh cells)?

[editor note: please reword the specific section to discuss other options and limitations]

We acknowledge that the putative HSPC in Figure 7 is not identified unequivocally based on its own HSPC-specific marker (e.g., cd41, Runx1+23, or draculin). We have been careful to describe unlabeled HSPCs as “putative”. We have emphasized that the putative HSPC in question is described based on morphology alone. For this dataset we focused on localization of the dbh:gfp+ cells to the kidney niche. Unfortunately, we cannot use the peroxidase signal generated in Workflow #1 and #2 at the same time because one relies on APEX2 and the other on DRAQ5 to create high contrast in the EM datasets and would generate overlapping signals.

Remark on essential Point 7: the answer to that point is well taken and understandable. In addition, since the strategy is to use mosaicism (hence reduced number of labelled HSPCs), this decreases the chance to capture the cells that immobilize in the perivascular, more distant niche (the HSPCS proximal to glomeruli appear to be more clusterized, increasing the chance to capture some of them). The authors have added in the discussion (lines 400-401 and not 392-394) that they are working on a labelling approach that would make it possible. I believe it is fair considering this point beyond the scope of this study.

We appreciate the reviewer acknowledging the technical challenges associated with aspects of the study.

For the reviewers’ information, we made one additional change that was not raised in the reviews. After careful consideration, we have revised the terminology of the “glial-like” cell to “ganglion-like cell” as it is a more accurate prediction of the dbh:gfp+ cell identity based on the available literature (1-4). The tyrosine hydroxylase/dopamine β-hydroxylase positive cells in that region of the embryo have been described as the “cervical sympathetic ganglion complex” (3), “superior cervical ganglion” (5), or “sympathetic cervical complex” made up of “neural crest-derived peripheral neurons, cranial ganglion neurons and glia” (1).

References

1. Stewart RA, Arduini BL, Berghmans S, George RE, Kanki JP, Henion PD, Look AT. Zebrafish foxd3 is selectively required for neural crest specification, migration and survival. Developmental Biology. 2006;292(1):174-88. doi: 10.1016/j.ydbio.2005.12.035. PubMed PMID: 16499899.

2. Stewart RA, Look AT, Kanki JP, Henion PD. Development of the peripheral sympathetic nervous system in zebrafish. Methods in cell biology. 2004;76:237-60. PubMed PMID: 15602879.

3. An M, Luo R, Henion PD. Differentiation and maturation of zebrafish dorsal root and sympathetic ganglion neurons. Annals of the New York Academy of Sciences. 2002;446(3):267-75. doi: 10.1002/cne.10214.

4. Guo S, Brush J, Teraoka H, Goddard A, Wilson SW, Mullins MC, Rosenthal A. Development of noradrenergic neurons in the zebrafish hindbrain requires BMP, FGF8, and the homeodomain protein soulless/Phox2a. Neuron. 1999;24(3):555-66. Epub 1999/12/14. doi: 10.1016/s0896-6273(00)81112-5. PubMed PMID: 10595509.

5. Zhu S, Lee J-S, Guo F, Shin J, Perez-Atayde AR, Kutok JL, Rodig SJ, Neuberg DS, Helman D, Feng H, Stewart RA, Wang W, George RE, Kanki JP, Look AT. Activated ALK collaborates with MYCN in neuroblastoma pathogenesis. Cancer cell. 2012;21(3):362-73. doi: 10.1016/j.ccr.2012.02.010. PubMed PMID: 22439933; PMCID: PMC3315700.

https://doi.org/10.7554/eLife.64835.sa2

Article and author information

Author details

  1. Sobhika Agarwala

    Center for Stem Cell and Regenerative Medicine, Department of Pharmacology, College of Medicine, University of Illinois at Chicago, Chicago, United States
    Present address
    Department of Cell & Regenerative Biology, School of Medicine and Public Health, University of Wisconsin-Madison, Madison, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Contributed equally with
    Keun-Young Kim
    Competing interests
    No competing interests declared

  2. Keun-Young Kim

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – review and editing
    Contributed equally with
    Sobhika Agarwala
    Competing interests
    No competing interests declared

  3. Sebastien Phan

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Software, Formal analysis, Visualization
    Competing interests
    No competing interests declared

  4. Saeyeon Ju

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared

  5. Ye Eun Kong

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared

  6. Guillaume A Castillon

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared

  7. Eric A Bushong

    Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6195-2433

  8. Mark H Ellisman

    1. Center for Research in Biological Systems, National Center for Microscopy and Imaging Research, University of California at San Diego, San Diego, United States
    2. Department of Neurosciences, University of California at San Diego School of Medicine, San Diego, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Project administration
    For correspondence
    mellisman@ucsd.edu
    Competing interests
    No competing interests declared

  9. Owen J Tamplin

    Center for Stem Cell and Regenerative Medicine, Department of Pharmacology, College of Medicine, University of Illinois at Chicago, Chicago, United States
    Present address
    Department of Cell & Regenerative Biology, School of Medicine and Public Health, University of Wisconsin-Madison, Madison, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing
    For correspondence
    tamplin@wisc.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9146-4860

Funding

National Heart, Lung, and Blood Institute (R01HL142998)

  • Owen J Tamplin

National Institute of Diabetes and Digestive and Kidney Diseases (K01DK103908)

  • Owen J Tamplin

American Heart Association (19POST34380221)

  • Sobhika Agarwala

National Institute of Neurological Disorders and Stroke (1U24NS120055-01)

  • Mark H Ellisman

National Institute of General Medical Sciences (R24 GM137200)

  • Mark H Ellisman

American Society of Hematology (Junior Faculty Scholar Award)

  • Owen J Tamplin

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by grants from the NIH NHLBI (R01HL142998), NIDDK (K01DK103908), American Society of Hematology (Junior Faculty Scholar Award; OJT), and the American Heart Association (Grant #19POST34380221; SA). The NCMIR is principally supported by grants from the NIH NINDS (1U24NS120055-01) and NIGMS (R24 GM137200). Light sheet imaging was performed at the Integrated Light Microscopy Core Facility at the University of Chicago with the help of Christine Labno and Vytas Bindokas, and at the UW-Madison Optical Imaging Core with Lance Rodenkirch. The authors wish to thank: Dr Daniela Boassa for her assistance in EM probe design with APEX2; Tom Deerinck, Steven Peltier, and Tristan Shone for technical advice with Focal CC; Iain A Drummond for feature identification in SBEM sections; Christian Mosimann for the draculin promoter 5’ entry vector; Willy Wong for depositing the SBEM datasets in The Cell Image Library.

Ethics

All experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committees at the University of Illinois at Chicago (Protocol ACC 19-051) and the University of Wisconsin-Madison (Protocol M006348).

Senior Editor

  1. Didier YR Stainier, Max Planck Institute for Heart and Lung Research, Germany

Reviewing Editor

  1. Cristina Lo Celso, Imperial College London, United Kingdom

Reviewer

  1. Anne Schmidt, CNRS UMR3738, Institut Pasteur, France

Version history

  1. Received: November 12, 2020
  2. Preprint posted: November 13, 2020 (view preprint)
  3. Accepted: July 4, 2022
  4. Accepted Manuscript published: August 9, 2022 (version 1)
  5. Version of Record published: August 19, 2022 (version 2)

Copyright

© 2022, Agarwala, Kim et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,257
    Page views
  • 661
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Sobhika Agarwala
  2. Keun-Young Kim
  3. Sebastien Phan
  4. Saeyeon Ju
  5. Ye Eun Kong
  6. Guillaume A Castillon
  7. Eric A Bushong
  8. Mark H Ellisman
  9. Owen J Tamplin
(2022)
Defining the ultrastructure of the hematopoietic stem cell niche by correlative light and electron microscopy
eLife 11:e64835.
https://doi.org/10.7554/eLife.64835

Categories and tags

Research organism

Further reading

    1. Developmental Biology
    2. Stem Cells and Regenerative Medicine

    Development: Shining a light on hematopoietic stem cells

    Anne Schmidt
    Insight

    A combination of light and electron microscopy has revealed further details about the location and interactions of hematopoietic stem and progenitor cells.

    1. Developmental Biology

    Dynamic compartmentalization of the pro-invasive transcription factor NHR-67 reveals a role for Groucho in regulating a proliferative-invasive cellular switch in C. elegans

    Taylor N Medwig-Kinney, Brian A Kinney ... David Q Matus
    Research Article

    A growing body of evidence suggests that cell division and basement membrane invasion are mutually exclusive cellular behaviors. How cells switch between proliferative and invasive states is not well understood. Here, we investigated this dichotomy in vivo by examining two cell types in the developing Caenorhabditis elegans somatic gonad that derive from equipotent progenitors, but exhibit distinct cell behaviors: the post-mitotic, invasive anchor cell and the neighboring proliferative, non-invasive ventral uterine (VU) cells. We show that the fates of these cells post-specification are more plastic than previously appreciated and that levels of NHR-67 are important for discriminating between invasive and proliferative behavior. Transcription of NHR-67 is downregulated following post-translational degradation of its direct upstream regulator, HLH-2 (E/Daughterless) in VU cells. In the nuclei of VU cells, residual NHR-67 protein is compartmentalized into discrete punctae that are dynamic over the cell cycle and exhibit liquid-like properties. By screening for proteins that colocalize with NHR-67 punctae, we identified new regulators of uterine cell fate maintenance: homologs of the transcriptional co-repressor Groucho (UNC-37 and LSY-22), as well as the TCF/LEF homolog POP-1. We propose a model in which the association of NHR-67 with the Groucho/TCF complex suppresses the default invasive state in non-invasive cells, which complements transcriptional regulation to add robustness to the proliferative-invasive cellular switch in vivo.

    1. Chromosomes and Gene Expression
    2. Developmental Biology

    Transcription: A hub of activity

    Virginia L Pimmett, Mounia Lagha
    Insight

    Imaging experiments reveal the complex and dynamic nature of the transcriptional hubs associated with Notch signaling.