Heterotypic interaction promotes asymmetric division of human hematopoietic progenitors Free

Asymmetric cell division is one evolutionary conserved strategy to generate the cellular diversity adopted by stem cells. Asymmetric divisions can lead to unequal inheritance of cell fate determinants (Tajbakhsh et al., 2009) or place the two daughter cells in distinct environments providing different fate determinants (Kiger et al., 2001). In most characterized stem cells so far, asymmetry relies on the polarization of the mother cell controlled by external cues of the cell microenvironment (Daley and Yamada, 2013; Song et al., 2002; Mesa et al., 2018). These cues induce centrosome positioning in interphase, and, in turn, orientation of the mitotic spindle along the polarity axis (Venkei and Yamashita, 2018). Paradigmatic of adult stem cells, hematopoietic stem and progenitor cells (HSPCs) self-renew and differentiate into lineage-determined daughter cells, which will give rise to all cell types of the hematopoietic system (Orkin and Zon, 2008). HSPCs are retained in the bone marrow within specific niches (Schofield, 1978), where external cues modulate their homeostasis (Wilson and Trumpp, 2006; Pinho and Frenette, 2019). However, the mechanism(s) leading to asymmetric division of HSPCs has not yet been well characterized. In vitro HSPCs can undergo asymmetric divisions (Beckmann et al., 2007; Ting et al., 2012; Görgens et al., 2014; Florian et al., 2018; Zimdahl et al., 2014; Vannini et al., 2019; Loeffler et al., 2019). In vitro (Wagner et al., 2007; Nakahara et al., 2019; Xiao et al., 2009; Yu et al., 2017) or in vivo (Calvi et al., 2003; Lo Celso et al., 2009; Bowers et al., 2015; Christodoulou et al., 2020) investigations have shown that stromal cells, including osteoblasts and endothelial cells, can modulate HSPC proliferation and differentiation. This suggests that stromal cells may play a role in controlling asymmetric HSPC divisions (Wu et al., 2007). Nevertheless, such a role still remains hypothetical. Classical in vitro co-cultures are unlikely to allow the mechanisms at play to be deciphered because they inadequately recapitulate the long-term confinement encountered by HSPCs in their niches. Moreover, in vivo imaging of HSPC–stromal cell interactions and subsequent cell divisions has not yet been achieved. To overcome such limitations, systems of intermediate complexity have been developed (Kräter et al., 2017; Bianco et al., 2019; Braham et al., 2019; Souquet et al., 2021; Glaser et al., 2022). We have recently set up a system of microwells as minimalist niches to show that human HSPCs have the capacity to interact in interphase with specific stromal cells of the bone marrow, and polarize by reorganizing intensively their intracellular architecture (Bessy et al., 2021). Combining this system with live-cell markers, we here investigate the impact of heterotypic interactions on the mode of division of HSPCs.

Non-adhesive microwells were used as culture model to overcome the limitations of classical co-culture systems for long-term imaging by preventing cells escaping from the wells (Fig. 1A,B, Fig. S1A-C, Movie 1). Microwells were coated with fibronectin, a major extracellular matrix protein in the bone marrow (Van Der Velde-Zimmermann et al., 1997; Zanetti and Krause, 2020), or seeded with either stromal cells of the hematopoietic niches (osteoblasts or endothelial cells) or skin fibroblasts as control cells (Fig. 1A).

Human cord blood CD34⁺ cells were used to perform experiments. CD34⁺ cells form a heterogenous population of stem and progenitor cells, but are considered as a global population representing the early steps of hematopoiesis and are referred to in the literature as HSPCs (Notta et al., 2011; Buenrostro et al., 2018).

CellTracker-labeled HSPCs were loaded at a density of one cell per well (Fig. 1A,B, Fig. S1B,C). CellTracker concentration was optimized to preserve cell viability, cell cycle progression and proliferation (Fig. S1D,E,G). Using this system, HSPCs could be tracked over a few generations (Fig. 1B, Movie 2). Upon culture, HSPCs were observed to enlarge between 20 and 40 h, indicative of cell cycle progression (Fig. 1E). In the three culture conditions, cell division occurred at around 40 h (Fig. 1C), followed by a second division around 20 h later (Fig. 1D), indicating that the different feeder cells seeded in the microwells had no effect on HSPC cell cycle kinetics.

HSPCs cultured on fibronectin or fibroblasts had a round morphology or exhibited a dorsal uropod, at 20 h and 40 h of culture (Fig. 1F,G, Fig. S2A). In contrast, at 20 h, more than 40% of HSPCs cultured on endothelial cells or osteoblasts were elongated and anchored on the feeder cell through a magnupodium (Fonseca et al., 2010; Bessy et al., 2021). Twenty hours later, prior to cell division, 60% of them had adopted this morphology (Fig. 1F-H, Fig. S2A). At 20 and 40 h of culture, the centrosome was found within the magnupodium, located close to the site of contact with the endothelial cell or osteoblast (Fig. 1G, lower panel, Fig. S2A, Movie 3), as previously described (Bessy et al., 2021). Accordingly, the cell polarity indices, defined as the ratio between the distance from the site of contact to the centrosome and the HSPC length (Fig. 1F,G) were significantly reduced in HSPCs interacting with endothelial cells and osteoblasts, compared with HSPCs cultured on fibronectin or fibroblasts (Fig. 1I).

The effect of HSPC polarization on the spatial distribution of Golgi and lysosomes was then analyzed. The Golgi apparatus was always close to the centrosome. In the case of HSPCs with a magnupodium, it was found extending within this structure (Fig. 1J,K). Lysosomes, labeled using cell-permeable and stable fluorogenic markers, were found tightly associated with the Golgi, extending within the magnupodium (Fig. 1L,M, Fig. S2B,C, Movie 4).

Taken together, these results indicate that the interaction of HSPCs with stromal cells of the hematopoietic niches induces HSPC polarization, marked by the formation of an elongated magnupodium. Within this structure, the centrosome is located proximal to the site of interaction. The Golgi apparatus and lysosomes are confined and extend toward the cell body (Fig. 1N). Such remarkably polarized architecture is stable and maintained during cell cycle progression, until mitosis onset.

We then investigated the impact of HSPC polarization on the subsequent mitosis. Cells were fixed at around 40 h of culture, corresponding to the peak of first mitosis (Fig. 1D). In all conditions, HSPCs were found to become rounded at mitosis onset (Fig. S3A). At prophase, in HSPCs cultured on endothelial cells or osteoblasts, centrosomes were significantly closer to the site of contact, compared with HSPCs cultured on fibroblasts or on fibronectin (Fig. S3A,B). At metaphase and anaphase, the centrosomes were located at the opposite spindle poles, thus defining the spindle axis. In HSPCs cultured on fibronectin, the spindle was oriented parallel to the bottom of the well (Fig. 2A-C, Fig. S3C,D), as previously documented (Zimdahl et al., 2014). Similarly, the spindle of HSPCs cultured on fibroblasts was oriented parallel to the surface of contact (Fig. 2A,C, Fig. S3C,D, second column). In contrast, the metaphase spindle of HSPCs interacting with endothelial cells and, more prominently, with osteoblasts was found to be oriented perpendicularly to the cell surface (Fig. 2A,C, third column, Movie 5). This perpendicularity was maintained at anaphase (Fig. S3C,D, fourth column).

These results show that the interaction of HSPCs with endothelial cells and, more drastically, with osteoblasts promotes the positioning of the spindle perpendicular to the surface of the stromal cell.

Centrosome positioning during interphase by external cues, and subsequent spindle orientation during mitosis, are hallmarks of asymmetric cell division (Venkei and Yamashita, 2018). Given that in human HSPCs asymmetric cell division has been shown to be associated with an unequal segregation of lysosomes into daughter cells (Loeffler et al., 2022), we investigated the effect of heterotypic interactions on lysosome inheritance in HSPC siblings.

The cell-permeable and stable fluorogenic marker LysoBrite^TM, was used to label and track HSPC lysosomes by live imaging. Upon division, CellTracker dye, used to label the whole cell, was always homogeneously segregated. In contrast, both equal and unequal LysoBrite segregations could be observed (Fig. 2D, Fig. S3E). The lysosomes were stably inherited as the LysoBrite signals were stable in time for both daughter cells (Fig. S3F). The lysosome inheritance ratio, measured as the ratio between lysosomes in the daughter cells, was significantly higher in mother HSPCs interacting with endothelial cells and osteoblasts (Fig. 2E). Setting 1.5 as a threshold value to discriminate symmetric (<1.5) versus asymmetric (≥1.5) inheritance (Loeffler et al., 2019), we determined that 60% of the HSPCs interacting with endothelial cells and osteoblasts underwent asymmetric lysosome inheritance, compared with 40% in the case of HSPCs cultured on fibroblasts or fibronectin (Fig. S3G).

Lysosome dynamics were further assessed with cell division geometry, during the mitosis of HSPCs interacting with osteoblasts. In HSPCs dividing parallel to the osteoblast surface, lysosomes were equally distributed to the opposite cell poles. In contrast, in HSPCs dividing perpendicularly, lysosomes were found to be enriched in one of the cell poles. This enriched pole was proximal to the site of contact (Fig. 2F-H).

Taken together, these results show that HSPC interaction with stromal cells of the niches – and particularly with osteoblasts – promotes during mitosis the redistribution of lysosomes associated with spindle orientation and the asymmetric segregation of lysosomes into the daughter cells. The proximal daughter cell inherits most of the mother lysosome pool.

Is there a causal link between HSPC polarization during interphase and the asymmetry of division? To address this question, HSPC polarization, cell division geometry and lysosome segregation were monitored using video microscopy in HSPCs interacting with osteoblasts, from mother cell interphase to daughter cell interphase (Fig. 3A,B). Lysosome polarization indices were measured during interphase. Based on the distribution of lysosomes in polarized HSPCs (Fig. 1M), a threshold value of 0.35 was set to discriminate polarized (<0.35) versus unpolarized (>0.35) lysosome distributions. Lysosome polarization indices were also measured during mitosis, at metaphase/anaphase (Fig. 3C), and at late telophase when daughter cells were individualized (Fig. 3D).

HSPCs that had an unpolarized lysosome distribution during interphase were mostly found to divide parallel to the osteoblast surface (Fig. 3A). During mitosis, lysosomes were redistributed within the whole rounded cell (Fig. 3A,C, blue box) and were eventually inherited symmetrically in the daughter cells (Fig. 3A,D, blue box, Movie 6). In contrast, HSPCs that had a polarized distribution of lysosomes during interphase, with lysosomes clustered in the magnupodium, mostly divided perpendicularly to the osteoblast surface (Fig. 3B). Lysosomes remained clustered close to the site of contact during mitosis (Fig. 3B,C, pink box). They were eventually asymmetrically segregated (Fig. 3B,D, pink box, Movie 7).

HSPC polarization during interphase involves the SDF-1 (CXCL12)/CXCR4 ligand/receptor couple (Bessy et al., 2021). To test whether SDF-1/CXCR4 was also involved in HSPC asymmetric division, we investigated the effect of the CXCR4 antagonist AMD3100 on HSPC division upon interaction with osteoblasts. AMD3100 was used at 50 µM, a concentration that did not impact survival, proliferation, or division kinetics (Fig. S3H-J). In these conditions, HSPC polarization was reduced during interphase (Fig. 3E). Moreover, during mitosis, the asymmetric inheritance of lysosomes was significantly decreased in AMD3100-treated cells (Fig. 3F,G). This result suggests that SDF-1/CXCR4 is involved both in HSPC polarization during interphase and in the asymmetry of HSPC division. It reinforces the causal link between HSPC polarization and further asymmetric division.

Taken together, these results show that HSPC interaction with bone marrow stromal cells is a bona fide external cue through SDF-1/CXCR4 driving asymmetric HSPC division. Interestingly, not all HSPCs respond to this cue and some undergo symmetric division. Nevertheless, once the heterotypic interaction has induced HSPC polarization with the formation of a magnupodium and the clustering of lysosomes toward the site of contact, the subsequent division is mostly marked by perpendicular spindle positioning, maintenance of the asymmetric distribution of lysosomes and final asymmetric inheritance of this organelle in the siblings.

Lysosome inheritance has been shown in human HSPCs to be predictive of cell surface marker expression in the daughter cells. In particular, CD34 expression is decreased in the daughter cell inheriting more mother cell lysosomes, whereas CD33 expression is not affected (Loeffler et al., 2022). To characterize further the daughter cells generated, we thus compared CD34 and CD33 expression in daughter cells of HSPCs cultured either on osteoblasts, which promote asymmetric divisions, or on fibronectin, as a control.

Fluorescently labeled CD34 and CD33 antibodies were added to the microwells after HSPC division (Fig. 4A) to quantify their distribution in the siblings (Fig. S4A-C). For both markers, asymmetric and symmetric distributions could be observed (Fig. 4B,C). For CD33, the ratios measured between the daughter cells (referred to as the sister ratios) were not significantly different upon culture on osteoblasts or on fibronectin (Fig. 4D). In contrast, the CD34 sister ratio was significantly higher upon culture with osteoblasts than on fibronectin (Fig. 4E). Threshold values of 1.5 and 1.25 were set to discriminate siblings with asymmetric (>1.5) and symmetric (<1.25) distributions. As far as CD33 was concerned, the two experimental conditions generated similar populations with one-third of the siblings exhibiting an asymmetric CD33 distribution (Fig. 4F). In contrast, CD34 asymmetric distributions were twice as common (41%) upon culture on osteoblasts compared with fibronectin (21%; Fig. 4G). This result indicates that heterotypic interaction with osteoblasts increases the level of CD34 asymmetric distribution in HSPC daughter cells.

Both markers were then simultaneously analyzed in the siblings (Fig. 4H). The resulting density maps in the two experimental conditions shed light on differences in sibling heterogeneity landscapes. The whole population of siblings obtained upon culture on fibronectin was found to be concentrated around low ratio values, indicative of a homogenous population. In contrast, the population resulting of HSPC culture on osteoblasts was more spread out: 75% was similar to the whole population obtained on fibronectin, whereas 25% of siblings emerged as a distinct population, with higher ratio values (Fig. 4I).

These results indicate that heterotypic interaction with osteoblasts boosts, upon mitosis, the heterogeneity of the progeny.

We here demonstrate that heterotypic interaction with bone marrow stromal cells is a genuine external cue that drives bona fide asymmetric division of human HSPCs. In the absence of polarizing external cues (i.e. cultured on fibronectin or skin fibroblasts), a fraction of HSPCs can undergo asymmetric division, indicating that HSPCs possess intrinsic properties that make them prone to asymmetry (Görgens et al., 2014; Zimdahl et al., 2014). However, heterotypic interaction with stromal cells of the vascular or endosteal niche impacts quantitatively and qualitatively the division by creating a bias toward asymmetry and by generating siblings that are more drastically divergent.

This division mode relies on a role of the centrosome, a conserved feature of asymmetric divisions in solid tissues (Venkei and Yamashita, 2018) as well as in lymphocytes (Chang et al., 2007). During interphase, the centrosome gets positioned in the vicinity of the contact site, closely associated with the Golgi apparatus and the lysosomes. Such organization is reminiscent of lymphocytes engaged in immune synapses (De La Roche et al., 2016). Nonetheless, whereas lymphocytes spread on their target cell, HSPCs interact with the stromal cell through an elongated and narrow magnupodium where the Golgi and lysosomes are confined (Bessy et al., 2021). Furthermore, whereas the immune synapse is maintained for minutes (Douanne and Griffiths, 2021), the magnupodium is maintained for hours. Its assembly may thus represent an efficient strategy developed by the small-sized HSPCs to sustain a robust and long-term polarization.

HSPC subsequent asymmetric division gives rise to two daughter cells positioned proximally and distally to the stromal cell. The proximal cell, positioned where the magnupodium was anchored, inherits most of the mother lysosomes. As a consequence, it has a reduced commitment to differentiation compared with the distal cell (Ito et al., 2016; Liang et al., 2020; Loeffler et al., 2022). The division thus leads to a spatial control of the position of the siblings relative to the stromal cell, which is a conserved feature of asymmetric divisions in other stem cell niches (Venkei and Yamashita, 2018). This property may account for in vivo observations of the organization cells into spatial clusters during blood cell differentiation (Zhang et al., 2021; Wu et al., 2024). It also may play an important role in competition mechanisms at play within the limited space of the hematopoietic niches, in physiological (Miao et al., 2022) and pathological (Colmone et al., 2008; Glait-Santar et al., 2015) contexts.

Our results indicate that the ligand/receptor pair SDF-1/CXCR4 is involved in the control of HSPC asymmetric division, similarly to what has been proposed for developing T cells (Pham et al., 2015). This original role could explain how the SDF-1/CXCR4 signaling pathway contributes to HSPC homeostasis (Kahn et al., 2004), in addition to its well-documented role in HSPC migration and retention in the niches (Lapidot et al., 2005). Similarly, SDF-1/CXCR4 dysfunction could directly affect HSPC asymmetric divisions in malignant contexts (Spoo et al., 2007; Nervi et al., 2009; Zimdahl et al., 2014).

Stem cell asymmetric division is classically considered as a binary process leading to the generation of daughter cells of distinct identities (Kuang et al., 2007; Cicalese et al., 2009). The early steps of hematopoiesis are now considered to occur as a continuum (Buenrostro et al., 2018), with the slow and progressive emergence of distinct cell populations (Tak et al., 2021). Our data support the idea that heterotypic interactions driving asymmetric divisions of stem and progenitor cells do generate siblings with different differentiation potentialities, rather than distinct identities. Such non-stereotypical asymmetric divisions are likely to participate in maintaining the plasticity of the early steps of hematopoiesis.

It is important to note that these results have been obtained on a global population of CD34⁺ cells, which encompasses hematopoietic stem cells, multipotent progenitors and common myeloid or lymphoid progenitors. Some aspects of asymmetric division, such as niche specificities, efficiency of asymmetry, siblings' heterogeneity etc., may differ among these subpopulations. Our results pave the way to further investigations, based on the use of microfabricated niches, to decipher these differences in physiological and pathological contexts.

The mold design and fabrication were performed as previously described (Bessy et al., 2021): microwell shape, size, and arrangement were drawn using the software CleWin and etched in the chrome layer onto a quartz photomask (Toppan Photomask). A wafer with microstructures was made on silicon. Silicon wafers were coated with a first 5-μm-thick layer of resin (SU8-3005; MicroChem; CTS), exposed to UV light at 23 mW/cm² (UV KUB2; Kloe) for 5 s and with a 50-μm-thick second layer (SU8-3050; MicroChem; CTS) exposed to UV light through the quartz mask at 23 mJ/cm² for 8 s in order to create microwells. After development (Developer SU8; MicroChem; CTS), only the exposed structures remained. The wafer was baked at 150°C for 2 h and coated with gas-phase trichloro(perfluorooctyl)silane (Sigma-Aldrich). A negative mold of the silicon wafer was created using PDMS and treated with silane in the same way as the wafer. A positive mold of PDMS was made from the first mold. This positive mold is hereafter referred to as the PDMS stamp.

Glass coverslips were coated with a solution containing 3-(trimethoxysilyl) propylmethacrylate (Sigma-Aldrich), acetone and ethanol, in a ratio of 1:0.5:50 for 10 min. They were then baked at 120°C for 1 h. The PDMS stamp underwent plasma treatment for 30 s and was immediately apposed on the coverslip. A fresh solution of 20% 37.5/1 acrylamide/bisacrylamide (Euromedex), 1% ammonium persulfate, tetramethylethylenediamine (Sigma-Aldrich), 1% photo initiator (2-hydroxy-2-methylpropiophenone; Sigma-Aldrich) in MilliQ water was drawn by capillary action between the PDMS stamp and the glass coverslip. The sample was exposed to UV light at 23 mW/cm² for 5 min (Fig. S1B). After exposure, the PDMS stamp was removed in MilliQ water and washed with 70% ethanol under UV light for 1 h for sterilization. The samples were incubated in sterile PBS overnight to remove potential traces of remaining toxic compounds.

All human umbilical cord blood samples collected from normal full-term deliveries were obtained after mothers' written and informed consent, following the Helsinki's Declaration and Health Authorities. Human umbilical cord blood samples were obtained from the Cord Blood Bank of the Saint-Louis Hospital by French national law (Bioethics Law 2011-814) and under declaration to the French Ministry of Research and Higher Studies.

Mononuclear cells were collected using Ficoll separation medium (Eurobio). CD34⁺ cells were further selected using Miltenyi magnetically activated cell sorting (MACS) columns (Miltenyi Biotech) according to the manufacturer's instructions. CD34⁺ cells were then either put in culture or frozen at −80°C in IMDM medium (Gibco) supplemented with 10% fetal bovine serum (FBS) and 10% DMSO (WAK Chemie Medical GmbH).

The cell lines used were hFOB (osteoblasts; CRL-11372, ATCC), HUVEC-XL (endothelial cells; 191027, Lonza) and BJ (skin fibroblasts; CRL-2522, ATCC), all cultured in DMEM-F12 medium (Gibco) supplemented with 10% FBS and 1% antibiotics (AA; Gibco, 15240062). Cells were tested monthly for Mycoplasma contamination using MycoStripTM 50 (InvivoGen, rep-mysnc-50). Experiments were only performed with Mycoplasma-negative cells.

The glass bottom of the microwells was coated with 10 μg/ml of fibronectin diluted in PBS for 5 min at 37°C. Microwells were gently rinsed twice in PBS, and a droplet of 40 µl of supplemented DMEM F-12 was placed on top of the microwells, then 5000 osteoblasts, endothelial cells, or skin fibroblasts in 5 µl volume (1000 cells/µl) were directly added to the DMEM-F12 droplet in a zigzag manner. After a 30-min incubation period to allow the seeding of osteoblasts and skin fibroblasts by gravity, the dishes were filled with DMEM-F12. Following a 2-hour interval to facilitate cell spreading, 5000 HSPCs were directly seeded onto the microwells using the same method to obtain around one HSPC per well. After a 30-min incubation period to enable HSPC seeding, IMDM supplemented with 10% FBS, 1% AA, 100 ng/ml human SCF (Peprotech, 300-07), 20 ng/ml human IL-3 (Peprotech, 300-23) and 10 ng/ml human G-CSF (Peprotech, 200-02) was added to mark the starting point (t0) of the experiment (Fig. 1A, Fig. S1C).

In specific experiments, in addition to control non-treated cells, cells were treated with the CXCR4 antagonist AMD3100 (Sigma-Aldrich). HSPCs were seeded and incubated with 50 µM of AMD3100.

Coverslips with the microwells were pasted to the bottom-left 35 mm plastic dish, before the sterilization step. The medium was constantly changed every 24 h.

For live imaging labeling, HSPCs were incubated with 20 µM 6-{[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]hexanoyl}sphingosine (NBD C6-Ceramide; Invitrogen) for 30 min at 4°C for Golgi staining or with Cytopainter (Abcam) for Golgi and nucleus staining. HSPCs were incubated with 2 µM Orange CellTracker (Thermo Fisher Scientific, C34551) for 30 min or/and 1:250 LysoBrite Deep Red (Interchim, 22646) or LysoBrite NIR (Interchim, 22641) for 50 min, both at 37°C. HSPCs were rinsed and pelleted twice with HBSS (Gibco) with 10 mM HEPES (Thermo Fisher Scientific). Labeled HSPCs were finally washed in supplemented IMDM with 10 mM HEPES and, after centrifugation (90 g for 3 min), seeded in the microwells and cultured in Phenol Red-free IMDM (Thermo Fisher Scientific) with 10 mM HEPES with cytokines, FBS and AA. Images were taken every 30 min for transillumination and every 2 h for laser illumination.

HSPCs were imaged using transillumination for 46 h. After 46 h and following the peak of cell divisions, anti-CD33-PE (555450, BD Biosciences) and anti-CD34-APC (55824, BD Biosciences) antibodies were added in a 1:60 ratio in 0.05% fetal human serum and PBS with 2 mM EDTA for 20 min in the microwells.

Cells were fixed in PBS with 4% methanol-free paraformaldehyde (Electron Microscopy Sciences) for 15 min at 37°C after 20 h or 40 h of culture. Cells were permeabilized in 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 20 min. Coverslips were neutralized with a solution of NH₄Cl (Sigma-Aldrich) for 10 min. The following primary antibodies and dilutions were used: rabbit anti-pericentrin (ab4448, Abcam; 1/500); human anti-giantin (A-R- H#03 TA10 hFc, Institut Curie; 1/200). Primary antibodies were incubated for 1 h. The following secondary antibodies and dilutions were used: Alexa Fluor 568-conjugated goat anti-human and 647-conjugated goat anti-rabbit (A-11011 and A-21244, respectively, Life Technologies; 1:1000). Alexa Fluor 488-conjugated phalloidin (A12379, Molecular Probes) was used to label F-actin. Finally, cells were incubated with DAPI (Sigma-Aldrich) for 10 min to stain the nucleus. The coverslips were mounted with Mowiol (Sigma-Aldrich).

For the measurement of polarization index and mitotic spindle orientation, images were taken using a Nikon Ti-eclipse microscope equipped with a spinning disk (Yokogawa-CSU-X1) with a 60×, 1.5 lens amplification, 1.4-NA oil objective on an electron-multiplying charge-coupled device camera (Retiga R3). MethaMorph was used as acquisition software. z-stacks of 0.5 µm were taken and a camera binning of 2×2 was used.

For live-cell imaging with trans and laser illumination light, an Olympus IX83 microscope equipped with a PECON CellVivo incubation system controlling temperature (37°C) and CO₂ concentration (5%) was used. Images were acquired with a 20×0.30-NA air objective on an ORCA-Flash4.0 Lite (Hamamatsu) camera and using MicroManager 1.4.21 software. A camera binning of 2×2 was used during the acquisition.

For detailed live-cell imaging of lysosome polarization and asymmetric inheritance with trans and laser illumination lights (Figs 3F, 4, Fig. S2B,C), a Nikon Ti2 Eclipse equipped with a Prime BSI Express CMOS camera (Photometrics) was used with a Nikon CFI Plan Fluor 60×/0.75 NA oil objective with 1.5 lens amplification. Okolab incubation system was used to control temperature (37°C) and CO₂ concentration (5%). Images were acquired using MicroManager 1.4.21 software. To improve the signal/noise ratio of LysoBrite and reduce laser intensity and exposition time, HSPCs were pre-stimulated for 20 h with the cytokine cocktail, then labeled with LysoBrite, and added to the microwells with osteoblasts, as previously described. After 15 h (35 h after the cytokine stimulation), microwells were positioned at the microscope and imaged (10 z-stacks with 4 µm step) every 15 min with a 2×2 camera binning.

The same method was used for the analysis of Golgi polarization (taking the closest point of the Golgi to the cell membrane as ‘B’) and prophase centrosome polarization (polarization index ≤0.3 was taken as a proximal centrosome and >0.3 as a distal centrosome). To be considered as a bona fide magnupodium, the structure was required to be an elongated protrusion (longer than 2 µm) with a centrosome located at its tip.

For SuperPlot representation (Lord et al., 2020), each color represents a biological replicate. The median of each replicate, the mean of the medians, and the s.e.m. are presented. SuperPlots were only generated in the case of biological replicates of more than 20 cells. The normality of data distribution was tested by the Shapiro–Wilk test for further statistical test selection. The significance of the difference between populations was tested using nonparametric (Kruskal–Wallis) ANOVA and Mann–Whitney U test for samples that were not normally distributed, or parametric unpaired t-test and one-way ANOVA for normally distribution samples. All statistical analyses were performed with Prism software (GraphPad).

Mitotic cell images were resliced to visualize the centrosomes and the point of interaction with the bottom of the microwell or the osteoblast/fibroblast. For the mitotic spindle angle, the angle generated between both centrosomes with respect to the image x-axis was subtracted from the angle generated for the point of interaction with respect to the image x-axis. Graphs were generated with R (https://www.r-project.org).

A line-scan from the contact site to a more distal point was used to assess the lysosome localization with respect to the Golgi in maximum projected images. Graphs were generated with Prism software (GraphPad).

Cells were tracked manually until division. To automate the daughter cell area selection, CellTracker images were processed with Gaussian Blur (sigma=2), Subtract Background (rolling=50), and then thresholding (Otsu dark) to create a mask for the daughter cell area. Background signal was selected manually inside the microwell for each image. Lysosome inheritance ratios were calculated in untreated images using the first time point following cell division, by dividing the brightest daughter cell by its sister, using the sum of LysoBrite pixel fluorescence values (CTCF). Ratios were therefore ≥1. Ratios <1.5 and ≥1.5 were respectively selected as indicative of symmetric and asymmetric lysosome inheritance. When both daughter cells were very close in the first time point after division, the following time point was analyzed. Manual selection of the daughter area was carried out when the automatized selection failed. Data were analyzed with Fiji and presented as SuperPlots. Statistical analyses were performed with Prism software (GraphPad).

Lysosome polarization in the mother cell was analyzed as described above for the time-lapse imaging before chromosome condensation, visualized by phase images. Metaphase/anaphase was analyzed in the time-lapse images, which allows the visualization of the chromosome orientation to determine mitotic spindle orientation with respect to the interaction point. In cases in which the chromosome orientation was not clear, the visualization of a cell-side attachment from metaphase to telophase was taken as a perpendicular division. Lysosome inheritance was measured after the division in the first time frame, which allows the visualization of the whole two daughter cell volumes.

Gaussian fit curves were created based on the CTCF values of HSPCs that were stained or not stained for the two CD marker signals; 95% confidence intervals of non-stained intestines were selected to determine the thresholds for positive CD34 and CD33 signals. Single mother cells were tracked manually until their division by transillumination to annotate the division time. To automate the daughter cell area selection, CD34 fluorescence images were processed with Gaussian Blur (sigma=2), Subtract Background (rolling=50) and then thresholding (Otsu dark) to create a mask for the daughter cell area. The background region was selected manually inside the microwell for each microwell. When the CD marker CTCF of both daughter cells was below the threshold, the cells were discarded. When one daughter cell was below the threshold, its CTCF was automatically converted to the value threshold and both sister cells were compared. CD marker production ratios were calculated after antibody incubation, by dividing the brightest daughter cell by its sister. Ratios were therefore always ≥1. Ratios ≤1.2 and ≥1.5 were respectively selected as indicative of symmetric and asymmetric distributions. Ratios between 1.2 and 1.5 were indicative of intermediate distribution. In cases in which both daughter cells were overlapping, the next time point was analyzed. Manual selection of the daughter area was carried out when the automatized selection failed. Data were analyzed with Fiji and presented as SuperPlots, or with R (https://www.r-project.org). In Fig. 4I, the corresponding density map was generated using the ‘ggdensity’ package (https://jamesotto852.github.io/ggdensity/, https://github.com/jamesotto852/ggdensity/). Statistical analyses were performed with Prism software (GraphPad).

We thank Sandrine Moutel and the platform TabIP of the Curie Institute (Paris) for sharing the human anti-Giantin antibody.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.203088.reviewer-comments.pdf