Angela E. Frimberger, University of Massachusetts Cancer Center – NRI Building, 55 Lake Avenue North, Worcester, MA 01655, USA. E-mail: Angela.Frimberger@umassmed.edu
Haematopoietic stem cells (HSCs) have been extensively characterized regarding in vivo engraftment, surface epitopes and genetic regulation. However, little is known about the homing of these rare cells, and their intrinsic motility and membrane deformation capacity. We used high-speed optical-sectioning microscopy and inverted fluorescent videomicroscopy to study highly purified murine lineage-negative, rhodamine-low, Hoechst-low HSCs over time under various in vitro conditions. We discovered extremely rapid motility, directed migration to stromal cells and marked membrane modulation. High resolution images with three-dimensional reconstruction showed the general presence of microspikes. Further, pseudopodia (proteopodia) were observed that were induced by stromal-derived factor-1 and steel factor. Proteopodia were directed towards and were quenched by stromal cells, at times bridged HSCs, and could rapidly retract or detach from cells. Proteopodia were also observed in vivo with homed HSCs in frozen sections of murine spleen, lung and heart. This is the first demonstration that HSCs are both fast and highly malleable in phenotype.
All body tissues arise from stem cells, and embryonic and tissue stem cells have become the focus of much study. A number of different stem cell systems have been discovered, including neural (Mehler & Kessler, 1999), hepatic (Thorgeirsson, 1996), gastrointestinal epithelium (Karam, 1999), prostate (Foster & Ke, 1997) and connective tissue (Bruder et al, 1994), but the best characterized is the haematopoietic system (Spangrude, 1991, 1992, 1994; Quesenberry, 1992; Quesenberry et al, 1997). Lymphohaematopoietic stem cells (HSCs) have been defined by their capacity to engraft and repopulate in vivo. They have been characterized as to surface epitopes, gene expression, and in vitro and in vivo multilineage cell production (Broxmeyer & Kim, 1999; Cooper & Spangrude, 1999; Quesenberry et al, 1999; Phillips et al, 2000). Recently, it has become clear that both murine and human HSCs express a variety of cytokine receptors and adhesion proteins. However, most HSC studies are performed using cell populations rather than on the level of the individual cell and, thus, little is known about the intrinsic motility, surface morphology and membrane deformability of these rare cells and the effect of these characteristics on homing.
HSC homing is the phenomenon by which stem cells infused intravenously specifically extravasate and engraft in the bone marrow and not in other organs. Homing is believed to rely primarily on adhesion molecule interactions between stem cells and stromal cells and extracellular matrix, but remains poorly defined. Recent studies indicate an inherent or rapidly induced capacity for homing after in vivo infusion of purified HSCs. Homing appears to be complete within 3 h (Cerny et al, 2000). We have defined a similar time frame for adhesion of engraftable stem cells to in vitro Dexter stroma (unpublished data), indicating that in vitro homing to pre-established complex stroma can serve as a model for in vivo homing. Moreover, much interest has been focused recently on mobilization of HSC populations from the bone marrow into the peripheral blood using cytotoxic drugs and cytokines (Papayannopoulou, 2000). Other studies have indicated relatively rapid induction of migration with exposure to the cytokines stromal-derived factor-1 (SDF-1) and steel factor (Gotoh et al, 1999). Stem cell mobilization is used clinically to increase the number of circulating HSCs for collection and transplantation, and the mechanisms of mobilization are relevant to clinical application, as well as basic biology. Nevertheless, the capacity for motility of individual HSCs, particularly in relation to a complex stroma, has not been well documented. Furthermore, despite the molecular characterization of HSCs, their three-dimensional surface morphology has not been closely studied. The morphology of HSCs is traditionally considered to vary between that of a transitional lymphocyte or a blast cell, however, a recent report demonstrated two different types of pseudopodia on a stem cell-enriched population and a haematopoietic progenitor cell line (Francis et al, 1998), indicating that the surface morphology of HSCs is more complex than had been thought.
As discussed above, the engraftment and molecular phenotype of HSCs is relatively well characterized, however, it has recently become clear that the phenotype is also highly plastic as to engraftment and adhesion receptor expression. Engraftable stem cells stimulated to transit the cell cycle by cytokines [interleukin-3 (IL)-3, IL-6, IL-11 and steel factor] experience a reversible loss of engraftment in late S/early G2 phase of the cell cycle (Habibian et al, 1998). Further, the adhesion receptor profile of murine haematopoietic progenitors and cell lines changes with cytokine incubation and cell cycle transit, resulting in changes in fibronectin adhesion and possibly other effects that may affect homing and engraftment (Becker et al, 1999). This phenotypical plasticity is in contrast to previous concepts depicting the stem cell as a quiescent, sessile cell possessing only unrealized potential. We have now studied highly purified [lineage-negative, rhodamine-dull, Hoechst-dull (lin−rholowHolow)] murine lymphohaematopoietic stem cells for their capacity to move, interact with other cells and alter their surface morphology.
Materials and methods
These studies were approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee. Six- to 8-week-old (18–22 g) BALB/c mice were purchased from Taconic Laboratories (Germantown, NY, USA). Mice were maintained in conventional clean conditions at the Animal Facility of the University of Massachusetts Medical Center, Biotech II building, and were given mouse food and acidified water ad libitum. All animals were acclimatized for at least 1 week prior to experimental use.
Fresh whole bone marrow was harvested by flushing bones from female BALB/c mice in phenol red-free Dulbecco's modified Eagle's medium (DMEM) low glucose (Gibco BRL, Grand Island, NY, USA) with penicillin 100 U/ml, streptomycin 100 μg/ml (Gibco), fungizone 0·0125 μg/ml (Gibco), 1 × 10−7 mol/l hydrocortisone sodium succinate (Upjohn, Kalamazoo, MI, USA) and 20% preselected horse serum (Lot AFG5429, Hyclone, Logan, UT, USA). Pooled harvested cells were counted on a Neubauer haemocytometer using crystal violet and suspended to a concentration of 4 × 106 per ml, then seeded into six-well plates (3·8 ml/well) with 25 mm round glass coverslips. Cultures were incubated at 33°C in 5% CO2 in humidified air, and fed weekly by removal of half the supernatant medium and replacement with fresh medium.
High proliferative potential colony-forming cell assay
Cells to be assayed were plated in 35 mm plastic culture dishes (Falcon) using a double-layer soft agar nutrient medium. Underlayers were 1·0 ml of 0·5% agar (Difco, Detroit, MI, USA) with α-modified Eagle's medium (α-MEM) (Gibco), with 20% fetal calf serum (FCS) (Hyclone) supplemented with glutamine (Gibco) and growth factors at the following final concentrations: colony-stimulating factor 1 (CSF-1, 5000 U/ml), granulocyte colony-stimulating factor (G-CSF, 5 ng/ml), granulocyte-macrophage CSF (GM-CSF, 2·5 ng/ml), interleukin-1α (IL-1α, 250 U/ml), interleukin-3 (IL-3, 100 U/ml), stem cell factor (SCF, 100 U/ml) and basic fibroblast growth factor (bFGF, 5 ng/ml) at optimal concentrations. Three to five plates were prepared for each experiment. The cells were suspended at a concentration of 200 cells per ml in a 0·5-ml overlay (for a total of 100 cells per plate) of 0·3% agar in α-MEM with 20% FCS. The plates were incubated at 37°C in 5% O2, 10% CO2 and 85% N2 at 100% humidity. After 14 d, the colonies were scored by the proportion of high proliferative potential colony-forming cells (HPP-CFCs) (very dense colonies over 0·5 mm diameter and moderately dense colonies over 1 mm diameter), low proliferative potential colony-forming cells (other cell clusters and colonies of over 50 cells) and total number of colonies.
Haematopoietic stem cell sorting
We isolated murine lineage-negative, rhodamine-dull, Hoechst-dull (lin−rholowHolow) cells using a modification of the method previously described (Wolf et al, 1993). This population has high engraftment capacity and, for the experiments reported here, averaged greater than 50% 7-factor HPP-CFCs; it therefore represents a highly enriched HSC population. Briefly, fresh whole bone marrow was harvested by crushing bones from BALB/c mice in phosphate-buffered saline with 5% heat-inactivated fetal bovine serum (PBS/5%HIFCS). Mononuclear cells were isolated using density gradient centrifugation and then incubated with rat origin lineage antibodies. Positive cells were depleted using anti-rat magnetic beads (Dynal A.S., Oslo, Norway). Remaining lineage-negative cells were stained with rhodamine 123 and Hoechst 33342 for 30 min at 37°C. Cells were washed and incubated twice for 15 min each at 37°C in stain-free buffer to allow dye efflux. Finally, cells were stained with propidium iodide to allow exclusion of non-viable cells on the sorter. Rhodamine-dull, Hoechst-dull cells were separated on a MoFlo high-speed cell sorter (Cytomation, Fort Collins, CO, USA). For each sort, a sample was cultured for HPP-CFCs.
For in vivo homing experiments (see below), lineage-negative, stem cell antigen-positive (lin−SCA+) cells were isolated. Lineage-negative cells were obtained as described above. The resulting lineage-negative cells were first stained with anti-rat fluoroscein isothiocyanate (FITC) to allow exclusion of any remaining positive cells, then with anti-SCA (Ly6A/E) phycoerythrin. Cells were also stained with propidium iodide to allow exclusion of non-viable cells on the sorter. Lin−SCA+ cells were separated on the MoFlo high-speed cell sorter.
Immediately after isolation, cells were stained with the fluorescent membrane stain PKH2 or PKH26 (Sigma, St. Louis, MO, USA), according to the manufacturer's instructions. Briefly, cells were washed in serum-free buffer and suspended in provided diluent at 2 × 107 cells/ml. The dye was diluted to 4 μl/ml of the provided diluent, then an equal volume of cell suspension was added to the dye and immediately mixed. Cells were incubated in dye for 30 s to 3 min, then an equal volume of serum was added to stop incorporation. Cells were washed four times in serum-containing media, then counted in trypan blue.
Sorted and stained cells were studied in several different conditions with and without the presence of pre-established stromal cell layers. In some experiments, cells were imaged immediately in media without cytokines or with stromal-derived factor-1 (SDF-1, 10 ng/ml), steel factor (10 ng/ml) or the two in combination. In other experiments, cells were imaged immediately in media without cytokines, and SDF-1, steel factor or the two in combination were added after 1 h. In a separate set of experiments, cells were incubated for 36–42 h in Teflon bottles in media containing IL-3 (50 U/ml), IL-6 (50 U/ml), IL-11 (50 ng/ml) and steel factor (50 ng/ml) before imaging.
Two different imaging systems were used to observe living HSCs over time under various conditions. In some experiments, cells were imaged on a custom-built, laser-illuminated, wide-field epifluorescence microscope equipped with a high-speed focus drive (1 μm/msec) for optical sectioning and a sensitive small format CCD camera that can acquire 128 × 128 pixel images at a maximum rate of 540 images/s (Rizzuto et al, 1998). Sixty-one plane three-dimensional (3D) images of fluorescent dye distributions in living cells were obtained in as little as 1–4 s with this system. For the period of the experiment, cells were maintained on a glass coverslip in a heated chamber, using gas exchange of the overlying air to maintain the pH of the media. Either plain glass coverslips or coverslips with pre-established stromal layers were used. These stromal layers were either stained with PKH-26 or unstained. Cells were observed in this system for up to 4 h with repeated two- and three-dimensional digital imaging. The raw 3D image data was digitally processed using an image restoration algorithm (Carrington et al, 1995) that reassigns out-of-focus photons to their true origin. Three-dimensional images were then reconstructed using the data analysis and visualization environment (dave) software. Other experiments used standard inverted fluorescence video microscopy (Nikon and Zeiss) with a mercury lamp for two-dimensional imaging. In these experiments, cells were placed on glass slides with glass coverslips, sealed with agar and inverted for observation in a heated, humidified compartment. In this system, a preparation of cells was observed and imaged for up to 30 min. The potential observation time in this system was shorter than for the custom-built microscope because photobleaching was more pronounced and the slide preparations were more prone to desiccation. In either system, the proportion of cells expressing proteopodia was determined by counting cells directly through the eyepiece, noting whether or not proteopodia were expressed and, if so, how many. For motility data, the digital system was used, and the x and y coordinates of the central point of the cell were recorded in conjunction with the time of the exposure, to calculate the distance travelled by the cell in the x-plane and y-plane, and the elapsed time between consecutive exposures within the same field.
In vivo homing experiments
C57BL/6 lin−SCA+ cells were stained with CFSE (5- and 6-carboxyfluorescein diacetate succinimidyl ester; Molecular Probes, Eugene, OR, USA), as previously described (Weston & Parish, 1990). Briefly, cells were resuspended at a concentration of 5 × 107 cells/ml in prewarmed PBS containing 1 μg/ml CFSE. Labelling was performed at 37°C for 15 min. Labelling was stopped by adding several volumes of ice-cold HBSS containing 10% FCS and the cells were recovered by centrifugation. Cells were resuspended in sterile saline for intravenous injection. Each animal received a total of 250 000 lin−SCA+ cells at a concentration of 1 × 106 cells/ml. At 3 h post transplant, the organs were removed and placed in O.C.T. Compound (Miles, Elkhart, IN, USA) and frozen at −20°C. Frozen sections were cut (8 μm) from each organ to examine for the presence of CFSE-positive cells. Analysis was performed using a Zeiss Axioplan 2 fluorescent microscope.
The level of statistical significance was set at 0·05. For determination of proteopod expression in cytokine-free vs. cytokine-containing conditions, data from three separate experiments were combined and the Cochran–Mantel–Haenszel test was used (SAS Procedures Guide, 1995). For determination of directionality of movement, a one-tailed Student's T-test was used to compare frequency and magnitude of movement in the negative vs. positive direction in the x-plane and in the y-plane separately.
In vitro homing and motility
In experiments using freshly isolated HSCs with pre-established stromal layers, the interaction of individual HSCs with individual stromal cells was readily apparent. HSCs were seen to briefly interact with a number of different stromal cells before establishing a longer-term and more intimate interaction with a particular stromal cell. This more intimate interaction lasted from 5 min up to the duration of the experiment (3 h or longer). We also recorded HSCs apparently in the process of ‘burrowing’ between the stromal cell and the coverslip (Fig 1). The HSCs also demonstrated rapid, directed motility between the various stromal cells with which they interacted (‘searching’). In one experiment, the velocity (mean ± SEM of 10 measurements) of a single HSC was 26·6 ± 2·1 μm/min, while in another experiment the velocity (mean ± SEM of 156 measurements) was 7·3 ± 0·4 μm/min, with a range of 0–19·4 μm/min, varying in clear relation to nearby cells. In one instance, we observed a maximum rate of movement of 160 μm/min. Furthermore, these measurements clearly showed directionality of movement. In the 156 measurements cited above, the total movement was 273 μm in the negative direction in the x-plane and 7 μm in the negative direction in the y-plane (Fig 2). Of the 156 measurements, 112 were negative and 17 were positive in the x-plane (the remainder were neutral), with total movement of 295·5 μm in the negative direction and 22 μm in the positive direction (significantly different; P < 0·0001). Among the same measurements, 58 were negative and 56 were positive in the y-plane, with total movement of 65·5 μm in the negative direction and 59·5 μm in the positive direction. Of 106 consecutive measurements when the cell was not interacting with stromal cells, 91 were negative (total 269 μm) and only four were positive (total 3 μm) in the x-plane (significantly different; P < 0·0001), while 48 were negative (total 55·5 μm) and 38 were positive in the y-plane (45 μm) for a net movement of 265·5 μm in the negative direction in x and 10·5 μm in the negative direction in y. Using both freshly isolated and cytokine preincubated cells, 82 ± 7% (mean ± SD of eight experiments) of HSCs had adhered to individual stromal cells (‘homed’) within 1 h of contact.
Morphology and membrane deformations: proteopodia
In experiments using both freshly isolated and cytokine preincubated cells, both with and without stromal cells, HSCs expressed various numbers and forms of membrane deformations (Figs 3 and 4). In all culture conditions, HSCs exhibited surface microspikes, occasionally with branching. We also observed cytokine-induced pseudopodial membrane extensions (proteopodia).
We observed cells and counted proteopodia-expressing and non-expressing cells for up to 4 h. In all experiments in which cells were exposed to SDF-1 or steel factor, proteopodia were seen on up to 17·5 ± 9% (mean ± SD of four experiments) of HSCs at any given time. Of expressing cells, 79% had 1–2 proteopodia, while the other 21% had 3–7 proteopodia. Proteopodia were either short, blunt and club-like, or long (up to 100 μm), slender and serpentine (Fig 3). The long proteopodia were themselves highly motile even when the cells were not in rapid motion. Proteopodia could form and retract rapidly or remain extended for at least 4·5 min, and we observed them in the process of being extended and retracted. In addition, we observed long proteopodia detaching from cells (Fig 4); whether this represents a variant of apoptotic vesiculation, another mechanism of membrane homeostasis, or some other phenomenon remains unclear. We also observed stem cells connected to each other by a proteopod ‘bridge’. The variety of proteopod behaviours observed in living cells are summarized schematically in Fig 5.
Cytokine induction of proteopodia
HSCs that had been incubated with steel factor, SDF-1 or a combination of both, formed more and larger proteopodia than cells not incubated in cytokines. When results of experiments in which cells from the same sorted population were counted under both conditions on the same day were compared, the percentage of cells expressing proteopodia was 0·7% in cytokine-free conditions and 7·7% in cytokines (P = 0·03; also see Fig 6). In all culture conditions, proteopodia were more likely to form and were larger when culture conditions began to deteriorate, either by evaporation or pH dysregulation (8·5–9·5) of the media.
Association of proteopodia with homing
In seven experiments in which proteopodia were seen in preparations containing stromal layers, we did not observe proteopodia on HSCs that had adhered to a stromal cell (‘homed’), but only on those that were motile and interacting with multiple stromal cells (‘searching’). Further, predominant proteopodia on these motile cells were consistently extended in the direction of a stromal cell.
Proteopodia in vivo
Using an alternate HSC separation consisting of lineage depletion and stem cell antigen staining (lin−SCA+), in conjunction with CFSE fluorescent labelling, we evaluated homing and morphology of infused HSCs in vivo. We observed proteopodia on CFSE-labelled lin−SCA+ cells in frozen sections of spleen, lung and heart taken from short-term recipients 3 h after cell infusion (Fig 7). In the 17 spleen sections counted, 124·1 ± 7·1 (mean ± SEM) labelled cells were observed per section; in 11 sections of lung, 66 ± 11 labelled cells per section were observed; and in 10 sections of heart, 0·6 ± 0·3 labelled cells were observed per section. In five spleen sections counted for proteopod expression, 58·7 ± 3·5% (mean ± SEM) of 570 CFSE-labelled cells expressed proteopodia. Similarly in lung sections, nearly half of the labelled cells were expressing proteopodia. Although bone marrow sections were prepared using a different method because of the need to decalcify for sectioning, no proteopod expression was found in homed CFSE-labelled cells seen in sections of bone marrow (Fig 8). These findings indicate that proteopodia exist in vivo after cell infusion.
These techniques allowed high resolution, two- and three-dimensional imaging of individual living stem cells over time, and our findings demonstrate the usefulness of such images in studies of stem cell biology and stem cell–stromal cell interactions. The images show both the rapid, directed motility and in vitro homing of individual HSCs, and the adhesive points of interaction between HSCs and stromal cells. Historically, HSCs are believed to home between stromal cells and the flask in Dexter culture and, in accordance with that belief, we recorded HSCs apparently in the process of ‘burrowing’ between the stromal cell and the coverslip (Fig 1). The finding of a high proportion (82%) of HSCs homing in vitro within 1 h is in agreement with recent studies indicating that homing after in vivo infusion of purified HSCs occurs within 3 h (Cerny et al 2000) and with other data from our group showing in vitro adhesion of engraftable HSCs to Dexter stroma in as little as 20 min (unpublished data). High resolution three-dimensional images of the adhesion interaction were obtained (Fig 3, Q). The speed of the motility that we observed, however, was unexpectedly great and the directionality of the movement was strongly significant.
Similarly, the finding of pseudopodia on a highly purified population of engraftable stem cells is previously unreported. Under optimal culture conditions, we observed pseudopodial membrane extensions that were induced (Fig 6) by SDF-1 and steel factor. These extensions were long and motile, could form and retract quickly or detach from cells, at times bridged HSCs, and were directed towards and quenched by stromal cells. In addition, more and larger pseudopodia were observed at extremes of pH (8·5–9·5) and with desiccation of the cultures. We have coined the term ‘proteopodia’ to describe these extensions, after the ancient Greek deity Proteus, the Old Man of the Sea, who could foretell the future and change shape at will. Francis et al (1998) have reported pseudopodia on a human haematopoietic stem cell-enriched population and on haematopoietic cell lines. That study did not observe highly purified stem cells, however, the pseudopodia described therein bear some similarities to the proteopodia we report here. In fact, the same group recently reported that podia are induced by increased osmolality (Oh et al, 2000), as are the proteopodia we report here. A variety of actin-based cell surface membrane projections have been described on different cells, including microvilli, pseudopodia, lamellipodia, microspikes and filopodia (Cooper, 1997). Lin−rholowHolow cells clearly show microspikes, but the proteopodia observed differ from these other projections, as outlined in Table I.
Table I. Characteristics of different membrane protrusions (Stossel, 1994 and Cooper, 1997; unless otherwise noted) compared with proteopodia.
Moderate width, bulbous
Phagocytosis, amoeboid movement
Redundant surface area to accommodate shape changes
Varied: either short and club-like or thin and very long
Periphery, varying in number
Directed movement to stroma
In both in vitro and in vivo settings, evidence was found to support the notion that proteopodia function in homing. In seven experiments in which proteopodia were seen in preparations containing stromal layers, we did not observe proteopodia on HSCs that had adhered to a stromal cell (‘homed’), but only on those that were motile and interacting with multiple stromal cells (‘searching’). Further, predominant proteopodia on these motile cells were consistently extended in the direction of a stromal cell. These observations suggest proteopodia function as an aid in homing. In addition, the finding of proteopodia in vivo on a large proportion of homing cells in various organs demonstrates that proteopodia can be visualized in vivo after cell infusion and suggests that they are not an artifact of in vitro conditions. Moreover, they concur with the observation that proteopodia were not seen on cells that had homed in vitro to bone marrow stromal cells. Proteopodia were a clearly observed and inducible phenomenon on HSCs and our findings suggest that proteopodia may play a role in the directed motility, homing and engraftment of HSCs, however, at present their actual role in these processes is not clearly established.
The formation and retraction of proteopodia and the degree of motility observed contradict the dogma of stem cells as quiescent, inactive, undifferentiated cells, and provide evidence for the active nature of stem cells with a phenotype that is both specific and plastic.
This work was funded by grants from the National Heart, Lung, and Blood Institute and the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health, and the Our Danny Cancer Fund. High-speed microscope and Data Analysis and Visualization Environment (DAVE) development were funded by grants from the National Science Foundation and National Heart, Lung, and Blood Institute. The image restoration algorithm used was created by Walter Carrington, PhD of the Biomedical Imaging Group. DAVE was created by Lawrence Lifshitz of the Biomedical Imaging Group. We wish to thank Steven Benoit, Jane Carlson, Bernice Fraioli, Houri Habibian, Ruud Hulspas, Lizhen Pang, Judy Reilly, and Allen Stering for technical assistance; Joanne Wuu for assistance with statistical evaluation; and Beth Mellor and Barry Davis for assistance with preparation of figures and video materials.