Microvasculature-directed thrombopoiesis in a 3D in vitro marrow microenvironment

Vasculature is an interface between the circulation and the hematopoietic tissue providing the means for hundreds of billions of blood cells to enter the circulation every day in a regulated fashion. The precise mechanisms that control the interactions of hematopoietic cells with the vessel wall are largely undefined. Here, we report on the development of an in vitro 3D human marrow vascular microenvironment (VME) to study hematopoietic trafficking and the release of blood cells, specifically platelets. We show that mature megakaryocytes from aspirated marrow as well as megakaryocytes differentiated in culture from CD34+ cells can be embedded in a collagen matrix containing engineered microvessels to create a thrombopoietic VME. These megakaryocytes continue to mature, penetrate the vessel wall, and release platelets into the vessel lumen. This process can be blocked with the addition of antibodies specific for CXCR4, indicating that CXCR4 is required for megakaryocyte migration, though whether it is sufficient is unclear. The 3D marrow VME system shows considerable potential for mechanistic studies defining the role of marrow vasculature in thrombopoiesis. Through a stepwise addition or removal of individual marrow components, this model provides potential to define key pathways responsible for the release of platelets and other blood cells.


Introduction
The adult human bone marrow releases nearly 500 billion cells into the blood each day [1,2]. Intravital imaging techniques have made it possible to visualize these complex processes in animal models, and have led to the identification of several pathways that mediate transmigration of cells from the marrow to the blood. These studies are largely conducted in rodents or zebrafish. However, the detailed interactions between the marrow vasculature and differentiated a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 blood cells, particularly at the terminal stages of maturation and blood cell release remain elusive for human cells. There is a growing appreciation that differences in scale between man and small animals are most likely not addressed by reiterating simple three dimensional cellular relationships to compensate for increased volume. Discrepancies in kinetics and outcomes in marrow regeneration between small animals and humans underscore this point [3][4][5][6]. Therefore, in vitro models are needed to model these phenomena and better understand human blood cell production and release. To address this, we have developed an engineered in vitro platform to approximate the vascular microenvironment (VME) and examine megakaryopoiesis.
The marrow is known for its complex architecture and diverse cell types. The vasculature, adipose tissue, fibroblasts, osteoblasts, osteoclasts, and hematopoietic cells have spatial relations that are critical for ordered blood cell production [7]. Recent studies, primarily from animal models, indicate that components of the VME can play more than one role in hematopoietic regulation [8][9][10][11][12][13]. For example, the endothelium contributes signals for lineage commitment, differentiation, and mobilization of progenitors [13][14][15]. In vitro work shows that human endothelial cells specifically support the development and differentiation of myeloid and megakaryocytic progenitors [16]. In small animals, hematopoietic stem cells (HSCs) localize in the perivascular space and, as they differentiate into megakaryocytes, they are in the perfect position to release platelets into the circulation [17][18][19][20][21]. In large animals, the difference in scale could mean that megakaryocytes may not be restricted to perivascular spaces alone. The release of platelets into the blood vessels requires that the megakaryocytes or some part thereof to come in contact with the vessel. The model presented here suggests that megakaryocytes migrate to achieve this end.
In vitro studies in liquid culture have proven useful for identifying cytokines and chemokines that contribute to hematopoietic regulation, cell proliferation, maturation, and motility [22,23]. Recent studies suggest the importance of physical factors that cannot be recapitulated in liquid culture, but can be approximated in 3D cultures [24][25][26][27]. A functional VME should include at the very least, architecture defined by the fibroblast, extracellular matrix [28,29], patent vasculature, and flow to facilitate processes such as platelet shedding [24][25][26]. Equally important for recapitulating a marrow VME is the inclusion of relevant cell populations in appropriate spatial relationships [26,30]. Systematically addressing these components should allow for an optimized vascular platform.
Here, we developed an in vitro microvessel system to investigate and identify critical components of the 3D marrow VME (Fig 1A and 1B). We show that megakaryocytes, either isolated from fresh human marrow aspirates or differentiated from hematopoietic progenitor cells in vitro, seeded into the matrix of the VME migrate over a two-week period to make contact with the vessel. Once in contact, they induce endothelial pore formation and release platelets into the lumen of the vessel. This closely approximates megakaryocyte behavior in vivo. Our study demonstrates the possibility of using such a 3D in vitro system to assemble the marrow microenvironment and examine complex hematopoietic processes.

Experimental methods
Bone marrow aspirates were collected under FHCRC IRB protocol 0999.209 stating that leftover specimens will be stored and used for an indefinite period of time. The cells isolated from aspirates were considered non-human subjects as no identifiable information was associated with the leftover specimen. Human umbilical cord blood was purchased from Bloodworks Northwest blood bank. Fresh peripheral blood was obtained under protocols and with written consent approved by the University of Washington (Protocol: HSD 45624).

Isolation of bone marrow aspirates and mononuclear cells
Marrow cells were obtained from discarded filters used to strain bone marrow following marrow aspiration from healthy donors in compliance with Institutional Review Board protocol, approved by the University of Washington and Fred Hutchinson Cancer Research Institute. The screen and filters containing marrow fibroblasts, stromal cells, bone pieces and fat cells were reverse perfused with 30 mL of phosphate buffered saline (PBS) and incubated for 10 min. Cells were dissociated from the screen and tubing with mechanical agitation. The resulting cell suspension was passed through a 20μm filter to separate out large cells, bone chunks, and fat globules from the smaller hematopoietic cells. Mononuclear cells, including megakaryocytes, from the filtrate were isolated through centrifugation with Ficoll-Paque (specific gravity 1.077) at 500 g for 30 minutes at room temperature. Both the unfiltered bone chunks, fat globules, large cells, and isolated bone marrow mononuclear cells (BMMCs) were resuspended and stored separately at 4C overnight in PBS with 10% fetal bovine serum (FBS) prior to use in vessel fabrication.
wt/vol (Hospira) was added to cord blood to a final concentration of 1.2%, and cells were processed by gravity sedimentation for 60 min. The leukocyte-enriched component was separated, centrifuged for 10 min at 300g and the supernatant was removed. The cell pellet was treated with ACK lysing buffer (Invitrogen) and washed with PBS. The cells were labeled with anti-CD34 antibody conjugated to magnetic microbeads (Miltenyi Biotec). The CD34+ fraction of cells was positively selected with an autoMACS separator, yielding over 90% CD34+ cell purity, as confirmed by flow cytometry (FACSCaliber).
Human cord blood or peripheral blood CD34+ cells were differentiated to megakaryocytes as described previously [31]. CD34+ cells were plated in 6-well plates at a density of 5x104 cells/ml and cultured in serum-free X-VIVO 10 medium supplemented with a cytokine combination consisting of IL3 (10 ng/mL), IL6 (10 ng/mL), SCF (10 ng/mL) (R&D Systems), and TPO (50 ng/mL, Peprotech) [31]. The suspension cultures were incubated at 37oC in a 5% CO2 humidified chamber. Media was changed after 7 days of culture. After 10 days, cells were collected and stained with PE conjugated CD41a antibody. CD41a+ megakaryocytes were sorted at >90% purity with a BD Biosciences FACSAria III sorter.

Washed platelets preparation
Fresh blood was drawn from healthy donors into 6mL ACD tubes (Solution B, BD Vacutainer) with written consent under protocols approved by the Institutional Review Board of the University of Washington. Washed platelets were isolated in a manner described previously [32]. Briefly, freshly drawn blood collected in 6mL ACD tubes was centrifuged at 120 x g for 15 minutes at RT medium acceleration and without brake. The platelet rich plasma (PRP) was transferred to a FACS tube using a transfer pipet. The PRP was centrifuged at 500 x g for 10 minutes at RT on slow brake and medium acceleration. The plasma was carefully removed leaving a pellet of platelets at the bottom of the tube. CGS buffer (13 mM Sodium Citrate, 120 mM Sodium Chloride,30mM Glucose, pH 6.5) was used to suspend the pellet gently and more CGS was added for a total volume of 10mL. Human recombinant PGI2 (Sigma-Aldrich) was added to the solution at 500ng/ml and the tube was inverted once to gently mix. A final centrifugation was performed at 400 x g for 10 minutes at RT with medium acceleration and slow brake. After the supernatant was removed, the platelets were resuspended in Tyrode's Buffer at half the volume of the original PRP volume.

Fabrication and culture of 3D engineered microvessels and marrow VME
A microfluidic network was built via soft lithography using collagen gel (6-7.5mg/ml), as described previously [33,34]. For marrow VME studies, isolated marrow cells and globules were mixed into the collagen at approximately 10 × 10 6 cells/mL. For permeability, migration, and maturation experiments, sorted megakaryocytes were added to the collagen at 10 6 cells/ mL, whereas for particle collection studies, unsorted megakaryocytes were added at 8-10 ×10 6 cells/mL. Cells were thoroughly mixed into collagen yielding a uniform distribution in devices.
To seed the devices, HUVECs were trypsinized and resuspended at a concentration of 5 x 10 6 cells/mL [33,34]. After removal of media from the inlet and outlet of the devices, 10uL of HUVEC suspension was added into the inlet of microvessel and allowed to attach at 37˚C for 15 min. After attachment, media was added to the inlet reservoir for perfusion culture. In bone marrow aspirate co-cultured vessels, endothelial networks were perfused with EGM and the non-inlet/outlet reservoirs were filled with X-VIVO 10 medium supplemented with IL3 (10 ng/mL), IL6 (10 ng/mL), SCF (10 ng/mL, R&D Systems), TPO (50 ng/mL, Peprotech) and EPO (2U/mL, Affymetrix eBiosciences). In megakaryocyte vessels, endothelial networks were perfused with EGM supplemented with 100ng/ml TPO. The media for all vessels was replenished every 12 hours. In gravity driven conditions, the flow rate peaked initially at approximately 10 μL/min and decreased with time until the inlet balanced with the outlet. This range of flow rates leads to a peak wall shear stress in the inlet or outlet vessels of approximately 10 dynes/cm 2 and an average of~0.1 dynes/cm 2 throughout the culture time.
For particle collection studies, megakaryocyte vessels were cultured under syringe pumpdriven flow starting 2 days after fabrication (Model 11 Plus, Harvard Apparatus). Syringes were connected to tubing (1/32"ID, 3/32"OD Silicon Tubing, McMaster) fit securely into the inlet with a tube-to-tube 90˚elbow connector. The flow rate was set at 3 μL/min so that the averaged wall shear stress in the inlet and outlet vessels remains 3 dynes/cm 2 . Perfusate was collected via outlet tubing connected to a FACS tube containing 500 μL ACD buffer (Solution B, BD Vacutainer) and 400 μL PBS (Lonza). Every 24 hours, perfusate with released particles and/or cells was collected and media was refilled.

Functional testing and FACS
Collected particles, whole blood, and washed platelets were analyzed using flow cytometry. Half of each sample was activated with 3U/mL thrombin for 5 minutes. All samples were fixed in 3.7% formaldehyde, washed in FACS buffer (2% fetal bovine serum in PBS), and stained for CD41a, IgG (BD Biosciences), DAPI, and 7AAD (Beckman Coulter) for 30 minutes at RT. The cells were washed and analyzed on FACS CANTO2. The number of particles collected from each vessel was calculated using AccuCount Ultra Rainbow Fluorescent Particles (Spherotech). Analysis was performed on FLOWJO. Fixed quiescent and activated particles were permeabilized with Triton X-100, stained with β-tubulin (1:100, Abcam) overnight, washed, and incubated with Alexa Fluor 488 for 1 hour. Particles were washed, resuspended into a 1% agarose solution, mounted on coverslips, and imaged with a Zeiss LSM 880 confocal microscope.

Immunofluorescence staining and confocal imaging
In situ fixation and immunofluorescence staining was carried out as described previously [33]. After 3 to 14 days of culture, co-cultured microvessels were fixed in situ by perfusion of 3.7% formaldehyde for 20 minutes, followed by three 15 minute washes with PBS. The devices were then perfused with blocking solution containing 2% bovine serum albumin (BSA) and 0.1% Triton X-100 (Invitrogen, Carlsbad, CA) before immunostaining. Primary antibody rabbit antihuman CD31 (Abcam), VE-Cadherin (Abcam), mouse anti-human ICAM1 (Abcam), or VCAM1 (Abcam) was diluted in the blocking solution and perfused through the vessel overnight at 4˚C. The devices were washed three times with PBS for 15 minutes each. The secondary antibody goat anti-rabbit Alexa Fluor 647 or Alexa Fluor 488 (Invitrogen) and nuclear counterstain Hoechst 33342 were then perfused through the vessel for one hour, and washed three times for 20 minutes each. Immunofluorescence z-stack images (step size 1-3μm) of microvessels were taken with a Nikon A1R confocal microscope with a 10x or 20x objective. Z-projections and cross sections were generated using Image J. Zoomed views of MKs on the vessel wall and 3D reconstructions of confocal images were generated using contour surface creation in Imaris.
Manual quantification of megakaryocyte migration and ploidy was performed in ImageJ. Migration was quantified using 5 image stacks (120 μm depth) each from 5 different vessels, with distance from the vessel wall normalized to the radius of each vessel. The number of lobes per megakaryocyte was quantified through manual lobe counting of Hoechst-stained MK nuclei from z-projected image stacks of 6 vessels. Data are presented as mean +/-SEM. Significant differences were determined with an unpaired Student's t-test, with significance considered at p < 0.05. The number of nucleus lobes in peripheral blood-derived megakaryocytes (Fig 2) and cord blood-derived megakaryocytes (S2 Fig) were reported separately.

Scanning electron microscopy imaging
The co-cultured devices were fixed in situ by perfusing 25% glutaraldehyde overnight before disassembly, where the vessel was opened to expose the luminal surface. The collagen was dehydrated in serial ethanol washes (50%, 70%, 85% and 100% ethanol) and critical point drying (Tousimis, SamDri-780). The vessels and matrices were then sputter coated with gold-palladium and analyzed by a FEI Sirion scanning electron microscope with an accelerating voltage of 5 kV, spot size 3.

Transmission electron microscopy imaging
Microvessels were fixed in half-strength Karnovsky's solution (2% paraformaldehyde/2.5% glutaraldehyde in 0.2 M cacodylate buffer). Microvessels were disassembled and fully immersed in the same fixative solution for several days. Samples were rinsed in 0.1 M cacodylate buffer then post-fixed using 2% OsO4 in 0.2 M cacodylate buffer followed by another rinse with 0.1 M cacodylate buffer. Sample dehydration was performed using immersions in graded solutions of ethanol, then propylene oxide (PO), before 1:1 PO/Epon 812 (Ted Pella Inc) immersion overnight. Fresh Epon 812 was then exchanged for 2 hours after which the blocks were cured for 48 hours at 60˚C. Ultrathin sections (70 nm) were cut from blocks using a diamond (Diatome US) blade on a Leica EMUC6 ultra-microtome and placed onto grids. Grids were stained with uranyl acetate for 2 hours and lead citrate for 5 minutes. Sections were imaged using a JEOL JEM-1400 Transmission Electron Microscope (JEOL Ltd.) using 100 kV acceleration voltage. Images were acquired with a Gatan Ultrascan 1000XP camera (Gatan, Inc.).

Measurement of microvessel permeability
To measure barrier function of HUVEC only, megakaryocyte co-cultured, and HUVEC with megakaryocyte-conditioned media vessels, 40kD FITC-Dextran (Sigma) was perfused through the microvessels in situ. Fluorescent confocal images were acquired at 1 frame/second for 10 minutes. The image sequences were analyzed with Matlab to estimate the permeability coefficient of dextran in collagen based on the model developed in a previous publication from our group [33].

Recapitulating a 3D human marrow VME
We engineered a 3D human marrow VME in type I collagen (7.5 mg/mL) using lithographic processes described previously (Fig 1A) [33]. We embedded cells from healthy human marrow aspirates around microchannel networks in a type I collagen gel (Fig 1B). HUVECs were seeded in the lumen of microchannels within the gel and cultured under perfusion. The endothelial cells formed a single-layer vessel with cobble-stone morphology and junctions at regions of cell-cell contact (Fig 1C). Immunofluorescence images suggested that co-cultured endothelial cells were in an activated state with varying degrees of CD31 surface expression (Fig 1C and 1D) and high VCAM-1 and ICAM-1 expression (Panels D-E in S1 Fig). The vessel walls also contained openings or pores, forming a discontinuous endothelial cell layer (arrows, Fig 1D), reminiscent of pores reported on the sinusoidal marrow endothelium [35,36]. Scanning electron microscopy allowed for an ultrastructural analysis of the whole VME ( Figure A  in S1 Fig) and close-up views of cell-cell interactions (Fig 1E and 1H). These ultrastructural analyses confirmed the presence of pores in the endothelium, ranging from less than a micron to several microns in diameter (Fig 1F, and Panels B-C in S1 Fig). These pores only appeared in hematopoietic cell co-culture. The luminal surface contained many platelet-like structures, ostensibly derived from mature megakaryocytes seeded in the matrix, including pro-platelet clusters ( Fig 1G) and string-like structures that appeared to be membrane-bound cytoplasmic beads (Fig 1H).

Megakaryocytes migrate towards the vessel as they mature
To obtain more mechanistic insights into the differentiation of specific hematopoietic lineages, we modified the marrow VME to specifically study thrombopoiesis. Human CD34+ progenitors, isolated from either umbilical cord blood or mobilized peripheral blood, were differentiated into megakaryocytes [31]. After 10 days of culture, 30-70% of the cells expressed the megakaryocyte marker CD41a (Figure A in S2 Fig). CD41a+ megakaryocytes were then purified by flow sorting and embedded in the type I collagen gel used to cast the microvessel platform (Fig 2A) [33]. The microvessels were cultured in endothelial growth medium supplemented with thrombopoietin (TPO) under gravity or syringe pump driven flow for 3 to 14 days. Due to the network geometry, the vessel branches further away from the inlet and outlet have wall shear stress approximately fifty folds lower than the inlet and outlet [32]. The range of flow conditions (averaged to 0.1 dynes/cm 2 in gravity driven conditions, and 3 dynes/cm 2 under syringe pump conditions) mimics the very low wall shear stress in the small arterioles of the marrow, estimated in previous literature to range from 0 to 4.6 dynes/cm 2 [37][38][39].
After 3 days of culture, endothelial cells formed junctions at regions of cell-cell contact, indicated by CD31 expression (Fig 2A and 2B). Though the majority of CD41a+ megakaryocytes remained in the matrix, some moved to the abluminal side of microvessel, and others appeared on the luminal surface of the vessel (Fig 2B) After 3 days of culture, megakaryocyte density increased over three fold near the vessel wall compared to the initial seeding density of 1 million cells/mL (Fig 3A.i and 3B and Figure D in S2 Fig). Concurrently, megakaryocyte density decreased to one fifth of the original density at a distance greater than 300 μm (3 times the vessel diameter) from the vessel wall (Fig 3B), suggesting that the megakaryocytes moved towards the vessel wall. Live imaging confirmed this phenomenon, as we observed megakaryocytes in the matrix actively migrating towards the vessel wall during culture (S1 Video, S2 Video, S3 Video). Canine megakaryocytes showed the same migration pattern. Megakaryocytes isolated from dog marrow were engineered to express GFP under the PF4 promoter and cultured in the same system [40]. After 3 days of culture, the majority of these megakaryocytes closely associated with the vessel wall (Panels A-C in S3 Fig). In contrast, a control human bone marrow stromal cell line (HS5) cultured in the matrix at the same cell density showed no significant change in cell density in relation to the vessel walls throughout culture (Fig 3B).
The maturation of megakaryocytes in vivo is driven primarily by thrombopoietin (TPO) [19] and stromal cell-derived factor 1 (SDF-1/CXCL12) [23], among other growth factors [41]. As they mature, megakaryocytes upregulate expression of CXCR4 and respond to CXCL12 (SDF-1) signaling, which directs their migration within the hematopoietic microenvironment [20,23,24,35,42,43]. To examine the role of CXCR4/CXCL12 signaling in the migration of megakaryocytes in our system, we studied the effect of a neutralizing CXCR4 antibody. In anti-CXCR4 treated cultures, megakaryocytes remained in the matrix with no migration after three days of culture (Fig 3Aii and 3B), suggesting that CXCR4/CXCL12 signaling is necessary for megakaryocyte migration and their interactions with the microvessels.
The presence of megakaryocytes in the matrix also affected microvascular permeability. FITC-conjugated 40-kDa dextran was perfused through the microvessels to estimate the permeability coefficient K of the endothelium, in an approach similar to one described previously ( Figure H in S3 Fig) [33]. The presence of megakaryocytes near the vessel increased microvessel permeability, with the permeability coefficient K = 0.31 ± 0.15 μm/s, nearly 10 fold higher than in microvessels without megakaryocytes (K = 0.032 ± 0.01 μm/s). Megakaryocyte conditioned media in HUVEC-only vessels decreased barrier function minimally (K = 0.11 +/-0.03 um/s, S4 Fig). Neutralizing CXCR4 antibodies restored the barrier function of the microvessels (K = 0.036 ± 0.014 μm/s, p<0.05) to a value similar to that of vessels without megakaryocytes ( Fig 3C) [33]. Scanning electron microscopy and confocal microscopy revealed that pores of 1-10 μm developed in the vessel wall during co-culture with megakaryocytes, similar to those seen in microvessels co-cultured with marrow aspirates. However, these pores are not seen in HUVEC-only vessels with or without conditioned media ( Figure I in S3 Fig). The pores or fenestrae likely account for the increased vessel permeability (Fig 3D).

Megakaryocytes penetrated the vessel wall and released platelet-like particles
In the marrow, megakaryocytes must migrate across or extend processes through the vessel wall into the lumen to release platelets [18,35,44]. However, it is unclear when or how megakaryocytes transmigrate through the endothelium, mainly due to the lack of access to the marrow and complex microenvironment. Our system allowed for close-up examination of the interaction of megakaryocytes and the vessel walls in real time. Megakaryocytes were observed to develop multiple processes that extended towards the vessel wall, migrated into the lumen, and released platelet-like particles (S2 Video). Confocal and electron microscopy revealed different stages of this process (Fig 4). Some megakaryocytes resided completely on the abluminal surface of the microvessel (Fig 4A). Some other megakaryocytes occupied both the abluminal and luminal space, apparently transmigrating through the endothelium while undergoing membrane demarcation (Fig 4B) and fragmentation into clusters or pro-platelet strings on the vessel wall (Fig 4C). Large megakaryocyte fragments or whole megakaryocytes were also found on the vessel walls (S5 Fig, S6 Fig), where they would be expected to finalize their maturation and fragment into platelets while in circulation. We next examined the capacity of our marrow VME to generate platelets. We embedded 7-10 × 10 6 /mL differentiated megakaryocytes (unsorted after 10 days of differentiation culture) in the collagen matrix for co-culture with HUVEC-lined microvessels under flow. After initial two days of culture, we collected the effluent perfusate every 24 hours (Fig 5A)  released particles/cells ranged from the size of microparticles to that of red blood cells, as defined by the forward-scatter plot of washed platelets and whole blood, along with plateletspecific CD41a and CD42b analysis (Fig 5B). The collected perfusate contained an average of 1.82 × 10 6 platelet-sized particles, regardless of granularity, with over 550,000 CD42b+ particles in approximately 2 mL per thrombopoietic VME device per day. Each device contained microvessels with a surface area of 0.53 cm 2 and a volume of 1.6 μL. Considering that megakaryocytes within 100 μm of the vessel wall were able to generate particles and release them into the circulation, we calculated the yield of CD42b+ platelet-sized particles per megakaryocyte per day to be approximately 27, without counting any particles adhered on the vessel wall. Cytoskeletal rearrangement was evident when particles were activated with 3U/mL thrombin for 5 minutes by immunofluorescence staining of CD41a and β-tubulin (Fig 5C and 5D).

Discussion
The bone marrow is essentially a liquid tissue with heterogeneous population and complex architecture. Stem cell and maturing components are not ordered in recognizable spaces as they are in epithelia. This poses a problem in visualizing hematopoietic differentiation, particularly in the context of large animals. In the past decade, advances have been made to visualize thrombopoiesis in mouse models via intravital microscopy. These studies show that marrow megakaryocytes released pro-platelets into the circulation with an estimated small number of platelets per megakaryocyte to meet the need for physiological platelet turnover [18,44,45]. The process of pro-platelet formation has been recapitulated in vitro, and also in this case, low numbers of platelet-like particles per megakaryocyte are produced [24,26,46]. Recently, attention has been paid to finding approaches or mechanisms to enhance the number of platelets generated per megakaryocyte. Under some stress conditions, marrow megakaryocytes were found to generate very large numbers of platelets to meet acute platelet need, potentially releasing them through rupture within the blood vessels, though this phenomena is under debate [21,47].
Here, we used hydrogel matrices composed of 0.75% type I collagen and 99.25% water to reconstruct part of marrow tissue and study the close interactions between megakaryocytes and microvessel walls. We showed that megakaryocytes from freshly aspirated marrow modified the endothelium of microvessels, interacted with vessel walls, and released platelet-like particles into the lumen. Although this system can benefit from further optimization and biological analysis, it validates the approach of using an engineered microvessel system to model the human marrow VME. We embedded purified CD41+ megakaryocytes in the engineered marrow VME, and for the first time, detailed the specific interactions of megakaryocytes with the vessel wall. The combination of three types of imaging data, namely live imaging, immunofluorescence, and scanning electron microscopic images, confirmed that megakaryocytes migrated towards the vessel and created pores in the endothelium, through which they either transmigrated or extended pro-platelets into the vessel lumen to release platelets. We showed that HUVECs, known to form continuous endothelial surfaces in both large vessels [48] and within in vitro microvessels [33], were modified by the surrounding megakaryocytes to become discontinuous and leaky. This structural phenotype is found in bone marrow endothelial cells, which are reported to be highly fenestrated [23,43,[49][50][51][52]. The mechanisms behind MK-induced pore formation have been suggested to arise from the matrix-degrading properties of the MK podosome, which are increased in the presence of endothelial basement membrane and SDF1 [53][54][55] In contrast, megakaryocyte conditioned media alone did not change the continuity of HUVEC microvessels. This suggests that the microenvironment can modulate specific endothelial cell phenotypes to adapt to local functional needs, which may be one source of endothelial cell heterogeneity [56,57].
We showed that megakaryocyte migration is mediated in part by CXCR4-CXCL12 signaling, as blocking CXCR4 led to reduced megakaryocyte migration. Though not the only chemoattractant responsible for MK migration in vivo, this platform provides opportunities to further investigate and modulate the paracrine signaling and cell-cell interactions in thrombopoiesis, particularly in a system that allows for the migration of cells from large distances away from vasculature. In addition, our system allows for the precise control of flow and perfusion of biochemical cytokines through the microvessels and particle collection from the outlets. Throughout our culture, flow has been maintained by gravity or syringe pump. The flow rate and resulting wall shear stress mimics a biologically relevant range, which has been reported as extremely low, yielding wall shear stress range between 0-4.6 dynes/cm 2 [37][38][39].
We showed different processes of platelet release, namely pro-platelet territories, plateletlike blebs, and large fragments found in the vessel lumen. Mature megakaryocytes normally release pro-platelets, which are long cytoplasmic extensions with a beaded structure that fragment into platelets in circulation [18,44,58]. We see CD41+ megakaryocytes localize around the vessel wall and extend processes through the endothelium to release particles. The generated platelet-like particles found on the vessel lumen or collected have similar morphology as seen via ultrastructure assessment from electron microscopy and marker expression from immunofluorescence microscopy. Though Nishimura et al show that megakaryocytes can rupture in the presence of inflammatory cues to meet acute platelet need, this phenomena is not widely observed [47,59]. Recent work by Itkin et al. identified different roles of distinct marrow blood vessel types in hematopoiesis: less permeable arterial vessels support a low reactive oxygen species environment, whereas more permeable marrow sinusoids promote the activation of hematopoietic progenitor cells and immature cell trafficking [15]. In our system, endothelial cells appear to be activated and display structural and functional changes, indicating an inflammatory environment. Megakaryocyte-induced permeability could provide sites for immature megakaryocytes or large fragments to migrate across the endothelium, in combination with the lack of mature basement membrane formation. In normal marrow, whole megakaryocytes typically do not transmigrate from the marrow into the venous blood in vivo. This may impact the release of megakaryocytes and/or platelets, leading to the multiple types of megakaryocyte fragmentation phenomena observed in our system.
The advantages of our system include controllable cellular composition, matrix, vascular structure, flow, and a 3D geometry, and preserves the capacity for high quality imaging. Nevertheless, there are still improvements that can address technical limitations present in our platform. For example, the vessel wall consists of a single layer of endothelium with a 100 μm diameter, which is larger than the marrow sinusoids. This diameter may induce different stresses and biophysical forces on the vessel walls and affect the cellular interactions between megakaryocytes and the endothelium. Future studies could examine the effect of vessel diameter on thrombopoiesis, ranging from 30 μm, the lower limit of our fabrication technique, to sub millimeter or examine the impact of vessel density on megakaryocyte migration. In addition, we observed large numbers of platelet-like particles adhered to the luminal wall, rather than flowing in the perfused media. It is unclear whether this occurs in vivo, and if so, whether physiological mechanisms exist to detach the platelets. It is possible that such adhesion may be an in vitro artifact due to high serum content in the media, which activates platelets after their release. In addition, the endothelial cells appeared to have become activated and fenestrated during culture. It would be interesting to re-examine this phenomenon with bone marrow sinusoid endothelial cells or endothelial cells that do not require serum in vitro to better mimic the marrow vasculature [60]. It is also expected that bone marrow endothelial cells have a unique phenotype, representing a less inflammatory and adhesive surface [15,61]. Adhesion blocking reagents can also be introduced through the vessel wall to allow for better release of particles. Finally, our collected platelet-like particles appeared to lack packaged granules present in human platelets, which could be improved by addition of blood proteins in the media or additional matrix components in biologically relevant gradients, such as laminin, fibronectin, or collagen IV. The production of proper granules could also lead to a more typical morphology in activated particles. These modifications to the culture system could improve the shortcomings of a collagen-based platform and yield more homogenous particles without preactivation. Nevertheless, our system demonstrated a functional in vitro human marrow VME that lends itself to future mechanistic studies on cell-cell and cell-matrix interactions in the marrow.