Embedding of cardiac spheroids within microvascularised microfluidic chips
Microfluidic chips for spheroid implantation were generated based on previously reported chips presenting three parallel microchannels separated by series of microposts12,28. The dimensions of the central channel were revised29 to allow simpler injection of the fibrin gel and implantation of the spheroid. In addition, one of the channels enabling injection of the gel in this central compartment was tilted at an angle of 30°, as this was found to limit risks of leakage and formation of bubbles within the central compartment. Microfluidic devices were prepared via photolithography and soft lithography, following established protocols29,30. Overall, the final design displayed a vascularised chamber (Fig. 1A, 2 in the device schematic), where a gel containing HUVECs or HUVECs/pericytes was injected through the corresponding tilted inlet, and a central 3 mm cell culture well where a cardiomyocytes-endothelial-fibroblast (CMEF) spheroid was cultured in a fibrin gel containing HUVECs or HUVECs/pericytes (Fig. 1A, 3 in the device schematic). The vascularised chamber was flanked by two lateral channels (Fig. 1A, 1 in the device schematic) connected to medium reservoirs. The setup was designed to enable opening of the vasculature onto the side channels, therefore allowing delivery of therapeutics to the CMEF spheroids through the formed microvasculature (Fig. 1 and Supplementary Fig. S1).
Impact of CMEF embedding on microvascularisation of the CMEF
The CMEF spheroids were cultured in three configurations (Fig. 1A) for 10 days. Spheroids were cultured in suspension in a ULA plate as control (referred to as “suspension”), in a ULA plate embedded in fibrin gel (referred to as “fibrin”) and in microfluidic chips, in fibrin gel and integrated within microvascularised networks (referred to as µFC). Bright field imaging (Fig. 1A) indicated that spheroids retained cohesion for at least 10 days in all culture configurations. While the spheroids in suspension retained their round morphology, spheroids embedded in fibrin (both in ULA plates and in the µFC) resulted in some cell scattering and migration in the surrounding matrix. Furthermore, confocal images in Fig. 1A show vascularised spheroids (stained with cardiac troponin T, cTnT) surrounded by a CD31+/NG2+ vascular network which connects the central well with the lateral medium channels (CD31 is an endothelial cell marker and NG2 is Neural/Glial Antigen 2, a marker associated with pericytes).
Four embedding conditions within µFCs were tested (Fig. 1B), in order to explore how these methodologies and associated integration of the spheroids within the hydrogel and vasculature, would impact on the architecture of the resulting constructs. The microvasculature was either composed of HUVECs monocultures or HUVECs/pericytes co-cultures. HUVECs were selected based on their broad availability and use in a range of microvascularised models (including for therapeutics testing)31. These cells were also found to display relatively good cell-to-cell homogeneity and displayed comparable phenotypes to other endothelial cell types, with greater variability at the single cell level than between cell types32. Pericytes are mural cells which have been shown to inhibit vessel hyperplasia and improve barrier properties of vascular networks33,34. These cells were previously found to improve the stability of microvascular networks in microfluidic chips, including when exposed to stressful culture conditions (serum starvation) or nanoparticles displaying toxicity35. CMEF spheroids were embedded at two time points: either simultaneously with vascular cells (early embedding) or 4 days after on chip vasculature formation (late embedding). Immunostaining and confocal microscopy did not indicate vascularisation of spheroids with the late embedding method, as shown in Fig. 1B and Supplementary Fig. 2A. While a vascular network formed in the chips, this failed to integrate with the spheroid and orthogonal cross-sections of confocal z-stack clearly indicate large gaps between the embedded spheroid and the microvascular bed (700 µm; Supplementary Fig. S2A). This is presumably due to the fibrin matrix used during embedding to keep the spheroid in place. Due to this gap, we were very often unable to image the spheroids in this configuration, as the objective focal depth was not extending far enough (this was particularly striking in the case of HUVECs/pericyte co-cultures, see Fig. 1B).
At early embedding, compared to HUVECs/pericytes co-cultures, HUVECs vasculatures were found to be denser and more interconnected (Fig. 1B). We also surprisingly found NG2+ cells wrapping around HUVECs mono-culture networks (Fig. 1B). The morphology of these networks resembled the morphology of HUVEC/pericyte networks previously studied35. NG2+ cells observed in HUVECs mono-cultures after spheroid embedding were proposed to originate from the spheroids, potentially corresponding to perivascular smooth muscle cells differentiated from cardiac endothelial cells, as previously reported36. On the other hand, HUVEC/pericyte networks seemed more disrupted and showed a weaker overlap of the vascular network and spheroid (Fig. 1B). This could result from the antagonistic effect of spheroid-derived NG2+ cells and the NG2+ pericytes introduced in the microvascular bed.
Other conditions that were explored included late embedding with the addition of extra HUVECs in the fibrin gel used to introduce the spheroid (Supplementary Fig. S2B), which led to comparable vascular networks but brought the spheroids further away from the max focal plane of the objective. We also investigated the impact of hypoxia in HUVEC/pericyte networks but found that the networks were severely disrupted (Supplementary Fig. S2B). Cultures in which VEGF was omitted from the medium resulted in disrupted vascular networks (Supplementary Fig. S2B). For the rest of study, models composed of HUVECs vasculatures will be designated as “H” and HUVECs/pericytes vasculatures as “HP”.
Morphological analysis of the microvasculatures and spheroids
Morphological analysis of the vasculatures formed after 10 days of co-culture (with embedded spheroids) indicated that H-vasculatures were better developed compared to HP-vasculatures (with both early and late embedding), with network areas more than double (Fig. 2; 350,000 ± 40,000 µm2 in H and 150,000 ± 30,000 µm2 in HP, p = 0.007). Furthermore, while most vessels were interconnected in H-vasculatures, this was not the case for HP co-cultures: as a result, the number of independent networks per areas of interest was 5 ± 2.8 and 84 ± 17 in H and HP early embedding respectively (Fig. 2, N of networks/area). The number of branches per network area was almost half in H compared to HP (2.6 10−3 ± 0. 6 10−3 and 4.6 10−3 ± 0.4 10−3 µm−2, respectively). The average branch diameter was instead not statistically different for H compared to HP, in early embedding (17 ± 3 µm and 10 ± 2 µm). This is in agreement with the hypothesis that NG2+ cells found in H vasculatures play similar functions to pericytes added directly as co-cultures in HP networks, preventing hyperplasia. This was not observed for H networks with late embedding, in agreement with the delaying of interactions between the network and spheroids and the increased gap separating the two compartments. Similarly, we found that the number of junctions per network area was lower in the H vasculature (Supplementary Fig. S3), although only statistically significant for H networks with late embedding, while the total branch length was similar in all conditions. The average branch length of H networks with late embedding was also maximum, with 63 ± 14 µm compared to 18 ± 1 µm for HP cocultures.
The morphology of CMEF spheroids changed upon embedding in a fibrin gel, with reduced roundness and increased scattering, but remained cohesive and did not disrupt (Fig. 3). While the spheroids in suspension had a compact morphology and smaller area (77,000 ± 4,100 µm2 at day 0 and 94,000 ± 4,600 µm2 at day 10), spheroids cultured for 10 days in fibrin or in the device showed larger areas (181,000 ± 8,400 and 168,000 ± 30.000 µm2, respectively). Accordingly, the perimeter of the spheroids was also increased after embedding (11,000 ± 1,800 and 7,400 ± 700 µm in fibrin and µFC respectively, compared to 1800 ± 170 at day 10 in suspension). As expected, circularity was much lower in the embedded spheroids (0.025 ± 0.008 and 0.044 ± 0.013 in fibrin and µFC respectively, compared to 0.41 ± 0.06 for day 10 suspensions). Aspect ratios also increased, while roundness and solidity decreased (supplementary Fig. S3).
Structure of CMEF spheroids embedded in µFCs
The architecture of the implanted CMEF spheroids was investigated next. Cytoskeletal, junctional and matrix expression were characterised to establish how spheroids integrated within surrounding networks. Spheroids in suspension, in fibrin and embedded within microvascular networks in µFCs (H with early implantation) were fixed and immunostained prior to confocal imaging. Vimentin was selected as a fibroblast marker37, although it is not exclusive to this cell type. Vimentin was found in all spheroids and conditions, in agreement with their initial composition (Fig. 4). In µFCs it was also localised with the microvascular networks, presumably as it regulates cell–cell adhesion integrity and blood vessel remodelling38,39. Neural/glial antigen 2 (NG2) has been used as a marker for mural cells, particularly pericytes and smooth muscle cells in the vasculature40,41. While there was a degree of overlap between NG2 and vimentin staining in spheroids alone (with and without fibrin, Fig. 4), in µFCs NG2+ cells aligned along the network as well as being associated with the spheroid body. In contrast, cardiac troponin T (cTnT) staining, a cardiomyocytes marker indicating maturation of the tissue42, was specifically associated with the spheroids and was apparently associated with the cytoskeleton, with evidence of striated pattern formation, demonstrating the integrity of the spheroids after vascularisation in µFCs.
Cytoskeletal markers were next examined. We observed striated patterns of F-actin and α-actinin cytoskeletons within spheroids (Fig. 5A), whether in suspension or when embedded in microvascularised µFCs. This suggests sarcomere formation43 in both conditions. Interestingly, striated α-actinin structures were also observed on “day 0” of spheroid cultures in suspension, but were less organised at this stage (Supplementary Fig. S4). A striated pattern was also noticeable in non-microvascularised fibrin gels, but less clear, presumably due to the reduced cohesion of these spheroids and the more apparent scattering of cells (Supplementary Fig. S4A). Non-muscle myosin II, which is required in the assembly of nascent myofibrils44, was also found in cardiac spheroids in suspension (Fig. 5A and S4A), and localised along vascular network in vascularised spheroids.
The junctional markers N-cadherin and β-catenin were also expressed in spheroids in suspension and in vascularised spheroids, both in the spheroid mass and associated with the vasculature (Fig. 5B and Supplementary Fig. S4B). CD31 staining further confirmed the vascularisation of spheroids in the device. A rudimentary vascularised structure was also visible in the spheroids in suspension (with or without fibrin gel), consistent with their composition. Expression of other proteins associated with the contractility machinery, myosin heavy chain (MHC) and myosin light chain (MLC), was also observed, further confirming the maturation of sarcomere structures (Supplementary Fig. S5). α-SMA, a marker used to identify vascular smooth muscle cells41 and myofibroblasts45, was expressed in the spheroids in suspension (day 0 and day 10, Supplementary Fig. S5) and in the device, in cells scattered along the vasculature. A cell population expressing PDGFRβ was also identified in the spheroids (Supplementary Fig. S5), further confirming the presence of mural cells in the spheroids, which are proposed to underly the stabilisation of microvascularised spheroids in the absence of pericytes.
Finally, we investigated extracellular matrix deposition. Cells in cardiac spheroids in suspension deposited laminin, fibronectin and collagen IV (Supplementary Fig. S6). This matrix was then highly remodelled once the spheroids were embedded in fibrin gels and in µFCs, where there was deposition by HUVECs too. Note that collagen IV was mainly deposited basally in microvascular networks, presumably by HUVECs, whereas fibronectin was also found in the peri-vasculature.
Longer term cultures (up to 25 days) were also explored. This resulted in vascularised spheroids with retained structure and presenting comparable expression of markers expected and observed at earlier time points (day 10; Supplementary Figs. S7 and S8). In particular, a striated pattern of α-actinin/F-actin can still be clearly seen in both spheroids, whether in suspension or embedded. Overall, these results confirmed the formation of well-structured vascularised spheroids in µFCs, displaying hallmarks of a striated contractile cytoskeleton typically associated with cardiomyocyte maturity.
Functional properties of vascularised cardiac spheroids
The functionality of the vascularised spheroids was next examined (early embedding in H-vasculature). The contractility of CMEF spheroids in suspension, in fibrin and in µFCs was monitored over 25 days of culture (Fig. 6a, Supplementary Fig. S9 and Supplementary Videos S1–4). Upon embedding spheroids in µFCs, contractility was perturbed but then restored over 48 h. This is likely due to adjustment to the new mechanical context and matrix remodelling within the new environment. Spheroids embedded in µFCs had comparable beating patterns as those kept in suspension (average over 10 days was 37 and 35 beats/min respectively). On day 5, rates were 45 ± 2.5 and 52 ± 2.5 beats/min in suspension and in µFCs respectively (statistically non-significant). Similar rates were also observed for spheroids embedded in fibrin and with late embedding in H-vasculatures in µFCs (Supplementary Fig. S9).
Perfusability of the spheroids was then investigated via a 10 kDa FITC-dextran assay (Fig. 6B and Supplementary Video S5). Dextran was added in medium in the lateral channels and allowed to perfuse into the network passively. Dextran can be seen to perfuse through the microvasculature, reaching all the way into the spheroid, even at day 10. The video recorded suggests that flow through the microvasculature may occur in response to repetitive contractions of the spheroids (although this remains to be demonstrated formally). To confirm lumen formation in the microvascular network formed, podocalyxin staining and confocal microscopy imaging was carried out (Fig. 6B). In agreement with the lumenal (apical) recruitment of podocalyxin46 and the perfusability of networks with FITC-dextran, networks displayed a clear continuous localisation of podocalyxin. Therefore, microvascularised cardiac spheroids in µFCs are perfusable and functional from a biomechanical point of view.
Application of vascularised cardiac spheroids for safety testing of therapeutics
Having demonstrated structural integrity, functionality and perfusability of CMEF spheroids embedded in µFCs, the application of these in vitro models for safety testing of therapeutics was explored. As a proof of concept, vandetanib was injected in vascularised spheroids through the µFCs (after 10 days co-culture), as a crude mimic of systemic delivery. In addition, this therapeutic was supplemented to spheroids cultured in suspension, for comparison. Vandetanib is a tyrosine kinase inhibitor used in the treatment of advanced stages of aggressive and symptomatic medullary thyroid cancer47. It targets vascular endothelial growth factor (VEGF) receptors and was proved to reduce tumour cell–induced angiogenesis in-vivo48. The recommended daily dose is 300 mg for adults and one of the common side effects is QT interval prolongation. In-vitro studies showed that vandetanib inhibited currents in cardiac action potentials49.
The potential adsorption of this therapeutic by the µFCs was first examined, as PDMS can rapidly lead to the absorption of hydrophobic compounds and the reduction of their concentration in microfluidic chips (Supplementary Fig. S10). To do so, solutions of vandetanib of known concentrations were incubated into µFCs for 30 min and aspirated prior to injection in HPLC. Concentrations of resulting solutions were determined by comparison of HPLC data to calibration curves generated from pristine solutions with defined concentrations. This data is gathered in Supplementary Fig. S10 and demonstrate that adsorption levels are below 4%.
The impact of vandetanib on spheroid beating was quantified at 1 and 10 µM (in medium containing 0.1% DMSO). We measured the beat rate before (time 0) and after treatment at different time points (Fig. 7 and Supplementary Videos S12–17). In suspension, spheroids beat rates dropped from 29 ± 3 to 23 ± 1 beats/min after 3 min incubation with 1 µM solutions. Normalised rates (0.81 beats/min) were lower than those recorded in empty carrier solutions (0.1% DMSO; 0.93 beats/min, p = 0.025, Supplementary Videos S6–11). Therefore, vandetanib resulted in a significant drop in beat rates. Similar observations were made at 10 µM concentrations, with normalised rates dropping to 0.78 beats/min, 3 min after treatment. However, at this higher concentration, spheroids shut down after 60 min, similarly to what is observed in literature50,51. Surprisingly, this did not occur with spheroids embedded in µFCs (Fig. 7). Although a drop in beating was observed for both concentrations (0.68 and 0.54 beats/min at 1 and 10 µM, respectively, compared to 0.81 and 0.78 beats/min in suspension), beat rates recovered after 60 min. In contrast to spheroids cultured in suspension, the difference in rate drop due to vandetanib was not significant compared to the carrier (0.64, 0.68 and 0.54 for carrier and vandetanib at 1 and 10 µM, respectively). Overall, these results indicate a greater susceptibility of cardiac spheroids to the carrier used for therapeutics exposure (DMSO) when embedded in µFCs. We note however that the kinetics of this initial transient drop in beat rate was comparable in suspension or within µFCs (at least at the first time point of our experiment), validating that fast diffusion occurs even through the microvasculature, to the spheroids tested. In addition, the impact of vandetanib on spheroids is negligible in these conditions, whereas this compound resulted in a transient impact followed by a severe disruption of cardiac beating at the highest concentration (10 µM) in suspension spheroids. Comparison of marker expression (cardiac, endothelial and NG2) following vandetanib treatment did not reveal any disruption of the spheroid or networks formed (Supplementary Fig. S11).
Therefore, the behaviour of implanted cardiac spheroids suggests a protective role of the vasculature on exposure of spheroids to therapeutics such as vandetanib, although the mechanism of this effect remains to be established. The concentrations selected are well within the range of concentrations typically used for this compound to quantify their impact on cardiac function in 2D and 3D cell culture systems (10–100 µM for vandetanib)51,52,53. However, these concentrations are significantly below those typically used for therapeutic effects in a clinical context (100–300 mg/day54,55). Therefore, increasing the complexity of in-vitro culture models, can more closely capture in-vivo response to therapeutics, and possibly fine chemical or nanomaterials, and mimic the impact of such compounds in a more realistic context. This may result in protective effects of such complex environments, as is presently observed, or identify secondary effects such as indirect cytotoxicity underpinned by metabolisation of therapeutics, as in the nephrotoxicity of hepatic metabolites56. Such understanding is essential to identify early on suitable concentration ranges for likely clinical efficacy, but also to more accurately establish whether side-effects are likely within these concentration ranges.