Biodegradable scaffold with built-in vasculature for organ-on-a-chip engineering and direct surgical anastomosis

AngioChip scaffold with branching interconnected lumens

The AngioChip scaffolds were constructed using a biodegradable elastomer, poly(octamethylene maleate (anhydride) citrate) (POMaC)^27,28 (Figure 1). POMaC was selected since it is UV-polymerizable, allowing rapid assembly under mild conditions, biodegrades by hydrolysis²⁷ (Supplementary Figure 1, 2), and is more elastic than FDA approved polyesters²⁹. Citric acid-based elastomers also have low thrombogenicity^30,31. Thin POMaC sheets were pre-patterned, in a scalable manner (Figure 1a), under UV illumination and stamped onto each other, layer-by-layer with precise alignment down to several microns, to form complex suspended microstructures and internal cavities. POMaC exhibited temporary and differential adhesion to glass (strong) and polydimethylsiloxane (PDMS) (weak) after photo-crosslinking, due to oxygen-induced inhibition of free radical polymerization on the surface of the PDMS³², which leaves a non-polymerized POMaC layer at the interface. Patterned POMaC sheets were robustly transferred, aligned and released from one substrate (PDMS) then bound to the POMaC structures supported by a glass substrate (Supplementary Figure 3), circumventing the challenge of printing biomaterials in mid-air³³. 3-D stamping enabled patterning of POMaC into various intricate structures from a 1-D tube (Figure 1e) to 2-D bifurcating conduits (Figure 1f) or a 3-D branching network (Figure 1g) mimicking a vascular bed within a fully interconnected lattice matrix, tailored to support the parenchymal cells (Figure 1c,d,f,gSupplementary Figure 4).

Internal networks branching in the x-y as well as y-z planes were perfusable through a single inlet and outlet (Figure 1b,f,g, Supplementary Figure 4, Movie 1). The smallest micro-channel in the network was 100μm by 50-100μm, with a wall thickness of 25-50μm. To improve the exchange of biomolecules and cell migration, 10μm micro-holes were patterned in the upper channel walls (Figure 1h-j, Supplementary Figure 5a,b). To increase porosity, 20μm micro-holes were patterned in the top and side channel walls (Figure 1k,l, Supplementary Figure 5c). To further enhance oxygen and nutrient exchange, nano-pores were incorporated into the bulk polymer by embedding and subsequently leaching out a porogen, confirmed by mass reduction (Figure 1m) and resulting in wrinkled nano-pores, as described³⁴ (Figure 1n).

For perfusion culture, the AngioChip scaffolds were installed in the main well between the inlet and outlet well of a customized bioreactor (Figure 1c, Supplementary Figure 6). Culture medium or endothelial cell (EC) suspension was perfused through the internal network driven by the liquid pressure head differences (Figure 1c, Supplementary Figure 7). This design removed the need for pumps, allowing access to both the parenchymal space and the internal vasculature using simple tools (e.g. micropipettes), and enabling facile tissue removal (Figure 1b). ECs were cultured within the internal network while the parenchymal cells were cultured within the lattice with native ECMs, allowing tissue remodelling (Figure 1d).

Tunable elasticity and permeability of AngioChips

The hydrolytically degradable scaffold lattice (Supplementary Figure 8a,b) was composed of multiple layers of meshes in the parenchyma connected by vertical posts (50μm diameter) (Figure 1i). This feature provided 100% interconnectivity (Figure 1g), facilitating cell seeding and enabling formation of tissues in both the x-y and y-z plane. The Young's modulus of the adult human myocardium was reported to vary from 10-20kPa (relaxed) to 200-500kPa (contracted)^35-38, while for adult liver it varied from 0.6-2.0 kPa for healthy and up to 20kPa for fibrotic livers³⁹. By varying the UV and heat exposure, monomer ratio and porogen concentration we were able to achieve the Young's Modulus of POMaC bulk polymer ranging from 53 ± 8 to 1423 ± 651kPa (Figure 2d, Supplementary Figure 8c, Supplementary Table 1-4).

The geometry and density of the AngioChip parenchymal lattice were further varied to fine-tune the scaffold mechanical properties to resemble the anisotropic properties of the adult rat ventricular myocardium as in design B (Figure 2a-c, e, Supplementary Figure 9-10, Supplementary Table 5). Anisotropy was enabled by the rectangular shape of the mesh, with higher spatial density for struts oriented in the long-edge direction (LD) than in the short-edge direction (SD) (Figure 2e). Both the effective elasticity (E) and the ultimate tensile strength (UTS) increased with lattice density. Further design iterations can yield scaffolds with mechanical properties tailored for specific applications (e.g. human myocardium or human liver) by tuning the lattice structure and the elasticity of the base material (Figure 2d).

The limited permeability of synthetic polymers hinders the success of other biodegradable microfluidic scaffolds^24,25. Although some biocompatible polymers are permeable to O₂/CO₂⁴⁰, they have a limited permeability for small molecules and proteins⁴¹ necessitating introduction of porosity to enhance permeability (e.g. as in transwell membranes, 9.3×10⁻⁵cm/s for 4kDa molecules⁴²). We found the cell-free internal AngioChip network with 10μm micro-holes was permeable to both small and large molecules (Figure 2f,g). It was >4 times more permeable for large molecules (70kDa TRITC-dextran) than the one without the 10μm micro-holes, >2 times more permeable than EC-covered network and more permeable than mammalian venules and capillaries in vivo (Figure 2f). The permeability of the EC-covered network was higher than the permeability of mammalian venules ((0.15±0.05)×10⁻⁶cm/s) for 70 kDa FITC-Dextran⁴³ and comparable to that of capillaries⁴⁴ to proteins (4.3×10⁻⁶cm/s). High permeability of the cell-free network allowed the EC coating to be the dominating transport resistance (Figure 2f), similar to how the ECs govern the permeability of blood vessels in vivo. The live cell tracker dye, carboxyfluorescein diacetate (CFDA, 557Da), was perfused through the network surrounded by cardiac cells, staining the live cells in the parenchyma (Figure 2h), consistent with convection-diffusion transport and metabolic conversion within the tissue. There was no significant leakage of the culture media during perfusion into the parenchymal space even in the absence of the endothelial cell coating (Supplementary Figure 8d). The AngioChip burst pressure was comparable to that of the rat femoral vein (Figure 2i, Supplementary Movie 2) and ~7 fold higher than the normal systolic blood pressure in a rat (130mmHg) or a human (120mmHg), indicating the network will be sufficient to withstand blood perfusion in the peripheral circulation.

Lumen endothelialisation with on-demand sprouting and cell trans-migration

Upon endothelialization with human umbilical vein ECs (HUVECs), CD31 immuno-staining revealed a confluent endothelium on the luminal network surface (day 2, Figure 3ad; Supplementary Figure 11a-c, day 7 Supplementary Figure 11e-h) with VE-cadherin expressed at the cell-cell junctions (day 2, Supplementary Figure 11d). The ECs physically covered the micro-holes on the vessel wall (Figure 3b,c, Supplementary Figure 11a-c), conformally and confluently coating the lumen even at the branch points (Supplementary Movie 3). In response to an angiogenic stimulus, thymosin β4, ECs migrated through the 20μm micro-holes into the parenchymal space, a first step of angiogenesis (Figure 3e-g). Endothelial network permeability dynamically increased in response to this biological stimulus (Figure 3h), a feature unique to living cells that cannot be reproduced using the polymer material alone. Human whole blood was perfused through the AngioChip network with or without EC coating at 15dynes/cm² (~5μL/min; Re, 0.023) (Figure 3i). The AngioChip network was designed so that the ECs in the first and second order branches experienced the same shear stress. Without an EC coating, more platelets bound to the network surface (Figure 3i-lSupplementary Figure 12) and became activated, as indicated by their extended pseudopodal morphology (Figure 3j). Attached platelets exhibited a trend to spread according to the blood flow pattern and accumulated more at the stagnation regions of branches and turns (Figure 3j, Supplementary Figure 12). Perfused human monocytes, THP-1, exhibited accumulation and adhesion in the network in response to an inflammatory stimulus, TNF-α, (Figure 3m-p), subsequent migration along the endothelialized surface (Figure 3q) and trans-migration through the 10μm micro-holes on the vessel walls, into the parenchymal space (Figure 3r). The AngioChip was versatile enough to also enable migration of Raw264.7 macrophages (Supplementary Figure 13).

Urea secretion and drug metabolism in vascularized liver

Primary rat hepatocytes mixed with 10% primary rat fibroblasts (to facilitate ECM remodelling and gel compaction⁹) were seeded into the parenchymal space of an endothelialized AngioChip scaffold, resulting in the aggregation of viable cells (Figure 4a-e, Supplementary Figure 14a). Hepatocytes (albumin stained) distributed throughout the lattice and around the vessel network, while ECs (CD31 stained) coated the inner lumen of the network (Figure 4a-c). Hepatic tissues were challenged with terfenadine, an antihistamine withdrawn from the market due to cardio-toxicity⁴⁵ (Figure 4f-g). Terfenadine is generally metabolized in the liver, to non-cardio-toxic fexofenadine, by the enzyme cytochrome P450 CYP3A4 isoform⁴⁵. Liquid chromatography–mass spectrometry (LC-MS) revealed the presence of fexofenadine in the outlet well (Figure 4g).

Entirely human liver-AngioChips were engineered using human embryonic stem cell (hESC) derived hepatocytes⁴⁶, human mesenchymal stem cells (hMSCs) as a supporting population and HUVECs for inner lumen coating. High-density culture resulted in the formation of junctions between hepatocytes (Figure 4j), positive staining for albumin (Figure 4k) and bile canaliculi (Supplemental Figure 15), that were similar in appearance to those obtained from a collagen sandwich (Figure 4h,i), a commonly used control. Cell density was 0.6±0.2×10⁸cells/cm³ at day 7 with cells present throughout the AngioChip volume and CD31⁺ ECs coating the inner lumen (Figure 4l,m). Secretion of urea per cell from endothelialized AngioChip tissues was quantified and shown to be higher than that of collagen sandwich control (Figure 4n-o).

Scalable assembly of contractile vascularized cardiac tissue

Cardiac tissues were created from either hESC-derived (Figure 5a-w, Supplementary Figure 14b, 18) or neonatal rat cardiomyocytes (Figure 5x-z). To provide evidence of tissue-level organization, entire AngioChips based on entirely human cells (hESC-derived cardiomyocytes, 10% hMSCs and HUVECs) were imaged using confocal microscopy. Elongated cell bundles were present throughout the entire volume (Figure 5a-f, Supplementary Figure 14b) including the scaffold interior (Figure 5l-m, Supplementary Figure 16). Although gel compaction was possible without the addition of supporting cells, tissues were more compacted and aligned when 10% MSCs were used (Figure 5c,e). Cardiac cell density, estimated from the histology sections, was 2.3±0.8×10⁸cells/cm³. A condensed tissue formed within 5 days (Figure 5i, Supplementary Figure 14b, 16); a time frame consistent with previously published studies of non-vascularized cardiac tissues that relied on gel compaction (5 days for human⁵ and 7 days for rat⁴⁷). Synchronous macroscopic contractions were observed as early as day 4 and the electrical excitability parameters of both rat and human tissues fell within the standard range for non-vascularized rat constructs⁴⁸ and both rat^9,47 and human micro-tissues⁵ (Figure 5j, k). AngioChip cardiac tissues contracted macroscopically and compressed the scaffold at each beat (Supplementary Figure 17b) without collapsing the internal vessel network while being perfused (Supplementary Movie 4,5). The contractile protein sarcomeric α-actinin and the structural protein F-actin were visible in the elongated cells (Figure 5d,f; Supplementary Figure 16). Striated cardiac tissue bundles of the AngioChip were similar to those present in the 3D Biowire construct, a control that previously improved cardiomyocyte structural maturation⁵ (Figure 5g). Cardiomyocytes cultivated in monolayers, a standard control, exhibited cross-striations, but a lack of overall orientation and a less anisotropic overall structure (Figure 5h) compared to those cultured in AngioChips (Figure 5f). At day 7, ECs coated the vessel lumen, while cardiomyocytes distributed throughout the lattice and packed around the vessels (Figure 5l-m). Human cardiac tissue exhibited impulse propagation across the entire tissue (4.8±1.3cm/s, n=7), without conduction block (Figure 5n, Supplementary Movie 6). These values are lower than conduction velocities (voltage) we reported for Biowires previously⁵, as AngioChips were cultivated for a shorter period of time and without electrical stimulation. Conduction velocity can be further improved by electromechanical stimulation^5,49. Within 30-40min of application, the tissues showed the expected positive chronotropic response to epinephrine (10μM, Figure 5o,p, Supplementary Figure 17A).

Engineering millimetre-scale thick tissues was possible using two scalable approaches: 1) by fabricating multilayer scaffolds (1.58mm thick, Figure 5q, Supplementary Figure 20a) via the 3D stamping technique with one inlet and one outlet (Figure 5q-w, Supplementary Figure 20) or 2) by stacking individual single-layer AngioChip tissues followed by the cultivation under perfusion to create a single multi-layer thick tissue with multiple inlets and outlets (Supplementary Figure 19). Fibrin gel or Matrigel hydrogel were independently used demonstrating versatility of hydrogel choices for parenchymal cell seeding. The resulting cardiac tissue was 1.75-2mm thick with a high density of elongated cells throughout the cross-section (Figure 5s-u, Supplementary Figure 20g,h) and visible cross-striations in the cells from the scaffold interior even after only 3 days in culture (Supplementary Figure 20i). Although the tissue was based on human cardiomyocytes, it was comparable to the thickness of a physiological structure, the adult rat heart left ventricle (Figure 5r).

The inner lumen of the AngioChip scaffold in the thick tissues were fully endothelialized, with organization of endothelial tubule-like structures in the parenchymal space to provide additional level of endothelialization (Figure 5v-w, Supplementary Figure 20f,j-l). In some instances these tubular structures clearly connected with the endothelial cells sprouting, through the 20μm micro-holes, from the inner luminal structures of the AngioChip scaffolds (Supplementary Figure 20f,j-l), demonstrating the direct interaction of the cells in the two compartments.

Rat cardiac tissue (Figure 5x) exhibited higher viability under perfusion compared to the non-perfused controls, which developed a necrotic core at day 7 (Figure 5y). The most cell death (i.e. lactate dehydrogenase release) occurred within the first 3 days and medium perfusion helped mitigate cell death (Figure 5z) on day 3, clearly demonstrating the benefit of perfusion even for the single-layer AngioChip tissues.

Direct surgical anastomosis of AngioChips to host vasculature

AngioChip scaffolds were connected to the femoral vessels on the hindlimbs of adult Lewis rats, in artery-to-artery (Figure 6a) and artery-to-vein (Figure 6b) mode, to demonstrate two different configurations of direct surgical anastomosis. The inlet and outlet (inner dimensions of 100μm×200μm and outer dimensions of 300μm×400μm) were connected to femoral vessels (Figure 6) with surgical cuffs. Similar citric acid-based polymers²⁸ have been shown to be antithrombotic in vascular grafts and to support EC growth in vivo^30,31. The animals were only heparinized during surgery. In both configurations, blood perfusion was established immediately, even in the absence of ECs (Figure 6a-b), although artery-to-artery mode was more technically challenging due to the higher pressure (Supplementary Movie 7). Blood pulsation was also observed (Supplementary Movie 7), more noticeably in the artery-to-artery configuration. Erythrocytes were only observed in the networks of tissues implanted with direct anastomosis (Figure 6c,d,g,h,k, Supplementary Figure 21,22). One week after the implantation of the AngioChip cardiac tissues, native angiogenesis also took place around the implants (Supplementary Figure 23). The presence of smooth muscle actin (SMA) positive cells was merely 2% in the isolated neonatal rat heart cells⁴; the significant SMA staining (Figure 6e,i, Supplementary Figure 24) suggested the penetration of mural cells or myofibroblasts into the implanted tissues consistent with the healing response⁵⁰ (Figure 6l). Troponin T immuno-staining demonstrated some elongated cardiomyocytes intertwining within the lattice of the AngioChips (Figure 6f, j, Supplementary Figure 25). AngioChip cardiac tissues endothelialized with Lewis rat primary vein ECs were also implanted via direct anastomosis in the artery-to-vein configuration. One week later, ECs coating the lumen were observed in 50% of the micro-channels from three different implants (2/5, 4/5, 1/5 of visible lumen for implant 1,2,3, respectively) (Figure 6m-p). Histologically, 85% of AngioChip lumen were blood clot-free after 1 week in vivo (Supplementary Figure 21, 22). Specifically, in non-EC coated AngioChips 3/3, 3/3, 4/5 and 3/5 of visible lumen were blood clot-free in the four implants and in the EC coated groups 3/5, 5/5, 5/5 were blood-clot free in the three implants. The channel walls did not degrade appreciably in one week, indicating that longer time is required, as biodegradation and biocompatibility studies showed POMaC polymer discs persisted for at least 5 weeks in vivo (Supplementary Figure 26).

Outlook

AngioChip provides a powerful platform technology for cultivation of vascularized tissues that overcomes key limitations in the fields of tissue engineering (scalable production of mm-sized tissues), organ-on-a-chip engineering (precise placement of endothelial and parenchymal cells, in a simple to operate format) and in vivo tissue implantation (direct anastomosis). This technology liberates the cells from the requirement to be in direct contact with drug-absorbent PDMS in a closed channel system for organ-on-a-chip application or to rely on hydrogels as the sole carrier that is too fragile to surgically connect into the hosts’ circulation for in vivo implantation. Here, we effectively decoupled the material choice for the engineered vessel network from the material choice for the parenchymal-space, allowing us to control the initial architecture of the vasculature and establish immediate perfusion in vitro and in vivo while sustaining the extensive remodelling of parenchyma. This micro-engineering approach provided no delay for tissue endothelialization, as perfusable endothelial network was achieved within one day, prior to parenchymal cell seeding. The new 3-D stamping technique allowed us to handle polymer sheets as thin as 25μm with 10-20μm holes to create a vessel wall that is merely 2-3 cells thick. Thus paracrine signalling between the ECs and the parenchyma, that usually decays significantly within a very short distance (~10 cells)⁵¹, can be sustained.

The thin channel walls in combination with the nano-pores and micro-holes were the key features that allowed effective molecular exchange, cell extravasation in a vascularized 3-D tissue model, and physiologically relevant modes for delivery of test drugs by convection-diffusion. The addition of 10μm micro-holes covering only 0.5% of the total network surface area increased the permeability of the network by more than 4 times compared to the micro-hole free scaffolds. The resulting permeability was higher than that of the native vasculature for EC-free AngioChips, ensuring that the polymer wall did not inhibit transport and enabling subsequent EC-coating to act as a governing resistance. AngioChip networks with 20μm micro-holes in the top and sidewalls increased the coverage of micro-holes to 3.2% of the surface area and guided the endothelial sprouts into the parenchyma. To further fine-tune the vessel permeability to match the unique environment in different organs, organ specific ECs should be used. AngioChip also enables fine-tuning of the elasticity for specific organs, a factor important for preventing long term inflammation and fibrosis ⁵², using a rectangular (for anisotropic⁵³) or square (for isotropic⁵⁴) lattice in the parenchymal-space, difficult to achieve with hydrogels²¹.

The AngioChip platform enables facile integration of different tissues on a single device by linking multiple AngioChips in series (Supplementary Figure 27). Conventional microfluidic systems require bulky external setups that make integration difficult. Closed chip configuration is incompatible with the current practices in biological laboratories and pharmaceutical industry, which rely heavily on open access for liquid dispensing with micro-pipetting. Our platform, resembling a standard multi-well plate, maintains an open configuration so that both the parenchymal space and the internal vasculature can be accessed with pipetting and allows different media to be used in each compartment, thus facilitating co-culture.

Long-term patency of the AngioChip networks should be determined in future in vivo studies. Immobilization of heparin onto the inner luminal surface with existing methods^50,55,56, could also enhance long-term blood vessel patency. Appropriate scaffold degradation rate is critical⁵⁰; therefore, long-term degradation of the AngioChip in vivo should be examined in the future and fine-tuned for a specific application by adjusting the citric acid content on the polymer-chain²⁷ to enable native mural cells and ECM to gradually take over the role of the synthetic polymer vessel wall. Future studies should determine if the original blood vessels will remain, upon complete AngioChip biodegradation, and if the endothelialized vasculature will exhibit appropriate vasodilation and vasoconstriction properties in vivo. Other non-thrombogenic, photo-crosslinkable scaffold materials, with tissue-specific elasticity, could be explored. Electrical stimulation during cardiac culture on AngioChips should be used to further synchronize tissue contractions and mature cells⁵. The proof-of-concept for the use AngioChip in organ-on-a-chip engineering provided here opens the door for future studies aimed at answering complex biological questions in vitro and in vivo.

In summary, the AngioChip was used to generate both in vitro cardiac and hepatic tissue models with defined vasculature and in vivo implants with direct surgical anastomosis. Uniquely, this platform could enable direct and rapid translation of in vitro testing results to in vivo validation and the development of effective regenerative strategies.

METHODS

Supplementary Material