Isabelle
Martinier
a,
Florian
Fage
a,
Alshaba
Kakar
a,
Alessia
Castagnino
b,
Emeline
Saindoy
c,
Joni
Frederick
b,
Ilaria
Onorati
c,
Valérie
Besnard
c,
Abdul I.
Barakat
b,
Nicolas
Dard
c,
Emmanuel
Martinod
c,
Carole
Planes
c,
Léa
Trichet
*a and
Francisco M.
Fernandes
*a
aLaboratoire de Chimie de la Matière Condensée de Paris, Sorbonne Université, 4 place Jussieu, 75005 Paris, France. E-mail: lea.trichet@sorbonne-universite.fr; francisco.fernandes@sorbonne-universite.fr
bLadHyX, CNRS, Ecole Polytechnique, Institut Polytechnique de Paris, Palaiseau, France
cLaboratoire Hypoxie & Poumon, Assistance Publique–Hôpitaux de Paris, Hôpitaux Universitaires Paris Seine-Saint-Denis, Hôpital Avicenne, Chirurgie Thoracique et Vasculaire, Université Paris 13, Sorbonne Paris Cité, UFR Santé, Médecine et Biologie Humaine, Bobigny, France
First published on 6th May 2024
Synthetic tubular grafts currently used in clinical context fail frequently, and the expectations that biomimetic materials could tackle these limitations are high. However, developing tubular materials presenting structural, compositional and functional properties close to those of native tissues remains an unmet challenge. Here we describe a combination of ice templating and topotactic fibrillogenesis of type I collagen, the main component of tissues’ extracellular matrix, yielding highly concentrated yet porous tubular collagen materials with controlled hierarchical architecture at multiple length scales, the hallmark of native tissues’ organization. By modulating the thermal conductivity of the cylindrical molds, we tune the macroscopic porosity defined by ice. Coupling the aforementioned porosity patterns with two different fibrillogenesis routes results in a new family of tubular materials whose textural features and the supramolecular arrangement of type I collagen are achieved. The resulting materials present hierarchical elastic properties and are successfully colonized by human endothelial cells and alveolar epithelial cells on the luminal side, and by human mesenchymal stem cells on the external side. The proposed straightforward protocol is likely to be adapted for larger graft sizes that address ever-growing clinical needs, such as peripheral arterial disease or tracheal and bronchial reconstructions.
In tubular tissues, epithelial cells line the luminal surface, whereas different cell types of mesenchymal phenotype populate the inner wall.2 In addition to the nature and spatial distribution of cells, tubular tissues display a hierarchical arrangement of the extracellular matrix (ECM) components at several length scales, dictating their mechanical properties to a large extent. Type I collagen, the major component of the connective tissues in mammals, is widely responsible for the multiscale mechanical properties of native tissues.3 Locally, it provides the mechanical support necessary for cell adhesion, migration and proliferation. At the macroscopic level, collagen dominates the tubular tissues’ viscoelastic response to the vast range of mechanical loading and ensures their dimensional stability. Moreover, in vitro as in vivo, type I collagen can be degraded and synthesized by different types of cells, providing a dynamic environment that synthetic counterparts cannot provide. The preceding arguments confirm the relevance of type I collagen as the central component to develop scaffold-based grafts4 and to recapitulate tissue-like architectures and fibrillar structures, thought to be key for attaining adequate mechanical properties and cell colonization.5
Here, we introduce a new strategy based on ice templating to tackle the elaboration of materials that are fully composed of type I collagen. Ice templating (equally named freeze casting or ice-segregation-induced self-assembly), a technique initially developed for the elaboration of porous ceramics6 and adapted to produce scaffolds based on biopolymers,7–9 offers the possibility to precisely tune the materials porosity by playing on different parameters such as solution composition, freezing kinetics, and temperature gradient.10 Coupling ice templating and topotactic fibrillogenesis, a technique inducing the formation of native-like collagen fibrils during thawing, enables self-assembly of type I collagen to occur, generating biomimetic constructs while maintaining the macroporous patterns defined by the ice templating technique.11,12 During freezing, the formation of ice crystals induces an increase in concentration of the collagen in solution, due to the insolubility of most solutes in ice.13 The local concentration of collagen attained by this process reaches the same range of those observed in biological tissues. The increase in concentration during freezing coupled to the lyotropic behavior of collagen in solution give rise to highly organized domains, which, upon fibrillogenesis, recapitulate the multiscale hierarchical architecture of collagen in native tissues.
In this work, we propose to use this technique to provide tubular materials applicable for tissue engineering purposes. In particular, we explored the effects of processing parameters such as the thermal conductivity of the cylindrical molds and fibrillogenesis routes on the textural and mechanical properties of the biomimetic tubular materials. In addition to the structural characterization we have performed in vitro colonization studies with a variety of cell lines to understand the effect of the materials’ processing routes on the colonization by human umbilical vein endothelial cells (HUVECs), adenocarcinomic human alveolar basal epithelial cells (A549), as well as human mesenchymal stem cells (hMSC).
The materials reported here combine, for the first time, the hierarchical organization found in native ECM—from the molecular scale up to the tissue level—using type I collagen, the mechanical performance in the range of native tissues as well as the capacity to be colonized by endothelial, epithelial and mesenchymal cell types. These findings open an exciting pathway towards the development of new biomimetic tubular tissue grafts that promise to find applications in pathologies such as peripheral arterial disease (PAD) or tracheal and bronchial reconstructions, for which the currently available materials fail.
Fig. 1 The internal texture of type I collagen materials obtained by ice templating is determined by the thermal conductivity difference between the inner and outer parts of the freezing molds. (A) Scheme depicting the ice nucleation and growth events in collagen-based tubular materials leading to the formation of an interface inside the walls of the tubes. SEM analysis of transverse sections of a selected material after freeze-drying showing the concentric position of the interface, highlighted with a red dotted line. (B) SEM images of longitudinal sections of freeze-dried collagen materials. The red dotted lines illustrate the position of the interface between the luminal (left) and external (right) surfaces. All scale bars are represent 500 μm. (C) Influence of the difference in molds’ thermal conductivities on the position of the interface formed in between the two ice growth regions. Numbers 1 to 6 refer to the SEM longitudinal sections shown in B for specific pairs of materials (displayed in ESI Table 1†). The interface position is defined as the percentage of the distance to the lumen with respect to the total wall thickness of the matrix. |
For the freezing conditions chosen here (dipping speed and mold geometry), and for the materials mentioned above, the position of the interface, xi, scales linearly with Δλ and can be adjusted between 5.6% and 62.5% of the tube wall as observed in Fig. 1B and C. The relevance of this interface is multiple since it is expected to play a role in the materials’ mechanical performance, the permeability towards cells and fluids, as well as in the regulation of the cellular migration path.
Samples were ice templated under equivalent boundary conditions (dipping speed, air and bath temperature) using both mold pairs (Cond and Ins). After lyophilization, longitudinal and transversal sections of the scaffolds were imaged by SEM to quantify the orientation and dimensions of the pores, as shown in Fig. 2.
Pore orientation analysis on longitudinal sections (Fig. 2A & B.a.1), showed that the pores are perpendicular (ca. 90°) to the longitudinal axis of the tube for the Cond freezing conditions (Fig. 2A.a.2). Conversely, Ins conditions favor the formation of axial pores, oriented at 12° with respect to the longitudinal axis of the tube (Fig. 2B.a.2). These results confirm the radial thermal gradient imposed by the outer conductive material across the collagen solution. At the initial moment of freezing, ice nucleates in contact with the outer mold and grows towards the lumen, as confirmed by the SEM observation of the transversal cross-section of Cond (Fig. 2A.b.1). Ins freezing conditions favor the axial thermal gradient imposed by the dipping process, yielding a different pore organization, as shown on both SEM longitudinal and transversal cross-sections (Fig. 2B.b.1).
Pore orientation was analyzed on a quarter of each tubular material cross-section (highlighted in gray on Fig. 2A & B.b.1) using the OrientationJ plugin in FIJI16 (Fig. 2A & B.b.2). In Cond materials, the continuous change of orientation of the pores, observed by the various colors, corresponds to a radial organization. On the contrary, Ins materials feature pores that are organized in coherent domains that follow one major orientation along the longitudinal axis of the tube. Higher magnification images on other specific parts of the tubular walls confirm this observation (Fig. 2A & B.b.3). The striking difference in the internal porosity of these materials is determined solely by the thermal gradient imposed by the mold walls.
However, the size of the pores observed transversally and at the luminal surface appears to be independent from the molds’ thermal conductivity difference. In fact, pores exhibit comparable dimensions with a median Feret diameter of 25.75 μm and 9.77 μm in Cond scaffolds, and 20.62 μm and 11.36 μm in Ins scaffolds, for transversal sections and luminal surfaces respectively (Table 1). In contrast, pore formation and dimensions at the external surfaces significantly relate to the thermal conductivity of the outer mold. The median Feret diameter and percentage of surface accessible porosity are both lower in Cond, with a Feret value almost 3 times higher for Ins (45.83 μm vs. 17.32 μm), and a percentage of surface accessible porosity 1.5 times higher for Ins scaffolds than for Cond scaffolds (48% vs. 63%) (Fig. 2A & B.d,12 and Table 1). These differences are explained by the ice nucleation conditions imposed by the various pairs of mold materials. Since freezing occurs later under Ins conditions, a wide supercooling zone is formed before freezing. This zone is however minimal in conductive conditions owing to an earlier freezing process, as found in Cond.
Freezing conditions | |||
---|---|---|---|
Analyzed zone | Measurement | Cond | Ins |
Transversal section | Median Feret/μm | 25.75 ± 8.43 | 20.62 ± 2.74 |
Median MinFeret/μm | 8.71 ± 0.77 | 9.14 ± 0.80 | |
Surface accessible porosity/% | 59 | 79 | |
Luminal surface | Median Feret/μm | 9.77 ± 0.26 | 11.36 ± 1.09 |
Median MinFeret/μm | 5.26 ± 0.12 | 4.13 ± 0.37 | |
Surface accessible porosity/% | N/A | 53.57 | |
External surface | Median Feret/μm | 17.32 ± 2.15 | 45.83 ± 5.97 |
Median MinFeret/μm | 8.69 ± 0.86 | 18.09 ± 1.39 | |
Surface accessible porosity/% | 48 | 63 |
The lyophilization of ice templated collagen materials was performed exclusively to observe these textural features, in dry conditions under the SEM. The hydrated fibrillar hydrogels were prepared by a topotactic fibrillogenesis process following ice templating, to induce the self-assembly of collagen molecules into native-like fibrils. The challenge associated with this transformation was to maintain the control over the porosity generated by ice templating, since the self-assembly requires temperatures that initiate the melting of ice.
Confocal images of the scaffolds confirm the ability to induce the topotactic fibrillogenesis of collagen scaffolds via the ammonia gas-phase route (noted Cond-NH3 or Ins-NH3). Pores formed during freezing remain open and interconnected, and the orientation imposed during the ice-templating process is maintained (Fig. 4A.1.a & 2.a). Conversely, the liquid-phase fibrillogenesis pathway induces either the collapse of the pores in the case of Cond-PBS samples, (Fig. 4B.1.a) or a strong reduction of their size for Ins-PBS samples (Fig. 4B.2.a).
Cond-NH 3 and Ins-NH3 feature similar pore size in transversal sections (median Feret diameter (Feret) × median minimum Feret diameter (MinFeret)) of (16.12 ± 1.17 μm × 6.31 ± 0.39 μm) and (18.76 ± 1.57 μm × 8.05 ± 0.64 μm), respectively (Fig. 4b.1-2 for A.1 and A.2, and Table 2). Compared to lyophilized samples, the pore size decreases during fibrillogenesis with a reduction of the Feret diameter of 37% and 9% for Cond and Ins conditions, respectively. As ammonia reaches the surface of ice crystals, a competitive process takes place. Thawing of ice crystals induced by the cryoscopic depression at the ammonia-ice interface competes with the self-assembly of collagen molecules. The swelling of the collagen walls is a direct consequence of ice thawing being favored over the fibrillogenesis. In such a case, a partial dilution of the collagen walls in the liquid water generated by the ice melting occurs, before the pH change induces the formation of the fibrillar collagen gel. This leads to a smaller pore size in the hydrogel than in the dried scaffold. We hypothesize that the lower pore diameter reduction for Ins-NH3 in comparison to Cond-NH3, is due to its unidirectional pore structure. It may favor a faster diffusion of ammonia molecules that promote collagen fibrillogenesis over the melting of ice. On the other hand, the isotropic texture of Cond samples may induce a slower diffusion of ammonia, which leads to faster ice melting and subsequent increased dilution of collagen walls. This observation is confirmed by the proportion of pores giving access to the inner wall in Fig. 4A.1.c.1 and A.2.c.1, where the respective pore coverage for Ins-NH3 and Cond-NH3 decreases from 35% to 24%.
Freezing and fibrillogenesis conditions | |||||
---|---|---|---|---|---|
Analyzed zone | Measurement | NH3 | PBS (10×) | ||
Cond | Ins | Cond | Ins | ||
Transversal section | Median Feret/μm | 16.12 ± 1.17 | 18.76 ± 1.57 | N/A | 3.78 ± 0.27 |
Median MinFeret/μm | 6.31 ± 0.39 | 8.05 ± 0.64 | N/A | 1.78 ± 0.14 | |
Median aspect ratio | 3.04 ± 0.10 | 2.33 ± 0.08 | N/A | 1.84 ± 0.05 | |
Surface accessible porosity/% | 24 | 34 | N/A | 31 | |
Luminal surface | Median Feret/μm | 10.41 ± 2.49 | 16.00 ± 2.25 | 5.18 ± 0.26 | 5.98 ± 0.60 |
Median MinFeret/μm | 5.56 ± 1.51 | 7.82 ± 1.18 | 3.16 ± 0.16 | 3.32 ± 0.37 | |
Median aspect ratio | 2.03 ± 0.07 | 1.97 ± 0.97 | 1.91 ± 0.06 | 1.47 ± 0.04 | |
Surface accessible porosity/% | 53 | 21 | <1 | 5 | |
External surface | Median Feret/μm | 5.04 ± 0.27 | 19.35 ± 3.28 | 4.78 ± 0.54 | 1.43 ± 0.12 |
Median MinFeret/μm | 2.68 ± 0.15 | 10.22 ± 1.54 | 3.43 ± 0.35 | 0.79 ± 0.07 | |
Median aspect ratio | 1.99 ± 0.02 | 1.78 ± 0.10 | 1.33 ± 0.09 | 1.46 ± 0.02 | |
Surface accessible porosity/% | 50 | 28 | <1 | 2 |
In contrast, PBS (10×)-fibrillated scaffolds demonstrate reduced pore accessibility in the transversal plane (Fig. 4B.1.b.1 and B.2.b.1). Surfaces have a smooth appearance, and regardless of the ∼20 μm periodic artifacts due to the sample cutting process, the method fails to retain the porous structure (Fig. 4B.1.a and B.2.a). Cond-PBS samples are devoid of pores at the observable length scale. In Ins-PBS, observable pores are scarce close to the lumen, even at higher magnification (Fig. 4B.1.b.2). In the center, small pores can be witnessed in between the formed fibrils (Fig. 4B.2.b.1), with a median Feret diameter of 3.78 ± 0.27 μm and an accessible pore surface of 30.5% (Fig. 4B.2.b.2). Overall, the conditions imposed by the fibrillogenesis in PBS (10×) favor the melting of ice crystals prior to the self-assembly of collagen. The architecture of ice crystals plays an important role in the preservation of their shape and the subsequent porosity. Unidirectional pore structures such as those found in Ins promote faster diffusion of the PBS ions, slightly shifting the kinetics in favor of collagen self-assembly which favors the preservation of the porous structure.
The pore size on the luminal surface is not altered by the gas-phase fibrillogenesis process (Fig. 4c.1-2, d.1-2 for A.1 and A.2). Pores in the lumen present Feret diameters of 10.41 μm and 16.0 μm for respectively Cond-NH3 and Ins-NH3 (Table 2), versus 9.77 μm and 11.36 μm in lyophilized scaffolds (Table 1). Pores of the external surface are 2 to 4 times smaller than those of the lyophilized scaffolds, with a mean Feret diameter of 5.04 μm and 19.35 μm for respectively Cond and Ins, compared to 17.32 μm and 45.83 μm in the dry state. However, the degree of surface accessible pores differs from what was observed for lyophilized scaffolds: Cond provides the highest pore coverage with 53% at the lumen and 50% at the external surface, compared to 21% and 28% for Ins. We can hypothesize that the mismatch between the conservation of the pore size and the reduction of surface accessible is due to a contraction of all pores following fibrillogenesis. The smallest pores observed after freeze-drying may disappear or become too small to be detected, while the largest pores may be reduced due to a slight collagen wall swelling between the dry and the hydrated forms. In contrast, fibrillogenesis in PBS (10×) leads to a significant reduction of the pore size at both the luminal and external surfaces (Fig. 4c.1-2, d.1-2 for B.1 and B.2). At the lumen, pore size is reduced by a factor of ∼2 for both Cond- and Ins-PBS., while on the external surface, size is reduced by factors of 4× and 32×, respectively. More importantly, the degree of accessible pores exceptionally decreases in Cond-PBS. Less than 1% of the luminal and external surfaces is covered by pores, which hinders access of cells to the inner wall of the material upon cellularization. In Ins-PBS, 5% and 2% of the luminal and external surfaces are covered by pores, corresponding to a 90% size reduction in comparison to lyophilized scaffolds. PBS (10×) route results in the formation of more rounded pores, characterized by a mean aspect ratio ranging between 1.58 and 1.8, in contrast to NH3 fibrillogenesis that yields values ranging from 2.05 to 2.30. As in any temperature-dependent process, the conditions of topotactic fibrillogenesis determine the final textural characteristics of the scaffolds. The gas-phase route favors the self-assembly of the collagen molecules that stiffen the walls surrounding the ice crystals and subsequently melt the ice crystals to reveal the pores. The resulting materials correspond thus to a fibrillar gel sample whose texture is the fingerprint of the ice crystals formed during ice templating. Conversely, the liquid-phase fibrillogenesis favors first the thawing of ice-crystals, resulting in a dilution of the collagen molecules initially concentrated in the interstices and a partial or complete loss of the texture defined by ice. Shifting the kinetics in favor of self-assembly can enhance the preservation of the porous network, as seen above for anisotropic structures obtained in Ins conditions. Such structures lead to faster diffusion rate of the fibrillation media, associated with increased likelihood of keeping the pores and their interconnectivity. However, the evolution of some textural features—such as the reduction of pore size—induced by liquid route topotactic fibrillogenesis presents advantages. Specifically, the pore dimensions on the luminal surface of tubular constructs obtained through PBS (10×) topotactic fibrillogenesis closely resemble those observed in the tracheal basal membrane, which display an average diameter of 1.76 μm.18 In this sense, the liquid-phase fibrillogenesis route described here can be seen as a supplementary tool for fine-tuning the porosity of the materials and matching the textural characteristics of a given native tissue. Noteworthy, for both fibrillogenesis routes, the shape and structure of the tubular materials remained stable for at least two years.
Ordered structures revealed by the birefringence signal emerge from two factors that are difficult to disentangle. On the one hand, molecular crowding can lead to the formation of mesophases, and in some case, to the appearance of pre-fibrillar structures able to produce birefringence signal.20 On the other hand, since samples are exposed to physicochemical conditions meant to induce fibrillogenesis—ammonia vapors and PBS (10×) solution—even collagen solutions displaying an isotropic molecular arrangement may lead to fibrillar structures able to produce birefringent signal. At the observed scale here, polarized optical microscopy provides clues to understand the global orientation of the collagen molecules within the tubular samples, but is insufficient to describe the formation of biomimetic fibrillar domains.
Ultra-thin longitudinal sections of the scaffolds observed under TEM demonstrate the existence of fibrillar structures for each ice templating/topotactic fibrillogenesis condition, similar to the motifs found in the ECM of arterial tissue. Observed at higher magnification, fibrils present a D-banding pattern along their axis, with distances below the typical 67 nm striations due to their off-plane orientation with respect to the cross-section plane (Fig. S3†). The existence of these features confirms the fact that both fibrillogenesis routes induce the formation of biomimetic fibrils, and follow an arrangement similar to that of native tissues. Exposed to ammonia vapors, collagen molecules form a dense network of small fibrils (Fig. 5-A.1.b and A.2.b), both transversal and parallel to the observation plane. Fibrils in Cond-NH3 are smaller and lie in a looser arrangement than those in Ins-NH3, revealing that a simple change of the thermal conductivity of the molds during ice-templating can affect collagen fibril formation. Fibrils in Ins-NH3 form compact arched patterns, surrounded by denser regions. Conversely, PBS (10×) promotes the formation of a higher number of large striated fibrils, in less tightly packed domains. These results confirm the observations of the texture observed under the confocal microscope, for which the swelling during fibrillogenesis led to narrower—or absent—pores. We hypothesize that during the partial melting of the ice crystals, concentrated collagen solution in the interstitial space is rediluted, which increases the mobility of the molecules and enables the formation of a greater number of supramolecular structures. The high compaction of the fibril network in Cond-NH3, Ins-NH3 and Cond-PBS, is similar to those of native tissues, suggesting a local concentration ranging between 50 and 200 mg mL−1.21 The diversity of fibril sizes and arrangement observed here provide a valuable tool to elaborate materials that mimick specific features of a variety of tissues’ ECM. To the best of our knowledge, no other tubular biomaterial has reached such a close resemblance to native tubular tissues in terms of collagen supramolecular and network organization.
Although the preceding results confirm the relevance of our elaboration strategy in obtaining biomimetic materials, they do not provide direct proof of the stabilization effect brought by the fibrillogenesis processes in cold conditions. We have measured the thermal stability of the different materials using Differential Scanning Calorimetry (DSC). The temperature of the endothermic peaks characterizes the denaturation processes of each sample (Fig. S4†). Samples fabricated in -PBS are more stable, with a mean denaturation temperature (Td) 2.6 °C higher than with -NH3 (Table 3). Such a difference can be explained by the prevalence of larger and more mature fibrils in these samples, leading to higher stability. The structural variations induced by the conductive or insulating mold do not significantly alter the Td. In all conditions, the denaturation temperature is occurs at higher temperature than for non-fibrillar materials (ca. 37 °C).
NH3 fibrillogenesis | PBS (10×) fibrillogenesis | |||
---|---|---|---|---|
Cond | Ins | Cond | Ins | |
T d ± SD/°C | 48.73 ± 3.41 | 46.76 ± 0.52 | 49.55 ± 1.09 | 51.11 ± 0.99 |
Two major differences standout regarding the influence of the fibrillogenesis pathways on the collagen texture and microstructure. Ammonia treatment enables the maintenance of the porosity and of the collagen walls’ concentration. It leads to the formation of small fibrils, tightly packed in a complex 3D arrangement according to the TEM and PLOM images. In contrast, PBS (10×) favors the thawing of ice that results in less concentrated collagen domains. These display larger and more mature fibrils than in NH3, which increase the thermal stability. Regardless of the mold materials conditions, it is evident that a modification of the physico-chemical parameters induces variations in the size and arrangement of fibrils. Unlike other existing tubular biomaterials, we offer here conditions that facilitate the stabilization of the liquid crystal-like mesophases, a key feature of the ECM of native tissues. Given the similarities of our materials to the native cellular microenvironment, we expect them to efficiently promote in vivo cellularization after implantation, a necessary process for the long-term biointegration.
The long-term stability of the tubular collagen samples was assessed one year after storage at 4 °C in 1× sterile PBS solution and compared it with a freshly prepared sample. Measured by DSC, both samples display a denaturation temperature around 46 °C, confirming the absence of denaturation of the fibrillar structure during storage (Fig. S5A†). In addition, visual inspection of the samples confirmed their macroscopic integrity (Fig. S5B†).
The resistance to macroscopic deformation in both the longitudinal and circumferential directions under quasi-static loading was evaluated for the four different types of obtained tubular materials (Fig. 6A and B). In the longitudinal direction, Ins-NH3 samples exhibited the highest stiffness, with a Young's modulus (E) of approximately 58 ± 39 kPa (mean ± SD). Conversely, the circumferential stiffness of Ins-NH3 samples, as well as the longitudinal and circumferential stiffness of Cond-NH3 samples, were notably lower, ranging between 6 and 10 kPa. These findings indicate that the matrices display anisotropic elastic properties under macroscopic loading, with greater rigidity observed along the longitudinal direction of Ins-NH3 samples compared to others. Under Ins conditions, the process of ice templating involves the uniaxial growth of ice crystals, which alignment is preserved through ammonia fibrillogenesis. The longitudinal orientation of pores explains the greatest Young's modulus obtained for Ins-NH3, and is coherent with superior elastic properties in the direction of crystal growth but poorer elastic properties in the transverse directions. In contrast, the thermal gradient imposed in Cond leads to radially oriented collagen walls and pores. As a consequence, the circumferential and longitudinal directions are perpendicular to the porous structure and show a relatively low Young's modulus. The analogy to the elastic properties of composite materials, here with a collagen matrix populated by zero Young's moduli inclusions—the pores—allows us to rationalize this behavior. One may invoke a Rule of Mixtures (ROM) behavior in case the pores and the walls are aligned with the traction direction, whereas an Inverse Rule of Mixtures (IROM)22 may explain the lower moduli when the traction is applied perpendicular to the main pore and wall directions.
Contrary to the results obtained using NH3 as a fibrillogenesis medium, PBS (10×) (Fig. 6B) provides the tubular materials with isotropic elastic properties. Regardless of the orientation of the pores initially developed through ice templating, elastic moduli remain close to 20 kPa. This behavior is ascribed to the fact that PBS (10×) favors faster ice melting over the self-assembly of collagen, resulting in poorly preserved porosity, small pore size and low anisotropy compared to NH3-fibrillated samples (Fig. 4). Instead of having a composite material with highly distinct mechanical properties in the longitudinal and circumferential directions, as seen in NH3-fibrillated samples, we obtain a material with similar mechanical properties in both deformation directions. This difference in elastic properties depending on the direction of stress between the two types of fibrillogenesis could allow mimicking a large range of tissues with orthotropic properties (such as bone, cartilage, tendon…) as well as tissues with isotropic properties (adipose tissue, connective tissue, spongy bone…).
The elastic properties discussed above describe the behavior of the materials at the tissue level. However, their local elastic properties—at the cell length scale—are equally important to ascertain their relevance as tissue grafts. The local Young's modulus was measured in compression on the luminal and transverse surfaces of -PBS and -NH3 samples using an Atomic Force Microscope (AFM). Unlike the tensile measurements that provided the macroscopic stiffness of the samples, these local stiffness measurements focused on the stiffness of the walls within the samples. Results depicted in Fig. 6 demonstrate that the -NH3 samples exhibited a significantly higher average stiffness of approximately 300 kPa compared to the -PBS samples, which had an average stiffness of around 50 kPa. Furthermore, the ammonia-fibrillated samples exhibited a broader distribution, characterized by increased variability in the 25th and 75th percentile values when compared to the -PBS fibrillated samples. Additionally, the violin plots suggest the co-existence of two populations of elastic local environments for the -NH3 samples, while the -PBS sample displays a Gaussian distribution indicative of a single population. Taken together, these results confirm the observations arising from the microscopy techniques (TEM and confocal microscopy) that suggested a partial dilution of the collagen walls during (10×) PBS-induced topotactic fibrillogenesis, whereas NH3 tends to preserve highly concentrated collagen domains in the internal walls of the scaffolds. The ability to tune the collagen concentration in our scaffolds through different fibrillogenesis protocols presents an interesting aspect. Cells are significantly impacted by the collagen concentration in their microenvironment,23 directly influencing the stiffness and mechanical properties of the surrounding extracellular matrix, as demonstrated by our AFM experiments. This crucially affects cellular adhesion, migration, differentiation, and tissue development. Consequently, tailoring local collagen concentration with our different fibrillogenesis protocols could result in adapted substrates for grafting at different locations with diverse surrounding cell types.
The trends of epithelial cellularization observed for these materials using A549 cells were equally confirmed by seeding HUVECs. As depicted in Fig. 7B, Cond surfaces generally provide a more favorable substrate for HUVECs adhesion than Ins surfaces with the cells exhibiting a more homogeneous distribution and more complete surface coverage regardless of the cell culture time and the fibrillogenesis pathway (with the exception of Cond-PBS at day 3). Cond-NH3 cell density follows that of the biomimetic control at days 3, 5 and 7. At day 7 cell densities are steady compared to the previous time point, suggesting that confluence has been attained.
Materials prepared by PBS (10×) fibrillogenesis display slower cellularization kinetics as compared to materials fibrillated in ammonia. Despite the initial differences, the cell density in PBS (10×) fibrillogenesis materials does evolve positively over time as can be observed from Fig. 7 and 8. These results demonstrate that it is possible to direct the cell–material interactions by an adequate choice of the mold thermal conductivity parameters and the fibrillogenesis route. The highest density was obtained for Cond, which coincides with the low porosity of the lumen previously observed under confocal microscopy. Combined with NH3 fibrillogenesis, Cond-NH3 surfaces present the highest cell densities at early cell culture time points, but as mentioned earlier cell density appears to decrease once confluence is attained.
One of the key works describing the elaboration of collagen-based materials for tubular constructs is that of Weinberg and Bell24 who proposed a gel-based approach to generate vascular models by casting a suspension of cells in a collagen solution inside a tubular mold. The resulting gel was compacted through a cell-mediated process to reach materials that demonstrated the ability to restore the function and, to some extent, the structure of vascular tissues.25 However, the materials they reported required mechanical reinforcement by a Dacron mesh before application due to insufficient mechanical resistance. The tubular constructs reported here reach the elastic properties of native arteries—with an elastic modulus up to 60 kPa, which is comparable to measurements performed on native canine arteries as reported by Holzapfel26—without using synthetic polymers. Use of biomimetic tubular materials that completely rely on biopolymers present in the ECM reduces the risk of foreign body reaction. Moreover, since the materials developed here are pre-concentrated during the freezing step, it is expected that the dimensional stability is increased with respect to the results of Hirai et al. where wall thinning was observed.25 Collagen-based materials obtained by ice templating coupled to topotactic fibrillogenesis have already demonstrated the ability to withstand long term cell culture experiments (up to 2 months) without any sensible dimensional variation due to the local increase in concentration.12 Li et al. have equally reported the elaboration of collagen-based materials with the goal of producing implantable tubular materials strategy based on drying and crosslinking collagen to modify the elastic properties of the tubular constructs.27 In a different approach, dense collagen sheets obtained after plastic compression28 as well as collagen-elastin composite matrices29 were rolled around a cylindrical mandrel to obtain tubular materials, the obtained materials demonstrated mechanical properties close to those of native tissues, with encouraging results obtained in vivo.29 Despite their promising results, these techniques fail to produce materials that display both the fibrillar organization of native tissues and the macroporosity required for successful cell colonization. The cellularization experiments reported in the present manuscript allow for the double cellularization of the biomimetic materials, with the adhesion of epithelial (A549) or endothelial cells (HUVEC) on the luminal side of the constructs. Due to the porosity generated by our process, hMSC were able to penetrate the tubes’ walls from the exterior (adventitial) side, opening an exciting path to the full cellularization of the materials, favoring homeostasis due to enhanced diffusion of nutrients and waste.12
Other self-standing porous collagen materials obtained by lyophilization of collagen solutions followed by physical30 or chemical31 cross-linking have been reported. Koens et al. elaborated tubular porous materials displaying three layers,32 but one month after implantation thrombus formation was observed and associated to the porosity of the luminal wall. In comparison, the materials reported here feature a topological control over the porosity. Indeed, a non-porous luminal surface is required to foster the formation of an endothelium on the internal wall, while an oriented porous structure in the external part of the material ensures effective cell colonization from the adjacent tissues.
Compared to other methods reported in the literature, our approach enables to obtain materials that display a controlled and hierarchical structure at different length scales. Because the process of ice templating has been already applied in industrial settings, it is reasonable to expect that it is likely to be adapted for larger scale graft production.
Here we designed a fabrication strategy that relies on type I collagen in solution that is further processed to obtain biomimetic materials. If the processes to elaborate collagen-based materials starting from collagen in solution are widespread, other options are available. L'Heureux and colleagues have developed materials based on TESA (tissue engineering by self-assembly) where cell sheets develop into dense biomimetic constructs applicable in vascular reconstruction.33,34 Their strategy yields materials with a compositional complexity that is determined by the cells’ production of ECM components, and thus goes beyond the initial composition of the formulation. The same strategy has been applied to the elaboration of biomaterial threads that could be woven into tubular constructs equally applicable as vascular grafts.35 Decellularized matrices push the complexity of the graft composition even further since they rely on the removal of the cells from native tissues but preserve the composition (to a given extent) and the architecture of native tissues. The matrices developed in the present work do not reach the level of compositional complexity found in materials obtained from cell sheet or decellularization technologies. However, because they are solely composed of major structural protein of ECM, they enable to isolate the factors triggering cells’ response by removing what could be called “biological noise” arising from cytokines, extracellular vesicles or other biologicals found in more complex compositions. In this sense, our materials could serve as instrumental models to characterize the role of the ECM architecture in tubular tissue reconstruction, and pave the way to obtain off-the-shelf tubular grafts presenting appropriate dimensions, structural and mechanical properties, while requiring only transient cellular preconditioning before implantation, if any.
Ice-templating has proven versatile as a shaping technique to fine-tune the materials porosity by adjusting the thermal boundary conditions imposed by molds. This simple method allows orienting pores in either longitudinal or radial directions, a tool to tailor the accessibility of cells to the scaffold. The topotactic fibrillogenesis conditions provide further control over the porosity while triggering the self-assembly of collagen molecules into biomimetic fibrils. Our materials offer cellular micro-environments closely resembling those of biological tissues, a lacking feature in materials currently used in the clinic. In particular, one material stands-out as a potential graft, Cond-NH3, which meets the requirements initially set regarding the structure of tubular tissues. It provides a luminal surface that promotes a rapid endothelium formation, as well as a porous outer layer that enables cell colonization.
Using only type I collagen and in the absence of any cytotoxic cross-linkers, the elastic properties at both the tissue and the cellular length scales, reach those of native tissues. This exceptional combination of native composition and mechanical properties, is given by our precise control over the porosity and orientation. It directs the anisotropic versus isotropic character of the materials’ elastic properties, offering the opportunity to mimic mechanical behaviors of a wide range of biological tissues.
Interactions between the materials and different cell lines – ECs and MSCs – demonstrate their potential to serve as biologically relevant substrates. On the basis of hMSCs migration experiments, ammonia fibrillogenesis proves more suitable to preserve pores and to facilitate cell colonization, ensuring a successful graft implantation. Epithelial and endothelial cells adhere well to the materials and proliferate until confluence is reached in all conditions, demonstrating their potential to promote endothelialization in vivo.
Overall, we introduce a new family of tubular collagen materials, exhibiting a unique combination of high concentration and porosity, with unprecedented resemblance to native tissues. Through their textural and compositional features, they promote both cell adhesion on the luminal surface and cell migration from the external surface, associated with mechanical performances suitable for functional replacement. We anticipate that these are central to a new range of grafts, whose biomimetic features secure assimilation by the body of the patient.
Beyond their role as tissue grafts, our materials are expected to act as organotypic models of tubular tissues. These may answer a vast range of open questions—from cell–cell and cell–matrix interactions up to drug treatment response—bridging current gaps for improved patient care in respiratory, vascular, gastrointestinal and urinary tracts pathologies.
Concentrated collagen is introduced in between two cylindrical molds (inner mold: 3 × 5 mm; outer mold: 8 × 10 mm: length: 50 mm) whose ends are hermetically sealed by caps. For removal of air bubbles, sample is centrifuged at 1 rpm and 10 °C for 10 minutes. The sample is placed onto a home-made set-up that allows a continuous and regulated dipping of the sample in liquid nitrogen. It is composed of: 2 optical axes, 4 linear rod rail support guides, a lead screw, a DC motor (12 V), a speed controller, a transformer and an ABS 3D-printed part linking the support guide and the lead screw (Fig. S1†).
The sample is ice-templated at a dipping speed of 10 mm min−1. Once frozen, the outer mold is quickly removed and a topotactic fibrillogenesis condition is applied. At low temperature, two conditions are compared: (a) exposure to ammonia vapors or (b) PBS bath (10×: 1.37 M, KCl 26.8 mM, Na2HPO4 80.7 mM, and NaH2PO4 14.7 mM). Pre-fibrillogenesis for (a) consists of exposing the sample to ammonia vapors at 0 °C in an ice-water bath for 48 h, followed by removal of ammonia using distilled water vapors for 24 h at 37 °C in a heat room. For (b), the sample is simply maintained in PBS (10×) at −3 °C in a thermostatic bath for 72 h. For both (a) and (b), fibrillogenesis is completed by keeping the sample in PBS (5×: NaCl 685 mM, KCl 13.4 mM, Na2HPO4 40.35 mM, and NaH2PO4 7.35 mM) for 14 days at room temperature.
Pore orientation was quantified using the ImageJ plugin OrientationJ. Pore surface coverage was determined as the percentage of accessible pores on a given surface.
Samples were fixed at days 3 and 11 after seeding with 4% formaldehyde in PBS (1×) for 16 hours at 4 °C, then rinsed in PBS. After permeabilization with 0.1% Triton-X100 in PBS (1×) for 10 minutes and BSA-blockage in 3% BSA/0.1% Tween-20 in PBS, samples were stained with 1 μg mL−1 TRITC-conjugated phalloidin (Sigma) for 15 min at room temperature. They were rinsed several times in PBS (1×), stained with DAPI and placed in aqueous mounting medium (Citifluor AF1) for confocal microscope analysis.
After 3 weeks of culture, samples were fixed with 4% formaldehyde in PBS (1×) for 16 hours at 4 °C, then rinsed in PBS. Samples were successively transferred in 15%, 30% sucrose in PBS (1×) and then in 50% OCT/30% sucrose mixture at 4 °C until equilibration. Finally, samples were progressively embedded in 100% OCT on dry ice and conserved at −80 °C until cryosections were performed. Ultrathin cryo-sections (10 μm-thick) were cut and post-fixed in ice-cold acetone for 15 minutes. Samples were then stained with haematoxylin and eosin dies for 5 minutes and placed in aqueous mounting medium for microscope analysis.
This work was supported by Agence Nationale de la Recherche (ANR) under grant agreement ANR-20-CE19-0029. IM acknowledges the PhD fellowship from the Physics and Chemistry of Materials’ graduate school (ED397).
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3bm01808c |
This journal is © The Royal Society of Chemistry 2024 |