Unravelling Form and Function: Improved function of engineered cardiac tissue through extra-cellular anisotropy

Cardiac tissue engineering is a promising therapeutic option for myocardial repair after injury, however, so far engineered heart patches have shown limited translational utility due to poor electrical integration and tissue contractility. Emerging research suggests that scaffolds that recapitulate the three-dimensional structure of the native myocardium improve physiological function. Complex scaffold fabrication remains a technical challenge and the isolated impact of scaffold architecture on tissue function and cellular physiology is poorly understood. Here, we provided a direct comparison between isotropic and aligned collagen scaffold morphologies where all confounding physio-mechanical features, such as strut wall thickness and surface roughness are conserved. This enabled the independent and systematic assessment of the effects of pore macro-architecture on global tissue function and cellular maturation. We seeded our scaffolds with embryonic stem cell derived cardiomyocytes (hESC-CM) and measured tissue function through calcium signal transduction and dynamic contractile strain. The aligned tissue constructs facilitated improved signalling synchronicity and directional contractility. We further examined the influence of scaffold macrostructure on intercellular organization and intracellular development. Cells on aligned constructs conformed to the orientation of the scaffold macro-structure and were found to have phenotypic and genetic markers of increased maturity. Our results isolate the influence of scaffold macro-structure on engineered tissue function at multiple length scales. These findings inform the design of optimized cardiac tissue and expand the potential for engineered tissue in regenerative and model medical systems by reducing the gaps in tissue functionality that limit their utility.


Generation of anisotropic and isotropic collagen scaffolds by directional icetemplating
Identification of the functional influence of scaffold structure is technically challenging due to the inherent link between manufacturing methodology and scaffold structure.
To systematically study the isolated functional effects of architectural anisotropy in engineered heart tissue we decided to focus on three-dimensional collagen scaffolds generated by directional ice templating. Directionally freeze-cast scaffolds are characterized by an inherent structural asymmetry (Figure 1a (Figure 1d). We exploited this feature to produce thin scaffold discs (8 mm diameter; 500-800 µm thickness) that were dominated by the architecture present on the circular face of the structure.
Scaffolds sliced along the transverse plane of the parent scaffold were characterized by aligned pores (aligned scaffolds) whereas the ones sliced along longitudinal plane were mainly characterized by non-aligned circular pores (isotropic scaffolds). Due to the use of a common parent scaffold all other physio-mechanical properties such as pore size, strut wall thickness, permeability, interconnectivity, and surface roughness were maintained across scaffold conditions this allows us to systematically investigate the influence of structure on construct performance.

Scaffold architecture does not affect cell viability and distribution.
Once generated the scaffolds were populated with hESC-CM (over 90% Troponin-T positive Figure 1e) and cultured for 7 days. At day 7, cellular viability and density were assessed via Alamar Blue fluorescence viability assay and by calculating nucleus density on immunofluorescence images of cellularized scaffolds, respectively. Our results showed no significant difference in cell survival or cell distribution between the two scaffold architectures ( Figure 1 f & g, 5.3x10 6 +/-2.5x10 6 vs 5.0x10 6 +/-1.5x10 6 arbitrary fluorescence units and 0.002+/-0.0004 vs 0.002+/-0.0006 cells/µm 2 on aligned and isotropic scaffolds respectively). We therefore assume that any observed functional differences are not due to preferential survival on a particular scaffold type.

Scaffold anisotropy improves contractility.
To analyse the contractility of the cellularized patches we utilised optical strain analysis to conduct a full field spatial-temporal assessment of construct deformation without constraining or disrupting the engineered tissue (Supplementary videos 1-2) 46,53,54 .
Scaffold architecture was found to dramatically influence deformation profiles and, thus, resultant principal strains (ε1 and ε2). Principal strain dynamics for both conditions occurred concurrently during contraction, however, isotropic constructs produced strains with equal and opposite magnitudes, indicating no net surface area change during the contraction (Figure 2a). The average maximal contractile strain (ε1) was -0.018+/-0.008 with a maximal inotropic strain rate of ~0.1 s -1 and lusitropic strain rate of ~0.05 s -1 (Figure 2b). Spatial analysis of each principal strain at peak contraction further illustrates deficient contractile function as large variability in magnitude and direction was observed across the construct surface (Figure 2c & d). In contrast principal strains of aligned constructs indicated a net negative change in surface area during contraction, a finding more consistent with functional cardiac tissue 53 ( Figure   2e). The average maximal contractile strain magnitude for ε1 (0.145+/-0.039) was ten times greater than both the orthogonal component, ε2 (0.014+/-0.023), as well as the principal strains of isotropic constructs (0.018+/-0.008 and 0.015+/-0.001 for ε1 and ε2 respectively) ( Figure 2 i-j). A similar relationship was observed in the strain rate profile ( Figure 2f). The contractile function of aligned constructs was further confirmed through spatial analysis of peak contraction demonstrating coordinated full field principal strains during contraction (Figure 2g & h). Furthermore, the structural anisotropy was found to direct the deformation such that ε1 was oriented parallel to scaffold alignment (Figure 2h). Scaffold alignment dramatically impacted construct contractility such that aligned scaffolds facilitated increased contractility indicated by To understand how closely the anisotropic constructs resemble the native cardiac tissue, we next compared the dynamic strain profiles for each architectural condition with previously reported strain dynamics of in vivo myocardial deformation 53 . Three key characteristics of the strain rate profile have been identified: a global minimum during peak systole (SRs), a global maximum with reduced magnitude during diastole (SRe) and a small maximum during the isovolumetric contraction of diastole (SRa) 53 .
The directional deformation dynamics produced by aligned constructs are consistent with the deformation profiles of native cardiac tissue in vivo, where ε1 is shown to be 2.5 times greater than ε2 53 . Similarly, with the exception of the intermediate peak due to isovolumetric contraction (SRa), the shape of the strain rate profile produced by aligned constructs is consistent with the strain rate produced by native myocardium.
SRs and SRe were found to be -0.04 and 0.025, approximately 44% and 83% of physiologically recorded values respectively 53 (Figure 2 l). Taken together, we demonstrated that aligned constructs not only enhance engineered construct contractility but also recapitulate the strain and strain rate characteristics of in-vivo cardiac tissue. These results are of particular relevance because the coordinated deformation between engineered and host tissue is paramount to ensure mechanical engraftment at the host/graft interface and has been one of the major functional limitations of engineered heart tissue 37,39,40 .

Electrical connectivity is improved in anisotropic scaffolds.
To explore the impact of long-range scaffold order on electrical conduction of

Scaffold anisotropy promotes cell alignment.
To place these functional differences between scaffold architectures in context and further understand the impact of extracellular order on engineered tissue development we performed orientation and coherence measurements of seeded cardiomyocytes.
Quantitative Fourier analysis of immunofluorescence micrographs of actin cytoskeletal structure (Phalloidin) was used to evaluate cellular orientation and intercellular alignment. Cells seeded onto isotropic scaffolds exhibited no preferential orientation direction ( Figure 5 a-d), whereas the cellular population of aligned constructs had a more uniform orientation and conformed to the extracellular macrostructure ( Figure 5 e-h) as demonstrated by increased cellular coherence (0.236+/-0.029 for aligned and 0.155+/-0.039 isotropic p=0.008) and reduced orientation variance (227.54+/-146.73 for aligned and 2.5x10 3 +/-1.6x10 3 for isotropic p=0.012) (Figure i-k). Homogeneity of cellular directionality can help to explain the improved contractile performance of aligned constructs, as coordinated, directional cell shortening increases tissue level deformation 46,58-60 .

Scaffold anisotropy increases cardiomyocyte maturation.
We next investigated in more detail the cardiomyocyte intracellular structural machinery, namely sarcomeres and gap junctions, to elucidate whether macrostructural tissue order influences phenotypic cellular development. We assessed sarcomere development by looking at sarcomeric α-Actinin staining. Cells Taken together these observations support the notion that macro-structural alignment facilitates phenotypic maturation of hESC cardiomyocytes which in turn promotes the formation of engineered heart tissue that more closely recapitulates the native myocardium. We further placed in context the observed functional differences between macrostructural morphologies by assessing cellular organization and phenotypic maturity.

Discussion
Unsurprisingly given the dramatic functional differences, it was found that the phenotypic development and intercellular organization were all positively influenced by increased architectural order of the engineered construct. The superior contractility of aligned constructs was supported by intercellular directional coordination of the cardiomyocyte population and by improved gap junction and sarcomere development 46 . Molecular analyses also suggested cellular and sarcomeric maturity as there was a shift from foetal sarcomeric proteins to the more mature forms. The improved calcium signalling dynamics and implied electrical conductivity were also supported by findings at multiple length scales. At the intracellular level the increased gap junction development observed in aligned constructs will improve intercellular signal propagation and electrical conductance. At a molecular level the increased expression of the ryanodine receptor helps to explain the increased responsivity and intracellular calcium handling efficiency and further validate a more mature phenotype.
While these results clearly support our claim that macro-structural extracellular order alone can improve engineered cardiac tissue function, it is difficult to ponder a direct mechanism by which long-range alignment influences gene expression when local surface and mechanical characteristics at a micro-and nano-scale are conserved.
We instead propose an indirect mechanism by which phenotypic cellular maturity is enhanced.
Bouchard and colleagues found that the application of increasing external electromechanical stimulation, gradually, hastened the maturation process of pluripotent stem cell derived-CM 64 . We hypothesise that long range anisotropic architecture facilitates a more coordinated contractile behaviour via globally organized cellular orientation. This enhanced and uniform contraction is effectively an autoloading system, in which the amplified contractile force, facilitated by structural alignment, serves to cyclically stimulate the tissue construct and enhance the rate of cellular maturation 64 . Therefore as maturity increases, so too does the magnitude of the external stimulation which, in turn, further encourages maturity, creating an autologous loop that parallels the work done by Bouchard et al. 2019 64 . The effect is significantly reduced for the isotropic structure, as no long-range cellular order is present.
Several approaches have been developed to produce 3D engineered cardiac tissue, however, recently the focus has shift towards the study of macroscopic scaffold architecture and its effects on cellular phenotype 17,45,68 . Our study demonstrates, through a direct and systematic comparison, that anisotropic extracellular structure enhances the functional biomimetic capacity of engineered heart tissue at multiple length scales. Despite the necessity for further study to fully deconvolute the relationship between tissue structure and cellular maturation, our work shows how the application of macro-scale structural order directs engineered tissues towards a phenotype resembling the native myocardium and informs the design of an optimal tissue engineered myocardium for regenerative medicine and disease modelling applications.

Scaffold production
Collagen slurry preparation A 1 w.t.% suspension of insoluble type I bovine dermal collagen (Devro) was prepared in 0.05 M acetic acid solution (Sigma-Aldrich UK). The mixture was left at 4°C to swell for 24 hours and homogenized in a blender at 22,000 rpms for 6 minutes. Gas was removed from the solution using a vacuum chamber (VirTis SP Scientific Wizard 2.0); the pressure was ramped from 750 torr to 2000 mtorr in 10 minutes. The slurry was allowed to habituate to room temperature (25°C).

Directional Ice-templating
Collagen slurry (9 mL) was loaded into a cylindrical polycarbonate mould (30 mm height, 20 mm internal diameter, 40 mm external diameter) with a copper base (2 mm thickness). The mould was placed onto a PID temperature controlled cold finger cooled with liquid nitrogen and programmed to hold at -10°C for 1 minute followed by cooling at a rate of 0.2°C min -1 . The top of the mould was exposed to the ambient environment.
After solidification, moulds were dried in a freeze drier (VirTis SP Scientific Wizard 2.0) at 0°C under a vacuum of less than 100 mtorr for 20 hours.

Cross linking
Cross-linking was carried out using a ratio of 5:2:20 EDC:NHS:COOH groups in collagen was used to cross link at 5% of the standard (5:2:1) 69,70 . Cross-linking reagents were dissolved in 95% ethanol and scaffolds were soaked for 2 hours within the mould. Scaffolds were washed (5 x 5 min) with deionised water.
After cross-linking scaffolds were freeze dried (VirTis SP Scientific Wizard 2.0) with a cooling rate of 0.2°C min -1 to a primary freezing temperature of -20°C. Drying occurred at 0°C under a vacuum of less than 100 mtorr for 20 hours.

Slicing
An 8 mm biopsy punch and sliced with a straight razor to a thickness of 500-700 mm.
Aligned structures were cut such that the circular face of the scaffold was parallel to the longitudinal plane of structural alignment. isotropic scaffolds were cut such that the circular face of the scaffold was parallel to the transverse plane of structural alignment as shown in Figure \ref{fig:MethodsScaffoldSlice}.

Scaffold Imaging
Scanning electron microscopy (SEM) micrographs were taken of scaffolds prior to cross linking. Collagen scaffolds were-sputter coated with gold for 2 min at a current of 20 mA. All micrographs were taken using a JEOL 820 SEM, with a tungsten source, operated at 10 kV.
X-ray micro-computed tomography (mCT) images (Skyscan 1172) were taken of each scaffold with a voltage of 25 kV, current of 138 mA and a pixel size of 5.46 mm.
Reconstructions of mCT images were performed with NRecon software by Skyscan.

Scaffold Analysis
Reconstructions were divided into nine volumes of interest (2.5 x 2.5 x 6.5 mm 3 ) dispersed across the bottom, middle, and top of the structure. Pore size analysis was applied to each transverse slice within the regions of interest. ImageJ software was used to binarize and watershed transverse slices and particle analysis was employed to compile pore size data. The pore sizes were analysed and visualized in MATLAB R2018a.
Fast Fourier Transform analysis was used to assess pore alignment according to the method laid out by Ayres et. al 2008 52 . 2D fast Fourier transform analysis was performed and radial sums of the resultant transform were collected in ImageJ. Pixel intensity for each radial direction was normalised by the minimum and plotted in MATLAB R2020a. The degree of preferential alignment was termed eAOP and utilised to compare between samples.

Cell Differentiation
CH9 hESCs were maintained and differentiated according to protocols described by

Scaffold Cellularization
Scaffolds described in section 0 were sterilised in 70% EtOH for 30 min. The EtOH was removed by PBS washing 3x5 minutes prior to scaffold conditioning with cell culture media (CDM BSA) for 1 hour in preparation for cell seeding. Cardiomyocytes were dissociated using TrypLE (Life technologies) and seeded at a density of 2 x 10 6 cells per scaffold in CDM BSA supplemented with ROCK inhibitor 1 µM.

Viability
PrestoBlue Cell Viability Reagent (Thermo Scientific) was added to culture media according to the manufacturer's instructions after 7 days of culture. Cells were incubated with the dye for 4 hours. Media was then sampled and fluorescence at 560 nm was analysed using VICTOR Multilabel Plate Reader (Perkin Elmer). Media containing PrestoBlue incubated in empty wells was used as background control.

Strain Analysis
Bright field videos were recorded on an Axiovert inverted microscope (Zeiss) using a Sony LEGRIA camera. Strain analysis was performed on bright field video samples with Ncorr digital image correlation software run on MatlabR2020a. The scaffold structure under bright field provided a reliable speckle pattern with sufficient contrast for analysis. A subset radius of 30 pixels and spacing of 5 pixels was used with the high strain option enabled. To avoid error due to global translation, the reference image was redefined after each beat, while the scaffold was in a relaxed state.
Principal strain calculations were performed with MatlabR2020a. Principal angle characterization was performed using circular analysis of diametrically bimodal circular distributions 73 .

Calcium Dynamic Analysis
At day 7 after seeding, Fluo-4 AM (10 µg ml -1 , Life technologies) was added to the cell culture media for 30 min at 37ºC.Scaffolds were then transferred in Tyrode's buffer and videos were recorded either with no stimulation or while pacing at frequency of 1 and 1.5 Hz using c-PACE EM pace (IONOPTIX). Videos were recorded on an Axiovert inverted microscope (Zeiss) using a Sony LEGRIA camera.
Video analysis was performed in MatlabR2020a. Fluorescence intensity was normalized and mean intensity was plotted against time. Both intensity peak frequencies and Fast Fourier Transform analysis were used to calculate pulse rate.
For samples that did not exhibit spatial deformation during calcium fluorescence, pulse rates were also calculated for each pixel to indicate global signalling uniformity.
Individual pulse times were recorded for each pixel and the temporal signalling uniformity in space was visualised through isochrones in MatlabR2020a.

Immunocytochemistry
Cell-seeded constructs were washed once in PBS then fixed for 1 hour with 4% PFA.
Incubation with primary antibody (diluted accordingly) was then performed. Constructs were then washed in PBS and incubated overnight with the appropriate secondary antibody, or phalloidin where appropriate, overnight. Constructs were then washed and stained with DAPI (Sigma, 1 µg ml -1 ) for 1 hour prior to imaging. Micrographs were obtained using an SP-5 confocal microscope (LEICA). Primary (I) and secondary (II) antibodies are listed in Table 1. Table 1 Primary and secondary antibodies

Cell density
Dapi stained nuclei were counted with particle analysis in ImageJ. The cell density for 200 µm 2 regions of interest was calculated in MatlabR2020a for each scaffold.

Cellular Alignment
F-actin staining was used to characterise cellular spreading and cytoskeletal alignment. The F-actin orientation and coherence of cardiomyocytes after 7 days of culture was measured for 50 µm 2 sections (27 measurements were taken per scaffold) with the OrientationJ plugin for ImageJ. The intra-scaffold variance was calculated for each individual scaffold.

Sarcomere development
Individual sarcomere chains were isolated from confocal images showing a-actinin such that the sarcomere band spanned the height of the region of interest. Banding intensity was characterize 61,62 . Fluorescence intensity was normalised by the minimum fluorescence (f0) such that, fnorm=(f/f0)-1. The mean fluorescence intensity signal was plotted along the length of the sarcomere chain and the relative prominence of each intensity peak was measured in MatLabR2020a to calculate sarcomere intensity.
Sarcomere width was defined as the signal wavelength.

Gap junction density
Immunofluorescence staining of Connexin-43 was used to visualise gap junction structures through fluorescence microscopy. Mature gap junctions were counted with particle analysis in ImageJ. The gap junction density per nucleus was calculated in MatlabR2020a for each scaffold. All primers were designed to span an intron-exon junction, and are listed in Table 2.

RNA extraction, retrotranscription and RT-qPCR
The relative expression of mRNA was obtained using the DCt method.

Open data
The original data from this paper is available at doi XXX.XXX.XXXX (site to be added)  c. d.

Figure 3
Live Fluo-4 AM calcium staining was performed on immature cardiomyocytes derived from H9 hESCs after 7 days of culture, video recordings of fluorescence dynamics were used to assess the temporal and spatial signalling uniformity a & b Mean fluorescence intensity in time for cardiomyocytes on isotropic and aligned scaffolds, respectively. c Pulse rate for all isotropic and aligned samples (aligned N=5; isotropic N=8). d-g Pulse rate in space and associated histogram for  f. e. d.
i. j. k.   a. i. j. l. n.

Figure 6
Intra-cellular structures and gene expression. a-f hESC-CM stained for sarcomeric α-actinin (red) after 7 days on (a-b) isotropic scaffolds and (c-d) aligned scaffolds; scale bars represent 20 µm. b & d representative quantification of sarcomere organization through relative intensity peak prominence along a single sarcomere chain. e relative intensity peak prominence (sarcomere intensity). f sarcomere length for cells on aligned (N=4) and isotropic (N=3) scaffolds. g-k hESC-CM stained for Dapi (blue) Troponin (Green) and Connexin (red) after 7 days on (g-h) isotropic scaffolds and (i-j) aligned scaffolds; scale bars represent 50 µm. k gap junction density for all isotropic (N=5) and aligned (N=4) samples. l-m qPCR quantification of relative expression of (l) RYR and (m-n) MYH7 to MYH6 and TTNI3 to TTNI1 expression ratios (N=3).