STRUCTURAL PHENOTYPING OF MYOCYTES

CROSS-REFERENCES TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No.

62/118,806 filed on February 20, 2015, the contents of which are incorporated by reference herein in its entirety.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR

DEVELOPMENT

This invention was made with government support under grant nos. HL 100408 and TR000522, awarded by the National Institutes of Health (NIH). The government has certain rights in the invention.

BACKGROUND OF THE INVENTION

In response to widespread efforts to commercialize differentiated stem cells (Brower, 1999, Nat Biotechnol 17: 139-142), the U.S. Food and Drug Administration (FDA) established a set of regulations and guidelines for manufacturing and quality control evaluation of human cellular and tissue-based products derived from stem cells (Current Good Tissue Practice (CGTP) and Additional Requirements for Manufacturers of Human Cells, Tissues, and Cellular and Tissue-Based Products (HCT/Ps). Food and Drug

Administration Center for Biologies Evaluation and Research (2011)). The recommendations outlined for evaluating differentiated stem cell phenotype were developed specifically to address patient safety concerns, such as tumorigenicity and immunologic incompatibility due to the initial focus of the industry on regenerative medicine applications (Fink, 2009, Science 324: 1662-1663). Concerns over patient safety may have slowed the commercialization of regenerative therapies (Fox, 2011, Nat Biotechnol 29:375-376), but the use of industrial stem cell-based products for in vitro research, particularly pharmaceutical screening applications (Rubin, 2008, Cell 132:549-552; Wobus and Loser, 2011, Arch Toxicol 85:79-117) is a promising goal that can potentially be reached in the near term.

Due to the mandate to test all drug compounds for potential adverse effects on the heart, in vitro cardiac toxicity screening is a particularly important application that has prompted the development of commercial stem cell-derived cardiac myocytes by a number of companies (Webb, 2009, Nat Biotechnol 27:977-979). In this context, the focus of quality assurance shifts from patient safety concerns to the development and adoption of measures that ensure these cells reliably mimic cardiac myocytes found in vivo. Unfortunately, no standardized guidelines currently exist for the comprehensive evaluation of structure, function and gene expression profile in stem cell derived myocytes. As a result, it is unclear whether the various stem cell-derived myocyte cell lines on the market exhibit comparable performance to one another, or if any of them accurately recapitulate the characteristics of native myocytes.

Several efforts have been reported in the emerging field of structural phenotyping for the integration of image acquisition, processing, and analysis to assess the response of cells and tissues to various challenges (Eliceiri et al, 2012, Nature Methods, 9: 697-710). All these methodologies are predicated on the assumption that cell shape is an important indicator of the cell pathophysiological state and rely on: i) image processing algorithms for the extraction of morphological features and ii) machine learning strategies to mine the cell morphology data (Crane et al, 2012, Nature Methods, 9: 977-980; Jones et al., 2009, Proc Natl Acad Sci USA, 106: 1826-1831; Treiser et al, 2010, Proc Natl Acad Sci USA, 107: 610- 615).

Many intracellular structures, such as the contractile cytoskeleton in striated muscles, are also predictors of cell function (Feinberg et al., 2007, Science, 317: 1366-1370).

Additionally, while a cell specifies along the myocyte lineage (Mummery et al., 2012, Circ Res, 111 : 344-358; Qian and Srivastava, 2013, Circ Res, 113: 915-921), it also progresses through myofibrillogenesis as force-generating units, known as sarcomeres, self-assemble along the actin cytoskeleton (Grosberg et al, 2011, PLoS Comp Biol, 7: el001088; Parker et al., 2008, Circ Res, 103: 340-342). Moreover, the contractile proteins of mature myocytes are continuously turned over and their spatial organization remodeled to adapt to

pathophysiological stimuli (McCain et al., 2013, Proc Natl Acad Sci USA, 110, 9770-9775; Sun et al, 2012, Sci Transl Med, 4: 130ral47; Wang et al, 2014, Nature Medicine, 20: 616- 623). Therefore while the presence of contractile proteins is necessary for myocytes function (Cahan et al, 2014, Cell, 158: 903-915; Mummery et al, 2012, Circ Res, 111 : 344-358), it is not sufficient to define the developmental stage (Sheehy et al, 2014, Stem Cell Reports, 2: 282-294), the health status (Wang et al, 2014, Nature Medicine, 20: 616-623) or the functional capabilities of these cells (Feinberg, 2007, Science, 317: 1366-1370).

Thus, there is a need in the art for a quality assessment and phenotype classification routine that involves relevant measurement parameters that are representative of the structural and functional phenotype of myocytes, including stem cell-derived myocytes. The present invention satisfies these needs. BRIEF DESCRIPTION OF THE DRAWINGS

The following detailed description of preferred embodiments of the invention will be better understood when read in conjunction with the appended drawings. For the purpose of illustrating the invention, there are shown in the drawings embodiments which are presently preferred. It should be understood, however, that the invention is not limited to the precise arrangements and instrumentalities of the embodiments shown in the drawings.

Figure 1, comprising Figure 1 A through Figure IE, is a comparison of mES, miPS and neonate gene expression profiles on isotropic and anisotropic ECM substrates. Figure 1A depicts culturing (i) mES, (ii) miPS and (iii) neonate myocytes on substrates with a uniform coating of FN resulted in isotropic cellular arrangement. Figure IB are volcano plots showing statistical comparisons of qPCR measurements of cardiac genes between (i) mES and neonate isotropic monolayers, and between (ii) miPS and neonate isotropic monolayers that reveal significant differences for a number of genes (two-tailed T-test, n = 3 for all conditions, points on the plot colored green or red represent genes with p < 0.05). Figure 1C depicts culturing (i) mES, (ii) miPS and (iii) neonate myocytes on substrates with micro-contact printed lines of FN that were 20 μιτι wide and spaced 4 μιτι apart resulted in anisotropic cellular arrangement in all three cell types. Figure ID are volcano plots showing statistical comparisons of qPCR measurements of cardiac genes between (i) mES and neonate anisotropic monolayers, and between (ii) miPS and neonate anisotropic monolayers that reveal slightly fewer genes demonstrating significant differences than in the isotropic cultures (two-tailed T-test, n = 3 for all conditions, points on the plot colored green or red represent genes with p < 0.05). Figure IE depicts hierarchical clustering of mean 2 ^"ACt values for a select panel of genes representing the components of the sarcomere, ion channel subunits, and genes commonly used to deduce ventricular vs. atrial identity that revealed a distinct separation between the neonate expression profile and the expression profiles of the mES and miPS engineered tissues. Scale bars = 100 μιτι.

Figure 2, comprising Figure 2A through Figure 2E, is a comparison of myofibril architecture in mES, miPS and neonate engineered tissues. Immunofluorescence visualization of sarcomeric a-actinin in Figure 2A depict isotropic monolayers of (i) mES, (ii) miPS, and (iii) neonate myocytes and in Figure 2B depict anisotropic monolayers of (i) mES, (ii) miPS, and (iii) neonate myocytes revealed the pattern of sarcomere organization adopted by each cell type in response to geometric cues encoded in the ECM. Immature premyofibrils (red arrows) were observed exclusively in mES and miPS engineered tissues. Quantitative evaluation of sarcomeric a-actinin immunofluorescence micrographs allowed statistical comparison of sarcomere organization and architecture. As shown in Figure 2C, orientational order parameter (OOP) was used as a metric of global sarcomere alignment within the engineered tissues and showed that anisotropic neonate tissues exhibited significantly greater overall sarcomere alignment than the mES and miPS anisotropic tissues. No significant differences in global sarcomere alignment were observed between the isotropic mES, miPS and neonate tissues. Figure 2D is a comparison of z-line spacing that revealed the neonate anisotropic tissues exhibited significantly greater sarcomere length than both the mES and miPS anisotropic tissues. As shown in Figure 2E, from the measurements of sarcomere length, sarcomere packing density was calculated for anisotropic tissues of each cell type. All three cell types exhibited significantly different sarcomere packing densities from the two other cell types, indicating that each type of myocyte gave rise to a unique sarcomere packing density. All results presented as mean ± standard error of the mean. Statistical tests used were either ANOVA (* = p < 0.05), or ANOVA on ranks († = p < 0.05). Scale bars = 10 μιη.

Figure 3, comprising Figure 3A through Figure 3H, is a comparison of electrical activity in mES, miPS and neonate engineered tissues. As depicted in Figure 3A, patch clamp recordings taken on isolated mES, miPS and neonate myocytes exhibited action potentials (AP) with both (i) ventricular-like, and (ii) atrial-like profiles. As depicted in Figure 3B, characterization of the AP traces revealed no significant differences between the three cell types, but the mES and miPS myocytes exhibited an equal proportion of ventricular-like (mES-v, miPS-v) and atrial-like (mES-a, miPS-a) AP traces, whereas the neonates exhibited primarily ventricular-like (neonate-v) AP profiles. As depicted in Figure 3C the

electrophysiological characteristics of anisotropic (i) mES, (ii) miPS, and (iii) neonate tissues were assessed using optical mapping and the photovoltaic dye RH237. As depicted in Figure 3D, comparison of conduction properties between the mES, miPS and neonate tissues revealed no significant differences in either longitudinal (LCV) or transverse (TCV) conduction velocity. As depicted in Figure 3E, evaluation of optical AP duration in anisotropic tissues revealed no significant differences in APD50, but a significant difference in APD90 between mES and neonate tissues was observed. As depicted in Figure 3F, comparison of Ca ²⁺ transients measured in anisotropic tissues revealed that the 50% decay time (CaT50) of the miPS tissues was significantly lower than the both the mES and neonate tissues, but the 90% decay time (CaT90) of both the mES and miPS tissues was significantly lower than the neonate tissues. As depicted in Figure 3G, patch clamp recordings were collected on isolated mES, miPS, and neonate myocytes to measure and compare (i) L-type, and (ii) T-type Ca ²⁺ current densities elicited at various holding potentials. As depicted in Figure 3H, patch clamp recordings of maximum Ca ²⁺ current density in isolated mES, miPS and neonate myocytes revealed a significant difference in total Ca ²⁺ current density (TOT) between the neonate and mES myocytes. No significant differences in L-type Ca ²⁺ current density (LCC) were observed, but a significant difference in T-type Ca ²⁺ current density (TCC) was observed between the neonate and mES myocytes. All results presented as mean ± standard error of the mean. Statistical test used was ANOVA (* = p < 0.05). Scale bars = 20 μηι.

Figure 4, comprising Figure 4A through Figure 4D depicts the comparison of contractile performance in mES, miPS, and neonate engineered tissues. As shown in Figure 4A, contractile performance of anisotropic mES, miPS, and neonate tissues was assessed using the muscular thin film (MTF) assay wherein the radius of curvature of the MTFs at (i) diastole and (ii) peak systole were used to calculate contractile stress. As depicted in Figure 4B, the radius of curvature of the MTFs was used to calculate and compare the temporal contractile strength profiles of anisotropic mES (green), miPS (red), and neonate (blue) tissues. As depicted in Figure 4C, comparison of MTF contractile output revealed that neonate anisotropic tissues generated significantly greater diastolic, peak systolic, and twitch stress than both the mES and miPS tissues. Depicted in Figure 4D, is a graphical representation of action potential morphology (black solid line), calcium transient morphology (blue dotted line), and contractility profile (red dotted line) during a typical excitation-contraction cycle of the mES, miPS, and neonate engineered anisotropic tissues. Statistical test used was ANOVA (* = p < 0.05).

Figure 5 is an integrated visual comparison of mES, miPS, and neonate experimental measurements. Strictly Standardized Mean Difference (β) values were computed for mES- and miPS-derived myocytes relative to the neonate cardiac myocytes from the mean and sample standard deviations collected for each experimental measurement. Descriptions for each abbreviation listed in the right-hand column can be found in Table 2. These β values were organized by measurement type (i.e. gene expression, myocyte architecture, electrophysiology, contractility) and plotted to allow comparison. Negative β values indicate measurements with higher relative magnitude in the neonate cardiac myocytes, whereas positive β values indicate measurements that were higher in the mES/miPS myocytes relative to the neonate cardiac myocytes. Figure 6, comprising Figure 6A through Figure 6F, is an evaluation of myocyte morphology. Figure 6A depicts isotropic cultures of (i) mES, (ii) miPS, and (iii) neonate cardiac myocytes fixed and immunostained for the presence of sarcomeric a-actinin (red), F- actin (green), and chromatin (blue). Cardiac myocytes were identified by the presence of sarcomeric a-actinin positive z-lines, and the boundaries of fully spread, mono-nucleated myocytes were manually traced using the polygon tool in Image! The total number of pixels contained within each traced polygon was used to calculate cellular aspect ratio (Figure 6B), and the total spread surface area (Figure 6C) for each cell type. Similarly, the voltage sensitive dye RH237 used for optical mapping experiments allowed identification of myocyte boundaries in anisotropic monolayers of mES (Figure 6D(i)), miPS (Figure 6D(ii)), and neonate cardiac myocytes (Figure 6D(iii)). The total number of pixels contained in each manually traced outline was used to calculate aspect ratio (Figure 6E), and total spread surface area (Figure 6F) for each type of myocyte. All results presented as mean ± standard error of the mean. Statistical tests used was ANOVA on ranks († = p < 0.05). Scale bars = 20 μηι.

Figure 7, comprising Figure 7A through Figure 7C, depicts sarcomere structural characterization. Image processing flow in Figure 7A: sarcomeric a-actinin immunographs were deconvolved, projected onto a single 2D image and then processed with a tubeness operator before further processing. In Figure 7B, the orientations of sarcomeric a-actinin positive pixels were detected with a structure tensor method, color coded using the hsv digital image representation (Figure 7B(i)) and finally displayed into a histogram (Figure 7B(ii)) of the normalized occurrences of each orientation. In Figure 7C, the sarcomere length and the overall regularity of the cytoskeletal structure were detected processing the immunograph 2D Fast Fourier Transform algorithm. The detected power spectrum (Figure 7C(i)), (for representation purpose a gamma correction of 0.1 was applied) was then integrated and normalized by the total energy. In Figure 7C(ii), the sarcomere packing density was defined as the area under the signal peaks (red curve) whose location related with the sarcomere length.

Figure 8, comprising Figure 8A through Figure 8D, depict ratiometric Ca ²⁺ transient measurements. In Figure 8A, anisotropic tissues were loaded with Fura-red and 20 lines

(white box, direction indicated by the white arrow) were scanned in dual-excitation mode at 405 nm (Figure 8A(i)) and 488 nm (Figure 8A(ii)); the sampling frequency was 250 Hz. Scale bars = 15 μιτι. In Figure 8B, the background-subtracted averaged number of photons collected with excitation at 405 nm (blue) and 488 nm (green) in each frame was used to obtain 2 signals proportional to the elevation of the cytoplasmic calcium in the tissue. In Figure 8C, the ratio of these signals is an improved measurement of the calcium transient as bleaching and other artifacts are automatically corrected for. To further improve signal quality, 4-6 steady-state transients (grey box) were averaged (Figure 8D) and the following quantities were calculated: diastolic level (grey box), peak level (*), time to peak (T2P) and the duration of the calcium transient at 50% (CaT50) and 90% (CaT90) decay.

Figure 9, comprising Figure 9A through Figure 9E, show representative

cardiomyocytes from neonate mouse (Figure 9A, pCM), and mouse (Figure 9B) or human (Figure 9C) induced pluripotent stem cells (respectively miCM and hiCM) derived cardiomyocytes on square fibronectin islands. Figures 9(A-C)(i) show overlays of a -actinin (gray) and chromatin (blue) representative immunographs. Figures 9A-C(ii) show Hue Saturation Value (hsv) representation of the a-actinin channels in Figures 9A-C(iii): the Hue channel was used to color-code each sarcomeric a-actinin positive pixel with its detected orientation. Figures 9A-C(iii) show 2D Fourier power spectra corresponding to micrographs in Figures 9A-C(i) (for representation purposes only the images have a γ-correction of 0.1). Figure 9D is a ID representation of the 2D power spectrum in Figure 9A(ii) (blue curve) and non-linear fitting with periodic (red curve) and aperiodic (black curve) components. The sarcomeric packing density is obtained from the area under the periodic component (red shaded area). Figure 9E is a quantitative analysis of cytoskeletal organization with nuclear eccentricity E, orientational order parameter (OOP) and sarcomeric packing density (ε). Data are represented as mean ± standard error of the means, n=3 for each condition.

Figure 10, comprising Figure 10A through Figure IOC, is a schematic representation of myofibrillogenesis. Figure 10A depicts the actin (green) cytoskeleton self-assembles during cell spreading; sarcomeric α-actinin (red) is initially diffuse in the cytoplasm. As shown in Figure 10B, during cytoskeleton maturation, sarcomeric α-actinin localizes along the actin bundles in puncta known as Z-bodies, either at discrete locations or in relatively long stretches. As shown in Figure IOC, when myofibrillogenesis is complete, sarcomeric a- actinin is localized in a regular lattice composed by Z-disks, ultra-structural units that signal the extremities of the sarcomere. The nuclear chromatin is indicated in blue.

Figure 11, comprising Figure 11 A and Figure 1 IB depicts migratory fibroblast

(Figure 11 A) patterned on square fibronectin islands (Figure 1 IB) exhibited an actin cytoskeletal structure characterized by the presence cortical actin at the cell borders and ring- like actin stress fibers in the perinuclear region. Actin (green), vimentin (red), chromatin (blue). Scalebar: 15 μιτι.

Figure 12 depicts intrinsic cytoskeletal bias in primary cardiomyocytes (i) and murine (ii) or human (iii) induced pluripotent stem cell derived cardiomyocytes. When cultured on substrates uniformly coated with fibronectin, cardiomyocytes assume a pleomorphic shape, sustained by a cytoskeletal architecture that is the sole expression of the cell intrinsic bias, as there are no engineered boundary conditions. Under those circumstances, mononucleated (chromatin signal is encoded in blue channel) pCMs (i) and miCMs (ii) showed polarized myofibrils exhibited periodic striation of actin (green) and sarcomeric a-actinin (red). On the contrary hiCMs (iii) showed diffuse cortical actin and ring-like myofibrils. Scale bar: 15 μιτι.

Figure 13, comprising Figure 13A through Figure 13D, depicts the metrics of contractile architecture to characterize the progression of myofibrillogenesis. Figure 13A) Schematic representation of a sarcomere (Figure 13A-i) and of the distribution of a-actinin (red) during myofibrillogenesis: in the cytoplasm (Figure 13A-ii), along the actin (green) filament in the form of Z-bodies (Figure 13A-iii) and incorporated into the Z-disks (Figure 13A-iv). Figure 13B) Algorithmic detection of the orientation and periodic registration of a- actinin positive structures using the image spatial (coordinates x,y) and Fourier (coordinates u,v) domains. Figure 13C) Color-coded orientations (Figure 13C-i, from the inset of synthetic image Figure 13 A-iii) displayed into a histogram (Figure 13C=-ii) that can be fitted to identify orientations belonging to Z-disks (red) and Z-bodies (black). In parallel, the 2D

Fourier power spectrum (Figure 13C-iii) was integrated into a ID curve (Figure 13C-iv) and fitted to identify the contribution of periodically spaced Z-disks (red) and aperiodic Z-bodies (black). Figure 13D) α-actinin immunostains (white) of mono-nucleated (DAPI, blue) primary (mpCM, Figure 13D-i) and murine (miCM, Figure 13D-ii) or human (hiCM, Figure 13D-iii) induced pluripotent stem cell-derived cardiomyocytes. The color coded

representation of the α-actinin orientation in the inset is reported below the image. The positive semi-plane for the Fourier transform is reported on the right of each image. Scale bar 20 μιη. See also Figure 16 and Figure 17.

Figure 14, comprising Figure 14A through Figure 14C, depicts the structural phenotyping of stem cell-derived cardiomyocytes. Figure 14A) α-actinin (white) and chromatin (blue) images of rpCMs at 6 (Figure 14A-i), 24 (Figure 14A-ii), and 48 hrs (Figure 14A-iii) as well as hiCMs at 72 hrs (Figure 14A-iv) after seeding with color-coded orientations and Fourier representations. Scale bar 25 μιτι. Figure 14B) Scatter plot showing how the metrics of myofibrillar architecture quantitatively captured the progression of myofibrillogenesis in rpCM tissues from differentiated (6 hr, brown squares) to immature (24 hrs, red circles) and finally mature (48 hrs, green triangles) myocytes. In contrast, the hiCM tissues (orange diamonds) exhibited a relatively immature myofibrillar organization. Figure 14C) A dataset comprising -120 sarcomeric a-actinin images per conditions (insets in panels Figure 14A i-iv) was acquired and the features extracted from this dataset were utilized to train several classifiers to distinguish the classes of differentiated (D), immature (I) and mature (M) myocytes. The classifiers assigned only -29% of the 118 hiCM images to the class of myocytes with a mature structural phenotype, with the rest showing differentiated or immature contractile architectures. Results are shown as Mean ± SEM. See also Figure 18.

Figure 15 is a series of schematics to illustrate the various aspects of

myofibrillogenesis that each one of the metrics (described in Table 5) specifically captures.

Figure 16, comprising Figure 16A through Figure 16C, provides the results of experiments which demonstrate the effect of common imaging artifacts on feature extraction. Scientific images, specifically those acquired through automatic imaging (Bakal et al, 2007, Science, 316: 1753-1756; Collinet et al, 2010, Nature, 464: 243-249; Jones et al, 2009, Proc Natl Acad Sci USA, 106: 1826-1831; Whitehurst et al., 2007, Nature, 446: 815-819), can be affected by several different types of noise and artifacts. Demonstrated herein, on the synthetic images from Figure 13, is the effect of out-of-focus blurriness (Figure 16A-i), salt- and-pepper noise (Figure 16A-ii) or poor contrast (Figure 16A-iii). The metrics were robust to the effect of such noise sources, as demonstrated by the orientational order parameter (OOP) and sarcomeric energy scores. While the exact numerical values were different (Figure 16B-i), the sarcomeric energy could statistically distinguish mature (M) myocytes from the others (Figure 16B-ii). Furthermore, the OOP values were significantly different between mature (M), differentiated (D) and immature (I) cardiomyocytes (Figure 16B-iii). Results are presented as mean ± SEM and analyzed with the ANOVA test (£>-value<0.05). Moreover, the pCM image in Figure 13 was analyzed using default settings (Figure 16C-i), as well as after the application of moderate (Figure 16C-ii) and severe (Figure 16C-iii) down- sampling; salt-and-pepper noise (Figure 16C-iv); and moderate (Figure 16C-v), acute (Figure 16C-vi) and severe (Figure 16C-vii) blurring. The performances of OOP, sarcomeric packing density (SPD) and sarcomere length (SL) was then compared (Figure 16C-viii) and it is demonstrated that only severe down-sampling and salt-and-pepper noise induced a greater than 20% change in the numerical score assigned to the image. Figure 17, comprising Figure 17A through Figure 17E, depict the results of experiments where the metrics of myofibrillar organization were used to characterize the maturation of primary and stem cell derived cardiomyocytes. Images showing chromatin (blue) and sarcomeric a-actinin (white) in murine primary cardiomyocytes (mpCM, Figure 17A-i) and murine (miCM, Figure 17B-i) or human (hiCM, Figure 17C-i) induced pluripotent stem cell derived cardiomyocytes were processed to detect and color-code the principal orientations of aSA positive structures (Figure 17A-ii, Figure 17 B-ii and Figure 17C-ii respectively for mpCMs, miCMs and hiCMs). Note that the HSV digital image representation was employed here, where Hue and Saturation channels encode orientation and coherency (Rezakhaniha et al, 201 1, Biomechanics and modeling in mechanobiology, 11 : 461-473) respectively, while the Value channel encodes the preprocessed (Sato et al, 1998, Medical image analysis, 2: 143-168) sarcomeric a-actinin image. (x,y) indicates a Cartesian system of coordinates for the spatial domain. Scalebar: 20 μιτι. The normalized Fourier spectra of the Value channel were reported for mpCMs (Figure 17A-iii), miCM (Figure 17B-iii) and hiCM (Figure 17C-iii). (u, v) and (ω, i9) respectively indicate a Cartesian and polar system of coordinates for the Fourier domain. Radial integration of Figure 17A-iii lead to the ID representation in Figure 17D (green dashed line) that was further fitted to identify the periodic (red) and aperiodic (black) components needed to derive the sarcomere packing density (SPD). (Figure 17E) Note that SPD could discriminate the maturation of the cell contractile cytoskeleton in this dataset where the nuclear eccentricity (Bray et al, 2010,

Biomaterials, 31 : 5143-5150 (e) and the traditional orientational order parameter (Sheehy et al., 2012, Biomechanics and Modeling in Mechanobiology, 1 1 : 1227-1239) (OOP) fell short due to the central symmetry in the cell geometry (Grosberg et al, 2011 , PLoS Comp Biol, 7: el 001088; Sheehy et al, 2012, Biomechanics and Modeling in Mechanobiology, 11 : 1227- 1239). Results are mean ± SEM, and were analyzed with ANOVA (p<0.05)

Figure 18, comprising Figure 18A through Figure 18D, depict the results of experiments using a myofibrillogenesis dataset. Figure 18A) Representative images from the dataset used in this study: columns from left to right represent sarcomeric α-actinin images collected at 6 hr, 24 hr and 48 hrs after seeding of rat primary cardiomyocytes (rpCMs) as well as commercially available human induced pluripotent stem cell derived myocytes

(hiCMs). Scale bar is 20 μιτι. The set of images in the red inset were color-coded to highlight the orientation of each a-sarcomeric positive pixel in the image, demonstrating how more mature myocytes have aligned sarcomeres sharing the same color. Figure 18B) 2D Fourier power spectra for the same set of image in the red inset in Figure 18A. Note how increasingly more mature myocytes display more intense periodic signal in the Fourier domain. Figure 18C) Schematic representations of the classifiers used to analyze this dataset: naive Bayes (NB, Figure 18C-i), neural network (NN, Figure 18C-ii) and tree bagging (TB, Figure 18C- iii). Figure 18D) Error rates for training (black) and classification (gray) as measured in ten randomly seeded and optimized iterations for each classifier. Results are presented as mean ± SEM.

Figure 19 is a flowchart of an exemplary method of the present invention, using a machine learning algorithm to provide an un-biased classification of the phenotype of one or more cells from a test group of cells.

Figure 20, comprising Figure 20A through Figure 20C, depicts the results of experiments quantifying skeletal myoblast actin cytoskeletal and nuclear architecture (Figure 20 A) Representative (i) anisotropic healthy (ii) anisotropic DMD, and (iii) isotropic engineered human skeletal muscle myoblast tissue immunostained for f-actin (black) and nuclei (blue). Scale bar is 25 μιη. (Figure 20B) The actin OOP is plotted for healthy and DMD anisotropic myoblast tissues as well as healthy isotropic myoblasts. The bars represent the mean OOP ± standard error, n=5 coverslips, 10 fields of view (318.5 μιτι x 318.5 μπι) per sample, for each condition. ** indicates p<0.05 relative to both the DMD anisotropic and healthy isotropic conditions. * indicates p<0.05 relative to the healthy isotropic condition. (Figure 20C) The myoblast nuclear OOP is plotted for healthy and DMD anisotropic myoblast tissues as well as healthy isotropic myoblasts. The bars represent the mean OOP ± standard error, n=5 coverslips, 3 fields of view (318.5 μιη χ 318.5 μιτι) per sample, >700 nuclei analyzed for each condition. ** indicates p<0.05 relative to both the DMD anisotropic and healthy isotropic conditions. * indicates p<0.05 relative to the healthy isotropic condition.

Figure 21, comprising Figure 21A thorough Figure 21C, depicts the results of experiments quantifying changes in actin cytoskeletal and nuclear architecture during myotube maturation. (Figure 21 A) Representative (i-iii) day 3 of differentiation healthy anisotropic, DMD anisotropic, and healthy isotropic myotubes and (iv-vi) day 6 of differentiation healthy anisotropic, DMD anisotropic, and healthy isotropic engineered human skeletal muscle myotubes immunostained for f-actin (black) and nuclei (blue). Scale bar is 50 μιη. (Figure 21B) The actin OOP is plotted for healthy and DMD anisotropic engineered tissues as well as healthy isotropic engineered tissues. The bars represent the median OOP ± standard error, n=5 coverslips, 10 fields of view (318.5 μιτι x 318.5 μιτι) per sample, for each condition. # indicates p<0.05 relative to the isotropic condition on the corresponding day. * indicates p<0.05 relative to another day within a condition. (Figure 21 C) The nuclear OOP is plotted for healthy and DMD anisotropic engineered tissues as well as healthy isotropic engineered tissues. The bars represent the median OOP ± standard error, n=5 coverslips, 3 fields of view (318.5 μιτι x 318.5 μπι) per sample, >400 nuclei analyzed for each day, for each condition. * indicates p<0.05 relative to another day within a condition, # indicates p<0.05 relative to all conditions on the corresponding day. ** indicates p<0.05 relative to the isotropic condition on the corresponding day.

Figure 22, comprising Figure 22A and Figure 22B, depicts the results of experiments quantifying sarcomere packing density during maturation. (Figure 22A) Representative (i-iii) day 3 of differentiation healthy anisotropic, DMD anisotropic, and healthy isotropic myotubes and (iv-vi) day 6 of differentiation healthy anisotropic, DMD anisotropic, and healthy isotropic engineered human skeletal muscle myotubes immunostained for sarcomeric a-actinin (black) and nuclei (blue). Scale bar is 50 μιτι. (Figure 22B) The sarcomere packing density (SPD) is plotted for each condition. The bars represent the average SPD ± standard error, n=5 coverslips, 10 fields of view (318.5 μιτι x 318.5 μιτι) per sample, for each condition. # indicates p<0.05 relative to the other condition on the corresponding day. (

DETAILED DESCRIPTION

It is to be understood that the figures and descriptions of the present invention have been simplified to illustrate elements that are relevant for a clear understanding of the present invention, while eliminating, for the purpose of clarity, many other elements found in typical platforms for assessing quality of biological cell lines. Those of ordinary skill in the art may recognize that other elements and/or steps are desirable and/or required in implementing the present invention. However, because such elements and steps are well known in the art, and because they do not facilitate a better understanding of the present invention, a discussion of such elements and steps is not provided herein. The disclosure herein is directed to all such variations and modifications to such elements and methods known to those skilled in the art.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, the preferred methods and materials are described.

As used herein, each of the following terms has the meaning associated with it in this section. The articles "a" and "an" are used herein to refer to one or to more than one (i.e., to at least one) of the grammatical object of the article. By way of example, "an element" means one element or more than one element.

"About" as used herein when referring to a measurable value such as an amount, a temporal duration, and the like, is meant to encompass variations of ±20%, ±10%, ±5%, ±1%, and ±0.1% from the specified value, as such variations are appropriate.

As used herein, a "potent" cell refers to any cell that is capable of at least some differentiation. Also as used herein, a "differentiated cell" refers to any cell that has at least partially differentiated from a potent cell. Further, a "target cell" refers to the cell that the differentiated cell is being compared to, in determination of how closely the differentiated cell resembles the target cell according to at least one measurable metric.

Throughout this disclosure, various aspects of the invention can be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the invention. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subranges as well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 2.7, 3, 4, 5, 5.3, 6 and any whole and partial increments therebetween. This applies regardless of the breadth of the range.

Description

The present invention includes a system and method for quality assessment and phenotyping of myocytes, myocyte progenitor cells, and stem cell-derived cells. In one embodiment, the stem cell-derived cells are at least partially differentiated cells. In another embodiment, the stem cell-derived cells are specialized cells. In certain embodiments, the myocytes are of skeletal muscle, smooth muscle, or of cardiac muscle (i.e., cardiomyocytes).

In one embodiment, the present invention allows a user to identify differences in one or more properties of differentiated cell tissues versus the target cell tissues that have important implications for their utility. In one embodiment, the present invention allows a user to identify the commercial differentiated cell product lines that are most suitable for their needs, and of the underlying potent cells producing these differentiated cells. The present invention allows users to focus their efforts to improve cell differentiation protocols, and further serves as a robust quality control procedure for ensuring that batches of potent cells reach the desired differentiated cell phenotype.

In one embodiment, the present invention allows a user to identify the phenotype of a cell based on one or more structural metrics of the cell. The present invention thus allows a user to examine the effect of a test compound or test condition on the phenotype of a cell. For example, in one embodiment, the present invention provides a screening method of identifying a compound or condition of interest, by determining if a test compound or test condition alters the phenotype of a cell from a first phenotype to a second phenotype. For example, the screening method may be used to identify compositions or conditions which promote cell differentiation, cell proliferation, cell survival, cell death, or the like.

In one embodiment, the quality assessment of a cell or the phenotype classification of a cell is accomplished using a machine learning algorithm. For example, in certain embodiments, an algorithm is trained to provide an assessment or classification based on one or more parameters of known, target, or standard cells. The algorithm can then use the training to assign quality assessments or phenotype classification to test cells based on the detected parameters of the test cells.

Method for Calculating a Quality Index

As contemplated herein, quality assessment is made by calculating a quality index based on at least one measurable metric which may include, without limitation, factors pertaining to one or more of genetic, electrophysiological, structural, and contractile information expressed as a numerical value. In some embodiments, the metric may relate to cytoskeletal organization, such as the sarcomere packing density of cardiomyocytes. It should be appreciated that the system and method of the present invention is not limited to these particular metrics, but instead may include any measurable metric of a cell or cell phenotype, provided such metric can be expressed as a value or score. Further, the quality index may be calculated based on just one metric, or it may be calculated based on a plurality of metrics. As contemplated herein, any combination of metrics may be used, the number and type of metrics being used generally depending on the type of differentiated cells being evaluated, or any other factors as determined by the user of the present invention.

These measurements are further made against a target cell, or alternatively against pre-calculated values for a target cell, such as a set of standard values to a target cell type. For example, the system and method of the present invention can assess the quality of stem cell derived myocytes, based on the integration of genetic, electrophysiological, structural, and contractile measurements, coupled with comparison against values for these measurements that are representative of the ventricular myocyte phenotype. In this embodiment, the efficacy of this procedure can be evaluated using commercially-available murine ES- (mES) and iPS- (miPS) derived myocytes compared against neonatal mouse ventricular myocytes (neonate).

While the present invention is focused primarily on stem cell-derived myocytes, it should be appreciated that the present invention is not limited to a particular cell type. Rather, the present invention allows for the calculation of a quality index for any type of biological cell, population of cells or tissue, as derived from any type of cell having the ability to differentiate, such as a stem cell, a progenitor and the like.

To determine how closely the test cells (e.g., differentiated cells) match the phenotype of the target cells (or predetermined target cell values), the present invention integrates at least one measured metric of the test cells, and calculates the difference, referred to herein as the "normalized residue," between the at least one measured metric of the test cells against the target cells or predetermined target cell values. For example, in one embodiment, for each experimental measurement, these values may be normalized, such as to the interval [0,1] and calculated the strictly standardized mean difference (denoted herein as β) according to the following: where μ represents mean and σ represents standard deviation, to evaluate the magnitude of difference, taking into account the variance in the measurements, between the test cells and the target cells. This allows for determination of the effect size for each experimental measurement and for identification of the parameters that show the greatest degree of similarity and difference from the target cell tissues. The normalized residues, or β values, may be used from each experimental measurement to calculate the mean squared error (MSE) versus the target cell tissues according to the following:

to define a single value that represents the total difference between the test cells and target cells based on the measurements performed. Accordingly, the MSE may be used herein as a quality index to provide a numeric score of how closely the test cells match one or more characteristics of the target cells. The combination of measurable metrics employed allows a user of the system and method of the present invention to pin-point specific differences in one or more properties of engineered differentiated cell tissues versus the target cell tissues that have important implications for their utility in in vitro assays of tissue function. Further, this "quality index" not only allows users to identify the commercial differentiated cell product lines that are most suitable for their needs, it also provides insight to the source of the underlying potent cells producing these differentiated cells. The system and method of the present invention allows users may better understand where to focus their research and development efforts to improve their differentiation protocols, and further serves as a robust quality control procedure for ensuring that batches of potent cells released to customers faithfully recapitulate the desired differentiated cell phenotype.

As contemplated herein, calculation of the MSE may include a mechanism by which to weight each information item or measurable component for any metric, and to calculate a value that is determinative of that metric. In one embodiment, the lower the MSE value, the closer the test cells are to the target cells. It should be appreciated that the values designated for each information item may vary according to the metric being measured. Further, the number or combination of information item categories will also effect the values designated. Depending on the application, one or more MSE scores may be set as a threshold value, where a score of equal to or above a designated value is indicative or predictive of quality. Alternatively, final score ranges can be used to designate categories of quality. It should be appreciated that the system of the present invention is not limited to any predetermined value, number, scale or other nomenclature for the MSE.

For example, as described in Example 1 herein, the β values presented in Figure 5 resulted in an MSE score of 4.95 between the mES (differentiated cells) and neonate cardiac myocytes (target cells), whereas the miPS myocytes (differentiated cells) resulted in an MSE score of 3.60 from the neonate cardiac myocytes (target cells). In this example, the miPS myocytes exhibited a global phenotype that was slightly closer to the neonate cardiac myocytes than the mES-derived myocytes, although both the mES and miPS myocytes demonstrated substantial differences from the neonate cardiac myocytes for a number of characteristics.

Phenotype classification The present invention provides a system and method for classifying the phenotype of a cell. For example, the phenotype of a myocyte, cardiomyocyte, skeletal myocyte, stem cell- derived myocyte, myocyte precursor cell, and the like can be identified base on one or more structural, functional, electrophysiological, or gene expression parameters of the cell. In one embodiment, the phenotype of the cell can be identified on one or more structural metrics of the cell. For example, Table 5 and Figure 15 provide a set of 1 1 structural metrics that can be used to provide a quality assessment or phenotype classification of a cell. In certain aspects, providing an assessment or classification based on one or more structural metrics of the cell allows for a quicker and higher throughput analysis, as it would not require labor intensive or lengthy analysis of cell function or gene expression.

In one embodiment, the structural metrics are based upon sarcomere organization in the cell. For example, one or more structural metrics can be identified or calculated based on observation of components of the sarcomere, including but not limited to a-actinin and actin. For example, in certain embodiments, the cell is contacted with a reagent (e.g., antibody, dye, etc.) which allows for visualization of the sarcomere organization in the cell. For example, in one embodiment, the cell is stained with an anti-a-actinin primary antibody and a fluorescent secondary antibody. In certain embodiments, the cells are fixed, stained, and imaged. In another embodiment, the cells are not fixed, thereby allowing for real-time observation of sarcomere organization. In certain embodiments, other cell components (e.g., DNA, nucleus, etc.) may be labeled to provide additional information regarding the morphology of the cell. Cells may be imaged using any suitable imaging system and method known in the art, including but not limited to fluorescent microscopy, confocal microscopy, two-photon microscopy, and the like. Images may be processed using any software platform known in the art, including but not limited to ImageJ and MATLAB. For example, the software platform can be used to identify sarcomeric positive (i.e. a-actinin positive) pixels, which can then be used to calculate one or more of the structural metrics or parameters described herein.

Exemplary algorithms for calculation of the structural metrics, including the orientational order parameter and sarcomere packing density are described elsewhere herein. Machine Learning

In certain embodiments, phenotype classification of one or more cells may be performed by a machine learning algorithm. These machine learning algorithms can use one or more of the cell's functional, structural, or genetic metrics recited herein to effectively and efficiently classify the cell as having a particular phenotype. For example, the algorithm can identify a test cell as being an immature cell, mature cell, differentiated cell, healthy cell, diseased cell, and the like.

In accordance with one or more embodiments of the present invention, it will be understood that the types of models used for machine learning, using the methods provided herein, is not necessarily limited. For example, the machine learning models may include random forest, support vector machines, discriminant analysis, neural networks, artificial neural networks, naive Bayes classifier, decision trees, bootstrap aggregation of decision trees (Tree bagging), genetic algorithm, and k-nearest neighbors. The utilization of exemplary machine learning models for the structural phenotyping of cells is described elsewhere herein.

In certain embodiments, the algorithm is initially trained with one or more known datasets. For example the algorithm may be trained with images or data from cells of a known or pre-determined phenotype. The algorithm may be trained with datasets from an immature cell, mature cell, differentiated cell, healthy cell, diseased cell, and the like. In certain embodiments, the training dataset may be obtained from one or more cells of the same species as the test cells for which the algorithm will be used for classification. In certain embodiments, the training data set may be obtained from one or more cells of a different species as the test cells. For example, it is described herein that since myofibrillogenesis is a well-conserved process, the classification of human-derived cells can be done via training of the algorithm against non-human standards.

An exemplary method of the present invention is depicted in Figure 19, where a machine learning algorithm 100 is used for the phenotypic classification of test cells. In certain embodiments, the method comprises training 10 of machine algorithm 100, using a training dataset comprising images and/or extracted metrics from cells having an established phenotype. Machine learning algorithm 100 can then be used to classify 20 one or more cells of a test dataset comprising extracted metrics and/or images of cells from one or more test groups. The method can thus be used for the un-biased classification of each image in the test dataset. Screening

The present invention provides a method for identifying a composition or condition which alters the phenotype of a cell. For example, in certain embodiments the method comprises contacting a cell with a test composition and classifying the phenotype of the cell, using the methods described herein. The cell may be cultured with the test composition for a defined time period prior to phenotype classification. For example, in certain embodiments, the cell may be cultured with the test composition for about 1 second, 5 seconds, 10 seconds, 30 seconds, 1 minute, 2 minutes, 5 minutes, 10 minutes, 30 minutes, 1 hour, 2 hours, 6 hours, 12 hours, 24 hours, 3 days, 7 days, 2 weeks, 1 month, 3 months, or longer. It can be determined if the test composition alters the phenotype of the cell as compared to a similar cell which is not cultured with the test composition. Aside from the particular composition or condition being screened, the cell may be cultured using any standard culture conditions or cell culture media known in the art.

The method can be used to determine if the test composition promotes the differentiation, proliferation, maturation, survival, or death of the cell. For example, in one embodiment, an immature myocyte may be cultured with a test composition to determine if the test composition induces the maturation of the cell into a mature myocyte. In one embodiment, a diseased cell may be cultured with a test composition to determine if the composition improves the health of the cell. In one embodiment, the method comprises identifying a test compound that does not alter the phenotype of the cell. The determination of the phenotypic change of the cell may be conducted by evaluating one or more of the metrics described herein, for example by calculating a quality index or via a machine learning algorithm.

In certain embodiments, the method comprises screening for culturing conditions which may promote a change in phenotype. For example, the screened culture conditions may include particular culture times, temperature, humidity, pC02, p02, and the like.

In certain embodiments, the method is a high throughput method, where a plurality of test compositions or conditions are screened. For example, in certain embodiments, a library of compositions are screened, where each composition of the library is individually contacted to one or more cells in order to identify which compositions induce or do not induce a change in phenotype of the cell.

System Platform

As contemplated herein, the present invention includes a system platform for performing the aforementioned methods for quality assessment of differentiated cells derived from potent cells. In some embodiments, the system of the present invention may operate on a computer platform, such as a local or remote executable software platform, or as a hosted internet or network program or portal. In certain embodiments, only portions of the system may be computer operated, or in other embodiments, the entire system may be computer operated. As contemplated herein, any computing device as would be understood by those skilled in the art may be used with the system, including desktop or mobile devices, laptops, desktops, tablets, smartphones or other wireless digital/cellular phones, televisions or other thin client devices as would be understood by those skilled in the art. The platform is fully integratable for use with any additional platform and data output that may be used, for example with the measurement of a particular metric.

For example, the computer operable component(s) of the system may reside entirely on a single computing device, or may reside on a central server and run on any number of end-user devices via a communications network. The computing devices may include at least one processor, standard input and output devices, as well as all hardware and software typically found on computing devices for storing data and running programs, and for sending and receiving data over a network, if needed. If a central server is used, it may be one server or, more preferably, a combination of scalable servers, providing functionality as a network mainframe server, a web server, a mail server and central database server, all maintained and managed by an administrator or operator of the system. The computing device(s) may also be connected directly or via a network to remote databases, such as for additional storage backup, and to allow for the communication of files, email, software, and any other data formats between two or more computing devices. There are no limitations to the number, type or connectivity of the databases utilized by the system of the present invention. The communications network can be a wide area network and may be any suitable networked system understood by those having ordinary skill in the art, such as, for example, an open, wide area network (e.g., the internet), an electronic network, an optical network, a wireless network, a physically secure network or virtual private network, and any combinations thereof. The communications network may also include any intermediate nodes, such as gateways, routers, bridges, internet service provider networks, public-switched telephone networks, proxy servers, firewalls, and the like, such that the communications network may be suitable for the transmission of information items and other data throughout the system.

Further, the communications network may also use standard architecture and protocols as understood by those skilled in the art, such as, for example, a packet switched network for transporting information and packets in accordance with a standard transmission control protocol/Internet protocol ("TCP/IP"). Any of the computing devices may be communicatively connected into the communications network through, for example, a traditional telephone service connection using a conventional modem, an integrated services digital network ("ISDN"), a cable connection including a data over cable system interface specification ("DOCSIS") cable modem, a digital subscriber line ("DSL"), a Tl line, or any other mechanism as understood by those skilled in the art. Additionally, the system may utilize any conventional operating platform or combination of platforms (Windows, Mac OS, Unix, Linux, Android, etc.) and may utilize any conventional networking and

communications software as would be understood by those skilled in the art.

To protect data, an encryption standard may be used to protect files from unauthorized interception over the network. Any encryption standard or authentication method as may be understood by those having ordinary skill in the art may be used at any point in the system of the present invention. For example, encryption may be accomplished by encrypting an output file by using a Secure Socket Layer (SSL) with dual key encryption. Additionally, the system may limit data manipulation, or information access. For example, a system administrator may allow for administration at one or more levels, such as at an individual reviewer, a review team manager, a quality control review manager, or a system manager. A system

administrator may also implement access or use restrictions for users at any level. Such restrictions may include, for example, the assignment of user names and passwords that allow the use of the present invention, or the selection of one or more data types that the subservient user is allowed to view or manipulate.

As mentioned previously, the system may operate as application software, which may be managed by a local or remote computing device. The software may include a software framework or architecture that optimizes ease of use of at least one existing software platform, and that may also extend the capabilities of at least one existing software platform. The application architecture may approximate the actual way users organize and manage electronic files, and thus may organize use activities in a natural, coherent manner while delivering use activities through a simple, consistent, and intuitive interface within each application and across applications. The architecture may also be reusable, providing plug-in capability to any number of applications, without extensive re-programming, which may enable parties outside of the system to create components that plug into the architecture. Thus, software or portals in the architecture may be extensible and new software or portals may be created for the architecture by any party.

The system may provide software applications accessible to one or more users to perform one or more functions. Such applications may be available at the same location as the user, or at a location remote from the user. Each application may provide a graphical user interface (GUI) for ease of interaction by the user with information resident in the system. A GUI may be specific to a user, set of users, or type of user, or may be the same for all users or a selected subset of users. The system software may also provide a master GUI set that allows a user to select or interact with GUIs of one or more other applications, or that allows a user to simultaneously access a variety of information otherwise available through any portion of the system.

The system software may also be a portal or SaaS that provides, via the GUI, remote access to and from the system of the present invention. The software may include, for example, a network browser, as well as other standard applications. The software may also include the ability, either automatically based upon a user request in another application, or by a user request, to search, or otherwise retrieve particular data from one or more remote points, such as on the internet or from a limited or restricted database. The software may vary by user type, or may be available to only a certain user type, depending on the needs of the system. Users may have some portions, or all of the application software resident on a local computing device, or may simply have linking mechanisms, as understood by those skilled in the art, to link a computing device to the software running on a central server via the communications network, for example. As such, any device having, or having access to, the software may be capable of uploading, or downloading, any information item or data collection item, or informational files to be associated with such files.

Presentation of data through the software may be in any sort and number of selectable formats. For example, a multi-layer format may be used, wherein additional information is available by viewing successively lower layers of presented information. Such layers may be made available by the use of drop down menus, tabbed folder files, or other layering techniques understood by those skilled in the art or through a novel natural language interface as described hereinthroughout. Formats may also include AutoFill functionality, wherein data may be filled responsively to the entry of partial data in a particular field by the user. All formats may be in standard readable formats, such as XML. The software may further incorporate standard features typically found in applications, such as, for example, a front or "main" page to present a user with various selectable options for use or organization of information item collection fields.

The system software may also include standard reporting mechanisms, such as generating a printable results report, or an electronic results report that can be transmitted to any communicatively connected computing device, such as a generated email message or file attachment. Likewise, particular results of the aforementioned system can trigger an alert signal, such as the generation of an alert email, text or phone call, to alert a manager, expert, researcher, or other professional of the particular results. Further embodiments of such mechanisms are described elsewhere herein or may standard systems understood by those skilled in the art.

Accordingly, the system of the present invention may be used for calculating a quality index of a cell or for the phenotype classification of a cell. In certain embodiments, the system may include a software platform run on a computing device that calculates the normalized residue, such as a strictly standardized mean difference (β), between a test cell and a targeted cell, and calculates a mean squared error (MSE) versus the target cell to define a value that represents the total difference between the test cell and targeted cell based on at least one measured metric of the test cell. In one embodiment, the system may include a software platform run on a computing device that performs a machine learning algorithm to classify the phenotype of a test cell based on at least one metric of the test cell.

EXPERIMENTAL EXAMPLES

The invention is now described with reference to the following Examples. These

Examples are provided for the purpose of illustration only and the invention should in no way be construed as being limited to these Examples, but rather should be construed to encompass any and all variations which become evident as a result of the teaching provided herein.

Without further description, it is believed that one of ordinary skill in the art can, using the preceding description and the following illustrative examples, make and utilize the present invention and practice the claimed methods. The following working examples therefore, specifically point out the preferred embodiments of the present invention, and are not to be construed as limiting in any way the remainder of the disclosure. Example 1 : Assessment of Stem Cell Derived Myocyte Differentiation

In order to develop quality assurance standards for assessing stem cell-derived myocyte differentiation, it is necessary to first establish a set of characteristics that reliably define cardiac myocyte identity. In one example such characteristics may include evaluation of form and function that give rise to the contractile properties of cardiac myocytes in the healthy, post-natal heart. In addition to measuring the expression of cardiac biomarker genes (Ng et al, 2010, Am J Physiol Cell Physiol 299:C1234-1249; Bruneau, 2002, Circ Res 90:509-519), the organizational characteristics of the contractile myofibrils (Feinberg et al, 2012, Biomaterials 33:5732-5741), the electrical activity that regulates myofibril contraction (Kleber and Rudy, 2004, Physiol Rev 84:431 -488), and the contractile force output of the myofibrils directly (Alford et al, 2010, Biomaterials 31 :3613-3621) were also examined. Since human ventricular myocytes are not readily available, commercially-available murine ES- (mES) and iPS- (miPS) derived myocytes were used, and these were compared against ventricular myocytes isolated from neonatal mice. Accordingly, the following example demonstrates the utility of comparing stem cell-derived myocytes and isolated cardiac myocytes possessing the desired phenotype using a multi-factorial comparison of high level myocardial tissue architectural and functional characteristics.

The following materials and methods were used in Example 1. Stem cell-derived myocyte culture

CorAt murine ES- and iPS-derived myocytes were cultured according to instructions, and with culture reagents supplied by the manufacturer (Axiogenesis, Cologne, Germany). Briefly, cells were cultured in T25 flasks pre-coated with 10 mg/ml fibronectin (FN) (BD Biosciences, Bedford, MA) in puromycin-containing culture media at 37°C and 5% CO2 for 24 hours, and in media that does not contain puromycin thereafter. After 72 hours, cells were dissociated with 0.25% trypsin and seeded onto micro-contact printed substrates at densities of 100,000/cm ² . Cells were cultured for 2 days on micro-contact printed substrates prior to experimentation. Neonatal mouse ventricular myocyte culture

Neonatal mouse ventricular myocytes were isolated from 2-day old neonatal Balb/c mice using procedures approved by the Harvard University Animal Care and Use Committee. Briefly, excised ventricular tissue was incubated in a 0.1% (w/v) trypsin (USB Corp., Cleveland, OH) solution cooled to 4°C for approximately 12 hours with agitation.

Trypsinized ventricular tissue was dissociated into cellular constituents via serial exposure to a 0.1% (w/v) solution of collagenase type II (Worthington Biochemical, Lakewood, NJ) at 37° C for 2 minutes. Isolated myocytes were maintained in a culture medium consisting of Medium 199 (Invitrogen, Carlsbad, CA) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS), 10 mM HEPES, 20 mM glucose, 2 mM L-glutamine, 1.5 vitamin B- 12, and 50 U/ml penicillin and seeded at a density of 200,000 cells/cm ² . From the second day of culture onward, the FBS concentration was reduced to 2% (v/v), and medium was exchanged every 48 hours. Myocytes were cultured for 4 days on micro-contact printed substrates prior to experimentation. Fabrication of micro-contact printed substrates

Silicone stamps designed for micro-contact printing were prepared. Photolithographic masks were designed in AutoCAD (Autodesk Inc., San Rafael, CA), and consisted of 20 μηι wide lines separated by 4 μηι gaps to impose a laminar organization on the myocytes.

Polydimethylsiloxane (PDMS, Sylgard 184, Dow Coming, Midland, MI) was used to fabricate stamps with the specified partem. Stamps were incubated with 50 μg/mL FN (BD Biosciences, Bedford, MA) for one hour. Glass coverslips were spin-coated with PDMS and treated in a UV-ozone cleaner (Jelight Company, Inc., Irvine, CA) immediately prior to stamping with FN. After transfer of the FN pattern to the surface of the PDMS-coated coverslips, they were incubated in 1% (w/v) Pluronic F127 (BASF, Ludwigshafen, Germany) to block cell adhesion to un-stamped regions.

"Heart-on-a-Chip" substrate fabrication

Engineered cardiac tissue contractile performance was measured using a custom muscular thin film based platform. Briefly, the "heart-on-a-chip" substrates consisted of glass coverslips selectively coated with a thermo-sensitive sacrificial polymer, Poly(N- isopropylacrylamide) (PIPAAm, Poly sciences, Inc., Warrington, PA), and with a second layer of PDMS. The thickness of the PDMS layer was found to be in the range of 10-18μιη for all "heart chips" used in this study (Dektak 6M, Veeco Instruments Inc., Plainview, NY).

"Heart-on-a-Chip" contractility experiments

During contractility experiments, samples were submerged in Tyrode's solution (mM, 5.0 HEPES, 5.0 glucose, 1.8 CaCl ₂ , 1.0 MgCl ₂ , 5.4 KC1, 135.0 NaCl, and 0.33 NaH ₂ PO ₄ . All reagents were purchased from Sigma Aldrich, St. Louis, MO). Rectangular films were cut out with a razor blade, and the bath temperature was decreased below the PiPAAm transition temperature, making possible for the MTF to bend away from the glass. Video recording of the deformation of each film were processed to obtain the time-course (Alford et al, 2010, Biomaterials 31 :3613-3621) of the tissue-generated stresses. The peak systolic and diastolic stresses were calculated as the average of the maxima and minima of the stress profile during 10 cycles at a pacing of 3 Hz, and twitch stress was defined as the difference between peak systolic and diastolic stresses.

Immunohistochemical labeling Samples were fixed in 4% (v/v) paraformaldehyde with 0.05% (v/v) Triton X-100 in PBS at room temperature for 10 minutes. Cells were incubated in a solution containing 1 :200 dilutions of monoclonal anti-sarcomeric a-actinin antibody (A7811, clone EA-53, Sigma Aldrich, St. Louis, MO), polyclonal anti-fibronectin antibody (F3648, Sigma-Aldrich, St. Louis, MO), 4',6'-diamidino-2-phenylindole hydrochloride (DAPI, Invitrogen, Carlsbad, CA), and Alexa Fluor 633-conjugated phalloidin (Invitrogen, Carlsbad, CA) for one hour at room temperature. Samples were then incubated in 1 :200 dilutions of Alexa Fluor 488-conjugated goat anti-mouse IgG and Alexa Fluor 546-conjugated goat anti-rabbit IgG secondary antibodies (Invitrogen, Carlsbad, CA) for 1 hour at room temperature. Labeled samples were imaged with a Zeiss LSM confocal microscope (Carl Zeiss Microscopy, Jena, Germany).

Evaluation of sarcomere structure

Analysis of sarcomeric structural characteristics of striated (cardiac and skeletal) mycoytes was conducted, after de-convolving acquired confocal Z-stacks of sarcomeric a- actinin fluorescence micrographs with Mediacy Autoquant (MediaCybernetics, Rockville, MD), on custom-designed ImageJ (NIH) and MATLAB (Mathworks, Natick, MA) software. Fluorescence micrographs were first pre-processed to highlight the filamentous structure of the cytoskeleton using a "tubeness" operator. This operator replaced each pixel in the image with the largest non-positive eigenvalue of the image Hessian matrix. The orientations of sarcomeric a-actinin positive pixels were determined using an adapted structure-tensor method and the orientational order parameter (OOP), a measure of the global alignment of the sarcomeres, was calculated from the observed orientations. The orientations observed in the micrographs were color-coded using the HSV digital image representation (Figure 7B(i)) where the Hue channel was used for orientation, the Saturation channel for pixel coherency (i.e. a measure of local contrast), and the Value channel for the pre-processed image. The normalized occurrence of the orientations that demonstrated a coherency higher than a given threshold (sub-threshold pixels were not color-coded) could then be displayed in a histogram (Figure 7B(ii)). Two components could be easily distinguished: blue-green coloration in (Figure 7B(i)) corresponded to pixels localized to Z-disks (black curve in Figure 7B(ii)), while red-yellow pixels were associated with long stretches of Z-bodies (red curve in Figure 7B(ii)). The sarcomere length and the overall regularity of the z-lines was determined by processing the fluorescence images with a 2D Fast Fourier Transform algorithm (the power spectrum of the image in Figure 7B(i) is reported in Figure 7C(i) with a gamma correction of 0.1 to improve visualization). To further analyze the Fourier representation without introducing user-bias, the power spectrum was then radially (cardiac myocytes) or longitudinally (skeletal myocytes) integrated and normalized by the total area under the ID curve. The previous step yielded a ID profile (blue curve in Figure 7C(ii)) that could be fitted with aperiodic (3, black line in Figure 7C(ii)) and periodic (red line in Figure 7C(ii)) components. The parameters {a, b, c] in (1) characterize the decaying exponential chosen to model the effect of noise and non-regularly distributed structures in the image, while the parameters {ω ₀ , a _k , 5 ] in (2) represent respectively, the wavenumber that corresponds to the sarcomere length, the amplitude and the width of the Gaussian peaks chosen to model the periodic peaks. The sarcomere packing density was defined as the area under the periodic component (shaded red in Figure 7C(ii)).

Γ _αρ (ω, 7 _αρ ) = a + be ^~COi ; γ _αρ = {a, b, c} (1)

( ω-]ιω ₀ \ ²

Γ _ρ (ω, 7 _ρ ) =∑l ₌₁ a _k e ^\ ¾ > ; γ _αρ = {ω ₀ , ¾, ¾} (2)

To optimize the fitting procedure, wo of approximately 2 micron and wo of approximately 3 micron was utilized as starting points for cardiac and skeletal myocytes, respectively. These numbers correspond to traditionally measured values for sarcomere lengths in cardiac and skeletal muscle tissues in vivo

Planar patch clamp electrophysiological recordings

Planar patch clamp experiments were conducted as previously described. Briefly, cells were cultured on fibronectin (BD Biosciences, Bedford, MA) coated T25 flasks for 5 days, then isolated using .25% trypsin (Invitrogen, Carlsbad, CA), re-suspended in

Extracellular Buffer Solution (EBS: mM, 140 NaCl, 4 KC1, 1 MgCl ₂ , 2 CaCl ₂ , 5 D-Glucose monohydrate, 10 Hepes, pH 7.4) to a final concentration of 1,000 cells / μί, and allowed to equilibrate for 5 minutes in EBS. The electronics were calibrated in the presence of EBS and Intracellular Buffer Solution (IBS: mM, 50 KC1, 10 NaCl, 60 K-Fluoride, 20 EGTA, 10 Hepes, pH 7.2) prior to flowing cells into the chamber. 5 of cell suspension was then introduced into the chip and the negative pressure automatically adjusted to produce a final seal resistance greater than 1 GOhm. During current clamp experiments, cells were subjected to 10 trains of 10 current pulses at 3 Hz; the current amplitude was set to 1.5 times the threshold for Action Potential (AP) generation. When the signal reached steady state, 10 APs were averaged yielding a representative trace for the calculation of action potential duration indicators. During voltage clamp experiments cells were kept in buffers containing TTX (10 μΜ), Nifedipine (10 μΜ), 4-AP (1 mM) and TEA (20 mM) purchased from Sigma Aldrich (St. Louis, MO). The membrane potential subjected to 2 voltage clamp protocols, first the membrane potential was held to a value of -90 V for 250 ms and then stepped from -70 to +40 mV in 10 mV steps for 250 ms, thus eliciting the total Ca ²⁺ current (TOT). Second, from the same holding potential, cells were stepped from -40 to +40 mV, a range in which mostly the L-type Ca ²⁺ current (LCC) is active. The T-type component (TCC) was then calculated as the difference between TOT and LCC. Optical mapping of electrophysiological properties

Samples were incubated in 4 μΜ RH237 (Invitrogen, Carlsbad, CA) for 5 minutes and washed 3 times with Tyrode's solution, prior to recording. Temperature of the bath solution was maintained at approximately 35° C using a digital temperature controller (TC- 344B, Warner Instruments, Hamden, CT) for the duration of the experiment. 10 μΜ

Blebbistatin (EMD Millipore, Billerica, MA) was added to minimize motion artifacts during recording of electrical activity. Samples were paced at 3 Hz with a 10 ms biphasic pulse at 10-15 V delivered using an SD-9 stimulator (Grass Technologies, Warwick, RI) and a bipolar, platinum point electrode placed approximately 300 - 500 μιτι above the sample and 1 - 2 mm from the top right comer of the field of view (FOV). Imaging was performed using a Zeiss Axiovert 200 epifluorescence microscope (Carl Zeiss Microscopy, Jena, Germany) equipped with an X-cite Exacte mercury arc lamp (Lumen Dynamics, Mississauga, Ontario). Illumination light was passed through a 40X/1.3 NA objective (EC Plan-NEOFLUAR, Zeiss, Jena, Germany) and a band-pass excitation filter (530-585 nm). Emission light was filtered at 615 nm with a long-pass filter, and focused onto the 100 x 100 pixel chip of a high speed MiCAM Ultima CMOS camera (Scimedia, Costa Mesa, CA). Images were acquired at 1000 frames per second from 250 x 250 μιτι fields of view. Post-processing of the raw data included reduction of drift induced by photobleaching by subtracting a linear fit of the baseline, applying a 3 x 3 pixel spatial filter to improve signal to noise ratio, and exclusion of saturated pixels. Activation time was calculated as the average maximum upstroke slope of multiple pulses over a 2 - 4 second recording window. Longitudinal and transverse conduction velocities (LCV and TCV) were calculated through a linear fit of the activation times along the horizontal and vertical axes of each FOV respectively. Optical action potential traces were calculated as the average of multiple pulses, while adjusting the offset of each pixel caused by different activation times. Ratiometric measurement of cardiomyocyte calcium transients

20 working aliquots of acetoxymethyl (AM) Fura Red (Invitrogen, F-3021) were obtained reconstituting 50 μg of the desiccated dye in 100 of Pluronic F-127 (20% solution in DMSO; Invitrogen, P-3000MP). Working aliquots were stored in the freezer and used within the week. Dye loading of myocytes was performed by exposing the cells for 20 minutes to a solution composed from a single working aliquot diluted in 2 mL of media. After dye loading, cells were kept in Tyrode's solution for 5 minutes, washed 3 times, and mounted on a coverslip holder for confocal imaging. Tissues were imaged using a Zeiss LSM LIVE (Carl Zeiss Microscopy, Jena, Germany) confocal microscope and a 40x objective equipped with an environmental chamber to ensure a constant physiological temperature in the bath of 37° C. Tissues were field stimulated at 3 Hz using the same equipment adopted in MTF experiments. Dual excitation ratiometric recordings were performed by rapidly switching (through an acousto-optical tunable filter) excitation laser lights at 405 nm and 488 nm and by collecting the corresponding emissions through a high-pass filter with cutoff at 546 nm. The 405 nm excitation offers an estimated 16% higher absorbance than what was recently reported for a 457 nm excitation light, while reducing the overlap between the Ca ²⁺ - bound and Ca ²⁺ -free excitation spectra. To maintain a high enough acquisition speed (250 fps), the recordings were constrained to 20 lines, oriented perpendicular to the main axis of the cells and ensuring minimal intersection with nuclei (white box Figure 8A). After background subtraction (performed in FIJI ⁴² ), two signals were obtained (Figure 8B): one (blue line) that increases with the Ca ²⁺ elevation corresponding to excitation at 405 nm, and one (green line) that shows an opposite trend and corresponded to the 488 nm excitation wavelength. The ratiometric representation of the calcium transient was taken as the ratio of the 405 nm and 488 nm signals (black trace in Figure 8C). Four consecutive transients at steady state were further averaged to create a representative single transient (Figure 8D) that was used to extract the following quantities: diastolic level (grey box), peak level (*), time to peak (T2P) and the duration of the Ca ²⁺ transient at 50% (CaT50) and 90% (CaT90) decay using Matlab (Mathworks, Natick, MA).

RT-qPCR gene expression measurements

Total RNA was collected in triplicate from both isotropic and micropatterned anisotropic samples using a Strategene Absolutely RNA Miniprep kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. Genomic DNA contamination was eliminated by incubating the RNA lysates in DNase I digestion buffer at 37°C for 15 minutes during the RNA purification procedure. The quantity and purity of RNA lysates was assessed using a Nanodrop spectrophotometer (Thermo Scientific, Wilmington, DE). Purified total RNA lysates with OD 260/280 ratios greater than 1.7 were used for RT-qPCR measurements. Complementary DNA strands were synthesized for genes of interest using an RT2 first strand synthesis kit (Qiagen Inc, Valencia, CA) and custom pre-amplification primer sets (Qiagen Inc, Valencia, CA). 500 ng of total RNA were used from each lysate for each first strand synthesis reaction. Expression levels for specific genes of interest (Table 3 and Table 4) were measured using custom RT2 Profiler RT-PCR arrays (Qiagen Inc, Valencia, CA) and a Bio-Rad CFX96 RT-PCR detection system (Hercules, CA). Statistical analysis of RT-qPCR threshold cycle data was carried out with the web-based RT2 Profiler PCR Array Data Analysis Suite version 3.5 (Qiagen Inc, Valencia, CA) according to published guidelines. Statistical analysis

All data are summarized as mean ±standard error of the mean. Data were first tested for normality (Shapiro-Wilk) and equal variance (Levene Median test). Based on the results from these tests, either 1-way ANOVA or ANOVA on Ranks were adopted to establish statistical difference between the groups. Pairwise comparisons were then assessed using either Dunn's or Tukey or Holm-Sidak methods as post-hoc tests. In the figures the significance of statistical tests (p-value) is indicated as follows: * = p < 0.05, ** = p < 0.001 for 1-way ANOVA and for† = p < 0.05,†† = p < 0.001 ANOVA on ranks.

The influence of tissue architecture on the contractile performance of engineered myocardium in vitro was previously reported (Feinberg et al, 2012, Biomaterials 33:5732-

5741). From this, characterization of the mES and miPS myocytes is made by evaluating their response to geometric cues encoded in the ECM, and measuring the expression of genes that are commonly used to delineate the cardiac myocyte lineage (Maltsev et al., 1994, Circ Res 75:233-244; Chin et al, 2009, Cell Stem Cell 5: 111-123; Sartiani et al, 2007, Stem Cells 25: 1136-1144). Culturing mES (Figure 1 A(i)) and miPS (Figure lA(ii)) myocytes on a substrate coated uniformly with fibronectin (FN) gave rise to monolayers with an isotropic cellular arrangement similar to the arrangement observed when neonate ventricular myocytes (Figure lA(iii)) were cultured in a similar manner. Moreover, mES (Figure 6A(i)), and miPS (Figure 6A(ii)) and neonate (Figure 6A(iii)) myocytes all assumed a pleomorphic morphology when cultured sparcely on isotropic FN (Figure 6B), even though the neonate cardiac myocytes displayed a smaller surface area than the mES and miPS myocytes (Figure 6C). Comparison of the expression profiles for isotropic mES (Figure lB(i)) and miPS (Figure lB(ii)) derived tissues versus the neonate tissues revealed a number of significant differences associated with ion channel subunits and components of the sarcomere. In particular, the mES tissues exhibited significantly higher expression of the L-type Ca ²⁺ channel subunit Cacnald (4.9 fold, p<0.05), as well as the T-type subunits Cacnalg (9.0 fold, p<0.05) and Cacnalh (42.2 fold, p<0.05) versus neonate tissues. Isotropic mES tissues also showed significantly lower expression ofIrx4 (-9.1 fold, pO.001), Myl2 (-3.2 fold, p<0.05), mdMyl3 (-3.8 fold, p<0.01) commonly associated with the ventricular myocyte phenotype (Ng et al, 2010, Am J Physiol Cell Physiol 299:C1234-1249), and significantly higher expression of the atrial marker genes Myl4 (40.2 fold, pO.001), andM /7 (24.5 fold, p<0.01) than the neonate isotropic tissues. In contrast, the miPS isotropic tissues showed significant differences in expression for Cacnald (5.7 fold, p<0.05), Cacnalh (27.9 fold, p<0.001), (14.1 fold, p<0.05) and Myl7 (11.1, p<0.05) versus the neonate isotropic tissues. These observations suggest that the miPS-derived myocytes exhibited an expression profile that more closely resembled the profile of the neonate ventricular myocytes than the mES-derived myocytes.

Based on previous studies, it was recognized that the gene expression profile of cardiac myocytes changed as a function of the tissue architecture within which they are embedded. Laminar, anisotropic myocardium was engineered from mES (Figure lC(i)), miPS (Figure lC(ii)), and neonate cardiac myocytes by culturing them on micro-contact printed FN, where the cells spontaneously formed cell-cell junctions and aligned with the geometric cues within the matrix to form a contiguous tissue of high aspect ratio cells (Figure 6D and Figure 6E). After several days in culture, the expression profiles of these engineered tissues were measured and compared. Comparison of the expression profiles for anisotropic neonate and mES tissues (Figure lD(i)) revealed a number of differences associated with Ca ²⁺ channel subunits, such as the L-type Ca ²⁺ channel subunit Cacnald (37.5 fold, p<0.0001), as well as the T-type subunits Cacnalg (20.2 fold, p<0.05), and Cacnalh (23.8 fold, p<0.05). Additionally, the mES anisotropic tissues showed significantly lower expression of the ventricular marker Irx4 (-7.7 fold, p<0.05), and significantly higher expression of the atrial markers Myl4 (254.8 fold, pO.01), wA MyU (104.0 fold, pO.01) versus the neonate tissues. In contrast, the miPS anisotropic tissues exhibited significant differences from the neonate tissues (Figure lD(ii)) for the Ca ²⁺ channel subunits Cacnald (36.9 fold, p<0.05) and Cacnalg (6.6 fold, p<0.05), as well as the atrial myosin light chain kinase gene M /4 (105.5 fold, p<0.01). Hierarchical clustering of neonate, mES, and miPS gene expression measurements revealed a distinct separation of the expression profiles for isotropic and anisotropic tissues, regardless of myocyte type (Figure IE). Moreover, the expression profiles for mES and miPS myocytes in both the isotropic and anisotropic cellular configurations clustered closer to each other than to the neonate tissues, suggesting that the mES and miPS myocytes exhibited global transcriptional profiles that were unique from the neonate expression pattern, despite differences in the relative expression profiles between the mES and miPS tissues.

One of the defining features of the native myocardium is the laminar arrangement of cardiac myocytes that serves to organize and orient the contractile sarcomeres to facilitate efficient pump function (Bruneau, 2002, Circ Res 90:509-519). The ability of mES and miPS engineered tissues to self-assemble myofibrils with alignment comparable to neonate ventricular myocytes were evaluated using image analysis software of the present invention. Immunofluorescence micrographs of sarcomeric a-actinin allowed for visualization of the orientations of the z-lines outlining the lateral edges of sarcomeres and to quantitatively assess sarcomere organization in the engineered tissues. Visualization of global z-line registration in isotropic monolayers of mES (Figure 2A(i)), miPS (Figure 2A(ii)) and neonate (Figure 2A(iii)) myocytes revealed random orientation patterns. In contrast, the anisotropic mES (Figure 2B(i)), miPS (Figure 2B(ii)), and neonate (Figure 2B(iii)) tissues demonstrated a greater degree of uniaxial z-line registration. To quantify the differences in global sarcomere organization between the mES, and miPS tissues, versus the neonate tissues (Figure 2C), a metric known as the orientational order parameter (OOP) was utilized, which is commonly used to characterize the alignment of liquid crystals, and ranges from zero (random organization) to one (perfect alignment). The anisotropic neonate tissues exhibited a significantly higher OOP value than both the mES and miPS tissues, suggesting that both types of stem cell-derived myocytes were unable to generate myofibrils with the same degree of global sarcomere alignment as the neonate myocytes. Isotropic tissues had low OOP values, due to the random organization of the cardiac myocytes. Measurement of registered z- line spacing also revealed that the anisotropic mES and miPS tissues displayed significantly shorter sarcomere lengths than the neonate tissues (Figure 2D). Moreover, quantification of "sarcomere packing density," i.e. the proportion of a-actinin localized to z-lines indicative of the presence of fully-formed sarcomeres, showed that the anisotropic neonate tissues exhibited significantly higher sarcomere packing density than the mES and miPS tissues. Taken together, these analyses revealed that the mES- and miPS-derived myocytes responded to ECM cues in a similar manner to the neonate myocytes, but exhibited sarcomere organization reminiscent of immature pre-myofibrils observed in embryonic cardiac myocytes (Dabiri et al, 1997, Proc Natl Acad Sci USA 94:9493-9498; LoRusso et al, 1997, Cell Motil Cytoskeleton 37: 183-198).

The electrical activity of cardiac myocytes regulates the initiation of myofibril contraction and is commonly measured as an indicator of myocyte identity and functionality (Kleber and Rudy, 2004, Physiol Rev 84:431-488; Maltsev et al, 1994, Circ Res 75:233-244; Weinberg et al, 2010, Methods Mol Biol. 660:215-237). Planar patch clamp recordings were used to compare and contrast the action potential characteristics of isolated mES, miPS and neonate myocytes. Two different demographics of cell types were identified, demonstrated by action potential morphology (AP). Most neonate myocytes mostly demonstrated ventricular-like APs (Figure 3A(i)) whereas mES- and miPS-derived myocytes exhibited APs that were evenly distributed between the ventricular-like (Figure 3A(i)) and satrial-like (Figure 3A(ii)) morphologies. Both the mES- and miPS-derived myocytes primarily exhibited APs as shown in Figure 3A(ii), whereas the neonate ventricular myocytes demonstrated APs illustrated in Figure 3A(i). Analysis of AP characteristics, such as maximum voltage (Vmax), action potential duration at 50% repolarization (APD50), and action potential duration at 90% repolarization (APD90), revealed that the mES and miPS myocytes exhibited roughly equal incidences of atrial-like and ventricular-like APs, whereas the neonate cardiac myocytes displayed ventricular-like AP characteristics (Figure 3B). In addition to AP characterization, the electrical conduction properties of the anisotropic mES (Figure 3C(i)), miPS (Figure 3C(ii)), and neonate (Figure 3C(iii)) tissues were measured using optical mapping and the voltage-sensitive fluorescent dye RH-237 (Weinberg et al, 2010, Methods Mol Biol. 660:215-237; Bursac et al, 2002, Circ Res 91:e45-54; Thomas et al, 2000, Circ Res 87:467-473) to evaluate the ability of the stem cell-derived myocytes to form the electromechanical syncytium that typifies the myocardium (Kleber and Rudy, 2004, Physiol Rev 84:431-488). No significant differences in the longitudinal (LCV) or transverse (TCV) conduction velocities were observed between the mES, miPS and neonate tissues (Figure 3D). AP duration measurements revealed no significant differences at 50% repolarization (APD50), but a significant (p<0.05) difference was observed at 90% repolarization (APD90) between the neonate and mES anisotropic tissues (Figure 3E). Ca ²⁺ plays a crucial role in coupling myocyte excitation and contractile activity (Bers, 2002, Nature 415: 198-205), therefore, the Ca ²⁺ transient activity in engineered anisotropic tissues, as well as the Ca current profiles of isolated mES, miPS and neonate myocytes were measured. Ca ²⁺ transients measured in anisotropic tissues revealed a significantly (p<0.05) shorter 50% decay time (CaT50) in the miPS, but not the mES tissues, as compared to the neonate, and significantly (p<0.05) shorter 90% decay time (CaT90) in both the mES and miPS tissues versus the neonate tissues (Figure 3F). Planar patch clamp recordings of L- (Figure 3G(i)) and T- (Figure 3G(ii)) type Ca ²⁺ current profiles revealed significantly (p<0.05) higher total (TOT) and T-type (TCC) maximum Ca ²⁺ current densities in the neonate myocytes versus the mES-derived, but not the miPS-derived myocytes (Figure 3H). Taken together, these data suggest that the mES and miPS myocytes possessed

electrophysiological properties similar to neonate cardiac myocytes, aside from differences in funny current and voltage-gated Ca ²⁺ channel subunit expression illustrated in Figure 1.

With the muscular thin film (MTF) contractility assay, it is now possible to assess the diastolic (Figure 4A(i)) and systolic (Figure 4A(ii)) function of engineered myocardium directly (Alford et al., 2010, Biomaterials 31 :3613-3621 ; Grosberg et al, 2011, Lab Chip 11 :4165-4173; Feinberg et al, 2007, Science 317: 1366-1370). Using the "heart-on-a-chip" MTF assay, the stress generation profiles of the anisotropic mES, miPS and neonate tissues were measured (Figure 4B), and their contractile performance compared. The anisotropic neonate tissues generated significantly (p<0.05) higher diastolic, peak systolic, and twitch stress than both the mES and miPS tissues (Figure 4C), with observed values for the neonate tissues within the range measured for isolated murine papillary muscle strips (Stuyvers et al, 2002, J Physiol 544:817-830; Kentish et al, 2001, Circ Res 88: 1059-1065; Gao et al, 1998, J Physiol 507(Pt 1): 175-184). The results of the contractility measurements clearly show a functional deficit in the mES- and miPS-derived myocytes that was not apparent in the electrophysiological measurements. The combined output of the electrophysiological, calcium transient and contractile force experimental measurements were used to create graphical representations of the excitation-contraction coupling profiles of the mES (Figure 4D(i)), miPS (Figure 4D(ii)), and neonate (Figure 4D(iii)) engineered tissues that clearly illustrate the similarities and differences in the excitation-contraction coupling amongst the cell types. These data illustrate that the miPS-derived myocytes are qualitatively more similar to the neonate myocytes than the mES-derived myocytes.

To determine how closely the mES- and miPS-derived myocytes matched the phenotype of the neonate ventricular myocytes, a novel numerical method was developed to integrate the set of gene expression, morphology, electrophysiology, and contractility experimental measurements collected on each cell population, and calculate the difference between the unknown and target cell populations. For each experimental measurement, the values were normalized to the interval [0,1] and calculated the strictly standardized mean difference (β) (Zhang, 2007, Genomics 89:552-561) between each unknown population (i.e. mES, miPS) and the neonate target population as follows: where μ represents mean and σ represents standard deviation, to evaluate the magnitude of difference, taking into account the variance in the measurements, between the stem cell-derived myocytes and the neonate cardiac myocytes (Figure 5). This allowed for determination of the effect size for each experimental measurement when comparing the mES and miPS to the neonate tissues, and to identify the parameters that show the greatest degree of similarity and difference from the target neonate ventricular myocyte tissues.

The β values were then used from each experimental measurement for the mES and miPS tissues and the mean squared error (MSE) versus the neonate tissues was calculated as follows:

MS£ = ±∑? ₌₁ ft ² (4) where n is the total number of experimental measurement β values included in the calculation, to evaluate the differences observed for each measurement category (i.e. the β values for gene expression, morphology, electrical activity, contractility used to calculate category-specific MSE values), as well as define a single MSE value calculated from all of the experimental measurements from all categories combined, that represents the total difference between the stem cell-derived and neonate cardiac myocytes based on the measurements performed (Table 1). The strictly standardized mean difference (β) values computed for each experimental measurement were used to calculate mean squared error (MSE) values for each of the major measurement categories, as well as all of the

measurements combined, in the comparisons of the mES (MSE^s), and miPS (MSE _m jps) engineered tissues to the neonate engineered tissues. Table 1 - mean squared error values calculated for each group of measurements in the comparison of the mES- and miPS-derived myocytes to the neonate ventricular myocytes.

Measurement Category MSE _mES MSE _mi p _S

Gene Expression 5.69 4.25

Morphology 1.30 1.48

Electrophysiology 1.16 0.57

Contractility 6.32 2.95

All Measurements 4.95 3.60

A lower MSE value indicates a better match to the neonate target phenotype, with an MSE value of zero indicating a perfect match.

It was found that the miPS tissues exhibited lower MSE values than the mES tissues for every measurement category, except morphology. In addition, the overall MSE values calculated from all of the experimental measurements combined revealed a lower MSE for the miPS engineered tissues than those comprised of mES-derived myocytes. This suggests that the miPS-derived myocytes exhibited a global phenotype that was slightly closer to the neonate cardiac myocytes than the mES-derived myocytes, although both the mES- and miPS-derived myocytes demonstrated substantial differences from the neonate cardiac myocytes for a number of characteristics.

Descriptions for each abbreviation listed in the right-hand column of Figure 5 can be found in Table 2. Provided in Table 4 are descriptions of the experimental measurements used for the above calculation, the means and standard deviations for these measurements used in the calculation, and each step in the process of arriving at the β values identified in Figure 5. The descriptions of the measurements and genes listed in Table 4 are provided in Table 2. The steps of the calculation proceed along the columns from left to right, where the last two columns contain the final β values for each experimental parameter. At the bottom of last two columns of Table 4 is the MSE calculation, representing the quality index "score" for each of the two stem cell-derived myocyte cell lines that were tested.

Accordingly, a quality control standard rubric for assessing stem cell-derived cardiac myocytes is shown. Using the experimental measurements described above and isolated neonatal ventricular myocytes as the reference phenotype, a "quality index" was developed that utilizes the magnitude and variance of these measurements to provide a numeric "score" of how closely the stem cell-derived myocytes match the characteristics of the neonatal cardiac myocytes. The combination of gene expression, morphological evaluation, electrophysiological, and contractility measurements employed allow a user of the system and method of the present invention to pin-point specific differences in the structural and functional properties of the mES and miPS engineered tissues versus the neonate tissues that have important implications for their utility in in vitro assays. Further, this "quality index" not only allows researchers to identify the commercial stem cell-derived myocyte product lines that are most suitable for their needs, it serves the stem cell industry as a quality assurance system for ensuring that batches released to customers faithfully recapitulate the desired phenotype.

Table 2. List of major experimental measurement categories.

gene expression Tnnt2 Troponin T2, cardiac

gene expression Ttn Titin

gene expression Myh6 Myosin, heavy polypeptide 6, cardiac muscle, alpha gene expression Myh7 Myosin, heavy polypeptide 7, cardiac muscle, beta

gene expression Myl2 Myosin, light polypeptide 2, regulatory, cardiac, slow gene expression My Myosin, light polypeptide 3

gene expression Myl4 Myosin, light polypeptide 4

gene expression Myl7 Myosin, light polypeptide 7, regulatory

gene expression Cacnalc Calcium channel, voltage-dependent, L type, alpha 1C subunit gene expression Cacnald Calcium channel, voltage -dependent, L type, alpha ID subunit gene expression Cacnalg Calcium channel, voltage -dependent, T type, alpha 1G subunit gene expression Cacnalh Calcium channel, voltage -dependent, T type, alpha 1H subunit gene expression Kcnel Potassium voltage-gated channel, Isk-related subfamily, member 1

gene expression Kcne2 Potassium voltage-gated channel, Isk-related subfamily, gene 2 gene expression Kcnd2 Potassium voltage-gated channel, Shal-related family, member

2

gene expression Kcnd3 Potassium voltage-gated channel, Shal-related family, member

3

gene expression Kcnh2 Potassium voltage-gated channel, subfamily H (eag-related), member 2

gene expression Kcnj2 Potassium inwardly -rectifying channel, subfamily J, member 2 gene expression Kcnj3 Potassium inwardly -rectifying channel, subfamily J, member 3 gene expression Ken) 11 Potassium inwardly rectifying channel, subfamily J, member

11

gene expression Kcnjl2 Potassium inwardly -rectifying channel, subfamily J, member

12

gene expression Kcnjl4 Potassium inwardly -rectifying channel, subfamily J, member

14

gene expression Kcnql Potassium voltage-gated channel, subfamily Q, member 1 gene expression Scn5a Sodium channel, voltage -gated, type V, alpha

gene expression Slc2al Solute carrier family 2 (facilitated glucose transporter), member 1

gene expression Slc2a2 Solute carrier family 2 (facilitated glucose transporter), member 2

gene expression Slc8al Solute carrier family 8 (sodium/calcium exchanger), member 1 gene expression Hcnl Hypeφolarization-activated, cyclic nucleotide -gated K+ 1 gene expression Hcn3 Hypeφolarization-activated, cyclic nucleotide -gated K+ 3 gene expression Hcn4 Hypeφolarization-activated, cyclic nucleotide -gated K+ 4 gene expression Gjal Gap junction protein, alpha 1

gene expression Gja5 Gap junction protein, alpha 5

gene expression Atpla2 ATPase, Ca++ transporting, cardiac muscle, fast twitch 1 gene expression Atp2a2 ATPase, Ca++ transporting, cardiac muscle, slow twitch 2 gene expression Ryr2 Ryanodine receptor 2, cardiac

gene expression Ckm Creatine kinase, muscle

Table 3: Custom RT-qPCR array gene list

Gene Symbol | Refseq # Gene Description Hey 2 NM 013904 Hairy/enhancer-of-split related with YRPW motif 2

Irx4 NM 018885 Iroquois related homeobox 4 (Drosophila)

BmplO NM 009756 Bone morpho genetic protein 10

Gata4 NM 008092 GATA binding protein 4

Myocd NM 145136 Myocardin

Nkx2-5 NM 008700 NK2 transcription factor related, locus 5 (Drosophila)

TbxS NM 011537 T-box 5

Nppa NM 008725 Natriuretic peptide type A

Actal NM 009606 Actin, alpha 1, skeletal muscle

Adralb NM 007416 Adrenergic receptor, alpha lb

Adra2a NM 007417 Adrenergic receptor, alpha 2a

Actcl NM 009608 Actin, alpha, cardiac muscle 1

Actnl NM 134156 Actinin, alpha 1

Actn2 NM 033268 Actinin alpha 2

Pin NM 023129 Phospholamban

Tnnt2 NM 011619 Troponin T2, cardiac

Tin NM 011652 Titin

Myh6 NM 010856 Myosin, heavy polypeptide 6, cardiac muscle, alpha

Myh 7 NM 080728 Myosin, heavy polypeptide 7, cardiac muscle, beta

Myl2 NM 010861 Myosin, light polypeptide 2, regulatory, cardiac, slow

Myl3 NM 010859 Myosin, light polypeptide 3

My NM 010858 Myosin, light polypeptide 4

Myl7 NM 022879 Myosin, light polypeptide 7, regulatory

Cacnalc NM 009781 Calcium channel, voltage-dependent, L type, alpha 1C subunit

Cacnald NM 028981 Calcium channel, voltage-dependent, L type, alpha ID subunit

Cacnalg NM 009783 Calcium channel, voltage-dependent, T type, alpha 1G subunit

Cacnalh NM 021415 Calcium channel, voltage-dependent, T type, alpha 1H subunit

KcnaS NM 145983 Potassium voltage-gated channel, shaker-related subfamily, member

5

Kcnel NM 008424 Potassium voltage-gated channel, Isk-related subfamily, member 1

Kcne2 NM 134110 Potassium voltage-gated channel, Isk-related subfamily, gene 2

Kcnd2 NM 019697 Potassium voltage-gated channel, Shal-related family, member 2

KcncB NM 019931 Potassium voltage-gated channel, Shal-related family, member 3

Kcnh2 NM 013569 Potassium voltage-gated channel, subfamily H (eag-related), member 2

Kcnj2 NM 008425 Potassium inwardly -rectifying channel, subfamily J, member 2

Kcnj3 NM 008426 Potassium inwardly -rectifying channel, subfamily J, member 3

Kcnjll NM 010602 Potassium inwardly rectifying channel, subfamily J, member 11

Kcnjl2 NM 010603 Potassium inwardly -rectifying channel, subfamily J, member 12

Kcnjl4 NM 145963 Potassium inwardly -rectifying channel, subfamily J, member 14

Kcnql NM 008434 Potassium voltage-gated channel, subfamily Q, member 1

Scn5a NM 021544 Sodium channel, voltage -gated, type V, alpha

Slc2al NM O 11400 Solute carrier family 2 (facilitated glucose transporter), member 1

Slc2a2 NM 031197 Solute carrier family 2 (facilitated glucose transporter), member 2

Slc8al NM 011406 Solute carrier family 8 (sodium/calcium exchanger), member 1

Hcnl NM 010408 Hypeφolarization-activated, cyclic nucleotide-gated K+ 1

Hcn3 NM 008227 Hypeφolarization-activated, cyclic nucleotide-gated K+ 3

Hcn4 NM 001081192 Hypeφolarization-activated, cyclic nucleotide-gated K+ 4

Gjal NM 010288 Gap junction protein, alpha 1

Gja5 NM 008121 Gap junction protein, alpha 5

Atpla2 NM 178405 ATPase, Na+/ + transporting, alpha 2 polypeptide

Atpla3 NM 144921 ATPase, Na+/ + transporting, alpha 3 polypeptide

Atp2al NM 007504 ATPase, Ca++ transporting, cardiac muscle, fast twitch 1 Atp2a2 NM 009722 ATPase, Ca++ transporting, cardiac muscle, slow twitch 2

Ryr2 NM 023868 Ryanodine receptor 2, cardiac

Ckm NM 007710 Creatine kinase, muscle

Acs 15 NM 027976 Acyl-Co A synthetase long-chain family member 5

Ptk2 NM 007982 PT 2 protein tyrosine kinase 2

Ilk NM 010562 Integrin linked kinase

ctgf NM 010217 Connective tissue growth factor

Itgal NM 001033228 Integrin alpha 1

Itga2 NM 008396 Integrin alpha 2

Itga4 NM 010576 Integrin alpha 4

Itga5 NM 010577 Integrin alpha 5 (fibronectin receptor alpha)

Itgav NM 008402 Integrin alpha V

Itgbl NM 010578 Integrin beta 1 (fibronectin receptor beta)

Itgb3 NM 016780 Integrin beta 3

Abra NM 175456 Actin-binding Rho activating protein

Rhoa NM 016802 Ras homolog gene family, member A

Cdc42 NM 009861 Cell division cycle 42 homolog (S. cerevisiae)

Racl NM 009007 RAS-related C3 botulinum substrate 1

Rockl NM 009071 Rho-associated coiled-coil containing protein kinase 1

Rock2 NM 009072 Rho-associated coiled-coil containing protein kinase 2

Rndl NM 172612 Rho family GTPase 1

Vcl NM 009502 Vinculin

Ctnnbl NM 007614 Catenin (cadherin associated protein), beta 1

Aifml NM 012019 Apoptosis-inducing factor, mitochondrion-associated 1

AtpSj NM 016755 ATP synthase, H+ transporting, mitochondrial F0 complex, subumt

F

Hsp90abl NM 008302 Heat shock protein 90 alpha (cytosolic), class B member 1

Hspa2 NM 008301 Heat shock protein 2

Hsphl NM 013559 Heat shock 105kDa/110kDa protein 1

Bcatl NM 007532 Branched chain aminotransferase 1, cytosolic

Ch25h NM 009890 Cholesterol 25-hydroxylase

Itpr2 NM 019923 Inositol 1,4,5-triphosphate receptor 2

Τ 2 NM 009367 Transforming growth factor, beta 2

Notchl NM 008714 Notch gene homolog 1 (Drosophila)

PouSfl NM 013633 POU domain, class 5, transcription factor 1

Nanog NM 028016 Nanog homeobox

Sox2 NM O 11443 SRY-box containing gene 2

Gapdh NM 008084 Glyceraldehyde-3 -phosphate dehydrogenase

Actb NM 007393 Actin, beta

Example 2: Metrics of Cvtoskeletal Organization to Identify the Structural Phenotvpes of Stem Cell Derived Cardiomyocytes

As demonstrated in herein, human induced pluripotent stem cell derived myocytes exhibited qualitatively and quantitatively underdeveloped contractile cytoskeletons with respect to murine primary and stem cell derived cardiomyocytes when exposed to in-vivo like experimental conditions. This is consistent with the notion that human stem cell derived cardiomyocytes may require longer time in culture or ad-hoc conditioning to fully mature, and suggests that metrics of cytoskeleton architecture can be utilized to quantitatively monitor this process. Accordingly, in addition to the metric parameters described in Example 1, a new metric of cytoskeletal organization, the sarcomere packing density, has been developed to further distinguish architectural phenotypes in establishment of the quality index used in the system and method of the present invention.

Demonstrated herein is a novel metric, the sarcomere packing density, that quantifies the presence of fully formed sarcomeres and provides an estimate for the maturity of the contractile cytoskeleton. The question was asked whether this metric could be utilized to perform structural phenotyping of stem cell derived cardiomyocytes. To answer this question immunocytochemistry analysis was performed of the cell cytoskeleton in primary (neonate mouse) and commercially available human and murine induced pluripotent stem cell derived cardiomyocytes cultured on engineered substrates that recapitulate the chemo-mechanical properties of the native microenvironment (McCain et al., 2012, Proc Natl Acad Sci USA

109:9881-9886). The experiments of Example 2 revealed that the sarcomere packing density numerically quantifies the inability of human induced pluripotent stem cell derived cardiomyocytes to assemble the kind of contractile cytoskeleton observed in murine primary and stem cell derived cardiomyocytes under the same experimental conditions.

The following materials and methods were used in Example 2. In brief, cell suspensions of primary cardiomyocytes (pCMs) were directly obtained from primary neonate mouse harvest while cultures of human (iCells from Cellular Dynamics International, Madison, WI) and murine (CorAt from Axiogenesis, Cologne, Germany) induced pluripotent stem cell derived cardiomyocytes (respectively hiCMs and miCMs) were obtained following the manufacturers' guidance.

All cell types were seeded on polyacrylamide gels engineered (McCain et al., 2012, Proc Natl Acad Sci USA 109:9881-9886) to a nominal substrate stiffness of 13 kPa and decorated with micro-contact printed fibronectin islands (BD Biosciences, Bedford, MA). Cells were cultured on the substrates with regular media exchanges for 72 hour and subsequently fixed and stained with primary antibodies: Alexa633-phalloidin (A22284 Invitrogen), DAPI (D3571 Invitrogen), anti-mouse sarcomeric a-actinin (A881 Sigma) and anti-human fibronectin (F3648 Sigma); and secondary antibodies: GAM-alexa546 (A21143 Invitrogen) and GAR-alexa488 (A11008 Invitrogen). Mono-nucleated, fully spread single cells were imaged with a confocal line scanning microscope (Zeiss LSM510 live).

Micrographs were preprocessed in FIJI (Schindelin et al, 2012, Nature Methods 9:676-682) to detect filamentous cytoskeletal structures (Sato et al., 1998, Medical Image Analysis 2: 143-168) and their orientations (Rezakhaniha et al., 2011, Biomech Model Mechanobiol 11 :461-473). Finally, Matlab (Mathworks, Natick, MA) circular statistics (Berens, 2009, Journal of Statistical Software 31 : 1-21) and image processing toolboxes were used to extract the quantitative metrics.

Micro-contact printing

Traditional photolithographic techniques were utilized to prepare

polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning) stamps. In particular, masks bearing the desired square (50x50 um) features were designed in AUTOCAD (Autodesk, Inc.) and fabricated at the Harvard University Center for Nanoscale Systems (CNS, NNIN, Cambridge, MA). Using a mask-aligner (ABM Inc.) UV-light was shine through the custom- made mask into a silicon wafer (Wafer World) that had been spin-coated with SU-8 3005 photoresist (MicroChem Corp). The wafer was then developed in propylene glycol methyl ether acetate and utilized to cast PDMS stamps.

Cell Culture Substrates

Polyacrylamide gels were engineered as previously described (McCain et al, 2012, Proc Natl Acad Sci USA 109:9881-9886). In particular to obtain a substrate stiffness of 13 kPa, the concentrations of streptavidin-acrylamide/bis were adjusted to a ratio of 7.5/0.3%. A 30 uL drop of polyacrylamide solution was added to a 25mm activated coverslip and temporarily sandwiched with a 18mm non-activated one. To transfer fibronectin islands, the thin hydrogel film was left to dry at 37°C for 10 mins, sterilized with a UV-ozone cleaner (Jelight Company, Inc.) and then micro-contact printed using fibronectin cross-linked with biotin via Sulfo-NHS-LC-Biotin (Pierce).

Primary Harvest Ventricular myocytes were isolated from day 2 neonate Balb/c mice according to procedures approved by the Harvard University Animal Care and Use Committee. In brief, animals were sacrificed and ventricles removed and incubated in cold (4°C) 0.1% (w/v) trypsin (USB Corp., Cleveland, OH) solution for approximately 12 hours. Ventricular tissue was further exposed to serial treatments (2 minutes each) of 0.1% (w/v) warm (37° C) collagenase type II (Worthington Biochemical, Lakewood, NJ) solution. Isolated neonate ventricular cardiac myocytes were seeded onto the engineered substrates at a density of 20,000 cells/cm ² and maintained in culture medium consisting of Medium 199 (Invitrogen, Carlsbad, CA) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS), 10 mM HEPES, 20 mM glucose, 2 mM L-glutamine, 1.5 μΐ. vitamin B-12, and 50 U/ml penicillin for the first 48 hours. After that, FBS concentration was switched to 2%.

Stem cell culture

Human and murine induced pluripotent stem cells derived cardiomyocytes (hiCM and miCMs) were kindly provided by Cellular Dynamics Inc. (CDI, Madison, WI) and

Axiogenesis (CorAt-iPS, Cologne, Germany). Cells were cultured in accord with manufacturers' recommendations. In particular, while hiCMs were seeded in 6-well plates in the presence of vendor-provided plating media, miCMs were enriched in T-25 flasks pre- coated with 10 mg/ml fibronectin (FN) (BD Biosciences, Bedford, MA) in the presence of manufacturer provided selection medium containing puromycin. After 72 hours, both cell types were dissociated with 0.25% trypsin-EDTA solution (Invitrogen, 25200-072) and re- seeded onto the engineered substrates at a density of 10,000 cells/cm ² .

Image Preprocessing

Preprocessing steps were performed using the ImageJ-based FIJI platform (Schindelin et al, 2012, Nature Methods 9:676-682). In particular, the following plugins were utilized: i) the tubeness plugin was used to highlight the filamentous structure of sarcomeric a-actinin positive pixels (Sato et al, 1998, Medical Image Analysis 2: 143-168); ii) the OrientationJ plugin (Rezakhaniha et al, 2011, Biomech Model Mechanobiol 11 :461-473) was used to calculate the orientations of each sarcomeric a-actinin positive pixels.

The sarcomere packing density

Force generation in striated muscle is associated with the vectorial summation of the contributions from all force generating units (Parker and Ingber, 2007, Philos Trans R Soc Lond B Biol Sci 362: 1267-1279) known as sarcomeres. Sarcomeres are ~2 μπι long linear assemblies of cytoskeletal proteins whose concerted action generate a quantum of force parallel to the orientation of the sarcomere (McCain and Parker, 2011, Pflugers Arch 462:89- 104). A common way to detect sarcomeres and their formation is via fluorescent

immunolabeling of sarcomeric α-actinin (red in Figure 9A(i)). This protein appears (Dabiri et al., 1997, Proc Natl Acad Sci USA 94:9493-9498; Parker et al, 2008, Circulation Research 103:340-342) to be diffuse in the cytosol during differentiation, then to assemble into puncta, known as Z- bodies (Sparrow and Schock, 2009, Nat Rev Mol Cell Biol 10:293-298), during the early phases of myofibrillogenesis and to localize to a regular lattice formed by Z-disks in mature myocytes (Figure 10). The distance between two Z-disks is the sarcomere length. Here, a novel quantitative metric of cytoskeleton organization is presented, the sarcomere packing density, whose value increases as more sarcomeric α-actinin positive pixels are localized in periodically spaced Z-disks.

To calculate the sarcomeric packing density, the Fourier transform of the pre- processed sarcomere α-actinin micrograph K(x, y) , was considered

and in particular its 2D power spectrum P(u, v) = \ F(u, v) | ² (where u and v are the coordinates of the Fourier domain and / ^' indicates the complex unit). Figure 9Aiii shows the power spectrum for the sarcomeric α-actinin micrograph in Figure 9Ai. In this representation, each pixel corresponds to a planar wave traveling across the spatial domain with frequency and orientation given by the pixel polar coordinate (ω, Θ) and power given by the pixel intensity P{u, v) . By radially integrating the expression for the signal energy (E, eq 6) a ID representation was obtained, Γ(ο) (blue traces in FigurelOD), that exhibited periodic peaks modulated by a monotonicall decreasing noise term.

To represent the periodic (f p , red curve in Figure 9D) and aperiodic (Γ up , black curve in Figure 9D) components, the following expression was considered ; γ) = Υ _ρ ω; γ _ρ ) + Υ _αρ (ω; γ _ρ )

By fitting the function Γ (co; to the data Γ(ί») the values were determined for the set of parameters γ = {a, b, a _k , b _k , co ₀ J . These parameters were utilized to determine the sarcomere length SL = co ^'1 and the sarcomere packing density (ε) ε = ^Υ _ρ {ω; γ _ρ άω γ {ω; γ)άω (8)

D I «

In particular the integration domain £> at the numerator of eq 8 can be chosen so that only non-overlapping peaks are considered, further reducing the effect of artifacts and noise.

Structural phenotyping of primary and human induced pluripotent stem cell derived cardiomyocytes

To showcase the ability of the sarcomere packing density to characterize the maturation of the cytoskeletal architecture in striated muscle, it was asked whether it could quantify the ability of human and murine induced pluripotent stem cells (respectively hiCMs and miCMs) to replicate the contractile cytoskeletal architecture observed in primary cells (pCMs). pCMs and iCMs were cultured on microcontact-printed hydrogels that mimic the native chemo-mechanical microenvironment and compared and contrasted their sarcomeric a-actinin organization. Qualitatively, the control pCMs showed mature cytoskeleton architecture (Figure 9A(i)): the actin bundles (green) were uniformly distributed throughout the cytosol and displayed clear striations in correspondence of the Z-disks, where most of the a-actinin (red) signal localized; moreover, the cell nucleus (blue) was minimally deformed as expected for the particular cell geometry. Similarly, the cell cytoskeletal in miCMs was (Figure 9B(i)) marked by striations of the actin bundles and regularly-arranged sarcomeric a- actinin positive Z-disks, although few regions displaying less dense packing of the myofibrils (white arrows) or Z-bodies were observed. In contrast, hiCMs exhibited actin and a-actin striations solely in the perinuclear region and arranged in ring-like myofibrils (red arrow in Figure 9C(i)). Moreover, at the cell periphery, the actin and a-actinin signals were diffuse (white arrows) and resembled the cortical architecture (Lauffenburger and Horwitz, 1996, Cell 84:359-369) observed in migratory cells (Figure 11). To quantify these differences, the experiment was restricted to 3 indicators: the nuclear eccentricity (e), an indicator adopted in existing structural phenotyping platforms; the orientational order parameter (OOP), that was previously utilized (Feinberg et al, 2012, Biomaterials 33:5732-5741) to estimate how similar the detected sarcomere orientations (Figure 9Aii, Figure 9Bii and Figure 9Cii) are; and the Fourier transform based (Figure 9Aiii, Figure 9Biii and Figure 9Ciii) sarcomere packing density (ε), that assesses the degree of development a contractile cytoskeleton. As shown in Figure 9E, while pCMs and hiCMs showed similar nuclear morphology

(respectively e=0.439±0.0812 and e=0.564±0.0796) and insignificantly different sarcomere alignment (respectively OOP=0.393±0.0980 and OOP=0.240±0.0749), pCMs did exhibit a significantly (p=0.001, n=3) higher density of well-formed sarcomeres (ε=0.324±0.016) in relation to that of hiCMs (ε=0.127±0.0217). Consistently with qualitative observations, while no significant differences between miCMs and pCMs or hiCMs were observed in the nuclear morphology (e=0.449±0.0422) and global sarcomere orientation (OOP=0.345±0.0224), the presence of periodically arranged sarcomeres (ε=0.262±0.0203) was significantly higher with respect to hiCMs (p=0.003, n=3) and similar to the level observed in pCMs (p=0.066, n=3).

Taken together these data suggest that pCMs and miCMs can be distinguished from hiCMs not only qualitatively, on the basis of structural hallmarks, such as cortical actin and ring-like myofibrils, but also quantitatively through a biophysically-sound metric, the sarcomere packing density, that permits a rigorous statistical classification.

Genetic, epigenetic and environmental factors all contribute to the pathophysiological state of cells and tissues. Recently, image processing and machine learning algorithms have been applied to correlate changes in cell morphology to underlying alterations of the genome (Crane, et al, 2012, Nature Methods 9: 977-980), expressome (Collinet et al., 2010, Nature 464:243-249) or proteome (Permian et al, 2004, Science 306(5699): 1194-1198) of the preparations. Here, the palette of morphometric features utilized in these studies has been extended, introducing a novel metric of cytoskeletal organization: the sarcomere packing density. As demonstrated herein, this metric can effectively distinguish the structural phenotypes of primary and stem cell derived cardiomyocytes using standard statistical tests. Notably, all myocytes considered in this study were positive for sarcomeric a-actinin suggesting that they would have been clustered in the same group based on the sole presence of this protein (Mummery et al, 2012, Circulation Research 111 :344-358) or its transcript (Chin et al, 2009, Cell Stem Cell 5: 111-123).

In previous methods, Fourier analysis has been adopted to estimate the sarcomere length. The automatic approach demonstrated herein offers significant advantages in that it considers the cytoskeleton within the entire cell, reducing the user-bias (Eliceiri et al, 2012, Nature Methods 9:697-710) introduced by manual selection of regions of interest in the spatial (Lundy et al, 2013, Stem Cells Dev 22(14): 1991-2002) or Fourier (Wei et al, 2010, Circulation Research 107:520-531) domains. Moreover, the algorithm to calculate this metric not only yields a better estimate of the sarcomere length but also reveals the relative presence of well-formed sarcomeres. By normalizing the energy of the periodic component to the total energy of the sarcomeric a-actinin immunograph, a cytoskeletal signal-to-noise ratio can be estimated that is independent of the cell size and is bound by the interval [0, 1]; a desirable property for many machine-learning algorithms (Shamir et al., 2010, PLoS Comput Biol 6:el000974).

In this study, metrics of cytoskeletal architecture were used to address the ability of human and murine induced pluripotent stem cell derived cardiomyocytes to assemble a contractile cytoskeleton similar to that observed in primary ventricular myocytes when subjected to engineered extracellular matrix guidance. When unconstrained, cells tend to assume a morphology dictated by their intrinsic cytoskeletal biases. For example, pCMs and miCMs tend to have pleomorphic shapes sustained by polarized cytoskeletal architectures, while hiCMs assumed ring-like cytoskeletal structures (Figure 12). It was previously observed that, on centrally symmetric islands, primary cells could either respond to the ECM cues or retain their natural polarity (Grosberg et al, 2011, PLoS Comput Biol 7:el001088) depending on the cell mechano-transduction ability (Sheehy et al, 2012, Biomech Model Mechanobiol. 11(8): 1227-39). Based on these considerations, a square pattern was chosen, and it was observed that while the cytoskeletal architecture in pCMs and miCMs conformed to the provided boundary conditions, hiCMs retained the ring-like myofibril structure that typified their pleomorphic structural phenotype, suggesting that pathways regulating mechano-transduction (Sheehy et al, 2012) may be engaged differently in the immature hiCMs than in the mature pCMs and miCMs. This is consistent with the notion (Mummery et al., 2012, Circulation Research 111 :344-358) that stem cells need to traverse a hierarchy of cardiac progenitor cells to become mature myocytes: in-vivo this process occurs over -260 days in human and -12 days in mouse (Sissman, 1970, Am J Cardiol 25: 141-148). This suggests that longer time in culture may be beneficial in obtaining mature hiCMs, a fact further supported by a recent study (Lundy et al, 2013, Stem Cells Dev 22(14): 1991-2002) where hiCMs cultured for longer than 100 days, showed strong evidence of structural and functional maturation.

Taken together, these considerations suggest that efforts for post-differentiation maturation strategies should be undertaken, to recapitulate, and possibly accelerate the natural maturation of stem cell derived cardiomyocytes in-vitro. In this context metrics of cytoskeletal architecture, integrated with traditional phenotyping methods (Beqqali et al., 2006, Stem Cells 24: 1956-1967; He et al, 2003, Circulation Research 93:32-39), can enable quantitative characterization of the phenotype of iCMs at each development phase, and proves a valuable quality control tool for stem cell derived cardiomyocytes production (Fox, 2011, Nat Biotechnol 29:375-376). Example 3: Automated Structural Phenotyping of Striated Muscle Cells

Described herein is a system and method of structural phenotyping based on classical image feature detection to elucidate the molecular mechanisms behind genetically or pharmacologically induced changes in cell morphology. As described herein, a set of eleven metrics was developed to capture the increasing sarcomere organization that occurs intracellularly during striated muscle cell development. To test these metrics, the localization of the contractile protein a-actinin was analyzed in a variety of primary and stem cell derived cardiomyocytes, the striated muscle cells present in the heart. Further, these metrics were combined with data mining algorithms to unbiasedly score the phenotypic maturity of human induced pluripotent stem cell derived cardiomyocytes.

Here, a set of eleven metrics were designed (Table 5 and Figure 15) that intrinsically score myocyte structural phenotypes by the increasing degree of organization and alignment that sarcomeres acquire during myofibrillogenesis. These metrics were utilized to score the phenotypic maturity of primary and stem cell-derived cardiomyocytes based on the degree of sarcomeric structural organization observed in α-actinin images.

The materials and methods employed in these experiments are now described. Substrate Engineering Photolithographic masks bearing desired features were drawn in AUTOCAD

(Autodesk, Inc.) and fabricated at the Harvard University CNS (NNIN, Cambridge, MA). UV-light was shone through the mask into a silicon wafer (Wafer World), previously spin- coated with SU-8 3005 photoresist (MicroChem Corp). The wafer was then developed in PGMEA (Sigma) and utilized to cast polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning; 10: 1 ratio) stamps.

PDMS was also spin-coated on 25 mm glass coverslips to a thickness of

approximately 25 μηι and cured overnight at 65°C. Coverslips were treated for 8 minutes in the UVO cleaner before coming into contact with the fibronectin (50 μg/mL, BD Biosciences, Bedford, MA) inked PDMS stamps.

Alternatively, a 15 drop of polyacrylamide solution (streptavidin-acrylamide/bis ratio 7.5/0.3%) was sandwiched between a 25 mm activated coverslip and a 18 mm non- activated one. To transfer fibronectin islands, the hydrogel was dried (37°C, 10 minutes); brought into contact with stamps inked with biotinylated fibronectin (Pierce) and finally sterilized with UV exposure (15 min).

Primary Harvest

Ventricular myocytes were isolated from day 2 neonate Balb/c mice and Sprague Dawley rats according to procedures approved by the Harvard University IACUC. Isolated ventricles were incubated in cold (4°C) 0.1% (w/v) trypsin (USB Corp., Cleveland, OH) solution for approximately 12 hours. Ventricular tissue was further exposed to serial treatments (2 minutes each) of 0.1% (w/v) warm (37°C) collagenase type II (Worthington Biochemical, Lakewood, NJ) solution. Isolated rat and mouse cardiomyocytes were seeded onto the engineered substrates at a density of 10,000 and 20,000 cells/cm ² , respectively, and maintained in culture medium consisting of Medium 199 (Invitrogen, Carlsbad, CA) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS), 10 mM HEPES, 20 mM glucose, 2 mM L-glutamine, 1.5 μΐ _^ vitamin B-12, and 50 U/ml penicillin for the first 48 hours. The FBS concentration was then reduced to 2%. Stem cell culture

Human and murine induced pluripotent stem cell derived cardiomyocytes (hiCM and miCMs) were kindly provided by Cellular Dynamics Inc. (CDI, Madison, WI) and

Axiogenesis (CorAt-iPS, Cologne, Germany) respectively. Cells were cultured as per manufacturers' recommendations. In particular hiCMs were seeded in the presence of vendor-provided plating medium; miCMs were positively selected after plating onto 10 mg/ml fibronectin. After 72 hours, both cell types were dissociated with 0.05% trypsin- EDTA solution (Invitrogen, 25200-072) and re-seeded onto the engineered substrates at a density of 10,000 cells/cm ² .

Immunocvtochemistry and Imaging

Cells were treated with 4% (v/v) paraformaldehyde and 0.05% (v/v) Triton X-100 in PBS at room temperature for 10 minutes before being incubated with anti-sarcomeric a- actinin (A781 1, Sigma) and anti-fibronectin antibody (F3648, Sigma) for one hour at room temperature (1 :200 dilution). Samples were further treated with DAPI (Invitrogen), Alexa Fluor 633-conjugated phalloidin (A22284, Invitrogen) and Alexa Fluor 488-conjugated goat anti-mouse IgG and Alexa Fluor 546-conjugated goat anti-rabbit IgG secondary antibodies (Invitrogen, Carlsbad, CA) for 2 hours at room temperature. Labeled samples were imaged with a Zeiss LSM confocal microscope (Carl Zeiss Microscopy, Jena, Germany) equipped with a EC Plan-Neofluar 40x/1.30 oil DIC M27 objective. 1024x1024 pixels per image were acquired for a final pixel size of 160 nm, much smaller than the sarcomere length (Figure 16C).

Image processing and future extraction

Preprocessing was performed with ImageJ/FIJI (Schindelin et al, 2012, nature

Methods, 9: 676-682). The tubeness plugin was used to highlight the filamentous structure of sarcomeric a-actinin positive pixels (Sato et al, 1988, Medical image analysis, 2: 143-168). The OrientationJ plugin (Rezakhaniha et al, 201 1 , Biomechanics and modeling in mechanobiology, 11 : 461-473) was used to calculate the orientations of each sarcomeric a- actinin positive pixels. MATLAB (Mathworks, Natick, Massachusetts) was also adopted for feature extraction.

Orientational Order Parameter

The structure tensor method (Rezakhaniha et al, 2011 , Biomechanics and modeling in mechanobiology, 1 1 : 461-473) generates a set of orientations {ϋ^ ϋ ₂ ,■■■ , i9 _N } whose frequency of occurrence were plotted in a histogram, such as in Figure 13Ci, Figure 13Cii and Figure 17. The orientational order parameter (OOP) was calculated using the mean resultant vector from circular statistics (Berens et al, 2009, Journal of Statistical Software, 31 : 1 -21) OOP = - N I ¹ ∑ ^Lj y]_ e ^w J

- (9) where ί = V— 1 is the complex unit, e is Euler's number (-2.71) and i9 _; - is the y-th orientation in the set {d^ i9 ₂ , ... , ϋ _Ν }. The sum of unit vectors in equation 9 is bound by 0 (for a set of randomly oriented vectors) and 1 (for a set of perfectly aligned vectors). Further, the orientation histogram was fitted with the following linear mix of Von Mises Distributions _f ro _, . a _, . a ,Λ _„ exp^ cosQ?-/^)] _{(Λ Λ} exp[£ ₂ cosQ9- * ₂ )] , .

(β; μ!, ¾, μ ₂ , ¾, χ) = y + (1 - y) (10)

Where μ _1>2 and δ _{1 2} represent the localization and spread parameters for the Z-disks and Z- bodies peak respectively, γ indicates the fraction of orientations allocated into the Z-disk peak and / ₀ is the modified Bessel function of order 0. Then, 1000 orientations from the two fitted Von Mises probability density distributions were sampled, and OOP ₁ and OOP ₂ for the Z-disk and Z-body peaks were calculated, respectively. In addition, a weighted version of the OOP: wOOP = γ * OOP _t that quantifies both the presence (y) and the alignment (OOP _t ) of the Z-disks in the image was introduced.

Sarcomere Packing Density

To calculate the degree of spatial organization of sarcomeres, the Fourier power spectrum P(u, v) of the pre-processed sarcomeric a-actinin image I(x, y) was first considered f P(u, v) = \F(u, v) \ ²

(11)

[F(U, V) = // ₂ 7(x, y) exp[i 2π ixu + yv)] dxdy

Equation 11 utilizes the Fast Fourier Transform algorithm to establish a correspondence between the spatial domain of the image l(x, y) and the Fourier domain where the power spectrum P(u, v)is defined. The signal energy (E _T0T ) was sampled and integrated along 1024 directions to obtain a ID representation (Γ(ω)) that exhibits periodic peaks (subscript p), in correspondence of spatial frequencies that are integer multiple of the principal harmonic ω ₀ , modulated by an aperiodic noise term (subscript ap). Ετοτ = Μ _Ά2 P(u, v)dudv = J ₀ ^∞ [ω J ^" _ ⁹ ° ₀ Ρ(ω, ϋ)άϋ] άω = ™[Τ(ω)]άω (12)

Equation 12 states that the total energy in the image can be expressed integrating the 2D power spectrum P (u, v) or alternatively, through the integration of its ID

representation Γ(ω) . To approximate this ID function, we chose the relationships in equation 13, where f _ap is a decaying exponential and f _p is the sum of 3 Gaussian peaks. f (ω; ξ) = Γ _ρ (ω; ξ _ρ ) + Τ _αρ (ω; ξ _αρ )

Γ _αρ (ω; ξ _αρ ) = a exp(- ω/b) ; ξ _αρ = {a, b] (13) Γ _ρ (ω; ξ _ρ ) = ∑l ₌₁ a _k exp[- (ω - k i ₀ ) ² /b _k ] ; ξ _ρ = {a _k , b _k , ω ₀ } _{¾=1 2ι3}

The function was fitted to the data using Matlab "lsqnonlin" function that uses the Trust Region Reflective algorithm (Coleman and Li, 1996, Siam J Optimization, 6: 418-445). By this approach, the parameters were able to be estimated, which were then used to calculate (see Equation 14) the sarcomere length (SL), the area under the periodic component E _p (or sarcomeric energy) as well as its ratio with the total area (total energy), a quantity termed herein as sarcomeric packing density (SPD).

SPD = E _p /E _T0T (14)

SL = ω ₀ ^_1 Machine Learning

The naive Bayes, neural network and tree bagging classifiers were implemented using Matlab built-in functions. Ten random iterations were seeded to ensure the results were stochastically robust. For the naive Bayes and tree bagging classifiers, training was performed with a 10-fold cross-validation test. For the neural network, 70%, 15% and 15% of the rpCMs dataset were used for the training, validation and testing phases, respectively.

The machine learning problem which was attempted to be solved could be divided in 2 parts. In the training phase, it was aimed to solve the following direct problem: given a set of classes C = {D, I, M] (respectively differentiated, immature and mature myocytes) and a set of features F= {F ₁₍ F ₂ , ... , F _M ] (related to the myofibrillar architecture of cells pertaining to those classes, such that e.g. F ₁ = SPD, F ₂ = OOP, ... ) an algorithm was sought that mapped a given combination of features / = {f _lt f ₂ , ... , f _M ] to a single class C = c. In the classification phase, it was aimed to solve the inverse problem: given a new set of features, that was not used during training, the algorithm was asked to assign it to one of the available classes.

A dataset was collected comprising sarcomeric a-actinin digital images from rpCMs obtained at 6, 24 and 48 hrs after seeding. Based on a-priori knowledge (Dabiri et al, 1997, Proc Natl Acad Sci USA, 94: 9493-9498; Parker et al, 2008, Cir Res, 103: 340-342; Sheehy et al, 2014, Stem Cell Reports, 2: 282-294) these time points were chosen to represent the classes of differentiated, immature and mature myocytes, respectively. Several classifiers were then implemented and trained (see Table 6) on this dataset. The classifiers were then utilized to classify features measured on independently acquired sarcomeric α-actinin images from hiCMs samples. In particular the number of times the myofibrillar architecture of hiCMs was not classified as mature was counted.

Since different classifiers are based on different assumptions and rely on different stochastic algorithms (Sun et al, 2012, Sci Transl Med, 4: 130ral47), three different frameworks and ten different randomly seeded iterations were tested, to ensure the classification was robust to the choice of a specific machine learning strategy and particular initialization. Classifier 1 : Naive Baves

In the framework of Bayesian classification, a classifier is based on a conditional model for the probability that a certain set of features belongs to a given class (Figure 18C-i). Under the naive hypothesis of conditional independence between features and once a suitable prior is selected, all model parameters can be derived applying the maximum likelihood estimation algorithm on the training dataset. Kernel distributions was chosen as priors, since they only require the random variables to be continued and not normally distributed, and randomly seeded ten different iterations of the Naive Bayes classifier. A 10- fold cross-validation test was adopted to determine the performance of the classifier. Classifier 2: Neural Network

In the framework of neural network a classifier is a network that possesses: i) an input layer, with as many neurons as there are features; ii) at least one hidden layer, with a number of neurons that can be optimized; and iii) an output layer, with as many neurons as there are classes. The neurons η _έ and n _; - are connected through a weight w _t and training the network is equivalent to assign the weights w _t such that when the input layer receives the set of features {F _t , F ₂ , ... , F _M ] pertaining to the class C, the output node associated with C exhibits the highest value (Figure 18C-ii). Matlab Neural Network toolbox was utilized to design a perceptron network and the number of hidden neurons (from 1 to 20) was optimized for 10 random iterations. The network was also trained using the back-propagation algorithm adopting 70% of the dataset for Training, 15% for Validation and 15% for final Testing.

Classifier 3 : Tree Bagging

Tree bagging stands for bootstrap aggregation of decision trees. Bootstrap aggregation is an ensemble meta-algorithm that optimally subdivides the entire dataset and uses each part to train a simpler classifier (in this case a binary decision tree). The final classification is obtained by voting: that is, if the maj ority of trees has assigned the set of features

{F ₁₍ F ₂ , ... , F _M ] to the class C = c than C = c will be the result of the global classification. The number of decision trees to be used is thus the parameter optimized in the

implementation.

Binary decision trees are simple yet powerful machine learning algorithms. For each feature {F ₁₍ , ... , F _M ] the algorithm chooses thresholds {t^ , ... , t _M ] and an ordering scheme, such that the set of features can be traveled from "the root to the leaves" coherently. The value of a given feature F _k is considered, compared with the relative threshold and a decision is made: either to examine the next feature in the tree F ₊₁ , or to assign a specific class to that combination (Figure 18C-iii).

Ten random initialization were performed of the algorithm. The optimal number of trees (in the range 5-100) was selected and each decision tree was then trained using the Gini's diversity index. A 10-fold cross-validation test was adopted to determine the performance of the classifier.

The results of the experiments are now described. Quantitative Analysis of the Contractile Cvtoskeleton in Striated Muscle Cells

Sarcomeres are ~2 μιτι long ultra-structures delimited by Z-disks that are rich in the contractile protein a-actinin (red in Figure 13 A). The localization of this protein can be taken to indicate the maturity of cardiac myocytes (Grosberg et al, 2011 , PLoS Comp Biol, 7: el001088): in differentiating cells, a-actinin is diffuse in the cytoplasm (Figure 13A-ii); in immature myocytes, it appears as a fibrous structure or as aperiodically-spaced puncta known as Z-bodies (Figure 13 A-iii); and in mature myocytes, a-actinin localizes into the Z-disks regular lattice (Figure 13A-iv). Therefore, the analysis (Figure 13B) was focused on identifying how regularly-spaced and well-aligned sarcomeric α-actinin positive structures were in the images. Each sarcomeric α-actinin positive pixel was first associated with the orientation (color-coded in Figure 13C-i) of its local neighborhood. Then, a bimodal distribution was fitted to the resulting orientation histogram (red and black curves in Figure 13C-ii). This enabled the extraction of several metrics: the global orientational order parameter (Grosberg et al, 2011, PLoS Comp Biol, 7: el001088) (OOP) a value that ranges from 0 to 1 as contractile elements become more aligned; separate OOPs for the two fitted modal distributions, representing Z-disks and Z-bodies, as well as their relative presence. Further, the image power spectrum was radially integrated (Figure 13C-iii), yielding a ID representation (Figure 13C-iv) that highlights the relative importance of each spatial frequency. In particular, the peaks (red curve) represent α-actinin positive elements arranged at a distance on the order of the sarcomere length and therefore become more prominent as sarcomerogenesis progresses. Through non-linear fitting, this component was separated from the aperiodic contribution of Z-bodies and other image artifacts (black curve). Relevant metrics in this case included the area (shaded in red) under the periodic component, the total area under the data curve, and their ratio, a quantity termed herein as the sarcomere packing density (SPD). Taken together, this set of metrics has a direct biophysical interpretation: substantially more mature myofibrillar architectures exhibit a regular lattice of well-oriented Z-disks, resulting in elevated values of SPD and OOP. Additionally, this analysis is robust to common imaging artifacts such as out-of-focus blurriness, salt-and-pepper noise or poor contrast (Figure 16C).

Quantitative Analysis of the Contractile Cvtoskeleton in Murine Primary and Stem Cell Derived Single Cardiomyocytes

To test the analysis tool, it was asked whether the ability of human and murine induced pluripotent stem cell-derived cardiomyocytes (hiCMs and miCMs, respectively) to replicate the contractile cytoskeletal architecture observed in murine primary cardiomyocytes (mpCMs) in vitro (Parker et al, 2008, Cir Res, 103: 340-342) could be quantified.

Qualitatively, it was observed that mpCMs (Figure 13D-i) and miCMs (Figure 13D-ii) three days after seeding on square fibronectin islands showed mature myofibrillar architecture, characterized by uniformly distributed sarcomeric a-actinin rich striations. Conversely, hiCMs (Figure 13D-iii) exhibited sparse Z-disks solely in the perinuclear region and arranged in ring-like myofibrils (red arrow in Figure 13D-iii). In addition, close to the hiCM membrane, the actin and sarcomeric a-actinin signals were diffuse (yellow arrows) and resembled the cortical architecture observed in immature and/or migratory cells (Parker et al., 2008, Cir Res, 103: 340-342; Sheehy et al., 2012, Biomechanics and Modeling in

Mechanobiology, 11 : 1227-1239). Quantitatively, while local regions of aligned Z-disks could be detected (color-coded insets below the panels in Figure 13D) the Fourier analysis clearly demonstrated a globally reduced periodicity in the sarcomere distribution of hiCMs than observed in miCMs and mpCMs (insets on the right of the panels in Figure 13D).

Consistently, the SPD measured in mpCMs and miCMs were 2 times higher than in hiCMs (Figure 17). Notably, all myocytes considered in this study were positive for sarcomeric a- actinin, suggesting that they would have been clustered in the same group by traditional assays detecting the presence of this protein or its transcript (Cahan et al, 2014, Cell, 158: 903-915; Mummery et al, 2012, Circ Res, 111 : 344-358).

Quantitative Analysis of the Maturity of the Contractile Cvtoskeleton in Human Stem Cell Derived Cardiomyocvtes

It has been previously shown that extending time in culture (Lundy et al., 2013, Stem Cells Dev, 22: 1991-2002; McCain et al., 2014, Biomaterials, 35: 5462-5471) could be beneficial for obtaining hiCMs with a more mature phenotype than observed here. However, evaluating the quality of mass-produced stem cell derived myocytes requires an extensive structure-function characterization and a direct comparison against myocytes exhibiting a post-natal phenotype (Sheehy et al, 2014, Stem Cell Reports, 2: 282-294). It was reasoned that, since the process of myofibrillogenesis is highly conserved across species (Sissman, 1970, American Journal of Cardiology, 25: 141-148), a pre-screening tool could be designed that estimates the effectiveness of maturation strategies by integrating the described metrics of myofibrillar architecture with machine learning algorithms for structural phenotyping. While available platforms require a user-selected training set, a set of images that an "expert" assigns to all phenotypic classes (Eliceiri et al, 2012, Nature Methods, 9: 697-170), the ability to recapitulate myofibrillogenesis in vitro using primary cardiomyocytes from neonate rats (rpCMs) (Agarwal et al, 2013, Advanced Functional Materials, 23: 3738-3746; Parker et al., 2008, Cir Res, 103: 340-342) was utilized to create such a training set. First, rpCMs and hiCMs were cultured as engineered tissues that mimic the native architecture of the myocardium (Sheehy et al, 2014, Stem Cell Reports, 2: 282-294) (Figure 14A and Figure 18). Secondly, the sets of features (Table 5 and Figure 15) extracted from images collected at 6 (Figure 14A-i), 24 (Figure 14A-ii), and 48 hr (Figure 14A-iii) after seeding were assigned to the classes of differentiated, immature, and mature myocytes, respectively. Digital images were collected from more than 100 cells (insets, Figure 14A and Figure 18) in each condition. Thirdly, the automatically annotated dataset was utilized to train a simple naive Bayes classifier as well as two more advanced supervised learning strategies based on neural networks and tree bagging (Figure 18C). Three machine learning strategies were selected that operate under various assumptions (Table 6) to demonstrate that the classification was not biased (Eliceiri et al, 2012, Nature Methods, 9: 697-170) by the choice of one specific algorithm. Finally, the three classifiers were asked to confirm whether or not hiCMs (Figure 14A-iv) possess a mature structural architecture.

Table 5: List of metrics of sarcomere organization developed, integrated or updated for this study. Additionally, see Figure 15 for a schematic representation of the role of each parameter. At the tissue level, as rpCMs in culture underwent myofibrillogenesis, it was observed that wOOP and SPD increased as expected. In comparison, hiCMs scored values consistent with their immature myofibrillar organization (Fig IB). Further, all classifiers trained on the rpCMs dataset failed to recognize a mature myofibrillar architecture in the majority of hiCMs images (Figure 18D). Specifically, -70% of the 118 hiCM images were not classified as mature by the naive Bayes classifier; -71% by the neural network classifier and -77% by the tree bagging classifier. For example, the neural network classified -40% of hiCMs as differentiated myocytes and -31% as immature myocytes (Figure 14C).

Interestingly, 29% of the cells embedded in anisotropic hiCM tissues did display mature myofibrillar architectures, suggesting a pool of hiCM with enhanced myogenic potential may exist (Hartjes et al, 2014, Stem Cells, 32: 2350-2359). Thus, this approach i) provided quantitative metrics for the organization of the contractile cytoskeleton of primary and stem cell derived cardiomyocytes and ii) utilized this information to unbiasedly and robustly quantify their maturation.

Table 6: Machine learning algorithms adopted for the analysis of the myofibrillogenesis dataset (Figure 18)

Metrics to characterize the phenotvpe of cardiomyoctves

Quantitative methods to characterize the contractile cytoskeleton of striated muscle cells have been previously proposed. For example, the analysis of the orientation of intracellular elements has been conducted adopting mean orientations (Rao et al, 2013, Biomaterials, 34: 2399-2411). Unfortunately, the specific circular statistics tests (Berens et al., 2009, Journal of Statistical Software, 31 : 1-21) required to compare and contrast these mean orientations are only rarely adopted. By utilizing the OOP values in the range 0-1, classical statistical tools, such as ANOVA (Figure 17) were able to be employed.

Additionally, approaches based on non-linear fitting of multiple Von Mises distributions have been proposed (Rezakhaniha et al, 2011, Biomechanics and modeling in mechanobiology, 11 : 461-473); in the context of fluorescence bioimaging though, these methods may suffer from over-fitting issues, given the large numbers of parameters that are needed to identify multiple distributions. Here the analysis was restricted to only two distributions, centered on the two biophysically-relevant principal directions of Z-disks and Z-bodies that are orthogonal to one another. Thus, the number of fitting parameters was able to be limited, largely reducing the risk of over-fitting. Fourier analysis has also been previously considered in the estimation of sarcomere length (Lundy et al, 2013, Stem Cells Dev, 22: 1991-2002; Wei et al, 2010, Circ Res, 107: 520-531). The automatic approach presented herein offers significant advantages in that the contractile cytoskeleton within the entire cell is considered, reducing the user-bias (Eliceiri et al, 2012, Nature Methods, 9: 697-710) introduced by manual selections in the spatial (Lundy et al., 2013, Stem Cells Dev, 22: 1991-2002) or Fourier (Wei et al, 2010, Circ Res, 107: 520-531) domains. Moreover, the present algorithm not only yields an estimate of the sarcomere length across the entire spatial extension of the cell/tissue but also reveals the relative presence of well-formed sarcomeres irrespectively from the direction of their alignment. When the myofibrils are highly aligned (Figure 14A- iii), the Fourier spectrum exhibits peaks along one principal direction; conversely, when sarcomeres are well aligned along many directions (Figure 13D-i), the Fourier spectrum shows a circular partem. By integrating across all directions in the Fourier domain, the sarcomeres translational periodicity across all directions in the spatial domain was able to be quantified. Finally, the method proposed here for the calculation of the sarcomere packing density significantly improves previous efforts (McCain et al, 2014, American Journal of Physiology: Heart and Circulatory Physiology, doi: 10.1152/ajpheart.00799.2013; Wang et al., 2014, Nature Medicine, 20: 616-623). By normalizing the energy of the periodic component to the total energy of the sarcomeric a-actinin image, a signal-to-noise ratio is estimated that is bound in the unit interval, a desirable property for many machine-learning algorithms (Shamir et al, 2010, PLoS Comp Biol, 6: el000974).

In conclusion, eleven metrics have been developed to characterize the structural phenotype of primary and stem cell-derived cardiomyocytes in a way that is biophysically related to their functional proficiency (Feinberg, 2007, Science, 317: 1366-1370). Moreover, by engineering myocyte shape and tissue architecture, a myofibrillogenesis dataset was generated that allows structural phenotyping of stem cell-derived cardiomyocytes in an unbiased fashion and that is largely robust to the choice of a specific machine learning strategy. Finally, while assessing the quality of human pluripotent stem cell-derived myocytes remains critical, to date their maturation has been sub-optimally estimated

(Mummery et al, 2012, Circ Res, 111 : 344-358; Sheehy et al., 2014, Stem Cell Reports, 2: 282-294) as healthy human myocytes are not readily available. Since myofibrillogenesis is an extremely well-conserved physiological process (Sissman, 1970, American Journal of Cardiology, 25: 141-148), this method allows for a quantitative characterization of myocytes maturation that naturally overcomes this limitation.

Example 4: Platform for Structural and Functional Phenotyping of Developing Healthy and Dystrophic Muscle

In the present study, the presently described algorithm was used to quantify nuclear and cytoskeletal changes and contractile strength during myoblast fusion and maturation in healthy and Duchenne muscular dystrophy (DMD) human engineered tissues.

Human Skeletal Muscle Culture

Human skeletal muscle myoblasts (Lonza, Walkersville, MD) were purchased and cultured in a growth media consisting of M199 culture media medium (GIBCO, Invitrogen, Carlsbad, CA) supplemented with 10% heat inactivated fetal bovine serum, 10 mM HEPES, 0.1 mM MEM nonessential amino acids, 20 mM glucose, 1.5 μΜ vitamin B-12, and 50 U mL ^"1 penicillin (GIBCO). The myoblasts derived from a Duchenne muscular dystrophy patient were purchased from DV Biologies (Costa Mesa, CA) and were cultured in the same growth media. Cells were plated and expanded in T-75 culture flasks (Coming, Corning, NY) prior to cell seeding. Prior to cell plating, T-75 culture flasks were incubated with 1 wt% gelatin from porcine skin (Sigma Aldrich, St. Louis) in a cell culture incubator for 30 minutes. The 1% gelatin solution was subsequently replaced with grow media in preparation for cell plating. The skeletal muscle myoblasts were allowed to proliferate over the course of 2-3 days and were seeded prior to reaching 70% confluence. Cell culture media was replenished after 48 hours after 3 rinses with sterile phosphate-buffered saline. For myoblast seeding, cells were removed from the cell culture flask using 0.25% of trypsin-EDTA (GIBCO), seeded onto chips at a density of 62,000 cells cm ^"2 . The myoblasts formed confluent monolayers of tissue after 48 hours in growth media. Next, to induce differentiation of the myoblasts, the growth media was replaced with a differentiation media consisting of DMEM-F12 (Lonza), supplemented with 2% horse serum (GIBCO). Differentiation media was replenished every 48 hours. Experiments were performed during a period ranging from 3 days culture in differentiation media to 6 days culture in differentiation media.

Immunohistochemistry

Engineered human skeletal muscle tissues were fixed by treatment with 4% (v/v) paraformaldehyde for 15 minutes (16% stock diluted in phosphate-buffered saline, Electron Microscopy Sciences, Hatfield, PA) and permeabilized with 0.05% (v/v) Triton X-100 in PBS at room temperature for 5 minutes. The samples were then washed with phosphate- buffered saline prior to incubation with primary antibodies. The samples were incubated with a primary antibody solution consisting of a 1:200 dilution of monoclonal anti-sarcomeric a- actinin antibody (A7811, clone EA-53, Sigma Aldrich, St. Louis, MO) in phosphate-buffered saline for 2 hours at room temperature. After 3X rinsing with room temperature phosphate- buffered saline, the engineered tissues were incubated with a secondary solution consisting of 1 :200 dilutions of Alexa Fluor 546-conjugated goat anti-mouse IgG secondary antibody (Invitrogen, Carlsbad, CA), Alexa Fluor 488-conjugated Phalloidin (Product# A12379 Invitrogen, Carlsbad, CA), and 4', 6-diamidino-2-phenylindole (DAP I) (Catalog #D1306, Invitrogen, Carlsbad, CA) for 1 hour at room temperature. Fluorescence microscopy was performed using a Zeiss LSM confocal microscope (Carl Zeiss Microscopy, Jena, Germany) using the EC Plan-Neofluar 40x/1.3 oil DIC M27 objective.

Quantitative Analysis of Nuclear Architecture Nuclear eccentricity and orientation were quantified using a custom image analysis software (Bray et al., 2010, Biomaterials, 31 : 5143-50). Briefly, an ellipse-fitting algorithm was utilized to trace the perimeter of nuclei in immunofluorescence images, enabling measurement of major and minor axes lengths, and eccentricity. In order to measure orientation angle, an orientation vector was drawn through the longitudinal axis of the fitted ellipse. To quantify global nuclear alignment, the orientational order parameter (OOP) was calculated using the mean resultant vector of the measured orientation angles from circular statistics (Pasqualini, 2015, Stem Cell Reports: 4, 340-7; Berens, 2009, J Stat Softw, 31 : 1- 21). Measurements were repeated for every nucleus within a field of view, excluding overlapping nuclei as well as nuclei not fully within the field of view. Statistical analysis to compare nuclear data was performed using SigmaPlot™ 12.0 software (San Jose, CA). Data from each condition was analyzed using Kruskal-Wallis One Way ANOVA on Ranks and compared pairwise using Dunn's Method. Results with p-values less than 0.05 were considered statistically significant.

Quantitative Analysis of Actin Cvtoskeleton and Sarcomere Organization

Custom image analysis software was utilized (Pasqualini, 2015, Stem Cell Reports: 4, 340-7) to analyze the organization of the actin cytoskeleton and sarcomeres. Briefly, for analysis of the actin cytoskeleton, the orientation of actin fibers was determined using the OrientationJ plugin (Rezakhaniha et al, 2012, Biomechanics and modeling in

mechanobiology, 11 : 461-73) and the actin OOP was calculated for each sample. The global OOP was calculated for each sample by analyzing ten fields of view (318. 5 μιτι by 318.5 μιτι). Five samples were compared for each condition on each day of comparison. Statistical analysis was performed using SigmaPlot™ 12.0 software (San Jose, CA). Data from each condition was analyzed using Kruskal-Wallis One Way ANOVA on Ranks and compared pairwise using Dunn's Method. Results with p-values less than 0.05 were considered statistically significant.

To quantify the expression of sarcomeric a-actinin, the projected area of the pixels positive for a-actinin immunosignal was measured relative the projected area of the pixels positive for phalloidin immunosignal using custom MATLAB software (MathWorks, Natick, MA). Custom analysis was used to determine the relative presence of sarcomeres as periodic structures in the a-actinin stainings (Pasqualini, 2015, Stem Cell Reports: 4, 340-7). In brief, the 2D Fourier transform of the a-actinin immunostaining is radially integrating in a ID signal that contains the relative importance of each spatial frequency. This signal was then fitted, to separate the aperiodic contribution of the z-bodies from the periodic contribution of the z-disks. The ratio of the areas under the periodic and aperiodic components, named sarcomeric packing density, was used compare the relative assembly of sarcomeres among the conditions.

Healthy and DMD primary human myoblasts were seeded on isotropic or anisotropic fibronectin micropatterned constructs. To quantify the resulting myoblast tissue architecture, healthy and DMD primary human myoblasts were seeded on to the micropatterns and cultured the tissues in growth media for 2 days (day -2 to day 0) to allow for formation of a confluent tissue. Immunohistochemistry was utilized to identify the actin cytoskeleton and nuclei within the engineered tissue (Figure 20A i-ii). The orientation of the actin fibers was measured using a method based on fingerprint identification algorithms and the orientational order parameter (OOP) (Grosberg et al., 2011, PLoS Comp Biol, 7: el001088; Pasqualini, 2015, Stem Cell Reports: 4, 340-7), a score of global anisotropy ranging from 0, representing random organization, to 1, representing parallel alignment was calculated. Both of the healthy and DMD anisotropic conditions exhibited more highly aligned actin cytoskeletons compared to the healthy isotropic condition (Figure 20B). However, the actin cytoskeleton of the anisotropic DMD myoblasts was less organized compared to the healthy myoblasts cultured on the same micropattern (Figure 20B).

In order to further quantify differences in subcellular architecture within the engineered tissues, the nuclear alignment and eccentricity was analyzed, which has been previously found to be sensitive measurements of anisotropy and cellular aspect ratio among highly organized tissues in smooth muscle (Alford et al, 2011, Integrative Biology:

Quantitative Biosciences from Nano to Macro, 3: 1063-70). To quantify nuclear alignment, a previously described method (Bray et al, 2010, Biomaterials, 31 : 5143-50) was used, which utilizes an ellipse-fitting tool to outline the perimeter of each nucleus within a field of view. Orientation vectors were drawn through the long axis of the fitted ellipses and the orientational order parameter was calculated using these orientation angles. It was found that both of the anisotropically -patterned healthy and DMD myoblasts tissues exhibited significantly higher nuclear alignment compared to the isotropic condition; however, the healthy anisotropic myoblasts had significantly greater nuclear alignment compared to the anisotropically patterned DMD myoblasts (Figure 20C). These data suggest that actin cytoskeletal and nuclear architecture of human skeletal myoblasts are dependent on the geometrical constraints imposed by the extracellular matrix, similar to results in a number of other cell types such as endothelial cells (Versaevel et al., 2012, Nature Communications, 3: 671), cardiac cells (Bray et al, 2010, Biomaterials, 31 : 5143-50), smooth muscle cells (Alford et al, 2011, Integrative Biology: Quantitative Biosciences from Nano to Macro, 3: 1063-70;Ye et al, 2014, Integrative Biology: Quantitative Biosciences from Nano to Macro, 6: 152-32). However, the DMD myoblasts exhibited decreased actin, nuclear alignment, and nuclear eccentricity compared to healthy myoblasts cultured on the same micropattern suggesting DMD myoblasts may not polarize to the extent of healthy myoblasts because nuclear eccentricity correlates with cellular aspect ratio and basal tone in smooth muscle cells (Alford et al, 2011, Integrative Biology: Quantitative Biosciences from Nano to Macro, 3: 1063-70).

Quantifying Changes in Cvtoskeletal Architecture During Myotube Maturation

Muscle formation and repair result from a dramatic transformation of mono- nucleated, non-striated myoblasts to multi-nucleated, striated myofibers via the process of myoblast fusion and myofibrillogenesis. Myoblast fusion and myofibrillogenesis require disassembly and reassembly of cytoskeletal components, transport and redistribution of myonuclei as well as de novo formation of sarcomeres, the contractile unit of myofibers (Doberstein et al, 1997, Joumal of Cell Biology, 1249-61 ; Fulton et al, 1981 ; Journal of Cell Biology: 91 : 103-12). Here, it was asked whether structural changes of the actin cytoskeleton and nuclei that occur during myoblast fusion and myotube maturation could be quantified. Moreover, it was examined whether anisotropic DMD myotubes would exhibit less organization relative to anisotropic healthy myotubes. To test this hypothesis, the actin OOP and nuclear OOP was quantified after 3 and 6 days in differentiation media. In the anisotropically-pattemed healthy and DMD myoblasts, increases in the alignment of the actin cytoskeleton from day 0 to day 6 was observed, as myoblasts fused to become myotubes of larger aspect ratios relative to mononucleated myoblast tissues (Figure 21A i-ii, iv-v, Day 0 data from Figure 20B was included for illustrative purposes). Similarly, a significant increase in actin alignment occurred in the isotropically patterned myoblasts from day 0 to day 6 in response to switching to a differentiation medium on day 0, but the tissues cultured on the isotropic patterning remained significantly less aligned relative to the tissues cultured on anisotropic patterns (Figure 21 A iii,vi; and Figure 21B). These data confirm that alignment of myoblasts in an end to end manner is a spontaneous, critical step for myoblast fusion in human skeletal myoblasts, consistent with studies performed in other mammalian systems (Swailes et al, 2004, Journal of Anatomy, 381-91). Importantly, this finding suggests that DMD myoblasts can fuse and form a highly organized actin cytoskeleton within myotubes.

Quantifying Changes in Nuclear Morphology During Myotube Maturation

Next, it was asked whether nuclear alignment and shape corresponded with cytoskeletal organization during myoblast fusion and myotube maturation in healthy and DMD tissues. In the present study, a significant increase in nuclear alignment from day 0 to day 3 and 6 was observed in the anisotropically patterned healthy and DMD myoblasts (Figure 21 C, Day 0 data from Figure 20C was included for illustrative purposes) as myoblasts fused to form myotubes. Taken together, these data suggest that nuclear and actin cytoskeletal architectures are dependent on the geometry of the ECM, but may decouple during fusion, consistent with previous work suggesting myonuclear transport occurs via microtubules and microtubule associated motors (Elhanany-Tamir et al, 2012, Journal of Cell Biology, 198: 833-46; Folker et al, 2012, Development, 139: 3827-37).

Quantifying Myofibrillogenesis. Myotube Number, and Myotube Diameter During

Maturation

In addition to remodeling the actin cytoskeletal and nuclear architecture during myoblast fusion and myotube maturation, fused myoblasts undergo myofibrillogenesis. This is the biological process that endows skeletal muscle with the structural framework for its contractile function. Briefly, premyofibrils form at the edges of the myotube in the form of Z- bodies, rich in punctate sarcomeric a-actinin (Dabiri et al, 1997, Proc Natl Acad Sci USA, 94: 9493-9498). Over time, these Z-bodies fuse to form Z-disks that represent the boundaries of the contractile unit of the myofiber, called sarcomeres (Dabiri et al, 1997, Proc Natl Acad Sci USA, 94: 9493-9498). Here, it was asked whether the rate of myofibrillogenesis was decreased in anisotropic DMD tissue compared to anisotropic healthy tissue. To answer this question, the abundance and ordered structure of sarcomeric α-actinin of healthy and DMD engineered muscle cultured on the 15 x 2 μιτι pattern as well as the isotropically patterned healthy engineered muscle was quantified utilizing image analysis software for analyzing sarcomeric structure in maturing cardiomyocytes (Pasqualini, 2015, Stem Cell Reports: 4, 340-7). To measure the abundance of sarcomeric a-actinin, the projected area of sarcomeric α-actinin immunosignal relative to the area of f-actin immunosignal of images acquired using immunohistochemistry and confocal microscopy was measured. Significantly greater projected area of sarcomeric α-actinin immunosignal was observed in the healthy anisotropic condition compared to the DMD condition and healthy isotropic condition on both day 3 and 6 (Figure 22Ai-vi, Figure 22B). Next, to determine whether the Z-bodies were undergoing organized assembly into Z-disks indicative of mature sarcomeric structure, the fraction of periodically distributed sarcomeric a-actinin immunosignal was quantified, which is termed sarcomeric packing density (Pasqualini, 2015, Stem Cell Reports: 4, 340-7). It was found that healthy engineered muscle had significantly greater SPD compared to DMD engineered muscle at the same time points. Hence, these data suggest that DMD myoblasts do not form sarcomeres as efficiently as healthy myoblasts.

In the studies presented herein, the presently described algorithm was employed to analyze the differences in cytoskeletal and nuclear remodeling that occur during myoblast fusion and myotube maturation due to differences in myoblast architecture. It was found that DMD myoblasts form significantly fewer and smaller myotubes compared healthy myoblasts on the same micropattern.

The disclosures of each and every patent, patent application, and publication cited herein are hereby incorporated herein by reference in their entirety.

While this invention has been disclosed with reference to specific embodiments, it is apparent that other embodiments and variations of this invention may be devised by others skilled in the art without departing from the true spirit and scope of the invention. The appended claims are intended to be construed to include all such embodiments and equivalent variations.