Morphological restriction of human coronary artery endothelial cells substantially impacts global gene expression patterns


  • Jessica M. Stiles,

    1. Department of Biomedical Sciences, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, El Paso, TX, USA
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Robert Pham,

    1. Klipsch School of Electrical and Computer Engineering, New Mexico State University, Las Cruces, NM, USA
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Rebecca K. Rowntree,

    1. Department of Biomedical Sciences, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, El Paso, TX, USA
    Search for more papers by this author
  • Clarissa Amaya,

    1. Department of Biomedical Sciences, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, El Paso, TX, USA
    Search for more papers by this author
  • James Battiste,

    1. Department of Neurology, University of Texas Southwestern Medical Center, Dallas, TX, USA
    Search for more papers by this author
  • Laura E. Boucheron,

    1. Klipsch School of Electrical and Computer Engineering, New Mexico State University, Las Cruces, NM, USA
    Search for more papers by this author
  • Dianne C. Mitchell,

    1. Department of Biomedical Sciences, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, El Paso, TX, USA
    Search for more papers by this author
  • Brad A. Bryan

    Corresponding author
    1. Department of Biomedical Sciences, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, El Paso, TX, USA
    • Correspondence

      B. A. Bryan, Center of Excellence in Cancer Research, Paul L. Foster School of Medicine, Texas Tech University Health Sciences Center, 5001 El Paso Drive, MSB1 Room 2111, El Paso, TX 79905, USA

      Fax: +1 915 783 5222

      Tel: +1 915 783 5235


    Search for more papers by this author


Alterations in cell shape have been shown to modulate chromatin condensation and cell lineage specification; however, the mechanisms controlling these processes are largely unknown. Because endothelial cells experience cyclic mechanical changes from blood flow during normal physiological processes and disrupted mechanical changes as a result of abnormal blood flow, cell shape deformation and loss of polarization during coronary artery disease, we aimed to determine how morphological restriction affects global gene expression patterns. Human coronary artery endothelial cells (HCAECs) were cultured on spatially defined adhesive micropatterns, forcing them to conform to unique cellular morphologies differing in cellular polarization and angularity. We utilized pattern recognition algorithms and statistical analysis to validate the cytoskeletal pattern reproducibility and uniqueness of each micropattern, and performed microarray analysis on normal-shaped and micropatterned HCAECs to determine how constrained cellular morphology affects gene expression patterns. Analysis of the data revealed that forcing HCAECs to conform to geometrically-defined shapes significantly affects their global transcription patterns compared to nonrestricted shapes. Interestingly, gene expression patterns were altered in response to morphological restriction in general, although they were consistent regardless of the particular shape the cells conformed to. These data suggest that the ability of HCAECs to spread, although not necessarily their particular morphology, dictates their genomics patterns.




human coronary artery endothelial cell




transforming growth factor




Regulation of the vascular system is essential for tissue growth and homeostasis, and aberrant vascular signalling has been implicated in a vast number of diseases, such as cancer, diabetes, arthritis, macular degeneration and cardiovascular disorders [1]. The majority of research examining endothelial function has focused on the effects of secreted growth factors and cytokines such as vascular endothelial growth factor, fibroblast growth factor, transforming growth factor (TGF)β and a host of other molecules on endothelial cell signalling and physiology. Although these factors undoubtedly play a critical role in regulating cardiovascular development, function and disease, a growing number of studies indicate that endothelial physiology, as well as that of many other cell types, is directed by an intimate combination of physical, chemical and biological cues present in the tissue microenvironment [2, 3]. Over a century ago, physical cues were hypothesized to play important roles in tissue development and there are no better examples in the human body than the deleterious effects of microgravity on bone structure [4] and hypertension on cardiovascular function [5]. However, almost all organisms have evolved specific structures that are tailored to respond to nano- and macroscale physical forces whereby cells are able to detect and respond to external forces through mechanically induced conformational or organizational changes in cellular molecules, such as stretch-sensitive ion channels, G protein coupled receptors, tyrosine kinase receptors, cadherins and integrins located on the plasma membrane and in cell-to-cell and cell-to-extracellular matrix junctions [6].

Over the past decade, a large number of studies have manipulated endothelial tension, compression and shear stress aiming to determine how mimicking blood flow affects endothelial function [7]. Despite the progress made in this area, many of the mechanisms regulating how extrinsic mechanical stresses affect endothelial physiology remain unknown, and the implications of such studies are primarily limited to extrapolations of how lumenal blood flow from normal, hypertensive and sclerotic conditions affects endothelial cells. A wealth of primarily qualitative evidence suggests that cell morphology-specific regulation of mechanotransduction is essential for cellular fate decisions such as proliferation, apoptosis, differentiation and quiescence [8-12]. For example, restriction of endothelial cell spreading using micropatterned substrates induces cell cycle arrest and apoptosis [8]. Alternatively, cell proliferation increases when cell spreading is allowed, whereas cells preferentially undergo differentiation in a moderately spread state. Endothelial migration is significantly more guided and regulated on narrower adhesive surfaces than on larger ones and geometric cues have been shown to modulate endothelial differentiation [13]. Other cells types may show distinct phenotypes solely on morphological alterations. For example, it has been reported that human stem cells can be directed to osteogenic or adipogenic developmental lineages by simply manipulating cell shape and thereby altering cellular mechanics [14], although more recent follow-up studies conducted in a separate laboratory suggest that adipogenic potential is not dependent on cell geometry [15]. Previously reported data obtained in our laboratory and others indicate that alterations in cell shape and cytoskeletal dynamics are capable of markedly overriding external mitogenic signalling [12, 16]. This suggests that, as opposed to a model in which cell proliferation, death and differentiation are largely independent of cell shape, these processes are coordinately regulated and modulated by cellular mechanics. Thus, the local differentials in growth factors, biochemistry and internal and external mechanical stress synergize to modulate the specificity that drives tissue heterogeneity during development, normal function and disease.

Cell shape changes have been associated with nuclear shape remodelling [11, 17-19]. It has been hypothesized that the transduction of mechanical information through cytoskeletal/nuclear coupling results in alterations that modulate chromosomal architecture and subsequent accession of transcription factors to their target genes [20-24]. Indeed, recent work has demonstrated that large-scale changes in cell shape induce alterations in chromosome condensation leading to marked effects on cell proliferation [25]. Thus, distinct cellular morphologies may drive the patterning of unique cytoskeletal architectures that govern global gene expression [26]. Despite these findings, it is not known how cell shape and its effects on cytoskeletal structures modulate global transcriptional patterns.

Although the normal surface of arteries is smooth, atherosclerotic arteries are characterized by irregular arrangement of endothelial cells, compromised monolayer integrity, irregular protrusions in the shape of scales or plates, and altered endothelial cell geometry [27, 28]. Thus an understanding of how endothelial cell shape changes affect cellular function may shed light on the deregulation of endothelial cells during aberrant states such as hypertension, arteriosclerosis and coronary artery disease. In the present study, we examined the global gene expression changes that occur when human coronary artery endothelial cells (HCAECs) are shape and spread restricted by micropatterning into reproducibly unique cellular morphologies that are distinctive in polarization, morphological angularity and actin cytoskeleton patterning. Given the wealth of data suggesting that cell shape and cytoskeletal patterning can alter cellular physiology across a large number of cell types, we specifically investigated whether unique alterations in these cellular properties are capable of modulating global gene expression changes in endothelial cells. Our data demonstrate that geometric restriction induces dramatic alterations in the HCAEC transcriptome, although these changes are independent of the exact cell shape and/or actin orientation assumed by the cell.


Quantitative analysis of cell shape-induced cytoskeletal and nuclear changes in HCAECs

To determine how cell shape alterations regulate the endothelial transcriptome, we must first utilize a system that manipulates cellular morphology at the same time as consistently maintaining all other growth variables. Accordingly, we seeded HCAECs on collagen I-coated spatially defined micropatterns, allowing cells to adapt to reproducible large (1600 μm2) geometric patterns, including a disc, crossbow, H, Y and L (Fig. 1A). We specifically utilized this cell type because endothelial cells of the coronary artery are constantly exposed to cyclic mechanical changes from blood flow during normal physiological processes and disrupted mechanical changes as a result of abnormal blood flow, cell shape deformation and loss of polarization during coronary artery disease. The size of the micropattern was specifically chosen because we tested micropatterns restricting the cells to either 700 or 1100 μm2; however, at these sizes, the cells failed to reproducibly conform to the intended shape (data not shown). Moreover, larger micropatterns would allow multiple cells to attach to one micropattern, thus dramatically affecting reproducibility of cell shape. As a control for nonrestricted morphology, cells were also plated at subconfluent levels on the chip in an area coated in collagen I. These patterns were specifically chosen for their ability to alter cell polarization (because this affects stress fibre architecture and nuclear orientation) [25] and the angularity of the cells' morphologies. Disk-shaped cells adopted a round morphology with obtuse cellular edges and random polarization. Crossbow and H-shaped cells exhibit a combination of obtuse and acute edges and become strongly and moderately polarized, respectively. Y- and L-shaped cells were dominated by acute angles, with strong polarization in the Y-shaped cells and no polarization in the L-shaped cells.

Figure 1.

HCAEC growth on micropatterned substrates. (A) Representative bright field images of HCAECs grown on a nonrestrictive collagen I-coated substrate (normal-shaped) or on collagen I-coated micropatterns (Y shape is represented). Although the micropatterned cells were spatially restricted, HCAECs were seeded at low densities in the nonrestricted controls to ensure minimization of cell-to-cell contacts. (B) Representative immunofluorescence (IF) and surface rendering (RF) images of nonrestricted and micropatterned (H, X-bow, disc, L and Y) HCAECs stained for actin (red), phospho-focal adhesion kinase (green) and nuclei (blue).

To quantify how endothelial cell shape drives actin cytoskeleton patterning, we performed immunofluorescent confocal imaging of normally-shaped and micropatterned HCAECs labelled with rhodamine-conjugated phalloidin (which stains the actin cytoskeleton), phosphorylated focal adhesion kinase (which stains cellular attachments to the extracellular matrix) and 4′,6-diamidino-2-phenylindole (DAPI) (which highlights the nucleus) (Fig. 1B). For a full understanding of the quantitative differences in the actin cytoskeletal orientation of each immunofluorescent image, we implemented algorithms to separate the structures of interest from the remainder of the image, thus allowing us to describe the image quantitatively rather than using standard qualitative methods. Accordingly, we employed techniques for linear feature extraction to segment and obtain orientation and length of the actin fibres from each image. The techniques included preprocessing the images to enhance foreground elements, actin fibre detection using fiberscore [29] and filtering including thresholding and mathematical morphology (Fig. 2). Figure 2B provides a more suitable input image to fiberscore for detection because the actin fibres are brighter and display higher contrast. Figure 2C,D shows the correlation and orientation outputs of fiberscore and is used for further analysis of length and orientation, respectively.

Figure 2.

Cytoskeletal image processing. Actin cytoskeleton images were processed as described in the Materials and methods. The processed images for a X-bow-shaped cell are shown. (A) Original immunofluorescence image in greyscale. (B) Preprocessing: contrast-limited adaptive histogram equalization. (C) Correlation image result from detection with fiberscore. (D) Orientation image result from detection with fiberscore. (E) Postprocessing: threshold. (F) Postprocessing: skeleton.

We first statistically analyzed the actin fibre orientations using images similar to those shown in Fig. 2D to quantitatively illustrate that HCAECs conforming to one micropatterned shape are indeed unique in cytoskeletal organization compared to those of another micropatterned shape. Immunofluorescent actin images from each shape were tiled into grid regions and the two-sample Kolmogorov–Smirnov (KS) test [30] was utilized to determine whether fibre orientation between cell shapes is truly unique and reproducible in structure (Table 1). High scores (closer to 1.0) occur when actin fibre orientations are largely dissimilar between cells and were observed across shape to shape comparisons. With the exception of normal-shaped cells (which demonstrated high actin orientation variability), we find relatively low rejection scores when comparing all the individual cells with their underlying cumulative tiling, meaning that cells of the same shape have fibre orientations more similar to each other than to other shapes. Note that these comparisons are not symmetric (e.g. comparing X-bow to disk yields slightly different scores than disk to X-bow). This asymmetry is a result of the fact that the orientation of individual images is being compared to the cumulative histogram of a specific shape; we are thus comparing individual disk image fibre orientations with the cumulative X-bow orientations and vice versa. These comparisons will yield similar but not identical results. These findings strongly validate the idea that the cellular morphologies induced by the micropatterned substrates result in reproducibly unique actin orientations between cell shapes. The analysis of shapes using the tiled grid regions, however, shows similarities in certain regions of a cell between shapes. Detailed analysis of the dominant and second dominant angles in actin orientation between cell shapes revealed that (a) crossbow-shaped cells have more contribution from actin angles close to 0° along the horizontal projection and angles oriented in opposite directions when comparing the widest regions of the crossbow with the narrowest regions; (b) disk-shaped cells have a more uniform distribution of actin angle orientations; (c) H-shaped cells have more contributions from angles close to 0° along the vertical centre; and (d) Y- and L-shaped cells display non-uniform orientation distributions each with a different dominant angle (Fig. 3A–C). This 3 × 3 tiling is applied in the same manner to all images; the consistency in the KS test results indicate the robustness of the results with respect to this choice of tiling. Note that in regions where it appears that there are no fibres and thus no orientation information (e.g. Y-shape left upper and lower corners) as a result of image variation, we do obtain a small amount of orientation information, as shown in Fig. 3C. We then analyzed the median fibre length using images similar to Fig. 3C between normal and micropatterned HCAECs using the previously described modified fiberscore analysis. As indicated in Fig. 3D, the median fibre length (± SEM) for normal-shaped HCAECs was significantly greater (6.84 ± 0.9 μm) than for crossbow- (2.9 ± 0.1 μm), disk- (3.3 ± 0.2 μm), H- (2.6 ± 0.2 μm), Y- (2.9 ± 0.3 μm) and L- (4.3 ± 0.4 μm) shaped cells. Thus, these data strongly indicate that actin orientation and length are truly unique between each cell shape and, if genomic alterations are truly shape and actin confirmation dependent, this model system is sufficient in both design and reproducibility to identify those changes.

Table 1. Correlation of actin fibre orientation between each shape. The data presented are the mean scores of the output via a two-sample KS test (scale of 0 to 1 where 1 completely rejects the null hypothesis of the test)
 CrossbowDiskH-cellY-cellL-cellNormal cell
Normal cell0.9510.890.950.880.86
Figure 3.

Quantification of actin fibre orientation and length in normal and micropatterned HCAECs. (A) fiberscore: orientation heatmaps depicting actin orientation for normal and micropatterned HCAECs in relation to their cellular axis. An angle starting at 0° is coincident with the x-axis and increases in a counter clockwise direction to 180°. (B) Representative 3 × 3 tiling of a crossbow-shaped HCAEC orientation heatmap. (C) Dominant and second dominant fibre angles in each grid block for each shape. The dominant and second dominant angles are calculated by combining angle information in all images of the particular shape tile. Angles are detected using the restricted angle resolution of fiberscore. (D) fiberscore: correlation quantification of the median fibre length of the actin fibres from normal and micropatterned HCAECs. At least 11–14 images were analyzed from each condition.

Using shape-engineered endothelial cells on circular, square and various rectangular adhesive micropatterns mimicking elongated bipolar shapes, Versaevel et al. [25] indicated that cell elongation and spreading is a key parameter of nuclear deformation and this process is absolutely dependent on lateral compressive forces generated by an actomyosin-mediated mechanism. It was further demonstrated that cell elongation leads to successive changes in the level of chromatin condensation as the nuclear shape index is decreased. To test whether changes in cell shape in general (as opposed to solely cell elongation, as shown previously) [25] induce nuclear deformation, we analyzed top and side images of the nuclei from normal and micropatterned HCAECs using confocal microscopy (at least 40 nuclei per condition). The prototypical HCAEC nucleus is ~ 15–18 μm long by 5–8 μm high and maintains a distinctive oval appearance (Fig. 4A, left), whereas deformed nuclei show variability from this norm, as shown in Fig. 4A (middle and right). Although irregularity in nuclear shape occurred relatively infrequently in normal-shaped cells (~ 6% of the cells exhibited nonprototypical nuclei), the percentages were significantly higher in the micropatterned HCAECs, ranging from just over 20% of the L- and Y shaped cells to approximately three-quarters of the population in disc shaped cells (Fig. 4B).

Figure 4.

Micropatterning of HCAECs increases the incidence of nuclear deformation. (A) Confocal top and side images of DAPI-stained HCAEC nuclei. The prototypical normal nucleus is shown in the left panel, whereas examples of deformed nuclei are shown in the middle and right panels. (B) Percentage of the cell population exhibiting a deformed nucleus. At least 40 nuclei were counted for each condition.

Morphological restriction in HCAECs results in large-scale changes in endothelial global gene transcription independent of the unique shape adopted

Distinct micropattern-mediated alterations in cell shape have been shown to affect lineage specification in mesenchymal progenitor cells [14], although less is known regarding how changes in cell morphology affect terminally differentiated cell types (such as an endothelial cells). Thus, we sought to address two questions: (a) does morphological restriction affect endothelial global transcription and (b) does a distinct cellular morphology uniquely affect endothelial global transcription. Using the reproducible micropatterning system described above, we can effectively address both questions.

We performed whole genome microarray analysis on total RNA collected from nonrestricted and micropatterned HCAECs cultured on 96-well collagen I-coated micropatterned plates and grown in standard growth media. The nonrestricted cells were grown at low confluence to minimize cell-to-cell contacts. Our data revealed large-scale alterations in gene expression as a result of HCAEC morphological restriction. As shown in Fig. 5A,B and Table 2 statistically relevant gene expression changes were equal or greater than two-fold in magnitude (P < 0.05) in at least one of the cell shapes compared to normally-shaped HCAECs cells cultured on the same micropatterned plate. The complete data set is publically available via the Gene Expression Omnibus ( (accession number GSE43349). These results provide strong evidence that restricting cell shape induces changes in the global transcriptional patterns of endothelial cells.

Table 2. Two-fold or greater alterations in gene expression compared to normal-shaped coronary artery endothelial cells (standard growth media)
Gene symbolGene nameAccession numberX-bowDiscHLY
TMEM100Transmembrane protein 100, TV2 NM_018286.2 6.456.36.47
PTGS2Prostaglandin-endoperoxide synthase 2 NM_000963.1 4.33.743.94.1
IRF6Interferon regulatory factor 6 NM_006147.2
ALPLAlkaline phosphatase, liver/bone/kidney, TV1 NM_000478.3
C8orf4Chromosome 8 ORF 4 NM_020130.3
HEY1Hairy/enhancer-of-split related with YRPW motif 1, TV2 NM_001040708.1
BMFBcl2 modifying factor, TV2 NM_033503.3 33.332.93
BMFBcl2 modifying factor, TV4 NM_001003943.1
LOC730525Hypothetical protein XM_001126202.1 32.82.734
SEMA3GSemaphorin 3G NM_020163.1
HSD17B11Hydroxysteroid (17β) dehydrogenase 11 NM_016245.2
F2RL3Coagulation factor II (thrombin) receptor-like 3 NM_003950.2
TOX2TOX high mobility group box family member 2, TV4 NM_001098796.1
C20orf100TOX high mobility group box family member 2 NM_032883.1
TOX2TOX high mobility group box family member 2, TV1 NM_001098797.1
SPRY1Sprouty homologue 1, antagonist of FGF signalling (Drosophila), TV1 NM_005841.1
SPRY1Sprouty homologue 1, antagonist of FGF signalling (Drosophila), TV2 NM_199327.1
ZBTB16Zinc finger and BTB domain containing 16, TV2 NM_001018011.1
TMEM140Transmembrane protein 140 NM_018295.2
NPTX1Neuronal pentraxin I NM_002522.2
SMAD7SMAD family member 7 NM_005904.2
ANKRD1Ankyrin repeat domain 1 (cardiac muscle) NM_014391.2
CXCR4Chemokine (C-X-C motif) receptor 4, TV1 NM_001008540.1
SYNMSynemin, intermediate filament protein, TVB NM_015286.5
HLXH2.0-like homeobox NM_021958.2
EFNB2Ephrin-B2 NM_004093.2
TNFAIP8L3Tumour necrosis factor, α-induced protein 8-like 3 NM_207381.2
NEDD9Neural precursor cell expressed, develop. down-regulated 9 , TV2 NM_182966.2
GDF15Growth differentiation factor 15 NM_004864.1
CALCRLCalcitonin receptor-like NM_005795.4 2.11.821.72
RDXRadixin, TV3 NM_002906.3
MMP10Matrix metallopeptidase 10 (stromelysin 2) NM_002425.1
CMTM8CKLF-like MARVEL transmembrane domain containing 8 NM_178868.3 221.81.82
C13orf15Regulator of cell cycle NM_014059.2
NDRG4NDRG family member 4 NM_022910.1
LOC100132564Hypothetical protein XM_001713808.1
CRYABCrystallin, alpha B NM_001885.1
RRAGDRas-related GTP binding D NM_021244.3
IL10Interleukin 10 NM_000572.2
LOC100129211Hypothetical protein XM_001718981.1
GRAPGRB2-related adaptor protein NM_006613.3
C8orf45Chromosome 8 open reading frame 45 NM_173518.2
PDGFBPlatelet-derived growth factor beta (oncogene homolog), TV1 NM_002608.1
LOC100190986Nuclear pore complex interacting protein pseudogene NR_024456.1
PGFPlacental growth factor NM_002632.4 1.821.61.51.6
LOC100132247Nuclear pore complex interacting protein related gene NM_001135865.1
FAM175AFamily with sequence similarity 175, member A NM_139076.2
PDGFBPlatelet-derived growth factor beta (oncogene homolog), TV2 NM_033016.1
LOC440353Nuclear pore complex interacting protein pseudogene NR_002603.1
KIAA1751KIAA1751 NM_001080484.1
LOC613037Nuclear pore complex interacting protein pseudogene NR_002555.2 1.61.421.52.1
MAGT1Magnesium transporter 1 NM_032121.4
ZNF738Misc_RNA, partial miscRNA XR_040185.1
DMC1DMC1 dosage suppressor of mck1 homolog NM_007068.2
LOC729978Similar to LOC339047 protein, TV2 XM_001723016.1
LOC23117KIAA0220-like protein, TV16 XM_933834.2
LOC100132585Hypothetical protein XM_001722111.1
LOC440348Nuclear pore complex interacting protein-like 2NM_001018059.
LOC440345Hypothetical protein, TV6 XM_933717.1
LOC728809Hypothetical LOC728809 XM_001719546.1
TRIM13Tripartite motif containing 13, TV4 NM_001007278.1
IMAGE:2760091 3NCI_CGAP_Lu28 Homo sapiens cDNA clone IMAGE:2760091 3 AW276479
CATSPER2Cation channel, sperm associated 2, TV4 NM_172097.1
MCART1Mitochondrial carrier triple repeat 1 NM_033412.1
NLRP8NLR family, pyrin domain containing 8 NM_176811.2
LOC255167Uncharacterized LOC255167 NR_024424.1
DDX51DEAD (Asp-Glu-Ala-Asp) box polypeptide 51 NM_175066.2
C21orf55Chromosome 21 ORF 55 NM_017833.2
LOC90586Amine oxidase, copper containing 3 pseudogene NR_002773.1
LOC100130168Hypothetical protein XM_001719127.1
MAPK8IP3Mitogen-activated protein kinase 8 interacting protein 3, TV2 NM_001040439.1 1.421.61.61.7
ZNF682Zinc finger protein 682, TV1 NM_033196.2
ZNF486Zinc finger protein 486 XM_371152.3 1.31.411.22.1
SULT1A1Sulfotransferase family, cytosolic, 1A, phenol-preferring, member 1, TV3 NM_177530.1
LOC100128510Hypothetical protein XM_001715759.1
LOC653994Similar to eukaryotic translation initiation factor 4H, TV2 XM_944429.1 −1.3−2.2−1.3−1.51
LOC648024Similar to eukaryotic translation initiation factor 4A, TV1 XR_018316.1 −1.3−1.8−1.6−2.1−1.3
NDUFA8NADH dehydrogenase (ubiquinone) 1α subcomplex, 8 NM_014222.2 −1.4−1.3−1.7−2.0−1.4
TNPO1Transportin 1, TV2 NM_153188.2 −1.4−2.3−1.5−1.7−1.3
SNAP23Synaptosomal-associated protein, 23 kDa, TV1 NM_003825.2 −1.4−2.0−1.4−1.6−1.3
TCEA1Transcription elongation factor A (SII), TV2 NM_201437.1 −1.4−2.0−1.6−1.8−1.4
ALCAMActivated leukocyte cell adhesion molecule NM_001627.2 −1.4−1.4−1.6−2.0−1.8
TCEAL8Transcription elongation factor A (SII)-like 8, TV2 NM_001006684.1 −1.4−1.5−1.7−2.0−1.4
TMEM189-UBE2V1TMEM189-UBE2V1 readthrough transcript, TV2 NM_003349.4 −1.4−2.0−1.5−1.7−1.3
LOC730052Misc_RNA (LOC730052) XR_016054.2 −1.4−2.0−1.4−1.7−1.3
TXNDC5Thioredoxin domain containing 5 (endoplasmic reticulum), TV1 NM_030810.2 −1.4−2.0−1.4−1.5−1.6
BCL2L1BCL2-like 1, nuclear gene encoding mitochondrial protein, TV1 NM_138578.1 −1.5−2.0−1.3−1.6−1.3
EIF4G2Eukaryotic translation initiation factor 4γ, 2, TV1 NM_001418.3 −1.5−2.2−1.5−2.0−1.5
TCP1T-complex 1, TV1 NM_030752.2 −1.5−2.1−1.4−1.8−1.4
CCT6AChaperonin containing TCP1, subunit 6A (zeta 1), TV1 NM_001762.3 −1.5−2.1−1.5−1.9−1.4
LOC644063Similar to heterogeneous nuclear ribonucleoprotein K XR_016547.1 −1.5−2.2−1.6−2.0−1.3
LSM5LSM5 homologue, U6 small nuclear RNA associated (Saccharomyces cerevisiae) NM_012322.1 −1.5−1.6−2.0−1.9−1.8
FEZ2Fasciculation and elongation protein zeta 2 (zygin II), TV1 NM_005102.2 −1.5−1.7−1.8−2.1−1.6
C14orf149Chromosome 14 ORF 149 NM_144581.1 −1.5−1.6−1.8−2.0−1.9
LOC728059Misc_RNA XR_015606.1 −1.5−2.4−1.6−2.3−1.7
THOC4THO complex 4 XM_001134346.1 −1.5−1.8−1.6−2.0−1.5
LYPLA1Lysophospholipase I NM_006330.2 −1.5−1.9−1.8−2.1−1.5
EDG1Endothelial differentiation, sphingolipid G-protein-coupled receptor, 1 NM_001400.3 −1.5−1.5−1.5−2.2−1.6
LOC648695Similar to retinoblastoma binding protein 4, TV5 XM_944246.2 −1.5−2.2−1.8−2.2−1.7
MALLMal, T-cell differentiation protein-like NM_005434.3 −1.5−1.3−1.7−2.0−1.7
ZFAND6Zinc finger, AN1-type domain 6 NM_019006.2 −1.5−2.2−1.6−1.8−1.6
ADKAdenosine kinase, transcript variant ADK-short NM_001123.2 −1.5−1.6−1.9−2.0−1.5
ZYXZyxin, TV1 NM_003461.4 −1.5−1.4−2.0−1.7−1.7
PAPSS23′-phosphoadenosine 5′-phosphosulfate synthase 2, TV1 NM_004670.3 −1.5−1.5−1.7−2.1−1.5
G3BP2GTPase activating protein (SH3 domain) binding protein 2, TV3 NM_203504.1 −1.5−1.6−1.6−2.1−1.6
LOC100130561Similar to high-mobility group protein 1-like 10, TV2 XM_001723189.1 −1.5−2.1−1.6−1.9−1.5
HIGD1AHIG1 hypoxia inducible domain family, member 1A, TV1 NM_001099668.1 −1.6−2.0−1.8−2.1−1.7
EPB41L3Erythrocyte membrane protein band 4.1-like 3 NM_012307.2 −1.6−1.7−1.6−2.0−1.8
IARSIsoleucyl-tRNA synthetase, TV short NM_002161.3 −1.6−1.5−1.7−2.0−1.5
RRAS2Related RAS viral (r-ras) oncogene homologue 2 NM_012250.3 −1.6−1.9−1.8−2.0−1.3
RANBP1RAN binding protein 1 NM_002882.2 −1.6−1.6−2.1−2.0−1.7
NOL6Nucleolar protein family 6 (RNA-associated), TV γ NM_139235.3 −1.6−1.3−2.0−1.7−1.5
C18orf55Chromosome 18 ORF 55 NM_014177.1 −1.6−1.7−1.7−2.1−1.6
CSE1LCSE1 chromosome segregation 1-like (yeast) NM_001316.2 −1.6−1.6−1.8−2.0−1.6
TIMM23Translocase of inner mitochondrial membrane 23 homologue NM_006327.2 −1.6−1.7−1.8−2.1−1.6
FHL2Four and a half LIM domains 2, TV4 NM_201557.2 −1.6−1.6−1.8−2.0−1.6
AP1S1Adaptor-related protein complex 1, sigma 1 subunit, TV4 NM_057089.2 −1.6−1.5−1.9−2.1−1.4
HNRPA2B1Heterogeneous nuclear ribonucleoprotein A2/B1, TV B1 NM_031243.1 −1.6−1.5−1.8−2.0−1.8
CCNCCyclin C, TV2 NM_001013399.1 −1.6−2.0−1.7−1.9−1.5
PTPLAD1Protein tyrosine phosphatase-like A domain containing 1 NM_016395.2 −1.6−1.7−1.8−2.1−1.5
HNRNPKHeterogeneous nuclear ribonucleoprotein K, TV2 NM_031263.2 −1.6−1.8−1.5−2.0−1.5
HAT1Histone acetyltransferase 1, TV1 NM_003642.2 −1.6−1.7−2.0−1.9−1.7
PSME3Proteasome (prosome, macropain) activator subunit 3, TV1 NM_005789.2 −1.6−1.3−2.0−2.0−1.6
HIGD1AHIG1 hypoxia inducible domain family, member 1A, TV1 NM_001099668.1 −1.6−1.7−1.8−2.1−1.6
ARMCX3Armadillo repeat containing, X-linked 3, TV2 NM_177947.2 −1.6−2.0−1.5−2.0−1.6
LOC100128266PREDICTED: Misc_RNA XR_038984.1 −1.6−2.1−1.8−2.1−1.7
DCBLD2Discoidin, CUB and LCCL domain containing 2 NM_080927.3 −1.6−1.6−2.1−1.9−1.8
SMSSpermine synthase NM_004595.2 −1.6−1.9−1.8−2.0−1.7
TPM3Tropomyosin 3, TV1 NM_152263.2 −1.6−1.9−1.8−2.1−1.4
LOC653884Similar to FUS interacting protein (serine-arginine rich) 1 XM_936240.1 −1.6−1.9−1.6−2.0−1.5
ATP5G1ATP synthase, mitochondrial Fo complex, subunit C1, TV2 NM_001002027.1 −1.6−1.5−1.9−2.0−1.7
SDCBPSyndecan binding protein (syntenin), TV2 NM_001007067.1 −1.6−2.0−1.6−2.1−1.7
MCM6Minichromosome maintenance complex component 6 NM_005915.4 −1.6−1.7−2.0−2.0−1.7
BRIX1BRX1, biogenesis of ribosomes, homologue (S. cerevisiae) NM_018321.3 −1.6−1.7−1.7−2.0−1.7
RPL29Ribosomal protein L29 NM_000992.2 −1.6−2.1−2.0−2.2−1.6
LOC644330Similar to tropomyosin 3 isoform 2 XR_017492.1 −1.6−2.0−1.8−1.9−1.6
LPXNLeupaxin NM_004811.1 −1.6−1.6−1.9−2.0−1.8
LOC100130506Hypothetical protein XM_001724500.1 −1.6−1.7−2.0−2.0−1.7
DDX21DEAD (Asp-Glu-Ala-Asp) box polypeptide 21 NM_004728.2 −1.6−1.6−1.7−2.0−1.5
LDHALactate dehydrogenase A NM_005566.1 −1.6−1.6−1.9−2.0−1.8
LOC642590Misc_RNA XR_016251.2 −1.6−1.7−2.0−1.9−1.7
FKBP14FK506 binding protein 14, 22 kDa NM_017946.2 −1.6−1.7−1.9−2.1−1.6
NME1NME/NM23 nucleoside diphosphate kinase 1, TV2 NM_000269.2 −1.6−1.8−2.1−2.1−1.7
AHNAKAHNAK nucleoprotein, TV1 NM_001620.1 −1.6−1.6−2.0−1.8−1.4
CKS2CDC28 protein kinase regulatory subunit 2 NM_001827.1 −1.6−1.5−2.0−2.1−1.7
CYCSL1Cytochrome c, somatic-like 1 on chromosome 6 NR_001561.1 −1.6−2.0−1.9−2.1−1.7
LOC646347Misc_RNA XR_017680.1 −1.7−1.9−1.9−2.1−1.8
WDR4WD repeat domain 4, TV2 NM_033661.3 −1.7−1.5−1.9−2.2−1.6
ALDH1A3Aldehyde dehydrogenase 1 family, member A3 NM_000693.1 −1.7−1.7−2.0−2.4−1.9
CLINT1Clathrin interactor 1 NM_014666.2 −1.7−1.6−1.8−2.0−1.6
GNG12Guanine nucleotide binding protein (G protein), γ 12 NM_018841.4 −1.7−2.5−1.6−2.1−1.6
TOMM5Translocase of outer mitochondrial membrane 5 homologue, TV1NM_001001790.2−1.7−1.7−2.0−1.9−1.9
MPZL2Myelin protein zero-like 2, TV1 NM_005797.2 −1.7−1.8−1.9−2.2−1.7
DUSP14Dual specificity phosphatase 14 NM_007026.2 −1.7−1.6−1.9−2.1−1.9
IDH1Isocitrate dehydrogenase 1 (NADP+), soluble NM_005896.2 −1.7−2.0−1.7−1.9−1.8
CYTL1Cytokine-like 1 NM_018659.2 −1.7−1.8−2.0−2.1−1.6
MLKLMixed lineage kinase domain-like NM_152649.1 −1.7−1.6−1.8−2.0−1.6
CTHRC1Collagen triple helix repeat containing 1 NM_138455.2 −1.7−1.8−1.8−2.0−1.6
C6orf173Chromosome 6 ORF 173 NM_001012507.1 −1.7−1.6−2.1−1.8−1.8
MGC40489Hypothetical protein XR_016048.1 −1.7−1.8−1.7−2.0−1.7
KDELR3KDEL endoplasmic reticulum protein retention receptor 3, TV1 NM_006855.2 −1.7−1.6−1.8−1.8−2.0
TNFSF4Tumour necrosis factor (ligand) superfamily, member 4 NM_003326.2 −1.7−1.7−1.7−2.1−1.9
AURKAAurora kinase A, TV5 NM_198436.1 −1.7−1.7−2.1−1.9−1.9
SMSSpermine synthase NM_004595.2 −1.7−2.0−1.8−2.0−1.7
RND3Rho family GTPase 3 NM_005168.3 −1.7−1.6−2.0−1.9−1.7
CLDN5Claudin 5 (transmembrane protein deleted in velocardiofacial syndrome) NM_003277.2 −1.7−1.4−1.8−2.1−1.9
EDN1Endothelin 1 NM_001955.2 −1.7−1.7−2.0−1.7−1.7
PVRL3Poliovirus receptor-related 3 NM_015480.1 −1.7−1.6−2.0−2.1−1.8
LOXLysyl oxidase NM_002317.3 −1.7−1.9−1.9−2.0−1.6
ICMTIsoprenylcysteine carboxyl methyltransferase NM_012405.3 −1.7−1.6−2.0−1.8−1.7
PRDX3Peroxiredoxin 3, nuclear gene encoding mitochondrial protein, TV1 NM_006793.2 −1.7−1.9−1.8−2.1−1.6
TUBB6Tubulin, β6 class V NM_032525.1 −1.7−2.2−1.6−1.8−1.6
VAMP5Vesicle-associated membrane protein 5 (myobrevin) NM_006634.2 −1.7−1.9−1.9−2.1−1.6
MORF4L2Mortality factor 4-like 2 NM_012286.1 −1.7−1.8−1.9−2.2−1.7
NOP56NOP56 ribonucleoprotein homologue (yeast), TV1 NM_006392.2 −1.7−1.8−1.9−2.1−1.7
HNRPKHeterogeneous nuclear ribonucleoprotein K, TV3 NM_031263.1 −1.7−1.8−1.8−2.0−1.9
RNF121Ring finger protein 121, TV1 NM_018320.3 −1.7−1.4−2.0−2.0−1.8
KDELC2KDEL (Lys-Asp-Glu-Leu) containing 2 NM_153705.4 −1.7−1.7−1.9−2.2−1.8
FJX1Four jointed box 1 (Drosophila) NM_014344.2 −1.7−1.6−2.0−2.1−1.9
DNMT1DNA (cytosine-5-)-methyltransferase 1 NM_001379.1 −1.7−1.4−1.8−2.0−1.7
LOC729779Misc_RNA (LOC729779) XR_019592.2 −1.7−2.0−1.8−1.6−1.7
FABP5Fatty acid binding protein 5 (psoriasis-associated) NM_001444.1 −1.7−1.6−1.9−2.0−1.7
ZDHHC6Zinc finger, DHHC-type containing 6 NM_022494.1 −1.7−1.8−1.9−2.2−1.7
IL1RL1Interleukin 1 receptor-like 1 (IL1RL1), TV2 NM_003856.2 −1.7−1.9−1.8−2.0−1.7
EBNA1BP2EBNA1 binding protein 2 NM_006824.1 −1.7−1.8−2.1−2.1−1.6
TFDP1Transcription factor Dp-1 NM_007111.3 −1.7−1.6−1.8−2.1−1.7
PAICSPhosphoribosylaminoimidazole succinocarboxamide synthetase, TV2 NM_006452.3 −1.7−1.7−2.0−2.2−1.6
CISD1CDGSH iron sulfur domain 1 NM_018464.2 −1.7−1.7−2.2−2.1−1.7
LOC100129086Similar to HIG1 domain family, member 1A XM_001725669.1 −1.7−2.1−2.0−2.1−1.7
POLE4Polymerase (DNA-directed), ε4, accessory subunit NM_019896.2 −1.8−1.8−2.0−2.0−1.6
FER1L3Fer-1-like 3, myoferlin (Caenorhabditis elegans), TV2 NM_133337.1 −1.8−1.5−1.9−2.0−1.9
PVRL3Poliovirus receptor-related 3 NM_015480.1 −1.8−1.9−2.2−2.2−1.9
RANBP1RAN binding protein 1 NM_002882.2 −1.8−2.1−2.2−2.3−1.8
RAB11ARAB11A, member RAS oncogene family NM_004663.3 −1.8−1.5−1.8−2.0−1.7
SLC38A1Solute carrier family 38, member 1, TV1 NM_030674.3 −1.8−1.7−2.1−2.0−2.0
IL8Interleukin 8 NM_000584.2 −1.8−1.9−2.1−2.0−1.7
LOC100132715Misc_RNA XR_039129.1 −1.8−1.5−1.9−2.0−1.6
LOC644330Similar to tropomyosin 3 isoform 2 XR_017492.1 −1.8−2.4−2.0−2.1−1.7
ZNF185Zinc finger protein 185 (LIM domain) NM_007150.2 −1.8−1.6−1.9−2.0−1.7
COL13A1Collagen, type XIII, α1, TV19 NM_080815.2 −1.8−1.6−2.1−2.0−1.8
PKD2Polycystic kidney disease 2 (autosomal dominant) NM_000297.2 −1.8−1.6−1.9−2.0−2.0
MAGED1Melanoma antigen family D, 1, TV2 NM_006986.3 −1.8−1.7−1.9−1.9−2.3
POLE3Polymerase (DNA directed), ε3 (p17 subunit) NM_017443.3 −1.8−1.6−2.0−2.1−1.7
CORO1CCoronin, actin binding protein, 1C, TV1 NM_014325.2 −1.8−1.5−1.8−2.0−1.8
LOC652481Similar to mitochondrial import inner membrane translocase subunit Tim23 XM_941942.1 −1.8−2.2−1.9−1.9−1.7
SLFN11Schlafen family member 11 NM_152270.2 −1.8−1.4−2.0−1.9−1.8
PRNPPrion protein (PRNP), TV3 NM_001080121.1 −1.8−1.7−1.9−2.0−2.2
FRMD6FERM domain containing 6 NM_152330.2 −1.8−1.8−2.1−2.2−1.9
PTS6-pyruvoyltetrahydropterin synthase NM_000317.1 −1.8−1.8−1.9−2.0−1.5
PECIEnoyl-CoA δ isomerase 2 (ECI2), TV1 NM_006117.2 −1.8−2.4−2.2−2.5−1.9
MGAT2Mannosyl-glycoprotein-acetylglucosaminyltransferase, TV2 NM_001015883.1 −1.8−2.1−1.6−2.1−1.6
ATP6V0E2ATPase, H+ transporting V0 subunit e2, TV1 NM_145230.2 −1.8−1.5−2.0−1.9−1.8
RPL6Ribosomal protein L6, TV1 NM_001024662.1 −1.8−2.0−1.9−2.2−1.8
CGNL1Cingulin-like 1 NM_032866.3 −1.8−1.8−2.2−2.3−2.0
LDHALactate dehydrogenase A, TV2 NM_001135239.1 −1.8−1.8−2.0−2.1−1.9
PGK1Phosphoglycerate kinase 1 NM_000291.2 −1.8−1.9−1.9−2.2−1.8
CCND3Cyclin D3 NM_001760.2 −1.8−1.6−2.0−2.0−2.0
SFRS2Serine/arginine-rich splicing factor 2 NM_003016.3 −1.8−1.7−2.3−2.2−1.9
F2RL1Coagulation factor II (thrombin) receptor-like 1 NM_005242.3 −1.8−1.8−1.8−2.2−1.8
PLSCR4Phospholipid scramblase 4 NM_020353.1 −1.8−1.7−1.8−2.1−1.6
KDELR3KDEL endoplasmic reticulum protein retention receptor 3, TV2 NM_016657.1 −1.8−2.0−2.0−2.2−1.9
LOC653226Similar to signal recognition particle 9 kDa protein (SRP9) XM_927451.2 −1.8−2.2−1.8−2.0−1.5
LOC387882Hypothetical protein NM_207376.1 −1.8−1.8−2.1−2.1−1.7
PPM1FProtein phosphatase, Mg2+/Mn2+ dependent, 1F NM_014634.2 −1.8−1.4−1.9−2.1−1.7
PRICKLE1Prickle homologue 1 (Drosophila) NM_153026.1 −1.8−1.4−2.0−1.7−1.8
TSPAN5Tetraspanin 5 NM_005723.2 −1.8−1.7−2.0−2.2−1.6
PDCD6IPProgrammed cell death 6 interacting protein NM_013374.3 −1.8−1.7−1.8−2.2−1.9
EFEMP1EGF-containing fibulin-like extracellular matrix protein 1 NM_004105.3 −1.8−3.0−1.8−2.2−1.7
CDC20Cell division cycle 20 homologue (S. cerevisiae) NM_001255.2 −1.8−1.9−2.1−1.8−2.0
LOC642590Misc_RNA XR_037021.1 −1.8−1.8−1.8−2.2−1.7
PRKAG2Protein kinase, AMP-activated, γ2 noncatalytic subunit, TVb NM_024429.1 −1.9−1.9−2.0−2.1−2.0
MRPL39Mitochondrial ribosomal protein L39, TV1 NM_017446.3 −1.9−1.9−1.9−2.2−1.7
TRAM2Translocation associated membrane protein 2 NM_012288.3 −1.9−1.5−2.0−2.1−1.8
B4GALT5UDP-Gal:βGlcNAc β 1,4- galactosyltransferase, polypeptide 5 NM_004776.2 −1.9−1.8−2.2−2.4−2.3
TUBA1ATubulin, α1a NM_006009.2 −1.9−2.0−1.8−2.3−1.9
KPNA2Karyopherin α2 (RAG cohort 1, importin α 1) NM_002266.2 −1.9−2.3−2.0−2.2−1.9
FER1L3Fer-1-like 3, myoferlin (C. elegans) (FER1L3), TV1 NM_013451.2 −1.9−1.9−1.8−2.0−2.0
NLGN1Neuroligin 1 NM_014932.2 −1.9−1.9−2.3−2.5−2.0
ALDH3A2Aldehyde dehydrogenase 3 family, member A2, TV2 NM_000382.2 −1.9−1.9−1.8−2.1−1.8
LOC732007Similar to phosphoglycerate mutase 1 XR_015684.1 −1.9−1.9−1.8−2.2−2.0
C21orf63Family with sequence similarity 176, member C NM_058187.3 −1.9−1.7−2.2−1.9−1.7
MSRB3Methionine sulfoxide reductase B3, TV1 NM_198080.2 −1.9−2.1−1.7−2.0−1.7
PLXNA2Plexin A2 NM_025179.3 −1.9−1.5−2.1−1.9−1.9
UCHL3Ubiquitin carboxyl-terminal esterase L3 (ubiquitin thiolesterase) NM_006002.3 −1.9−2.0−2.3−2.3−2.0
MT1GMetallothionein 1G NM_005950.1 −1.9−1.4−2.0−1.8−1.8
NEXNNexilin (F actin binding protein), TV1 NM_144573.3 −1.9−2.2−2.1−2.1−1.9
CRIM1Cysteine rich transmembrane BMP regulator 1 (chordin-like) NM_016441.1 −1.9−2.5−2.0−2.4−1.9
LOC644774Similar to phosphoglycerate kinase 1 XM_927868.1 −1.9−2.3−2.1−2.4−2.0
UBE2TUbiquitin-conjugating enzyme E2T (putative) NM_014176.2 −1.9−1.9−2.0−2.0−1.8
LOC441019Hypothetical LOC441019 XM_498969.2 −1.9−1.5−2.1−2.1−2.1
PGAM1Phosphoglycerate mutase 1 (brain) NM_002629.2 −1.9−2.4−1.8−1.9−1.6
LPHN2Latrophilin 2 NM_012302.2 −1.9−1.6−2.1−2.3−1.9
EHD4EH-domain containing 4 NM_139265.2 −1.9−1.6−1.9−2.0−1.7
MYOFMyoferlin, TV1 NM_013451.3 −1.9−1.8−1.9−2.2−1.9
PTTG1Pituitary tumour-transforming 1 NM_004219.2 −1.9−2.2−2.1−2.1−1.8
TUBA1CTubulin, α1c NM_032704.2 −1.9−2.0−1.9−2.2−1.9
ANXA2Annexin A2, TV2 NM_001002857.1 −1.9−2.8−2.2−2.5−1.8
FILIP1LFilamin A interacting protein 1-like, TV3 NM_001042459.1 −1.9−1.8−2.1−1.9−2.0
TRIP6Thyroid hormone receptor interactor 6 NM_003302.2 −1.9−1.8−2.1−1.9−1.9
GIMAP7GTPase, IMAP family member 7 NM_153236.3 −1.9−1.9−2.1−2.6−2.0
PECIEnoyl-CoA δ isomerase 2, TV1 NM_006117.2 −1.9−1.9−2.1−2.1−1.9
TMEM14ATransmembrane protein 14A NM_014051.3 −1.9−2.1−2.1−2.3−2.3
CALD1Caldesmon 1 (CALD1), TV5 NM_033140.2 −2.0−2.0−2.3−2.1−2.0
LOC402221Similar to actin α1 skeletal muscle protein XM_938988.1 −2.0−2.0−1.8−2.2−2.2
CCND2Cyclin D2 NM_001759.2 −2.0−1.7−1.9−2.1−2.1
PRNPPrion protein (PRNP), TV2 NM_183079.2 −2.0−2.0−2.2−2.2−2.4
FRMD6FERM domain containing 6, TV2 NM_152330.3 −2.0−2.0−2.0−2.1−1.9
EFHD2EF-hand domain family, member D2 NM_024329.4 −2.0−1.7−2.2−2.2−2.0
AADACL1Arylacetamide deacetylase-like 1 NM_020792.3 −2.0−2.3−2.2−2.4−2.2
TGM2Transglutaminase 2, TV1 NM_004613.2 −2.0−1.8−2.3−2.1−1.9
CAV2Caveolin 2 (CAV2), TV1 NM_001233.3 −2.0−2.7−2.3−2.6−2.0
NNMTNicotinamide N-methyltransferase NM_006169.2 −2.0−2.1−2.2−2.2−2.0
UAP1UDP-N-acteylglucosamine pyrophosphorylase 1 NM_003115.3 −2.0−1.6−2.2−2.0−1.9
TJP2Tight junction protein 2 (zona occludens 2), TV2 NM_201629.1 −2.0−1.8−2.2−2.0−2.0
AURKAAurora kinase A, TV3 NM_198434.1 −2.0−1.9−2.2−2.1−2.1
CSTF3Cleavage stimulation factor, 3′ pre-RNA, subunit 3, 77 kDa, TV2 NM_001033505.1 −2.0−2.2−2.1−2.3−1.9
PTPLAProtein tyrosine phosphatase-like, member A NM_014241.3 −2.0−1.9−2.1−2.3−2.0
CAV1Caveolin 1, caveolae protein, 22 kDa NM_001753.3 −2.0−2.0−2.3−2.3−1.9
EXT1Exostosin 1 NM_000127.2 −2.0−1.7−2.0−2.4−2.2
CCNA2Cyclin A2 NM_001237.2 −2.0−1.9−2.1−1.9−1.9
CD59CD59 molecule, complement regulatory protein, TV2 NM_000611.4 −2.0−1.5−2.0−2.1−2.1
TUBB2CTubulin, β4B class IVb NM_006088.5 −2.0−1.9−2.2−2.4−2.4
SFRS3Splicing factor, arginine/serine-rich 3 NM_003017.3 −2.0−2.0−2.1−2.2−2.0
RANRAN, member RAS oncogene family NM_006325.2 −2.0−2.2−2.3−2.4−2.0
ADAM9ADAM metallopeptidase domain 9, TV1 NM_003816.2 −2.0−2.8−2.0−2.3−1.9
LRP8Low density lipoprotein receptor-related protein 8, TV3 NM_017522.3 −2.0−1.9−2.2−2.2−2.2
MELKMaternal embryonic leucine zipper kinase NM_014791.2 −2.0−2.0−2.1−2.3−2.0
GALNT10Polypeptide N-acetylgalactosaminyltransferase 10, TV2 NM_017540.3 −2.0−1.9−1.9−2.0−1.9
CBX6Chromobox homologue 6 NM_014292.3 −2.0−1.6−2.3−2.5−2.0
CALM1Calmodulin 1 (phosphorylase kinase, δ) NM_006888.3 −2.0−1.7−2.3−2.4−2.1
PTTG1Pituitary tumour-transforming 1 NM_004219.2 −2.1−1.9−2.1−2.2−1.8
IL8Interleukin 8 NM_000584.2 −2.3−2.9−3.0−3.2−2.5
IL1RL1Interleukin 1 receptor-like 1, TV2 NM_003856.2 −2.1−2.0−2.5−2.5−1.9
FZD4Frizzled homologue 4 (Drosophila) NM_012193.2 −2.1−1.6−2.0−2.1−2.1
GLCEGlucuronic acid epimerase NM_015554.1 −2.1−2.0−2.5−2.7−2.4
UBE2CUbiquitin-conjugating enzyme E2C, TV6 NM_181803.1 −2.1−2.1−2.0−2.1−1.9
FAM176AFamily with sequence similarity 176, member A, TV1 NM_001135032.1 −2.1−2.1−2.1−2.2−2.2
ICAM2Intercellular adhesion molecule 2, TV1 NM_001099786.1 −2.1−2.1−2.4−2.8−2.1
TGM2Transglutaminase 2, TV2 NM_198951.1 −2.1−2.0−2.5−2.4−2.0
EPHA2EPH receptor A2 NM_004431.2 −2.1−1.7−2.0−2.1−2.0
FEN1Flap structure-specific endonuclease 1 NM_004111.4 −2.1−2.2−2.6−2.5−2.2
ATP1B1ATPase, Na+/K+ transporting, β1 polypeptide NM_001677.3 −2.1−2.0−2.1−2.6−2.1
ODZ3Odz, odd Oz/ten-m homologue 3 (Drosophila) NM_001080477.1 −2.1−1.9−2.3−2.5−2.1
FILIP1LFilamin A interacting protein 1-like, TV1 NM_182909.2 −2.1−1.8−2.1−2.1−1.9
NMT2N-myristoyltransferase 2 NM_004808.1 −2.1−2.1−2.4−2.5−2.3
PHACTR2Phosphatase and actin regulator 2, TV1 NM_001100164.1 −2.1−1.9−2.2−2.5−1.9
TUBA1BTubulin, α1b NM_006082.2 −2.1−1.9−2.4−2.1−2.3
C20orf127Chromosome 20 ORF 127 NM_080757.1 −2.1−1.8−2.7−2.6−2.0
NPFFR2Neuropeptide FF receptor 2, TV1 NM_004885.1 −2.1−2.1−2.3−2.2−2.3
LIMA1LIM domain and actin binding 1 NM_016357.3 −2.2−2.1−2.2−2.2−1.9
BASP1Brain abundant, membrane attached signal protein 1 NM_006317.3 −2.2−2.0−2.4−2.5−2.2
TNFRSF12ATumour necrosis factor receptor superfamily, member 12A NM_016639.1 −2.2−1.9−2.6−2.2−2.1
KRT7Keratin 7 NM_005556.3 −2.2−1.8−2.2−2.2−2.0
NCAPGNon-SMC condensin I complex, subunit G NM_022346.3 −2.2−2.1−2.2−2.3−2.4
CCNA1Cyclin A1 NM_003914.2 −2.2−2.4−2.4−2.5−2.5
DIO2Deiodinase, iodothyronine, type II, TV3NM_001007023.2−2.2−2.0−2.1−1.9−2.2
DDAH1Dimethylarginine dimethylaminohydrolase 1 NM_012137.2 −2.2−2.1−2.8−2.6−2.4
CAV1Caveolin 1, caveolae protein, 22 kDa NM_001753.3 −2.2−2.4−2.4−2.7−2.4
TYMSThymidylate synthetase NM_001071.1 −2.2−2.3−2.4−2.2−2.0
GRB14Growth factor receptor-bound protein 14 NM_004490.2 −2.2−2.1−2.5−2.4−2.1
CAV2Caveolin 2, TV1 NM_001233.3 −2.2−2.5−2.3−2.6−2.2
MGLLMonoglyceride lipase (MGLL), TV1 NM_007283.5 −2.2−1.8−2.1−2.2−2.1
FILIP1LFilamin A interacting protein 1-like, TV2 NM_014890.2 −2.2−1.8−2.6−2.5−2.1
CEP55Centrosomal protein 55 kDa NM_018131.3 −2.3−2.2−2.2−2.4−2.4
CALD1Caldesmon 1, TV3 NM_033157.2 −2.3−2.8−2.5−2.3−1.9
UBE2CUbiquitin-conjugating enzyme E2C, TV3 NM_181800.1 −2.3−2.4−2.4−2.7−2.2
MTEMetallothionein E NM_175621.2 −2.3−2.0−3.2−2.4−2.5
MCM4Minichromosome maintenance complex component 4, TV1 NM_005914.2 −2.3−2.2−2.5−2.6−2.3
FABP4Fatty acid binding protein 4, adipocyte NM_001442.1 −2.3−2.1−2.2−2.3−2.5
PLOD2Procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2, TV2 NM_000935.2 −2.3−2.2−2.6−2.8−2.5
TXNRD2Thioredoxin reductase 2, nuclear gene encoding mitochondrial protein NM_006440.3 −2.4−2.2−2.9−2.8−2.4
LDLRLow-density lipoprotein receptor (familial hypercholesterolaemia) NM_000527.2 −2.4−2.2−2.7−2.6−2.5
GIMAP4GTPase, IMAP family member 4 NM_018326.2 −2.5−2.2−2.6−2.9−2.8
PRC1Protein regulator of cytokinesis 1, TV2 NM_199413.1 −2.5−2.1−2.2−2.3−2.2
MGLLMonoglyceride lipase, TV1 NM_007283.5 −2.5−2.8−2.8−2.8−2.4
FKSG30Actin-like protein NM_001017421.1 −2.5−2.4−2.3−2.6−2.2
ALDH1A3Aldehyde dehydrogenase 1 family, member A3 NM_000693.2 −2.5−2.6−2.6−2.9−2.6
CYR61Cysteine-rich, angiogenic inducer, 61 NM_001554.3 −2.5−2.1−2.5−2.3−2.7
MAD2L1MAD2 mitotic arrest deficient-like 1 (yeast) NM_002358.2 −2.5−2.7−2.7−2.6−2.6
CCL15Chemokine (C-C motif) ligand 15, TV1 NM_032964.2 −2.5−2.2−2.5−2.3−2.3
S1PR3sSphingosine-1-phosphate receptor 3 NM_005226.2 −2.5−2.0−2.5−2.4−2.5
C6orf105Chromosome 6 ORF 105 NM_032744.1 −2.5−3.1−2.7−2.9−2.6
TACSTD2Tumour-associated calcium signal transducer 2 NM_002353.1 −2.6−2.2−2.6−3.1−2.7
MT1EMetallothionein 1E NM_175617.3 −2.7−2.2−3.1−2.7−2.2
PLOD2Procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2, TV1 NM_182943.2 −2.7−3.1−2.9−3.4−2.7
STC2Stanniocalcin 2 NM_003714.2 −2.7−2.2−3.3−3.0−2.9
SDPRSerum deprivation response (phosphatidylserine binding protein) NM_004657.4 −2.8−3.1−3.2−3.7−3.0
PLOD2Procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2, TV2 NM_000935.2 −2.8−2.4−2.9−3.0−2.9
LOC399942Similar to tubulin α-2 chain (α-tubulin 2), TV5 XM_934471.1 −3.0−3.3−2.9−3.0−2.8
CXCL1Chemokine (C-X-C motif) ligand 1 NM_001511.1 −3.1−3.0−3.3−3.2−2.9
UHRF1Ubiquitin-like with PHD and ring finger domains 1, TV1 NM_001048201.1 −3.2−2.7−3.4−3.1−3.5
PTGER4Prostaglandin E receptor 4 (subtype EP4) NM_000958.2 −3.3−2.5−3.4−3.9−3.4
MGC87042Similar to six transmembrane epithelial antigen of prostate XM_001128032.1 −3.4−2.9−3.7−3.8−3.4
TOP2ATopoisomerase (DNA) II α 170 kDa NM_001067.2 −3.5−3.3−3.4−3.5−3.4
LOC399959Mir-100-let-7a-2 cluster host gene (nonprotein coding) NR_024430.1 −3.6−3.0−4.2−4.0−3.9
STEAP1Six transmembrane epithelial antigen of the prostate 1 NM_012449.2 −3.6−3.5−3.6−4.0−3.7
BMP4Bone morphogenetic protein 4, TV3 NM_130851.1 −3.6−2.8−4.3−3.8−3.8
LOC158376Hypothetical protein XM_001129749.1 −3.9−3.3−3.1−3.6−3.6
DKK1Dickkopf 1 homologue (Xenopus laevis) NM_012242.2 −5.2−4.0−6.2−6.0−5.5
RGS4Regulator of G-protein signalling 4 NM_005613.3 −7.6−6.5−9.9−10.2−9.0
Figure 5.

Global changes in gene expression between normal and micropatterned HCAECs. (A) Hierarchical clustering and heatmap representation of the 361 genes differentially expressed in the shape-restricted cells compared to the nonrestricted controls. The colour-coded scale (blue–green = down-regulation; orange–red = up-regulation) for the normalized fold changes is indicated at the bottom. Details for the regulated genes are provided in Fig. 2 and are publically available via the Gene Expression Omnibus (accession number GSE43349). Only genes with expression levels regulated above a two-fold change (P < 0.05) compared to the nonrestricted cells are shown. (B, C) Profile plots correlating the gene expression levels based on normalized signal intensities of probe sets between nonrestricted and micropatterned HCAECs under (B) standard growth conditions or (C) after 48 h of serum starvation. (D) Profile plot comparison of the gene expression intensity changes only between each micropattern condition.

Although seeding density was controlled in these experiments to minimize cell-to-cell contact (particularly in the control samples where cellular interactions are possible), it is probable that the use of rich growth media encourages the proliferation of the nonrestricted cells but, because shape restriction has been shown to inhibit proliferation [8], is unable to do so in the shape-restricted cells. This could potentially induce bias in the interpretation of the data from the unrestricted HCAECs as a result of differences in cell cycle progression or cell-to-cell contacts that arise between the mother and daughter cells following mitosis. To address this potential concern, we performed the same micropatterning experiment as described above, except the cells in both the nonrestricted and shape-restricted conditions were serum starved for 48 h before RNA collection to block cell proliferation, thus eliminating variables such as cell-to-cell contact, cell cycle differences, etc. A comparison of the profile plots of nonrestricted versus shape-restricted HCAECs grown in standard growth conditions or subsequent to serum starvation yielded similar results, indicating that, regardless of growth conditions, proliferation or cell-to-cell contact, morphological restriction induced significant changes in the global gene expression profiles (Fig. 5C). The complete microarray data set for the serum starvation experiment is publically available via the Gene Expression Omnibus (accession number GSE44168).

Previous data collected from mesenchymal progenitor cells conforming to micropatterns that induced the cells to form obtuse versus acute morphological angles suggested that cell morphology controls lineage specification. Although endothelial cells are terminally differentiated, we aimed to determine whether such distinct morphological as well as polarity changes might influence the endothelial transcriptome. By excluding the nonrestricted conditions from the analysis, we compared the gene expression changes between only the micropatterned endothelial cells adhering to cross-bow, disc, H, L and Y shapes to examine whether distinct cellular morphology can affect endothelial gene expression patterns. As shown in Fig. 5B,C and Table 2, gene expression changes did not significantly vary based on the particular shape, actin patterning or polarity to which the cells conformed. Indeed, statistical analysis of the genomic data set failed to reveal a single two-fold or greater (P < 0.05) alteration in gene expression between any of the cell shapes. These data suggest that, unlike mesenchymal stem cells whose phenotype can be modulated by cellular angularity, endothelial cells grown under these unique geometric constraints do not differ in their global gene expression patterns. Cumulatively, our data indicates that morphological constraint, rather than cellular angularity and polarity, alter the global transcriptome under these conditions.

Pathway analysis of the morphology induced transcriptome changes

We next implemented a systems level approach to understand how geometric constraint may affect the overall cellular phenotype. Our initial analysis reported above included two-fold or greater changes in gene expression, yet, for this network analysis, we broadened our microarray data set (from the standard growth condition experiment) to include the 1.4-fold or greater statistically relevant (P < 0.05) changes in gene expression. This cut-off was selected not only to refrain from limiting our network analysis to solely the highest expression changes, but also to take into account transcriptional changes that were less pronounced but still relevant with regard to modulating cellular physiology. This resulted in ~ 8% of the human genome experiencing changes in gene expression (642 up-regulated genes and 1218 down-regulated genes). We then performed metacore pathway analysis of these gene expression changes to predict significant alterations in major cellular processes, including cell cycle regulation (P < 3.3 × 10−8) (Table 3), cytoskeletal dynamics and cell adhesion (P < 4.2 × 10−5) (Table 4), glycolysis/gluconeogenesis (P < 2.7 × 10−4) (Table 5), TGFβ signalling (P < 1.6 × 10−3) (Table 6) and wingless-type (Wnt) signalling (P < 1.6 × 10−3) (Table 7). Because TGFβ signalling has been shown to play a major role in arteriosclerotic disease progression, we confirmed our microarray data utilizing quantitative PCR to detect the shape-induced alterations in mRNA expression levels of the TGFβ signalling genes SMAD6, SMAD7 and TGFB2, as well as several genes reportedly involved in the atherosclerotic process, including LPL, MMP1, KDR, ITGA2, ACE, BIRC3, IL1R1, ICAM1, HEY1, BCL2, CSF2, APOE, PDGFB, BCL2A1, CCL2 and LDLR (Fig. 6).

Table 3. Fold changes in mRNA expression levels of genes involved in cell cycle progression
Table 4. Fold changes in mRNA expression levels of genes involved in cytoskeletal dynamics and cell adhesion
Table 5. Fold changes in mRNA expression levels of genes involved in glycolysis and gluconeogenesis
Table 6. Fold changes in mRNA expression levels of genes involved in TGFβ signalling
Table 7. Fold changes in mRNA expression levels of genes involved in Wnt signalling
Figure 6.

Quantitative PCR confirmation of microarray data. Confirmatory quantitative PCR was performed on 19 genes whose expression was shown to be altered in the microarray data. Relative quantification (RQ) values are shown for each gene expression change. cDNA was obtained from normal and crossbow shape cells grown under standard culture conditions.


The interplay between the physical, chemical and biological cues to which cells are constantly exposed modulates processes ranging from those as broad as cellular lineage determination to those as subtle as the functional nuances between two adjacent cells. Despite the number of studies addressing this area of research, the molecular mechanisms by which these cues synergize is largely unknown. It has been reported that cellular morphology and cytoskeletal angularity greatly influence progenitor lineage specification [14] and that changes in cell shape influence chromatin condensation via nuclear deformation [25]. In the present study, we aimed to determine whether morphological changes in coronary artery endothelial cells could affect the global patterns of gene expression. Understanding how cell shape change affects the coronary artery endothelial cell transcriptome may allow us to better understand the molecular aberrations that underlie coronary artery disease. The present study made use of micropatterned growth substrates that force cells to conform to precise geometric shapes. Although micropatterned cell growth has been utilized in a limited number of studies, there is little evidence that such techniques consistently lead to morphological and cytoskeletal patterns that are highly reproducible and truly unique between different micropatterns. We utilized pattern recognition algorithms and statistical analysis to confirm that cells conforming to the crossbow, disk, H, L or Y shapes had truly reproducible cellular morphology and cytoskeletal architecture unique for each cell shape adopted. Given that most analysis of cytoskeletal organization in the available literature is qualitative in nature, this algorithm can be extensively used in the future to provide quantitative interpretations of the differences in both static (as we have analyzed) and dynamic cytoskeletal structures between two or more treatment groups.

Upon demonstrating the reproducibility and applicability of micropatterns to control cellular morphology, we utilized microarray technology to analyze how morphological restriction and unique cellular morphologies affect the HCAEC transcriptome. Our data indicate that morphological restriction (i.e. ability of the cell to spread) is a major regulator of endothelial gene expression patterns, as demonstrated by large-scale changes in gene expression after morphological restriction of HCAECs. Our data indicate that morphological restriction via micropattern adherence greatly increases the incidence of nuclear deformation in HCAECs. Given that large-scale cell shape changes results in a drastic condensation of chromatin as a result of lateral compressive force-induced nuclear orientation shifts and deformation [25], it is possible that restricting cell spreading affects the dynamic genome architecture in the nuclear space, thus regulating gene expression by modulating the geometric constraints that regulate dynamic chromatin positioning. We suspect that shape-induced gene expression changes are more complex than simply a consequence of nuclear deformation given that the transcriptome between each of the micropatterned shapes was remarkably similar, whereas the level of nuclear deformation varied drastically between the individual micropatterns. Indeed, although distinct cell shapes and cytoskeletal patterning have been reported to regulate mesenchymal progenitor lineage determination and endothelial cell chromatin condensation [14, 25], we were very surprised to discover that shape induced gene expression patterns were remarkably constant across all altered cellular morphologies tested relative to each other. Moreover, considering a recent study suggesting that cell geometry does not regulate the adipogenic differentiation of mesenchymal stem cells [15], further follow-up studies are needed to determine how cellular geometry affects the phenotype of different cell types. Our data do not necessarily contradict the report of shape-induced differentiation in mesenchymal progenitor cells [14] but, instead, suggest that there are varying levels of responsiveness to morphology driven cellular outputs between different cell types (mesenchymal progenitor versus coronary artery endothelial cells). Cummulatively, our data suggest that the ability of HCAECs to spread (but not necessarily their particular morphology) dictates their genomics patterns. These data build on and corroborate the findings reported in earlier work indicating that endothelial spreading regulates cell fate decisions between proliferation and death [8, 11].

Bioinformatics analysis of the microarray data revealed that the largest functional groupings of genes whose expression was altered upon morphological restriction were those involved in cell cycle regulation (30 genes) and cytoskeletal dynamics/cell adhesion (34 genes). Within the identified cell cycle regulators, a number of genes were strongly involved in spindle assembly, cell cycle phase transition, nucleocytoplasmic transport of cyclins and cyclin-dependent kinases, and chromosome condensation. With the exception of one gene (H1F0, which encodes for a histone protein), the expression the identified cell cycle-related genes was down-regulated, including the major cell cycle promoters CDK6, CCNA1, CCNB2, CCND2 and CCND3. Considering the previously proposed impact of cell shape on chromosome condensation, we were intrigued at the down-regulation of genes involved in DNA accessibility, including condensin (NCAPG), topoisomerase II α (TOP2A), histone H3 (H3F3B) and histone H1 (H1F0). These particular changes could have a role in modulating global gene expression, lineage specification and the cellular physiology of endothelial cells and their progenitors. In mesenchymal progenitor cells, it has been reported that shape-induced contraction enhances c-Jun N-terminal kinase and extracellular-related kinase 1/2 activity in conjunction with wingless-type signalling [14]. Pathway analysis of the microarray data from the shape confirmed that HCAECs revealed shape-induced alterations in the expression of genes involved in Wnt signalling (up-regulation of TCF4 and down-regulation of RUVBL2, SNAI2, FZD4 and DKK1) and an up-regulation in JUN expression, indicating that similar changes in these signalling pathways likely occur when the endothelial cell morphology is altered. Additionally, the expression of several genes encoding members of the TGFβ signalling cascades was altered upon changes in HCAEC shape, including the ligands BMP2, BMP4 and TGFB2, the type II receptor BMPR2, and the signalling effectors SMAD6 and SMAD7. Given that aberrant TGFβ signalling is critically implicated in the progression of coronary artery disease and arteriosclerosis [31], it is possible that endothelial cell shape changes could initiate and/or exacerbate disease progression via alterations in the expression of key genes involved in these processes.

Materials and methods

Cell culture and treatments

Primary cultures of human coronary artery endothelial cells (HCAECs; < 5 passages; #PCS-100-020; ATCC, Manassas, VA, USA) were cultured in vascular cell basal media (#PCS-100-030; ATCC) supplemented with 0.2% bovine brain extract, 5 ng·mL−1 human epidermal growth factor, 10 mm l-glutamine, 0.75 units·mL−1 heparin sulfate, 1 μg·mL−1 hydrocortisone, 50 μg·mL−1 ascorbic acid, 2% fetal bovine serum and pen/strep. For serum starvation experiments, HCAECs were cultured in vascular cell basal media (#PCS-100-030; ATCC) supplemented with 10 mm l-glutamine, 0.75 units·mL−1 heparin sulfate, 1 μg·mL−1 hydrocortisone, 50 μg·mL−1 ascorbic acid and pen/strep for 48 h before RNA collection. For cell shape patterning, collagen I-coated coverslips and 96-well plates with micropatterns surrounded by non-adhesive surfaces (Cytoo Inc., Grenoble, France) were seeded with ~ 5000 or 50 000 HCAECs per well and coverslip, respectively, in accordance with the manufacturer's instructions. For the control, cells were seeded at low density approximately equal to that seen in the micropatterned conditions (to minimize cell-to-cell contacts) on collagen I-coated coverslips and 96-well plates. For all experiments, disc, crossbow, H, Y, and L adhesive micropatterns (1600 μm2) plus controls were contained on the same chip or plate to reduce experimental variability.


Micropatterned coverslips (Cytoo Inc.) were fixed in fresh 4% paraformaldehyde, blocked in 5% BSA plus 0.5% Tween-20, and incubated with 1 : 200 phospho-FAK (#3283; Cell Signaling, Danvers, MA, USA) antibody, 1 : 350 rhodamine-conjugated phalloidin (Cytoskeleton Inc., Denver, CO, USA) and 1 : 1000 DAPI. Anti-phospho-FAK was labelled with a FITC-conjugated secondary antibody and immunofluorescent images were captured in 0.1-μm Z-stacks using a C2SI scanning laser confocal microscope (Nikon, Tokyo, Japan). Images were equivalently processed in nikon elements 3.2, surface rendering images were obtained using imaris, version 6.0 (Bitplane AG, Zurich, Switzerland) and three-dimensional deconvolution was performed using Autoquant X3 (Media Cybernetics, Inc., Bethesda, MD, USA).

Quantification of actin fibre length

For each analysis, 11–14 images of each shape from the actin immunofluorescent images were utilized. Images were initially preprocessed by implementing contrast-limited adaptive histogram equalization, which enhances the contrast of the image in small regions rather than as a whole [32] (Fig. 2B). Images are rotated to have consistent orientation of the micropattern for all analyses. For automatic detection of actin fibres, we utilized the fiberscore algorithm reported by Lichtenstein et al. [29], which bases the segmentation of fibres on the probability that a pixel neighbourhood belongs to a fibre. The output of the fiberscore algorithm comprises a correlation image (Fig. 2C), which indicates pixels with higher probability of belonging to a fibre, and an orientation image (Fig. 2D), which indicates the orientation of the fibre at each pixel location. To remove fibres from the resulting fiberscore output that are less correlated than other image regions, we performed a two-step post processing method: (a) remove pixels with correlation values below a predetermined threshold (Fig. 2E) and (b) skeletonize the fibre structures with combinations of the basic morphological operations erosion and opening [32] (Fig. 2F). The skeletonization process removes repetitive information within each detected fibre. Individual and median fibre lengths were obtained by measuring the processed fibre length in the skeletonized images.

Quantification of actin fibre orientation

For analysis of actin fibre orientation, each image was divided into nine separate tiles in the form of a 3 × 3 grid, thus providing information on where in the cell certain distributions of angles occur. Tiling allows for the option of local subcellular measurements of actin orientation, at the same time as gathering all information in the tiles provides a measure of the entire cell. For quantitative analysis of the 3 × 3 tiling, we implemented the two-sample KS test [30] to compare cell images within a single shape in terms of overall fibre orientation distributions. We used the KS test in two different methods to calculate the amount of difference between the distributions of fibre angles. In the first method, we compared the entire individual image to the cumulative tiling, providing a measure of the overall global difference in fibre distributions. The second method compared an individual image with the cumulative shape image on a tile-by-tile basis, providing a local measure of the difference between individual cell distributions and the cumulative distributions. This tile-by-tile comparison is used to pinpoint similar regions between cell shapes that can be result in less uniqueness in global shape comparisons. Both methods count the number of null hypothesis rejections (at a significance level of 0.05) and normalize according to the number of KS tests.

Gene expression analysis

For each shape tested, as well as the nonrestricted controls, ~ 5000 HCAECs were grown in each well of a 96-well micropatterned plate. This was replicated in 16 independent wells per shape to minimize experimental error. Total RNA for each shape was isolated using the Purelink RNA Micro kit (Invitrogen, Grand Island, NY, USA) after 24 h of the cells adhering to the substrate. The isolated RNA from the replicates (5000 cells per shape multiplied by 16 independent replicates) were pooled, amplified and biotin-labelled using an Illumina TotalPrep RNA Amplification Kit (Illumina, San Diego, CA, USA). Some 750 ng of biotinylated aRNA was then briefly heat-denatured and loaded onto expression arrays to hybridize overnight. Following hybridization, arrays were labelled with Cy3-streptavidin and imaged using the Illumina ISCAN. Intensity values were transferred to genespring gx software (Agilent Technologies Inc., Santa Clara, CA, USA) and data were filtered based on the quality of each call. Statistical relevance was determined using analysis of variance with a Benjamini Hochberg false discovery rate multiple testing correction (P < 0.05). Data were then limited by fold change analysis to statistically relevant data points demonstrating a two-fold or greater change in expression. Omics pathway analysis was performed with metacore software (GeneGo, San Diego, CA, USA). Microarray data were publically deposited in Gene Expression Omnibus (standard growth conditions = accession number GSE43349; Serum starvation conditions = accession number GSE44168). For confirmation of microarray results, RNA from normal- and crossbow-shaped cells was converted to cDNA using the Verso cDNA kit (Thermo-Scientific, Waltham, MA, USA) and quantitative PCR was performed using SYBR Green probes (Invitrogen) with an ABI7900HT real-time PCR instrument (Invitrogen).


Support for the present study was provided by a National Heart, Lung and Blood Institute grant HL098931 and internal support to B.A.B., a NASA EPSCoR award to New Mexico State University, and internal support from New Mexico State University to L.B.