|Home | About | Journals | Submit | Contact Us | Français|
The directed formation of complex three-dimensional (3D) tissue architecture is a fundamental goal in tissue engineering and regenerative medicine. The growth of cells in 3D structures is expected to influence cellular phenotype and function, especially relative cell distribution, expression profiles, and responsiveness to exogenous signals; however, relatively few studies have been carried out to examine the effects of 3D reaggregation on cells from critical target organs, like the heart. Accordingly, we cultured primary cardiac ventricular cells in a 3D model system using a serum-free medium to test the hypothesis that expression profiles, multicellular organizational pathways, tissue maturation markers, and responsiveness to hormone stimulation were significantly altered in stable cell populations grown in 3D versus 2D culture. We found that distinct multi-cellular structures formed in 3D in conjunction with changes in mRNA expression profile, up-regulation of endothelial cell migratory pathways, decreases in the expression of fetal genes (Nppa and Ankrd1), and increased sensitivity to tri-iodothyronine stimulation when compared to parallel 2D cultures comprising the same cell populations. These results indicate that the culture of primary cardiac cells in 3D aggregates leads to physiologically relevant alterations in component cell phenotype consistent with cardiac ventricular tissue formation and maturation.
Tissues and organs depend on complex, three-dimensional (3D), multi-cellular structures to function. Reestablishing these tissue architectures ex vivo or in situ is a central goal in tissue engineering and regenerative medicine. The heart is a principal target for research in these areas, but the effects of 3D organization on the cells that make up the heart are not well understood. Encouragingly, several groups1–12 have described the self-organization and neovascularization of 3D heart cell reaggregates in vitro. These studies demonstrate that populations of primary cardiac cells can reestablish aspects of tissue-level architecture when grown in supportive culture systems; however, the effects of 3D culture on the molecular phenotype of the component cells remain largely unexplored. With mounting evidence that multi-cellular organization has profound effects on cell function,13,14 there is a need to characterize associations between multi-cellular structure and cell phenotype ex vivo so that issues of cell responsiveness to different culture conditions can be included in the design of advanced scaffolds for tissue engineering.
Primary neonatal rat cardiac ventricle cells (NRCVCs) represent a useful and accessible model system for tissue engineering and molecular studies. Isolated NRCVCs comprise a mixed population of cells representative of the intact tissue and include 75–80% cardiac muscle cells (cardiomyocytes [CMs]), which are widely recognized to be nonproliferative, with the remainder made up largely of endothelial cells (ECs) and cardiac fibroblasts,15–18 which are highly proliferative. The growth of such mixed cell populations in serum-containing media often leads to fibroblast overgrowth. Serum-free media, on the other hand, can severely restrict cell proliferation, and in the present study, NRCVCs were grown in a custom serum-free medium to investigate phenotypic differences associated with 3D versus two-dimensional (2D) culture systems. Specifically, we investigated alterations in multi-cellular architecture, differences in gene expression profile, and changes in hallmarks of ventricular tissue development and maturation, including (i) the distribution of ECs within 3D aggregates, (ii) the maturational status of the ventricular myocytes detected as decreases in fetal gene expression, and (iii) the responsiveness of the cellular system to hormonal stimulation by triiodothyronine (T3).
NRCVCs were isolated using Neonatal Cardiomyocyte Isolation System kits (Worthington Biochemical, Lakewood, NJ) as in previous studies by our group.5 Viable cells were enumerated by Trypan-blue exclusion (Invitrogen, Carlsbad, CA). To support the inclusion of the same cell sub-populations in 3D and 2D cultures, Nunclon™ Δ-Surface tissue culture polystyrene (TCPS) were used in both cases for cell seeding: microcarriers were used as seeding surfaces to initiate 3D cultures and Nunclon Δ-Surface plates were used for 2D (Nunc/Thermo-Fisher Scientific, Rochester, NY). To encourage cell–cell interactions and to reduce cell proliferation by contact inhibition, cultures were generally inoculated at a high density (1×106 cells per 4.8cm2 of TCPS per mL of medium) although in some experiments, cultures were inoculated at a medium (0.5×106 cells per 4.8cm2) or low (0.2×106 cells per 4.8cm2) density to encourage fibroblast proliferation as noted in the text. In addition, a defined serum-free medium that had been used previously by our group (AI-1 medium)5,19 was employed to limit mitogenic effects from serum. AI-1 medium comprised a 1:1 mixture of Dulbecco's modified Eagle's and Ham's F-12 media, to which was added 1× modified Eagle's medium (MEM) nonessential amino acids, 0.1× MEM vitamin solution, 2.45mg/mL sodium bicarbonate, 100U/mL penicillin, 100U/mL streptomycin (all from Invitrogen), 20μg/mL sodium ascorbate, 0.25mg/mL fetuin, 0.0835mg/mL calcium chloride, 0.001μg/ml sodium selenate, 1.35mg/mL additional dextrose, 0.01μg/mL epidermal growth factor, 75pg/mL tri-iodothyronine (unless otherwise indicated in the text), 0.05μg/mL hydrocortisone, 0.05mU/mL insulin (all from Sigma–Aldrich, St. Louis, MO), 5μg/mL holo-transferrin (BD Biosciences, Bedford, MA), 10mg/mL bovine serum albumin (BSA; Roche Diagnostics, Indianapolis, IN), and 1μM palmitate (Sigma-Aldrich) preconjugated with 0.12mg/mL fatty-acid-free BSA (Roche Diagnostics) as carrier. In addition to the use of fetuin in the medium, TCPS surfaces were coated with human fibronectin (BD Biosciences; 1μg per cm2) to encourage cell adhesion in the absence of serum. For 3D cultures, microcarriers were suspended in polytetrafluoroethylene VueLife™ vessels (American FluorSeal, Gaithersburg, MD) by rotation as in our previous work.5 Polytetrafluoroethylene culture vessels were rotated about their long axes at 30rpm to suspend cells and microcarrier beads without excessive fluid shear, which can be damaging to cells.20 Cultures were given a complete change of medium 24h after initiation and at 96-hour intervals thereafter. Cultures were assessed throughout to assure spontaneous beating, which is characteristic of healthy NRCVC cultures.
Fluorescence microscopy was carried out using methods previously established by our group.5,6,21 In brief, parallel 2D and 3D culture samples were fixed in buffered, 2% paraformaldehyde (for sarcomeric myosin, vimentin, and BrdU staining) containing 0.1% Triton X-100 (Sigma–Aldrich) or −20°C methanol (for platelet/endothelial cell adhesion molecule [PECAM] staining), blocked with 3% BSA in phosphate-buffered saline. Samples were stained with bis-benzimide (Hoechst 33258, Sigma–Aldrich) to detect nuclei and anti-sarcomeric myosin heavy chain (MyHC; monoclonal A4.1025, Developmental Studies Hybridoma Bank, Iowa City, IA, to detect CMs), anti-PECAM (BD-Pharmingen, Franklin Lakes, NJ, to detect EC clusters), or anti-vimentin (monoclonal V9; Sigma-Aldrich, to detect mesenchymal cells) with a Texas-Red-conjugated secondary antibody (Jackson Immuno Research, West Grove, PA). Selected samples were also stained with Alexa-Fluor-488-conjugated phalloidin (Invitrogen) to detect filamentous actin, which is dramatically elevated in CMs due to the presence of sarcomeres. Viable cell content was estimated using Cell Titer Blue Assays (Promega, Madison, WI) following the manufacturer's recommendations. To estimate cell proliferation by BrdU (5-bromo-2-deoxyuridine; Sigma-Aldrich) incorporation, cells were labeled with 100μM BrdU for 6h on day 4 of culture. DNA was then denatured with 2N HCl for 30min at 37°C followed by neutralization in 0.1M borate buffer (pH 8.0) to expose BrdU. Anti-BrdU antibody conjugated to Alexa Fluor 594 (Invitrogen) was used for observation. BrdU-positive cells were enumerated in 10 randomly selected microscope fields at 200× total magnification. Light microscopic observations were carried out using an Evolution QEi, 12-bit, cooled CCD camera (Media Cybernetics, Silver Spring, MD) mounted to an Olympus model BX-60 epi-fluorescence microscope operated using Image Pro Plus software (Media Cybernetics); fluorescence images were pseudo-colored for presentation. Scanning electron microscopy (SEM) was performed on a JEOL 840A scope operating at 10kV. Samples for SEM were fixed at 4°C with 2% glutaraldehyde in 0.1M cacodylate buffer, rinsed and dehydrated in graded acetonitrile solutions,22 critical point dried, and sputter coated with gold before observation.
Biochemical assays for creatine kinase, glucose-6-phosphate dehydrogenase, malate dehydrogenase, nicotinamide adenine dinucleotide (NAD)-dependent cytochrome-c reductase, and pyruvate kinase (PK) were carried out as previously described.21,23,24 Briefly, cells were rinsed and lysed to release cytosolic and mitochondrial enzymes. Specific activities were measured using direct or coupled assays and expressed in Units per mI of homogenate, with 1U being defined as the amount of enzyme that reduces 1nmol of substrate (NAD, NAD phosphate [NADP], or cytochrome c as appropriate for the enzyme) per mg of protein per minute. Protein and DNA levels in homogenates were determined using BCA (Pierce Chemical, Indianapolis, IN) and Hoechst 33258 (Sigma-Aldrich), and specific activities were calculated as Units of enzyme per mg of protein; the amount of protein per DNA was calculated as an index of hypertrophy.
Levels of MyHC, which is expressed in CMs but not in other cardiac-derived cells, and filamentous actin (f-actin), which is more abundant in CMs than in other cells due to the presence of sarcomeres, were quantified fluorimetrically and normalized to nucleic acid staining. To accomplish this, samples were fixed, rinsed, and stained as above to detect MyHC (A4.1025 antibody with a Texas-Red-conjugated secondary), f-actin (Alexa-Fluor-488-conjugated phalloidin), and nuclei (Hoechst 33258). Fluorescence levels were quantified using a Cytofluor 4000 fluorescence reader (Applied Biosystems, Foster City, CA).
Real-time quantitative polymerase chain reaction (PCR) was carried out using a single-color BioRad My-iQ detection system and iScript One-Step reverse transcription PCR Kits with SYBR Green (BioRad, Hercules, CA). Primers are listed in Table 1. RNA was isolated using TRI Reagent (Molecular Research Center, Cincinnati, OH). For 2D cultures, 1.0mL TRI Reagent was added to each 35mm dish. Three-dimensional aggregates were collected in 15mL conical tubes, and the medium was aspirated; 1mL of TRI Reagent was added, and the sample vortexed. The TCPS support beads floated to the surface, and the volume of liquid below was collected in sterile Eppendorf tubes. The manufacturer's directions were followed for further purification of RNA. For analysis of α- and β-MyHC expression, multiplex reverse transcription PCR was employed due to the significant homology between the nucleotide sequences and the large size of the PCR products needed to discriminate the two isoforms. Briefly, 1 μg of RNA was used for each 20μL reaction, which consisted of 5mM MgCl2, 2.5mM dNTP mix, 20units of RNAsin, 12.5μg/mL Oligo dT 15, and 40units of avian myeblastosis virus reverse transcriptase (all from Promega). Ten microliters of the reverse transcription reaction was subsequently used in 50μL PCRs with Taq polymerase (5000 μ/mL; Invitrogen), and 25pmol/μL of each primer as listed in Table 1. Reactions containing all four primers were run for 30 PCR cycles. The product was separated on 1% agarose gels and observed by ethidium bromide staining. The fluorescence intensity of each band was quantified using an Eagle-Eye gel documentation system (Stratagene, LaJolla, CA).
Expression data derived from Affymetrix chips used in a previous experiment5 were analyzed to identify differences in expression associated with 2D versus 3D culture using The Institute for Genomic Research Multiple Experiment Viewer (TMeV) version 4.3.02 from the TM4 microarray software suite.25 Briefly, cells were grown in 2D and 3D culture for 1, 4, and 6 days in two separate experiments generating 12 separate samples. RNA was isolated using TRI Reagent (Molecular Research Center) with bromochloropropane and purified using RNEasy kits (Qiagen, Valencia, CA). Double-stranded cDNA was prepared using a T7-oligo-dT primer (T7-dT24, Integrated DNA Technologies, Coralville, IA) and SuperScript Double Stranded cDNA Synthesis Kits (Invitrogen). T7-polymerase in vitro transcription and biotin-labeling were performed using BioArray HighYield RNA Transcription Kits (Enzo Life Sciences, Farmingdale, NY). Labeled cRNAs were hybridized to GeneChip Rat Genome U34A (RG-U34A) Arrays (Affymetrix, Santa Clara, CA), washed, and stained with R-phycoerythrin streptavidin (Molecular Probes) using the GeneChip Fluidics Station 400 and the EukGE-WS2 fluidics protocol. Signals were acquired using an Affymetrix GeneArray Scanner, and numeric data were derived from raw data images and normalized using Microarray Suite 5 software (Affymetrix). Signals on each chip were mean centered to a value of 500 based on a scaling factor calculated from the middle 96% of all signal values. The entire data set of 8799 test probesets across all samples was loaded into TMeV for analysis.
Atrial natriuretic peptide (ANP), which exhibits decreasing expression during ventricular CM maturation, was detected in cellular extracts as previously described26 using a high-sensitivity enzyme immunoassay for rat ANP containing amino-acids 1–28 (Bachem, San Carlos, CA) following the manufacturer's recommendations. This assay detected total ANP including pre-pro-ANP, pro-ANP, and ANP. Expression of Ankrd1, which encodes cardiac adriamycin response protein (CARP; also called cardiac ankyrin repeat protein), another marker of immature CMs, was assessed by Western blot using a polyclonal antibody generously provided by Dr. Yimin Zou.27 Cultures were rinsed, total protein was extracted using lithium dodecyl sulfate-containing sample buffer, and protein content was determined by bicinchonninic acid (BCA) assay. Equivalent amounts of protein (10μg) were separated on NuPage 4–12% Bis-Tris PAGE gels (Invitrogen) in NuPage MOPS Running Buffer (Invitrogen). Proteins were transferred to nitrocellulose membranes (GE-Amersham, Piscataway, NJ) in NuPage Transfer Buffer (Invitrogen) at 30V for 1h. Membranes were rinsed and blocked for 1h in 8% nonfat dry milk (BioRad) and incubated overnight in 1:1000 polyclonal anti-CARP antibody followed by horseradish peroxidase–conjugated goat anti-rabbit secondary (BioRad). Bands were detected using enhanced chemi-luminescence (ECL) reagents from GE-Amersham (Piscataway, NJ) and Kodak Biomax-Light film (Rochester, NY).
Populations of isolated NRCVCs contain all the cells found in the intact tissue in the same relative proportions.15,16,28 Since the adhesion of cells is affected by the physico-chemical nature of the culture substrate, we sought to assure that the same subpopulations of cells were present in the 2D and 3D systems employed here. Cultures were initiated on identical TCPS surfaces coated with fibronectin. Nonadherent cells were then removed by medium exchange. Performing this TCPS-based collection of cells using micro-carrier beads within rotating bioreactors resulted in the same degree of total cell adhesion and CM-specific adhesion as that seen in control tissue culture plates (Fig. 1A). Three-dimensional cultures on TCPS beads and 2D cultures on TCPS plates collected on day 6 of culture were nearly identical in their sarcomeric MyHC, filamentous actin (f-Actin), and total protein content (Fig. 1B). Intermediary metabolic enzyme activities in extracts collected from day 6 cultures were also evaluated and found to be nearly identical in the two systems (Fig. 1C). These data indicate that the constituent cells comprising the 3D/bioreactor cultures and the 2D/plate cultures were the same.
Fibroblast overgrowth is a critical issue in NRCVC culture, and we sought to limit cell proliferation by using high seeding densities and by employing a serum-free medium. To verify that cell proliferation was held in check in the serum-free medium, Cell Titer Blue assays were used to estimate changes in the number of cells present in cultures over time. As seen in Figure 2A, the total number of cells present in NRCVC cultures grown without serum remained relatively constant over a full 13 days, whereas cells grown with 10% fetal calf serum added to the AI-1 medium continually increased in number. To specifically quantify proliferation, cells were exposed to BrdU (Fig. 2B). NRCVCs seeded at a high density and those grown in the serum-free medium had significantly attenuated levels of BrdU staining compared to cells grown with serum or at a lower initial density. Human vascular smooth muscle cells (T/G-HA vSMC, CRL-1999, American Type Culture Collection, Manassas, VA), which were assayed as a positive control for a proliferative cell type, demonstrated the highest degree of BrdU incorporation.
As seen in Figure 3, the cells grown in bioreactor cultures for 6 days self-organized 3D masses that remained associated with the beads but were elevated from the microcarrier surfaces. These elevated cell masses approached a depth of 100microns, and appeared between beads in all multi-bead aggregates; on average, clusters contained 8.2 (+5.2) TCPS beads based on a random sampling of 100 clusters. Overall, we saw no signs of necrosis within clusters. Three-dimensional architectures were typified by an exterior layer of ECs similar to the cardiac luminal epithelium. The same cells collected on TCPS in static, 2D cultures did not form such structures and were typified by isolated islands of ECs surrounded by other cell types in a mosaic-like pattern. In addition, observation of cultures at the SEM level indicated dramatic differences such that the superficial layer of ECs in 3D closely resembled the cardiac lumen in the appearance of cell junctions and surfaces, whereas 2D cultures exhibited prominent nuclei. The emergence of an EC sheet on the exterior of 3D aggregates that are principally made up of CMs was a striking and unexpected finding. In typical culture systems, nonmuscle cells like ECs adhere to TCPS surfaces more rapidly than CMs.29,30 This phenomenon forms the basis of the widely employed strategy of preplating to remove non-CMs from NRCVC suspensions before culture.31–33 We investigated the possibility that the absence of serum resulted in a reversal of the typical relative adhesion order using timed adhesion assays in which cells were allowed to adhere for the time indicated then were rinsed and allowed to spread overnight. Staining for MyHC and nuclei was carried out, and CM (MyHC-positive) versus non-CM (MyHC-negative) cells were enumerated. As seen in Figure 4, CMs adhered more slowly than non-CMs in our serum-free system just as they do in other cell culture systems.
The appearance of ECs on the surface of 3D reaggregate cultures principally containing CMs suggested that cell migration may have taken place; accordingly, we assessed the activation of a critical EC migratory pathway in 3D versus 2D culture. Upregulation of Id3 and the associated Id1, which acts in concert with Id3 to control EC migration,34,35 is considered necessary and sufficient to mediate bone morphogenetic protein-induced EC migration, which functions via attenuation of thrombospondin (Thbs).36 We, therefore, measured BMP, Id, and Thbs expression in our culture systems using real-time quantitative PCR. Fold differences were calculated using the ΔΔCt method with Tnni3 (i.e., cardiac troponin I) as the loading reference within each sample and expression in freshly isolated cells as the baseline. As shown in Figure 5A, Bmp-2 levels increased slowly in the 2D control cultures reaching a peak on day 4, whereas Bmp-2 level increased dramatically on day 1 of 3D cultures and returned to near control levels on days 2 and 3 before peaking again on day 4. Id3 levels (Fig. 5B) in the 2D controls stayed low until day 4 when they started to rise, reaching a peak on day 5; in 3D, Id3 peaked on day 1 then again on day 5. Thbs levels (Fig. 5C) in the 2D samples were elevated on day 1, peaked on day 2, and then declined; Thbs in 3D cultures, on the other hand, remained low on day 1, consistent with the elevated Id3, but peaked on day 2. These differences occurred within samples that had consistent levels of the Tnni3 housekeeping gene throughout the experiment (Fig. 5D) and suggest that BMP-induced EC migratory pathways were active in early NRCVC cultures grown in 3D. This interpretation was supported by the observed loss of 3D structure formation when Noggin, a BMP antagonist, was included in the culture medium (Fig. 5E, F). Thus, genes in EC migratory pathways were differentially activated in 3D culture, and these pathways may have played central roles in the formation of 3D aggregate structure, especially during the initial phases of culture establishment. It should be noted that BMPs, including BMP-2, also play important roles in CM differentiation and cardiac morphogenesis in vivo,37 and the upregulation of Bmp-2 expression early in 3D culture may exert effects on multi-cellular organization beyond the potential activation of EC migration.
To assess general differences in gene expression patterns between 2D and 3D cultures, we employed a genomic analysis approach. Previous work in our lab had been carried out to compare bioreactor-cultured NRCVCs with parallel cells grown in standard tissue culture dishes over the same time period. We found differential expression of genes associated with tissue morphogenesis and blood vessel formation.5 Data from that study were reanalyzed here to identify genes that were differentially expressed at statistically significant levels between the 2D and 3D arms of the study. RNA from days 1, 4, and 6 of 2D and parallel 3D cultures were analyzed from two separate experiments (i.e., 12 RNA samples) using Affymetrix RG-U34A Arrays. The data were compared using a nonparametric, Two Factor Mack-Skillings Test to separate time-dependent and culture-dependent differences. This approach identified 158genes (Table 2) represented by 167 probesets that were significantly different in 3D versus 2D culture (p<0.01). Based on a figure of merit analysis, K-Means clustering was carried out on these probesets to identify six clusters among the culture-type significant genes (Figure 6). Our analysis demonstrates that fundamental differences in the molecular phenotype of primary cardiac cells exist between our 3D and 2D culture systems.
In heart tissue, EC structure and CM structure formation, maturation, and maintenance are linked, especially in the developing fetus.38–40 Since comparisons of our 2D and 3D culture systems demonstrated alterations in EC distribution and EC migratory pathways, we further investigated the effects of 3D culture on NRCVC maturational status by evaluating protein markers associated with immature tissue.
ANP is an endocrine mediator that is highly expressed in mature cardiac atria. In cardiac ventricles, ANP expression decreases with CM maturation so that mature CMs express very little ANP.26 The process of ANP production differs between atrial and ventricular CMs. In both cell types, Nppa (the gene encoding ANP) mRNA is translated into a 152-amino acid pre-pro-peptide, which is cleaved to form the 126-amino acid pro-ANP. Atrial CMs store pro-ANP in granules from which ANP is released via a regulated pathway when intermittent stretch increases in the tissue, usually due to elevated blood volume. In contrast, ventricular CMs secrete ANP via a constitutive pathway in which ANP is released shortly after synthesis, and pro-ANP is generally not stored in granules.41,42 Significant production of ANP by ventricular CMs occurs only in immature tissue and is only reexpressed in adult tissue that is hypertrophic. Since immature and hypertrophic ventricular cells actively produce ANP, they contain elevated steady state levels of immunoreactive ANP compared to mature ventricular cells. We used this pattern of ANP production to assess the status of our NRCVCs, and as seen in Figure 7A, cells grown in 3D culture had substantially depressed ANP levels, which is consistent with a more mature CM phenotype.
Similar to Nppa, expression of the Ankrd1 gene, which encodes cardiac ankyrin repeat protein (CARP),27,43 is also down-regulated in the adult heart44 but reexpressed in cardiac pathogenesis.45,46 Western blotting for CARP expression at days 1, 4, and 6 of 2D versus 3D culture indicated that Ankrd1 expression was substantially reduced in 3D culture, further supporting a difference in NRCVC maturational status associated with 3D culture (Fig. 7B).
Finally, we investigated whether culture in 3D affected the physiologic responsiveness of the component cells to exogenous stimulation. Previous work by Armstrong and coworkers established that chick embryonic CMs can proliferate and respond to growth factor stimulation when cultured in 3D, but not when grown in 2D.47 In those studies, factors associated with hyperplastic growth of the embryonic myocardium were used to assess effects on cell proliferation. In the present study, responsiveness to T3, which is associated with postnatal CM phenotype in mammalian tissue, was assessed. Early postnatal, ventricular maturation in vivo is typified by a conversion in the MyHC isoform composition of CMs so that the relative level of α-MyHC (i.e., Myh6) increases while β-MyHC (i.e., Myh7) decreases. This conversion occurs in conjunction with increasing levels of circulating T3, and T3 has been shown to antithetically regulate α- and β-MyHC expression.48 The switch in MyHC isoform in response to T3 elevation occurs rapidly in mature, native tissue with complete conversion seen in 24h when hypothyroid rats are made replete with T3.49 Accordingly, we investigated the responsiveness of 3D tissue-like constructs to T3 exposure by evaluating the level of α- and β-MyHC present 24h after a switch from T3-free to T3-containing media. Our serum-free culture system allows us to control the specific level of hormone present, and as can be seen in Figure 8, 2D and 3D cultures both respond to increasing hormone exposure by reexpressing α-MyHC. The cells grown in 3D culture, however, switched their expression to a much greater degree in the time-frame of the experiment. This increased rate of conversion to α-MyHC is consistent with what is seen in mature, native tissue, suggesting that 3D culture elicits a shift in hormone responsiveness to a more mature, native tissue level.
We found that NRCVCs cultured in 3D aggregates comprising the same cell population as parallel 2D cultures exhibit subtle but physiologically important alterations in phenotype. Cells cultured in 3D self-assembled into a tissue-like conformation with a superficial layer of ECs resembling the luminal epithelium of native heart tissue. Three-dimensional culture was associated with altered gene expression profiles relative to 2D, and 3D culture resulted in the differential activation of EC migratory pathways. Three-dimensional cultures exhibited decreased ANP and CARP protein levels indicative of improved tissue maturation, and 3D aggregates responded more rapidly to T3 exposure than parallel 2D cultures, consistent with a mature, native-tissue phenotype. These data indicate that 3D culture in the serum-free medium results in the reorganization and phenotypic alteration of the component cells into structures and behaviors similar to those seen in vivo.
Three-dimensional culture systems have significantly advanced our understanding of fundamental relationships between tissue-level organization and component cell function,13,14 and substantial progress has been made in the use of 3D methods to investigate the formation of cardiac tissue.1–12 The phenotypic adaptations of cells when organized in engineered tissues, however, have not been well elucidated. To study aspects of ex vivo cardiogenesis, we dissociated functional tissue and obtained a mixed cell population representing the original tissue's composition but lacking its multi-cellular organization. It is noteworthy that this cell population is the target population for stem cell and other cell procurement strategies in cardiovascular regenerative therapies and that understanding the controlled assembly of tissue level structures by this population may be critical to the ultimate success of cardiac tissue engineering efforts. In the present study, the formation of 3D aggregates is not entirely surprising as other investigators have used similar low-shear suspension culture models to generate 3D aggregates. The presented work, however, is distinguished by the use of a serum-free medium and highly controlled culture conditions, which severely restrict cell overgrowth (Figs. 1 and and2)2) and which allow the identification of phenotypic differences between component cells in parallel 3D and 2D models.
The up-regulation of EC migratory pathways (Fig. 5) and the appearance of a luminal endothelial sheet in the 3D cultures (Fig. 3) were striking and unexpected given the propensity of CMs to adhere to surfaces later than other cells (Fig. 4). One possibility is that fluid shear in our 3D suspension-culture system activated the CMs or ECs, which are known to be sensitive to mechanical stimulation.50–54 The level of fluid shear determined for systems like the one employed here, however, is low (<<10 dynes/cm2),55,56 and the shear level may not be sufficient to trigger a cellular response. In addition, fluid shear is continuously present throughout the duration of 3D culture, but the Bmp-2, Id3, and Thbs1 expression levels increase and then decrease in both systems (Fig. 4). Finally, Bmp-2 expression, which may drive EC migration, is reportedly unaffected by shear stress in other experiments using similar cells,57 and preliminary studies in our lab to induce Bmp-2 expression in 2D cultures by low level shear exposure have been unsuccessful (data not shown). Thus, although a role for fluid shear cannot be completely ruled out in EC redistribution, it is unlikely that shear contributes to differences in Bmp-2 expression. A second possibility is that differences in Bmp-2 RNA level and EC migration result from coupled interactions between CMs and ECs activated by the 3D contacts permitted in the suspension culture system. BMP-2 is a potential mediator of coordinated EC and CM structure formation in vivo,38 and CM function is suspected of driving aspects of cardiac morphogenesis during normal development via BMP.58 This is an attractive alternative, but further experiments with 3D models are needed to clarify potential multi-cellular interactions and the mechanisms driving multi-cellular structure formation in vitro.
Interestingly, researchers using NRCVC reaggregate culture systems based on serum-containing media do not report exteriorized ECs,7–12 and we initially suspected that the absence of serum in our system may have resulted in a reversal of relative adhesiveness such that CMs adhered first forming a core of cells onto which ECs subsequently adhered. Experiments shown in Figure 4, however, indicate that CMs adhere later than other cells in our serum-free medium; thus, the appearance of a superficial epithelial layer is not easily accounted for by early CM adhesion to TCPS. Our observations suggest that NRCVCs possess a latent ability to organize an EC sheet after culture initiation and that this ability is accessed in 3D/bioreactor cultures but not in 2D/plate cultures when a serum-free medium is used.
Genomic expression analysis using Affymetrix Chips indicated that a number of genes were differentially expressed in 3D versus 2D culture and that these clustered into six groups (Fig. 6). Evaluation of either the entire 167 genes or the genes in each of the six clusters separately was carried out using the online database for annotation, visualization, and integrated discovery (DAVID) informatics resource from the National Institutes of Health.59,60 After controlling for false discovery using Benjamini-Hochberg corrections, DAVID analysis revealed no thematic relationships among the differentially expressed genes beyond ontologies associated with muscle contraction and muscle development (not shown). Nonetheless, there were physiologically significant differences in gene expression associated with 3D culture, and several of the genes identified in the genomic screening, including Bmp-2 (Cluster 2; also in Fig. 5), as well as Nppa and Ankrd1 (Clusters 1 and 6, respectively; also in Fig. 7), were analyzed elsewhere in this article. Determining the functional significance of the other expression differences will require additional molecular and proteomic anlayses.
The observed differences in ANP and CARP levels suggest the acquisition of a different and possibly more mature cell phenotype in 3D versus 2D culture. ANP expression from the Nppa gene has been well studied. ANP is expressed in immature ventricular CMs, but expression decreases over time. In the mature heart, ANP expression is largely restricted to atrial CMs except under pathologic conditions associated with increased ventricular CM stretch (i.e., preload) and hypertrophy.61,62 Since there is no preloading of CMs in our system and no evidence for hypertrophy (Fig. 2), the decrease in ANP is suggestive of CM maturation in the 3D cultures. Similarly, differences in CARP expression from the Ankrd1 gene, which is developmentally down-regulated but reexpressed in hypertrophic ventricular tissue,44,63 suggest that 3D culture encourages CM maturation. As with Bmp-2, a possible explanation for these differences lies in alteration of cell–cell contacts in 3D. Interestingly, recent work on signaling mechanisms involving Notch pathways, which are triggered by specific cell–cell interactions, has shown that both ANP64,65 and CARP66 levels are controlled by the Notch signaling intermediate Hey-2. Thus, differences in ANP and CARP may be associated with altered cell–cell interactions and the acquisition of a layered, tissue-like geometry in 3D.
Ventricular CMs respond to changes in T3 levels by altering the ratio of α- and β-MyHC.48In vivo, this isotype switch occurs rapidly when T3 is given to hypothyroid animals.49 Developmentally, the complement of α- and β-MyHC in rodent cardiac ventricles shifts from ~50% α at birth to nearly 100% α by 3 weeks postnatal.67 Thus, the rapid and nearly complete conversion of MyHC mRNA to the α form when 3D cultures were given 3nM T3 (Fig. 8) is consistent with a shift in hormone sensitivity and the adoption of a more mature CM phenotype.
Taken together, our data indicate that the growth of cardiac ventricular cells in 3D resulted in subtle but significant alterations in cell organization and function. In particular, 3D culture was associated with differences in gene expression, cell migratory signaling, tissue maturation, and responsiveness to hormonal stimulation; these are all differences that impact directly on the design of functional, 3D cardiac tissue. The mechanisms driving alterations in cell function require further investigation but likely involve specific cellular interactions in the 3D environment. This concept is supported by our studies, which emphasized 3D cell masses and did not employ 3D scaffolds. Consideration of the cell interactions that drive the functional alteration of cells and tissues may impact heavily on scaffold design and strategies for tissue formation and may prove critical as the field approaches clinical applications for complex 3D tissues.
The authors thank Michael Bruner, Therese McLaughlin, and Jaimee Militar for their laboratory assistance and helpful discussions. This work was supported by funds from the Nemours Foundation, the National Aeronautics and Space Administration (NAG9-1339), and the National Center for Research Resources at the National Institutes of Health (1P20-RR020173-01).
No competing financial interests exist.