Annals of Biomedical Engineering

, Volume 35, Issue 8, pp 1414–1424

An In Vitro System to Evaluate the Effects of Ischemia on Survival of Cells Used for Cell Therapy

  • Bryce H. Davis
  • Thies Schroeder
  • Pavel S. Yarmolenko
  • Farshid Guilak
  • Mark W. Dewhirst
  • Doris A. Taylor

DOI: 10.1007/s10439-007-9301-2

Cite this article as:
Davis, B.H., Schroeder, T., Yarmolenko, P.S. et al. Ann Biomed Eng (2007) 35: 1414. doi:10.1007/s10439-007-9301-2


Maintaining cell viability is a major challenge associated with transplanting cells into ischemic myocardium to restore function. A likely contributor to significant cell death during cardiac cell therapy is hypoxia/anoxia. We developed a system that enabled quantification and association of cell survival with oxygen and nutrient values within in vitro constructs. Myoblasts were suspended in 2% collagen gels in 1 cm diameter × 1 cm deep constructs. At 48 ± 3 h post-seeding, oxygen levels were measured using microelectrodes and gels were snap-frozen. Bioluminescence metabolite imaging and TUNEL staining were performed on cryosections. Oxygen and glucose consumption and lactate production rates were calculated by fitting data to Fick’s second law of diffusion with Michaelis–Menten kinetics. Oxygen levels dropped to 0 mmHg and glucose levels dropped from 4.28 to 3.18 mM within the first 2000 μm of construct depth. Cell viability dropped to approximately 40% over that same distance and continued to drop further into the construct. We believe this system provides a reproducible and controllable test bed to compare survival, proliferation, and phenotype of various cell inputs (e.g., myoblasts, mesenchymal stem cells, and cardiac stem cells) and the impact of different treatment regimens on the likelihood of survival of transplanted cells.


Myoblast Ischemia Stem cell Cardiomyoplasty Myocardial infarction 


The use of adult stem and progenitor cells to repair pathophysiological consequences of myocardial infarction in the human heart12,15,17,20,22,31,37,41,43,51 has become widespread. However, a major obstacle with this technology is the failure of cells to survive long term.29,35,48 Previous studies suggest that more than 70% of the cells injected during cellular cardiomyoplasty die within the first 3–4 days after transplantation.1,26,40,46,53 While the exact mechanisms responsible for cell death are unknown, it is hypothesized that the ischemic microenvironment surrounding injected cells is a major contributor to the poor survival rate.30,54 However, very few researchers have directly examined the role of ischemia on the survivability of transplanted myoblasts, mesenchymal stem cells, or bone marrow mononuclear cells.26,42 One reason for this limited research is that determining the impact of ischemia and low-nutrient concentration, per se, on survival of transplanted cells is difficult to distinguish in an animal model, especially in the face of impacts from inflammatory cells, cytokines, and matrix metalloproteases, to name a few. Because of these in vivo complexities, there is a need to create an in vitro model system to directly evaluate the impact of ischemia on the survival of cells being transplanted.

In these types of studies, an in vitro model system holds several advantages over in vivo animal models. First, in vitro models enable control of environmental factors such as inflammation, cell rejection, cytokines, etc., on transplanted cell survival, thereby isolating the role of ischemia. Second, the use of an in vitro model allows easy and reproducible changes in the availability of oxygen and nutrients to the cells, permitting a more accurate determination of the thresholds of oxygen or nutrients a particular cell type needs to survive. Finally, in vitro oxygen and nutrient consumption rates of the transplantable cells are easily quantified within a model system, whereas accurate measurement of in vivo metabolism is difficult to discern from that of native, host cell metabolism. The ability to measure these consumption rates is crucial to understand the role of ischemia on cell death because injected cells are competing against each other as well as host cells, for limited oxygen and nutrients in the ischemic/hypoxic infarct. By understanding the metabolic demands of the injected cells, we can potentially calculate the levels of glucose and oxygen available to cells injected into myocardial infarct scar. With better estimates of these oxygen and nutrient values and viability of cells at those values (measured in vitro), it should be possible to devise injection strategies designed to maximize transplanted cell survival. Further, precise calculation of oxygen and nutrient consumption rates of these cells under the gradient of conditions experienced in 3-D constructs or infarct should provide valuable information for computational models designed to predict the viability of cells transplanted in vivo in both animals and in patients. To date, a limitation of many in silico analyses of non-cardiac cells transplanted into a cardiac milieu is that they utilize only approximations of cell oxygen and glucose consumption rates (OCR and GCR) and lactate production rates (LPRs) because few, if any, reliable measurements exist quantifying these parameters in an infarct, or infarct-like, ischemic milieu.

Although little data exist on the specific oxygen and nutrient metabolism and viability of progenitor cells, OCR and GCR of various other cell types have been examined, both in vivo and in vitro in non-cardiac tissues. The current gold-standard technique to quantify oxygen levels and consumption rates in tissues or constructs uses probes, such as recessed-tip glass microelectrodes,38 placed into tissues, constructs, or into the media of cell culture vessels.27,36,50 Microelectrodes were used to monitor oxygen in the current study. Glucose and other metabolites can be measured in a number of ways, but a new, potentially powerful technique that shows promise in tissues and 3-D cultures is bioluminescence metabolite imaging (BMI).32 BMI involves combining frozen sections of the tissue of interest with an enzyme solution that links lactate or glucose to a light producing luciferin–luciferase system. This technique has been previously used to examine tissues as diverse as tumors39,49 and hearts.32 Finally, nitro blue tetrazolium chloride (NBT) and Terminal deoxynucleotidyl Transferase Biotin-dUTP Nick End Labeling (TUNEL) have been used extensively, in histology sections, to examine cell viability and apoptosis, respectively.18,19,21,33,34,52

The goal of this project was to develop an in vitro system to quantify the effects of ischemia on the survival of transplanted cells by seeding porcine myoblasts in type I collagen constructs. This system enabled manipulation of individual parameters to evaluate their impact on cell survival in an environment with hypoxia and glucose conditions encompassing those found in myocardial infarct scar. The 3-D collagen gel model system was designed such that oxygen and nutrient levels were controlled at the surface of the gel and diffusive transport, coupled with oxygen and nutrient consumption by the cells, enabled a gradient of nutrients and oxygen to exist throughout the construct: this gradient was designed to encompass values measured in infarcted pig heart (manuscript in progress). The constructs were cultured for 48 h, with media changed every 12 h, to allow the constructs to reach steady-state. Recessed-tip oxygen microelectrodes were used to measure oxygen gradients23 and bioluminescence imaging was used to measure the gradient of glucose and lactate through the gels. Cell viability and apoptosis were measured using NBT and TUNEL staining, respectively. OCR and GCR were calculated using a custom algorithm written in MATLAB.


Skeletal Muscle Cell Preparation

A 500–700 mg biopsy of the semi-membranosus skeletal muscle was obtained from the hindlimb of swine under general anesthesia (3% isoflurane) (n = 6 pigs). The tissue was mechanically dissected into 1 mm3 pieces and washed twice with Dulbecco’s phosphate buffered saline (PBS), as previously described by our group.45 The tissue was then plated in growth medium containing 10% fetal bovine serum and 0.5% gentamicin (10 mg/mL). After 3 days, tissue fragments were triturated, and additional growth media was added. Fragments were removed after 5 days. The myoblasts were expanded a total of 28 days post-biopsy, 6–8 passages, while maintaining less than 70% confluence. This technique is used regularly in our laboratory to maintain greater than 80% desmin positive myoblasts.43

Gel Construction

Gel constructs, 1 cm diameter × 1 cm depth, were composed of myoblasts suspended in type-I collagen, the primary component of myocardial infarct scar. Creation of collagen gel constructs required 4× concentrated Dulbecco’s modified eagle’s medium (DMEM) to achieve the desired initial glucose concentration within the construct. The 4× DMEM was obtained via addition of double distilled H2O to low glucose DMEM powder (Invitrogen, Carlsbad, CA, catalog # 31600-034). A 2 mg/mL collagen mixture was prepared on ice by combining high concentration rat tail collagen type I (BD Biosciences, San Jose, CA, catalog # 354249) with 4× DMEM, and NaOH was added to neutralize pH to 7.0.

Cells were trypsinized, washed twice in PBS, and re-suspended in growth medium. The cell solution was then mixed with the collagen mixture to obtain a final concentration of 2 million cells/mL gel. This initial concentration lead to a final concentration of approximately 2.5 million cells/mL in the gels at 48 ± 3 h, the time chosen for measurement in order to obtain pseudo-steady state conditions. After mixing, the gel solution was pipetted into a well of a standard 24 well tissue culture plate (Falcon, BD Biosciences, San Jose, CA, catalog # 353047). Gels were placed in a standard tissue culture incubator for 45 min to solidify. Growth media was then added (500 μL) to the top of the gel. Growth media was replaced every 12 h for 48 ± 3 h. For quantification of cell viability, a subset of gels had growth media containing 0.05% NBT for 6–8 h prior to snap freezing. As an extra perturbation to the system, after gel formation, a set of constructs (n = 3) were placed in a tissue culture incubator maintained at 5% oxygen for the 48 h preceding measurements.

Oxygen Measurements

Oxygen tension throughout the gels was measured using recessed-tip, gold-plated oxygen microelectrodes. The microelectrodes were calibrated before and after the experiments using a saline-filled tonometer alternatively perfused with mixtures of nitrogen gas containing 0.0, 2.5, 5.0, and 15% oxygen, as described previously.2 The apparatus used for oxygen measurements is shown in Fig. 1. To maintain proper pH in the gel, 500 μL of new media, identical to the growth media but containing 25 mM HEPES buffer (Invitrogen, Carlsbad, CA, catalog # 12430-054), was placed in the chamber on the gel. A waterbath maintained a constant temperature of 37.0 ± 0.5 °C in the gel by use of a feedback thermocouple placed adjacent to the dish containing the gel construct.
Figure 1

Schematic diagram of the setup used for measurement of oxygen levels throughout the constructs. Oxygen was measured throughout the gel construct using gold-plated microelectrodes. A circulating waterbath was used to keep the gel constructs at 37 °C, and a Faraday cage (signified by the dotted line) was used to protect the probe from extraneous electrical noise

The oxygen microelectrode was introduced into the collagen gel and positioned at a depth of 3000 μm from the gel surface using a micromanipulator (Model MO-102, Narishige International, East Meadow, NY). At this position, the microelectrode output corresponded with an oxygen tension of 0 mmHg. The system was then allowed to stabilize for 15 min. While continuously recording, the microelectrode was then retracted towards the surface at 50 μm intervals every 60 s. This process was repeated three times per gel (n = 6 gels). For measurements from gels subjected to hypoxia during the system perturbation (n = 3 gels), the micromanipulator, culture plates, and waterbath were enclosed by a plastic container. Oxygen tension within the chamber was maintained at 5% oxygen by use of a ProOx hypoxia chamber (BioSpherix Ltd., Redfield, NY) during measurements.


Following oxygen measurements, media from the top of the gel was removed, and to minimize distortion caused by freezing, the constructs were snap-frozen within the well using liquid nitrogen, easing comparison of oxygen measurements to histology and bioluminescence data. Following freezing, gels were removed from the wells and kept at −80 °C until cryosectioning. Gels were surrounded with Tissue-Tek optimal cutting temperature (OCT) compound (Sakura Finetek, Torrance, CA) to allow reproducible sectioning. The constructs were then cut into two hemi-cylinders so that microsections would encompass the full depth of the constructs. Sections were then cut to a thickness of 20 μm for bioluminescence imaging (three sections per construct) and histology (four to five sections per construct), as shown in Fig. 2.
Figure 2

A representation of the gel constructed to measure cell viability and oxygen and nutrient metabolism. Cells were seeded to obtain 2.5 million cells/mL at the time of measurement (48 ± 3 h post-seeding). Following oxygen measurements, gels were snap-frozen and cryosections were taken through the center plane of the gel construct. Bioluminescence imaging allowed measurement of glucose and lactate levels within the construct, and TUNEL staining and NBT labeling allowed examination of cell viability through the gel. Data were then directly compared between cell viability and the availability and metabolism rates of oxygen and glucose

Cell Viability (TUNEL + NBT)

A subset of gels (n = 3 control and n = 3 hypoxia) was exposed to 0.05% NBT (Molecular Probes, Invitrogen, Carlsbad, CA, catalog # N-6495) growth medium for 6–8 h prior to snap freezing. NBT is cell-permeable and reduces to a blue formazan product by viable cells. Therefore, blue-colored cells were considered viable at the time of snap freezing and uncolored cells were considered dead. To measure the contribution of apoptosis to cell death, TUNEL analysis was performed on cryosections (Roche Applied Sciences, Indianapolis, IN, catalog # 11684817910). A propidium iodide (PI) secondary stain provided verification of the total number of cells from all sections. A custom MATLAB algorithm was written to count the total number of cells (PI positive), live cells (NBT positive), and apoptotic cells (TUNEL positive) across each gel cryosection. A total of 10, non-overlapping regions on 4–5 cryosections of each gel were examined.

Bioluminescence Metabolite Imaging

For bioluminescence imaging, cryosections were freeze-dried in a Labconoco Freeze Dry System (Labconoco Co., Kansas City, MO) and stored at −80 °C. Imaging was performed as described previously.32 In brief, successive central sections of the gels were inverted and placed in a chamber filled with a solution linking the substrate of interest (glucose or lactate) to light emission through a bacterial luciferase.50 Data were acquired on a Zeiss Axioskip 2 microscope in a black box with an Andor Technology DV465C-FI CCD camera. Images, at 15× magnification (resulting in pixels dimensions of 15 × 23 μm), were accumulated through photon counting of the emitted light. Carefully controlled glucose and lactate standards were also acquired using the same procedure. Calibration with these standards allowed conversion of emitted light intensity to the associated metabolite concentration. A custom MATLAB (The Math Works Inc., Natick, MA) algorithm was then used to find the average pixel intensity and convert it to the average concentration of the metabolite of interest at each depth from the surface of the gel (n = 3 images per gel). The total number of pixels used to determine each average value per depth (approximately 400 pixels per value) varied depending on the image, as bubbles were excluded from the analysis. The analysis averaged together all ‘relevant’ (identified as within the gel but without a bubble) pixels for each 100 ± 10 μm depth. Results were plotted as average intensity of three bioluminescence images per gel vs. depth for later analysis regarding glucose consumption and lactate production.

Calculating Oxygen and Glucose Consumption

The oxygen and glucose consumption rate (OCR and GCR, respectively) and LPR of myoblasts in the gel constructs were calculated using a custom MATLAB algorithm. These calculations are similar to oxygen and glucose consumption calculations previously performed in tissues and tissue engineered constructs.27,36,47 In brief, the data collected via oxygen microelectrodes and bioluminescence imaging were fit to Fick’s second law of diffusion:
$$ k\frac{{dP}} {{dt}}\, = \,Dk\frac{{\partial ^2 P}} {{\partial x^2 }} - Q, $$
where k is the Bunsen solubility coefficient of oxygen (k = 1 × 10−9 mol/cm3 mmHg at 37 °C, k = 1 for glucose and lactate, as they are already in molar concentrations), P is the oxygen tension (or glucose/lactate concentration), D is the diffusion coefficient in the gel (Doxygen = 2 × 10−5 cm2/s,6Dglucose = 5 × 10−6 cm2/s and Dlactate = 9.0 × 10−6 cm2/s24,25), t is time, x is depth into the gel, and Q is the consumption rate (oxygen or glucose) or lactate production rate.
For oxygen and glucose, Q was modeled using first order Michaelis–Menten kinetics:
$$ Q\, = \,\frac{{PV_{{\hbox{O}}_2 \max } }} {{P_{\hbox{m}} \, + \,P}}, $$
where Pm is the oxygen or glucose concentration and \( Q = {\raise0.5ex\hbox{$\scriptstyle 1$} \kern-0.1em/\kern-0.15em \lower0.25ex\hbox{$\scriptstyle 2$}}V_{{\hbox{O}}_2 \max }\), the maximum rate of oxygen or glucose consumption per unit volume of the construct. Therefore, combining Eqs. (1) and (2) and setting the equation to steady state conditions, we obtain
$$ {\hbox{0}}\,{\hbox{ = }}\,Dk\frac{{\partial ^2 P}} {{\partial x^2 }}\, + \,\frac{{PV_{{\hbox{O}}_2 \max } }} {{P_{\hbox{m}} \, + \,P}}. $$
For calculating LPRs, lactate data were fit using an exponential equation, and Q (LCR) was solved using central finite differences,
$$ Q(x)\, = \,Dk\frac{{P(x - 1)\, - \,2P(x)\, + \,P(x + 1)}} {{\Delta x^2 }} $$
By fitting the oxygen microelectrode and bioluminescence imaging data to the resulting equation using a MATLAB non-linear least squares curve-fitting algorithm (lsqcurvefit.m) and numerical partial differential equation solver (bvp4c.m), we obtained the average OCR, GCR, and LPR at depths throughout the construct. Cell viability per depth was calculated by fitting to the equation:
$$ V(x)\, = \,Ae^{( - Bx)}, $$
where V is viability, and A and B were solved by fitting a decaying exponential36 to a plot of NBT cell viability vs. depth measured in the constructs. OCR, GCR, and LPR per cell were calculated by multiplying solved Q (Eqs. 2 and 4) and viability values and dividing by the average cell density of each 50 μm of construct depth.

Statistical Analysis

Data are expressed as the mean ± standard error of the mean. To assess the success of the curve-fitting algorithms to calculate consumption and/or production terms, R2 values and the root mean squared (RMS) of the residuals of the fit were calculated. Comparisons of oxygen, glucose, and lactate levels and consumption (or production) with gel depth were performed using one-way ANOVA with Fisher’s PLSD post-hoc analysis between individuals, when ANOVA showed significance. Significance was set at p < 0.05. A two-sample t-test of correlation coefficient with a Fisher’s z-transformation was used to test whether staining of NBT and TUNEL were similarly correlated with depth into the gel constructs. To test the success of the hypoxia system perturbation, differences in oxygen levels, oxygen consumption rate, and viability vs. depth were compared in control and hypoxia constructs using two-way ANOVA. When significance (p < 0.05) was found, Fisher’s PLSD test was used to find pairwise differences. When significance of interaction coefficient allowed, oxygen values and consumption rates and viability were split by depth to determine differences between control and hypoxia results at each depth.


Cell Viability Results

As indicated in Fig. 3, the gel constructs successfully provided an oxygen/nutrient gradient that transitioned from full cell viability to extensive apoptosis—similar to infarcted myocardium.26,40,46,53 Cell death increased with depth, such that 70+% of cells were non-viable by 2750 μm into the gel (Fig. 3). TUNEL and NBT staining were similar (p ≥ 0.05, ANOVA), suggesting that the majority of cell death was due to apoptosis, rather than necrosis. A two-sample t-test for correlation coefficient with a Fisher’s z-transformation showed that NBT and TUNEL staining were similarly correlated with depth (p > 0.5).
Figure 3

Percentage of NBT positive cells (♦) and TUNEL negative cells (◊) as a function of construct depth. NBT staining suggests that cell death begins immediately within the gel, but 70% cell death does not occur until approximately 2800 μm depth. The R2 values for exponential curve fits were 0.934 for NBT data and 0.862 for TUNEL data

Curve Fitting

Curve fitting for oxygen consumption resulted in an R2 of 0.994 and an RMS of the residuals less than 2% of the maximum oxygen level. For glucose consumption, curve fitting of control conditions resulted in an R2 of 0.897 and an RMS of residuals of less than 4.5% of the maximum glucose levels. Finally, for lactate production calculations, curve fitting resulted in an R2 of 0.902 and an RMS of the residuals less than 4% of the maximum lactate level. The curve fit to the viability data had an R2 of 0.934 for NBT data and 0.862 for TUNEL data. For the system perturbation, the curve fit for oxygen data had an R2 of 0.928 with an RMS of residuals of less than 7.5% of the maximum.

Oxygen Measurements and Oxygen Consumption Rate (OCR)

Plots of oxygen tension vs. construct depth (Fig. 4) showed a decrease from 56.3 ± 3.53 mmHg at the gel surface to 0 mmHg at 2000 μm. The OCR of the myoblasts dropped from 1.04 ± 0.19 fmol/min/cell at the surface to 0.0 fmol/min/cell at 2000 μm into the gel. Oxygen levels had significantly decreased from surface conditions by 100 μm into the gel (p < 0.05); however, cellular oxygen consumption rates remained high. OCR was significantly reduced from surface consumption rates beginning at 1500 μm depth into the constructs (p < 0.05, Fig. 4).
Figure 4

Mean oxygen profile (♦) ± SEM and mean oxygen consumption rates (◊) as a function of depth. Significant differences in oxygen level at each point from the surface of the gel are marked with * (p < 0.05). Differences in oxygen consumption rate from the rate at the surface of the gel are marked with ** (p < 0.05)

Glucose Levels and Glucose Consumption Rate and Lactate Levels and Lactate Production Rate

A representative bioluminescence image data from which glucose and lactate concentrations were calculated is shown in Fig. 5. Average glucose profiles ± the standard error of the mean (SEM) of the six constructs were plotted vs. gel construct depth (Fig. 6). Glucose levels dropped from 4.83 ± 0.29 to 2.51 ± 0.30 mM through the 3000 μm depth. Glucose levels were significantly decreased from surface conditions by 1100 μm depth into the gel (p < 0.05). The GCR of the myoblasts increased from 1.67 ± 0.18 fmol/h/cell at the surface of the gel to 4.26 ± 0.37 fmol/h/cell at 3000 μm depth. However, GCR did not significantly increase from surface GCR until 1500 μm into the constructs (p < 0.05, Fig. 6).
Figure 5

Representative pseudo-colored image used to calculate glucose and lactate levels within the gels

Figure 6

Mean glucose profile (♦) and mean glucose consumption rates (◊) of cells. Differences in glucose level at each point from the surface of the gel are marked with * (p < 0.05, Fishers PLSD). Differences in glucose consumption rate from the rate at the surface of the gel are marked with ** (p < 0.05, Fishers PLSD)

Lactate levels increased from 2.74 ± 0.25 mM at the gel surface to 4.84 ± 0.21 mM at 3000 μm depth, while LPR increased from 1.05 ± 0.09 to 7.53 ± 0.80 fmol/h/cell over the same distance. Lactate levels significantly increased from surface conditions to 300 μm distance into the gel, while LPR increased significantly from surface production rates to 1500 μm into the constructs (p < 0.05, Fig. 7).
Figure 7

Mean lactate profile (♦) and mean lactate production rates (◊) of cells. Differences in lactate level at each point from the surface of the gel are marked with * (p < 0.05, Fishers PLSD). Differences in lactate production rate from the rate at the surface of the gel are marked with ** (p < 0.05, Fishers PLSD)

The combined data show that over 75% of cells died by 3000 μm depth into the constructs. At this depth, mean oxygen levels dropped below 0.1 mmHg, but glucose levels remained above 2 mM.

System Perturbation

Oxygen tension, oxygen consumption, and cell viability were quantified in constructs subjected to a 4-fold decrease in oxygen tension (i.e., ∼5% external oxygen tension) during both culture and oxygen measurements. Oxygen tension decreased from a high of only 14.7 ± 0.25 mmHg at the gel surface to 0 mmHg at a depth of 1700 μm, significantly lower than control constructs at each measured depth (p < 0.05). At the gel surface, the myoblast OCR was 0.99 ± 0.11 fmol/min/cell, similar to OCR in control constructs. OCR under 5% oxygen culture conditions decreased to 0.0 fmol/min/cell by 1700 μm into the construct, significantly shallower than in control constructs (p < 0.05, Fig. 8a). Further, viability was significantly reduced in constructs following hypoxia system perturbation, when compared to controls (p < 0.05, Fig. 8b).
Figure 8

(a) When split by depth, oxygen tension for constructs under hypoxia (▲) was significantly reduced compared to control conditions (♦) for the first 1200 μm, and OCR under hypoxia perturbation (△) was significantly reduced from OCR in control constructs (◊) between 1000 and 1500 μm (p < 0.05, Fishers PLSD). Differences from control plates are marked for oxygen tension (*) and OCR (**). (b) Viability of myoblasts vs. depth, as measured by NBT staining for control (♦) and hypoxia (▲) conditions. Two-way ANOVA with Fisher’s PLSD showed significant differences in viability vs. depth (p < 0.05). When split by depth, viability in hypoxia plates was significantly reduced compared to control plates until 1700 μm depth. Statistically significant differences (p < 0.05, Fishers PLSD) are marked for hypoxia vs. control conditions (*)


The goal of this project was to develop an in vitro system to test the role of oxygen and nutrient deprivation on survival of myoblasts, cells portrayed as relatively “ischemia-resistant,” and used for transplantation into ischemic myocardium. To achieve this goal, a number of hurdles present in currently used systems had to be overcome. Previously, cellular oxygen and glucose metabolism was quantified primarily in 2-D culture systems or in stirred cell suspensions. In these studies, oxygen consumption under hypoxia has typically3,10,13 been measured as a difference between the oxygen concentrations at an input and an output of a cell culture area. Similarly, glucose consumption has commonly been examined by measuring the average drop in glucose in media over time.3,10,16 Examination of cell viability and metabolism at low levels of oxygen and glucose is problematic in these systems because these methods are incapable of examining pseudo-steady state conditions that mimic in vivo scenarios. When very low glucose levels were desired, media had to be changed often to maintain a consistent level of glucose to the cells. Unfortunately, media changes are complicated by a desire to keep oxygen levels low in the cell culture system. A flow system could be used, but a filter system must be added to keep floating dying or dead cells within the system to ensure accurate viability measurements. The use of the novel 3-D culture system described in this study overcomes these limitations. This system allows a range of oxygen and glucose levels to be supplied to the cells of interest by cellular metabolism and diffusion. The absolute ranges achievable are then controlled by gel dimensions, cell density, and glucose and oxygen concentrations supplied at the surface of the gel.

After the cells reached pseudo-steady state (48 h post-seeding), we used proven techniques of gold-plated oxygen microelectrode measurements and metabolic bioluminescence imaging, combined with post-hoc modeling and histology, to measure the metabolism and viability of myoblasts. The system was robust, producing consistent data, even at low levels of oxygen and nutrients. The maintenance of residuals of less than 5% of the measured values in the curve-fitting algorithm indicated good accuracy in the curve-fitting algorithm used to calculate OCR, GCR, and LPRs. The OCR of myoblasts measured at the construct surface fit well with values found in the literature.11,28 Further, the oxygen consumption rate of myoblasts remained constant over a wide range of oxygen tension and was reduced only in the presence of very low oxygen levels. The Pm value in the Michaelis–Menten kinetic model for oxygen fit was 2.86 ± 0.88 mmHg, which was up to five times larger than published values for many tissues,9,44 but well within the range of Pm published for myocardial muscle cells.7,8,36 The maintenance of OCR over a range of oxygen tension may indicate the importance of providing adequate oxygen supply to preserve cell viability. Previous papers have shown that low oxygen can stimulate hypoxia inducible factors, such as HIF-1α, which affect metabolism of cells under hypoxia4 and can lead to growth arrest and apoptosis via the p53 pathway.5,14 GCR and LPR rose with increasing distance into the gel. These increases were likely due to a shift toward anaerobic metabolism caused by limited oxygen levels deep within the gels. The three- to five-fold increase in GCR and LPR agree well with increases reported for other mammalian cells under 20% oxygen vs. severe hypoxia.3 The gradient of ischemia achieved within the gel correlated with cell viability at each depth. Therefore, our experimental setup enabled determination of where along the oxygen/nutrient gradient cells begin to die due to ischemic conditions.

This system is also designed to allow side-by-side comparison of a variety of cell types and conditions with potential for cell transplantation. Not only can the system be used to test which cells are “best suited” to survive in ischemic environments similar to infarcted myocardium, but this setup also provides a reproducible and high throughput method to test variables designed to improve the ability of cells to survive ischemic shock. Several conditions can be examined at once, using cells from the same animal, allowing for better evaluation of techniques attempting to increase viability of the transplanted cells. In order to produce a variety of stress conditions, growth media could be replaced with media containing increased or decreased levels of glucose, and the oxygen level in the incubator could be altered to create a high or low oxygen environment at the gel surface. To show this ability to customize parameters, we examined oxygen tension, OCR, and viability of myoblasts within a construct cultured in an incubator maintained at 5% oxygen. As expected, oxygen tension within the construct was significantly reduced (p < 0.05). OCR was similar at the gel surface but was significantly reduced from 1000 to 1500 μm (p < 0.05). Finally, cell viability was significantly reduced within control constructs from the gel surface down to 1700 μm depth (p < 0.05).

The major limitation of the system presented here is the inability to account for all of the complexities of the in vivo infarct milieu. While this setup provides a more specific study of the role of hypoxia and glucose starvation on cell survival, it neglects other potential contributors to cell metabolism and function, such as cytokines and growth factors that could change cell metabolism and eschew our results. A further limitation is that in our setup, cells are evenly distributed throughout the simulated infarct. In animals or patients, however, the distribution is more complicated: cells are injected in pockets with much higher cell density. In these discrete areas of high cell density, oxygen, glucose, and lactate levels could change quickly, potentially changing metabolism over a very short scale. The reason we chose to use evenly distributed cells at low cell density was to provide greater accuracy in our measurements of oxygen and nutrient consumption to validate the system. Likely, determining how results obtained using this simplified ischemic infarct system will impact cell viability in a more complicated in vivo infarct, even ignoring other causes of cell death, will require the use of computer modeling. Ongoing modeling studies utilizing the data obtained here and accounting for multiple injection geometries, injection locations, and cell densities are underway to better understand how ischemia impacts injected cell survival.

Future studies will focus on two areas: a side-by-side comparison of the viability of various cell types used or proposed for cardiac cell therapy—including bone marrow, myoblasts, and cardiac or other tissue derived stem cells—to determine resistance to ischemic cell death followed by studies in which we alter cell or milieu-dependent parameters in an attempt to increase cell survival, such as decreasing the glucose or oxygen consumption rates of the transplanted cells. By creating an in vitro system where we can alter either milieu or cells, we expect to be able to develop ways to improve the survivability of cells in an ischemic environment similar to myocardial infarct scar, and thereby help to maximize the improvements seen following cell therapy for myocardial infarction.


The described 3-D system can be used to measure the metabolism and survival of cells under a variety of oxygen and glucose levels likely encountered after transplantation into infarcted myocardium. Oxygen and glucose levels within the system ranged from values prevalent in uninjured tissue to levels lower than those found in infarcted myocardium. Relevant parameters of the system (e.g., available oxygen tension) could be reproducibly altered and the effect on cell viability measured. Comparison between previously published data and values measured at the surface of our system indicate that the techniques we used to measure oxygen tension, glucose concentration, OCR, GCR, and LPR within the construct were valid. It is our hope that this system will be used to study metabolism and viability of a host of transplantable cell types under a range of conditions in an attempt to improve viability of cells transplanted during cardiac cell therapy.


This work was supported in part by NHLBI/National Institutes of Health awards to Dr. Taylor (R-01 HL-63346, HL-63703). We would also like to thank Robert Nielsen and Zahid N. Rabbani for their help.

Copyright information

© Biomedical Engineering Society 2007

Authors and Affiliations

  • Bryce H. Davis
    • 1
  • Thies Schroeder
    • 2
  • Pavel S. Yarmolenko
    • 2
  • Farshid Guilak
    • 1
    • 3
  • Mark W. Dewhirst
    • 2
  • Doris A. Taylor
    • 1
    • 4
  1. 1.Department of Biomedical EngineeringDuke UniversityDurhamUSA
  2. 2.Department of Radiation OncologyDuke University Medical Center DurhamUSA
  3. 3.Department of Surgery, Division of Orthopedic SurgeryDuke University Medical Center DurhamUSA
  4. 4.Center for Cardiovascular Repair, Department of Physiology, College of MedicineUniversity of MinnesotaMinneapolisUSA

Personalised recommendations