Gradients of Po2 between capillary blood and mitochondria are the driving force for diffusional O2 delivery in tissues. Hypoxic microenvironments in tissues that result from diffusional O2 gradients are especially relevant in solid tumors because they have been related to a poor prognosis. To address the impact of tissue O2 gradients, we developed a novel technique that permits imaging of intracellular O2 levels in cultured cells at a subcellular spatial resolution. This was done, with the sensitivity to O2 ≤3%, by the O2-dependent red shift of green fluorescent protein (AcGFP1) fluorescence. Measurements were carried out in a confluent monolayer of Hep3B cells expressing AcGFP1 in the cytoplasm. To establish a two-dimensional O2 diffusion model, a thin quartz glass slip was placed onto the monolayer cells to prevent O2 diffusion from the top surface of the cell layer. The magnitude of the red shift progressively increased as the distance from the gas coverslip interface increased. It reached an anoxic level in cells located at ∼220 μm and ∼690 μm from the gas coverslip boundary at 1% and 3% gas phase O2, respectively. Thus the average O2 gradient was 0.03 mmHg/μm in the present tissue model. Abolition of mitochondrial respiration significantly dampened the gradients. Furthermore, intracellular gradients of the red shift in mitochondria-targeted AcGFP1 in single Hep3B cells suggest that the origin of tissue O2 gradients is intracellular. Findings in the present two-dimensional O2 diffusion model support the crucial role of tissue O2 diffusion in defining the O2 microenvironment in individual cells.
- oxygen sensing
oxygen transport from air to the respiratory enzyme in mitochondria depends on two distinct mechanisms: convection and diffusion. In the final step of the oxygen cascade, i.e., from capillary blood to mitochondria, oxygen is transported exclusively by diffusion according to the potential difference called the Po2 gradient. Thus Po2 in individual cells is independently determined by capillary blood Po2 and Po2 gradients in tissue.
Fick's law of diffusion indicates that the Po2 gradient per unit diffusion length is the product of flux density and oxygen diffusion resistance in tissue. A number of factors, including mitochondrial respiration, the number and spatial distribution of mitochondria in a cell, the number of cells for a capillary, and the topology of the oxygen diffusion path determine oxygen flux density and oxygen path length in tissue. Thus Po2 gradients at a given metabolic condition vary considerably according to the tissue of interest. Of particular importance are solid tumors in which increases in diffusion distances due to disrupted vascular network architectures produce significant hypoxic loci (diffusion-limited oxygen delivery; 9, 24). Intratumor hypoxia may lead to a more malignant phenotype, at least in part, through a signaling via hypoxia inducible factor-1 (HIF-1, see Ref. 19 for review). Furthermore, hypoxia may endow an increased resistance to treatment (15, 18). Thus hypoxia is the hallmark of solid tumors as well as an important therapeutic target.
The magnitude and the physiological relevance of Po2 gradients in tumor tissue have been discussed using mathematical models of oxygen diffusion (10, 23). Subsequently, high spatial resolution imaging of oxygen in in vivo tumor tissue was reported using various optical techniques. Helmlinger et al. (13) determined profiles of interstitial Po2 between two adjacent vessels in human tumor xenografts at 10 × 10 μm resolution using phosphorescence quenching of albumin-bound metalloporphyrin. Intervessel interstitial Po2 profiles exhibited a variety of shapes such as flat, monotonic, and parabolic that were not previously predictable using simple diffusion models. Recently, spatial distribution of hypoxic cells (Po2 < 10 mmHg) in tumor tissue was determined using immunohistochemical staining of exogenous hypoxia markers such as nitroimidazoles (see Ref. 5 for review). These studies demonstrated hypoxic cells at a distance beyond ∼100 μm from the supplying vessels, indicating diffusional gradients of oxygen. Whereas the spatial resolution is high enough to locate hypoxic cells in tissue, histochemical techniques using exogenous or endogenous hypoxia markers in biopsy samples do not bring quantitative information regarding the magnitude of oxygen deprivation. Furthermore, these in vitro techniques do not allow repeated hypoxia imaging. Thus we undertook the present study to establish an in vitro two-dimensional oxygen diffusion model in cultured cells in which oxygen distribution is imaged with a subcellular spatial resolution using a green fluorescent protein (GFP).
Cell culture and transient expression of AcGFP1.
Human hepatoma cell line Hep3B was cultured and maintained in DMEM containing 10% fetal calf serum on a low-fluorescent polystyrene disk (Cell Disk LF, Sumilon, Japan) of 13.5 mm diameter and 0.1 mm thickness, which was placed in 35-mm culture dishes. For transient expressions of a mutant GFP (AcGFP1) in cytosol and in mitochondria, we introduced expression vectors pAcGFP1-C1 and pAcGFP1-Mito (Clontech, Madison, WI), respectively, using a linear polyethylenimine derivative (jetPEI, Polyplus-transfection, France).
Imaging of intercellular gradients of oxygen.
This experiment was conducted in Hep3B cells expressing AcGFP1 in the cytoplasm (pAcGFP1-C1 vector). At confluent density, DMEM was replaced with HEPES-buffered DMEM without phenol red (PromoCell). Immediately before the experiment, the cell disk was transferred to an airtight measuring cuvette, and a rectangular quartz glass coverslip (19 × 3.5 mm, 0.6 mm thickness) was gently placed onto the monolayer Hep3B cells. Supplementary 20 μl HEPES-buffered DMEM without phenol red was added to portions of the cell disk not covered with the coverslip. The measuring cuvette was placed on the stage of an inverted microscope with ×10 object lens (0.30 NA; IX71, Olympus, Japan) and gassed with humidified gas containing 0%-21% O2 in N2 at 3 ml/min. The temperature was regulated at 37°C. Because oxygen diffusion from the top and bottom surfaces of the cell layer was virtually abolished, this may serve as a two-dimensional model of tissue oxygen diffusion.
At least 60 min were allowed to elapse until the steady state was established. The first AcGFP1 fluorescence images (denoted as G1 and R1 for green and red channels, respectively) were acquired using a 16-bit CCD camera (SV512, PixelVision, Tigard, OR). Excitation/emission wavelengths for G1 and R1 were 475 ± 10/510 ± 12 nm and 525 ± 23/595 ± 30 nm, respectively. Exposure durations for G1 and R1 were 0.2 and 0.6 s, respectively. The cells were then illuminated with 475 ± 10 nm light (100-W mercury arc lamp, neutral density filter removed) for 220 s (PA, photo activation). Immediately after PA, the second AcGFP1 fluorescence measurements (G2 and R2 for green and red channels, respectively) were conducted. Finally, a phase contrast image was recorded to determine the distance of cells from the gas-coverslip boundary. Fluorescence images were stored in a computer. After background subtractions, cellular fluorescence was identified using image-processing software (IPLab, Scanalytics, Rockville, MD), and average intensity was calculated for each individual cell.
Imaging of intracellular gradients of oxygen.
Intracellular heterogeneities of oxygen were assessed at a subcellular spatial resolution in Hep3B cells expressing AcGFP1 in mitochondria (pAcGFP1-Mito vector). In contrast to transiently expressed AcGFP1 using pAcGFP1-C1 vector in which the fluorescent molecule freely diffuses in the cytoplasm, AcGFP1 molecules may be bound in the mitochondrial membrane. If movement of mitochondria in the cell can be ignored, mitochondria-targeted AcGFP1 could report oxygen-dependent red shift of AcGFP1 at the location of mitochondria, allowing hypoxia imaging with a subcellular spatial resolution.
The red shift of AcGFP1 in mitochondria was imaged with the technique mentioned above except for the following. First, imaging was conducted with a ×60 object lens (0.70 NA) where 1 pixel on the computer display corresponded to 0.17 μm. Second, green and red fluorescence imaging was conducted with 0.6- and 1.8-s exposures, respectively. Third, duration for PA was 420 s. In some experiments, mitochondrial respiration was stimulated by an uncoupler of oxidative phosphorylation (1 μM CCCP) to increase oxygen flux.
Because of slight but unavoidable changes in the position of cell/mitochondria between image acquisitions before and after PA, it was difficult to generate images representing changes in the G or R fluorescence before and after PA (i.e., G1/G2 image or R1/R2 image). Instead, we generated an image by normalizing the R2 image by the corresponding G2 image (R2/G2) that semiquantitatively represents hypoxic red shift of GFP fluorescence.
Red shift of AcGFP1 fluorescence in hypoxia.
Figure 1 illustrates changes in the R channel image following PA in Hep3B cells expressing AcGFP1 in mitochondria (×60 object lens). In the aerobic cells (10% O2, no coverslip), PA for 420 s resulted in similar reductions in the G and R fluorescence (G2/G1 = 0.32, R2/R1 = 0.31), presumably due to photo bleaching of the fluorophore. In contrast, PA in the anaerobic condition (superfusion with <0.001% O2 in N2, no coverslip) significantly augmented the R fluorescence, whereas the G fluorescence was reduced (R2/R1 = 1.66, G2/G1 = 0.43). Magnitude of anaerobic red shift (R2/G2) increased with PA duration where it plateaued for PA = 5–7 min (data not shown).
Figure 2 indicates that increases in the R fluorescence after PA (220 s) reflect the cellular oxygen level in Hep3B cells expressing AcGFP1 in the cytoplasm (×10 object lens). The R2/G2 value was 0.074 ± 0.003 at 10% O2 (means ± SD, n = 27), whereas it increased to 0.078 ± 0.004 at 3% O2 (n = 40), 0.101 ± 0.017 at 1% O2 (n = 23), and 0.220 ± 0.074 at <0.001% O2 (n = 50). In Hep3B cells treated with 2 mM KCN, R2/G2 value was 0.071 ± 0.004 (n = 13) at 10% O2 while it increased to 0.205 ± 0.048 (n = 10) at <0.001% O2. Increases in R2/G2 were statistically significant for superfusion O2 levels ≤3% (Welch's t-test, comparisons with 10% O2). Similar to cytoplasmic AcGFP1, Hep3B cells expressing AcGFP1 in mitochondria demonstrated significant hypoxic red shift following a PA of 420 s. The average value for R2/G2 at <0.001% O2 was 0.430 ± 0.114 (n = 10), while it was 0.074 ± 0.002 (n = 4) at 10% O2.
We examined the effect of the mitochondrial redox state on the red shift (oxidative redding, see discussion). Reductions of the mitochondrial respiratory chain by 1 μM CCCP in abundant oxygen in Hep3B cells did not provoke red shift (R2/G2 = 0.074 ± 0.002, n = 4) in mitochondrial AcGFP1. This is also true for aerobic Hep3B cells expressing AcGFP1 in cytoplasm in which the respiratory chain was fixed to the reduced state by 2 mM KCN (R2/G2 = 0.071 ± 0.004, n = 13).
We also examined whether the R fluorescence is affected by changes in pH. This was conducted in AcGFP1 (protein) in PBS of various pH values. At pH values of 6, 7, 7.4, and 8 at 36°C, R/G values did not change significantly.
Imaging of intercellular gradients of oxygen.
Figure 3 shows representative data indicating changes in the magnitude of the red shift in Hep3B cells expressing AcGFP1 in the cytoplasm. The measuring cuvette was gassed with 3% O2 in N2. Cells expressing sufficient AcGFP1 are indicated as a–i (left, phase contrast image). In contrast that R2/G2 values were comparable to those for aerobic PA in cells near the gas-coverslip interface (a–c), R2/G2 progressively increased as the distance from the boundary increased (d–i).
Figure 4 summarizes the red shift as functions of the distance from the gas-coverslip boundary. Cells near the center of the coverslip (1,750 μm from the boundary) are assumed anoxic and the R2/G2 value for these cells is indicated in means ± SD (0.306 ± 0.075, n = 6). When the cuvette was gassed with 1% O2 (∼7 mmHg), R2/G2 increased progressively with the depth and reached the level comparable to the anoxic cells at 223 μm inside the coverslip. Thus the calculated average gradient of oxygen in this two-dimensional tissue model was 0.03 mmHg/μm. Similarly, when the cuvette was gassed with 3% O2 (∼21 mmHg), R2/G2 reached the anoxic level at 687 μm from the gas-coverslip boundary, indicating the similar O2 gradient of 0.03 mmHg/μm. It is of note that the oxygen gradient was considerably reduced when the mitochondrial respiration was suppressed with 2 mM KCN.
Imaging of intracellular gradients of oxygen.
We further extended the present technique to address the intracellular gradients of oxygen. Figure 5 demonstrates the inter- and intracellular gradients of the red shift in Hep3B cells treated with 1 μM CCCP. The cuvette was gassed with 3% O2 where oxygen diffuses from the left to the right in Fig. 5. The left cell (Fig. 5, A/A′) was located at 314 μm inside from the gas-coverslip boundary, whereas the right cell (B/B′) was at 360 μm from the boundary. In the upstream cell (A/A′), following 420 s PA, R/G value increased from 0.076 to 0.168, indicating a hypoxic red shift. In contrast, the downstream cell (B/B′) showed much larger red shift, and R/G values increased from 0.085 to 0.641. These results are consistent with the existence of large intercellular oxygen gradients along the direction of oxygen flow. Furthermore, imaging of R2/G2 in the right cell demonstrated significant intracellular gradients of the red shift consistent with the direction of oxygen flow, indicating intracellular gradients of oxygen. We conducted 10 experiments in which three were discarded due to technical problems. Significant intracellular gradients in R2/G2 that are consistent with the direction of oxygen diffusion were demonstrated in four experiments.
Imaging of hypoxia with a subcellular spatial resolution using GFP.
Elowitz et al. (7) first reported that photoactivation of enhanced GFP (EGFP) with a strong blue light turns the green fluorescence red. They used this technique to define the diffusional movement of EGFP molecules in the cytoplasm of living bacteria. Recently, our group extended this technique for imaging hypoxia/anoxia in cultured mammalian cells, acutely isolated single cardiomyocytes, crystalloid-perfused isolated heart, and blood-perfused kidney and liver (21). Although this technique depends on sufficient expression of GFP, it offers, ideally, a spatial resolution at the single molecule level and is particularly suitable for imaging hypoxia in living tissue.
Although the precise mechanism remains to be elucidated, green-to-red photoconversions have been reported in various GFPs either in the presence of electron acceptors (oxidative redding, Ref. 3) or in anaerobic conditions (anaerobic redding, Ref. 7). In the present study, we focused on the anaerobic redding of AcGFP1 as a tool for in vivo oxygen imaging. It is presumable that anaerobic redding in vivo may also be affected by changes in the redox state in mitochondria (oxidative redding) that are associated with decreases in cellular oxygen level. This might be particularly relevant to the hypoxic red shift in mitochondria-targeted AcGFP1. However, we demonstrated in Hep3B cells that reductions of mitochondrial respiratory chain by respiratory inhibitors in abundant oxygen did not provoke red shift. Thus, in the present study, observed red shift in cytoplasmic and mitochondrial AcGFP1 primarily reflects decreases in oxygen concentration in the microenvironment of respective molecules.
As shown in Fig. 2, increases in R2/G2 were significant at 3% cuvette O2 (compared with 10% cuvette O2) but were more clearly detected at cuvette O2 levels ≤1% (∼7 mmHg). Fortunately, this sensitivity range substantially overlaps Po2 ranges representing mild to severe hypoxia in vital organs. Furthermore, this detection range covers oxygen levels found in hypoxic tissue area in various solid tumors (≤2.5 mmHg; Ref. 24) and is thus physiologically and pathophysiologically relevant.
Oxygen gradients in in vivo tissue model.
Because the present in vivo tissue model is simplistic, Fick's law of diffusion may be readily applicable. If we adopt values for the oxygen consumption rate (9.6 nmol·min−1·mg protein−1) and cell density (0.13 mg/cm2) reported in monolayer Hep3B cells in a culture dish (14) and Krogh's diffusion constant (Ko2) determined in hamster retractor muscles (9.5 × 10−10 ml O2·cm−1·s−1·mmHg−1; Ref. 2), Fick's law of diffusion would yield tissue oxygen gradients with a magnitude of ∼0.06 mmHg/μm at 37°C. If the Ko2 value determined in tumor tissue (4.2 × 10−10 ml O2·cm−1·s−1 mmHg−1; Ref. 11) is used, then the Po2 gradient would be 0.13 mmHg/μm at 37°C. As shown in Fig. 4, the red shift reached the anoxic level at ∼220 μm and ∼690 μm from the gas-coverslip boundary at 1% and 3% cuvette O2, respectively. Thus the average tissue oxygen gradient was 0.03 mmHg/μm in the present model. The value is compatible with the theoretical predictions, suggesting that the present model accurately represents two-dimensional diffusion of oxygen. Together, in monolayer Hep3B cells (cell diameter 20–25 μm) at resting metabolic rate, Po2 drops slightly less than 1 mmHg per cell.
Oxygen gradients and cellular hypoxic sensing.
In a range of oxygen at which HIF-1α is stabilized, either pharmacological or genetic inhibition of mitochondrial respiration has been demonstrated to prevent HIF-1α accumulation (Ref. 4 for HIF-2α, 6, 12), while PGC-1α mediated mitochondrial biogenesis upregulates HIF-1α (16). These observations strongly suggest that mitochondria are involved in the regulation of HIF-1α induction in mild hypoxia. Of particular interest is the hypothesis that mitochondrial respiration critically defines intracellular oxygen levels and thereby HIF-1α stability through changes in diffusional oxygen gradients (12, 22, 25). In fact, elevations of cellular oxygen levels following mitochondrial inhibition, presumably as a result of dissipation of diffusional oxygen gradients, have been demonstrated in cultured cells (6, 12, 17), whereas increases in mitochondrial mass resulted in the decrease of oxygen levels (16). Although this hypothesis clearly explains why mitochondrial inhibition downregulates HIF-1α in cultured cells in mild hypoxia and further addresses the physiological role of diffusional oxygen gradients, this has not been proven in vivo.
Mitochondrial respiration sets both extracellular and intracellular gradients of oxygen. In the case of cultured cells in which the above-mentioned experiments were carried out, oxygen diffuses from the atmosphere to the cell through a layer of culture medium. The oxygen diffusion distance is several hundred times longer in culture medium than in a cell, whereas flux density and oxygen diffusion resistance may be the same in culture medium and in monolayer cells. Consequently, in conventional cultures, the majority of oxygen gradients resides in the extracellular medium, suggesting that changes in mitochondrial respiration affect cellular oxygen levels mainly through changes in the extracellular oxygen gradient depending on the thickness of the medium layer outside the cell (14, 28). Thus the role of mitochondrial respiration in cellular oxygen sensing in vivo cannot be addressed until direct measurements of inter- and intracellular oxygen gradients are carried out.
The present technique utilizing GFP as a probe for intracellular oxygen satisfies these demands. The results in monolayer Hep3B cells at a resting metabolic rate unequivocally demonstrated considerable intercellular oxygen gradients (Figs. 3 and 4). The origin of the oxygen gradient is most likely intracellular because at increased oxygen flux we detected significant intracellular heterogeneities of the red shift (Fig. 5). Furthermore, abolition of mitochondrial respiration by 2 mM KCN decreased intercellular oxygen gradients and elevated intracellular oxygen in cells distant from the oxygen source (Fig. 4). Although the two-dimensional cell culture model significantly differs in many aspects from the three-dimensional in vivo tissue (26), present results strongly indicate the impact of diffusional oxygen gradients upon the oxygen microenvironment of individual cells.
The induction of HIF-1α at low Po2 was demonstrated to suppress mitochondrial respiration in cultured cells (1, 8, 17, 20, 27). In in vivo conditions, decreases in mitochondrial respiration would reduce diffusional oxygen gradients in tissue and thereby elevate Po2 at HIF-1α. Present results support the critical role of mitochondrial respiration in defining oxygen levels in individual cells. Thus the tissue oxygen gradient would comprise a negative feedback regulatory mechanism for HIF-1α induction and HIF-mediated gene expressions. Because tissue oxygen gradients increase in proportion to the diffusion distance, these regulatory mechanisms may be particularly important in diffusion-limited solid tumor cells in which the oxygen diffusion path is considerably longer than in normal tissue due to abnormal microcirculatory architectures (24).
The present technique potentially offers a wide range of applications. With the use of subcellular localization vectors, GFP can be expressed in a specific organelle or subcellular structure. Thus targeted GFP expression in a specific organelle may allow us to explore differences in the oxygen microenvironment among these subcellular structures with different functions. Because measurement of the red shift can be performed repeatedly (21), another potential application may be in vivo tracking of the oxygenation status in GFP-expressing cancer xenografts in immunodeficient mice that have undergone various interventions such as irradiation and anticancer drug treatment.
In summary, we evaluated the feasibility and validity of our newly proposed technique for imaging oxygen levels in GFP-expressing cultured cells. With this technique, we provided evidence for intercellular and intracellular oxygen gradients. Furthermore, using an accurately defined two-dimensional tissue oxygen diffusion model, we experimentally confirmed the feasibility of previous simple mathematical models for tissue oxygen diffusion. On the basis of these data, we addressed the impact of diffusion-limited oxygen delivery in the physiology and pathophysiology of solid tumor tissues.
No conflicts of interest, financial or otherwise, are declared by the author(s).
We thank to Dr. Jerry D. Glickson for reading the manuscript.
- Copyright © 2010 the American Physiological Society