This paper reports a detailed analysis of the effect of low oxygen conditions (hypoxia) on the reporter green fluorescent protein (GFP). It questions the feasibility of using GFP for gene expression studies under tumor conditions. Hypoxia is a characteristic of both experimental and clinical tumors. Several important factors are pointed out which need to be considered when using GFP as reporter gene. GFP fluorescence is the final product of a long and complex pathway involving transcription, translation, and posttranslational modifications. All of these steps may be affected by the availability of oxygen. We show specifically that cellular GFP fluorescence decreased with reduced oxygenation, anoxia virtually eliminated fluorescence and protein levels, and fluorescence recovery after anoxia required 5–10 h of reoxygenation. In conclusion, GFP appears to be a good marker gene to study location or movement of proteins or cells but should be used with great caution as a reporter of gene expression under tumor conditions.
The progressive development of reporter gene technology has greatly contributed to the study and understanding of cellular events associated with signal transduction and gene expression. Several genes, with easily measurable phenotypes distinguishable above a background of endogenous proteins, are used commonly as reporters in a broad range of applications, including gene transfer and expression studies (1, 2). The most widely used reporter genes encode: (a) the bacterial enzyme β-galactosidase (3); (b) the bacterial enzyme chloramphenicol acetyltransferase (4); (c) the bioluminescent protein luciferase (also known as aequorin or monooxygenase) from firefly (Photinus pyralis) or the sea pansy (Renilla reniformis; Ref. 5); and (d) the GFP3 from jellyfish (6). Because the first three reporters require exogenously added substrates and/or cofactors, they are of limited use in living organisms. GFP, however, has no such requirements (7).
The GFP from the jellyfish Aequorea victoria is a 238 amino acid polypeptide, which is highly fluorescent and stable in many assay conditions (6). Reports on its sequence (8) and studies on its expression in heterologous systems (7) made it a unique reporter gene. Applications for which GFP has been used successfully include monitoring the transfer and expression of genes in living cells and tissues, subcellular location and protein movement within living cells by fusion to genes of interest (9), and location and fate of labeled cells within whole organisms, to trace, e.g., metastasis (10). Mutagenized GFP variants with improved fluorescence intensity and spectral qualities (11) and with reduced half-life for studies of transient gene expression (12) have increased the use of GFP in a variety of biological applications. GFP shows low toxicity and no interference with normal cellular activities and is easily detectable by fluorescence microscopy and quantifiable by FACS analysis (13).
A major advantage in using GFP as an in vivo marker protein is the lack of requirement for exogenous substrates or cofactors to produce the active fluorescent molecule (7). However, GFP requires molecular oxygen to catalyze the posttranslational cyclization to form the protein’s fluorophore (14). This requirement may be a problem when GFP is used as a reporter in biological systems where oxygen is limiting.
Low oxygen tension (hypoxia) is a common feature of both experimental and clinical tumors. Tumor hypoxia arises from insufficient and abnormal blood supply and is the result of an imbalance in oxygen delivery and consumption (15). Originally, diffusion-limited hypoxia, resulting from large intercapillary distances, was perceived as the sole cause of tumor hypoxia. However, hypoxic cells can also arise from perfusion-driven changes in oxygen supply. Such cells are subjected to rapid and reversible changes in oxygenation (16, 17).
This tumor environment provides some unique opportunities for therapy, especially gene therapy. The expression of several genes important for tumor growth and spread, including those encoding growth factors (e.g., vascular endothelial growth factor), oncoproteins, and transcription factors, has been shown to be induced by hypoxia (18, 19). The cellular response to hypoxia consists of two main components, namely, the HIF-1-dependent transcriptional regulation and a hypoxia-dependent stabilization of certain mRNAs. HIF-1 is a heterodimeric nuclear transcription factor consisting of HIF-1α, the oxygen-sensitive subunit, and HIF-1β. The transcription factor HIF-1 binds to its recognition sequence, the hypoxia regulatory element (HRE), in the vicinity of oxygen-sensitive genes. It is common to all mammalian cells, tissues, and organs tested to date with high abundance in human tumors (20, 21, 22). The use of this oxygen-sensitive gene regulation system has been proposed for targeted gene therapy (23).
Vascular endothelial growth factor-promoter-regulated GFP fluorescence in wound healing and in tumor formation has been demonstrated in vivo (24). However, little is known about how GFP fluorescence is affected by tumor conditions. The aim of the present study was to determine the feasibility of using GFP under low and variable oxygenation conditions, which are prevalent in solid tumors. Specifically, we have determined the effects of simulated tumor conditions on GFP fluorescence, GFP protein levels, and GFP mRNA levels in vitro and carried out a preliminary analysis of GFP fluorescence in solid s.c. tumors.
MATERIALS AND METHODS
Cell Line and Growth Media.
The human bladder carcinoma cell line T24 (European Collection of Cell Cultures, Salisbury, UK; Refs. 25 and 26) and its transfected derivatives were used. The cells were maintained in DMEM (Life Technologies, Inc., Paisley, UK) supplemented with 10% FCS (Sigma Chemical Co., Gillingham, UK), 2 mm l-glutamine (Life Technologies, Inc.), 100 units/ml penicillin, and 100 μg/ml streptomycin (Sigma Chemical Co.). Cycloheximide (Sigma Chemical Co.) at 0.3 mm was used to inhibit protein synthesis.
Cell proliferation and cell viability were monitored by cell counting using a hemocytometer and trypan blue (Sigma Chemical Co.) exclusion staining.
To mimic the heterogeneous oxygenation of solid tumors, cells were plated in 6-cm oxygen-impermeable dishes (Permanox; Nalge Nunc International) and maintained at 37°C under various oxygen conditions: (a) incubator: humidified air, 5% CO2; (b) anaerobic glove cabinet (DON Whitley Scientific, Ltd., Shipley, UK): 90% N2, 5% H2, 5% CO2 with palladium catalyst; and (c) air-tight Perspex boxes flushed continuously with a humidified gas mixture containing <0.0005, 0.02, 0.1, 0.3, 1, 2, 5, or 95% O2 and 5% CO2, balance N2 (BOC Gases, London, UK).
In the experiments involving anoxia, the samples were not exposed to oxygen until their final analysis by performing manipulations in the anaerobic glove cabinet and keeping tubes tightly closed outside the cabinet.
The media pH was measured in air (21% O2) and in anoxia (0% O2) and found to be the same (pH 7.5).
All DNA manipulations were performed according to standard procedures (27) using restriction enzymes, T4 DNA ligase, Mung Bean nuclease, and buffers according to manufacturer’s instructions (Life Technologies, Inc., Rosche and New England BioLabs, Herts, UK).
A modified version of the plasmid pd2EGFP-N1 (Clontech Laboratories, Basingstoke, UK), named pCONGFP, containing the 9-27 gene promoter (a low expressing, physiologically relevant promoter; Ref. 28) instead of the CMV promoter was constructed as follows. The CMV promoter was removed from pd2EGFP-N1 by AsnI-NheI restriction, followed by Mung Bean nuclease digestion and ligation to produce pΔCMVd2EGFP-N1. The plasmid pDW9-27CD2 (kindly provided by Dr. G. Stark) was linearized with HindIII and blunted, and the 9-27 gene promoter was excised by EcoRI digestion. This fragment was ligated to the pΔCMVd2EGFP-N1, which had been restricted with XhoI, blunted, and EcoRI digested to produce pCONGFP.
A competitor DNA template, to be used in measurements of steady-state levels of d2EGFP mRNA by competitive RT-PCR (see description below) was constructed by excision of a 32-bp fragment from pd2EGFP-N1 by BcgI digestion, followed by Mung Bean nuclease treatment and religation.
Transfection was performed according to the method described by Hart et al. (29). Briefly, T24 cells at 5 × 104 cells/well were transfected in air in 24-well plates for 5 h with complexes containing, at a ratio of 0.75:4:1 by weight, lipofectin reagent (Life Technologies, Inc.), integrin-binding peptide (Institute of Child Health, London), and plasmid DNA (1 μg of DNA/well final concentration of pd2EGFP-N1, pΔCMVd2EGFP-N1, or pCONGFP) in OptiMEM (Life Technologies, Inc.).
Transfected cells were maintained in complete DMEM containing Geneticin (G-418 sulfate; Life Technologies, Inc.) at the concentration of 0.5 mg/ml active drug. After 30–60 days from transfection, G-418 resistant clones were tested for the expression of the reporter protein d2EGFP (excitation maximum 488 nm, emission maximum 507 nm) by FACS analysis (FACScan; Becton Dickinson, Cowley, Oxfordshire, UK). Cells were scored positive if they showed fluorescence above controls transfected with pΔCMVd2EGFP-N1. In particular, a clone, stably transfected with pCONGFP and expressing d2EGFP, was selected and named C17.
Experimental Conditions in Vitro.
C17 stable transfectants and T24 untransfected controls were plated in 6-cm Permanox dishes at 105 or 2.5 × 105 cells/dish, allowed to attach for 24 h in air, and then, after replacement of the oxygenated media with preconditioned anoxic one, either: (a) exposed for 16 h to different oxygen conditions (Figs. 1 and 2); (b) exposed for 0–24 h to normoxia or anoxia and then reoxygenated for up to 24 h (Fig. 3); or (c) treated with cycloheximide in normoxia, anoxia, or during reoxygenation (Fig. 4).
Green fluorescence was monitored by FACS analysis, protein levels were monitored by Western blot analysis, and d2EGFP mRNA levels were monitored by competitive RT-PCR.
Green fluorescence was monitored by FACS analysis, and the results were recorded as means of the main GFP peak. Fluorescence, expressed in arbitrary units, was recorded as the ratio of the fluorescent signal produced by transfected cells compared with untransfected cells. On the fluorescence scale, untransfected T24 cells have a fluorescence of 10 units, whereas T24 cells transfected with promoter-less constructs showed a range between 9 and 20 units.
Western Blot Analysis.
Western blot analysis was performed essentially as described previously (30). Cell monolayers were lysed in Triton X-100 lysis buffer, and equal quantities of proteins from total cell lysates were separated by SDS-PAGE (12% polyacrylamide gels; Invitrogen-Novex, Groningen, Netherlands). Proteins were then transferred to nitrocellulose membranes (Micron Separations, Inc., Westborough, MA) using a Pharmacia-Biotech semidry blotter. Immunoblotting was performed with primary monoclonal anti-GFP antibodies, which detect all GFP variants (1:1000; Clontech Laboratories), and secondary peroxidase-conjugated goat antimouse immunoglobulins (1:2000; Dako, Ely, UK), according to manufacturer’s instructions. Detection of immunoreactive bands was performed using the enhanced chemiluminescence technique (ECL kit; Amersham Pharmacia Biotech, Amersham, UK). The bands were analyzed by densitometry using Visilog software (Noesis, Leshlis, Coutaboeuf, France).
To confirm equal loading, the blots, probed previously for GFP, were briefly washed, re-blocked, and probed with anti-actin monoclonal antibodies (1:1000; Sigma Chemical Co.).
Total RNA was extracted from cells using RNAzol B (Biogenesis, Ltd, Poole, Dorset, UK), according to manufacturer’s instructions. A glycogen solution (Rosche) added to the samples at the concentration of 20 μg/ml was used to enhance the precipitation of RNA in isopropanol. Possible residual genomic DNA contamination was removed by RNase-free DNaseI treatment (Sigma Chemical Co.), whereas proteins were removed by phenol: chloroform: isoamyl alcohol extraction (Life Technologies, Inc.). RNA concentration and purity were estimated by reading the absorbance at 260 and 280 nm; RNA integrity was demonstrated by 1% agarose gel electrophoresis.
First-strand cDNAs were synthesized from total RNA preparations using SuperScript II RNase H− reverse transcriptase and Oligo (dT)12–18 Primer (Life Technologies, Inc.), according to the manufacturer’s instructions. A reaction without reverse transcriptase was performed to control for genomic DNA contamination. All PCR amplifications were performed using Reddy-Load PCR Mix (Advanced Biotechnologies, Ltd., Epsom, UK) in a Mastercycler Gradient Thermo cycler (Eppendorf, Cambridge, UK). Primers were custom made by Life Technologies, Inc. For competitive PCR, a fixed amount of cDNA was mixed with serial dilutions of competitor DNA and subjected to PCR amplifications using the following primers for d2EGFP: forward, 5′ CGA CGT AAA CGG CCA CAA GTT CAG 3′ and reverse, 5′ GTC CTC CTT GAA GTC GAT GCC CTT 3′ at 0.5 μm each.
The expected product sizes were 339 bp for the endogenous d2EGFP and 303 bp for the competitor DNA. The following thermocycling conditions were used: 94°C for 5 min, followed by 30 cycles of amplification consisting of 94°C for 1 min, 48°C for 1 min, 72°C for 1 min, and a final 72°C extension for 10 min. Aliquots of PCR reactions were electrophoresed through an ethidium bromide-stained 4% MetaPhor Agarose gel (FMC Bioproduct, Rockland, ME) in 1 × Tris-Acetate-EDTA. The bands were analyzed by densitometry using Visilog software (Noesis).
The “house-keeping gene” β-actin was amplified as a control using the following primers: forward, 5′ TCA TCA CCA TTG GCA ATG AG 3′ and reverse,5′ CAC TGT GTT GGC GTA CAG GT 3′. The thermocycling and electrophoresis conditions used were the same as those described for the amplification of d2EGFP. The expected product sizes for β-actin of 250 bp (template: genomic DNA) and 155 bp (template: cDNA) were detected (results not shown).
Several technical challenges were encountered when optimizing the electrophoretic separation of competitive PCR products. Good resolution and separation between bands is essential to analyze band intensity by densitometry. Separation of the competitor band (303 bp) from the endogenous d2EGFP test band (339 bp) on an agarose gel, even at the high percentages of 2.5–3%, was found to be insufficient. The use of 4% MetaPhor gel improved separation and resolution of the bands but also revealed a third band situated between the test and the competitor band. This band has been demonstrated to be a hybrid consisting of a mixture of test and competitor DNA by several means: (a) alkaline denaturing 4% MetaPhor gel electrophoresis eliminated the hybrid (but reduced resolution compared with a native gel, results not shown) and (b) when PCRs were carried out separately on the test and on the competitor samples, they produced the expected bands of 339 and 303 bp; when these were mixed, subjected to a total protein phenol extraction (to remove Taq polymerase from previous PCR) followed by a final PCR cycle, a hybrid band of the predicted size was detected on 4% MetaPhor gel (results not shown). It is therefore clear that in competitive PCRs, where PCR products of very similar sequences are generated, the appearance of a hybrid band on high separation matrice gel electrophoresis cannot be eliminated. Because the extra band is necessarily a 1:1 hybrid of test and competitor DNA, which removes an equal amount of each pool (test and competitor), the two bands representing these two species can be compared with each other (quantified) while ignoring the hybrid.
Formula used for quantification:
Number of copies of d2EGFP mRNA/cell = [μg total RNA/μg total RNA used for RT] × 20*× [number of copies of competitor equal to cDNA**/number of cells used in extraction]
* 1/20th of total cDNA was used for competitive PCR
** determined from MetaPhor gel
Mice and Tumors.
Female severely combined immunodeficient mice from our pathogen-free colony were used. T24 tumors were initiated by s.c. injection of 106 viable and stably transfected T24 cells, highly expressing GFP (under the transcriptional control of the strong CMV promoter, pEGFP-N1; Clontech Laboratories), in 0.1 ml DMEM under the skin on the left flank. Mice were used for experiments at 30–40 days after inoculation, when tumors reached ∼40 mm3 in volume, as described below. Treatment protocols were carried out in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986 and with approval from the Ethical Review Committee of the Gray Laboratory Cancer Research Trust.
Experimental Protocol in Vivo.
Immunohistochemical detection of hypoxia in the tumors was performed using the hypoxia marker pimonidazole (31). Pimonidazole (Hypoxyprobe-1; Natural Pharmacia International, Inc., Belmont, MA) was injected i.p. (60 mg/kg in 0.2 ml PBS) into tumor-bearing mice. Tumors were excised 90 min thereafter, frozen, and stored at −80°C for cryosectioning. Sections (10 μm) were cut at three different levels throughout the tumor. An epifluorescence microscope (Eclipse TE200; Nikon United Kingdom, Ltd., Kingston Upon Thames, UK) with a narrow band filter (500–510 nm) and equipped with a custom made imaging system was used to visualize GFP fluorescence. Tumor sections were subsequently fixed in cold acetone for 10 min, stained for pimonidazole according to manufacturer’s instructions, and viewed under transmitted light.
Significance tests were carried out on the data groups using ANOVA followed by the Student t test for individual pairwise comparisons, with values of P < 0.01 considered as significant.
Effect of Different Oxygen Tensions on GFP Fluorescence, Protein, and mRNA Levels.
GFP fluorescence in human bladder carcinoma cells, T24 stably transfected with pCONGFP (clone C17), was analyzed under a variety of oxygen tensions, ranging from anoxia (catalyst-induced anoxia and gassing with 95% N2 and 5% CO2) through hypoxia (0.02, 0.1, 0.3, and 1% O2) to physiological oxygen conditions (2 and 5% O2), normoxia (21% O2), and finally, hyperoxia (95% O2; Fig. 1 a). Green fluorescence was quantified by FACS analysis, and results were recorded as fluorescence units compared with untransfected controls (which have been demonstrated to have a similar background fluorescence as cells stably transfected with the promoter-less construct, pΔCMVd2EGFP-N1). Cells under physiological oxygen tensions and hyperoxic conditions showed significantly increased fluorescence up to 161–174% of normoxia. This result was unexpected and suggests that several competing factors determine net fluorescence through this oxygen range (see “Discussion”). Cells under hypoxia showed fluorescence similar to that of cells under normoxia, except for 0.02% O2 where significantly reduced fluorescence was observed (P < 0.01). Severe catalyst-induced anoxia reduced GFP fluorescence significantly (33% of normoxia) compared with all other oxygenation conditions (P < 0.01). Normoxic fluorescence was less than both physiological and hyperoxic conditions.
To assess whether the variations in fluorescence observed under different oxygen tensions were because of changes in protein expression or fluorophore formation, Western blot analysis was performed. Four different oxygen conditions were investigated: catalyst-induced anoxia (0% O2), hypoxia (0.1% O2), physiological O2 (2% O2), and normoxia (21% O2). C17 cells were exposed for 16 h to the different oxygenation conditions, and whole cell lysates were used for Western blot analysis (Fig. 1 b). A band of about 31 kDa corresponding to the d2EGFP was detected. A band of 42 kDa, corresponding to the internal control actin, was present at similar intensity in all conditions analyzed, confirming that equal amounts of total proteins had been loaded. Quantification of chemiluminescent signal by scanning densitometry showed similar GFP protein levels under physiological oxygenation compared with air (88% of normoxic value). GFP was reduced under hypoxia (41% of normoxic value), and little or no protein was detected under anoxia.
To analyze if the reduced protein levels were attributable to a reduction in the GFP mRNA pool, competitive RT-PCR studies were carried out in parallel to the protein assays. No significant differences in the steady-state levels of d2EGFP mRNAs were found under the four oxygen conditions analyzed (Fig. 2).
Growth characteristics recorded after 16 h incubation under the four different oxygen conditions showed a reduction of viable cells under hypoxia (73%) and anoxia (52%) compared with air (defined as 100%), with no difference under physiological conditions. The reduction in cell numbers was attributable to a reduction in growth rather than an increase in cell death. C17 cells showed a total cessation of growth under anoxia, with anoxia-related cell death up to 13% after 31 h under anoxia (results not shown).
GFP Fluorescence Over Time in Normoxia and Anoxia.
The effect of oxygen deprivation and reoxygenation on GFP fluorescence was analyzed over time. Preplated C17 cells were exposed to normoxic or anoxic conditions, and the green fluorescent signal was monitored by FACS analysis every 2 h for up to 12 h of incubation and at 24 h (Fig. 3 a). No significant fluctuations in fluorescence were detected in normoxia over the period analyzed. However, cells incubated in anoxia showed, after an initial 2-h delay, a significant reduction of fluorescence over the first 8 h to background levels.
It is known that in a cell-free system maintained in air at room temperature, a synthetic GFP precursor molecule rearranges to form an active fluorophore able to emit green fluorescence under near-UV light in ∼5 h (32). To observe the rate of fluorescence recovery, C17 cells were moved from long-term anoxia to air, and fluorescence was monitored by FACS analysis (Fig. 3, a and b). Results showed that 5–10 h of reoxygenation were necessary for cells to recover to the levels of fluorescence present in cells constantly maintained in normoxia. Parallel Western blot analysis showed an increase in GFP protein associated with reoxygenation (Fig. 3 c).
C17 cells incubated in the presence of cycloheximide to inhibit protein synthesis in normoxia showed a loss of fluorescence over time, which paralleled the loss of fluorescence in cells under anoxia (Fig. 4,a). Cells placed under anoxic conditions while being treated with cycloheximide showed a similar rate of loss of fluorescence, showing that the treatments were not additive. When anoxic cells were reoxygenated in the presence of cycloheximide, they showed a rapid but transient increase in fluorescence (Fig. 4,b). It is unlikely that this is because of a new protein synthesis and may therefore be indicative of a small pool of noncyclised GFP. Cells under long-term anoxia (after 16 h of anoxia) treated with cycloheximide showed no further reduction in fluorescence (Fig. 4 b).
GFP Fluorescence and Hypoxia in Solid Tumors.
To determine the correlation between GFP fluorescence and hypoxia in solid tumors, seven s.c. T24 tumors initiated by injection of stably transfected T24 cells expressing GFP were analyzed for green fluorescence and assessed for hypoxia using the hypoxia marker pimonidazole. Distribution and intensity of GFP fluorescence differed both between tumors and within a tumor, suggesting different levels of GFP expression in different tumor areas (Fig. 5, a and c). Subsequent staining of the same tumor sections for pimonidazole showed that staining for pimonidazole was also heterogenous (Fig. 5, b and d). Although some tumors showed diminished fluorescence in areas of pimonidazole staining (Fig. 5, a and b), others showed GFP fluorescence overlapping into areas of hypoxia (Fig. 5, c and d).
Our study reports the detailed studies of the effect of simulated tumor conditions on the reporter GFP. It questions the feasibility of using GFP for gene regulation studies in an in vivo system. Our findings identify several important factors that need to be considered when analyzing gene expression by monitoring GFP fluorescence. Final fluorescence is used as a surrogate marker for a long and complex pathway starting with the rate of transcription, translation, and posttranslational modifications, which all contribute to the resulting emission of green fluorescence. For gene regulation studies, conclusions about the rate of transcription are drawn by analyzing the strength of the fluorescence signal. Only by the careful analysis presented here could one possibly approach this difficult approximation.
It is known that the formation of the GFP final fluorophore, a posttranslational modification resulting from spontaneous cyclization and oxidation of the protein’s amino acids -Ser65 (or Thr65)-Tyr66-Gly67-, requires molecular oxygen (14). Our data now provides information on the amount of oxygen required to visualize the fluorophore in live cells. A general dose response to oxygenation was found, where higher oxygen concentrations resulted in higher fluorescence and lower concentrations resulted in reduced fluorescence (Fig. 1,a), but statistical analysis showed that only catalyst-induced anoxia resulted in a significant reduction of fluorescence signal compared with all other oxygenation conditions. However, normoxia (21% O2) is not a realistic oxygen concentration under physiological conditions, where median pO2 values range from 24–66 mmHg (3.1–8.7% O2; Ref. 33). Importantly for in vivo studies, fluorescence under radiobiological hypoxic conditions (<0.3% O2) was significantly lower (by more than 50%) compared with physiological conditions (2 and 5% O2). Normoxic fluorescence is lower than the signal detected under both physiological and hyperoxic conditions, the reason for which remains unclear. It may be possible that the reasons for increased fluorescence compared with normoxia may be different, in that physiological oxygen concentrations may be optimal for fluorescence, whereas hyperoxia may induce a stress response resulting in increased fluorescence. Additionally, although Western analysis showed that hypoxia and anoxia reduced GFP protein levels, physiological conditions did not increase the levels compared with normoxia, which does not correlate with the fluorescence data (Fig. 1,b). The variation in protein level with oxygenation is likely to depend on both the rate of translation (which is known to be reduced under severe hypoxia, Ref. 34), as well as protein half-life. The destabilized GFP used in the in vitro experiments described here has the mouse ornithine decarboxylase amino acid residues 422–461, which contain a Proline, Glutamic Acid, Serine, Threonine (PEST) amino acid sequence targeting proteins for degradation (12) fused to the COOH terminus. The cycloheximide results presented here (Fig. 4,a) showed that a similar rate of protein degradation occurred in air and anoxia, indicating that reduced protein half-life may not be the main reason for the loss of GFP under anoxia. The variation in fluorescence and protein levels appears not to be attributable to variations in the GFP mRNA pool. Using competitive RT-PCRs, no significant difference in the GFP species (14–21 copies per cell) between different oxygenation conditions was detected (Fig. 2).
Interestingly, a transient increase in fluorescence was detected when cells were treated with cycloheximide during reoxygenation (Fig. 4,b). This may indicate the build-up of a small pool of nonfluorescent protein under anoxia, which only requires oxygen to form the fluorophore. However, nonfluorescent GFP appears not to be the main reason for a lack of fluorescence under anoxia, because Western analysis showed a significant loss of protein as well (Figs. 1,b and 3,c). It should be noted that the limit of detection for GFP using Western analysis might differ from the level of detection by FACS analysis. An alternative explanation for the transient increase in fluorescence during cycloheximide-treated reoxygenation may be a delay in the inhibition of protein synthesis. This appears unlikely judging from the immediate effect seen in air and anoxia (Fig. 4 a).
For this in vitro analysis, a physiologically relevant promoter was chosen rather than a strong viral promoter such as CMV, which may have saturated the fluorescence signal. The 9-27 promoter was used previously as the backbone for the insertion of hypoxia-regulated elements in a feasibility study of transcriptional regulation for gene therapy (23). In the previous study, as well as the current one, the 9-27 promoter demonstrated low transcriptional activity and did not respond to hypoxic conditions. These qualities make it a good candidate for further optimization of gene therapy constructs and for gene expression studies under physiological conditions. Fluorescence microscopy requires a much higher level of expression of GFP than the 9-27 promoter permitted. Hence, in the in vivo work described here, we have used a CMV promoter to drive GFP synthesis in the T24 tumor. Using this promoter, GFP fluorescence was visible in all analyzed tumors. However, the distribution and intensity of fluorescence differed between tumors and within a tumor, suggesting different levels of GFP expression in different tumor areas.
Reporter genes other than GFP were considered for this study. Other genetic reporters in general use, such as luciferase, β-galactosidase, and chloramphenicol transferase, are enzymes and thereby require substrates, such as luciferin, O-nitrophenyl-β-d-galactopyranoside, and chloramphenicol, respectively, to produce detectable products (27). This makes their use as real-time reporters for gene expression in vivo more difficult. However, apart from GFP, luciferase is currently the only other marker gene used for real-time live imaging (35, 36). Luciferase was therefore considered carefully.
Luciferase is a monomeric protein that does not require posttranslational modification, and therefore, its activity can be assessed immediately after translation (37). Although luciferase has successfully been used in vivo (36), it has added complications for use under tumor conditions. Luciferase requires molecular oxygen and ATP at the time of activation of luciferin to produce luminescence, both of which are in limited supply in most solid tumors (38). The use of luciferase for gene regulation studies involving cyclic AMP-modulating agents has been questioned in the past (39). It was shown that cyclic AMP-elevating agents could lead to a significant change in luciferase activity independently of a transcriptional activation of promoter elements. It appears, therefore, that the modified GFP used in this study, with its shortcomings, is still superior to other reporter genes currently available.
In addition, our preliminary in vivo results demonstrate that distribution of GFP is heterogeneous throughout the tumor, suggesting different levels of GFP expression in different tumor areas. The heterogeneity of GFP expression appeared to correlate with pimonidazole staining in some tumors, but not in others, implying that GFP can still be detected in oxygen conditions that are sufficient to bioreductively activate pimonidazole. On the other hand, GFP expression was absent in areas that were not positive for pimonidazole staining. Therefore, factors other than hypoxia may influence GFP expression, such as tumor energy status, tumor infiltration by normal cells, and genetic instability of some stably transfected cells, resulting in loss of functional incorporation of the GFP gene. In addition, tumor cells will be in different cell cycle phases and may therefore exhibit different levels of GFP. All these issues further support our concern that the use of GFP as a reporter gene for real-time in vivo studies of gene regulation is limited.
In conclusion, it appears doubtful that GFP could be used effectively as a reporter gene for real-time in vivo studies of gene regulation in tumors. Specifically: (a) cellular GFP fluorescence reduces with reduced oxygenation; (b) fluorophore formation, depending on the variant used, was shown to take from 1–5 h (14, 40), and GFP fluorescence in our system was regained over 5–10 h of reoxygenation; (c) even the PEST-destabilized version of GFP has a half-life which is unsuitable to detect rapid changes in gene expression; and (d) low oxygen reduced GFP protein levels significantly. Therefore, GFP should be used with great care as a reporter of gene expression under tumor conditions. Still, GFP is a clearly good marker gene to study location and movement of proteins or cells and can be used effectively in normal tissues under physiological oxygenation conditions.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported by the Cancer Research Campaign Grant SP2292/0102.
The abbreviations used are: GFP, green fluorescent protein; d2EGFP, destabilised enhanced GFP; RT-PCR, reverse transcription-PCR; FACS, fluorescence-activated cell sorting; HIF-1, hypoxia-inducible factor 1; CMV, cytomegalovirus.
We thank Mick Woodcock for help with FACS, Kevin Prise for assistance with Visilog software, George Wilson and Frances Daley for help with the pimonidazole experiments, Gemma Lewis for contribution to the in vivo experiments, and Richard Foxon for critical discussions.