Cancer Research Targets Telomeres
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bundy, J. G.
Right arrow Articles by Brindle, K. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bundy, J. G.
Right arrow Articles by Brindle, K. M.
[Cancer Research 66, 7606-7614, August 1, 2006]
© 2006 American Association for Cancer Research

Cell, Tumor, and Stem Cell Biology

Metabolic Consequences of p300 Gene Deletion in Human Colon Cancer Cells

Jacob G. Bundy1, N. Gopalakrishna Iyer2, Michelle S. Gentile2, De-En Hu1, Mikko Kettunen1, Ana-Teresa Maia2, Natalie P. Thorne2, James D. Brenton2, Carlos Caldas2 and Kevin M. Brindle1

1 Department of Biochemistry and 2 Cancer Genomics Program, Department of Oncology, Hutchison/Medical Research Council Research Centre, University of Cambridge, Cambridge, United Kingdom

Requests for reprints: Kevin M. Brindle, Department of Biochemistry, University of Cambridge, 80 Tennis Court Road, Cambridge CB2 1GA, United Kingdom. Phone: 44-1223-333674; Fax: 44-1223-766002; E-mail: kmb{at}mole.bio.cam.ac.uk.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Metabolite profiling using 1H nuclear magnetic resonance (NMR) spectroscopy was used to investigate the metabolic changes associated with deletion of the gene for the transcriptional coactivator p300 in the human colon carcinoma cell line HCT116. Multivariate statistical methods were used to distinguish between metabolite patterns that were dependent on cell growth conditions and those that were specifically associated with loss of p300 function. In the absence of serum, wild-type cells showed slower growth, which was accompanied by a marked decrease in phosphocholine concentration, which was not observed in otherwise isogenic cell lines lacking p300. In the presence of serum, several metabolites were identified as being significantly different between the two cell types, including glutamate and glutamine, a nicotinamide-related compound and glycerophosphocholine (GPC). However, in the absence of serum, these metabolites, with the exception of GPC, were not significantly different, leading us to conclude that most of these changes were context dependent. Transcript profiling, using DNA microarrays, showed changes in the levels of transcripts for several enzymes involved in choline metabolism, which might explain the change in GPC concentration. Localized in vivo 1H NMR measurements on the tumors formed following s.c. implantation of these cells into mice showed an increase in the intensity of the peak from choline-containing compounds in the p300– tumors. These data show that NMR-based metabolite profiling has sufficient sensitivity to identify the metabolic consequences of p300 gene deletion in tumor cells in vitro and in vivo. (Cancer Res 2006; 66(15): 7606-14)


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
The nuclear protein p300 is a transcriptional coactivator, which possesses intrinsic acetyltransferase activity and is closely related to cyclic AMP-responsive element binding protein (CREB)–binding protein (1). The two proteins share many biological roles but are not isofunctional. Previous data suggest that both p300 and CREB-binding protein are involved in a number of different pathways, which affect cell cycle control, apoptosis, differentiation, and proliferation. Despite the large volume of published data, analysis of p300 function has not been straightforward for several reasons. First, there are a large number of cellular targets known to be associated with it. Second, it is often difficult to separate the functional specificity of p300 and CREB-binding protein. In fact, few studies even attempt to make the distinction despite mounting evidence that the homologues do play separate roles. Supporting this fact, it is evident that p300 and CREB-binding protein are mutated in different epithelial tumor types, suggesting different roles in carcinogenesis (2–5). Thus, it would be useful to have a comprehensive method for assessing the overall global effect of p300 gene deletion on cellular physiology. This would both define a phenotype for p300 deletion and give possible new insights into the cellular function of the protein.

Transcriptomic analysis using cDNA microarrays could be useful both in detecting changes in the overall transcription network consequent to p300 deletion and in identifying novel p300 targets and pathways. However, identifying changes at the transcript level may be insufficient to delineate the global cellular effect of p300 disruption. There may be changes in cell physiology that are not represented at the level of the transcriptome, either because of posttranslational modification of proteins or because of the metabolic compensation that can result from network rearrangements (6). Comprehensive studies have shown a low degree of correlation between mRNA levels and actual protein concentrations (7, 8).

An alternative strategy for global cellular analysis involves the various techniques that can be collectively termed metabolic profiling or metabolomics. Metabolomics specifically refers to the comprehensive analysis and statistical interpretation of the complement of small-molecule metabolites in a biological sample. As such, it could be used as a complementary functional genomics technique that allows a direct assessment of the effects of genetic changes on cellular physiology and can add a great deal to the interpretation of transcript profile data (e.g., refs. 9–11). The two major analytic platforms for metabolomics are mass spectrometry and nuclear magnetic resonance (NMR) spectroscopy (12).

NMR-based methods have been important in oncology because of the potential of using in vivo magnetic resonance spectroscopy for noninvasive diagnosis; previous work has included attempts to define biochemical markers associated with different tumor types, as well as the use of pattern-recognition methods for automated diagnosis (e.g., refs. 13–15; reviewed in ref. 16). Furthermore, it has been postulated that metabolomic data might be used to generate new testable hypotheses about biological mechanism (e.g., in understanding the mechanism of drug action or the function of a specific gene product). Unfortunately, this has remained difficult thus far, as there is neither a general model that comprehensively relates metabolite profile changes to genomic/proteomic information nor any widely applicable method for generating hypotheses from metabolomic data. However, based on experiments in the yeast Saccharomyces cerevisiae, it has been suggested that the congruence of metabolite patterns in strains deleted for two different genes might indicate that these genes affect similar areas of cellular metabolism. Therefore, if the function of one of the genes is already known, then the function of the second unknown gene might be inferred (17). The identity of the specific metabolites perturbed by gene deletion might also lead to testable hypotheses on mechanism. For example, tumors deficient for hypoxia-inducible factor 1ß were found to have reduced ATP concentrations. Changes in the concentration of other metabolites, particularly a reduction in glycine, led to the hypothesis that the increased glycolytic activity of cancer cells may be used to provide extra glycine as an essential substrate for ATP (purine) synthesis (18, 19).

The aims of this study were twofold: first, to determine whether metabolic profiling had sufficient specificity to differentiate between a pair of isogenic cell lines that differ by a single gene; second, to define a metabolic signature that could define p300 gene deletion under differing contexts, which might give new insights into p300 function, and to compare this signature with transcriptomic data. In these experiments, we used p300-deficient (p300–) cells, which had been previously generated by homologous recombination–mediated gene targeting in the colorectal carcinoma cell line HCT116 (3).


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Maintenance and growth of cell lines. HCT116 (American Type Culture Collection, Manassas, VA) and its derivatives were cultured in McCoy's 5A medium with 10% FCS and penicillin/streptomycin (Invitrogen, Paisley, United Kingdom). p300 gene targeting was done as previously described (3). The single expressed p300 allele in HCT116 was targeted by homologous recombination resulting in cells null for any expressed p300 protein. p300– stock cultures were propagated in the presence of 500 µg/mL G418 (Sigma, Cambridge, United Kingdom). All experiments were done in the absence of antibiotic selection using three independently targeted p300– clones. Serum depletion was carried out by growing cells to 40% confluence, then replacing the medium with serum-free DMEM (Invitrogen). Cells were incubated in serum-free medium for a further 48 hours before harvesting.

Tumor implantation. HCT116 and p300– cells (from a single clone; 1 x 106) were implanted s.c. in severe combined immunodeficient mice. When they had reached a size of ~150 mm3, the tumors were examined using localized 1H NMR and 31P NMR spectroscopy and then excised for metabolite extraction. Procedures were carried out under the authority of project and personal licenses issued by the Home Office, United Kingdom.

Metabolite extraction. Cells grown in serum-enriched medium were harvested at 70% confluence and cells grown in serum-free medium were harvested 48 hours after medium change. Ten 14-cm plates were extracted sequentially using the same 10-mL aliquot of ice-cold 6% HClO4 and the acid extract combined to give a single replicate. Combined extracts were centrifuged to remove cell debris and the supernatant neutralized using 2 mol/L K2CO3. The solution was then left for 30 minutes on ice before centrifuging to remove precipitated KClO4. The supernatant was lyophilized and then dissolved in 0.65 mL of 0.1 mol/L phosphate buffer (pH 7.0) containing 0.5 mmol/L sodium trimethylsilyl-2,2,3,3-tetradeuteroproprionate in 2H2O, and 0.6 mL of the supernatant was transferred to a 5-mm NMR tube. All samples for analysis were obtained in triplicate for each experimental condition (i.e., three separate samples each for HCT116 and the three independently derived p300– clones). For 31P NMR, the samples used for 1H NMR were combined for each cell line, treated with Chelex-100 ion exchange resin, lyophilized, and then dissolved in 4 mL of 50 mmol/L triethanolamine buffer (pH 8.0) containing 5 mmol/L EDTA.

NMR spectroscopy. Spectra were acquired using a Varian INOVA spectrometer equipped with a 9.4-T Oxford Instruments magnet. The resonance frequency for 1H was 400 MHz. For the one-dimensional 1H spectra, 64 transients were acquired across a spectral width of 8 kHz into 32K data points using the NOESY pulse sequence: relaxation delay-90-t1-90-tm-90-acquire; relaxation delay = 1.5 seconds, t1 = 4 ms, tm = 150 ms. The water resonance was suppressed by low power irradiation during the tm and relaxation delay periods. An additional delay of 1 second was added before relaxation delay, which, with an acquisition delay of 1.95 seconds, gave a total recycle time of 4.45 seconds. The sample temperature was 303 K and samples were spun at 16 Hz. The spectra were multiplied by an exponential apodization factor equivalent to 0.6-Hz line broadening before Fourier transformation, phased manually, and referenced with respect to sodium trimethylsilyl-2,2,3,3-tetradeuteroproprionate at {delta} = 0.0 ppm.

Proton-decoupled 31P spectra were acquired using a 90° pulse and a 30-second relaxation delay; typically 4K to 8K transients were acquired per sample. The spectra were processed with a 2-Hz exponential apodization function and chemical shifts were referenced to phosphocreatine ({delta} = 0.0 ppm).

For in vivo 1H spectroscopy, animals were anesthetized by i.p. administration of 0.39 mg/kg fentanyl citrate, 12.5 mg/kg fluanisone, and 5 mg/kg midazolam and immobilized in a quadrature volume coil (Varian, Inc., Palo Alto, CA; length 8 cm, diameter 4 cm). Body temperature was maintained by a flow of warm air through the magnet bore during the experiment. Localized 1H spectra were acquired using a STEAM pulse sequence incorporating asymmetrical excitation pulses, outer-volume saturation, and CHESS-type water saturation pulses before the first 90° pulse (1 ms) and during the mixing time. Spectra were collected into 32K complex data points, with a spectral bandwidth of 8 kHz, and are the sum of 512 transients, collected in 32 blocks of 16 transients, with echo time = 10 ms, mixing time = 50 ms, and repetition time = 4 seconds. The position of the voxel, which had a volume of ~30 µL (3 x 3 x 3 mm), was selected from a multislice pilot spin-echo image (repetition time = 500 ms; echo time = 30 ms; field-of-view, 3.5 x 3.5 cm; slice thickness, 2 mm; 256 data points, 128 increments with 2 transients per increment; 11 contiguous slices) and was chosen to avoid contamination from surrounding tissue. Signal intensities were normalized to the total spectral integral between 0.5 and 4 ppm.

Data analysis. The 1H spectra of cell extracts were divided into bins of equal width and the total signal within each bin integrated, thus converting the spectra into a numerical object suitable for multivariate data analysis. A bin width of 0.01 ppm was used and the spectra integrated between {delta} = 9.5 and 0.5 ppm, excluding the region 5.0 to 4.5 ppm, which contained the suppressed water proton resonance, resulting in each spectrum being reduced to a vector of length 850. The binned spectra were normalized such that the total spectral integral was set to be equal for all the spectra. The data were then transformed by log (ni + 0.3); the value 0.3 was chosen to remove the overall correlation between signal intensity (i.e., metabolite concentration) and variance, as discussed in ref. 20. The mean-centered data were then analyzed by principal component analysis using the covariance matrix.

Microarray analysis. Microarray experiments were done using the CodeLink UniSet Human 20K Bioarray platform (GE Healthcare). Briefly, RNA was obtained from HCT116 and two p300– clones (clones p300–1 and p300–2), grown in serum-containing medium to 70% confluence. Double-stranded cDNA and subsequent biotin-labeled cRNA were synthesized from 2 µg of total RNA using the CodeLink Expression Assay Kit, following the instructions of the manufacturer. Hybridizations with CodeLink Bioarrays were done for 20.5 hours at 37°C using 10 µg of fragmented cRNA. Bioarrays were stained with Cy5-streptavidin (GE Healthcare) and scanned using an Agilent G2565BA scanner (Agilent Technologies, Geneva, Switzerland). Four technical replicate arrays were analyzed for HCT116 and three for each of two p300– clones. Data analysis was done using the R statistical package and environment for statistical computing and graphics (R: A Language and Environment for Statistical Computing). We used the R software package called LIMMA3 for various exploratory graphics and the analysis of differentially expressed genes (21). The final analysis was based on log 2 transformed foreground intensities (with no background correction) after quantile normalization across all arrays for the single-channel data from each array. A linear model was used to combine replicate arrays and appropriate contrasts were constructed to estimate the overall differential expression for each gene between the p300– knockout and wild-type samples (22). The empirical Bayes method was used to obtain a B-statistic for ranking the genes for evidence of differential expression (22) and a volcano plot was used to visualize the results for differentially expressed genes.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
NMR analysis of cells grown in serum-containing medium. NMR analysis was done on HCT116 and three independently targeted p300– clones grown in serum-enriched medium and in exponential phase growth (70% confluent). Extracts of these cells showed multiple resonances, some of which could be assigned to specific metabolites (Fig. 1 ). Principal component analysis was used to compare metabolite profiles from the different cell types. The spectra were binned, as described in Materials and Methods, and the intensity of each bin normalized to the total spectral integral. This eliminates the effects of differences in cell number and possible differences in the efficiency of metabolite extraction. Based on principal component analyses of the bin intensities, the cell lines clustered into two distinct groups, which differed in their p300 status (Fig. 2A ). (Principal components 1 and 2 accounted for 46% of the variance in the data set. Lower principal component were also examined but no patterns related to p300 status were observed, and hence only the results for principal components 1 and 2 are shown.) Therefore, metabolite profiling had sufficient specificity to differentiate p300 status in otherwise isogenic cell lines.


Figure 1
View larger version (8K):
[in this window]
[in a new window]
  		Figure 1. Representative 1H NMR spectrum of a perchloric acid extract of HCT116 cells. Assigned resonances are labeled on the plot. A, aromatic region of the spectrum; B, aliphatic region of the spectrum. The vertical scale of the aromatic region is increased 5-fold compared with the aliphatic region.

 

Figure 2
View larger version (8K):
[in this window]
[in a new window]
  		Figure 2. Principal component analysis scores plots showing separation of HCT116 wild-type and p300– cells. A, cells grown in serum-enriched media. B, cells grown in serum-free media. Open circles, HCT116 wild-type; filled symbols, p300– cells. The different symbols for p300– cells represent three independently targeted deletion clones. C, analysis of combined data from both normal and serum-free conditions. Ellipses show the spread of points; the center of the ellipse marks the mean; and the size of the ellipse represents the SD (the three p300– clones are considered as a single group in this analysis). The "HCT116" and "p300–" ellipses represent the cells grown on serum-enriched medium; the "HCT116sf" and "p300– sf" ellipses represent the cells grown on serum-free medium.

 
Principal component analysis loading plots were used to determine which metabolites seemed to be specifically affected by p300 deletion (Fig. 3A ). p300– cells showed an increase in a nicotinamide-related compound (based on the pattern of the aromatic 1H resonances), as well as a decrease in the concentrations of glutamate and glutamine (compared with wild-type HCT116 cells).


Figure 3
View larger version (13K):
[in this window]
[in a new window]
  		Figure 3. Principal component analysis loading plots showing variables associated with differences between HCT116 wild-type and p300– cells. A, cells grown in serum-enriched media. Loading plot for principal component 1 in Fig. 2A. B, plot showing the effect of growth in serum-free media. Loading plot for principal component 1 in Fig. 2C. C, combined data from growth of cells in serum-rich and serum-free conditions. Loading plot for principal component 2 in Fig. 2C. Glu, glutamate. Gln, glutamine.

 
NMR analysis of cells grown in serum-free medium. To determine whether changes in metabolite profiles caused by p300 deletion were context dependent, NMR analyses were conducted on cell lines grown under serum-free conditions. Previous experiments had shown that HCT116 wild-type and p300– cells have radically different growth characteristics, depending on the presence or absence of FCS in the growth medium. The doubling times of HCT116 were 18.8 ± 0.6 hours with serum and 43.4 ± 2.2 hours in serum-free medium, whereas for p300– cells, the removal of serum has much less of an effect on growth rate, where the doubling times were 28.0 ± 2 and 34.6 ± 1.7 hours, respectively.

In this set of experiments, principal component analysis showed that NMR spectra obtained from HCT116 wild-type cells still clustered separately from p300– cells, along principal component 1 (Fig. 2B). When principal component analysis plots from both sets of experiments (in serum-enriched and serum-free media) were combined, a distinct pattern emerged (Fig. 2C). The cell lines were clearly separated along the principal component 1 axis according to the composition of the growth medium (serum-enriched versus serum-free). The loading plot for principal component 1 in Fig. 2C is shown in Fig. 3B, which shows that phosphocholine (PCho) was significantly higher in cells grown in the presence of serum. In addition, the samples were separated along the principal component 2 axis according to p300 status. This implies that there is a metabolite profile or signature defining p300 deletion, which is independent of the overall metabolic conditions imposed by the different growth medium. In addition, principal component analysis shows very clearly that the relative metabolic effects of p300 deletion are greater under serum-free conditions (Fig. 2C). The loading plot for the principal component 2 axis in Fig. 2C is shown in Fig. 3C, which shows that the N-methyl peak from choline-containing compounds [PCho and glycerophosphocholine (GPC)] is significantly higher in the p300– cells.

Changes identified in the loading plots were verified by visual inspection of the NMR spectra (Fig. 4 ). This revealed that the changes in the choline region of the spectrum observed on the loading plot were due to a singlet resonance, at {delta} = 3.24, which was significantly increased in p300– cells. This singlet, which we have assigned to GPC, was partially overlapping with the singlet resonance from PCho. Their identities were confirmed by adding the respective pure compounds to extracts and acquiring further NMR spectra. In addition, 31P NMR spectra of the pooled samples confirmed that the GPC peak (singlet at 2.85 ppm) was increased in the p300– compared with control samples (Fig. 4C). Multivariate analysis did not identify this increase in GPC, as the bin resolution was insufficient to distinguish the GPC resonance from the larger singlet resonance of PCho. To provide a quantitative measure of the differences between the cell lines, 1H N-methyl resonances from PCho and GPC were integrated manually. Considering all the p300– clones as a group, our data showed that for cells grown in the presence of serum, the concentration of GPC was increased 2.1-fold as a consequence of p300 gene deletion. This increase was significant (P = 0.00017, t test). PCho levels were unaffected (P = 0.47).


Figure 4
View larger version (14K):
[in this window]
[in a new window]
  		Figure 4. 31P and 1H NMR spectra of perchloric acid extracts of HCT116 wild-type and p300– cells. The 1H spectra, showing the N-methyl resonances from choline-containing compounds, were acquired from extracts prepared from HCT116 wild-type (A) and p300– cells (B) grown in the presence of serum. Each plot represents an independent experiment. The GPC resonance (3.245 ppm) is visible as a shoulder on the larger PCho resonance (3.235 ppm). The 31P spectra (C) were acquired from wild-type and p300– cells grown under the specified conditions. sf, cells that had been grown in serum-free medium.

 
Metabolic changes caused by serum deprivation were also immediately apparent from visual inspection of the spectra. The N-methyl resonances from both PCho and GPC were greatly reduced in HCT116 parental cells under serum-free conditions. In contrast, there were no differences in PCho and GPC concentrations for the p300– cells under these different growth conditions. These differences were readily apparent in the 31P spectra of the pooled cell extracts used for the 1H NMR experiments (Fig. 4C). These 31P spectra also suggest that glycerophosphoethanolamine (at 3.45 ppm) is elevated in the p300– cells relative to controls in the presence of serum but is decreased in serum-free conditions.

NMR analysis of xenografted tumors in severe combined immunodeficient mice. To determine if the metabolite signature for p300 deletion was maintained in tumor xenografts, HCT116 and p300– cells were implanted to form s.c. tumors in severe combined immunodeficient mice. GPC was not detectable in localized 31P spectra obtained from the tumors (data not shown) and was only detectable in the 31P spectra of four tumor extracts (three of eight p300– tumor extracts and one of six HCT116 control tumor extracts; data not shown). Principal component analysis of the 1H spectra of the tumor extracts failed to reveal a difference between the p300– and HCT116 tumors (data not shown). However, the localized in vivo 1H spectra revealed a marked difference, with the relative intensity of the peak at 3.2 ppm from choline-containing compounds being significantly higher in the p300– tumors (P = 0.0068, t test) as compared with the HCT116 control tumors (Fig. 5 ). There was also a decrease in the lipid methylene signal at 1.3 ppm in the p300– tumors but no significant differences in the intensities of the lipid signals at 0.9, 2.0, and 2.2 ppm.


Figure 5
View larger version (7K):
[in this window]
[in a new window]
  		Figure 5. Localized 1H spectra acquired from HCT116 and p300– tumors. Spectra were obtained using a STEAM sequence from HCT116 (A) and p300– tumors (B). The spectra shown are the mean ± SE (displayed as a gray area about the mean) of 7 HCT116 wild-type and 10 p300– tumors (5 tumors each of two different p300– clones).

 
Gene expression analysis of cells grown in serum-containing medium. Previous data had suggested that choline metabolism plays an important role in cancer cells. Given that p300– cells in vitro consistently seem to have high levels of GPC and that p300– tumors in vivo show changes in the relative levels of choline-containing compounds, we decided to analyze the expression levels of genes involved in choline metabolism in the isogenic pair of cell lines. We had previously done microarray experiments using the CodeLink UniSet Human 20K Bioarray platform to compare gene expression levels of HCT116 and p300– cells. These experiments were done using mRNA from HCT116 and two p300– clones, grown to 70% confluence in serum-containing medium, similar to the first set of NMR profiling experiments. Using this data set, we examined the expression levels of genes for enzymes involved in choline metabolism that were represented on this platform (see Supplementary Table S1). An empirical Bayes method was used to determine all genes significantly different between the two cell lines, and these are displayed graphically in a volcano plot (Fig. 6A ). This data set shows that the transcript levels of the following enzymes were significantly increased in p300– cells compared with parental HCT116 cells: choline kinase, CTP:phosphocholine cytidylyltransferase 1A, choline phosphotransferase 1, phospholipase A2 group VII, and phospholipase A2–activating protein. Importantly, these enzymes represent a major component of the choline metabolic pathway (Fig. 6B).


Figure 6
View larger version (16K):
[in this window]
[in a new window]
  		Figure 6. Differential expression of genes affecting choline metabolism in p300– and wild-type HCT116 cells. A, volcano plot showing gene expression profiling of all genes on the array. Genes encoding enzymes involved in choline phospholipid metabolism that showed significantly different expression levels between p300– and HCT116 wild-type cells are in highlighted in bold. The X-axis represents M, which is a measure of differential expression between p300– and HCT116 cells; the Y-axis represents B, which is the statistical likelihood of the difference. B, pathways of choline metabolism. Metabolites are shown in boxes and the enzymes responsible for their interconversion are shown beside the connecting arrows. Vertical arrows, significant up-regulation of the transcript for the enzyme in p300– cells (based on the microarray experiments). For those enzymes shown in italics, there was no significant difference in transcript levels between the wild-type HCT116 and p300– cells.

 

   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
Microarray profiling has shown that >300 transcripts are differentially expressed in HCT116 cells as a result of p300 deletion, and these genes include a number of metabolic enzymes.4 However, it is not currently possible to use these data to make predictions of the metabolic consequences of p300 loss. Hence, a comprehensive metabolic assessment—metabolomics—is necessary to fully delineate the cellular phenotype resulting from p300 gene deletion. There is still debate over which analytic techniques are most suitable to define the metabolite complement of a tissue (23, 24). NMR is relatively insensitive compared with mass spectrometry and, consequently, only gives information on a relatively small subset of the most abundant cellular metabolites. Nonetheless, it is evident that the patterns of these metabolites may be sufficient to distinguish between cellular perturbations secondary to genetic modifications, drug interaction, or changes in substrates (16, 17, 23, 25–28). Despite the limitations of NMR-based metabolite profiling, it had sufficient sensitivity to detect the effects of deletion of a single gene, p300, in an otherwise isogenic background and consistently clustered the wild-type HCT116 cells and the p300– clones.

A possible explanation for the sensitivity of NMR-based metabolite profiling, sometimes even to subtle perturbation, may lie in the interconnectedness of cellular networks in general and metabolic networks in particular (29). Interestingly, the most highly connected metabolites include a relatively large number that are detectable by NMR in vivo or in cell or tissue extracts (30). Thus, NMR-based metabolite profiling may be sensitive to specific perturbations because these are propagated throughout the entire metabolic network via the highly connected, and NMR-detectable, nodes. However, a corollary of this is that it is difficult to relate the observed metabolic changes to the specific perturbation that lies at its origin, or, in other words, it is difficult to unambiguously assign a mechanism based on metabolomic data alone. There are some obvious exceptions to this general statement, when the altered metabolites have a specific and clearly understood biological role with respect to the perturbation; for example, glutathione concentrations were shown to increase in Escherichia coli cells in response to oxidative stress (31). Less obvious inferences may also be possible: NMR-based profiling of hypoxia-inducible factor 1ß knockout tumors was used to relate lowered ATP concentration to a decrease in the concentrations of metabolites that serve as precursors for purine biosynthesis (18, 19).

Defining the metabolic profiles of the parental HCT116 cells and all three independently targeted p300– clones showed that there were a number of metabolites that were significantly different between these two groups. These differences included an increase in a nicotinamide-related compound and a decrease in the concentrations of glutamate and glutamine. The decrease in glutamate and glutamine concentrations suggests that mitochondrial oxidation of these substrates may be increased. However, the transcript levels of glutamate dehydrogenase (a mitochondrial enzyme responsible for glutamate and glutamine utilization) were reduced in p300– cells,4 which is inconsistent with this notion. Instead, the changes in glutamate and glutamine levels may reflect some wider perturbation of the metabolic network that has no direct connection with p300 function, as glutamate is one of the most highly connected compounds in cellular metabolism.

A different approach to the problem was afforded by conducting similar experiments in cells grown in serum-free medium (Fig. 2). This adjustment of growth conditions was expected to allow better definition of those metabolic changes that are more specifically related to p300 function. That medium composition has a greater metabolic effect than p300 gene deletion is perhaps not surprising. The lack of growth factors and hormones in the medium is likely to have profound metabolic consequences. Regardless of the variability imposed by medium composition, our data show that the two isogenic cell lines still segregated into two distinct clusters. In this case, however, we found that several "apparent biomarkers" of p300 deletion are themselves context dependent, where some of the metabolic differences observed in serum-containing medium were no longer evident in the serum-free medium. For example, in serum-free medium the resonances from the nicotinamide-related compound and glutamine/glutamate no longer had high loadings in principal component analysis (Fig. 3B).

The increases in GPC concentration, however, were maintained under both serum-enriched and serum-free growth conditions, implying that GPC is a genuine biomarker of p300 deletion in the cultured cells. The N-methyl resonances of GPC and PCho (the only resonances clearly visible in the one-dimensional 1H spectrum) are very close in frequency (chemical shift; see Fig. 4) and therefore were not resolved by the spectral binning used to digitize the spectra for principal component analysis. Therefore, we integrated the GPC and PCho resonances manually. The increases in GPC levels were confirmed by acquiring 31P NMR spectra from the pooled extracts used for the 1H NMR experiments, in which the resonances of GPC and PCho are well resolved (Fig. 4C).

GPC is produced from phosphatidylcholine by the sequential actions of phospholipase A2, to give 1-acylglycerophosphocholine, and then lysophospholipase 1, to give GPC (see Fig. 6). GPC and glycerophosphoethanolamine are generally considered to be markers of phospholipid breakdown (32). Previously it has been shown that GPC levels were reduced in breast cancer cells compared with normal human mammary epithelial cells, and that this was correlated with decreased transcript levels for phospholipase A2 and lysophospholipase (33). An increase in PCho was similarly correlated with an increase in choline kinase levels. Thus, these previous results clearly pointed to an increased flux from choline to phosphatidylcholine but a decreased flux from phosphatidylcholine to GPC; instead, there was an increased flux from phosphatidylcholine back to PCho via phospholipase C. In the current study, we observed an increase in GPC but not PCho in the p300– cells, coupled with a general increase in transcription levels for choline-metabolizing enzymes (choline kinase, CTP:phosphocholine cytidyltransferase, choline phosphotransferase 1, phospholipase A2–activating protein, and phospholipase A2 group VII; see Supplementary Table S1 and Fig. 6). There were no significant changes in phospholipase D or phospholipase C transcript levels. The most parsimonious explanation, consistent with the above observations, is that the overall flux of choline metabolism from choline to GPC through phosphatidylcholine was increased as a result of p300 deletion. p300 acts as a transcriptional cofactor for hypoxia-inducible factor 1 and it has previously been shown that tumors derived from a hepatoma line lacking hypoxia-inducible factor 1 also exhibited an increase in GPC levels (18). Changes in the levels of disparate metabolites are thought to reflect the interconnectedness of metabolic pathways (30); however, the results of this study and a previous study (33), in which changes in GPC concentration have been observed, suggest that in the case of the choline metabolites, this interconnectedness may be mediated at the level of the transcription network.

Differences in choline metabolism were also observed in vivo in the tumors formed following s.c. implantation of HCT116 and p300– cells. Localized 1H spectra showed that the peak at 3.2 ppm, which is largely from choline-containing compounds, was higher in the p300– tumors as compared with the HCT116 controls. There was also a change in lipid composition, with a significant reduction in the methylene peak at 1.3 ppm in the p300– tumors (see Fig. 5). These differences observed in vivo in the tumor levels of choline-containing compounds and in lipid composition could potentially be used for future in vivo NMR diagnostic applications, such as detecting p300 mutations or therapeutic p300 inhibition (34).

Although GPC seemed to be a robust biomarker of p300 deletion in HCT116 cells, it is unlikely to be useful in the detection of p300 mutation or inhibition in vivo. GPC was not detectable by in vivo 31P NMR measurements on p300– tumors, and although a N-methyl resonance was observed in the in vivo 1H NMR spectra at 3.2 ppm, the GPC and PCho resonances were not resolved. Moreover, we were unable to detect a difference in GPC levels in tumor extracts (data not shown), although this may reflect heterogeneity of the tumors (the in vivo data were acquired from an ~30 µL voxel) or inadequate tumor extraction leading to irreproducible metabolite profiles. The latter problem may be addressable in future studies by using magic angle spinning of solid tumor samples. Furthermore, GPC is unlikely to be a unique biomarker of p300 deletion because changes in GPC levels have been observed in response to many different cellular stresses, including low pH, cytotoxic compounds, and radiation damage (32, 35–40).

There is a substantial body of work showing that choline-containing compounds, in particular PCho, are elevated both in tumors and in cell cultures derived from tumors (reviewed in ref. 35). PCho and choline compounds are considered to be proliferation markers, and there is often a switch from a high to a low GPC/PCho ratio during carcinogenesis. In addition, high PCho is also correlated with a high invasiveness phenotype (32). We observed here that PCho levels were correlated with growth when the cell lines were grown in different media. For HCT116 wild-type cells, there was a dramatic reduction in PCho levels together with reduced growth rates as a result of serum deprivation (Fig. 4). A reduction in PCho has also been observed in MCF7 breast cancer cells treated with tumor necrosis factor to induce cell cycle arrest and apoptosis; in this case, the reduction in PCho levels was caused by a simultaneous reduction in choline transport and an increase in CTP:phosphocholine cytidyltransferase activity (41). It is possible that a similar mechanism was operating in this current study. Conversely, PCho levels were unaffected when p300– cells were cultured in serum-free medium, where there was a much more modest decrease in growth rate. A possible explanation for increased PCho levels in these cells could be up-regulated choline kinase activity, the transcript levels for which were observed to increase in p300– cells in serum-containing medium. Choline kinase transcription has previously been observed to correlate with 31P NMR–detectable PCho in breast cancer cells (33) and has been suggested to be a potential therapeutic target (42, 43).

Global profiling "omic" techniques often pick up many variables that respond to a biological perturbation, many of which reflect downstream or correlated, but noncausative, differences. Indeed, it has been shown that the transcripts most responsive to a gene deletion are generally those that show the highest overall variability, and hence would be very poor biomarkers (44); the same is likely to be true of metabolites. If simple experimental modulations, such as changing medium composition, can be used to prioritize metabolites (or mRNAs or proteins) for further study by selecting those that are affected under the greatest range of physiologic conditions, this approach could be of great general significance (45).

In summary, data presented here have shown that NMR-based metabolic profiling has the power to discriminate single gene alterations and identify biomarkers, which potentially can function as surrogates for the genetic event. Moreover, we have shown that these surrogates can be detected noninvasively in vivo. This approach could be extended to identify markers that correlate with common cancer mutations (e.g., p53), with potential diagnostic applications. Because it is often difficult to distinguish the functions of p300 and CREB-binding protein, it would also be interesting to compare the metabolic effects of CREB-binding protein and p300 deletion.


   Acknowledgments
 
Grant support: Cancer Research UK, BBSRC, and National Medical Research Council (Singapore) Medical Research Fellowship (N.G. Iyer). J.D. Brenton is a Cancer Research Senior Clinical Fellow.

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.


   Footnotes
 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

J.G. Bundy and N.G. Iyer contributed equally to this work.

Current address for J.G. Bundy: Biomedical Sciences Division, Imperial College, Sir Alexander Fleming Building, London SW7 2AZ, United Kingdom. Current address for N.G. Iyer: Department of Surgical Oncology, National Cancer Centre, 11 Hospital Drive, Singapore 169610.

3 http://bioinf.wehi.edu.au/limma/. Back

4 N.G. Iyer and C. Caldas, unpublished data. Back

Received 8/25/05. Revised 5/10/06. Accepted 5/26/06.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Giordano A, Avantaggiati ML. p300 and CBP: partners for life and death. J Cell Physiol 1999;181:218–30.[CrossRef][Medline]
  2. Iyer NG, Ozdag H, Caldas C. p300/CBP and cancer. Oncogene 2004;23:4225–31.[CrossRef][Medline]
  3. Iyer NG, Chin SF, Ozdag H, et al. p300 regulates p53-dependent apoptosis after DNA damage in colorectal cancer cells by modulation of PUMA/p21 levels. Proc Natl Acad Sci U S A 2004;101:7386–91.[Abstract/Free Full Text]
  4. Ward R, Johnson M, Shridhar V, van Deursen J, Couch FJJ. CBP truncating mutations in ovarian cancer. Med Genet 2005;42:514–8.[Free Full Text]
  5. Kishimoto M, Kohno T, Okudela K, et al. Mutations and deletions of the CBP gene in human lung cancer. Clin Cancer Res 2005;1:512–9.
  6. ter Kuile BH, Westerhoff HV. Transcriptome meets metabolome: hierarchical and metabolic regulation of the glycolytic pathway. FEBS Lett 2001;500:169–71.[CrossRef][Medline]
  7. Gygi SP, Rochon Y, Franza BR, Aebersold R. Correlation between protein and mRNA abundance in yeast. Mol Cell Biol 1999;19:1720–30.[Abstract/Free Full Text]
  8. Anderson L, Seilhamer J. A comparison of selected mRNA and protein abundances in human liver. Electrophoresis 1997;18:533–7.[CrossRef][Medline]
  9. Urbanczyk-Wochniak E, Luedemann A, Kopka J, et al. Parallel analysis of transcript and metabolic profiles: a new approach in systems biology. EMBO Rep 2003;4:989–93.[CrossRef][Medline]
  10. Hirai MY, Yano M, Goodenowe DB, et al. Integration of transcriptomics and metabolomics for understanding of global responses to nutritional stresses in Arabidopsis thaliana. Proc Natl Acad Sci U S A 2004;101:10205–10.[Abstract/Free Full Text]
  11. Tohge T, Nishiyama Y, Hirai MY, et al. Functional genomics by integrated analysis of metabolome and transcriptome of Arabidopsis plants over-expressing an MYB transcription factor. Plant J 2005;42:218–35.[CrossRef][Medline]
  12. Goodacre R, Vaidyanathan S, Dunn WB, Harrigan GG, Kell DB. Metabolomics by numbers: acquiring and understanding global metabolite data. Trends Biotechnol 2004;22:245–52.[CrossRef][Medline]
  13. Beckonert O, Monnerjahn K, Bonk U, Leibfritz D. Visualizing metabolic changes in breast-cancer tissue using H-1-NMR spectroscopy and self-organizing maps. NMR Biomed 2003;16:1–11.[CrossRef][Medline]
  14. Bradamante S, Barchiesi E, Pilotti S, Borasi G. High-resolution 1H NMR spectroscopy in the diagnosis of breast cancer. Magn Reson Med 1998;8:440–9.[CrossRef]
  15. Sitter B, Sonnewald U, Spraul M, Fjosne HE, Gribbestad IS. High-resolution magic angle spinning MRS of breast cancer tissue. NMR Biomed 2002;15:327–7.[CrossRef][Medline]
  16. Griffin JL, Shockcor JP. Metabolic profiles of cancer cells. Nat Rev Cancer 2004;4:551–61.[CrossRef][Medline]
  17. Raamsdonk LM, Teusink B, Broadhurst D, et al. A functional genomics strategy that uses metabolome data to reveal the phenotype of silent mutations. Nat Biotechnol 2001;19:45–50.[CrossRef][Medline]
  18. Griffiths JR, McSheehy PMJ, Robinson SP, et al. Metabolic changes detected by in vivo magnetic resonance studies of HEPA-1 wild-type tumours and tumours deficient in hypoxia-inducible factor-1ß (HIF-1ß): evidence of an anabolic role for the HIF-1 pathway. Cancer Res 2002;62:688–95.[Abstract/Free Full Text]
  19. Griffiths JR, Stubbs M. Opportunities for studying cancer by metabolomics: preliminary observations on tumors deficient in hypoxia-inducible factor 1. Adv Enzyme Regul 2003;43:67–76.[CrossRef][Medline]
  20. Purohit PV, Rocke DM, Viant MR, Woodruff DL. Discrimination models using variance-stabilizing transformation of metabolomic NMR data. OMICS 2004;8:118–30.[CrossRef][Medline]
  21. Smyth GK. Limma: linear models for microarray data. In: Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W, editors. Bioinformatics and computational biology using R and bioconductor. New York: Springer; 2005.
  22. Smyth GK. Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 2004;3:Article3. Epub 2004 Feb 12.
  23. Griffin JL. Metabonomics: NMR spectroscopy and pattern recognition analysis of body fluids and tissues for characterization of xenobiotic toxicity and disease diagnosis. Curr Opin Chem Biol 2003;7:648–54.[CrossRef][Medline]
  24. Robertson DG. Metabonomics in toxicology: a review. Toxicol Sci 2005;85:809–22.[Abstract/Free Full Text]
  25. Griffin JL. Metabolic profiles to define the genome: can we hear the phenotypes? Philos Trans R Soc Lond B Biol Sci 2004;359:857–71.[Abstract/Free Full Text]
  26. Krishnan P, Kruger NJ, Ratcliffe RG. Metabolite fingerprinting and profiling in plants using NMR. J Exp Bot 2005;56:255–65.[Abstract/Free Full Text]
  27. Lindon JC, Holmes E, Bollard ME, Stanley EG, Nicholson JK. Metabonomics technologies and their applications in physiological monitoring, drug safety assessment and disease diagnosis. Biomarkers 2004;9:1–31.[CrossRef][Medline]
  28. Nicholson JK, Connelly J, Lindon JC, Holmes E. Metabonomics: a platform for studying drug toxicity and gene function. Nat Rev Drug Discov 2002;1:153–61.[CrossRef][Medline]
  29. Brindle KM. Metabolomics: Pandora's box or Aladdin's cave? The Biochemist 2003;25:15–7.
  30. Brindle KM. Magnetic resonance spectroscopy and imaging—powerful tools for functional genomics. In: Kholodenko BN, Westerhoff HV, editors. Metabolic engineering in the post genomic era. Norwich: Horizon Biosciences; 2004.
  31. Tweeddale H, Notley-McRobb L, Ferenci T. Assessing the effect of reactive oxygen species on Escherichia coli using a metabolome approach. Redox Rep 1999;4:237–41.[CrossRef][Medline]
  32. Podo F. Tumour phospholipid metabolism. NMR Biomed 1999;12:413–39.[CrossRef][Medline]
  33. Glunde K, Jie C, Bhujwalla ZM. Molecular causes of the aberrant phospholipid metabolism in breast cancer. Cancer Res 2004;64:4270–6.[Abstract/Free Full Text]
  34. Gilles RJ, Morse DL. In vivo magnetic resonance spectroscopy in cancer. Annu Rev Biomed Eng 2005;7:287–326.[CrossRef][Medline]
  35. Ackerstaff E, Glunde K, Bhujwalla ZM. Choline phospholipid metabolism: a target in cancer cells? J Cell Biochem 2003;90:525–53.[CrossRef][Medline]
  36. Milkevitch M, Shim H, Pilatus U, et al. Increases in NMR-visible lipid and glycerophosphocholine during phenylbutyrate-induced apoptosis in human prostate cancer cells. Biochim Biophys Acta 2005;1734:1–12.[Medline]
  37. Farber SA, Slack BE, Blusztajn JK. Acceleration of phosphatidylcholine synthesis and breakdown by inhibitors of mitochondrial function in neuronal cells: a model of the membrane defect of Alzheimer's disease. FASEB J 2000;14:2198–206.[Abstract/Free Full Text]
  38. Roman SK, Jeitner TM, Hancock R, Cooper WA, Rideout DC, Delikatny EJ. Induction of magnetic resonance-visible lipid in a transformed human breast cell line by tetraphenylphosphonium chloride. Int J Cancer 1997;73:570–9.[CrossRef][Medline]
  39. Street JC, Mahmood U, Matei C, Koutcher JA. In vivo and in vitro studies of cyclophosphamide chemotherapy in a mouse mammary carcinoma by 31P NMR spectroscopy. NMR Biomed 1995;8:149–58.[Medline]
  40. Mahmood U, Alfieri AA, Ballon D, Traganos F, Koutcher JA. In vitro and in vivo 31P nuclear-magnetic-resonance measurements of metabolic changes post radiation. Cancer Res 1995;55:1248–54.[Abstract/Free Full Text]
  41. Bogin L, Papa MZ, Polak-Charcon S, Degani H. TNF-induced modulations of phospholipid metabolism in human breast cancer cells. Biochim Biophys Acta 1998;1392:217–32.33.[Medline]
  42. de Molina AR, Banez-Coronel M, Gutierrez R, et al. Choline kinase activation is a critical requirement for the proliferation of primary human mammary epithelial cells and breast tumor progression. Cancer Res 2004;64:6732–9.[Abstract/Free Full Text]
  43. Rodriguez-Gonzalez A, de Molina AR, Fernandez F, et al. Inhibition of choline kinase as a specific cytotoxic strategy in oncogene-transformed cells. Oncogene 2003;22:8803–12.[CrossRef][Medline]
  44. Hughes TR, Marton MJ, Jones AR, et al. Functional discovery via a compendium of expression profiles. Cell 2000;102:109–26.[CrossRef][Medline]
  45. Brown AJP, Planta RJ, Restuhadi F, et al. Transcript analysis of 1003 novel yeast genes using high-throughput northern hybridizations. EMBO J 2001;20:3177–86.[CrossRef][Medline]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bundy, J. G.
Right arrow Articles by Brindle, K. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bundy, J. G.
Right arrow Articles by Brindle, K. M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online