This study provides the first evidence that cell lines derived from later-stages of the premalignant condition Barrett's esophagus have increased glycolysis compared to early-stage BE, while preserving mitochondrial function. Furthermore, we found that later-stage BE cell lines modulate their oxidative phosphorylation with an elevated Crabtree effect. Both of these phenotypic alterations are hallmarks of cancer metabolism and suggest that premalignant BE cell lines evolved their energy metabolism in response to variable oxygen and glucose levels.
We found that ECAR was highest in the three BE cell lines derived from patients with the highest genomic instability (CP-B, CP-C and CP-D) and lowest in the cell line derived from a patient with relatively more stable genome (CP-A), suggesting that a glycolytic phenotype was selected during the process of neoplastic progression. The BE cell line with the highest genomic instability, CP-D, was more glycolytic than the other BE cell lines, which was consistent with it having the greatest ECAR decrease following 2-DG addition and lowest ECAR increase following 2,4-DNP addition. However, mitochondrial function was conserved during CP-D's ‘glycolytic switch’ since it was able to compensate for glycolytic inhibition by upregulating oxidative phosphorylation. Although CP-C was more glycolytic than CP-A, it displayed a high level of OCR, which did not increase in response to glycolytic inhibition. This greater reliance on oxidative phosphorylation made CP-C more insensitive to glycolytic inhibition than CP-D. Similar to CP-D however, ECAR in CP-C was more insensitive to uncoupling by 2,4-DNP than CP-A, which is consistent with the higher baseline ECAR in CP-C. Insensitivity of glycolysis rates in CP-D to mitochondrial inhibition was also confirmed using the ATPase inhibitor oligomycin (Table S4
). All cell lines maintained increased OCR levels in the absence of exogenous glutamine and glucose, indicating that mitochondrial upregulation is not specifically dependent on exogenous glutamine substrate (data not shown).
OCR in the BE cell lines did not correlate with ECAR, indicating that the changes in the two metabolic pathways evolved independently. The lack of OCR increase in CP-A, CP-B and CP-C cell lines in response to glycolytic inhibition suggests that ATP/ADP levels did not fall below a threshold that would stimulate mitochondrial activity, suggesting that oxidative phosphorylation supplied the majority of energy in these cell lines. Likewise, in response to 2,4-DNP-mediated uncoupling, OCR increases were not significantly different in CP-A and CP-C, and marginally higher in CP-D, suggesting that their mitochondria maintained a similar level of mitochondrial reserve capacity. This was also confirmed in an independent experiment in which oligomycin-treated BE cells had similar percentage OCR decreases, which corresponds to a similar contribution of ATP-generating coupled-respiration to total cellular OCR (Table S4
). CP-B, the BE cell line which displayed the highest OCR, also displayed the highest OCR increase in response to 2,4-DNP, indicating that this cell line had the highest mitochondrial reserve capacity and/or utilized different available mitochondrial substrates. Increased mitochondrial reserve capacity has been shown to be dependent on substrate concentration in cancer cell lines but also differs based on mitochondrial enzyme activity in normal tissues 
. Despite evidence that would suggest mitochondrial differences between the BE cell lines, mitochondrial mass assessed by flow cytometric assays were not found to be different (Figures S2).
Finally, CP-C and CP-D cell lines demonstrated a greater change in OCR via Crabtree effect than the less glycolytic CP-A and CP-B cell lines, suggesting adaptation towards a more cancer-like phenotype. Increased Crabtree effect in BE cells predicts a greater survival advantage to adapt to conditions of glucose and oxygen fluctuation such as gastric reflux-induced ulceration or ischemia.
Difficulties associated with OE-33 cell adhesion prevented normalization of data collected for this EA cell line for direct comparison to the BE cell lines. However, based on normalization estimated from a limited data set, OE-33 had the highest levels of ECAR out of the cell lines tested and had percent ECAR that decreased as much as CP-D following 2-DG addition, indicating that it is a significantly glycolytic cell line (Text S1
). OE-33 also had the highest levels of OCR and was sensitive to oligomycin, indicating a glycolytic cell line with functional mitochondria.
Given our findings, we propose a model in which BE begins with a metabolism which is largely dependent on oxidative phosphorylation, as seen in CP-A, but then progresses through a metabolic phenotype which is intermediate between normal and cancer (). This intermediate stage displays increased glycolysis while mitochondria remain functional. As the cells progress, later stages of BE display a more pronounced Crabtree effect, which enables mitochondrial downregulation in response to substrate. Mitochondrial activity and uncoupling subsequently increase as the tissue progresses to EA.
Barrett's esophagus progression to esophageal adenocarcinoma involves an intermediate metabolic stage with increased glycolysis and functional mitochondria.
Several mechanisms are likely to contribute to the metabolic changes in the BE cell lines. Both copies of TP53 are mutated and/or altered in CP-B, CP-C, and CP-D (Table S1
), which may be of significance since loss of TP53 has been reported to upregulate glycolysis and downregulate mitochondrial activity to produce the Warburg effect in several cell lines 
. Notably the TP53-wild type BE cell line, CP-A, which is more representative of the genotype of BE patients that do not progress to cancer, demonstrated lower levels of glycolysis. A number of hypoxia-response genes have also been reported in BE tissue: glucose transporter Glut-1 
and pyruvate kinase isoform M2 
, both known to be associated with elevated glycolysis, vascular endothelial growth factor (VEGF) and erythropoietin 
, which are upregulated by hypoxia-induced transcription factors HIF-1α and HIF-2α respectively. The genomic analysis performed in our study revealed that CP-C and CP-D have a single copy loss involving HIF-1α, however, the functional status of the remaining allele is unknown (Table S6
). Overexpression of HIF-1α has been reported in other cancers 
and may result from degradation-insensitive forms of this protein or increases in mTOR activity 
. CP-D was also found to contain an amplified PHDGH gene (Human Cancer Genome Atlas), which is related to serine biogenesis and suggests an alternate use of the upregulated glycolytic pathway in this cell line 
. The identification of a glycolytic phenotype in BE should bring to attention the increased importance of glycolytic and hypoxia-resistance pathways, which are routinely investigated in cancers but may also become altered earlier in progression.
Since endogenous reactive oxygen species (ROS) are produced by mitochondria, the mitochondrial genome is particularly susceptible to ROS-mediated damage. Although CP-A had the fewest nuclear genome alterations (Figure S1
), it displayed the most mitochondrial genome mutations of all the cell lines (Table S7
). Most of the CP-A mutations were heteroplasmic which indicates that multiple different clones were present. However, this diversity may also indicate that the predominantly oxidative phosphorylation metabolism in this cell line generates higher ROS levels, consistent with findings by Chen et al 
. Remarkably, we found that the cell line displaying the most nuclear genome alterations, CP-D, had the fewest number of mitochondrial genome mutations (Table S7
), consistent with CP-D maintaining active mitochondria and suggesting that CP-D was effective at suppressing ROS in the mitochondria and/or underwent a mitochondrial genetic sweep which shifted energy metabolism away from oxidative phosphorylation.
Although the Warburg and Crabtree effects have been commonly associated with cancer progression, higher glycolysis and Crabtree effect have also been observed in actively proliferating non-cancerous cells, such as stem cells and lymphocytes 
. Since BE undergoes frequent damage and repair, higher rates of growth in this tissue could be promoted by elevated glycolysis which, through the pentose-phosphate pathway, provides higher levels of anabolic substrates and NADPH, an important cofactor responsible for glutathione regeneration which suppresses ROS. Chen et al reported that CP-C had lower toxicity than CP-A to exogenous ROS species
, although glutathione levels were not different and the differences in toxicity may be due to CP-A's more sensitive wild type p53, which has intact apoptotic signaling. Although we did not study ROS levels in the BE cell lines, altering metabolism may be a mechanism for preventing higher levels of mutation during BE progression. Interestingly, of the four patients from whom the cell lines were derived, the patient from whom the most mitochondrially-active BE cell line (CP-B) was developed was the only patient in whom EA was detected.
Given the preliminary nature of our findings, it is important to determine if the metabolic changes found in BE cell lines are also found in biopsies from patients. Although tissue culture is clearly a model system, all of the metabolic measurements of ECAR and OCR were done at confluency, a crude approximation of the structure of cells in tissue. Also, the BE cell lines in this study were immortalized using hTERT, avoiding potential metabolic effects of other immortalization methods that affect p53, Rb, and Myc activity such as SV40 large T antigen. However, since the BE and EA cell lines used in this study were derived from different patients, further analysis of longitudinally collected samples within the same patient would more directly address these questions. The most promising support for our findings in vivo
are from BE tissue quantitative fluorescent imaging which displays elevated NAD(P)H levels representative of glycolytic tissue 
, suggesting that our observation in BE cell lines may translate to BE segments in patients.