Preservation of high glycolytic phenotype by establishing new acute lymphoblastic leukemia cell lines at physiologic oxygen concentration

Abstract

Cancer cells typically exhibit increased glycolysis and decreased mitochondrial oxidative phosphorylation, and they continue to exhibit some elevation in glycolysis even under aerobic conditions. However, it is unclear whether cancer cell lines employ a high level of glycolysis comparable to that of the original cancers from which they were derived, even if their culture conditions are changed to physiologically relevant oxygen concentrations. From three childhood acute lymphoblastic leukemia (ALL) patients we established three new pairs of cell lines in both atmospheric (20%) and physiologic (bone marrow level, 5%) oxygen concentrations. Cell lines established in 20% oxygen exhibited lower proliferation, survival, expression of glycolysis genes, glucose consumption, and lactate production. Interestingly, the effects of oxygen concentration used during cell line initiation were only partially reversible when established cell cultures were switched from one oxygen concentration to another for eight weeks. These observations indicate that ALL cell lines established at atmospheric oxygen concentration can exhibit relatively low levels of glycolysis and these levels are semi-permanent, suggesting that physiologic oxygen concentrations may be needed from the time of cell line initiation to preserve the high level of glycolysis commonly exhibited by leukemias in vivo.

Introduction

Cancer cell lines grown in vitro are routinely used as models intended to represent cancer cells in vivo. The great majority of cell lines are established in supra-physiologic levels of oxygen (O₂) by combining 5% CO₂ with room air to achieve an O₂ concentration of approximately 20%. In contrast, the mean O₂ concentration in normal tissues of the human body is 7% [1], and the O₂ concentration in poorly oxygenated regions of cancerous tissues can be <2% [2]. Supra-physiologic levels of O₂ might have deleterious effects, as they have been reported to increase the generation of oxygen radicals that may, in turn, increase the rate of cellular oxidative damage, DNA strand breaks, and mutations [3].

Many available leukemia cell lines demonstrate DNA microsatellite instability, indicating loss of DNA mismatch repair activity [4,5]. Such cell lines, and especially those that have been extensively expanded over decades, are likely to have accumulated significant genetic and epigenetic changes under non-physiologic selection pressures since their isolation from the patient. In patients, leukemia cells chiefly reside in the bone marrow, a tissue compartment normally containing an oxygen concentration of ~5% O₂ [6]. In recent decades, few studies have reported the culturing of leukemia cell lines at or near physiologic O₂ concentration [7,8]. In a report comparing healthy primary T lymphocytes cultured ex vivo in 5% versus 20% O₂, cells in 5% O₂ exhibited less intracellular oxidation and a higher ratio of intracellular glutathione to oxidized intracellular glutathione, indicative of reduced oxidative stress [9]. However, detailed studies comparing the cellular and molecular consequences of initiating cancer cell lines at atmospheric versus physiologic levels of O₂ are lacking. In particular, during the process of establishing continuously growing cell lines from biological or clinical specimens, it has not been defined whether the employment of different O₂ concentrations results in transient metabolic phenotypes that can be fully reversed by changing the O₂ concentration, or instead selects for metabolically distinct, dominant subpopulations whose metabolic phenotype cannot be fully reversed.

At physiologic oxygen tensions in vivo, nearly all types of normal, healthy cells metabolize the majority of their glucose through the mitochondrial oxidative phosphorylation pathway. In conditions of low oxygen tension (hypoxia), normal cells decrease mitochondrial oxidative phosphorylation while increasing their uptake of glucose, metabolizing most of the glucose through the glycolysis pathway into biosynthetic intermediates such as lipids, amino acids, and nucleotides, as well as adenosine triphosphate and lactate [10]. In cancer cells, however, a heavy reliance on increased glucose uptake and glycolysis exists not only in hypoxia but also at physiological and mildly aerobic oxygen tensions. This increased utilization of glycolysis relative to oxidative phosphorylation in cancer cells, commonly referred to as the Warburg effect [11,12], is considered a biochemical hallmark of cancer.

The glycolytic phenotype is observed in the majority of aggressive human cancers, thereby providing a valuable clinical imaging tool employing positron-emission tomography to selectively identify cancer cells via their increased uptake of the glucose analog ¹⁸fluorodeoxyglucose [13]. The mechanistic reason(s) for the predilection of most cancers to employ aerobic glycolysis remain(s) controversial. It has been proposed that higher levels of aerobic glycolysis reflect a need of malignant cells to circumvent mitochondrial production of reactive oxygen species [14] or a need to metabolize greater amounts of biosynthetic intermediates [10] rather than a need to increase total adenosine triphosphate production.

As in most cancer types, glycolysis is heavily utilized in childhood acute lymphoblastic leukemias (ALL) [15]. Here we compared the in vitro growth, survival, drug resistance, and molecular characteristics of cell lines established directly from pediatric ALL patients in physiologic versus atmospheric oxygen levels.

Methods and materials

Patients and cell lines

Bone marrow and peripheral blood samples were collected from 12 pediatric ALL patients over a 2 year period. Informed consent to utilize excess clinical material for research purposes was obtained from patients' parental guardians. All research involving human subjects was reviewed and approved by the Children's Hospital Los Angeles and the Texas Tech University Health Sciences Center committees for protection of human subjects under all applicable guidelines.

Cell lines were established from patient material and maintained in Iscove's modified Dulbecco's medium (IMDM) (BioWhittaker Inc., Walkersville, MD) supplemented with 20% heat-inactivated (56 °C for 30 min) fetal bovine serum (FBS), 3 mM L-glutamine, 5 μg/ml insulin, 5 μg/ml transferrin, and 5 ng/ml selenous acid (BD Biosciences, San Jose, CA). From the time of their initial isolation, cell cultures were maintained at 37 °C in humidified incubators containing either 20% O₂ (95% room air, 5% CO₂) or 5% O₂ (achieved by automated addition of nitrogen gas) using Thermo Forma 3130 incubators (Thermo Fisher Scientific, Waltham, MA) with standard 180 l nitrogen tanks (AirGas, Radnor Township, Pennsylvania). The accuracy of oxygen concentrations reported by the incubator sensors was routinely confirmed using a Fyrite O₂ gas analyzer (Bacharach, Inc., New Kensington, PA).

Reagents

All antibodies used in this study are listed in Supplementary Table 1. For glucose and lactate assays, glucose-free/pyruvate-free IMDM was custom ordered from Gibco/Life Technologies (Carlsbad, CA). d-(+)-glucose was from Sigma-Aldrich (St. Louis, MO). CFSE was from Molecular Probes (Eugene, OR). Vincristine, etoposide, and dexamethasone were from Sigma-Aldrich (St. Louis, MO); l-asparaginase was from Merck (Whitehouse Station, NJ); doxorubicin was from the NCI, NIH (Bethesda, MD); 4-hydroxycyclophosphamide (4-HC), the active metabolite of cyclophosphamide, was from the Duke Comprehensive Cancer Center, Department of Medicine (Durham, NC).

Cell cycle analysis by bromodeoxyuridine (BrdU) incorporation

Phases of the cell cycle were measured using the allophycocyanine (APC) BrdU Flow Kit (BD Biosciences) according to the manufacturer's protocol, with the exception that incubation of cells in 7-AAD was performed overnight to achieve smaller CVs. Plated cells were allowed 24 h to adhere prior to a 30 min incubation with BrdU, permeabilization, and addition of 7-AAD. Data was acquired on an LSRII flow cytometer and analyzed using FACSDiva software (BD Biosciences). During analysis, cell aggregates were excluded by gating tightly on the pulse area versus pulse height dot plot of the 7-AAD parameter.

Cell division assay by CFSE staining

To examine the number of cell divisions [16], cells were loaded with 10 μM carboxyfluorescein diacetate, succinimidyl ester (CFSE) (Molecular Probes, Eugene, OR). COG-LL-317 and COG-LL-319 cells were incubated for 10 min whereas COG-LL-332 cells required 1 min. Next, staining was quenched by incubating with five volumes of ice-cold culture medium for 5 min, followed by three washes. Cells were then cultured for six days, washed, and analyzed on an LSRII cytometer using a 525 nm band pass filter.

Immunoblotting

A total of 15 × 10⁶ cells were harvested and lysed in RIPA buffer (Upstate, Charlottesville, VA) for protein extraction. 20 μg of protein per lane was fractionated on 10–20% Tris–Glycine gels (Invitrogen, San Diego, CA), transferred to nitrocellulose membranes, and probed with primary antibodies, followed by incubation with the appropriate HRP-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Membranes were incubated with West Pico Chemiluminescent Substrate (Pierce, Rockford, IL) and visualized on an autoradiography film (Denville Scientific, Metuchen, NJ).

Polymerase chain reaction

To examine the possibility of Epstein–Barr Virus DNA in the cell lines, PCR was performed by extracting genomic DNA from cell pellets using PureLink™ Genomic DNA Purification kit (Invitrogen, Carlsbad, CA). PCR involved primers that amplify a 210 bp fragment of the Epstein–Barr Encoded RNA gene from the Epstein–Barr Virus genome [17]. Forward and reverse primers (5′-CCCGCCTACACACCAACTAT-3′ and 5′-AGTCTGGGAAGACAACCACA-3′) were obtained from Integrated DNA technologies (Coralville, IA). PCR was carried out using a GeneAmp™ 9700 PCR System (Applied Biosystems, Foster City, CA) with 2.5 units GoTaq™ DNA polymerase (Promega, Madison, WI). Reaction tubes of 50 μl contained 1X PCR Buffer, 1.5 mmol/L MgCl₂, 400 μmol/L dNTPs, 20 pmol forward primer, 20 pmol reverse primer, and 100 ng template DNA. PCR was begun with a 95 °C incubation period for 2 min, followed by 35 cycles at 94 °C for 30 s, 60 °C for 30 s, and 72 °C for 1 min, with a final time extension of 9 min at 60 °C. As positive controls, the lymphoblastoid cell lines SMS-KCL and SMS-SAL were used, previously established by infecting blood samples with live Epstein–Barr Virus [18]. Jurkat cells, known to not express Epstein–Barr Virus [19], served as a negative control.

Short tandem repeat (STR) genotyping

To verify the identity of each cell line, short-tandem-repeat (microsatellite DNA) genotyping was employed using the AmpF STR Identifiler™ kit according to the manufacturer's protocol (Applied Biosystems, Foster City, CA). The gender-specific locus ameolgenin was amplified along with the following microsatellite loci: D8S1179, D21S11, D7S820, CSF1P0, D3S1358, TH01, D13S317, D16S539, D2S1338, D19S433, vWA, TPOX, D18S51, D5S818, FGA. STR profiles were compared to those of other cell lines using a searchable database at www.COGcell.org.

Xenografts from established cell lines

Immunodeficient 6–10 week-old male NOD/SCID mice were γ-irradiated with 250 cGy to eliminate residual immune cells, and then injected in the lateral tail vein with 5 × 10⁶ leukemia cells in 100 μl of serum-free phosphate buffered saline (PBS). Mice were humanely sacrificed when they became moribund, and spleens were removed for analyses. The frequency of xenotransplanted cells was determined by double staining for human and mouse CD45 and quantified by flow cytometry. Animal experiments were conducted under a protocol approved by the Animal Care and Use Committee, Children's Hospital Los Angeles.

Mycoplasma detection

All cell lines were grown without antibiotics to allow ready detection of mycoplasma, for which they tested negative using the Cambrex MycoAlert detection assay (Fisher Scientific, Pittsburgh, PA) and Hoechst DNA staining (American Tissue Culture Collection, Manassas, VA).

Glucose consumption

Glucose levels were measured using a colorimetric assay (Cayman Chemical Company, Ann Arbor, MI) based on the glucose oxidase-peroxide reaction, according to the manufacturer's protocol. Cells were pelleted and re-suspended at a concentration of 10⁶ cells/ml in glucose-free/pyruvate-free IMDM without addition of pyruvate but supplemented with 1.0 mg/ml glucose rather than the 4.5 mg/ml present in routine IMDM medium, to prevent the assay from being overwhelmed by a high level of glucose. The medium was also supplemented with 10% FBS. Cells were incubated for 24 h, pelleted, and then 5 μl of supernatant was used. Absorbance was monitored at 570 nm, using a VersaMax tunable microplate reader (Molecular Devices, Sunnyvale, CA). Glucose consumption (%) was calculated as follows: 1 – (glucose level in experimental)/(glucose level in medium alone) × 100.

Lactate production

Lactate levels were measured using a colorimetric assay (BioVision Research Products, Mountain View, CA) based on the lactate oxidase reaction, according to the manufacturer's protocol. To minimize background signals, FBS levels were reduced to 5%. Cells were pelleted and re-suspended at a concentration of 10⁶ cells/ml in glucose-free/pyruvate-free IMDM supplemented with 5% FBS, 4.5 mg/ml glucose, but no pyruvate. Cells were incubated for 24 h, pelleted, and then 1 μl of supernatant was used. Absorbance was monitored at 510 nm. To correct for the influence of lactate in FBS, values from controls receiving IMDM containing 5% FBS were subtracted from experimental values.

Cytogenetic analysis

Karyotyping involved standard clinical GTG banding techniques [20] according to the International System for Human Cytogenetic Nomenclature [21]. Twenty metaphases were completely analyzed for each cell line.

Immunophenotyping

Fc receptors were blocked using 20 μl of 100% human serum (Omega Scientific, Tarzana, CA) for 10 min prior to incubation with antibodies for 45 min. Data was acquired on a Becton Dickinson LSRII flow cytometer. Dead cells and debris were excluded according to their increased staining with DAPI (0.5 ng/ml final concentration) and low forward scatter properties. At least 30,000 cells were analyzed for each sample. Percent positive was calculated as [% positive for experimental–% positive for isotype-matched control]. As some surface markers on hematopoietic cells can be modulated transiently by cryopreservation [22-26], thawed cell lines were incubated for ≥ 6 weeks before staining. Patient bone marrow samples were immunophenotyped and analyzed in a clinical pathology laboratory, identifying fixed leukemia cells after gating on CD45 expression and side scatter characteristics.

Gene expression profiling

RNA was hybridized to the U133 Plus 2.0 GeneChip oligonucleotide microarray from Affymetrix (Santa Clara, CA), according to the manufacturer's protocol. Probe sets' fluorescence intensity values were normalized using the invariant probe set method and modeling the perfect match minus mismatch algorithm as implemented in the dChip program [27]. Gene expression values obtained from these sets were log(base 10) transformed before analysis. Log-transformed gene expression values were standardized to a mean value of 0 and a standard deviation of 1. A cut-off value of 1.5 was chosen as an approximate minimum change that might be expected to have a measurable biologic effect. Data were analyzed with dChip and Genetrix (Epicenter Software, Pasadena, CA) software [28]. Raw data files are available at: http://www.cogcell.org/suppdata/.

Real-time RT-PCR

Matching primers and probes for selected genes were designed using Primer Express (Perkin-Elmer Applied Biosciences, Foster City, CA). The probe was labeled with a reporter fluorescent dye, FAM (6-carboxy-fluoescein), on the 5′ nucleotide and a quenching dye, Iowa Black (Integrated DNA Technologies, CoralVille, 1A), on the 3′ nucleotide. The amplicon spanned at least two exons to avoid amplification of genomic DNA. Primer sequences are available at http://www.cogcell.org/suppdata. Total RNA was extracted using TRIzol reagent (Invitrogen) and further purified with an RNeasy Midi Kit (Qiagen). All reactions were performed using the ABI 7700 Sequence Detection System (Perkin-Elmer Applied Biosciences).

Cell enumeration

Enumeration of viable cells employed a Vi-CELL XR automated cell counter (Beckman Coulter, Inc., Miami, FL) which utilizes trypan blue dye exclusion. Each cell line was seeded into 24 wells on 24-well plates at 0.3 × 10⁵, 3 × 10⁵, and 1 × 10⁵ cells/ml for 317, 319, and 332 cells, respectively. Every day for eight days, triplicate wells were harvested and evaluated individually for cell number and % cell viability. Doubling time was determined by converting cell viability data to a log(2) scale and plotting the time required for doubling.

Plating efficiency and drug cytotoxicity

Plating efficiencies and drug cytotoxicity were measured using a DIMSCAN fluorescence-based digital imaging microscopy system, as described previously [29]. For plating efficiencies, cells were plated at 4-fold serial dilutions in 200 μl/well of culture medium in 96-well plates and incubated for six days, then loaded with fluorescein diacetate (FDA) (5 μg/ml final concentration) for 20 min. FDA is not retained in dead cells, enabling discrimination of viable from dead cells. Background fluorescence was quenched by addition of eosin Y (1 μg/ml final concentration) plus digital thresholding [29]. For drug cytotoxicity assays, 15,000 cells in 100 μl/well were seeded into wells of 96-well plates 24 h before adding 100 μl of drugs to each well, followed by an incubation for 72 h and loading with FDA, as above. A minimum of six replicate wells were run for each treatment group.

Statistics

The statistical significance of differences in cell cycle phase frequencies was assessed using the paired samples, two-tailed t-test. In experiments examining the effect of switching cell cultures from one O₂ condition to the other, a parametric exponential decay model was used to regress the percent cells in S-phase (y) on time in weeks (x). The model is

$$y=\{\begin{array}{cc}{\alpha }_o+({\alpha }_h-{\alpha }_o)(1-\text{exp}(-{\beta }_{o}x))+\gamma e\hfill & \text{for cell line in}20\%{\mathrm{O}}_2\hfill \\ {\alpha }_h+({\alpha }_o-{\alpha }_h)(1-\text{exp}(-{\beta }_{h}x))+\gamma e\hfill & \text{for cell line in}5\%{\mathrm{O}}_2\hfill \end{array}\phantom{\}}$$

where e is a 0/1 indicator indexing the different experiments (0=Experiment 1, 1=Experiment 2). α_o and α_h are the zero-time intercept parameters for 317 and 317h cells, respectively, whereas β_o and β_h are the exponential decay parameters. In using this model, we are assuming that % S-phase in 317 cells that were switched to physiologic O₂ conditions will approach the zero-time value of 317h cells, and similarly that the % S-phase in 317h cells that were switched to atmospheric O₂ conditions will approach that of 317 cells. γ is a parameter that reflects the overall difference in S-phase percent between cohorts 1 and 2. The model was fit via non-linear unweighted least squares [30] to the combined 42 S-phase % values in cohort 1 (triplicates at 14 time points) and 30 S-phase % values in cohort 2 (triplicates at 10 time points). Statistical computations were performed with the NL function in Stata 9.2 software (StataCorp LP, College Station, TX).

For drug cytotoxicity data of each of the 6 drugs examined, two-way error-weighted analysis of variance (ANOVA) was used, with transformation of fluorescence readings to the natural log scale before ANOVA.

RT-PCR quantitative values were derived from CT values of each of the target genes that were normalized to the housekeeping gene β2 microglobulin using the Delta CT method. The gene expression score “– Delta CT” was defined as “–(CT_{Target Gene}–CT_{Housekeeping Gene})”. Linear mixed effect models were used to determine differences in gene expression levels between cells cultured in 5% versus 20% oxygen conditions, with the effects from cell lines, oxygen conditions, and the interaction between cell lines and oxygen conditions evaluated as fixed effects, and variations in plates evaluated as random effects. Analysis was performed for the four target genes separately. Statistical computation was performed using Stata software (StataCorp, College Station, TX).

Results

Characteristics of new cell lines

Utilizing freshly obtained specimens from 12 patients, we succeeded in establishing three pairs of pediatric ALL cell lines, one member of each pair cultured from the time of its isolation in 20% O₂ while the other member in 5% O₂ (Supplementary Table 2). COG-LL-317 (hereafter referred to as 317) and COG-LL-332 (332) cells were derived from patients with T cell ALL; COG-LL-319 (319) cells were from a patient with pre-B cell ALL defined according to surface marker staining and a low level of surface immunoglobulin. The names of cell lines established in 5% O₂ were appended with the letter “h”; e.g., 317h cells were established in 5% O₂ and correspond to 317 cells established from the same patient in 20% O₂. Full karyotypes are shown in Supplementary Fig. S1. The cell lines were negative for Epstein Barr Virus (EBV) gene expression (Supplementary Fig. S2) and have proliferated robustly in continuous culture at all time-points examined (more than 1 year).

The genetic identity of the cell lines was authenticated by genotyping 15 short tandem repeat (STR) loci and 1 gender-specific locus. STR genotyping demonstrated a positive match between each cell line and corresponding donor patient material (Supplementary Table 3, left columns). The STR profiles of these new cell lines were unique from those of all other cell lines in our laboratory, and unique in an STR database of >3000 cell lines that we maintain online (www.COGcell.org). STR analyses comparing relatively low passages (0–5) to high passages (120) demonstrated complete profile stability for 317 and 319 cell lines at each examined passage. In 332 cells, comparison to donor patient material revealed two microsatellite alterations by passage 43 at both 5% and 20% O₂ (Supplementary Table 3, right columns). Acquisition of microsatellite alterations in 332 cells was not limited to in vitro conditions, since two microsatellite alterations were also observed after the first round of xenografting the original patient material directly into NOD/SCID mice (data not shown).

The potential of these newly established cell lines to grow in vivo was confirmed when 5 × 10⁶ 317h, 319h, or 332h cells were injected into the tail veins of young, immunodeficient NOD/SCID mice (4 mice/cell line), which subsequently survived for means of 6.5, 9, and 4.5 weeks, respectively. Upon sacrifice, all xenografted mice contained enlarged spleens comprised almost entirely (95–98%) of cells expressing human CD45 (Supplementary Fig. S3).

Sensitivity to chemotherapeutic drugs

The drug sensitivities of cells established and maintained in 20% versus 5% O₂ levels differed, but in an unpredictable and cell line-dependent manner. For example, whereas 319h and 332h cells exhibited greater sensitivity to the DNA-damaging agents doxorubicin, etoposide, 4-hydroxycyclophosphamide (4-HC), and vincristine than did 319 and 332 cells at the highest three doses examined, 317h cells exhibited less sensitivity than 317 cells (P<0.05, Fig. 1), with the caveat that 317 and 317h cells were no longer within the 4 log range of the assay at the highest dose of doxorubicin. A similarly complex pattern was observed for the non-DNA damaging drug l-asparaginase and the steroid dexamethasone. All three cell line pairs were negative for expression of the drug efflux proteins MDR1 (ABCB1) and MRP1 (ABCC1), and slight O₂ concentration-related changes in expression of BCRP1 (ABCG2) did not correlate with changes in drug sensitivity (data not shown).

Fig. 1 –

Effects of O₂ conditions on drug dose response curves. 317 (top row), 319 (middle row), and 332 (bottom row) cell line pairs were treated with 0.03–30 nM doxorubicin (DOX), 0.3–30 nM etoposide (ETOP), 0.01–10 mg/ml 4-hydroxycyclophosphamide (4-HC), 0.1–100 ng/ml vincristine (VCR), 0.01–10 IU/ml l-aparaginase (ASP), and 0.01–10 mM dexamethasone (DEX). With the exception of the highest dose of each drug, the selected concentrations represent clinically achievable doses. Multi-log cytotoxicity was measured using a digital image microscopy system. Fractional survival of treated versus untreated control cells is shown on log scale. Cells established in physiologic O₂ levels, black circles; cells established in atmospheric O₂ levels, gray circles; cells established in a 20% O₂ incubator (atmospheric O₂ levels) but switched to a 5% O₂ incubator (physiologic levels) for eight weeks, open triangles. The prolonged, eight week interval was chosen to allow sufficient time to observe gradual effects. Each condition had 6 replicates and bars represent standard deviation.

To examine whether drug sensitivities could be changed by switching the O₂ concentration, cells established in atmospheric O₂ were moved to culture in a 5% O₂ incubator for eight weeks prior to determining dose response curves. With a few exceptions (notably etoposide), switching from atmospheric to physiologic O₂ concentration generally induced greater drug sensitivity, rather than inducing a pattern similar to that of cells established in physiologic O₂ (Fig. 1, triangles). These findings demonstrate that different and unpredictable drug sensitivities can exist in ALL cells established and maintained in atmospheric versus physiologic O₂ concentrations, and suggest that switching the O₂ concentration of cultured cells from 20% to 5% tends to increase the sensitivity of ALL cells to most chemotherapeutic drugs rather than mimicking the sensitivity of cells established in 5% O₂.

Immunophenotype

The immunophenotypes of the cell line pairs were compared with their xenografts and with those of original clinical specimens evaluated at diagnosis. Little difference was observed between 317 and 317h cells or between 319 and 319h cells (Fig. 2). In contrast, while 332 cells lacked expression of CD2 and clearly expressed CD33 on ~40% of cells, 332h cells expressed a high level of CD2 but lacked CD33. This latter profile resembled those of xenotransplanted cells and original patient material. This finding demonstrates that expression of specific cell surface markers can be affected by the oxygen concentration in which a cell line is established.

Fig. 2 –

Immunophenotype of in vitro leukemia cell lines, in vitro cells passaged as xenografts in vivo, and original patient samples. Immunophenotypes of leukemia cell lines established and maintained in 20% O₂ (top row) or 5% O₂ (second row) were examined using flow cytometry. To determine the immunophenotype of cell lines xenotransplanted into NOD/SCID mice (third row), cells from cultures in 5% O₂ (317h, 319h, 332h) were tail-vein injected, and viable cells from mouse spleens were examined in three-color analyses, gating on human cells according to negative staining with anti-mouse CD45 mAb and positive staining with anti-human CD45 mAb (as in Fig. S3), with immunophenotype evaluated in the third color. Immunophenotypes of clinical samples, which included a very low percentage of normal bone marrow cells, are also presented (bottom row). Results represent means±standard error of three independent experiments performed for xenografts and for each cell line under both culture conditions. Asterisks indicate statistical significance from cells established and maintained in 5% O₂ (P<0.05). Patient clinical specimen data was from routine clinical testing performed at diagnosis.

Differential rates of cell expansion

The growth characteristics of the cell populations established in both O₂ conditions were examined. Using an automated viable cell counter, all three cell lines established in 20% O₂ were observed to expand their numbers more slowly than parallel cultures established in 5% O₂ (P<0.001) (Fig. 3A). Similarly, serial dilution experiments using digital imaging microscopy showed decreased plating efficiency for the three cell lines maintained in 20% O₂ relative to 5% O₂ (P<0.001) (Fig. 3B). Furthermore, all three cell lines in 20% O₂ exhibited a reduced number of cell divisions, as determined in 6-day CSFE assays (Fig. 3C).

Fig. 3 –

Effects of cell culture oxygen concentration on growth kinetics. (A) Population growth of cell lines was measured using an automated trypan blue staining system. 317 cells were plated in fresh medium at 0.3 × 10⁵ cells/ml, 319 cells at 3 × 10⁵ cells/ml, and 332 cells at 1 × 10⁵ cells/ml in 24-well plates (1 ml/well), and triplicate wells per cell line were measured on each of the days shown. Open circles, 20% O₂; closed circles, 5% O₂. The experiments shown are representative of at least three independent experiments per cell line. (B) Cell plating efficiency was measured in a clonogenic assay of serially diluted cells using fluorescence-based digital image microscopy to quantify viable cells. Serially diluted cells were incubated for six days and then stained with fluorescein diacetate (FDA). (C) Cell divisions were quantified by staining with CFSE, incubating for six days, and determining fluorescence using flow cytometry. For each cell line, voltages on photo-multiplier tubes (PMTs) were kept constant for cultures in parallel O₂ conditions. (D) Cell cycle analysis according to incorporation of the thymidine analog bromodeoxyuridine (BrdU) into DNA as cells enter S-phase. Continuously growing cells were plated at 1 × 10⁵, 4 × 10⁵, and 2 × 10⁵ cells/ml for 317, 319, and 332 cells, respectively. (E) Left bar graphs show the means of frequencies of cells in S-phase, performed as in Fig. 2D, with bars representing standard deviation from 3–5 independent experiments per cell line; open bars, 20% O₂; filled bars, 5% O₂. Right bar graphs show the mean frequencies of sub-G1 events. (F) Effect of switching the cell line oxygen conditions on proliferation rate. 317 cells (triangles) were switched from a 20% O₂ incubator to a 5% O₂ incubator, and 317h cells (squares) from 5% to a 20% O₂ incubator. BrdU incorporation was examined every seven days for thirteen weeks in the first cohort (black markers) and nine weeks in the second cohort (gray markers). Regression lines were calculated using a parametric exponential decay model as described in Materials and methods, and predicted a convergence of the regression lines at 30 weeks (not shown). Error bars represent the mean of triplicate wells analyzed at each time point for each cell line.

The BrdU incorporation assay was used to simultaneously examine the frequency of cells in S-phase (as a measure of proliferation) and the fraction of cells manifesting sub-G1 DNA content (as a measure of cell death frequency). All three cell lines established in 20% O₂ exhibited reduced frequencies of cells incorporating BrdU relative to parallel cultures in 5% O₂ (Fig. 3D), at a significance level of P<0.05 (Fig. 3E, first column of bar-graphs). In addition, cell lines established in 20% O₂ exhibited modestly higher frequencies of sub-G1 events than parallel cultures in 5% O₂ (P<0.05) (Fig. 3E, second column of bar-graphs). These findings demonstrate that lower rates of population growth observed in 20% O₂ culture conditions were due to both decreased cell cycling and increased spontaneous cell death.

To determine whether the effects of establishment oxygen conditions were reversible, aliquots of 317h cells were moved to culture in a 20% O₂ incubator and conversely 317 cells were moved to culture in a 5% O₂ incubator. Unexpectedly, only a modest change in the frequency of cells exhibiting BrdU incorporation was observed in either cell culture condition over a 13-week period (Fig. 3F). This demonstrates that the effects of switching O₂ conditions can be gradual, and suggests that population growth rates of ALL cell lines are not readily changed by altering the O₂ conditions in which they were established.

Differential expression of glycolysis-related genes

Levels of reactive oxygen species (ROS) can affect multiple cellular processes including proliferation, but ROS levels measured by dichlorofluorescein diacetate staining and flow cytometry did not correlate with the observed differences in cell line expansion (data not shown). Therefore, to explore the molecular differences between the cell lines established in the two culture conditions, we used whole genome expression analysis. DNA oligonucleotide microarrays containing 54,000 probe sets (~ 38,500 transcripts) identified 69 probe sets (representing 59 genes) whose average expression exhibited a greater than 1.5-fold difference between the two oxygen conditions for all three cell line pairs (P<0.005). Of these genes, 48 exhibited lower expression in 20% O₂ while 11 exhibited higher expression (Fig. 4A). As expected for cell lines maintained in 20% O₂, lower expression was observed for vascular endothelial growth factor (VEGF) and BNIP3. Interestingly, among the probe sets exhibiting lower expression in 20% O₂ were six sets representing four genes implicated in glycolysis, pyruvate dehydrogenase kinase isoenzyme 1 (PDK1), lactate dehydrogenase A (LDHA), triosephosphate isomerase 1 (TPI1), and aldolase C (ALDOC). Lower expression of these genes in cell lines established in 20% O₂ was confirmed by real-time RT-PCR (Fig. 4B). The positions of the products of these genes in the glycolytic pathway are shown in Fig. 4C, with PDK1 and LDHA holding notably pivotal positions.

Fig. 4 –

Effects of cell culture oxygen concentration on gene expression. (A) Whole genome analysis providing hierarchical clustering of genes differentially expressed under 5% and 20% O₂ conditions. Columns represent RNA expression of the indicated cell lines (two passages per cell line). Each row represents a probe set (some genes are recognized by more than one probe set). Names of glycolysis-regulating genes are denoted in blue. The heat map indicates a high (red) or low (green) level of expression, according to the scale shown at the bottom. (B) mRNA expression of glycolysis-regulating genes determined by RT-PCR. Results represent the mean and standard deviation of at least two independent experiments per gene product. Expression relative to 317 cells is shown, except in the cases of ALDOC and LDHA expression in 332 cells, which were examined in separate experiments. Asterisk (*) denotes P<0.01. (C) Abbreviated representation of the glycolysis pathway.

Lower protein expression of PDK1, LDHA, TPI1, and ALDOC in cell lines established in 20% O₂ was confirmed by immunoblotting (Fig. 4, lanes 1–2, 5–6, 9–10). The transcriptional regulator of these genes, HIF-1α, as well as glucose transporter-1 (Glut-1), a glycolysis-related transcriptional target of HIF-1α, also exhibited lower expression in cells established and maintained in 20% O₂ (Fig. 5). Interestingly, switching the O₂ conditions for an 8-week interval resulted in only partial reversal of protein expression patterns, generally resulting in intermediate expression levels (Fig. 4, lanes 3–4, 7–8, 11–12). Examining proteins known to regulate glycolysis with HIF-1α, PFKFB3 showed modestly lower expression in cells established and maintained in 20% O₂, but little or no differences were observed in expression of PKM1 and PKM2 (data not shown), consistent with direct transcriptional activation by HIF1α as a main mechanism affecting expression of PDK1, LDHA, TPI1, ALDOC, and Glut-1. Together, these findings suggested that aerobic glycolysis was operating at a lower level in ALL cell lines established in atmospheric rather than physiologic O₂, and demonstrate that the expression levels of glycolysis-regulating proteins can be semi-permanently affected by the oxygen tensions in which cell lines are established.

Fig. 5 –

Effects of cell culture oxygen concentration on expression of glycolysis-related proteins. Cell culture lysates were subjected to immunoblotting. For each cell line, the first two lanes represent cultures established in 20% and 5% O₂, while the second two lanes (gray font) represent the same cell lines eight weeks after switching their culture conditions from 20% to 5% O₂, or from 5% to 20% O₂. Actin staining was used to control for loading.

Differential glucose consumption and lactate production

To confirm that glycolysis was reduced in cells established in atmospheric oxygen conditions, relative glucose consumption was examined. After equivalent numbers of cells were incubated in fresh culture medium for 24 h, measurement of residual glucose in the medium indicated two- to three-fold lower levels of glucose consumption (P<0.005) in cells established in 20% O₂ relative to 5% O₂. Switching the oxygen conditions of cultures for eight weeks resulted in only intermediate changes in the levels of glucose consumption, with the exception of 319 cells which exhibited no detectable change (Fig. 6A).

Fig. 6 –

Effects of cell culture oxygen concentration on glucose consumption and lactate production. (A, B) White and black bars represent cultures established in 20% or 5% O₂, respectively, while light gray and dark gray bars represent the same cell lines eight weeks after their O₂ conditions were switched to 5% O₂ or 20% O₂, respectively. (A) Glucose consumption. Twenty-four hours after plating 10⁶ cells/ml in glucose-free/pyruvate-free IMDM supplemented with 1 mg/ml glucose and 10% FBS, culture medium was examined for glucose concentration levels. (B) Lactate production. Twenty-four hours after plating 10⁶ cells/ml in glucose-free/pyruvate-free IMDM supplemented with 4.5 mg/ml glucose and 5% FBS, cell culture media was examined for lactate concentration. Asterisk (*) denotes P<0.02; n.s.=not significant. Results are representative of three independent experiments.

To further characterize the function of glycolysis under atmospheric O₂ conditions, lactate levels were assayed in culture medium after cells were incubated in fresh medium for 24 h. Lower (P<0.01) levels of lactate (2.5-fold to 9-fold) were observed in cultures established in 20% O₂ relative to 5% O₂ (Fig. 6B). Switching the oxygen culture conditions for eight weeks resulted in only intermediate changes in the levels of lactate production (Fig. 6B). These findings confirmed that aerobic glycolysis was functionally lower in cell lines established at atmospheric O₂ concentration, and indicated a semi-permanence of the metabolic phenotype established at cell line initiation.

Discussion

Acquisition of genetic and epigenetic alterations in cell lines in vitro represents an obstacle for cancer research. Supra-physiologic levels of O₂ may result in generation of oxygen radicals that can induce oxidative damage, DNA strand breaks, and mutations [3], and may thereby provide a non-physiologic selection pressure potentially leading to permanent alterations in cancer cell lines that render them less representative of the original cancers. For example, the genome of HeLa cervical carcinoma cells has been reported to be relatively stable over time compared with that of other cancer cell lines; however, in this relatively stable cell line, there was evidence for multiple somatic mutations acquired between different HeLa clones in vitro within just a limited number of gene regions examined [31]. These observations suggest that conventional methods for culturing cells may be suboptimal, and improvements to conditions in which cell lines are established and maintained to better mimic conditions in the cancer microenvironment may be warranted.

A few studies have reported the culturing of leukemia cell lines at or near physiologic O₂ concentration [7,8]. Initiation in 5% O₂ of a cell line from a 12-year old T-ALL patient and three cell lines from other types of cancer was reported previously [32]; however, it was not reported whether those lines were established simultaneously in atmospheric O₂ and therefore no comparison of the effects of establishing and maintaining cell lines in physiologic versus atmospheric O₂ levels was available in that report. The cell lines described here provide this comparison, they provide models for biological and preclinical therapeutic studies at physiologic O₂, and they may be useful for defining additional molecular and biologic differences that may exist between ALL cell lines established at atmospheric versus physiologic O₂.

The findings reported here demonstrate that establishment of pediatric ALL cell lines at physiologic O₂ concentration can result in higher proliferation and glycolytic function than cell lines established from the same patients at atmospheric O₂ concentration. These findings suggest that the common practice of initiating ALL cell lines in atmospheric O₂ levels may compromise the metabolic phenotype that was established by the leukemias in vivo. One potential explanation for diminution of glycolysis resulting from cell line initiation in hyperoxic culture conditions is that atmospheric O₂ levels might have permitted diversion of cellular energy production away from glycolysis towards mitochondrial oxidative phosphorylation, a change which became semi-permanently fixed in the cell population through a currently undefined mechanism. An alternative explanation that easily clarifies the observed semi-permanence of the resulting metabolic phenotypes is that establishment of each cell line in atmospheric O₂ concentration selected for a subpopulation of initiating blast cells that preferentially utilized pyruvate and O₂ as substrates to drive oxidative phosphorylation, whereas establishment in physiologic oxygen tension selected for the growth of a subpopulation of cells that utilized higher levels of glucose and glycolysis. Consistent with the latter explanation, different subpopulations of cancer cells employing different levels of glycolysis have been suggested previously [33,34].

Rapidly proliferating cells are believed to frequently exhibit heightened sensitivity to DNA-damaging agents. Also, a subset of genes that regulate cell proliferation was reported to correlate with drug sensitivity in pediatric ALL [35]. It was therefore conceivable that all three cell lines established and maintained in 5% O₂ would exhibit a greater response to the DNA-damaging agents doxorubicin, etoposide, 4-hydroxycyclophosphamide, and vincristine. However, we observed that drug sensitivities of cells established in 20% versus 5% O₂ varied in a unpredictable, cell line-dependent manner. Whereas 319h and 332h cells exhibited greater sensitivity to DNA-damaging agents, 317h cells exhibited less sensitivity. This latter result might reflect the previous observation that hypoxia can result in heightened apoptosis resistance in some cell types [36-38]. Given the unpredictable drug responses observed, determination of the aspects affecting drug sensitivities of cell lines established and maintained in disparate O₂ conditions requires further study in additional experimental models using targeted approaches.

Taken together, the findings presented here suggest the possibility that many existing ALL cell lines may be comprised of subpopulations that do not optimally represent the dominant metabolic phenotypes of the leukemias from which they were derived. Cell lines established at physiologic O₂ levels and exhibiting a high glycolytic phenotype likely provide better representations of the majority of acute lymphoblastic leukemias growing in patients. In conclusion, the current results suggest that, in order to better model the in vivo metabolic states of the majority of acute lymphoblastic leukemia, new in vitro cell line models should be established directly from patient blasts in culture conditions employing physiologic O₂ concentrations.