Unconjugated bilirubin induces apoptosis in colon cancer cells by triggering mitochondrial depolarization



Bilirubin is the principal end product of heme degradation. Prompted by epidemiologic analyses demonstrating an inverse correlation between serum bilirubin levels and cancer mortality, we examined the effect(s) of bilirubin on the growth and survival of colon adenocarcinoma cells. Adenocarcinoma cell monolayers were treated with bilirubin over a range of bilirubin:BSA molar ratios (0–0.6), and viability was assessed colorimetrically. Apoptosis was characterized by TUNEL assay, annexin V staining and caspase-3 activation. The mechanism(s) by which bilirubin induces apoptosis was investigated by Western blotting for cytochrome c release, assaying for caspase-8 and caspase-9 activation and for mitochondrial depolarization by JC-1 staining. The direct effect of bilirubin on the membrane potential of isolated mitochondria was evaluated using light-scattering and fluorescence techniques. Bilirubin decreased the viability of all colon cancer cell lines tested in a dose-dependent manner. Cells exhibited substantial apoptosis when exposed to bilirubin concentrations ranging 0–50 μM, as demonstrated by an 8- to 10-fold increase in TUNEL and annexin V staining and in caspase-3 activity. Bilirubin treatment evokes specific activation of caspase-9, enhances cytochrome c release into the cytoplasm and triggers the mitochondrial permeability transition in colon cancer monolayers. Additionally, bilirubin directly induces the depolarization of isolated rat liver mitochondria, an effect that is not inhibited by cyclosporin A. Bilirubin stimulates apoptosis of colon adenocarcinoma cells in vitro through activation of the mitochondrial pathway, apparently by directly dissipating mitochondrial membrane potential. As this effect is triggered at concentrations normally present in the intestinal lumen, we postulate a physiologic role for bilirubin in modulating colon tumorigenesis. © 2004 Wiley-Liss, Inc.

Bilirubin, a major product of heme catabolism, is generated by heme oxygenase–mediated oxidation of heme to form biliverdin, followed by biliverdin reductase catalyzed reduction to the physiologic isomer bilirubin IXα. Bilirubin is cleared from the circulation by the liver, where it is conjugated by the bilirubin-specific 1A1 isoform of UGT (UGT1A1). The resultant bilirubin mono- and diglucuronides are actively secreted across the canalicular membrane of the hepatocyte into the bile by MRP2. In some nonmammalian vertebrates (e.g., birds, reptiles, amphibians), heme degradation terminates at the heme oxygenase step,1 with the biliverdin end product efficiently eliminated in the bile and urine.2, 3 The evolutionary development of the energetically costly process of bilirubin production and elimination suggests that this bile pigment may serve an important physiologic function.

As bilirubin is a potent antioxidant4, 5, 6, 7 and oxidative stress is considered to be a major contributing factor to carcinogenesis,8 it has been proposed that bilirubin may protect against the development of malignancy.9, 10 In support of this hypothesis, a large Belgian population study11 and an unpublished Veterans Administration analysis12 found that individuals with high baseline serum bilirubin concentrations exhibit significantly reduced rates of cancer-related mortality. Additionally, a case-control study by Ching et al.10 reported an odds ratio for breast cancer of 0.5 for subjects in the highest vs. the lowest quartile of serum bilirubin.

While there have been a handful of in vitro studies demonstrating cytotoxic effects of bilirubin on neuroblastoma,13, 14 hepatoma,15, 16 fibrosarcoma17 and Ehrlich ascites tumor cells,18 there is scant information regarding the mechanism underlying these effects. Bilirubin inhibits osteoblast proliferation19 and stimulates apoptosis in rat brain neurons,20, 21, 22, 23 astrocytes,24 fibroblasts24 and bovine brain endothelial cells.25 However, many studies of bilirubin-induced apoptosis have been conducted at high molar ratios of bilirubin to albumin or under serum-free conditions. Due to the low aqueous solubility at neutral pH,26 substantial aggregation of bilirubin is likely to have occurred during the course of these experiments, making the physiologic relevance of the findings uncertain.27

Although biliary bilirubin is predominantly in the form of mono- and diglucuronides,28 significant deconjugation occurs as a result of spontaneous hydrolysis29 as well as through the action of mucosal30, 31 and bacterial32 β-glucuronidases. Since unconjugated bilirubin normally is present within the intestinal lumen,33 we examined the modulatory effects of this bile pigment on colon cancer cell growth and survival. Our data indicate that physiologic concentrations of bilirubin induce apoptosis in colon adenocarcinoma cell lines through a mechanism that appears to involve direct depolarization of cellular mitochondria.


BDT, bilirubin ditaurate (ditaurobilirubin) ; BrdU, brominated deoxyuridine triphosphate ; CCCP, carbonyl cyanide 3-chlorophenylhydrazone; COX, cyclooxygenase; DTT, dithiothreitol; LDH, lactate dehydrogenase; MRP2, multidrug resistance-associated protein 2; PI, propidium iodide; PTP, permeability transition pore; TdT, terminal deoxynucleotidyl transferase; TNF-α, tumor necrosis factor-α; TUNEL, TdT-linked UTP nick-end labeling; UGT, uridine diphosphate glucuronosyltransferase.


Unconjugated bilirubin (bilirubin IXα), biliverdin (biliverdin IXα) and BDT (BDT · 2Na) were obtained from Porphyrin (Logan, UT). Bilirubin IXα was further purified according to the method of McDonagh and Assisi,34 to eliminate potential lipid contaminants. The resultant bilirubin crystals were desiccated under vacuum in a light-protected vessel until use, with purity assessed by absorbance in chloroform solution (ϵ453 nm = 60,000–62,000 M–1 · cm–1). Extinction coefficients for commercial biliverdin (ϵ376 nm = 43,000 M–1 · cm–1) and BDT (ϵ451 nm = 42,000 M–1 · cm–1) were determined in methanol and water, respectively. DMSO was purchased from Fisher Scientific (Pittsburgh, PA). NS398, a COX-2-specific inhibitor, was obtained from Biomol (Plymouth Meeting, PA). TdT was purchased from GIBCO BRL (Carlsbad, CA). FITC-conjugated anti-BrdU antibodies were obtained from Phoenix Flow Systems (San Diego, CA). Fluorescent probes JC-1 and rhodamine-123 were obtained from Molecular Probes (Eugene, OR). TNF-α (human recombinant) and CCCP were purchased from Sigma (St. Louis, MO).

Cell lines and cell culture techniques

HCT15, HCT116, SW480 and LoVo colon adenocarcinoma cell lines were obtained from the ATCC (Manassas, VA). HT29 and CaCo2 colon cancer cells were kindly provided by Dr. M. Cohen (Cincinnati Children's Hospital, Cincinnati, OH) and Dr. R. Giannella (University of Cincinnati), respectively. Cell monolayers were maintained at 37°C in 5% CO2 in ATCC-recommended growth media (RPMI-1640 with L-glutamine and 25 mM HEPES for HCT15 cells, DMEM with 10 mg/ml human apotransferrin and 1% penicillin/streptomycin for HT29 cells, McCoy's 5A with 1.5 mM glutamine for HCT116 cells, DMEM with 1% penicillin/streptomycin for CaCo2 cells, Ham's F-12 with 2 mM glutamine for LoVo cells, Leibovitz's L-15 with L-glutamine for SW480 cells). Fresh medium was added at the start of each treatment period. Cell culture products were obtained from Fisher Scientific or GIBCO BRL.

Solubilization and addition of reagents

Stock solutions of bilirubin, biliverdin, BDT and NS398 were freshly prepared in DMSO immediately prior to each experiment and an aliquot (≤1% v:v) added to the cell culture medium. To exclude potential confounding effects related to the use of DMSO as a solubilizing agent, all experiments included a vehicle control group. At the low concentrations utilized, DMSO had no demonstrable effect on the viability of any of the cell lines studied. Furthermore, we employed an alternative method for bilirubin solubilization involving dissolution of bilirubin in 50 mM potassium phosphate (pH 12), as previously described.35, 36 Addition of bilirubin solubilized in this manner in an aliquot comprising ≤1% of the total volume did not alter the pH of the culture medium.37 Using this alternative dissolution method, viability results were identical to those obtained using DMSO as the bilirubin vehicle. To insure adequate pigment solubilization during the incubation period, the culture medium contained 20% FBS in all cell experiments, with the exception of studies in which TNF-α was added to cell monolayers. As TNF-α is inactivated by serum, this cytokine was dissolved in PBS (pH 7.4) containing 1% BSA and experiments were conducted in serum-free medium. All experiments were performed in a manner to insure that bilirubin was protected from exposure to light.

Cell viability assay

Cell viability was assessed using the Cell Titer 96 Non-Radioactive Cell Proliferation Assay kit (Promega, Madison, WI), which is based on the metabolism of MTT. Cell monolayers were harvested at approximately 70% confluence and seeded in 96-well plates at a density of 2.0 × 105 cells/ml. Following a 24 hr recovery period, monolayers were incubated for the indicated time intervals in the presence of bilirubin (0.5–50 μM), biliverdin (50 μM), BDT (50 μM) or an equivalent volume of the DMSO vehicle. This was accomplished by dissolving bilirubin in DMSO and adding a 100 μl aliquot of this stock solution to 10 ml of fresh culture media containing 20% FBS to produce the appropriate final bilirubin concentration. Cells were then overlaid with 100 μl of bilirubin-containing medium and incubated for the indicated time intervals. The number of viable cells was determined colorimetrically at 570 nm using a 96-well plate reader (SpectraMax 340PC; Molecular Devices, Palo Alto, CA). To account for potential interference with the viability assay,38 in these and all other experiments, data were corrected for the intrinsic absorbance of bilirubin (and other added compounds) by subtracting the absorbance of control incubations performed in the absence of cells.

Semiquantitative RT-PCR analysis

Semiquantitative RT-PCR was performed on whole-cell RNA using the Retroscript PCR kit (Ambion, Austin, TX). Primer pairs specific for human UGT1A1 (5′-GGT GAC TGT CCA GGA CCT ATT GA-3′ and 5′-TAG TGG ATT TTG GTG AAG GCA GTT-3′) and biliverdin reductase (5′-GGA AGA GAC CAA GAT GAA TA-3′ and 5′-CAA GAG TTC AAC ATG CTC-3′) were utilized for amplification,39, 40 with the rig/S15 small ribosomal subunit protein (5′-TTC CGC AAG TTC ACC TAC C-3′ and 5′-CGG GCC GGC CAT GCT TTA CG-3′) serving as control. Reactions were subjected to 35 cycles at 95°C for 1 min, 58°C for 1 min and 72°C for 2 min.

Measurement of cellular LDH release

LDH release into the culture medium was assessed using the CytoTox 96 Non-Radioactive Cytotoxicity Assay kit (Promega). Cells were seeded in a 96-well plate at 2.0 × 105 cells/well. After a 24 hr recovery period, monolayers were incubated in the presence or absence of 50 μM bilirubin or an equal volume of DMSO vehicle, as described above for the assessment of cell viability. Culture medium was sampled after 24 and 48 hr of treatment and assayed for LDH activity by measuring absorbance at 490 nm. Maximum cellular LDH release was determined by addition of 10% Triton X-100, to induce complete cell lysis.

TUNEL assay

Cell monolayers were grown in T75 flasks to a concentration of approximately 1 × 107 cells/flask. A 100 μl aliquot of DMSO vehicle (control) or a freshly prepared bilirubin stock solution solubilized in DMSO was added to 10 ml of culture medium containing 20% FBS to produce a final concentration 50 μM bilirubin, and the entire volume was then overlaid onto the cells. After 48 hr of treatment, cells were harvested and fixed in 1% methanol-free formaldehyde, followed by 70% ethanol. Permeabilized cells were incubated in the presence of TdT and BrdU overnight at 22°C and then with FITC-conjugated anti-BrdU antibodies for 1 hr. Total cellular nucleic acids were labeled with PI for 30 min prior to analysis by flow cytometry (EPICS-XL; Coulter, Hialeah, FL). Control cells were treated in an identical manner except that BrdU was excluded from the incubation medium.

Annexin V staining

Annexin V staining was performed using the ApoTacs system (Biosource, Camarillo, CA). Cells were harvested 24 or 48 hr following addition of 50 μM bilirubin, 100 μM NS398 or DMSO vehicle to the culture medium, using the same approach outlined above for the TUNEL assay. Monolayers were washed twice with PBS and stained with annexin V–FITC, according to the manufacturer's instructions. Counterstaining with PI was performed immediately prior to flow cytometry.

Caspase activation assays

Cellular caspase activity was measured using caspase-specific assay kits from R&D Systems (Minneapolis, MN). Monolayers were grown to confluence in serum-containing medium and then incubated in the presence of 50 μM bilirubin, 250 ng/ml TNF-α or the appropriate vehicle controls for 12 (caspase-8) or 48 (caspases 3 and 9) hr. Cells treated with TNF-α or PBS vehicle (but not bilirubin or DMSO vehicle) were repeatedly washed and then incubated in serum-free medium immediately prior to treatment, to prevent TNF-α inactivation. Cell monolayers were harvested and the lysates assayed for caspase activity by cleavage of specific peptide-conjugated substrates (DEVD for caspase-3, LEHD for caspase-9 and IETD for caspase-8), as detected by changes in optical density at 405 nm.

Determination of cytochrome c release into the cytosol

Following incubation in the presence or absence of 50 μM bilirubin, 100 μM NS398 or DMSO vehicle for the indicated time intervals, cells were harvested and washed in PBS. The cell pellet was suspended in 5 volumes of ice-cold 0.25 M sucrose, 10 mM KCl, 1.5 mM MgCl, 20 mM HEPES, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 0.1 mM PMSF, 1 μg/ml aprotinin, 10 μM leupeptin and 1 μg/ml pepstatin (pH 7.5) and subjected to 40 strokes with a Dounce homogenizer.41 Cell homogenates were centrifuged serially at 1,000g for 10 min, 10,000g for 30 min and 100,000g for 1 hr at 4°C. Supernatant (cytosol) was harvested and the protein concentration determined using the Bio-Rad (Hercules, CA) assay. Samples containing 50 μg protein were subjected to SDS-PAGE on a 4–12% precast polyacrylamide gel (Invitrogen, Carlsbad, CA). Murine antihuman cytochrome c primary antibodies (eBioscience, San Diego, CA) were used for Western blotting (1:500 dilution). Bands were detected by enhanced chemiluminescence (Amersham, Piscataway, NJ) with goat antimouse IgG conjugated to horseradish peroxidase (Pierce, Rockford, IL).

Detection of mitochondrial depolarization in cultured cells

Cell monolayers were incubated in the presence of 50 μM bilirubin or DMSO vehicle and then harvested. Following a PBS wash, cells were suspended in culture medium containing the fluorescent probe JC-1 (10 μg/ml) for 10 min and then analyzed by flow cytometry.42

Isolation of rat liver mitochondria

Rat liver mitochondria were prepared according to the method of Constantini et al.43 Briefly, male Sprague-Dawley rats (250–300 g) were anesthetized and the livers rapidly excised and placed in ice-cold isolation medium (210 mM mannitol, 70 mM sucrose, 5 mM HEPES, 1 mM EGTA, pH 7.5). Organs were washed twice, minced and subjected to 10 strokes of a Dounce homogenizer. The homogenate was centrifuged (650g) for 10 min at 4°C, following which the supernatant was decanted and centrifuged at 7,700g for 10 min. The pellet was washed once, resuspended in isolation medium without EGTA and utilized immediately.

Assessment of the permeability of isolated mitochondria

Rat liver mitochondria (0.2 mg protein/ml) were suspended in 2.5 ml of 0.1 M KCl, 0.1 mM sucrose, 10 mM TRIS-HCl and 3 μM rotenone (pH 7.4) at room temperature in a continuously stirred cuvette. Small aliquots of calcium chloride (50 μM) and succinate (10 mM) were added to the mitochondrial suspension immediately prior to treatment with depolarizing agents. The time-dependent decrease in light scattering, indicative of large-amplitude mitochondrial swelling,44 was continuously monitored using an Aminco-Bowman II fluorescence spectrophotometer (SLM Instruments, Inc., Rochester, NY), with excitation and emission wavelengths at 620 nm.45 Confirmatory studies of mitochondrial membrane potential were performed by monitoring changes in rhodamine fluorescence (excitation 488 nm, emission 525 nm) according to the method of Notario et al.46 Experiments were conducted as outlined above, except that mitochondria were suspended in buffer containing 5 μM rhodamine-123 prior to addition of calcium. Mitochondrial depolarization is reflected by a time-dependent increase in rhodamine fluorescence intensity. All data are corrected for bilirubin absorbance.47

Statistical analyses

Data were analyzed using a computer-based statistical package (Statistix 7; Analytical Software, Tallahassee, FL) with differences between mean values assessed by ANOVA with Scheffe's comparison.


Bilirubin decreases colon adenocarcinoma cell viability

The effect of bilirubin on the viability of 6 colon adenocarcinoma cell lines (HCT15, HT29, HCT116, SW480, LoVo, CaCo2) was tested by incubating monolayers in the presence of unconjugated bilirubin (0–50 μM) for 48 hr. As the 20% FBS present in the culture medium corresponds to a BSA concentration of approximately 90 μM,48 the ratio of bilirubin to BSA is estimated to range 0–0.6 in these experiments. At the end of the treatment period, the number of viable cells was assessed colorimetrically by the metabolism of MTT. As shown in the left panel of Figure 1, a dose-dependent decrease in the number of viable cells was observed, although the magnitude of the effect differed between the various cell lines. HCT15, HCT116 and SW480 cells were most sensitive to bilirubin treatment, exhibiting only 10% of the number of cells present in the control group at the 50 μM dose. However, 50 μM bilirubin resulted in a more modest reduction in cell number (approx. 40%) in the HT29, CaCo2 and LoVo cell lines. Notably, mRNA levels for UGT1A1, as determined by RT-PCR, were over 4-fold higher in HT29 vs. HCT15 cells, while message levels for biliverdin reductase were similar between the 2 cell lines (Fig. 1, right panel). We speculate that the increased expression of this primary bilirubin-conjugating enzyme may explain, at least in part, the greater resistance of the HT29 cell line to bilirubin cytotoxicity.

Figure 1.

Effect of bilirubin on colon cancer cell viability. Monolayers of HCT15 (black bars) and HT29 (gray bars) colon adenocarcinoma cells were incubated for 48 hr in the presence of bilirubin (0.5–50 μM), biliverdin (BV, 50 μM) or DMSO vehicle (0 μM). At left, cell viability is plotted relative to vehicle-treated controls. Bars represent the average (±SD) of 5 replicate experiments. At right, representative blots of RT-PCR analyses of whole-cell RNA isolated from HCT15 and HT29 monolayers incubated for 24 hr in the presence or absence (NT) of bilirubin (50 μM) or an equivalent volume of DMSO vehicle. Primer pairs specific for UGT1A1 and biliverdin reductase (BVR) were employed, with amplification of a 361 bp region of the rig/S15 ribosomal subunit protein serving as control. *p < 0.0001 vs. vehicle.

Neither the bilirubin precursor biliverdin (Fig. 1, left panel) nor the chemically stable BDT (data not shown) altered the number of viable cells present at 48 hr, suggesting that the observed responses are bilirubin-specific. The additional finding that HCT15 monolayers incubated in the presence of bilirubin exhibit a marked decrease in the number of viable cells over time (Fig. 2, left panel) supports the notion that bilirubin induces cell death as opposed to suppressing growth. The gradual decline in bilirubin absorbance in the culture medium suggests that, at the highest (50 μM) concentration utilized, approximately 3% degradation occurs by 12 hr, 8% by 24 hr and 15% over the course of 48 hr incubation (Fig. 2, right upper panel). Additionally, the time-dependent emergence of a small absorbance peak at 450 nm indicates that approximately 10% of added biliverdin undergoes catalyzed conversion to bilirubin by 48 hr of incubation (Fig. 2, right lower panel).

Figure 2.

Time course of bilirubin-mediated effects on the number of viable colon cancer cells. Left panel depicts the number of viable HCT15 cells grown in medium alone (black bars) or in the presence of DMSO vehicle (white bars) or 50 μM bilirubin (gray bars) at the indicated time intervals. Bars reflect mean cell number (±SD) relative to untreated controls (n = 5). At right, the differential absorption spectrum of the culture medium from HCT15 cells incubated with bilirubin (50 μM, upper panel) or biliverdin (50 μM, lower panel) vs. an equivalent volume of DMSO vehicle obtained at 0, 12, 24 and 48 hr. *p < 0.0001 vs. control and vehicle treatment, **p < 0.0001 vs. control and vehicle treatment and vs. bilirubin treatment at 24 hr.

Mechanism of bilirubin-induced cell death

Cellular release of LDH is generally considered to be a hallmark of necrotic cell death.49, 50 To elucidate whether bilirubin cytotoxicity is due to induction of cellular necrosis, release of LDH from bilirubin-treated HCT15 cells was assessed at 24 and 48 hr. For these studies, HCT15 monolayers were incubated in the presence of bilirubin (50 μM) or DMSO vehicle, and the concentration of LDH in the medium was determined by colorimetric assay (Fig. 3). Maximal LDH release was measured at the end of the treatment period by inducing complete cell lysis with 10% Triton X-100. No significant differences in the level of LDH in the culture medium were observed between the treatment groups at either time point, indicating that bilirubin does not cause substantial cellular necrosis. In addition, the finding that neither bilirubin nor DMSO significantly alters maximal LDH activity following cell lysis indicates that, at the concentrations utilized, these compounds do not interfere with the assay. To determine whether bilirubin induces cellular apoptosis, we performed a TUNEL assay in HCT15 monolayers exposed to 50 μM bilirubin for 48 hr. FACS analysis demonstrated an increased number of BrdU-positive cells in bilirubin-treated vs. vehicle-treated monolayers (Fig. 4), indicating that bilirubin causes cellular DNA fragmentation, a characteristic terminal event in the apoptotic cascade. The lack of effect of bilirubin on the PI staining pattern of HCT15 cells provides additional evidence against bilirubin-mediated cell cycle arrest. We further speculate that the nonsignificant trend toward increased levels of LDH in the medium of bilirubin-treated HCT15 cells (Fig. 3) reflects enzyme release from late apoptotic cells (see below).

Figure 3.

LDH release from colon cancer cells treated with bilirubin. HCT15 cells were incubated in medium alone (Control), medium plus DMSO vehicle or medium plus 50 μM bilirubin (n = 5). LDH activity in the culture medium was quantified at 24 (gray bars) and 48 (white bars) hr and plotted relative to maximum LDH release from control cells treated with 10% Triton X-100 (black bars).

Figure 4.

TUNEL assay of bilirubin-treated colon cancer cells. HCT15 cell monolayers were treated with 50 μM bilirubin or DMSO vehicle for 48 hr and analyzed by TUNEL assay, with total cellular nucleic acids detected by PI staining. Cells incubated in the presence of DMSO (left upper panel) or bilirubin (left lower panel) but without BrdU treatment served as negative controls. The marked increase in FITC staining of cells treated with bilirubin (41% positivity, right lower panel) compared to vehicle control (4% positivity, right upper panel) is indicative of bilirubin-induced apoptosis.

In intact cells, phosphatidylserine is restricted to the inner hemileaflet of the plasma membrane, while apoptotic cells are unable to maintain this normal phospholipid asymmetry. Based on the affinity of annexin V for phosphatidylserine, bilirubin-induced apoptosis was further verified by staining cell monolayers with FITC-conjugated annexin V. Concomitant staining with PI was performed to assess plasma membrane integrity as cells in the earliest phases of apoptosis are able to exclude PI, while necrotic or late apoptotic cells become readily permeable to this fluorescent probe.51 The simultaneous measurement of annexin V–FITC and PI fluorescence intensity by FACS permits a distinction (albeit rough) to be made between viable cells (annexin-negative, PI-negative) and those undergoing early apoptosis (annexin-positive, PI-negative) or late apoptosis/necrosis (annexin-positive, PI-positive). HCT15 cell monolayers were incubated in the presence of DMSO vehicle, bilirubin (50 μM) or NS398 (100 μM), a COX-2-specific inhibitor which induces apoptosis in this cell line.52 Cells were harvested at 24 and 48 hr, stained with PI and FITC-conjugated annexin V and analyzed by flow cytometry. As displayed in Figure 5, treatment with either bilirubin or NS398 was associated with a significant increase in the percent of annexin V-positive cells at each time point. Both treatments also caused a time-dependent shift from annexin V positivity alone (right lower quadrant) to combined positivity for PI plus annexin V (right upper quadrant), consistent with progression from early to late apoptosis over the subsequent 24 hr of incubation. These findings, which are concordant with the TUNEL assay results (Fig. 4), indicate that bilirubin induces apoptosis in HCT15 colon cancer cells.

Figure 5.

Annexin V staining of colon cancer cells treated with bilirubin and NS398. HCT15 monolayers were incubated in the presence of DMSO vehicle, bilirubin (50 μM) or the COX-2 inhibitor NS398 (100 μM). Cells were harvested at 24 (upper panels) or 48 (lower panels) hr, stained with annexin V–FITC (horizontal axis) and analyzed by flow cytometry. PI staining, a marker of cell membrane integrity, is displayed on the vertical axis. The percent of total cells in each quadrant is indicated.

Caspases are cysteine proteases that mediate many of the cellular effects of apoptosis. Through several pathways, apoptotic stimuli result in cleavage and activation of effector caspases, of which caspase-3 is generally held to be the most pivotal. To provide additional verification of bilirubin-induced apoptosis, we examined whether bilirubin causes activation of caspase-3. For these studies, TNF-α, a known activator of the death receptor pathway in colon cancer cells,53, 54 was utilized as a positive control. HCT15 cell monolayers were incubated in the presence of bilirubin (50 μM), TNF-α (250 ng/ml) or the appropriate vehicle controls for 48 hr. An 8-fold increase in caspase-3 activation was observed in cell lysates from both bilirubin- and TNF-α-treated cells compared to the corresponding vehicle-treated controls (Fig. 6). Taken together, these data indicate that bilirubin decreases the viability of HCT15 adenocarcinoma cells through induction of apoptosis.

Figure 6.

Bilirubin and TNF-α cause activation of caspase-3 in colon cancer cells. HCT15 cell monolayers were incubated for 48 hr in the presence of PBS, TNF-α (250 ng/ml), DMSO vehicle or bilirubin (50 μM). Incubations with TNF-α and PBS vehicle were performed in the absence of serum, while treatment with bilirubin and DMSO was conducted in the presence of 20% FBS. Caspase-3 activity in cell lysates was assayed by cleavage of the specific substrate peptide DEVD, resulting in increased absorbance at 405 nm. Bars reflect the mean (±SD) of 5 experiments, with data corrected for intrinsic absorbance of the treatments. *p < 0.0001 vs. control.

Pathway of bilirubin-induced apoptosis

Caspase activation traditionally has been considered to be mediated via 2 principal routes: the death receptor (extrinsic) pathway and the mitochondrial (intrinsic) pathway.55 Activation of the extrinsic apoptotic pathway, which is triggered by ligation of death receptors (e.g., Fas) in response to an external ligand, is characterized by increased activity of initiator caspase-8.54 The intrinsic pathway is provoked by various extracellular or intracellular stimuli that ultimately cause release of cytochrome c from mitochondria, resulting in activation of caspase-9.56, 57 Induction of either pathway produces a cascade effect through subsequent activation of effector caspases.55, 56 We sought to distinguish whether bilirubin initiates apoptosis by way of the intrinsic or extrinsic pathway by examining for activation of caspase-8 vs. caspase-9. HCT15 cell monolayers were incubated in the presence of bilirubin (50 μM), TNF-α (250 ng/ml) or the corresponding vehicle control. The differential activation of caspases 8 and 9 in cell lysates was determined colorimetrically, employing an experimental design similar to that described above for caspase-3. As anticipated, lysates from HCT15 cells treated with the death receptor activator TNF-α exhibited a significant increase in caspase-8 activity at 12 hr (Fig. 7, left panel). In contrast, incubation of cell monolayers in the presence of bilirubin did not cause activation of caspase-8. Conversely, a caspase-9-specific substrate peptide was cleaved to a much greater extent by lysates from bilirubin-treated HCT15 cells compared to cells incubated for 48 hr in the presence of TNF-α or vehicle alone (Fig. 7, right panel). These data suggest that, in contradistinction to TNF-α, bilirubin induces apoptosis via the mitochondrial pathway.

Figure 7.

Differential activation of caspases 8 and 9 by bilirubin and TNF-α. HCT15 cell monolayers were treated as outlined in Figure 6, except that substrates specific for caspase-8 (left panel) and caspase-9 (right panel) were employed. Bars represent the mean (±SD) of 5 replicate experiments. *p < 0.0001 vs. control.

As cytochrome c is normally sequestered within mitochondria, increased levels in the cytoplasm are indicative of enhanced mitochondrial permeability, an early event in the intrinsic apoptotic pathway.57 We assessed for cytochrome c release in HCT15 cells by incubating monolayers in the presence or absence of bilirubin (50 μM), NS398 (100 μM) or DMSO vehicle. Cells were harvested at various time intervals, and the concentration of cytochrome c in the cell cytosol was determined by Western blotting. At 12 hr, an approximately 8-fold increase in the concentration of cytochrome c was detected in cytosol isolated from bilirubin-treated cells vs. DMSO-treated controls (Fig. 8), while NS398, a known stimulator of the mitochondrial pathway,52 caused a more modest 3-fold increase in cytochrome c levels at the same time point. The bilirubin-induced increase in cytosolic cytochrome c persisted to 24 hr of incubation. These findings provide further support that bilirubin triggers apoptosis in HCT15 cells via activation of the mitochondrial pathway.

Figure 8.

Western blot analysis of cytochrome c release. HCT15 monolayers were incubated in the presence of DMSO vehicle (D), 50 μM bilirubin (Br) or 100 μM NS398 (NS). Cells were harvested at 12 (left lanes) and 24 (right lanes) hr, and the cytosolic fraction (50 μg protein) was subjected to Western blotting with an antihuman cytochrome c primary antibody (upper panels). Corresponding densitometric results plotted relative to 12 hr DMSO-treated controls are displayed in the lower panel.

It is generally held that the intrinsic apoptotic pathway is initiated by a breach in mitochondrial integrity,57, 58 such that the usually impermeable inner mitochondrial membrane becomes permeable to the nonspecific passage of ions and small molecules, causing complete loss (depolarization) of the transmembrane potential. We employed the fluorescent dye JC-1, which exhibits potential-dependent accumulation in mitochondria, to determine the effect of bilirubin on mitochondrial permeability in intact cells. Following 12 hr exposure to bilirubin or DMSO vehicle, HCT15 cells were labeled with JC-1 and subjected to flow cytometry. A decrease in the ratio of red to green fluorescence was observed in bilirubin-treated cells vs. DMSO-treated controls (Fig. 9), indicative of mitochondrial depolarization.51, 59 Thus, using a variety of complementary techniques, our data indicate that bilirubin-induced apoptosis occurs through specific activation of the intrinsic mitochondrial pathway.

Figure 9.

Effect of bilirubin on mitochondrial membrane potential in intact cells. HCT15 monolayers were incubated for 12 hr in the presence of DMSO vehicle (upper panels) or 50 μM bilirubin (middle panels). Following labeling of the cells with JC-1, green (left panels) and red (right panels) fluorescence intensity was quantified by flow cytometry. Lower panels display the results of bivariate analysis of JC-1 fluorescence, with red and green fluorescence intensity plotted along the vertical and horizontal axes, respectively. Control experiments demonstrated no effect of bilirubin on JC-1 fluorescence in the absence of incubation.

Effect of bilirubin on the permeability of isolated mitochondria

Collapse of the mitochondrial transmembrane potential, typically mediated by the opening of the large-conductance PTP, appears to be an early and irreversible event in the extrinsic pathway of apoptosis.60 Based on our previous work demonstrating that bilirubin is capable of transporting protons across model and biologic membranes,35 we speculated that bilirubin might induce apoptosis by directly depolarizing cellular mitochondria and, hence, may not require activation of the PTP. To test this hypothesis, we examined the effect of bilirubin on the permeability of isolated rat liver mitochondria. Changes in mitochondrial volume were assessed by light scattering,44 with CCCP, a known mitochondrial uncoupler,61 employed as a positive control. As expected, addition of 20 μM CCCP to energized mitochondria produced an abrupt decrease in light scattering (Fig. 10a), indicative of large-amplitude swelling caused by induction of the mitochondrial pore transition.61 Additionally, pretreatment with 2 μM cyclosporin A, an inhibitor of the PTP,62 blocked CCCP-mediated mitochondrial depolarization (Fig. 10a).

Figure 10.

Detection of large-amplitude swelling in isolated rat liver mitochondria. The time-dependent decrease in light scattering (λ = 620 nm) of isolated rat liver mitochondria (0.2 mg protein/ml), reflective of large-amplitude swelling, was monitored continuously at room temperature. Mitochondria were energized with calcium chloride (50 μM) and succinate (10 mM) immediately prior to addition of depolarizing agents. (a) Addition (arrow) of CCCP (20 μM), but not DMSO vehicle, promptly induces mitochondrial swelling, a process that is inhibited by preincubation with 2 μM cyclosporin A (CsA + CCCP). (b) Bilirubin (BR, 20 μM) causes a more gradual increase in mitochondrial volume, a phenomenon that is unaffected by pretreatment with cyclosporin A (CsA + BR). Preincubation with CCCP (CCCP + BR) completely abolishes the effect of bilirubin. Similarly, addition of CCCP (arrow) following induction of large-amplitude swelling by bilirubin does not evoke further changes in mitochondrial volume (BR + CCCP). (c) Tracings were recorded following addition of 20 μM bilirubin in the presence (+mito) or absence (–mito) of mitochondria. (d) Bilirubin (0–50 μM) causes a dose-dependent increase in the rate and magnitude of mitochondrial swelling compared to an equivalent volume of DMSO vehicle.

Addition of bilirubin to energized rat liver mitochondria also induced an increase in large-amplitude swelling (Fig. 10b). However, the finding that cyclosporin A did not inhibit this effect suggests that bilirubin does not cause depolarization by activating the mitochondrial transition pore. The facts that bilirubin-mediated mitochondrial swelling was abolished by pretreatment with CCCP and, conversely, that the absorbance changes induced by CCCP were abrogated by pretreatment with bilirubin indicate that the observed effects reflect mitochondrial depolarization (Fig. 10b). This conclusion is further supported by the absence of time-dependent changes in absorbance when bilirubin was added to the incubation buffer in the absence of mitochondria (Fig. 10c). Finally, as shown in Figure 10d, bilirubin-mediated mitochondrial swelling was dose-dependent.

These results were validated by assessing for bilirubin-induced depolarization of isolated mitochondria utilizing rhodamine-123, a membrane-permeable dye that exhibits potential-dependent fluorescence.46, 63 Energized mitochondria sequester rhodamine-123, which manifests as a decrease in fluorescence intensity (Fig. 11, left panel). Addition of 20 μM CCCP resulted in rapid mitochondrial depolarization and a concomitant increase in rhodamine fluorescence. In contrast, treatment with 20 μM bilirubin produced an initial diminution, followed by a more gradual increase in rhodamine fluorescence intensity (Fig. 11, right panel), findings which are consistent with transient hyperpolarization preceded by sustained mitochondrial depolarization. That these data reflect true alterations in mitochondrial potential are validated by the finding that pretreatment of energized mitochondria with CCCP abolished the fluorescence changes induced by bilirubin (Fig. 11, right panel). Moreover, addition of BDT, which we previously showed is incapable of spontaneous transmembrane diffusion,35 did not cause the hyperpolarization phenomenon observed with bilirubin treatment. As bilirubin and BDT exhibit similar inner filter effects, these findings provide further support that the modulation of rhodamine fluorescence is due to bilirubin-specific induction of changes in mitochondrial potential. Taken together, these data indicate that bilirubin exerts a direct effect on mitochondrial permeability through a mechanism that does not appear to involve opening of the PTP.

Figure 11.

Depolarization of isolated mitochondria monitored by changes in rhodamine fluorescence. Isolated rat liver mitochondria (0.2 mg protein/ml) were suspended in buffer containing 5 μM rhodamine-123, and time-dependent changes in rhodamine fluorescence (excitation 488 nm, emission 525 nm) were continuously monitored at room temperature. Following addition of calcium chloride (50 μM), mitochondria were energized with succinate (10 mM, arrow), which causes a decrease in rhodamine fluorescence intensity (left panel). Addition of CCCP (20 μM, arrow), but not DMSO vehicle, produces an abrupt increase in fluorescence, indicative of mitochondrial depolarization. At right, 20 μM bilirubin (BR) first induces a transient decrease in rhodamine fluorescence, followed by a marked increase in fluorescence intensity, changes that are not observed with the DMSO vehicle. Pretreatment with CCCP abrogates bilirubin-induced fluorescence changes (CCCP + BR), while 20 μM BDT fails to cause mitochondrial hyperpolarization.


Our studies demonstrate that bilirubin, at physiologic concentrations (0–50 μM), reduces the viability of colon adenocarcinoma cell lines in vitro. Using several complementary techniques, including TUNEL assay, annexin V staining and measurement of caspase-3 activation, we have shown that bilirubin stimulates apoptosis in HCT15 colon adenocarcinoma cells. We have further demonstrated that, consistent with the findings of Rodrigues et al. in rat brain neurons,20, 64 bilirubin-induced apoptosis is mediated via the intrinsic (mitochondrial) pathway, as evidenced by cytochrome c release into the cytosol, cellular mitochondrial depolarization (detected by JC-1 staining) and specific activation of caspase-9. This effect appears to be specific to the unconjugated form of bilirubin as the metabolic precursor, biliverdin, and the stable conjugate, BDT, did not alter cell viability. These observations are consistent with the findings of Cowger et al.,65 who showed that 25 μM bilirubin, but not biliverdin, blocks cellular ATP generation by L-929 murine fibroblasts. As colon cancer cells express message for biliverdin reductase, the absence of a cytotoxic effect of biliverdin suggests that cellular biliverdin reductase activity is insufficient to produce enough bilirubin to induce apoptosis. In support of this hypothesis, we estimate from our absorption spectra (Fig. 2, right lower panel) that HCT15 cell monolayers incubated for 48 hr in the presence of 50 μM biliverdin generate approximately 5 μM bilirubin in the medium, a concentration which does not cause a significant level of cell death.

Investigators have previously shown that addition of albumin to the culture medium protects cells against bilirubin toxicity.13, 17 In this regard, it is notable that our studies demonstrate substantial cytotoxic effects of bilirubin on HCT15 cells at molar ratios of bilirubin:BSA that are well below unity. These findings are in line with data reported by Chuniaud et al.,24 who showed that half-maximal lysis of rat fibroblasts and astrocytes at 48 hr occurred at bilirubin:human serum albumin ratios of 0.75 and 0.85, respectively. While cellular apoptosis was not directly assessed, these authors also demonstrated impaired cellular mitochondrial activity at these bilirubin concentrations. In contrast, Rodrigues et al.20 detected only weak bilirubin effects on rat brain neurons at a bilirubin:BSA ratio of 0.5. These investigators performed the majority of analyses of bilirubin-induced apoptosis at a bilirubin:albumin molar ratio of 3,64 using concentrations at which bilirubin would be anticipated to precipitate from solution.20 Moreover, as opposed to bilirubin-mediated neuronal cell death, which is triggered by stimulation of the NMDA receptor66 and prevented by NMDA receptor antagonists,22 both glutamate and NMDA inhibitors reduce the viability of HT29 adenocarcinoma cells.67 These disparities in the cellular response may be a result of nonphysiologic concentrations of bilirubin utilized or, alternatively, may reflect distinct mechanisms of bilirubin action in different cell lines. In support of the latter hypothesis are reports of differing sensitivities of neural cells to bilirubin cytotoxicity.13, 14

To insure that bilirubin remained solubilized throughout the incubation period, cells were grown in the presence of 20% FBS. Due to the substantial affinity of BSA for bilirubin,68, 69, 70 even at the highest quantities of total bilirubin utilized in our studies (50 μM), the free bilirubin concentration in the medium is likely to be very low. Based on the significant cytotoxic effects of bilirubin observed at 25 μM concentrations (Fig. 1), it is estimated that approximately 7 × 109 bilirubin molecules per cell are required to induce cell death. If one assumes that only free bilirubin is available for cellular uptake, then this value is closer to 1 × 107. Notably, since unconjugated bilirubin in the intestine is not albumin-bound, free bilirubin concentrations within the bowel lumen are likely to be substantially higher than in the serum.

It is generally held that the extrinsic apoptotic pathway is initiated by a breach in mitochondrial integrity, resulting in the collapse of the mitochondrial transmembrane potential.60 Since the original observation by Day in 1954 that bilirubin inhibits the respiratory activity of rat brain homogenate,71 a number of studies have examined the effect of bilirubin on mitochondrial function. At the cellular level, Cowger et al.65 demonstrated that bilirubin, when added to the culture medium at concentrations ranging from 5–250 μM (in the absence of albumin), uncouples oxidative phosphorylation in L-929 cells with a potency over 16 times greater than biliverdin. Similarly, Chuniaud et al.24 showed that mitochondrial activity, as assessed by MTT metabolism, in cultured rat brain astrocytes and human fibroblasts is impaired by bilirubin and that this effect occurs at bilirubin:human serum albumin ratios ranging 0.5–1.5. With regard to isolated mitochondria, several investigators have demonstrated that bilirubin, at concentrations exceeding 50 μM, is capable of uncoupling oxidative phosphorylation72, 73, 74 and inducing mitochondrial swelling.73, 75 Stumpf et al.76 further showed that bilirubin, at concentrations of 12–24 μM, specifically induces increased inner mitochondrial membrane conductance without altering the proton motive force. These authors postulated that uncoupling observed at higher bilirubin concentrations was due to inhibition of dehydrogenases and electron transport. That bilirubin is capable of directly causing depolarization of isolated mitochondria, as has previously been proposed by Rodrigues et al.,64 is supported by our light-scattering and rhodamine fluorescence data. Our finding that bilirubin-induced large-amplitude swelling is not blocked by cyclosporin A, an established inhibitor of the mitochondrial PTP,58, 77 in conjunction with data from Rodrigues et al. indicating that bilirubin alters mitochondrial lipid polarity,23, 78 further suggests that bilirubin is capable of exerting a direct effect on mitochondrial membrane integrity.

The above findings also suggest a potential mechanism by which bilirubin induces apoptosis. We35 and others79 have shown that unconjugated bilirubin diffuses rapidly through model and native membrane vesicles and that this represents an important mechanism by which bilirubin is taken up into cells.37 Once inside the cytoplasm, bilirubin likely associates with mitochondrial membranes, as well as other organelles, as a result of high-affinity binding to membrane lipids.36, 74, 80 Because bilirubin traverses phospholipid bilayers as the uncharged diacid species,35 the physiologic bilirubin dianion must acquire 2 protons prior to diffusion across the mitochondrial membrane. These protons are subsequently released from the bilirubin diacid at the opposing bilayer hemileaflet.35 Consequently, we propose the following potential mechanisms by which bilirubin may directly alter mitochondrial permeability: (i) hyperpolarization of the inner mitochondrial membrane induced by bilirubin-mediated transduction of protons from the cytosol into the intercristal space or (ii) dissipation of the mitochondrial proton gradient as a consequence of bilirubin traversing both the inner and outer mitochondrial membranes. Support for the former hypothesis is derived from the observation that bilirubin induces a decrease in rhodamine fluorescence immediately prior to depolarization of isolated mitochondria (Fig. 11b), consistent with a transient increase in mitochondrial potential. We speculate that this initial mitochondrial hyperpolarization event reflects the transduction of protons from the cytosol into the intermembrane space by bilirubin as it diffuses across the outer mitochondrial membrane. Indeed, transient hyperpolarization has been proposed as a possible mechanism whereby the intrinsic apoptotic pathway is activated.81 This theory holds that the resultant decrease in pH within the intermembrane space causes protonation of weak acids, which freely diffuse into the matrix, producing an increase in mitochondrial osmolality and subsequent depolarization.57 Alternatively, since protons generated in the mitochondrial matrix are actively transported into the intermembrane space to establish an electrochemical potential, bilirubin might reasonably be expected to dissipate this pH gradient as it traverses both the inner and outer membranes.

Given the absence of an effect of the bilirubin precursor biliverdin and the bilirubin conjugate BDT on colon adenocarcinoma cell viability (Fig. 1), it appears that it is the unconjugated form of bilirubin that specifically modulates tumorigenesis. Bilirubin is secreted into bile as the 1-O-acylglucuronide conjugate,28, 82 which partially undergoes acyl migration in the gallbladder and intestine to form 2-, 3- and 4-O-acyl isomers.83, 84 Deconjugation of bilirubin 1-O-acylglucuronide, but not the other positional isomers,84 occurs within the intestinal lumen both by spontaneous nonenzymic hydrolysis29 and through the action of mucosal and bacterial β-glucuronidases30, 31 present primarily in the distal small bowel and colon, where luminal bilirubin levels are the highest.32, 85 Plasma unconjugated (but not conjugated) bilirubin also enters the gut via passive diffusion across the intestinal mucosa.33, 86, 87 While bacteria-mediated catabolism of luminal bilirubin to urobilinogens can occur,32, 88 the concentrations of unconjugated bilirubin utilized in our experiments are concordant with levels normally present within the bowel. Indeed, the average daily rate of bilirubin excretion in human stool has been measured at 60 nmol/kg body weight.33 Assuming an average stool volume in a 70 kg person of 100 ml/day, the normal fecal bilirubin concentration is estimated to be approximately 42 μM. Although a sizeable proportion of intraluminal bilirubin may be unavailable for cellular uptake because it is precipitated as insoluble calcium salts89 or bound to dietary fiber90 and lipid,91, 92 the concentration of unconjugated bilirubin in the colon under physiologic conditions is within the range in which we have demonstrated inhibitory effects on colon cancer cell viability.

Unconjugated hyperbilirubinemia (> 1 mg/dl, or 17 μM) occurs under conditions where heme breakdown is accelerated and/or hepatic bilirubin conjugating activity is impaired. While markedly elevated serum bilirubin levels (>20 mg/dl, or 340 μM) are associated with an increased risk of neurologic injury (kernicterus) in newborns, more modest elevations appear to be entirely innocuous. One example is Gilbert's syndrome, a common inherited defect in hepatic bilirubin metabolism characterized by serum levels of unconjugated bilirubin that are less than 4 times normal.93 This benign condition is caused by a polymorphism in the promoter TATA element of the gene encoding UGT1A1,94 resulting in decreased enzyme expression.95 Due to passive permeation from the plasma, daily bilirubin excretion in the stool rises to a mean of 130 nmol/kg in individuals with Gilbert's syndrome and to 300 nmol/kg in those with Crigler-Najjar (an inherited disorder characterized by a near total absence of hepatic UGT1A1 activity),33 corresponding to stool concentrations of 90 and 230 μM, respectively. We speculate that this may lead to a reduced susceptibility to colorectal carcinoma in affected individuals.

However, if bilirubin is able to enter all cells by passive diffusion,35, 37 why is it that individuals with these conditions do not manifest intestinal toxicity? We speculate that malignant cells are more susceptible to bilirubin-induced apoptosis because of a decreased ability to secrete and/or metabolize the bile pigment. In this regard, of the colon cancer cell lines tested, the more well-differentiated (e.g., CaCo2, SW480) exhibit greater resistance to bilirubin toxicity. These data are consistent with the findings of Notter and Kendig,14 who demonstrated that the N2AB-1 murine neuroblastoma cell line is less sensitive to bilirubin toxicity following drug-induced morphologic differentiation. Since cellular bilirubin levels are likely determined not only by extracellular bilirubin concentrations but also by intracellular binding, export and metabolism of bilirubin, we further propose that resistant cells possess more efficient protective mechanisms. Support for this hypothesis is derived from our finding that bilirubin-resistant HT29 cells express higher levels of UGT1A1 message compared to the more sensitive HCT15 cell line (Fig. 1, right panel). These results also are in line with prior studies demonstrating that HT29 cells exhibit higher UGT activity than bilirubin-susceptible HCT116 cells.96 Other investigators have shown increased levels of glutathione S-transferases (the principal intracellular bilirubin binding protein) in more well-differentiated CaCo2 cells.97 Notably, we observed no adverse effects on growth pattern, weight gain or food consumption in mice receiving as much as 1 mg oral bilirubin daily for 7 weeks (data not shown), supporting the benign nature of bilirubin with respect to the normal intestinal mucosa. While bilirubin has generally been regarded as little more than a metabolic by-product of heme degradation, epidemiologic analyses suggesting a protective effect of bilirubin against cancer-related death, when considered in conjunction with the in vitro data presented, support the hypothesis that bilirubin functions as an endogenous inhibitor of colon tumorigenesis.


S.D.Z. was supported by grant DK063954 from the NIH, an Ohio Division Supported Pilot Research Grant from the American Cancer Society and a Cancer Prevention Research Grant from the Cancer Research and Prevention Foundation. P.K. was the recipient of an American Liver Foundation Pamela Denmark Student Research Fellowship award. We gratefully acknowledge Ms. F. Yu, Dr. J. Johnson and Ms. E. Kopras for expert technical assistance.