Improved glycemic control with colesevelam treatment in patients with type 2 diabetes is not directly associated with changes in bile acid metabolism§


  • Gemma Brufau,

    Corresponding author
    1. Departments of Pediatrics , Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    • Center for Liver, Digestive and Metabolic Diseases, Department of Pediatrics, University Medical Center Groningen, Hanzeplein 1, 9713 EZ Groningen, The Netherlands
    Search for more papers by this author
    • fax: +31-503611746

  • Frans Stellaard,

    1. Departments of Laboratory Medicine, Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Kris Prado,

    1. Kinemed, Inc., Emeryville, CA
    Search for more papers by this author
  • Vincent W. Bloks,

    1. Departments of Pediatrics , Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Elles Jonkers,

    1. Departments of Laboratory Medicine, Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Renze Boverhof,

    1. Departments of Laboratory Medicine, Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Folkert Kuipers,

    1. Departments of Pediatrics , Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    2. Departments of Laboratory Medicine, Center for Liver, Digestive and Metabolic Diseases, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Elizabeth J. Murphy

    1. Kinemed, Inc., Emeryville, CA
    2. Department of Medicine, University of California San Francisco, San Francisco, CA
    Search for more papers by this author

  • Potential conflict of interest: Kinemed, Inc., represented in this study by Elizabeth Murphy, is the recipient of a research support from Daiichi Sankyo Inc., in the form of an investigator initiated trial grant to fund this trial. University of Groningen, represented by Folkert Kuipers, received a research grant supported by Daiichi Sankyo to study the effects of colesevelam on bile acid metabolism in mice. Both Dr. Murphy and Kuipers have acted as consultants to Daiichi Sankyo within the past year. Dr. Murphy is no longer a consultant.

  • The goal of this mechanistic study is to evaluate the interaction between differences in glucose control and differences in bile acid kinetics. Data reported here are focused on the underlying physiologic effects of the drug on bile acid kinetics and not on clinical outcomes or treatment for a medical condition.

  • §

    Trial is registered at “Effects of colesevelam on bile acid kinetics”


Bile acids (BAs) are essential for fat absorption and appear to modulate glucose and energy metabolism. Colesevelam, a BA sequestrant, improves glycemic control in type 2 diabetes mellitus (T2DM). We aimed to characterize the alterations in BA metabolism associated with T2DM and colesevelam treatment and to establish whether metabolic consequences of T2DM and colesevelam are related to changes in BA metabolism. Male subjects with T2DM (n = 16) and controls (n = 12) were matched for age and body mass index. BA pool sizes and synthesis/input rates were determined before and after 2 and 8 weeks of colesevelam treatment. T2DM subjects had higher cholic acid (CA) synthesis rate, higher deoxycholic acid (DCA) input rate, and enlarged DCA pool size. Colesevelam resulted in a preferential increase in CA synthesis in both groups. CA pool size was increased whereas chenodeoxycholic acid and DCA pool sizes were decreased upon treatment. Fasting and postprandial fibroblast growth factor 19 (FGF19) levels did not differ between controls and diabetics, but were decreased by treatment in both groups. Colesevelam treatment reduced hemoglobin A1C by 0.7% (P < 0.01) in diabetics. Yet, no relationships between BA kinetic parameters and changes in glucose metabolism were found in T2DM or with colesevelam treatment. Conclusion: Our results reveal significant changes in BA metabolism in T2DM, particularly affecting CA and DCA. Colesevelam treatment reduced FGF19 signaling associated with increased BA synthesis, particularly of CA, and resulted in a more hydrophilic BA pool without altering total BA pool size. However, these changes could not be related to the improved glycemic control in T2DM. (HEPATOLOGY 2010;)

Bile acids (BAs) are amphipathic molecules synthesized from cholesterol in the liver. In addition to the well-established role of BAs in dietary lipid absorption and cholesterol homeostasis, they also act as signaling molecules. BAs activate the nuclear receptor farnesoid X receptor (FXR), act as ligands for the G-protein–coupled bile acid receptor TGR5 and activate mitogen-activated protein kinase pathways.1 Through activation of these diverse signaling pathways, BAs regulate their own enterohepatic circulation at two main sites: in the small intestine via FXR–fibroblast growth factor 19 (FGF19) signaling and in the liver via FXR–short heterodimer partner signaling. Via these pathways, BAs are also thought to contribute to the maintenance of triglyceride, cholesterol, energy, and glucose homeostasis.1, 2

BA sequestrants form nonabsorbable complexes with BAs in the gastrointestinal tract, thereby preventing their ileal reabsorption and promoting fecal excretion. BA synthesis is stimulated at the expense of plasma low-density lipoprotein (LDL) cholesterol concentrations. BA sequestrants have been used for more than 30 years in the treatment of dyslipidemias, by reducing LDL cholesterol by 9%-28%, increasing triglycerides 1%-28%, and increasing high-density lipoprotein (HDL) cholesterol 0%-9%.3 More recently, BA sequestrants have also been shown to be effective in improving glycemic control in patients with type 2 diabetes mellitus (T2DM).4 It has been speculated that the mechanism by which BA sequestrants improve glycemic control in T2DM is through modifications in BA metabolism2; however, this has yet to be proven. Moreover, it is not known whether alterations in BA metabolism seen with T2DM may contribute to the metabolic derangements associated with the disease. Several previous studies have addressed BA kinetics in diabetes,5-9 but results have been inconsistent. These older studies tended to have heterogeneous subject populations, including patients with type 1 and patients with type 2 diabetes, and analytical techniques available at the time limited simultaneous quantification of all relevant BA parameters in a single study. Therefore, the first goal of this study was to clearly delineate the changes in BA metabolism associated with T2DM. The second objective was to establish the effects of colesevelam (a BA sequestrant) on BA metabolism and to relate those effects to improvements in glucose and lipid metabolism. For this purpose, we quantified kinetic parameters of cholic acid (CA), chenodeoxycholic acid (CDCA), and deoxycholic acid (DCA) in male T2DM subjects and age-matched and body mass index (BMI)-matched control subjects before and after 2 and 8 weeks of colesevelam treatment.


BA, bile acid; BMI, body mass index; 13C-DCA, 24-13C-deoxycholic acid; CA, cholic acid; CDCA, chenodeoxycholic acid; D4-CA, 2,2,4,4-2H4-cholic acid; D4-CDCA, 2,2,4,4-2H4-chenodeoxycholic acid; DCA, deoxycholic acid; FGF19, fibroblast growth factor 19; FTR, fractional turnover rate; FXR, farnesoid X receptor; GLP-1, glucagon-like peptide 1; HbA1C, hemoglobin A1C; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; LDL, low-density lipoprotein; LXR, liver X receptor; MS, mass spectrometry; OGTT, oral glucose tolerance test; SHP, short heterodimer partner; TG, triglyceride; T2DM, type 2 diabetes mellitus.

Patients and Methods


The protocol was approved by the RCRC Institutional Review Board (Austin, TX) and written informed consent was obtained for all procedures. Studies were performed at Diabetes and Glandular Research Associates (San Antonio, TX) and Clinical Pharmacology of Miami (Miami, FL). Male subjects with T2DM (n = 16) and nondiabetic healthy control subjects (n = 12), age 40-60 years and with BMI of 25-35 kg/m2, were recruited. Subjects with fasting triglycerides (TGs) >5.65 mM, HDL cholesterol < 1.55 mM, blood pressure ≥145/95, or a history of autonomic neuropathy, renal insufficiency, or gastroparesis were excluded. To minimize the presence of insulin resistance in the healthy control group, subjects with impaired fasting glucose (>5.6 mM), impaired glucose tolerance (>7.8 mM), elevated fasting insulin (>17.0 μIU/mL), TG ≥1.7 mM, or HDL cholesterol >1.0 mM were excluded. Subjects with diabetes who had been treated with insulin within 6 months or with a thiazolidinedione at any time were excluded. Goal hemoglobin A1C (HbA1C) at enrollment was 6.7%-10.0%. Diabetes was treated with diet alone in nine subjects and with glipizide in seven subjects. No subject in either group was on a lipid-lowering therapy within 3 months prior to enrollment. Baseline characteristics are shown in Table 1. Four subjects in the T2DM group withdrew prior to completion of the study due to personal matters.

Table 1. Baseline Characteristics of Healthy and T2DM Subjects
  1. Data are given as median (interquartile range).

Age, years (range)49 (44-58)52 (40-59)
Race, n (%)  
 White9 (75%)14 (86%)
 Black2 (17%)2 (14%)
 Native American1 (8%)0 (0%)
Ethnic group, n (%)  
 Non-Hispanic8 (67%)2 (13%)
 Hispanic4 (33%)14 (87%)
Body weight (kg)85.7 (72.7-106.1)85.6 (74.5-111.4)
BMI (kg/m2)28.9 (25.2-32.0)30.6 (25.2-35.5)
Fat-free mass (%)73.7 (65.3-78.0)70.1 (64.0-84.6)

Study Design.

Subjects were admitted to the research unit for a study at baseline and after 2 and 8 weeks of drug treatment (Supporting Information Fig. 1). Subjects were given an ad libitum standardized low-fat evening meal (≤30% fat). Two hours later, a single baseline blood sample was obtained for BA kinetics measurement and postprandial FGF19 level. Subjects then received an oral BA solution with 50 mg each of 2,2,4,4-2H4-CA, 2,2,4,4-2H4-CDCA, and 24-13C-DCA, dissolved in 150 mL of 1% sodium bicarbonate. Subsequently, blood samples were obtained 1.5-2.5 hours after meals at ∼13, 17.5, 24, 37, 46, 63, 70, and 91 hours after BA administration. On day 2, a single fasting blood sample was obtained for lipid profile, HbA1C, and FGF19 levels. Body composition was determined by bioelectrical impedance using a Tanita body composition analyzer. On day 4, a 75-g oral glucose tolerance test (OGTT) was performed with blood collections for glucose and insulin at 0, 1, and 2 hours.

Drug Treatment.

After the baseline study, subjects received six tablets of colesevelam (Daiichi Sankyo, Inc.) daily (3.75 g/day) for 8 weeks: three with lunch and three with dinner. The one subject taking a dinnertime sulfonylurea received all six tablets with lunch. During inpatient admissions, subjects received all six tablets with the morning meal on day 1 to avoid interference with the BA kinetic analyses. Pill counts were performed at week 2 and 8 to assess compliance.

Bile Acid Kinetic Studies.

The BA kinetic parameters assessed were pool sizes, fractional turnover rates (FTR), and synthesis rates/input rates for the three major BAs (CA, CDCA, and DCA). The kinetic model is a one-compartment model with synthesis (input) and turnover as equal influx and efflux. Isotope enrichments were measured in plasma samples collected over time after an oral bolus of labeled BAs. Conjugated BAs were extracted,10 deconjugated, derivatized, and analyzed by gas chromatography/mass spectrometry (GC/MS) as described.11 Linear regression of log enrichment, plotted against time, gives the pool size (y-intercept) and the FTR (slope) for each BA. Synthesis rate was obtained by multiplying the pool size by FTR.11 The fraction of CA turnover that is transferred to the DCA pool is calculated by dividing the DCA input rate by CA synthesis.

Unconjugated CA concentrations were measured in postprandial plasma by liquid chromatography tandem MS (LC-MS/MS) as a marker of bacterial overgrowth.12 Total plasma BA concentrations were assessed using an enzymatic fluorometric assay.13


Insulin was measured using a two-site immunoenzymometric assay; lipid profile, including direct LDL cholesterol, was assayed using an automated chemistry-immunoanalyzer. Insulin resistance was assessed using the homeostasis model assessment of insulin resistance (HOMA-IR) and the Matsuda-DeFronzo insulin sensitivity index (ISOGTT). A sandwich enzyme-linked immunosorbent assay kit was used for detection of FGF19 in serum (R&D Systems, Minneapolis, MN).

Statistical Analysis.

All subjects who completed the first inpatient stay were included in the analysis. Baseline differences between patients with diabetes and control patients were evaluated using the Mann-Whitney U test. Because of missing data points due to subject drop out, mixed effects logistic regression models were used to test differences. Condition (controls, T2DM), time (baseline, 2-week, 8-week), and the condition by time interaction were fixed. Subject effect was considered random. Because these models assume a normal distribution of residuals, bootstrapping was used to confirm significant findings.

Pooled within-group Pearson correlations were used to look at associations across both groups that are not affected by differences in group means. Pearson correlations were also done within each group. Significance was set at P < 0.05.


Subject Characteristics.

Subjects with diabetes and control subjects did not differ in age, weight, BMI, or percentage fat-free mass (Table 1). Subjects with diabetes had higher fasting glucose, fasting insulin, HbA1C, fructosamine, and 2-hour OGTT glucose (Table 2). They were also significantly more insulin resistant as assessed by both HOMA-IR and ISOGTT. Total cholesterol and LDL cholesterol concentrations were not significantly different between groups; however, the diabetic subjects had higher TG and lower HDL cholesterol levels.

Table 2. Changes in Biochemical Parameters from Baseline to Week 8 of Colesevelam Treatment in Healthy and in T2DM Subjects
Parameter BaselineWeek 2 DifferenceWeek 8 Difference
  1. Data are given as median (interquartile range).

  2. *P <0.05, **P < 0.01, and ***P <0.001 versus baseline.

  3. #P <0.05, ##P< 0.01, and ###P < 0.001 versus control group.

Fasting glucose (mM)Control5.1 (4.2-5.7)−0.2**0
 T2DM8.7 (6.3-15.5)# # #−1.3***−1.1***
Fasting insulin (mIU/L)Control7.3 (5.1-22.2)−0.02+5.37**
 T2DM13.2 (5.4-32.9)# #+4.59−2.84
HbA1C (%)Control5.5 (5.1-6.1)NA0.0
 T2DM8.6 (6.3-10.6)# # #NA−0.7***
Fructosamine (μmol/L)Control184 (95-233)−4+2
 T2DM328 (245-504)# # #−16−31**
2-hour OGTT (mM)Control5.0 (2.2-7.3)+0.2−0.5
 T2DM16.6 (9.7-25.6)# # #−2.7−1.7
HOMA-IRControl1.6 (1.1-4.8)−0.1+1.2*
 T2DM6.2 (2.1-16.6)# # #+0.5−2.1***
ISOGTTControl5.3 (2.7-10.3)+1.7*+0.2
 T2DM2.5 (1.1-6.0)# # #+0.5+0.3
Total cholesterol (mM)Control4.8 (3.3-5.8)−0.1−0.3*
 T2DM5.0 (3.7-6.4)−0.4**−0.0
LDL cholesterol (mM)Control2.7 (1.8-3.9)−0.2−0.3*
 T2DM3.3 (2.3-7.2)−0.7***−0.3
HDL cholesterol (mM)Control1.3 (0.7-1.5)+0.0+0.0
 T2DM1.0 (0.6-1.3)#0.0+0.0
Triglycerides (mM)Control1.2 (0.5-2.6)+0.1+0.0
 T2DM2.7 (1.1-5.6)# # #+1.1***+0.4*

Effect of Colesevelam Treatment on Glucose and Lipids.

Medication compliance, based on pill counts, was 99% (range: 83%-101%). Six of 24 subjects (25%) reported gastrointestinal adverse events (constipation, gas, nausea, diarrhea, abdominal cramping) during colesevelam administration but none necessitated discontinuation of treatment.

In patients with T2DM, colesevelam treatment reduced fasting glucose at 2 and 8 weeks (1.3 mM and 1.1 mM, respectively; P < 0.001) and at 8 weeks reduced HbA1C by 0.7% (P < 0.001) and fructosamine by 31 μM (P < 0.01; Table 2). Insulin resistance as measured by the HOMA-IR was reduced at 8 weeks in patients with T2DM. No significant changes were seen in 2-hour OGTT glucose or ISOGTT. A significant increase in fasting insulin was seen at 8 weeks in the control group, resulting in an increase in HOMA-IR, and thus fasting insulin was no longer significantly higher in the diabetic group at the end of the study. All other glucose-related parameters were higher in the diabetic group throughout the study.

There was an approximately 10% reduction in LDL cholesterol in both groups at 8 weeks of treatment. This was significant in the T2DM group at 2 weeks and in the control group at 8 weeks (Table 2). Reductions in total cholesterol were more modest and again were significant in the T2DM group at 2 weeks and in the control group at 8 weeks. TG levels were significantly increased after 2 (+1.1 mM, P < 0.001) and 8 weeks (+0.4 mM, P < 0.05) in T2DM. No changes were seen in HDL cholesterol levels.

BA Kinetics in T2DM.

The absolute DCA pool size was 33% larger in patients with T2DM than in control patients (P < 0.01; Fig. 1C, week 0). This increase was counterbalanced by a nonsignificant decrease in CDCA pool size (Fig. 1B), resulting in no changes in total pool size (Fig. 1D). No differences were seen for CA pool size (Fig. 1A). These differences resulted in significant change in the percent contribution of individual BAs to the total pool size in patients with T2DM, with a higher percent from DCA (37% versus 28% in controls; P < 0.05) and a lower percent from CDCA (32% versus 42% in controls, P < 0.001) (Fig. 1E). CA synthesis was 40% higher in patients with T2DM (P < 0.05) than in control patients, but no differences in CDCA synthesis were seen (Fig. 2). In line with the increase in CA synthesis, the DCA input rate was 54% higher in the diabetic group than in the control group (P < 0.01) at baseline (Fig. 2D). However, the fraction of CA turnover that was transferred to the DCA pool was not different between groups (Fig. 2E). No differences in FTR were seen for any of the BAs studied.

Figure 1.

Bile acid pool size in control (black circles) and T2DM (white squares) subjects at baseline and after treatment with colesevelam. (A) CA pool, (B) CDCA pool, (C) DCA pool size, (D) total pool, (E) relative pool, and (F) hydrophobicity marker. Values are estimates ± 95% confidence interval. In (E), white bars represent CA, grey bars are CDCA, and black bars are DCA. *P < 0.05 and **P < 0.001 versus baseline, #P < 0.05 and ##P < 0.001 versus control.

Figure 2.

Bile acid synthesis in control (black circles) and in T2DM (white squares) subjects at baseline and treatment with colesevelam. (A) CA synthesis, (B) CDCA synthesis, (C) total BA synthesis, (D) DCA input rate, and (E) fraction of CA transferred to the DCA pool. Values are estimates ± 95% confidence interval. *P < 0.05 and **P < 0.001 versus baseline, #P < 0.05 and ##P < 0.001 versus control.

Effects of Colesevelam Treatment on BA Kinetics.

As early as 2 weeks, colesevelam treatment had significant effects on BA metabolism. CA pool size was approximately doubled after 2 weeks of treatment in both groups (P < 0.001; Fig. 1A). In contrast, CDCA and DCA pool sizes were significantly decreased in both groups (Fig. 1B,C). This decrease was significant for the T2DM group only at 2 weeks, and at 8 weeks values had begun to return to baseline. In the control group, values remained low. Given the increased DCA pool size at baseline in the T2DM group, these changes resulted in even greater differences in DCA pool size at 8 weeks (e.g., four-fold higher levels in patients with T2DM). The differences in individual BA pool sizes balanced each other such that no differences in total BA pool size were seen with treatment in diabetic subjects at any point. In controls, there was a trend toward a slightly lower total pool size, which was significant at 2 weeks (Fig. 1D).

Figure 2E shows the relative contribution to the total pool of the major bile salt species. Comparing treatment versus baseline, the percentage of CA in both groups increased from approximately 30% at baseline to 65% after treatment (P < 0.001). The relative contribution of CDCA to the BA pool size decreased in controls from ∼40% to ∼20% and from ∼30% to ∼20% in patients with T2DM (P < 0.001). The contribution of DCA to the total BA pool also decreased after colesevelam treatment from ∼40% to ∼20% in patients with T2DM and from ∼30% to ∼10% in controls (P < 0.001).

The ratio of CA pool size to the sum of CDCA and DCA pool sizes was used as a marker of the hydrophobicity of the BA pool. At baseline, there was no difference in this marker between the groups; however, colesevelam treatment increased this ratio five-fold in both groups (Fig. 1F), indicating a substantial decrease in the hydrophobicity of the BA pool.

Colesevelam treatment doubled CA synthesis in both groups (Fig. 2A). CDCA and total BA synthesis were also significantly increased in both groups (Fig. 2B,C). DCA input, which was already lower in the control group at baseline, decreased further with colesevelam treatment leading to larger differences between the two groups at 8 weeks (Fig. 2D). At 2 weeks, there was a significant reduction in the fraction of CA turnover due to DCA production in both groups but in the T2DM group, this returned to normal at 8 weeks, whereas it remained low in the control group (Fig. 2E).

Despite the significant changes seen in CA pool size and synthesis, there were no consistent changes in CA FTR. However, for CDCA and DCA, the combination of large decreases in pool sizes and increases in synthesis/input rates were associated with highly significant three-fold increases in FTR in both groups for both BAs (Supporting Information Table 1).

No differences were found between the diet-treated and sulfonylurea-treated diabetic subjects with respect to BA kinetic measurements (Supporting Information Table 1).

FGF19 and Plasma BAs.

At baseline, no differences in fasting or postprandial FGF19 levels were seen between groups. Colesevelam treatment decreased fasting and postprandial FGF19 levels in both groups (P < 0.001; Fig. 3). Strong negative correlations were seen between fasting FGF19 levels and BA synthesis (Table 3) only in controls at baseline. No significant correlations were seen with postprandial FGF19 levels or between BA synthesis and FGF19 levels in patients with T2DM. During treatment with colesevelam, no significant correlations were seen (data not shown).

Figure 3.

FGF-19 and plasma BA concentrations in control (black circles) and in T2DM (white squares) subjects at baseline and after treatment with colesevelam. (A) Fasting FGF19, (B) fed FGF19, (C) fasting BA, and (D) fed BA. Values are estimates ± 95% confidence interval. *P < 0.05 and **P < 0.001 versus baseline, #P < 0.05 and ##P < 0.001 versus control.

Table 3. Correlation Coefficients (r) Between BA Kinetic Parameters and FGF19 Levels at Baseline
ParameterFasting FGF19Fed FGF19
  1. *P <0.05, **P <0.01, ***P <0.001.

CA synthesis−0.76***0.25−0.57−0.13
CA FTR−0.57*−0.04−0.38−0.13
CDCA synthesis−0.67*0.03−0.49−0.6
CDCA FTR−0.53***−0.36*−0.34−0.12
DCA FTR−0.52***0.11−0.340.02
Total synthesis−0.73***0.17−0.54−0.11

At baseline, plasma BA concentrations were lower in the fasting compared to the fed state (in controls, 9.7 μM versus 14.7 μM and in T2DM, 8.1 μM versus 13.2 μM; P < 0.05), and there were no significant differences between groups (Fig. 3). Colesevelam resulted in a significant reduction in fed BA concentrations in both groups (Fig. 3D) and in fasting BA levels in controls (Fig. 3C). No reduction in fasting BA was seen in T2DM with colesevelam treatment.

Fed BA levels were negatively correlated with BMI for the entire cohort at baseline (R = −0.50, P = 0.007) and after 2 weeks of colesevelam treatment (R = −0.66, P < 0.0001). In the control group, a significant negative correlation was also noted between TSH (thyroid stimulating hormone) and fed (R = −0.87, P = 0.005) and fasting (R = −0.72, P = 0.04) plasma BA concentrations. No such correlation was observed in the T2DM group (R = 0.10 and 0.26, respectively; nonsignificant).

No significant differences were seen in postprandial plasma levels of unconjugated CA (as a marker of bacterial overgrowth) between controls and T2DM (0.59 μM [0.18-1.28] versus 1.08 μM [0.23-2.27], respectively).

Correlations Between BA Kinetic and Metabolic Parameters.

No significant correlations were seen at baseline or after treatment in either group between any of the BA kinetic parameters, the hydrophobicity marker, or plasma BA concentrations and glucose/insulin levels, markers of insulin resistance, or markers of glycemic control. In addition, despite significant changes in both BA and glucose parameters after colesevelam treatment in patients with T2DM, no correlations were seen between changes in these parameters.

Significant correlations were observed between plasma TG concentrations and CA synthesis and total BA synthesis throughout the study (Fig. 4). Moreover, changes in absolute TG levels from week 0-8 correlated with week 0-8 changes in CA synthesis (r = 0.59, P < 0.01), CDCA synthesis (r = 0.47, P < 0.05), and total BA synthesis (r = 0.47, P < 0.05) suggesting a relationship between these parameters. Correlations were substantially stronger in patients with T2DM than in the controls subjects (data not shown). No correlations were observed between TG levels and DCA input rates. However, significant correlations were seen at baseline between TG levels and DCA FTR (r = 0.45, P < 0.01) and CDCA FTR (r = 0.42, P < 0.0001).

Figure 4.

Correlations between TG levels and CA and total BA synthesis in control (black circles) and T2DM (white squares) subjects.

No correlations were found between BA parameters and LDL cholesterol, HDL cholesterol, or total cholesterol.


New insights into the important role of BA in signaling and metabolic control have renewed interest in these molecules to better understand basic physiology and their potential as therapeutic targets. In view of their broad role in control of cholesterol, TG, and glucose metabolism,2 we studied whether altered BA metabolism contributes to the pathogenesis of diabetes and may be a target for treatment of the disease. Supporting this hypothesis, treatment with BA sequestrants improves lipid profiles and glycemic control in patients with T2DM.4 We present data showing significantly altered BA metabolism in T2DM and further modulation of BA metabolism during treatment with colesevelam, a BA sequestrant. However, we were unable to unearth the hypothesized direct links between changes in glucose metabolism and changes in BA metabolism, suggesting that the role of BA sequestrants in glucose lowering may be via alternate mechanisms.

Using a state-of-the-art stable isotope dilution technique, we provide the first description of a complete quantification of BA kinetics in subjects with T2DM. The comparison group was controlled for confounding factors known to alter BA metabolism (e.g., age, sex, obesity)14-16 and was more insulin-sensitive than the T2DM group despite similar BMI. We found that T2DM is associated with significant alterations in CA and DCA metabolism. Subjects with T2DM had an increased CA synthesis rate, an increased DCA input rate, a correspondingly larger DCA pool size, and a trend toward a smaller CDCA pool size. Total pool size was unchanged but relative contributions of individual BA differed with a significant increase in percentage from DCA and decrease in percentage from CDCA. Previous studies5-9 have also shown altered BA metabolism with diabetes. However, these studies were restricted to measures of relative BA composition alone or total BA synthesis without data on individual BA species and often had mixed populations with both type 1 and type 2 diabetes, which made interpretation difficult and results inconsistent.5-9 For example, an increase in percentage DCA was observed in one study8 but not in others.6, 9 Similarly, a decrease in percent contribution by CDCA was seen in some9 but not other studies.6, 8

Effects of colesevelam on BA metabolism were profound. Colesevelam forms nonabsorbable complexes with BAs in the gastrointestinal tract, stimulating their fecal excretion and removal from the enterohepatic circulation (see review by Goldfine4). It has thus been speculated that colesevelam treatment would result in a decrease in the total BA pool size. However, we found that colesevelam treatment did not change total BA pool size in either diabetic or healthy subjects.

Of specific interest, colesevelam treatment resulted in profound changes in the composition of the BA pool. CA pool size (the most hydrophilic BA in humans) was, surprisingly, increased after colesevelam treatment, whereas DCA and CDCA pool sizes were decreased, corresponding with changes previously seen with cholestyramine.17 DCA, which was increased with T2DM, is a secondary BA formed by bacterial deconjugation and 7α-dehydroxylation of CA in the proximal colon and has long been implicated in the pathogenesis of gallstones,18 a disease highly prevalent in patients with T2DM. A number of studies,18, 19 but not all,20 shows that the percentage of DCA in bile is higher in patients with gallstone disease, which has been attributed to bacterial overgrowth in the small intestine. This will result in increased conversion of CA to DCA and in enhanced colonic DCA absorption due to reduced colonic motility.21 We measured the concentration of unconjugated CA in postprandial plasma as a marker of bacterial overgrowth12 and found no differences between groups, suggesting the increase in DCA pool size found in patients with T2DM is not explained by bacterial overgrowth. Although patients with autonomic neuropathy and gastroparesis were excluded from the study, without a direct measure of intestinal transit time, we cannot definitively state that a longer transit time in T2DM was not involved. Nonetheless, the fact that colesevelam treatment reduced the hydrophobicity of the BA pool suggests that BA sequestrants may decrease risk of gallstone disease. In support of this, cholestyramine treatment in hamsters fed a lithogenic diet reduced the incidence of gallstones to 0% from 100%.22

FGF19 expression is induced in an FXR-dependent manner in ileal epithelial cells in response to BA release into the intestinal lumen upon feeding. The profound decreases in CDCA and DCA but not CA pool size observed with colesevelam suggest that colesevelam binds to CDCA and DCA to a greater extent than CA in the intestine, in agreement with in vitro binding studies.23 Colesevelam thus results in a smaller influx of the most potent FXR ligands,24 CDCA and DCA, into ileal enterocytes. This presumably results in less activation of FXR, leading to the large decrease in plasma FGF19 levels observed. FGF19 circulates through the blood to act on hepatocytes to reduce BA synthesis2 by suppressing the expression of the rate-limiting enzyme cholesterol 7α-hydroxylase (cytochrome P450 7A1 [CYP7A1]).25 Although there is much data linking FGF15 (the murine ortholog of FGF19) to BA synthesis in rodents, human data is limited. Using a surrogate marker of BA synthesis in healthy controls, an inverse relationship was seen with FGF19 levels during a prolonged fast and a relationship was suggested during feeding.26 Our data in controls support a negative effect of fasting FGF19 levels on CA synthesis, CDCA synthesis, and total BA synthesis in humans. Correlations with fed FGF19 levels, while present, were not statistically significant. A single postprandial time point was used and more frequent sampling may be required to better represent the postprandial FGF19 peak. The lack of correlation in T2DM, where BA synthesis rates are increased, is intriguing. Although this study does not allow for a mechanistic explanation, one could hypothesize that in T2DM there is either a disruption in FGF19 signaling or a factor driving BA synthesis that overwhelms the signal from FGF19.

Colesevelam treatment resulted in a significant increase in BA synthesis as early as 2 weeks after treatment, with a preferential increase in CA synthesis. This together with the low affinity of colesevelam for CA helps explain the increased CA pool size. It is intriguing that the changes in BA synthesis caused by colesevelam treatment were mainly directed to increase CA rather than CDCA synthesis especially in the control group. Several lines of evidence suggest that enzymes involved in BA metabolism (e.g., CYP7A1 and CYP8B1 [the enzyme that controls the CA:CDCA ratio]) are differentially regulated.2, 27 Intestine-specific FXR−/− mice have increased CA synthesis with no changes in CDCA synthesis, suggesting a role for FGF15/FGF19 in regulation of CYP8B1.28 However, the loss of correlation between FGF19 levels and BA synthesis seen here after colesevelam treatment suggests other factors may be more important regulators of BA metabolism under these conditions.

Treatment with colesevelam for 8 weeks had the anticipated effects on glucose and lipid levels with a 0.7% reduction in HbA1C and a 1.1 mM reduction in fasting glucose in T2DM and a ≈10% reduction in LDL cholesterol in both groups. TG levels were significantly increased in T2DM. We wondered whether metabolic changes observed in T2DM before and after colesevelam were related to the observed changes in BA metabolism. We found no correlation between parameters of BA metabolism and glucose metabolism, suggesting that the changes in glucose parameters are not directly BA-mediated. One possible explanation could be that colesevelam interferes with glucose absorption. However, data from mouse studies suggest that the glucose-lowering effect of colesevelam is not due to impaired glucose absorption (unpublished data). Recently, several studies have shown an association between the glucose-lowering effects of BA sequestrants and increased glucagon-like peptide 1 (GLP-1) levels in patients with T2DM.29-32 It has been shown that colesevelam treatment resulted in decreased plasma glucose levels and increased GLP-1 release in patients with T2DM29 and in a rat model of diet-induced obesity.30 Similarly, Zucker diabetic rats showed normalized plasma glucose and HbA1C levels after cholestyramine treatment, and these effects were accompanied with an increase GLP-1 release after glucose stimulation.31 Accordingly, the FXR–LXR–short heterodimer partner pathway may not be involved in the glucose-lowering effects of BA sequestrants. A potential mechanism may be that increased BA concentration in the intestine activates TGR5, resulting in the release of incretins such as GLP-1.33 Future studies are needed to gain more insight into these potential mechanisms.

Consistent with previous studies,34, 35 TG levels were positively correlated with BA synthesis. Moreover, changes in BA synthesis were correlated with changes in TG levels, suggesting a common regulatory pathway. Whether these changes are related to increased TG production or decreased clearance remains to be established, although animal and in vitro data suggest the former. Herrema et al.36 proposed a mechanism involving the FXR–LXR–SREBP-1c (sterol regulatory element binding protein-1c) cascade pathway, leading to increased TG production driven by BAs levels. Furthermore, FGF19 has been shown to inhibit hepatic fatty acid synthesis.37 Dedicated studies clearly are needed to completely understand the mechanistic relationship between TG levels and BA synthesis in humans.

In summary, BA metabolism was altered in T2DM, particularly affecting CA and DCA metabolism with an increase in DCA pool size. BA synthesis was inversely correlated with plasma FGF19 in control patients but not in patients with T2DM. Colesevelam treatment led to an increase in BA synthesis, unchanged total BA pool size, and differential effects on individual BA resulting in a decreased hydrophobicity of the circulating BA pool. Treatment also improved glycemic control in subjects with diabetes. Yet, under conditions of profound perturbations in both glucose metabolism (T2DM) and BA metabolism (BA sequestrant treatment), we found no correlation between markers of insulin resistance/glucose metabolism and BA metabolism. Therefore, a firm link between BA and glucose metabolism in T2DM remains elusive, and future studies will be necessary to determine the mechanism for glucose lowering by BA sequestrants.


This study was supported by an investigator-initiated trial grant by Daichii Sankyo, Inc. The authors thank Ido Kema, Claude van der Ley, and Paul van Dijk for plasma sterol measurements and Alan Bostrom for statistical assistance.