Divergence in aerobic capacity impacts bile acid metabolism in young women

Abstract

Liver adaptations may be critical for regular exercise and high aerobic capacity to protect against metabolic disease, but mechanisms remain unknown. Bile acids (BAs) synthesized in the liver are bioactive and can putatively modify energy metabolism. Regular exercise influences BA metabolism in rodents, but effects in humans are unknown. This study tested whether female subjects screened for high aerobic capacity (Hi-Fit, n = 19) [peak oxygen consumption (V̇o_2peak) ≥45 mL·kg⁻¹·min⁻¹] have increased hepatic BA synthesis and different circulating BA composition compared with those matched for age and body mass with low aerobic capacity (Lo-Fit, n = 19) (V̇o_2peak ≤35 mL·kg⁻¹·min⁻¹). Diet patterns, activity level, stool, and blood were collected at baseline before participants received a 1-wk standardized, eucaloric diet. After the 1-wk standardized diet, stool and blood were again collected and an oral glucose tolerance test (OGTT) was performed to assess insulin sensitivity and postprandial BA response. Contrary to our hypothesis, serum 7α-hydroxy-4-cholesten-3-one (C4), a surrogate of BA synthesis, was not different between groups, whereas Hi-Fit women had lower fecal BA concentrations compared with Lo-Fit women. However, Lo-Fit women had a higher and more sustained rise in circulating conjugated BAs during the OGTT. Hi-Fit women showed a significant post-OGTT elevation of the secondary BA, lithocholic acid (a potent TGR5 agonist), in contrast to Lo-Fit women where no response was observed. A 1-wk control diet eliminated most differences in circulating BA species between groups. Overall, the results emphasize the importance of using a standardized diet when evaluating BAs and indicate that regular exercise and aerobic capacity modulate BA metabolism under postprandial conditions.

NEW & NOTEWORTHY Women with contrasting exercise and aerobic capacity levels show clear differences in bile acid (BA) metabolism. Women with low aerobic capacity (Lo-Fit) have increased circulating conjugated BAs post oral glucose tolerance test (OGTT), whereas women with high aerobic capacity (Hi-Fit) display a transient increase. Hi-Fit women show an increase in the secondary BA, lithocholic acid, during the OGTT not seen in Lo-Fit women. Differences in circulating BA species between Hi- and Lo-Fit women possibly contribute to differences in insulin sensitivity and energy regulation via different signaling mechanisms.

INTRODUCTION

Aerobic capacity, also termed cardiorespiratory or aerobic fitness, is the maximal capacity to utilize oxygen systemically (V̇o_2peak) during maximal exercise effort. Aerobic capacity is impacted by genetics, age, and exercise habits. Low aerobic capacity and chronic physical inactivity are strong predictors of chronic disease and early mortality; in contrast, regular exercise and high aerobic capacity provide robust protection (3, 26). Mechanisms explaining this phenomenon need to be fully characterized (29, 39). Independent of body weight, physical inactivity and low aerobic capacity increase the likelihood of fatty liver disease, whereas high levels of physical activity and aerobic capacity reduce the frequency of diagnosis (9, 41). These findings suggest a profound effect of aerobic capacity and exercise on liver metabolism. Our studies utilizing a rodent model bred for divergent intrinsic aerobic capacity replicate outcomes observed in human patients and have begun to shed light on hepatic-driven mechanisms. For instance, high-capacity runner (HCR) rats, bred for high aerobic capacity, are protected from fatty liver, whereas low-capacity runner (LCR) rats, bred for low aerobic capacity, are highly susceptible to diet-induced fatty liver (27–29, 32, 38). These findings track with markers of higher hepatic mitochondrial function (greater respiration and fat oxidation) and reduced de novo lipogenesis in HCR versus LCR rats (37, 38). Similar findings have been found in rodents provided running wheels versus those that remain sedentary (37).

Bile acids (BAs) were once thought to serve exclusively as a facilitator of fat digestion and an end point for cholesterol catabolism, but their function is now being expanded to many other facets of metabolism. For example, BAs bridge the mitochondrial oxidation of substrate to lipid synthesis. Excess substrate carbon in the mitochondria during postprandial conditions is exported into the cytosol, e.g., as citrate, that can be converted to acetyl CoA. Acetyl CoA is then trafficked to either cholesterol and BA synthesis or toward de novo lipogenesis. Factors that control trafficking of acetyl CoA toward cholesterol and BA are largely driven by feedback regulation via enterohepatic recycling of BA and how much BA is excreted in feces.

Several lines of evidence suggest a role for BA metabolism in both fatty liver and the protective effects of exercise on liver metabolism. Treatment with BA sequestrants, a drug that binds and blocks BA reabsorption in the intestine, increases fecal loss of BAs and reduces liver fat storage in mice (46). Importantly, the increased fecal loss of BAs with sequestrant drugs causes increased hepatic expression of cholesterol 7α-hydroxylase (CYP7a1), the rate-limiting step for converting hepatic cholesterol into bile acids. This response is required to maintain systemic BA levels, as typically only 5% are lost in the feces and 95% are reabsorbed and recycled back to the liver via enterohepatic circulation. Overexpression of CYP7a1 also provides protection against steatosis through a similar mechanism (19). Finally, mice that undergo chronic exercise via voluntary wheel running also showed increases in both fecal bile acid and cholesterol excretion (21). The latter results suggest that high aerobic capacity and regular exercise divert postprandially increased citrate away from de novo lipogenesis and toward bile acid synthesis and fecal excretion.

Additionally, recent research is starting to identify changes in specific types of BA species that correlate with an individual’s metabolic health status. Obese and insulin-resistant conditions have been linked to an increase in the ratio of cholic acid (CA) to chenodeoxycholic acid (CDCA)-derived BAs (16). The only exercise and BA study we are aware of in healthy human participants revealed that acute exercise increases circulating lithocholic acid (LCA), a secondary BA derived from CDCA and potent G protein-coupled bile acid receptor (TGR5) agonist (30).

It is unknown whether humans with high or low fitness display differences in total bile acid excretion and/or circulation and if these track with insulin sensitivity or glucose tolerance. To determine whether an aerobic exercise-BA association exists in humans, we tested the hypothesis that female subjects screened for high aerobic capacity due to regular exercise (Hi-Fit, n = 19) (V̇o_2peak ≥45 mL·kg⁻¹·min⁻¹) have increased hepatic BA synthesis and altered composition compared with those matched for age and body mass with low aerobic capacity and low levels of regular exercise (Lo-Fit, n = 19) (V̇o_2peak ≤35 mL·kg⁻¹·min⁻¹).

METHODS

Ethical approval.

The protocol was approved by the Institutional Review Board (IRB00006196) at the University of Kansas Medical Center (KUMC; STUDY00140444). Informed oral and written consent was obtained from each subject before their participation in the study. Participant screening took place at the University of Kansas Medical Center main campus in the Research in Exercise and Cardiovascular Health (REACH) Laboratory at Hemenway Life Sciences Center. All other visits took place at the University of Kansas Clinical Research Center (KUMC CRC) located in Fairway, Kansas.

Participants.

Healthy, young female participants were recruited from the KUMC campus and nearby communities using flyers, email, and word of mouth. Previous research suggests that estrogen signaling is a strong regulator of energy metabolism within the liver (13, 33, 42). Therefore, we only studied female participants to avoid potential sex-based differences that may lead to heterogenous outcomes and diminish power. Prior to coming in the laboratory for a more formal screening, participants were verbally screened for health and behavior to check inclusion/exclusion criteria. Participants eligible for the formal screening (visit 1) were between the ages of 18 and 35, nonsmokers, and weight stable (no greater than 5% change ≤3 mo), did not follow a restrictive diet, and had no history of chronic diseases or gastrointestinal surgeries that would interfere with bile acid metabolism. Formal screening included a maximal oxygen consumption (V̇o_2max) exercise test to exhaustion and health history questionnaire. To be enrolled in the study, participants’ V̇o_2max results had to be either ≥45 mL·kg⁻¹·min⁻¹ (Hi-Fit) or ≤35 mL·kg⁻¹·min⁻¹ (Lo-Fit). These specific cutoffs were chosen based on the American College of Sports Medicine’s classification of excellent to superior fitness or fair to poor fitness, respectively, according to the specific age range of the study (Guidelines for Exercise Testing and Prescription, ninth edition). We screened 89 participants, a total of 39 subjects were enrolled, and only 38 completed the study. One participant dropped out due to a job relocation, leaving n = 19 in the Hi-Fit group and n = 19 in the Lo-Fit group.

Experimental design.

Once determined eligible for enrollment, study subjects were provided a 3-day food diary, activity monitor, and materials for stool collection. About 1 wk after visit 1, subjects returned for a blood draw following an overnight fast (>10 h) (visit 2) and brought back a completed 3-day food diary (23), a frozen stool sample, and an activity monitor. Following visit 2, participants were supplied with a standardized diet for the next 7–8 days. During the last few days of the standardized diet, subjects provided a second stool sample and brought this to visit 3. On visit 3, subjects completed another fasting blood draw paired with an oral glucose tolerance test (OGTT) and a dual-energy X-ray absorptiometry (DEXA) body composition scan.

Demographic measures.

Height was measured during formal screening (visit 1) using a stadiometer in the REACH laboratory. Weight was measured during screening (visit 1), before diet (visit 2), and postdiet (visit 3) using a calibrated body weight scale in the REACH laboratory and at KUMC CRC. Participants were responsible for recording daily food intake for 3 days (2 weekdays and 1 weekend) only at baseline. Diet records were analyzed for micro- and macronutrient content using Nutrient Data System for Research (NDSR, version 2016). Current activity level of participants was tracked and measured using accelerometers (ActiGraph GT3X). Activity data were interpreted by averaging total moderate-to-vigorous physical activity (MVPA) as categorized by the Actilife software using the Freedson Adult MVPA cut points (12). To be included in the analysis, a minimum wear time of ≥ 8 h over a period of 3 days (2 weekdays + 1 weekend day) was required.

Maximal oxygen consumption (V̇o_2max) was determined by indirect calorimetry using a Bruce exercise test protocol (4) on a motor-driven treadmill in the REACH laboratory as performed by the Thyfault laboratory previously (23). Body composition via DEXA was measured postintervention (visit 3) to quantify fat and lean body mass at KUMC CRC. Participants also provided information on general health history, medication usage (including birth control), and lifestyle habits using questionnaires.

Standardized diet.

Meals were prepared and packed-out at the KUMC CRC Metabolic Kitchen with support from the KUMC Dietetics and Nutrition Department. The diet was designed to be eucaloric for each participant (based on Harris-Benedict prediction) and consisted of a macronutrient distribution per day: 30–35% fat, 15–20% protein, and 45–50% carbohydrate, along with 15–25 g/day of fiber (see Table 1). Before obtaining meals, participants were provided with a menu for the 7-day standardized diet and queried about food tolerance issues. Any foods identified by the participant were replaced with another food item of approximately equal nutritive value. Participants who were not taking any dietary supplement that could alter bile acid or lipid metabolism, primarily any probiotics/prebiotics or omega-3 supplements, started the standardized diet immediately after visit 2. Participants who were taking supplements that could potentially confound results stopped taking the supplements for a 1-wk washout period before starting the standardized diet. During the standardized diet, participants were directed to consume all the prepared food and avoid any other nonstudy-provided foods. Participants could consume additional calorie-free beverages. A food record was given to the participant to record their percent food intake of the provided meals and any additional beverages consumed. Participants arrived at KUMC CRC ~1–2 times during the 7–8-day feed to receive pack-out meals.

Table 1.

Self-reported energy and nutritional intake values from both Hi-Fit and Lo-Fit participants before and following the 7-day control diet

	Hi-Fit		Low-Fit
	Prediet	Postdiet	Prediet	Postdiet
Energy, kcal	2,136.56 + 116.54	2,521.73 + 64.14*	1,774.62 + 96.59†	2,065.93 + 48.14*
Total fat, g	85.07 + 4.98	97.47 + 2.62*	73.78 + 4.74	77.03 + 2.25
SFAs, g	26.69 + 1.94	27.30 + 0.68	23.95 + 1.86	22.23 + 0.59
MUFAs, g	32.15 + 1.94	39.07 + 1.19*	25.36 + 1.77†	29.97 + 0.94*
PUFAs, g	18.48 + 1.34	23.21 + 0.57*	18.04 + 1.40	18.51 + 0.55
Total carbohydrate, g	259.15 + 20.36	305.51 + 7.99*	199.55 + 15.61†	256.19 + 4.73*
Total protein, g	89.00 + 5.11	119.29 + 2.91*	70.39 + 4.12†	97.78 + 2.45*
Methionine, g	1.96 + 0.13	2.60 + 0.06*	1.58 + 0.11†	2.11 + 0.06*
Cystine, g	1.23 + 0.07	1.54 + 0.04*	0.94 + 0.06†	1.28 + 0.03*
Glutamic acid, g	17.01 + 0.94	23.57 + 0.59*	13.96 + 0.82†	19.44 + 0.45*
Glycine, g	3.71 + 0.23	4.71 + 0.12*	2.87 + 0.17†	3.77 + 0.11*
Cholesterol, mg	355.06 + 37.63	228.91 + 4.86*	267.67 + 32.43	185.05 + 5.85*
Total dietary fiber, g	28.52 + 1.97	25.09 + 0.68	16.95 + 1.13†	20.78 + 0.44*
Soluble dietary fiber, g	8.16 + 0.54	8.42 + 0.28	4.86 + 0.33†	6.84 + 0.14*
Insoluble dietary fiber, g	20.15 + 1.54	16.49 + 0.40*	12.01 + 0.88†	13.77 + 0.30*
% Calories from fat	35.42 + 1.47	30.30 + 0.08*	37.03 + 1.62	30.17 + 0.12*
% Calories from SFAs	11.18 + 0.72	9.40 + 0.06*	12.08 + 0.65	9.37 + 0.05*
% Calories from MUFAs	13.40 + 0.63	11.21 + 0.02*	12.76 + 0.72	11.07 + 0.04*
% Calories from PUFAs	7.56 + 0.35	6.99 + 0.03	8.94 + 0.64	7.05 + 0.05*
% Calories from carbohydrate	45.94 + 1.91	50.37 + 0.13*	43.60 + 2.11	50.14 + 0.09*
% Calories from protein	16.63 + 0.49	19.38 + 0.07*	15.98 + 0.63	19.73 + 0.07*

Values are means ± SE. Significance was determined by independent samples t test for changes between groups and paired t test for within group; Hi-Fit, women with high aerobic capacity; Lo-Fit, women with low aerobic capacity; MUFAs, monounsaturated fatty acids; PUFAs, polyunsaturated fatty acids; SFAs, saturated fatty acids.

^†P < 0.05 compared with baseline of Hi-Fit;

^*P < 0.05 compared with baseline of same group; n = 18–19 participants.

Stool collection and sample processing.

Two stool samples were collected, once before initiation of the standardized diet, between visits 1 and 2, and again at the end of the 7-day standardized diet (right before visit 3). Collection occurred at two different time points to determine whether nonstandardized diet conditions impact fecal bile acids in either group. Protocols were given to participants instructing them on the proper handling of stool specimens, including storage in freezer using conical tubes and placement in a cooler with ice pack. In brief, participants were given a screw-top fecal container with a spoon, a disposable stool collection device that sat across the toilet, a plastic specimen biohazard bag, a silver insulated envelope, and an ice pack. The stool collection device was fixed to the toilet via adhesive tape. The participant’s stool was collected into the paper portion of the device, followed by the participant using the spoon to scoop stool into the fecal container. The fecal container was then closed with a screw-top lid and placed into the plastic specimen bag, which was then placed into the silver insulated envelope next to an ice pack and stored in the participant’s freezer until it was transported to the laboratory and stored at −80°C. For each sample, weighed feces (~2 g) was pulverized in a cryogenic grinding vial in liquid nitrogen using the 6875 Freezer/Mill cryogenic grinder (SPEX SamplePrep; Metuchen, NJ) according to the manufacturer’s instructions as follows: the grinding vial was precooled in liquid nitrogen for 1 min before the precooled samples in liquid nitrogen were ground. Samples were ground for 40 s at a rate of 10 cycles per second, followed by 1-min cool time. Powdered samples were collected in 50-mL conical tubes and stored in the −80°C freezer.

Blood collection.

Serum bile acid, glucose, and lipid measurements were conducted in samples collected before diet intervention (visit 2) and postdiet intervention (visit 3). Measurements were taken at two different time points to determine whether nonstandardized diet conditions impact metabolic measures in either group or in both groups. To measure serum metabolites, participants were fasted overnight (>10 h). A catheter was placed in an arm vein, and a baseline blood sample (≤100 mL) was collected.

An OGTT was performed at visit 3 to assess metabolic health via glucose tolerance and predicted insulin sensitivity as done previously by our laboratory (24, 25). Participants fasted overnight (>10 h) before entering the KUMC CRC in the morning, where an in-line catheter was placed in an arm vein and a baseline blood sample (≤50 mL) was collected. Participants then consumed a glucose tolerance beverage (Trutol 75, Thermo Scientific) containing 75 g of glucose. For the next 2 h, they were asked to lie still on a reclined chair while a small blood sample was collected (≤1.5 mL) every 30 min to measure glucose, insulin, and bile acids. Blood samples were collected into serum separator tubes at 0, 30, 60, 90, and 120 min. Blood sat at room temperature for 30 min and was then centrifuged at 3,000 g for 10 min at 4°C. Serum was frozen at −80°C until subsequent analysis.

Serum analysis.

Serum glucose was determined using the glucose oxidase method (Sigma, St. Louis, MO) or a glucose analyzer (YSI), serum insulin was measured by ELISA kit (Alpco, Salem, NH), and nonesterified fatty acid (NEFA) was measured by spectrophotometric assay (Wako, Neuss, Germany). The areas under the curve (AUCs) for insulin and NEFAs (inverted AUC because these metabolites drop in concentration post-OGTT) were calculated using the trapezoidal method. Fibroblast growth factor 19 (FGF19) concentrations were determined by ELISA (Quantikine Human FGF-19 ELISA kit, R&D Systems Inc., Minneapolis, MN) according to the manufacturer’s instructions.

LC-MS analysis of bile acids.

Glycine-conjugated, taurine-conjugated, and nonconjugated BA concentrations were quantified by the Arkansas Children’s Nutrition Center Metabolomics and Analytical Chemistry Core in fasting serum using LC-MS methodology. Serum samples (50 μL) were prepared as previously described (22). Powdered feces (100 mg) was resuspended in 2-mL tubes containing 500 µL aqueous methanol plus 100 µL internal standard (CA-D4, DCA-D4, GCDCA-D9, LCA-D4, 50% MeOH, final concentration 5 µg/mL) and homogenized using a Precellys 24 homogenizer (Bertin Corp., Rockville, MD) at 5,300 rpm for two 30-s cycles. The homogenates were centrifuged at 10,000 g and −4°C for 10 min. The supernatant was transferred into 2-mL microcentrifuge tubes. The collected supernatant was evaporated to dryness under a nitrogen stream and then reconstituted in 100 µL of 50% MetOH. Chromatic separation was performed on an UltiMate 3000 UHPLC system fitted with an Acquity BEH C18 column, 100 × 2.1 mm, 1.7 μm, (Waters, Milford, MA). A flow rate of 200 mL/min was used: mobile phases A (10 mM ammonium acetate, 0.01% acetic acid in water) and B (0.01% acetic acid in methanol) with the 30-min gradient as follows: 0–1 min, 25% B; 1–21 min, 95% B; 21–26 min, 95% B; 26–27 min, 25% B, 27–30 min; 25% B. Identification was carried out on a Q-Exactive high-resolution accurate mass (HRAM) spectrometer, and data were acquired by an ESI-Full-MS scan and analyzed using Xcalibur 4.0 and TraceFinder 3.3 software. The following settings were used: nitrogen as sheath, auxiliary, and sweep gas was set at 45, 10, and 2 units, respectively; resolution, 70,000 full width at half maximum; automatic gain control target, 1e6 ions; maximum injection time, 200 ms; scan range, 275–750 m/z; spray voltage, 2.70 kV; capillary temperature, 320°C; and source temperature, 425°C. Individual bile acids were identified by exact mass and retention time as shown in Table 2. For each BA, peak areas, normalized to their deuterated internal standard, were used to quantitate; calibration curves (0–5,000 nM) showed linearity >0.99.

Table 2.

Bile acids measured, exact mass, retention time, and internal standards for mass spectrometry analysis of bile acid species

Bile Acid	Exact Mass (M-H)	RT (min)	ITSD
CA	407.28030	15.78	CA-D4
CDCA	391.28538	17.64	DCA-D4
DCA	391.28538	17.91	DCA-D4
UDCA	391.28538	14.72	DCA-D4
LCA	375.29047	19.42	LCA-D4
GCA	464.30176	13.81	GCA-D4
GCDCA	448.30685	15.49	GCDCA-D9
GDCA	448.30685	15.91	GCDCA-D9
GUDCA	448.30685	12.67	GCDCA-D9
GLCA	432.31193	17.26	LCA-D4
TCA	514.28440	13.67	GCA-D4
TCDCA	498.28948	15.31	GCDCA-D9
TDCA	498.28948	15.69	GCDCA-D9
TUDCA	498.28948	12.56	GCDCA-D9
TLCA	482.29457	17.04	LCA-D4

CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; GCA, glycocholic acid; GDCA, glycodeoxycholic acid; GCDCA, glycochenodeoxycholic acid; GLCA, glycolithocholic acid; GUDCA, glycoursodeoxycholic acid; ITSD, internal standards; LCA, lithocholic acid; RT, retention time; TCA, taurocholic acid; TCDCA, taurochenodeoxycholic acid; TDCA, taurodeoxycholic acid; TLCA, taurolithocholic acid; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid.

LC-MS analysis of 7α-hydroxy-4-cholesten-3-one (C4).

The quantification of C4 was determined in fasting serum samples with the use of LC-MS as previously described (22). Serum samples (30 μL) were spiked with the internal standard, 7α-hydroxy-4-cholesten-3-one-d7 (C4-D7) at 9 nM final concentration; lipids were extracted using a modified Bligh and Dyer method (2). Chromatographic separation was performed on an UltiMate 3000 UHPLC system (Thermo Scientific) using an Acquity BEH C18 column (100 × 2.1 mm, 1.7 μm; Waters) (22); identification was carried out on a Q-Exactive HRAM spectrometer (Thermo Scientific); and data were acquired by positive heated electrospray ionization (HESI+) targeted-SIM scan and analyzed using Xcalibur 4.0 and TraceFinder 3.3 software (22). Calibration curves, 0–65 nM, showed linearities with >r² = 0.997. The limits of detection for C4 and C4-D7 were 0.97 nM and 0.58 nM, respectively. The limit of quantitation for both compounds was 1.9 nM. C4 concentrations were corrected by using percentage recovery of the internal standard.

Data analysis.

Differences in fecal bile acids and serum bile acid species poststandardized diet, along with AUCs from the OGTT were compared between the Hi-Fit and Lo-Fit groups using an independent samples t test. To test whether there were differences between groups for changes in serum measurements during the OGTT, we performed a repeated measures two-way ANOVA (time × group). SigmaPlot (V.13) and SPSS (IBM Analytics, New York, NY) were used to perform statistical analysis, and Prism was used to plot graphs. Statistical significance was set at P < 0.05, and all data were expressed as means ± SE.

RESULTS

Participant characteristics.

Characteristics of the Hi- and Lo-Fit participants are summarized in Table 3. Age, body mass index, height, and weight were nearly identical between groups, whereas V̇o_2max was the greatest differentiating factor (P < 0.0001). As expected, MVPA, assigned diet kilocalorie, and body fat and fat-free mass percentage, along with absolute fat-free mass, were also significantly different between Hi-Fit and Lo-Fit groups. Although not intentional, the Hi-Fit group experienced a significant loss of body weight during the standardized diet, an effect not found in Lo-Fit (P < 0.05). This unintentional weight change was likely due to an underestimation of caloric needs because of a higher activity level (see Table 3). The standardized diet lowered total serum BA in Lo-Fit women but not Hi-Fit women (Fig. 1A). Fecal BA concentrations were lowered by the diet intervention in both groups (significant diet effect, Fig. 1B). There was a main effect of Hi-Fit women having lower fecal BA concentrations regardless of diet status compared with Lo-Fit women (Fig. 1B). C4 and FGF19 concentrations were not significantly affected by diet in either group, nor were there differences across group for these measures (Fig. 1, C and D).

Table 3.

Characteristics of healthy, normal weight women divergent in aerobic capacity

Variable	Hi-Fit	Lo-Fit
n	19	19
Age, yr	25.63 ± 0.81	25.95 ± 1.00
BMI	22.33 ± 0.50	22.64 ± 0.51
Height, cm	168.37 ± 1.98	166.23 ± 1.92
Weight, kg	63.47 ± 2.09	62.69 ± 2.07
V̇o2max, mL·kg−1·min−1***	48.89 ± 0.68	32.68 ± 0.72
MVPA avg, min/day**	43.61 ± 4.20	23.71 ± 3.02
% Body fat***	24.61 ± 0.83	35.07 ± 1.13
Fat free mass, kg**	48.08 ± 1.55	40.95 ± 1.24
FFM/total mass, %***	0.76 ± 0.01	0.66 ± 0.01
Assigned kcal/day***	2,505.26 ± 63.74	2,031.58 ± 53.56
∆ Weight*	−0.68 ± 0.19	−0.05 ± 0.16

Values are means ± SE. Significance was determined by independent samples t test. BMI, body mass index; FFM, fat free mass; Hi-Fit, women with high aerobic capacity; Lo-Fit, women with low aerobic capacity; MVPA, moderate-to-vigorous physical activity.

^*P < 0.05,

^**P < 0.001,

^***P < 0.0001.

Fig. 1.

Diet intervention significantly reduces serum and fecal bile acid (BA) concentrations. Fasting serum collected pre- and postdiet standardization in women with high aerobic capacity (Hi-Fit) and low aerobic capacity (Lo-Fit). A: serum total BA (mM). B: fecal BA (mmol/g). C: serum 7α-hydroxy-4-cholesten-3-one (C4; nM). D: serum fibroblast growth factor 19 (FGF19; pg/mL). Significance was determined by two-way RMANOVA (fitness × diet) followed by Student–Newman–Keuls post hoc analysis, *P < 0.05 between groups (Hi-Fit vs. Lo-Fit); n = 17–19 participants.

Insulin sensitivity.

Hi-Fit and Lo-Fit women demonstrated similar changes in blood glucose during a 75-g OGTT administered following the weeklong diet intervention phase (Fig. 2A) but had significantly different insulin responses. As expected, Hi-Fit women had lower serum insulin throughout the OGTT compared with Lo-Fit women (Fig. 2B) and lower insulin AUC (data not shown), reflective of higher peripheral insulin sensitivity. We also measured serum nonesterified fatty acids (NEFAs) at baseline and during the OGTT. Upon ingestion of a mixed meal, serum NEFA concentration is suppressed as lipolysis is inhibited by increased insulin secretion (11). The Lo-Fit group displayed a higher NEFA concentration during the OGTT versus Hi-Fit, suggestive of modestly higher lipolysis in the fasted state (Fig. 2C), although the change in absolute NEFA from time 0 to time 120 was the same for both groups.

Fig. 2.

Higher fit women display greater insulin sensitivity. Serum measurements at 0, 30, 60, 90, and 120 min post 75-g oral glucose tolerance test (OGTT). A: serum glucose (mg/dL). B: serum insulin (uIU/mL). C: serum nonesterified fatty acids (NEFAs; meq/L). Significance was determined by two-way RMANOVA (fitness × time) followed by Student–Newman–Keuls post hoc analysis; for changes over time within groups, P < 0.05, a < b < c; *P < 0.05 between groups [women with high aerobic capacity (Hi-Fit) vs. women with low aerobic capacity (Lo-Fit)] at each time point; n = 17–19 participants.

Serum bile acids.

Glycine-conjugated, taurine-conjugated, and unconjugated bile acids were measured in overnight-fasted serum and during the OGTT. The sums of total, conjugated, and unconjugated BA concentrations and individual BA concentrations were not different between Hi-Fit and Lo-Fit women postdiet intervention in the overnight-fasted state (Fig. 3, A and B). In Lo-Fit women, total conjugated BA concentrations increased by threefold at 30 min post-OGTT, and concentrations remained high through the remainder of the time course (Fig. 4, A and C). Hi-Fit women also displayed an increase in total conjugated BA concentrations at 30 min post-OGTT (~2.2-fold), but in contrast to Lo-Fit women, concentrations had returned to baseline by 60 min onward (Fig. 4, A and C). In contrast, Hi-Fit women had an ~40% increase in serum total unconjugated BA at 30 min post-OGTT (Fig. 4, B and D). This transient rise was followed by a return to (or somewhat below) baseline concentrations by 60 min onward (Fig. 4, B and D). Lo-Fit women did not show any significant changes in total unconjugated BA concentrations throughout the OGTT (Fig. 4, B and D).

Fig. 3.

Serum bile acid (BA) profile in the overnight-fasted state following 1-wk controlled diet intervention in women with high and low aerobic capacity (Hi-Fit and Low-Fit, respectively). A: serum total, unconjugated, and conjugated BAs (uM). B: serum BA species (uM). n = 18–19 participants. CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; GCA , glycocholic acid; GDCA, glycodeoxycholic acid; GCDCA, glycochenodeoxycholic acid; GLCA, glycolithocholic acid; GUDCA, glycoursodeoxycholic acid; LCA, lithocholic acid; TCA, taurocholic acid; TCDCA, taurochenodeoxycholic acid; TDCA, taurodeoxycholic acid; TLCA, taurolithocholic acid; TUDCA, tauroursodeoxycholic acid; UDCA, ursodeoxycholic acid.

Fig. 4.

Oral glucose tolerance test (OGTT)-associated bile acid (BA) serum concentrations are different between women with high and low aerobic capacity (Hi-Fit and Lo-Fit, respectively). Fasting serum measurements at 0, 30, 60, 90, and 120 min post 75-g OGTT. A: serum total conjugated BAs (µM). B: serum unconjugated BAs (µM). C: percent change of serum-conjugated BAs. D: percent change of unconjugated BAs. Significance was determined by two-way RMANOVA (fitness × time) followed by Student–Newman–Keuls post hoc analysis; for changes over time within groups, P < 0.05, a < b < c; *P < 0.05 between groups (Hi-Fit vs. Lo-Fit) at each time point; n = 18–19 participants.

We next considered the patterns of individual BA species following the OGTT both as a relative percent change (Fig. 5) and absolute value (Table 4). Hi-Fit women trended toward higher concentrations of the secondary BA, deoxycholic acid (DCA), at 30 min post-OGTT in comparison to Lo-Fit women, but this did not reach statistical significance (Fig. 5A). Percent changes in LCA concentrations were higher at 30, 60, 90, and 120 min post-OGTT in Hi-Fit women compared with Lo-Fit women (Fig. 5B). However, the opposite was true when evaluating absolute value, as the concentration of LCA was significantly higher in Lo-Fit women at 0, 30, 60, 90, and 120 min post-OGTT, although there were no changes over time (Table 4). Serum ursodeoxycholic acid (UDCA) concentrations post-OGTT were highly variable; we observed no significant differences between Hi-Fit and Lo-Fit women (Fig. 5C). The primary BA, CA, did not significantly change following OGTT irrespective of group (Fig. 5D). Serum CDCA concentrations in both Hi-Fit and Lo-Fit women decreased following OGTT (Fig. 5E).

Fig. 5.

Higher fit women have an increase in secondary unconjugated bile acids (BAs) post oral glucose tolerance test (OGTT). Fasting serum measurements at 0, 30, 60, 90, and 120 min post 75-g OGTT. A: percent change in deoxycholic acid (DCA). B: percent change in lithocholic acid (LCA). C: percent change in ursodeoxycholic acid (UDCA). D: percent change in cholic acid (CA). E: percent change in chenodeoxycholic acid (CDCA). Significance was determined by two-way RMANOVA (fitness × time) followed by Student–Newman–Keuls post hoc analysis; for changes over time within groups, P < 0.05, a < b < c; *P < 0.05 between groups [women with high aerobic capacity (Hi-Fit) vs. women with low aerobic capacity (Lo-Fit)] at each time point; n = 18–19 participants.

Table 4.

Absolute quantities of unconjugated bile acids during oral glucose tolerance test

Group	Time					P Values
	0	30	60	90	120	Fitness	Time	Interaction
DCA
Hi-Fit	0.34 + 0.07a	0.39 + 0.06a	0.33 + 0.05a,b	0.31 + 0.05b	0.29 + 0.05b	0.188	<0.001	0.295
Lo-Fit	0.28 + 0.05	0.25 + 0.04	0.22 + 0.03	0.21 + 0.03	0.21 + 0.03
LCA
Hi-Fit	0.002 + 0.003	0.011 + 0.001	0.008 + 0.001	0.006 + 0.001	0.006 + 0.001	<0.001	0.389	0.136
Lo-Fit	0.019 + 0.004*	0.018 + 0.004*	0.017 + 0.002*	0.017 + 0.003*	0.016 + 0.003*
UDCA
Hi-Fit	0.037 + 0.01	0.062 + 0.01	0.031 + 0.01	0.042 + 0.01	0.031 + 0.01	0.933	0.193	0.15
Lo-Fit	0.042 + 0.01	0.040 + 0.01	0.045 + 0.01	0.049 + 0.01	0.038 + 0.01
CA
Hi-Fit	0.063 + 0.02	0.051 + 0.01	0.041 + 0.01	0.036 + 0.01	0.343 + 0.01	0.285	0.077	0.214
Lo-Fit	0.037 + 0.01	0.036 + 0.01	0.032 + 0.01	0.032 + 0.01	0.039 + 0.01
CDCA
Hi-Fit	0.088 + 0.02a	0.068 + 0.01b	0.046 + 0.01b	0.036 + 0.01b	0.033 + 0.01b	0.182	<0.001	0.957
Lo-Fit	0.116 + 0.03a	0.081 + 0.02b	0.073 + 0.02b	0.057 + 0.01b	0053 + 0.01b

Values are means ± SE. CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; Hi-Fit, women with high aerobic capacity; LCA, lithocholic acid; Lo-Fit, women with low aerobic capacity; UDCA, ursodeoxycholic acid. Significance was determined by two-way RMANOVA (fitness × time) followed by Student–Newman–Keuls post hoc analysis; for changes over time within groups, P < 0.05, a < b < c;

^*P < 0.05 between groups (Hi-Fit vs. Lo-Fit) at each time point; n = 18–19 participants.

Bile acid production markers.

Cholesterol 7α-hydroxylase (CYP7a1) is the rate-limiting enzyme in BA production. Its expression has a diurnal rhythm and increases with meal consumption (14). The quantification of serum 7α-hydroxy-4-cholesten-3-one (C4) is a clinically proven surrogate measure of CYP7a1 activity and reflects the overall rate of hepatic BA synthesis (1, 15). In the current study, we found no significant difference in C4 levels between the Hi-Fit and Lo-Fit participants during the OGTT (Fig. 6A).

Fig. 6.

Bile acid (BA) synthesis marker serum 7α-hydroxy-4-cholesten-3-one (C4) is unaffected by aerobic capacity. Serum measurements at 0, 30, 60, 90, and 120 min post 75-g oral glucose tolerance test. A: C4 (nM). B: serum fibroblast growth factor 19 (FGF19; pg/mL). C: serum cholecystokinin (CCK; pg/mL). Significance was determined by two-way RMANOVA (fitness × time) followed by Student–Newman–Keuls post hoc analysis; *P < 0.05 between groups [women with high aerobic capacity (Hi-Fit) vs. women with low aerobic capacity (Lo-Fit)] at each time point; n = 17–19 participants.

CYP7a1 activity is under negative feedback regulation from BA returning to the liver, which triggers transcriptional repression (7). The nuclear receptor farnesoid X receptor (FXR) is primarily responsible for facilitating the repressive effect of BA on CYP7a1 within both the liver and intestines. In the intestine, BA activation of FXR induces fibroblast growth factor 19 (FGF19), which is released into the hepatic portal vein and binds with hepatic fibroblast growth factor receptor 4 (FGFR4) to inhibit CYP7a1. Aerobic capacity did not influence fasting or post-OGTT FGF19 levels, and blood concentrations were not strongly impacted by the OGTT either at specific time points (Fig. 6B) or AUC (data not shown).

Cholecystokinin (CCK) is a peptide hormone synthesized and secreted in the upper small intestine by enteroendocrine cells. Upon ingestion of a meal, CCK is rapidly released into circulation and acts to stimulate gallbladder contraction (BA release) and delay gastric emptying (6). Because BA synthesis (as marked by C4) was not different between the Hi-Fit and Lo-Fit cohorts of women, we measured CCK to determine whether gallbladder contraction may play a role in the increased circulating conjugated BA in the Lo-Fit women. Although there were no significant group differences in the post-OGTT CCK rise, the Lo-Fit women displayed higher CCK concentrations overall (Fig. 6C).

Fecal bile acids.

About 5% of total BAs are lost through fecal excretion and must be replenished by a compensatory increase in hepatic synthesis. We also examined fecal BA concentration to estimate total fecal BA loss. Because post-OGTT serum-conjugated BAs were higher in the Lo-Fit women, we reasoned that reabsorption/recycling might also be higher in this group postprandially on a day-to-day basis. If true, this could lead to less conjugated BA being excreted in feces in the Lo-Fit women. However, fecal total BA concentration was not different between the Hi-Fit and Lo-Fit groups (Fig. 7A), and there were no differences in conjugated BA and other species (Fig. 7B). It is important to note that although concentration was not different, we did not collect a total volume over a specified time period, so total fecal BA loss/day could not be determined.

Fig. 7.

Fecal bile acid concentration and species profile is not influenced by aerobic capacity. Fecal bile acid concentration post 1-wk controlled diet intervention. A: total bile acids (umol/g). B: bile acid species (umol/g). n = 18–19 participants [women with high aerobic capacity (Hi-Fit) vs. women with low aerobic capacity (Lo-Fit)]. CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; LCA, lithocholic acid; UDCA, ursodeoxycholic acid.

DISCUSSION

Currently, very little is known about how exercise or aerobic capacity modifies BA metabolism, despite BAs playing an important role in liver and whole body metabolism. In rodents, regular exercise has been shown to increase fecal BA and cholesterol loss (21). Interestingly, CYP7a1 overexpression and BA sequestrants are known to protect against hepatic steatosis (19, 46). These observations, in aggregate, have led us to hypothesize that higher amounts of exercise and/or fitness promote fecal BA loss, which in turn would channel liver cytosolic acetyl-CoA toward activated cholesterol and BA synthesis and away from de novo lipogenesis. Of note, differences in circulating BA species have been observed after an acute bout of exercise in healthy male subjects (30). In the current study, we aimed to understand how regular exercise corresponding with a high aerobic capacity impacts overall BA metabolism in both postabsorptive and OGTT (postprandial) conditions. Counter to our hypothesis that Hi-Fit women would display increased BA synthesis and fecal BA loss, we found BA synthesis measured as serum 7α-hydroxy-4-cholesten-3-one (C4), a clinically proven surrogate of bile acid production, was not different between Hi-Fit and Lo-Fit women both fasting and during the OGTT. Also, fecal BA concentrations did not differ between fitness groups. However, Lo-Fit women showed a sustained increase in circulating conjugated bile acids following an OGTT, whereas Hi-Fit women showed increased conjugated BA at only 30 min post-OGTT. Hi-Fit women displayed a sustained elevation of the secondary BA, lithocholic acid (LCA), a potent TGR5 agonist, during the entire OGTT. Together, these data suggest that differences in exercise and aerobic capacity do lead to changes in BA metabolism and postprandial circulating concentrations of select metabolites. The impact of these changes is unknown but could, in theory, differentially modulate energy metabolism in the postprandial state via bioactive effects of BAs.

Aside from potential exercise effects, it is well known that macronutrient composition, specifically fat intake, has a heavy influence on BA metabolism (10). Additionally, dietary supplementation (e.g., prebiotics/probiotics) can have indirect effects on BA metabolism by altering gut microbiota (22). Results from the current study confirm the need to control for food and supplement intake when studying BAs, as all major BA measurements became normalized in both Hi- and Lo-Fit women after consumption of a standardized diet. Specifically, Lo-Fit participants showed a significant decrease in total serum BA after the 1-wk control diet.

There are two major receptors BAs are known to activate that can induce changes in energy metabolism. First, aside from playing a pivotal role in mediating the feedback regulation of BA synthesis, the hepatic nuclear receptor farnesoid X receptor (FXR) has triglyceride-lowering effects within both the liver and plasma (18). Multiple mechanisms are thought to be responsible for the lipid-lowering effects of FXR, but the primary mechanism is through downstream inhibition of the lipogenic transcription factor’s steroid response element-binding protein (SREBP)-1c and carbohydrate response element-binding protein (ChREBP) (5, 45). Within the liver, FXR can also directly increase fat oxidation through the induction of peroxisome proliferator-activated receptor-α (PPARα) (34) or indirectly by activating the intestinal hormone FGF19 to signal back to the liver via the portal vein (40). The conjugated form of the primary BA, chenodeoxycholic acid (CDCA), is the most potent agonist of FXR (36). A decrease in the ratio of CDCA to the other primary BA, cholic acid (CA), and its derivatives is now being associated with obesity and insulin resistance (16). In the current study, although we saw reduced insulin sensitivity in the Lo-Fit group, this did not correspond to a decreased ratio of CDCA to CA-derived BAs (data not shown) or plasma FGF19 levels, either in a fasting or postprandial state. It is possible that these ratios are more germane to insulin resistance associated with obesity than with lower insulin sensitivity found in young and otherwise healthy Lo-Fit women studied herein.

The G protein coupled receptor (TGR5) is the second receptor activated by BAs and primarily regulates energy metabolism through its expression in human skeletal muscle and intestines. When induced by BAs in the muscle, TGR5 increases energy consumption via its coexpression with the thyroid hormone receptor, type 2 iodothyronine deiodinase (D2) (44). In the intestines, TGR5 activation promotes glucose tolerance by inducing glucagon-like peptide-1 (GLP-1) release from intestinal L-cells (35). LCA, a secondary BA derived from CDCA, is the most potent TGR5 agonist (18). Of the individual BAs measured in the current study, Hi-Fit women showed increased LCA in response to the OGTT, whereas Lo-Fit women showed no response but had higher overall absolute levels. Previous research in healthy men has demonstrated increased blood LCA after exposure to an acute bout of exercise (30). The correlation between LCA and exercise, whether acute or chronic, suggests that improved insulin sensitivity and fitness-associated changes in energy metabolism could in part be through specific BA signaling either in the intestine or skeletal muscle. However, it should be noted that the mechanistic effects of BA to modify energy metabolism have been developed in murine models and have largely not been tested in humans. Thus, more work is needed to determine whether BA truly drives energy homeostasis differences across the fitness spectrum in human subjects.

Previous research, along with the current study, supports the notion that aerobic capacity is important for higher insulin sensitivity and overall metabolic health (31). Specifically, aerobic capacity and regular exercise are crucial for liver health and are known to protect against hepatic steatosis (37) while also tracking with higher insulin sensitivity and protection against type 2 diabetes (8, 17). In a rodent model specifically bred for high and low aerobic capacity (HCR/LCR), we have shown that a high aerobic capacity is associated with increased hepatic mitochondrial oxidative capacity and fat oxidation, factors associated with protection against hepatic steatosis and insulin resistance (8, 17, 29, 38). However, the mechanisms driving these effects are yet to be defined. Exercise studies in mice show an upregulation in both cholesterol turnover and fecal BA excretion (20, 21). In agreement with these findings, unpublished data from our laboratory show that rats bred for high aerobic capacity have increased hepatic mRNA expression of CYP7a1, the rate-limiting enzyme in bile acid production, along with increased fecal total BA excretion. Interestingly, previous results show a unique connection between increased BA synthesis, increased BA fecal loss, and hepatic mitochondrial function. Mice given BA sequestrants show increased hepatic PGC-1a expression, a powerful regulator of mitochondrial biogenesis and function, and evidence of increased mitochondrial bioenergetic function (43, 46).

Despite these findings in rodent models, herein we did not see a difference in BA synthesis (C4) or fecal BA concentration between Hi- and Lo-Fit women. Interestingly, there was a sustained increase in circulating conjugated bile acids throughout the OGTT in Lo-Fit women. This increase in circulating conjugated bile acids in Lo-Fit women could be due to several reasons. Conjugation of BAs is important for facilitating their reuptake in the small intestine (ileum) and contributing to feedback regulation via enterohepatic circulation. The increase in conjugated BAs during the OGTT in Lo-Fit women could mean that they are reabsorbing more from the small intestine to be recycled, whereas in Hi-Fit women, a greater proportion of conjugated BAs are reaching the colon, where they interact with gut bacteria to become unconjugated BAs. On the other hand, the increase in circulating conjugated BAs in Lo-Fit women could also be reflective of greater BA release from the gallbladder, which would align well with the increase in serum CCK also seen in the Lo-Fit women. If we were to speculate that a greater proportion of BAs in the Hi-Fit women were to reach the colon and become unconjugated, this would mean that less would be reabsorbed, as unconjugated BA transport occurs mostly by passive absorption. If this were the case, then we might also assume an increase in BA synthesis, an effect we did not see according to our C4 data. In summary, although fitness level did not impact apparent fasting or post-OGTT BA synthesis rate, there were clear differences in other aspects of BA metabolism in Hi-Fit versus Lo-Fit women.

There were several experimental limitations that likely effected our results. A primary reason why there may be no difference in fecal BA content between the Hi- and Lo-Fit women was that we did not collect total fecal volume over a specified time period; rather, we only measured concentration in one stool. By not collecting total volume over a 24–48-h time period, we were not able to correct for potential fecal volume differences between the groups, ultimately affecting net bile acid excretion. It should also be noted that the kilocalorie requirements of the Hi-Fit women were on average 500 kcal/day greater than the Lo-Fit women, so it is likely that fecal volume would have also been substantially increased. We also required that all participants ceased exercising at least 4 days before each blood draw to minimize the impact of acute exercise on metabolic outcome measures. This was done to ensure that differences between the groups were solely due to contrasting levels of aerobic capacity, rather than the lasting effects of an acute exercise bout. It is possible that cessation of exercise several days before collection masked any BA differences inherent to regular exercise. Whether energy intake is equal to or greater than energy requirements (energy balance) may also influence BA outcomes, as excess substrate provides the fuel source for cholesterol and BA synthesis. The Hi-Fit group had subtle but significant weight loss on the control diet, which may have impacted outcomes, although large differences in BA measures were not seen pre- versus poststandardized 7-day diet in this group. A lack of change in BAs within the Hi-Fit group is interesting to note, as habitual exercise volumes likely change gut microbiome and therefore might have also played a role in BA profile. We did not evaluate gut microbiota and therefore cannot speak to the level of influence this might have had on BA measures. Finally, to truly assess the power of exercise and aerobic capacity on energy metabolism, an important outcome measure would be to perform liver imaging, but due to ethical and financial concerns, this was not incorporated into the research design.

In conclusion, we found clear differences in BA metabolism between women with contrasting exercise and aerobic capacity levels. Results from the current study demonstrate that Lo-Fit women have a sustained increase in circulating conjugated BAs post-OGTT, whereas Hi-Fit women showed only a transient increase followed by a return to baseline in later stages of the OGTT. Hi-Fit women also showed sustained elevations of the secondary BA, LCA, during the OGTT not seen in the Lo-Fit women. The differences in circulating BA species between Hi- and Lo-Fit women could possibly contribute to their differences in insulin sensitivity and energy regulation via different signaling mechanisms. Elevated LCA in Hi-Fit women may improve insulin sensitivity and energy expenditure via induction of the receptor TGR5. However, an alternative view must be acknowledged: insulin sensitivity and signals associated with exercise and fitness drive the differences in BA metabolism and not vice versa. This study offered a unique opportunity to observe metabolic differences between healthy Hi-Fit and Lo-Fit women and provides a useful framework for future study design to build within this line of research. Ultimately, more research is needed to confirm whether BA metabolism plays a role in mediating metabolic health associated with exercise and aerobic capacity and to determine if improved or deteriorating metabolic health drive within-person changes in BA homeostasis.

GRANTS

This work was supported by US Department of Agriculture Agricultural Research Service Projects 6026-51000-010-05S and 6026-51000-012-06S. J. P. Thyfault is also supported by Veterans Affairs-Merit Grant 1I01BX002567 and NIH Grants R01 KD121497 and R01 AR071263.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.M., K.E.M., S.H.A., and J.P.T. conceived and designed research; A.M., J.L.W., S.A.B., H.L., and K.E.M. performed experiments; A.M., K.D., H.L., K.E.M., and J.P.T. analyzed data; A.M., K.E.M., S.H.A., and J.P.T. interpreted results of experiments; A.M., K.E.M., and J.P.T. prepared figures; A.M. and K.E.M. drafted manuscript; A.M., J.L.W., K.D., S.A.B., K.E.M., S.H.A., and J.P.T. edited and revised manuscript; A.M., K.E.M., S.H.A., and J.P.T. approved final version of manuscript.