Quantification of common and planar bile acids in tissues and cultured cells

DOI 10.1194/jlr.D120000726 Copyright © 2020 Shiffka et al. Published under exclusive license by The American Society for Biochemistry and Molecular Biology, Inc.. NASH and NAFLD (9–12), diabetes (13–15), metabolic disease (16), and drug-induced liver injury (17). Understanding the particular changes in the BA pool that occur in these conditions can aid in diagnostic and prognostic assessments thereof. Moreover, a deeper awareness of the mechanisms that give rise to the perturbations in the BA pool during disease can assist in the understanding of hepatic and gastrointestinal (GI) function and pathology. The biosynthetic pathway of BAs is highly complex, resulting in an exceptionally diverse pool of BAs mostly within the liver, GI tract, bile, plasma, urine, and feces (2, 5). BAs and their derivatives can also accumulate in other tissues, especially in illness (1, 3, 5, 18). In health, the BA pool is mainly made up of conjugated primary BAs; however, in disease and following injury, dysregulation in BA synthesis and/or homeostasis can result in the accumulation of unusual and intermediate species (1, 3, 5, 19–21). Thus, the total number of possible BA derivatives that circulate in disease is orders of magnitude more diverse than in health, demonstrating the need for a more sensitive quantification method that captures this high heterogeneity in the multiple compartments of the BA pool. This phenomenon of distortion in the BA pool has already been demonstrated in numerous studies in human patients in the disease states mentioned above as well as in several animal models (10, 22–28). Nevertheless, there remain multiple subsets of BAs that have yet to be characterized in much detail, if at all. Further investigation is needed to determine whether the appearance of these molecules in disease is related to disease progression, either as instigators or byproducts, and whether they can be used to aid in clinical detection and prognoses. Because it is still unclear exactly why, how, and which of these BA species develop in various disease states, investigating them is critical for understanding the molecular mechanisms behind these conditions. Unfortunately, due to their high degree of structural similarity, most detection methods are insufficient to simultaneously detect and differentiate the large variety of BAs (29, 30). One such case is the planar BAs, a category that differs in their 3D structure, resulting in a “flat” or “planar” conformation contrary to the more typical “bent” or “twisted” shape of the steroid backbone (1, 25, 31). Abstract Bile acids (BAs) have been established as ubiquitous regulatory molecules implicated in a large variety of healthy and pathological processes. However, the scope of BA heterogeneity is often underrepresented in current literature. This is due in part to inadequate detection methods, which fail to distinguish the individual constituents of the BA pool. Thus, the primary aim of this study was to develop a method that would allow the simultaneous analysis of specific C24 BA species, and to apply that method to biological systems of interest. Herein, we describe the generation and validation of an LC-MS/MS assay for quantification of numerous BAs in a variety of cell systems and relevant biofluids and tissue. These studies included the first baseline level assessment for planar BAs, including allocholic acid, in cell lines, biofluids, and tissue in a nonhuman primate (NHP) laboratory animal, Macaca mulatta, in healthy conditions. These results indicate that immortalized cell lines make poor models for the study of BA synthesis and metabolism, whereas human primary hepatocytes represent a promising alternative model system. We also characterized the BA pool of M. mulatta in detail. Our results support the use of NHP models for the study of BA metabolism and pathology in lieu of murine models. Moreover, the method developed here can be applied to the study of common and planar C24 BA species in other systems.

The planar BAs are of interest due to their resurgence in the circulating BA pool in several types of liver disease, but the high degree of similarity to the more abundant BAs make them challenging to detect in vivo (1,(23)(24)(25)31). Historically, BA heterogeneity was detected with GC-MS (8,23,(32)(33)(34). Recently, several research groups have taken advantage of the high specificity and selectivity of LC-MS/ MS to reliably detect and quantify these molecules among their more typical counterparts (18,30,(35)(36)(37). The study herein sought to build upon and extend previous methodologies and simultaneously characterize BAs in vitro and in vivo.
BA metabolism is often studied using animal models that assume similarity to human systems. Murine BA metabolism is well studied and known to have several discrepancies when compared with that of humans, but mouse and rat models are still used most often (5,(38)(39)(40)(41)(42)(43)(44)(45). Nonhuman primate (NHP) models represent an attractive alternative because of their increased genetic resemblance to man; however, BA metabolism in NHPs has not been as well characterized (46,47). A secondary purpose of this study was to apply the method developed herein to establish a detailed baseline profile of the BA pool in relevant tissues and biofluids in NHPs, for the comparison to injury models in this animal and to humans. Similarly, cultured cell systems of human origin are also popular in studying liver physiology and pharmacology, but these models, too, are poorly detailed with regard to BA synthesis and metabolism. Examples of primary and immortalized cell systems were examined. Herein, the development, validation, and application of a sensitive UPLC-MS/MS method for the simultaneous quantification of abundant BAs, as well as uncommon and planar mammalian BA species, is demonstrated in human cell lysates and cell culture media and in NHP liver tissue, bile, and plasma.

Immortalized cell lines and culture conditions
COS-1 and HepG2 cells were purchased from ATCC (Manassas, VA). HuH-7 cells were a generous gift from Dr. Hongbing Wang (University of Maryland, Baltimore, MD). COS-1, HuH-7, and HepG2 cells were cultured in DMEM supplemented with 10% FBS, penicillin (100 IU/ml), and streptomycin (100 g/ml) (Life Technologies, Inc., Rockville, MD). For the analysis of basal BA production in immortalized cell lines, cells were plated at a density of 2.0 × 10 6 cells per 10 cm culture dish in DMEM containing no supplements and in hepatocyte culture medium (HCM) and maintained at 37°C with 5% CO 2 for 48 h (Lonza, Basel, Switzerland). Human primary hepatocytes (HPHs) were obtained from BioIVT (Baltimore, MD). Hepatocytes at 90% viability were seeded at 1.5 × 10 6 cells/well in a 6-well collagen-coated plate in InVitroGRO CP medium (BioIVT). After overnight attachment at 37°C in a humidified atmosphere of 5% CO 2 , the culture medium was changed to complete Williams' E medium. The cells were overlaid with 0.25 mg/ml Matrigel (Corning Inc., Corning, NY) for another 24 h before the medium was switched to HCM. The medium was then aspirated and stored at 20°C until further preparation for LC/MS-MS injection. Before lysis, cells were rinsed with cold PBS. If not further processed immediately, culture dishes were stored at 80°C. then centrifuged at 500 g for 10 min. Four milliliters of each supernatant were transferred to fresh glass culture tubes and dried down under N 2 flow. Extracts were resuspended in 100 l 1:1 ACN:water with 0.01% FA, resulting in a final IS concentration of 50 ng/ml, and transferred to LC-MS vials and then stored at 20°C until injection. For cells, 1 ml RIPA buffer with 5 ng of each IS was added to each plate or well following a PBS wash. Culture dishes were rocked at 4°C for 1 h and then scraped. Total protein was estimated at this point using the Bradford assay. Lysates were added to 4 ml 100% ACN in glass culture tubes and vortexed quickly and then centrifuged at 2,000 rpm for 10 min. Four milliliters of each supernatant were transferred to fresh glass culture tubes and dried down under N 2 flow. Extracts were resuspended in 100 l 1:1 ACN:water with 0.01% FA, resulting in a final IS concentration of 50 ng/ml, and transferred to LC-MS vials and then stored at 20°C until injection.
Plasma and bile preparation. ISOLUTE PLD+ phospholipid depletion columns were filled before use with 400 l ACN with 1% FA as protein crash solvent. Five microliters of 1 ng/l IS solution and 100 l plasma aliquot were added to the column and vortexed. Flow-through was then collected into glass culture tubes by applying approximately 4 psi positive pressure by nitrogen gas. Eluates were further dried down by nitrogen flow and then resuspended in 100 l 1:1 ACN:water with 0.01% FA for LC-MS/MS injection. If resuspensions were not injected that day, samples were stored at 80°C. For bile samples, the protocol was performed similarly, save that the aliquoted bile was diluted 1:9 or 1:1,000 with water and 5 l of 1 ng/l IS or 100 l of 50 ng/l IS, respectively to each dilution, were added before being applied to the ISOLUTE PLD+ column. Additionally, the bile that was diluted 1:1,000 with water preextraction was diluted 1:100 postextraction by resuspending in 100 l 1:1 ACN:water with 0.01% FA and then diluting 10 l into 1,000 l of 1:1 ACN:water with 0.01% FA (resulting in a final dilution factor of 1:100,000) for the analysis of highly abundant analyte species. For all dilutions, final injections contained 50 ng/ml of each IS.
Liver sample preparation. Approximately 50 mg sections of liver initially removed from the central right lobe of naïve rhesus macaques were preweighed and added to Fisherbrand™ prefilled bead mill tubes (2ml, 1.4mm beads) with 750 l 50% methanol. Tissue sections were homogenized using a Precellys 24 tissue homogenizer run at 6,500 rpm for 10 s; tubes were then centrifuged at 500 g for 1 min. 600 l of each homogenate was aspirated and spiked with IS (1 ng each IS per 10 mg tissue section weight), then transferred to a new tube with 3 ml ice-cold alkaline ACN (ACN with 5% v/v NH 4 OH, equal to 1.15M NH 4 OH) and vortexed for 5 s. Homogenates were shaken at ambient room temperature (23 ± 2°C) for 1 h, then centrifuged at 5,000 g for 10 min. 3 ml of each supernatant was applied to ISOLUTE PLD+ columns. Approximately 4 p.s.i. positive pressure was applied to collect flow-through into glass culture tubes. Flow-through fractions were dried under N 2 flow and then resuspended in 1:1 ACN:water with 0.01% FA proportionally to starting tissue section weight and IS added. Final injections thus contained 0.5 mg of liver tissue per l of resuspension. The use of a basic crash solvent in this procedure has the disadvantage of dehydrating 7-hydroxy groups in 7-hydroxy-3-oxo-4-ene BA structures, causing loss of at least one analyte of interest during sample preparation (52).

LC-MS/MS
Chromatography. The LC-MS/MS method used was adapted from (35). LC-MS/MS analysis was performed on a Waters I-Class UPLC coupled to a Waters Xevo TQ-XS triple quadrupole mass spectrometer (Waters Corporation, Milford, MA). Separation was effected by a Waters ACQUITY BEH (ethylene bridged hybrid) C 18 UPLC column (150 × 2.1 mm, 1.7 m) using gradient elution with the following mobile phases: water with 0.01% FA (solvent A) and ACN with 0.01% FA (solvent B). The gradient was as follows: 25-40% solvent B in 12 min and then 40-75% solvent B in 14 min. Solvent B was increased to 100% over 30 s and held for 2 min to wash the column. Return to 25% B was accomplished over 30 s and allowed to reequilibrate over 4 min. The column was maintained at 55°C with a flow rate of 350 l/min. Two microliters of sample were injected per run.
ESI-MS/MS. Detection was performed in negative ion mode using ESI. Source conditions were as follows: capillary voltage was 2.50 kV, and cone voltage was 10 V. The desolvation temperature was set to 500°C and the source temperature to 150°C. The desolvation gas flow operated at 800 l/h, the cone gas flow at 150 l/h, and the collision gas flow at 0.15 ml/min. The Q1 low mass resolution was set to 3.06 and the high mass resolution to 14.84. The Q3 low mass resolution was set to 2.73 and high mass resolution to 15.11. Ion guide offsets were set to 3 V and 0.3 V. Both the entrance and exit potentials were set to 1.0. Dwell time for each analyte was 7.0 ms. BA species were detected using scheduled multiple reaction monitoring wherein each BA is detected with multiple reaction monitoring according to a unique precursor to product ion m/z transition or selected ion monitoring where a single ion is detected in both Q1 and Q3 with no fragmentation (supplemental Table S3). Each BA was further identified by its chromatographic retention time (supplemental Table S3).
Data processing. Data were processed using Waters MassLynx and TargetLynx software (version 4.1 SCN 901). Analyte responses were determined by the peak area of the BA, and analyte concentrations were determined by calibration curves constructed for each BA standard by plotting the ratio of the response for each BA to the IS response against the nominal concentration. Linearity of calibration curves was assessed through linear regression analysis using a weighting factor of 1/x. Calculated concentrations of analytes were converted from nanograms per milliliter to nanograms per milligram of starting liver tissue by dividing by a factor of 500 to reflect the theoretical 500 mg of liver tissue in every 1 ml of final prepared resuspension used for injection. Bile measurements were multiplied by their dilution factors (i.e., either ×10 or ×100,000) to reflect the initial dilution with water and postextraction dilution in mobile phase.

Method validation
Method validation studies were designed by following the FDA's Guidance for Industry for Bioanalytical Method Validation for sensitivity, linear range, precision, accuracy, recovery, and bench-top stability. Because BAs are endogenous compounds and blank matrix is not feasible to obtain, the nonendogenous IS compounds were used as surrogates to validate the LC-MS/MS method. Intraand inter-day precision was determined by injection of the same set of prepared plasma samples spiked with IS 4 and 24 h apart, respectively. Similarly, plasma spiked with IS was left at room temperature (23 ± 2°C) for 4 h before performing sample preparation and compared with plasma spiked with IS prepared freshly in order to assess benchtop precision. Identical validation experiments were conducted for bile, liver tissue, and cultured cells (COS-1 cells used for validation experiments) and cell media.

Statistics
Student's t-test analyses were performed using GraphPad Prism software (version 8.3.0).

Method development
Sample preparation optimization. Initially, the protocol from Han et al. (35) was duplicated for use in plasma; however, this resulted in poor extraction efficiency (5-15%) with the HybridSPE-Phospholipid 96-well plates and continuously had issues with the wells becoming clogged with biological samples. However, a similar sample preparation method provided by another vendor, i.e., the ISOLUTE PLD+ protein and phospholipid removal plate, which provided better recovery but exhibited high variability between wells. Subsequently, it was determined that using the ISO-LUTE PLD+ phospholipid depletion columns yielded the most robust extraction [i.e., highest and most consistent recovery and lowest coefficient of variation values and utilized this method for the remainder of our experiments in biofluids and tissue. The protocol developed for use with these columns required less than half the time; it was also less labor-and resource-intensive than other methods (35). Biotage ISOLUTE PLD+ protein and phospholipid depletion columns were utilized for plasma, bile, and liver sample preparation. Cultured cells and media did not require SPE procedures. Table S3). CA-d 4 was used as an IS for the unconjugated BAs (LCA, CDCA, DCA, UDCA, CA, ACA, 7-hydroxy-3oxo-chol-4-en-24-oic acid, IALCA, 3-oxo-chol-4-enic acid, and 3-oxo-CA), and GCDCA-d 4 was used as an IS for the conjugated BAs (GLCA, GDCA, GUDCA, GCA, TLCA, TDCA, TCDCA, and TCA). BAs were resolved with a reverse-phase gradient UPLC separation using a 150 mm column with a sub-2 m particle size C 18 stationary phase with an ACN/water/FA-based mobile phase. Retention times ranged from 7.9 min (TCA) to 22.0 min (LCA) (supplemental Table S3). BAs that fragmented well were detected according to an m/z transition (supplemental Table S3). Some BAs exhibited poor fragmentation efficiency and were more suitably detected by monitoring the precursor ion in both Q1 and Q3 (LCA, CDCA, DCA, UDCA, CA, ACA, and IALCA). Those BAs that shared a similar nominal mass and utilized the same selected reaction monitoring transitions were chromatographically resolved ( Fig. 1).

Method validation
BA sensitivity and linearity were evaluated using authentic standards for each BA. Because BAs are endogenous species, ISs (CA-d 4 and GCDCA-d 4 ) were used as nonendogenous BA surrogates to evaluate intra-and inter-day precision and accuracy, apparent recovery, and benchtop stability (supplemental Table S2).
Sensitivity and linearity. Sensitivity and linearity were determined by evaluating calibrants ranging in concentration from 0.05 to 2,500 ng/ml. The limit of detection (LOD) is defined as those points within the calibration curves with a signal-to-noise ratio (S/N) greater than 3. The lower limit of quantification (LLOQ) is defined as the lowest point on the calibration curve whose S/N is greater than 10 and whose calculated value is 20% of the nominal value. The LOD for BAs in this assay ranged from 0.05 to 10 ng/ml with most BAs having LODs of 0.05-0.5 ng/ml ( Table 1). LLOQs for the various BAs are represented by the lower limit of the calibration curves in Table 1 and were between 0.1 and 10 ng/ml with most BAs having LLOQs having 0.5 ng/ml. The most sensitive LLOQ was TLCA (0.1 ng/ ml), whereas the least sensitive LLOQ was for 3-oxo-chol-4enic acid (10 ng/ml). The upper limit of quantification was 2,500 ng/ml for all BAs included in this assay. Linearity was evaluated via regression coefficients (r 2 ) for calibration curves that exceeded 0.99, ranging from 0.990 to 0.999 (Table 1).
Precision and accuracy. Intra-day and inter-day instrument precision and accuracy were evaluated for plasma, bile, liver, cell lysate, and cell media. Intra-day precision for all matrices evaluated was less than 15%, ranging from 2.4% to 14.6% (supplemental Table S2). Inter-day precision was also less than 15%, ranging from 2.6% to 14.5%. (supplemental Table S2). Intra-and inter-day accuracy for all matrices ranged from 80.4% to 119.5% and from 88.2% to 113.3%, respectively.
Recovery. Apparent recovery of CA-d 4 and GCDCA-d 4 in cell lysates was 74.1% and 66.0%, respectively, and in cell medium was 98.8% and 85.5%, respectively, with excellent reproducibility for each (less than ±8.4%). Bile and liver had similar recovery for CA-d 4 and GCDCA-d 4 , 100% for both in bile and 73% for both in liver tissue. Variability in recovery was less for liver (less than ±4.5%) than in bile (less than ±28.9%). Plasma had different levels of recovery for CA-d 4 and GCDCA-d 4 with CA-d 4 having 45.5 ± 13% recovery and GCDCA-d 4 having 114.9 ± 28.6% recovery (supplemental Table S2).
Bench-top stability. Bench-top stability was evaluated for plasma, bile, liver, cell lysate, and cell media at room temperature (23 ± 2°C) for 4 h. BA stability in each of the matrices was between 2.8% and 13.6% deviation of the nominal value, indicating acceptable stability (defined as 15%) over this time period at these conditions in each of these matrices (supplemental Table S2).

Application
In order to demonstrate the utility of this method, several in vitro and in vivo model systems frequently used to study hepatic function were designated for investigation. The normal BA pools of two immortalized cell lines, as well as HPHs, were determined. The BA levels in the liver tissue, plasma, and bile of healthy M. mulatta, a NHP laboratory model, were also established.
BA composition of cultured cell systems. Hepatic cell systems commonly used include two immortalized hepatoblastoma cell lines, HepG2 and HuH-7, as well as HPHs. HepG2 and HuH-7 cells are routinely grown in DMEM, whereas HPHs require a supplemented medium such as HCM. The basal levels of BAs in the immortalized cells grown in both types of media were determined for comparison to HPHs ( Tables  2, 3). Additionally, both types of medium with and without COS-1 growth were analyzed to ensure that no exogenous BAs were present. For the most part, slightly more BAs were present in HCM than in unsupplemented DMEM, though this difference was only statistically significant when examining individual BAs, not total BAs (TBAs) as a whole. This was most obvious in the cases of CDCA and its taurine-conjugate, TCDCA, and was true in both the measured medium aliquots and cell lysates (Tables 2, 3).
Both the immortalized and the primary cells exhibited levels of secondary BAs (namely, UDCA, LCA, and DCA and their conjugates) to some extent, though this was much more pronounced in the HepG2 and HuH-7 cell lines and especially true for DCA and its amidated conjugates (Tables 2, 3; supplemental Fig. S1). In HPHs, these species were only detectable in conjugated forms (G-and T-amidated) in the culture medium (supplemental Fig. S2). Secondary BAs BA composition in plasma, bile, and liver tissue from a NHP model. Naïve NHP plasma, bile, and liver tissue were examined to determine detailed baseline information of the BA pool for later comparison to liver and gut injury models in these animals as well as in humans. In plasma, CDCA and its taurine and glycine conjugates predominated with the unconjugated BAs as the most abundant. Hence, within plasma, unconjugated BAs were the most abundant species, totaling an estimated average of 73% of TBAs with 69% of this being CDCA (supplemental These values were determined using authentic standards dissolved in mobile phase. LOD is defined as analyte peak area having S/N >3. LLOQ is represented as the lower end of the linear range and is defined as analyte peak area having S/N >10 and deviation from the nominal value less than ±20%. The r 2 value reflects fit to linear regression with 1/x weighting. Determined concentration is displayed as mean ± SD. a Values are presented as picograms per microgram of total protein corresponding to cells grown to confluence in n = 3 10 cm dishes over 48 h containing 10 ml total medium per plate.
b Values are presented as picograms per microgram of total protein corresponding to cells grown to confluence in n = 6 wells of a 6-well plate over 48 h containing 2 ml total medium per well per 24 h.
c Indicates at least one value <LLOQ and, thus, not included (n < 3 for HepG2 and HuH-7 cells). Fig. S3A). Glycine-conjugated and taurine-conjugated BAs made up averages of approximately 10% and 17% of the plasma BA pool, respectively, again being mostly conjugates of CDCA (supplemental Fig. S3A). Total circulating BAs ranged from 1.17 to 2.58 g/ml and averaged 1.82 g/ml ( Table 4). BA profiles within NHP bile and liver tissue were very similar to one another. In both, conjugated BAs, specifically TCDCA and TCA, dominated the BA pool. In bile, TCDCA occupied an average of nearly half of BA species (40.5% of TBAs) at 180.7 mg/ml (Table 4, supplemental Fig. S3B). TCDCA concurrently made up an average of approximately one-fourth of the BA pool in liver (26.5% of TBAs) at 17.9 ng/mg tissue (supplemental Fig. S3C). On average, TCA comprised 21% (89 mg/ml) of the BA pool in bile and 29.1% in liver tissue (32 ng/mg tissue). The succeeding Determined concentration is displayed as mean ± SD. a Values are presented as picograms per microgram of total protein corresponding to cells grown to confluence in n = 3 10 cm dishes over 48 h containing 10 ml total medium per plate.
b Values are presented as picograms per microgram of total protein corresponding to cells grown to confluence in n = 6 wells of a 6-well plate over 48 h containing 2 ml total medium per well per 24 h.
c Indicates at least one value <LLOQ and, thus, not included (n < 3 for HepG2 and HuH-7 cells). most abundant BA species were also very similar between liver and bile. In bile, the next most prominent BA species were, in descending order: GCA (13.8% of TBAs), GDCA (9.8%), GCDCA (6.3%), and TDCA (5.4%) (supplemental Fig. S3B). In liver, the next most abundant BAs were GCA (18.7% of TBAs), GCDCA (9.8%), GDCA (6.4%), and TDCA (5.7%) (supplemental Fig. S3C). In total, conjugated BAs made up most of both the bile and liver BA pools in similar ratios, with taurine-conjugated species comprising 68.6% (bile) and 62.3% (liver) and glycine-conjugated species making up 31.4% (bile) and 36.3% (liver) of TBAs, on average. Thus, conversely to plasma, unconjugated BAs were minor constituents and totaled less than 1.5% of TBAs in liver and 0.004% of TBAs in bile, reflecting the high proficiency of the healthy hepatic conjugative enzyme systems. TBAs in bile ranged between 378 and 485 mg/ml (Table 4), an average of roughly 430 mg/ml. In liver, TBAs were highly variable and ranged between 19.5 and 175.5 ng/mg tissue, averaging 97.5 ng/mg tissue. The high variability is a consequence of this study's small sample size and ad libitum feeding; additional samples and control of fasting/feeding status would provide a better indication of TBA concentration, as well as individual BA concentration, in healthy M. mulatta liver and bile.

LC-MS/MS methodology
A LC-MS/MS assay was developed for quantification of BA concentrations in a variety of cell systems and relevant biomatrices. These studies illustrate the validation of a simplified rapid sample preparation in comparison to recent methods. Using this method, BA isomers were reliably separated; however, additional specificity for these and similar molecules could be gained by using an instrument with higher resolution capabilities. Note that the first baseline level assessment for planar BAs, including ACA, in cell lines, biofluids, and tissue under healthy conditions is included. This single assay and simple sample preparation methods are highly adaptable and could be easily applied to a variety of experimental conditions relevant to BA homeostasis and pathology. This methodology could also be applied to human clinical samples or other animal models with the addition of species-specific BAs. Furthermore, the sensitivity of this method allows for the quantification of even low abundance BA species. A significant limitation of this method, however, is that its application for measurement of C27, sulfated, and glucuronidated BA species has not yet been assessed.

Cell culture BA quantification
Unsupplemented DMEM proved to be sufficient for the cellular production of BAs to allow measurement of the BA pool and total viability in immortalized HepG2 and HuH-7 cell lines for the time period examined (48 h); however, slightly more BAs were present per total protein in both cell lines when incubated in supplemented medium (i.e., HCM) (Tables 2, 3). In both media and cell lines, taurine conjugation was slightly more prevalent than glycine conjugation (supplemental Fig. S1), which is directly contrary to the case in the HPHs examined herein (supplemental Fig.  S2C) and in humans in vivo, in which glycine conjugation predominates greatly (53,54). This is especially interesting because neither DMEM nor HCM include taurine in their formulations, whereas glycine is included at a concentration of 30 mg/l (in DMEM) to 50 mg/l (in HCM), indicating that all three of these cell types are synthesizing taurine to some degree. Biosynthesis of taurine occurs in vivo in the liver and represents a sulfur excretion mechanism, so it is to be expected that hepatic and hepatocyte-derived cells are capable of this reaction; however, it is unexpected that the immortalized cells studied herein preferentially amidate BAs with the limited taurine instead of the more available glycine molecule that is favored in vivo (58). The altered amidation pattern in immortalized cells must be a result of modified cell metabolism, such as impedance in uptake in one or both of these amino acids or altered functionality and/or selectivity in the BA-CoA:amino acid N-acyltransferase (BAAT), the enzyme responsible for BA conjugation (1,3,5,55). Indeed, qRT-PCR analysis has demonstrated that BAAT mRNA is undetectable in HepG2 and HuH-7 cell lines in comparison to HPHs (59). This is supported by this study, which demonstrated approximately equal amounts of conjugated and unconjugated BAs produced per total protein content in these immortalized cells (supplemental Fig. S1). Thus, BA levels in HepG2s and HuH-7s indicate that the cells are capable of BA production and conjugation but are deficient in comparison to HPHs and the healthy whole liver. Supporting this, HepG2 and HuH-7 cells presented impaired BA synthesis as a whole, producing only up to 6% of the TBAs produced by HPHs (Tables 2, 3). This is also in agreement with previous studies, which show very low or undetectable levels of mRNAs associated with BA synthetic enzymes CYP7A1, CYP7B1, and CYP8B1 in HepG2 and HuH-7 cells (59). Conversely, these cell lines express CYP27A1 mRNA, the rate-limiting enzyme of the acidic or alternative pathway of BA synthesis, at levels 130-155% those of HPHs (1,59). This, combined with the higher proportions of CDCA and its conjugates, adds evidence to the theory that the acidic or alternative pathway of BA synthesis is more active in immortalized cells relative to primary cells; however, BA synthesis as a whole is still very depressed, which is likely due to decreased expression of the succeeding enzymes in the BA synthesis pathway (60,61).
Of the BAs produced in immortalized cells, the amidated BAs were found almost exclusively in the culture medium, whereas the unconjugated BAs, CA and ACA, made up nearly all of the BA species within cells (supplemental Fig. S1). These distinct extracellular and intracellular profiles indicate that BA transport mechanisms selective for amidated BAs, such as BSEP, MDR2/3, and MRP, are still active within HepG2 and HuH-7 cells (1). Earlier studies demonstrated that, while BSEP and MDR2/3 mRNA levels are marginal or are undetectable in HepG2 and HuH-7 cells, MRP1 and MRP4 mRNA is nearly as abundant or many times more abundant than in HPHs (59). Hence, HepG2 and HuH-7 cells are more capable of exporting than they are at synthesizing amidated BAs, which is underpinned by both the high proportion of unconjugated BAs within the cell lysates and by the production of the secondary BAs, LCA and DCA, which are presumed to be produced in vivo by intestinal bacteria (Tables 2, 3; supplemental Fig. S1) (1,3,5,55,56). Earlier studies investigating the production of BAs by cultured cell systems did not probe for secondary BAs and enzymatically deconjugated BAs before analysis; thus, to the best of our knowledge, this study offers the most detailed characterization of BA production in HepG2 and HuH-7 cells to date (60,62,63). The production of secondary BAs is presumably due to the buildup of unconjugated CA and CDCA within the cells: when insufficiently conjugated and exported, these molecules are subject to dehydroxylation, which is likely performed by the same enzymes that rehydroxylate LCA and DCA in the normal liver, as this enzymatic reaction is reversible (64). The presence of secondary BAs in immortalized and primary cells calls into question the assumption that LCA and DCA are generated exclusively from the activity of intestinal bacteria in vivo. Further studies are needed to determine whether hepatic synthesis of "secondary" BAs contributes significantly to the BA pool and whether this contribution is altered during disease.
As far as we are aware, this study is the first to observe the presence of planar BAs in human-derived cultured cells, though planar BAs have been discovered in both healthy and compromised individuals (30,(65)(66)(67). A significant portion of the unconjugated BAs present in HepG2 and HuH-7 cell lysates was ACA, the 5-epimer of CA. The planar BAs represent a subset of BAs that retain the flat 3D structure of cholesterol instead of the bent or twisted structure of typical mammalian BAs (1,24,25,31). This conformation can either be due to an -positioning of the C5 hydrogen atom, as in ACA, or a double bond between C4 and C5, as in 7-hydroxy-3-oxo-chol-4-en-24oic acid (1,19,25,31). These molecules are abundant in pregnancy and during the first few months of life but gradually disappear after infancy and are not usually detectable in healthy adults (25,31). In certain cases of liver or GI disease, however, these species reappear among circulating BAs (31). The mechanisms of the recurrence of planar BAs have yet to be fully elucidated, but it is thought that their synthesis is due to immature hepatic enzyme systems that can be found in the underdeveloped hepatocytes of the fetus and in the dysregulated cells of the injured liver (31). Alternatively, it is possible that these species are isomerized from their more typical counterparts, either by hepatic or microbial enzymes (31). The production of planar BAs by immortalized cells, however, substantiate the theory that proliferating hepatocytes must dedifferentiate in order to repopulate the injured liver and, therefore, express less specialized or incomplete metabolic enzymes, resulting in the production of "fetal" BA species (68). Moreover, previous studies in HepG2 cells detected high levels of additional BA precursors and intermediates for which we did not probe (62).
Though HPHs seem to produce fewer TBAs compared with the whole healthy human liver, these cells at least appear to produce similar ratios of each BA and amidate in similar patterns (supplemental Fig. S2) (69,70). Glycine amidation predominated the BA pool greatly, but it should be noted that HCM does not provide any exogenous taurine, as stated above, which is presumably why HPHs produced such a low proportion of taurine-conjugated BAs. Also, this study examined BA production only in the 48 h following perfusion, so it should also be considered that the amount of BAs generated changes over time and does not reach maximal production until day 6 (i.e., 144 h) (60,63). Finally, only one donor was assayed in this study; further investigations would tell whether these findings are true of all HPH cultures. Primary cells therefore offer a relatively good model of human liver BA synthesis and metabolism, as long as the above limitations are considered (60,69,70). Conversely, the HepG2 and HuH-7 cell lines make very poor models for the study of BA synthesis and metabolism, as they do not produce nearly as many or the same pattern of BAs as HPHs or whole human liver.

NHP BA quantification
NHP models are the most relevant and preferred model systems currently available in drug discovery and development due to the high degree of phylogeny and homogeneity to humans (71). Despite this, BA synthesis and metabolism in these animals remains poorly characterized. Herein, the BA pools of several compartments were quantified in an accepted laboratory model, the NHP.
Naïve NHP plasma BAs were dominated by unconjugated CDCA, distantly followed by TCDCA and GCDCA and then by GCA and TCA (Table 4, supplemental Fig.  S3A). These results conform with previous studies indicating the prevalence of unconjugated BAs in rhesus macaque plasma; however, they disagree with previous studies regarding individual BAs, finding a higher percentage of unconjugated CDCA instead of DCA (46,47,72). Circulating TBAs concentration in NHP plasma proved comparable to that in human plasma as reported in scientific literature, i.e., an average of 1.82 g/ml in NHPs compared with 1.01-2.13 g/ml in humans (Table 4) (35,53,70). It is important to note, however, that the composition of the BA pool was altered in naïve NHP plasma, yielding a higher CDCA:CA ratio than is usual in human plasma. NHP plasma also presented a high proportion (73%) of total unconjugated BAs, specifically CDCA, whereas human plasma BAs are mostly (56-59%) glycine-amidated (supplemental Fig. S3A) (35,46,53). Yet, this has been shown to change with diet and fasting state: the glycine:taurine ratio in humans can change with diet and cooking habits, and the ratio of unconjugated to amidated BAs is increased in the fasting state compared with the nonfasting state (35,53).
The preponderance of CDCA compared with CA that occurred in plasma was also reflected in naïve macaque bile and liver, which demonstrated similar profiles when compared with one another (supplemental Fig. S3B, C). In these tissues, the BA pools were dominated by a high proportion of taurine-conjugated primary BAs, CDCA and CA. As has been reported in human bile and liver, an overwhelming majority of BAs in naïve macaque liver and bile were amidated, with only a very small fraction escaping conjugation (supplemental Fig. S3B, C) (1,3,5,19,70,73). Though interindividual variability in this study was relatively high due to small sample size and ad libitum feeding, it is apparent that naïve NHP liver produced comparable TBAs to healthy human liver as reported in previous studies, i.e., an average of 97.5 ng/mg of tissue versus 7.7-29.5 ng/mg of tissue in humans (Table 4) (70,73). Although, as in plasma, NHP liver shows a notable difference in regard to the glycine/taurine ratio, preferring taurine to glycine amidation, which is converse to that in humans (supplemental Fig. S3B, C) (3,35,70).
In terms of TBAs, the NHP offers a high resemblance to humans in plasma, liver tissue, and bile (35,46,53,54,69,70,73). The individual members of the BA pool in these tissues, however, are distinct between NHPs and humans. The greatest differences observed in this study include: increased CDCA/CA ratio in the NHP BA pool, greater proportion of unconjugated compared with conjugated BAs within plasma, and the preference for taurine versus glycine amidation. Additionally, though this study did not evaluate sulfated BAs, at least one previous study has reported marked differences between the extent of this BA elimination pathway in NHPs versus humans (46,47). Overall, however, this laboratory animal represents a highly relevant model for the study of BA synthesis, metabolism, and pharmacology with relatively minor disparities in comparison to other models.

CONCLUSIONS
The optimization, validation, and application of a sensitive and selective LC-MS/MS method for the simultaneous detection and quantification of numerous BAs in cultured cell systems and several biomatrices are described herein. The immortalized hepatoblastoma cell lines HepG2 and HuH-7 were studied for comparison with cultured HPHs and previous research examining whole liver in vivo (69,70). Results from these studies demonstrate that these immortalized cell systems are poor indicators of healthy hepatic BA metabolism; however, HPHs are a promising alternative. Although, before using these cells as models, it should be established whether supplementing culture medium with exogenous taurine would present a similar BA profile to that produced in vivo. In addition, more HPH donors should be examined to ensure that these findings are universal.
These experiments also support the use of NHP models in favor of murine models when studying BA synthesis and metabolism, especially for comparison to human systems. Laboratory rats and mice offer the advantages of low costs, rapid growth, and straightforward handling and use, but their low homology to man makes correlation to human systems challenging and usually inaccurate (38)(39)(40)(41)(42)(43)45). In BA research, the use of murine models is additionally com-plicated by the utilization of a set of unique BA epimers, the muricholic acids (38,39,41,45). NHPs offer a more homologous, suitable system for the study of BA synthesis and metabolism, both in physiology and in illness, that much more closely approximates that in healthy adult humans. Further characterization in these species, however, to elucidate such processes as BA elimination (i.e., renal versus biliary), sulfonation and glucuronidation, and signaling for comparison to man, should be pursued to establish physiological parameters that may be altered in relevant pathologies The continued study of NHP BA synthesis with regard to C27 BA precursors and intermediates should prove valuable in aligning this laboratory model with human conditions. Additionally, the study of conjugated planar BAs (such as TACA and GACA) both in laboratory models and in human studies would be of increased relevance due to the propensity of BAs to be amidated or otherwise conjugated in most biological systems.
In summary, these studies have proven the utility of a validated method for the quantification of common and planar C24 BAs in a variety of systems. Detailed baselines have been established herein for multiple model systems for the study of BA production and metabolism, which have implications for the selection and use of cell and animal models for the study of BA metabolism and pathology.

Data availability
All data are contained within the article.