Quantification of All-Trans Retinoic Acid by Liquid Chromatography–Tandem Mass Spectrometry and Association with Lipid Profile in Patients with Type 2 Diabetes

Retinoic acids are vitamin A metabolites that have numerous essential functions in humans, and are also used as drugs to treat acne and acute promyelocytic leukemia. All-trans retinoic acid (atRA) is the major occurring metabolite of retinoic acid in humans. This study provides a sensitive and specific liquid chromatography–tandem mass spectrometry approach in order to quantify atRA in human plasma samples. The isolation of atRA by hyperacidified liquid–liquid extraction using hexane and ethyl acetate resulted in a recovery of 89.7 ± 9.2%. The lower limit of detection was 20 pg·mL−1, and 7 point calibration displayed good linearity (R2 = 0.994) in the range of 50–3200 pg mL−1. Selectivity was guaranteed by the use of two individual mass transitions (qualifier and quantifier), and precision and accuracy were determined intraday and interday with a coefficient variation of 9.3% (intraday) and 14.0% (interday). Moreover, the method could be used to isolate atRA from hyperlipidemic samples. Applying this method to plasma samples from patients with poorly controlled Type 2 diabetes significantly decreased atRA plasma levels as compared to those of the healthy controls. In addition, atRA concentrations were highly associated with increased low-density lipoprotein (LDL) and decreased high-density lipoprotein (HDL) cholesterol levels.


Introduction
Vitamin A is absorbed in the small intestine as retinol, and its metabolites have various physiological functions during development and for the maintenance of cellular homeostasis. Retinol can be oxidized to retinaldehyde, which is essential for vision, and further to retinoic acids, which play key roles in the regulation of the immune system, cell growth, and cell differentiation. Although several different retinoic acids are synthesized in cells, all-trans retinoic acid (atRA) is the major active metabolite of retinol, which mediates the majority of physiological functions [1]. AtRA is a lipophilic substance that acts through binding to DNA-bound nuclear receptors. This allows for the regulation of over 500 genes depending on cell type [2]. Physiological atRA levels are maintained by regulated synthesis through aldehyde dehydrogenases and degradation by enzymes of the cytochrome P450 family [3]. In clinical practice, atRA is used for the treatment of acne (topical application) and acute promyelocytic leukemia (systemically). In blood, atRA undergoes a rapid turnover, and only degradation products, such as 4-oxo-atRA, can be detected in urine [4,5]. It is likely that the metabolism of retinoids is affected or may display a causal role in the etiology of various diseases [1,3]. For example, the production rate of atRA is regulated by food intake and insulin [6], while atRA inhibits progression of diabetic nephropathy in a diabetes mouse model by preventing nuclear translocation of NF-κB in glomeruli and proximal tubules [7]. This indicates the potential role of atRA in the pathogenesis of diabetes mellitus and diabetic complications, which warrants further mechanistic investigations and association studies to determine whether the circulatory levels of atRA levels would be useful as biomarkers predicting diabetic complications.
Several studies over the last 30 years addressed atRA plasma levels in humans using different technical approaches [8][9][10][11][12]. However, such approaches were used for therapeutic drug monitoring in acute promyelocytic leukemia or for determining atRA plasma concentrations in healthy volunteers with less specific and sensitive techniques than what is currently possible [8,10,[13][14][15][16].
This study aimed at establishing a highly sensitive state-of-the art method in order to quantify plasma concentrations of atRA and to apply it to the analysis of patients with Type 2 diabetes. Here, we report a robust and simple extraction assay for atRA that can be applied to hyperlipidemic plasma samples, a common preanalytical challenge in laboratory medicine, particularly with respect to obese or diabetic patients [17,18]. The presented method allows for sufficient recovery, and sensitive and specific quantification. In addition, given atRA instability, this study aimed to assess the conditions under which atRA levels remain constant either in the blood or after extraction.

Fragmentation and Mass Transitions
The acquisition of atRA was achieved in the positive ionization mode of mass spectrometry. In the negative ionization mode of mass spectrometry, the abundance of precursor and product ion was~10-fold lower than that of positive ionization (Supplementary Figure S1). The precursor and product ion scan for atRA revealed fragments of 283.1 m/z (qualifier) due to water loss, and 122.9 m/z (quantifier) caused by the loss of the complete side chain of the carbonyl groups ( Figure 1A). This was confirmed by the fragmentation patterns of atRA-d5. However, the smaller fragment of 126.8 m/z (quantifier) displayed a loss in deuterated hydrogen ( Figure 1B). olites 2021, 11, 60 2 of 11 enzymes of the cytochrome P450 family [3].
In clinical practice, atRA is used for the treatment of acne (topical application) and acute promyelocytic leukemia (systemically). In blood, atRA undergoes a rapid turnover, and only degradation products, such as 4-oxo-atRA, can be detected in urine [4,5]. It is likely that the metabolism of retinoids is affected or may display a causal role in the etiology of various diseases [1,3]. For example, the production rate of atRA is regulated by food intake and insulin [6], while atRA inhibits progression of diabetic nephropathy in a diabetes mouse model by preventing nuclear translocation of NF-κB in glomeruli and proximal tubules [7]. This indicates the potential role of atRA in the pathogenesis of diabetes mellitus and diabetic complications, which warrants further mechanistic investigations and association studies to determine whether the circulatory levels of atRA levels would be useful as biomarkers predicting diabetic complications.
Several studies over the last 30 years addressed atRA plasma levels in humans using different technical approaches [8][9][10][11][12]. However, such approaches were used for therapeutic drug monitoring in acute promyelocytic leukemia or for determining atRA plasma concentrations in healthy volunteers with less specific and sensitive techniques than what is currently possible [8,10,[13][14][15][16].
This study aimed at establishing a highly sensitive state-of-the art method in order to quantify plasma concentrations of atRA and to apply it to the analysis of patients with Type 2 diabetes. Here, we report a robust and simple extraction assay for atRA that can be applied to hyperlipidemic plasma samples, a common preanalytical challenge in laboratory medicine, particularly with respect to obese or diabetic patients [17,18]. The presented method allows for sufficient recovery, and sensitive and specific quantification. In addition, given atRA instability, this study aimed to assess the conditions under which atRA levels remain constant either in the blood or after extraction.

Fragmentation and Mass Transitions
The acquisition of atRA was achieved in the positive ionization mode of mass spectrometry. In the negative ionization mode of mass spectrometry, the abundance of precursor and product ion was ~10-fold lower than that of positive ionization (Supplementary Figure S1). The precursor and product ion scan for atRA revealed fragments of 283.1 m/z (qualifier) due to water loss, and 122.9 m/z (quantifier) caused by the loss of the complete side chain of the carbonyl groups ( Figure 1A). This was confirmed by the fragmentation patterns of atRA-d5. However, the smaller fragment of 126.8 m/z (quantifier) displayed a loss in deuterated hydrogen ( Figure 1B).

Chromatography
The efficient separation of atRA was achieved using a reverse-phase C18 column. The positive ionization of the precursor ions was enhanced by the addition of formic acid. The use of a simple biphasic gradient composed of water and methanol resulted in a chromatogram with a stable peak shape and reproducibility, and a low noise signal ( Figure 2C,D). However, to avoid nonspecific peaks ( Figure 2C; RT,~9 min) due to a mild carry-over effect, five consecutive blanks composed of water and methanol (1:1) were run every 50 samples.

Chromatography
The efficient separation of atRA was achieved using a reverse-phase C18 column. The positive ionization of the precursor ions was enhanced by the addition of formic acid. The use of a simple biphasic gradient composed of water and methanol resulted in a chromatogram with a stable peak shape and reproducibility, and a low noise signal ( Figure 2C,D). However, to avoid nonspecific peaks ( Figure 2C; RT, ~9 min) due to a mild carry-over effect, five consecutive blanks composed of water and methanol (1:1) were run every 50 samples.

Analytical Specificity
The purity verification of atRA and atRA-d5 was provided by the supplier company. The two product ions with the highest intensity were selected as quantifier and qualifier (Table 1). Using artificial human plasma (Biseko ®, Dreieich, Germany) without endogenously produced atRA, no coeluting compounds were found, which interfered with the detection of either quantifier or qualifier. However, the possibility that any medications could interfere with our assay could not be excluded. Following the optimization of multiple reaction monitoring (MRM), several injections of spiked artificial plasma showed that the fragmentation patterns for qualifier and quantifier of atRA and atRA-d5 were reproducible and stable over a long period of time with acceptable accuracy and precision (n ≥ 100 injections, Table 2).

Analytical Specificity
The purity verification of atRA and atRA-d5 was provided by the supplier company. The two product ions with the highest intensity were selected as quantifier and qualifier (Table 1). Using artificial human plasma (Biseko ®, Dreieich, Germany) without endogenously produced atRA, no coeluting compounds were found, which interfered with the detection of either quantifier or qualifier. However, the possibility that any medications could interfere with our assay could not be excluded. Following the optimization of multiple reaction monitoring (MRM), several injections of spiked artificial plasma showed that the fragmentation patterns for qualifier and quantifier of atRA and atRA-d5 were reproducible and stable over a long period of time with acceptable accuracy and precision (n ≥ 100 injections, Table 2).

Linearity and Determination Limits
For the 7 point calibration, the ratio between atRA and atRA-d5 was used for quantification. The obtained calibration coefficients (R 2 ) for 7 point calibration (50-3200 pg·mL −1 ) were > 0.99 ( Figure 1A). The achieved lower limits of detection and quantification were 20 and 50 pg·mL −1 , respectively ( Figure 1A,B).

Recovery, Precision, and Accuracy
The extraction recovery of atRA-d5 in plasma was 89.7 ± 9.2% (Table 2). Methanol precipitation and acidification with hydrochloric acid resulted in the best recovery as compared to other precipitation or acidification reagents such as trichloroacetic acid, ethanol, acetonitrile, or formic acid (data not shown). The precision of the replicate analyses was evaluated for 50, 600, and 2500 pg·mL −1 . The coefficient of variation (CV) was intraday at 9.3% and interday at 14.0% (Table 2). Accuracy achieved a mean value of 96.5% on intraday basis and a mean of 101.2% on interday basis ( Table 2). The long-term accuracy and precision of artificial plasma (n ≥ 100 samples spiked with varying atRA concentration) were 94.0 ± 14.2% and 17.8% CV, respectively ( Table 2).

Stability
Following storage in darkness under various conditions, samples were quantified and compared to the quantification directly after extraction. The storage of plasma for 1 week at 4 • C was associated with a significant reduction in atRA (~51%; Table 3). Postprocessing stability in an assay buffer revealed that atRA was degraded when stored for 1 week at 4 • C. Several freeze-thaw cycles were not associated with a significant reduction in atRA (~14% reduction; Table 3). Furthermore, exposure to 254 nm UV light for 3 h was associated with~40% reduction in atRA (Supplementary Table S1).

Clinical Application
This method was developed to quantify atRA in human plasma samples. A total of 14 plasma samples had triglyceride levels above 300 mg·dL −1 . The best recovery in those samples was achieved following strong acidification with HCl (10 M), liquid-liquid extraction with a combination of ethyl acetate/hexane (1:1), and the incubation of the sample at 4 • C for 20 min. No acidification and the usage of different extraction solvents were associated with poor recovery in those hyperlipidemic samples (data not shown). For a constant extraction procedure, this method was applied to all plasma samples.
Using the developed LC-MS/MS method, plasma levels of atRA were found to be in the range of 0.57 to 2.76 ng·mL −1 in all patients. A cohort of patients with poorly controlled Type 2 diabetes displayed hyperlipidemia, increased C-reactive protein, and blood glucose concentration, as well as excessively increased HbA1c (Table 4). In patients with poorly controlled Type 2 diabetes, atRA levels were significantly lower, with 1.38 ± 0.42 ng·mL −1 , as compared to healthy individuals with 1.77 ± 0.45 ng·mL −1 (p < 0.01; Figure 3A). In addition, plasma atRA levels were strongly associated with the lipid profiles of the whole cohort (n = 44), especially for high-density lipoprotein (HDL; r = 0.598; p < 0.001) and low-density lipoprotein (LDL) cholesterol (r = −0.469; p < 0.01), and also for body-mass index (BMI) and triglyceride concentration ( Figure 3B-F). In a cohort of patients with well-adjusted Type 2 diabetes, atRA levels decreased (1.47 ± 0.37 ng·mL −1 ), but this was statistically not significant (p = 0.053 vs. controls; Supplementary Table S2). Table 4. Mean baseline characteristics of control and diabetic patient cohorts. All parameters were determined prior to collection. Data are mean ± SD; *** p < 0.001, vs. controls. All other characteristics were not significant (p > 0.05).

Discussion
This study offers a sensitive and specific liquid chromatography, tandem mass spectrometry (LC-MS/MS) approach for the quantification of atRA in hyperlipidemic human plasma samples. Within this context, this method for the first time describes the usage of 122.9 m/z for the quantification of atRA in biological matrices, which may be a result of a more efficient fragmentation in the collision cell as compared to that of other studies [6,8,19]. As previously described, the separation and ionization of atRA were achieved using a reverse-phase C18 column and formic acid [8,9,14,20]. With 20 and 50 pg·mL −1 as the lower limits of detection and quantification, this method is more sensitive than other published methods in the field that use LC-MS and high-performance liquid chromatography (HPLC) approaches [9][10][11]15,16]. Ex vivo, atRA is chemically unstable, and readily oxidized and isomerized when exposed to heat, light, or oxygen [15,16],

Discussion
This study offers a sensitive and specific liquid chromatography, tandem mass spectrometry (LC-MS/MS) approach for the quantification of atRA in hyperlipidemic human plasma samples. Within this context, this method for the first time describes the usage of 122.9 m/z for the quantification of atRA in biological matrices, which may be a result of a more efficient fragmentation in the collision cell as compared to that of other studies [6,8,19]. As previously described, the separation and ionization of atRA were achieved using a reverse-phase C18 column and formic acid [8,9,14,20]. With 20 and 50 pg·mL −1 as the lower limits of detection and quantification, this method is more sensitive than other published methods in the field that use LC-MS and high-performance liquid chromatography (HPLC) approaches [9][10][11]15,16]. Ex vivo, atRA is chemically unstable, and readily oxidized and isomerized when exposed to heat, light, or oxygen [15,16], which reflects the necessity for stability analysis. Pre-and postprocessing analyses revealed that atRA had sufficient short-term stability, but degraded when stored for 1 week at 4 • C, which is in line with previous studies on atRA stability [21]. Exposure to 254 nm UV light for 3 h or longer was associated with a significant reduction in atRA. Therefore, the usage of amber reaction tubes and the avoidance of direct sunlight during analysis are strongly recommended.
As a lipophilic compound, atRA presents challenges during extraction, particularly in patients with an altered lipid profile displayed by increased triglycerides, LDL, and total cholesterol. With plasma levels of atRA between 0.57 and 2.76 ng·mL −1 in all patients, our results were in line with those of other studies in healthy individuals, where 0.3-5 ng·mL −1 of atRA was found to be present in the plasma and serum [8][9][10]15,16]. In patients with poorly controlled Type 2 diabetes, lower atRA levels than those of healthy individuals were found. However, a cohort of patients with well-adjusted Type 2 diabetes displayed only a minor decrease in atRA. Future studies should address whether atRA could be causally involved in the progression of diabetes mellitus.
In all analyzed individuals, the plasma atRA levels were strongly associated with lipid profiles and body-mass index (BMI). This confirms a recent study in which atRA was claimed to be a predictive marker for the development of metabolic syndrome, which is strongly associated with the development of Type 2 diabetes and cardiovascular disease [22]. Experimental diabetic model systems showed that atRA can block adipocyte differentiation [23]. Furthermore, atRA affects insulin secretion, which is not surprising given the ubiquitous expression of the retinoic X receptor [24]. Whether the reduction in atRA plasma concentration in dyslipidemic patients is the cause or effect of lipid dysregulation remains unclear and should be addressed in further studies.

Preparation of Calibration Standards
Stock solutions of atRA and atRA-d5 were prepared in dimethyl sulfoxide at concentrations of 500 (atRA) or 100 µg·mL −1 (atRA-d5) and stored at −80 • C. Working solutions of atRA were prepared in methanol in the range of 500-100 µg·mL −1 and stored at −20 • C. Calibration standards of atRA were 50, 100, 200, 400, 800, 1600, and 3200 pg·mL −1 with the addition of 500 pg·mL −1 of atRA-d5 for each calibrator. All stock and working solutions were kept in amber reaction vials to avoid any degradation by ultraviolet light.

Sample Collection and Extraction
Ethylenediaminetetraacetic acid (EDTA) plasma samples were obtained from 20 healthy controls and 39 patients with Type 2 diabetes. All participants were in a fasting state, and patients with Type 2 diabetes were enrolled if diagnosis had been established according to the guidelines of the German Diabetes Association [26]. The study was ap-proved by the ethics committee of Heidelberg University Hospital (Project Identification Code S-383/2016). All patient material and data were acquired with formal written informed consent and in agreement with the guidelines of the ethics committee as previously described [19]. The plasma supernatant was directly aliquoted following centrifugation (5 min at 5000× g; 4 • C) of the whole blood and frozen at −80 • C. Samples were thawed once, and liquid-liquid extraction was carried out under reduced light condition in amber tubes. The internal standard (100 pg of atRA-d5) was added to 200 µL of plasma, which was acidified with 5 µL of hydrochloric acid (10 M). Afterwards, 400 µL of methanol was added for protein precipitation. After centrifugation (10 min at 10,000× g; 4 • C), the supernatant was transferred into a new amber tube and 300 µL of hexane and 300 µL of ethyl acetate was added, which was then mixed for 10 s. The mixture was left for 20 min at 4 • C in darkness, and the aqueous and organic phase was separated by centrifugation (10 min at 10,000× g; 4 • C). The upper organic phase was evaporated using a vacuum concentrator (Eppendorf Concentrator Plus) at room temperature. The residue was resuspended in 60 µL of methanol/water (2:1) and transferred into an HPLC vial (Waters, Eschborn, Germany).

Chromatography
All analyses were performed on a Waters ® Acquity UPLC class I system (Waters, Eschborn, Germany) equipped with a binary solvent-delivery system with an online degasser and a column manager containing a column oven connected to a UPLC autosampler. AtRA was separated by reverse-phase LC on a Waters ® Acquity BEH C18 column (1.7 µM, 2.1 × 50 mm) at a flow rate of 0.2 mL·min −1 and a column temperature of 20 • C. During analyses, all samples were stored in the autosampler at a temperature of 4 • C, and the injection volume for each sample varied between 10 and 30 µL. Solvent A consisted of 0.1% formic acid in water, and Solvent B was 0.1% formic acid in methanol. For each run, gradient elution was performed. Solvent A was decreased from 80% to 0% (0-4 min), remained isocratic at 0% (4-6 min), and increased back to 80% (6-15 min). The column eluent was directed into the MS, and analyses were performed using MassLynx XS software.

Mass Spectrometry
AtRA detection was performed on a XEVO TQ-S tandem quadrupole mass spectrometer (Waters ® ) equipped with an electrospray ionization source (ESI) operated in positive ion mode. Analyte detection was performed using multiple reaction monitoring (MRM). Source parameters were set as follows: capillary voltage, 3.5 kV; desolvation temperature, 350 • C; desolvation gas flow, 850 L·h −1 ; source temperature, 200 • C; cone gas flow, 250 L·h −1 ; collision gas flow, 0.15 mL·min −1 ; nebulizer gas flow, 5 bar. Cone and collision voltage were individually optimized for atRA and atRA-d5 and are summarized in Table 1. Acquisition and quantification were completed with MassLynx 4.1 and TargetLynx 2.7.

Validation Procedure
The method was validated for selectivity, matrix effects, linearity, lower limits of detection (LLOD) and quantification (LLOQ), recovery, stability, precision, and accuracy (intra/interday). Seven-point calibration was performed using linear regression. For all concentration calculations, the area ratio of atRA/atRA-d5 was plotted against nominal calibrator concentration. Blank samples (without analytes) were measured for each calibration curve to ensure reliability. For the determination of recovery, LLOD, LLOQ, precision, and accuracy, 200 µL of spiked artificial plasma was used. Recovery, precision, and accuracy were validated in an intraday assay using three different concentrations (50, 600, 2500 pg·mL −1 ), which were measured in quadruplicates. For interday variability, the same concentrations (50, 600, 2500 pg·mL −1 ) of spiked artificial plasma were quantified on 3 consecutive days in quadruplicates. LLOD and LLOQ were determined with a signalto-noise ratio of 6.4 (LLOQ) and 3.2 (LLOD). Stability was validated in human artificial plasma (preprocessed) and in assay buffer (postprocessed) at various temperature levels and for different durations by spiking 200 µL of plasma sample with 1000 pg of atRA.
AtRA stability was also assessed for exposure to UV light (254 nm) using transparent tubes. Matrix effects were defined as a suppression of or increase in signal intensity for the chosen MRMs (matrix effects while ionization) or as an increase or decrease in recovery of atRA-d5 (matrix effect while extraction).

Clinical Chemistry
Blood was drawn under fasting conditions as described above, and samples were immediately processed in the accredited Central Laboratory of Heidelberg University Hospital. Plasma levels of total cholesterol, HDL cholesterol, triglycerides, blood glucose, and C-reactive protein were analyzed with clinical chemistry automation (AdviaXPT ® 2400 chemistry analyzer, Siemens Healthineers, Erlangen, Germany) according to the appropriate standard operating protocol. HbA1c quantification was performed on whole-blood EDTA samples, and analyses were executed using a VARIANT II HbA1c analyzer (BioRad, Hercules, CA, USA) that utilizes cation-exchange high-performance liquid chromatography. LDL cholesterol was determined using the Friedewald formula as previously described [27].

Statistical Analysis
Statistical-data analysis was performed using GraphPad Prism 7 (GraphPad Software Inc., San Diego, CA, USA). All data are expressed as mean values ± standard error (SE), and were analyzed for significance using an unpaired t test with Welch's correction. Linear regression and Pearson's correlation coefficient were used to study the association of atRA and BMI, total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides.

Conclusions
In this study, atRA was quantified in human-plasma samples using a robust and sensitive LC-MS/MS method that achieved convincing analytical parameters of accuracy, precision, recovery, and stability. Within this context, atRA was isolated from hyperlipidemic human-plasma samples, and results showed a decrease in atRA levels in poorly controlled patients with Type 2 diabetes, and a strong correlation with LDL and HDL cholesterol. In summary, this study adds valuable content for further analytical determinations of atRA using LC-MS/MS, and indicates altered metabolism of retinoids in patients with diabetes mellitus.
Supplementary Materials: The following are available online at https://www.mdpi.com/2218-198 9/11/1/60/s1. Figure S1: Comparison of ionization mode of mass spectrometry for atRA (100 pmol direct injection on column), Table S1: Assessment of atRA stability upon UV-exposure, Table S2: Mean baseline characteristics and atRA levels of the control group (as shown in Table 4) and patients with well-adjusted / poorly controlled type 2 diabetes. Funding: This research was funded by the Deutsche Forschungsgemeinschaft, project number 394046768 for the SFB1366 project C4 to A.F., and SFB 1118 and GRK 1874 to T.F. and P.N.).

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by Heidelberg University Hospital (S-383/2016, September 2016).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy and/or ethical concerns.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.