Analysis of Vitamin A and Retinoids in Biological Matrices

Abstract

Vitamin A signaling pathways are predominantly driven by the cellular concentrations of all-trans-retinoic acid (atRA) as the main mechanism of retinoid signaling is via activation of retinoic acid receptors. atRA concentrations are in turn controlled by the storage of vitamin A and enzymatic processes that synthesize and clear atRA. This has resulted in the need for robust and highly specific analytical methods to accurately quantify retinoids in diverse biological matrices. Tissue-specific differences in both the quantity of retinoids and background matrix interferences can confound the quantification of retinoids, and the bioanalysis requires high performance instrumentation, such as liquid chromatography mass-spectrometry (LC-MS). Successful bioanalysis of retinoids is further complicated by the innate structural instability of retinoids and their relatively high lipophilicity. Further, in vitro experiments with retinoids require attention to experimental design and interpretation to account for the instability of retinoids due to isomerization and degradation, sequential metabolism to numerous structurally similar metabolites, and substrate depletion during experiments. In addition, in vitro biological activity is often confounded by residual presence of retinoids in common biological reagents such as cell culture media. This chapter identifies common biological and analytical complexities in retinoid bioanalysis in diverse biological matrices, and in the use of retinoids in cell culture and metabolic incubations. In addition, this chapter highlights best practices for the successful detection and quantification of the vitamin A metabolome in a wide range of biological matrices.

1. Vitamin A and retinoids

Vitamin A (retinol) is a dietary fat-soluble vitamin obtained from animal sources as retinyl esters (RE) while retinaldehyde can be synthesized from provitamin A β-carotene (Figure 1). Retinol is metabolized in the body (Figure 1) to retinyl esters for storage and to the important active signaling molecule all-trans retinoic acid (atRA) (Blaner et al., 2016; Harrison & Dela Seña’s, 2015). atRA and its metabolite 4-oxo-RA, are potent transcriptional activators of nuclear retinoic acid receptors (RARs) (Larange & Cheroutre, 2016; Topletz et al., 2015). atRA is believed to be responsible for the biological activity of vitamin A, such as regulating expression of genes important in reproduction and development (Clagett-Dame & Knutson, 2011; Duester, 2008; Hogarth et al., 2011), in energy homeostasis (Blaner, 2019; Chen & Chen, 2014; Zhao et al., 2012), in maintenance of epithelia (McCullough et al., 1999), in regulation of cell cycle and differentiation (Liu et al., 2014; Mezquita & Mezquita, 2019) and in immunity (Grizotte-Lake et al., 2018; Iwata et al., 2004; Larange & Cheroutre, 2016). The fact that the activation of RARs is driven by changes in ligand (atRA) concentrations, has resulted in a need for robust analytical methods to detect and quantify retinoids. Historically, retinoid analysis has relied on the signature UV absorbance of retinol and retinyl esters at the absorbance maximum between 325–383 nm (Napoli, 1986), and the UV and fluorescence based measurements of retinol have been possible due to the high circulating concentrations (1–3μM) of retinol (Gatti et al., 2000; Schmidt et al., 2003). Similarly, in the liver the concentrations of retinol and retinyl esters, predominantly in the form of retinyl palmitate, range from 1.2–161 nmol/g liver and 12.6 nmol/g-42.8 μmol/g liver, respectively (Zhong, Kirkwood, et al., 2019), making their UV detection feasible. In contrast, atRA concentrations are over 100-fold lower in circulation and tissues with measured concentrations around 1–10 nM making quantitative measurements of atRA challenging. Additionally, in several matrices multiple isomers of RA are present, such as 13-cisRA (Figure 1), requiring robust chromatographic separation to distinguish structurally similar analytes (Arnold et al., 2012). Likely due to these challenges, atRA and its metabolites have historically been rarely measured in animal or human tissues and plasma. The advances in liquid chromatography- mass spectrometry (LC-MS) technologies over the past 10–15 years have, however, enabled the detection and quantification of atRA in various tissues (Kane et al., 2005; Kane, Folias, Wang, et al., 2008). In addition, improvements in LC column quality and ultra-high-pressure liquid chromatography (UHPLC) have allowed development of methods that reproducibly separate the isomers of RA. With these advances, measurement of retinoid concentrations in tissues has become feasible for a trained analytical or bioanalytical chemist, and methods to quantify a variety of retinoids including retinol, retinyl esters (RE), atRA and its isomers, 4-oxo-RA and stereoisomers of 4-OH-RA have been developed (Arnold et al., 2012; Jones et al., 2015; Shimshoni et al., 2012; Zhong, Kirkwood, et al., 2019). However, even with the profound analytical advancements, sample handling and preparation and sample quantity are critical determinants for successful bioanalysis of retinoids. This chapter will cover retinoid bioanalytical methods with an emphasis on best-practices to minimize inherent concerns related to analyte instability, biological matrix effects and ion suppression, matrix and/or reagent interferences, and substrate depletion.

Figure 1:

Vitamin A metabolic pathway with key retinoid metabolite structures and eznymes contributing to the specific steps of atRA biosynthesis.

2. Stability of retinoids in stock solutions and samples

2.1. Stability of all-trans retinoic acid in solutions

Light sensitivity (instability) (Napoli, 1986) and sensitivity to heat and oxidation are well documented characteristics of retinoids (Runge & Heger, 2000). To avoid degradation, retinoids in powder form should be stored in a desiccator at −80°C in dark amber ampules or vials under inert gas. Similarly, stock solutions of retinoids should be stored in dark amber vials at −80°C preferably under inert gas. Since retinoids are rapidly degraded and practically insoluble in aqueous solutions, stocks should be prepared in organic solvent for maximum long-term stability. The reported solubility of atRA in common organic solvents is summarized in Table 1. As retinoids have an extensive conjugated double bond structure (Figure 1), they have a signature UV absorbance peak at wavelengths 325–383 nm. This absorbance enables the use of HPLC-UV to monitor the stability of stock solutions and simultaneously detect isomerization and/or formation of auto-oxidation products (Figure 2) (Napoli, 1986). Due to the sensitivity of retinoid stocks to handling, the stability of stock solutions should be monitored routinely before use. Figure 2A–B show the stability of atRA at −80°C in the three common solvents often used in biological analyses. While atRA was stable in ethanol and methanol, it degraded in DMSO over 3 weeks as shown by the emergence of degradation products in the chromatograms. This strongly suggests that if atRA is used in cell culture in DMSO, the solutions should be prepared fresh for treatments. Overall, researchers should confirm the stability of their retinoid stocks if stock solutions are stored for longer than a month. The degradation of atRA stocks was also assessed in aqueous buffers and 5–30% degradation was observed after 30 minutes at 37°C (Figure 2C) suggesting that retinoids should not be kept in aqueous solutions and buffers for prolonged periods of time. Further, as previously shown (Kane & Napoli, 2010), atRA degraded rapidly to various isomerization products after just 1 hour of light exposure in clear vials. Isomerization was minor after 1 hour of light exposure when retinoids were kept in amber vials, but significant isomerization was detected after 6 hours even in amber vials (Figure 2D). The rapid light-induced degradation of atRA is consistent with the observed instability in human serum when samples were not protected from light during collection (Arnold et al., 2012) and shows that work with retinoids should always be conducted under yellow lights.

Table 1:

Solubility of all-trans-retinoic acid in common organic solvents as reported by major chemical suppliers.

	Sigma	Cayman Chemicals	Santa Cruz	Toronto Research Chemicals
DMSO	40 mg/mL	20 mg/mL	25 mg/mL	Slightly
Methanol	--	--	--	Slightly
Ethanol	2.7 mg/mL*	0.5 mg/mL	10 mg/mL	--
Acetone	--	--	30 mg/mL	--
Chloroform	Slightly	--	40 mg/mL	Slightly
Ethyl Acetate	--	--	--	Slightly
Dimethylformamide	--	20 mg/mL	20 mg/mL	--

^*95% ethanol, -- solubility not reported

Figure 2: Stability of atRA in organic solvents, aqueous buffers and with light exposure.

HPLC-UV chromatograms of stock solutions of 1 mg/mL of atRA in ethanol, methanol or DMSO analyzed at 0 weeks (A) and after 2 weeks (B) of storage in −80°C in amber vials are shown. Panel (C) shows the analysis of atRA stability in aqueous buffers. The stability of atRA (100 nM) was monitored over 30 minutes at 37°C (common incubation temperature) in five different aqueous buffers: 100 mM potassium phosphate buffer, pH 7.4 (Kpi); Kpi with 0.5 mM dithiothreitol (DTT); Kpi with 0.1mM ethylenediaminetetraacetic acid (EDTA); phosphate buffered saline (PBS); and 10mM Hepes, 100mM potassium chloride, 0.1mM EDTA, 0.5mM DTT, pH 7.4 (HEDK). The effect of light exposure on atRA (1 mg/mL in ethanol) stability in amber or clear glass vials is shown in (D). The solutions of atRA were directly exposed to ambient light (standard laboratory fluorescent lights) for an hour or six hours and the formation of degradation products and isomerization was monitored by HPLC-UV. In panels A, B, and D the Y-axis baselines are offset for visual comparison of the appearance of degradation products.

Serum-free conditions are commonly used in cell culture, but consistent with the instability in aqueous solutions, retinoids have been reported to be unstable and prone to isomerization in the absence of protein-binding in culture (Sharow et al., 2012; Tsukada et al., 2002). Specifically, a recent study quantified the recovery of retinoids at 24 hours in cell-free incubations in serum-containing and serum-free culture medium (Sharow et al., 2012). In the absence of serum, atRA recovery was <30% of the starting concentration, with more than half of the total RA observed as isomers. Furthermore, retinol recovery was low (<5% recovery) in the cell-free incubations lacking serum protein due to the formation of unidentified degradation products. The recoveries of all the analytes were significantly higher in parallel cell-free incubations using serum-supplemented medium. Similarly, supplementation of serum-free medium with 6 mg/mL BSA resulted in improved recovery of all retinoids at 24 hours, although atRA appeared to be more stable in these experiments than retinol (Sharow et al., 2012). Taken together, these experiments suggest that cell culture media should be prepared fresh prior to experiments and retinoids should not be stored in cell culture media.

2.2. Stability of retinoids in biological matrices

While the stability of retinoids in aqueous solutions is compromised by light and storage conditions, biological matrices such as serum described above may stabilize retinoids from isomerization and oxidation allowing for storage of biological samples without significant compromise in sample quality. For example, only minor (<20%) light-induced retinoid depletion and isomerization was observed in human liver homogenates when monitored for up to 1 hour (Zhong, Kirkwood, et al., 2019). This contrasts with human serum which is susceptible to light-induced retinoid degradation (Arnold et al., 2012). In addition, for the analysis of human or mammalian blood, serum, or plasma the initial sample exposure to ambient or clinic lighting during sample collection should be minimized. As blood draws under yellow lights are usually not feasible, it is best practice to protect samples from light by wrapping the tubes with aluminum foil and/or using covered ice containers. The initial light protection is likely more significant to retinoid recovery than the sample preparation time needed for isolating serum or plasma and the ultimate storage time of the samples. As shown in Figure 3A, the time from the blood draw to the separation of light protected plasma by centrifugation did not impact the concentration (peak area ratio relative to the corresponding internal standard) of retinol, atRA, or 13cisRA. This finding agrees with similar studies which have demonstrated that retinol measurements from plasma are not impacted significantly by sample preparation spin-delay (Abraham et al., 2019; Craft et al., 1988; Cuerq et al., 2015; Hankinson et al., 1989).

Figure 3: The stability of retinol, atRA and 13-cisRA in blood and plasma.

Panel (A) shows the effect of time from blood draw to separation of plasma from blood on retinoid stability. The blood was drawn to heparin vacutainer tubes (green top), immediately light protected with foil, and inverted. Blood samples were kept on ice for 0, 15, 30, and 60 minutes post-blood draw, followed by centrifugation for 10 min at 1,000 × g to separate plasma. Panels (B-D) show the analysis of storage conditions on retinol (B), atRA (C) and 13-cisRA (D) recovery from human plasma samples. For these analyses aliquots of plasma samples collected above were stored at −20°C or −80°C for up to 90 days. At designated time points, aliquots were thawed and analyzed by LC-MS/MS. For analysis, samples were spiked with an IS mixture (atRA-d₅, 13-cisRA-d₅, and ROL-d₈), protein precipitated with acetonitrile (1:1 v/v), centrifuged and analyzed as described in the protocols section. Analyte peak areas were normalized to their respective IS. Data is shown as means ± SD for three replicates per time point and peak area ratios are presented as % of t = 0 days. Shaded regions denote ±15% around t=0 starting measurement to indicate range of acceptable analytical variability.

Due to the inherent lability of retinoids in aqueous solutions, the stability of retinoids in biological samples upon storage is often of concern. Yet, as shown in Figure 3B–D, when human plasma samples were stored at −20°C or −80°C up to three months the storage temperature and time had minimal impact on the stability of the three analyzed retinoids. Similar data has been reported demonstrating that serum retinol was stable for five months when stored at −20°C and for 28 months at −70°C (Craft et al., 1988). If sample stability is of concern, stable-labeled retinoids, such as atRA-d₅, can be added to samples prior to storage to monitor for isomerization or degradation in the samples during storage (Arnold et al., 2012). However, this approach may result in challenges in quantitative analyses due to the presence of isotope labelled compounds in the samples. Notably, when we repeated an analysis of atRA concentrations in a set of liver samples stored in −80°C for one year, the concentrations measured after a year-long storage were within analytical variability, further suggesting that retinoids in biological matrices are stable when properly stored and handled.

3. Considerations in the bioanalysis of retinoids in tissue

3.1. Analytical considerations

Quantification of endogenous tissue atRA concentrations is challenging, due to the low concentrations of atRA, poor ionization and extensive fragmentation in the MS, and need for powerful separation due to existence of retinoic acid isomers (Arnold et al., 2012; Jones et al., 2015; Zhong, Kirkwood, et al., 2019). In contrast to atRA, retinol and retinyl ester concentrations are relatively high in tissues, especially in the liver, making analysis and quantification feasible by a variety of methods. A major challenge in the development of bioanalytical methods for retinoid analysis is the lack of biological tissues devoid of retinoids that could be used as a blank matrix. Due to the lack of a blank matrix, the identity of the presumed retinoid peaks even in spiked samples cannot be easily confirmed and the recovery of retinoids or magnitude of matrix effects in given tissue cannot be determined. Modern LC-MS methods have improved the confidence in peak identification via incorporation of MS³ analysis (MS/MS/MS) into analyte identification in addition to multiple reaction monitoring (MRM) (Arnold, Kent, Hogarth, Schlatt, et al., 2015; Jones et al., 2015; Zhong, Kirkwood, et al., 2019). This is critical as several matrix components elute close to or simultaneously with atRA or its isomers in LC-MS/MS runs and show identical MS/MS transitions with atRA. As illustrated in Figure 4 for several tissues, other matrix components can easily be mis-analyzed as atRA unless careful attention is paid to peak identification and confirmation. This is less of a concern in LC-MS analysis of retinol and retinyl esters due to their much higher concentrations than atRA in most tissues (Figure 5).

Figure 4. Detection of atRA and 13-cisRA in various mouse tissues by different mass spectrometric methods.

Panels (A-D) show the detection of endogenous atRA and 13-cisRA and their corresponding deuterium labeled internal standards (atRA-d₅ and 13-cisRAd₅) by multiple reaction monitoring (MRM) from liver (A), spleen (B), small intestine (C) and skin (D) of mice fed normal chow. The blue lines show the internal standard trace (transition m/z 306>208) and black lines show the endogenous atRA and 13-cisRA trace (transition m/z 301>205). Panels E-H show the corresponding MS³ traces for liver (E), spleen (F), small intestine (G) and skin (H) for the samples analyzed in A-D. All samples were processed with the liquid-liquid extraction method described in section 6. The identification and confirmation of the correct peak for atRA in MRM chromatograms for quantification relies on IS retention time together with the atRA peak shown in MS³ chromatogram.

Figure 5. Detection of retinyl palmitate and retinol in various mouse tissues by LC-MS/MS.

The detection of retinyl palmitate and retinyl oleate from mouse liver (A), small intestine (B) and spleen (C) is shown with the black trace depicting MS/MS transition m/z/ 269>93 and the blue trace depicting transition m/z 269>95. The internal standard (retinyl palmitate-d₄) is shown as a red trace depicting MS/MS transition m/z 273>94. The detection of retinol from mouse liver (D), small intestine (E) and spleen (F) is shown with the black trace depicting MS/MS transition m/z 269>93 and the blue trace depicting transition m/z 269>95. The internal standard (retinol-d₈) is shown as a red trace depicting MS/MS transition m/z 277>98. All samples were prepared using liquid-liquid extraction as described in section 6.

Different biological tissues may contain different endogenous compounds like hormones and lipids as is clearly evident by comparison of the interfering peaks within the chromatograms from different tissues (Figure 4). Therefore, accurate identification and quantification of retinoids relies heavily on good separation of these interfering peaks from the target analyte peak. Even within the same organ but from different donors, the lipid content and concentrations of various endogenous compounds is expected to vary, causing matrix related differences in retinoid recovery, observable interferences in chromatograms and ion suppression occurring in the MS. For example, in three independent experiments, we compared the recovery/ion suppression of isotope-labeled internal standards (IS), atRA-d₅ and 13cisRA-d₅, between serum, liver, skin, and colon tissues (Figure 6). The recovery/ion suppression following extraction of atRA-d₅ and 13cisRA-d₅ was similar between liver and serum, but the overall signal for atRA-d₅ and 13cisRA-d₅ was 30–50% lower in skin and colon when compared to serum (Figure 6). Similarly, the signal for retinyl palmitate-d₄ in the colon and small intestine was about 30% of the signal for the same concentration in serum. In contrast, the retinol-d₈ signal was similar regardless of the matrix (Figure 7), demonstrating that the matrix effect is retinoid specific. These tissue-dependent differences in the signals are likely due to ion suppression caused by coeluting matrix components but could also be due to different absolute recovery of the retinoids in the different tissues. Hence, to correct for this variability in ion suppression/recovery, use of isotope labelled internal standards (e.g. retinol palmitate-d₄) and LC-MS/MS analysis is highly recommended. Chemically different internal standards such as retinyl acetate will not correct for tissue specific ion suppression due to the different retention times and likely will not have identical extraction characteristics as endogenous retinyl esters. Isotope labeled internal standards are compatible with LC-MS/MS methods, are structurally identical with the target analyte, and thus, serve as powerful tools in correcting for matrix effects. This concept was shown in a previous study demonstrating that by using an isotope-labeled internal standard, the slopes of retinoid standard curves built with charcoal treated blank human serum and mouse liver homogenates were similar (Zhong, Kirkwood, et al., 2019). In addition, retention time similarity between the target retinoid peak and the isotope labeled internal standard peak is helpful in identifying the analyte peak for quantification, especially when multiple interfering peaks are present in the chromatograms (Figure 4).

Figure 6. The impact of tissue matrix on retinoic acid detection.

Each panel represents an independent experiment in which atRA and 13-cisRA were measured in mouse tissues (liver, colon and skin). Known amount of atRA and 13-cisRA were spiked into charcoal treated serum to make standard curves and quality control (QC) samples for quantification. Each open circle represents a target tissue sample obtained from an individual mouse and each filled circle represents a RA-spiked serum sample. Within one independent measurement same amount of IS mixture (atRA-d₅ and 13cisRA-d₅) was added into all samples (RA-spiked serum samples and tissue samples). Samples were processed with a liquid-liquid extraction method as described in section 6 and analyzed with LC-MS/MS methods described in Table 3. The LC-MS injection volume was the same for all samples. The data is reported as the percentage that the IS peak area in each individual tissue sample is of the average IS peak area in the RA-spiked serum samples. In the absence of any matrix effect and with uniform recovery from each tissue the percentage value should be 100% on average. Panels A-C are for atRA-d₅ and panels D-E are for 13cisRA-d₅. Blue bars indicate mean values for each group.

Figure 7. The impact of tissue matrix on retinol and retinyl palmitate detection.

Each panel represents an independent experiment in which retinol and retinyl palmitate were measured in mouse tissues (colon and small intestine (SI)). Known amount of retinol and retinyl palmitate was spiked into charcoal treated serum to make standard curves and quality control (QC) samples for quantification. Each open circle represents a target tissue sample obtained from an individual mouse and each filled circle represents a retinoid-spiked serum sample. For each independent measurement the same amount of IS mixture (retinol-d₈ and retinyl palmitate-d₄) was added into all samples (retinoid-spiked serum samples and tissue samples). Samples were processed with a liquid-liquid extraction method as described in section 6 and analyzed with LC-MS/MS methods described in Table 3. The LC-MS injection volume was the same for all samples. The data is reported as the percentage that the IS peak area in each individual tissue sample is of the average IS peak area in the retinol and retinyl ester-spiked serum samples. In the absence of any matrix effect and with uniform recovery from each tissue the percentage value should be 100% on average. Panels A and B are for retinol-d₈ and panels C and D are for retinyl palmitate-d₄. Blue bars indicate mean values of normalized IS peak area.

3.2. Biological considerations

Endogenous retinoids are most commonly measured in mouse and human tissues. Mouse models are often used for retinoid-related studies and retinoids have been detected and quantified in a variety of mouse tissues including serum, liver, adipose, kidney, heart, spleen, skin, brain and testis (Arnold et al., 2012; Arnold, Kent, Hogarth, Schlatt, et al., 2015; Obrochta et al., 2014; Schmidt et al., 2003; Zhong, Hogarth, et al., 2019). Mouse liver retinyl palmitate concentration is the highest among all the retinoids in various tissues, at the range of 100–2,000 nmol/gram liver while retinyl palmitate concentrations in other mouse tissues are about 0.1–15 nmol/gram tissue. Tissue retinol concentrations are in the range of 0.2–400 nmol/gram and liver retinol is the highest among all tissues. atRA concentrations are about 1–30 pmol/gram tissue while other retinoic acids and downstream metabolites, such as 13cisRA and 4-oxo-atRA, are not detected in mouse organs (Obrochta et al., 2014; Zhong, Hogarth, et al., 2019). Based on these concentrations, 50–100 mg of tissue is usually the minimum quantity needed for measurement of atRA concentrations while retinol and retinyl palmitate can often be measured from smaller quantities (Table 2).

Table 2:

Recommended sample preparation methods and sample quantities for the analysis of retinoids in tissues.

Tissue	Target retinoid	Recommended amount of tissue	Sample preparation method
Liver	Retinol and retinyl palmitate	10–40 mg (mouse and human)	Acetonitrile precipitation of proteins
Liver	Retinoic acid	100–120 mg (mouse); 50–100 mg (human)	One-round extraction with hexanes from acidified tissue homogenate
Liver	4-oxo-retinoic acid	50–100 mg (human)	Ethyl acetate extraction from acidified tissue homogenate
Serum	Retinol, retinyl palmitate, and retinoic acid	50–100 μL (mouse)	Two-round extraction with hexanes from first basic and then acidified tissue homogenate
Serum	Retinol and retinoic acid	50–100 μL (human)	Acetonitrile precipitation of proteins
Testis	Retinol, retinyl palmitate, and retinoic acid	30–60 mg (mouse)	Acetonitrile precipitation of proteins
Other organs such as heart, spleen, skin	Retinol, retinyl palmitate, and retinoic acid	120–140 mg	Two-round extraction with hexanes from first basic and then acidified tissue homogenate

For human tissues, retinoid concentrations have mainly been quantified in serum and plasma (Arnold et al., 2012; Jones et al., 2015; Kane et al., 2005; Kane, Folias, & Napoli, 2008; Napoli, 1986). In addition, retinoid concentrations have been reported in human liver and testis (Arnold, Kent, Hogarth, Griswold, et al., 2015; Zhong, Kirkwood, et al., 2019), while data for other human tissues is lacking. Compared to mouse liver, human liver has similar retinol and retinyl palmitate concentrations, but higher 13cisRA and atRA concentrations (Zhong, Kirkwood, et al., 2019). Similarly, human serum retinol levels are similar to those in mice but atRA and 13cisRA concentrations are higher (Arnold et al., 2012). In human serum, 9cisRA and 9,13-di-cis-RA are also detected. Furthermore, the downstream metabolites of atRA and 13cisRA, 4-oxo-atRA and 4-oxo-13cisRA, have been detected in human liver and serum and their concentrations were either similar or higher than parent RA concentrations. In contrast, neither 4-oxo-RA metabolite was detected in mouse tissue (Zhong, Kirkwood, et al., 2019) with the available tissue quantities.

Other than tissue and species differences, animal strain, sex and diet are also important factors that can contribute to variations in retinoid concentrations. For example, mouse strain and diet were shown to impact retinoid concentrations in mouse tissues (Obrochta et al., 2014). In another study, serum retinol and liver atRA concentrations were found to be significantly lower in female than male mice (Zhong, Hogarth, et al., 2019). However, sex differences in retinoid concentrations were not observed in human livers in a study including 50 healthy liver samples (Zhong, Kirkwood, et al., 2019). Taken together, when conducting retinoid analysis in biological tissues, species, tissue type, strain, sex and diet should be considered when assessing the amount of sample needed and in selection of the appropriate analytical and sample preparation methods (described in Section 6).

4. Considerations for the use of retinoids in in vitro biological experiments

4.1. The presence of retinoids in biological reagents

The use of retinoids in cell culture experiments is vital for understanding the basics of retinoid biology and signaling pathways, but there is little consensus for the proper procedure for their use in vitro. A common source of variation between studies is the degree of serum supplementation in cell culture medium varying from serum-free to 5% or 10% fetal bovine serum (FBS). While this variability in the FBS content may affect retinoid stability as described above in section 2.1, it can also alter the overall treatment conditions. FBS is a chemically ill-defined biological matrix that includes a wide range of serum proteins, fatty acids, hormones, and cytokines (Yao & Asayama, 2017). In addition, standard FBS usually contains measurable amounts of retinol, in the range of 10–50 nM for a 10% FBS containing media (Baltes et al., 2004; Randolph et al., 1997; Sharow et al., 2012). In sensitive experiments, this contamination with retinol may drive or alter biological responses. For example, one study estimated that 5% FBS-supplemented medium contained 25 nM retinol but lacked atRA. Yet, this media retinol content resulted in steady-state intracellular content of retinoic acid of 25–50 nM (Randolph et al., 1997), a concentration within normal biologically active range. Whether this concentration is sufficient to confound cell culture experiments is not well defined. In HepG2 cells with baseline treatment and culture in the presence of 5nM atRA, RA was still found to dose-dependently induce CYP26A1 and RARβ (Tay et al., 2010). One may argue that in vivo cells will never be in a retinoid free environment, and therefore culturing cells at some level of baseline retinoids may be advised. Yet, variability in media retinol concentrations may cause unnecessary variability in experimental results of retinoid signaling and confound dose-response experiments. Thus, contaminating retinol in reagents should be considered when interpreting the vast range of biological responses downstream of vitamin-A signaling pathways.

Fatty-acid free, or charcoal-stripped FBS is a retinoid depleted alternative for cell culture experiments that may help to overcome the lot-to-lot variation in baseline retinoid medium content. Serum alternatives, such as neonatal calf serum, may have different retinoid content and should be quantified using HPLC-UV or LC-MS methods described herein prior to use. Thus, the routine use of these products in variety of applications such as sample preparation buffers, blocking of antibody non-specific binding, flow cytometry cell sorting buffers, and sample tube coating deserves consideration to ensure that interference is minimized. Best practices may include opting for charcoal-stripped or light-exposed reagents to minimize residual retinol background signals and downstream interference such as with retinoid quantification from flow-cytometry isolated cells.

4.2. Retinoid uptake and depletion in cell culture

Retinoids are generally assumed to have high permeability across cell membranes (Noy, 1992b), and this is likely true for retinol and for atRA at acidic pH when atRA is unionized. However, the pKa of atRA is 4.76 (PubChem) and therefore in physiological pH atRA is negatively charged, likely limiting the permeability of atRA across membranes (Noy, 1992a). Indeed, in Caco-2 monolayers, atRA was shown to have relatively poor permeability (Jing et al., 2017), suggesting that atRA may need active transport for uptake into cells. Another process that may limit apparent cell permeability and uptake of atRA to cells is protein binding in the media. Apart from maintaining the stability of retinoids in solution, FBS supplementation and specifically albumin concentration, appears to be a critical determinant of both retinoid uptake into cells and the rate of metabolism/isomerization (Noy, 1992b, 1992a; Randolph et al., 1997; Rundhaug et al., 1987; Topletz et al., 2015; Tsukada et al., 2002). Thus, when using retinoids in vitro, investigators should consider the potential impact of binding proteins on the data.

4.3. In vitro metabolic studies

The enzymes of the CYP26 family are the most important contributors to the clearance of atRA (Isoherranen & Zhong, 2019). However, recombinant CYP26 enzymes are notoriously difficult to express and purify in vitro, and at present no reports exist of successful expression of recombinant CYP26 enzymes in E. coli or in yeast. However, recombinant CYP26 enzymes have been successfully expressed and isolated in microsomal preparations from insect cells (Lutz et al., 2009; Topletz et al., 2012; Zhong et al., 2018), while others have used mammalian cells for recombinant expression of CYP26s (Gomaa, Bridgens, Aboraia, et al., 2011; Gomaa, Bridgens, Veal, et al., 2011; Krivospitskaya et al., 2012). atRA metabolism has also been extensively studied in human liver microsomes (HLMs) and in variety of recombinant CYPs including CYP3A4, CYP3A7 and CYP2C8 (Isoherranen & Zhong, 2019; Marill et al., 2000; Thatcher et al., 2010; Topletz et al., 2019). The best characterized metabolite of atRA is 4-OH-atRA (Figure 1), which is essentially always detected in CYP incubations, but detection of additional metabolites such as 18-OH-atRA, and 4-oxo-atRA is also often reported. With recombinant CYP26s, identified primary atRA metabolites include hydroxylation in the 4-, 16- and 18- carbons (Figure 8). However, with CYP26 enzymes, formation of sequential metabolites from atRA such as 4-oxo-atRA, and tentatively identified 4,16- and 4,18-diOH-atRA is also prevalent (Lutz et al., 2009; Shimshoni et al., 2012; Topletz et al., 2015). The characterization of sequential metabolism of retinoids in vitro remains challenging due in part to the lack of reference standards for the sequential metabolites and assessment of the biological role of such metabolites requires “made to order” or “in house” synthesis which can be resource intensive. Furthermore, a typical incubation reaction can generate several primary and sequential metabolites. Therefore, methods to achieve chromatographic separation of multiple structurally similar retinoid metabolites is critical for HPLC-UV analysis (Figure 5) as well as LC-MS/MS due to the identical molecular weights of the isomeric hydroxylation products. Due to the high affinity of atRA to CYP26 enzymes and the high catalytic activity, substrate and primary metabolite depletion are major concerns in kinetic studies with atRA. The depletion of atRA can be >20% within 2 minutes when incubations are done at concentrations around K_m with just 2 nM of CYP26A1, resulting in inaccurate kinetic estimates if parameters are not corrected for substrate depletion. Adjusting protocols for lower enzyme concentrations is one approach to address substrate and metabolite depletion but the ability to detect product formation at low enzyme concentrations becomes a limitation with current quantification methods. As such, substrate depletion approach may be the most appropriate for determining accurate kinetic constants for atRA in in vitro incubations, and this approach has been used in several studies (Lutz et al., 2009; Shimshoni et al., 2012; Topletz et al., 2015). Retinoid instability and sensitivity to auto-oxidation also contributes to challenges with in vitro incubations.

Figure 8: Detection of the metabolites of atRA formed by cytochrome P450 enzymes.

Panel A shows the HPLC-UV trace of atRA incubated with CYP26B1 in the absence of cofactor (NADPH). For panel B the metabolites were generated by incubating 500nM atRA with 5nM of recombinant CYP26B1 reconstituted with P450 reductase and in the presence of NADPH as described in section 6.3. Primary metabolites were identified as 4-OH-, 16-OH-, and 18-OH-atRA based on previous experiments although no reference standard for 16-OH-RA exists. The di-hydroxylation sequential metabolites M2, M3, and M4 have been tentatively identified but their structures have not been confirmed via synthetic standards. Acitretin was used as an internal standard. (Lutz et al., 2009; Thatcher et al., 2011).

5. Extraction methods for retinoids from biological matrices

All steps should be performed under yellow light. In general, for the majority of tissues liquid-liquid extraction is recommended, while the acetonitrile precipitation method can be used for certain analyses such as detecting retinoids in human plasma or serum (see Table 2).

5.1. Equipment

1. Workspace with Yellow Light

2. Borosilicate Glass Tubes (16× 150mm, 15mL)

3. Borosilicate Glass Pasteur Pipettes

4. Universal Glass Homogenizer with Glass Pestle

5. Handheld Electric Drill

6. Bead Mill Homogenizer with ceramic beads

7. Analytical Balance

8. Benchtop Centrifuge

9. Microfuge tubes (1.7 mL)

10. Benchtop Vortex

11. Nitrogen Evaporator

12. UHPLC coupled to high end Tandem Mass Spectrometer

13. Amber Autosampler vials and caps

14. Ascentis Express RP Amide column (2.7 μm; 150 mm × 2.1 mm; Sigma Aldrich)

15. Ascentis Express RP-Amide, Guard Cartridge (2.7 μm; 5mm × 2.1 mm; Sigma Aldrich)

5.2. Chemicals and reagents

1. Blank human serum (DC Mass Spect Gold MSG 4000, Golden West Biologics)

2. Powdered retinoid stocks (Sigma Aldrich; purity ≥98%)

3. Stable-labeled retinoid internal standards (Cambridge Isotope Laboratories, Inc. or Toronto Research Chemicals)

   1. All-trans Retinoic Acid-d₅
   2. 13-cis Retinoic Acid-d₅
   3. Retinyl palmitate-d₄
   4. Retinol-d₈

4. 0.25 M Potassium Hydroxide (KOH) dissolved in Ethanol

5. Acetonitrile (Optima LC/MS Grade)

6. Methanol (Optima LC/MS Grade)

7. Water (Optima LC/MS Grade)

8. Formic acid (Optima LC/MS Grade)

9. 4N Hydrochloric Acid (HCl)

10. Hexanes (Certified ACS)

11. Ethyl Acetate (Optima)

12. Ethanol (200 proof)

13. 0.9% NaCl solution

14. Nitrogen Gas

5.3. Two-round liquid-liquid extraction method for retinol, RE and RA

5.3.1 Weigh tissues; recommended tissue weights are listed in Table 3.

Table 3: Representative LC-MS/MS parameters for retinoid analysis including target ion transitions and instrument specific ionization parameters.

(Topletz et al., 2015; Zhong, Kirkwood, et al., 2019)

Ionization	Analyte	MRM m/z (MS3 m/z)	*CE/DP/EP (V)	*GS1/2	LC Conditions
APCI Positive ion	Retinol	269>93, 95	30/35/4	70/70	Sample injection: 4μL Column: Amide column at 40°C, 0.4 mL/min LC Conditions: A) H20 +0.1%FA B) ACN +0.1% FA 0–2 min at 60% B Gradient 60–66% B by 9.2 min Gradient 66–100% B by 13 min Return to starting conditions at 23 min
	Retinol-d8	277>98, 102	30/35/4	70/70
	Retinyl palmitate	269>93, 95	30/35/4	70/70
	Retinyl palmitate-d4	273>94, 98	30/35/4	70/70
	All-trans-retinoic acid and 13-cis-retinoic acid	301>205 (301>205>159)	17/80/10	80/0	Sample injection: 20μL Column: Amide column at 25°C, 0.4 mL/min LC Conditions: A) H20 +0.1%FA B) ACN +0.1% FA 0–3 min at 70% B Gradient 70–78% B by 12 min Increase to 100% B for 3 min Return to starting conditions at 15 min
	All-trans-retinoic acid-d5 and 13-cis-retinoic acid-d5	301>205 (306>208>162)	17/80/10	80/0
	4-oxo-all trans retinoic acid and 4-oxo-13cis-retinoic acid	315>137, 147, 241	29,26,20/50/10	60/70	Sample injection: 20μL Column: Amide column at 25°C, 0.5 mL/min. LC Conditions: A) H20 +0.1%FA B) ACN +0.1% FA 0–2 min at 40% B Gradient 40–67% B by 9 min Increase to 100% B for 2 min Return to starting conditions
	4-oxo-all trans retinoic acid-d3	318>137	29/50/10	60/70
ESI Negative ion	All-trans-retinoic acid and 13-cis-retinoic acid	299>255	−22/−140/−7	60/60	Sample injection: 10μL Column: C18 column at 40°C, 0.35 mL/min. LC Conditions: A) H20 +0.1%FA B) ACN +0.1% FA 0–3 min at 10% B Gradient 50–95% B by 7.6 min Return to starting conditions at 10.1 min
	All-trans-retinoic acid-d5 and 13-cis-retinoic acid-d5	304>260	−22/−140/−7	60/60
	4-hydroxy retinoic acid	315>253	−28/−80/−7	60/60
	4-hydroxy retinoic acid -d6	321>259	−28/−80–7	60/60
	4-oxo-retinoic acid	313>269	−22/−95/−5	60/60
	4-oxo- retinoic acid-d3	316>272	−10/−100/−5	60/60

^*APCI, atmospheric pressure chemical ionization; ESI, electrospray ionization; CE, collision energy; DP, declustering potential; EP, entrance potential; GS, Gas

5.3.2 Homogenize with 5x tissue weight of saline with universal glass homogenizer on ice

5.3.3 Transfer tissue homogenate to glass tubes using glass Pasteur pipettes and keep glass tubes on ice

5.3.4 Spike in internal standard to tissue homogenates

5.3.5 Add 2 mL of 0.25 M KOH dissolved in ethanol and vortex

5.3.6 Add 10 mL of hexanes to extract

5.3.7 Centrifuge at low speed for 5 mins

5.3.8 Transfer upper organic layer (retinol and RE fraction) to a new glass tube and evaporate at 25°C under gentle N₂ flow; Dry glass tubes are kept on ice until reconstitution

5.3.9 Add 120 μL 4 M HCl to the lower aqueous layer and vortex

5.3.10 Add 10 mL hexanes to extract

5.3.11 Centrifuge at low speed for 5 mins

5.3.12 Transfer upper organic layer (RA fraction) to a new glass tube and evaporate at 25°C under gentle N₂ flow; Dry glass tubes are kept on ice until reconstitution

5.3.13 Reconstitute retinol and retinyl ester fraction with 100–300 μL acetonitrile and transfer to amber MS vials for LC-MS/MS analysis

5.3.14 Reconstitute RA fraction with 60 μL acetonitrile, transfer to amber MS vials and add 40 μL water for LC-MS/MS analysis

5.4. One-round liquid-liquid extraction method for RA

Note: An ethyl acetate extraction (Section 5.6) is necessary for the extraction of RA metabolites as this one-round liquid-liquid extraction procedure will not extract the 4-oxo-RA and 4-OH-RA metabolites

5.4.1 Prepare samples the same as in steps 5.3.1 to 5.3.4 in method 5.3

5.4.2 Add 2 mL of acetonitrile+1% formic acid to homogenates and vortex

5.4.3 Add 10 mL of hexanes to extract

5.4.4 Centrifuge at low speed for 5 mins

5.4.5 Transfer upper organic layer (RA fraction) to a new glass tube and evaporate at 25°C under gentle N2 flow; Dry glass tubes are kept on ice until reconstitution

5.4.6 Reconstitute RA fraction with 60 μL acetonitrile, transfer to amber MS vials and add 40 μL water for LC-MS/MS analysis

5.5. Acetonitrile precipitation method

5.5.1 Weigh tissues (recommended tissue weights are listed in Table 2).

5.5.2 Homogenize with 4–5x tissue w/v saline with universal glass homogenizer on ice or with bead mill homogenizer using ceramic beads.

1. For liver tissue: Homogenization/dilution with a saline volume equivalent to 25x tissue weight is recommended.

2. For serum: Saline dilution is not needed

5.5.3 Transfer tissue homogenates to microfuge tubes and keep on ice

5.5.4 Add internal standard and gently vortex

5.5.5 Add ice-cold acetonitrile

1. For liver tissue: The ratio of acetonitrile to tissue homogenates is recommended to be 15:1 to 20:1

2. For non-liver tissues: The ratio of acetonitrile to tissue homogenates is recommended to be 1 to 1. Intensely vortex and leave on ice for 10 minutes

5.5.6 Centrifuge at 18,000 × g at 4°C for 30 mins

5.5.7 Transfer supernatant to amber MS vials for LC-MS/MS analysis. If tissue is liver do the following:

1. For retinol measurement: Transfer supernatant to amber MS vials for LC-MS/MS analysis

2. For retinyl ester measurement: Transfer a 10-fold diluted supernatant to amber MS vials for LC-MS/MS analysis

5.6. Ethyl acetate extraction for 4-oxo-RA

The following steps can be added after 5.3.12 or 5.4.5 to extract 4-oxo-RA

5.6.1 Add 10 mL of ethyl acetate to aqueous layer to extract

5.6.2 Add 2 mL of deionized water to facilitate phase separation

5.6.3 Centrifuge at low speed for 20 mins

5.6.4 Transfer upper organic layer (4-oxo-RA fraction) to a new glass tube and evaporate at 25°C under gentle N₂ flow; Dry glass tubes are kept on ice until reconstitution

5.6.5 Reconstitute 4-oxo-RA fraction with 60 μL acetonitrile, transfer to amber MS vials and add 40 μL water for LC-MS/MS analysis

5.7. LC-MS/MS instrumentation and conditions

While proper handling and extraction of retinoids is a critical component for optimal quantitation of retinoids, analytical sensitivity is heavily dependent on the instrumentation and separation by liquid chromatography. While many options exist for highly abundant retinoids, such as for serum retinol measurements (Gatti et al., 2000; Kane, Folias, & Napoli, 2008; Napoli, 1986; Schmidt et al., 2003), other analytes require high performance instrumentation and extraction methods (Arnold et al., 2012; Jones et al., 2015; Kane et al., 2005; Kane, Folias, Wang, et al., 2008; Kane & Napoli, 2010; Shimshoni et al., 2012; Topletz et al., 2015; Zhong, Kirkwood, et al., 2019). Currently, MS/MS using a triple-quadrupole instruments provides the highest sensitivity for detecting RA derivatives and metabolites (Kane & Napoli, 2010). Such instruments provide options for multiple-reaction monitoring, ion fragmentation energies, and ion selection. In addition, MS³ is valuable for confirming the identity of RA and its isomers and can be set up to monitor product ions of RA from the 2^nd precursor ions (Zhong, Kirkwood, et al., 2019). Analyte separation can be obtained by coupling MS to a high-performance LC (HPLC) or ultra-high performance LC (UHPLC). The optimal analytical method is unique for each compound, and multiple LC-MS/MS methods are needed in order to quantify and separate common retinoids in a single sample. Generally, retinol, retinyl esters, atRA, 13cisRA, and 4-oxo-RA isomers can be quantified using atmospheric pressure chemical ionization (APCI) and positive-ion mode (Zhong, Kirkwood, et al., 2019). The analytes can be separated using an amide column, such as an Ascentis Express RP Amide column (2.7 μm; 150 mm × 2.1 mm) with a corresponding pre-column and mobile phase consisting of (A) water with 0.1% formic acid (FA) and (B) and ACN with 0.1% FA. However, RA isomers, retinol and retinyl esters, and 4-oxo-RA require different flow rates, mobile phase gradients, and collision energies among other parameters and thus require separate methods, some of which are summarized in Table 3. RA metabolites such as 4-OH-RA can be separated using a robust C18 column in either positive or negative electrospray ionization (Topletz et al., 2015). In addition, if using stable-labeled internal standard such as atRA-d₅, the corresponding transitions need to account for the impact of the deuterium label in increasing the monitored m/z.

6. In vitro characterization of all-trans-retinoic acid metabolism

6.1. Equipment

1. Yellow Light Workspace

2. Borosilicate Glass Tubes (16× 150mm, 15mL)

3. Borosilicate Glass Pasteur Pipettes

4. Analytical Balance

5. Benchtop Centrifuge

6. Microfuge tubes (1.7 mL)

7. Benchtop Vortex

8. 37°C Shaking Water Bath

9. Water Bath Nitrogen Gas Evaporator

10. HPLC coupled to UV or Mass Spectrometer

11. Amber Autosampler vials and caps

12. Zorbax Extended C18 column, 2.1 × 100 mm, 3.5 μM (Agilent)

13. SecurityGuard Cartridge C18, 4.0 × 3.0 mm (Phenomenex)

6.2. Chemicals and reagents

1. All-trans retinoic acid stock in methanol (100x of final substrate concentration)

2. 100 μM acitretin stock in methanol (internal standard for HPLC-UV)

3. 10 μM 4-oxo-d₃ (internal standard for LC-MS/MS)

4. 4OH-atRA and 4oxo-atRA stocks in methanol for standards

5. CYP26 microsomal fractions quantified by CO-difference spectra, human liver microsomes (HLMs) with protein content quantified, or commercial insect cell expressed recombinant CYPs

6. Purified recombinant rat cytochrome P450 reductase (CPR for recombinant enzyme incubations, not needed for human liver microsomes or commercial insect cell microsomes co-expressed with reductase)

7. Acetonitrile (Optima HPLC Grade)

8. Ethyl acetate (Analytical grade)

9. Nitrogen gas

10. 100 mM potassium phosphate buffer, pH 7.4 (Kpi)

6.3. Metabolic incubation protocol

6.3.1 Prepare microsomal solutions:

• For recombinant enzyme microsomes: Prepare a 500 pmol/mL : 1000 pmol/mL (1:2) solution of CYP26 to CPR in Kpi; vortex and allow to sit at room temperature for 5 min before putting back on ice.

• For HLMs: Prepare a 20 mg/mL HLMs solution (optimized for reaction volume)

• For commercial recombinant enzymes: Prepare a 1000 pmol/mL supersomes solution

6.3.2 For each incubation reaction, pipette buffer, CYP26/CPR solution, and atRA into a tall 15mL borosilicate glass tubes on ice in the following volumes:

• For recombinant enzyme microsomes:

  • 880 μL of Kpi
  • 10 μL of CYP26/CPR solution
  • 10 μL of atRA stock in methanol (100x of final incubation concentration)
• For HLMs:

  • 880 μL of Kpi
  • 10 μL of HLMs
  • 10 μL of atRA stock in methanol (100x of final incubation concentration)
• For commercial recombinant enzyme microsomes:

  • 885 μL of Kpi
  • 5 μL Supersomes
  • 10 μL of atRA stock in methanol (100x of final incubation concentration)

6.3.3 Vortex well and pre-incubate at 37°C for 5 min in a shaking water bath

6.3.4 Initiate reactions with 100 μL of 10 mM NADPH (or Kpi for no cofactor control) for a final volume of 1mL

6.3.5 Vortex and incubate reaction

• For recombinant CYP26 enzyme microsomes: 2 min incubation

• For HLMs: 30 min incubation

• For commercial recombinant enzymes: 10 min incubation

6.3.6 After the indicated time in 6.3.5, add 4 mL of ethyl acetate to stop the reaction, vortex well and keep on ice

6.3.7 Add 3 μL of 100 μM acitretin (LC-UV) or 2 μL of 5 μM 4-oxo-d₃ (LC-MS/MS) and vortex well

6.3.8 Centrifuge at low speed for 15 min for phase separation

6.3.9 Transfer upper organic layer to a separate borosilicate glass tube using a glass Pasteur pipette

6.3.10 Evaporate organic fraction at 25°C under gentle nitrogen flow

6.3.11 Keep evaporated tubes on ice

6.3.12 Reconstitute in 150 μL of methanol and transfer to MS vials

6.3.13 Store samples at −20°C until analysis via HPLC-UV or LC-MS/MS

7. Summary

Recent advancements in instrument sensitivity and extraction methods using stable-labeled internal standards have dramatically improved the quantitative analysis of retinoids in a wide range of human and rodent tissues. However, the biological variation of tissue and serum retinoids and variability in interfering matrix components and ion suppression poses a challenge for the analysis of retinoids from complex matrices, such as liver tissue, via mass spectrometry. Furthermore, in vitro experiments with retinoids have unique considerations that can impact experimental interpretation. For example, routine use of BSA and/or FBS can dramatically contribute to the baseline intracellular retinoid content in cells, which may be present in adequate quantities to elicit a biological response for sensitive measures. In addition, protein binding plays a critical role in the stability of retinoids in media and impacts the uptake and metabolic elimination in cell culture, which in some instances can result in retinoid depletion. Substrate depletion is also a major concern in characterizing metabolite kinetics from enzyme incubations and should be considered in the design of experiments and interpretation of results. In conclusion, the successful bioanalysis of vitamin A and retinoids requires careful handling and preparation of stocks, reagents, and samples throughout both experimental and analytical procedures.