A cell-based assay for retinaldehyde dehydrogenase activity: Retinoid quantification as an alternative to current fluorescence-based approaches

Abstract

The biologically active metabolite of Vitamin A, retinoic acid, is essential for regulating immune tolerance, development, and metabolism. A key regulator of retinoic acid signaling is its synthesis by retinaldehyde dehydrogenase, whose expression is tightly regulated and cell-type specific. Current cell-based assays for retinaldehyde dehydrogenase activity rely on fluorescent aldehyde substrates, which lack specificity, limiting their accuracy and interpretability. Here, we developed a sensitive, cell-based assay that directly quantifies retinaldehyde dehydrogenase activity by measuring a panel of retinoids, including all-trans-retinoic acid, using liquid chromatography-mass spectrometry. Employing cultured conventional dendritic cells, we demonstrate that retinoic acid synthesis is time-, substrate-, and enzyme-dependent. Compared to fluorescence-based assays, our assay avoided artifactual signals influenced by cell density and provided a direct, quantitative measure of enzymatic activity in the context of broader retinoid metabolism. This assay offers additional practical advantages, including flexibility in sample processing and compatibility with other downstream metabolite analyses. Together, our protocol provides a robust, specific, and functionally relevant approach that complements existing fluorescence-based approaches to study retinoic acid biosynthesis in immune cells and beyond.

The biologically active form of Vitamin A, retinoic acid, regulates gene expression via nuclear retinoic acid receptors and plays essential roles in immune homeostasis, embryonic development, and cellular metabolism (1, 2, 3). For instance, all-trans-retinoic acid (ATRA) is one of the most potent regulators of immune tolerance: conventional dendritic cells type 1 (cDC1s), particularly in the mesenteric lymph nodes, secrete retinoic acid to induce regulatory T-cells and maintain immune tolerance in mucosal tissues (4, 5, 6, 7, 8, 9).

Given its potent and pleiotropic effects, retinoic acid signaling is tightly regulated by ligand availability (10, 11, 12). First, retinoic acid must be synthesized by irreversible oxidation of its precursor, retinal (Fig. 1A). This reaction is catalyzed by retinaldehyde dehydrogenases (RALDHs), members of the aldehyde dehydrogenase 1A family, whose expressions are tightly regulated. For instance, RALDH2 expression is restricted to specific immune subsets such as gut lamina propria and mesenteric CD103-expressing cDCs (13). This ensures localized and context-specific control of retinoic acid production and signalling.

Figure 1.

Establishment of an LC-MS-based assay for retinoid quantification.A, schematic of the retinoid synthesis pathway. B, neat standards of the retinoids were subjected to the LC-MS protocol as outlined in the Experimental Procedures. The peak intensity for each metabolite (MS transition) has been normalized to the respective maximum signal, then overlaid to demonstrate the successful separation of these metabolites. C and D, a calibration curve was created by spiking Medium B (recipe in Experimental Procedures) with a mixture of neat retinoid standards, followed by serial dilution with Medium B to generate a 12-point calibration curve (including the zero standard). 250 μl of each standard was extracted, and lipid extracts were subjected to LC-MS. The X-axis denotes the total amount of metabolite in the 250 μl standard. Analysis of precision (CV, coefficient of variation), accuracy (back-calculated value as a proportion of the real value, expressed as a percentage), and selectivity were conducted as outlined in the Experimental Procedures. These parameters were used to determine the dynamic range (cut-offs defined in the Experimental Procedures), shaded in (C) and depicted as a calibration curve in (D). Data in C presented as mean ± S.E. from 3 separate runs, with each dot representing the average of 5 injections performed within each run. Data in (D) presented as mean ± SD from 5 technical replicates, from a representative experiment, where the Y-axis units reflect peak areas that have been normalized to the internal standard (acitretin) and background signal (naïve media), as described in the Experimental Procedures. The data for ATRA is depicted in C and D, and other metabolites in Table S2.

Despite the biological importance of RALDHs, cell-based measurement of their activity remains a technical challenge. A commonly used assay employs fluorescent aldehyde substrates, such as AldeFLUOR (14) or AldeRed (15). These are oxidized by endogenous aldehyde dehydrogenases (ALDHs), which leads to their retention within cells for detection by flow cytometry. However, these probes are not specific for RALDH (16) and do not directly report on the production of retinoic acid, its downstream metabolism, or its efflux. This limits their utility for studying the control of retinoic acid signalling.

To overcome these limitations, we developed a cell-based assay that directly measures RALDH activity by quantifying the endogenous production of ATRA. Intact cells are briefly incubated with retinal, followed by lipid extraction and quantification of ATRA by liquid chromatography-mass spectrometry (LC-MS). Using the cDC1 cell-line, MutuDC1940 (17), we demonstrate this assay outperforms fluorescence-based methods by providing both functional specificity and biochemical relevance. This offers a robust and accurate tool for studying retinoic acid production in live immune cells and in vitro contexts.

Results

An LC-MS-based assay for retinoid quantification

The most abundant retinoic acid, ATRA, is synthesized by oxidation of retinal (Fig. 1A) by RALDH enzymes, a key regulatory node in the retinoid metabolism pathway. ATRA can subsequently be converted into degradation products such as 4-hydroxy- and 4-oxo-retinoic acid. To examine this pathway, we developed a targeted LC-MS/MS method to quantify these metabolites, as well as other retinoic acids (e.g., 9-cis-retinoic acid (9cRA), acitretin) and retinoid precursors (retinol, retinyl-acetate). We employed reversed-phase LC to resolve these retinoid species (Fig. 1B), which were then detected by multiple reaction monitoring (MRM) mode via tandem MS (Table S1 for optimised MRM conditions). MS/MS fragmentation spectra of these retinoids in cell extracts closely matched those of authentic standards, confirming analyte identity in the LC-MS/MS assay for this biological matrix (Fig. S1). This pipeline exhibited a wide dynamic range for each retinoid species (Fig. 1, C and D for ATRA and Table S2 for all metabolites), determined by evaluating the precision, accuracy, and selectivity of calibration curves. Combined with this assay’s specificity for each metabolite (Fig. 1B), this assay was used to study the kinetics of retinoid production.

Detection of RALDH activity in cultured cDC1s

Hence, we established a cell-based assay for RALDH activity. The general principle involved a short-term incubation with all-trans-retinal (retinal), quenching with cold methanol, extraction with methyl-tert-butyl ether (MTBE), and retinoid quantification by LC-MS/MS (Fig. 2A). We employed the MutuDC1940 cDC1 cell line (17), evaluating RALDH kinetics in both suspension and adherent formats. In both cases, we observed ATRA production increased linearly with time (Figs. 2B and S2A). Higher substrate concentrations increased ATRA production, with a plateau reached with 10 to 20 μM of substrate (Figs. 2C and S2B). The EC₅₀ was 3.0 ± 0.4 μM (suspension) and 5.9 ± 0.6 μM (adherent), in concordance with isolated RALDH2 enzyme having a Km of 0.66 μM for retinal (18) and the observed increase in Km when evaluating enzymes in intact cells (19). The levels of ATRA were substantially less than the initial retinal concentration (e.g., at 119.8 ± 15.3 nM ATRA for suspension cells incubated with 10 μM retinal for 15 min, where the ATRA concentration was determined by absolute ATRA levels and assay volume of 250 μl), confirming the substrate was not being depleted within this short timeframe.

Figure 2.

Detection of RALDH activity in cultured dendritic cells.A, schematic of the RALDH activity assay, as detailed in the Results and Experimental Procedures. RAL, all-trans-retinal; MTBE, methyl-tert-butyl ether. B and C, mutuDC1940 cells in either suspension or adherent formats (see Experimental Procedures) were incubated with either 10 μM RAL for the indicated times (B) or indicated concentrations of RAL for 15 min (C). Lipids were extracted and all-trans-retinoic acid (ATRA) was measured by LC-MS. ATRA levels are presented relative to the 20 min timepoint (B) or 10 μM RAL (C). Data presented as mean ± S.E. from 3 to 4 separate experiments. Absolute ATRA levels for B and C are presented in Fig. S2, A and B, respectively, with each experiment exhibiting similar trends despite between-experiment variability in baseline ATRA production. D, mutuDC1940 cells were treated for 20 h with 1 μM 9-cis-retinoic acid (9cRA), then subjected to Western blotting. Data presented as mean ± SE from 4 separate experiments. ∗p < 0.05 by t test. E, mutuDC1940 cells were treated for 20 h with 1 μM 9cRA, then assayed for RALDH activity in the suspension cell format (as in B and C), utilising the indicated RAL concentrations for 20 min. Data presented as mean ± S.E. from 4 separate experiments. ∗, p < 0.05 by t test. F, mutuDC1940 cells were treated with 9cRA and assayed for RALDH activity as in (E), except pre-incubated with the indicated concentrations of 673A, Win18446 (Win446), and N,N-diethylaminobenzaldehyde (DEAB) for 5 min prior to the addition of 10 μM RAL. Data presented as mean ± S.E. from 3 separate experiments. ∗p < 0.05; ∗∗p < 0.01 vs 9cRA condition (second column) by Dunnett's test. G, mutuDC1940 cells were incubated in suspension in Medium B with 10 μM RAL for 20 min (i.e., simulating the RALDH activity assay in E and F), after which tubes were chilled on ice prior to centrifugation at 300g and 4 °C for 5 min. The supernatant (media) was collected, then cells were resuspended in 250 μl of naïve, ice-cold Medium B, and both media and cells were subjected to ATRA analysis (A). In parallel, samples containing both cells and media (without fractionation) were analysed to determine total ATRA levels. The ATRA content of each fraction is presented as a portion of total levels. Data presented as mean ± S.E. from 5 separate experiments. H–J, mutuDC1940 cells were treated with 9cRA and assayed with 10 μM RAL as in (F). Absolute levels of retinoids were quantified in Fig. S2C. The levels of selected retinoids are presented here, normalised to the ATRA levels in the control-treated cells. ROL, retinol; 4OHRA, 4-hydroxy-retinoic acid. This dataset overlaps with G in a subset of experiments, where ATRA production data for the control-treated cells in H were used to determine total ATRA levels in G. In H, data presented as mean ± S.E. from 3 separate experiments. ∗p < 0.05, by t test. In I, data presented as the response to 9cRA (log-scaled fold-change in levels for 9cRA-versus control-treated cells). In J, ROL and ATRA levels are depicted relative to the sum of their abundances in control-treated cells. 9cRA treatment did not significantly change the total production of ATRA and ROL (p > 0.05, t test). There was no significant difference between the relative increase in ATRA and decrease in ROL with 9cRA treatment (p > 0.05, t test). In H–J, dots of the same colour depict individual datapoints derived from the same experiment across all three panels. K and L, Selected samples from F (cells treated with 9cRA, 9cRA + 1 μM Win446, or 9cRA + 10 μM DEAB) were analyzed for ROL (K) and 4OHRA (L) abundance. M, schematic depicting the changes in net flux in response to 9cRA, as detailed in the Results.

Next, we evaluated whether this assay could detect changes in RALDH activity. First, since RALDH2 (Aldh1a2) is an RXR/RAR target gene (13, 20), we stimulated Aldh1a2 expression using 9cRA (Fig. 2D). Like ATRA, this retinoic acid is also a ligand of RXR/RAR (21) but can be resolved from endogenous ATRA with our LC protocol (Fig. 1B). Consequently, 9cRA treatment increased ATRA production at both sub-maximal (2.5 μM) and near-maximal (10 μM) retinal concentrations (Fig. 2E). Conversely, briefly pre-incubating with RALDH-specific (673A, Win18446) or pan-ALDH (DEAB) inhibitors reduced ATRA production (Fig. 2F). Thus, our cell-based assay is sensitive to RALDH activity.

Examining how changes in RALDH activity impact broader retinoid metabolism in cultured cDC1s

Since we are performing the assay in live, intact cells, this provides an opportunity to evaluate the impact of RALDH activity in the context of the broader retinoid pathway (Fig. 1A). Conversely, ATRA can be effluxed or degraded, and retinal can be metabolized independently of RALDH—these can all influence ATRA levels and apparent RALDH activity.

First, we considered ATRA efflux. We fractionated cells and conditioned media prior to extraction, which revealed that a majority of the ATRA was effluxed into the media (Fig. 2G). Thus, extracting the entire reaction (media and cells, rather than the cells alone) was necessary to accurately quantify total ATRA production.

Next, we evaluated how the retinoid pathway responded to 9cRA treatment. It is important to note that 9cRA can non-enzymatically, reversibly isomerize to ATRA in situ, in the presence of protein thiols (22). To mitigate this confounding effect, cells were washed and resuspended in fresh media prior to retinal incubation, with little 9cRA detected upon retinal incubation (Fig. S2C). In this experiment, the major products included ATRA, retinol, and 4-hydroxy-retinoic acid (Fig. S2C). Of these, 9cRA treatment reduced retinol production, and increased ATRA and 4-hydroxy-retinoic acid production (Figs. 2H and S2C). We visualized these data as fold-changes (9cRA/control) to compare the responses of adjacent metabolites (Fig. 2I). This revealed a ‘crossover’ point between retinol and ATRA (from log-scale negative to positive), which corresponds to an altered site of regulation between these metabolites (23, 24). Notably, like RALDH, the enzyme that catalyzes reduction of retinal into retinol (retinal reductase, DHRS3) is known to be upregulated by retinoic acid signaling (25). However, 9cRA treatment increased ATRA levels and reduced retinol levels by a similar magnitude (Fig. 2J). Thus, while retinol is a dominant fate of retinal (Fig. S2C) that increases substantially over endogenous levels in a substrate- and time-dependent fashion (Fig. S2D), 9cRA leads to a net shift in retinal flux towards ATRA synthesis. This was corroborated by the increase (albeit not statistically significant) in retinol levels upon RALDH inhibition (Fig. 2K).

The first degradation product of ATRA, 4-hydroxy-retinoic acid, was also a major fate in this assay (Fig. S2C). When cell-free controls were spiked with comparable quantities of ATRA (10, 100 nM), 4-hydroxy-retinoic acid was detected at substantially lower levels than in the experimental samples (<5% of ATRA in cell-free controls versus ∼35% in control-treated cells). Further, 4-hydroxy-retinoic acid changed similarly to ATRA in response to 9cRA treatment (Fig. 2I). The major enzyme involved in ATRA degradation, CYP26, is also known to be transcriptionally induced by retinoic acids (26). Although CYP26 inhibition significantly reduced 4-hydroxy-retinoic acid and increased ATRA levels, these fold-changes were not substantial (<20%) (Fig. S2E). Nevertheless, 4-hydroxy-retinoic acid production was sensitive to RALDH inhibition (Fig. 2L). Together, this suggests 4-hydroxy-retinoic acid is generated independently of CYP26, as a function of higher RALDH activity (and ATRA production).

Overall, these data indicate that 9cRA leads to net flux of retinal away from reduction to retinol treatment towards RALDH-dependent ATRA production and degradation (Fig. 2M). This highlights the importance of considering neighboring reactions in a cell-based assay and demonstrates that in our cellular context, changes in ATRA levels were primarily a function of ATRA synthesis (i.e., RALDH activity).

ATRA production is more specific for RALDH activity than existing fluorescence-based assays

Next, we compared our assay to the conventional fluorescence-based assay. The latter involves incubating cells with a dye consisting of a fluorescent moiety ligated to an aldehyde group (e.g., AldeFLUOR, AldeRed-588A (15)). The aldehyde is oxidized by endogenous ALDHs, generating a carboxylic acid that causes the dye to be retained within cells for detection by flow cytometry (Fig. 3A). Here, we evaluated how this assay responds to changes in RALDH activity. RALDH inhibition should narrow the fluorescence distribution and lower median fluorescence intensity (MFI) (Fig. 3B).

Figure 3.

ATRA production is more specific for RALDH activity than existing fluorescence-based assays.A, schematic of the fluorescence-based ALDH assay (with AldeRed depicted as an example), as detailed in the Results. B, Hypothetical result, whereby if RALDH contributes to AldeRed oxidation (A), then subsequent analysis of AldeRed retention by flow cytometry should show a decrease in peak width (coefficient of variation, CV) and median fluorescence intensity (MFI) upon RALDH inhibition. C, mutuDC1940 cells were treated as described in Figure 2F, except Medium B was replaced with the assay buffer from the AldeRed assay kit (Merck, catalogue #SCR150) for the ATRA production assay. Data presented as mean + S.E., from 3 to 4 separate experiments for each condition. 9cRA, 9-cis-retinoic acid; Win446, Win18446; DEAB, N,N-diethylaminobenzaldehyde. D and E, mutuDC1940 cells were subjected to the AldeRed retention assay as described in the Experimental Procedures, using either 10⁶ cells/ml and the indicated AldeRed concentrations (D) or 0.61 μg/ml AldeRed with the indicated cell densities (E). Data presented as representative histograms (left) and MFI relative to the highest AldeRed concentration (D) or lowest cell density (E), as mean ± S.E. (right), from 3 to 4 separate experiments per condition. F, mutuDC1940 cells were subjected to the AldeRed retention assay with the indicated AldeRed concentrations and cell densities, with or without 10 μM DEAB. Data presented as representative histograms (left) and the effect of DEAB treatment on MFI (top right) and CV (bottom right) as mean ± S.E., from 4 separate experiments. The control-treated data in this dataset (F) overlaps with E in a subset of experiments. G, mutuDC1940 cells were treated for 20 h with or without 1 μM 9cRA, then subjected to the AldeRed retention assay in the presence or absence of 1 μM Win446 or 10 μM DEAB. The AldeRed retention assay was performed with 0.61 μg/ml AldeRed at either low (0.2 × 10⁶ cells/ml) or high (1 × 10⁶ cells/ml) cell density. Data presented as representative histograms (left) and MFI relative to the control (Ctrl) condition, as mean + S.E. (right), from 3 to 6 separate experiments for each condition. H and I, conditions from G without RALDH inhibition, presented as representative scatter plots (H), with the AldeRed^high population enclosed by a box in each plot (H) and presented as population frequency (AldeRed^high/eGFP⁺) (I). Data in (I) presented as mean + S.E., from 3 to 6 separate experiments for each condition.

Since our cell line constitutively expresses eGFP (27), we utilized the AldeRed reagent to avoid spectral overlap (15). This reagent is commercially available as a kit (Merck catalogue #SCR150). Although this kit utilizes an (undisclosed) assay buffer that likely differs from our buffer (Medium B, see Experimental Procedures), ATRA production was still sensitive to 9cRA stimulation and RALDH inhibition in this kit’s buffer (Fig. 3C). Further, we observed that fluorescence intensity was sensitive to AldeRed concentrations (Fig. 3D) and cell density (Fig. 3E)—notably, the highest cell density matches the conditions in the suspension format of our assay (Figs. 2A and 3C).

Next, we evaluated the response to RALDH inhibition under these conditions. The narrowing of fluorescence distribution was consistently observed across all conditions (Fig. 3F, bottom right). In contrast, there was a markedly variable effect on MFI (Fig. 3F, top right) where DEAB only lowered MFI at the lowest cell density. Although ATRA production was sensitive to changes in RALDH activity (increased with 9cRA, decreased with inhibitors) at higher cell density (Fig. 3C), similar effects on MFI were only observed at a lower density in the fluorescence-based assay (Fig. 3, G–I). Together, this demonstrates that indirect fluorescence-based assays require validation after optimization.

Discussion

Here, we developed a specific, cell-based assay for RALDH activity. Using a rapid LC-MS/MS-based protocol that can resolve a range of retinoid species (Fig. 1), we demonstrated that ATRA production was time-, substrate-, and RALDH-dependent (Fig. 2). This assay outperformed current fluorescence-based methods for detecting changes in RALDH activity (Fig. 3).

It is important to note that the activity measured by our assay reflects the functional production of ATRA in live, intact cells rather than the intrinsic catalytic properties of isolated RALDH enzymes (18) or RALDH activity in homogenates (28). Cellular RALDH activity is influenced by substrate uptake, cofactor (NAD⁺) availability, compartmentalization, competing metabolic pathways, and product efflux—factors absent from purified-enzyme assays and known to shift apparent Km and Vmax in situ (19). We demonstrated that upon acute retinal incubation in our cell culture system, ATRA levels reflect RALDH activity by using biologically relevant perturbations, such as substrate dose, incubation time, RALDH inhibitors, and 9cRA induction. Parenthetically, although media washout minimized the contribution of 9cRA isomerization to measured ATRA levels (Fig. S2C), 9cRA activates both RAR and RXR (21); thus, future studies could employ selective agonists that do not interconvert with ATRA such as LG100268 (29) or bexarotene (30) (both RXR-selective agonists that lack the polyene backbone of retinoic acid) to dissect transcriptional effects on RALDH activity. Furthermore, increased RALDH activity impacted neighboring reactions, reducing retinal conversion to retinol and increasing ATRA degradation. Thus, our assay enables the study of RALDH activity in the context of broader retinoid metabolism.

Our protocol overcomes several limitations in existing fluorescence-based approaches for measuring RALDH activity. First, fluorescence-based approaches cannot distinguish RALDH from other ALDHs, whereas here we employed a RALDH-specific substrate (retinal). Second, fluorescence-based approaches cannot distinguish between intracellular substrate and product (e.g., reduced and oxidized AldeRed, respectively), leading to binary output (“ALDH-High”, “ALDH-Low” (15)). In contrast, our assay distinguishes between substrate (retinal) and product (ATRA) (Fig. 1B) and quantifies changes in ATRA levels over time (Fig. 2B).

Third, related to the above points, the indirect nature of fluorescence-based assays can generate spurious effects—here, we observed that MFI only responded as expected to RALDH modulated at lower cell densities (Fig. 3F). RALDH inhibitors did not increase background fluorescence, i.e. in the absence of AldeRed (data not shown). It is likely that at higher cell densities, AldeRed dye becomes limiting (Fig. 3E), masking changes in AldeRed oxidation and retention. In contrast, our assay provides a direct and quantitative readout of RALDH activity, enabling us to demonstrate that retinal was not limited in the acute timeframe of our assay. Thus, our assay yielded consistent ATRA production kinetics in multiple cell formats (Fig. 2, A and B), and avoided artefactual signals observed in fluorescence-based methods (Fig. 3).

Fourth, there are multiple stopping points in the protocol prior to LC-MS/MS analysis (e.g., after quenching with methanol or solvent evaporation). Parenthetically, solvent evaporation also enables analytes to be concentrated based on the required detection sensitivity. Fifth, our protocol can be multiplexed with other assays—for instance, the aqueous phase could be used for isolating polar metabolites. Finally, our LC-MS/MS-based pipeline could be applied to retinoid quantification in serum or tissue samples, offering utility beyond measuring cellular RALDH activity alone.

On the other hand, a limitation of our assay is that it operates best with homogenous cell populations. Thus, fluorescence-based approaches could be initially used to screen heterogenous cell populations for ALDH-positive cells, which could be isolated and subjected to our assay to confirm changes in RALDH activity. In this way, our assay can complement existing approaches. Furthermore, at present, our assay does not resolve the activity of individual RALDH isoforms, which could be addressed by selective overexpression or knockdown of these isoforms depending on the cellular context being studied (as in (28)). In addition, future studies should consider retinol esterification as another product of retinal incubation. Here, acute retinal incubation led to a marked accumulation of retinol within minutes (Fig. S2D), consistent with rapid retinal reduction and retinol esterification likely not being a kinetically dominant fate under these assay conditions—this could be further validated by the quantification of retinyl esters (e.g., retinyl palmitate, stearate, linoleate, and oleate (31)). Although retinyl esters can generate the same in-source fragment ion as retinol (32), quantitative interpretation requires ester-specific calibration due to differences in detector response. Last, differences in retinal uptake could also impact ATRA production - although this is unlikely to account for our findings, since substantial amounts of retinal were reduced to retinol (Fig. S2C), this could be explored in future work by measuring the import of retinal analogs that are poorly metabolized by RALDH (33).

Together, we present a RALDH assay that overcomes key limitations of fluorescence-based methods. Acute incubation with excess substrate (retinal) allows the evaluation of RALDH activity in isolation of other ALDHs, paired with a robust LC-MS/MS-based assay for unambiguous quantification of ATRA. This assay is compatible with multiple sample formats and can be adapted to broader applications in retinoid analysis. Importantly, patterns in RALDH activity can be cross-referenced to steady-state retinoid levels, which are a function of precursor availability, RALDH activity, and downstream ATRA degradation—such experiments would place RALDH function within its physiological context. While best suited for homogenous cell populations, our assay complements existing fluorescence-based screens, studying RALDH function with greater precision.

Experimental procedures

Chemical reagents

Unless otherwise specified, the following compounds were purchased from MedChemExpress: all-trans-retinoic acid (catalogue #HY-14649), acitretin (Ro 10–1670, catalogue #HY-B0107), all-trans-retinal (catalogue #HY-W004500 or Sigma Aldrich catalogue #R2500), 9-cis-retinoic acid (catalogue #HY-15128), all-trans-4-oxo-retinoic acid (catalogue #HY-107494A), talarozole (R115866, catalogue #HY-14531), 4-hydroxy-retinoic acid (catalogue #HY-125904), Win18446 (catalogue #HY-W011094), 673A (catalogue #HY-122912), retinol (Sigma-Aldrich catalogue #95144), and retinyl acetate (catalogue #HY-N0679). All compounds were initially resuspended in DMSO. N,N-diethylaminobenzaldehyde (DEAB) was obtained from the AldeRed assay kit (Merck-Millipore catalogue #SCR150) and concentrated by solvent evaporation under N₂ gas and resuspension in DMSO.

Cell culture

MutuDC1940 cells (17) were obtained from Applied Biological Materials (catalogue #T0528). Cells were maintained in Medium A, consisting of RPMI1640 (catalogue #61870127, with GlutaMAX supplement) supplemented with 10% (v/v) foetal bovine serum (catalogue #10099141), 10 mM HEPES (pH 7.4, catalogue #15630080), 55 μM β-mercaptoethanol (catalogue #21985023), and 100 U/ml penicillin/streptomycin (catalogue #15140122), all sourced from Thermo Fisher Scientific. Cells were cultured at 37 °C and 5% (v/v) CO₂. Cell density was closely monitored to prevent overgrowth and passages 20 to 40 were used for experiments. Adherent cells were passaged by lifting cells with PBS at 37 °C. Quality control was conducted by regular phenotypic analysis for dendritic cell identity (e.g., cytokine production, capacity to induce regulatory T-cells, cell marker analysis by flow cytometry).

ATRA production assay

Cell pretreatments

Cells were seeded at a density of 1 million cells in 2 ml of Medium A per well in 6-well plates. After at least 2 h (to maximize cell adherence), cells were treated by the addition of 1/20 vol of medium. For 9cRA treatment, this 21× solution was prepared by resuspending 9cRA stock (or DMSO solvent for the control treatment) in Medium B, which was identical to Medium A except that fetal bovine serum was replaced with 0.5% (w/v) fatty acid-free bovine serum albumin (Sigma-Aldrich catalogue #A7030). This solution was incubated at 37 °C for at least 30 min to facilitate 9cRA-albumin conjugation, prior to addition to cells.

ATRA assay

Following treatment, cells were subjected to the assay outlined below, unless otherwise specified in the figure legends. The assay was performed by incubating cells in a total volume of 250 μl of Medium B – in general, this involved incubating cells in 150 μl of medium, followed by the addition of 50 μl of 5× solution of RALDH inhibitors, equilibration for 10 min, then addition of 50 μl of 5× solution of retinal. After 20 min, cells were quenched by the addition of 300 μl pre-chilled (−20 °C) methanol containing 0.1 mg/ml butylated hydroxytoluene (to prevent lipid oxidation) and 1 μM acitretin (retinoid internal standard). Samples were stored at −20 °C until extraction. This protocol was applied differently based on the cell format.

For the suspension cell format, plates were spun at 300g for 5 min to collect both adherent and non-adherent cells. Cells were lifted with PBS, pelleted at 300g for 5 min, and resuspended in Medium B. Cells were dispensed into 2 ml round-bottom tubes (Eppendorf catalogue #30123344), held in a pre-warmed (37 °C) tube rack, which was gently shaken after additions of inhibitor and retinal. Following the assay, cells were quenched by adding pre-chilled methanol directly to each tube.

For the adherent cell format, plates were spun at 300g for 5 min to collect both adherent and non-adherent cells. Media was aspirated, and cells were washed once with Medium B before being incubated in 150 μl of Medium B per well. The plate was gently rocked following the addition of the inhibitor and retinal. Following the assay, cells were quenched by the addition of pre-chilled methanol, scraped, and transferred to 2 ml round-bottom tubes.

In both cases, Medium B was preincubated in a sterile flask at 37 °C and 5% CO₂ for at least 30 min before being used for treating cells or suspending inhibitors and retinal. Furthermore, tubes (suspension) and plates (adherent) were incubated under similar conditions with lids open (tubes) or removed (plates) in the CO₂ incubator to maximise exposure to warm air containing 5% CO₂. For experiments involving AldeRed assay buffer, equilibration and incubation steps were performed at 37 °C without exposure to 5% CO₂. Retinoids were incubated in their respective, pre-equilibrated assay media at 37 °C for at least 30 min before cell treatment (ATRA assay) or serial dilution (calibration standards), to facilitate albumin conjugation.

Last, the following controls were incubated and extracted in parallel to the above samples in each experiment: (i) cell-free controls, each consisting of 200 μl of naïve media mixed with 50 μl of 5× retinal, and (ii) calibration standards, consisting of naïve media supplemented with retinoid standards, serially diluted with naïve media to yield at least 6 calibration points.

Retinoid isolation by organic two-phase extraction

Lipids were isolated from cell extracts by extraction with MTBE. This relies on phase separation upon mixing MTBE, methanol, and water in a 10:3:2.5 (v/v/v) ratio (34). Cell extracts contained 250 μl aqueous solution and 300 μl methanol, to which 1 ml of MTBE was added. This mixture was briefly vortexed and centrifuged at 16,000g and 4 °C for 15 min. For each sample, 500 μl of supernatant (lipid phase) was transferred to 1.5 ml tubes and dried down at 28 °C under N₂ gas. Handling only an aliquot of the lipid phase enabled faster drying, reducing sample degradation and increasing throughput whilst being controlled for by the internal and calibration standards. Performing the drying in 1.5 ml conical-bottom tubes (Eppendorf catalogue #30123328 or Axygen catalogue #MCT-175-C) enabled easier pelleting following sample reconstitution in the next section. Due to the light-sensitive nature of retinoids, this entire procedure was performed in the dark where possible.

Retinoid quantification by LC-MS/MS

Sample preparation

Dried lipid extracts were reconstituted in 60 μl of acetonitrile/water (3:1, v/v) and resuspended by ultrasonication for 1 min at room temperature. Samples were centrifuged at 16,000g and 4 °C for 10 min, after which 40 μl of supernatant was transferred to SureSTART 0.3 ml glass microvials (performance level 3, Thermo Fisher Scientific) for LC-MS analysis.

Retinoid separation by LC

Metabolites were resolved by LC using a 1290 Infinity II ultra-high performance LC system (Agilent) with an InfinityLab Poroshell 120 EC-C18 column (Agilent, 2.7 μm particle size, 2.1 mm internal diameter × 100 mm length, with matched guard column attached). Buffer A consisted of water with 0.1% (v/v) formic acid and Buffer B consisted of acetonitrile with 0.1% (v/v) formic acid. Applying a gradient of 75% to 98% B, using acetonitrile as the solvent for Buffer B was superior to methanol and methanol/acetonitrile (1:1) for resolving ATRA and 9cRA (data not shown). The final gradient consisted of: 75% B, 0 min; 75% B, 1 min; 98% B, 6 min; 98% B, 7 min; 75% B, 7.5 min; 75% B; 8.5 min; with a flow rate of 0.5 ml/min throughout. The autosampler temperature was 4 °C, column temperature was 30 °C, and injection volume was 3 μl.

LC pre- and post-treatment

Before sample analysis, the LC column was conditioned with initial LC conditions (75% B) at 0.5 ml/min for 10 min, followed by at least 3 injections with water/acetonitrile (1:3, v/v) and 5 ‘dummy sample’ injections using the calibration samples. After the batch run, the LC column was washed prior to storage as described previously (35).

Sample analysis

MS analysis was performed using an Agilent 6470 QQQ with Jet Stream Technology Ion Source operated in positive-ion mode. General source parameters were adopted from a published UPHLC-MS/MS workflow (35): gas temperature, 200 °C; gas flow, 11 L/min; nebulizer pressure, 40 psi; sheath gas temperature, 400 °C; sheath gas flow, 12 L/min; capillary voltage, 3000 V for positive mode; and nozzle voltage at 500 V for positive mode. MRM transitions for retinoids were optimized and calibrated using metabolite standards, with transition parameters described in Table S1. Data acquisition was performed with a 10 ms dwell time for each transition in standard MRM mode.

Data analysis

LC-MS/MS data were extracted using MassHunter Qualitative Analysis (Agilent) and Skyline (36). Peak areas were normalized to those of the internal standard, acitretin (example using ATRA as analyte in Equation 1).

$$\text{normalised}\_{\text{ATRA}}_{\text{sample}}={\text{ATRA}}_{\text{sample}}\times {\text{acitretin}}_{\text{average}}/{\text{acitretin}}_{\text{sample}}$$ (1)

and adjusted by the lowest calibration standard, naïve media containing no retinoids (example using ATRA as analyte in Equation 2).

$$\text{adjusted}\_{\text{ATRA}}_{\text{sample}}=\text{normalised}\_{\text{ATRA}}_{\text{sample}}-\text{normalised}\_{\text{ATRA}}_{\text{naive}\_\text{media}}$$ (2)

These adjusted values were converted to absolute amounts (e.g., pmol) using the calibration standards before the background signal (e.g., residual contaminants in the retinoid stocks) was removed using cell-free controls. ATRA production was then normalized to the time of incubation with retinal and cell number, the latter obtained either by the seeding density (suspension format) or cells treated in parallel (adherent format).

Analyte identification

To confirm analyte identity in the biological matrix (cells in culture media) following extraction, product ion (MS/MS) spectra were acquired for each retinoid. The same LC conditions and MS parameters were used as MRM quantification (Table S1), except product ions were scanned from m/z 50 to 330 with a step size of 0.1 and scan time of 500 ms. Spectra from authentic standards were compared with those obtained from cell extracts spiked with 250 pmol of each standard (per 550 μl suspension of cells, media, and MeOH supplemented with BHT) prior to extraction with MTBE. Matching dominant product ions and conserved fragmentation patterns were observed across standards and biological samples (Fig. S1).

Calibration quality control

Specificity was confirmed by the resolution of individual metabolite species by LC-MS/MS, using both individual standards (Fig. 1B) and a combined standard mix. To determine the dynamic range of each metabolite, a series of calibration standards was spiked into assay media and subjected to the analyses above. These were performed in parallel to experimental samples to evaluate LC-MS/MS performance under typical batch sizes. Precision and accuracy were evaluated from 5 injections of each calibration standard (37), with the dynamic range defined by both coefficient of variation (for precision) and relative error (for accuracy) being <15% (or 20% at the lower limit of quantification). This entire procedure was performed on 3 separate occasions, from which within-run and between-run variability were evaluated (Table S2). This data was also used to confirm selectivity, whereby matrix samples (i.e., media without any calibration standards added) generated less than 20% of the signal at the lower limit of quantification (37).

Flow cytometric analysis of ALDH activity by AldeRed retention

Cells were seeded at a density of 1 million cells in 2 ml of Medium A per well in 6-well plates. After overnight treatment, cells were lifted with PBS at 37 °C, pelleted by centrifugation at 300g for 5 min at 4 °C, and resuspended in cold Medium C (verapamil-supplemented AldeRed Assay Buffer). The routine assay used 1 million cells/ml with 0.61 μg/ml AldeRed reagent, with variations to these parameters specified in individual figure legends. For experiments involving inhibitors (e.g., DEAB), the cell suspension was incubated with the respective inhibitors for 5 min on ice prior to the addition of AldeRed reagent. After the AldeRed reagent was added, cells were incubated at 37 °C in the dark. After 35 to 50 min, cells were washed by being pelleted via centrifugation at 300g for 5 min at 4 °C and resuspended in fresh, ice-cold Medium C, before being subjected to flow cytometric analysis. Samples were kept on ice and acquired within 30 min using an LSR Fortessa flow cytometer (BD Biosciences). FlowJo software (version 10.10.0) was used for data analysis.

Western blot analysis

Following treatment, plates were spun at 300g for 5 min to collect both adherent and non-adherent cells. After the media was aspirated, cells were washed once with ice-cold PBS and lysed at room temperature with PBS supplemented with 1% (w/v) SDS. Lysates were scraped, needle-homogenized, and centrifuged at 16,000g for 20 min at room temperature. The protein content of the clarified supernatant was quantified by the Pierce bicinchoninic acid assay (Thermo Fisher Scientific). Cell lysate (10 μg) was resolved by SDS-PAGE, transferred to polyvinylidene difluoride membranes (BioRad), and immunoblotted as described previously (38). Primary antibodies detecting RALDH2 (catalogue #83805) and β-actin (clone D6A8, catalogue #8457) were obtained from Cell Signaling Technology. Detection was performed using either (i) chemiluminescence using secondary antibodies conjugated to horseradish peroxidase (Jackon Immunoresearch), incubation with Immobilon chemiluminescent reagent (Merck-Millipore) and imaging using the iBright CL1500 imager (Thermo Fisher Scientific), or (ii) fluorescence using secondary antibodies conjugated to fluorophores (LI-COR) and imaging using the Odyssey CLx (LI-COR). Densitometric analysis was performed using ImageJ software (39) or LI-COR Image Studio.

Data availability

Unless otherwise indicated, the data described here is contained within the manuscript. Unpublished data can be shared upon request by contacting the corresponding author (S. N., severine.navarro@qimrb.edu.au).

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: S. N.’s research is supported by the CSL Research Accelerator Initiative.

Reviewed by members of the JBC Editorial Board. Edited by Qi-Qun Tang