Ethanol promotes differentiation of embryonic stem cells through retinoic acid receptor-γ

Ethanol (EtOH) is a teratogen, but its teratogenic mechanisms are not fully understood. The alcohol form of vitamin A (retinol/ROL) can be oxidized to all-trans-retinoic acid (RA), which plays a critical role in stem cell differentiation and development. Using an embryonic stem cell (ESC) model to analyze EtOH's effects on differentiation, we show here that EtOH and acetaldehyde, but not acetate, increase differentiation-associated mRNA levels, and that EtOH decreases pluripotency-related mRNAs. Using reporter assays, ChIP assays, and retinoic acid receptor-γ (RARγ) knockout ESC lines generated by CRISPR/Cas9 and homologous recombination, we demonstrate that EtOH signals via RARγ binding to RA response elements (RAREs) in differentiation-associated gene promoters or enhancers. We also report that EtOH-mediated increases in homeobox A1 (Hoxa1) and cytochrome P450 family 26 subfamily A member 1 (Cyp26a1) transcripts, direct RA target genes, require the expression of the RA-synthesizing enzyme, aldehyde dehydrogenase 1 family member A2 (Aldh1a2), suggesting that EtOH-mediated induction of Hoxa1 and Cyp26a1 requires ROL from the serum. As shown with CRISPR/Cas9 knockout lines, the retinol dehydrogenase gene Rdh10 and a functional RARE in the ROL transporter stimulated by retinoic acid 6 (Stra6) gene are required for EtOH induction of Hoxa1 and Cyp26a1. We conclude that EtOH stimulates stem cell differentiation by increasing the influx and metabolism of ROL for downstream RARγ-dependent transcription. In stem cells, EtOH may shift cell fate decisions to alter developmental outcomes by increasing endogenous ROL/RA signaling via increased Stra6 expression and ROL oxidation.

Here we provide evidence that EtOH causes ESC differentiation by increasing RA synthesis by Aldh1a2 following uptake of ROL from the medium by the Stra6 transporter and ROL oxidation by Rdh10. Downstream RA signaling is then dependent on RAR␥-mediated transcription via direct RARE activation in primary RA-responsive genes. Elucidating the precise mechanisms underpinning the interactions between EtOH and RA-mediated transcription during ESC differentiation enhances our fundamental understanding of several disease phenotypes during development.
Transcripts increased by Ͼ2-fold by 40 mM EtOH treatment included those of the homeotic (Hox) family (Hoxa1, Hoxb1, Hoxa5) and Cdx1 ( Fig. 1C; Fig. S1C). Because several transcripts increased by EtOH are direct RAR/RXR transcriptional targets (22)(23)(24), we then analyzed additional primary RA target genes, RAR␤2 and Cyp26a1. We detected increases after 48 h of EtOH treatment (Fig. 1C) compared with vehicle-treated ESCs. We additionally measured transcripts of RA-responsive genes in another ESC WT line, CCE, to rule out any AB1 ESC linespecific effects of EtOH. We found that EtOH also increased transcript levels of RA-responsive genes in CCE cells, and that doses of 40 and 80 mM EtOH elicited similar effects (Fig. S1D). We used either 40 or 80 mM EtOH in subsequent experiments. EtOH treatment did not increase transcript levels of lineagespecific genes that were also unaffected by RA treatment at 48 h, such as Fgf5 (ectoderm) and Sox17 (endoderm) in ESCs (Fig. S1E). Thus, EtOH increases transcript levels of specific RA target genes rather than affecting a broad differentiation phenotype.
To probe for additive effects of EtOH and retinoids we added EtOH to ESCs that were also treated with 1 M RA or ROL. We used Hoxa1, Cdx1, and Hnf1␤ as readouts for both RA responsiveness and ESC differentiation and did not detect additional increases in transcript levels compared with RA/ROL-treated cells alone at 48 h (Fig. S1F), suggesting that transcript induction of differentiation-associated genes by RA and EtOH converges on the same pathway.
To determine whether EtOH increases Hoxa1 mRNA levels by enhancing mRNA stability or by increasing transcription, we treated CCE ESCs with EtOH or 1 M RA for 48 h, isolated RNA immediately from some wells, and added 2 g/ml of actinomycin D to other wells for 30, 90, or 240 min to block transcription. The differences in the derivatives of the linear regression lines between untreated and EtOH-treated WT ESCs were Ϫ0.034 Ϯ 0.09 (p ϭ 0.76) for Hoxa1 (Fig. 1F) and Ϫ0.043 Ϯ 0.04 (p ϭ 0.54) for Cyp26a1 (Fig. 1G). The absence of major changes in halflives of both Hoxa1 and Cyp26a1 mRNAs between vehicletreated and EtOH-treated ESCs suggests that the increases in transcript levels upon EtOH treatment do not primarily result from increased mRNA stability in the presence of EtOH.

RARE activation is necessary for ethanol-mediated Hoxa1 transcription in embryonic stem cells
To determine whether a functional RARE is required for signaling by EtOH we next performed a transient transfection in ESCs using Hoxa1-lacZ minigene reporter constructs in which lacZ was cloned into the Hoxa1 coding sequence (22). We used two different constructs; one contained an enhancer with an intact RARE (WT, AGTTCA) and the other contained an RARE that was inactivated by mutation (Hoxa1-lacZ muRARE, AaTTac). We treated these transfected ESCs with vehicle (0.1% DMSO), EtOH (40 mM), or RA (0.5 M) for 24 h. We observed a 1.5 Ϯ 0.15-fold (p ϭ 0.034) increase in ␤-gal activity in the EtOHtreated and a 1.9 Ϯ 0.28-fold (p ϭ 0.036) increase in RA-treated WT ESCs transfected with the construct harboring an intact WT RARE (Fig. 2C). We did not observe any increase in ␤-gal activity in either EtOH-or RA-treated lysates from WT cells transfected with the Hoxa1-lacZ muRARE construct. These results show first, that the effects of EtOH occur at the transcriptional level and second, that there is a requirement for a functional RARE to mediate EtOH-induced transcriptional effects on Hoxa1.

Ethanol promotes stem cell differentiation via RAR␥
Enrichment of histone 3 lysine acetylation (acetyl-H3) allows RAREs to become more accessible for the RAR/RXR complex to bind and induce transcription. We performed chromatin immunoprecipitation (ChIP) assays using an antibody against the H3K27ac modification, which identifies transcriptionally active enhancers (32), to examine histone acetylation patterns in chromatin near RAREs of genes which exhibited mRNA increases by EtOH. Both Hoxa1 and Cyp26a1 contain at least one RARE at enhancers, whereas RAR␤2 contains a RARE near its proximal promoter (18). Genes from the EtOH-treated WT ESCs exhibited Ͼ1.5-fold H3K27ac enrichment near RAREs compared with vehicle-treated ESCs (2.1 Ϯ 0.25-fold, p ϭ 0.01, Hoxa1; 2.7 Ϯ 0.54-fold, p ϭ 0.036, RAR␤2; 1.6 Ϯ 0.07-fold, p ϭ 0.001, Cyp26a1) (Fig. 2D). These increases in H3K27ac chromatin marks upon EtOH treatment suggest that the chromatin near the RAREs is in a configuration in which transcription is activated.

Ethanol increases transcripts associated with retinol metabolism
To determine whether RA is a required intermediate for the EtOH-mediated increases in differentiation-associated genes, such as Hoxa1 and Cyp26a1, we first measured transcript levels of several genes required for RA synthesis from ROL. ROL is primarily metabolized to retinaldehyde by retinol dehydrogenase-10 (Rdh10) (7). Using semiquantitative RT-PCR, we showed that transcripts of Rdh10, but not Rdh5 or Rdh11, were increased by EtOH in WT ESCs (Fig. 3A). By RT-qPCR analysis we also showed EtOH-associated increases in transcript levels of the RAR␥ target gene Rdh10 (1.7 Ϯ 0.12-fold, p ϭ 0.004) and the intracellular ROL transporter Rbp1 (Crbp1) (6.7 Ϯ 1.3-fold, p ϭ 0.011) (Fig. 3, B and C) in WT ESCs. Crabp2, which transports RA to the nucleus (33), displayed increased transcript levels (5.28 Ϯ 1.1-fold, p ϭ 0.018) in WT ESCs upon EtOH addition. Rbp1 and Crabp2 exhibited regulation by RAR␥, because the RAR␥E8 Ϫ/Ϫ ESC line showed attenuated increases in these transcripts by EtOH compared with those in WT ESCs (Fig.  3C). Transcripts for the retinaldehyde reductase, Dhrs3, but not Dhrs4, were increased by EtOH treatment (15.7 Ϯ 3.1-fold, p ϭ 0.009) in WT ESCs (Fig. 3D). Importantly, Dhrs3 stabilizes the Rdh10-containing retinoid oxidoreductase complex (34). These data show that EtOH increases mRNAs of key genes that metabolize ROL to retinaldehyde. In contrast, the Aldh1a2 mRNA level was not increased by EtOH in WT ESCs ( Fig. S3; Fig. 1A).

Aldh1a2 is required for ethanol-mediated transcriptional changes
Because EtOH increased transcripts of genes involved in RA synthesis and nuclear transport, we ablated Aldh1a2 activity using CRISPR/Cas9 targeted to two sequences in intron and exon 5 to generate an Aldh1a2E5 Ϫ/Ϫ ESC line (Fig. 4, A and B). The absence of Aldh1a2 prevented the EtOH-mediated Hoxa1 and Cyp26a1 transcript increases observed in WT cells (Fig.  4C). These data indicate that metabolism of retinaldehyde to RA is required for EtOH to increase Hoxa1 and Cyp26a1 transcripts.

Ethanol treatment did not cause detectable increases in RA levels in embryonic stem cells
We measured RA levels in EtOH-treated ESCs using reversed phase HPLC-tandem MS to determine whether increased intracellular RA levels correlated with the increases in Hoxa1 and Cyp26a1 transcripts. Using a triple quadrupole mass spectrometer, we detected a peak for 20 pmol of an RA standard at a retention time of 3.5 min (Fig. S5A). Calibration curves were generated for RA with a limit of detection of 40 fmol and a lower limit of quantitation of 382 fmol (95 nM for 4 ϫ 10 6 cells, where 1 l volume ϭ 1 ϫ 10 6 cells and 4 l ϭ injection volume) for the transition m/z 301.2 3 123.1, and 341 fmol (85 nM for 4 ϫ 10 6 cells) for a secondary m/z 301.2 3 159.1 transition (Fig. S5, B and C). We also generated a calibration curve for 4-oxo-RA, a metabolite of RA (Fig. S5D). RA levels in ESCs treated with either vehicle or EtOH were too low to detect, but we detected an RA peak in cells treated with exogenous RA for 8 h (Fig. 4D; Fig.  S5E). We observed a second peak at retention time 3.25 min for transition m/z 301.2 3 159.1, but this peak did not correspond to any known RA isomer or metabolite (Fig. S5, F-H).
To increase the sensitivity for RA detection, we next treated WT ESCs with EtOH for 48 h and switched to a medium containing high vitamin A (ϩ0.5 M ROL) 6 h prior to collecting lysates. This medium contained a 5-10-fold higher ROL concentration than that in standard 10% serum-containing medium (0.05-0.1 M). In ESCs cultured in 0.5 M ROL we could measure intracellular RA above the sensitivity threshold of the mass spectrometer, but we still detected no changes in RA levels in EtOH-treated ESCs compared with vehicle-treated cells (Fig. 4E).
RA is oxidized to 4-oxo-RA (36), so we measured 4-oxo-RA as a surrogate for RA and observed a downward trend in 4-oxo-RA levels after EtOH addition that was not statistically significant (Fig. 4F). Thus, we did not observe increases in intracellular RA levels by MS after EtOH addition.

Stra6 is necessary for ethanol-dependent increases in Hoxa1 and Cyp26a1, but not Dhrs3 transcripts
Stra6 is a ROL transporter that is expressed in some, but not all, tissues and cell types (37,38), and its high expression is associated with a ROL requirement for differentiation (39). If EtOH induces RA-mediated transcription by enabling increased entry of ROL into ESCs for oxidation to RA, then the loss of Stra6 function should abrogate this effect. We first mea-

Ethanol promotes stem cell differentiation via RAR␥
sured Stra6 mRNAs in WT and RAR␥ Ϫ/Ϫ ESCs. Stra6 transcripts were elevated by 18.4 Ϯ 2.8-fold (p ϭ 0.003) in WT ESCs treated with EtOH, but were not elevated in the absence of RAR␥ (Fig. 5A). We saw effects of EtOH on the long and short Stra6 isoforms similar to those we observed with 1 M RA, with the long isoform increased to a greater extent by both EtOH and RA treatment (Fig. S6).

Rdh10 is required for ethanol-dependent increases in differentiation-associated genes
To address the potential requirement for Rdh10-dependent oxidation of ROL for EtOH-mediated increases in differentiation-associated transcripts we used CRISPR/Cas9 to generate deletions in both alleles in exon 2 of the Rdh10 gene (Fig. 5, G  and H).

Discussion
The effects of EtOH on RA levels and signaling are highly debated; either potentiation (15) or inhibition (12)(13)(14)16) of RA signaling in cell culture and animal models has been reported. Early studies relied on indirect assessment of RA activity or addition of exogenous ROL (13,40), as a sensitive method of detecting RA levels was lacking until more recently. Recent studies have found that fluctuations in retinoid levels following EtOH administration often vary in a sex-or tissue-specific manner (15,16). For example, Kim et al. (16) showed that retinyl esters (REs) were depleted in lungs of adult rats from dams fed 6.7% alcohol between embryonic day 7 and 21, with decreased levels in the ventral prostates and livers of males only. RA levels were not measured in this study, however, and depletion of retinyl esters could imply increased transport and utilization of retinoid stores for RA production in other tissues.
Another study used both RE and RA levels as readouts for retinoid activity, revealing a complex physiological response to EtOH (15). Using a 6.5% EtOH-containing diet in mice for 1 month, Napoli and colleagues (15) showed that RE levels were unchanged in the brain and increased in kidneys and testis, yet hippocampal and cortex RA levels were increased by 20-fold and 2-fold, respectively, kidney RA levels were unchanged, and serum and testis RA levels were also increased.
The contextual relationship between EtOH and RA may also be influenced by developmental stage, contributing to differences in the literature. Shabtai et al. (41) demonstrated in Xenopus embryos that a deficiency in Aldh2 expression during gastrulation may create a competition for limiting amounts of Aldh1a2 enzyme and diminish RA production. Using a zebrafish model of high-dose (100 mM) EtOH exposure during gastrulation, addition of RA partially rescued some toxic effects on anteroposterior axis formation, ear development, and craniofacial cartilage defects but exposure to low-dose (1 nM) RA alone or with EtOH recapitulated other FASD-like developmental defects (12). Exposure to pharmacological doses of retinoids, such as through use of the prescription acne medication isotretinoin, also causes severe birth defects resembling an

Ethanol promotes stem cell differentiation via RAR␥
FASD-like phenotype (42). Hence, retinoid teratogenicity is complex, as RA is central to cell differentiation and organismal development (7), and phenotypes present similarly whether low or high levels of RA are present (9 -11, 42). The effects of EtOH are equally complex and are associated with both increases and decreases in RA levels in accordance with tissue physiology as well as gene expression patterns at different developmental stages.
Our use of ESCs allowed us to determine mechanistically how retinoid signaling in pluripotent stem cells, representing the most primitive stage of development, is affected by EtOH exposure. Additionally, we generally used a dose of EtOH (40 mM) that is representative of a concentration that will be present in the bloodstream of a binge drinking adult (20) to analyze the effects in stem cells without subjecting the cells to concentrations that may be potentially lethal in humans and induce a variety of secondary toxic events.

mRNAs of differentiation genes are increased and pluripotency factor mRNAs are decreased in embryonic stem cells treated with ethanol
Prior studies have shown that EtOH delays or interferes with proper differentiation along specific lineages in cell culture models of directed differentiation (3,43,44). Thus, we probed the acute effects of EtOH on selected self-renewal and differentiation-associated genes in undifferentiated ESCs. We detected decreases in some pluripotency marker transcripts, and increases in several differentiation-related transcripts ( Fig. 1C;  Fig. S1B). The loss of pluripotency in EtOH-treated cells was confirmed using alkaline phosphatase staining (Fig. S1A). Addition of 1 M RA to cultured ESCs directly increases mRNAs of many lineage factors to cause differentiation along a parietal endoderm (epithelial) lineage (24). We have previously shown that Hoxa1 and Cyp26a1 protein levels are correlated with their mRNA levels (45,46). We showed here that EtOH addition to cultured ESCs induced transcripts of several differentiationassociated genes, which was recapitulated by administering the EtOH metabolite AcH but not by acetate, suggesting that either EtOH or AcH is responsible for these increases in differentiation-associated mRNAs (Fig. 1, D and E). We speculate that the inhibitory effect of acetate on EtOH-mediated transcript induction of some differentiation-associated genes (Fig. 1E) may result from contributions by acetate-derived acetyl groups to histone acetylation modifications. Moussaieff et al. (47) have demonstrated that 1 mM acetate treatment of ESCs delays endodermal differentiation via its conversion to acetyl-CoA and subsequent transfer of these acetyl groups to histone lysines to maintain transcriptionally active chromatin for pluripotency-related genes.

RAR␥ binding to RAREs is necessary for ethanol-induced increases in mRNA levels of differentiation-associated genes
The activation of genes associated with differentiation by RA via RARs is well-characterized (7,19). Activation of RAR-controlled transcriptional hubs in stem cells produces localized effects within RA-controlled chromosomal regions in factories of related differentiation genes containing RAREs that configure to their proper spatial position for transcriptional effects (17,48). RAR␥ is an essential transcription factor in RA-dependent differentiation of ESCs (27,49,50). Some functional redundancy exists among the three types of RARs in ESCs (26,51,52), but only RAR␥ was demonstrated to mediate F9 embryonic carcinoma cell differentiation and override activity of other RARs (49). Additionally, the loss of RAR␥, but not RAR␣, was associated with differentiation defects and altered Hoxa1 expression (28,53), which are likely caused by the dynamics of RAR subtype-binding patterns following ligand activation. Both RAR␣ and RAR␥ occupy a large number of sites genomewide during ESC differentiation (50). However, whereas RAR␣ is enriched 24 -48 h after RA signaling commences to sustain differentiation, RAR␥ initiates differentiation via direct activation of primary response genes (22,28,50,53). We showed here that the increases in mRNAs induced by EtOH were prevented by ablation of RAR␥, implicating direct RAR␥/RXR-mediated signaling in promoting transcriptional effects of EtOH on differentiation genes (Fig. 2, A and B; Fig. S2).

Ethanol induction of differentiation-associated transcripts in embryonic stem cells depends on Aldh1a2
We demonstrate here that EtOH treatment of ESCs likely increases intracellular ROL from the serum to generate RA to activate transcription. This transcriptional effect of EtOH requires Aldh1a2 expression, as genetic ablation of Aldh1a2 prevented EtOH-mediated increases in Hoxa1 and Cyp26a1 transcripts (Fig. 4C).
Despite our inability to detect differences in RA levels between EtOH-treated and untreated ESCs (Fig. 4, D and E), depleting medium of ROL caused abrogation of EtOH-mediated transcriptional effects (Fig. 4C). Precedence for potent RA activity in the absence of detectable RA increases by MS is found in the literature. For example, Blaner and colleagues (54) demonstrated that Lrat (lecithin-retinol acyltransferase) ablation in the livers of mice was associated with increases in several RA response genes despite no detectable changes in RA levels measured by a highly sensitive LC-MS protocol. This is in line with our own findings, as an Lrat-deficient state in the liver mimics the natural state of ESCs, which are not equipped for ROL storage as esters (27). Excess retinaldehyde that is not oxidized to RA for downstream transcription would instead be converted back to ROL by Dhrs3 to maintain homeostasis (34).
Restoration of Hoxa1 and Cyp26a1 transcript induction by EtOH occurred upon adding ROL back into ROL-depleted medium, showing a retinoid requirement and implying enhanced sensitivity to available ROL in the presence of EtOH (Fig. 4C). To determine the mechanism underlying increased sensitivity to available ROL by EtOH, we measured mRNAs of genes in the ROL metabolism pathway and found increases in several, including Rbp1, Crabp2, Rdh10, and Dhrs3 (Fig. 3). Rdh10 and Dhrs3 exist in a bifunctional complex to ensure that RA levels are tightly controlled (34). Although increasing the Rdh10 level in the presence of ROL proportionally increases detectable RA levels, an increase in both protein components, Rdh10 and Dhrs3, of the oligomeric complex prevents overall levels of RA from rising intracellularly (34). In our study, both Rdh10 and Dhrs3 mRNAs increase following EtOH treatment, with larger increases in Dhrs3, consistent with higher levels of

Ethanol promotes stem cell differentiation via RAR␥
Dhrs3 being required for fine-tuning retinoid oxidoreductase complex activity (34).
In addition, although Cyp26a1 transcript levels were elevated by EtOH, we do not think that Cyp26a1 is a major contributor to the lack of detectable changes in RA levels after EtOH addition, as levels of 4-oxo-RA, a common polar metabolite formed from Cyp26a1 oxidation of RA, were not increased but rather trended downward (Fig. 4F). This finding is consistent with a model of EtOH causing enhanced sensitivity of ESCs to low amounts of RA generated from ROL metabolized by the retinoid oxidoreductase complex (model, Fig. 6).

Stra6-dependent retinol uptake from the medium facilitates efficient conversion of retinol to retinoic acid by Rdh10 for signaling in ethanol-treated embryonic stem cells
The Stra6 transporter, which facilitates ROL intracellular uptake, exhibited increased mRNA levels following EtOH treatment (Fig. 5A), and loss of Stra6 RARE function was sufficient to abrogate EtOH-mediated increases in Hoxa1 and Cyp26a1 transcripts (Fig. 5D). Stra6 has "gatekeeper" functions in ESCs; in the absence of EtOH we speculate that the "gate" remains closed and ROL cannot enter the cells in high enough quantities to facilitate signaling. Given that ESCs express only a very low level of Lrat for ROL storage as retinyl esters (27), ROL entering the cells via the Stra6 transporter should be preferentially oxidized to RA. This suggests that EtOH may exert more toxicity via greater signaling through the RA pathway in cell types that do not express much Lrat. Because RA levels were not increased despite functional effects on expression of differentiation-related genes, it is likely that a steady influx of ROL through Stra6 followed by ROL conversion to RA via Rdh10 occurs, with effi-cient usage of newly synthesized RA to trigger nuclear signaling and subsequent differentiation through RAR␥-mediated transcription (model, Fig. 6). Our findings in Rdh10-null ESCs further support this model. The activation of differentiation-associated genes by EtOH was completely abrogated in Rdh10-null ESCs (Fig. 5I). These results suggest that ROL is preferentially oxidized by Rdh10 upon EtOH treatment. Despite the failure of ROL to induce differentiation-associated mRNAs in the absence of a functional Stra6 RARE, induction of these mRNAs in Rdh10-null ESCs by EtOH was similar to that in WT. This suggests that once ROL is imported into ESCs it can still signal in the absence of oxidation by Rdh10, possibly via its efficient intracellular conversion to 4-oxoretinol, which serves as a direct ligand for RARs (55,56).

Conclusions
Our findings collectively improve our understanding of the mechanisms by which EtOH metabolism affects RAR␥ signaling and differentiation in stem cells. We have demonstrated that EtOH causes stem cell differentiation via the activation of RA:RAR␥-mediated transcription in pluripotent stem cells. We propose a model of enhanced ROL uptake in EtOH-treated ESCs, whereby EtOH causes Stra6-dependent ROL uptake into ESCs, followed by its conversion to RA by Rdh10 and Aldh1a2. RA is then transported to the nucleus to bind RAR␥ for RA:RAR/RXR-mediated transcription (Fig. 6). Because ESCs represent an early stage in a dynamic cascade of events in early embryogenesis, they serve as a good model for studying EtOH stem cell toxicity. Our lab has previously shown that exogenous RA stimulates target gene transcription in doses as low as 100 pM (24), and thus EtOH effects, via changes in RA signaling, can potentially greatly shift the trajectory of cell fate decisions to alter developmental outcomes. Our data raise the exciting possibility that stem cell-related complications of EtOH exposure may be amenable to manipulation of RAR target genes for the prevention of EtOH-associated toxicities and diseases.

Experimental procedures
Cell culture and reagents AB1, CCE, RAR␤ ϩ/Ϫ ␥ Ϫ/Ϫ , RAR␥E8 Ϫ/Ϫ , and Stra6 RAREϪ/Ϫ ESCs were cultured as described previously (24,26,29,37,45). Cells were treated with 95% EtOH; 1 mM AcH (Calbiochem); 1 mM sodium acetate (Sigma), pH ϭ 7.4; 0.1, 0.5, or 1 M ROL (Sigma); and all-trans-RA (Sigma) at concentrations of 0.1 or 1 M dissolved in 100% DMSO. AcH was aliquoted from a freshly opened bottle and tubes were stored at Ϫ20°C for no more than 2 months. Each aliquot was immediately discarded after being added to the medium. Retinoids were prepared in dim light from a 1 mM stock solution. We used 1 M RA for experiments to differentiate ESCs along an extraembryonic endoderm lineage, and 0.5 M as a positive control for stimulating RARE activation in the ␤-gal assay. 0.1% DMSO was added to each treatment group not containing RA. ESCs were seeded in 6-well plates and harvested simultaneously for treatments conducted 72, 48, or 24 h prior to collecting lysates. Reagents were changed twice daily ϳ12 h apart, with the final reagent change completed 8 h prior to harvest. EtOH was used at concentrations of 40 mM and 80 mM in various experiments. 103 units/ml of leu- Figure 6. Model for ethanol regulation of stem cell differentiation via activation of RA signaling. ROL, in complex with Rbp4, is a substrate for Stra6, which imports ROL into the cell. EtOH increases the mRNAs of Stra6 and several genes in the RA synthesis pathway, including Rdh10, Dhrs3, Rbp1, and Crabp2. Intracellular ROL is either converted to 4-oxoretinol via Cyp26a1 or presented to the RA synthesis machinery upon binding to Rbp1. ROL is not stored as retinyl esters in ESCs, as Lrat is not expressed in these cells. The Rdh10/Dhrs3 complex oxidizes ROL to retinaldehyde, which serves as a substrate for Aldh1a2-catalyzed oxidation to all-trans-RA. Newly formed RA is then transported to the nucleus by Crabp2, where it activates the RAR␥/RXR transcriptional complex to stimulate expression of RA-responsive genes necessary for ESC differentiation.

Ethanol promotes stem cell differentiation via RAR␥
kemia inhibitory factor was added to medium for all experiments, including medium in which KnockOut TM SR (Gibco) replaced ESC-grade fetal calf serum.

␤-gal assay
␤-gal assays were performed as described previously (22). CCE cells were grown in 6-well plates and transfected the following day with 2-3 g of either WT Hoxa1 minigene-lacZ or Hoxa1-lacZ muRARE constructs. A pGL3-luciferase construct with an upstream SV40 promoter was simultaneously transfected at 0.1-0.2 g (15:1 ratio sample:control) to normalize ␤-gal activity to luciferase expression. Cells were treated 48 h after transfection with DMSO (0.05%), 40 mM EtOH, or 0.5 M RA for 24 h with a reagent change 8 h prior to harvest. RA doses above 0.5 M did not cause additional stimulation of reporter assays. Cells were collected in TEN buffer and sonicated to prepare lysates for the ␤-gal assays.
Generation of Aldh1a2E5 ؊/؊ and Rdh10E2 ؊/؊ lines CRISPR constructs for Aldh1a2E5 Ϫ/Ϫ line creation were generated using the pX461, pSpCas9n(BB)-2A-GFP nickase vector (Addgene #48140). The CRISPRevolution Synthetic RNA kit was used in generating Rdh10E2 Ϫ/Ϫ cells (Synthego, San Francisco, CA). Single cell dilutions were plated and grown for 1 week. Colonies were subsequently grown in 24-well plates and harvested in PBS. Clones were genotyped for genome editing by PCR amplification followed by restriction digestion. Clones lacking the restriction site were Sanger sequenced on both alleles and double-positive knockout clones were expanded in culture (see supporting "Materials and methods").

ChIP assays
Experiments were performed as described previously (18) with described antibodies (see supporting "Materials and methods"). Primers used for PCR analyses are detailed in Table S1.

Retinoid extraction and HPLC-tandem MS analysis
Lysates were collected in PBS and extracted using 50% acetonitrile/butanol and saturated K 2 HPO 4 (57). The organic phase was vacuum dried and reconstituted in 100% acetonitrile before loading. Retinoid separation was conducted using HPLC (Agilent, Palo Alto, CA) and Jet Stream electrospray ionization in positive ion mode. Reversed phase HPLC-tandem MS analysis was performed as described previously (58) (see supporting "Materials and methods").

Statistical treatment of the data
Statistical analysis was conducted on at least three independent biological replicates for each experiment using GraphPad Prism 7.0 software. The mean Ϯ S.E. was determined. Analysis of variance (ANOVA) was used to determine statistical significance within sets of three or more groups, and Student's t test was used to compare two independent populations. A twotailed p value Ͻ0.05 was considered statistically significant.