SWI/SNF complexes act through CBP-1 histone acetyltransferase to regulate acute functional tolerance to alcohol

SWI/SNF chromatin remodeling genes are required for normal acute responses to alcohol in C. elegans and are associated with alcohol use disorder in two human populations. In an effort to discover the downstream genes that are mediating this effect, we identified SWI/SNF-regulated genes in C. elegans. To identify SWI/SNF-regulated genes in adults, we compared mRNA expression in wild type and swsn-1(os22ts) worms under conditions that produce inactive swsn-1 in mature cells. To identify SWI/SNF-regulated genes in neurons, we compared gene expression in swsn-9(ok1354) null mutant worms that harbor a neuronal rescue or a control construct. RNA sequencing was performed to an average depth of 25 million reads per sample using 50-base, paired-end reads. We found that 6813 transcripts were significantly differentially expressed between swsn-1(os22ts) mutants and wild-type worms and 2412 transcripts were significantly differentially expressed between swsn-9(ok1354) mutants and swsn-9(ok1354) mutants with neuronal rescue. We examined the intersection between these two datasets and identified 603 genes that were differentially expressed in the same direction in both comparisons; we defined these as SWI/SNF-regulated genes in neurons and in adults. Among the differentially expressed genes was cbp-1, a C. elegans homolog of the mammalian CBP/p300 family of histone acetyltransferases. CBP has been implicated in the epigenetic regulation in response to alcohol in animal models and a polymorphism in the human CBP gene, CREBBP, has been associated with alcohol-related phenotypes. We found that cbp-1 is required for the development of acute functional tolerance to alcohol in C. elegans. We identified 603 transcripts that were regulated by two different SWI/SNF complex subunits in adults and in neurons. The SWI/SNF-regulated genes were highly enriched for genes involved in membrane rafts, suggesting an important role for this membrane microdomain in the acute alcohol response. Among the differentially expressed genes was cbp-1; CBP-1 homologs have been implicated in alcohol responses across phyla and we found that C. elegans cbp-1 was required for the acute alcohol response in worms.


Background
Alcohol is a widely used and abused drug. The significant majority of adults in the United States uses alcohol, and the misuse of alcohol has serious societal impacts [1]. Disturbingly, problems associated with alcohol misuse appear to be getting worse; for example, the rate of deaths associated with alcohol use has risen substantially in the last two decades, particularly among groups that have traditionally lagged in alcohol-related mortality [2].
The liability to abuse alcohol is strongly influenced by an individual's genetics [3], although the specific genes involved are as yet incompletely understood. An individual's level of response (LR) to alcohol manifests earlier than alcohol problems, and is a strong predictor of the likelihood to develop alcohol use disorder (AUD) [4][5][6]. LR to alcohol is a composite assessment that consists of both physiological measures, such as body sway, as well as subjective measures of intoxication. A lower sensitivity to the depressing effects of alcohol early in adulthood is correlated with a higher lifetime incidence of alcohol problems [7][8][9]. Identifying the genes that modulate the LR to alcohol is the goal of our work.
We use the nematode Caenorhabditis elegans as a model for understanding the acute physiological response to alcohol (ethanol), and for identifying genes that are likely to affect LR in humans. We previously found that SWI/SNF chromatin remodeling complex genes were required for normal acute ethanol response behaviors in C. elegans [10]. SWI/SNF are multi-protein complexes that modify chromatin structure by regulating the interaction of histones and DNA at particular genes, occluding or revealing transcription factor binding sites (reviewed in [11]); these actions result in an epigenetic modulation of target gene expression. SWI/SNF complexes share common enzymatic core proteins, and different complexes have different accessory subunits that confer their sequence specificity (reviewed in [12,13]). In worms, most SWI/SNF complex members were involved in regulating acute ethanol response behaviors, but different SWI/SNF complexes affected different behaviors: BAF subunits were required for normal initial sensitivity to ethanol, while PBAF subunits were required for acute functional tolerance to ethanol [10]. We found that allelic variation in several SWI/SNF complex members was associated with alcohol dependence in different adult human populations [10,14]. Allelic variation in different SWI/SNF complex members was also associated with antisocial behavior in adolescents, a phenotype that is strongly predictive of future alcohol use problems [14].
Together, these data strongly argue that the SWI/SNF complex plays a functional role in human AUD, and support the investigation of the biological underpinnings of this role. Because SWI/SNF acts to regulate transcription, the downstream effectors through which it modulates the acute ethanol response are of significant interest. Here, we use transcriptional profiling in C. elegans to identify these downstream modulators. We took advantage of two genetic reagents, one that localizes expression of swsn-9 to neurons using tissue-specific rescue of a swsn-9 null allele, and one that temporally restricts function of swsn-1 to adults using a temperature-sensitive allele of swsn-1 [15]. swsn-9 encodes a PBAF subunit, while swsn-1 encodes a core subunit that is predicted to be present in both BAF and PBAF; both genes are required for acute functional tolerance to ethanol [10]. We found that 603 genes were regulated by the SWI/SNF complex in adults and in neurons. Among these genes was cbp-1, encoding a histone acetyltransferase. We found that animals carrying loss of function alleles of cbp-1 had defects in acute ethanol response behaviors, indicating that cbp-1 is an important SWI/SNF target for LR to ethanol.

SWI/SNF-regulated genes in adults
We previously showed that swsn-1 is required during adulthood for the development of acute functional tolerance (AFT) to ethanol [10]. swsn-1(os22ts) is a temperature sensitive allele of swsn-1; os22ts is an amino acid substitution that causes the protein encoded by swsn-1(os22ts) to be functional at 15°C, but destabilized so that its function is lost at 25°C [15]. For these experiments, we reared N2 and swsn-1(os22ts) at the permissive temperature (15°C) from the time the eggs were laid through the fourth larval stage (L4), at which time we shifted the worms to the restrictive temperature (25°C) and grew them to adulthood (Fig. 1a). In wild type worms, SWSN-1 is expressed both during development and in adulthood. In swsn-1(os22ts) mutant worms, this temperature shift protocol produces functional SWSN-1 during development and inactive SWSN-1 in adulthood. We found that worms lacking SWSN-1 during adulthood failed to develop AFT, while wild type worms reared under the same conditions developed AFT to ethanol [10]. This result suggests that the SWI/ SNF complex regulates gene expression in adults that is necessary for AFT in these animals. To identify genes that are regulated by SWSN-1 in adults, we performed RNA sequencing on wild type and swsn-1(os22ts) mutant worms that were subjected to the temperature shift protocol. Five biological replicates were generated on different days; mutant and wild type worms were reared in parallel, thus limiting the effect of environmental variation on our comparison. We isolated RNA from young adult worms approximately 6 h after the temperature shift, and at a time when they were beginning to lay eggs. RNA sequencing libraries were prepared and sequenced by the Genomic Services Laboratory at Hudson Alpha (Huntsville, AL).
We assessed the correlation between biological replicates using principal component analysis and found that our biological replicates grouped together and the two genotypes cluster in distinct groups (Fig. 1b). The first two principal components accounted for 95% of the variance in the dataset, with principal component one Fig. 1 SWI/SNF-regulated genes in adults. a A temperature-sensitive allele, swsn-1(os22ts), was used to demonstrate that SWI/SNF complexes are required in adult tissues for AFT [10]. Diagram (top) shows the temperature shift regimen and table (bottom) summarizes alcohol response behaviors from this experiment. b Principal component analysis with gene expression plotted relative to the first two principal components (PC1 and PC2). Wild type and swsn-1(os22ts) replicates are most similar to replicates of the same genotype. c Volcano plot shows genes that are upregulated (red) and down-regulated (blue) in swsn-1(os22ts) mutants. Dashed lines indicate the FDR and fold change cutoffs (FDR ≤ 0.05 and fold change ≥1) (variation between sample types) accounting for 92% of the variance. We analyzed differential gene expression using DESeq2 [16] and found that 6813 genes were differentially expressed between swsn-1(os22ts) mutant and wild type worms (FDR ≤ 0.05) (Additional file 1). A volcano plot shows the wide distribution of differentially expressed genes (DEGs) (Fig. 1c). Approximately equal numbers of genes had increased (3210) and decreased (3603) expression in swsn-1(os22ts) mutants when compared to wild type. This suggests that the SWI/SNF complex both activates and represses gene expression in adult tissues, which is consistent with previous observations of SWI/SNF regulatory activity [17][18][19][20].

SWI/SNF-regulated genes in neurons
We previously showed that swsn-9(ok1354) mutant worms were unable to develop AFT to ethanol and that, if we expressed swsn-9::GFP in all neurons under the control of the ric-19 promoter, we could restore AFT to swsn-9(ok1354) mutants [10]. This result indicates that the SWI/SNF complex regulates gene expression in neurons that is important for AFT.
Our previous experiments used extra-chromosomal arrays to show that swsn-9 was required in neurons. Extrachromosomal arrays are multicopy DNA concatamers that are transmitted with varying frequency through mitosis and meiosis [21]. In order to have uniform transgene expression and normal diploid gene dosage, we generated single copy insertions of the neuronal expression construct Pric-19::swsn-9::GFP using the Mos-1 mediated single copy insertion (MosSCI) technique [22]. The MosSCI targeting vector contains a selectable marker; therefore we also generated single copy insertions of the empty vector as a control (Fig. 2a). We crossed these insertions into swsn-9(ok1354) and tested them for rescue of the swsn-9(ok1354) AFT defect (Fig. 2b). Wild type worms develop AFT, while swsn-9(ok1354) worms fail to develop AFT. We found that single copy insertions of the empty vector did not rescue AFT, while single copy insertions containing Pric-19::swsn-9::GFP did rescue the AFT defect of swsn-9(ok1354) null mutants.
In order to identify SWI/SNF-regulated neuronal genes, we performed RNA sequencing on swsn-9(ok1354) mutants harboring either the empty vector control or the neuronal swsn-9::GFP rescue construct. We collected five biological replicates on different days and rescue and control worms were reared in parallel. We isolated total RNA from young adult worms of each genotype. RNA sequencing libraries were prepared and sequenced by the Genomic Services Laboratory at Hudson Alpha (Huntsville, AL).
We assessed the correlation between biological replicates using principal component analysis and found that our biological replicates grouped together (Fig. 2c). The first two principal components accounted for 75% of the variance in the dataset, with principal component one (variation between sample types) accounting for 57% of the variance. We analyzed differential gene expression using DESeq2 [16] and found that 2412 genes were differentially expressed between swsn-9(ok1354) mutants with and without neuronal rescue (FDR ≤ 0.05) (Additional file 1). A volcano plot shows the distribution of DEGs (Fig. 2d). Consistent with our swsn-1(os22ts) data, we found that similar numbers of genes were downregulated (1157) as were up-regulated (1255).

SWI/SNF-regulated genes, in neurons, and in adults
The intersection of these two datasets identifies genes that are differentially expressed in neurons and in adults, in SWI/SNF mutants relative to wild type. Because we isolated mRNA from whole adult animals and we expect that the relevant gene expression changes will be in a small number of cells, we have not used any fold-change cutoff for this analysis. We identified 603 genes that were differentially expressed in the same direction in both swsn-1(os22ts) and swsn-9(ok1354) mutants compared to wild type (FDR ≤ 0.05); 393 genes were upregulated and 210 were down-regulated (Fig. 3a). These 603 genes are excellent candidates for mediators of the SWI/SNF effect on AFT.

GO term enrichment analysis
We performed an analysis to determine if particular functional classes of genes are overrepresented in our DEGs. We found that gene ontology (GO) biological process terms associated with metabolic processes, such as "mitochondrial ATP synthesis coupled electron transport" (GO:0042775) and "carboxylic acid catabolic process" (GO:0046395), were overrepresented in the swsn-1(os22ts) DEGs (Additional file 2). By contrast, we found a wide range of overrepresented GO biological process terms for the swsn-9(ok1354) DEGs, ranging from "regulation of cell shape" (GO:0008360) to "translation initiation" (GO:0006413) to "mitochondrial gene expression" (GO:0140053) (Additional file 2). Interestingly, we found that the GO biological process term "fatty acid metabolic process" (GO:0006631) was overrepresented in both the swsn-1(os22ts) and swsn-9(ok1354) DEGs, with 2.4 fold enrichment in the swsn-9(ok1354) DEG set and 1.8 fold enrichment in the swsn-1(os22ts) DEGs. Finally, we performed gene ontology GO term overrepresentation tests for genes at the intersection of our two datasets (Additional file 2). We found 7.5 times as many genes with the GO cellular process term "membrane rafts" (GO:0045121) and 3.4 times as many genes with the GO biological process term "innate immune response" (GO:0045087) as would be expected for a gene list of this size (Fig. 3b).
Comparison of swsn-1 and swsn-9 regulated genes SWI/SNF complexes are composed of multiple subunits that combine to form functionally distinct complexes [reviewed in [12,13]]. Our two RNA sequencing datasets utilized different SWI/SNF mutants: swsn-1 encodes a core subunit that is present in all SWI/SNF complexes, while swsn-9 encodes an accessory subunit that is specific to the PBAF complex in worms (Fig. 4a) [10,23]. We have an opportunity with these datasets to compare the effect of loss of different SWI/SNF subunits on gene a Constructs for single-copy rescue of swsn-9(ok1354) mutants in neurons. The empty MosSCI vector (top) serves as a negative control and contains the unc-119+ selectable marker. The neuronal rescue plasmid (bottom) expresses SWSN-9::GFP under control of the ric-19 promoter; it also contains the unc-119+ selectable marker. b Animals were tested on 0 mM and 400 mM exogenous ethanol. Relative speeds were calculated as treated over untreated speeds. The development of AFT is indicated by a statistically significant recovery of speed between 10 and 30 min of exposure. All genotypes on the same graph were tested simultaneously on the same plates. Wildtype (WT) animals develop AFT, while swsn-9(ok1354) mutants fail to develop AFT (n = 6). A single copy insertion of Pric-19::SWSN-9::GFP (rescue) was able to rescue the AFT defect of swsn-9(ok1354) mutants, while a single copy insertion of the empty MosSCI vector (control) did not rescue the AFT defect. Error bars indicate S.E.M. Paired one-tailed Student's t tests were used for statistical comparisons of speeds at 10 and 30 min. Oneway ANOVA, with Tukey's multiple comparison test, was used to compare initial sensitivity across the strains; no significant difference was observed. For indicated comparisons: ns, not significant; **p ≤ 0.01 c Principal component analysis with gene expression plotted relative to the first two principal components (PC1 and PC2). swsn-9(ok1354) mutant replicates containing the rescue or control constructs are most similar to replicates of the same genotype. d Volcano plot shows genes that are up-regulated (red) and down-regulated (blue) in swsn-9(ok1354) mutant neurons. Dashed lines indicate the FDR and fold change cutoffs (FDR ≤ 0.05 and fold change ≥1) expression. We examined the 2000 most variable genes in wild type, swsn-1(os22ts), and swsn-9(ok1354) using kmeans clustering (Fig. 4b-d; Additional file 3). As expected, we identified clusters containing genes that were up-or down-regulated in swsn-1(os22ts) (clusters A, D, and E) and swsn-9(ok1354) mutants (clusters B and G). Note that clusters B and G contain genes that are regulated by swsn-9 and also genes that respond to temperature, since the wild type and swsn-1(os22ts) worms were shifted from 15°C to 25°C prior to gene expression analysis, whereas swsn-9(ok1354) mutants were not. We also identified clusters of genes that are regulated similarly in both swsn-1(os22ts) and swsn-9(ok1354) mutants (clusters C and F). These genes are regulated by the PBAF complex in adult neurons. Cluster C contains genes that are normally activated, while Cluster F contains genes that are normally repressed by PBAF.
We asked if our PBAF target genes were regulated by PBAF in other systems. Chowdhury et al. identified 2464 genes that were differentially expressed in a renal cell carcinoma model upon re-expression of PBRM-1 [24]. We identified human homologs of the genes in clusters C and F using the DRSC Integrative Ortholog Prediction Tool (DIOPT) [25], retaining any genes with a DIOPT score greater than two. This resulted in lists of 336 and 312 unique human homologs of the genes in clusters C and F, respectively. We compared these lists of human genes with the DEGs found in the Chowdhury study and found that 11.2% of the cluster C and F homologs were regulated by PBRM-1 in cancer cell lines (Additional file 3).
Human homologs of SWI/SNF-regulated genes are associated with alcohol-related phenotypes We are interested in finding SWI/SNF-regulated genes that are mediating the effects of the SWI/SNF complex on human alcohol-related phenotypes. We identified human homologs of the 603 worm genes identified at the intersection of our two datasets using DIOPT [25]. This resulted in a list of 874 unique human genes. We generated a list of human genes for which there was a reported gene disease association in the literature using DisGenNet [26]. We included disease terms related to alcohol or substance misuse, consumption, and withdrawal. Among the 874 human homologs of our SWI/ SNF-regulated genes we found that 84 (9.6%) are linked a Differential gene expression analysis identified 6813 genes with differential expression between swsn-1(os22ts) and wild type adult worms (purple) and 2412 genes between swsn-9(ok1354) mutants with and without neuronal rescue (blue). Of the 603 DEGs that were regulated in the same direction in both comparisons, 392 were up-regulated, and 210 were down-regulated. b PANTHER GO biological process and cellular compartment terms enriched in this gene set; there were no significant GO molecular function terms. GO terms are plotted against the fold enrichment relative to the expected number of genes in lists of these sizes to one or more alcohol or substance use phenotypes in the DisGeNET database (Additional file 4). Next, we used the human gene list to query PubMed for publications linking these genes to problematic alcohol use (search terms include "alcoholism", "alcohol use disorder", or "alcohol dependence") or to alcohol more broadly ("alcohol" or "ethanol"). We found 418 genes with one or more alcohol publications and 224 genes b t-SNE map shows distinct clustering for seven k-means clusters. c Description of the seven clusters. d Heatmap of clustered genes; genotypes are indicated on the bottom with one or more alcohol phenotype publications (Additional file 4). We sought to identify epigenetic regulators among our SWI/SNF-regulated genes because epigenetic regulators, like SWI/SNF, often work together. We queried PubMed for any citations linking these genes to epigenetic regulation and found 518 genes with one or more PubMed citations containing the search term "epigenetic". Finally, we generated a list of 195 genes for which there is literature evidence for both alcohol misuse phenotypes and epigenetic mechanisms of action (Additional file 4). These 195 genes were ranked based on the number of citations linking the genes to epigenetics; the top twenty genes are reported in Table 1. These are our top candidate genes for future study.
The histone acetyltransferase CBP-1 is dose-dependently required for AFT Among our top candidate genes is the C. elegans gene encoding CBP, cbp-1, which is similarly homologous to two closely related human genes, CREBBP and EP300. Both CREBBP and EP300 appear in our top 20 genes related to human alcohol problems and epigenetics (Table 1), strongly supporting our interest in cpb-1. CBP homologs have well described roles in the nervous system response to alcohol in animal models [27][28][29][30][31], and CREBBP is associated with addiction in humans [32]. We reasoned that if SWI/SNF is regulating ethanol response behaviors in part through its regulation of cbp-1, then disrupting function of cbp-1 itself should alter ethanol response behaviors. cbp-1 expression was decreased relative to wild-type in swsn-1(os22ts) and swsn-9(ok1354) mutants (Fig. 5b), both of which are defective in acute functional tolerance (AFT), so we predicted that a loss of cbp-1 should decrease or eliminate AFT. We therefore tested two different strong loss-of-function alleles of cbp-1 in our behavioral assays (Fig. 5a). Mutants carrying homozygous cbp-1 loss-of-function alleles die, so we tested cbp-1 heterozygotes. We found that cbp-1(ok1491) heterozygotes did not develop AFT (Fig. 5c). We found that cbp-1(bm2)/balancer heterozygotes had reduced AFT, whereas cbp-1(bm2)/+ heterozygyotes had no AFT to ethanol (Fig. 5d). This supports a model in which SWI/SNF acts, at least in part, through its regulation of cpb-1 in ethanol response behaviors.

Discussion
The SWI/SNF chromatin remodeling complex is required in C. elegans for the development of acute functional tolerance to ethanol [10]. In this study, we identified the set of transcripts that are regulated by the SWI/SNF chromatin remodeling complex in adult neurons. To identify the transcripts under the regulatory control of swsn-9 in neurons, we generated a strain in which we provided swsn-9 expression only in neurons in an otherwise swsn-9(ok1354) mutant background (Fig. 1). By comparing the transcriptome of these animals to that of swsn-9(ok1354) mutant animals, we could find transcripts regulated by the SWI/ SNF complex in neurons. We found that 2412 genes were differentially expressed in this sample, suggesting that these genes are under the control of a SWSN-9 containing SWI/SNF complex in neurons. To identify transcripts that are under the regulatory control of swsn-1 in adults, we used the special temperature sensitive allele swsn-1(os22ts) to provide SWSN-1 function throughout development, then eliminate SWSN-1 function in adults (Fig. 2). We compared the transcriptomes of animals with and without SWSN-1 function in adulthood to identifiy candidate AFT  mediators. We found that 6813 genes are under regulatory control by swsn-1 in adults. These effects on gene expression may be direct or indirect; for example, some may be the result of SWI/SNF's regulation of one or more transcriptional repressors or activators that themselves regulate expression of the genes. We observed that more than two times as many genes were regulated by swsn-1 than by swsn-9 in our study; this was consistent with our expectations because SWSN-9 is an accessory subunit that is predicted to be found only in PBAF complexes, while SWSN-1 is a core subunit that is predicted to be present in both BAF and PBAF complexes. Further, this difference in the number of regulated genes may also be due in part to the fact that our swsn-9(ok1354) dataset is confined to neuronally expressed genes, whereas the swsn-1(os22ts) dataset includes all tissues. Interestingly, in addition to different numbers of genes, we also found that the swsn-1 and swsn-9 datasets consisted of mostly non-overlapping sets of genes. If these two SWI/SNF complex subunits were acting together for most or all gene regulation by PBAF, then we would expect that the swsn-9 neuronallyregulated dataset would largely be a subset of the swsn-1-pan tissue regulated dataset. In fact, we observed the contrary; most of the genes in each dataset do not appear in the other dataset.
There are several possible explanations for this lack of regulatory overlap. One possibility is that, in addition to acting together in some cases, these two subunits also participate in separate complexes that target different genes. This would imply that SWSN-9 can act with other as yet undetermined proteins to regulate transcription independent of its known function with SWSN-1. This would be a non-canonical function of SWSN-9, and while our data do not exclude this possibility, we do not favor this hypothesis. A second possibility is that swsn-1(os22ts) may be non-null at 25°C, so that it does provide some function in our restrictive conditions. In such a case, we would see DEGs disrupted by the swsn-9(ok1354) null mutation that are not disrupted by swsn-1(os22ts) because of its residual activity, but those genes nevertheless do require the function of SWSN-1. We favor a third explanation: in our experimental paradigm, swsn-1(os22ts) function is only disrupted in the adult, and its function is more normal during development. In contrast, swsn-9(ok1354) is non-functional throughout development. We hypothesize that the swsn-9-specific DEGs are regulated by SWI/SNF before adulthood. We do not observe their dysregulation by the swsn-1(os22ts) allele because in this work it functions normally during development. Thus, the swsn-9-specific DEGs represent developmental gene regulation defects that persist into adulthood. Our datasets provide a rich resource with which to examine the functions of these SWI/SNF target genes.
Although it remains possible that swsn-1 and swsn-9 regulate AFT through independent mechanisms, because we found that loss of either swsn-9 or swsn-1 function causes similar defects in AFT [10], we favor the idea that they act together in the SWI/SNF complex for AFT. Therefore, our primary interest was in genes that are regulated by both of these subunits. We found that a total of 603 genes were regulated in the same direction by both swsn-9 and swsn-1. These are candidate AFT mediators. To begin to characterize these genes, we first looked at overrepresented classes of genes, with the goal of identifying biological processes that are likely to be involved in AFT.

SWI/SNF regulates the membrane microenvironment
We were particularly intrigued by our observation that the most overrepresented category in the intersection of the swsn-9 and swsn-1 regulated genes are those annotated to be involved in membrane rafts, which are a form of membrane microarchitecture characterized by particular lipid comoposition, thick bilayer and the presence of cholesterol. Membrane microachitecture can provide a platform for regulating the activity of many membrane proteins, both directly and indirectly. For example, membrane characteristics such as lipid makeup, bilayer thickness, and cholesterol content have all been shown to modulate the activity of the BK channel (reviewed in [33,34]), which is a well characterized ethanol target (reviewed in [35,36]). In addition, membrane microdomains such as lipid rafts can regulate the access of proteins to effector molecules, thereby affecting their activity [37,38]. Importantly, we have previously demonstrated that cholesterol, a key component of lipid rafts, is required for the development of AFT [39]. In addition, we have shown that levels of the long chain ω-3 polyunsaturated fatty acid, eicosapentaneoic acid (EPA), tune the levels of AFT; EPA is a component of lipid rafts [40,41]. Taken together, these data point to a probable role for rafts in AFT. Our gene expression data implicate SWI/SNF complexes in the regulation of membrane raft components.
These data have led us to propose a molecular model for the development of AFT in which the ethanol sensitivity of ethanol target proteins, such as BK channels, can be modulated by changing the membrane milleu in which they reside. SWI/SNF regulation of factors that modify lipid rafts may be required if one of the mechanisms of AFT is for ethanol target proteins to be moved into or out of lipid rafts to normalize their function if they are exposed to ethanol. AFT in our system is observable within 30 min, and such modulation could be very quick, since it would not rely on transcription or changes in protein abundance. If there are appropriate lipid raft components available, then the system could be poised to respond to ethanol quickly when it is encountered.
SWI/SNF acts through cbp-1 to regulate AFT Generally, gene families have expanded in more complex organisms, so our 603 candidate C. elegans genes are homologous to 874 human genes. As a first analysis of our dataset, we asked if any of the homologs of the genes in our candidate dataset have been implicated in alcohol response behaviors in other studies, by searching PubMed for coincidence of the human gene name and alcohol response search terms such as 'alcohol dependence' and 'alcohol use disorder,' and then further narrowed our search by filtering these by a mention of epigenetic regulation. This narrowed our list to 195 genes (Additional file 4). We noted that among the top 20 of these genes (Table 1), two, CREBBP (encoding CBP) and EP300 (encoding p300), are homologs of the same C. elegans gene, cbp-1. Both CBP and p300 have been shown to be involved in ethanol effects in several systems [28,30,[42][43][44]. cbp-1 is also interesting to us because CBP is known to act in concert with SWI/SNF complexes in epigenetic modulation of gene transcription [45,46]. Together, these make cbp-1 an excellent candidate mediator of the SWI/SNF effect on AFT. We tested strains heterozygous for two different loss of function mutations in cbp-1 and found that cbp-1/+ animals had defects in the development of AFT, demonstrating that this SWI/SNF target gene is important in this process. We favor a model in which SWI/SNF can modulate the transcription of cbp-1, and, CBP-1 and the SWI/SNF complex act in concert to regulate the transcription of effector genes that are required for the development of AFT. CBP-1 may also interact with other transcription factors to regulate the expression of effector genes. Intriguingly, both p300 and CBP have direct interactions with human CTBP family members in complex regulatory relationships [47][48][49][50]. CTBP1 and CTBP2 are transcriptional repressors that are homologous to C. elegans ctbp-1. We previously found that ctbp-1 is required for normal AFT in worms [39], further suggesting a mechanism by which cbp-1 may play a role in AFT. Our ultimate goal is to identify the direct effectors of AFT, to further our molecular understanding of this important aspect of the physiological effects of alcohol.

Conclusions
This work describes the transcriptional targets of two subunits of the SWI/SNF chromatin remodeling complex in C. elegans. We generated one dataset of genes whose transcription is regulated by SWSN-9 in neurons. The second dataset is genes whose transcription requires SWSN-1 in adult tissues. Both of these genes are required for the development of acute functional tolerance, so we examined the intersection of these datasets, and identified 603 genes whose regulation is dependent on SWSN-1 in adults and SWSN-9 in neurons. Among these genes we expect to find the effectors of AFT. Pathway enrichment analysis identified genes involved in membrane rafts, suggesting that the regulation of membrane microarchitecture may play an important role in AFT. The candidate gene cbp-1 is homologous to mammalian genes that have been implicated in ethanol respose behaviors, and we found that cbp-1 function is required for normal AFT. This dataset provides a rich resource with which to examine the mechanisms underlying the development of acute functional tolerance to ethanol.

cbp-1 genetic analysis
Two cbp-1 alleles were used in this study. Each allele was backcrossed at least six times to wild type (N2) worms before testing. Homozygous cbp-1 loss-offunction alleles are lethal; therefore they were backcrossed alternately to N2 and a genetic balancer chromosome and they were tested as heterozygotes over a wild-type or balancer chromosome. To generate cbp-1(lf)/+ worms, balanced strains were crossed with wild type males and cbp-1/+ worms were identified in the F1 generation by the absence of a GFP marker on the balancer chromosome. During backcrossing, individuals were genotyped by PCR using primers flanking the mutations; deletions were detected by their shorter PCR product. All primers used in this study are listed in Additional file 5.

RNA sample collection
Five replicates were performed on different days; strains that were compared to one another were collected in parallel. We isolated mRNA from young adult worms in order to minimize embryonic transcripts in our samples. Wild-type and swsn-1(os22ts) strains (N2 and HS304) were reared at 15°C through the fourth larval stage (L4) and then shifted to 25°C; swsn-9(ok1354) strains (RA577 and RA629) were maintained at 20°C continuously. Approximately 250 L4 worms were placed onto a plate containing OP50 and allowed to develop until most of the population was at the young adult stage and had not yet begun to lay eggs. At this time, any remaining L4 worms were removed from the plates and young adult worms were washed from the plate with M9 medium. Each replicate (n = 1) consisted of this pool of approximately 250 worms. The worms were rinsed once with M9 and stored in Trizol (Ambion, Carlsbad, CA) at − 80°C until RNA preparation.

RNA sequencing and analysis
Total RNA was isolated using the miRNeasy kit with oncolumn DNase I digestion (Qiagen, Venlo, Netherlands). RNA integrity was assessed on four representative samples using the Experion Automated Electrophoresis Station (Bio-Rad, Hercules, CA). All samples had RQI values ranging from 9.8 to 10.
Indexed sequencing libraries were prepared from polyA(+) selected RNA and sequenced at the Genomic Services Lab at Hudson Alpha (https://gsl.hudsonalpha. org/index). The libraries were sequenced as 50-base, paired-end reads, to an average read depth of 25 million reads per sample using an Illumina HiSeq v4 2500 (Illumina, San Diego, CA). We used FastQC (https://www. bioinformatics.babraham.ac.uk/projects/fastqc/) t o examine the raw RNA sequencing data and found that they were of high quality and did not contain Illumina adapter sequences. Sequences were aligned to the C. elegans genome (Ensembl genome assembly release WBcel325) using Tophat2 version 2.1.1 [59], with Bowtie2 version 2.2.9 as the alignment algorithm. The GTF option was used to provide Tophat with a set of gene model annotations and the following parameters were specified (max-multihits 1, mateinner-dist 200, −I 18000 -I 40). Aligned reads were sorted and indexed using SAMtools [60]. Gene-based read counts were obtained using HTSeq version 0.6.1 [61], using the Caenorhabditis_elegans.WBcel235.86.gtf annotation file with the union overlap resolution mode. Differential expression was determined using DESeq2 version 1.18.1 with alpha set to 0.05 [16]. FPKM (Fragments Per Kilobase of exon per Million fragments mapped) values were obtained using Cufflinks version 2.2.1 [62]. Mean FPKM values are reported, any values for which the FPKM status is FAIL were excluded. To examine the variance among our samples and replicates, we performed principal component analysis on regularized log transformed data using the rlogTransformation and plotPCA functions in DESeq2 [24]. Pearson's correlation matrices were generated from read counts using iDEP [63] with no filtering for genes with minimum counts per million (CPM) and normalization using log 2 (CPM) (Additional file 5). Volcano plots were generated from read counts using iDEP [63] with parameters set to match our DESeq2 analysis (no CPM filtering, alpha = 0.05, fold change > 1). GO term enrichment for the differentially expressed genes (DEGs) was determined using the statistical overrepresentation test in PANTHER [64][65][66]. Gene lists were compared to all C. elegans genes in PANTHER using the GO biological process, cellular component, and molecular function datasets and Fisher's exact test with false discovery rate (FDR) correction (FDR < 0.05). Overrepresented GO terms were clustered into hierarchical groups; only the most specific subclass is reported. k-means clustering of gene expression was performed using iDEP [63], with filtering to remove genes with fewer than 0.5 CPM in at least five samples. The top 2000 most variable genes were chosen based on the distribution of standard deviations for all genes and k was set (k = 7) such that it produced distinct and largely non-overlapping clusters when visualized using a t-distributed stochastic neighbour embedding (t-SNE) map [67].

PubMed database searches
We created two web-based applications to facilitate searching the NCBI PubMed database with a large list of genes (www.geneinvestigator.com). The advantage of these applications is that they loop through the gene list, searching PubMed one gene at a time, thus providing search results on a gene by gene basis. They include a fully customizable search expression that is concatenated with the gene name using the Boolean operator AND. The search expression can include any number of terms linked by the Boolean operators AND, OR, and NOT. The 'counts' application returns a table with the number of citations matching each gene. The 'retrieve' application retrieves the twenty most recent citations for each gene.

Locomotion assays
Locomotion tracking and analysis was performed as described previously [68,69], with minor modifications. We treated the animals with 400 mM exogenous ethanol. The waxy cuticle of C. elegans is resistant to the passage of many chemicals, including ethanol [70,71] so that in these experiments, the tissue concentration of ethanol is approximately 12% of the exogenous concentration, or approximately 48 mM [71]. Worms continue to accumulate ethanol over the course of the assay, so that there is a slightly higher concentration at 30 min relative to 10 min; behavioral improvement in this time is not due to pharmacokinetic effects, but rather to the development of acute functional tolerance to ethanol [71]. Ten first day adult worms were placed in copper rings that had been melted into the surface of agar plates containing 0 mM or 400 mM ethanol. Groups were matched so that worms from the same population were always treated with 0 mM and 400 mM ethanol in parallel. The locomotion of the worms was recorded for 2 min at 10 min of exposure and again at 30 min of exposure. The average speed was calculated for each group of worms (a treatment group of 10 worms yielded (n = 1)) using Image Pro Plus software (Media Cybernetics, Inc., Rockville, MD). Six trials were performed for each genotype. Relative speeds were calculated (treated speed/ untreated speed × 100) for the 10 and 30 min exposure time points. Initial sensitivity to ethanol is measured by the relative speed of the worms after 10 min of exposure. Development of acute functional tolerance (AFT) was determined as a statistically significant difference between the relative speed of the animals after 10 min of ethanol exposure and the relative speed of the same animals after 30 min of ethanol exposure. One-tailed paired Student's t tests were used for statistical comparisons of the relative speeds at the two time points. Comparisons of initial sensitivity and recovered speed across strains were made using 1-way ANOVA, with Tukey's multiple comparison post-hoc test. Only animals that were tested simultaneously on the same plates were compared to each other.
Additional file 3: k-means clustering of wild type, swsn-1, and swsn-9 gene expression data. Included in this file are: 1-Distribution of gene standard deviations across all three datsets. 2-k-means clustering of the top 2000 most variable genes (k = 7). 3-Results of DIOPT analysis for cluster C genes. 4-Results of DIOPT analysis for cluster F genes. 5-Cluster C and F genes compared with PBRM1 targets identified in [46]. 6-GO biological process terms that are overrepresented in each of the clusters. 7-GO cellular component terms that are overrepresented in each of the clusters. 8-GO molecular function terms that are overrepresented in each of the clusters.
Additional file 4: Analysis of human homologs of SWI/SNF-regulated genes. Included in this file are: 1-Results of DIOPT analysis for all genes in the intersection between swsn-1(os22ts) and swsn-9(ok1354) datasets. 2-Disease terms related to alcohol or substance use in the DisGeNET database. 3-Human homologs with gene disease associations (GDA) for alcohol/substance use terms on DisGeNET. 4-Citations on PubMed linking the human homologs to alcohol, alcohol-related diseases/diagnoses, and epigenetics. 5-Genes for which there are PubMed citations for alcohol phenotypes and epigenetics. 6-Top 20 genes from the PubMed analysis mapped back to C. elegans genes. 7-PubMed citations for the top 20 genes; limit of 20 citations returned.