Differential Regulation of Alcohol Consumption and Reward by the Transcriptional Cofactor LMO4

Abstract

Repeated alcohol exposure leads to changes in gene expression that are thought to underlie the transition from moderate to excessive drinking. However, the mechanisms by which these changes are mobilized to a maladaptive response that leads to alcohol dependence are not well understood. One mechanism could involve the recruitment of transcriptional co-regulators that bind and modulate the activity of several transcription factors. Our results indicate that the transcriptional regulator LMO4 is one such candidate regulator. Lmo4-deficient mice (Lmo4gt/+) consumed significantly more and showed enhanced preference for alcohol in a 24-hour intermittent access procedure. shRNA-mediated knockdown of Lmo4 in the nucleus accumbens (NAc) enhanced alcohol consumption whereas knockdown in the basolateral amygdala (BLA) decreased alcohol consumption and reduced conditioned place preference to alcohol. To ascertain the molecular mechanisms that underlie these contrasting phenotypes, we carried out unbiased transcriptome profiling of these two brain regions in wild type and Lmo4gt/+ mice. Our results revealed that the transcriptional targets of LMO4 are vastly different between the two brain regions, which may explain the divergent phenotypes observed upon Lmo4 knockdown. Bioinformatic analyses revealed that Oprk1 and genes related to the extracellular matrix (ECM) are important transcriptional targets of LMO4 in the BLA. Chromatin immunoprecipitation (ChIP) revealed that LMO4 bound Oprk1 promoter elements. Consistent with these results, disruption of the ECM or infusion of NorBNI, a selective kappa opioid receptor (KOR) antagonist, in the BLA reduced alcohol consumption. Hence our results indicate that an LMO4-regulated transcriptional network regulates alcohol consumption in the BLA.

Introduction

Alcohol use disorder (AUD) is a chronic relapsing neuropsychiatric condition characterized by compulsion to seek and take alcohol, loss of control in limiting intake, and the emergence of a negative emotional state when alcohol is unavailable. A key mechanism in the transition from moderate to excessive alcohol use involves gene expression-dependent changes in neuronal structure and function (1-7). Recent studies have leveraged advances in next generation sequencing to identify genes and gene networks that are altered by chronic alcohol consumption (1, 3-5, 8, 9) and contribute to alcohol dependence (7). A major gap in our knowledge is understanding how these diverse transcriptional changes are orchestrated into the maladaptive response that characterizes AUD.

One mechanism could be through transcription factors, which act as “master switches” of gene expression. Alcohol recruits some of these transcription factors by increasing their activity or expression, for example by altering neuronal activity or activating key signaling pathways (7). Another mechanism could involve transcriptional co-regulators that interact with different transcription factors to activate or repress gene expression. To date only two transcriptional co-regulators have been shown to affect alcohol consumption, namely the cAMP response element-binding protein (CREB) (6) and the PPAR gamma coactivator A (10). Here we identify a third transcriptional regulator Lim-Only 4 (LMO4) that regulates alcohol consumption,

LMO4 does not bind DNA directly, but instead regulates gene expression by functioning as a scaffold for DNA-binding proteins, such as transcription factors and regulatory co-factors (11). LMO4 can also act as a calcium–sensitive transcriptional co-activator to regulate neuronal gene expression in an activity-dependent manner (12). Interestingly, LMO4 is enriched in brain regions that play key roles in addiction, including the nucleus accumbens (NAc) and basolateral amygdala (BLA) (13). LMO4 in the NAc has been shown to modulate the rewarding and psychomotor properties of cocaine (13), and in mice with reduced levels of LMO4 in the BLA, reward-paired cues elicit approach to the reward port but do not serve as conditioned reinforcers to support new learning (14). These results suggest that LMO4 regulates neural processes that function to imbue reward-paired cues with motivational value and support behaviors characteristic of addiction.

Here we investigated the role of LMO4 and LMO4-dependent gene networks in regulating alcohol consumption in mice. Our results indicate brain region-specific roles for LMO4; in the BLA LMO4 enhances alcohol consumption, whereas in the NAc it reduces alcohol consumption. To understand the molecular basis for this differential regulation, we performed an unbiased transcriptome analysis and found that transcriptional targets of LMO4 are strikingly different in these two brain regions. Since reducing LMO4 in the BLA suppressed drinking, which is a result of potential translational value, we further investigated transcriptomic changes in this region and their causal contributions to alcohol drinking. Bioinformatics analysis identified genes related to the extracellular matrix (ECM) and the kappa opioid receptor (Oprk1) as important LMO4 targets. Disruption of the ECM or injection of a kappa opioid receptor antagonist into the BLA reduced the high levels of alcohol consumption observed during an intermittent alcohol access procedure. Hence, our results indicate an important role for LMO4-regulated transcriptional programs in the BLA in promoting excessive alcohol consumption.

Materials and Methods

Subjects:

We used 8- to 12-week-old, male, heterozygous Lmo4gt (Lmo4gt/+) mice and their wild type littermates on a C57BL6/J background for behavioral and biochemistry experiments, and 8- to 12-week-old C57BL/6J mice for shRNA and drug infusion experiments. Mice were group housed until surgery, after which they were housed individually. Mice used for drinking studies, behavioral experiments, and tissue collection for RNAseq studies were housed in 12h/reverse dark cycle with lights off at 11AM and lights on at 11PM. Mice were individually housed for drinking studies. For all other studies mice were group housed with ad lib access to food and water. All procedures were approved by the University of Texas at Austin Institutional Animal Care and Use Committee.

Two-bottle choice intermittent access alcohol consumption:

We used the intermittent access (IA) ethanol drinking procedure described by Hwa and colleagues (15). Briefly mice were singly housed and given 24-h access to escalating concentrations of alcohol (3%, 6%, and 10%) three days per week (Monday, Wednesday, and Friday) in the first week. Subsequently, alcohol concentrations were increased to 20% (v/v) ethanol three days per week for four weeks. Alcohol bottles were made available three hours after the onset of the dark cycle. The position of alcohol and water bottles were alternated every session. The amount of alcohol (g/kg) consumed and preference for ethanol, calculated by dividing the amount of ethanol containing water consumed by total fluid consumed was measured every 24h. Consumption of sucrose and quinine were tested as described previously (10). For studies of KOR, mice were injected i.p. with 5 mg/kg U50, 488 (Tocris) dissolved in sterile saline 6 weeks after the start of the drinking procedure. Alcohol consumption as measured 3 hours after U50, 488 injection.

Conditioned place preference for alcohol:

Conditioned place preference (CPP) was measured using a two-chambered apparatus as described (16). Please refer to supplementary methods for detailed procedure.

Stereotaxic surgeries, canula implantation, and drug delivery:

Male C57BL/6J mice were infused bilaterally with 1 μl of lentivirus (10 ⁷–10 ⁸ pg/μl) using a Hamilton syringe, as described previously (17, 18).The coordinates for BLA were A/P −1.6, M/L ±3.25, D/V −4.5 from skull. The coordinates for the NAc were A/P +1.43, M/L ±0.9, and D/V −4.55 from skull. After injection, mice were allowed to recover for 2 weeks before commencing alcohol consumption or CPP experiments. A separate group of mice were implanted with stainless steel guide cannulae (Plastics one) aimed bilaterally at the BLA (AP −1.6, ML ±3.25, DV −3.5 from skull). Cannulas were affixed to the skull with dental acrylic (H00325, Coltene, Cuyahoga Falls, OH). Cannula-implanted rats were administered the antibiotic cefazolin (100 mg/kg, s.c.) and singly housed during a 1–2-week recovery period before experiments. Drugs were infused using an injector whose tip protruded 1mm past the guide cannula. chABC was dissolved in 1XPBS +0.01%BSA at a concentration of 50U/ml and a volume of 0.5μl was administered per side. Control mice received infusions of vehicle (PBS+0.01%BSA). Cannulated mice were allowed to consume alcohol for 6 weeks in the IA procedure before infusing ChABC. Alcohol consumption was measured 24h after ChABC infusion. NorBNI was dissolved in saline at a concentration of 5mg/ml and a volume of 0.5μl (2.5μg/side) was administered per side. Control mice received 0.5μl saline infusions. Cannulated mice were allowed to consume alcohol for 6-7 weeks before infusion of Nor-BNI. Alcohol consumption was measured 24h after Nor-BNI infusion.

RNA Extraction:

Detailed procedures are described in supplementary methods.

RNA sequencing, read alignment, and analyses:

Detailed procedures are described in supplementary methods.

Quantitative PCR (QPCR)

RNA was extracted from wild type and Lmo4gt/+ mice as outlined above. The extracted RNA was amplified using the Taqman PreAmp kit from Applied Biosystems (Thermo Fisher Scientific, Rockford, IL) according to manufacturer’s instruction. Sixty nanograms of RNA were reverse transcribed using the High Capacity Reverse Transcription kit from Applied Biosystems (Thermo fisher Scientific). The preamplification reaction comprised of 15ng of reverse transcribed cDNA, 0.2X Taqman probe, and 1X Taqman PreAmp Mastermix. The preamplified product was diluted 1:20 in 1X TE buffer and 2.5μl of the preamplified product was used in a real time PCR amplification reaction which also contained 5μl of 1X SSoAdvanced Universal Probes Mastermix (Biorad, Hercules, CA), 0.5μl of 20X Taqman primer/probe mix in a total volume of 10μl. Relative mRNA levels of target genes were determined using BIORAD software as previously described (19, 20). Data were normalized to multiple endogenous control genes including GusB, Tfrc, Tubb2b, and Rplp0. Catalog numbers of Taqman gene expression assays used are listed in Table S3.

Chromatin immunoprecipitation

ChIP was performed essentially as outlined (21). Detailed procedure is described in supplementary methods.

In situ hybridization

We used RNAScope (Advanced Cell Diagnostics, Newark, CA) to perform fluorescence in situ hybridization and examine colocalization between Lmo4, Oprk1, Sulf2, Drd1, and Drd2. Detailed procedures are described in supplementary methods.

Results

LMO4 plays a dynamic role in alcohol consumption and reward

In a 24h intermittent access, two-bottle choice drinking procedure (IA) (13), Lmo4gt/+ mice, which express 50% less LMO4 than wild type counterparts (13), showed enhanced consumption and preference for alcohol (Figs. 1A, B) compared with wild type littermates [F_{Genotype x Day} (11, 308) = 2.61, P = 0.003]. No differences were observed in total fluid intake between the two genotypes (Fig. 1C). In contrast, there were no genotypic differences in alcohol intake in separate cohorts of mice that underwent 4h-limited, IA access (Fig. 1D) (22, 23) or 24h-continuous access, two-bottle choice (data not shown) procedures. These results suggest that LMO4 only affects high levels of ethanol intake achieved by the 24h-access IA drinking procedure. Two-bottle choice sucrose (14) saccharin, and quinine consumption (Fig. 1E, F) were not different between wild type and Lmo4gt/+ mice.

Figure 1: Lmo4 haploinsuffficiency increase alcohol consumption and preference.

A) Lmo4gt/+ mice consumed more ethanol than wild type mice in the 24h IA procedure. *P < 0.05, Bonferroni test, n = 15/genotype. B) Lmo4gt/+ mice also showed enhanced ethanol preference, *P < 0.05, Bonferroni test, n = 15/genotype. C) Total fluid intake was not different between the two genotypes. D) Ethanol consumption did not differ between the two genotypes in a 4h limited access procedure. E) Saccharin and F) quinine consumption were not different between the two genotypes (n = 5-6/genotype)

Since LMO4 in the NAc promotes behavioral sensitization to cocaine (13), we tested whether it influences alcohol consumption. Lmo4 is expressed in both D1 and D2 receptor expressing neurons in the NAc (Fig. 2A). Knockdown of NAc LMO4 by RNA interference (13) modestly enhanced alcohol consumption [F_{shRNA x Day} (14, 322) = 1.942, P = 0.0218; Fig. 2C], but did not increase alcohol preference (Fig. 2D). There was no effect of NAc knockdown on total fluid, sucrose, or quinine consumption (Fig. 2E-G). In contrast, LMO4 knockdown in the dorsolateral striatum did not alter alcohol consumption or preference (Fig. S1A and B).

Figure 2: Knockdown of LMO4 in the NAc increases alcohol consumption.

A) In situ hybridization revealed that Lmo4 is expressed in both D1 and D2 containing neurons in the NAc. Scale bars 50μm. B) Representative image of lentivirus injection into the NAc. C) Knockdown of LMO4 in the NAc increased alcohol consumption. *P < 0.05, Bonferroni test. D) Knockdown of LMO4 in the NAc did not change preference or E) total fluid intake. F) Saccharin and G) quinine consumption were not different between the two treatments (n = 11/group).

We have previously found that LMO4 knockdown in the BLA impairs cue-reward learning (14). Several studies point to a role for the BLA in regulating both appetitive and consummatory aspects of alcohol reward (24). Hence, we knocked down LMO4 in the BLA by RNA interference and examined effects on alcohol consumption. Surprisingly we found that BLA knockdown reduced alcohol consumption [F_{shRNA x Day} (14, 226) = 1.739, P = 0.0483; Fig. 3B] and preference [F_shRNA (1, 19) = 5.074, P = 0.0363; Figs. 3C). No changes were observed in total fluid, saccharin, or quinine intake (Fig. 3D-F). Given that LMO4 in the BLA supports cue-reward learning (14), we predicted that Lmo4 knockdown in the BLA might reduce conditioned place preference (CPP) for alcohol. We used DBA/2J mice in these experiments as they develop robust CPP to ethanol (25). As predicted, LMO4 knockdown in the BLA prevented ethanol CPP [F_{Conditioning x shRNA} (1, 15) = 5.417, P = 0.0343; Fig. 3G]. In summary, knockdown of LMO4 in the BLA reduces both alcohol consumption and reward, a phenotype opposite to that of global haploinsufficiency or selective knockdown in the NAc.

Figure 3: Knockdown of LMO4 in the BLA reduces alcohol consumption and conditioned place preference.

A) Representative image of lentivirus injection into the BLA. Ethanol consumption (B) and preference (C) are reduced by Lmo4 knockdown in the BLA. *P < 0.05, n = 10-11/group. D) Total fluid intake was not different between the two groups. LMO4 knockdown in the BLA did not alter sucrose (E) or quinine (F) consumption. (G) LMO4 knockdown reduced ethanol CPP. E) *P < 0.05.

Transcriptional targets of LMO4

Since knockdown of LMO4 in the BLA and the NAc produced opposite effects on alcohol consumption, we predicted that the transcriptional targets of LMO4 in the BLA and NAc are different. To investigate this hypothesis, we carried out unbiased transcriptome profiling in wild type and Lmo4gt/+ mice using RNAseq. This experiment was performed in Lmo4gt/+ mice rather than after knockdown with shRNA because of technical challenges involved in isolating and extracting RNA from virally transduced neurons. Furthermore, since Lmo4gt/+ mice are heterozygous, we anticipated that developmental compensation would be minimal.

We identified 1,063 genes that were differentially expressed (P < 0.05) between wild type and Lmo4gt/+ mice in the BLA (see Table BLA_NAC_DEgenes_withsymbol.xlsx in supplementary Data). Of these, 480 were significantly upregulated while 583 were significantly downregulated in Lmo4gt/+ mice compared with wild type mice (Fig. 4A and B). These results confirm that LMO4 can both activate and repress transcription (11). We chose 9 of these differentially expressed genes for validation by an orthogonal method, QPCR. Criteria for selecting these genes included expression level in the BLA (normalized counts that were greater than 100), log₂ fold change of 0.2 or greater, and association with alcohol-related phenotypes as determined by the INIA Texas Gene Expression Database and Gene Weaver. We successfully validated that 7 out of 9 genes (*P < 0.05, unpaired t-test) were differentially expressed (Fig. 4D and Fig. S2A and B). The fold changes obtained by QPCR matched the fold changes observed by RNAseq (Fig. 4E)

Figure 4: Transcriptome analysis identifies ECM-related genes as transcriptional targets of LMO4 that regulate alcohol consumption.

A) MA plots depicting log₂ fold changes in the BLA and the NAc. B) ~ 1000 genes were differentially expressed in the BLA and NAc of wild type and Lmo4gt/+ mice of which some were upregulated and others were downregulated. C)There were only about 48 genes that were common between the two brain regions. D) We validated six out of nine genes in the BLA by QPCR . *P < 0,05, unpaired t-test. n = 9-11 mice/group/brain region. Expression relative to the housekeeping gene Gus B is shown. E) Heatmap comparing log₂ fold changes obtained by QPCR and RNAseq. F) WGCNA analysis identified 20 modules of which 3 were significantly correlated with LMO4 knockdown. Of these, the yellow module was particularly intriguing as it contained Lmo4. Module eigen gene analysis revealed that genes in this module are downregulated upon Lmo4 knockdown. G) QPCR validation of 4 of the 5 differentially expressed ECM genes by QPCR. Expression relative to Gusb is shown. P < 0.05, n = 9-11/genotype. H) Colocalization of Lmo4 and Sulf2 mRNA in the BLA. Arrows point to cells expressing both Lmo4 and Sulf2. Scale bar 50μm. J) Infusion of ChABC into the BLA transiently reduced alcohol (J) but not water (K) consumption (**P < 0.01, n = 14-15 mice/group). I) PNNs regenerated 1 week after ECM disruption in the BLA (***P < 0.001, *P < 0.05, n = 3 - 4 mice/group).

We next performed RNAseq on NAc tissue to determine if the transcriptional targets of LMO4 in the NAc differ from those in the BLA. We identified 984 differentially expressed genes (P < 0.05, see Table BLA_NAC_DEgenes_withsymbol.xlsx in supplementary Data) in the NAc, of which 455 were upregulated and 529 were downregulated in Lmo4gt/+ mice compared with wild type mice (Fig. 4B). We selected 11 genes for validation using QPCR. We were able to confirm the differential expression of 6 of these 11 genes (Fig. S3A, B, and C). Interestingly, we observed only 48 genes that were common between the NAc and the BLA (Fig. 4C). Most were similarly up or downregulated in both of these brain regions with the exception of six genes (Table 1). These results indicate that the vast majority of transcriptional targets of LMO4 are very different in these two brain regions. Only a few LMO4 regulated genes are common to both brain regions, and of these few, some are regulated in opposite directions. These regional differences in transcriptional landscape are likely to underlie the divergent alcohol consumption phenotypes we observed after knockdown of LMO4 in the BLA versus the NAc.

Table 1:

List of genes which were differentially expressed in both the BLA and NAc but were regulated in an opposite manner in the two brain regions.

Gene Name	BLA	NAc
NPAS1	Up	Down
Smoc1	Up	Down
Alox 5	Down	Up
Prdm8	Up	Down
Igfbp2	Up	Down
B4galnt3	Down	Up

Because the BLA drinking phenotype was more robust and reducing alcohol consumption is a therapeutic goal for AUD treatment, we focused our attention on identifying mechanisms by which LMO4 regulates BLA gene expression. We took a systems approach using weighted gene co-expression network analysis (WGCNA) (26) to identify gene networks and cellular processes that were responsive to LMO4 downregulation in the BLA. Dendrograms representing the intercorrelation among genes were split into 20 modules with minimum module size of 100 using the dynamic tree cut method with a cutting height of 0.99 (Fig. S4A). We used modulepreservation statistic to ensure that the networks generated were robust (Fig. S8). Of these, only three modules (yellow, greenyellow, and purple) were significantly correlated with LMO4 deficiency (Fig. S4B). The yellow module was particularly intriguing as it contained Lmo4. Genes in this module were significantly downregulated in Lmo4gt/+ mice compared with wild type mice (*P < 0.05, Fig. 4F). Gene ontology analysis using both Genemania (Table S1) and DAVID (not shown) revealed that the yellow module was significantly enriched for genes related to the ECM (Table S1). Further the yellow module also contained the Oprk1 gene, which encodes the kappa opioid receptor (KOR). Of the ten ECM-related genes that were enriched, five were differentially expressed in our RNAseq data. We were able to validate differential expression for 4 of these 5 genes by QPCR (Fig. 4G and Fig. S5). These genes were collagen subunit 5a2 (Col5a2), hyaluronan and proteoglycan link protein 2 (Hapln2), and the matrix remodeling enzymes a disintegrin and metalloproteinase with thrombospondin motif 2 (Adamts2) and sulfatase 2 (Sulf2). Consistent with this result, we found that Sulf2 and Lmo4 mRNA are co-expressed in BLA pyramidal neurons (Fig. 4H).

Genes regulated by LMO4 in the BLA promote alcohol drinking

If LMO4 regulation of gene expression in the BLA is causally related to alcohol drinking, then manipulation of its targets should also affect drinking. In the brain the ECM is organized into loose ECM that surrounds the parenchyma of neurons, and highly specialized ECM structures called perineuronal nets (PNNs) (27). PNNs ensheath the soma, distal dendrites, and proximal axon segments of both pyramidal and interneurons in the BLA (28). In adult brain, PNNs regulate synaptic plasticity and play a role in fear and drug memories (29, 30). If LMO4 regulated transcription is important for the maintenance of PNNs, then disrupting them might reduce ethanol consumption. To examine this hypothesis, we disrupted the PNNs in the BLA by microinjecting the ECM degrading enzyme chondrotinase ABC (29). We cannulated the BLA of new groups of mice and subjected them to the IA drinking procedure for 4-5 weeks to establish a stable baseline level of drinking. We then infused ChABC 24h before ethanol bottles were re-introduced. Control mice received infusions of vehicle (PBS+0.01%BSA). We found that ChABC significantly reduced alcohol consumption 24h post infusion [F_{Time x Drug interaction} (2, 54) = 3.568, P=0.0350; Fig. 4J). We found a significant main effect of Time on water consumption [F_Time (2,54) = 6.258, P = 0.0036] but no Time x Drug interaction (Fig. 4K). However, the effect of ChABC was transient, and alcohol drinking returned to baseline levels 72h post infusion (4J). To determine the reason behind this transient effect, we examined the time course of PNN degradation and reappearance. We found that ChABC infusion caused near complete dissolution of PNNs in the BLA 24h after infusion, but they reappeared almost completely 1 week later (Fig. 4I). These results suggest that ChABC suppression of alcohol consumption is short-lived in part because PNNs can rapidly reassemble.

The LMO4-containing module also included Oprk1, the gene encoding the KOR, which is particularly intriguing because KORs are known regulators of alcohol consumption (31-36). The abundance of Oprk1 transcripts was reduced by approximately 50% in Lmo4gt/+ mice compared with wild type mice. Within the BLA, KOR is expressed on pyramidal projection neurons and regulates nicotine seeking behavior (37). In situ hybridization revealed that 50% of Lmo4 expressing neurons also express Oprk1. (Figs. 5A-C). We performed ChIP to determine if LMO4 physically interacts with Oprk1. To find structural evidence of this interaction, we used existing open chromatin data on ENSEMBL, including DNAse 1 hypersensitivity sites and ATAC-Seq (assay for transposase accessible chromatin-Seq) data to identify putative promoter/enhancer regions of Oprk1. We then designed primers to amplify 150bp segments spanning this region. Normal rat IgG was used as a non-specific antibody control. We established that the LMO4 antibody was able to immunoprecipitate LMO4 crosslinked to chromatin by performing chromatin immunoprecipitation (ChIP) followed by western blot analysis (Fig. 5D). We next performed ChIP followed by QPCR from amygdala punches and found significant enrichment of four Oprk1 promoter sequences with the LMO4 antibody but not with rat IgG, suggesting that LMO4 binds in a complex with the Oprk1 promoter (*P < 0.05, Fig. 5F). Based on these results we conclude that Oprk1 is a novel transcriptional target of LMO4 in the BLA.

Figure 5: BLA KORs promote alcohol consumption.

A) Co-localization of Oprk1 and Lmo4 mRNA in the BLA. We used dual fluorescence RNAScope to ascertain colocalization of Oprk1 and Lmo4 expression in the BLA. B) Higher magnification image of colocalization is shown. Scale bar 20μm. C) Colocalization was quantified using Image J. Approximately 50% of Lmo4 positive cells overlap with Oprk1. Arrows point to expression overlap. D) The LMO4 antibody but not a nonspecific rat IgG was able to immunoprecipitate LMO4 from formaldehyde crosslinked forebrain lysates. E) Schematic of Oprk1 promoter region with binding sites for QPCR primers. F) Significant enrichment of some of KOR promoter elements with LMO4 antibody compared to Rat IgG control. Mean +/− SEM from three separate experiments is shown. *P < 0.05, n = 3/group. Nor-BNI infusion into the BLA significantly reduced alcohol (G) but not water (H) consumption. *P>0.05, n = 12-14/group. I) The effect of U50,488 on alcohol consumption 3hrs after injection was measured and found to be significantly attenuated in shLmo4-injected mice compared to shCon. Dotted line represents amount of alcohol consumed in vehicle-injected mice. * P < 0.05, one-tailed paired student’s t-test, n = 9/group.

We next assessed the contribution of KORs in the BLA to IA drinking. We infused a KOR antagonist, norbinaltorphimine (nor-BNI) into the BLA of mice that underwent the IA procedure for 6 weeks and found that it reduced alcohol consumption by almost 50% 24h after infusion [F_{Drug x Time} (2, 48) = 5.343, P = 0.0080; Fig. 5G]. Drinking continued to remain low in nor-BNI infused mice 72h after infusion while water consumption remained unchanged (Fig. 5G-H). We found a significant main effect of Time on water consumption [F_Time (2,48) = 4.641, P = 0.0144] but no Time x Drug interaction (Fig. 5H). To determine whether KORs function downstream of LMO4 in the BLA to regulate alcohol consumption, we examined the ability of U50,488, a KOR agonist to enhance drinking in mice with LMO4 knockdown in the BLA. We found that while U50, 488 enhanced alcohol consumption 3 hours after injection in shCon-injected mice, its effects were significantly attenuated in shLMO4 injected mice [Fig. 5I, *, p<0.05, one-tailed student’s t-test]. Together these findings imply that reduced KOR expression contributes to decreased alcohol consumption after LMO4 knockdown in the BLA.

Discussion

Our results identify a dynamic role for the transcriptional regulator LMO4 in alcohol consumption. We observed contrasting effects of global and brain-region specific downregulation of LMO4 on this phenotype. Global haploinsufficiency of LMO4 resulted in a selective enhancement of alcohol consumption and preference. While regional knockdown of LMO4 in the NAc also enhanced alcohol drinking, the effect was small and was not associated with increased preference. In contrast, knockdown of LMO4 in the BLA reduced alcohol consumption and preference. The drinking phenotype in Lmo4gt/+ mice is the net effect of reduced LMO4 in different brain regions as well as effects of Lmo4 haploinsufficiency during development, although we suspected that developmental compensation was minimal since Lmo4gt/+ mice still have one functional Lmo4 allele. Our brain region-specific results are reminiscent of those for BDNF where attenuation of BDNF-TrkB signaling in different brain regions (38) or different cell types within the same brain region (39) has contrasting effects on cocaine CPP.

A plausible explanation for brain-region specific effects on behavior could be that the transcriptional targets of LMO4 are different in the BLA and NAc. Indeed, we found evidence that this is the case. Unbiased transcriptome profiling of BLA and NAc revealed very few overlapping genes indicating that LMO4 transcriptional targets differ widely between these two regions. Our results are consistent with physiological observations that LMO4 knockdown has differential effects on neuronal firing in different brain regions. For example, knockdown of LMO4 in single-minded-one neurons in the hypothalamus reduces excitability (40) whereas knockdown in BLA pyramidal neurons increases excitability (18). How increased excitability in BLA pyramidal neurons reduces drinking and whether LMO4 affects neuronal excitability in the NAc should be determined in future studies.

We focused our functional validation of LMO4 targets on the BLA because: 1) the LMO4 drinking phenotype in the BLA was stronger than in the NAc, 2) a manipulation that decreases drinking may reveal mechanisms that are amenable to translation, and 3) LMO4 in the BLA regulates cue-reward learning and dopamine responses in BLA pyramidal neurons (8). Further, the mechanism by which LMO4 enhances alcohol consumption in the NAc may be more complex as LMO4 is expressed in both D1 and D2-containing neurons, which are known to regulate alcohol consumption in opposing manner (41). Future studies will selectively knock down LMO4 in D1 or D2 neurons in the NAc and examine effects on alcohol consumption.

Among transcriptional targets, we were intrigued by Oprk1 since its expression was reduced by approximately 50% with LMO4 downregulation. Oprk1 is a known regulator of alcohol consumption (31, 32, 35), and the dynorphin/KOR system is thought to contribute to the development of a negative affective state that promotes excessive alcohol drinking (33, 34). Mice lacking KORs show decreased alcohol consumption (42) and KOR antagonists decrease alcohol consumption in two–bottle choice and binge drinking procedures (32) as well as operant alcohol self-administration in alcohol-dependent animals (35). We found here that approximately half of LMO4-expressing cells in the BLA express KORs and that KORs are direct transcriptional targets of LMO4 in these cells. Moreover, microinfusion of nor-BNI into the BLA reduced alcohol consumption and preference, and LMO4 knockdown in the BLA attenuated the ability of a KOR agonist to enhance drinking. These results suggest that LMO4-mediated expression of KORs in the BLA is necessary for high levels of alcohol intake in mice.

In addition to analysis of differential gene expression, we used WGCNA (26) to identify gene networks and cellular processes that were responsive to LMO4 downregulation in the BLA. We found that the network module most significantly correlated with LMO4 knockdown was enriched in genes related to the ECM. Several studies suggest that manipulating ECM components can alter the cellular and behavioral response to drugs of abuse including alcohol (43-47). Alcohol in particular is a potent regulator of some ECM proteins. Alcohol exposure and withdrawal increases the levels and activity of several extracellular proteases including matrix metalloproteinase-9 (MMP-9) and MMP-9 inhibition reduced escalation of alcohol consumption (48, 49) and tissue plasminogen activator (tPA) (45). Chronic ethanol exposure increases the density of PNNs and levels of critical PNN constituents such as aggrecan, brevican, and phosphacan in the insular cortex (50). However, the ECM is a supramolecular assembly of many proteins and the effects of alcohol on many other ECM constituents are unknown.

Of the 10 ECM–related genes found in the yellow module, we found that five were significantly differentially expressed in Lmo4gt/+ mice compared with wild type mice. Three out of these five are involved in the formation and maintenance of PNNs. Hapln2 is part of PNNs that surround the nodes of Ranvier (51) and its expression is enriched in neurons. The PNN constituent reelin is a substrate of ADAMTS2 in the brain. Sulf2 encodes an ECM enzyme that removes 6-O-Sulfate groups from heparan sulfate, which is a major component of PNNs and modulates heparan sulfate-dependent signaling (52). As we hypothesized, enzymatic dissolution of the ECM in the BLA led to a transient reduction in alcohol consumption suggesting that the ECM in the BLA contributes to LMO4 regulation of alcohol consumption. The transient nature of this reduction indicates that BLA neurons rapidly adapt to the loss of ECM. Our results also reveal that once dissolved, the ECM in the BLA regenerates rapidly, although the time course of regeneration does not fully explain the transient nature of the alcohol consumption phenotype. One possibility is that neurons rapidly adapt to the physiological changes that accompany ECM dissolution and such neuroadaptations could underlie the normalization of alcohol consumption following a transient decrease.

To our knowledge ours is the first study to use an unbiased whole genome approach to identify transcriptional targets of LMO4 in the brain. Elucidating transcriptional targets of LMO4 is particularly challenging since LMO4 does not directly bind DNA but instead relies on interactions with DNA-binding transcription factors. A comprehensive study of LMO4-interacting proteins in the brain is lacking. This being the case, it is difficult to know which of the differentially expressed genes detected by our RNAseq analysis are direct transcriptional targets of LMO4. Despite the behavioral phenotypes in Lmo4gt/+ and wild type mice with BLA knockdown of LMO4 being opposite, we were able to use Lmo4gt/+ mice to identify differentially expressed genes in the BLA that contribute to the phenotype of reduced alcohol consumption. One of these differentially expressed genes, Oprk1, encodes the KOR, a known regulator of alcohol consumption. Furthermore, systemic administration of KOR antagonists reduce drinking in animals (31, 32, 35). This finding validates our approach of using Lmo4gt/+ mice to identify brain-region specific mechanisms that guide drinking behavior in mice.

In summary, our results identify a novel role for the transcriptional regulator LMO4 in differentially regulating alcohol intake. Future studies will investigate different brain regions and neuronal subpopulations to determine whether there exist LMO4 transcriptional signatures that could be used to query databases such as the LINCS (Library of Integrated Cellular Signatures) 1000 database and identify small molecule therapeutics with the potential to reduce alcohol consumption.