Behavioral actions of alcohol: phenotypic relations from multivariate analysis of mutant mouse data

Authors

  • Y. A. Blednov,

    1. Waggoner Center for Alcohol and Addiction Research, Section on Neurobiology, Institute for Neuroscience, Institute for Cell and Molecular Biology, University of Texas at Austin, TX, USA
    Search for more papers by this author
  • R. D. Mayfield,

    1. Waggoner Center for Alcohol and Addiction Research, Section on Neurobiology, Institute for Neuroscience, Institute for Cell and Molecular Biology, University of Texas at Austin, TX, USA
    Search for more papers by this author
  • J. Belknap,

    1. Portland Alcohol Research Center, Department of Behavioral Neuroscience, Oregon Health & Science University, Portland, OR, USA
    Search for more papers by this author
  • R. A. Harris

    Corresponding author
    1. Waggoner Center for Alcohol and Addiction Research, Section on Neurobiology, Institute for Neuroscience, Institute for Cell and Molecular Biology, University of Texas at Austin, TX, USA
    Search for more papers by this author

R. A. Harris, Waggoner Center for Alcohol and Addiction Research, Section on Neurobiology, Institute for Neuroscience, Institute for Cell and Molecular Biology, University of Texas at Austin, TX 78712, USA. E-mail: harris@mail.utexas.edu

Abstract

Behavioral studies on genetically diverse mice have proven powerful for determining relationships between phenotypes and have been widely used in alcohol research. Most of these studies rely on naturally occurring genetic polymorphisms among inbred strains and selected lines. Another approach is to introduce variation by engineering single-gene mutations in mice. We have tested 37 different mutant mice and their wild-type controls for a variety (31) of behaviors and have mined this data set by K-means clustering and analysis of correlations. We found a correlation between a stress-related response (activity in a novel environment) and alcohol consumption and preference for saccharin. We confirmed several relationships detected in earlier genetic studies, including positive correlation of alcohol consumption with saccharin consumption and negative correlations with conditioned taste aversion and alcohol withdrawal severity. Introduction of single-gene mutations either eliminated or greatly diminished these correlations. The three tests of alcohol consumption used (continuous two-bottle choice and two limited access tests: drinking in the dark and sustained high alcohol consumption) share a relationship with saccharin consumption, but differ from each other in their correlation networks. We suggest that alcohol consumption is controlled by multiple physiological systems where single-gene mutations can disrupt the networks of such systems.

One goal of behavioral genetics is to determine which behaviors have either common or distinct genetic determinants. Perhaps the most successful approach for determining relationships between behavioral traits has been the use of recombinant inbred (RI) mice. The GeneNetwork/WebQTL site is particularly powerful and allows correlation of behaviors measured in many laboratories during the past decade (Chesler et al. 2004; Philip et al. 2010), and the Mouse Phenome Database also contains extensive physiological data from inbred and RI strains (Grubb et al. 2009). However, there are limitations to these database approaches, including a preponderance of RI mice derived solely from C57BL/6J and DBA/2J and the necessity of combining data collected at different sites at different times. A study on the reproducibility of behavioral phenotypes measured at different sites indicated considerable variability in some behavioral measures (Wahlsten et al. 2003), although alcohol consumption by mice is a remarkably stable and reproducible trait (Wahlsten et al. 2006). The data in this study were collected in the same facility over several years, as indicated in the publication dates. A range of mutant mice has been studied for alcohol responses by many laboratories, but most mutants were tested for only one or a few behaviors (Crabbe et al. 2006).

As a part of the Integrated Neuroscience Initiative on Alcoholism (INIA-West), we have measured behavioral effects of alcohol, as well as several other behaviors, in null mutant mice under identical laboratory conditions. Altogether, this has yielded data on 31 behaviors in 37 different mutant mouse lines as well as their wild-type (WT) control lines. Specifically, we studied 32 mutant lines with single-gene deletions, two lines with knock-in mutations and three transgenic overexpressing mutant lines (Tables 1 and S1). Because the WT mice mainly represent mixed genetic backgrounds that are somewhat different for each line, this also provides a useful set of data for evaluation of behavioral relationships among other WT strains. We used two main statistical approaches, correlation and K-means clustering, to evaluate the relation between the different behaviors in WT and mutant mice. It should be noted that this approach provides behavioral, not genetic, correlations. Most prominent in WT mice was the correlation between two-bottle choice (2-BC) ethanol consumption and saccharin consumption, response to novelty, conditioned taste aversion (CTA) and severity of acute alcohol withdrawal. We also examined the difference between WT and mutants for the traits and found that response to novelty, 2-BC ethanol consumption and CTA was altered by many of the mutations; however, limited access, one-bottle ethanol consumption tests [(drinking in the dark (DID) and scheduled high alcohol consumption (SHAC)] were more resistant to mutation.

Methods and materials

Most of the data used in the analyses were obtained from published studies listed in Table 1. Data from several mutants have not been published and information about these mutants is given in the following sections. Behavioral testing of all mutants was carried out as described in the individual published papers (Table 1). A brief summary of each behavioral measure is given in the following sections.

Table 1.  Summary of the behavioral effects of mutations in male mice
Gene symbolReferenceNoveltyAWPLORRCTADIDSHAC Two-bottle choice 
  1. Genes mutated are shown by gene symbol followed by genetic background if the mutant was tested on more than one background. More details are given in Tables S1–S3. Reference denotes the sources for data previously published. Sources for all mutants are given in Table S1. The upward arrows indicate that a behavior is increased by the mutation; downward arrows indicate a decrease in the mutants and 0 is not changed. The statistical analysis is based on t-tests for AWP, LORR, DID and SHAC and ANOVA for novelty, CTA and two-bottle choice. The 2-BC tests involve multiple concentrations and the ANOVA determined a genetic difference across all concentrations. The raw data are given in Table S2.

  2. ↓, decreased behavior in comparison with wild type; ↑, increased behavior in comparison with wild type; 0, no change in comparison with wild type; NA, not tested.

   4.0 g/kg3.8 g/kg (g/kg – four sessions)(g/kg – six sessions)Preference for saccharinPreference for quininePreference for EtOHEtOH (g/kg)
Gad2 (B6N5) Blednov et al. (2010) 000NA0000
Gad2 (129N1) Blednov et al. (2010) 0000000
Gad2 (129N2) Blednov et al. (2010) 000000
Pdyn Blednov et al. (2006) 0000000
Faah Blednov et al. (2007) 0000000
Tas1r3 Blednov et al. (2008) 00000
Il6 (B6xSJL-TG)Not published000000
Il6 (B6-TG)Not published00000000
Gabra2 (B6N1) Boehm et al. (2004c) 00000000
Gabra2 (B6N2)Not published000000
Nos1Not published000000
SncaNot published000000000
Gabra5 Boehm et al. (2004c) 0000000
Prkcb1Not published000NA00000
Trpm5 Blednov et al. (2008) 0000
Grm5 Blednov and Harris (2008) 00000000
Penk-rsNot published00NA000
Ccr2 Blednov et al. (2005) 00NA000
Ccr5 Blednov et al. (2005) 000NA00000
Htr3aNot published0000000
Fyn Boehm et al. (2003) 00NANA000
Fyn (TG) Boehm et al. (2004a,b)000NANA0000
Gabra1 Blednov et al. (2003a,b)0NA0000
Gabra1 (SHLA-KI) Werner et al. (2006), Borghese et al. (2006)000000
Gabrb2 Blednov et al. (2003a,b)NA00000
Glra1 (S267Q-KI) Findlay et al. (2003) 000NANA0000
Glra1 (S267Q-TG) Findlay et al. (2002, 2005)0NANANA000
Glra1 (oscillator)Not published0NANA00000
Trpv1 Blednov and Harris (2009) 0000000
Ccl2 Blednov et al. (2005) 00NA00000
Ccl3 Blednov et al. (2005) 00NA00
Drd3Not published0NA0NANANA00
Kcnj6 (B6x129) Blednov et al. (2001, 2002)0NA000
Kcnj6 (B6N6)Not published00NANA000
Grm4 Blednov et al. (2004) 00NANA00000
GrprNot published00NANA00
Gnat3 Blednov et al. (2008) 0000

Two-bottle choice

Mice were given continuous access to water or a solution containing ethanol in ascending concentrations of 3%, 6%, 9% or 12% v/v. Each concentration was presented for 4 days and consumption was calculated as both amount of alcohol (g/kg/day) (2BAE) and as preference (2BPRE). For the correlation analyses, the data are designated by the ethanol concentration; for example, 2BAE12 indicates the amount of ethanol consumed from the 12% solution in the 2-BC test. The amount of water consumed in conjunction with ethanol consumption was measured (e.g. 2BH2OE12) as was the total fluid intake (e.g. 2BTI12). For clustering and correlation analyses, we focused on the 12% ethanol solution. Consumption of 0.03 and 0.06 mm quinine (2BPRQN1 and 2BPRQN2, respectively) and 0.033% and 0.066% saccharin (2BPRSAC1 and 2BPRSAC2, respectively) were also measured in the 2-BC test and presented as preference values.

Limited access alcohol consumption (DID and SHAC)

Two tests were used for consumption of ethanol from a single bottle (no choice). DID allows consumption of a 15% ethanol solution for 2 h at the beginning of the dark phase each day for 3 days and 4 h on day 4. For correlation and clustering purposes, the sum of ethanol consumed during all four sessions was calculated. SHAC involves 30 min of alcohol consumption (5% of ethanol solution) daily followed by periods of water consumption that vary from 3.5 to 7.5 h. The total consumption of ethanol for all six drinking sessions (six experimental days) was used for correlational analyses. Alcohol consumption was measured as g/kg/day and was denoted as DID-E and SHAC-E. Consumption of water by the SHAC mice was denoted as SHAC-H2O. DID was tested with 19 mutants (and their WTs), whereas all other behaviors were tested with 37 mutants.

Conditioned taste aversion

Mice were restricted to 2 h of water per day and given 1 h access to a saccharin solution every other day for 10 days. Immediately after access to saccharin, mice received injections of saline or ethanol [(2.5 g/kg intraperitoneal (i.p.)]. Reduced consumption of the saccharin solution is used as a measure of CTA. For correlation purposes, CTA was calculated as a difference between area under the curve for the saline-treated group and area under the curve for the ethanol-treated group for each genotype separately.

Duration of loss of righting reflex

Mice were injected with an ethanol dose of 3.4 or 3.8 g/kg i.p. and duration of loss of righting reflex (LORR) was defined as the length of time from when the mouse was placed in a supine position until it was able to right itself.

Handling-induced convulsions after acute ethanol (AWN and AWP)

Mice were scored for severity of handling-induced convulsions (HICs) before (basal value) and after injection of 4 g/kg ethanol i.p. Scores were obtained every hour until HIC returned to basal levels. This resulted in a biphasic curve, with initial values being negative or less than basal (anticonvulsant effect) and later values being positive or above basal (acute withdrawal). The areas of these phases were determined and used to calculate values for acute withdrawal negative (AWN) and acute withdrawal positive (AWP).

Motor response to a new environment (novelty)

Mice were placed in an unfamiliar experimental cage and motor activity was measured every 10 min for 3 h. Total response was calculated as area under the curve and was determined separately for each genotype.

Mutant mice

As noted in Table 1, information for many of the mutant mice has been published and has not been repeated here. Unless noted, all mice were produced by heterozygous breeding and WT littermates were used as controls. All procedures were approved by the Institutional Animal Care and Use Committee and adhered to NIH Guidelines. The University of Texas facility is AAALAC accredited. Details for the mutant mice, which are not described in our publications are summarized in Table S1 and are provided in the following sections.

Grpr

Mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and homozygous (females) or hemizygous (males) for the X-linked Grprtm1Jfb-targeted mutation were generated and described by Hampton et al. (1998).

Kcnj6 (B6N6)

GIRK2 null mutant mice were generated by Signorini et al. (1997). Mice from the original mixed background 129v × C57Bl/6 were backcrossed to C57Bl/6J inbred mice for six generations.

Drd3

Mice were purchased from Jackson Laboratories. Generation of mice with the Drd3tm1Dac mutation was described by Accili et al. (1996).

Glra1 (oscillator)

Mice with this spontaneous mutation (spasmodic oscillator Glra1nmf11) were purchased from Jackson Laboratories and were on a C57BL/6J genetic background.

Snca

Null α-synuclein mutant mice were purchased from Jackson Laboratories and were originally described by Abeliovich et al. (2000). Breeding pairs of Snca null mice were maintained on the original mixed C57BL/6 × 129/SvJ genetic background by heterozygous breeding.

Nos1

Mice were purchased from Jackson Laboratories and homozygous mutants for Nos1tm1Plh were originally generated and described by Huang et al. (1993).

Gabra2 (B6N2)

GABA α2-subunit knock-out mice were provided by Paul Whiting, Elisabeth Garrett and Thomas Rosahl at Merck Sharp and Dohme (Harlow, UK). These mice were described by Boehm et al. (2004c, 2006). For this study, they were backcrossed one more time with C57Bl/6J mice.

Il6 (B6-TG) and Il6 (B6xSJL-TG)

Mice with astrocyte-specific transgenic overexpression of Il6 were generated on mixed B6xSJL genetic background and described by Campbell et al. (1993). Another colony was generated by backcrossing mutant mice for at least six generations with C57Bl/6J.

Penk-rs

Mice were purchased from Jackson Laboratories and the deletion on the preproenkephalin gene (Penk−/−; B6.129-Penk-rstm1Pig) was described by Konig et al. (1996). These mice were originally generated using a 129-derived R1 embryonic cell line and then backcrossed onto a C57BL/6J (B6) background for 10 generations.

Htr3a

Mice were purchased from Jackson Laboratories and the Htr3atm1Jul/J-targeted mutation was described by Zeitz et al. (2002).

Prkcb

These mice were generated by Leitges et al. (1996).

Mathematical analyses

K-means clustering was used to visualize the relationships among behaviors and mutant mice. The behavioral data were first summarized in two ways: (1) a behavioral difference score was calculated by subtracting the WT animal scores from the mutant animal scores (difference score = mutant score − WT score) and (2) difference scores were then summed across multiple testing sessions, resulting in a cumulative difference score for each behavior. The cumulative difference score simplifies the data by capturing and summarizing behavioral data in a session-wide manner. The resulting cumulative difference scores were then standardized by z-score transformation, and missing values were imputed for behavioral tests performed in at least 80% of the different mutant mice. z-Scores were calculated independently for each behavioral measure. K-means clustering was then performed using k = 5. The number of clusters was based on a principal component analysis, indicating that the first five eigenvalues explain approximately 75% of the variance in the data set (Fig. S1).

Correlation matrices were generated (using the same z-scored data described for the K-means clustering) to evaluate the relationships between alcohol drinking and behavior using JMP®, Version 7, SAS Institute Inc. (Cary, NC, USA, 1989–2007). The raw data used for the correlation analyses are presented in Tables S2 (males) and S3 (females). The pairwise correlations (r2) are presented individually for male and female WT and mutant mice in Tables S4–S11. Correlations were considered significant if P < 0.05 (not corrected for multiple comparisons). This significance criterion is appropriate when the goal is not hypothesis testing but rather generating new hypotheses suitable for future hypothesis testing experiments. Most of the present work is exploratory in nature; thus, a relatively relaxed significance criterion is acceptable to point the way to future research efforts. However, this approach does not correct for the many correlations calculated and reported in the following sections. To deal with this shortcoming, we also calculated the false discovery rate (FDR) for each pairwise correlation, a commonly used correction for multiple testing. These were calculated according to the method of Benjamini and Hochberg (1995) and are presented in the Supporting information next to the corresponding P value. For this approach, the significance threshold was set at FDR = 0.05, which is the proportion of false positives expected according to the null hypothesis (r = 0) among all correlations judged to be statistically significant. Thus, all significant correlations meeting or exceeding this criterion have a 5% chance of being false positives.

Results

The differences between WT and mutant mice for 31 behaviors tested in 37 lines are summarized in Table 1 (males) and complete data for males and females are given in Tables S2 and S3. Both male and female mice were used for all tests; the data from males are shown in the main text and data from females (and males) are in the Supporting information. Table 1 demonstrates that some behaviors were much more likely to be changed by mutations than others and this is summarized as follows (number of mutants with changes/number of mutants tested, percent with changes): novelty (27/37, 73%), AWP (5/36, 14%), LORR (8/37, 22%), CTA (15/28, 54%), DID (1/18, 6%), SHAC (3/35, 9%), saccharin 2-BC (9/37, 24%), quinine 2-BC (8/37, 22%), ethanol 2-BC preference (19/37, 51%) and ethanol 2-BC intake (19/37, 51%). Thus, response to novelty, CTA and 2-BC ethanol drinking was the most sensitive to these mutations. It should be noted that the genetic background of the mice is predominantly C57Bl/6J and substrains of 129 mice, giving some limitation to the range of genetic diversity available.

To simultaneously evaluate and visualize the relationship between the behaviors and the mutations, we used K-means clustering to generate a heat map of these relationships (Fig. 1). The number of behavioral clusters (five) was based on principal component analysis (Fig. S1). For the 2-BC data, we focused on consumption of the 12% solution. There are five clusters of behaviors: the first captures ethanol and saccharin consumption; the second contains water and total fluid consumption; the third contains quinine preference; the fourth combines LORR, alcohol withdrawal severity, CTA and response to novelty and the fifth has ethanol and water consumption in the SHAC test together with an anticonvulsant measure. The first cluster (alcohol and saccharin consumption) is consistent with expectations and published literature (Bachmanov et al. 1996; Belknap et al. 2008; Yoneyama et al. 2008). The other behavioral clusters show that consumption of quinine, water and ethanol in the SHAC test are distinct behaviors not related to each other or to 2-BC ethanol consumption. The three behaviors produced by injection of ethanol (LORR2, AWP and CTA) cluster together and, surprisingly, this cluster also includes motor response to novelty (which is measured without injection of ethanol). The clustering of genetic mutations (y-axis) produced five groups, the largest of which has 13 mutants and is driven mainly by a consistent increase in 2-BC alcohol and saccharin drinking, a consistent decrease in water consumption and a somewhat less consistent decrease in SHAC alcohol consumption. This cluster contains genes involved in GABA (Gad2 and Gabra1), glycine (Glra1) and neuroimmune signaling (Ccr2, Il6 and B6-TG). The three taste mutants (Gnat3, Tas1r3 and Trpm5) are clustered with Kcnj6 (B6N6) and this cluster is driven by decreased consumption of ethanol (2-BC and SHAC) and saccharin and increased consumption of water. It is interesting to note that genetic background can be as important as the specific gene. For example, the Kcnj6 and Il6 (B6-TG) mutations were each tested on different genetic backgrounds and the change in background placed them in different clusters. In contrast, Gabra2 mutants on two different backgrounds (B6N1 and B6N2) were assigned to the same cluster and two Gad2 mutants (129N1 and 129N2) are in the same cluster, whereas Gad2 (B6N5) is in a different group. Results in Fig. 1 are for males and results from females are given in Fig. S2 and show many of the relationships noted for males. For example, 2-BC ethanol and saccharin consumption cluster together and LORR2, AWP and CTA form another cluster, while quinine consumption is a separate cluster for females. The gene clustering for females is similar (but not identical) to males with a 19-gene cluster containing GABA, glycine and neuroimmune genes driven by increased alcohol and saccharin consumption and the taste genes clustered together and driven by alcohol, saccharin and quinine consumption (Fig. S2). Obvious biological implications of the gene clusters are limited. For both males and females, three taste-related genes (Gastducin, Tlr3 and Trpm5) cluster together, but other genes that were expected to be related are in different clusters. For example, GABA-related genes are found in all five clusters for females and in four different clusters for males (Fig. 1).

Figure 1.

K-means clustering of behavioral changes resulting from mutations. Red represents an increase in mutant relative to wild type, green represents a decrease and black signifies no difference between mutant and wild type. Behaviors and genotypes that are significantly related are shown as blocks or clusters. Data are from male mice; results from females are shown in Fig. S1. Details on mutations and behaviors are given in Tables 1 and S1.

We examined the relationship among behaviors for WT mice and, separately, for mutant mice. As noted previously, there is genetic diversity among the WT strains due to background differences, as well as the genetic changes produced by mutation in the other strains. We focused on alcohol consumption and generated correlation matrices to evaluate relationships between alcohol drinking and other behaviors for WT and, separately, for male mutant mice (Figs. 3–5). All correlation values represent r2. Complete results of correlational analyses for males and females (WT and mutant) are presented in Tables S4–S11. Examples of the scatterplots underlying the correlations are given in Fig. 2 for 2-BC alcohol consumption (2BAE12), saccharin consumption (SAC1), CTA and AWP for WT mice (upper panels) and mutant mice (lower panels). In the WT mice, consumption of ethanol in the 2-BC choice test showed the strongest correlation with saccharin drinking and significant correlations with measures of alcohol withdrawal severity, response to novelty and alcohol CTA (Fig. 3a). Several other novel correlations were found: (1) the positive correlation between stress response (novelty) and preference for saccharin, (2) the negative correlations between stress response and CTA as well as severity of ethanol-induced withdrawal and (3) the positive correlation between CTA and ethanol-induced LORR. As discussed below, the correlation between 2-BC ethanol consumption and consumption of sweet solutions, CTA and alcohol withdrawal severity found in previous studies (Bachmanov et al. 1996; Green & Grahame 2008; Metten et al. 1998) were confirmed in our WT mice. Figures 2 and 3 present correlation data using a 12% ethanol solution (very similar correlations were obtained for 6% and 9% ethanol, Tables S4–S7).

Figure 2.

Scatterplots showing correlation of alcohol consumption (2-BC test with 12%, 2BAE12) with saccharin consumption (SAC1), conditioned taste aversion (CTA) or acute alcohol withdrawal severity (AWP) for male mice. Upper panels (a–c) are for the different wild-type mice and lower panels (d–f) are for different mutant mice. Values are given as z-scores representing the mean from each strain. Statistically significant (P < 0.05) correlations include the linear regression (solid line) and the 95% confidence intervals (dotted line). The coefficient of determination (r2) corresponding to those in Figs. 3 and 4 and are (a) 0.64, (b) 0.19, (c) 0.32 and (d) 0.27. The mutants that represent apparent outliers in (d) and (f) include Tas1r3 and Glra1 (S267Q-KI) (d) and Gabrb2 (f).

Figure 3.

Relationships among two-bottle choice drinking and other behaviors for wild-type (upper panel) and mutant (lower panel) mice. Significant correlations (P < 0.05) are noted by the arrows and connecting lines with the coefficient of determination (r2) given on each line connecting the traits; dotted lines represent positive correlations and solid lines are negative correlations. Data are from male mice; results from females are in Fig. S3. Abbreviations used in Figs. 3–5 and S3–S5 are 2BAE, two-bottle choice, amount of alcohol consumed denoted by the concentration of alcohol (2BAE12 is amount of 12% solution consumed); AWN, alcohol withdrawal negative, acute anticonvulsant measure; AWP, alcohol withdrawal positive, acute alcohol withdrawal severity; CTA, conditioned taste aversion; DID, drinking in the dark or limited access drinking; LORR, duration of loss of righting reflex produced by 3.8 g/kg ethanol; NOVELTY, motor response to a novel situation; SAC1/SAC2, preference for 0.033/0.066% saccharin; SHAC, sustained high alcohol consumption; QN1/2, preference for 0.03/ 0.06 mm quinine. Raw data for all behaviors are given in Tables S2 and S3 and all pairwise correlations are in Tables S4–S11.

In the mutant mice, the relationships are remarkably different and few of these correlations are maintained (Figs. 2d, e, f and 3b). Ethanol consumption remains linked with saccharin, but the coefficient of determination (r2) dropped from 0.53 for the WT to 0.19 for the mutants. Similar results were found in WT and mutant female mice (Fig. S3a and b). Compared with male mice, WT females showed a stronger positive link with stress response but no association with CTA. Ethanol consumption also remained linked with saccharin and stress response in female mice, but r2 values dropped from 0.57 to 0.20 for the WT and from 0.57 to 0.33 for the mutants.

The DID test differs from the continuous 2-BC drinking in several aspects, including the use of only one bottle containing the alcohol solution (no water choice), a limited access period and a higher concentration of ethanol. The DID in WT male mice was positively correlated with 2-BC drinking and also with saccharin consumption and response to novelty, as noted for the 2-BC in Fig. 2, and negatively correlated with measures of anticonvulsant activity of ethanol (AWN) (Fig. 4a). Most of these correlations were disrupted by mutation (Fig. 4b). Like DID, the SHAC test uses a single bottle of ethanol and limited access, and also has a period of water deprivation, which increases alcohol consumption and uses a lower concentration of alcohol (5% instead of 15% in DID test). In the WT mice, consumption in this test was not correlated with 2-BC drinking but was related to the amount of water consumed in the SHAC test and CTA and weakly to saccharin consumption and LORR (Fig. 5a). Mutations removed the correlation with saccharin and weakened the correlations with CTA and SHAC water (Fig. 5b). Very similar patterns of correlations were shown for the WT and mutant female mice (Figs. S4 and S5).

Figure 4.

Relationships among limited access drinking (DID) and other behaviors for wild-type (upper panel) and mutant (lower panel) mice. Significant correlations are noted by the arrows and connecting lines with the coefficient of determination (r2) given on each line; dotted lines represent positive correlations and solid lines are negative correlations. Details are given in legend to Fig. 3. Data are from male mice; results from females are in Fig. S4.

Figure 5.

Relationships among limited access (SHAC) drinking and other behaviors for wild-type (upper panel) and mutant (lower panel) mice. Significant correlations are noted by the arrows and connecting lines with the coefficient of determination (r2) given on each line; dotted lines represent positive correlations and solid lines are negative correlations. Details are given in legend to Fig. 3. Data are from male mice; results from females are in Fig. S5.

Discussion

Continuous 2-BC alcohol consumption has been examined for its relationship with other behaviors in inbred strains and selected lines (see Green & Grahame 2008). One key aspect of the present analysis is that the use of WT mice with a range of mixed backgrounds provides behavioral correlations that are very similar to those obtained with other genetic approaches. For example, a relationship between CTA and alcohol consumption was measured for selected lines of rats and mice (Green & Grahame 2008), and alcohol withdrawal severity was negatively correlated with alcohol consumption in heterogeneous stock, inbred and RI mouse strains (Belknap et al. 2008; Hitzemann et al. 2009; Metten et al. 1998). Our data show a robust positive correlation between saccharin and alcohol consumption similar to that in studies of inbred strains, hybrid mice and taste-deficient mutants (Bachmanov et al. 1996; Blednov et al. 2008; Yoneyama et al. 2008). We also observed a positive correlation between response to novelty and 2-BC alcohol consumption. To our knowledge, the relationship between response to novelty and alcohol consumption in rodents has not been noted in any publications, but a GeneNetwork analysis of the ‘BXD Published Phenotypes' shows correlations between 2-BC alcohol consumption and a number of measures of response to novelty in the open field. (Data in Record ID 10153 for alcohol preference in males is correlated with measures of activity in a novel open field given in Record IDs 11349, 11414 and 11863.) Although this was a different test of novelty than was used in our studies, it indicates the generality of the association between alcohol consumption and the novelty response. In addition, the clinical literature suggests a relationship between novelty seeking and alcohol craving and relapse in alcohol-dependent patients (Evren et al. 2012).

In view of the consistent evidence from multiple genetic models that CTA, withdrawal severity and consumption of sweet solutions strongly influence alcohol consumption, it is remarkable that our mutant mice retained only a correlation between saccharine and alcohol consumption and even this correlation was much weaker than that in the WT mice. We conclude that the relationships between alcohol consumption and other phenotypes are maintained by multiple brain signaling systems, which are quite sensitive to disruption by single-gene changes. This is consistent with genomic studies, which suggest that genetic variation in alcohol consumption in mice is related to altered expression of many different transcripts (Mulligan et al. 2006; Tabakoff et al. 2008). One consideration is the choice of mutations because if we used only mutants that disrupted alcohol consumption, then we would see reduced correlations with the other behaviors. However, only about half of the mutants showed changes in alcohol consumption, and many (73%) showed changes in a nonalcohol behavior, response to novelty. The mutants were selected on the basis of genetic, biochemical or physiological data suggesting that they might be of interest in alcohol research, but the range of functions of these genes (taste receptors, neurotransmitter receptors, ion channels, chemokines and enzymes) is quite diverse.

In contrast to continuous 2-BC, the DID and SHAC models of alcohol consumption have not undergone detailed genetic analysis. One study of 12 inbred strains showed a correlation between DID and 2-BC consumption as well as with alcohol withdrawal severity and CTA (Rhodes et al. 2007). In our WT mice, we also found that consumption of ethanol in DID and 2-BC tests were correlated. In contrast with 2-BC consumption, ethanol intake in these limited access tests showed negative correlations with either LORR and CTA (SHAC test) or AWN (DID test). Both LORR and AWN result after administration of ethanol in high doses (3.8 g/kg for LORR and 4.0 g/kg for AWN), which leads to high concentrations of ethanol in blood. It should be noted that both tests with limited access to alcohol were developed for producing relatively high levels of blood ethanol concentrations (Finn et al. 2005; Rhodes et al. 2007). Our data indicate that the determinants of DID and SHAC drinking are somewhat different from 2-BC in that consumption in DID and SHAC is not related to withdrawal severity (AWP). SHAC, similar to 2-BC but not DID, was related to CTA. A relationship with saccharin consumption is a common denominator for all three tests, but 2-BC ethanol consumption may be more affected by single-gene mutations than DID or SHAC.

The robustness of the relationships between alcohol consumption with CTA and withdrawal severity raises the question of how the linkage between these traits might have arisen. An evolutionary argument is often proposed in which attraction to fermenting fruit evolved as a strategy to obtain food (Dudley 2002) and a corollary is that consumption of alcohol would be promoted by reduced aversion and reduced physical dependence (hangover) on alcohol. There is evidence that Drosophila are attracted to overripe fruit, in part, due to ethanol (Zhu et al. 2003) and that ethanol may be rewarding in this species (Kaun et al. 2011) and wild treeshrews consume fermented floral nectar containing alcohol (Wiens et al. 2008). In contrast, some birds and bats reduce intake of food when it contains alcohol (Mazeh et al. 2008; Sanchez et al. 2008). The concentration of alcohol found in fermenting fruit is very low and it is unlikely that mammals could achieve blood ethanol concentrations from such food sources that correspond to the ethanol doses used to produce behaviors such as CTA and acute alcohol withdrawal severity. However, it is interesting to note that fruit bats show a seasonal variation in alcohol preference, which is dependent on ambient temperature, possibly to modulate the hypothermic effects of alcohol (Korine et al. 2011).

An alternative possibility is that rodent ethanol consumption is an epiphenomenon that is due to an evolved preference for sweet tastes. Indeed, deletion of proteins critical for sweet perception markedly reduces alcohol consumption (Blednov et al. 2008). But this only complicates understanding of the relationship among traits, because we must now ask why ethanol CTA and withdrawal severity are related to an evolutionary development of preference for sugars. It is possible that ethanol phenotypes, such as CTA and acute withdrawal, are more general indicators of brain reward, activation and plasticity and it is these meta-processes that are related to frugivory. As discussed above, our observation that the relationships among these phenotypes is disrupted by multiple single-gene deletions may suggest a complex, and perhaps indirect, relationship between these traits.

In summary, we provide a novel meta-analysis of multiple behaviors in a range of WT and mutant mice and confirm and extend the relationships between alcohol phenotypes. A unique aspect of this meta-analysis is that all 31 behaviors were tested in the same laboratory. Specifically, 2-BC ethanol consumption is correlated with a number of other traits (response to novelty, 2-BC saccharin, CTA and alcohol withdrawal severity) and is frequently altered by null mutation or other genetic manipulations. In contrast, limited access DID or SHAC tests have fewer and weaker correlations with other behaviors and appear more resistant to mutation. These differences may be important for medication development as they suggest that these three tests of alcohol consumption may have different genetic foundations and different pharmacological responses.

Acknowledgments

Supported in part by grants from the NIH/NIAAA Integrated Neuroscience Initiative on Alcoholism (INIA-West; AA13520) and NIH Grant AA06399. We thank Drs. John Crabbe, George Koob and Jody Mayfield for their helpful comments on the manuscript.

Ancillary