Considering first the g/kg consumption of ethanol during 24 hr two-bottle choice, these results suggest that the genes influencing two-bottle preference for ethanol are generally somewhat distinct from those that lead HDID-1 mice to drink large amounts during a limited access, single bottle test with 20% ethanol in the circadian dark. This appears to be true at EtOH concentrations between 3% and 25% v/v. At concentrations of 30 and 40%, HS mice drank significantly greater amounts of ethanol than did HDID-1.
When we consider the preference ratio data in this experiment, the picture becomes a little more complicated. Neither genotype showed particularly high preference for ethanol vs water in Experiment 1; no average preference ratio exceeded 37% (see ). In comparison, after similar two-bottle tests of escalating concentrations of ethanol, several inbred strains of the C57/C58 lineage typically show preference ratios > 90% for ethanol concentrations between 3 and 12–15% (Yoneyama et al., 2008
; Wahlsten et al., 2006
). In Experiment 1, the differences in preference between HDID-1 and HS mice were small, were only seen at two concentrations, and were not in the same direction. Ethanol preference ratios do not always follow consumption data exactly, and are well known to show an inverted-U relationship with increasing concentration (McClearn, 1968a
). This is thought to reflect the joint influences of motivation to experience ethanol's post-ingestive effects [e.g., (Belknap et al., 1977
)] and the different taste sensations elicited by different concentrations [e.g. (Belknap et al., 1993
; Bachmanov et al., 2003
We earlier reported a fairly strong genetic correlation between g/kg ethanol intake on the 4th
day of the standard DID test (Rhodes et al, 2007
) and g/kg intake on the final two days of a brief two-bottle preference study (Belknap et al., 1993
). Yet, in the current experiments, we found only modest evidence supporting such a correlation (i.e., higher preference in HDID than HS only at 9%). There are several potential reasons for this. Data from Belknap et al (1993)
were taken from the final two days of a protocol where male mice first drank 3% ethanol or water for 3 days, then 6% ethanol or water for 3 days, and finally, 10% ethanol or water for 4 days. There were 7 inbred strains in common between the two studies, including C57BL/6J, and the strain mean correlation was r = 0.70 (P
= .08). Removal of the C57BL/6J strain from this correlation lowered it to r = 0.30 (NS). Thus, the correlation among inbred strain means could have been a false positive due to low N. However, the Belknap et al data are quite representative of other inbred strain ethanol preference data, and correlate with other extensive data sets with r values exceeding 0.90 (Wahlsten et al, 2006
). We have since tested 17 additional inbred strains for DID, and there are now 14 strains in common between the Belknap et al (1993)
data set and our DID data (Crabbe et al., unpublished data and Rhodes et al., 2007
). The genetic correlation across 14 strains remained r = 0.70, P
< 0.01); removal of the C57BL/6J strain from the analysis lowered the correlation coefficient slightly, but not significantly (r = 0.59, P
< 0.05). Thus, the preponderance of evidence from inbred strains suggests genetic overlap between DID for 20% and two-bottle preference for 10% ethanol.
An additional consideration is the genetic animal model employed to estimate the strength of the genetic association between DID and preference. All inbred mouse strains lack heterozygosity at all genes. Hence, dominance genetic variation (which can reflect the interactive effects between an individual’s pair of alleles at a single gene) is absent in such populations. Dominance has been shown to influence continuous access two-bottle preference drinking (Blednov et al., 2005
) and DID limited access drinking (Phillips et al, 2010
). In the HDID selected lines, dominance may have played a role in either type of drinking, but has not been assessed.
Our test for genetic correlation using a selected line was able to ask the question only unidirectionally: selection for genes affecting DID appeared to have only modest effects on preference drinking. The parallel experiment would be a test of mice selectively bred for high vs low preference (e.g., HAP vs LAP mice, see Grahame et al, 1999a
, Oberlin et al., 2010
) for DID. HAP mice drink significantly more ethanol than LAP mice during sessions where access to alcohol or water is limited to 2-hr periods, a situation resembling the 2-bottle version of the DID test (Grahame et al, 1999b
). In late 2003, we tested male HAP-1 and LAP-1 mice (generation 46, obtained from Dr. Nick Grahame) using a modified 4 day, 20%, 1-bottle DID test (daily 4 hr access, with hourly readings). Consistent with the hypothesis that high preference-drinking genotypes would consume more ethanol in the DID test, HAP-1 mice drank more ethanol than LAP-1 mice and had higher BECs (7.8 ± 1.1 vs. 2.5 ± 0.4 g/kg and 0.96 ± 0.15 vs. 0.15 ± 0.05 mg/ml, respectively). This relatively high intake of HAP mice in the DID procedure has been confirmed by others (N. Grahame, S. Boehm II, unpublished studies using only HAP genotypes). Therefore, long-term selective breeding for high vs. low preference in a 2-bottle choice continuous access procedure has produced lines of mice that differ in DID. There are other more technical reasons why estimates of genetic correlation may be found to be significant when selecting for Trait A and examining Trait B, but not vice versa, and why experiments with selected lines may not give the same result as those with inbred strains (for discussion, see Crabbe et al, 1990
; Crabbe, 1999
; Henderson, 1989
). Overall, there is reasonable evidence for some genetic overlap between the two traits.
In extended tests with the range of low to moderate ethanol concentrations (e.g., 3% – 15%), g/kg intake tends to be maintained approximately stably at a genotype-specific value, and the animals appear to adjust their relative preference to maintain an approximate g/kg dose level of ethanol, although this level does vary modestly with concentration between 12 and 24% ethanol (McClearn, 1968a
). In Experiment 1, mice were tested only using a series of ascending concentrations, so carryover effects could play a role. These results in fact tended to show a different pattern -- small and gradual increases in g/kg intake but relatively stable preference ratios as concentrations increased. As the highest concentration (≥ 30%) were introduced, HDID-1 mice showed the expected reductions in preference, while the HS/Npt mice did not. In fact, the HS mice showed increasing preference ratios as well as intakes over time. For an extended discussion of the role of concentration in absolute and relative alcohol intake, see (McClearn, 1968a
The basis for the very high consumption of EtOH solutions ≥30% in HS/Npt control mice is unknown. This intake is in the range of C57BL/6J mice, a genotype known for high EtOH preference. A possibility suggested by an anonymous reviewer is that the earlier experience drinking lower ethanol concentrations affected the ability of the HS mice to discriminate among the offered flavors toward the end of the experiment. That is, they may have developed the taste indifference revealed in Experiment 2 (see discussion below) earlier than the HIDI-1 mice. On the other hand, they did show significant avoidance of these high ethanol concentrations relative to water. Preference ratios were 37, 31, and 22% for the 30, 35 and 40% ethanol concentrations, respectively. The HS mice were able to achieve very high intakes of the highest ethanol concentrations with low preference scores in Experiment 1 because they also drank significantly more water than did the HDID-1 mice (). When naïve mice of these genotypes were offered 15% ethanol vs water for two hrs/day, consumption, but not preference, was equivalent, at least across the first several weeks (). Gradually, however, both genotypes increased their ethanol consumption and this increase from Block 7 over Block 1 was greater in HDID-1 mice than in HS mice (91% vs. 45%, respectively). Preference was clearly greater in HDID-1 than HS mice throughout Experiment 4 (74 – 89% vs. 51 – 71%, respectively), and in contrast to the results of Experiment 1, both genotypes showed significant preference for 15% ethanol during the 2 hr period (all blocks for HDID-1 mice but only 4/7 blocks for HS mice; see ).
HS mice drank significantly more water and total fluid than did HDID-1 mice during some experiments. In Experiment 1 with 24 hr access (), the line difference in water intake was about 1 ml on average (range 0.1 – 1.7 ml across the different ethanol concentrations). This accounted for the average difference in total fluid intake in this experiment. In Experiment 4 with 2 hr access periods (), the line difference in water intake was about 0.2 ml on average (range 0.12 – .26 ml) and that for total fluid was slightly less (0.15 ml). Interestingly, the genotypes did not differ in either water or total fluid intake during Experiment 3 (), and so this finding may be specific to two-bottle choice experiments for water vs ethanol but not water vs other tastants. We do not know the basis for this difference, but it appears to be a consistent correlated response, suggesting a genetic basis shared with high ethanol DID. The genetic relationship is inverse, with the HDID genotype drinking less water than the control HS stock. The difference does not appear to be because the HDID-1 line is drinking less water or fluid than do “normal” mice, as a survey of inbred strain data from the Mouse Phenome Database showed that mice of 7 of the 8 strains that make up the HS/Npt drink water in about the same range (1.82 – 7.68 ml/30 g body weight; Seburn, 2010
, MPD:Seburn1:9238). Furthermore, data from our recent inbred strain survey using the DID test also shows that the range of 2 hr water intake in the HDID-1 line is consistent with that of the progenitor strains (0 – 1.45 ml/30 g body weight in 2 hr; Crabbe et al., unpublished data). The difference is not due to obvious differences in body size, as the genotypes do not differ in body weight and the intakes we report were indexed as ml/ 30 g. Some difference in body composition could be involved, or a difference in the numerous hypothalamic and extrahypothalamic peptides and hormones involved in water balance and fluid intake regulation. For example, in one report, deletion of the arginine vasopressin 1 receptor gene produced mice with high ethanol preference (Sanbe et al, 2008
), although this was not seen by a different group that deleted the same gene (Caldwell et al., 2006
Experiments 2 and 3 were undertaken to test the hypothesis that differences in taste preference were important for the differences in ethanol intake between genotypes. However, the original tastant data (Experiment 2) revealed a major problem, as they offered no evidence that either genotype either preferred or avoided any tastant at any concentration. Experiment 3, however, provided clear evidence that naive HDID-1 and HS/Npt mice did not differ in their preference for sweet solutions, nor in their avoidance of the bitter solutions of quinine. Because the outcomes for both a caloric (sucrose) and non-caloric (saccharin) solution were similar, the differences in ethanol intake between genotypes at high concentrations are not likely to be driven by calorie seeking.
Why did Experiment 2 fail completely? We considered two possible explanations. First, the concentrations chosen could have been inappropriate. If the tastants were imperceptible to both genotypes, they would have had no basis for discriminating between the two bottles and 50% preference ratios would be expected. A second possibility was that the exposure to very high ethanol concentrations may have affected in some way the subsequent tastant tests. For example, the ingested ethanol might have denatured proteins in the mouth or damaged taste bud cells. Some other carryover effects from the high ethanol concentration drinking may have affected results. For example, the taste of high concentration ethanol may have conditioned an aversion to sucrose and saccharin.
We deemed the first hypothesis unlikely given the use of these concentrations in the literature with multiple genotypes of mice (Blednov et al., 2005
; Crabbe et al., 2006
; Lush, 1981
; Lush, 1984
; Lush, 1989
). We also consulted with Alex Bachmanov of the Monell Chemical Senses Institute, an expert on mouse taste, and he provided us with unpublished tastant preference drinking data that suggested we should have seen strong preference for sucrose and saccharin. Data for quinine were more sparse. We therefore pilot-tested some other concentrations for the three tastants, leading to our choices for Experiment 3.
The second hypothesis, suggested to us by our colleagues John Belknap and Aaron Janowsky, was conceded as a possibility by Alex Bachmanov, who pointed out that taste bud cells turn over in about 1–2 weeks. We have no physiological data to bring to bear on this hypothesis. Dr. Bachmanov also pointed out that exposure to drinking the highly-preferred tastant sucrose in particular could taint further taste experiments with the same mice. Based on this advice, we tested mice in Experiment 3 in the order deemed by Dr. Bachmanov least likely to lead to carryover effects, and we added a water-only washout between tastant tests. Carryover effects, however, could not easily explain the failure of the mice to avoid quinine solutions in Experiment 2. They also could not explain the fact that we saw no evidence of an effect of order of testing during Experiment 2.
There are substantial data in both mice and rats suggesting that there is some degree of coordinate genetic influence on preference for ethanol and sweet taste (Lemon et al., 2004
; Kampov-Polevoy et al., 1999
; Belknap et al., 2008
). Our results suggest that two-bottle 24-hour access ethanol preference and preference for sweet solutions are not genetically correlated in the HDID-1 and HS genotypes. Genotypic differences in ethanol preference were clear throughout many weeks of limited access testing (Experiment 4), but the lack of a genotypic difference in preference for either sucrose or saccharin was clear at the outset of testing (Experiment 3). It is possible, however that long-duration tests for sucrose or saccharin preference could reveal the emergence of greater sweet preference in HDID-1.
Many studies have tested the effects of genetic engineering of either a knockout or an over expressing transgenic for a target gene on ethanol preference. When these studies were reviewed in 2006, 79 genes had been targeted and tested for ethanol preference, and most had also assessed the possibility that sucrose or saccharin or quinine preference had been affected (Crabbe et al., 2006
). Many of those studies used the same mice tested for ethanol preference to assess preference for other tastants, much as we did for Experiments 1 and 2. We conclude that attempts to control EtOH preference studies for taste sensitivity and preference or avoidance of other tastants should be cautiously interpreted if tastants are offered after high EtOH concentrations. The safest procedure would be a completely between-subjects design for each tastant, starting with naive mice, or a serial test with naïve mice in which sucrose preference is left to last, as in our Experiment 3.
We doubt that any of the ethanol drinking data reported in this set of studies is of substantial pharmacological significance. By the end of Experiment 4, HDID mice were showing a high preference for 15% ethanol (90%, similar to that shown by C57BL/6J mice) and were ingesting about 1.7 g/kg during each 2 hr test. In a test of two-bottle DID with HDID-1 mice from the 9th
selected generation, they drank about 3 g/kg in the 4 hr test but reached average BECs of only 0.12 mg/ml (Crabbe et al, 2009
). Mice from the later generations reported here most likely drank more during their 2 hr tests than had the 9th
generation animals reported earlier, but we doubt that they drank enough more to achieve pharmacologically significant blood levels. They were also only drinking 15% ethanol in Experiment 4 rather than 20% when tested in S9. Intakes of 1.5 – 2.0 g/kg ethanol in a 2 hr limited access two-bottle preference study resemble those reported for C57BL/6J (Rhodes et al., 2007
) and are less than the 3 g/kg intakes seen in HAP mice (Grahame et al., 1999b
), genotypes with known high preference for ethanol. Estimates of BEC in HAP mice after 2 hr access with water available also reveal modest BECs, averaging less than 0.5 mg/ml, although a few mice reached BECs greater than 1.0 mg/ml. While HDID mice are willing to ingest sufficient ethanol in the single bottle DID test to reach behavioral intoxication (Crabbe et al, 2009
), it remains to be tested whether the current generation of mice will reach average BECs > 1.0 mg/ml when water is an alternative. On the other hand, BECs in the range between 0.5 and 1.0 mg/ml can have significant behavioral effects on mice. BECs in this range induce locomotor stimulation in FAST (Shen et al., 1995
) and a genetically heterogeneous (HS) stock of mice (Lessov and Phillips, 1998
) and facilitated performance on an accelerating rotarod task in some inbred strains (Rustay et al, 2003
Finally, the difference in genetic outcomes between Experiment 1 and 4 may reflect the different exposure models employed. There was a clear difference in preference between genotypes during 2 hr daily limited access, with an absolute difference in intake developing after a few weeks. Perhaps the daily, but limited, access to alcohol across the circadian day led to greater intake in HDID-1. The temporal pattern in which alcohol is offered clearly affects intake in rodents. Early studies with rats showed that offering 10% ethanol every other day for 24 hr periods led to an escalation in intake (Wise, 1973
), and this pattern has been employed more recently to increase intakes of higher concentrations as well (Simms et al., 2008
). The role of frequency and duration of ethanol access in HDID mice remains to be explored.