These experiments, and prior work, allow us to draw several tentative conclusions about the effect of IA on ethanol consumption. First, when access is limited to 4 h per day, the effect of IA to increase ethanol consumption is not seen in mice. Although there was a significant difference between IA and CA groups of HDID-1 male mice in Experiment 1 during the first week after IA was initiated, this appeared to result more from a reduction in the CA group than an increase in the IA group (see Fig. ). No hint of a difference in intake between IA and CA groups in Experiment 2 was seen in C57BL/6J mice, even though IA can lead to increased access in this genotype when offered in 24 h-periods (Experiment 5; Hwa et al., 2011
; Melendez, 2011
). We did not test periods of intermittent availability between 4 and 24 h, and only tested four genotypes, so we cannot determine where in this interval the effect begins to appear for mice. We also cannot rule out the possibility that 4 h-access MWF would be effective in female C57BL/6J mice, or in rats.
A second conclusion, drawn from the literature, is that prior experience with ethanol is not necessary for IA to yield increased intakes. Wise's original study and several since have found this (Wise, 1973
; Melendez, 2011
), although some of them lack a CA comparison group (Pinel and Mucha, 1975
; Simms et al., 2008
A third conclusion we draw is that neither the HDID-1 nor the HDID-2 selectively bred mouse line responded to IA by increasing their ethanol intake vs. CA groups; neither did the HS controls. This suggests that if there are genetic contributors to IA escalation, they differ from those leading to high BEC during limited access drinking in the dark. There are no current systematic data that suggest that IA drinking is genetically influenced, but it would be very surprising if it were not, as all other ethanol drinking phenotypes we know of show such influences (Crabbe et al., 2010
). A study with the Sardinian Preferring rat selected line offered either 10 or 20% ethanol vs. water, either with CA or IA-MWF, for 3 weeks. The IA groups drank more ethanol than the CA groups, and their intake increased during the first three drinking sessions, but not thereafter (Loi et al., 2010
). Simms et al. (2008
) also found that MWF access enhanced intake vs. CA in three rat genotypes, Wistar, Long Evans (LE) and the selectively bred P rat. Intakes appeared to increase more in the Wistar and LE rats than in P rats; P rats started at higher intake levels . This study reported increased intake of either 10 or 20% ethanol; however, there were no CA comparison groups for the Wistar and P rats.
The insensitivity of HDID and HS mice appears to be true in both sexes. Because the genetically heterogeneous HS control line for this selection experiment also failed to show escalated drinking with IA, there may be a crucial gene or genes that is absent in this entire population. Although we cannot rule this out, we consider this scenario unlikely for two reasons. First, it implies a simple genetic structure to IA drinking (i.e. only a few genes responsible). Other kinds of ethanol drinking have shown complex genetic structures (multigenic or polygenic, Crabbe et al., 2010
). Furthermore, most murine genes are probably segregating in this HS/Npt cross, which was derived from intercrossing eight inbred strains (including C57BL/6J) that represent four distinct mouse lineages (Petkov et al., 2004
). The DID phenotype for which HDID-1 and -2 mice were selected is genetically somewhat distinct from standard two-bottle preference drinking (Rhodes et al., 2007
; Crabbe et al., 2011
), but it will be necessary to test more genotypes for the effects of intermittency, both to establish its sensitivity to the genetic background and to determine its genetic correlation with other traits.
Although entirely speculative, there is the possibility that elevated drinking with IA is only seen in mice of the C57BL/6J or a closely related lineage. This possibility is suggested by the lack of an effect in either male or female HS/Npt mice. Its likelihood is diminished by the fact that multiple rat genotypes have shown IA elevation. On the other hand, the robust effect of cycles of dependence and withdrawal to escalate drinking in mice (Lopez and Becker, 2005
; Griffin et al., 2009
) has been reported principally in C57BL/6J mice. A recent study (Lopez et al., 2011
) showed modest escalation after withdrawal in HAP-2 male mice, but not in HAP-2 females or either LAP-2 males or females (Lopez et al., 2011
We examined our HS data to see whether a subset of HS mice showed substantial elevation by first creating an index to quantify individual differences in the elevation due to IA. We treated CA and IA groups, and males and females separately. We took each individual's Day 8 g/kg intake (the first day when offered 20% ethanol) and subtracted it from its average intake during Week 4 (see Figs and ). We then computed the mean and standard deviation increase for each CA group. Using the raw increase score for each IA animal, we then expressed it as a difference from the mean increase score for the appropriate CA group. Finally, we standardized this index by dividing by the standard deviation of the CA group mean. The mean standardized effect of IA in HS females was 1.49, and for males was −0.17. These standardized scores were normally distributed with standard deviations of 1.57 and 0.82, respectively. Of the total population of 17 HS mice, 2 members of the IA groups showed increases of greater than 2 SD. Assuming a prior probability of a frequency of 2%, the presence of 2/17 scores so extreme was significant by the binomial test (P < 0.05). Both animals were females, and the 2 of 8 proportion was significant for the females only as well (P < 0.01). Thus, we tentatively conclude that there was a subset of HS mice (at least of HS females) that was particularly responsive to the effect of IA under the conditions we employed here. To determine whether these putative ‘responders’ were enriched for alleles from the C57 lineage would require detailed genotyping. The main conclusion is that most HS animals did not noticeably escalate intake under IA conditions. There are clearly large individual differences in the degree of escalation in the published rat studies as well.
C57BL/6J mice of both sexes show increased intake when offered IA to alcohol; thus, Experiment 5 provided a replication of the main finding of Hwa et al. (2011
). We remain puzzled by the difference in absolute intake between the mice in our study and those of Hwa et al
. The difference was apparent in both CA and IA groups, of both sexes. While patterns of genotypic influence on behavioral traits can certainly be shown to differ in different laboratory environments (Crabbe et al., 1999
; Chesler et al., 2002
), we reproduced nearly all the aspects of the Hwa et al
. method we could. We purchased mice from the same supplier (though they had to travel far further to reach Portland) and tested them at about the same age (9–10 weeks). The experimental protocol they used was reproduced as faithfully as possible in our laboratory. We could not duplicate their food, bedding, temperature, humidity, air quality or tap water. A reviewer of the current manuscript pointed out that we used 200 proof ethanol vs. 95%, so the dehydrating agents in our alcohol could have reduced our intakes through an effect on taste or through some other mechanism. And, of course, different experimenters performed the study in our laboratory than in theirs. Any of these factors could have affected the results. The only additional, similar study we know of that employed mice also used C57BL/6J males. Adult male mice were offered 15% ethanol vs. water either EOD or daily for 2 weeks. Intake across the 14 drinking sessions of the CA group remained stable at ~8 g/kg/day, while intake EOD began to increase on the third session and reached a plateau on sessions 4–7 of ~14 g/kg/day. A second experiment showed the same effect in adolescent mice: at this age EOD intake escalated to reach ~17 g/kg/day (Melendez, 2011
). However, the Melendez procedure did not acclimate the animals to lower concentrations of ethanol at the beginning.
What is the source of the intermittent escalation in drinking? Few studies that have reported on the phenomenon have tried to explain it mechanistically, either biologically or theoretically. Holloway et al.
explored several variables that influenced the degree of escalation in rats, which they found to include alcohol concentration, nature of prior experience with alcohol, frequency and regularity of periodic access and individual differences (Holloway et al., 1984
). They and all others who have looked also reported that the escalations did not persist if animals were switched from IA to CA—in our studies, we did not examine persistence. Pinel made the interesting suggestion that the effect of IA was not to escalate intake but rather that the effect of CA was to inhibit normal escalation of intake (Pinel and Huang, 1976
). He reached this conclusion because he and others showed that intermittency accompanied a higher intake of saccharin and quinine solutions as well as alcohol (Wayner et al., 1972
; Pinel and Huang, 1976
). Because these compounds did not share reinforcing effects, addictive potential or pharmacological effects, he concluded that the only common factor must be taste-related. He hypothesized that an ‘inhibitory factor’ that dissipated with time must be the explanation for the greater levels of intake in IA groups than CA groups, where dissipation could not be completed between drinking sessions.
Two additional unknowns are the length of the period of access to ethanol (e.g. here 4 vs. 24 h) and the length of the period between sessions of IA to ethanol that show effective increases due to intermittency. One study (Hargreaves et al., 2009
) offered beer (increasing to 4.4% ethanol) during two 1-h sessions daily or every third day. In this study, intermittency did not lead to intake greater than seen in a CA group. Other early studies explored the effect of longer periods of IA. Wayner et al. (1972
) saw increased intake with once each 3-day IA. Sinclair's group first gave rats 30 days of two-bottle preference testing, and then offered ethanol EOD, or once each 3 or 4 days. All three IA intervals yielded further increases in g/kg/day intake (Sinclair and Bender, 1979
). Neither of these studies had a CA comparison group. Holloway tested the effects of offering ethanol every 2nd, 3rd or 5th day, or once a week, and found the longer intervals to be the most effective (Holloway et al., 1984
). More recently, in a study in male C57BL/6J mice, adult animals were first given 6 weeks of two-bottle preference testing as a choice between 10% ethanol and water with 24 h CA (Melendez et al., 2006
). Thereafter, ethanol was withheld for 6 days in one group, while the other group continued to have an ethanol–water choice. Ethanol was returned for 24 h after the abstinence period, and the cycle of 6 days off, one day on was repeated 10 times. The group offered ethanol only once a week showed escalating intakes beginning in the 2nd week and eventually was consuming 18 g/kg/day, while the CA group maintained a stable intake of ~10–11 g/kg/day. The author characterized this result as an example of an alcohol deprivation effect (ADE). The ADE was first reported in rats (Sinclair and Senter, 1967
) and is a well-known phenomenon that has been replicated in many species. It is taken by many as a model for relapse drinking (for reviews, see Sinclair, 1972
; Lê and Shaham, 2002
Another remaining puzzle is that we do not know how long it takes to develop increased intake under IA conditions. A recent study offering 20% ethanol vs. water on MWF only for 3–4 months included a CA access comparison group. After this initial period, the IA-MWF group was drinking more ethanol than the CA group; however, no data were presented describing their acquisition of drinking during the initial 3–4 months (Hopf et al., 2010
). Our data (Experiments 4 and 5), Hwa et al. (2011
), and Melendez (2011)
suggest that approximately a week of IA is required to see the effect, but this may only be true for paradigms that employ gradually escalated ethanol concentrations. Other studies with or without acclimation have shown the increase either virtually immediately or only after several days (e.g. Wise, 1973
; Pinel and Mucha, 1975
; Pinel and Huang, 1976
; Pinel et al., 1976
; Melendez, 2011
). With long intervals of intermittency (one week), both immediate and delayed increases have been reported (Holloway et al., 1984
; Melendez et al., 2006
Another variant of this paradigm has come to be termed ‘multiple scheduled access’ (MSA). In the initial report (Murphy et al., 1986
), male P rats were given access to 10% ethanol and water either continuously (24 h/day, 7 days/week) or during a single 4 h period of access each day, or during 4, 1 h access periods, spaced 2 h apart during the dark cycle. Animals given four separated exposures drank more during their total 4 h of access than animals given all 4 h continuously: both groups drank less (g/kg/day) than animals continuously exposed to ethanol for 24 h/day. The development of the groups’ drinking across days was not reported. A later study offered female rats concurrent access to 15 and 30% ethanol either continuously or MWF—i.e. offered MSA. All animals always had water access. By the third week of MSA, these animals were drinking as much on a g/kg/day basis during their total 3 h access as the CA, 24-h group (Bell et al., 2006a
). Subsequent studies using this method have shown similar increases with intermittent MSA (Bell et al., 2006b
). In these studies, ethanol intake remained less on a g/kg/day basis in animals receiving only 3 × 1 h access than in animals offered CA. A recent study compared CA with intermittent MSA in adolescent and adult P rats. Remarkably, in this study, adolescent rats drank more during the 3 h MSA periods than CA adolescents did over the whole 24 h access period (Bell et al., 2011
). Also, we only offered a single, uninterrupted access period each day (of 24 h length in most studies), so it is unclear how to compare our results with those studies that saw increased intake in rats offered 2, 3 or 4 × 1 h access to two different ethanol concentrations. These MSA studies led to increased intake vs. a single 2, 3 or 4 h exposure.
Finally, we reiterate several limitations of these results. We do not know whether periods of access longer than 4 h but less than 24 h would lead to greater intake. It may be imprudent to generalize characterizations of the effect across murine and rat species. While EOD and MWF schedules of intermittency seem equivalent in rats, once/week also works in mice, and we have no directly comparable mouse data for such schedules as the MSA procedure. We have explored a limited number of genotypes. Although the effect seems to be present with or without choice of ethanol and water, it is unclear what its parameters might be if multiple ethanol concentrations were to be offered simultaneously (for example, ethanol intake increases as a direct function of the number of ethanol bottles offered in addition to a water bottle) (Tordoff and Bachmanov, 2003
). Although the effect seems to appear early during repeated intermittent offerings, we have little data systematically exploring this parameter. We are more comfortable in believing that both sexes of both species show an IA increase in drinking, but this does not necessarily imply that the underlying mechanisms are the same. In conclusion, further work will be needed to determine whether this method can yield substantial and repeated levels of self-intoxication.