Mice lacking paternal expression of imprinted Grb10 are risk‐takers

Abstract The imprinted genes Grb10 and Nesp influence impulsive behavior on a delay discounting task in an opposite manner. A recently developed theory suggests that this pattern of behavior may be representative of predicted effects of imprinted genes on tolerance to risk. Here we examine whether mice lacking paternal expression of Grb10 show abnormal behavior across a number of measures indicative of risk‐taking. Although Grb10 +/p mice show no difference from wild type (WT) littermates in their willingness to explore a novel environment, their behavior on an explicit test of risk‐taking, namely the Predator Odor Risk‐Taking task, is indicative of an increased willingness to take risks. Follow‐up tests suggest that this risk‐taking is not simply because of a general decrease in fear, or a general increase in motivation for a food reward, but reflects a change in the trade‐off between cost and reward. These data, coupled with previous work on the impulsive behavior of Grb10 +/p mice in the delayed reinforcement task, and taken together with our work on mice lacking maternal Nesp, suggest that maternally and paternally expressed imprinted genes oppositely influence risk‐taking behavior as predicted.


| INTRODUCTION
The imprinted genes Grb10 and Nesp affect impulsive choice behavior in opposing direction. 1,2 Mice lacking paternal Grb10 (Grb10 +/p ) prefer a larger, but delayed reward to a smaller, but more immediate reward in the delayed reinforcement task (DRT). 1 In contrast, mice lacking maternal Nesp (Nesp m/+ ) choose a more immediate, smaller reward over a larger, but delayed reward in the DRT. 2 These behavioral findings, coupled with colocalization of expression of Nesp and Grb10 in a number of brain regions 3,4 and cell types, 1 has led to the suggestion that they may have an antagonistic effect on the control of behavior. 1,5 This fits with the general idea that genomic imprinting evolved as a consequence intragenomic conflict between maternallyand paternally derived alleles arising as a consequence of kinselection. 6,7 Recently a theoretical basis for how imprinted genes may influence risk-taking behavior has been proposed. 8 The theory comes from an extension of a model of bet-hedging, where an allele that leads to reduced mean reproductive success can be favored by selection if the allele also leads to a sufficiently large reduction in reproductive variance. An intragenomic conflict arises because the trade-off between selection on mean and variance is different for maternally and paternally inherited alleles. When reproductive variance is higher in males (as it is for most mammals) selection favors reduction of reproductive variance more strongly for paternally inherited alleles. When an allele is maternally inherited, selection more strongly favors increased mean reproductive success, even at the cost of increased reproductive variance. Following the "loudest voice prevails" principle, 9 this predicts that paternally expressed imprinted genes will promote risk-averse, variance-reducing behaviors, while maternally expressed imprinted genes will promote risk-tolerant, varianceincreasing behaviors.
Wilkins and Bhattacharya suggest that the opposing pattern of behavior shown by Grb10 +/p and Nesp m/+ mice in the DRT supports this idea. The DRT is generally considered a measure of impulsive choice 10,11 ; Grb10 +/p mice are more likely to choose the larger reward and are interpreted as less impulsive. However, in a naturalistic environment, delay introduces the possibility a reward will be lost to a competitor (loss of opportunity), or that the individual will be exposed to risk of predation before receipt of the reward (cost in death or injury). [12][13][14][15] Choosing a delayed, larger reward in the DRT may therefore indicate not only less impulsive, but also more risky behavior. 16 The idea that the choice of the more immediate, but smaller reward, displayed by Nesp m/+ mice in the DRT may reflect less risktaking is supported by their behavior in other tasks. Nesp m/+ mice show altered reactivity to, and are less willing to explore, novel environments. 3 The propensity to explore a novel environment is regarded as a good index of risk-taking behavior and has been used as such in a number of studies. 17 However, the data for the behavior of Grb10 +/p mice being indicative of an increased willingness to take risk are more equivocal. Here we address this, by examining the behavior of Grb10 +/p mice in a novel environment. We also directly tested risktaking behavior using the Predator Odor Risk-Taking (PORT) task. The PORT task was developed by us, 18 and has been used by others in both mice 19,20 and rats, 21 to specifically assess "real-life" risky situations where there is a trade-off between a cost (risk of predation) and a food reward. Although Grb10 +/p mice show no difference from wild type (WT) animals in their willingness to explore a novel environment, their behavior on the PORT task is indicative of increased risk-taking.
Follow-up tests suggest that this risk-taking is not simply because of a general decrease in fear, or a general increase in motivation for a food reward, but reflects a change in the trade-off between cost and reward.

| Animals
All procedures were conducted in accordance with the UK Animals The Grb10 null line was maintained on an F1-hybrid B6CBA F1/crl line from Charles River. Because of potentially confounding metabolic phenotypes associated with Grb10 m/+ mice, 22 comparisons of Grb10 +/p were made with WT littermate controls. Subjects were male mice aged between 3 and 6 months during testing. All mice were housed in single-sex, environmentally enriched cages (cardboard tubes, shredmats, chew sticks) of 2-5 adult mice per cage. Cages were kept in a temperature-and humidity-controlled animal holding room (21 ± 2 C and 50 ± 10% respectively) on a 12-h light-dark cycle (lights on at 7:00 h, lights off at 19:00 h). Standard laboratory chow was available ad libitum, but during the progressive ratio (PR) experiment water was restricted to 2 h access per day. This regime maintained the subjects at ≈90% of free-feeding body weight and motivated the animals to work for the food reward used in the task. All testing occurred during the light phase. Two separate cohorts were used. Cohort 1 (WT, n = 11; Grb10 +/p , n = 13) undertook locomotor activity (LMA), novelty place preference (NPP), milk consumption test, PORT task, acoustic startle with a minimum of 3 days between each test. Cohort 2 (WT, n = 14; Grb10 +/p , n = 11) undertook milk consumption testing and the PR, again with a minimum of 3 days between each test. All animals within a cohort were used throughout testing but a small number of animals were removed or lost (because of an inability to perform training stages or death) as testing progressed.

| Locomotor activity
LMA was measured using a battery of 12 activity cages, each measuring 21 × 21 × 36 cm. The activity cages were clear Perspex boxes containing two transverse infrared beams 10 mm from the floor, spaced equally along the length of the box, linked to an Acorn computer using ARACHNID software (Cambridge Cognition Ltd., Cambridge, UK). Activity was measured for 2-h sessions in the dark, over three consecutive days; data were collected in 5-min bins throughout each session. Testing took place at the same time every day. The cages were thoroughly cleaned after each animal, using 1% acetic acid solution. The main measure was "runs", recorded when the animal broke the two infrared beams consecutively.

| Novelty place preference
NPP was assessed using an apparatus consisting of two adjacent boxes with an opening in the middle, which could be occluded by a door.
These two arenas were made distinct by the color (black or white) and the texture of the floor (smooth plastic or sandpaper). In the first stage of the test the door was closed and the animal was placed in one side of the box for 1 h and allowed to habituate to that side. Then the mouse was taken out, the door removed and the mouse put back in the same side and allowed to explore both the habituated side and the novel side for a total of 30 min. The side to which the animal was habituated, and in which the sand-paper was placed, was pseudo-randomly allocated to avoid confounding the results. The arena was thoroughly cleaned after each animal, using 1% acetic acid solution.
The movement of each subject was tracked using a camera mounted approximately 2 m above the test arena, connected to EthoVision Observer software (Noldus Information Technology, Netherlands). Behavioral measures, obtained automatically, included the duration of time spent in each zone, frequency of entries, and latency of first entrance into the novel zone. The time spent in and the number of entries into the novel compartment were measured automatically by a video tracking system, using Noldus software.

| PORT task
The PORT task was conducted using the same methods and apparatus as previously described 18 and full details can be found in the Supporting information. Briefly, following habituation to the apparatus, animals were trained to leave the start chamber, traverse the middle chamber and to collect a food reward in the third chamber of the apparatus. For training trials, clean standard mouse bedding (wood shavings) was distributed evenly over the floor of the middle chamber.
Trial length was set to 10 min but was terminated when the subject had traversed the apparatus and was observed to have collected the reward. The mouse was then removed and placed in a holding box until the start of the next trial. In the test trials the middle chamber bedding was mixed with either "self-odor bedding" (bedding taken from the mouse's home cage), or "predator-odor bedding" (wood shavings mixed with a synthetic predator cue, 2,4,5-trimethylthiazoline [TMT]; Contech Inc., Canada). The main measurements taken in each trial was the latency to leave the start chamber.

| Predator odor enhanced acoustic startle response
The predator odor enhanced acoustic startle response (POE-ASR) was assessed in two separate test sessions, a week apart, immediately following a 10-min exposure to either untainted wood-shaving bedding (control condition) or fox odor-tainted bedding, mixed at the same concentration of TMT as used in the PORT task (see above). The order of odor presentation was counter-balanced between mice. ASR was monitored using SR-Lab apparatus (San Diego Instruments, USA), according to the previous method used 18 (see Supporting information for full details).

| Condensed milk test
In order to increase motivation and performance of the mice in the PR task mice were placed on a schedule of water restriction. Water was maintained at a 2 h regime for the duration of the experiment, and food was available ad libitum at all times (apart from when in chambers). Body weight was monitored prior to water restriction, and throughout the first 10 days of restriction. Once body weight had stabilized (>90% free-feeding weight) the mice were habituated to the condensed milk reward used in the operant tasks. This was carried out using the condensed milk test (CMT) to check for preference of condensed milk over water as described previously 18 (see Supporting information for full details). The amount drunk and preference for condensed milk over water were measured on five successive days.
This was to prevent mice having a neophobic reaction to the reward during the experiment and also to test for any differences in consumption and/or acquisition of a preference.

| Progressive ratio
All the sessions of the PR task were performed in 9-hole operant chambers (Cambridge Cognition Ltd, UK) modified for use in mice, as described previously. 23 For the PR task, only the central nose-poke hole was used. The mice were presented with a visual stimulus (light) recessed into the holes and were trained to respond to this stimulus with a nose-poke as recorded by infra-red beams spanning the hole. Conditioned reinforcement (CRf: one nose poke required for reward delivery) was carried out for 5 days. Following this, a PR schedule was carried out. Here, the number of nose pokes required to receive a reward ascended linearly every four trials for three (FR4) sessions. FR4 sessions were followed by three FR2 sessions, where the number of nose pokes required to receive a reward ascended linearly every two trials. These PR sessions were followed by three CRf sessions.
A number of measures were taken, including number of rewards received, latency to first nose-poke and latency to collect the reward.
During the PR sessions, an additional measure, the maximum number of nose pokes an animal was willing to make to receive a reward, was deemed the "breakpoint" (BP) and was the main indication of the animal's motivation to work for a reward.      F I G U R E 1 Locomotor activity (LMA) and novelty place preference (NPP) behavior in in Grb10 +/p and wild type (WT) littermates. As the LMA session progress, activity reduces (A), a pattern also seen across consecutive days (B). However, there were no activity differences detected between Grb10 +/p and WT mice. Similarly, in the NPP test all animals showed a preference in the proportion of time spent in the novel environment, but there were no differences absolute time spent in the novel chamber between Grb10 +/p and WT mice (C). This was supported by other measures in the NPP test, including number of entries into (D) and latency to first enter (E) the novel environment. Data are mean values ±SEM. # (p < 0.05) and ## (p < 0.01) indicate within subject (factors BIN or DAY) differences acquisition, the mice were trained to cross the apparatus to collect a reward, passing through the middle chamber which had plain wood shavings on the floor (Supporting information, Figure S1). Both Grb10 +/p and WT mice spontaneously demonstrated this behavior, and there was no difference between Grb10 +/p and WT mice (main effect of GENOTYPE, F 1,22 = 3.78, p = 0.072, partial η 2 = 0.20). These data suggest that habituation and task acquisition were equivalent between Grb10 +/p and WT mice.

| Statistical analyses
As expected, 18,19,21 the introduction of a predator odor into the middle chamber of the test arena significantly increased the overall latency to leave the start chamber, relative to the presence of a control odor (Figure 2 (Figure 2(B)). Furthermore, while all but one of the 14 WT animals showed an increase in the time in the Start chamber following the introduction of fox oor, only half (5/10) of Grb10 +/p mice showed a similar slowing of the latency to leave the Start chamber.
As previously, 18 we confirmed that the presence of the predator odor induced an equivalent fear response in the Grb10 +/p and WT mice. Exposure to fox odor caused a significantly enhanced (24% increase) acoustic startle response (ASR) relative to prior exposure to control bedding (Figure 3(A), (B), main effect of ODOR, F 1,22 = 11.34, p = 0.003, partial η 2 = 0.34), an increase that was equivalent in Grb10 +/p and WT mice (main effect of GENOTYPE, F 1,22 = 2.05, p = 0.166, partial η 2 = 0.09).

| Motivation for the food reward is not altered in Grb10 +/p mice
We performed the CMT to examine consumption and preference for a novel palatable foodstuff, namely 10% condensed milk. The total F I G U R E 2 Wild type (WT) and Grb10 +/p behavior in Predator Odor Risk-Taking (PORT) task. (A) All but one WT animal showed an increase in latency to leave the Start chamber when a predator odor (fox) was introduced into the middle chamber of the apparatus, relative to a control odor (wood shavings from the mouse's home cage). (B) In contrast, only 5/10 Grb10 +/p mice showed an increase upon introduction of the predator odor, and the overall magnitude of change in latency to leave the Start chamber was reduced. Representative traces from single trials of a WT mouse with control bedding (C) and fox odor (D); and a Grb10 +/p mice with control bedding (E) and fox odor (F) volume of milk consumed increased across successive days ( Figure 4 (A); main effect of DAY, F 3.10,68.2 = 33.28, p < 0.001, partial η 2 = 0.602), but did not differ between Grb10 +/p and WT mice (main effect of GENOTYPE, F 1,22 = 0.34, p = 0.565, partial η 2 = 0.015). All animals showed an initial aversion to the novel foodstuff (preference of less than 50%), but with subsequent exposures acquired a preference for the milk reward over water (preference of approximately 75%) ( Figure 4(B); main effect of DAY, F 2.66,58.6 = 25.63, p < 0.001, partial η 2 = 0.54). However, there was no difference between Grb10 +/p and WT mice in either their overall preference (main effect of GENOTYPE, F 1,22 = 1.79, p = 0.194, partial η 2 = 0.075) or in the rate at which their preference was acquired (interaction between GENO-TYPE and DAY, F 2.66,58.6 = 1.23, p = 0.305, partial η 2 = 0.53). These data are from the animals that went on to the PR test (cohort 2), but a similar pattern of results was seen in animals who went on to be tested in the PORT task (cohort 1, see Supporting information, Figure S2).
We then examined the motivation to work for a palatable solution using a PR task. Mice were initially trained to respond on a CRf schedule for five sessions. During the CRf sessions, subjects were able to carry out a maximum of 100 trials, which is equal to 100× 22 μl F I G U R E 3 Wild type (WT) and Grb10 +/p behavior in Predator Odor Risk-Taking task. Acoustic startle response in both WTs (A) and Grb10 +/p mice (B) showed an equivalent increase following pre-exposure predator odor in comparison to control odor F I G U R E 4 Palatable food consumption and progressive ratio behavior in Grb10 +/p and wild type (WT) littermates. Consumption (A) and preference (B) for 10% condensed milk increased with successive sessions but was not different between Grb10 +/p and WT littermates. In the PR task imposition of the FR4 (number of nose pokes required to receive a reward ascends linearly every four trials) and FR2 (number of nose pokes required to receive a reward ascends linearly every two trials) reduced the total number of trials relative to conditioned reinforcement (one nose poke required for a reward delivery), but there was no difference between Grb10 +/p and WT littermates (C). Similarly, the breakpoints at FR4 and FR2 were also equivalent between Grb10 +/p and WT mice. ## indicates within subject (factor DAY) differences p < 0.01 rewards within each 30 min test session. Analysis of the initial CRf stage revealed that both genotypes achieved the required level of performance per session, showing no effect of genotype (t 1,19 = 0.80, p = 0.44; data not shown). Following CRf training, subjects were switched to the two PR schedules: first FR4, followed by FR2. The BP was defined as the maximum number of nose pokes an animal was willing to make to receive a reward and is an indication of the animal's motivation to work for a reward. To demonstrate the effects of the imposition of the PR schedule, performance during the PR sessions were compared with the average of the three CRf sessions following PR testing (Figure 3(C)). Imposition of the PR schedule led to a signifi-

| DISCUSSION
Grb10 is currently a unique example of an imprinted gene in which the different parental alleles show distinct patterns of expression and have distinct physiological functions. 4 We have previously demonstrated that mice lacking a paternal copy of Grb10 have altered behavior, 24 including a higher tolerance of delayed rewards in a DRT. 1 One suggestion is that these changes reflect a role for Grb10 in regulating risk-taking behavior broadly. 8 We examined this in a number of tests showing that, although Grb10 +/p mice explore a novel environment to the same extent as their WT littermates, they are more risktaking on the PORT task. This is the direction of effects predicted from our previous analysis on the DRT 1 and by evolutionary theory, 8 and taken together these data suggest that Grb10 normally influences the cost versus risk assessment and acts to reduce risk-taking.
To test whether loss of paternal Grb10 expression would influence risk-taking we took a broad approach. First, we examined the propensity of Grb10 +/p mice to explore novelty, both in terms of basic LMA and habituation to a new environment, and also an explicit test of investigation of a novel environment. In both tests, the behavior of Grb10 +/p mice was equivalent to WT littermates. These data were supported by the condensed milk test, which was used to assess consumption and preference for a palatable substance but can also be regarded as a measure of food neophobia. The prediction from the behavior of Grb10 +/p mice on the DRT 1 would be that they will be more willing to take risks and therefore explore a novel environment more (or more quickly). It is possible that these tests are not sensitive enough, or that there is a ceiling effect, to detect such a "positive" change. Nevertheless, it seems that across novelty domains, Grb10 +/p mice appear to behave normally.
We also used an explicit test of risk-taking to examine the behavior of Grb10 +/p mice, namely the PORT task. This task was developed by us, 18 and has been used by others to assess risk-taking in both mice 19,20 and rats. 21 The PORT task examines a more ecologically valid aspect of risk-taking, namely the trade-off between a food reward and the risk of predation in obtaining that reward. Behavior in the task is sensitive to changes that affect this balance, such as the presence of a predator odor, or a reduction in the value of the reward. 18 In the PORT task, Grb10 +/p mice are quicker than WT littermates to leave the start chamber in the presence of a predator odor (fox) in order to obtain the food reward. There were no differences in habituation or acquisition of the task, and no difference in latencies in control trials, suggesting that the Grb10 +/p mice are more willing to take risks. Importantly, the difference in behavior in the PORT task is not because of changes in either fear, or motivation for palatable food alone, as predator odor enhanced acoustic startle and behavior in a PR task were equivalent between Grb10 +/p and WT mice. This suggests that loss of paternal Grb10 alters the point of trade-off between a obtaining a food reward and the risk of predation in obtaining that reward.
The direction of effects in the PORT task, where Grb10 +/p mice show increased risk-taking, is consistent with findings from the DRT.
Here, Grb10 +/p mice were more willing to wait for a large, but delayed, food reward. Although not developed as a direct test, behavior in discounting tasks such as the DRT have been correlated with, 25 and have been used as a proxy measure for, risk-taking. 16,26,27 Taken together, these data suggest that in the brain paternal Grb10 normally influences the cost versus risk decision-making, and acts to make mice more risk-averse. A complementary pattern is observed in mice lacking maternal Nesp, which show reduced exploration of a novel environment 3 and decreased tolerance of delay in the DRT. 2 This suggests opposite effects of Grb10 and Nesp on risk-taking behaviors. This suggests opposite effects of Grb10 and Nesp on risk-taking behaviors.
Interestingly, these genes show a strong degree of colocalisation, 1 including in neurons of the dorsal raphé nucleus and locus coeruleus, two brain areas known to modulate risk-taking behaviors. 28,29 Moreover, these patterns are consistent with the predicted direction of effects of imprinted genes on risk-related behaviors. 8 According to this model of bet-hedging and genomic imprinting, maternally and paternally expressed imprinted genes have conflicting influences on risk-tolerance as a consequence of differences in reproductive variance between males and females. When reproductive variance is higher in males (as it is for most mammals) then paternally expressed imprinted genes like Grb10 will promote risk-averse, variancereducing behaviors, and maternally expressed imprinted genes like Nesp will promote risk-tolerant, variance-increasing behaviors.
Opposing phenotypic effects of oppositely imprinted genes is a common feature of evolutionary models of genomic imprinting. The kinship theory of imprinting 30 attributes the evolution of imprinting to an intragenomic conflict arising from differences in the inclusivefitness effects of maternally and paternally inherited alleles. This has been most thoroughly studied in the context of pre-natal growth effects in mammals, where the fact that a female may have offspring by more than one male means that an offspring's demands on maternal resources have a greater adverse effect on an offspring's matrilineal kin than on its patrilineal kin. In this setting, theory predicts that paternally expressed imprinted genes will increase fetal growth, while maternally expressed imprinted genes will restrict growth, a pattern that has largely been borne out in data.
Although risk-related behaviors have not been formally modeled, the kinship theory has also been extended to behavioral phenotypes in a variety of ways. [31][32][33] It is possible that these sorts of risk-related imprinted genes could result from asymmetric inclusive-fitness effects. For example, we noted that a delayed reward can be viewed as a risky reward if there is a chance that the reward will be taken by a competitor. If that competitor is a conspecific, and the species has male-biased dispersal, the competitor would be more closely related, on average, to the focal individual's maternal genes than their paternal genes. Those maternal genes might then favor more risk tolerance, since the fitness consequences of losing the reward would be partially offset by the indirect fitness benefit to matrilineal kin. To distinguish between these different types of explanations would require both formal modeling and a characterization of the risk-related effects of imprinted genes in a broader range of taxa. In the example outlined above, where imprinted genes affect risk because of sex-biased patterns of dispersal, we would predict risk effects to covary with dispersal patterns across taxa. The bet-hedging model, 8 which is based on sex differences in reproductive variance, would predict more consistency in the pattern of imprinted-gene effects on risk in mammals, since males have higher reproductive variance except in rare cases.
Here we test the idea that the imprinted gene Grb10 is involved in modulating risk-taking behavior. Although not true across all domains, mice lacking paternal Grb10 do show increased risk-taking on the PORT task. This, coupled with previous work on a DRT, suggest that this idea is broadly correct. Taken together with our previous work on another imprinted gene, Nesp, these data suggest that maternally and paternally expressed imprinted genes oppositely influence risk-taking behavior. This "parliament of the mind," 33 caused by opposing parental genomes pulling impulsive choice and risk-taking in different directions, is an additional factor that should be taken into consideration when considering apparently irrational or sub-optimal choice behaviors. 34