Btbd3 expression regulates compulsive-like and exploratory behaviors in mice

BTB/POZ domain-containing 3 (BTBD3) was identified as a potential risk gene in the first genome-wide association study of obsessive-compulsive disorder (OCD). BTBD3 is a putative transcription factor implicated in dendritic pruning in developing primary sensory cortices. We assessed whether BTBD3 also regulates neural circuit formation within limbic cortico-striato-thalamo-cortical circuits and behaviors related to OCD in mice. Behavioral phenotypes associated with OCD that are measurable in animals include compulsive-like behaviors and reduced exploration. We tested Btbd3 wild-type, heterozygous, and knockout mice for compulsive-like behaviors including cage-mate barbering, excessive wheel-running, repetitive locomotor patterns, and reduced goal-directed behavior in the probabilistic learning task (PLT), and for exploratory behavior in the open field, digging, and marble-burying tests. Btbd3 heterozygous and knockout mice showed excessive barbering, wheel-running, impaired goal-directed behavior in the PLT, and reduced exploration. Further, chronic treatment with fluoxetine, but not desipramine, reduced barbering in Btbd3 wild-type and heterozygous, but not knockout mice. In contrast, Btbd3 expression did not alter anxiety-like, depression-like, or sensorimotor behaviors. We also quantified dendritic morphology within anterior cingulate cortex, mediodorsal thalamus, and hippocampus, regions of high Btbd3 expression. Surprisingly, Btbd3 knockout mice only showed modest increases in spine density in the anterior cingulate, while dendritic morphology was unaltered elsewhere. Finally, we virally knocked down Btbd3 expression in whole, or just dorsal, hippocampus during neonatal development and assessed behavior during adulthood. Whole, but not dorsal, hippocampal Btbd3 knockdown recapitulated Btbd3 knockout phenotypes. Our findings reveal that hippocampal Btbd3 expression selectively modulates compulsive-like and exploratory behavior.

central nose poke port on a fixed-ratio 1 schedule in a 30-minute session. The central nose poke port was lit continuously until the animal nose-poked for reward. Once the mouse retrieved the reward, a variable intertrial interval (ITI) was utilized. Once animals consistently responded in the central port, they were trained in the Go/No-Go paradigm. In Phase III of training, mice were trained to respond in the central port during a timed stimulus presentation, in which the central port was lit and active for 8 seconds, then 5 seconds, then finally 3 seconds, in five, 30-minute sessions for each stimulus duration. Mice that did not reach criterion (20 successful trials within a session) by the end of the fifth training session at the 3second stimulus interval were excluded from analysis for the Go/No-Go paradigm. Mice were also punished for incorrect responses (responses in unlit ports), for omissions (lack of response during the stimulus window), and for premature responses (responses prior to initiation of the trial indicated by light in the central port). Punishment consisted of a 4 second time out with the house light on. Animals had to retrieve reward from the magazine to initiate the subsequent trial. Initially, a variable ITI of 5 -17 seconds was used. However, this resulted in high levels of premature responding; thus the ITI was shortened to 3 -7 seconds during Phase III of training and was maintained throughout Go/No-Go testing.
Mice then underwent 15 30-minute Go/No-Go test sessions. A "Go" trial consisted of the central port being lit during the 3-second stimulus presentation, as during training except that responses in incorrect ports were not punished. "No-Go" trials consisted of the central port being lit concomitantly with an additional, green, light just above the central port for the 3-second stimulus interval, creating a compound stimulus. Mice were rewarded for refraining from responding to the compound stimulus, and were punished with a time out for responding (a "false alarm"). The first five days were training for the No-Go stimulus. The remaining ten days of testing were used for analysis. The false alarm rate provided the primary outcome measure of response inhibition, with a high false alarm rate indicating impairments in action restraint, a type of impulsivity 1 . Premature responding provided the secondary response inhibition measure. The sensitivity index d-prime (d') measures how well animals learn the task by comparing the hit rate and false alarm rate.
Progressive Ratio Breakpoint. Following completion of Go/No-Go testing, mice moved on to the PRBP task. First, mice underwent a refresher training session at Phase II. The next day, mice were tested in the 60-minute PRBP test session, as previously described 2 . Briefly, mice were required to respond in the central nose poke port progressively more times in order to earn a reward in the following steps: 1,2,4,7,11,16,22,29,37,46, 56 and 67 responses required for reward. The "breakpoint," defined as the highest number of responses the animal would perform to obtain a reward, was the primary outcome measure. All mice that reached criterion for Go/No-Go testing, regardless of the days to criterion, were included in the PRBP and PLT analyses, as days to criterion had no effect on breakpoint (F(2,43) = 1.03; p = .37).
Probabilistic Learning Task. Finally, animals underwent PLT testing. First, animals underwent one day of training using a modified version of the Phase II training session, in which all five ports were lit and active for earning reward. The following day, mice underwent PLT testing in 3 blocks of 60 trials for a maximum 60-minute session. Two ports were lit for each trial, and were counterbalanced across groups.
Within a block, one port was the "target" port and the other the "non-target." The target port was commonly rewarded (90, 80, or 70% of the trials for blocks 1, 2, and 3 respectively) and the non-target port was uncommonly rewarded (10, 20, or 30% of the trials). After completing the 60 trials for block 1 (90 or 10% reward probabilities), the two previously lit ports were extinguished and two new ports were lit, one with 80% and the other with a 20% reward probability. Similarly, for the third block, the previously lit ports were extinguished and two new ports were lit, one with 70% and one with 30% reward probability. "Win-stay" and "lose-shift" strategies on the target and non-target ports were the primary outcome measures, where "win-stay" denotes a trial in which an animal returns to the same port in which they were rewarded on the previous trial. "Lose-shift" refers to a trial in which an animal shifts responding to the opposite port after receiving punishment on the previous trial. Accuracy was also a primary outcome measure, defined as the percentage of trials in which the animal responded on the target port.
Olfactory Dis/Habituation. The olfactory dis/habituation test was used to assess olfactory sensation as previously described 4 . Mice were placed in a cage. Cotton swabs were dipped in odorant and lowered into the cage for 60 seconds, followed by a 2-minute intertrial interval. Each odorant had 3 trials in a row to habituate the animal. Then a novel odorant was introduced and repeated for a total of 3 trials. The number of sniffs was recorded. The first odorant was always water, followed by isoamyl acetate or ethyl acetate in a counterbalanced fashion.
Olfactory Memory. The olfactory memory test was used to measure the memory retention of a familiar odor as previously described 4 . Stimuli were presented in the same setup as for olfactory dis/habituation. Animals were first exposed to ethyl vanillin for a period of 4 minutes to habituate them to the stimulus.
An hour later, animals were exposed to ethyl vanillin again for a second trial of 4 minutes. If the animal remembers the odorant, they are expected to sniff the stimulus less during the second trial than the first.
Whisker Brushing. Whisker brushing was performed as previously described 5 . Briefly, animals were scruffed in one hand while the other brushed the distal end of the whiskers on each side of the face.
Turning the face during or just after whisker brushing was considered indicative of a response. Footprint Test. The footprint test was used to assess motor coordination and balance as previously described 6 . Animals were placed in a corridor (70 cm long) lined with paper. First, animals underwent a habituation phase (2, 10 minute sessions) to train them to run the corridor. A chunk of milk chocolate was placed at the end of the corridor. The end of the corridor was covered to make it dark. Animals were scruffed and front paws were painted one color and hind paws were painted a different color. Animals were then placed at the beginning of the corridor and allowed to explore the corridor. If the animal reached the end of the corridor and the chocolate reward, they were placed back at the start with a fresh piece of chocolate at the end. In the test phase, the corridor was lined with fresh paper and a piece of chocolate. The animal's paws were painted and then the animal was placed at the start. As soon as the animal reached the end of the corridor or turned around in the corridor, the animal was removed and the trial was over. The test phase was repeated for a total of three trials or until a clear set of footprints with the animal moving in a straight path was obtained. Output measures were stride length (distance between footfalls with the same foot), overlap (distance between the center of the plantar of the fore and hind limb on the same stride), and base width (distance between the fore or hind feet on the same stride).
AAV Generation and Delivery. The AAV2/8 serotype was used because of its efficient gene delivery to the central nervous system of neonatal mice 7,8 . Cetalomegalovirus (CMV) was selected as the promoter because of its transduction efficiency in the central nervous system in neonatal rodents 9 and long-term expression 10 . A woodchuck posttrascriptional regulatory element (WPRE) was included to enhance gene expression 11 . The P2A peptide permitted separate expression of Cre recombinase and tdTomato 12 .
TdTomato was used as the reporter because of its brightness and photostability 13 .
Mice were initially assigned to a group based on visual estimation of sex on postnatal day 2. A few supplemental mice received viral infusions for each group due to anticipated exclusion of a subset of animals due to misplaced infusions. Postnatal day 2 mice were removed from the home cage and placed on a heating pad. Animals were cryoanesthetized on ice prior to surgery. The head was placed on a chin rest custom fitted into a stereotaxic device (Stoelting, Wood Dale, IL, USA). Virus was administered using a syringe pump (Harvard Apparatus, Holliston, MA, USA) holding a 10 μl syringe (#1701, Hamilton Company, Reno, NV, USA) attached to plastic tubing and a 33-gauge cannula (Plastics One, Roanoke, VA, USA) that was secured into the stereotaxic device. Coordinates were measured relative to lambda. Infusions were performed bilaterally at a rate of 0.1 μl/min for a total infusion of 0.25 μl per side.
The cannula was left in place for an additional minute for diffusion of virus prior to slowly drawing up the cannula. Animals were then removed from the stereotaxic device, and paws were tattooed for identification before returning to the heating pad. Once fully recovered from anesthesia, animals were rolled in soiled bedding from the home cage and returned to the home cage. for processing. Brains were cut coronally at 200 μm section thickness on a vibratome. Sections were collected in 6% sucrose, blotted dry, and mounted onto gelatin-coated slides. Slides were then alkalinized in 18.7% ammonia. Next, slides were developed in Dektol (Kodak, Rochester, NY) and fixed in Kodak Rapid Fix. Slides were dehydrated in ethanol and cleared in xylene before coverslipping. Images were taken of each selected neuron at 600× magnification using a motorized microscope (Axioplan 2, Carl Zeiss, LLC, USA) and a camera (DXC-390, Sony Corporation, Tokyo, Japan) and reconstructed using Neurolucida software (MBF Bioscience, Williston, VT) to obtain the complete dendritic tree. Threedimensional analysis of the reconstructed neurons was performed using NeuroExplorer software (MBF Bioscience). Sholl analysis was performed on both apical and basal dendrites for pyramidal neurons.
Spine density was assessed for all dendrites except on mediodorsal thalamus spiny stellate neurons due to insufficient resolution in the limited dendritic tree. Measurements were averaged across neurons within a subject, and this averaged measurement for each animal was used as the unit of analysis for statistical testing.

Barbering
Mantel-Cox log rank tests revealed significant effects of fluoxetine treatment compared to vehicle pooled across genotypes (Figure 1c Open Field. In constitutive Btbd3 WT, HT, and KO mice (n = 17 female/27 male WT, 34 female/27 male HT, and 11 female/20 male KO mice), there was no effect of genotype on spatial d (Supplementary Figure   1h). However, a genotype by bin interaction (F(16,1040) = 2.57; p<.0001) in a repeated measures ANOVA and post hoc tests revealed that Btbd3 KO mice traveled a greater total distance than WT or HT mice in bins 1-4 ( Figure 2a). However, post hoc ANOVAs showed main effects of bin for each genotype for distance traveled. Furthermore, no difference in locomotor habituation was found between genotypes, assessed as the percent distance traveled in bin 9 relative to bin 1 (F(2,133) = 0.822; p = .44).
In Btbd3 whole hippocampal KD mice, a viral condition by bin interaction was identified in a repeated measures ANOVA for total distance traveled (Figure 4d; F(8,560) = 4.22; p<.0001). Post hoc tests revealed higher activity levels in mice receiving AAV-Cre mice than AAV-tdTomato within each bin.
Post hoc ANOVAs revealed a main effect of bin within each viral condition. Thus, consecutive bins were compared within each viral condition to examine habituation to novelty (Figure 4d). A significant difference was found between distance traveled in bins 1 and 2, and between bins 2 and 3, within control mice. However, there were no differences in distance traveled between any two consecutive bins within Btbd3 KD mice. Next, habituation was compared between viral conditions. AAV-Cre mice habituated less than AAV-tdTomato mice, as indicated by a greater percent total distance traveled in bin 9 relative to