Corresponding author: W. H. Meck, Department of Psychology and Neuroscience, Genome Sciences Research Building II—3rd Floor, 572 Research Drive—Box 91050, Duke University, Durham, NC 27708, USA. E-mail: email@example.com
Interval timing within the seconds-to-minutes range involves the interaction of the prefrontal cortex and basal ganglia via dopaminergic–glutamatergic pathways. Because the secreted protein brain-derived neurotrophic factor (BDNF) is able to modulate dopamine release as well as glutamatergic activity, we hypothesized that BDNF may be important for these timing mechanisms. Recently, the calcium-responsive transcription factor (CaRF) was identified as an important modulator of BDNF expression in the cerebral cortex. In this study, a strain of Carf knockout mice was evaluated for their ability to acquire the ‘Start’ and ‘Stop’ response thresholds under sequential and simultaneous training conditions, using multiple (15-second and 45-second) or single (30-second) target durations in the peak-interval procedure. Both Carf+/− and Carf−/− mice were impaired in their ability to acquire timed response thresholds relative to Carf+/+ mice. Additionally, control mice given microinjections of BDNF antisense oligodeoxynucleotide to inhibit protein expression in the prefrontal cortex showed timing impairments during acquisition similar to Carf mice. Together, these results suggest that the inhibitory processes required to update response thresholds and exert temporal control of behavior during acquisition may be dependent on CaRF regulation of genes including Bdnf in cortico-striatal circuits.
Interval timing is fundamental to the generation of patterns of behavior, including those found in optimal foraging, motor control and decision making (Buhusi & Meck 2005; Meck 2003). A common procedure for studying the temporal control of behavior is the peak-interval (PI) procedure, in which subjects learn to set independent ‘Start’ (S1) and ‘Stop’ (S2) response thresholds for the control of a timed response sequence centered on a target duration (Buhusi & Meck 2009). Thus, the PI procedure offers the opportunity not only to examine the acquisition of response sequences that involve the inhibition of responding as a function of time since signal onset but also changes in the accuracy and precision of timing as a function of behavioral and physiological variables (Balci et al. 2009b; Buhusi et al. 2009; Church et al. 1994; Jin & Costa 2010; Matell et al. 2003, 2004, 2006, 2011; Meck & Church 1984). In the PI procedure, one typically observes a 'high state' of responding defined by the S1 and S2 time points that is centered on or around the target duration. At steady-state performance, the width of this 'high state' is regulated such that the ratio of its width to the associated target duration is referred to as the Weber fraction (WF), which is usually constant across a range of target durations and represents the scalar property of interval timing, i.e. variability in the placement of the S1 and S2 response thresholds increases in proportion to the length of the target duration (Church et al. 1994; Gibbon et al. 1997).
Timing and time perception in the seconds-to-minutes range involves the interaction of the basal ganglia and the prefrontal cortex via dopaminergic–glutamatergic pathways (Coull et al. 2011; Jones & Jahanshahi 2011; Matell et al. 2003). Manipulations of these dopaminergic systems are able to modify interval timing by altering clock speed and other properties of the internal clock (Balci et al. 2009a, 2010; Cheng et al. 2006a,b, 2007; Coull et al. 2011; Drew et al. 2003, 2007; Höhn et al. 2011; Lake & Meck 2013; Lustig & Meck 2005; Meck 1983, 1996, 2006b; Meck et al. 2012; Ward et al. 2009). Brain-derived neurotrophic factor (BDNF) modulates dopamine (DA) levels, release and uptake (Blochl & Sirrenberg 1996; Hyman et al. 1991). The BDNF is expressed in glutamatergic pyramidal neurons in the cortex as well as dopaminergic neurons in the brainstem, both of which synapse on GABAergic medium spiny neurons (MSNs) of the striatum. The BDNF itself is transported to the striatum where it activates TrkB receptors on MSNs. The majority of BDNF protein within the striatum arises from the cortex (Baquet et al. 2004). Cortical BDNF can influence cortical-striatal circuit function by modulating release of glutamate in the striatum (Berglind et al. 2009). Furthermore, striatal MSNs are particularly sensitive to BDNF-dependent regulation of dendritic complexity and spine density (Rauskolb et al. 2010) and have been shown to be crucial to the coincidence-detection processes underlying interval timing (Allman & Meck 2012; Buhusi & Oprisan 2013; Coull et al. 2011; Lustig et al. 2005; Matell & Meck 2004; Oprisan & Buhusi 2011). Precise temporal and spatial control of BDNF expression is essential to its function.
Bdnf mutations that reduce either overall levels of BDNF or BDNF secretion have substantial effects on brain development and plasticity (Chen et al. 2006; Egan et al. 2003; Genoud et al. 2004). Transcriptional control is a major mechanism of BDNF regulation (Lyons & West 2011; McDowell et al. 2010; Tao et al. 2002). One important modulator of Bdnf transcription is the calcium-responsive transcription factor (CaRF—Pfenning et al. 2010; Tao et al. 2002; West 2011; West & Greenberg 2011). Although CaRF is expressed throughout the brain, Carf knockout (KO) mice show a selective reduction of BDNF expression only in the cortex, but not in the striatum or hippocampus (McDowell et al. 2010). At present, it is unknown what impact the loss of CaRF function might have on interval timing. Consequently, mice with Carf gene deletions were evaluated for their ability to acquire S1 and S2 response thresholds using multiple target durations (15-second and 45-second) as well as a single target duration (30-second) in order to evaluate the contribution of CaRF signaling to the acquisition of timed response sequences in the PI procedure (Jin et al. 2009; MacDonald et al. 2012). In addition, antisense BDNF oligonucleotide (BDNF-ODN) microinjections into prefrontal cortex of C57BL/6J mice were used to evaluate the effect of blocking BDNF expression on the acquisition of S1 and S2 response thresholds in a 30-second PI procedure (Bosse et al. 2012; Dluzen et al. 2002; Lau et al. 1998; Lee et al. 2004).
Materials and methods
Experiment 1 used 20 male mice: 5 Carf+/+ wild-type (WT), 10 Carf−/+ heterozygous (Het) and 5 Carf−/− KO littermates. Experiment 2 used 40 male mice: 5 Carf+/+ WT, 7 Carf−/+ Het, 8 Carf−/− KO littermates, 10 C57BL/6J mice (Charles River Laboratories, Raleigh, NC, USA) given vehicle microinjections (CON) and 10 C57BL/6J mice given antisense BDNF oligonucleotide microinjections (BDNF-ODN). All Carf mice were generated at Duke University. The establishment of the Carf−/− strain was reported in McDowell et al. (2010). These mice were maintained on a mixed 129S4/C57BL6 background and were bred as female Carf+/− × male Carf+/− crosses to yield Carf+/+, Carf+/− and Carf−/− littermates for all the experiments reported in this study. Mouse pups were weaned at 3 weeks of age and adult mice in both experiments were housed four mice per cage in a climate-controlled animal colony with a 12:12 light/dark cycle (lights on at 0700 h and off at 1900 h). Standard rodent chow (5001—Purina LabDiet®, St. Louis, MO, USA) and water were available ad libitum in the home cages throughout the experiments, except during the food-restricted period of behavioral testing described below. Mice were assigned to one of the 10 lever boxes and trained in two behavioral sessions with different genotypes or microinjection conditions distributed across daily sessions that were conducted 7 days/week as described below. During behavioral training, mice were maintained at 85–90% of their normal body weight. All experiments were conducted under a protocol approved by the Duke University Institutional Animal Care and Use Committee in accordance with the National Institutes of Health guidelines for the care and use of animals. The mice were approximately 3 months old when the experiments began and were tested during the light phase of their light/dark cycle at approximately the same time each day.
Mice were anesthetized with a ketamine/xylazine cocktail (100/10 mg/kg, ip) and placed in a mouse stereotaxic apparatus (David Kopf Instruments, Tujunga, CA, USA). Bilateral guide cannula (33-gauge, Plastics One, Roanoke, VA, USA) was implanted in the medial prefrontal cortex (A/P 1.7, M/L ±0.5 and D/V −2.5 mm; Paxinos & Franklin 2008). Stainless-steel stylets were placed in the guide cannulae to maintain patency. To minimize pain and distress during the first 48 h post-surgery, mice received subcutaneous injections of buprenorphine (0.5 mg/kg every 12 h).
The BDNF antisense oligonucleotide sequence (ODN; 5′-TCT TCC CCT TTT GGT-3′; Integrated DNA Technologies, Coralville, IA, USA) or phosphate-buffered saline vehicle (VEH) was infused through the bilateral cannulae at a rate of 0.50 µl/min and 1 nmol/µl per side. Infusion cannulae were attached with polyethylene tubing (PE50; Plastics One) to a 10-µl Hamilton syringe (Reno, NV, USA), which was controlled by a microinfusion pump (KDS 100; KD Scientific, New Hope, PA, USA). Injection cannulae were left in place for 30 seconds after infusion to allow for diffusion of drug away from cannula tip. Microinjections were given to the CON and BDNF-ODN groups of mice 15 min prior to the beginning of sessions 1–5 of PI training as described below.
The experimental apparatus consisted of 10 matching lever boxes (Model ENV-307A; Med Associates, St. Albans, VT, USA) housed in sound-attenuating chambers (Model ENV-021M; Med Associates). The dimensions of each lever box were 21.59 × 17.78 × 12.70 cm. The ceiling, side walls and door of each box were made from clear Plexiglas. The front and back walls were stainless-steel panels and the floor was made of parallel stainless-steel bars. The front wall of each box contained left and right retractable levers; a food cup was located between the levers and a cue light was located directly above the food cup. A pellet dispenser delivered 20-mg grain-based food pellets (Research Diets, Inc., New Brunswick, NJ, USA) into the food cup. The back wall of each box contained a house light (14-W, 100 mA) directed toward the ceiling. The operant chambers were controlled by the Med-PC IV software package. The fan was on throughout the session. An IBM-PC compatible computer attached to an electronic interface (Model DIG-700 and SG-215; Med Associates, Inc.) was used to control the experimental equipment and record the data. The time of each lever press was recorded to an accuracy of 10 milliseconds and placed into 1-second time bins.
Experiment 1: sequential acquisition of S1 and S2 response thresholds with 15-second and 45-second target durations
Pretraining (PT) of lever pressing (PT sessions 1–12). All mice were given 12 daily sessions of lever-press training in which both the left and right lever were present, but one of these levers was randomly selected with equal probability to be retracted and inserted in a 1-second cycle for 3 seconds prior to the delivery of a food pellet once every 60 seconds, thus constituting a conditioned stimulus (CS)–unconditioned stimulus (US) interval. In addition to the free food delivered as a function of these CS–US pairings, a food pellet was delivered for every lever press. Sessions ended after the mouse received 60 food pellets (free + earned) or 60 min had passed, whichever came first. All mice acquired the lever press response as indicated by the receipt of 60 food pellets in less than 30 min.
15-second and 45-second fixed-interval (FI) training (FI sessions 1–15). In Experiment 1, a dual 15-second and 45-second FI training schedule was used. Fixed-interval trials were signaled by the onset of a white-noise signal and the appropriate lever(s) was primed for reinforcement at the associated target duration(s). The target duration used on each trial (15-second or 45-second) was randomly selected with equal probability and no external cue was given to indicate which lever/duration was in effect. In all cases, the first response following the selected target duration resulted in the delivery of a food pellet, signal termination and the onset of a random duration intertrial interval (ITI), range 30–90 seconds. The assignment of target durations to response levers was counterbalanced both within and across groups of mice.
15-second and 45-second PI training (PI sessions 1–20). Peak-interval training was used to assess the accuracy and precision with which mice timed the target duration(s). Sessions consisted of two trial types: FI trials (as described above) and unreinforced probe trials. During probe trials the lever was extended and the white-noise signal turned on for a minimum of three times the target duration plus an additional random amount of time with a mean of 20 seconds, uniformly distributed. No food was available for lever pressing on these unreinforced probe trials. Fixed-interval and probe trials were ordered randomly with 50% probability each. Thus, one of the two target durations (15-second or 45-second) was presented in conjunction with non-reinforced probe trials in a random sequence. No external cue was provided to indicate which, if any, lever/target duration would be selected for reinforcement on any trial. Mice were free to respond on the lever(s) at any time during the session, although only responses made to the appropriate lever following the target duration were reinforced. See Matell et al. 2004, 2006 for additional procedural details concerning this type of PI procedure.
Experiment 2: simultaneous acquisition of S1 and S2 response thresholds with a 30-second target duration
Pretraining of lever pressing (PT sessions 1–15). All mice were initially given 12 daily sessions of lever-press training in which one response lever was available (left or right—counterbalanced across groups). This lever was normally present in the chamber, but retracted and inserted in a 1-second cycle for 3 seconds prior to the delivery of a food pellet once every 60 seconds, thus constituting a CS–US interval. In addition to the free food delivered as a function of these CS–US pairings, a food pellet was delivered for every lever press. Sessions ended after the mouse received 60 food pellets (free + earned) or 60 min had passed, whichever came first. All mice acquired the lever press response as indicated by the receipt of 60 food pellets in less than 30 min. The last three sessions of pretraining consisted of random-interval training with a mean of 30 seconds (range of 3–90 seconds, uniformly distributed) separating the last reinforced response from the next priming of reinforcement. This random interval training was designed to facilitate the transition from free-operant responding to the discrete-trial procedure used in 30-second PI training.
30-second PI training (PI sessions 1–25). In Experiment 2, a single target duration of 30 seconds was used. The basic idea was to provide a simpler discrimination for the Carf mice to learn than the multiple 15-second and 45-second target durations (each associated with a different response lever) used in Experiment 1. In addition to the 30-second target duration, mice were allowed to acquire and express S1 and S2 response thresholds simultaneously. What this means is that mice were transitioned to the PI procedure immediately following pretraining rather than requiring an initial phase of FI training in which a S1 response threshold is acquired before the presentation of unreinforced probe trials and the subsequent acquisition of the S2 response threshold (see Balci et al. 2009b). Peak-interval training was conducted in a manner similar to that described for Experiment 1 in which trials were signaled by the onset of a white-noise signal and reinforcement was delivered for the first response following the 30-second target duration. The delivery of reinforcement was also associated with termination of the white-noise signal and the onset of a random duration ITI, range 30–90 seconds, uniformly distributed. Mice in the BDNF-ODN and VEH groups were given bilateral microinjections of BDNF ODN or phosphate-buffered saline into the prefrontal cortex prior to PI sessions 1–5 as described above. No injections were given thereafter.
Individual mean peak functions for each target duration were fit using a Gaussian curve with the addition of a linear ramp function to account for right-tailed skew. These fits accounted for over 90% of the variance for all groups of mice and did not reliably differ as a function of treatment condition. The Gaussian fits were used to obtain peak time (a measure of accuracy), peak spread (a measure of precision) and peak rate (a measure of motivation), as previously described (Cheng & Meck 2007; Church et al. 1994; Matell et al. 2006). Peak time divided by peak spread at the 50th percentile was used as a measure of the relative standard deviation or sensitivity to time—also referred to as the WF or coefficient of variation (CV) (Church et al. 1994; Gibbon et al. 1997). The scalar property of interval timing predicts a constant WF across multiple target durations.
A rate index representative of the mean S1 and S2 response thresholds was also determined for FI and PI response functions averaged over blocks of sessions. This rate index was calculated by taking the response rate in a specified interval (20% of the target duration) just prior to or after the observed peak time as a ratio of overall response rate within the first (S1) or second (S2) half of the trial as defined by the target duration. Higher values of S1 and S2 indicate sharper FI or PI timing functions and better duration discrimination (cf. Agostino et al. 2011a,b; Cheng & Meck 2007). The alpha level was set at P < 0.05 for all statistical analyses.
Experiment 1: sequential acquisition of S1 and S2 response thresholds with 15-second and 45-second target durations
Carf gene deletions do not impair acquisition of lever pressing
All mice acquired reliable lever pressing within 12 sessions, and there were no between-group differences in the number of lever presses at the end of pretraining, F2,17 = 0.64, P > 0.05—followed by Fisher's least significant difference (LSD) tests, ns. There were, however, modest differences in the speed with which lever pressing was acquired. All Carf−/− and half of the Carf+/− groups each required an average of 10 sessions to reach the criterion level of lever pressing, whereas the remaining Carf+/− mice and all of the Carf+/+ groups required 12 sessions.
Carf−/− and Carf-+/− mice have a deficit in the acquisition of S1 response thresholds during FI training
Acquisition of temporal control during FI training was evaluated using the S1 rate index in all Carf groups for 15-second and 45-second target durations. Mean proportions of maximal response rate as a function of signal duration for FI sessions 1–3, 4–6, 7–9 and 10–12 (three-session blocks) are illustrated in Fig. 1a,b for 15-second and 45-second target durations, respectively. All mice acquired temporal control of responding as a function of the elapsing signal duration across the 15 sessions of FI training. However, KO and Het mice were impaired in their ability to acquire S1 response thresholds relative to WT mice. Data for the first block of three sessions were omitted from further analysis because of the low response rates occurring in those sessions following the shift from pretraining (in which every lever press was reinforced) to FI discrete-trial training in which only the first lever press after the target duration was reinforced on each trial. For the acquisition of temporal control of responding, there were significant effects of genotype, F2,17 = 3.67, P < 0.05; target duration, F1,17 = 15.01, P < 0.01; session block, F2,34 = 9.83, P < 0.001 and the target duration × session block interaction, F2,34 = 6.83, P < 0.01. The S1 rate index is plotted as a function of session blocks 2–4 for 15-second FI and 45-second FI training in Fig. 1c. Post hoc Fisher's LSD tests comparing the S1 rate index during session blocks 2 and 3 indicated significant differences between KO and WT mice for 15-second FI and 45-second FI training, P's < 0.05—with this effect being most evident during session block 2 for the 45-second FI condition as illustrated in Fig. 1d. In contrast, post hoc Fisher's LSD tests showed non-significant effects of genotype for session block 4 as mice neared asymptotic levels of responding, P's > 0.05. Overall, across all session blocks, there was a trend for the S1 rate index to be higher for WT mice (Carf+/+) and lower in the KO mice (Carf−/−), placing the Het mice (Carf+/−) in between as illustrated in Fig. 1c.
Carf−/− mice have a deficit in the acquisition of S2 response thresholds during PI training
The mean proportions of maximal response rate plotted as a function of signal duration (seconds) during PI sessions 3–4, 6–7 and 15–16 for the 15-second PI condition and PI sessions 6–7 for the 45-second PI conditions are presented in Figs. 2 and 3, respectively. Acquisition of temporal control during PI training was evaluated using the S2 rate index as a function of Carf genotype for 15-second and 45-second target durations. For both the 15-second and 45-second target durations, WT (Carf+/+) mice rapidly acquired a S2 response, producing a Gaussian-shaped response function centered at the expected times of reinforcement by sessions 15–16. In contrast, Het (Carf+/−) and KO (Carf−/−) mice were impaired in their acquisition of the S2 response for both the 15-second and 45-second target durations as shown by significant effects of genotype, F2,17 = 4.29, P < 0.05; session block, F2,34 = 42.16, P < 0.001 and the genotype × session block interaction, F4,34 = 5.73, P < 0.05 for the 15-second target duration and significant effects of genotype, F2,17 = 3.65, P < 0.05; session block, F2,34 = 47.04, P < 0.001 and the genotype × session block interaction, F4,34 = 3.03, P < 0.05 for the 45-second target duration. These effects were most evident during the intermediate sessions of PI training (sessions 6–7) as illustrated in Figs. 2d and 3 for 15-second PI and 45-second PI training, respectively—post hoc Fisher's LSD tests, P's < 0.05. Although there was a significant effect of target duration on peak rate with the 15-second PI function having a mean peak rate of 58.11 ± 8.06 resp/min and the 45-second PI function having a mean peak rate of 44.06 ± 9.56 resp/min, F1,17 = 11.31, P < 0.01, there were no significant differences in peak time or peak rate as a function of genotype, session block or their interaction, P's > 0.05.
Carf gene deletions do not violate the scalar property of interval timing at steady-state performance
The mean CV averaged over all mice was 0.48 ± 0.03 for the 15-second PI functions and 0.52 ± 0.04 for the 45-second PI functions during PI sessions 16–20. An analysis of variance (anova) indicated that CVs did not differ significantly either within or between treatment conditions, F2,17 = 1.04, P > 0.05—indicating that the temporal control of behavior conformed to the scalar property of interval timing.
Experiment 2: simultaneous acquisition of S1 and S2 response thresholds with a 30-second target duration
Carf gene deletions do not impair acquisition of lever pressing
All mice acquired reliable lever pressing within 12 sessions, and there were no between-group differences in the number of lever presses at the end of pretraining, F2,17 = 0.64, P > 0.05—followed by Fisher's LSD tests, ns.
Carf−/− and Carf-+/− mice have a deficit in the acquisition of S1 and S2 response thresholds during PI training
The mean proportions of maximal response rate plotted as a function of signal duration (seconds) for Carf mice during PI sessions 1–5 and PI sessions 21–25 for the 30-second PI condition are presented in the upper left and right panels of Fig. 4, respectively. The acquisition of temporal control during PI sessions 1–5 was evaluated using S1 and S2 rate indexes as shown in Fig. 5. A repeated-measures anova conducted on the S1 and S2 rate indexes during PI sessions 1–5 using Carf genotype and response threshold (S1 and S2) as main factors confirmed significant effects of genotype and response threshold on the rate index measure, F2,17 = 3.58, P < 0.05 and F2,17 < 4.63, P < 0.05, respectively. The genotype × response threshold interaction was non-significant, F2,17 = 0.38, P > 0.05. Post hoc Fisher's LSD tests showed that the S1 rate index was significantly higher than the S2 rate index for all groups of mice, P's < 0.05, and that both the S1 and S2 rate indexes for Carf+/+ mice were significantly higher, P's < 0.05, than those for Carf+/− and Carf−/− mice—which did not differ from each other, P > 0.05. The difference between S1 and S2 response thresholds continued to be significant during PI sessions 21–25, F2,17 = 6.24, P < 0.05, but there were no longer any significant differences in the S1 and S2 rate indexes as a function of Carf genotype, F2,17 = 0.64, P > 0.05 nor was the genotype × threshold interaction significant, F2,17 = 0.31, P > 0.05. In terms of the mean peak functions, there were no significant differences in peak time (31.45 ± 5.04, 33.29 ± 6.72 and 32.58 ± 6.18 seconds) or peak rate (61.02 ± 8.12, 58.29 ± 10.45 and 65.02 ± 9.33 resp/min) as a function of the Carf+/+, +/−, −/− genotypes, respectively.
Mice given BDNF-OCN microinjection into prefrontal cortex have a deficit in the acquisition of S1 and S2 response thresholds during sessions 1–5 of PI training
The mean proportions of maximal response rate plotted as a function of signal duration (seconds) for BDNF-ODN and CON mice during PI sessions 1–5 and PI sessions 21–25 for the 30-second PI condition are presented in the lower left and right panels of Fig. 4, respectively. The acquisition of temporal control during PI sessions 1–5 was evaluated using S1 and S2 rate indexes as shown in Fig. 5. A repeated-measures anova conducted on the S1 and S2 rate indexes during PI sessions 1–5 using injection type (BDNF-ODN or vehicle) and response threshold (S1 or S2) as main factors confirmed significant effects of both factors on the rate index measure, F1,18 = 18.64, P < 0.001 and F1,18 = 37.94, P < 0.0001, respectively. The injection type × response threshold interaction was non-significant, F1,18 = 0,19, P > 0.05. Post hoc Fisher's LSD tests showed that the S1 rate index was significantly higher than the S2 rate index for both groups of mice, P's < 0.05, and that both the S1 and S2 rate indexes for the CON group were significantly higher, P's < 0.05, than those for the BDNF-ODN group, P > 0.05. The difference between S1 and S2 response thresholds continued to be significant during PI sessions 21–25, F1,18 = 5.86, P < 0.05, but there were no longer any significant differences in the S1 and S2 rate indexes as a function of injection type, F1,18 = 1,38, P > 0.05. The injection type × response threshold interaction was also non-significant, F1,18 = 0.15, P > 0.05. In terms of the mean peak functions, there were no significant differences in peak time (33.35 ± 6.1, 30.6 ± 5.4 and 31.45 ± 6.3 seconds) or peak rate (65.48 ± 6.32, 59.74 ± 9.36 and 63.87 ± 8.47 resp/min) as a function of the injection types, respectively.
Histological assessment showed that all cannulae were located close to or within the prefrontal region of the cortex. Injectors were clearly visible as glial scar tracts terminating within the prefrontal cortex. Moreover, behavioral training of the CON and BDNF-ODN mice following the discontinuation of microinjections showed that BDNF antisense ODN did not cause permanent damage to the frontal cortex that may have interfered with timing performance.
Overall, our results indicate that Carf+/− and Carf−/− mice are selectively impaired in the acquisition of both S1 and S2 response thresholds when reproducing target durations in the multiseconds range. In contrast to this deficit in the temporal regulation of response thresholds, no reliable differences were observed in the acquisition of lever pressing, peak response rates or in the accuracy of the reproduced target durations as a function of genotype, arguing for similar motivational and memory function among groups. The observation that Carf+/− and Carf−/− mice exhibited similar levels of impairment for threshold setting in Experiment 1 (sequential acquisition of S1 and S2 response thresholds for multiple target durations) and Experiment 2 (simultaneous acquisition of S1 and S2 response thresholds for a single target duration) suggests a common source of influence. The acquisition of timed performance typically reflects a reduction in non-scalar sources of variability and the proportional adjustment of response thresholds as a function of training and target duration (e.g. Cheng & Meck 2007; Church et al. 1991, 1994; Gallistel et al. 2004; Gibbon 1977; MacDonald & Meck 2004; Meck & Church 1984). Such proportional adjustment of S1 and S2 response thresholds is a major hallmark of scalar timing and reflects the property of time-scale invariance (e.g. Buhusi & Oprisan 2013; Buhusi et al. 2009; Gallistel et al. 2004; Gibbon 1977; Gibbon & Church 1992; Gibbon et al. 1984). The observation that all groups of Carf mice showed similar levels of variability at steady-state performance (as measured by the CV) suggests an impairment in the acquisition/updating of response thresholds and not an alteration in the basic oscillatory mechanisms supporting scalar timing in memory (Buhusi & Meck 2005; Gibbon et al. 1984; Matell & Meck 2004; Merchant et al. 2013).
It has been shown that rats and mice learn to stop responding during unreinforced probe trials over the course of PI training rather than acquiring a S2 response threshold during earlier FI training (Balci et al. 2009b; MacDonald et al. 2012—see also Meck & Church 1984). This asymmetry in the acquisition of S1 and S2 response thresholds is nicely illustrated in this study by contrasting the results from Experiments 1 and 2—which indicate that the lack of reinforcement on probe trials (i.e. partial extinction) is required to learn when to stop responding in the PI procedure. Moreover, it has recently been hypothesized that differences in S1 and S2 response thresholds reflect separate striatal mechanisms for initiating and terminating responding, respectively (Cheng & Meck 2007; Coull et al. 2011). Using intracerebral infusions of the protein synthesis inhibitor anisomycin, MacDonald et al. (2012) showed that the acquisition of S1 response threshold depends, in part, on protein synthesis in the dorsal striatum, but not the ventral striatum. Conversely, disruption of protein synthesis in the VS, but not the DS, impairs acquisition of the S2 response. The conclusion is that the dorsal and ventral regions of the striatum function as a competitive neural network that encodes the temporal boundaries marking the beginning and end of a timed response sequence—see MacDonald & Monchi (2011). Taken together, our data suggest that the inhibitory processes required to apply response thresholds and ‘fine tune’ the temporal control of behavior are, at least in part, dependent on CaRF, likely through the activity-dependent regulation of protein synthesis in cortico-striatal circuits known to be critical for this type of timed performance (Buhusi & Meck 2005; Hong et al. 2005; Höhn et al. 2011; MacDonald & Meck 2005; MacDonald et al. 2012; Matell & Meck 2004; McDowell et al. 2010; Meck et al. 2008; Merchant et al. 2013; West 2011; West & Greenberg 2011; West et al. 2002).
Impairments in the acquisition of S1 and S2 response thresholds of a magnitude similar to Carf mice were also observed for mice given BDNF-ODN microinjections into the prefrontal cortex during acquisition of the 30-second PI procedure. As a consequence, our findings allow us to hypothesize about a potential role for CaRF-dependent BDNF activity in cortico-striatal circuits involved in interval timing, given that BDNF is an important modulator of DA levels in striatal MSNs—both of which have been shown to be involved in timing in the seconds-to-minutes range (Coull et al. 2011; Meck 2006a,b). The BDNF also regulates glutamatergic neurotransmission in the prefrontal-cortex accumbens pathway (Agostino et al. 2011a,b; Berglind et al. 2009). Animal models showing reduced BDNF expression in cortical and subcortical afferents, e.g. DA transporter KO mice (Fumagalli et al. 2003), transgenic mouse models of Huntington's disease (Samadi et al. 2013) or infusion of antisense BDNF blocking oligonucleotides and/or anisomycin into specific brain regions to inhibit protein synthesis (Lee et al. 2004; MacDonald et al. 2012), have been shown to impair memory consolidation in a manner consistent with the findings reported here for the acquisition of S1 and S2 response thresholds in PI timing procedures (see MacDonald et al. 2012). These findings support a role of BDNF in the protein-synthesis-dependent phase of learning involving intermediate-term memory (e.g. 24 h) as opposed to short-term memory (e.g. 3 h) as described by Lee et al. (2004). As a consequence, BDNF released into the striatum from cortical afferents may be involved in the signal transduction pathway that leads to the regulation of timed response sequences (Abidin et al. 2008; Goggi et al. 2002; Kim et al. 2013; Yoshida et al. 2003).
To address the possibility that dysregulation of BDNF expression in cortical afferents may impair interval timing, we studied the temporal control of behavior in Carf mice that have deficient expression of cortical BDNF owing to deletion of the gene encoding the transcription factor CaRF. The CaRF is a neuronal, calcium-regulated, transcriptional activator of Bdnf (Shieh et al. 1998; Tao et al. 2002). The Carf KO mice are born at the expected Mendelian ratios and exhibit no overt physical or behavioral phenotypes that would limit their performance on behavioral tasks (McDowell et al. 2010). In this study, Carf+/+, Carf+/− and Carf−/− mice acquired lever pressing for food reinforcement in a normal fashion using standard conditioning procedures. In contrast, Carf+/− and Carf−/− mice showed impairments in their ability to set S1 and S2 response thresholds during the acquisition of FI and PI timing tasks. These effects on the acquisition, but not asymptotic timed performance, suggest that the lack of this transcription factor leads to deficiencies in cortico-striatal function, which may be related to the reduced expression of Bdnf mRNA transcripts and BDNF protein in the cerebral cortex of Carf−/− mice. Given that patterns of cortical oscillations are detected by MSNs in the dorsal striatum, alteration in cortical projections may be responsible for the observed impairments in the response thresholding of target durations (Coull et al. 2011; Matell & Meck 2004). Alternatively, because CaRF is a transcription factor that has other target genes in addition to BDNF, a combination of molecular mechanisms may be responsible for the observed impairments/enhancements in the temporal control of behavior as a function of aging and other factors (Boger et al. 2011; Cheng et al. 2006b, 2008; Meck & Williams 1997a,b, 2003). Prenatal or postnatal choline supplementation, for example, increases BDNF levels in the brains of adult rats and mice (Glenn et al. 2007; Nag et al. 2008) and reduces non-scalar sources of variance in the setting of response thresholds for rats trained in the PI procedure (Cheng & Meck 2007).
Importantly, our previous studies have shown that Carf−/− mice exhibit a series of deficits that have been linked to BDNF expression in cortical projection neurons (McDowell et al. 2010). For example, although Carf−/− mice show normal spatial learning in the Morris water maze and normal context-dependent fear conditioning, they have an enhanced ability to find a new platform location on the first day of reversal training in the water maze and they extinguish conditioned fear more slowly than their WT Carf+/+ littermates, both of which are influenced by changes in cortical BDNF levels (Gorski et al. 2003). In addition, Carf−/− mice show normal short-term and long-term memory in a novel object recognition task, but exhibit impairments during the remote memory phase of testing, which is considered to be a reliable test of cortical function (Frankland & Bontempi 2005).
In summary, the present results indicate impairments in the neural processes involved in setting both S1 and S2 response thresholds for duration discriminations in the multiple-seconds range in both Carf+/− and Carf−/− mice as well as control mice given BDNF-ODN microinjections into the prefrontal cortex. These results suggest the relevance of transcription factors, calcium signaling and BDNF in the regulation of start/stop signals in cortico-striatal circuits during the acquisition of temporally controlled response sequences (cf. Jin & Costa 2010). Further investigation using a broader array of behavioral and molecular techniques will provide more details of the mechanisms involved in the temporal control of behavior and determine the degree to which these effects differ between constitutive and conditional genetic models (see Agostino et al. 2011a,b; Balci et al. 2008, 2010; Cheng et al. 2011; Drew et al. 2007; Farrell 2011; Gourley et al. 2009; Harrison et al. 2012; Meck et al. 2012; Tucci 2012).