Variable behavior and repeated learning in two mouse strains: Developmental and genetic contributions

Highlights

• Two mouse strains were compared on tasks relevant to behavioral rigidity.

• Variability measures differed by on strain, contingency, and age.

• BALB/c mice were more sensitive to variability contingencies than C57Bl/6 mice.

• BALB/c mice acquired more complex behavior chains than C57Bl/6 mice.

• Degree of behavioral variability in adolescence did not affect adult learning.

Abstract

Behavioral inflexibility is often assessed using reversal learning tasks, which require a relatively low degree of response variability. No studies have assessed sensitivity to reinforcement contingencies that specifically select highly variable response patterns in mice, let alone in models of neurodevelopmental disorders involving limited response variation. Operant variability and incremental repeated acquisition (IRA) were used to assess unique aspects of behavioral variability of two mouse strains: BALB/c, a model of some deficits in ASD, and C57Bl/6. On the operant variability task, BALB/c mice responded more repetitively during adolescence than C57Bl/6 mice when reinforcement did not require variability but responded more variably when reinforcement required variability. During IRA testing in adulthood, both strains acquired an unchanging, performance sequence equally well. Strain differences emerged, however, after novel learning sequences began alternating with the performance sequence: BALB/c mice substantially outperformed C57Bl/6 mice. Using litter-mate controls, it was found that adolescent experience with variability did not affect either learning or performance on the IRA task in adulthood. These findings constrain the use of BALB/c mice as a model of ASD, but once again reveal this strain is highly sensitive to reinforcement contingencies and they are fast and robust learners.

Introduction

Response variability can range from repetitive and rigid (low variability) to unpredictable and stochastic (high variability) (Neuringer and Jensen, 2012). While often thought to be the result of insufficient experimental control, variability is an operant dimension of behavior that is sensitive to reinforcement contingencies much like force, topography, or location and its presence is imperative for the acquisition of behavior. Behavioral rigidity is can be viewed as a trans-disease process (Bickel and Mueller, 2009) associated with autism spectrum disorders (ASD), damage to the prefrontal cortex, and schizophrenia, among other syndromes. ASD represent especially interesting examples because although their etiology remains unknown, some degree of abnormally rigid behavior, or a behaviorally rigid endophenotype, unifies disorders along this spectrum (Chadman et al., 2009; Devlin and Scherer, 2012; Ey et al., 2011; Lam et al., 2008). Non-human research in this area has focused on assessing whether genetic manipulations produce a behaviorally rigid phenotype consistent with ASD (Bechard and Lewis, 2012; Crawley, 2007; Ehninger et al., 2008; Ey et al., 2011; Silverman et al., 2010; Wöhr and Scattoni, 2013), but identification of animal models of behavioral rigidity as it pertains to ASD has proven particularly difficult (Bechard and Lewis, 2012; Chadman et al., 2009; Crabbe et al., 1999; Crawley, 2008, 2004; Crawley et al., 1997; Moy et al., 2006; Moy and Nadler, 2008; Roullet and Lassalle, 1995).

Behavioral assays used to select genetic models often include tasks that assess deficits in two fundamental aspects of behavioral variability. First, observations of unstructured behavior in a home cage (DeVisser et al., 2006; Pearson et al., 2011; Powell et al., 1999; Spruijt and DeVisser, 2006), an open field (Moy et al., 2007; Roullet and Crawley, 2011; Wöhr and Scattoni, 2013), or an un-baited maze (Hölter et al., 2015; Kas et al., 2014; Moy et al., 2007; Tanimura et al., 2008), have been used to assess the degree of behavioral variability that occurs in the absence of any constraints on or support for response variability imposed by reinforcement contingencies. Second, behavioral flexibility necessary for acquisition is often assessed with spatial reversal learning in open field arenas (Amodeo et al., 2012; Bechard and Lewis, 2012; Cheh et al., 2006; Colacicco et al., 2002), T-mazes (Lewis et al., 2007; Moy et al., 2007, 2006; Roullet and Crawley, 2011; Tanimura et al., 2008), Y-mazes (Pletnikov et al., 2002), radial arm mazes (Rendall et al., 2016), Morris water mazes (Cheh et al., 2006; Moy et al., 2007; Nakatani et al., 2009; Rendall et al., 2016; Roullet and Crawley, 2011; Tanimura et al., 2008), and Barnes mazes (Nakatani et al., 2009). In these tasks, navigation to a particular spatial location is positively or negatively reinforced and upon mastery of this contingency the spatial location that produces reinforcement is reversed. How readily animals navigate to a newly-reinforced spatial location following the reversal is a measure of behavioral flexibility. Behavioral rigidity is indicated by slow acquisition and flexibility by rapid acquisition.

Mastery of reversal tasks inherently requires a minimal degree of variability. At the beginning of a reversal, extinction of the previously reinforced response induces an increase in spatial variability, but extinction-induced variability is modest and fleeting (Antonitis, 1951; Eckerman and Lanson, 1969). As the animal contacts the new contingency, navigating to the now reinforced location, navigation variability again decreases. Continued variation of response location is incompatible with mastery of the reversal contingency, often defined as near exclusive responding in the post-reversal location. Here, variability is a byproduct of reinforcement contingencies and not the essential component of the behavior under investigation. Reversal tasks cannot definitively measure rigidity of operant behavior because they do not assess the extent to which reinforcement can select highly variable behavior; an important distinction. A preparation is required that generates variability explicitly in a robust, reproducible way and in such a manner that it can be increased or decreased by simple experimental manipulations. This distinction between variability as a byproduct of reinforcement contingencies and as an essential component of the behavior is an important one.

Operant variability interventions increase the variability of verbal and non-verbal responses in humans with ASD (Esch et al., 2009; Miller and Neuringer, 2000) although the degree of variability observed following training does not approach that observed in matched controls (Miller and Neuringer, 2000). Such behavioral interventions have been increasingly implemented in clinical settings to improve responding and promote acquisition of new behavior for individuals with ASD (Esch et al., 2009; Koehler-Platten et al., 2013; Rodriguez and Thompson, 2015).

A third type of behavioral rigidity, sensitivity to reinforcement of response variations, should be considered when assessing and selecting animal models of the behavioral rigidity endophenotype. A failure to assess the upper-limit of behavioral variability could result in false-positive conclusions regarding animal models of ASD. Ideally, animal models of this endophenotype should show deficits in three fundamental aspects of behavioral variability relative to comparison strains 1) decreased spontaneous variations, 2) deficits acquiring a novel operant, and 3) insensitivity to reinforced variations. Operant variability is particularly appealing in regard to this third aspect of the behavioral rigidity endophenotype.

The BALB/c strain models social deficits (Brodkin, 2007; Ferrari et al., 2005; Popova, 2006), increased anxiety (Brodkin, 2007; Lalonde and Strazielle, 2008; Moy et al., 2007), decreased spontaneous variations (Kalueff and Tuohimaa, 2005; Lalonde and Strazielle, 2008; Moy et al., 2008), and some neurobiological abnormalities (Chandana et al., 2005; Chugani et al., 1999) associated with ASD, but there is conflicting evidence as to whether this strain also models the behavioral rigidity endophenotype that impairs the acquisition of a novel operant. Some have reported impaired reversal learning in the BALB/c strain relative to the C57Bl/6 strain (Crawley et al., 1997; Roullet et al., 1993; Roullet and Lassalle, 1995), while others have found the opposite (Moy et al., 2007). Tasks such as incremental repeated acquisition (IRA), which provide a more in-depth analysis of the second aspect of the behavioral rigidity endophenotype, could be useful for clarifying conflicting evidence. IRA is particularly appealing because it assesses how readily animals acquire familiar and novel response chains that are increasingly complex.

Currently, there are no published studies characterizing sensitivity to reinforcement of response variations in the BALB/c, C57Bl/6, or any other mouse strain. Doing so is important because the C57Bl/6 strain is commonly used as a comparison strain in ASD research (Moy et al., 2008, 2007) and is a common background strain for many genetic models (Crawley et al., 1997). Here, strain differences were assessed on several behavioral tasks: autoshaping, operant variability threshold procedure, and IRA. Initial testing on autoshaping and operant variability threshold procedures occurred in adolescence. Once mice reached adulthood, they were tested on IRA and re-tested on the same operant variability threshold procedure (see Fig. 1 for timeline of experimental events).

This approach allowed for the assessment of three primary goals. The first was to determine whether the BALB/c strain modeled invariant responding and reduced sensitivity to reinforcement of highly variable behavior, relative to the comparison C57Bl/6 strain. The second was to identify age-dependent changes in response variations (Mason, 1993). Adolescence is an important period of neurodevelopment, in particular the dopaminergic system (Spear, 2000), which has been shown to play an important role in learning and reinforcement (Berridge and Robinson, 1998). The third was to determine if establishing a history of invariant (ANY) or highly variable (VAR) responding during adolescence promoted acquisition and/or learning increasingly complex response chains on IRA. Previous research using an operant variability task similar to the one used here, demonstrated reinforcement of highly variable, but not moderately variable or invariant responding facilitated the concurrent acquisition of a difficult-to-learn response sequence (Grunow and Neuringer, 2002).