The Impact of Hippocampal Lesions on Trace Eyeblink Conditioning and Forebrain-Cerebellar Interactions

Abstract

Twenty-five years ago Behavioral Neuroscience published a pivotal paper by Moyer, Deyo and Disterhoft (1990) that described the impaired acquisition of trace eyeblink conditioning in rabbits with complete removal of the hippocampus. As part of the Behavioral Neuroscience celebration commemorating the 30th anniversary of the Journal, we reflect upon the impact of that study on understanding the role of the hippocampus, forebrain, and forebrain-cerebellar interactions that mediate acquisition and retention of trace conditioned responses, and of declarative memory more globally. We discuss the expansion of the conditioning paradigm to species other than the rabbit, the heterogeneity of responses among hippocampal neurons during trace conditioning, the responsivity of hippocampal neurons following consolidation of conditioning, the role of awareness in conditioning, how blink conditioning can be used as a translational tool by assaying potential therapeutics for cognitive enhancement, how trace and delay classical conditioning may be used to investigate neurological disorders including Alzheimer's Disease and schizophrenia, and how the two paradigms may be used to understand the relationship between declarative and nondeclarative memory systems.

Eyeblink Conditioning in the Rabbit

Eyeblink conditioning became one of the more widely used paradigms to study learning and memory during the 1970s as the rabbit model of the paradigm was developed and analyzed using behavioral psychophysics to optimize the paradigm (Schneiderman, Fuentes, Gormezano, 1962; Disterhoft, Kwan, Lo, 1977). A neurobiological analysis of the learned behavior began with an examination of the hippocampus because of the impact that hippocampal ablation had on acquisition of new declarative memories in humans that had temporal lobe resections for the treatment of intractable epilepsy, e.g. patient H.M. (Scoville & Milner, 1957). Consistent with the results of the lesion, neurophysiological recordings from rabbit hippocampus during delay eyeblink conditioning (where the unconditioned stimulus overlaps and coterminates with the conditioning stimulus) revealed learning-specific increases in the amplitude and time-course of extracellularly recorded activity from CA1 pyramidal neurons (Berger, Alger, Thompson, 1976), as expected for a structure mediating memory.

However, since others had shown that neither the dorsal hippocampus nor cortex were required for acquisition of simple delay conditioning (Oakley & Russell, 1972; Norman, Buchwald & Villablanca, 1977), the investigation of the essential pathway mediating blink conditioning extended beyond the hippocampus. The essential pathway was examined from the periphery into the central nervous system (CNS), i.e. from the retractor bulbi (RB) muscle (which pulls the eye back into the socket and causes extension of the nictitating membrane) to the accessory abducens nucleus (which has the motor neurons that innervates RB; Disterhoft, Quinn, Weiss, Shipley, 1985), and beyond (Gonzalez-Joekes & Schreurs, 2012). The essential pathway was also examined by recording activity from several sites within the central nervous system (CNS) while looking for neuronal activity that mirrored the amplitude and time course of the conditioned response (CR). The cerebellum was one site that revealed “neuronal models” of CRs (McCormick et al., 1981, 1982).

The findings that ablation of the lateral cerebellum abolished behavioral CRs and that activity in the cerebellar nucleus developed concomitantly with expression of CRs (McCormick et al., 1982) led to an intensive investigation during most of the 1980s on the role of the cerebellum in mediating delay EBC. However, the scientific wind turned back to the hippocampus in 1986 when three papers examined the role of the hippocampus in trace conditioning, where a stimulus free gap separates the two conditioning stimuli. The decade concluded with the submission of the Moyer et al. paper which was published in 1990.

The three papers examining the role of the hippocampus were by Port et al. (1986), Solomon et al. (1986), and James et al. (1987). All of them examined the effect of dorsal hippocampal lesions upon trace conditioning of the blink response in rabbit. The reports by Port and by James failed to show a significant deficit in conditioning. In fact, the learning curves and peak amplitudes of the CR presented by James et al. were almost identical between rabbits that had received the dorsal hippocampal lesion and those that had received a cortical control lesion. Their rabbits were conditioned with a 90 dB burst of white noise as the conditioning stimulus (CS) and an AC periorbital shock of 2.5 mA for 50 ms as the unconditioned stimulus (US). The study by Solomon et al. conditioned rabbits with a 1kHz, 85 dB tone and a 3 psi puff of air to the cornea after the rabbits received an aspiration lesion of the dorsal hippocampus. Their results showed a severe impairment for acquiring trace CRs. The different outcomes of the two studies may have been due to partial lesion effects, i.e. the ventral hippocampus was still intact (Weiss, Bouwmeester, et al. 1999), or a fortuitous consequence of the US modality, i.e. years later it was shown that lesions of the prefrontal cortex impaired trace conditioning when the US was an airpuff to the cornea, but not when the US was an electric shock to the periorbital region (Oswald, Maddox, Tisdale, Powell, 2010).

Given the different results of those three reports, Moyer et al. (1990) examined the effect of hippocampal lesions that included the dorsal and ventral regions on acquisition of trace CRs using a tone conditioning stimulus and an airpuff unconditioned stimulus. Importantly, they also compared 300 ms and 500 ms trace intervals upon acquisition rates, to determine if the imposed memory load affected hippocampal dependence. The results indicated that hippocampal dependence increased with memory load, i.e. lesions severely impaired acquisition of trace CRs when the trace interval was 500 ms, and had variable results when a 300 ms interval was used. The “trace 500” paradigm appears to have become the standard for examining hippocampal-dependent learning in the rabbit, and delay conditioning is often used after trace conditioning to demonstrate that the brainstem and cerebellum are intact and functioning (Weiss, Bouwmeester, Power & Disterhoft, 1999; Tseng, Guan, Disterhoft, Weiss, 2004) suggesting that higher brain centers, e.g. hippocampus, prefrontal cortex (PFC), and thalamus (Halverson, Poremba & Freeman, 2008; Steinmetz, Buss, Freeman, 2013) are critical regions for mediating the trace CR.

The modality of the US is less standard among laboratories. Some laboratories use a puff of air (or nitrogen) to the cornea, and some laboratories use a periorbital shock. The shock US appears to be more aversive and likely activates neuromodulators more so than does an airpuff. The neuromodulators may compensate for the loss of hippocampal-prefrontal cortical activation in lesion studies by either modulating the PFC directly, or modulating direct projections from the ventral hippocampus to PFC (Rosene & Van Hoesen, 1977; Witter et al., 1989). The loss of this modulation might explain why complete hippocampal lesions (those that include the ventral hippocampus) or lesions of the prefrontal cortex, particularly the caudal anterior cingulate cortex (cACC), prevent acquisition of trace EBC with an airpuff US (Kronforst-Collins & Disterhoft, 1998; Weible, McEchron, Disterhoft, 2000).

CA1 Responsivity During Delay and Trace Conditioning

Neuronal responses from CA1 hippocampus during delay conditioning are quite remarkable (Berger & Thompson, 1978; Berger, Rinaldi, Weisz & Thompson, 1983). Neurons from rabbits presented with pairs of tones and airpuffs exhibited increased activity that developed first in response to the airpuff and then gradually increased activity during the CS-US interval over the course of the session, as CRs were developing, and the increased activity preceded the behavioral response in time. This increase began within the first few trials of stimulus presentations. The correlation of neuronal activity and behavior changed with learning such that early in training neuronal activity correlated more with the unconditioned response, and later in training the activity correlated more with the conditioned response. These changes did not occur in rabbits given explicitly unpaired presentations of tones and puffs in random order, even on trials with a large unconditioned response, and not when a spontaneous blink was present. These data are strong evidence for a learning-specific change in the activity of CA1 neurons recorded from conditioned rabbits.

However, since the hippocampus is not required for delay conditioning, as previously mentioned, recordings were taken from CA1 during hippocampally-dependent trace conditioning in search for activity patterns that might reveal the nature of the role that the hippocampus plays in conditioning and memory consolidation (Solomon et al., 1986; Weiss, Kronforst-Collins & Disterhoft, 1996; Weible, O’Reilly, Weiss & Disterhoft, 2006). The results of Solomon et al. were based on multiunit data and suggested that there was more diversity in the response profiles than had been found with delay conditioning. The report by Weiss et al. analyzed single neuron data from rabbits that achieved at least 60% CRs, i.e. the rabbits had already passed the initial acquisition phase. Those data confirmed and catalogued the wide variety of responses that were found in CA1 during trace conditioning, suggesting that memory demanding trace conditioning requires the hippocampus to do critical computations that are not necessary for simple delay conditioning, and that inhibitory processes (reflected as a rate-decreasing response) are just as involved as excitatory processes (McKay, Oh, Disterhoft, 2013). The report by Weible et al. examined the hippocampus more thoroughly during acquisition. They recorded from the dorsal and ventral hippocampus of both sides of the brain (relative to the eye being stimulated). A variety of responses was again found, but this study also reported that the ipsilateral hippocampus exhibited significant increases in activity early in training while the contralateral hippocampus exhibited greater response magnitudes later in training. They also reported that the ventral hippocampus was less involved in terms of the percentage of responsive neurons (8% vs 69%), and more involved during the post-trial period than during the CS-US interval. This may reflect anatomical connections with the hypothalamic-pituitary-adrenocortical axis (Jacobson & Sapolsky, 1991; Herman et al., 1992) and the amygdala (Petrovich, Canteras & Swanson, 2001).

An examination of the diversity of responses was taken a step further by McEchron, Weible and Disterhoft (2001). They used a hierarchical clustering analysis to segregate neurons by differences in their response profiles and found six different categories. Interestingly, the two most frequent profiles were classified as rate-decreasing activity and accounted for 52% of the profiles. More recently we have begun to use support vector machine (SVM) and general linear modeling (GLM) approaches to classify neurons according to their response profiles and have found that the SVM approach can predict conditioned responses and their onset latency based on neuronal activity (Lawlor, Hattori, Weiss, Disterhoft & Kording, 2013). We expect that these statistical approaches will reveal interesting changes in network properties that were previously unrecognized.

Hippocampus and Memory Consolidation

The hippocampus is critically involved in acquisition of new memories, as was exemplified most frequently by the case of patient H.M. However, given that HM had intact episodic memories for events prior to his surgery, the hippocampus has been considered to act as a temporary memory buffer with permanent memory residing elsewhere, most likely in the cerebral cortex (Squire & Alvarez, 1995). The effect of this consolidation process was tested with trace EBC in the rabbit after Kim and Fanselow (1992) had demonstrated a time-limited role for the hippocampus in contextual fear conditioning. Kim, Clark, and Thompson (1995) lesioned the hippocampus either one day or one month after rabbits had attained a criterion level of performance on trace EBC. The rabbits lesioned one day after criterion were severely impaired when tested after a seven day recovery period (they did not exhibit more than about 20% of trials with conditioned responses), but rabbits that were not lesioned until one month after attaining criterion returned to pre-lesion performance levels after a slight decrease in performance on the first session following the one month hiatus. This is strong evidence that the hippocampus is not required for memory recall after consolidation has occurred, at least as measured by performance rates during trace EBC.

The biophysical basis for the hippocampus not being required following memory consolidation is likely due to the transient nature of the learning specific reduction in the amplitude of the post-burst afterhyperpolarization (AHP) of CA1 and CA3 pyramidal neurons (Moyer, Thompson, Disterhoft, 1996; Thompson, Moyer, Disterhoft, 1996a), i.e. the amplitude of the post-burst AHP returned to a baseline level within seven days after rabbits achieved the learning criterion of at least 80% CRs within one session. Apparently, this is a sufficient period of hippocampal activation to support the formation of consolidated memory representation in other regions such as the neocortex. The transient nature of hippocampal involvement in eyeblink conditioning has also been demonstrated in rats (Takehara, Kawahara, Kirino, 2003) and mice (Takehara, Kawahara, Takatsuki, Kirino, 2002) that received aspiration lesions of the hippocampus at various times after learning was achieved.

Although the hippocampus is not required to express CRs in the consolidated state, it may still be involved in expressing the CR. For example, even though HM’s semantic memory was spared, he exhibited severe impairments in remote autobiographical memory (Steinvorth, Levine & Corkin, 2005). This suggests that the hippocampus may be doing important processing during recall of consolidated memory. To address this question directly we adopted a paradigm of training our animals to criterion (plus 3–5 days) and then giving them a 30 day training hiatus before testing (and recording single neuron activity) for retention of remotely acquired memory. We examined this activity in the hippocampus after consolidation (Hattori, Chen, Weiss & Disterhoft, 2014) and also inferred hippocampal involvement by looking at the prefrontal cortex after consolidation (Hattori, Yoon, Disterhoft & Weiss, 2014). The results were dramatic and surprising, i.e., hippocampal pyramidal neurons were found to be even more responsive after the training hiatus than before the hiatus. This activity may reflect reactivation or reconsolidation processes. Strong reactivation would be needed if the memory had degraded over time (Rudy, Biedenkapp, O’Reilly, 2005). Furthermore, activity in the prelimbic cortex was also found to be significantly greater after consolidation than before consolidation and may reflect hippocampal-prefrontal interactions (Hattori, Yoon, et al., 2014) and entorhinal-prefrontal interactions (Takehara-Nishiuchi, Bared, Morrissey, 2012).

Eyeblink Conditioning in the Rat

The rabbit is an excellent animal model system for studying blink conditioning since it tolerates restraint well, has large eyes with few spontaneous eye movements, and it has a relatively large head that allows implants to be secured to the skull for studies using chronic recording and stimulation techniques. However, since other behavioral tasks are necessary to examine the more global nature of hippocampal-dependent memory, species that have more spontaneous behaviors than a rabbit would be useful. The rat is ideal for examining such behaviors, and since EBC, the hippocampus, and memory are very sensitive to the effects of age (Thompson, Moyer & Disterhoft, 1996b; Oh et al., 2010; Matthews et al., 2009; Gant & Thibault, 2009), we and others adapted the method of Skelton (1988) to condition rats while they were tethered and allowed to move about freely within a conditioning chamber. The freely moving rat model was used to examine the effects of aging (Weiss & Thompson, 1992; Knuttinen et al., 2001), the interaction of two hippocampally-dependent tasks (Kuo et al., 2006; Curlik et al., 2014), the effect of varying the trace interval (Weiss, Knuttinen et al., 1999; Bangasser et al., 2006; Walker & Steinmetz, 2008), and the effect of potential therapeutics to ameliorate age-related cognitive impairments (Burgdorf et al., 2011). Importantly, we tweaked the paradigm for trace conditioning in the rat (Weiss et al., 1999) and then showed that complete lesions of the hippocampus were necessary to prevent acquisition of trace EBC (Weiss, Bouwmeester, Power, & Disterhoft, 1999), just as Moyer et al. had found for the rabbit. Interestingly, the critical trace interval for hippocampal dependence was found to be as short as 250ms; an interval of 500ms did not support conditioning when the US was a puff of air to the cornea. In summary, the rat is an excellent experimental subject and when combined with protein knockdown techniques (short hairpin RNAs, small interfering RNAs, or microRNA) the rat model may be as good a model system as the mouse (see next section) to examine the molecular mechanisms of memory.

Eyeblink Conditioning in the Mouse

In order to understand molecular mechanisms mediating hippocampal function and conditioning, we and others adapted the freely moving rat methodology to the mouse so that we could use transgenic and knockout technologies to analyze the systems and cellular mechanism of learning and memory storage. This occurred at a time prior to the proliferation of gene interfering techniques. Our initial study examined one of the murine models for Alzheimer’s disease and the effect on hippocampal volume (Weiss et al., 2002). We found that the V717F transgene impaired acquisition and decreased hippocampal volume, as measured with magnetic resonance imaging. We also showed in C57Bl6 mice that ibotenic acid lesions of the hippocampus impaired acquisition in proportion to the volume of tissue destroyed, and that the critical trace interval was at least 250 ms, as also found for the rat (Tseng, Guan, Disterhoft & Weiss, 2004), i.e., the mouse would not acquire conditioned responses when the trace interval was 500 ms (as used with rabbits), but they were able to be conditioned with delay procedures. Others have used the mouse paradigm to examine deletion of hippocampal monoamine oxidase (Singh et al., 2013), involvement of CA3-CA1 synapses (Gruart, Munoz & Delgado-Garcia, 2006), the role of the CA3 NMDA receptors (Kishimoto et al., 2006), adaptive timing in cerebellum (Chettih et al., 2011), the role of nucleus reuniens as a common input to both hippocampus and medial prefrontal cortex (Eleore et al., 2011), and the role of cerebellar Purkinje neurons (Brown, Agelan & Woodruff-Pak, 2010; Koekkoek et al., 2005).

The tethered preparation was critical to using the mouse since our pilot work indicated that whole-body restraint was stressful for the mouse. However, Chettih, McDougle, Ruffolo, and Medina (2011) demonstrated good conditioning in the head-fixed mouse when the legs of the mouse were allowed to ambulate on top of a freely moving cylindrical treadmill. This simple modification appears to reduce the stress of restraint and allows conditioning to occur. The restrained mouse preparation designed by Medina and his colleagues is now being used in our laboratory, and can be used with genetically modified mice, especially mice engineered with the inducible Cre system that allows specific gene deletion, a technique more suited for mice at this time.

Eyeblink Conditioning in Humans

The declarative nature of memory can be inferred from experiments with non-human animal subjects, but it can only be examined directly in human subjects. The first report of eyeblink conditioning was done in humans by Cason (1922), two years before Pavlov’s lectures about conditioned reflexes to the Military Medical Academy in St Petersburg, Russia. Cason was attempting to dissociate the timing of voluntary responses from the timing of conditioned responses. His results indicated that the conditioned response occurred with a latency that was significantly shorter than the latency for the same person instructed to blink as quickly as possible when the sound was heard. This suggested that different circuitry was mediating the conditioned reflex, a result that we and others are still investigating to this day.

It was not until Clark and Squire (1998) interviewed their subjects about the stimulus contingencies in a conditioning session that the full potential of using human subjects was realized. They asked about the temporal relationship of the conditioning stimuli that was used in either delay or trace blink conditioning. Their results revealed that trace, but not delay conditioning required awareness of the temporal relationship between the conditioning stimulus and the unconditioned stimulus, i.e. amnesic patients acquired delay conditioning but not trace conditioning. Knuttinen et al. (2001) found similar effects due to age-related impairments in awareness. While there are some methodological concerns with the Clark and Squire study, i.e. the effects of aging and attention (LaBar & Disterhoft, 1998), the face validity of eyeblink conditioning as an assay for declarative memory and hippocampal function increased after seeing that the percent of trials with CRs and the participants’ correct predictions of when a US would be delivered. Their work also suggested that promoting awareness, by attempting to predict the onset of the US, facilitated trace conditioning (Manns, Clark, Squire, 2000), possibly due to activation of the hippocampus. In fact, greater hippocampal activation was found during trace eyeblink conditioning as compared to concurrently learned delay conditioning when assayed with functional magnetic resonance imaging (fMRI) in humans (Cheng et al., 2008). The mechanisms underlying facilitation by awareness may be related to the mechanisms mediating facilitated acquisition when trials are specifically presented only when rabbits are exhibiting hippocampal theta activity (Griffin et al., 2004; Asaka et al., 2005; Berry & Hoffmann 2011).

Trace Eyeblink Conditioning as a Tool for Translational Neuroscience

EBC as an Assay for Therapeutics

The hippocampus is essential for declarative memory and its function is impaired by aging and Alzheimer’s Disease. The hippocampal dependence of trace eyeblink conditioning, and its dependence on awareness of stimulus contingencies has made the paradigm useful for studying cognitive deficits associated with normal aging and age-related impairments, and the amelioration of those deficits by potential therapeutics. Several classes of therapeutics have been tested with blink conditioning: calcium channel blockers, inhibitors of acetylcholinesterase, allosteric potentiating ligands for nicotinic receptors, muscarinic agonists, and allosteric modulators of the NMDA receptor. Calcium channel blockers were examined since an imbalance in calcium homeostasis has been hypothesized to underlie age-related cognitive impairments (Landfield, 1987; Oh, Oliveira, Waters & Disterhoft, 2013). Nimodipine, a dihydropyridine, blocks L-type calcium channels and was found to facilitate learning in aging rabbits (Deyo, Straube, Disterhoft, 1989). Metrifonate, a cholinesterase inhibitor improves learning and retention in aging rabbits (Kronforst-Collins et al., 1997; Oh et al. 1999). Galantamine, a potent allosteric potentiating ligand of nicotinic acetylcholine receptors (and a weak competitive and reversible cholinesterase inhibitor) increases acetylcholine release, facilitated acquisition of trace eyeblink conditioning in aging rabbits, and increased the excitability of CA1 hippocampal pyramidal neurons (Weible, Oh, Lee, Disterhoft, 2004; Oh, Wu, Power, Disterhoft, 2006;). CI-1017, an agonist of the M1 muscarinic receptor also facilitated acquisition of trace EBC in aging rabbits and increased the excitability of CA1 hippocampal pyramidal neurons (Weiss, Preston, Oh, Schwarz, Welty & Disterhoft, 2000). The theoretical rationale for these studies was the proven effect of each of the compounds mentioned to increase excitability of CA1 hippocampal pyramidal neurons by reduction of the size of the slow, post-burst afterhyperpolarization, mediated by a calcium-activated outward potassium current which is increased in aging neurons, leading to age-associated impairments in cognition (Oh and Disterhoft, 2006). Finally, GLYX-13, an allosteric modulator of the NMDA receptor, was derived from a monoclonal antibody against NMDA receptors from the dentate gyrus (Moskal et al., 2005). Both the antibody (B6B21) and GLYX-13 were found to facilitate trace EBC in aging animals when administered into the ventricle or intravenously, respectively (Thompson, Moskal, Disterhoft, 1992; Moskal et al., 2005; Burgdorf et al., 2011).

EBC as an Assay for Neurological Disease

Aside from using blink conditioning to study learning and memory impairments due to normal aging and the effects of different therapeutics, the paradigm has been used to study disease related impairments in Alzheimer’s disease (AD) and schizophrenia. AD shares some overlap with the effects of age (since age is a major contributing factor for the disease), but Woodruff-Pak (1990, 2001, 2007) reports that individuals with AD are more impaired on EBC (especially delay EBC, due to degeneration of cerebellar Pukinje neurons) than are age-matched control subjects.

Factors other than aging that influence AD include high cholesterol and diabetes. Bernard Schreurs and the late David Sparks published several reports regarding the role of cholesterol, copper, and dietary treatments on AD-like pathology and behavioral conditioning in the rabbit (Sparks & Schreurs, 2003; Schreurs et al., 2007; Wang & Schreurs, 2010; Schreurs, 2013). The effects of copper and cholesterol are complicated and are affected by the timing of the treatment in relation to training and retention testing. However, further research with this dietary paradigm is important since rabbits fed a cholesterol enriched diet exhibit at least 12 pathological markers of AD (Sparks & Schreurs, 2003), and because rabbits have the same amino acid sequence for amyloid as do humans (Davidson et al., 1992). This makes the rabbit a non-transgenic model system for AD that excludes the side-effects of foreign DNA, as present in murine models of AD.

Spontaneous Alzheimer’s Disease, the most prevalent form of the disease, is considered by some to be due to a form of diabetes (“Type 3 Diabetes”) or an “insulin-resistant brain state” (de la Monte & Wands, 2008; Correia et al., 2011; Talbot et al., 2012). Several studies by William Klein and his colleagues support the hypothesis that defective insulin signaling is a major contributor to the pathology associated with AD (Zhao et al., 2009; De Felice et al.,2009; Bomfim et al., 2012), including the pathology found in rabbits that develop diabetes following an injection of Alloxan. Alloxan kills islet cells in the pancreas, induces a diabetic state due to the lack of insulin, and after an incubation period of about 15 weeks several pathological markers of AD are apparent including, a five-fold increase in Aβ40/Aβ42 in cortex and hippocampus, the generation of Aβ oligomers, and elevation of phospho-tau along neuronal tracts (Bitel et al., 2010, 2012). These pathologies in the rabbit suggest that hippocampal-dependent trace eyeblink conditioning should be a sensitive behavioral assay for the changes induced by a diabetic state in the rabbit.

Both trace and delay eyeblink conditioning are dependent on the cerebellum (Woodruff-Pak, Lavond & Thompson, 1985), but only trace conditioning is dependent on the hippocampus for acquisition of CRs. However, if the hippocampus malfunctions, as in AD or following an injection of scopolamine (a cholinergic antagonist), delay conditioning is also impaired, unless conditioning is done in an experimental animal in which the hippocampus was removed (Solomon et al., 1983). These data suggest that cerebellar-hippocampal interactions (and cerebellar-prefrontal interactions) are also important for delay conditioning and probably other cognitive tasks.

Schizophrenia is a disorder that Nancy Andreasen and colleagues termed a “cognitive dysmetria” (Andreasen, Paradiso & O’Leary, 1998) and is thought to be due to disturbances in a cortico-cerebellar-thalamic-cortical circuit (Sears, Andreasen & O’Leary, 2000). This circuit is expected to be active during EBC and could affect both trace and delay conditioning, if a malfunctioning circuit impairs delay conditioning as does a malfunctioning hippocampus. Brown et al. (2005) and Bolbecker et al. (2009) each examined the effect of schizophrenia in delay EBC and found that affected individuals exhibited lower rates of conditioning due to an increase in maladaptive short latency responses, an effect also seen following the hippocampal lesions of Moyer et al.

Hierarchical Memory Systems

Squire and Zola-Morgan (1991) proposed a hierarchical scheme for organizing memory into declarative (explicit) and nondeclarative (implicit) components. They included simple (delay) classical conditioning at the bottom of the tree, as part of the nondeclarative system and had it nestled in between the functional components of nonassociative learning and priming; the more complex trace conditioning paradigm was not included in their scheme. We propose instead that classical conditioning be placed in a more central position, next to procedural learning, and to bridge the divide between declarative and nondeclarative memory (Fig. 1). This scheme can be used to frame hippocampal and prefrontal interactions with the cerebellum as proposed by Weiss and Disterhoft (1996) and consistent with results using electrical stimulation of the prefrontal cortex (Kalmbach, Ohyama and Mauk, 2010) and recording data from the prelimbic cortex (Hattori, Yoon et al., 2014). The basic premise is that the hippocampus and prefrontal cortex have an innate property to form an initial association between the CS and US when they are presented with a trace interval separating them. The projection from prefrontal cortex to pons then facilitates the effectiveness of the CS into the cerebellum to facilitate conditioning that would otherwise be outside of the optimal timing for long term depression (LTD) to occur at the parallel fiber-Purkinje cell synapse. The cerebellar network eventually projects to the red nucleus and subsequently motor neurons mediating the blink while the hippocampal-reuniens-prefrontal network (Cassel et al., 2013; Hallock, Wang, Shaw & Griffin, 2013; Varela, Kumar, Yang, Wilson, 2014) activates the prelimbic cortex where working memory, awareness, and the declarative nature of the experience may be instantiated, just as neuronal activity in the dorsolateral prefrontal cortex of monkeys appears to reflect the memory for a cue that has been presented and removed during delayed matching to sample tasks (Bauer & Fuster, 1976; Quintana, Yajeya & Fuster, 1988; Fuster 2001).

Figure 1.

Domains of memory. This adaptation of a figure from Larry Squire and Stuart Zola-Morgan (1991) shows that trace conditioning can be positioned to act as a Rosetta stone (Thompson, 1988) for understanding hippocampal-cerebellar interactions that mediate conditioned adaptations of somatomotor responses.

Lesion/Inactivation Studies

The preceding hierarchy is based largely on functional considerations. A more mechanistic view requires anatomical localization, and cell type specificity would greatly add to an understanding of the signal processing that occurs along the pathways. Anatomical localization is often determined by removing a structure from the circuit. In regards to the circuitry mediating blink conditioning, the deep cerebellar nuclei appear to be essential for mediating both delay and trace conditioning (Woodruff-Pak & Disterhoft, 2008 for review). However, the complete essential forebrain circuitry that contributes to hippocampal-cerebellar interactions mediating trace conditioning is still unknown and is actively being investigated by us and others (Siegel et al., 2012). Aspiration, electrolytic, and neurotoxic lesions have been classical ways to remove a structure and determine the necessity of a region. Ibotenic acid lesions of primary somatosensory cortex were found to prevent and abolish whisker-signaled trace eyeblink conditioning (Galvez, Weible, Disterhoft, 2007), muscimol induced inactivation of secondary somatosensory cortex was found to abolish conditioning in rabbits that had consolidated their conditioning for one month (Ward, Flores, Weiss, Disterhoft, 2011), electrolytic lesions of the caudate were found to prevent acquisition and impaired improvements in responding among rabbits that had been allowed to consolidate their learning for a period of one month (Flores & Disterhoft, 2009, 2013), aspiration lesions of the caudal anterior cingulate region prevented acquisition of trace conditioning (Weible, McEchron, Disterhoft, 2000), and electrolytic lesions of the perirhinal or postrhinal corticies (but not lateral entorhinal cortex) significantly impaired acquisition of trace blink conditioning in rats (Suter, Weiss & Disterhoft, 2013). Inactivation of accessory thalamic nuclei were found to impair acquisition and retention of delay blink conditioning to either a visual or auditory conditioning stimulus and would be expected to have similar effects on trace conditioning (Halverson & Freeman, 2010a,b; Halverson, Lee, Freeman, 2010; Steinmetz, Buss, Freeman, 2013). More modern techniques include the cre-lox system (Nagy, 2000) which allows conditional gene deletion in mice, chemogenetic inactivation which uses a peripheral injection to activate a virus expressing designer receptors exclusively activated by designer drugs (DREADDs, Alexander et al., 2009), and optogenetic techniques (Lee et al., 2012) which allow precise temporal control over inactivation (which can prevent complications due to functional reorganization following prolonged inactivation, Goshen et al., 2011). We are currently investigating the effects of chemogenetic and optogenetic inactivation in rabbit and will be examining cell type specific inactivation in the mouse.

Eyeblink Conditioning: Past its prime, or primed for the future?

Some may wonder if eyeblink conditioning has passed its prime after nearly four decades of research. We argue instead that the paradigm is primed for the future, especially in regards to translational neuroscience research and treatments for normal age-related memory impairments (Cheng, Faulkner, Disterhoft & Desmond, 2010; Burgdorf et al., 2011; Bellebaum & Daum, 2004; Weible, Oh, Lee & Disterhoft, 2004), Alzheimer’s Disease (Woodruff-Pak, 2001), psychiatric diseases, e.g. schizophrenia and autism (Parker, Narayanan & Andreasen, 2014) and addiction. Nancy Andreasen has proposed that disrupted cerebellar-forebrain input to both the prefrontal cortex and ventral tegmental area contributes to a dysmetria of thought and schizophrenic behavior, and to autistic behavior. Thomas Gould has examined the differential effects of the addictive compound nicotine on delay and trace conditioning, albeit using fear conditioning instead of blink conditioning (Kenney & Gould, 2008; Gould, Feiro, Moore, 2004) and finds that trace conditioning is affected more than delay conditioning. Given our ability to use individually adjustable tetrodes to record many neurons simultaneously throughout the rabbit brain during conditioning (Hattori, Chen, et al., 2014; Hattori, Yoon, et al., 2014), and our ability to image the rabbit brain longitudinally without anesthesia or sedation (Miller et al., 2003, 2008; Schroeder, Weiss, Procissi, Disterhoft, 2012) puts the field in a very good position to understand the mechanisms underlying conditioning and memory for the associated stimuli, and the problems that arise to impair or prevent the creation, storage, or retrieval of memories.

Robert “Bob” Clark (2011) pointed out the irony that a simple reflex system like delay eyeblink conditioning can inform us so much about declarative memory by simply inserting a stimulus free trace interval between the conditioning stimuli. We agree with him and continue to build upon a solid foundation of neurobiological research that followed the Moyer et al. report demonstrating that acquisition of trace eyeblink conditioning requires the hippocampus.