Valproate reopens critical-period learning of absolute pitch

Introduction

Absolute pitch (AP), the ability to identify or produce the pitch of a musical sound without any reference point, has long fascinated musicians, music scholars, psychologists, and neuroscientists (Stumpf, 1883; Mull, 1925; Takeuchi and Hulse, 1993; Zatorre, 2003; Levitin and Rogers, 2005). Individuals who possess AP, constituting about 0.01% of the general population, are able to identify the pitch class, i.e., one of the 12 notes of the Western musical system, e.g., C, D, G#, of a sound with great accuracy (varying between 70–99%, depending on the task, as compared to 10–40% for non-AP individuals, Takeuchi and Hulse, 1993). Their errors are usually not more than a semitone away from the target sound, as compared to 3 or more semitones for non-AP individuals. AP possessors also make octave errors, i.e., they identify the pitch class, but not the pitch height correctly, labeling a C₄ (middle C) as C₅, an octave higher. Pitch class and pitch height identification are thus believed to be separate processes, and only the former constitutes a crucial test for AP (Takeuchi and Hulse, 1993; Levitin and Rogers, 2005), which AP possessors perform effortlessly and automatically. Their reaction times are faster, at least when responding correctly, than those of non-possessors (Miyazaki, 1990), suggesting that the former have direct access to pitch names in memory, whereas the latter might rely on relative pitch to calculate pitch class. In light of AP possessors' highly accurate and automatic identification of pitch class, it has been suggested that in addition to perceptual mechanisms, AP involves the association of (verbal) labels for pitch classes in long-term memory (Zatorre, 2003; Levitin and Rogers, 2005).

The central role of these associations is further supported by functional imaging studies showing the involvement of the left posterior dorsolateral frontal cortex and the planum temporale, known to be responsible for learning conditional associations (Zatorre et al., 1998; Ohnishi et al., 2001; Bermudez and Zatorre, 2005; Wilson et al., 2009). In addition, AP possessors have a larger planum temporale (Zatorre et al., 1998), with a left-right asymmetry involving a smaller right planum temporale and an increased leftward asymmetry in AP possessors (Keenan et al., 2001), and show hyperconnectivity in the temporal cortex (Loui et al., 2011), facilitating tone-label mapping.

Importantly, acquiring AP has a critical period (Levitin and Zatorre, 2003; Russo et al., 2003). A critical period is a fixed window of time, usually early in an organism's lifespan, during which experience has lasting effects on the development of brain function and behavior. The principles of critical period phenomena and neural plasticity are increasingly well understood both at the behavioral/experiential (Kleim and Jones, 2008) and at the molecular/cellular level (Hensch, 2005). Specifically, behaviorally induced plasticity in the healthy brain, typically after the end of the relevant critical period, can lead to improvement beyond normal or average performance levels. However, for many tasks, this requires targeted training—simple routine use is often insufficient. The factors known to influence the efficiency of such targeted training include the number of repetitions involved, the intensity of the training as well as the relevance or saliency of the stimuli or task trained. Importantly, such training-induced learning is quite specific to the trained task and to the underlying brain networks, although some transfer to other, related domains of knowledge or skills is sometimes possible. At the cellular level, critical periods close when maturational processes and experiential events converge to cause neuoro-physiological and molecular changes that dampen or eliminate the potential for further change (Hensch, 2005; Bavelier et al., 2010), thus imposing “brakes” on neuroplasticity. One of the epigenetic changes leading to decreased plasticity after the critical period involves the action of HDAC, an enzyme that acts as an epigenetic “brake” on critical-period learning (Morishita and Hensch, 2008; Qing et al., 2008). Research has shown that inhibition of HDAC can reopen critical-period neuroplasticity in adult mice to enable recovery from amblyopia (Putignano et al., 2007; Silingardi et al., 2010), and to facilitate new forms of auditory learning (Yang et al., 2012).

The age of onset of musical training has been shown to predict AP acquisition (Deutsch et al., 2009), indicating the presence of a critical period. AP is most typically seen in individuals who started musical training before 6 years of age (Levitin and Zatorre, 2003; Russo et al., 2003; Miyazaki and Ogawa, 2006). Indeed, the distribution of the age of first formal musical training in a large number of individuals with AP can be modeled with a gamma function with a mode at 4–6 years (Levitin and Zatorre, 2003). Training that begins after the age of 9 very rarely leads to AP, and there are no known cases of an adult successfully acquiring it (Brady, 1970; Ward and Burns, 1999; Levitin and Rogers, 2005). The appropriate type of input, i.e., training associating absolute pitches to labels, thus needs to be available before the end of the critical period for AP to develop. For most individuals, this is not the case, as the two major sources of auditory input during early development, language, and the Western musical tradition mainly rely on relative pitch, explaining why not all musically trained individuals have AP. In the absence of AP cues during the critical period, the perceptual system is reorganized, shifting weight from absolute to relative pitch information (Takeuchi and Hulse, 1993; Saffran and Griepentrog, 2001; Saffran, 2003, although see Trehub, 2003 for a somewhat different view).

AP is thus particularly interesting from a neuro-scientific perspective, as it provides a model for understanding the interaction of genes and experience on the development of neural and cognitive function (Zatorre, 2003). In the current study, we explored whether a reopening of the critical period was possible for AP learning in human adults. We sought to establish whether the administration of valproate (VPA), a commonly used anticonvulsant and mood stabilizer, known to inhibit HDAC and modulate the epigenome to promote neuroplasticity (Phiel et al., 2001; Schloesser et al., 2007; Machado-Vieira et al., 2011) would facilitate training naïve, non-musician adults on the identification of pitch classes in a classical AP task (Deutsch et al., 2006).

Previous studies (Meyer, 1899; Mull, 1925; Wedell, 1934; Hartman, 1954; Lundin, 1963; Cuddy, 1968; Russo et al., 2003) have shown that training improves adults' AP abilities only under restricted conditions. Three factors appear to play a particularly important role: participant's previous musical training/experience, whether a single tone or a series of tones are used for training, and the duration/intensity of the training. Notable improvement is achieved only when musically highly proficient participants are trained on a single note for extensive periods (from several weeks to several months) and tested on the recognition of this single target tone among several non-target ones (Mull, 1925; Cuddy, 1968). It is possible, however, that at least in some such cases, the improvement is actually due to greater familiarity with the particular task, tone series or procedure, rather than to a genuine increase in AP ability (Takeuchi and Hulse, 1993). Improvement is much more limited when musically untrained participants are tested (Cuddy, 1968), when participants are trained on a series of pitch classes rather than on just one pitch class (Cuddy, 1968), or when training is relatively short or less intense (Vianello and Evans, 1968). Prior musical training is particularly important, as musically proficient participants typically perform better than chance (Mull, 1925; Lundin, 1963) and better than musically naïve participants (Cuddy, 1968) on pitch identification already prior to AP training, it is thus not surprising that some AP improvement may be achieved in this population. A recent finding further shows that in addition to early musical training, the current musical environment also contributes to (the maintenance of) AP abilities, suggesting the presence of some residual plasticity for AP, at least in individuals in whom AP emerged during the early critical period (Wilson et al., 2012).

The use of a single tone vs. a series of tones for training is also highly relevant, as the two training methods might tap into different underlying pitch perception abilities or might represent tasks of varying degrees of complexity. In our current understanding, the identification of only one pitch class serving as an internal reference, sometimes referred to as quasi-AP or single tone AP, is qualitatively different from true AP, whereby the individual is able to identify a large number of pitch classes automatically and without an internal or external reference. Albeit often similar in the percentage of correct identifications, the two types of abilities can be distinguished on the basis of reaction times, as AP possessors have faster reaction times than absolute tuning tone processors (Miyazaki, 1990; Levitin and Rogers, 2005). It needs to be noted, however, that according to some authors (e.g., Cuddy, 1968) the single tone method might also lead to true AP eventually.

The length and intensity of training also affects performance (Brady, 1970). Fast improvement is observed for tones that are separated by large pitch distances, whereas more extended training is necessary for pitches that are closer together (Hartman, 1954). Not surprisingly, the latter are in general harder to learn and to discriminate than the former, with musically naïve participants not being able to achieve discrimination errors smaller than about 5 semitones with pitch classes that are separated by small distances (Wedell, 1934). Further, interference effects are sometimes observed when tone series are taught gradually, involving the introduction of new tones that fall in between already trained ones, decreasing pitch distance, and disrupting the subjective organization of the scale.

To assess whether taking VPA could reopen the opportunity for critical-period-like learning in adults, we conducted a randomized, double-blind, placebo-controlled study, in which 24 young adult males received either placebo or VPA treatment in a cross-over design with two treatment arms. Participants underwent 15 days of treatment during which they took capsules (VPA or placebo) each day. During the second week of treatment, participants observed training videos that taught participants to associate six pitch classes from the 12-tone Western musical system (e.g., C, D, E, F#/Gb, G#/Ab, A#/Bb) with six proper names (e.g., Sarah, David, Francine, Jimmy, Karen, Leo). We chose to use proper names instead of the actual note names to make the task equally novel and accessible for participants with and without any prior musical training, and to divert attention from the music theoretical aspect of the task. We acknowledge that this may interfere with existing knowledge of actual musical note names in the participants who had sufficient musical training, but given that there were few such participants in our study, we considered that the advantages of using proper names outweighed the potential negative influence of such interference. On day 15, participants were given a post-treatment assessment for AP in which they heard 18 synthesized piano tones and had to identify the proper name associated with the pitch class of each tone. After the first treatment arm, a washout period of 2 to 4 weeks elapsed. Eighteen out of the 24 original participants then entered the second treatment arm, which was similar to the first one, except that the drug (VPA or placebo) not received during the first arm was administered.

Given the difficulty of improving AP performance in adulthood, we hypothesize that in our task, even a small advantage in pitch class identification in the VPA as compared to the placebo group is suggestive of the reopening of plasticity, as musically naïve participants were trained for a relatively short time period on several pitch classes, conditions under which no existing study has shown any improvement in AP. The strong hypothesis is that there might be improvement in the VPA condition in both arms of the design. However, since new training is introduced in the second treatment arm, successful learning in the first arm might carry over to and interfere with learning in the second arm. Effects are thus more likely in the first arm only. This study is, therefore, intended as a proof-of-concept demonstration that critical-period-like AP learning may be at least partly restored by using a drug to remove the epigenetic brakes on neural plasticity.

Materials and methods

Material

Results

Discussion

Until now we had no mechanistic account of the neural processes underlying the critical period of AP. More generally, we have lacked human experimental models with which to measure the potential for a compound to facilitate neuroplasticity in the adult human brain. This study provides the “proof-of-concept” for the possibility to restore neuroplasticity using a drug by offering evidence for a possible effect of VPA on AP perception. In confirmation of our hypothesis, AP performance varied according to treatment condition. Normal male volunteers performed significantly better on a test of AP after 2 weeks of VPA treatment than after 2 weeks of placebo.

Certain aspects of our findings warrant further discussion. First, it was not possible to establish baseline performance on the AP task, as the association between the musical notes and the names was necessarily established during training. One possible option would have been to test participants' AP performance early in the regimen, while they were gradually reaching the full dose of VPA. We had decided not to implement such a baseline test, as no precise information was available regarding the time course of the effect of VPA under the current conditions, so AP training was only started on day 8, once participants have reached the full dose. The absence of a baseline test might introduce ambiguity into the interpretation of the effect. Indeed, there are identifiable individual differences in musical ability, and in the neuroanatomical structures supporting it (see Herholz and Zatorre, 2012 for a review), which might have led to participants with better AP abilities assigned to the VPA group in the first treatment arm by chance. However, there is evidence in our study that the significant effect of treatment for the AP task cannot be fully accounted for by a between-group difference in the ability to acquire AP. If such a difference were the driving force, then we would have expected that in the second treatment arm the participants in the potentially high-ability group (i.e., those who were in the VPA condition for the first treatment arm) would outperform the low-ability group (i.e., those who were in the placebo condition for the first treatment arm). However, this was not the case. Further, since our participants were musically naïve (participants who received musical training all started after the age of 7 years, with a mean of 12 years), the presence of complete or partial AP possessors in either of the groups is very unlikely. As reported above, there was no significant difference between the two groups in the second treatment arm. Further, all of the top scorers from the VPA condition in the first treatment arm completed the second treatment arm, eliminating the possibility that the participants who completed the second treatment arm were not representative of the group from the first treatment arm. This strengthens our interpretation that the VPA treatment led to the significantly higher performance in the VPA condition compared to placebo.

Second, the analysis of the crossover, i.e., of the 17 participants for whom we have data from both arms, revealed an order-dependent effect of treatment. For participants who took VPA first, AP performance was significantly higher after VPA treatment than after placebo. In contrast, for participants who initially took placebo, there was no such difference. It may be that carry-over effects impeded performance on the AP task in the second treatment arm. A memory conflict between the pitch classes and proper names used in the first treatment arm could have interfered with those used in the second. Because of this order effect, the most reliable comparison focused on the first treatment arm. Future research should aim to use a randomized controlled trial (RCT) design with treatment condition as a between-participants factor. Doing so would avoid any possible carry-over effects from one treatment arm to the next. Relatedly, it needs to be noted that we did not test how long the effect of the improvement in AP perception lasted. Future research will need to address this question by retesting participants after several days or weeks following the end of training.

Third, our study was not set up to measure reaction times. Participants were informed that they had 4 s available for their response with no instruction to do so rapidly. However, as AP possessors are known to recognize pitch classes faster than non-AP possessors do (Levitin and Rogers, 2005), reaction times constitute a relevant measure to use in future follow-up studies.

Fourth, the sizes of the groups tested were relatively small compared to some recent AP studies, which might explain why no significant correlation was observed between AP performance and training compliance, i.e., the amount of training received. However, when compared to studies using VPA with healthy participants, our sample sizes were quite appropriate (e.g., n = 12 in the placebo as well as in the VPA group in the (Bell et al., 2005a,b) study).

Fifth, we observed an effect of VPA on the Altman Self-Rating Mania Scale. Indeed, VPA is commonly used as a mood stabilizer to control mania in bipolar disorder (Macritchie et al., 2001). Our results in healthy participants therefore confirmed the clinical action of VPA, which may have an effect on sub-clinical levels of mania as well. Importantly, however, an improved or more stable mood in the VPA condition cannot explain the obtained results, as we would then expect general cognitive improvement in the other neurocognitive tasks administered, as well as greater AP performance in both treatment arms.

In sum, our study is the first to show a change in AP with any kind of drug treatment. The finding that VPA can restore plasticity in a fundamental perceptual system in adulthood provides compelling evidence that one of the modes of action for VPA in psychiatric treatment may be to facilitate reorganization and rewiring of otherwise firmly established pathways in the brain and its epigenome (Shen et al., 2008).

Valproic acid is believed to have multiple pharmacological actions, including acute blockade of GABA transaminase to enhance inhibitory function in epileptic seizures and enduring effects on gene transcription as an histone deacetlyase (HDAC) inhibitor (Monti et al., 2009). Of relevance here is the epigenetic actions of this drug, as enhancing inhibition does not reactivate brain plasticity in adulthood (Fagiolini and Hensch, 2000), but reopening chromatin structure does (Putignano et al., 2007). While systemic drug application is a rather coarse treatment, the effects may differ dramatically by individual cell type (TK Hensch and P Carninci, unpublished observations). VPA treatment mimics Nogo receptor deletion to reopen plasticity for acoustic preference in mice (Yang et al., 2012), suggesting a common pathway through the regulation of myelin-related signaling which normally closes critical period plasticity (McGee et al., 2005). Future work will address the cellular actions of VPA treatment in the process of reactivating critical periods. Future MRI studies will also be needed to establish whether HDAC inhibition by VPA induces hyperconnectivity of myelinated, long-range connections concurrent with renewed AP ability (Loui et al., 2011).

If confirmed by future replications, our study will provide a behavioral paradigm for the assessment of the potential of psychiatric drugs to induce plasticity. In particular, the AP task may be useful as a behavioral correlate. If further studies continue to reveal specificity of VPA to the AP task (or to tasks on which training or intervention is provided), critical information will have been garnered concerning when systemic drug treatments may safely be used to reopen neural plasticity in a specific, targeted way.