Overview

Neuroplasticity has a profound impact on the manifestation and progression of clinical symptoms in many psychiatric illnesses. Understanding the multiscale role that plasticity has on neuropathology is critical to creating improved, targeted therapies. Through experiments that directly manipulate plasticity to create testable, causal changes over time, and through cross-sectional, retrospective, and postmortem analysis that reveal accumulated differences in plasticity that arise naturally, it is possible to obtain valuable information about the mechanism-of-action underlying healthy and disordered synaptic plasticity. While there has been considerable progress toward understanding these mechanisms, the impact of plasticity on behavior and the best approaches to induce neuroplastic change for therapeutic gains still remain open and important questions. This review seeks to provide a contemporary overview of the science elucidating the mechanisms of plasticity, the current state-of-the-art in methods used to image plasticity and promote therapeutic gains, and to provide suggestions for profitable future directions.

Mechanisms of neural plasticity

Neural plasticity refers to the possibility of altering the strength of connections within the nervous system through experience or injury. Plasticity can act to reorganize either the structure or the function of neurons and is necessary not only for neural networks to acquire new capabilities, but also for them to remain robust and stable over time. A key element of plasticity revolves around the temporal coincidence of activity. So-called, spike-timing-dependent plasticity (STDP) is a Hebbian learning rule in which the modification of synaptic strengths depends on the relative timing of action potentials [1]. Within monosynaptic pairs of neurons, it has been shown that if an input spike from the presynaptic neuron occurs immediately before the postsynaptic neuron’s output spike, then that input becomes stronger, creating long-term potentiation (LTP). If the input spike occurs immediately after an output spike, however, that input is made weaker, creating long-term depression [2, 3] (LTD) (While STDP generally involves molecular alterations at the synapse they rely on wide variety of distinct mechanisms that can differ in different brain regions (e.g., LTP induced during learning is different in the hippocampus and amygdala), among different neurons in the same brain region (e.g., endocannabinoid and non-endocannabinoid LTD in projections from the striatum to the basal ganglia), or within the same types of neurons (e.g., hippocampal CA3 pyramidal neurons that converge from different afferent inputs)). This critical window of timing-dependency spans tens of milliseconds and has profound functional implications on brain function, creating a means for activity-dependent bidirectional modification of synaptic strength, and ultimately forming the physiological basis for learning and memory. The mechanism underlying STDP has been attributed to two different glutamate receptors that are commonly co-expressed, the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor and the N-methyl-D-aspartate (NMDA) receptor [4]. The NMDA is a glutamate receptor cation channel that is also widely referred to as a “coincidence detector”. Within this channel, coincidence is detected by the simultaneous presence of both membrane depolarization that vacates a channel-blocking magnesium ion, and the binding of its natural ligand, glutamate [4, 5]. The preexisting membrane depolarization of NMDA is in turn mediated through the coactivation of AMPA activation [6] to create reciprocal cellular mechanisms that enable long-term synaptic changes.

Beyond spike-timing-dependent plasticity that occurs rapidly at the synapses, other slower homeostatic processes occur over hours, days, or weeks to modify ion channel density, neurotransmitter release, or postsynaptic receptor sensitivity [7]. These processes are triggered in response to rapid activity-dependent changes and constitute a negative feedback loop, decreasing connectivity in response to high neuronal activity but increasing connectivity when activity drops [8]. Yet, another plasticity-based moderator of neuronal network communication is the growth of myelin, a multilayered membrane produced by oligodendrocytes that surrounds axons to increase the speed by which electrical signals propagate through the nervous system. While widely associated with critical developmental periods, there is also considerable evidence that activity-dependent myelination continues into adulthood [78]. CBT was found to yield decreases in both gray matter volume and BOLD responsivity in the amygdala, implying structural and functional level changes that decrease synaptic strength. The authors postulated that the diminished gray matter mediated the decreased responsivity and social anxiety in the CBT group, providing a specific mechanism by which plasticity may produce clinically meaningful effects in this treatment paradigm [78]. Conversely, a study of 13 patients with chronic pain who received CBT showed improved clinical measures as well as increased gray matter in the prefrontal and posterior parietal cortices as demonstrated through MRI VBM compared to 13 healthy controls [79]. These two studies taken together reinforce the premise for therapy-driven modulations of synaptic plasticity. Both studies led to clinical improvement, however through a decrease and increase in synaptic strength, respectively.

Prefrontal neural networks involved in motivation, cognition, social and emotional behavior exhibit learning-dependent plasticity. There has been considerable research on develo** engaging cognitive training to induce and maximize this plasticity, yielding several key takeaways. Namely, the mechanisms behind cognitive training induced neural plasticity are present throughout life and can impact higher-order cognition, but are also impacted by behavioral states and brain chemistry [80]. From a clinical perspective, data has demonstrated large, generalizable, and durable effects, with evidence of plasticity in frontal and sensory neural networks [80].

Individuals with schizophrenia are known to have accelerated cortical thinning, and by extension synapse loss, that is associated with illness severity. A recent study examined changes in cortical thickness in patients with recent onset schizophrenia randomly allocated to either targeted cognitive training or a computer game control intervention; whereas mean cortical thickness was not significantly changed in response to targeted cognitive training, individual increases in cortical thickness were predictive of increases in global cognition—an effect not seen in the control group [81]. Previous studies in patients with schizophrenia have established that cognitive training may impact structural and functional plasticity. Notably, 40 h of auditory and working memory targeted cognitive training were associated with changes in thalamic volume in patients with schizophrenia, suggestive of either a protective or enhancing effect on thalamic structure [82]. Of these two possibilities, a model for neural response to targeted cognitive training in patients with schizophrenia is one in which the training evokes general neuroprotection, as opposed to region-specific changes in neuroplasticity [81]. This is believed to be due to the increased cortical thickness, or prevented cortical thinning, seen in patients with schizophrenia who had undergone targeted cognitive training and exhibited corresponding increased global cognitive scores [81].

The left and right middle frontal gyri (MFG) of patients with schizophrenia are known to have signaling abnormalities, exhibiting both hyper and hypoactivation when performing working memory tasks [83, 84]. Intensive cognitive training in patients with schizophrenia increased fMRI signal efficiency of MFG compared to computer game control recipients with lasting effects present at 6 months post-intervention; notably, participants had improved working memory task performance, increased activation in the left MFG, and a significant association between improved task performance and right MFG signal [85]. Once more, changes in signaling efficacy seen in the cognitive training group lends itself to the idea that neural plasticity in humans can be induced in a measurable form via deliberate training.

Future research directions—challenges and opportunities

Understanding the roles of neuroplasticity in neuropsychiatric illnesses and their relationship to symptom development is critical to creating improved, targeted therapies. While there has been progress toward understanding the mechanisms of plasticity and innovations in therapies, there remain several challenges to studying this process. One of the major challenges is the multiscale nature in which the brain changes. Plasticity occurs on a spectrum from microscale changes at the synapse to morphological alterations that span the entire nervous system. They occur moment-to-moment and continually throughout the entire lifespan. Relating synaptic development to clinical improvement requires integration of multiple tools in conjunction, adding to the complexity of the evaluation, but creating new opportunities for discovery and development. Moreover, when comparing the brain before and after treatment, or even between healthy and diseased individuals, plasticity can depend on experience, time, and individual factors such as stress and genetics [15]. Interactions between these variables needs to be considered when attempting to delineate healthy and diseases brains. Therefore, this highly variable process complicates the generalizability of biomarkers and presents challenges to the field moving forward.

Although animal models can provide direct insight into neuronal alterations through invasive imaging and sampling, establishing valid animal models of psychiatric illnesses is difficult and means that findings remain speculative until validated in humans [86]. Animal models are also typically more homogenous in their symptomology and do not have co-morbid illnesses which tend to complicate disease etiology in humans. This is especially relevant when trying to correlate changes in plasticity to improvements in disease related behavior. On the other hand, most measures to assess plasticity in humans are noninvasive and are therefore not able to directly ascertain changes at a microscale, leaving open the possibility that they reflect indirect consequences of disordered plasticity and not the foundational pathology itself. In addition, due to the heterogenous nature of psychiatric illnesses and multiple factors influencing plasticity, establishing baseline healthy brain states, relative to disease states, is an important challenge. It is therefore challenging to assess the causes of maladaptive plasticity and distinguish these from their effects on behavior.

Despite these challenges, there is tremendous growth in the field with boundary-crossing innovations that allow greater resolution in space and time, and new approaches to link across multiple scales. In particular, there is a strong movement toward the development of predictive biomarkers of treatment efficacy that are critical for understanding not only how plasticity influences symptoms but determining who will benefit most from a given therapy. For example, by demonstrating correlations between specific markers of plasticity and individuals who respond to treatment, future clinical trials can use this information to predict whether a patient is likely to improve through changes in brain plasticity [87]. One example of a clinical biomarker currently in use is the presence of single nucleotide polymorphisms that correlate with greater efficacy of olanzapine in schizophrenia [88]. Epigenetic markers are also useful to predict drug efficacy such as the absence of methylated exon 4 of BDNF which is associated with reduced response to anti-depressants [89]. Not only would it be informative to have biomarkers of treatment response, but also of treatment progression [90]. To address some of the concerns with regards to disease heterogeneity and causes of dysregulation, it is crucial to have a better biological understanding of healthy neuroplasticity and how it can be dysregulated. Biomarkers would provide an objective measure of disease progression and pathogenesis, which could ultimately inform therapies.

Another profitable direction for future development is the individualization of treatment protocols to each patient with the goal of improving therapeutic gains. This is especially relevant to address the challenge of varied clinical profiles. For example, in most TMS protocols, patients are stimulated at the same brain region regardless of connectivity patterns or brain state during stimulation. Multiple studies, however, have shown that patients who respond best to TMS received stimulation at the node of the DLPFC most anti-correlated with the SGC [66]. Currently, there are investigations analyzing the difference between stimulating during resting-state, a more variable brain state between patients, versus during a task to create consistency or engage the network of interest [91]. Finally, combining therapies could provide another degree of personalization if patients respond better with just neurostimulation or with behavioral or pharmacological interventions as well. Biomarker development and improved therapeutic tools have the potential to transform patient outcomes through treatment personalization and objective measure of neuroplasticity.

Collectively, there is a profound need consistency in methodologies of interventions for psychiatric disorders. Currently response and remission rates remain low which may be due in part to the lack of proper controls, highly variable treatment parameters, and lack of consistency among clinical symptoms of patients [92]. By combining functional and structural imaging tools, therapies can be tailored to the individual. These and other innovations described in the preceding sections will continue to drive the field forward and to create more effective and efficient treatments