Myelination of parvalbumin interneurons: a parsimonious locus of pathophysiological convergence in schizophrenia

Abstract

Schizophrenia is a debilitating psychiatric disorder characterized by positive, negative and cognitive symptoms. Despite more than a century of research, the neurobiological mechanism underlying schizophrenia remains elusive. White matter abnormalities and interneuron dysfunction are the most widely replicated cellular neuropathological alterations in patients with schizophrenia. However, a unifying model incorporating these findings has not yet been established. Here, we propose that myelination of fast-spiking parvalbumin (PV) interneurons could be an important locus of pathophysiological convergence in schizophrenia. Myelination of interneurons has been demonstrated across a wide diversity of brain regions and appears highly specific for the PV interneuron subclass. Given the critical influence of fast-spiking PV interneurons for mediating oscillations in the gamma frequency range (~30–120 Hz), PV myelination is well positioned to optimize action potential fidelity and metabolic homeostasis. We discuss this hypothesis with consideration of data from human postmortem studies, in vivo brain imaging and electrophysiology, and molecular genetics, as well as fundamental and translational studies in rodent models. Together, the parvalbumin interneuron myelination hypothesis provides a falsifiable model for guiding future studies of schizophrenia pathophysiology.

Main

Schizophrenia is a chronically debilitating psychiatric disorder with a lifetime prevalence of ~1%.¹ Patients with schizophrenia classically exhibit a constellation of positive, negative and cognitive symptoms.² Although many theories have been proposed, the precise neurobiological mechanism underlying schizophrenia has remained elusive. The most widely described models have been the dopamine³ and glutamate hypotheses,4⁴ although in recent years models regarding interneuron dysfunction⁵ and myelination abnormalities⁶ have gained increasing support.

In this Perspective, we hypothesize that previous observations of interneuron dysfunction and myelination abnormalities in schizophrenia might converge on the altered myelination of fast-spiking parvalbumin (PV) interneurons. First, we summarize the major evidence supporting interneuron dysfunction and myelination abnormalities in schizophrenia. Next, we summarize electron microscopy and immunofluorescence studies that convincingly demonstrate interneuron myelination, which frequently occurs on fast-spiking PV interneurons. Finally, we discuss how impairments in myelination of PV interneurons could lead to consequent abnormalities in gamma synchronization and ultimately give rise to the symptoms which define schizophrenia.

Parvalbumin interneuron dysfunction in schizophrenia

Deficits in GABAergic signaling have been widely proposed as a fundamental pathophysiological mechanism underlying schizophrenia.⁷ More specifically, several recent lines of evidence including human postmortem studies, genetics and in vivo electrophysiological recordings in patients and translational mouse models have identified fast-spiking PV interneurons as the major GABAergic cell-type affected in schizophrenia (Table 1).

Table 1 Comparison of interneuron and myelination data for schizophrenia

Expression of GAD67—the predominant gamma-aminobutyric acid (GABA) synthesizing enzyme—has consistently been found to be reduced at both the messenger RNA and protein levels in several brain regions of patients with schizophrenia, a finding that has been well controlled for confounding factors.^{8, 9, 10, 11, 12, 13, 14} Downregulation of GAD67 messenger RNA has been reported in ~30% of dorsolateral prefrontal cortex interneurons^{15, 16} and entirely undetectable in ~50% of PV interneurons.¹⁷ Expression of PV messenger RNA^{18, 19, 20} and protein²¹ is also reduced in schizophrenia, while the neuronal density of cortical PV interneurons is unchanged^{22, 23, 24, 25} (but see also ref. 52). Since the expression of both PV and GAD67 are experience-dependent²⁶—and GAD67 and PV expression are highly correlated²⁶—their shared downregulation suggests a functional impairment of fast-spiking interneurons.²⁷ Morphologically, PV cell inputs onto pyramidal neurons have no discernible alterations²¹ suggesting a primary functional abnormality of PV interneurons. Consistent with these neuropathological findings, in vivo positron emission tomography (PET) imaging has demonstrated widespread alterations of cortical GABA transmission in schizophrenia, a finding that was most prominent in the subset of patients who were antipsychotic-naïve.²⁸ Altogether, these results provide compelling evidence of cortical PV interneuron dysfunction in schizophrenia.

PV interneurons are essential in generating cortical oscillations in the gamma range (~30–120 Hz), mediated by synchronized inhibition of large pyramidal cell ensembles.^{29, 30} Through rhythmic perisomatic inhibition onto surrounding pyramidal cells, synchronous ensembles of PV cells evoke high-frequency gamma oscillations in the cerebral cortex.^{31, 32, 33} Gamma synchrony has been shown to function critically across a range of cognitive functions, including working memory and attention³⁴ with well-replicated abnormalities in schizophrenia.^{5, 35} Abnormalities in other frequency bands such as theta and alpha have also been reported in schizophrenia, but the neural mechanisms underlying these frequencies remain less well understood.³⁵

Electroencephalographic studies in schizophrenia have shown a reduced amplitude and impaired phase locking of gamma band activity over frontal areas while assessing working memory and executive functioning tasks.³⁵ Although some studies have observed concurrent increases in gamma band activity at rest, this finding has been less well replicated.³⁵ Taken together, impairments of in vivo gamma oscillations in patients with schizophrenia are highly consistent with the PV interneuron abnormalities observed by postmortem histopathology.

The classical onset of schizophrenia occurs within a relatively narrow window of neurodevelopment, between ~18 and 25 years of age.² This late adolescent age of onset has often been attributed to the ongoing functional maturation of the brain during this neurodevelopmental critical period.² Specifically in late adolescence, rates of synaptic pruning and myelination become asymptotic for which impairments in these processes have been linked to the disease onset.² Notably, maturation of gamma band synchrony also occurs during late adolescence³⁵ which coincides developmentally with the clinical onset of schizophrenia.¹

In addition to in vivo brain imaging, electroencephalographic recordings and postmortem histopathology, molecular genetic studies of schizophrenia have also revealed an important contribution of interneuron dysfunction to the pathophysiology of schizophrenia. A recent genetic study of copy number variation has now provided causal evidence for GABAergic dysfunction in the etiology of schizophrenia.³⁶ In this study, Pocklington et al. performed a functional gene set analysis for enriched biological mechanisms using a large schizophrenia case-control dataset and found that copy number variations were significantly enriched in cases for genes responsible for inhibitory neurotransmission (in particular the GABA_A receptor complex), glutamatergic neurotransmission, long-term synaptic plasticity and associative learning. The genetic variant with the highest known risk for schizophrenia is the 22q11 microdeletion which has a penetrance of ~40%.^{37, 38} Transgenic mouse models have been generated to investigate the underlying neurobiology conferred by 22q11 microdeletion. Df(16)A mice harboring a 27-gene microdeletion syntenic to a 1.5 Mb region of human 22q11.2 exhibit similar brain abnormalities as found in human 22q11 microdeletion carriers, including cortico-cerebellar, cortico-striatal and cortico-limbic circuits.³⁹ Moreover, multiple different mouse models of 22q11 microdeletion have replicated a cell-type specific impairment in PV interneurons and disrupted local synchrony of neural activity, consistent with the deficit in gamma oscillations observed in schizophrenia.^{40, 41, 42}

Evidence for interneuron dysfunction in schizophrenia has also been supported by a wide variety of non-genetic rodent models.⁴³ The major examples include pharmacological NMDA receptor antagonism and neurodevelopmental immunological challenge, both of which consistently exhibit synaptic and network abnormalities reminiscent of schizophrenia pathophysiology. Specifically, these studies have identified electrophysiological changes in local microcircuit connectivity and synaptic plasticity, with alterations in excitation/inhibition balance and gamma band synchronization.

Taken together, the combination of genetic, postmortem, and in vivo electrophysiological and functional imaging results from human clinical studies of schizophrenia converge with translational rodent modeling to identify fast-spiking PV interneuron dysfunction as a major pathophysiological mechanism underlying schizophrenia etiology.

Myelination abnormalities in schizophrenia

Independent of PV interneuron alterations, myelination abnormalities have also been extensively implicated in schizophrenia through both in vivo brain imaging and postmortem assessments (Table 1). Numerous diffusion tensor imaging studies have been published for schizophrenia (reviewed in ref. 6), of which the overwhelming consensus has been the association of schizophrenia with globally decreased fractional anisotropy. Notably, the decrease in fractional anisotropy appears to become more severe with increasing age and illness duration.⁴⁴ Many of the early brain imaging studies of schizophrenia were performed in cohorts with extensive histories of psychotropic medication, inpatient hospitalization, smoking and medical co-morbidities, which could have a confounding deleterious influence on white matter integrity. Thus, an important question has been whether myelination abnormalities are already present in drug-naïve patients with first-episode schizophrenia who have never received psychotropic medication. Recently, several diffusion tensor imaging studies have been performed in such cohorts,^{44, 45, 46, 47, 48, 132} Synaptic input from local GABAergic interneurons has been shown to dynamically regulate OPC differentiation to oligodendrocytes.¹³³ OPCs receive strong GABAergic synaptic input from PV, and to a lesser extent from non-PV, interneurons.¹³² Notably, the peak neurodevelopmental period of interneuron-OPC connectivity (P10-P14) would thus position interneuron myelination precisely in the window following the initial onset of GABAergic burst firing, but before maturation of high-frequency gamma oscillations.¹³⁴ This also closely aligns with the timing of human frontal cortex oligodendrocyte development which plateaus in early adulthood and is highly distinct from white matter development in which oligodendrocytes have already reached their maximum number by ~5 years of age.¹³⁵ Moreover, in further contrast to white matter, frontal cortex gray matter exhibits a substantial turnover of oligodendrocytes and myelin that persists throughout adulthood.¹³⁵ Analogously, rodent studies have demonstrated that OPCs exhibit important distinctions in their physiology, proliferation and differentiation between gray and white matter in rodents.¹³⁶ Therefore, regional differences in human OPCs are also not unlikely.

Interestingly, direct contacts of interneurons onto OPCs¹³⁷ are only locally distributed, reaching a typical maximum distance of 50–70 μm,¹³² which is highly similar to the estimate for the maximal length of OPC processes. An interesting question remains why interneurons have such a restricted spatial localization of their connectivity onto OPCs, since PV cells establish synaptic contacts with pyramidal cells across a distance approximately six times larger.¹³⁸ One possibility is that OPCs utilize reciprocal synaptic input to regulate their proliferative drive. Alternatively, it may be that myelination preferentially occurs on proximal axonal segments, in close apposition to the observed localization of OPCs and allowing for rapid differentiation to oligodendrocytes with enhanced myelination plasticity.

Potential functions of interneuron myelination

PV interneurons function to synchronize pyramidal cell ensembles, and thereby generate high-frequency oscillations.¹³⁹ Since cortical PV axonal arborization is widely ramified and distributed over distances of up to 300 μm,¹³⁸ there might be considerable benefits of myelination for optimizing the fidelity of fast action potential transmission. Indeed, computational modeling has suggested a unique contribution of (interneuron) conductance delays in the dynamics of gamma frequency oscillations.¹⁴⁰ Evidence exists that nodes of Ranvier begin forming before the onset of myelination,¹¹⁷ a mechanism specific for GABAergic neurons, which enhances axonal conduction of action potentials without myelin. Thus, in addition to simply increasing the speed of action potential propagation, myelin could function to ensure the integrity of precisely timed action potentials, as has been proposed for myelinated excitatory axons.¹⁴¹ Myelin plasticity would then have the potential to support the local synchronization of action potentials necessary for generating high-frequency oscillations.¹⁴² Indeed, myelinated axons exhibit both higher conduction velocities and enhanced long-range coherence.¹⁴³ Although non-PV cortical interneuron subtypes (e.g., somatostatin, VIP) exhibit synaptic connectivity across similar distances,¹³⁸ their lack of influence in maintaining high-frequency oscillations is consistent with their absence of myelination. Furthermore, the activity-dependence of myelination¹⁴⁴ might permit dynamically regulated influences on the fidelity of fast action potential transmission and high-frequency oscillations.

Furthermore, myelin could provide metabolic and trophic support for energetically costly PV cells. PV cell characteristics, including high-frequency spiking and rapid action potential kinetics, require a particularly high energy utilization through predominantly mitochondrial oxidative phosphorylation.¹⁴⁵ Gamma band synchrony, closely linked to cognition, is highly sensitive to metabolic disruption. Furthermore, compared with pyramidal cells, PV cells exhibit high densities of mitochondria and expression of cytochrome c and cytochrome c oxidase, proteins crucial for the electron transport chain. Moreover, PV cell-specific disruption of cytochrome oxidase assembly leads to changes in PV cell intrinsic excitability, afferent synaptic input, and gamma/theta oscillations, as well as schizophrenia-related behavioral impairments in sensory gating and social behavior.¹⁴⁶

During gamma oscillations, peak oxygen consumption approaches the demand observed during seizures and mitochondrial oxidative capacity operates near its functional limit.¹⁴⁵ Metabolic and trophic support conferred by myelination^{147, 148} might therefore allow PV axons to optimize their energy utilization. Consistent with the importance of myelination in regulating axonal energy metabolism is the considerable discrepancy of mitochondria content (30-fold) in myelinated versus unmyelinated axons.¹⁴⁹ Myelin has been proposed to regulate axonal energy metabolism via the monocarboxylate transporter 1 channel.^{147, 148} Furthermore, the high-peak oxygen consumption of PV cells during gamma band synchrony could require the additional lactate provided by oligodendrocytes.

Taken together, the electrophysiological dynamics of fast-spiking PV interneurons, their dense branching onto pyramidal neurons requiring finely tuned temporally synchronized inhibition, and their high-energy consumption are likely interdependent mechanisms governed by PV interneuron myelination.

Implications for schizophrenia

Both interneuron dysfunction and myelination abnormalities have been independently proposed as important contributors to the underlying pathophysiology of schizophrenia. These mechanisms have each amassed convincing support from postmortem histopathology, in vivo imaging and electrophysiology, genetics and neurodevelopment (Table 1). However, neither hypothesis is capable of accounting for the full set of clinical research findings in schizophrenia. In contrast, interneuron myelination brings together both of these models, explains a more comprehensive portion of the existing data, and offers a well-defined falsifiable model.

Impairments of PV interneuron myelination could directly contribute to schizophrenia through several mechanisms. Impaired action potential fidelity, energy restrictions during highly-demanding cognitive tasks, aberrant axonal branching and a higher occurrence of ectopic action potentials could each independently, or in combination, disrupt inhibitory network function. Such changes to PV interneurons would likely result in abnormalities of local gamma synchronization, with a potential further impact on the integrity of long-range thalamocortical and cortico-striatal circuits, and striatal dopamine signaling, ultimately giving rise to schizophrenia symptoms.

In this Perspective, we have proposed the novel hypothesis that altered myelination of PV interneurons might function prominently in the pathophysiology of schizophrenia. However, many questions remain to be answered. At what point during development does interneuron myelination occur and to what extent does this coincide with the clinical symptoms of schizophrenia? Does interneuron myelination vary across brain regions? Is cortical interneuron myelination truly reserved for fast-spiking PV interneurons, or are non-fast-spiking interneurons (e.g., somatostatin, VIP) myelinated as well? How does the plasticity of PV interneuron myelination compare with that of glutamatergic axons? And perhaps most importantly, to what extent might PV interneuron myelination represent an etiological pathophysiology and therapeutic target for schizophrenia?

Future studies to examine the parvalbumin interneuron myelination hypothesis could be approached through a variety of methods. In particular, the most important experiments would include: (a) detailed histological assessments of subtype-specific interneuron axonal myelination in postmortem brain tissue from patients with schizophrenia, (b) corresponding functional studies in rodent models of schizophrenia to directly assess the causality of alterations in parvalbumin interneuron myelination on behavioral and electrophysiological phenotypes, (c) electrophysiological studies of rodent models with temporally-controlled and cell-type specific disruption of myelination and (d) functional genomic studies on the effect of schizophrenia risk variants on (interneuron) myelination, for example, by utilizing human induced pluripotent stem cells or genetically modified mice.