Extracellular Matrix Remodeling with Stress and Depression: Studies in Human, Rodent, and Zebrafish Models

Abstract

Emerging evidence suggests that extracellular matrix (ECM) alterations occur with stress. Specifically, increases in perineuronal net (PNN) deposition have been observed in rodents exposed to chronic corticosterone or persistent social defeat stress. The PNN is a specific form of ECM that is predominantly localized to parvalbumin (PV)-expressing inhibitory interneurons where it modulates neuronal excitability and brain oscillations that are influenced by the same. Consistent with a role for ECM changes in contributing to the depressive phenotype, recent studies have demonstrated that monoamine reuptake inhibitor type antidepressants can reduce PNN deposition, improve behavior, and stimulate changes in gamma oscillatory power that may be important to mood and memory. The present review will highlight studies in humans, rodents, and zebrafish that have examined stress, PNN deposition and/or gamma oscillations with a focus on potential cellular and molecular underpinnings.

Graphical Abstract

Stress and cortisol increase PNN levels. The PNN is a specialized form of ECM that is predominantly localized to PV-expressing GABAergic interneurons. Because increased PNN levels can enhance PV-expressing neuron mediated inhibition of pyramidal cells, stress associated increases in PNN deposition likely reduce E/I balance and gamma power to negatively affect mood and memory. Successful antidepressant therapy can counteract stress-related changes in PNN levels.

Introduction

Major Depressive Disorder (MDD) is considered a major cause of disability throughout the world (Belmaker & Agam, 2008). Clinically, MDD is a heterogeneous disorder that is characterized by persistent mood instability, anhedonia, insomnia, loss of energy, and cognitive dysfunction (MacQueen & Frodl, 2011). In addition to these clinical symptoms, structural magnetic resonance imaging (MRI) scans of patients suffering from depression show reduced grey matter volume in areas of the brain including the hippocampus, insula, thalamus and nucleus accumbens (Joshi et al., 2015; Ancelin et al., 2019). Moreover, studies show that affected patients have decreased left and right hippocampal volume as compared to controls, and that the hippocampal CA1 region is particularly affected (Cole et al., 2010; Cole et al., 2011).

Although clinical symptoms and structural changes in MDD patients have been relatively well investigated, less is known about the molecular mechanisms that underlie these changes. Some studies suggest that the onset of depression can be triggered by exposure to traumatic life events or stressors (Holmes & Rahe, 1967; Daley et al., 2000; Schoenfeld et al., 2017); in fact, about 50% of individuals diagnosed with clinical depression or anxiety report having had early stressful life events (Li et al., 2016). An early publication linking stressful life events to the onset of MDD dates back to the 1990s. McGonagle and Kessler conducted 1,755 face-to-face interviews from nonblack married individuals to investigate whether chronic stressful life events were associated with the development of depressive episodes (McGonagle & Kessler, 1990). Consistent with an earlier study (Costello, 1982), this report suggested that exposure to chronic stressors was linked to a higher incidence of depression. Moreover, the more exposure an individual had to diverse stressors (e.g loss of a loved one, financial difficulties, etc), the more severe were the symptoms.

Exposure to stress activates the hypothalamic-pituitary-adrenal (HPA) axis which is linked to the release of cortisol in humans, and corticosterone in mice. In zebrafish, inter-renal cells of the kidney, the homologue of the mammalian adrenal gland, represent the final component of this axis (Alsop & Vijayan, 2009). Studies have shown that humans suffering from depression have higher cortisol levels as compared to controls or those in remission (Hoifodt et al., 2019). Moreover, lower levels of HPA activation have been associated with better working memory in individuals with MDD (Zobel et al., 2004). Cortisol or corticosterone receptors are highly expressed in the hippocampus and thus activation of the HPA axis has the potential to cause hippocampal injury. Consistent with this, chronic stress and glucocorticoids reduce hippocampal dendritic complexity in rodents models (Watanabe et al., 1992; Magarinos & McEwen, 1995b; Conrad et al., 1999; Kleen et al., 2006; McLaughlin et al., 2007). The CA3 region may be among the first of hippocampal areas to show dendritic restructuring (Woolley et al., 1990; Magarinos & McEwen, 1995a), which may be reversed after cessation of chronic stress or glucocorticoid treatment (Conrad et al., 1999). A reduction in the number and/or arborization of hippocampal pyramidal cells, which release the excitatory neurotransmitter glutamate, might in turn be expected to reduce overall excitatory to inhibitory (E/I) balance [reviewed in (Alaiyed & Conant, 2019) and (Thompson et al., 2015)].

Stress and corticosterone have also been linked to increased deposition of perineuronal net (PNN) proteins in rodent models related to depression (Riga et al., 2017) (Alaiyed et al., 2019b). PNNs are a form of extracellular matrix (ECM) that is predominantly localized to PV-expressing, GABA-releasing interneurons, and thought to influence their ability to fire (Balmer, 2016; Favuzzi et al., 2017; Hayani et al., 2018; Tewari et al., 2018). Studies at the single cell level suggest that PNN digestion with chondroitinase, or attenuation via peri-tumoral associated proteolysis, increases membrane capacitance of PV expressing neurons and concomitantly reduces their firing frequency by 50% (Tewari et al., 2018). Consistent with this, other studies have reported diminished PV output following chondroitinase treatment (Balmer, 2016) and reduced GABA-mediated inhibition of medial prefrontal cortex (mPFC) pyramidal neurons following the same (Slaker et al., 2015). PV-expressing neurons that are lacking brevican, a PNN component, also receive less excitatory input, a finding that may be related to the ability of the PNN and brevican to localize GluAs (Frischknecht et al., 2009; Favuzzi et al., 2017). Complicating the issue, however, is a study in which lower currents were required to elicit an action potential in PV-expressing neurons after PNN disruption (Dityatev et al., 2007), and a study showing that stress-related increases in PNN deposition were associated with decreased sIPSCs in CA1 pyramidal cells (Riga et al., 2017). Differences in described effects at the single cell level likely derive from differences in systems investigated (i.e. cell culture versus slice) and/or subtle changes in physiological stimuli.

Though predominantly localized to PV-expressing neurons, PNNs have also been observed to surround subgroups of pyramidal neurons in CA2. PNNs that surround these pyramidal cells impair their ability to undergo long term potentiation (LTP) (Carstens et al., 2016). Attenuation of PNNs that surround CA2 pyramidal cells might thus additionally be expected to increase E/I balance.

Despite the complexity of PNN effects at the single cell level, effects at the level of neuronal population activity suggest that overall, reduced PNN deposition may increase E/I balance. PNN attenuation enhances gamma power (Lensjo et al., 2017; Alaiyed et al., 2019a; Alaiyed et al., 2020), an endpoint that is sensitive to altered E/I balance. Optogenetic excitation of pyramidal cells increases gamma power (Yizhar et al., 2011), as does reduced glutamatergic excitation of PV-positive interneurons (Racz et al., 2009). Moreover, in animal models of conditions such as fragile X syndrome and schizophrenia, reduced PNN levels are reported in conjunction with increases in E/I balance (Hirano et al., 2015; Lovelace et al., 2018; Dwir et al., 2019).

This review will discuss molecular players that may be important to reduced E/I balance in the background of stress and depression, and potential mediators of treatment-associated restoration of this balance. We will highlight the potential relationship between increased ECM/PNN deposition and altered neuronal population dynamics, including gamma oscillations and sharp wave ripples (SWRs), which are important to mood and memory. The discussion will additionally address what has been or may be learned not only from human and rodent, but also from zebrafish models.

I. Stress and depression induce changes in PNN core proteins

A recent study focused on deciphering the molecular mechanisms that lead to the onset of depressive episodes by looking at a stress paradigm in rats and examining specific PNN constituent changes. This study utilized a social defeat-induced persistent stress (SDPS) rat model in which rats were exposed to chronic stress and single-housed for a period of 2–3 months (Riga et al., 2017). Similar to human patients suffering from MDD, stressed rats showed impaired cognitive processing and synaptic plasticity, including reduced LTP in hippocampal CA1 pyramidal neurons. Interestingly, mass spectrometry analysis of the proteome of the dorsal hippocampus of these rats showed an increase in synaptic chondroitin sulfate proteoglycans including brevican and neurocan, which are critical components of the PNN (Riga et al., 2017; Bozzelli et al., 2018). Digestion of the PNN with chondroitinase improved plasticity in the hippocampus and rescued short-term spatial memory deficits (Riga et al., 2017).

In related work using mice, it has been shown that mRNA levels for PNN components are increased 4 hours after fear conditioning (Banerjee et al., 2017). In addition, mice exposed to chronic corticosterone (4 weeks) show increased PNN deposition surrounding fast spiking interneurons and increased levels of hippocampal brevican (Alaiyed et al., 2020). And though studies of stress-related changes in the human PNN are relatively lacking, post-mortem samples from bipolar patients show a change in the sulfation pattern of PNN components that influences their resistance to proteolysis (Pantazopoulos et al., 2015).

Stress paradigms used in mice, including unpredictable chronic stress exposure (UCS) (Ito et al., 2010), have recently been adapted for zebrafish (Piato et al., 2011). Though these models have not been used to explore neuronal plasticity in the zebrafish brain, the UCS model has been used to investigate memory, as well as physiological changes in CRF, cortisol and glucocorticoid receptor (GR) expression (Piato et al., 2011). Memory is impaired in stressed fish, and these fish show increased levels of both CRF and cortisol. mRNA levels of GRs are decreased in stressed zebrafish (Piato et al., 2011), a finding consistent with a previously reported negative correlation between cortisol and GR-expressing neurons in the murine hippocampus (Sapolsky et al., 2000). In other studies that have examined the relationship between cortisol and behavioral measures in zebrafish, an acute restraint stress model has been used (Ghisleni et al., 2012). Restraint-stress exposed fish show elevated cortisol levels as compared to controls and they spend more time swimming in the bottom portion of the tank. In addition, a longer latency to explore the upper portion of a novel tank has been linked to anxiety in this species (Egan et al., 2009)

Though a relationship between stress and PNN component expression in zebrafish has yet to be examined, zebrafish express critical PNN components. Chondroitin sulfate proteoglycans, and possibly neurocan and hyaluronan and proteoglycan link protein 1, are expressed in the zebrafish head during development (Becker & Becker, 2002; Kang et al., 2008). Zebrafish also express matrix metalloproteinase (MMP) paralogues known to remodel the PNN in rodents (Pedersen et al., 2015). For example, MMP-9, an MMP that has been particularly well-implicated in human and rodent pyramidal cell neuroplasticity and PNN remodeling (Nagy et al., 2006; Bach et al., 2018; Alaiyed & Conant, 2019a; Bach et al., 2019), is expressed in zebrafish brain.

II. Antidepressants and PNN modulation

Of relevance to stress-related increases in PNN deposition and potential effects on mood, are mouse and rat studies linking reduced PNN levels to antidepressant efficacy (Riga et al., 2017). Both venlafaxine and ketamine can reduce PNN levels in rodent models (Matuszko et al., 2017; Alaiyed et al., 2019a). Effects of venlafaxine depend in part on its ability to increase MMP-9 levels (Alaiyed et al., 2020), which may follow from increased activation of specific monoamine receptor subtypes (Bijata et al., 2017; Alaiyed et al., 2020). And though ketamine may more rapidly influence mood, likely through its ability to reduce the activity of PV-expressing interneurons by targeting GluNs on this cell population (Homayoun & Moghaddam, 2007; Picard et al., 2019), it is tempting to speculate that PNN attenuation could contribute to the slowly evolving effects of this compound. It should be noted, however, that excessive PNN attenuation could be maladaptive. For example, mice deficient for the PNN component neurocan display manic-like behavior (Miro et al., 2012).

The question of whether antidepressants reduce ECM deposition in zebrafish has yet to be examined. Similar to mammals, however, zebrafish have fully functional serotonin, norepinephrine, GABAergic, glutamatergic and dopaminergic systems [summarized in (Rico et al., 2011)]. Zebrafish also show behavioral responsiveness to antidepressant administration. For example, the ability of antidepressants (fluoxetine and nortriptyline) to prevent unpredictable stress-associated changes in cortisol levels and behavior has been demonstrated (Marcon et al., 2016). Moreover, antidepressants and mood stabilizers including fluoxetine (Abreu et al., 2014; Idalencio et al., 2015; Giacomini et al., 2016a; Giacomini et al., 2016b) have now been shown to reverse stress-related behavioral changes in varied zebrafish models. In addition, a recent paper demonstrates that the serotonin receptor 1A partial agonist buspirone shapes stress-dependent behavioral reactivity in zebrafish (Varga et al., 2020).

Of interest, zebrafish express sulfotransferases that may influence the sensitivity of PNN components to proteolysis (Habicher et al., 2015; Sahu et al., 2019). Knockdown of chondroitin-4-sulfotransferase in zebrafish, which generates neuroplasticity-restricting 4-sulphated chondroitin sulfate-glycosaminoglycans in rodents (Foscarin et al., 2017), accelerates regeneration following spinal cord injury (Sahu et al., 2019). This result is consistent with conserved mechanisms by which ECM affects neuroplasticity in diverse organisms. The availability of tools to manipulate cell-specific effector gene expression in zebrafish, such as the Gal4-UAS system (Kawakami et al., 2016), should allow for PNN components such as brevican to be expressed in neurons or glial cells in this model organism. Future studies could then evaluate PNN component effects on depression-related behaviors.

III. PNN levels modulate the activity of PV-expressing interneurons through varied mechanisms: relevance to E/I balance

There are multiple mechanisms by which the PNN facilitates firing of PV-expressing interneurons. In a recent study, it was shown that the PNN can reduce membrane capacitance of PV-expressing cells to allow better high frequency spiking (Tewari et al., 2018). Work from other investigators has demonstrated that an intact PNN can limit diffusion of GluAs along the PV-expressing cell membrane (Frischknecht et al., 2009), and thus better localize GluAs so that they can respond to released glutamate. In related work from the Rico group, deletion of the PNN component brevican was shown to reduce synaptic input to PV-expressing neurons and the expression of GluA1 in this cell type (Favuzzi et al., 2017). Since each PV-expressing interneuron contacts multiple pyramidal neurons (Andersen, 2006), a reduction in GABA release from this cell type might profoundly influence large scale E/I balance, an endpoint relevant to the depressive phenotype (Thompson et al., 2015; Page & Coutellier, 2019). Consistent with this, earlier work with mice deficient in the PNN component tenascin-R showed that these animals had elevated basal excitatory synaptic transmission and reduced perisomatic GABAergic inhibition (Bukalo et al., 2007). Moreover, recent work in cultured neurons showed that elimination of four PNN components was associated with an increase in excitatory synaptic components (Gottschling et al., 2019). Additionally consistent with reduced PV neuron-mediated inhibition of pyramidal cells in the setting of PNN attenuation, chondroitinase mediated PNN digestion increased c-fos expression in pyramidal cells of the insula during aversion-resistant self-administration of ethanol (Chen & Lasek, 2019). Increased E/I balance in the visual cortex has also been observed with environmental enrichment-associated reductions in PNN levels (Sale et al., 2007).

IV. E/I balance may influence neuronal population dynamics, including gamma power, which are important to mood and memory

It has been posited that a change in E/I balance can influence neuronal population dynamics (Bozzelli et al., 2018). In an early study using optogenetic techniques, activation of PFC pyramidal cells increased gamma frequency power (Yizhar et al., 2011). This is consistent with a link between the amplitude of sharp waves, which represent pyramidal cell depolarization, and gamma power (Sullivan et al., 2011). In contrast, selective activation of PV-expressing neurons decreases E/I balance (Yizhar et al., 2011). In another study, which expressed channel rhodopsin-2 in excitatory neurons of specific layers in the murine visual cortex, light stimulation generated 20–80 Hz gamma oscillations that propagated along the vertical axis. This study also suggested that while recurrent inhibition and a low E/I balance may be important to network stabilization within cortical layers (Ozeki et al., 2009), an E/I balance tilted more towards excitation, as was observed in upper cortical layers, might promote propagation of activity during gamma (Adesnik, 2018). And though discussion of E/I balance and gamma activity in a non-layer or non-location specific manner likely misses the nuanced nature of potential changes in the intact organism, there is indirect or correlative support for the possibility that relatively large scale increases in E/I balance, as affected by widespread reductions in perisomatic inhibition (Bukalo et al., 2007; Favuzzi et al., 2017), can enhance gamma activity (Lensjo et al., 2017; Alaiyed et al., 2019a). Ketamine, a rapid-acting antidepressant that stimulates pyramidal cell disinhibition by acting as a preferential antagonist for GluNs localized to PV expressing neurons (Homayoun & Moghaddam, 2007; Picard et al., 2019), has been shown to enhance gamma power (Ye et al., 2018; Manduca et al., 2020). Moreover, if the PNN components tenascin-C or –R act in a similar manner to brevican in terms of enhancing PV excitability, it is worth noting that mice deficient in these components additionally demonstrate increased gamma power (Gurevicius et al., 2004; Gurevicius et al., 2009).

Increased E/I balance in the setting of PNN attenuation may additionally influence the abundance of SWRs, events in which sequential replay of learning-associated neuronal assemblies occurs in a time-compressed manner to facilitate memory retrieval and consolidation (Ego-Stengel & Wilson, 2010; Buzsaki, 2015). Indeed, PNN disruption with hyaluronidase or chondroitinase has been shown to increase the abundance of hippocampal SWRs in murine hippocampus (Sun et al., 2018).

Since increased PNN deposition occurs in rodent models associated with chronic stress (Alaiyed et al., 2019b) (Riga et al., 2017), we recently examined the effects of corticosterone on SWR abundance and ex vivo gamma power. Consistent with a link between increased PNN deposition and altered neuronal population dynamics, corticosterone treated mice showed reduced SWR abundance, reduced gamma oscillation power and reduced working memory (Alaiyed et al., 2020). Of interest, the serotonin-norepinephrine reuptake inhibitor venlafaxine reversed corticosterone-associated increases in the PNN as well as SWR abundance, gamma power and working memory (Alaiyed et al., 2020).

In a related study of mice that were exposed to chronic restraint stress, restoration of gamma activity was observed in animals with spontaneous remission of depressive symptoms (Khalid et al., 2016). In yet an additional murine study, deficits in gamma activity were inversely correlated with behavioral despair (Sauer et al., 2015). Gamma oscillations have also been monitored during social interactions in the nucleus accumbens of rats that were exposed to chronic social defeat. This study observed a significant reduction of power increase in high gamma activity (61–90 Hz) for stress-exposed rats as compared to the non-stressed group (Iturra-Mena et al., 2019).

Human studies have examined gamma oscillations in the background of depression and anxiety, though there is substantial heterogeneity in terms of whether oscillatory activity was assessed during specific tasks or the resting state. In one electroencephalogram study, subjects with high scores on depression and anxiety questionnaires had reduced resting gamma in the anterior cingulate cortex (Pizzagalli et al., 2006). In another study, subjects suffering from depression showed reduced frontal cortex gamma when performing emotion-related tasks (Strelets et al., 2007). Moreover, increased gamma frequency synchronization has been observed with paroxetine, a selective serotonin reuptake inhibitor antidepressant, as well as with improvements on the Hamilton Depression scale (Arikan et al., 2018). A schematic overview of the potential effects of stress on E/I balance is shown in figure 1. Indeed, though human and animal studies of gamma oscillations and cognition often focus on phase amplitude coupling between gamma and theta, regional changes, and resting versus task-relevant changes in power, increases in gamma activity are generally linked to improved attention and short term memory (Montgomery & Buzsaki, 2007; Lisman, 2010; Magazzini & Singh, 2018). Both of these cognitive domains may be impaired in MDD (MacQueen & Frodl, 2011).

Figure 1.

To date, studies of stress and changes in PNN deposition or neuronal population dynamics have been performed in rodents (Khalid et al., 2016; Riga et al., 2017). Stress (middle panel) has been linked to increased PNN deposition (Riga et al., 2017) and reduced complexity of pyramidal dendrites, which in turn may reduce E/I balance to diminish the power of gamma oscillations (Khalid et al., 2016). PNN changes may also influence the abundance of SWRs (Sun et al., 2018). Antidepressant therapy (lower-most panel) may attenuate stress-related PNN deposition through its ability to increase MMP-9 levels (Tamasi et al., 2014; Alaiyed et al., 2019a), restore E/I balance, and normalize gamma power (Alaiyed et al., 2019a; Alaiyed et al., 2020).

The question of whether neuronal population dynamics are affected by stress or antidepressants in zebrafish has not been well investigated. Sharp waves (SWs) have, however, been detected in the zebrafish hippocampal homologue, the anterodorsal lobe of the lateral telencephalic pallium (Vargas et al., 2012). In preliminary work, we have also detected SW events in this region as shown in figure 2, and we hope to address the effects of chronic stress and potential antidepressant medications on SW frequency in future studies.

Figure 2.

SWR events recorded from the anterodorsal lobe of a 35 day old wild-type (Ekkwill strain) zebrafish following tricaine anaesthesia, decapitation, brain extraction and 2h recovery as described (Vargas et al., 2012). A Traces shown are consecutive local field potential (LFP) recordings. SWR events with low amplitude (red arrowheads) and high amplitude (blue arrowheads) occur spontaneously. B. One large SWR event (marked by a broken line box in A) is displayed with an expanded scale. Top trace: filtered between 0.1–200 Hz, Bottom trace, filtered between 20–200Hz to show gamma and ripple oscillations. C. Power spectrum of the SWR event (black line). In red we show the power spectrum of background noise, 1 second following the SWR event. During the SWR, power peaks in low gamma (20–39), high gamma (50–100), and ripples (120Hz) can be seen in black, with signal above that of background noise.

Summary and Future Directions

Human, rodent, and zebrafish studies suggest that stress is a potential contributor to the depressed phenotype. Moreover, stress-related changes in ECM expression and associated alterations in neuronal population activity may contribute to cognitive and affective changes seen in individuals suffering from depression. Future studies will be important to tease out the contribution of stress-upregulated corticosteroids to PNN component expression in specific glial cells and neurons, and to also examine PNN stability and proteolytic remodeling in depression models. Additional in vivo recordings of neuronal population activity in rodent brain, as a function of stress and antidepressant therapy, may additionally improve our understanding of the depressed phenotype and antidepressant efficacy. In particular, the use of multi-electrode recordings will be necessary to answer questions such as whether memory-relevant sequential assembly replay occurs during venlafaxine-enhanced SWRs. Finally, further work in zebrafish may allow for relatively rapid screening of antidepressants to assess their ability to influence ECM, E/I balance and/or behavioral alterations that occur with stress. Zebrafish represent an ideal model organism for in vivo imaging of the entire central nervous system, providing a window into large scale monitoring of neuronal activity relevant to depression-related changes in neuroplasticity.

Abbreviations:

ECM

extracellular matrix

PNN

perineuronal net

PV

parvalbumin

MDD

major depressive disorder

MMP

matrix metalloproteinase

HPA

hypothalamic-pituitary-adrenal

E/I

excitatory to inhibitory

SWR

sharp wave ripple

CRF

corticotropin-releasing factor

GR

glucocorticoid receptor

GluN

N-methyl-D-aspartate type glutamate receptor

GABA

Gamma aminobutyric acid

UAS

upstream activation sequence