Thehttp://nro.sagepub.com/ Neuroscientist
Metaplasticity in Human Cortex Florian Müller-Dahlhaus and Ulf Ziemann Neuroscientist published online 11 March 2014 DOI: 10.1177/1073858414526645 The online version of this article can be found at: http://nro.sagepub.com/content/early/2014/03/11/1073858414526645
Published by: http://www.sagepublications.com
Additional services and information for The Neuroscientist can be found at: Email Alerts: http://nro.sagepub.com/cgi/alerts Subscriptions: http://nro.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav
>> OnlineFirst Version of Record - Mar 11, 2014 What is This?
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
526645
research-article2014
NROXXX10.1177/1073858414526645The NeuroscientistMüller-Dahlhaus and Ziemann
Article
Metaplasticity in Human Cortex
The Neuroscientist 1–18 © The Author(s) 2014 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1073858414526645 nro.sagepub.com
Florian Müller-Dahlhaus1 and Ulf Ziemann1
Abstract Metaplasticity refers to the modification of plasticity induction (direction, magnitude, duration) by previous activity of the same postsynaptic neuron or neuronal network. In recent years evidence from animal studies has been accumulated that metaplasticity significantly contributes to network function and behavior. Here, we review the evidence for metaplasticity at the system level of the human cortex as investigated by non-invasive brain stimulation. These studies support the notion that metaplasticity is also operative in the human brain and is mostly homeostatic in nature, that is, keeping network activity within a physiological range. However, non-homeostatic metaplasticity has also been described, which can increase non-invasive brain stimulation–induced aftereffects on cortical excitability, or learning. Current evidence further suggests that aberrant metaplasticity may underlie some neurological and psychiatric diseases. Finally, first proof-of-principle studies show that the concept of metaplasticity can be harnessed for treatment of patients suffering from brain diseases. Keywords metaplasticity, non-invasive brain stimulation, transcranial magnetic stimulation, transcranial direct current stimulation, motor learning, motor cortex, long-term potentiation, long-term depression
Introduction The human brain has an amazing capacity to reorganize its structure and function to adapt to an ever-changing environment. It is now widely accepted that synaptic plasticity plays a critical role in processes of learning and memory. Various forms of synaptic plasticity have been described in the last decades (Nelson and Turrigiano 2008; Sjöström and others 2001), and it has become clear more recently that Hebbian-type synaptic plasticity, that is, long-term potentiation (LTP) and depression (LTD) is highly regulated itself. Metaplasticity is one prominent concept of activity-dependent plasticity regulation and encompasses changes in the synaptic and/or neuronal state, which shape the direction, magnitude and duration of future synaptic changes (Abraham and Bear 1996). Importantly, this modulation of the ability of neurons to express synaptic plasticity can occur in the absence of changes in the excitability of the stimulated network per se. Here, we review recent evidence for metaplasticity and its contribution to network function and behavior in the human brain. This evidence is derived from studies using non-invasive brain stimulation (NIBS), that is, transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS). TMS and tDCS are capable of inducing long-lasting changes in cortical excitability and connectivity in the awake, non-anaesthetized human brain. Although the cellular and molecular mechanisms of these LTP- and LTDlike effects are not well understood (Müller-Dahlhaus
and Vlachos 2013), first experimental evidence from in vitro studies suggests that repetitive magnetic stimulation (Vlachos and others 2012) and direct current stimulation (Fritsch and others 2010) can induce N-methyl-d-aspartate (NMDA)-receptor (NMDAR)- and brain-derived neurotrophic factor (BDNF)-dependent LTP, respectively. In humans, metaplasticity was tested by studying interactions between successive NIBS protocols, or between NIBS and behavioral interventions such as motor learning. These studies show that metaplasticity is operative on the level of the human cerebral cortex. They also demonstrate that aberrant metaplasticity may underlie certain neurological and psychiatric disorders. On a translational note, first promising studies suggest that metaplasticity can be successfully harnessed to modulate, or even restore, the ability of neural plasticity under pathological conditions.
Basic Concepts of Metaplasticity The basic idea of metaplasticity is that the threshold for activity-dependent synaptic plasticity is not static but 1
Department of Neurology and Stroke, Eberhard-Karls University Tübingen, Tübingen, Germany Corresponding Author: Ulf Ziemann, Department of Neurology and Stroke, and Hertie Institute for Clinical Brain Research, Eberhard-Karls University Tübingen, Hoppe-Seyler-Straße 3, Tübingen, 72076, Germany. Email: [email protected]
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
2
The Neuroscientist
Figure 1. Concept of metaplasticity. Please note that metaplasticity, that is, the modulation of Hebbian-type plasticity by prior neural activity (priming) can occur in the absence of changes in the excitability of the stimulated neuron/network per se. Red, metaplasticity; blue, induction of Hebbian plasticity in the naïve cerebral cortex; purple, expression of Hebbian plasticity under conditions of metaplasticity. For further explanations, please see section “Basic Concepts of Metaplasticity.”
dynamic, and a function of the integrated prior activity of the postsynaptic neuron (Abraham and Bear 1996) (Fig. 1). The time of metaplasticity expression in most studies was in the order of minutes to hours, but can last for days (Abraham and others 2001) and even weeks (Buschler and Manahan-Vaughan 2012). In addition, metaplasticity can be expressed over a broad range of spatial extents of the postsynaptic neuron. Whereas heterosynaptic metaplasticity can be global, that is, cell-wide, or confined to specific dendritic compartments, homosynaptic metaplasticity is only expressed at those synapses that participated in the initial bout of priming activity. Heterosynaptic metaplasticity is due to dendritic excitability changes (Harvey and Svoboda 2007) or modulation of synaptic tagging and capture mechanisms of plasticity-related proteins (Li and others 2014; Sajikumar and others 2009). It may also involve intercellular communication, possibly mediated by astrocytes (for a recent review, see Hulme and others 2014). In homosynaptic metaplasticity, group 1 metabotropic glutamate receptors (mGluRs) mediate facilitation and prolongation of subsequent LTP (Raymond and others 2000), whereas activation of NMDARs inhibits subsequent LTP (Huang and others 1992) and facilitates subsequent LTD (Christie and Abraham 1992). Of note, metaplasticity has been mostly described for excitatory synaptic neurotransmission, but recent evidence suggests that it may also account for plasticity regulation at inhibitory synapses (Edwards and others 2008; Fischer and others 1997). Metaplasticity thus regulates synaptic plasticity across space and time (for a recent review, see Hulme and others 2013). It serves to prolong the time window for associative interactions between neural events and may therefore underlie increased information encoding during repeated, spaced learning trials. However, it also ensures neuronal
and network stability, for example, by providing resistance against subsequent learning and thus enabling memory retention after periods of increased information encoding. The term homeostatic metaplasticity describes this neuroprotective effect to stabilize synaptic weights in neuronal networks while maintaining the capacity for synaptic plasticity. Homeostatic metaplasticity, as formalized in the Bienenstock–Cooper–Munro (BCM) theory of bidirectional synaptic plasticity (Bienenstock and others 1982), states that the synaptic modification threshold θM, that is, the threshold for induction of LTP versus LTD, is not stable but varies as a function of the integrated postsynaptic activity: it decreases at low levels of previous postsynaptic activity, favoring induction of LTP over LTD. Conversely, θM increases at high levels of recent postsynaptic activity, thereby favoring the probability of subsequent LTD over LTP (Fig. 2M). This sliding synaptic modification threshold thus enables maintenance of neuronal and network activity in a physiological range. Induction of BCM-like metaplasticity involves multiple mechanisms including NMDAR activation (Abraham and others 2001), and is expressed by changes in NMDAR subunit composition (Philpot and others 2007). More recently, modeling studies suggested an additional role for non-NMDAR mechanisms in the expression of metaplasticity such as changes in voltagegated ion channels (Narayanan and Johnston 2010). BCM-like homeostatic metaplasticity typically accounts for cell-wide, that is, heterosynaptic metaplasticity, but synapse-specific, that is, homosynaptic homeostatic metaplasticity has also been reported (Huang and others 1992).
Evidence for Metaplasticity in Human Cortex Evidence from NIBS-NIBS Studies Priming and Test Stimulation Applied to Primary Motor Cortex (M1). The most elaborate evidence for homeostatic metaplasticity in the human brain comes from a study by Hamada et al. (2008) who tested plasticity of corticospinal neurons in the primary motor cortex (M1), as indexed by changes in the amplitude of motor evoked potentials (MEPs) evoked by single-pulse TMS. They applied repeated quadro-pulse stimulation (QPS), that is, quadruplets were delivered every 5 seconds over a period of 30 minutes (Fig. 2A). Short interstimulus intervals between the quadruplet pulses of 1.5, 5, or 10 ms resulted in a long-lasting (>30 minutes) increase of MEP amplitude, that is, an LTP-like effect (Figs. 2B-D), whereas longer interstimulus intervals of 30, 50, or 100 ms resulted in an LTD-like decrease (Figs. 2E-G) (Hamada and others 2008). These QPS protocols were primed by a short
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
3
Müller-Dahlhaus and Ziemann
Figure 2. Homeostatic metaplasticity in a NIBS-NIBS study. (A) Stimulation protocol. Quadruplets (four TMS pulses) are applied in bursts of varying interstimulus intervals (1.5, 5, 10, 30, 50, 100 ms) every 5 seconds (0.2 Hz) over a period of 30 minutes (360 trains). This protocol is referred to as quadro-pulse stimulation (QPS) (Hamada and others 2007). (B-G) QPS-1.5ms, QPS-5ms, and QPS-10ms result in long-lasting (>30 minutes) increases in MEP amplitude, while QPS-30ms, QPS-50ms, and QPS-100ms result in a long-lasting MEP amplitude decrease. (H) A short (10 minutes) QPS-5ms protocol does not result in overt change in MEP amplitude and is used as priming protocol to test metaplasticity in the subsequent panels. (I, J) Priming with QPS-5ms leads to a shift of the LTP-like MEP increase induced by QPS-10ms (white circles in I) toward an LTD-like MEP decrease (black circles), and to an increase of the LTD-like MEP decrease induced by QPS-30ms (white and black circles in J without and with priming, respectively). (K, L) Priming with short (10 minutes) QPS-50ms leads to an increase of the LTP-like MEP increase induced by QPS-10ms (white and black circles in K without and with priming, respectively) and a shift of the LTD-like MEP decrease induced by QPS-30ms (white circles in L) to an LTP-like MEP increase (black circles). (M) Schematic summary of the results of priming with QPS-5ms (blue curve) versus priming with QPS-50ms (red curve) on the LTP threshold θM. Note the rightward shift of θM when compared with the non-primed curve (black curve) with QPS-5ms priming, that is, probability of LTP induction decreases, and the leftward shift with QPS-50ms priming, that is, probability of LTP induction increases (modified from Hamada and others 2008, with permission). LTD = long-term depression; LTP = long-term potentiation; MEP = motor-evoked potential; NIBS = noninvasive brain stimulation; TMS = transcranial magnetic stimulation
QPS-5ms protocol that, when given alone did not result in any overt change of MEP amplitude, that is plasticity of corticospinal neurons (Fig. 2H), but presumably increased activity in the stimulated network. This QPS5ms priming abolished the LTP-like plasticity induced by QPS-1.5ms and QPS-5ms (not shown) and even shifted the LTP-like effect induced by QPS-10ms toward MEP depression (Fig. 2I) (Hamada and others 2008). In addition, priming with short QPS-5ms increased the LTD-like effect induced by QPS-30ms (Fig. 2J) and resulted in trends toward increases of the LTD-effects induced by QPS-50ms and QPS-100ms (Hamada and others 2008). In contrast, short QPS-50ms priming, which did not result in MEP amplitude change when given alone (baseline 1
measurements in Figs. 2K and L), but presumably decreased activity in the stimulated network, increased the LTP-like plasticity induced by subsequent QPS-10ms (Fig. 2K) and shifted the LTD-like plasticity induced by QPS-30ms (Fig. 2L) and QPS-50ms toward an LTP-like MEP increase (Hamada and others 2008). Altogether, these data are fully compatible with homeostatic metaplasticity according to the BCM theory (Bienenstock and others 1982) as reviewed above: priming with short QPS5ms (i.e., high-frequency priming according to the short interstimulus interval of 5 ms in the quadruplet) resulted in a rightward shift of the threshold for induction of LTPlike plasticity, that is, an enhanced probability of LTDlike plasticity and decreased probability of LTP-like
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
4
The Neuroscientist
plasticity (blue curve in Fig. 2M). Vice versa, priming with short QPS-50ms (that is, low-frequency priming according to the long interstimulus interval of 50 ms in the quadruplet) resulted in a leftward shift of the threshold for induction of LTP-like plasticity, that is, decreased probability of LTD-like plasticity and increased probability of LTP-like plasticity (red curve in Fig. 2M). Similar observations in accord with the BCM theory have been made in a number of other studies (synopsis in Table 1). For example, low-frequency (1Hz) regular repetitive TMS (rTMS) typically results in LTD-like MEP amplitude decrease when given alone (Chen and others 1997). However, priming of 1-Hz rTMS with highfrequency (6 Hz) rTMS resulted in a stronger LTD-like MEP decrease compared with 1-Hz rTMS alone (Iyer and others 2003). A shortcoming of that study was that the effects of the 6-Hz priming protocol, when given alone, on MEP amplitude had not been tested. Therefore, it remained unclear to what extent any possible MEP change by the priming protocol has contributed to this enhancement of LTD-like plasticity by subsequent 1Hz rTMS. Furthermore, interactions of 6Hz rTMS priming with subsequent induction of LTP-like plasticity had not been tested. Paired associative stimulation (PAS) induces LTP-like or LTD-like changes in MEP amplitude (PASLTP, PASLTD), depending on the interstimulus interval between the electrical stimulation of a peripheral nerve and associated single-pulse TMS of the contralateral M1 (MüllerDahlhaus and others 2010; Stefan and others 2000; Wolters and others 2003). PAS has gained some interest because several features resemble those of spike-timingdependent plasticity as studied on a cellular level (MüllerDahlhaus and others 2010). The LTP-like increase in MEP amplitude induced by PASLTP was occluded by priming with PASLTP, while priming with PASLTD led to a trend toward stronger LTP-like plasticity by the subsequent PASLTP protocol when compared to a priming control condition (Müller and others 2007). Furthermore, effects on MEP amplitude induced by primed PASLTP were inversely correlated with those induced by priming PAS (Müller and others 2007). Because the two PAS protocols likely activated the same neuronal circuits, it was concluded that “homosynaptic-like” homeostatic metaplasticity in accord with the BCM theory underlay the observed PAS interactions. Theta-burst stimulation (TBS) induces an LTP-like increase or LTD-like decrease in MEP amplitude, depending on the pattern of stimulation: intermittent TBS (iTBS) results in LTP-like plasticity, continuous TBS (cTBS) leads to LTD-like plasticity (Huang and others 2005). Priming cTBS with iTBS resulted in enhanced LTD-like plasticity (Doeltgen and Ridding 2011). These data confirmed findings from one previous study (Todd and others 2009) that,
in addition, demonstrated that 2- and 6-Hz priming had no effect on the LTD-like plasticity effect induced by cTBS. The interactions between two successive TBS protocols were investigated more extensively in another study that showed that pairing of identical protocols (iTBS → iTBS, cTBS → cTBS) resulted in suppression of the non-primed TBS effects on MEP amplitude, while pairing of different protocols (cTBS → iTBS, iTBS → cTBS) enhanced the non-primed TBS effects on MEP amplitude (Murakami and others 2012). These results are in full compliance with homeostatic metaplasticity according to the BCM theory. The study by Murakami and colleagues also investigated for the first time systematically metaplasticity of inhibitory circuits by utilizing the fact that iTBS and cTBS lead to long-term increases and decreases in GABAAergic inhibition, respectively, as indexed by the paired-pulse TMS measure of short-interval intracortical inhibition (SICI; Huang and others 2005). While non-primed TBS was subthreshold for inducing significant changes of SICI, iTBS → iTBS resulted in SICI decrease, and cTBS → cTBS in SICI increase compared with the non-primed conditions (Murakami and others 2012). Importantly, the changes in SICI induced by priming TBS correlated with the changes in MEP induced by subsequent test TBS. These findings indicate that plasticity in M1 inhibitory circuits is also regulated by homeostatic metaplasticity, and that priming effects on GABAergic inhibition contribute to the homeostatic regulation of metaplasticity in excitatory circuits (Murakami and others 2012). Another study investigated the effects of the interval between two identical TBS protocols (iTBS → iTBS or cTBS → cTBS) (Gamboa and others 2011). The interactions were by and large homeostatic, with one notable exception: cTBS → cTBS resulted in an enhancement (rather than reduction or occlusion) of LTDlike plasticity with the longest tested interval of 20 minutes (Gamboa and others 2011). While the reasons for this latter non-homeostatic interaction remained unclear, a candidate mechanism might be consolidation of LTD through increases in de novo protein synthesis and gene transcription, as has been demonstrated in hippocampal slices for repeated spaced LTP-induction protocols (Nguyen and others 1994). Consolidation of LTP by protein synthesis and gene transcription is associated with resistance to depotentiation (Woo and Nguyen 2003). In accord with consolidation of synaptic plasticity, recent cTBS → cTBS experiments (10-minute interval between priming and test cTBS) demonstrated a non-homeostatic interaction with significant lengthening of the LTD-like MEP decrease >120 minutes (Goldsworthy and others 2012) that was resistant to de-depression by voluntary contraction or short iTBS (150 pulses) (Goldsworthy and others 2014). In one other study (Mastroeni and others 2013) the effects of polymorphisms of the BDNF gene (val66met vs. val66val) were investigated, since previous studies have shown,
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
5
Müller-Dahlhaus and Ziemann Table 1. Synopsis of NIBS-NIBS Studies to Test Metaplasticity in Human Cortex. Site of Priming/ Test Stimulation M1/M1 M1/M1 M1/M1 M1/M1
M1/M1
M1/M1
M1/M1 M1/M1
M1/M1
M1/M1 M1/M1
M1 / M1
Priming Stimulation
Test Stimulation
Effects on Test Stimulation
Reference
6Hz rTMS, 90%RMT, 1-Hz rTMS, 115% RMT,
Metaplasticity in Human Cortex Florian Müller-Dahlhaus and Ulf Ziemann Neuroscientist published online 11 March 2014 DOI: 10.1177/1073858414526645 The online version of this article can be found at: http://nro.sagepub.com/content/early/2014/03/11/1073858414526645
Published by: http://www.sagepublications.com
Additional services and information for The Neuroscientist can be found at: Email Alerts: http://nro.sagepub.com/cgi/alerts Subscriptions: http://nro.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav
>> OnlineFirst Version of Record - Mar 11, 2014 What is This?
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
526645
research-article2014
NROXXX10.1177/1073858414526645The NeuroscientistMüller-Dahlhaus and Ziemann
Article
Metaplasticity in Human Cortex
The Neuroscientist 1–18 © The Author(s) 2014 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1073858414526645 nro.sagepub.com
Florian Müller-Dahlhaus1 and Ulf Ziemann1
Abstract Metaplasticity refers to the modification of plasticity induction (direction, magnitude, duration) by previous activity of the same postsynaptic neuron or neuronal network. In recent years evidence from animal studies has been accumulated that metaplasticity significantly contributes to network function and behavior. Here, we review the evidence for metaplasticity at the system level of the human cortex as investigated by non-invasive brain stimulation. These studies support the notion that metaplasticity is also operative in the human brain and is mostly homeostatic in nature, that is, keeping network activity within a physiological range. However, non-homeostatic metaplasticity has also been described, which can increase non-invasive brain stimulation–induced aftereffects on cortical excitability, or learning. Current evidence further suggests that aberrant metaplasticity may underlie some neurological and psychiatric diseases. Finally, first proof-of-principle studies show that the concept of metaplasticity can be harnessed for treatment of patients suffering from brain diseases. Keywords metaplasticity, non-invasive brain stimulation, transcranial magnetic stimulation, transcranial direct current stimulation, motor learning, motor cortex, long-term potentiation, long-term depression
Introduction The human brain has an amazing capacity to reorganize its structure and function to adapt to an ever-changing environment. It is now widely accepted that synaptic plasticity plays a critical role in processes of learning and memory. Various forms of synaptic plasticity have been described in the last decades (Nelson and Turrigiano 2008; Sjöström and others 2001), and it has become clear more recently that Hebbian-type synaptic plasticity, that is, long-term potentiation (LTP) and depression (LTD) is highly regulated itself. Metaplasticity is one prominent concept of activity-dependent plasticity regulation and encompasses changes in the synaptic and/or neuronal state, which shape the direction, magnitude and duration of future synaptic changes (Abraham and Bear 1996). Importantly, this modulation of the ability of neurons to express synaptic plasticity can occur in the absence of changes in the excitability of the stimulated network per se. Here, we review recent evidence for metaplasticity and its contribution to network function and behavior in the human brain. This evidence is derived from studies using non-invasive brain stimulation (NIBS), that is, transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS). TMS and tDCS are capable of inducing long-lasting changes in cortical excitability and connectivity in the awake, non-anaesthetized human brain. Although the cellular and molecular mechanisms of these LTP- and LTDlike effects are not well understood (Müller-Dahlhaus
and Vlachos 2013), first experimental evidence from in vitro studies suggests that repetitive magnetic stimulation (Vlachos and others 2012) and direct current stimulation (Fritsch and others 2010) can induce N-methyl-d-aspartate (NMDA)-receptor (NMDAR)- and brain-derived neurotrophic factor (BDNF)-dependent LTP, respectively. In humans, metaplasticity was tested by studying interactions between successive NIBS protocols, or between NIBS and behavioral interventions such as motor learning. These studies show that metaplasticity is operative on the level of the human cerebral cortex. They also demonstrate that aberrant metaplasticity may underlie certain neurological and psychiatric disorders. On a translational note, first promising studies suggest that metaplasticity can be successfully harnessed to modulate, or even restore, the ability of neural plasticity under pathological conditions.
Basic Concepts of Metaplasticity The basic idea of metaplasticity is that the threshold for activity-dependent synaptic plasticity is not static but 1
Department of Neurology and Stroke, Eberhard-Karls University Tübingen, Tübingen, Germany Corresponding Author: Ulf Ziemann, Department of Neurology and Stroke, and Hertie Institute for Clinical Brain Research, Eberhard-Karls University Tübingen, Hoppe-Seyler-Straße 3, Tübingen, 72076, Germany. Email: [email protected]
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
2
The Neuroscientist
Figure 1. Concept of metaplasticity. Please note that metaplasticity, that is, the modulation of Hebbian-type plasticity by prior neural activity (priming) can occur in the absence of changes in the excitability of the stimulated neuron/network per se. Red, metaplasticity; blue, induction of Hebbian plasticity in the naïve cerebral cortex; purple, expression of Hebbian plasticity under conditions of metaplasticity. For further explanations, please see section “Basic Concepts of Metaplasticity.”
dynamic, and a function of the integrated prior activity of the postsynaptic neuron (Abraham and Bear 1996) (Fig. 1). The time of metaplasticity expression in most studies was in the order of minutes to hours, but can last for days (Abraham and others 2001) and even weeks (Buschler and Manahan-Vaughan 2012). In addition, metaplasticity can be expressed over a broad range of spatial extents of the postsynaptic neuron. Whereas heterosynaptic metaplasticity can be global, that is, cell-wide, or confined to specific dendritic compartments, homosynaptic metaplasticity is only expressed at those synapses that participated in the initial bout of priming activity. Heterosynaptic metaplasticity is due to dendritic excitability changes (Harvey and Svoboda 2007) or modulation of synaptic tagging and capture mechanisms of plasticity-related proteins (Li and others 2014; Sajikumar and others 2009). It may also involve intercellular communication, possibly mediated by astrocytes (for a recent review, see Hulme and others 2014). In homosynaptic metaplasticity, group 1 metabotropic glutamate receptors (mGluRs) mediate facilitation and prolongation of subsequent LTP (Raymond and others 2000), whereas activation of NMDARs inhibits subsequent LTP (Huang and others 1992) and facilitates subsequent LTD (Christie and Abraham 1992). Of note, metaplasticity has been mostly described for excitatory synaptic neurotransmission, but recent evidence suggests that it may also account for plasticity regulation at inhibitory synapses (Edwards and others 2008; Fischer and others 1997). Metaplasticity thus regulates synaptic plasticity across space and time (for a recent review, see Hulme and others 2013). It serves to prolong the time window for associative interactions between neural events and may therefore underlie increased information encoding during repeated, spaced learning trials. However, it also ensures neuronal
and network stability, for example, by providing resistance against subsequent learning and thus enabling memory retention after periods of increased information encoding. The term homeostatic metaplasticity describes this neuroprotective effect to stabilize synaptic weights in neuronal networks while maintaining the capacity for synaptic plasticity. Homeostatic metaplasticity, as formalized in the Bienenstock–Cooper–Munro (BCM) theory of bidirectional synaptic plasticity (Bienenstock and others 1982), states that the synaptic modification threshold θM, that is, the threshold for induction of LTP versus LTD, is not stable but varies as a function of the integrated postsynaptic activity: it decreases at low levels of previous postsynaptic activity, favoring induction of LTP over LTD. Conversely, θM increases at high levels of recent postsynaptic activity, thereby favoring the probability of subsequent LTD over LTP (Fig. 2M). This sliding synaptic modification threshold thus enables maintenance of neuronal and network activity in a physiological range. Induction of BCM-like metaplasticity involves multiple mechanisms including NMDAR activation (Abraham and others 2001), and is expressed by changes in NMDAR subunit composition (Philpot and others 2007). More recently, modeling studies suggested an additional role for non-NMDAR mechanisms in the expression of metaplasticity such as changes in voltagegated ion channels (Narayanan and Johnston 2010). BCM-like homeostatic metaplasticity typically accounts for cell-wide, that is, heterosynaptic metaplasticity, but synapse-specific, that is, homosynaptic homeostatic metaplasticity has also been reported (Huang and others 1992).
Evidence for Metaplasticity in Human Cortex Evidence from NIBS-NIBS Studies Priming and Test Stimulation Applied to Primary Motor Cortex (M1). The most elaborate evidence for homeostatic metaplasticity in the human brain comes from a study by Hamada et al. (2008) who tested plasticity of corticospinal neurons in the primary motor cortex (M1), as indexed by changes in the amplitude of motor evoked potentials (MEPs) evoked by single-pulse TMS. They applied repeated quadro-pulse stimulation (QPS), that is, quadruplets were delivered every 5 seconds over a period of 30 minutes (Fig. 2A). Short interstimulus intervals between the quadruplet pulses of 1.5, 5, or 10 ms resulted in a long-lasting (>30 minutes) increase of MEP amplitude, that is, an LTP-like effect (Figs. 2B-D), whereas longer interstimulus intervals of 30, 50, or 100 ms resulted in an LTD-like decrease (Figs. 2E-G) (Hamada and others 2008). These QPS protocols were primed by a short
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
3
Müller-Dahlhaus and Ziemann
Figure 2. Homeostatic metaplasticity in a NIBS-NIBS study. (A) Stimulation protocol. Quadruplets (four TMS pulses) are applied in bursts of varying interstimulus intervals (1.5, 5, 10, 30, 50, 100 ms) every 5 seconds (0.2 Hz) over a period of 30 minutes (360 trains). This protocol is referred to as quadro-pulse stimulation (QPS) (Hamada and others 2007). (B-G) QPS-1.5ms, QPS-5ms, and QPS-10ms result in long-lasting (>30 minutes) increases in MEP amplitude, while QPS-30ms, QPS-50ms, and QPS-100ms result in a long-lasting MEP amplitude decrease. (H) A short (10 minutes) QPS-5ms protocol does not result in overt change in MEP amplitude and is used as priming protocol to test metaplasticity in the subsequent panels. (I, J) Priming with QPS-5ms leads to a shift of the LTP-like MEP increase induced by QPS-10ms (white circles in I) toward an LTD-like MEP decrease (black circles), and to an increase of the LTD-like MEP decrease induced by QPS-30ms (white and black circles in J without and with priming, respectively). (K, L) Priming with short (10 minutes) QPS-50ms leads to an increase of the LTP-like MEP increase induced by QPS-10ms (white and black circles in K without and with priming, respectively) and a shift of the LTD-like MEP decrease induced by QPS-30ms (white circles in L) to an LTP-like MEP increase (black circles). (M) Schematic summary of the results of priming with QPS-5ms (blue curve) versus priming with QPS-50ms (red curve) on the LTP threshold θM. Note the rightward shift of θM when compared with the non-primed curve (black curve) with QPS-5ms priming, that is, probability of LTP induction decreases, and the leftward shift with QPS-50ms priming, that is, probability of LTP induction increases (modified from Hamada and others 2008, with permission). LTD = long-term depression; LTP = long-term potentiation; MEP = motor-evoked potential; NIBS = noninvasive brain stimulation; TMS = transcranial magnetic stimulation
QPS-5ms protocol that, when given alone did not result in any overt change of MEP amplitude, that is plasticity of corticospinal neurons (Fig. 2H), but presumably increased activity in the stimulated network. This QPS5ms priming abolished the LTP-like plasticity induced by QPS-1.5ms and QPS-5ms (not shown) and even shifted the LTP-like effect induced by QPS-10ms toward MEP depression (Fig. 2I) (Hamada and others 2008). In addition, priming with short QPS-5ms increased the LTD-like effect induced by QPS-30ms (Fig. 2J) and resulted in trends toward increases of the LTD-effects induced by QPS-50ms and QPS-100ms (Hamada and others 2008). In contrast, short QPS-50ms priming, which did not result in MEP amplitude change when given alone (baseline 1
measurements in Figs. 2K and L), but presumably decreased activity in the stimulated network, increased the LTP-like plasticity induced by subsequent QPS-10ms (Fig. 2K) and shifted the LTD-like plasticity induced by QPS-30ms (Fig. 2L) and QPS-50ms toward an LTP-like MEP increase (Hamada and others 2008). Altogether, these data are fully compatible with homeostatic metaplasticity according to the BCM theory (Bienenstock and others 1982) as reviewed above: priming with short QPS5ms (i.e., high-frequency priming according to the short interstimulus interval of 5 ms in the quadruplet) resulted in a rightward shift of the threshold for induction of LTPlike plasticity, that is, an enhanced probability of LTDlike plasticity and decreased probability of LTP-like
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
4
The Neuroscientist
plasticity (blue curve in Fig. 2M). Vice versa, priming with short QPS-50ms (that is, low-frequency priming according to the long interstimulus interval of 50 ms in the quadruplet) resulted in a leftward shift of the threshold for induction of LTP-like plasticity, that is, decreased probability of LTD-like plasticity and increased probability of LTP-like plasticity (red curve in Fig. 2M). Similar observations in accord with the BCM theory have been made in a number of other studies (synopsis in Table 1). For example, low-frequency (1Hz) regular repetitive TMS (rTMS) typically results in LTD-like MEP amplitude decrease when given alone (Chen and others 1997). However, priming of 1-Hz rTMS with highfrequency (6 Hz) rTMS resulted in a stronger LTD-like MEP decrease compared with 1-Hz rTMS alone (Iyer and others 2003). A shortcoming of that study was that the effects of the 6-Hz priming protocol, when given alone, on MEP amplitude had not been tested. Therefore, it remained unclear to what extent any possible MEP change by the priming protocol has contributed to this enhancement of LTD-like plasticity by subsequent 1Hz rTMS. Furthermore, interactions of 6Hz rTMS priming with subsequent induction of LTP-like plasticity had not been tested. Paired associative stimulation (PAS) induces LTP-like or LTD-like changes in MEP amplitude (PASLTP, PASLTD), depending on the interstimulus interval between the electrical stimulation of a peripheral nerve and associated single-pulse TMS of the contralateral M1 (MüllerDahlhaus and others 2010; Stefan and others 2000; Wolters and others 2003). PAS has gained some interest because several features resemble those of spike-timingdependent plasticity as studied on a cellular level (MüllerDahlhaus and others 2010). The LTP-like increase in MEP amplitude induced by PASLTP was occluded by priming with PASLTP, while priming with PASLTD led to a trend toward stronger LTP-like plasticity by the subsequent PASLTP protocol when compared to a priming control condition (Müller and others 2007). Furthermore, effects on MEP amplitude induced by primed PASLTP were inversely correlated with those induced by priming PAS (Müller and others 2007). Because the two PAS protocols likely activated the same neuronal circuits, it was concluded that “homosynaptic-like” homeostatic metaplasticity in accord with the BCM theory underlay the observed PAS interactions. Theta-burst stimulation (TBS) induces an LTP-like increase or LTD-like decrease in MEP amplitude, depending on the pattern of stimulation: intermittent TBS (iTBS) results in LTP-like plasticity, continuous TBS (cTBS) leads to LTD-like plasticity (Huang and others 2005). Priming cTBS with iTBS resulted in enhanced LTD-like plasticity (Doeltgen and Ridding 2011). These data confirmed findings from one previous study (Todd and others 2009) that,
in addition, demonstrated that 2- and 6-Hz priming had no effect on the LTD-like plasticity effect induced by cTBS. The interactions between two successive TBS protocols were investigated more extensively in another study that showed that pairing of identical protocols (iTBS → iTBS, cTBS → cTBS) resulted in suppression of the non-primed TBS effects on MEP amplitude, while pairing of different protocols (cTBS → iTBS, iTBS → cTBS) enhanced the non-primed TBS effects on MEP amplitude (Murakami and others 2012). These results are in full compliance with homeostatic metaplasticity according to the BCM theory. The study by Murakami and colleagues also investigated for the first time systematically metaplasticity of inhibitory circuits by utilizing the fact that iTBS and cTBS lead to long-term increases and decreases in GABAAergic inhibition, respectively, as indexed by the paired-pulse TMS measure of short-interval intracortical inhibition (SICI; Huang and others 2005). While non-primed TBS was subthreshold for inducing significant changes of SICI, iTBS → iTBS resulted in SICI decrease, and cTBS → cTBS in SICI increase compared with the non-primed conditions (Murakami and others 2012). Importantly, the changes in SICI induced by priming TBS correlated with the changes in MEP induced by subsequent test TBS. These findings indicate that plasticity in M1 inhibitory circuits is also regulated by homeostatic metaplasticity, and that priming effects on GABAergic inhibition contribute to the homeostatic regulation of metaplasticity in excitatory circuits (Murakami and others 2012). Another study investigated the effects of the interval between two identical TBS protocols (iTBS → iTBS or cTBS → cTBS) (Gamboa and others 2011). The interactions were by and large homeostatic, with one notable exception: cTBS → cTBS resulted in an enhancement (rather than reduction or occlusion) of LTDlike plasticity with the longest tested interval of 20 minutes (Gamboa and others 2011). While the reasons for this latter non-homeostatic interaction remained unclear, a candidate mechanism might be consolidation of LTD through increases in de novo protein synthesis and gene transcription, as has been demonstrated in hippocampal slices for repeated spaced LTP-induction protocols (Nguyen and others 1994). Consolidation of LTP by protein synthesis and gene transcription is associated with resistance to depotentiation (Woo and Nguyen 2003). In accord with consolidation of synaptic plasticity, recent cTBS → cTBS experiments (10-minute interval between priming and test cTBS) demonstrated a non-homeostatic interaction with significant lengthening of the LTD-like MEP decrease >120 minutes (Goldsworthy and others 2012) that was resistant to de-depression by voluntary contraction or short iTBS (150 pulses) (Goldsworthy and others 2014). In one other study (Mastroeni and others 2013) the effects of polymorphisms of the BDNF gene (val66met vs. val66val) were investigated, since previous studies have shown,
Downloaded from nro.sagepub.com at UNIV OF SOUTHERN CALIFORNIA on April 6, 2014
5
Müller-Dahlhaus and Ziemann Table 1. Synopsis of NIBS-NIBS Studies to Test Metaplasticity in Human Cortex. Site of Priming/ Test Stimulation M1/M1 M1/M1 M1/M1 M1/M1
M1/M1
M1/M1
M1/M1 M1/M1
M1/M1
M1/M1 M1/M1
M1 / M1
Priming Stimulation
Test Stimulation
Effects on Test Stimulation
Reference
6Hz rTMS, 90%RMT, 1-Hz rTMS, 115% RMT,
