Category:Synaptic Plasticity

INTRODUCTION

NEURAL ANALOGS OF LEARNING AND MEMORY

It is now generally accepted that many forms of adaptive behavior, including learning and memory, engender lasting physiological changes in the brain; reciprocally, neural plasticity among the brain’s synaptic connections provides the capacity for learning and memory. A major step toward linking synaptic changes with actual instances of learning-related behavioral modification was taken in the 1960s when Eric Kandel published a review entitled “Cellular Neurophysiological Approaches in the Study of Learning”, which emphasized the key to understanding memory formation is through studying neural dynamics at the synaptic level 6. Progress toward these goals wasn’t made, however, until after identifying a simple model organism that displayed adaptive behavior, and pinpointing the neurons involved with this process - Kandel’s team identified a gill withdrawal reflex in Aplysia californium (a sea hare) that could be modified by various forms of conditioning, and then mapped out the sensory and motor neurons involved in this process 5.

After working out these prerequisite issues Kandel and his colleagues were in a position to address many long-standing questions about memory physiology. In a series experiments published in the 1970s, they first detailed the neural correlates of habituation. Applying a mild mechanical or electrical stimulation to an Aplysia’s siphon elicits a pronounced retraction of the gill (i.e. the gill withdrawal reflex) along with excitatory postsynaptic potentials (EPSP) in the gill’s motor neuron; after repeated weak stimulation they found significant reductions in both gill withdrawal and EPSPs lasting for several minutes 7. When this procedure was done for five consecutive days, habituation lasted several weeks 8. Importantly, they could fully mimic these effects when a series of weak stimulations was applied directly to the siphon’s sensory neuron. This is an example of homosynaptic depression, so called because stimulation is applied directly to the pathway being depressed/habituated.

A similar approach was used to characterize sensitization. First they measured baseline gill withdrawal to a weakly stimulated siphon. Then they carried out a behavioral protocol applying a noxious stimulus to the tail, followed by weak stimulation of the siphon, which resulted in greater (sensitized) gill withdrawal compared to baseline 9. They reproduced these effects neurologically by strongly activating a tail interneuron followed by mild activation of a siphon sensory neuron, which resulted in greater gill motor neuron EPSPs and sensitized gill withdrawal 10. Like habituation, when this procedure was repeated across multiple days, sensitization lasted several weeks 11. Sensitization is underlied by heterosynaptic facilitation, as synaptic strength of one pathway (siphon > gill) was enhanced by a second pathway (tail > siphon) 12, 13 (see Fig. 2). In either case of short-term habituation or sensitization, forms of implicit learning, the modulating factor was clearly at the presynaptic terminal, such that after habituation the sensory neuron releases less glutamate neurotransmitter 14, and after sensitization it releases more glutamate neurotransmitter 15. Furthermore it was found that transmitter release was directly regulated by the amount of calcium that entered the presynaptic terminal in response to an action potential 15-17.

In subsequent studies conducted in the 1980s and early 1990s, Kandel’s group uncovered the mechanistic difference between short-term and long-term facilitation of this reflex. They found that short term facilitation only required the activation of local proteins like cAMP and PKA that aid in calcium entry into the presynaptic terminal 12, 13, 18. On the other hand, long-term facilitation required the local proteins, in this case PKA and MAPK, to travel up the axon to the nucleus (via retrograde transport) and activate gene transcription factors like CREB 19, 20. New RNA transcripts were then trafficked back down the axon (via anterograde transport) and translated into protein locally at the appropriate terminal, culminating in the growth of new synaptic connections between the sensory and motor neuron 21.

To summarize, these studies revealed that a simple reflex could be modified through learning, and stored as an implicit form of memory. Furthermore, the sensory, motor, and interneurons involved in this reflex circuit could be directly stimulated to reproduce learned behaviors. Perhaps the most striking aspect of these studies was the systematic demonstration that key proteins could be artificially introduced or blocked to precisely mimic natural learning and memory 18. The fundamental breakthrough, however, was that Kandel’s studies provided clear evidence that learning produced changes in synaptic strength, and these changes were a necessary and sufficient component of memory.

POSTSYNAPTIC LONG-TERM POTENTIATION

By the end of the 1980s Kandel’s studies using Aplysia all pointed toward a model where synaptic efficacies are regulated by presynaptic mechanisms; meanwhile the postsynaptic terminal seemed to be a relatively static structure with the purpose of interpreting the quanta of neurotransmitter released by the upstream neuron. However, a limitation of the invertebrate Aplysia model is that neuropsychological inquiries are restricted to simple implicit forms of memory. The mammalian brain, though, has the capacity to store and recall various forms of explicit memory, now thought to require the support of a relatively sophisticated neural circuitry like that found in the mammalian medial-temporal cortex, in a region called the hippocampus.

The hippocampus plays a central role in information integration and spatial memory 22, and its functional integrity is critical for learning and consolidating explicit memory 23. This limbic structure receives afferent projections from entorhinal cortex along the perforant pathway, first enervating granule cells in the dentate gyrus (DG), and continue on with mossy fibers to finally make connections with CA3 pyramidal cells. In turn, CA3 neurons project axons to CA1 pyramidal neurons along the Schaffer collateral pathway, in a final stop before information is routed through the subiculum and looped back to the entorhinal cortex (see Fig 3). All three hippocampal pathways have been used extensively for studying activity-dependent changes in synaptic strength, most notably long-term potentiation (LTP) and long-term depression (LTD). LTP and LTD refer to lasting changes in synaptic efficacies, where LTP is the stable enhancement of synaptic strengths, and LTD is the stable reduction in synaptic strengths. Given these broad definitions, it’s possible for functionally distinct forms of synaptic plasticity to be considered ‘LTP’ 24. Notably, different forms of long-lasting potentiation have been found in the hippocampus itself: at CA3 mossy fiber synapses potentiation is owed to presynaptic mechanisms 25, 26, while potentiation is owed to postsynaptic mechanisms at CA1 Schaffer collateral synapses 27-31. Here I will use the term long-term facilitation (or LTF) when synaptic efficacy-changes are presynaptically mediated (described above in the Aplysia model), and LTP when changes to synaptic efficacy are postsynaptically mediated (consistent with the established convention 13, 24).

The first published observation of LTP, credited to Bliss and Lomo in 1973, was at hippocampal perforant pathway synapses 32. In this study, the authors used an electrode to enervate perforant path axons with short bursts of high-frequency electrical stimulation (100 Hz for ~3 sec), and recorded EPSPs from downstream DG interneurons. This procedure resulted in potentiated connectivity that lasted several hours and was mediated by postsynaptic changes. In this way LTP is analogous to LTF, with three primary exceptions: (1) LTF stems from presynaptic changes while LTP stems from postsynaptic changes, as per definition 4, (2) LTF involves a di-synaptic circuit whereas LTP involves a mono-synaptic circuit 33, and (3) LTF doesn’t require postsynaptic calcium influx while LTP does (re-visited below) 4. Also unlike LTF, LTP has not been so eloquently linked to adaptive behavior; to-date however, LTP remains the most compelling cellular analog of learning and memory in a majority of cortical regions involved in explicit forms of memory 34-36.

A first step in establishing LTP as an analog of learning was the development of LTP-induction protocols that mimic firing patters during hippocampal-dependent learning. In the pioneering study by Bliss and Lomo, they stimulated neurons at 100 Hz for 3-4 seconds; while neurons do communicate in high-frequency (~100 Hz) “bursts” of action potentials, in vivo bursting activity is brief, only lasts for fractions of a second 37. This disparity prompted research on the development of LTP induction paradigms that were more in-line with innate firing patterns. Ensuing studies revealed that optimal hippocampal LTP induction involves using a priming stimulus followed 100-200 ms later by a brief 100 Hz burst (~10 pulses), in-phase with the ongoing theta rhythm of the hippocampus 38. Indeed, these temporal features of neural firing, together called “theta-bursting”, are typical of CA1 pyramidal neuron activity in rodents during the exploration of space and learning cues 39, 40.

Theta-bursts applied to Schaffer collaterals result in a 3-phase sequence of synaptic potentiation. The short-lasting early phase is referred to as post-tetanic potentiation (PTP), and is mediated by calcium-facilitated release of presynaptic neurotransmitters 41. This initial phase is analogous to long-term facilitation, and its effects decay rapidly as calcium is cleared from the presynaptic terminal over a ~2 min time-course 42. The subsequent phases, known as short-term potentiation (STP) and LTP, generally regarded as two parts of the same phenomenon, depend on the activation of postsynaptic glutamate receptors and calcium (see Fig. 4) 43.

In testing the effects of various compounds on Schaffer pathway transmission, Collingridge et al. (1983) found a pair of compounds that affected excitatory glutamatergic signaling in two functionally different ways: quisqualate globally depressed excitatory transmission, while amino-phosphonovalerate (APV) only blocked potentiation without affecting normal excitatory transmission 43. We now refer to the glutamate receptors that bind those selective antagonists as AMPA-type receptors (AMPAR) and NMDA-type receptors (NMDAR), respectively (after their selective agonist). Building off this and other work 44, Roger Nicoll’s team found that inhibiting postsynaptic calcium entry after a theta-burst fully blocked LTP; conversely, allowing calcium entry into the postsynaptic terminus (even in the absence of a presynaptic stimulus) was sufficient to induce LTP 4. Thus it was discovered that LTP was a postsynaptic process that required NMDAR activation and calcium influx to initiate, and the recruitment of (additional) postsynaptic AMPARs to manifest 45.

AMPA-TYPE GLUTAMATE RECEPTORS (AMPARs)

Hippocampal-dependent memory formation is generally thought to involve changes to the postsynaptic number of AMPARs 28, 29, 46, 47. AMPARs are the primary glutamate receptors that mediate fast-excitatory transmission in the central nervous system, and their expression-level at synapses is directly related to synaptic strength 48. Since their discovery, AMPAR trafficking and regulation has been of acute interest. A variety of electrophysiological and biochemical studies have revealed that AMPARs actively undergo endo/exocytotic exchange between intracellular compartments and the dendritic surface 28, 29, 49, 50, at extrasynaptic membrane locations outside of dendritic spines 51. Relatively recent experiments using techniques like fluorescence recovery after photobleaching (FRAP) and single particle tracking (SPT) methods have established that AMPARs diffuse from extrasynaptic sites into synapses via lateral surface diffusion 52; indeed AMPARs diffuse along the entire dendritic surface, constitutively trafficking in-and-out of synaptic areas 51-53 (see Fig. 5). Thus the quantity of AMPARs in a given synapse is not static, but instead, a dynamic steady-state. Given these exciting new discoveries, a major thrust in today’s work has been geared toward understanding how the cell regulates AMPAR surface diffusion in-and-around synapses, and characterizing AMPAR dynamics from biogenesis through channel conductance.

The AMPAR subunit-family is comprised of four unique subunit proteins, GluR1-GluR4 (aka GluA1-GluA4) that can intermix to form a single AMPAR 54, 55. These 105 kDa-subunits 56 share ~70% gene sequence homology, and may undergo post-transcriptional splicing that results in a long or short version of the intracellular carboxyl-terminus (Ctail) 57 (the implications of having Ctail-variants is discussed below; see Fig. 6a). Before assembling into functional AMPARs, GluR of the same subtype pair-up to form dimers; these dimers then converge with another GluR dimer to form a functional tetrameric AMPAR; thus, AMPARs are considered a ‘dimer of GluR dimers’. The two dimers in an AMPAR can be the same type, yielding a homomeric tetramer (e.g. GluR1-GluR1--GluR1-GluR1), or two different types that yield a heteromeric tetramer (e.g. GluR1-GluR1--GluR2-GluR2). It should be noted however that various factors restrict homomeric AMPARs from assembling in vivo at meaningful quantities (nevertheless, recombinant homomeric AMPARs have been used extensively as molecular probes) 58. Instead, endogenous AMPARs in the hippocampus are almost exclusively heteromeric pairs of GluR1 and GluR2 dimers (GluR1/2), with the remaining AMPARs consisting of GluR2/3 heteromers 59; in both cases GluR2 is present in the receptor. GluR2 pre-mRNA in the adult brain undergoes a “Q/R editing” process that results in the replacement of glutamine (Q) by arginine (R) at the M2 channel segment, yielding (R)-GluR2 60. Aside from gating Na+ in preference for K+, this edit has a fundamentally important consequence: (R)-GluR2 subunits are impermeable to calcium 61. As a result, endogenous AMPAR activation doesn’t itself elicit calcium influx and its associated molecular cascade. Indeed, disrupting Q/R editing has detrimental consequences - mice genetically engineered to express only unedited (Q)-GluR2 develop seizures and die early in development 62.

AMPAR Regulation and Trafficking Overview

Among the first assessments of the role of AMPARs in LTP was a manipulation that deleted the GluR1 subunit, which blocked hippocampal LTP induction 63. This finding was followed by numerous works focusing on unique properties of the different AMPAR subunits, with particular interest in the GluR Ctail differences. A line of work by Roberto Malinow and colleagues produced a compelling model of AMPAR trafficking during basal and LTP conditions. In this model the GluR2 subunit allows AMPARs to enter basal (non-potentiated) synapses, while the GluR1 subunit promotes synaptic-insertion of AMPARs during LTP 51 — a functional difference shown to depend on properties of their respective Ctails (strikingly, swapping their Ctails swaps their trafficking behavior). A separate line of work focused on the role of AMPAR auxiliary proteins known as transmembrane AMPAR regulatory proteins (TARPs). Similar to the GluR1 deletion, a particular TARP knock-out was found to hinder LTP. This prompted a line of work examining AMPAR regulation during baseline and LTP, with many more experiments aimed at identifying the intracellular binding partners of TARPs and GluR Ctails at synapses. Perhaps unsurprisingly, a variety of subsurface proteins shown to interact with GluR Ctails have recently been implicated in LTP. As it is, the regulation of AMPAR trafficking via Ctail/TARP interactions with synaptic proteins remains a topic of immense interest.

AMPAR Trafficking: GluR Ctails

surface trafficking: AMPARs diffuse laterally along the surface of dendrites and into synapses 64, but before this can happen they must first migrate from the ER to exocytotic zones along the dendrite, and then be inserted into the extrasynaptic membrane 51. Studies have revealed that GluR subtypes have different affinities for surface insertion and synaptic expression, mainly owed to their limited Ctail homology. To address several questions about the contribution of GluR Ctails on surface expression, Granger al’et Nicoll (2013) conducted a study using organotypic CA1 pyramidal cells with an AMPAR-null background (Gria1-3fl/fl [i]) and then expressed recombinant GluR subunits 65. Their experiments showed that membrane insertion of AMPAR is significantly reduced (by ~75%) when the Ctail is removed from GluR1 (GluR1ΔC) compared to full-length GluR1 in the null background. A similar effect was found with GluR2ΔC, in fact, the GluR2 Ctail may be slightly more efficient at promoting surface expression than the GluR1 Ctail 66. That said, there were no appreciable differences in surface levels between full-length GluR1 homomers in the null background vs. AMPAR heteromers in the wild-type cell (as recorded from somatic outside-out patches, or ‘OOPs’) suggesting that GluR1 homomeric AMPARs traffic to the somatic surface just as readily as GluR2-containing AMPAR heteromers.

basal synaptic trafficking: Interestingly, full-length GluR1 expressed in the null background had greatly reduced EPSCs (by 32%) compared to neighboring wild-type cells, suggesting GluR1 homomers don’t accumulate in basal synapses as readily as wild-type heteromers (which all contain GluR2) [ii] 59. As somatic surface expression was the same for recombinant GluR1/1 homomers and wild-type GluR1/2 heteromers, there are only a few ways to explain this finding: (1) GluR2 is important for synaptic insertion at basal synapses; (2) GluR1/1 surface expression only matches GluR1/2 surface expression in the soma and not at dendrites; (3) GluR1 is actively preventing insertion at basal synapses. Interpretation #1 is consistent with a prior study by Makino al’et Malinow (2009) that utilized fluorescence recovery after photobleaching (FRAP) methods — spines with GFP-tagged GluR1 homomers recovered to baseline levels just minutes after photobleaching, indicating that GluR1 quickly diffuses in-and-out of basal spines 51. Conversely, a substantial amount of GluR2 fluorescence (~20%) failed to recover within the 30-min assay, indicative of a slowly-diffusing or trapped fraction of GluR2 homomers in basal spines. Other studies have tested whether these effects are mediated by different binding sequences along the GluR Ctail. Recent work suggests that GluR Ctail sequences are ligands for PDZ domain proteins (further discussed below), and insertion of AMPARs into basal synapses may require the functional integrity of the PDZ-binding segment on the GluR2 Ctail 67 (but see 65). Interestingly, expression of GluR2 dissociated Ctails act as ‘dominant-negatives’ to depress basal transmission while the dissociated Ctail of GluR1 has no effect. Overall these data support a model where GluR2, but not GluR1, promotes receptor accumulation in basal synapses.

LTP synaptic trafficking: As mentioned above, the trafficking profile of GluR subunits change upon LTP induction. In the Makino al’et Malinow (2009) study, both GluR1 and GluR2 homomers showed the same (~20%) non-recovery after chemically-induced LTP, suggesting GluR1 synaptic anchoring ability increases during LTP (while GluR2 anchoring remains constant). The GluR1 Ctail may interact with several synaptic proteins transformed by the LTP signaling cascade, allowing activated synapses to accumulate GluR1-containing AMPARs 68. Indeed, the expression of GluR1 dissociated Ctails can block LTP 67. Perhaps the most striking finding from this work was the discovery that swapping the GluR1 and GluR2 Ctails flip-flops their trafficking profiles 67 -- discriminating evidence that differences in GluR1 and GluR2 trafficking behavior is at least partially mediated by their different Ctails.

AMPAR and SAP Interactions: MAGUK, GRIP, ABP

AMPARs are thought to accumulate at synapses by tethering to proteins anchored within the postsynaptic density (PSD). In the mid 1990s a scaffold-associated protein (SAP) called PSD-95, one of the most abundant PSD proteins, was found to interact with glutamate receptors 69. This finding hinted at a role for PSD proteins in regulating synaptic receptor numbers. PSD95 is a member of a superfamily of SAP known as membrane-associated guanylate kinases (MAGUK)[iii], that have a guanylate kinase-like domain (GK), an SH3 domain, and one or more PDZ domains (see Fig. 7). More than 150 proteins have PDZ domain motifs 70, which interact with PDZ-ligand sequences often found on c-terminal peptides, including the Ctail of glutamate receptors (as in Fig. 6b) 71, 72. PDZ-ligands are short amino acid sequences that assist in folding PDZ domains at a β-finger, promoting the assembly of signaling complexes (see Fig. 8). PDZ-domains are common components of scaffolding proteins like MAGUK, and are involved in organizing supramolecular signaling complexes 71 that localize enzymes with their substrates or cluster receptors at biological membranes 73. These clusters form a dense protein scaffolding that ultimately anchors to the actin cytoskeleton.

PDZ interactions, supramolecular clustering, and β-finger folding are complex processes beyond the scope of this review (but see 73 & 74), nonetheless, the emerging role of PDZ-AMPAR interactions represent an exciting new frontier in synaptic plasticity research. Here, I will limit the discussions to a relatively small group PDZ-containing proteins with putative involvement in synaptic plasticity, starting with the Dlg1-4 group of MAGUKs. The Dlg1-4 MAGUK proteins, individually known as SAP97, PSD93, SAP102, and PSD95 (respectively; shown in Fig. 7) are highly expressed at glutamatergic PSDs where they have been implicated in AMPAR synaptic targeting 72, 73. Studies have shown that as the synaptic levels of SAP97, PSD93, or PSD95 increase, so do AMPAR levels 75-79 (with similar, though less dramatic, results have been reported for SAP102 78, 80).

Given this finding, it’s somewhat inexplicable that SAP97 is the only MAGUK among them known to directly bind AMPARs. Indeed, the GluR1 Ctail directly binds a PDZ domain on SAP97 81, 82 (perhaps involving another synaptic protein called EBP 4.1N 83-85; see Fig. 6b), but to complicate things even further this interaction is shown to happen early in biogenesis, and not at synapses .

While GluR1 is the only subunit known to directly bind a MAGUK Dlg protein, GluR2 is an active binding partner with other PDZ proteins. GluR2 has a consensus sequence (S-V-K-I; see Fig 6a) shown to interact with specific PDZ domains on glutamate receptor interacting proteins (GRIP) 86-88, a splice variant of GRIP called AMPAR binding protein (ABP) 88, 89, and PICK1 90 scaffold proteins. The unique set of synaptic binding partners of GluR1 and GluR2 likely contribute to their different trafficking profiles during basal and potentiated transmission.

Given that (1) PDZ domain proteins help cluster receptors near biological membranes, (2) short carboxyl segments are optimal ligands for PDZ domains, and (3) the overexpression of each Dlg upregulated synaptic AMPARs, one would think GluR Ctails are prime candidates for directly binding the PDZ domain of all four of these MAGUKs. Clearly further study is needed to clarify the mechanism by which PSD93, PSD95, and SAP102 recruit synaptic AMPARs. As it is, there is some evidence that another class of proteins, known as transmembrane AMPAR regulatory proteins (TARPs), contribute to AMPAR trafficking by linking AMPARs to PSD93 and PSD95 MAGUK proteins 91.

TARPs - AMPAR AUXILIARY SUBUNITS

Aside from the interactions between GluR Ctails and MAGUK, there’s evidence that another class of proteins contribute to AMPAR trafficking, known as transmembrane AMPAR regulatory proteins (TARPs) 92, 93. Particular attention has been given to a TARP called stargazin (aka γ2 [iv]). A strain of γ2 knock-out (K.O.) mice called stargazer mice were found to lack the AMPAR-mediated fast-spiking component of mEPSCs[v] in cerebellar neurons, thought to contribute to the stargazer phenotype of absence epilepsy and ataxia. Indeed, granule cells in stargazer cerebellum did not respond to AMPA, even though cellular levels of GluR proteins and mRNA were normal 94, 95. This deficit was shown to be cerebellar-specific, as hippocampal currents were normal in stargazer mice, however the precise role of these stargazin-related TARPs remains controversial (see 96).

A report by Chen et al. (2000) suggested that stargazin/γ2 could associate with AMPARs and bind PSD95 via PDZ interactions 97 (see Fig. 8). This was a provocative finding because GluR Ctails are supposedly incompatable with the PDZ domains of PSD95; so it seemed peculiar that the most abundant synaptic protein, one with PDZ domains perfectly evolved for clustering membrane proteins, didn’t interact with AMPARs [vi]. In the forementioned study, stargazin was shown to co-immunoprecipitate with AMPAR subunits in COS cells - heavily with GluR4, and to a lesser degree, GluR1 and GluR2 [vii]. Chen and coworkers followed this up by expressing recombinant stargazin to the stg-mutant (γ2 knockout) mouse -- they found that full-length stargazin could fully rescuse cerebellar mEPSC[viii], while stargazinΔC (stargazin without the Ctail) rescues only non-synaptic currents in the cerebellum 98. In the hippocampus, expressing stargazinΔC in pyramidal neurons resulted in a marked decrease in both mEPSC amplitude and frequency (suggesting there was both a global reduction total surface AMPARs and a complete loss of AMPARs at some synapses).

Together, the experiments above show that cerebellar stargazin promotes AMPAR surface expression and synaptic accumulation, however only for the latter was stargazin’s PDZ domain important. The hippocampal stargazin story is less clear. On one hand, the total absence of stargazin (γ2 K.O.) had no effect on hippocampal EPSCs 95; on the other hand, the insertion of PDZ-lacking stargazin drastically reduced hippocampal mEPSC. What then, is the role of stargazin in the hippocampus?

The hunt to define a functional role of stargazin in the hippocampus, if any, has involved a range of experiments attacking the problem on many fronts. It’s now known that mice express eight stargazin-like proteins (γ1 through γ8), but only four of these isoforms (γ2, γ3, γ4, and γ8) have been show to interact with AMPAR and thus receive the ‘TARP’ designation 98, 99. One or more of the stargazin-family TARPs (γTARPs) bind all four GluR subunits 97, 100, and are found in all major brain regions 101, with γ8 being highly expressed in the hippocampus 98. In a γ8 knock-out model, extrasynaptic AMPARs in hippocampal neurons were drastically reduced, while GluR subunits pooled in the ER 102. In another study, overexpressed GluR1 remained in cell bodies, but when co-expressed with γ2, AMPARs were transported to dendrites 66. It therefore appears AMPAR distribution in hippocampal dendrites relies on γTARPs , suggesting they probably interact with AMPARs during the early stages of biogenesis 98. Indeed, FRET reveals γTARPs are in close proximity to GluR at the ER 103, perhaps assisting in subunit assembly or ER export 98, 104. After leaving the ER, tetrameric GluR complexes enter the Golgi system and are woven through a lipid bilayer to form functional AMPARs 105. Again at this stage, γTARPs may assist in AMPAR membrane-insertion or Golgi export by binding MAP1, a microtubule associated protein important for neurogenesis 106, though a functional corollary of γTARP - MAP1 interaction is lacking.

Two emerging roles for TARPs are (1) the enhancement of AMPAR conductance 96 and (2) protection of AMPAR from proteolysis 107, functions that may be in addition to trafficking roles. Evidence for the modulation of AMPAR biophysical properties comes from studies that exhibit AMPARs co-expressed with stargazin in oocytes, HEK cells, or hippocampal pyramidal neurons have reduced desensitization 66, 108-110, slower deactivation 108, 110, and enhanced recovery 109, 110. One of these studies provided solid evidence that TARPs promote AMPAR dendritic distribution but not surface expression, and also suggested that γ2 could protect against proteolysis. When γ2 was expressed in pyramidal neurons, it selectively prevented GluR1, but not GluR2 lysosomal degredation 66. Furthermore GluR2/3 heteromers were more rapidly digested by trypsin in stargazer extracts than control extracts from cerebellar lysates (it should be noted that GluR1 is not found in adult cerebellar granule neurons) 111. One possible explanation is that poorly assembled multimeric proteins have increased proteolysis susceptibility 112, and if γTARPs help assemble AMPARs, stoichiometric levels between AMPARs and γTARPs must be around 1:1+ for maximal protection. Thus, overexpression of GluR subunits, particularly recombinant subunits, may yield poorly assembeled AMPARs that are targeted for degredation.

DO CTAILS OR TARPS REGULATE AMPAR TRAFFICKING?

From the large body of work one gets the sense that nearly every operative step in the lifetime of an AMPAR depends on Ctails interactions with PDZ and other regulatory proteins, and yet all these same steps seem to involve TARP interactions 114. Moreover, how are subunit-specific trafficking effects observed when TARPs lack subunit specificity 115? The most straightforward answer would be if TARPs complex with AMPARs early on in biogenesis, where aid in tetramer formation, ER export, cytosolic transport, and surface expression. After the TARP-AMPAR complex reaches the surface they dissociated and part ways. Alternatively, the reverse scenario may be true, where MAGUK proteins help along the secretory pathway, transitioning AMPARs from the ER and Golgi systems out to dendritic arbors. As mentioned above, SAP97 is known to associate with GluR1 to a much greater extent in these organelles than at synapses 86. However, this may change later in development, after various other subunit-specific binding proteins like GRIP and ABP mobilize into synapses to accommodate subunit-specific trafficking (reviewed in 94). As it is, there seems to be overlap between PDZ-SAPs and TARPs in AMPAR maturation and cytosolic trafficking, and overlaps between GluR-Ctails and TARPs in synaptic tethering to PDZ-SAPs. The reason for these redundancies is unclear, but it could be related to the evolution of neurons in different brain structures. For example, without GluR1 (or the GluR1-Ctail), AMPARs cannot get to the dendritic surface of CA1 pyramidal neurons 59, 65; since the GluR1 subunit is not endogenously expressed in cerebellar granule neurons in vivo 112, 116-119 how then are AMPARs found in abundant quantities on the surface of these neurons? The answer appears to be stargazin - without it, cerebellar granule cells do not respond to AMPA 95, 96. Maybe AMPAR-Ctails and TARPs do have slightly redundant trafficking properties (at least in neurons that express GluR1), but this redundancy could have important consequences when it comes to neural network signaling. It has recently been proposed that AMPARs make multivalent interactions with synaptic proteins 120. Multivalency may play an key role in synaptic weight stability, and several features emerge when one considers that spatial-temporal stability is not all-or-none, particularly when it comes to influencing the synaptic steady-state number of receptors: (1) it could add stability to AMPAR-TARP-SAP complexes (has not been addressed),

three features emerge from consideration of the supramolecular assemblies that might promote multivalent interactions (Fig. 1). First, although the exact stoichiometry of TARP per AMPAR remains unclear, and may vary depending on cell type9–12, recent models of hippocampal neurons suggest the presence of either two or four units per receptor11. Hence, synaptic TARP-containing AMPAR complexes would present two or more identical PDZ domain–binding motifs from the auxiliary subunits. Second, postsynaptic MAGUKs, which are key elements of the PSD architecture13, share the same domain topology and in particular include a repeat of three clustered class I PDZ domains14,15 that provide anchoring sites for multiple binding partners13. Furthermore, the relative orientation of the ligand-binding grooves of the first two PDZ domains of PSD-95 favors accommodating multiple ligands originating from the same face, according to a NMR-derived structural model16. Functional studies in neurons indicate that the AMPAR–Stargazin complex preferentially binds to either or both of the first two PSD-95 PDZ domains4,17. Finally, the oligomerization properties of MAGUKs add another degree of complexity that allows for multivalent interactions. In particular, PSD-95α (predominant in neurons18) can form dimeric or multimeric assemblies via the N-terminal region17. Notably, this property is critical for the AMPAR’s synaptic function, as mutants that prevent oligomerization show effects on basal AMPAR currents4,17.

What would multivalency mean in the context of LTP? Perhaps this is a key property that makes endogenous LTP possible. In the Malinow model, GluR2 populates basal synapses and LTP elicits GluR1 synaptic influx; given that GluR1/2 heteromers compose over 80% of the AMPARs on the surface ^[1].

Surface Diffusion

%-------------###############################----------------%
%             Two Dimensional Realistic Brownian Motion
%-------------###############################----------------%

dimensions = 2;         % two dimensional simulation
tau = .1;               % time interval in seconds
time = tau * 1:N;       % create a time vector for plotting

k = sqrt(D * dimensions * tau);
dx = k * randn(N,1);
dy = k * randn(N,1);

x = cumsum(dx);
y = cumsum(dy);

dSquaredDisplacement = (dx .^ 2) + (dy .^ 2);
 squaredDisplacement = ( x .^ 2) + ( y .^ 2);

plot(x,y);
title('Particle Track of a Single Simulated Particle');

%-------------############################------------------%

clc; close all; clear all;
%===================================%
%		STARTING PARAMETERS
%-----------------------------------%

Ndots = 200;			% number of particles in ES
NSteps = 500;			% number of steps per trial

Sc = 1.0;				% scale of model
dT = 0.1;				% time step (s)
m = 2;                  % spatial dimensions

Des = (Sc*dT)* .13;		% D coef ES
Dsb = (Sc*dT)* .05;		% D coef Syn Basal
Dsp = (Sc*dT)* .01;		% D coef Syn LTP

k = sqrt(m*Des);		% sd of Des step-size distribution

XYL = zeros(2,Ndots);	% XY particle locations
XYS = zeros(2,Ndots);	% XY particle step sizes
XYLp = zeros(2,NSteps);	% preallocate matrix for trace dot

DendSz = [3 6] ./Sc;	% surface area (µm) scaled
SIZE.x = DendSz(1)/2;   % half enclosure size X-dim
SIZE.y = DendSz(2)/2;   % half enclosure size Y-dim



%===================================%
%			FIGURE SETUP
%-----------------------------------%

Flh = figure(1);
	[Flh] = FIG1FORMAT(Flh, SIZE);

Ph1 = scatter(XYL(1,:),XYL(2,:),5,[0 0 1]);
	[Ph1] = PLOT1FORMAT(Ph1, SIZE);


%===================================%
%			MAIN LOOP
%-----------------------------------%
for Nt = 1:NSteps 

	% generates step sizes
		XYS = (k * randn(2,Ndots));	

	% adds step to location
		XYL = XYL+XYS;	

	% rebound at enclosure walls
		[XYL] = ENCLOSE(Nt,XYL,SIZE.x,SIZE.y,Ndots); 

	% save step of first dot (for trace)
		XYLp(:,Nt) = XYL(:,1);
	
	% plot diffusion
		TracePlot(Nt,XYL,Ph1,XYLp,1)

end %[for loop]
%===================================%

REFERENCES

Wiki References

FOOTNOTES