Modeling signal transduction in classical conditioning with network motifs

Synaptic Mechanisms of Delay Eyeblink Classical Conditioning: AMPAR Trafficking and Gene Regulation in an In Vitro Model

An in vitro model of delay eyeblink classical conditioning was developed to investigate synaptic plasticity mechanisms underlying acquisition of associative learning. This was achieved by replacing real stimuli, such as an airpuff and tone, with patterned stimulation of the cranial nerves using an isolated brainstem preparation from turtle. Here, our primary findings regarding cellular and molecular mechanisms for learning acquisition using this unique approach are reviewed. The neural correlate of the in vitro eyeblink response is a replica of the actual behavior, and features of conditioned responses (CRs) resemble those observed in behavioral studies. Importantly, it was shown that acquisition of CRs did not require the intact cerebellum, but the appropriate timing did. Studies of synaptic mechanisms indicate that conditioning involves two stages of AMPA receptor (AMPAR) trafficking. Initially, GluA1-containing AMPARs are targeted to synapses followed later by replacement by GluA4 subunits that support CR expression. This two-stage process is regulated by specific signal transduction cascades involving PKA and PKC and is guided by distinct protein chaperones. The expression of the brain-derived neurotrophic factor (BDNF) protein is central to AMPAR trafficking and conditioning. BDNF gene expression is regulated by coordinated epigenetic mechanisms involving DNA methylation/demethylation and chromatin modifications that control access of promoters to transcription factors. Finally, a hypothesis is proposed that learning genes like BDNF are poised by dual chromatin features that allow rapid activation or repression in response to environmental stimuli. These in vitro studies have advanced our understanding of the cellular and molecular mechanisms that underlie associative learning.

Systematic analysis of noise reduction properties of coupled and isolated feed-forward loops

Cells can maintain their homeostasis in a noisy environment since their signaling pathways can filter out noise somehow. Several network motifs have been proposed for biological noise filtering and, among these, feed-forward loops have received special attention. Specific feed-forward loops show noise reducing capabilities, but we notice that this feature comes together with a reduced signal transducing performance. In posttranslational signaling pathways feed-forward loops do not function in isolation, rather they are coupled with other motifs to serve a more complex function. Feed-forward loops are often coupled to other feed-forward loops, which could affect their noise-reducing capabilities. Here we systematically study all feed-forward loop motifs and all their pairwise coupled systems with activation-inactivation kinetics to identify which networks are capable of good noise reduction, while keeping their signal transducing performance. Our analysis shows that coupled feed-forward loops can provide better noise reduction and, at the same time, can increase the signal transduction of the system. The coupling of two coherent 1 or one coherent 1 and one incoherent 4 feed-forward loops can give the best performance in both of these measures.

Cellular behavior can be affected by noise in molecular interactions. Signaling pathways should process noisy input signals and support cellular decision making by properly transducing the signals, while removing noise from them. Three component networks of feed-forward loops (FFLs) have been proposed to serve as ideal noise reducers, while linear pathways were shown to be good signal transducers. These signaling units do not work in isolation, so there is a possibility that a combination of various feed-forward loops can provide good noise reduction, while maintaining good signal transduction. To test this hypothesis, we have systematically tested the noise reducing and signal transducing capabilities of all possible combinations of feed-forward loops and compared them with the performance of individual FFLs. We built mathematical models of all these systems and compared their capabilities at reducing noise in the input signal while maintaining responses to meaningful changes in the incoming signal. We found that a combination of two copies of a special type of fully positive signaling FFLs is the best noise reducer, while a combination of two incoherent (one positive, one negative signal) FFLs can provide the best signal transduction. The combination of these two FFLs could provide good signal processing where both noise reduction and signal transduction are achieved.

Amyloid fibrils induce dysfunction of hippocampal glutamatergic silent synapses

Silent glutamatergic synapses lacking functional AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazoleproprionate) receptors exist in several brain regions including the hippocampus. Their involvement in the dysfunction of hippocampal glutamatergic transmission in the setting of Alzheimer's disease (AD) is unknown. This study demonstrated a decrease in the percentage of silent synapses in rats microinjected with amyloid fibrils (Aβ1-40 ) into the hippocampal CA1. Also, pairing low-frequency electric stimuli failed to induce activation of the hippocampal silent synapses in the modeled rats. Immunoblotting studies revealed a decreased expression of GluR1 subunits in the hippocampal CA1 synaptosomal preparation, indicating a potential reduction in the GluR1 subunits anchoring in postsynaptic density in the modeled rats. We also noted a decreased expression of phosphorylated cofilin, which regulates the function of actin cytoskeleton and receptor trafficking, and reduced expression of the scaffolding protein PSD95 in the hippocampal CA1 synaptosome in rats injected with Aβ1-40 . Taken together, this study illustrates dysfunction of hippocampal silent synapse in the rodent model of AD, which might result from the impairments of actin cytoskeleton and postsynaptic scaffolding proteins induced by amyloid fibrils.

MeCP2 regulates Tet1-catalyzed demethylation, CTCF binding, and learning-dependent alternative splicing of the BDNF gene in Turtle

MECP2 mutations underlying Rett syndrome cause widespread misregulation of gene expression. Functions for MeCP2 other than transcriptional are not well understood. In an ex vivo brain preparation from the pond turtle Trachemys scripta elegans, an intraexonic splicing event in the brain-derived neurotrophic factor (BDNF) gene generates a truncated mRNA transcript in naïve brain that is suppressed upon classical conditioning. MeCP2 and its partners, splicing factor Y-box binding protein 1 (YB-1) and methylcytosine dioxygenase 1 (Tet1), bind to BDNF chromatin in naïve but dissociate during conditioning; the dissociation correlating with decreased DNA methylation. Surprisingly, conditioning results in new occupancy of BDNF chromatin by DNA insulator protein CCCTC-binding factor (CTCF), which is associated with suppression of splicing in conditioning. Knockdown of MeCP2 shows it is instrumental for splicing and inhibits Tet1 and CTCF binding thereby negatively impacting DNA methylation and conditioning-dependent splicing regulation. Thus, mutations in MECP2 can have secondary effects on DNA methylation and alternative splicing.

DOI:http://dx.doi.org/10.7554/eLife.25384.001

Editor's Highlight: Computational Modeling of Plasma Vitellogenin Alterations in Response to Aromatase Inhibition in Fathead Minnows

In vertebrates, conversion of testosterone into 17β-estradiol (E2) is catalyzed by cytochrome P450 (CYP) 19A aromatase. An important role of E2 in oviparous vertebrates such as fish is stimulation of hepatic synthesis of the glycolipoprotein vitellogenin (VTG), an egg yolk precursor essential to oocyte development and larval survival. In fathead minnows (FHMs) (Pimephales promelas) exposed to the aromatase inhibitor fadrozole, plasma VTG levels do not change in concert with plasma E2 levels. Specifically, while plasma VTG and E2 levels both drop quickly when aromatase is first inhibited, the recovery of plasma VTG upon cessation of aromatase inhibition is substantially delayed relative to the recovery of plasma E2. We modified an existing computational model of the FHM hypothalamic-pituitary-gonadal axis to evaluate alternative hypotheses that might explain this delay. In the first hypothesis, a feedback loop involving active transport of VTG from the blood into the ovary is used. The activity of the transporter is negatively regulated by ovarian VTG. In the second hypothesis, a type 1 coherent feed-forward loop is implemented in the liver. This loop has 2 arms, both requiring E2 activation. The first arm describes direct, canonical E2-driven transcriptional induction of VTG, and the second describes an E2-driven intermediate transcriptional regulator that is also required for VTG synthesis. Both hypotheses accurately described the observed VTG dynamics. This result could be used to guide design of laboratory experiments intended to determine if either of the motifs, or perhaps even both of them, actually do control VTG dynamics in FHMs exposed to aromatase inhibitors.

Regulation of BDNF chromatin status and promoter accessibility in a neural correlate of associative learning

Brain-derived neurotrophic factor (BDNF) gene expression critically controls learning and its aberrant regulation is implicated in Alzheimer's disease and a host of neurodevelopmental disorders. The BDNF gene is target of known DNA regulatory mechanisms but details of its activity-dependent regulation are not fully characterized. We performed a comprehensive analysis of the epigenetic regulation of the turtle BDNF gene (tBDNF) during a neural correlate of associative learning using an in vitro model of eye blink classical conditioning. Shortly after conditioning onset, the results from ChIP-qPCR show conditioning-dependent increases in methyl-CpG-binding protein 2 (MeCP2) and repressor basic helix-loop-helix binding protein 2 (BHLHB2) binding to tBDNF promoter II that corresponds with transcriptional repression. In contrast, enhanced binding of ten-eleven translocation protein 1 (Tet1), extracellular signal-regulated kinase 1/2 (ERK1/2), and cAMP response element-binding protein (CREB) to promoter III corresponds with transcriptional activation. These actions are accompanied by rapid modifications in histone methylation and phosphorylation status of RNA polymerase II (RNAP II). Significantly, these remarkably coordinated changes in epigenetic factors for two alternatively regulated tBDNF promoters during conditioning are controlled by Tet1 and ERK1/2. Our findings indicate that Tet1 and ERK1/2 are critical partners that, through complementary functions, control learning-dependent tBDNF promoter accessibility required for rapid transcription and acquisition of classical conditioning.

Coincidence detection in a neural correlate of classical conditioning is initiated by bidirectional 3-phosphoinositide-dependent kinase-1 signalling and modulated by adenosine receptors

How the neural substrates for detection of paired stimuli are distinct from unpaired stimuli is poorly understood and a fundamental question for understanding the signalling mechanisms for coincidence detection during associative learning. To address this question, we used a neural correlate of eyeblink classical conditioning in an isolated brainstem from the turtle, in which the cranial nerves are directly stimulated in place of using a tone or airpuff. A bidirectional response is activated in <5 min of training, in which phosphorylated 3-phosphoinositide-dependent kinase-1 (p-PDK1) is increased in response to paired and decreased in response to unpaired nerve stimulation and is mediated by the opposing actions of neurotrophin receptors TrkB and p75(NTR) . Surprisingly, blockade of adenosine 2A (A2A ) receptors inhibits both of these responses. Pairing also induces substantially increased surface expression of TrkB that is inhibited by Src family tyrosine kinase and A2A receptor antagonists. Finally, the acquisition of conditioning is blocked by a PDK1 inhibitor. The unique action of A2A receptors to function directly as G proteins and in receptor transactivation to control distinct TrkB and p75(NTR) signalling pathways allows for convergent activation of PDK1 and protein kinase A during paired stimulation to initiate classical conditioning.

Identification of a functionally distinct truncated BDNF mRNA splice variant and protein in Trachemys scripta elegans

Brain-derived neurotrophic factor (BDNF) has a diverse functional role and complex pattern of gene expression. Alternative splicing of mRNA transcripts leads to further diversity of mRNAs and protein isoforms. Here, we describe the regulation of BDNF mRNA transcripts in an in vitro model of eyeblink classical conditioning and a unique transcript that forms a functionally distinct truncated BDNF protein isoform. Nine different mRNA transcripts from the BDNF gene of the pond turtle Trachemys scripta elegans (tBDNF) are selectively regulated during classical conditioning: exon I mRNA transcripts show no change, exon II transcripts are downregulated, while exon III transcripts are upregulated. One unique transcript that codes from exon II, tBDNF2a, contains a 40 base pair deletion in the protein coding exon that generates a truncated tBDNF protein. The truncated transcript and protein are expressed in the naïve untrained state and are fully repressed during conditioning when full-length mature tBDNF is expressed, thereby having an alternate pattern of expression in conditioning. Truncated BDNF is not restricted to turtles as a truncated mRNA splice variant has been described for the human BDNF gene. Further studies are required to determine the ubiquity of truncated BDNF alternative splice variants across species and the mechanisms of regulation and function of this newly recognized BDNF protein.

Two-stage AMPA receptor trafficking in classical conditioning and selective role for glutamate receptor subunit 4 (tGluA4) flop splice variant

Previously, we proposed a two-stage model for an in vitro neural correlate of eyeblink classical conditioning involving the initial synaptic incorporation of glutamate receptor A1 (GluA1)-containing α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid type receptors (AMPARs) followed by delivery of GluA4-containing AMPARs that support acquisition of conditioned responses. To test specific elements of our model for conditioning, selective knockdown of GluA4 AMPAR subunits was used using small-interfering RNAs (siRNAs). Recently, we sequenced and characterized the GluA4 subunit and its splice variants from pond turtles, Trachemys scripta elegans (tGluA4). Analysis of the relative abundance of mRNA expression by real-time RT-PCR showed that the flip/flop variants of tGluA4, tGluA4c, and a novel truncated variant tGluA4trc1 are major isoforms in the turtle brain. Here, transfection of in vitro brain stem preparations with anti-tGluA4 siRNA suppressed conditioning, tGluA4 mRNA and protein expression, and synaptic delivery of tGluA4-containing AMPARs but not tGluA1 subunits. Significantly, transfection of abducens motor neurons by nerve injections of tGluA4 flop rescue plasmid prior to anti-tGluA4 siRNA application restored conditioning and synaptic incorporation of tGluA4-containing AMPARs. In contrast, treatment with rescue plasmids for tGluA4 flip or tGluA4trc1 failed to rescue conditioning. Finally, treatment with a siRNA directed against GluA1 subunits inhibited conditioning and synaptic delivery of tGluA1-containing AMPARs and importantly, those containing tGluA4. These data strongly support our two-stage model of conditioning and our hypothesis that synaptic incorporation of tGluA4-containing AMPARs underlies the acquisition of in vitro classical conditioning. Furthermore, they suggest that tGluA4 flop may have a critical role in conditioning mechanisms compared with the other tGluA4 splice variants.