Cognitive and emotional information processing: protein synthesis and gene expression

Abstract

Recent findings suggest that functional plasticity phenomena such as long-term potentiation (LTP) and long-term depression (LTD) – cellular processes underlying memory – are restricted to functional dendritic compartments. It was also shown, however, that a relatively strong activation of a synaptic input can abolish compartment restrictions. Our data support these findings and we present one cellular pathway responsible for uncompartmentalization of the normally localized plasticity processes by the action of rolipram, an inhibitor of type 4 phosphodiesterases. In contrast with compartment-restricted information processing, uncompartmentalization requires transcription. In the search for system relevance of compartmentalization versus uncompartmentalization we describe firstly data which show that more cognitive information processing in rats' behaviour may follow rules of compartmentalization, whereas stressful, more life-threatening, inputs abolish compartment-restricted information processing involving transcription. Our findings allow us to suggest that consolidation of processes which take place during the cognitive event most probably depend on local protein synthesis, whereas stress immediately induces gene expression in addition, resulting in a compartment-unspecific up-regulation of plasticity-related proteins (PRPs), providing the entire neuron with a higher level of ‘reactiveness’. These data would provide a specific functional cellular mechanism to respond differentially and effectively to behaviourally weighted inputs.

Hippocampal CA1 pyramidal neurons receive their inputs in apical and basal dendrites from other hippocampal neurons and from various extrahippocampal structures including neuromodulatory inputs from the amygdala, ventral tegmental area and many others (Amaral & Kurz, 1985; Alkon et al. 1991; Amaral & Witter, 1998; Pikkarainen et al. 1999; Lisman & Grace, 2005). The information from these system-wide inputs, such as spatial, contextual, relational information or information containing system-relevant ‘decisions’, arrives as modulatory information. This must be integratively processed in target neurons to enable a final encoding, storage, integration or transmission of the information at/to the various networks in which the particular neuron takes part. Apical and basal dendrites of CA1 pyramidal neurons do not, however, just differ in their innervation but also in their morphology and physiological properties (Kaibara & Leung, 1993; Aika et al. 1994; Leung & Shen, 1995; Ishizuka et al. 1995; Çavus & Teyler, 1998; Pike et al. 2000; Inoue et al. 2001). Thus, many questions arise from these differences, such as whether system-relevant information processing and integration are restricted to single synapses or to dendritic compartments or to the whole neuron.

Functional compartmentalization: heterosynaptic requirements, protein synthesis, synaptic tagging and cross-tagging

Recent findings suggest that cellular information processing takes place rather in functional neuronal compartments than in single synapses (Frey, 1997; Frey, 2001; Alarcon et al. 2006; Govindarajan et al. 2006; Frey & Frey, 2007; Huang et al. 2006; Reymann & Frey, 2007; Sajikumar et al. 2007). The first hint for such an assumption has been given from findings that late-LTP and late-LTD in different dendritic compartments of CA1 pyramidal cells and granular cells of the dentate gyrus can be modulated by inhibitors and activators of aminergic or metabotropic glutamatergic receptors as well as the action of opioids which do not specifically innervate the same glutamatergic synapse where LTP or LTD was induced (for review see Frey, 1997, 2001; Sajikumar & Frey, 2004a; Navakkode et al. 2005, 2007). Furthermore, late-LTP and late-LTD require protein synthesis for their expression induced by heterosynaptic interactions within locally restricted dendritic compartments (Frey, 2001; Reymann & Frey, 2007; Frey & Frey, 2007; Sajikumar et al. 2007).

As far back as the 1980s we suggested that late-LTP (and later also late-LTD) requires concomitant activation of different transmitter systems (Frey et al. 1989, 1990, 1991b, 1997; Matthies et al. 1990; Frey & Morris, 1998a; Sajikumar & Frey, 2004a). The typical LTP/LTD experiment, involving a brief period of high-frequency stimulation, or a short period of low-frequency stimulation in the case of LTD, overlooks the more likely situation in the behaving organism, where the activation of a population of glutamatergic synapses at a dendritic compartment of a given neuron is likely to be accompanied by a dynamic activation of non-glutamatergic heterosynaptic inputs to that dendritic branch. The tetanization/low-frequency stimulation used in LTP/LTD studies, involving simultaneous field activation of hundreds of fibres, activates more than one kind of neurotransmitter input, and it is that cooperative action of inputs that induces late-LTP/LTD. We therefore favoured the notion that a time-dependent convergence of two or more events in a specific neuronal compartment is required for late-LTP/LTD. Given the effects of dopamine on learning (Flood et al. 1980), we investigated first whether dopamine is simultaneously released during LTP induction in the apical dendrites of CA1 neurons in the hippocampus, and whether it is required for late-LTP and late-LTD there (Frey et al. 1989, 1990, 1991b; Sajikumar & Frey, 2004a). The apical dendrites of the CA1 in the hippocampus are innervated by dopaminergic fibres that course through the mesolimbic pathway (Baulac et al. 1986), and there is evidence of the expression of the D5 receptor in CA1 pyramidal cells. The D5 receptor is related to the D1 dopamine receptor that is coupled to adenylyl cyclase. In addition to glutamate and possibly other neurotransmitters, dopamine levels increase during conventional LTP induction in the above region (Frey et al. 1990). To determine whether dopamine might be the additional activator necessary for late-LTP and/or late-LTD in this dendritic compartment, we showed that it plays a crucial role in the initiation of the mechanisms responsible for late-LTP as well as late-LTD in the apical dendrites of hippocampal CA1 neurons (Sajikumar & Frey, 2004a; Navakkode et al. 2007). When specific inhibitors of the dopaminergic D1 and D2 receptors were administered during tetanization or LTD induction, late-LTP/LTD was prevented (Frey et al. 1990, 1991b; Frey, 1997; Sajikumar & Frey, 2004a; Navakkode et al. 2007). Our results suggested that the influences of aminergic transmitters are not only modulatory. We have demonstrated that activation of dopaminergic inputs to hippocampal CA1 neurons plays a critical role, perhaps acting as a cellular switch to establish the late phase of LTP or LTD, since the time course of LTP/LTD decay after the application of dopamine antagonists is similar to that seen after protein synthesis inhibition. It was shown that repeated application or transient application (Huang & Kandel, 1995; Frey & Morris, 1998a; Sajikumar & Frey, 2004a; Navakkode et al. 2007) of a D1/D5 receptor agonist in CA1 synapses in adult animals induces a delayed potentiation or depression whose time courses are similar to those found after administration of dopamine. Interestingly, whether LTP or LTD is induced depends on only relatively small differences in the concentration of applied dopamine (Sajikumar & Frey, 2004a). Thus, repeated administration of 10 μm dopamine revealed late-LTD whereas 50 μm resulted in late-LTP if measured in apical CA1 dendrites. For the late event to occur we have recently shown that a synergistic glutamatergic input during dopamine application (Navakkode et al. 2007) is required which points to analog properties observed during electrical inductions of the late event. There it was shown that a threefold tetanization resulted in transient increased cAMP levels which were also dependent on NMDA receptor function (Frey et al. 1993). The fact that small differences in the concentration of the modulatory transmitter can induce opposite plasticity events suggests that one and the same synapse (or synapse population representing a functional input) can express both forms of the late form of either LTP or LTD.

With respect to the heterosynaptic requirements for inducing late-LTP/LTD, there is now growing evidence that in adult brains non-glutamatergic inputs, such as dopaminergic ones in the apical dendritic layer of CA1 pyramidal neurons, modulate the availability of plasticity-related proteins (PRPs) in synergism with NMDA receptor function (Frey et al. 1993; Navakkode et al. 2007), whereas the glutamatergic input alone mainly mediates the maintenance processes of LTP during early LTP as well as the setting of the tag (for review see Frey, 2001). Synergistic interactions with heterosynaptic inputs during establishment of early LTP/LTD as well as setting the tags cannot, however, be ruled out. Our findings on the requirement of dopamine for late-LTP in apical dendrites of hippocampal CA1 neurons have recently been supported by studies from the Morris laboratory in which some of the above experiments were reinvestigated (O'Carroll & Morris, 2004).

The question then arose how the above processes – induced by different functional synaptic and ‘neuromodulatory’ inputs – could systematically interact in a neuron to assure a differentiated but integrated information processing during memory formation. Our recent data and the results presented here suggest that instead of the formation of a memory trace within a single synapse, functional synaptic populations determined by local morphological characteristics as well as by the means of specific ‘neuromodulatory’ innervation, create dendritic functional compartments under normal circumstances. In addition to our findings, this hypothesis is supported by several other laboratories (Mel, 1993; Engert & Bonhoeffer, 1997; Kelleher et al. 2004; Polsky et al. 2004; Govindarajan et al. 2006).

In addition to the above findings with respect to the proposed dendritic functional compartments and the observation that electrically induced LTP/LTD is synaptic input specific, we have also identified new associative properties during LTP and LTD termed ‘synaptic tagging’ and ‘cross-tagging’ (Frey & Morris, 1997; Sajikumar & Frey, 2004a; Alarcon et al. 2006). How pre-existing or newly synthesized PRPs interact with specific, activated synapses expressing LTP (or, also LTD, e.g. see Sajikumar & Frey, 2004a), but not with non-activated synapses, is fundamental to the synapse specificity thought critical for information processing and memory formation. We explained synaptic input specificity by the concept of ‘synaptic tagging’, in which newly synthesized PRPs activated by heterosynaptic interactions bind to recently potentiated, glutamatergic ‘tagged’ synapses, thus maintaining LTP and input specificity (Frey & Morris, 1997, 1998a,b). We had also shown that stimulation that normally leads to early LTP could also induce late-LTP if a separate pathway had been strongly tetanized within a specific time window. Thus, tetanization of a pathway can induce a LTP with variable persistence as a function of the prior history of activation of the neuron (Frey & Morris, 1998b). The tag is transiently active with an expected half-life of about 30 min in the intact animal. It has also been suggested that the PRPs are characterized by a specific, relatively short half-life of about 1–2 h (see for review Korz & Frey, 2004; Sajikumar et al. 2005a). Only in the event of both processes, the synapse-specific tag as well as PRPs, being available, can the two interact and transform early-LTP into late-LTP at the stimulated synapses (Frey & Morris, 1998b). The existence of tag and PRP dynamics therefore determines an effective, functionally important time window during which a normally transient form of functional plasticity can be transformed into a long-lasting one. ‘Synaptic tagging’ can therefore also explain how a strong tetanus, producing protein synthesis-dependent LTP (late-LTP) in one pathway, can prolong the potentiation in an independent, weakly tetanized pathway that would normally have produced only early-LTP, but with a tag set (Frey & Morris, 1998b). The late-associative time window is not only determined by the mechanistic half-life of tags and PRPs. Synaptic tags or tag complexes cannot only be passively deactivated by cellular degradational processes but also actively through distinct electrical stimulation such as by depotentiating stimuli within 5–10 min after the setting of the tag by specific stimulation (Stäubli & Scafidi, 1999; Sajikumar & Frey, 2004b). We are currently studying the functional relevance of these processes in the intact animal (see also below). During the last few years, our results on tagging have been verified in various laboratories (Kauderer & Kandel, 2000; Dudek & Fields, 2002; Barco et al. 2002; Adams & Dudek, 2005; Young & Nguyen, 2005; Barco et al. 2005), and it was also shown that analog tagging processes can occur during animals' behaviour (Korz et al. 2001; Moncada & Viola, 2007) as well as in invertebrates' neurons (e.g. Martin et al. 1997). Furthermore, tagging was not only described for LTP but also LTD in mammals (Kauderer & Kandel, 2000; Sajikumar & Frey, 2004a; Sajikumar et al. 2005b) characterized by similar cellular and functional properties (Sajikumar & Frey, 2004a). Recently, the synaptic tagging model has been expanded to include functional interactions between LTP and LTD, referred to as ‘cross-tagging’ (Sajikumar & Frey, 2004a; Sajikumar et al. 2005b). ‘Cross-tagging’ describes the capability of late-LTP/late-LTD in one synaptic input (S1) to transform the opposite, protein synthesis-independent early-LTD/early-LTP in an independent synaptic input (S2) into its long-lasting form (Sajikumar & Frey, 2004a). Cross-tagging not only expands the repertoire of functional interactions between different synapses and afferent pathways, but also raises the following fundamental question: Do neuromodulatory brain structures determine local, functional compartments including the specificity of tags and PRPs? This question addresses the fundamental issue as to whether distinct functional neuronal compartments exist, or if induction of a late plasticity event affects the entire pyramidal neuron. The data presented here strongly support our hypothesis that distinct functional compartments exist under normal conditions which are characterized by specific properties and restrictions.

In sum, ‘tagging’ processes are required for an input-specific integration of sensory information as required for the formation of a protein synthesis-dependent late stage of hippocampal LTP (Frey, 1997; Frey & Morris, 1997, 1998a). In short, under particular circumstances a sensory afferent stimulation activates – by either LTP- or LTD-inducing stimuli to hippocampal apical CA1 dendrites – a transient synaptic tag and induces the local synthesis of PRPs (Frey, 2001). These PRPs can only bind to synapses with a tag set, thus guaranteeing synaptic input specificity. It was suggested that afferent signals (LTP or LTD) should be integrated at glutamatergic synapses for a transient period of time of up to 4 h maintaining LTP or LTD and expressing a tag during the first 30 min (Frey, 1997; Reymann & Frey, 2007). In parallel, the organism/system has the capacity to evaluate the relevance of this signal and if sufficiently important to send an associative, heterosynaptic signal via a ‘neuromodulatory’ input within a time window of 60 min to the hippocampal target neuron. This input activates the synthesis of PRPs which are able to bind to synaptic tags transforming the normally transient memory trace into a long-lasting one. Since extrahippocampal heterosynaptic ‘neuromodulatory’ inputs innervate only distinct parts of a neuron's dendrite within the hippocampus it can be assumed that these neuromodulatory inputs determine functional compartments (Frey & Frey, 2007). Interestingly, recent data have shown that synaptic tagging as well as cross-tagging is restricted to neuronal functional compartments under normal conditions, such as after more physiological stimulation (Alarcon et al. 2006; Sajikumar et al. 2007; Reymann & Frey, 2007; Frey & Frey, 2007). Under distinct circumstances, however, such as after very strong activation, the barrier of compartmentalized plasticity can be broken (Alarcon et al. 2006). The authors provided evidence that a fourfold tetanization of apical dendritic inputs in combination with a twofold tetanization of basal dendritic inputs led to the partial loss of the restriction of LTP-tagging to apical dendrites. Basal inputs with early LTP induction and tags set could then benefit from the PRPs whose synthesis was initiated by the apical event and thus express late-LTP.

Uncompartmentalization: gene expression

The question now arises as to whether such uncompartmentalization of functional hippocampal plasticity may fulfil any relevant functional role and it raises the question of which intracellular events are responsible for the uncompartmentalization.

We have therefore first searched for putative cellular pathways involved in uncompartmentalization. Thus, we identified that inhibition of the phosphodiesterase 4B3 (PDE4B3, for which a role as a process-unspecific PRP during LTP and LTD was shown (Navakkode et al. 2004; Navakkode et al. 2005; Reymann & Frey, 2007)) by rolipram is capable of breaking compartment-restricted LTP induced by a weaker, more physiological tetanization, which normally induces a compartment-restricted form (Alarcon et al. 2006; Sajikumar et al. 2007). If rolipram is applied together with compartmentalized LTP induction in a synaptic input rS1 (Fig. 1B, filled circles) in the stratum radiatum of hippocampal slices in vitro, compartmentalization of tagging processes to basal dendrites is abolished. If, within an effective time window (Frey & Morris, 1998b) early LTP is induced in separate synaptic inputs oS3 (Fig. 1B, filled triangles) and oS4 (Fig. 1B, open triangles) at the basal dendrite, these early LTPs are transformed into late-LTPs by means of synaptic tagging. Surprisingly, tagging is not restricted any longer to the apical dendrites but the tagging process spreads throughout the neuron, i.e. tags far away from the late-plasticity process can now benefit from PRPs whose synthesis was induced by the latter event. In a following set of experiments we have now electrically induced the strong plasticity event, i.e. late-LTP in a synaptic input oS3 in the stratum oriens (Fig. 1C), but also under the influence of rolipram. As can be seen in Fig. 1C, strong tetanization together with rolipram in oS3 also abolishes the restriction of tagging events to the basal dendrites which would be normally restricted to that compartment if a more physiological LTP had been induced (Alarcon et al. 2006) without boosting it by stronger tetanization and inhibited PDE4B3 function. Under these conditions, tagging does not only occur between basal inputs oS3 and oS4 but it spreads also to distant synapses in the apical dendrites rS1, where the early-LTP is now transformed into late-LTP by the capture of PRPs synthesized via the rolipram-boosted strong event in oS3. Late-LTP induced by a very strong tetanization which causes uncompartmentalization requires protein synthesis (Alarcon et al. 2006) in a similar way to rolipram-reinforced LTP (Navakkode et al. 2004). Interestingly, and in contrast with a more physiologically induced late-LTP (Fig. 2A and B), the boosted forms of late-LTP as well as their related processes of transcompartmentalized tagging/capturing of PRPs now also require gene expression in addition to translation (Fig. 1D and E). We tested the effect of two structurally different, irreversible mRNA synthesis inhibitors, actinomycin D (ACD, Fig. 1D) and 5,6-dichloro-1-β-d-ribofuranosyl benzimidazole (DRB, Fig. 1E). Application of these drugs before LTP induction and boosting by rolipram prevented all forms of late-LTP including compartmentalized as well as uncompartmentalized tagging/capturing of PRPs (Fig. 1D and E). In sum, unlike normal protein synthesis-dependent late-LTP (Frey, 2001; Sajikumar et al. 2005a; Reymann & Frey, 2007; Frey & Frey, 2007) rolipram-boosted uncompartmentalization is characterized by an immediate requirement of trans-cription. Conventional late-LTP under our conditions does not require transcription for its initial maintenance of about 8 h but depends on protein synthesis from local dendritic, pre-existing mRNA (Otani et al. 1989; Frey, 1997; Frey & Frey, 2007) (Fig. 2).

Figure 1. Uncompartmentalization of synaptic tagging by rolipram.

A represents the schematic location of the electrodes to stimulate two separate synaptic inputs rS1 and rS2 in stratum radiatum (red colour throughout) in rat hippocampal slices in vitro and the location of additional electrodes to stimulate independent basal synaptic input (oS3 and oS4, blue colour) to the CA1-pyramidal neuron. The location of the recording electrodes for the population spike (dotted lines) as well as the field EPSP is given. Furthermore, a third field EPSP recording electrode is placed in the area of the basal dendrites between stimulating electrodes oS3 and oS4. When the stratum oriens field EPSP was directly recorded, the electrode for recording the population spike (dotted electrode) was omitted. The analog traces near to the recording electrodes show representative examples of recorded potentials. B, after recording a baseline, early LTP was induced by a weak tetanus (WTET, open arrow) in a basal synaptic input oS3 (filled triangles). Thirty minutes later, WTET was applied to an independent basal synaptic input oS4 (open triangles). Forty five minutes after the first tetanization rolipram, a PDE4B3 inhibitor (black box), was applied to the bath medium for a duration of 30 min. One hour after WTET to oS3, late-LTP was induced in synaptic input rS1 of apical dendrites in the stratum radiatum (filled circles) using a strong tetanus (STET, filled arrow) under the influence of rolipram. The rolipram-boosted STET in rS1 was able to transform the normally early LTPs in oS3 and oS4 into late-LTP. The LTPs in all inputs were statistically significantly different from their baseline before tetanization during the entire experiment (Wilcoxon signed-rank test; P < 0.05). Insets show representative potentials for the given inputs 30 min before (dotted line), 30 min after (dashed line) and 6 h after (continuous line) tetanization. C, early LTP was induced by a weak tetanus (WTET) in the apical synaptic input rS1 (filled circles). Thirty minutes later, WTET was applied to an independent basal synaptic input oS4 (open triangles). Fourty five minutes after the first tetanization, rolipram, a PDE4B3 inhibitor, was applied to the bath medium for a duration of 30 min. One hour after WTET to rS1, late-LTP was induced in a separate and independent input of the basal dendrites oS3 (filled triangles) under the influence of rolipram. The rolipram-boosted STET in oS3 was also able to transform the normally early LTPs in oS4 and rS1 into late-LTPs. The LTPs in all inputs were statistically significantly different from their baseline before tetanization during the entire experiment (Wilcoxon signed-rank test; P < 0.05). D, after recording a baseline for 45 min, the irreversible mRNA-synthesis inhibitor actinomycin D (ACD, 25 μm; hatched box) was applied to the bath medium for 30 min before washout. Forty-five minutes after drug washout, early LTP was induced by WTET to oS3 (filled triangles), followed by WTET to oS4 30 min later (open triangles). Then, 15 min later WTET to oS4 rolipram was applied for 30 min. Thirty minutes after WTET to oS4 a STET under the influence of rolipram was applied to rS1. As can be seen from the graph, all late forms of LTP, including transformation of early LTP to late-LTP by synaptic tagging, were prevented by the mRNA-synthesis inhibitor. The potentiation in oS3 was statistically significantly different when compared with its own baseline before tetanization for up to 60 min, potentiation in oS4 for up to 220 min and in rS1 for up to 255 min (P < 0.05, Wilcoxon signed-rank test). E, similar experimental design as in D but instead of ACD a second, structurally different inhibitor 5,6-dichloro-1-β-d-ribofuranosyl benzimidazole (DRB, 100 μm) was applied. DRB revealed a similar outcome as after the application of ACD - all late LTP-forms were prevented. The potentiation in oS3 was statistically significantly different when compared with its own baseline before tetanization for up to 45 min, potentiation in oS4 for up to 180 min and in rS1 for up to 240 min (P < 0.05, Wilcoxon signed-rank test). The drugs had no effect on baseline recordings (Frey et al. 1996). We used 56 transverse hippocampal slices (400 μm), prepared from 56 male Wistar rats (7 weeks old), as previously described (Frey & Morris, 1997; Sajikumar et al. 2005a, 2007). Anisomycin (Sigma), a reversible protein synthesis inhibitor, was used at a concentration of 25 μm (dissolved in ACSF and 0.1% DMSO which had no effect on control recordings). The concentration used blocked at least 85% of incorporation of [³H]leucine into hippocampal slices (Frey et al. 1991a). Emetine (Tocris) was used at a concentration of 20 μm (dissolved in ACSF and 0.1% DMSO). Rolipram (Tocris), a type IV phosphodiesterase inhibitor, was used at a concentration of 0.1 μm dissolved in ACSF and 0.1% dimethylsulfoxide. The transcriptional inhibitors actinomycin D (Tocris, 25 μm) or DRB (100 μm; Sigma; dissolved in ACSF and 0.1% DMSO) were used in according experiments.

Figure 2. The effect of the mRNA synthesis inhibitor actinomycin D or the protein synthesis blocker anisomycin on tagging processes in hippocampal CA1 in vitro.

A, after the recording of a stable baseline, the irreversible mRNA blocker ACD was applied for 30 min and then washed out. Two hours after the recording of the baseline early LTP was induced in the basal input oS3 by a WTET (filled triangles). Forty-five minutes later late-LTP was induced in oS4 (open triangles) by STET. The drug did not influence late-LTP in oS4 and did not prevent synaptic tagging/capturing to oS3, i.e. in the basal dendritic compartment. Both inputs were potentiated over the recorded time period of 6 h (Wilcoxon signed-rank test, P < 0.05). B, after the recording of a stable baseline, the irreversible mRNA blocker ACD was applied for 30 min and then washed out. Two hours after the recording of the baseline early LTP was induced in the stratum radiatum, i.e. the apical dendrite of a CA1 neuron, rS1 by a WTET (filled circles). Forty-five minutes later late-LTP was induced in the apical rS2 (open circles) by STET. The drug did not influence late-LTP in rS2 and did not prevent synaptic tagging/capturing to rS1, i.e. in the apical dendritic compartment. Both inputs were potentiated over the recorded time period of 6 h (Wilcoxon signed-rank test, P < 0.05). C and D, similar experimental design as in A and B, respectively. A baseline was only recorded for 60 min, however, before first tetanization. Instead of actinomycin D, the reversible protein synthesis inhibitor anisomycin was applied 30 min before STET to oS4 (C) or rS2 (D), respectively, with a duration of 60 min before washout. As both graphs reveal, the late form of LTP as well as synaptic tagging is prevented in the apical as well as the basal dendritic compartments by the protein synthesis inhibitor (basal compartment: significant potentiation in oS3 up to 150 min (filled triangles) and in oS4 up to 205 min (open triangles) after tetanization; apical compartment: significant potentiation in rS1 up to 180 min (filled circles) and in rS2 up to 225 min (open circles) after tetanization; Wilcoxon signed-rank test, P < 0.05). Analog traces again represent potentials 30 min before (dotted line), 30 min after (dashed line) and 6 h after tetanization (continuous line). Hatched boxes represent drug application.

Is there any physiological relevance of uncompartmentalization? It has been shown that late plasticity events such as LTD in the CA1 or LTP in the hippocampal dentate gyrus can be enhanced, reinforced or prevented by the associative stimulation of heterosynaptic neuromodulatory inputs, including stress-related glucocorticoid receptor function (Xu et al. 1997, 1998; Frey et al. 2001; Li et al. 2003; Richter-Levin & Akirav, 2003; Korz & Frey, 2003, 2004; Uzakov et al. 2005). LTD and LTP may serve different memory storage mechanisms (Xu et al. 1997; Kemp & Manahan-Vaughan, 2004, 2007; Lemon & Manahan-Vaughan, 2006) whose underlying mechanisms and relation to behaviour are not yet fully understood. Our working hypothesis is that sensory-conditioning information can be automatically but transiently stored at glutamatergic synapses (see above). During that time a transient synaptic tag can be set within these synapses. Whether the information is consolidated into a long-term memory trace depends on evaluation of the information by the system within a time window of about 60 min (Frey, 1997, 2001; Korz & Frey, 2004; Reymann & Frey, 2007; Sajikumar et al. 2007). As mentioned above for the cellular event of LTP, if the organism ‘decides’ that the information is system-relevant it can send an associative heterosynaptic unconditioning signal to the neurons where the information is transiently stored and a tag is set within the effective time window of about 30 min after the automatic recording of the sensory information. Interestingly, prior system relevant states also seem to be able to prime the target neuron by activating the neuromodulatory input up to 30 min before the subsequent sensory information is automatically stored (Frey, 2001). The neuromodulatory inputs thereby regulate the level of available PRPs which can then bind to the tags, thus transforming the normally transient event into a long-lasting one. If both inputs, the glutamatergic input (conditioning stimulus) as well as the neuromodulatory one (unconditioning stimulus), are associatively activated within this time window, a long-term memory can then be formed (for review see Frey, 2001; Reymann & Frey, 2007; Sajikumar et al. 2007; Frey & Frey, 2007). Taking into consideration the above results and theoretical aspects, we were now interested to search for the impact on the system of translation-dependent compartmentalized neuronal information processing versus translation and transcription-dependent uncompartmentalized neuronal information processing.

We had recently shown that early LTP in the dentate gyrus could be reinforced into late-LTP if its induction was associatively paired with a more cognitive spatial (holeboard training, Fig. 3A, open circles) or stressful learning protocol (swim-stress in the watermaze, Fig. 3B, open circles) (Korz & Frey, 2003, 2004; Uzakov et al. 2005). Both forms of LTP reinforcement were translation dependent (Korz & Frey, 2003; Uzakov et al. 2005). The question remains whether LTP reinforcement would also require transcription. As mentioned above, we found no difference in early-LTP induction by the different behavioural tasks. Application of a transcription inhibitor had no effect on the LTP reinforcement in the holeboard task (Fig. 2A), suggesting that gene expression is not required for LTP reinforcement by the cognitive task. In strong contrast, the LTP reinforcement in the watermaze (Fig. 3B, open circles) was prevented by inhibition of transcription (Fig. 3B, filled circles). Watermaze swim-stress-induced LTP reinforcement requires transcription, in addition to translation. As shown in Fig. 3C, the swim in the watermaze caused a dramatic increase in the corticosterone response compared with the holeboard training, supporting our conclusion that a 2 min watermaze swim is quite stressful for the rat.

Figure 3. Behavioural LTP reinforcement and the requirement for mRNA synthesis in stress condition.

A represents the time course of early LTP reinforced into late-LTP by the associative interaction with a holeboard task (open circles). Application of ACD (72 μg (12 μl ACSF)⁻¹; 2 μl min⁻¹, an effective concentration taken from Manahan-Vaughan et al. 2000) into the ventricle 2 h before tetanization had no effect on LTP reinforcement in the holeboard (filled circles). LTP was statistically significantly different from baseline in both the treated as well as the control group up to 8 h after tetanus (Wilcoxon signed-rank test, P < 0.05). B shows the time course of LTP reinforcement by a 2 min swim episode in the watermaze (open circles) and its prevention 3 h post-tetanus (no statistical difference from baseline up to 6 h, Wilcoxon signed-rank test, P > 0.05; and statistically below baseline at time points 7 h and 8 h, Wilcoxon signed-rank test, P < 0.05) if ACD was applied 2 h before tetanization (filled circles). The control potentials, in contrast, remained elevated over the whole time course (P < 0.05, each time point). The lower parts in A and B show representative examples of the evoked potentials 30 min before (dotted line), 30 min after (dashed line) and 8 h after tetanization (continuous line). The bar graphs in C demonstrate the corticosterone response in blood samples 15 min after swimming (open box) or after the first or 10th trial in the holeboard (grey boxes). A statistically significant difference could be detected between samples (χ²= 8.06, d.f. = 2, P < 0.05), with elevated concentrations in the stressed group as compared with holeboard-trained animals (trial 1: U = 10, trial 10: U = 8, P < 0.05 each; Kruskal–Wallis H test with subsequent Mann–Whitney U tests).For the analysis of the combined electrophysiological/behavioural experiments 34 rats were used. Electrophysiology: Male rats (7 weeks old) were anaesthetized with Nembutal (50 mg kg⁻¹, i.p.). A monopolar recording electrode was implanted stereotaxically into the granule cell layer of the dentate gyrus (coordinates: AP −2.8 mm, L 1.8 mm from bregma, 3.2–3.5 mm ventral from dura) and a bipolar stimulation electrode into the perforant path (coordinates: AP −6.9 mm, L 4.1 mm, 2.2–2.5 mm ventral from dura) of the right hemisphere. Each electrode consisted of an insulated stainless steel wire 125 μm in diameter. During preparation, test pulses were delivered to optimize the population-spike amplitude (PSA). We used the PSA as a measure of LTP instead of the excitatory postsynaptic potential (EPSP) for several reasons. Working in vivo, with freely moving animals, and using a single recording electrode it is difficult to assess simultaneously the EPSP and the PSA. Our recording electrode was located in the hilus, and thus far away from the locus of EPSP generation. The potential in the recorded trace there, often mistakenly used as an EPSP, is therefore a mixture of different unknown fields including the EPSP component. Measurement of this component where possible revealed no differences, however, in the time course when compared with the PSA. Furthermore, the PSA represents the discharge of action potentials by the postsynaptic population, which is functionally the more relevant outcome of synaptic function. A cannula (coordinates: AP −0.8 mm, L 1.6 mm from bregma) was implanted into the lateral ventricle of the right hemisphere for pharmacological treatment. The animals were allowed at least 1 week to recover from surgery. For the experiments, rats were placed into a recording box (40 cm × 40 cm × 40 cm) and the electrodes connected to a swivel by a flexible cable. This allowed the freely moving animals ad libitum access to food and water. Biphasic-constant current pulses (0.1 ms per half-wave) were applied to the perforant path in order to evoke DG field potentials of about 40% of the maximum PSA. After a stable baseline had been registered for 1 h, LTP was induced by weak tetanic bursts (3 bursts of 15 pulses of 200 Hz with 0.1 ms duration of each stimulus and 10 s interburst interval, the same stimulus intensity as for PSA testing (200–400 μA)). Every 15 min after tetanization 5 test stimuli (10 s interpulse interval) were delivered and the mean values of field potentials were stored. Holeboard apparatus and procedure: The test apparatus consisted of a black board (1 m × 1 m) with 36 regularly arranged holes (6 cm in diameter and 8 cm deep) and transparent Plexiglas walls of 27 cm height around it (COGITAT by Cognitron GmbH, Göttingen, Germany). Technical equipment and furniture served as distal visual cues; additional cues were fixed on the outside of the Plexiglas walls. Photobeams were mounted at the surface, the middle and on the ground of each hole. The holes were baited with standard food pellets (dustless precision pellets, 45 mg, BioServ, Frenchtown, USA). Signals of the photobeams were registered, counted and stored on a PC by RatMemory V2.4 software. During experiments animals were transferred to the holeboard room into a start-box. The box was opened and the animals entered the test arena. Inspections were indicated by breakings of the surface beams; visits were counted if the beams in the middle of the holes were broken. Finding and removal of the pellets were indicated by breaks of the ground beam. A trial was automatically stopped after 2 min or when the animal had found all pellets. The time to find all pellets (latencies), the working memory errors (inspecting or visiting a hole that had been baited but that was already inspected or visited and the pellet picked up during a specific trial) and the reference memory errors (inspecting or visiting a hole that was unbaited) were counted. After each trial the animals were transferred back into the recording chambers. The board was cleaned with water if an animal urinated or defecated. Beneath the holeboard there was a second board on which food pellets were scattered randomly to avoid odour information from baited holes. All experimental and control animals received a spatial training on a fixed pattern of baited holes over 10 trials (5 trials on day 1, 4 trials on day 2 and the last trial on day 3), with a 20 min intertrial interval. Stress protocol: The circular water tank (1.82 m diameter and 58 cm height) was filled with latex-stained (Sakret, Gießen, Germany) opaque water up to a level of 38 cm with the temperature set to 25 ± 2°C. Before transference to the maze, the electrode connections were protected from water contact with Vaseline. One 2-min swim episode in the water tank served as an emotional challenge. After swimming, animals were towel-dried, transferred back into the recording chamber and LTP was recorded as described above. The swim path was recorded on a video camera and fed to a video-tracking (HVS image analyser, VP200) and analysis system (Watermaze by Richard Morris and Roger Spooner v. 2.17a) located in an adjacent room. Pharmacology: On the last day a 1 h baseline was performed, immediately after which either actinomycin D (72 μg dissolved in ACSF, 12 μl, 2 μl min⁻¹) or vehicle were injected. Two hours after injection and 15 min before the last trial or the stress event (between 11:15 and 11:30 h) animals received a weak tetanus. Holeboard-trained animals were provided with two food pellets per day at random time points during training and testing to avoid anticipation: access to water was ad libitum. Stress-hormone level determination: Stress hormones were determined in unprepared animals that received the same treatment as the prepared animals. Fifteen minutes after the last holeboard trial or the stress event, respectively, the animals were decapitated and trunk blood was collected in an Eppendorf tube and allowed to coagulate on ice for 30 min. Then the blood was centrifuged and the serum was stored at −20°C. The time from opening the recording chamber until finishing the blood sample did not exceed 20 s. Samples were analysed by a radio-immuno-assay (Korz & Frey, 2003).

In order to interpret the results of uncompartmentalization of plasticity events at the cellular level and the system's data, we suggest that under normal, low-stress conditions information processing takes place in neuronal nets restricted to functional neuronal compartments. The information is stored specifically within distinct networks representing a specific content. This form of consolidation of a memory trace does not require transcription but translation at least during the first 8 h. If, however, an acute stress event comes into play, the organism reacts by boosting its information processing systems: it sends a non-local signal via stress hormones to adequate neurons whose capability to handle afferent information is dramatically increased by the induction of transcription-dependent synthesis of PRPs which are not restricted to single neuronal compartments but can reach all activated synapses transforming a normally transient memory trace into a long-lasting one. Therefore, more neuronal networks – in which the given neuron interacts – can process and consolidate the more system-relevant information, induced by stress or life-threatening content, much more effectively, by losing, however, some of their content specificity.

Our data represent a first hint of how differentially weighted system-relevant information can be handled at the systems-neuronal level.