|Home | About | Journals | Submit | Contact Us | Français|
To explore how gene products, required for the initiation of synaptic growth, move from the cell body of the sensory neurons to its presynaptic terminals, and from the cell body of the motor neuron to its postsynaptic dendritic spines, we have searched for anterograde transport machinery in both the sensory and motor neurons of the gill-withdrawal reflex of Aplysia. We found that the induction of long-term facilitation (LTF) by repeated applications of serotonin, a modulatory transmitter released during learning in Aplysia, requires upregulation of kinesin heavy chain (KHC) in both pre- and postsynaptic neurons. Indeed, upregulation of KHC in the presynaptic neurons alone is sufficient for the induction of LTF. However, KHC is not required for the persistence of LTF. Thus, in addition to transcriptional activation in the nucleus and local protein synthesis at the synapse, our studies have identified a third component critical for long-term learning-related plasticity: the coordinated upregulation of kinesin-mediated transport.
In neurons that participate in learning and memory storage, individual synapses exhibit site-specific plasticity, which, in its long-term form, requires somatic transcription and translation as well as local protein synthesis at the activated and marked synapses (Martin et al., 1997; Kandel, 2001). Since the nucleus is an organelle shared by all the synapses of a neuron, the findings of synapse specificity raises the question: How can the products of transcription or translation reach a stimulated synapse without affecting unstimulated synapses? In principle there are at least three mechanisms whereby synapse-specific modification can occur: (1) selective targeting of mRNAs and proteins to the stimulated sites; (2) broad distribution of mRNAs and proteins to all synapses with productive utilization of these gene products by selective capture at the stimulated sites; and (3) local protein synthesis in response to stimulation at the active synapse. All three mechanisms require that mRNAs and proteins synthesized in the cell body be actively transported to the synapses.
Even though we now understand aspects of the nuclear events for activating long-term synaptic plasticity and of the local protein synthetic events involved in the maintenance of synapse specific facilitation (Kandel, 2001, Bailey et al., 2004), the molecular mechanisms that coordinate the nuclear and the synaptic events have not yet been delineated. In particular, we know very little about how gene products are transported from the cell body to synapses in response to learning-related activity, nor do we know what particular gene products are transported. We therefore sought to investigate these questions by first exploring whether anterograde molecular transport to synapses is regulated in response to repeated pulses of 5HT that induce long-term facilitation (LTF). We focused on the kinesin family of motor proteins because they are known to be involved in the routine anterograde transport of cargo from the cell body to the synapse (Vale et al., 1985; Goldstein and Yang, 2000) and therefore seemed good candidates for coordinating the learning-related dialogue between the cell body and synapses.
The kinesin superfamily (KIF) of molecular motors transports, in a microtubule and ATP-dependent manner, three types of cargos: organelles, mRNAs and proteins (Vale and Fletterick, 1997; Hirokawa, 1998). The kinesin transport machinery is composed of a heavy chain subunit (KHC) containing a conserved motor domain that attaches to the microtubule tracks and a light chain (KLC) subunit that is thought to confer regulatory specificity and specificity for binding of unique cargo (Goldstein and Philip, 1999; Goldstein and Yang, 2000).
To address the question of whether kinesins might be regulated in neurons in the context of memory storage, we have cloned from the Aplysia nervous system several isoforms of kinesins, and identified isoforms of both light and heavy chains that are highly enriched in neurons. We next found that these two isoforms are regulated in novel ways. Five spaced pulses of 5HT, a modulatory neurotransmitter released in the intact animal by sensitizing stimuli (Montarolo et al., 1986, Glanzman et al., 1989), required for long-term memory in Aplysia upregulate, as immediate response genes, expression of the neuronal kinesin isoforms ApKLC2 (Aplysia kinesin light chain 2) and ApKHC1 (Aplysia kinesin heavy chain 1). We found that this 5HT induced increase in kinesin levels is both necessary and sufficient for the induction of LTF. By contrast, kinesin upregulation is not critical for the persistence of LTF. We have also identified several cargo proteins associated with ApKHC. These include the proteins neurexin and neuroligin involved in de novo synapse formation during development, and piccolo and bassoon proteins required for the differentiation of the active zone.
We cloned three KLC isoforms and two KHC isoforms by degenerate PCR and by mining an Aplysia neuronal EST database (Supplementary results and Figure S1). We focused on ApKHC1 and ApKLC2 because they are highly expressed in neurons, and cloned their full-length cDNAs by screening Aplysia cDNA libraries.
To explore the possibility that these neuronal isoforms might have specific functions in experience-dependent plasticity and synaptic growth of neurons, we examined whether the levels of ApKLC2 and ApKHC1 are regulated by 5HT. We reasoned that for the long-term changes of synaptic strength, communication between the nucleus and the synapses might require an enhancement of axonal transport. If this were so, the requirement for transport might be accompanied either by an increase in the total number of motor molecules or by a modification of pre-existing motor molecules so that they could carry more cargo or carry it more efficiently. Alternatively, there might be an increase in the total number of cargos in response to 5HT or a modification of pre-existing cargos such that they now bind kinesin. The latter possibility has the advantage of providing specificity to the anterograde transport process. To test these possibilities, we first examined whether the mRNA levels of ApKLC2 and ApKHC1 change with application of 5HT.
We first examined the mRNA levels of ApKLC2 and ApKHC1 by semi-quantitative RTPCR. Each of the two pleural ganglia of the central nervous system (CNS) contains a large cluster of sensory neurons. We therefore isolated total RNA from pleural ganglia dissected from the CNS of Aplysia that were exposed to five pulses of 10 μM 5HT (0 minutes and 30 minutes after 5HT treatment). We found an upregulation of (~2 fold, p< 0.01, Student’s t test) ApKLC2 and ApKHC1 mRNA at thirty minutes after treatment with 5HT (Figure 1A and B). To further confirm the upregulation of kinesins by 5HT, we performed real-time RTPCR to quantify transcript level changes. Using Aplysia GAPDH levels to normalize the data, we found a ~4 fold increase in transcript levels [ApKHC1 (4.5 ± 0.8) and ApKLC2 (4.8 ± 0.7), p< 0.01, Student’s t test] in 30 minutes after application of 5HT (Figure 1C).
Is the increase in mRNAs induced by repeated pulses of 5HT specific to the presynaptic or the postsynaptic cell of the gill-withdrawal reflex neural circuit? To examine induction of mRNA at single cell level, we studied mRNA levels of ApKHC1 in sensory (SN) and motor (MN) neuron cocultures. RNA in situ hybridization experiments revealed that at 30 minutes following 5HT stimulation, there was an upregulation of ApKHC1 in both the sensory and motor neurons (fold increase in mean fluorescence intensity: SN 3.8 ± 0.4, MN 4.2 ± 0.5; p<0.01, Student’s t test; Figure 1D and E). These results suggest that long-term synaptic facilitation is associated with enhanced kinesin mediated transport in both pre- and postsynaptic cells.
We next asked: Are these increases in the mRNA levels of ApKLC2 and ApKHC1 induced by 5HT reflected in their protein levels? We isolated total proteins from the pleural ganglia treated with five pulses of 5HT (0 minutes and 30 minutes after treatment) and immunoblotted them using affinity-purified anti- ApKLC and ApKHC antibodies (Supplementary Figure S3). Following application of five pulses of 5HT, there was also an increase in the protein levels of ApKLC and ApKHC (fold increase: ApKHC 1.8 ± 0.2; ApKLC 1.6 ± 0.2; p< 0.01, Student’s t test; Figure 1F and G). We verified the increase in the protein levels induced by 5HT in sensory neurons using immunohistochemical analysis of KHC and KLC in sensory neurons (fold increase in mean fluorescence intensity: ApKHC 1.2 ± 0.02; ApKLC 1.6 ± 0.3; p< 0.05, Student’s t test; Figure 1H and I).
The rapid increase in their mRNA levels suggested to us that ApKLC2 and ApKHC1 might function as primary response or immediate early genes. The induction of immediate early genes is independent of new protein synthesis and only dependent on transcriptional upregulation by constitutively present transcriptional regulators such as CREB in Aplysia sensory neurons. To characterize the transcriptional induction of kinesins by 5HT, we focused on the expression of ApKHC1. We found that the induction of ApKHC1 is not affected by the treatment of pleural ganglia with protein synthesis inhibitors emetine or anisomycin but the induction was blocked by the exposure to transcriptional inhibitor Actinomycin D (Supplementary results, Figure S2).
The failure of protein synthesis inhibitors to block 5HT mediated upregulation of the kinesin mRNA suggests that a protein synthesis independent signal is responsible for this transcriptional upregulation. Since 5HT upregulates cAMP level (Bacskai et al., 1993), we asked whether an artificial increase in cAMP could induce ApKHC1 expression. We found that application of forskolin, a stimulator of cAMP levels, which in turn can activate CREB, resulted in increased levels of ApKHC1 mRNA in isolated sensory clusters, and this increase was blocked by the treatment with14-22 amide, a myristoylated catalytic site-specific inhibitor of PKA (Supplementary results, Figure S2). These results suggest that ApKHC1 is an immediate early gene and that its transcriptional induction depends on cAMP and PKA.
Kinesins are thought to be relatively abundant proteins. Hence, one might wonder why kinesins would be upregulated to meet the transport requirements of nerve cells during learning. One possibility is that kinesins have a short half-life and become limiting for anterograde transport of cargos. The other possibility is that more kinesin is required irrespective of its relative abundance in response to the remodeling and growth of new synapses that accompany LTF. To test these possibilities, we investigated the rate of synthesis of new ApKHC protein in unstimulated neurons by pulse-chase experiment using 35S-methionine labeling followed by coimmunoprecipitation of ApKHC and ApKLC. Contrary to the possibility that kinesins have short half-life, we found that both ApKHC and ApKLC showed two different half-lives, presumably relating to two different protein pools. One pool, which we could not measure precisely, is short-lived with half-lives of less than three hours (Table S2 and S3, and Supplementary results). The other pool, which we measured reliably, is long-lived (ApKLC: 16.5 hours, r2 =0.98 and ApKHC: 20 hours, r2 =0.98; Supplementary Figure S4, Table S1).
Does the increase in kinesin level induced by 5HT actually affect the anterograde trafficking of cargos? As a first step in understanding the effect of 5HT on cargo trafficking, we focused on vesicle trafficking in sensory neurons using video-enhanced contrast-differential interference contrast (VEC-DIC) microscopy (Schnapp et al., 1985). We recorded movement of vesicles for 10 minutes before and 30 minutes after the application of five pulses of 5HT. We observed in the sensory neurons, vesicles of varying sizes and mobility. For reliable quantitation of vesicle trafficking, we focused only on vesicles of ~300 nm size. These ~300 nm size vesicles move to the distal processes at a velocity of about 200-325 nm per second. In response to five pulses of 5HT, these vesicles showed an increase in their number, moving in the anterograde direction in sensory neurons (Figure 2A and B; % change: 5HT, 43 ± 4.5; untreated control, 10 ± 3.5; n=7, p<0.05, Student’s t test). Application for 60 minutes of 5 μM nocodozole (Grigoriev et al., 1999), an inhibitor of tubulin polymerization, blocked movement of these vesicles (data not shown) suggesting that transport of these vesicles depends on microtubules. Since only a fraction of the transport can be visualized with VEC-DIC microscopy, the ~40% increase in trafficking we observed may only represent a fraction of the net effect on vesicular transport.
Is kinesin-mediated transport necessary for long-term synaptic facilitation? To test this idea, we first examined the kinesin heavy chain. We used the sensory neuron-motor neuron coculture in which a single motor neuron is innervated by two sensory neurons (Montorolo et al., 1986). To specifically degrade ApKHC1 mRNAs, we used antisense oligonucleotides to deplete ApKHC1 (Supplementary Figure S5 and Supplementary results). We microinjected antisense oligonucleotides (50 μg/ml) into one of the sensory neurons in coculture using the other sensory neuron as an internal control. First, we examined whether ApKHC1 antisense microinjection has any effect on short-term facilitation (STF) of sensory to motor neuron synapses. Cells were treated with one pulse of 10 μM 5HT for five minutes 20 hours after oligo microinjection and excitatory postsynaptic potentials (EPSPs) were measured 10 minutes after 5HT application (Figure 3A; % change in EPSP amplitudes: 5HT alone 99.3 ± 15.2, n=8; basal −4.5 ± 4.9, n=9; 5HT + antisense oligo 100 ± 13.1, n= 7; antisense oligo alone −3.4 ± 3.2, n=8; sense oligo + 5HT 93 ± 10.7, n=8). The antisense ApKHC1 did not affect 5HT induced STF (n=7, p>0.05, Student’s t test).
We then explored ApKHC1 function in LTF. After 3 hours of oligo injection into sensory neurons, cultures treated with five pulses of 10 μM 5HT and EPSPs were measured 24 hours after 5HT application (Figure 3B; % change in EPSP amplitudes: 5HT alone 113.9 ± 27.9, n=8; basal −6.7 ± 7.4, n=9; 5HT + anti sense oligo −2.1± 6.8, n= 7; antisense oligo alone −5.9 ± 8.8, n=8; sense oligo + 5HT 99.8 ± 30.6, n=8). There was a significant reduction in mean EPSP amplitude in the sensory neuron that received the antisense ApKHC1 oligo and five pulses of 5HT exposure (p < 0.01, n= 7, Student’s t test). Microinjection of sense oligos (control for the antisense oligo injections) to the sensory neurons did not have any significant effect on the 5HTinduced increase in mean EPSP amplitudes measured at 24 hours. Taken together these results suggest that presynaptic ApKHC1 levels are necessary for the induction of LTF.
To determine that the kinesin light chain was also necessary for LTF, we used antisense oligos designed specifically to degrade ApKLC2 mRNAs (Supplementary Figure S5 and Supplementary results), and microinjected (50 μg/ml) into one of the sensory neurons in coculture using the other sensory neuron as an internal control. We first examined whether ApKLC2 antisense or sense (as control) oligo microinjection into sensory neuron has any effect on STF. The antisense or sense ApKLC2 oligo did not affect 5HT induced STF (n=7, p>0.05, Student’s t test, Figure 3C, % change in EPSP amplitudes: 5HT alone 95.3 ± 15.8, n=8; basal −9.0 ± 7.1, n=5; 5HT + antisense oligo 104.3 ± 12.8, n= 7; antisense oligo alone −10 ± 7.3, n=8; sense oligo + 5HT 93 ± 13.3, n=7).
We then studied ApKLC2 function in LTF by injecting antisense or sense oligos into sensory neurons (Figure 3D; % change in EPSP amplitudes: 5HT alone 49.0 ± 10.3, n=8; basal −6.1 ± 6.8, n=9; 5HT + antisense oligo −10.2 ± 13.3, n= 8; antisense oligo alone −7.0 ± 11.9, n=8; sense oligo + 5HT 57.7 ± 12.3, n=8). The significant reduction in mean EPSP amplitude in the sensory neuron that received antisense ApKLC2 oligo and five pulses of 5HT (p < 0.01, n= 8, Student’s t test) suggests that kinesin light chain levels in the presynaptic neuron are also necessary for the induction of LTF.
We then examined the significance of kinesin-mediated transport in postsynaptic neurons by depleting ApKHC1 in the motor neuron. We used Aplysia bifurcated neuronal cultures in which a single bifurcated sensory neuron is innervated by two spatially separated motor neurons (Martin et al., 1997). We injected antisense or sense ApKHC1 oligo into one motor neuron and used the other as an internal control.
At 24 hours after bath application of five pulses of 10 μM 5HT, we observed a significant reduction in the mean EPSP amplitudes (p < 0.05, n=5, Student’s t test) at the motor neuron–sensory neuron synapses that received antisense oligo injection (Figure 3E; % change in EPSP amplitudes: 5HT alone 81.6 ± 12.2, n=7; basal −5.8 ± 3.5, n=6; 5HT + anti sense oligo 10.4 ± 18.8, n= 5; antisense oligo alone −28.3 ± 9.1, n=5; sense oligo + 5HT 74.3 ± 12.5, n=6). These results suggest that postsynaptic kinesin levels are also critical for the induction of LTF.
Long-term memory storage has at least two phases, an induction phase and a persistence phase. We have recently found that these phases have distinct molecular requirements (Martin et al., 1997; Si et al., 2003). Local protein synthesis mediated by the cytoplasmic polyadenylation element binding protein (CPEB) is not required for the induction of LTF in Aplysia, but is critical for the persistence of LTF at 72 hours after 5HT stimulation. This would suggest that the initial steps in new synapse formation use proteins synthesized in the cell body and transported to the synapses. Consistent with this idea, we find that blocking of ApKHC1 either in the presynaptic or the postsynaptic neuron blocks induction of LTF. Once induced, does LTF at 72 hours require enhanced levels of kinesin? Dependence of the persistence phase of LTF on kinesin levels would suggest the requirement for a continuous, enhanced delivery of cargos from the cell body to the synapses, whereas independence on kinesin levels would suggest that once LTF is established, synapses become autonomous with respect to transport from cell body and hence elevated kinesin mediated transport is not necessary during persistence.
To discriminate between these possibilities we injected ApKHC1 antisense oligos into sensory neurons 24 hours after induction of LTF and measured EPSPs at 72 hours (Figure 3F). We found no significant difference (n=6, p>0.05, paired t test.) of mean EPSP amplitudes measured at 72 hours due to microinjection of ApKHC1 antisense oligos at 24 hours after 5HT application compared to controls that did not receive antisense oligonucleotides (% change in EPSP amplitudes: 5HT alone: @ 24 hours 91.9 ± 13.8, n=6; @ 72 hours 65.5 ± 17.2, n= 6; 5HT alone @ 24 hours before antisense oligo injection 101.8 ± 22.8, n=6; 5HT + ApKHC1 oligo injection at 24 hours: EPSP @ 72 hours 55.7 ± 18.2, n=6; untreated control : @ 24 hr 1.7 ± 9.5, n=6; @ 72 hours −4.4 ± 10.6, n=6). These results suggest that once LTF is induced at the synapses and the requisite gene products are delivered for immediate growth of new synaptic connections, elevated kinesin levels are not required for the maintenance phase of LTF.
If the kinesin level is a limiting factor for the induction of LTF, would an increase in kinesin level be sufficient to induce an increase in synaptic strength? To address this question, we examined the effect of an ectopic increase in the level of ApKHC1 on the strength of the synaptic connections between the sensory and motor neuron by measuring EPSPs. To overexpress ApKHC1 in sensory neurons, we cloned full-length ApKHC1 in frame with EGFP into a plasmid construct while overexpression of EGFP alone in cocultures was used as control. Because of the variability due to differences in efficiency of injection of plasmid constructs into cells and differences in time taken to first detect the fluorescence and varying expression levels, we arbitrarily took the time of injection as time zero for our analysis. Overexpression of ApKHC1-EGFP resulted in increase in EPSPs (Figure 4A & B) at 12 hours (n=12, p<0.0001, Student’s t test) after injecting the construct. This increase then decayed gradually over 72 hours (% change in EPSP amplitudes: ApKHC1-EGFP injections @12 hours 115.4 ± 23.2, n=12; @ 24 hours 67.9 ± 17.1 n=9; @48 hours 15.5 ± 13.5, n= 5; @ 72 hours 3.7 ± 17.4, n=4; EGFP injections: @12 hours 0.1 ± 3.6, n=7; @ 24 hours 4.0 ± 2.1, n=6; @48 hours −4.0 ± 3.4, n= 7; @ 72 hours −3.3 ± 7.5, n=7). Overexpression of EGFP alone did not enhance synaptic strength. Thus, an increase in kinesin levels alone is sufficient to cause an induction of long-term facilitation but it does not enhance maintenance as shown by a decay of EPSPs over the period of 72 hours. This supports the idea that kinesin mediated fast axonal transport is limiting and that this increase in kinesin level is sufficient for the induction of LTF, presumably by carrying cargos that are required for induction.
How does the enhanced transport result in these distinct forms of synaptic facilitation? To begin to address this question, we decided to analyze what molecules are transported by kinesin in these neurons and we examined the protein components of the kinesin cargo. Protein cargos are synthesized in the cell body and transported to the distal processes as multi-protein complexes or as vesicular components. Using biochemical and genetic methods, proteins such as tubulin (Terada et al., 2000), glutamate receptors (Setou et al., 2000), APP (Kamal et al., 2000), and JIP scaffolding proteins (Verhey, et al., 2001) have been found to be associated with kinesin. Recently several RNA binding proteins associated with kinesin have been identified using proteomics (Kanai et al., 2004).
To identify proteins in Aplysia neurons that are associated with the kinesin heavy chain, we used both proteomic and candidate approaches (Figure 5 and Table S4). With the proteomics approach, we tried to identify systematically the protein components of the kinesin complex by coimmunoprecipitation followed by mass spectrometric sequencing of protein bands. The lack of Aplysia genome sequence information limited our efforts to achieve a comprehensive list of protein cargos using the proteomics approach. However, mass spectrometry allowed us to reliably identify several proteins (Table S4 and Supplementary Figure S6).
The presence of the kinesin heavy chain and light chain in the immunoprecipitated complex indicated that we had successfully maintained the integrity of the kinesin transport complex isolated from Aplysia neurons (Figure 5B & C). Interestingly, we identified a Ca2+ activated K+ channel as part of the cargo complex, probably transported as a vesicular component. Recently a voltage gated K+ channel transported by Kif7 has been described (Chu et al., 2006). In the kinesin complex we also found the Ca2+ / calmodulin dependent protein kinase II, which has been described as important in various forms of synaptic plasticity (Frankland et al., 2001; Mayford et al., 1996), and the molecular motor, myosin heavy chain (MHC), that is involved in actin dependent dendritic transport of molecules. MHC has been previously identified to interact with kinesin (Ohashi et al., 2002). Finally we found that the cargo of the kinesin complex (Supplementary Table S4) also contains cytoskeletal components such as actin, tubulin and intermediate filament protein. Previously it had been shown that disruption of kinesin (Kif5a) caused abnormal neurofilament transport (Xia et al., 2003). Taken together these results suggest that ApKHC cargo contain molecules that are required for cytoskeletal remodeling and synaptic plasticity.
In the Candidate approach, we assumed that molecules that are important for synapse formation might be transported by kinesin. We therefore used specific antibodies to probe the ApKHC complex for the presence of such proteins. By western blotting we detected the presence of proteins required for de novo synapse formation during development, such as neurexin (Dean et al., 2003) and neuroligin (Ichtchenko et al., 1995; Scheiffele et al., 2000), and presynaptic active zone proteins piccolo (Fenster et al., 2000) and bassoon (tomDieck et al., 1998) associated with the kinesin complex (Figure 5C). However, proteins such as the transcriptional regulator Aplysia CREB2 were not found (Supplementary Figure S3b) in the complex. The presence of these synaptic proteins in the complex suggested that their sub-cellular distribution could be regulated by kinesin-mediated transport.
Is the upregulation of kinesin in response to 5HT accompanied by an increase in the synthesis of cargos for synapse formation? Indeed, the increase in the ApKHC level would not necessarily be accompanied by an increase in cargo. To explore this possibility, we examined whether the protein levels of neuroligin, neurexin, piccolo and bassoon are regulated independently by 5HT. We found, using western blotting, that the protein levels of neurexin, neuroligin, piccolo and bassoon are indeed upregulated in the pleural ganglia by five pulses of 5HT (Figure 5D, E and F; n=4, p<0.05, Student’s t test), suggesting not only that levels of kinesin motor but also levels of some of its cargos are upregulated in a coordinate fashion by 5HT.
We now focused on piccolo to investigate the physiological significance of cargo upregulation by 5HT. We cloned a fragment of the Aplysia homolog of piccolo based on the available sequence information in the Aplysia Neuronal EST database and designed antisense oligos to degrade piccolo mRNAs (Supplementary Figure S8 and Supplementary results). These oligos were microinjected into sensory neurons of the coculture as described above.
Piccolo antisense microinjection into sensory neurons did not affect STF (Figure 5G). EPSPs were measured following one pulse application of 5HT (% change in EPSP amplitudes: 5HT alone 85.17 ± 14.8, n=8; basal −2.5 ± 5.8, n=9; 5HT + anti sense oligo 101.4 ± 24.6, n= 7; antisense oligo alone 1.3 ± 6, n=8; sense oligo + 5HT 84.2 ± 14.5, n=8; sense oligo alone −4.2 ± 5.9, n=6) and did not show any significant difference (n=7, p>0.05, Student’s t test) between antisense or sense oligo injection in the presence of 5HT and 5HT application alone. Measurements of EPSPs at 24 hours show that microinjection of antisense oligo against piccolo into sensory neurons blocked the induction of LTF (p < 0.05, n = 6, Student’s t test) induced by application of five pulses of 10 μM 5HT (Figure 5H; % change in EPSP amplitudes: 5HT alone 73.4 ± 11.7, n=10; basal −9.8 ± 11.4, n=5; 5HT + anti sense oligo 26.1 ± 9.1, n= 14; antisense oligo alone −2.8 ± 13.4, n=14; sense oligo + 5HT 68.5 ± 18.3, n=6). We then injected piccolo antisense oligos into motor neurons to examine whether piccolo has any postsynaptic function. EPSP measurements after 24 hours of microinjection and 5HT application show that piccolo antisense oligo injection in motor neurons does not block induction of LTF (n=6, p>0.05, Student’s t test). These data suggest that piccolo levels in sensory neurons are critical for induction of LTF. Piccolo is thought to perform scaffolding functions at active zones along with bassoon, RIMs, CASK, CAST, Velis and Mints (Sudhof, 2004). Because long-term memory storage requires formation of new synapses, availability of the molecules involved in synaptic architecture such as the piccolo-bassoon complex at the presynaptic active zones is important for induction of LTF.
Wong et al. (2002) have suggested that the overexpression of kinesin in mouse brain leads to activation of a positive feedback loop which turns on CREB mediated transcription of some cargos. To test for this possibility in Aplysia, we examined whether overexpression of kinesin by itself could induce an increase in piccolo mRNA levels and whether long-term increase in EPSPs due to kinesin overexpression required CREB activity. We found that piccolo is a target of CREB and its mRNA levels in sensory neurons showed no increase at 12 hours but a modest increase of 1.4 ± 0.04 fold at 24 hours after ApKHC1 overexpression (Supplementary results, Figure S8).
We next asked whether CREB activity is required for the observed increase in EPSPs due to ApKHC1 overexpression. To specifically block CREB activity, we injected the CRE element to which phosphoCREB binds (Dash et al., 1990) into sensory neuron and also overexpressed ApKHC1-GFP. We used a mutant CRE (mCRE) element to which phosphoCREB cannot bind as a control. We first studied the effect of CRE injection on LTF at 12 hours and 24 hours. We found that CRE injection blocked the 5HT induced increase in EPSPs (Figure 6A) at both 12 hours (n=6; p<0.001, Student’s t test) and at 24 hours (n=8; p <0.001, Student’s t test). By contrast the control mCRE injection did not affect 5HT induced increase in EPSPs at 12 or 24 hours (% change in EPSP amplitudes: basal −5.2 ± 7.2, n=9; 5HT alone @ 12 hours 72 ± 11.6, n=7; 5HT + CRE @ 12 hours 9.4 ± 12, n= 7; 5HT + mCRE @ 12 hours 71.9 ± 11.2, n= 4; 5HT alone @ 24 hours + 66.3 ± 13.2, n=8; 5HT + CRE @ 24 hours 16 ± 13, n= 5; 5HT + mCRE @ 24 hours 68.2 ± 11.9, n= 4; CRE alone −5.4 ± 10.3, n=5; mCRE alone −2.8 ± 7.7, n=5). These results are consistent with the findings (Dash et al., 1990) that CREB activity is required for the LTF induced by 5HT.
We then blocked CREB activity in sensory neurons that overexpress ApKHC1 to test whether CREB is required for the long-term increase in EPSPs induced by overexpression of ApKHC1. CRE or mCRE was injected two hours after injection of ApKHC1-GFP construct and EPSPs were measured at 12 hours and 24 hours. Our results show that CRE injection into sensory neurons did not block the overexpression of ApKHC1 induced increase in EPSPs at 12 hours (n=5; p<0.0001, Tukey-Kramer multiple comparison test) (Figure 6B). By contrast CRE injection blocked the increase at 24 hours (n=5; p<0.001, Tukey-Kramer multiple comparison test; % change in EPSP amplitudes: basal @ 12 hours 5.7 ± 2.4, n=11; basal @ 24 hours 1.5 ± 2.7, n=11; KHC-GFP alone @ 12 hours 34.8 ± 2.8, n=8; KHC-GFP alone @ 24 hours 18.1 ± 1.3, n=8; KHC-GFP + CRE @ 12 hours 30 ± 2.1, n= 5; KHC-GFP + CRE @ 24 hours −4.7 ± 2.1, n= 8; KHC-GFP + mCRE @ 12 hours 31.2 ± 3.2, n= 6; KHC-GFP + mCRE @ 24 hours 22.8 ± 1.8, n= 5). These results suggest that the overexpression of ApKHC1 is sufficient to induce an increase in EPSPs at 12 hours without requiring CREB by increasing axonal transport of pre-existing cargos in the cell body. However, the EPSP increase at 24 hours is dependent on CREB at least at the higher levels of ApKHC overexpression achieved in our experiments and supports the idea proposed by Wong et al. that overexpression of KHC could lead also to the activation of CREB through a positive feed back loop. This is consistent with the results with piccolo. Our data further show that this positive feedback loop is a late event in the establishment of LTF induced by overexpression.
To further investigate the necessity and sufficiency of ApKHC1, we examined the effects of down regulation of piccolo mRNA levels on the increase in EPSPs induced by ApKHC overexpression. We injected sense or antisense piccolo oligonucleotides into sensory neurons two hours after injection of ApKHC1-GFP construct and measured EPSPs at 12 hours and 24 hours. We found that piccolo antisense or sense injection into sensory neurons does not block ApKHC1 overexpression induced increase in EPSPs at 12 hours (Figure 6C; n=8; p<0.0001, Tukey-Kramer multiple comparison test) but antisense piccolo oligonucleotides blocked the increase at 24 hours (n=7; p<0.001, Tukey-Kramer multiple comparison test; % change in EPSP amplitudes: basal @ 12 hours 7.6 ± 1.3, n=5; basal @ 24 hours 6.7 ± 1, n=5; KHC-GFP + sense Piccolo @ 12 hours 60.2 ± 29, n= 6; KHC-GFP + antisense Piccolo @ 12 hours 68.2 ± 5.7, n= 8; KHC-GFP + sense Piccolo @ 24 hours 24.3 ± 3.6, n= 6; KHC-GFP + antisense Piccolo @ 24 hours 6.3 ± 3, n= 7). These results are in agreement with the data that CREB mediated activation of transcription is not required for the increase in EPSP at 12 hours, but CREB is required for the EPSP increase at 24 hours. Consistent with these results, we found that piccolo knockdown block the EPSP increase at 24 hours in ApKHC1 overexpressing neurons or in neurons that are exposed to five pulses 5HT.
We then examined whether ApKHC1 is required for the EPSP increase at 12 hours by 5HT. We injected ApKHC1 antisense oligos into sensory neurons and exposed the cultures to five pulses of 5HT. The consequent down regulation of ApKHC1 by antisense ApKHC1 blocked the 5HT induced increase (p<0.001, Tukey-Kramer multiple comparison test; Figure 6D) in EPSPs at 12 hours suggesting that ApKHC1 function is critical for the EPSP increase at 12 hours (% change in EPSP amplitudes: 5HT alone 82.73 ± 11.6, n=8; basal −2.89 ± 3.61, n=7; 5HT + anti sense oligo 24.7 ± 4.9, n= 7; sense oligo + 5HT 73.9 ± 4.4, n=7).
In summary, these results suggest that the electrophysiological phenotype that we described for overexpression of ApKHC1 has at least two phases, a 12-hour and 24-hour phase. ApKHC1 is both necessary and sufficient to induce an increase in EPSPs at 12 hours. By contrast CREB activity is also required at 24 hours.
Despite the expanding knowledge of the nuclear and synaptic mechanisms that underlie learning-related synaptic growth (Kandel, 2001; Bailey et al., 2004, Flavell and Greenberg., 2008), we know little about how nuclear and synaptic processes are coordinated during learning and memory storage. We here demonstrate a new component of this coordination.
We have found that the kinesins that mediate fast axonal transport are upregulated rapidly by five pulses of 5HT, a physiological signal for long-term facilitation in the intact animal. This upregulation was unexpected for two reasons: first, this is the first evidence that a molecular motor can be regulated by a physiological signal. Second, the finding that upregulation of ApKHC1 is sufficient to initiate LTF implies that a major function of CREB- mediated transcription in long-term synaptic plasticity is not to provide new species of molecules but to increase the levels of pre-existing molecules so as to meet the demand of new synapse formation.
LTF involves the growth of new synaptic connections and is evident as new synaptic growth in both the presynaptic sensory neurons and postsynaptic motor cell. In this coordinated process, the initial signaling originates presynaptically in the sensory neuron. Thus, the local application of five pulses of 5HT to the cell body of the sensory neuron or overexpression of phosphoCREB in the sensory neuron is both necessary and sufficient to initiate the long-term process (Cassadio et al., 1999). But phosphoCREB by itself is not sufficient to lead to the growth of new synaptic connections. By contrast, five pulses of 5HT to both the sensory neuron cell body and sensory-motor neuron synapses not only initiate the long-term process but also leads to growth of new synaptic connections (Martin et al., 1997).
The finding that application of 5HT to the sensory neuron cell body can recruit transcription and initiate the long-term process raises the question: How can knockdown of kinesin in the postsynaptic neuron block this early expression of the long-term process in presynaptic neurons? This question raises the interesting possibility that there may be one or more permissive steps in the postsynaptic cell that need to be activated for the long-term process to be maintained both pre- and postsynaptically. Kinesin in the motor neuron might mediate one of these critical postsynaptic steps. In support of this, we find that neuroligin, a postsynaptic protein involved in synaptogenesis, is present in the kinesin cargo complex and its blockade also blocks the long-term process (Choi et al., abstract (131.19/D16) presented at Society of Neuroscience annual meeting 2007).
Previous studies have found that ectopic overexpression of a kinesin, KIF17, in mice improves spatial and working memory (Wong et al., 2002). These mice also showed increased levels of glutamate receptors. Conversely, in a model of Alzheimer’s disease, impairing axonal transport by reducing the dosage of a kinesin molecular motor protein enhanced the frequency of axonal defects and increased amyloid-beta peptide levels and amyloid deposition in mice (Stokin et al., 2005). Consistent with these studies, we now find that during learning in Aplysia, overexpression of the kinesin heavy chain as well as of the kinesin light chain occurs physiologically, in a regulated manner. This finding raises the question: Why does learning-related synaptic plasticity require an increase in kinesin levels in Aplysia neurons? One possibility is that kinesin levels are limiting in neurons for the induction of LTF. In support of this idea we found that ectopic overexpression of kinesin heavy chain alone in the sensory neuron resulted in an increase in synaptic strength similar to the effect of 5HT application. Conversely, inhibition of ApKHC1 either in the presynaptic sensory neuron or the postsynaptic motor neuron or inhibition of ApKLC2 in sensory neurons inhibits the establishment of LTF by 5HT.
Upregulation of the transport machinery-KHC and KLC ensure enhanced coordinated transport of molecules that are dependent on both KHC and KLC. Alternately the rate of anterograde transport can be regulated in a stimulus-dependent manner by posttranslational modification of microtubule tracks (Reed et al., 2006) or by modifications of microtubule binding proteins such as tau (Dixit et al., 2008). Binding to KLC is a prerequisite for the transport of certain cargos such as the Golgi complex (Gyoeva et al., 2000) and amyloid precursor protein (Kamal et al., 2000). Loss of KLC function leads to severe defects in neurons and eventually to cell death (Gindhart et al., 1998; Rahman et al., 1999). Taken together this data support the idea that kinesin-mediated transport of cargo to synaptic sites is a rate-limiting step in memory storage.
It is interesting that overexpression of ApKHC1 alone was sufficient to induce an increase in synaptic strength at 24 hours, whereas inhibition of one of its cargo-piccolo inhibited 5HT induced increase in EPSPs at 24 hours. Consistent with the findings of Wong et al. (2002) that overexpression of KIF17 led, in addition to increase in transport to increased levels of phosphoCREB that resulted in activation of several CREB targets such as NR2b, we found that piccolo, a protein cargo of kinesin, is a CREB target and is induced by ApKHC1 overexpression. However, down regulation of CREB activity or levels of piccolo mRNA did not block the increase in EPSPs at 12 hours due to ApKHC1 overexpression, suggesting that overexpression of kinesin and recruitment of pre-existing cargos are sufficient to produce an immediate increase in EPSPs during the first 12 hours. If KHC overexpression can activate CREB targets, why does ApKHC1 induce LTF overexpression not maintained at 72 hours? Overexpression of KHC mimics and recapitulates the 5HT-induced LTF when 5HT is applied only to the cell body. As found by Martin et al. (1997), 5HT application to only the cell body of sensory neuron leads to LTF that is not maintained. In addition, the maintenance of LTF requires a synaptically generated signal that allows marked synapses to utilize productively the cargos sent from the cell body. In the absence of this local signal, LTF induced by overexpression of KHC, as is in the case of CREB, cannot be maintained.
Since formation of new synapses is rapid and requires continuous interaction of both the presynaptic and the postsynaptic neuron, one would expect synaptogenesis to be highly coordinated. Two distinct mechanisms, alone or in combination, could contribute to the rapid coordinated formation of new synapses in response to learning: 1) proteins required for new synapse formation might already be present at the potential pre- and postsynaptic sites so that learning only triggers their assembly into the molecular complexes of the synaptic architecture, and 2) proteins important for new synapse formation might be transported in a coordinated fashion from the cell body to the synapses. Our findings that inhibition of ApKHC1 in either the pre- or postsynaptic neuron blocks induction of LTF rule out the first possibility acting alone and support the second possibility. Consistent with this view, overexpression of ApKHC1 alone was sufficient to induce an increase in EPSPs at 12 hours and inhibition of piccolo, a protein cargo of kinesin in presynaptic neurons, blocked the induction of LTF. These findings further suggest that protein cargos transported by kinesin are immediately needed for synapse formation.
In addition to the protein cargo required for the immediate initiation of LTF such as neurexin, neuroligin, piccolo, and bassoon, tubulin and actin, the kinesin cargo also contain mRNAs (Kanai et al., 2004). These RNAs seem to be needed only later at a time when local protein synthesis becomes critical for the maintenance (Martin et al., 1997 and Si et al. 2003).
Taken together these results suggest a model for a cellular program for learning that involves upregulation of synthesis and active delivery to the synaptic sites in both a coordinated and temporally regulated fashion of proteins and mRNAs. According to this view, the proteins carried by kinesin are utilized immediately for the induction of new synaptic connections required for memory storage and mRNAs are used subsequently for maintaining growth of these connections. This model illustrates the strategy whereby a neuron couples the three critical components, transcriptional activation in the nucleus, active molecular transport, and subsequent activation of local translation. This strategy, we define as “programmed learning response,” involves temporal modulation of several critical steps in a concerted manner to coordinate molecular events that lead to the formation of new synapses and enhanced synaptic transmission that are decisive for long-term memory storage. Figure 7 illustrates this model for a programmed learning response in the sensory neuron-motor neuron connections of gill-withdrawal neural circuit in Aplysia. Upregulation of kinesin heavy chain and light chain in sensory neurons and motor neurons is one of the critical components of this program: it facilitates the communication between genes and synapses. Unraveling the intricate details of this program will be required for the molecular understanding of learning induced synaptogenesis and long-term memory storage.
Details about the cloning of Aplysia kinesins, DNA constructs, oligonucleotides (PCR primers, antisense and sense oligoncleotides) antisera production and commercial antibodies used in this study are described in the supplementary information.
The central nervous systems were isolated from 80-100 gram Aplysia californica and kept in L15 media supplemented with salts and glutamine. The nervous systems were kept at 18 degree C for 14-16 hours before 5HT stimulation or any pharmacological treatments. 5HT (10 μM) treatment was carried out as described previously (Si et al., 2003). Pleural ganglia or sensory clusters were isolated for preparation of total RNA or proteins. The dissection was carried out in ice-cold seawater to minimize injury associated molecular changes and quickly frozen in dry ice or immediately lysed for RNA or protein isolation. When used, pharmacological agents (anisomycin, emetine, actinomycin D) were applied for 30-60 minutes before 5HT treatment.
Details of RNA and protein isolation, semi-quantitative RTPCR, Real-Time PCR, western blotting and quantitation are described in the supplementary information.
The in situ hybridization using digoxigeniin (DIG) labeled riboprobes was followed as described in Giustetto et al. (2003). Details of the generation of probes are available in the supplementary information. After hybridization, the sense and antisense RNA probes were visualized using a Fluorescent Antibody Enhancer kit (Roche, Basel, Switzerland) for DIG detection. The immunohistochemical analyses of sensory neurons were performed as previously described (Si et al., 2003). Affinity purified rabbit anti ApKHC or ApKLC antibodies were diluted 1:75 corresponding to 10 μg/ml. Alexa 488 or Texas Red conjugated anti-rabbit secondary antibodies (Invitrogen, CA) were used for visualization. Images were acquired using a Fluoview 1000 confocal microscope (Olympus, Germany) with 20X objective. In all the figures (except in Figure 5) only projection images are shown. For quantitation of the immunohistochemical data, we only focused on the cell body and major axons. The quantitative analysis of immunocytochemical data were carried out on confocal image stacks using METEMORPH.
Aplysia CNS were isolated and kept overnight at 17 °C in modified L15 media supplemented with glutamine. Pleural ganglia were isolated in ice-cold 1:1 artificial sea water-isotonic MgCl2 and lysed at 4 °C in a buffer containing 50 mM Tris pH 7.6, 1 mM EDTA, 150 mM NaCl, 0.5% NP40, protease inhibitors (Roche, Basel, Switzerland) and phosphatase inhibitors (Calbiochem, Darmstadt, Germany). All subsequent steps were carried out in a cold room. The lysate was spun down at 5400 x g for 15 minutes and affinity-purified ApKHC antibody was added to the supernatant and placed on a rotator for 12-14 hours. Protein A/G beads (Pierce, Rockford, IL) pre-equilibrated in lysis buffer were added to the supernatant and incubated for two hours. The bead-antibody complexes were centrifuged briefly (100 x g), removed the supernatant and washed three to four times in the lysis buffer. Complexes were analyzed on a 5-20% gradient SDS-PAGE. Details of processing of protein bands and mass spectrometric analysis are described in supplementary information.
The sensory neuron axons of the sensory-motor neuron culture were observed with video-enhanced differential interference contrast microscopy on an Axiovert 100TV (Carl Zeiss MicroImaging, Inc. Germany) microscope (Schnapp, 1985; Choquet et al., 1997). Video data were digitalized using FinalcutPro and vesicle sizes were measured using ImageJ (http://rsb.info.nih.gov/ij/).
Cell cultures and electrophysiology were performed essentially as described by Montorolo et al. (1986) and bifurcated cultures as described by Martin et al. (1997). Preparation of phosphoCREB and microinjection into nucleus were performed according to Cassadio et al. (1999). After basal EPSP measurements, the oligos (50 μg/ml) were pressure injected into the cells. For the ApKHC1 overexpression experiments, plasmids purified by CsCl gradient ultracentrifugation were microinjected (1 μg/ μl) into sensory neurons and EPSPs were measured after 12, 24, 48 and 72 hours of injection.
We thank Thomas Jessell, Steven Siegelbaum, Kausik Si, and Kandel lab members Joseph Rayman and Ilias Pavlopoulos for their critical comments on an earlier version of this manuscript. We are grateful to Peter Schieffele for the gift of anti-neurexin antibody. We thank Joun-Hun Kim for help with Piccolo antisense microinjection experiments, Aviva Olsavsky for help with the RTPCR experiments, and John Edwards of Columbia’s Genome Center for help with kinesin ESTs. Special thanks to Vivian Zhu for help with cell cultures, Hannah Cho for technical help, and Charles Lam for help with graphics. This work is supported by HHMI, and NIH grants P50 HG002806 and R01 MH075026.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.