BDNF Facilitates L-LTP Maintenance in the Absence of Protein Synthesis through PKMζ

Late-phase long term potentiation (L-LTP) is thought to be the cellular basis for long-term memory (LTM). While LTM as well as L-LTP is known to depend on transcription and translation, it is unclear why brain-derived neurotrophic factor (BDNF) could sustain L-LTP when protein synthesis is inhibited. The persistently active protein kinase ζ (PKMζ) is the only molecule implicated in perpetuating L-LTP maintenance. Here, in mouse acute brain slices, we show that inhibition of PKMζ reversed BDNF-dependent form of L-LTP. While BDNF did not alter the steady-state level of PKMζ, BDNF together with the L-LTP inducing theta-burst stimulation (TBS) increased PKMζ level even without protein synthesis. Finally, in the absence of de novo protein synthesis, BDNF maintained TBS-induced PKMζ at a sufficient level. These results suggest that BDNF sustains L-LTP through PKMζ in a protein synthesis-independent manner, revealing an unexpected link between BDNF and PKMζ.


Introduction
LTP in acute hippocampus slices has long been used as a model to study the cellular mechanisms underlying learning and memory. There are temporally distinct types of LTP: protein synthesisindependent early phase LTP (E-LTP) and protein synthesisdependent late phase LTP (L-LTP) [1,2,3], paralleling the two forms of memory -short-term and long-term memories [4]. While numerous studies have been done on E-LTP, much less is known about the mechanisms for L-LTP. The secreted trophic protein BDNF and intracellular signaling molecule PKMf are the two best-studied molecules; both are necessary and sufficient to maintain L-LTP [5,6,7,8,9]. BDNF through its presynaptic or postsynaptic TrkB receptor activates the downstream mitogenactivated protein kinase (MAPK), phosphatidylinositol 3-kinase (PI3K) and phospholipase C-c (PLC-c) pathways [10]. PKMf is a brain-specific, atypical isoform of protein kinase C. It is persistently active, due largely to the lack of regulatory domain and therefore second-messenger-independent [11]. BDNF and PKMf share several common characteristics in regulating hippocampal L-LTP. First, either perfusion of BDNF or intracellular introduction of PKMf directly facilitates synaptic transmission by promoting postsynaptic responses [9,12,13,14]. Second, inhibition of either BDNF or PKMf abolishes L-LTP [5,15]. Third, BDNF and PKMf could modulate the morphological changes of dendritic spines [12,16]. However, the relationships between the two molecules in regulating L-LTP remain unclear.
Substantial evidence suggests that the expression of BDNF gene is controlled by neuronal activity [17]. In the hippocampus, the BDNF mRNA levels in the CA1 region are rapidly increased in response to the L-LTP inducing tetanic stimulation [18,19]. Weak tetanic stimulation, which normally induces only E-LTP, could induce L-LTP as long as the BDNF levels are elevated [5]. With these results, one can hypothesize that strong tetani trigger the expression of BDNF which in turn enhances the synthesis of PKMf in the hippocampus, leading to L-LTP. However, application of BDNF could rescue L-LTP deficits even when protein synthesis is completely blocked [5]. These perplexing results raise the possibility that BDNF may increase the PKMf level not by enhancing its synthesis but by reducing degradation to achieve LTP maintenance.
The present study attempts to reveal a mechanistic link between BDNF and PKMf. We found that BDNF-related neuronal activities augmented PKMf expression but BDNF alone did not modulate steady-state PKMf protein level. Moreover, in the absence of protein synthesis, BDNF sustained L-LTP by maintaining activityinduced PKMf at a sufficient level. These results together suggest that BDNF-dependent L-LTP is mediated by PKMf, and explain how BDNF can maintain L-LTP even when protein synthesis is completely blocked.

PKMf mediates BDNF-dependent late phase LTP
Previous studies indicate that L-LTP can be further divided into the BDNF-dependent form which can be induced by theta burst stimulation (12 TBS) or a perfusion of cAMP analogs such as forskolin, and the BDNF-independent form which is triggered by the classic 4 sets of tetanus (46tetani) [20]. PKMf is known to mediate L-LTP induced by 4 tetani [21]. To determine whether PKMf is also responsible for BDNF-dependent form of L-LTP, we applied ZIP, a myristoylated PKMf-substrate peptide inhibitor (5 mM), well after the BDNF-dependent L-LTP was expressed. In 12 TBS-induced L-LTP, ZIP applied 1 hour after stimulation successfully reversed L-LTP (Fig. 1A, 1B, 9968% at 175-180 min, p,0.01 compared with control). In contrast, a scrambled ZIP peptide (5 mM) did not affect L-LTP maintenance (166613% at 175-180 min). Forskolin-induced L-LTP was induced by a 15-min perfusion of a combined forskolin (50 mM) and the phosphodiesterase inhibitor IBMX (30 mM). ZIP was applied 80 min after LTP induction when a stable L-LTP was fully established. Again, ZIP abolished L-LTP (Fig. 1C, 1D, 10366% for ZIP, 184618% for Scrambled ZIP, at 175-180 min, p,0.001). These results suggest that PKMf is required for the maintenance of BDNFdependent L-LTP.

Steady-State PKMf level is not maintained by BDNF
To investigate whether and how BDNF regulates PKMf expression, we tested steady-state PKMf expression in wild-type (WT) and homozygous BDNF knockout (KO) mice. At postnatal day 18 (P18), brain tissues were dissected and subjected to Western blot analysis. Surprisingly, there was no significant difference of endogenous PKMf expression between WT and KO mice in the cortex and hippocampus, respectively ( Fig. 2A, Cortex, p = 0.79; Hippocampus, p = 0.52).
A number of studies have demonstrated that BDNF promotes gene transcription and translation [22]. We next investigated whether exogenous BDNF treatment affected PKMf protein level in primary neuron culture. Embryonic neurons derived from WT, heterozygous (Het) and KO mice were cultured for 7 days (DIV 7) and then exposed to BDNF (100 ng/ml) or vehicle for 24 hours, and total PKMf protein level was measured by Western blot. Addition of BDNF did not cause any significant change in the levels of PKMf in WT or BDNF mutant genotypes (Fig. 2B, WT, p = 0.91; Het, p = 0.44; KO, p = 0.34). Thus, it appears that without substantial enhancement of neuronal or synaptic activities, BDNF does not alter the steady-state level of PKMf protein.

BDNF modulates activity-dependent PKMf levels to sustain L-LTP in the absence of protein synthesis
We have previously shown that treatment with BDNF is sufficient to rescue L-LTP impairment when protein synthesis is completely blocked [5]. We attempted to examine whether BDNF could modulate PKMf to sustain the L-LTP at this situation. Consistent with the previous study, L-LTP was fully established by  BDNF (200 ng/ml) despite of protein synthesis inhibition by anisomycin (40 mM). We next applied PKMf inhibitor ZIP at 1 hour after tetanus and found L-LTP was completely reversed (Fig. 3A-3B, 9466% for ZIP, 160615% for Scrambled ZIP, at 175-180 min, p,0.001). These results raise the possibility that BDNF regulates PKMf to ensure a sustained L-LTP through a protein synthesis independent mechanism.
To further characterize how BDNF regulates PKMf, we compared PKMf level in the hippocampal slices at different time points after 12 TBS. The 12 TBS group was normalized against control condition in which slices were not stimulated. The 12 TBS plus BDNF and anisomycin treatment groups were normalized against that without BDNF treatment. The PKMf signals on the Western blot were normalized to that of b-tubulin on the same lane. At the early stage of L-LTP (around 1 hour after tetanus), synaptic activation induced a small but statistically significant increase of PKMf level (Fig. 3C, 11665.8%, p,0.05). This elevation of PKMf level was protein synthesis dependent. Application of BDNF (200 ng/ml) together with 12TBS did not further increase PKMf level. At the late stage of L-LTP (around 3 hour after tetanus), the BDNF-treated slices exhibited a much higher level of PKMf compared with the one in the presence of anisomycin (Fig. 3C, 161.6620.5%, p,0.05). Thus, BDNF combined with strong tetanus could increase the steady-state level of PKMf when protein synthesis is completely blocked. These results suggest that in the absence of protein synthesis strong tetanic stimulation together with BDNF could somehow elevate PKMf protein level, which in turn is responsible for L-LTP maintenance.

Discussion
Although L-LTP is known to be dependent on translation, it has been puzzling why BDNF could rescue the L-LTP deficit in the presence of the protein synthesis inhibitor anisomycin [5]. Considering that anisomycin may elicit unspecific stress-related pathways [23], we previously also applied emetine (20 mM) to block protein synthesis. A similar rescuing effect of BDNF was observed in L-LTP impairment, validating the notion that BDNF promotes L-LTP maintenance in the absence of protein synthesis [5]. In the present study, we have revealed an unexpected role of PKMf in mediating this BDNF-dependent form of L-LTP. In the absence of protein synthesis, BDNF seems to sustain L-LTP by means of maintaining a sufficient level of activity-induced PKMf. These data provide a mechanistic link between BDNF and PKMf and suggest their critical role in the maintenance of L-LTP despite of protein synthesis inhibition.
Unlike the classic, tetanus-induced L-LTP, the cAMP or 12TBS induced L-LTP requires an increase in local concentration of dendritic proteins but not nucleus activity [24], and is dependent on BDNF [20]. Similar to the classic L-LTP, however, we now demonstrate that the BDNF-dependent L-LTP also requires PKMf. Interestingly, in primed L-LTP through type I mGluRs activation, neither suppression of BDNF nor PKMf alone could reverse L-LTP. But a co-inhibition of BDNF and PKMf completely abolishes its maintenance [25]. These results could be interpreted as PKMf acts either in parallel or synergistically with BDNF. However, we provide several lines of evidence suggesting that PKMf could be downstream of BDNF, at least in BDNFdependent L-LTP. First, application of the PKMf inhibitor ZIP after cAMP or 12 TBS reverses the BDNF-dependent LTP. Second, BDNF together with 12TBS increases hippocampal PKMf level. Finally, in the presence of anisomycin, BDNF rescue of L-LTP deficit could be reversed by ZIP.
In general, BDNF-TrkB signaling is crucial for activity-induced new protein synthesis [22]. Moreover, synthesis of PKMf is a common target of many signaling pathways in LTP induction, including the major BDNF downstream pathways, such as PI3kinase, MAPK, mTOR, etc [26,27]. However, we did not detect a difference of steady-state PKMf expression between WT and BDNF KO mice. Further, application of BDNF to cultured WT neurons did not increase PKMf protein level. We reasoned that BDNF may need to work together with high frequency neuronal activity to up-regulate PKMf through a protein synthesis independent mechanism. Indeed, we found that BDNF together with the L-LTP inducing 12TBS increases the PKMf protein level, even in the presence of anisomycin.
How BDNF maintains PKMf level without protein synthesis? One attractive hypothesis is that BDNF inhibits TBS-induced degradation of PKMf through the ubiquitin-proteasome system (UPS). A balance in protein synthesis and degradation has been implicated in the maintenance of long term plasticity, structurally and functionally [28]. When protein synthesis is inhibited, PKMf level decreases primarily through UPS-mediated degradation. Given that BDNF-TrkB signaling acts upstream of UPS coupling neuronal activity with protein turnover [29], it is possible that BDNF counters PKMf degradation to maintain L-LTP. Indeed, without BDNF treatment, PKMf level keeps low under anisomycin perfusion [30]. Moreover, there is a critical window for BDNF to rescue L-LTP impairment -no later than 10 minutes after tetanus [5]. An alternative hypothesis is that BDNF regulates PKMf protein translocation to the stimulated synaptic site. According to ''synaptic tagging'' theory, PKMf is suggested as a plasticity-related protein (PRP) that not only potentiate synaptic responses at strongly tetanized pathways, but also at weakly stimulated pathways as long as synaptic tags are set [31]. BDNF may facilitate PKMf translocation from cytoplasm to synaptic sites. Specifically, when protein synthesis is inhibited, a local shortage of newly synthesized PKMf may drive the need for PKMf translocation and BDNF may facilitate this process. Regardless, it is critical for BDNF to hold PKMf at a sufficient level within this window before it is completely consumed by dynamic neuronal activities.
Taken together, these results expand the range of BDNF modulation of long term plasticity beyond a protein synthesis dependent manner and provide a strong mechanistic link between BDNF and PKMf in the maintenance of L-LTP.

Ethics Statement
All experiments were approved by the National Institutes of Health (NIH) Animal Care and Use Committee and followed the NIH Guidelines ''Using Animals in Intramural Research''. The NICHD Animal Use Proposal Number is 07-020.

Primary Neuron Culture
Primary cortical neurons were cultured from embryos produced by crossing BDNF heterozygous animals. Each fetus (E18) was dissected carefully to prevent blood contamination. A tissue chunk was pinched off for genotyping. Cortices from fetuses of the same genotype were digested with trypsin, dissociated and plated together. At DIV 7, vehicle or BDNF (100 ng/ml) was applied to cultures for 24 hours.
Field excitatory postsynaptic potentials (fEPSP) were evoked in CA1 stratum radiatum by stimulating Schaffer Collateral with a bipolar tungsten electrode and recorded with ACSF-filled glass pipettes using an Axoclamp-2B amplifier (Axon Instruments, Sunnyvale, CA). Recordings with maximal fEPSP less than 1 mV or with substantial changes in the fiber volley were rejected. Baseline responses were set to ,40% of maximal response and were recorded for 15 minutes. Late phase long-term potentiation was induced by tetanic stimulation, which contains 12 bursts, each with 4 pulses at 100 Hz and an inter-burst interval of 200 msec.
The initial slope of the fEPSP was measured as an index of synaptic strength. Data was analyzed by Clampfit 9 (Molecular Devices, Sunnyvale, CA) and presented as mean 6 s.e.m..