Abstract
In many excitatory synapses, synaptic zinc is co-released with glutamate to modulate neurotransmission. Synaptic zinc modulates the responsiveness of auditory cortex to sound, and synaptic zinc levels and signaling are modulated by sensory experience, termed zinc plasticity. The mechanisms underlying zinc plasticity remain unknown. We discovered that high- and low-frequency electrical stimulation of dorsal cochlear nucleus synapses reduces and increases synaptic zinc signaling, respectively. This bidirectional zinc plasticity is evidenced by changes in zinc inhibition of AMPA and NMDA receptor activity. Increases and decreases in zinc signaling require activation of Group 1 metabotropic glutamate receptors (mGluRs). Activation of Group 1 mGluRs with a higher agonist concentration increases presynaptic zinc levels and postsynaptic zinc signaling, whereas activation with a lower concentration reduces zinc levels and signaling. Sound-induced zinc plasticity also requires Group 1 mGluRs. Our results reveal the mechanisms underlying synaptic zinc plasticity, elicited by either synaptic activity or sound experience.
Introduction
In many brain areas, including the neocortex, limbic structures, and the auditory brainstem, glutamatergic vesicles are loaded with zinc (Danscher & Stoltenberg, 2005; Frederickson et al., 2005). This pool of mobile, synaptic zinc is co-released with glutamate and is a major modulator of glutamatergic neurotransmission. Namely, synaptic zinc inhibits synaptic and extrasynaptic NMDA receptor (NMDAR) excitatory postsynaptic currents (EPSCs), and modulates AMPA receptor (AMPAR) EPSCs (Kalappa et al., 2015; 2017; Anderson et al., 2015; Vogt et al., 2000; Vergnano et al., 2014). Synaptic zinc inhibits AMPAR EPSCs at lower frequencies of synaptic activity, but enhances steady state AMPAR EPSCs during higher frequencies of activity by recruiting endocannabinoid signaling, reducing probability of release and enhancing synaptic facilitation (Kalappa et al., 2015; 2017; Perez-Rosello et al., 2013). In awake mice, synaptic zinc enhances the responsiveness (gain) of sound-evoked responses in auditory cortical principal neurons but reduces the gain of cortical interneurons (Anderson et al., 2017). Activity-dependent changes in gain are crucial for sensory processing, for they allow sensory systems to adapt to changing sensory environments and facilitate central recovery of sensory processing in response to peripheral damage (Atallah et al., 2012; Ohshiro et al., 2011; Dean et al., 2005; Watkins & Barbour, 2008; Rabinowitz et al., 2011; Natan et al., 2017; Chambers et al., 2016; Resnik et al., 2017).
Consistent with the role of synaptic zinc in gain modulation of sound-evoked responses, sensory experience bidirectionally modulates the levels of vesicular zinc and synaptic zinc signaling in several sensory brain areas (Nakashima & Dyck, 2009; Li et al., 2017; Kalappa et al., 2015; McAllister & Dyck, 2017). In the somatosensory cortex, whisker plucking increases zinc levels, whereas whisker stimulation reduces zinc levels (Brown & Dyck 2002; 2005). In the primary visual cortex, monocular deprivation increases vesicular zinc levels (Dyck et al., 2003). In the retina, optic nerve damage increases zinc levels, which in turn inhibit optic nerve regeneration and promote cell death (Li et al., 2017). In the dorsal cochlear nucleus (DCN), an auditory brainstem nucleus, exposure to loud sound reduces vesicular zinc levels and synaptic zinc signaling (Kalappa et al., 2015). Yet, the cellular and molecular mechanisms underlying the experience-dependent plasticity of synaptic zinc levels remain unknown. Elucidating these mechanisms is crucial for understanding how the brain adapts during normal sensory processing, and why it fails to properly adjust in sensory disorders associated with pathological central adaptation, such as in tinnitus (Auerbach et al., 2014).
To determine the mechanisms of zinc plasticity, we used electrophysiology and fluorescent imaging in the DCN, which contains parallel fibers (PFs) with high levels of synaptic zinc that are modulated by sound experience (Frederickson et al., 1988; Rubio & Juiz, 1998; Kalappa et al., 2015). We investigated these mechanisms in vitro, in response to electrical synaptic activation in brain slices, as well as in vivo, in response to loud sound exposure. Our results demonstrate bidirectional activity-dependent zinc plasticity, and reveal a novel mechanism of plasticity in central synapses that involves mGluR activation.
Results
Bidirectional activity-dependent zinc plasticity requires Group 1 mGluRs
To investigate the mechanisms underlying zinc plasticity, we first determined whether we could induce zinc plasticity in DCN PF synapses in mouse brain slices. In these synapses, application of the fast, high-affinity extracellular zinc chelator ZX1 potentiates AMPAR and NMDAR EPSCs via postsynaptic mechanisms (Anderson et al., 2015; Kalappa et al., 2015; 2017). Moreover, this potentiation is dependent on ZnT3, the transporter that loads zinc into synaptic vesicles (Palmiter et al., 1996; Cole et al., 1999), indicating that synaptic zinc inhibits AMPAR and NMDAR EPSCs. We therefore used the ZX1 potentiation of AMPAR and NMDAR EPSCs, as well as imaging of presynaptic zinc levels, as assays for monitoring zinc plasticity in the DCN.
Consistent with previous studies, we found that ZX1 potentiated postsynaptic PF AMPAR EPSCs in DCN cartwheel cells (CWCs), a class of inhibitory interneurons (Fig. 1 A-B) (Kalappa et al., 2015; 2017). To develop an in vitro model of zinc plasticity to study underlying mechanisms, we tested whether we can induce zinc plasticity by using patterns of synaptic activation that induce plasticity of glutamatergic signaling, such long-term potentiation and depression (LTP/LTD) in the DCN PF synapses (Tzounopoulos et al., 2004; 2007; Fujino & Oertel, 2003). We started by examining the effect of ZX1 on AMPAR EPSCs following high-frequency stimulation of PFs (HFS, 3 x 100 Hz for 1 sec, 10 sec inter-stimulus interval), which induces glutamatergic LTP (Fujino & Oertel, 2003). After applying HFS and obtaining a new stable baseline, we renormalized AMPAR EPSC amplitude to the new baseline, to quantify the amount ZX1 potentiation after HFS (Fig. 1 C). After HFS, ZX1 application did not potentiate AMPAR EPSCs (Fig. 1 C-D). The loss of ZX1 potentiation indicates a loss of zinc-mediated inhibition of AMPARs, suggesting that HFS caused a reduction in synaptic zinc signaling, thereby inducing zinc plasticity.
After establishing that HFS reduced zinc signaling, we then used this stimulation paradigm as a tool for studying the underlying mechanisms. NMDARs mediate the induction of LTP and LTD in the DCN and most central synapses (Malenka & Nicoll, 1993; Tzounopoulos et al., 2004; 2007; Fujino & Oertel, 2003). To test the role of NMDARs in the induction of zinc plasticity by HFS, we blocked NMDARs with APV (NMDAR antagonist, 50 μM; Fig. 2 A). As evidenced by the lack of ZX1 potentiation of AMPAR EPSCs after HFS, APV did not affect HFS-induced reductions in zinc signaling (Fig. 2 A, C), suggesting that NMDARs are not required for the induction of zinc plasticity.
Parallel fiber synapses in the DCN also exhibit glutamatergic plasticity that involves metabotropic glutamate receptor (mGluR) signaling (Fujino & Oertel, 2003). Furthermore, Group 1 mGluRs are expressed in CWCs and in the DCN molecular layer, which contains PF terminals (Wright et al., 1996; Bilak & Morest, 1998). We therefore tested whether Group 1 mGluR signaling is necessary for the induction of zinc plasticity. To test this hypothesis, we repeated the experiment shown in 2A, but now we blocked Group 1 mGluRs with MPEP (4 μM, mGluR5-selective antagonist) and LY367385 (100 μM, mGluR1-selective antagonist) (Fig. 2 B). Under these conditions, HFS did not affect ZX1 potentiation compared to control, no HFS, conditions (Fig. 2 B-C), suggesting that Group 1 mGluR activation is necessary for the induction of zinc plasticity by HFS.
Glutamatergic plasticity is bidirectional: synapses undergo LTP or LTD in response to high- or low-frequency stimulation, respectively (Malenka & Nicoll, 1993; Mulkey & Malenka, 1992; Fujino & Oertel 2003). To determine whether zinc plasticity is bidirectional, we tested whether low-frequency stimulation (LFS) increases zinc signaling (Fig. 3 A). Because the induction of zinc plasticity depends on mGluR activation, we used conditions that favor mGluR activation such as LFS (5 Hz, 3 min), blockade of NMDARs with APV, and high extracellular concentrations of divalent ions (4 mM Ca2+ and Mg2+; Oliet et al., 1997). Compared to control (no LFS), LFS increased the amount of ZX1 potentiation (Fig. 3 A, C). Note that control ZX1 potentiation in these conditions (Fig. 3 C) was slightly less, albeit not significantly different (p=0.11, unpaired t test), than previous control experiments performed in ACSF with 2.4/1.3 mM of extracellular Ca2+/Mg2+ (Fig. 1 D). This is likely due to reduced neuronal excitability (Oliet et al., 1997; Kalappa et al., 2015). Together, these results suggest that LFS increased synaptic zinc signaling, thus demonstrating that activity-dependent plasticity of zinc signaling is bidirectional: HFS reduces zinc signaling, whereas LFS increases zinc signaling.
We next tested whether Group 1 mGluR activation is necessary for the increased zinc signaling by LFS. Whereas LFS alone increased ZX1 potentiation compared to control, LFS in the presence of MPEP and LY367385 did not increase ZX1 potentiation (Fig. 3 B-C). This indicates that Group 1 mGluR activation is necessary for the induction of zinc plasticity by LFS. Together, these results reveal that activation of Group 1 mGluR signaling is necessary for the induction of bidirectional zinc plasticity by HFS or LFS.
Group 1 mGluR activation is sufficient to induce bidirectional zinc plasticity
Is activation of Group 1 mGluRs sufficient to induce zinc plasticity? Because Group 1 mGluRs are required for both increases and decreases in zinc signaling by different stimulation paradigms, we hypothesized that the direction of plasticity depends on differential activation of mGluRs during the induction protocols. To test this, we applied high or low concentrations of DHPG (Group 1 mGluR agonist, 50 μM or 5 μM). Consistent with previous studies, application of 50 μM DHPG caused LTD (Fig. 4 A; Huber et al., 2001; Wisniewski & Car, 2002; Snyder et al. 2001). After applying 50 μM DHPG, obtaining a new stable baseline, and then applying ZX1, we observed that the ZX1 potentiation of EPSCs was significantly increased compared to control (Fig. 4 A, C-D). This result indicates that a high concentration of DHPG increases zinc signaling: Group 1 mGluR activation is sufficient to induce zinc plasticity.
Because increases and decreases in zinc signaling induced by L/HFS depend on Group 1 mGluR activation (Fig. 2 and 3), we next tested whether application of a lower concentration of DHPG reduces zinc signaling. After applying 5 μM DHPG and obtaining a new stable baseline, ZX1 did not potentiate EPSCs (Fig. 4 B-D). This result indicates that activation of Group 1 mGluRs with a lower concentration of DHPG reduces zinc signaling. Together, these results demonstrate that Group 1 mGluR activation is sufficient to induce bidirectional zinc plasticity. Furthermore, the direction of zinc plasticity depends on the concentration of DHPG: 50 μM DHPG increases zinc signaling, whereas 5 μM DHPG reduces zinc signaling. These results are consistent with the notion that bidirectional zinc plasticity depends on differential activation of Group 1 mGluRs by either L/HFS or high/low concentrations of DHPG.
Group 1 mGluR activation bidirectionally modulates presynaptic zinc levels
We used activity-dependent changes in the amount of ZX1 potentiation of AMPAR EPSCs for assessing changes in zinc signaling. However, ZX1 potentiation is determined by the postsynaptic zinc-mediated inhibition of AMPAR EPSCs, as well as the amount of presynaptic zinc release (Kalappa et al., 2015). Because previous studies have demonstrated activity-dependent modulation of presynaptic zinc levels (Nakashima & Dyck, 2009; Kalappa et al., 2015), we hypothesized that the plasticity of zinc signaling is due, at least in part, to the modulation of presynaptic zinc levels. To quantify potential changes in presynaptic zinc levels, we used DA-ZP1 (0.5 μM), a fluorescent intracellular zinc sensor capable of tracking presynaptic zinc levels in PF terminals (Kalappa et al., 2015; Zastrow et al., 2016). To induce zinc plasticity, we applied DHPG, which, unlike synaptic stimulation, induces zinc plasticity in all terminals in the slice. We imaged DA-ZP1 fluorescence (Fig. 5 A) in control slices and in slices treated with DHPG (50 μM or 5 μM). Compared to the non-treated control slices, incubation of slices with 50 μM DHPG increased DA-ZP1 fluorescence in the DCN molecular layer, indicating increased presynaptic zinc levels in PF terminals (Fig. 5 B). In contrast, incubation of slices with 5 μM DHPG reduced DA-ZP1 fluorescence, indicating reduced zinc levels (Fig. 5 B). These results demonstrate that application of different concentrations of DHPG causes bidirectional modulation of presynaptic zinc levels. Furthermore, these results are consistent with our electrophysiological experiments: 50 μM DHPG increases zinc levels and ZX1 potentiation of EPSCs, whereas 5 μM DHPG reduces zinc levels and ZX1 potentiation of EPSCs. Although our results do not rule out potential postsynaptic mechanisms of zinc plasticity, our results demonstrate that zinc plasticity involves modulation of presynaptic zinc levels.
Zinc plasticity does not depend on AMPAR plasticity, and it affects zinc-mediated inhibition of NMDARs
Pharmacological or electrophysiological induction of zinc plasticity is consistent with modulation of presynaptic zinc levels, but postsynaptic modifications are also possible. Because PFs exhibit postsynaptic AMPAR LTP (Fujino & Oertel, 2003; Tzounopoulos et al., 2007; Zhao & Tzounopoulos, 2011), we therefore sought to determine whether ZX1 potentiation is correlated with the amount of AMPAR LTP. To test for this, we plotted the ZX1 potentiation versus the changes in AMPAR EPSC amplitude after HFS (Fig. 6 A). We found no correlation between AMPAR plasticity and ZX1 potentiation (Fig. 6 A). This result suggests that modulation of zinc signaling does not depend on the amount of AMPAR potentiation, and supports the notion that modification of postsynaptic AMPARs during LTP are likely not involved in zinc plasticity.
Our DA-ZP1 imaging results suggest that zinc plasticity involves modulation of presynaptic zinc signaling. Therefore, the induction of zinc plasticity should also affect postsynaptic NMDAR EPSCs, which are inhibited by zinc via direct high-affinity receptor binding (Paoletti et al., 1997; Vergnano et al., 2014). To test this hypothesis, we quantified the ZX1 potentiation of NMDAR EPSCs after inducing zinc plasticity with HFS. To monitor NMDAR EPSCs, we used a short train of presynaptic stimulation (5 pulses at 20 Hz) to activate extrasynaptic NMDARs, for NMDAR EPSCs recorded in somata of CWCs are mostly mediated by extrasynaptic NMDARs activated by glutamate spillover during this short train (Anderson et al., 2015). To avoid keeping CWCs at +40 mV for too long while recording NMDAR EPSCs, and to maintain the same induction protocol used in our previous experiments, we initially recorded AMPAR EPSCs at -70 mV, then applied HFS. Subsequently, we blocked AMPARs with DNQX (20 μM, AMPA/kainate receptor antagonist) and recorded at +40 mV to obtain a stable baseline of NMDAR EPSCs before applying ZX1 (Fig. 6 B). Consistent with our results on AMPAR EPSCs, after HFS, ZX1 no longer potentiated NMDAR EPSCs (Fig. 6 B, D), whereas ZX1 enhanced NMDARs EPSCs in interleaved control experiments where HFS was not applied, also consistent with previous results from DCN CWCs (Anderson et al., 2015). These results demonstrate that HFS-induced zinc plasticity reduces zinc-mediated inhibition of NMDARs. To determine whether this modulation shares the same mechanism with the zinc plasticity evidenced by changes in the ZX1 potentiation of AMPAR EPSCs, we tested whether Group 1 mGluR signaling is required. Indeed, application of MPEP and LY367385 blocked the observed zinc plasticity (Fig. 6 C-D), demonstrating that Group 1 mGluR signaling is required for HFS-induced plasticity of NMDAR zinc inhibition.
Finally, to control for potential non-specific effects of ZX1 on presynaptic zinc levels, perhaps through redistribution of zinc levels in different cellular compartments, we tested the effect of ZX1 on DA-ZP1 fluorescence. Indeed, application of ZX1 10 min prior and during DA-ZP1 application did not affect the fluorescence levels in the molecular layer of the DCN (Fig. 6 E). This result indicates that ZX1 does not affect presynaptic zinc levels, and is consistent with the notion that changes in the ZX1 potentiation of EPSCs after L/HFS detect activity-dependent changes in presynaptic zinc levels and signaling.
Together, our results indicate that pharmacological or synaptic activation of Group 1 mGluRs causes bidirectional synaptic zinc plasticity. Synaptic zinc plasticity modulates zinc inhibition of NMDARs and AMPARs, suggesting that it is, at least in part, expressed by activity-dependent changes of presynaptic zinc levels, regardless of the postsynaptic effector of zinc.
Sound-dependent zinc plasticity requires Group 1 mGluRs in vivo
Our experiments described here, using in vitro brain slice electrophysiology and fluorescent imaging in the DCN, point toward a mechanism of bidirectional plasticity of presynaptic zinc levels mediated by Group 1 mGluR signaling. We therefore hypothesized that mGluR activation may also be necessary for the sound-dependent zinc plasticity observed in the DCN (Kalappa et al., 2015). To test this hypothesis, we examined DA-ZP1 fluorescence, and ZX1 potentiation of EPSCs, in DCN slices from mice exposed to loud sound (116 dB, 4 hours). Consistent with previous studies (Kalappa et al., 2015), slices from noise-exposed (N.E.) mice exhibited reduced DA-ZP1 fluorescence in the molecular layer, indicating reduced presynaptic zinc levels, compared to slices from control, sham-exposed, mice (Fig. 7 A). To test whether Group 1 mGluRs are necessary for the sound-induced plasticity of zinc levels, we administered a systemic, blood brain barrier-permeable Group 1 mGluR antagonist (AIDA, i.p., 2 mg/kg) twice: 30 min before and 1.5 hours after beginning the noise exposure. Slices from N.E. mice treated with AIDA did not show changes in DA-ZP1 fluorescence (Fig. 7 A), suggesting that inhibition of Group 1 mGluR activity blocked the sound-induced reduction of presynaptic zinc levels. Moreover, because N.E. mice given vehicle (saline) treatment showed normal reductions in zinc levels, and AIDA treatment in sham-exposed mice caused no change in zinc levels (Fig. 7 A), these results suggest that Group 1 mGluR activation is necessary for sound-induced changes of presynaptic zinc levels, but does not affect baseline zinc levels.
To test whether Group 1 mGluRs are required for sound-induced reductions in synaptic zinc signaling, we quantified the ZX1 potentiation of AMPAR EPSCs in DCN slices from N.E. mice and N.E., AIDA-treated mice (Fig. 7 B-C). Consistent with previous studies (Kalappa et al., 2015), and the reduction of presynaptic zinc levels after noise exposure (Fig 7 A), ZX1 potentiation was not observed in slices from N.E. mice (Fig. 7 B-C). However, in slices from N.E., AIDA-treated mice, ZX1 potentiation was observed (Fig. 7 B-C). Together, these results indicate that Group 1 mGluRs are required for sound-induced plasticity of presynaptic zinc levels and zinc signaling. Furthermore, although AIDA treatment blocked zinc plasticity in DCN PF synapses (Fig. 7 A-C), it did not affect assays that are sensitive to presynaptic glutamate release probability, such as paired-pulse ratio (PPR) and coefficient of variation (CV) analysis (Fig. 7 D). This indicates that mGluR-dependent zinc plasticity specifically modulates presynaptic zinc levels and zinc signaling, without affecting presynaptic glutamate signaling in PFs. Furthermore, AIDA treatment did not affect sound-induced hearing loss in N.E. mice, quantified using Auditory Brainstem Responses (ABRs) (Fig. 7 E). ABRs reflect the synchronous activity, arising from the auditory nerve (Wave I), of auditory brainstem nuclei to the inferior colliculus (Waves II-V) in response to sound stimuli. Elevated ABR thresholds indicate increased hearing thresholds. In N.E. mice, because no ABRs were detected when recording from the exposed (ipsilateral) ear, we recorded ABRs from the non-exposed (contralateral) ear (Fig. 7 E). Moreover, similar ABR thresholds may be accompanied by differences in the suprathreshold response of Wave I, which could reflect differential degeneration of the auditory nerve (Kujawa & Liberman, 2009). AIDA treatment did not affect noise-induced changes in either ABR thresholds or Wave I amplitude (Fig. 7 F), thus indicating that the effect of AIDA on blocking zinc plasticity is not due to differential noise-induced hearing loss after AIDA treatment. Together, these results demonstrate that Group 1 mGluR signaling is required for sound-dependent zinc plasticity in vivo.
Discussion
Our results show that zinc plasticity is an experience- and mGluR-dependent process that bidirectionally modulates synaptic zinc levels and signaling in the DCN. Is this a general mechanism that applies to all synaptic zinc-containing brain areas? Synaptic zinc is present throughout the neocortex and other brain structures, such as the amygdala and the hippocampus (McAllister & Dyck, 2017). Moreover, synaptic zinc is modulated by sensory activity throughout the sensory cortex (McAllister & Dyck, 2017), shapes the gain of central sensory responses (Anderson et al., 2017), and when upregulated by optic nerve injury, it inhibits retinal ganglion cell survival and axon regeneration (Li et al., 2017). It is therefore likely, although not tested here, that the reported zinc plasticity mechanism is a general mechanism that dynamically modulates sensory processing for adaptation to different sensory environments and injury.
Whereas the exact synaptic, natural, and ethologically relevant stimuli that elicit zinc plasticity remain unknown, here we developed an in vitro and in vivo model for studying bidirectional zinc plasticity. This is a crucial step towards further elucidation of the detailed natural stimuli eliciting zinc plasticity, as well as the precise cellular and molecular mechanisms underlying the induction and expression of zinc plasticity. Moreover, our model will be useful for probing the behavioral consequences of zinc plasticity and their underlying synaptic mechanisms, which remain unknown.
Mechanisms of Group 1 mGluR-mediated zinc plasticity
Our results show that differential activation of Group 1 mGluRs, by either L/HFS or high/low concentrations of DHPG, determines the induction and direction of zinc plasticity. Both LFS and HFS induce mGluR-mediated glutamatergic plasticity in various central synapses. In the hippocampus, LFS induces mGluR-LTD (Oliet et al., 1997; Huber et al., 2000), but when NMDARs are blocked, mGluR-LTD is induced by HFS (Wu et al., 2001). In cerebellar climbing fibers, HFS induces mGluR-LTD (Brasnjo & Otis, 2001; Dzubay & Otis, 2002), whereas in DCN PFs, mGluR activation contributes to both HFS-induced LTP and LFS-induced LTD (Fujino & Oertel, 2003). Our results show that prolonged LFS increases zinc signaling, similarly to Group 1 mGluR activation with 50 μM DHPG; whereas, brief HFS reduces zinc signaling, similarly to activation with 5 μM DHPG. This suggests that prolonged LFS activates Group 1 mGluR signaling differently than brief HFS. Group 1 mGluRs, mGluR1 and mGluR5, are linked to the IP3-Diacylglycerol (DAG) signaling pathway, leading to intracellular rises in Ca2+ from intracellular stores (Abdul-Ghani et al., 1996; Conn & Pin, 1997; Kim et al., 2008). We therefore propose, albeit not tested here, that differential mGluR activation, by L/HFS, leads to subsequent release of different amounts or types of intracellular Ca2+ signals. Different Ca2+ signals may in turn activate diverse signaling pathways that ultimately lead to increased and decreased synaptic zinc levels. An analogue that comes to mind is the mechanism via which differential activation of NMDARs, by various levels of synaptic activity, leads to variable Ca2+ levels and signaling, ultimately determining the induction of both LTP and LTD (Malenka & Bear, 2004).
In the DCN, cartwheel cells (CWCs) express Group 1 mGluRs, particularly mGluR1, suggesting that the locus of induction of zinc plasticity is postsynaptic (Wright et al., 1996). Because zinc plasticity involves modulation of presynaptic zinc levels, this would suggest the presence of a retrograde signal from CWCs involved in the expression of zinc plasticity in PFs. Alternatively, the presence of mGluR1 on axon terminals in the DCN molecular layer may support a presynaptic locus of induction (Bilak & Morest, 1998). Because ZnT3 determines vesicular zinc levels (Palmiter et al., 1996; Cole et al., 1999), modulation of ZnT3 expression or function may underlie the expression of zinc plasticity. In the retina, optic nerve injury increases ZnT3 immunostaining, supporting that increases in ZnT3 expression mediate increases in synaptic zinc levels (Li et al., 2017). Yet, in the barrel cortex after whisker plucking, ZnT3 protein and mRNA levels remain unchanged despite increases in vesicular zinc content (Liguz-Lecznar et al., 2005; Nakashima et al., 2011). As such, changes in barrel cortex synaptic zinc levels are more consistent with a modulation of ZnT3 function. In the context of ZnT3 modulation, it is interesting that the vesicular glutamate transporter 1 (VGlut1) is co-targeted to synaptic vesicles with ZnT3, and increases ZnT3 zinc transport (Salazar et al., 2005). Because VGlut1 is highly expressed in the DCN molecular layer (Zhou et al., 2007), one hypothesis is that modulation of VGlut1 may modulate ZnT3 function in PF terminals. However, noise-induced zinc plasticity does not affect presynaptic glutamate dynamics (Fig. 7 D), suggesting that zinc and glutamate signaling can be independently modulated. Therefore, a change in ZnT3 expression is a more likely candidate for mediating decreased synaptic zinc levels after noise exposure in the DCN. While our results reveal a role for Group 1 mGluRs and changes in presynaptic zinc levels for the induction and expression of zinc plasticity, further support from additional experiments will be necessary to determine the detailed induction and expression mechanisms.
Implications of zinc plasticity for LTP and LTD
Low frequency stimulation typically induces glutamatergic LTD (Mulkey & Malenka, 1992; Fujino & Oertel, 2003), whereas HFS typically induces LTP (Malenka & Nicoll, 1993; Fujino & Oertel, 2003). Based on our findings, we propose that zinc plasticity by L/HFS may further enhance LTD and LTP. Namely, LFS increases zinc signaling, thereby increasing NMDA/AMPAR inhibition, which would further enhance the effects of glutamatergic LTD. HFS reduces zinc signaling, thereby reducing NMDA/AMPAR inhibition, which would further enhance LTP. Overall, zinc plasticity serves as a positive feedback mechanism to enhance the effects of LTP or LTD on glutamatergic synaptic transmission.
Zinc inhibits NMDARs and thus modulates the induction of NMDAR-dependent LTP and LTD in the hippocampus (Izumi et al., 2006; Takeda et al., 2009; Vergnano et al., 2014). As such, zinc plasticity may also contribute to ‘metaplasticity’, the modulation of subsequent LTP and LTD (Abraham & Tate, 1997). Based on our findings, HFS-dependent reductions in zinc signaling will promote subsequent NMDAR-dependent LTP and decrease subsequent NMDAR-dependent LTD. Conversely, increases in zinc signaling after LFS would inhibit subsequent NMDAR-LTP and enhance NMDAR-LTD. Together, zinc plasticity acts a positive feedback signal in NMDAR-dependent synaptic plasticity and metaplasticity.
In synapses expressing mGluR-dependent LTD (Oliet et al., 1997), zinc plasticity is also expected to act as a positive feedback signal, for LFS is expected to enhance zinc levels and further suppress glutamatergic transmission. However, in synapses expressing NMDAR-independent LTP, such as in hippocampal mossy fiber-CA3 synapses, zinc plasticity is not expected to act as a positive feedback. Zinc contributes to mossy fiber presynaptic LTP in response to HFS (Pan et al., 2011), and therefore if HFS reduces zinc levels in mossy fiber synapses, zinc plasticity will act as negative feedback signal for subsequent LTP induction. Taken together, we propose that the role of zinc plasticity in LTP and LTD depends on the specific mechanisms underlying LTP and LTD, but zinc plasticity mostly acts as a positive feedback signal, especially for the most abundant forms of LTP and LTD in the brain, which depend on NMDAR activation.
Clinical and translational implications of zinc plasticity
In the context of zinc plasticity as a positive feedback signal for NMDAR-dependent plasticity, it is interesting that exposure to loud sound -- known to induce tinnitus -- decreases synaptic zinc levels in the DCN. Such decreases could potentially lead to runaway excitation due to enhanced LTP and decreased LTD, and thus to pathological DCN hyperactivity associated with tinnitus (Tzounopoulos, 2008). Noise-induced pathological hyperexcitability through LTP/LTD-like mechanisms in the DCN PF synapses has been hypothesized and recently implicated in tinnitus treatment (Tzounopoulos, 2008; Marks et al., 2018), therefore suggesting that noise-induced reductions in synaptic zinc might contribute to tinnitus.
In addition to a potential involvement of zinc signaling in tinnitus, pathological neuronal zinc signaling is associated with numerous other neurological disorders, such as schizophrenia, Alzheimer’s disease, and autism. Variants in the ZnT3 gene are associated with schizophrenia (Perez-Becerril et al., 2016); Alzheimer’s patients display reduced levels of ZnT3 expression (Olesen et al., 2016); ZnT3KO mice exhibit cognitive deficits similar to Alzheimer’s and autism-like phenotypes (Adlard et al., 2010; Yoo et al., 2016); and a zinc ionophore rescues behavioral phenotypes in mouse models of autism (Lee et al., 2015). Therefore, there is converging evidence that pathological zinc signaling may play a significant role in the cognitive deficits observed in these neurological disorders. As such, understanding the cellular mechanisms that govern the plasticity of zinc levels is a crucial first step in determining the molecular pathways that, when disrupted, may lead to pathological zinc levels. Furthermore, elucidating these mechanisms, such as the role of Group 1 mGluR signaling identified here, may lead to novel targets and approaches for treating pathological zinc signaling in disease states.
Materials and Methods
Animals
Male or female ICR mice (Envigo) were used in this study, aged between postnatal day 17 (P17) to P28. All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh, Pittsburgh, PA.
Brain slice preparation for electrophysiology and fluorescent imaging
Mice were deeply anesthetized with isoflurane (3% in O2), then immediately decapitated and their brains were removed. Brain slices were prepared in artificial cerebrospinal fluid (ACSF, 34°C) containing the following (in mM): 130 NaCl, 3 KCl, 1.2 CaCl2·2H2O, 1.3 MgCl2·6H2O, 20 NaHCO3, 3 HEPES, and 10 D-Glucose, saturated with 95% O2/5% CO2 (vol/vol), pH = 7.25-7.35, ~300 mOsm. Using a Vibratome (VT1200S; Leica), coronal brain slices (210 μm thickness) containing the left dorsal cochlear nucleus (DCN) were cut, then placed in a chamber containing warm (34°C) ACSF, and incubated for 60 min at 34°C, then room temperature (no longer than 3 hours) before beginning electrophysiology or imaging experiments. Incubating ACSF was the same as cutting ACSF, except it was stirred with Chelex 100 resin (Bio-Rad) for 1 hour to remove contaminating zinc, then filtered using Nalgene rapid flow filters lined with polyethersulfone (0.2 μm pore size). After filtering, high purity CaCl2·2H2O and MgCl2·6H2O (99.995%; Sigma Aldrich) were added. All plastic and glassware used for these experiments were washed with 5% nitric acid.
Electrophysiology
Whole-cell recordings
DCN slices were transferred to the recording chamber and perfused with ACSF (1–2 mL/min), maintained at ~34°C using an inline heating system (Warner Instruments). Recording ACSF was the same as incubating ACSF (see above), except it contained 2.4 mM CaCl2·2H2O. Whole-cell recordings from cartwheel cells were performed using glass micropipettes (3–6 MΩ; Sutter Instruments). Cartwheel cells were identified by the presence of complex spikes in cell-attached configuration before break-in or in response to current injections in current-clamp mode after break-in (Zhang & Oertel, 1993; Manis et al., 1994; Tzounopoulos et al., 2004). Recording pipettes were filled with a potassium-based internal solution (except for Fig. 6, see below) containing the following (in mM): 113 K-gluconate, 4.5 MgCl2·6H2O, 14 Tris-phosphocreatine, 9 HEPES, 0.1 EGTA, 4 Na2ATP, 0.3 Tris-GTP, and 10 sucrose (pH = 7.25, 295 mOsm). For experiments shown in Fig. 6 B-D measuring NMDAR EPSCs, recordings were performed using a cesium-based internal solution containing the following (in mM): 128 Cs(CH3O3S), 10 HEPES, 4 MgCl2·6H2O, 4 Na2ATP, 0.3 Tris-GTP, 10 Tris-phosphocreatine, 1 EGTA, 1 QX-314, and 3 Na-ascorbate (pH = 7.25, 300 mOsm). Recordings were performed using ephus (Suter et al., 2010) and a MultiClamp 700B amplifier (Axon Instruments). Data were sampled at 10 kHz and low-pass-filtered at 4 kHz. Series resistance (Rs) and input resistance (Rm) were monitored during the recording period by delivering -5 mV voltage steps for 50 ms. Rs was calculated by dividing the -5 mV voltage step by the peak current generated immediately after the voltage step. Rm was calculated by dividing the -5 mV voltage step by the difference between the baseline and steady-state hyperpolarized current, then subtracting Rs. Data were excluded if Rs or Rm changed by more than 20% from the baseline period. EPSCs were evoked using an Isoflex stimulator (A.M.P.I., 0.1 ms pulses) through a glass ACSF-containing theta electrode to stimulate the zinc-rich parallel fibers. All EPSCs were recorded in the presence of SR95531 (20 μM, GABAAR antagonist) and strychnine (1 μM, GlyR antagonist). AMPAR EPSCs were recorded in voltage-clamp mode at -70 mV. For paired-pulse experiments, the inter-stimulus interval was 50 ms. NMDAR EPSCs were evoked by a 5-pulse stimulus train (20 Hz) (Anderson et al., 2015), recorded in voltage clamp mode at +40 mV, and in the presence of DNQX (20 μM, AMPA/kainate receptor antagonist). All drugs were always bath applied.
Induction of plasticity
High-frequency stimulation (HFS) consisted of 3 trains of 100 Hz pulses for 1 sec, with 10 sec between trains. For the experiments shown in Fig. 1 C, a subset of cells (n=5) were depolarized to -10 mV during each HFS train, while the other subset (n=6) were held at -70 mV during HFS. Because we observed no difference in the zinc plasticity (EPSC % baseline after ZX1) between these subsets (depolarized = 102.7±5.4%, non-depolarized = 108.8±8.6%, p = 0.58, unpaired t test), they were grouped together for subsequent analysis. For all other experiments, cells were voltage-clamped at -70 mV during HFS. For experiments measuring NMDAR EPSCs after HFS (Fig. 6 B-D), DNQX (20 μM) was added after HFS, then cells were voltage-clamped at +40 mV to record NMDAR EPSCs. Low-frequency stimulation (LFS) consisted of 5 Hz pulses for 3 min. During LFS, cells were held at -80 mV in current-clamp mode. To isolate mGluR-mediated plasticity, all LFS experiments were performed in the presence of APV (50 μM, NMDAR antagonist), and with external ACSF containing 4 mM CaCl2·2H2O and 4 mM MgCl2·6H2O (Oliet et al., 1997). For normalized EPSCs (% baseline), EPSC amplitudes were normalized to the average EPSC amplitude during the 5 min baseline period before L/HFS, DHPG, or ZX1 application. To quantify ZX1 potentiation after L/HFS or DHPG application, EPSC amplitudes were renormalized to the average EPSC amplitude of the new baseline period 5 min before ZX1 application. ZX1 potentiation (shown in bar graphs) was quantified as the percent increase in the average EPSC amplitude during the last 5 min of ZX1 application compared to the 5 min baseline period before ZX1 application.
Vesicular zinc imaging with DA-ZP1
After preparation and incubation of DCN slices (described above), slices were transferred to the imaging chamber and perfused with recirculating ACSF (2–3 mL/min) maintained at ~34°C. Imaging of presynaptic vesicular zinc levels in DCN parallel fibers was performed using DA-ZP1, a high-affinity, membrane permeable fluorescent zinc sensor (Zastrow et al., 2016). DA-ZP1 (0.5 μM) was added to the ACSF, and allowed to incubate for at least 5–10 min before imaging. In order to quantify the effects of DHPG (Fig. 5) or noise exposure (Fig. 7A) compared to controls, pairs of treated and control slices were always imaged simultaneously in the same imaging chamber. Images were acquired using an upright microscope (Olympus BX5) with epifluorescence optics through a 20x water immersion objective. Green fluorescent signals were isolated using a Pinkel filter set (Semrock LF488/543/625-3X-A-000) in response to excitation by an ephus-driven blue LED (M470L2; Thorlabs), and images were acquired using a CCD camera (Retiga 2000R, QImaging). Images consisted of 20 frames captured at 0.2 Hz, which were then averaged together and analyzed using custom routines in MATLAB (Mathworks). In order to quantify zinc-sensitive fluorescence in the DCN molecular layer, a region of interest (ROI) within each slice was selected containing the zinc-rich molecular layer or the zinc-lacking deep layer (Zastrow et al., 2016; Kalappa et al., 2015). Fluorescence intensity was averaged within each ROI, and the zinc-sensitive fluorescence for each slice was calculated by subtracting the deep layer fluorescence from the molecular layer fluorescence. Zinc-sensitive fluorescence in treated slices (DHPG in Fig. 5B or noise exposure in Fig. 7A) was normalized to the simultaneously imaged control, non-treated, slice. For DHPG experiments (Fig. 5), DHPG (5 μM or 50 μM) was added to ACSF in the incubating chamber at least 20 min before initiating imaging. For experiments testing potential effects of ZX1 on DA-ZP1 fluorescence (Fig. 6E), imaging of ZX1-treated or control slices was performed in the presence or absence, respectively, of ZX1 (100 μM) in the external ACSF. For ZX1-treated slices, ZX1 was added to the ACSF 10 min before adding DA-ZP1, then fluorescent images were acquired 10 min after DA-ZP1 application (20 min after ZX1 application).
Noise exposure
Noise exposure was performed based on previously published methods (Kalappa et al., 2015). Sham- or noise-exposed mice were anesthetized using 3% isoflurane during induction and 1–1.5% during maintenance. Noise-exposed mice were exposed for 4 hours to narrow bandpass noise at 116 dB sound pressure level (SPL), centered at 16 kHz with a 1.6 kHz bandwidth. Noise was presented unilaterally (left ear) through a pipette tip inserted into the left ear canal, with the other end attached to a calibrated speaker (CF-1; Tucker Davis Technologies). Sham-exposed mice underwent an identical procedure except without any noise exposure. For mice given intraperitoneal injections of AIDA (2 mg/kg) or vehicle (saline, 0.9% NaCl), one injection was given 30 min prior to exposure, and a second injection was given 2 hours later. After noise- or sham-exposure, ABRs were collected and mice recovered from anesthesia, then DCN slices were prepared (within 30 min after exposure).
ABRs
Auditory Brainstem Responses (ABRs) were measured based on previously published methods (Kalappa et al., 2015). ABRs were recorded immediately after noise- or sham-exposure. During ABR measurements, mice were anesthetized using 3% isoflurane during induction and 1–1.5% during maintenance. Mice were placed in a sound attenuating chamber and temperature was maintained at ~37°C using a heating pad. A subdermal electrode was placed at the vertex, the ground electrode placed ventral to the right pinna, and the reference electrode placed ventral to the left pinna (sham- or noise-exposed ear). For ABR measurements from contralateral ears of noise-exposed mice, the reference electrode was placed ventral to the right pinna (contralateral ear) and the ground electrode placed ventral to the left pinna. ABRs were detected in response to 1 ms click sound stimuli, presented through a pipette tip inserted into the ear canal, with the other end attached to the speaker (CF-1; Tucker Davis Technologies). ABRs were recorded in response to clicks presented in 10 dB steps, ranging from 0–80 dB SPL. 1 ms clicks were presented at a rate of 18.56/sec using System 3 software package from Tucker Davis Technologies, and ABRs were averaged 512 times and filtered using a 300–3,000 Hz bandpass filter. ABR threshold was defined as the lowest stimulus intensity which generated a reliable Wave 1 in the response waveform. Wave 1 amplitude was measured as the peak-to-trough amplitude of the first wave in the ABR waveform (latency ~2 ms), in response to 80 dB SPL clicks.
Drugs
All chemicals used for ACSF and internal solutions were purchased from Sigma-Aldrich. The following drugs were purchased from HelloBio: SR95531 hydrobromide, DL-AP5, DNQX disodium salt, MPEP hydrochloride, LY367385, and (S)-3,5-Dihydroxyphenylglycine (DHPG). Strychnine hydrochloride was purchased from Abcam. (S)-α-Methyl-4-carboxyphenylglycine (MCPG) and (RS)-1-Aminoindan-1,5-dicarboxylic acid (AIDA) were purchased from Tocris. ZX1 was purchased from STREM Chemicals.
Data analysis
All data analysis was performed using Matlab (Mathworks), Excel (Microsoft), or Prism 7 (GraphPad). For statistical comparisons, we used one-sample t tests or unpaired t tests (between groups) if the data passed the Shapiro-Wilk normality test. Welch’s correction was used if variances were significantly different (F test). For non-normally distributed data, we used Wilcoxon signed rank tests or Mann-Whitney tests (between groups). Significance levels are defined as p < 0.05. Group data are presented as mean ± SEM.
Funding
National Institute on Deafness and Other Communication Disorders: R01-DC007905 (T.T.); F31-DC015924 (N.W.V.).
Author Contributions
N.W.V. performed experiments and analyzed data. N.W.V. and T.T. designed experiments and wrote the paper.
Ethics
All methods for animal handling, anesthesia, and sacrifice were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh, Pittsburgh, PA. The approved IACUC protocol numbers used for this study were: 14125118, 14094011, 17071036, and 17127808.
Competing Interests
The authors declare that no competing interests exist.
Acknowledgements
We thank Drs. Stephen Lippard and Jacob Goldberg from Massachusetts Institute of Technology for generously providing DA-ZP1.