Growth differentiation factor-15 promotes glutamate release in medial prefrontal cortex of mice through upregulation of T-type calcium channels

Growth differentiation factor-15 (GDF-15) has been implicated in ischemic brain injury and synapse development, but its involvement in modulating neuronal excitability and synaptic transmission remain poorly understood. In this study, we investigated the effects of GDF-15 on non-evoked miniature excitatory post-synaptic currents (mEPSCs) and neurotransmitter release in the medial prefrontal cortex (mPFC) in mice. Incubation of mPFC slices with GDF-15 for 60 min significantly increased the frequency of mEPSCs without effect on their amplitude. GDF-15 also significantly elevated presynaptic glutamate release, as shown by HPLC. These effects were blocked by dual TGF-β type I receptor (TβRI) and TGF-β type II receptor (TβRII) antagonists, but not by a TβRI antagonist alone. Meanwhile, GDF-15 enhanced pERK level, and inhibition of MAPK/ERK activity attenuated the GDF-15-induced increases in mEPSC and glutamate release. Blocking T-type calcium channels reduced the GDF-15 induced up-regulation of synaptic transmission. Membrane-protein extraction and use of an intracellular protein-transport inhibitor showed that GDF-15 promoted CaV3.1 and CaV3.3 α-subunit expression by trafficking to the membrane. These results confirm previous findings in cerebellar granule neurons, in which GDF-15 induces its neurobiological effects via TβRII and activation of the ERK pathway, providing novel insights into the mechanism of GDF-15 function in cortical neurons.

GDF15 enhanced mEPSC frequency in mPFC by increasing glutamate release. It is generally accepted that changes in mEPSC frequency may be caused by modulation of presynaptic transmission, while changes in amplitude or decay time would suggest a different postsynaptic receptor function. We used HPLC to measure the glutamate concentration in the ACSF to confirm the effect of GDF-15 on glutamate release. Slices from the right and left hemispheres were incubated with ACSF in the presence or absence of GDF-15, and glutamate release was measured as ng/mg. GDF-15 significantly increased glutamate release in mPFC slices by Scientific RepoRts | 6:28653 | DOI: 10.1038/srep28653 147.05% (from 76.52 ± 7.70 to 189.60 ± 22.57 ng/mg, n = 12, p < 0.0001) (Fig. 3A). Consistent with the results for mEPSCs, this increase in glutamate release was reduced by LY2109761 and U0126, to 14.25% (n = 6, p = 0.526) and 12.71% (n = 6, p = 0.597), respectively (Fig. 3B,D). However, in the presence of SB431642, GDF-15 still increased glutamate release by 99.03% (from 64.95 ± 9.8 to 129.27 ± 19.9 ng/mg, n = 8, p = 0.0116) (Fig. 3C).
Overall, these results suggested that GDF-15 enhanced mEPSCs by increasing glutamate release in the mPFC, and activation of the TβRII/ERK pathway is required for the GDF15-mediated increase in glutamate release.

GDF-15 increased the release of glutamate through T-type calcium channels. Neurotransmitter
release is thought to be triggered by enhanced presynaptic calcium levels associated with voltage-gated calcium channels (VGCCs) 16,17 . Growing evidence suggests that T-type VGCCs play a key role in controlling neurotransmission near the rest potential and sustaining neurotransmitter release during mild stimulation 18 . We therefore investigated the association between T-type VGCCs and the GDF-15-induced increase in glutamate release using a specific calcium channel blocker. Treatment of mPFC slices with the T-type VGCC blocker NiCl 2 (100 μM) significantly eliminated the GDF-15-induced increase in mEPSC frequency to −2.3% (n = 18, p = 0.871) (Fig. 4A). Similar results were obtained with different T-type VGCC inhibitors, mibefradil (10 μM) and TTA-P2 (2 μM), which reduced the GDF-15-induced increases in mEPSC frequency to 4.6% (n = 17, p = 0.733) and 3.5% (n = 13, (F) the AMPA or NMDA receptors mediated mEPSC were separated by specific blockers. The left two panels show the representative recordings in presence of AMPA or NMDA receptors blockers; the right three panels show the frequencies, amplitudes and decay times of mEPSC in presence of AMPA or NMDA receptors blockers with or without GDF-15. Results are shown as means ± SEM. *p < 0.05 compared with corresponding control (without GDF-15) determined by one-way ANOVA. Results are shown as means ± SEM. *p < 0.05 compared with corresponding control (without GDF-15) determined by unpaired Student's t-test.
Overall, these data indicated that GDF-15 increased T-type VGCC activity via the TβRII/ERK pathway.

GDF-15 increased T-type VGCC activity through promoting Ca V 3.1 and Ca V 3.3 surface expression.
The effect of GDF-15 on mEPSCs was short-term, and 1 h was not long enough to influence protein transcription and translation. We therefore suspected that GDF-15 may promote the membrane trafficking of T-type VGCC proteins. Using specific antibodies, we confirmed that all three α-subunits of T-type VGCCs (Ca V 3.1, Ca V 3.2, and Ca V 3.3) were expressed on mPFC pyramidal neurons (Fig. 6A), but Ca V 3.1 was the most highly expressed and Ca V 3.2 the least expressed, consistent with a previous in situ hybridization study 22 . We then measured the effect of GDF-15 on the surface expression of T-type VGCCs using a membrane extraction kit. Western blotting showed that GDF-15 significantly increased the membrane expression of Ca V 3.1 and Ca V 3.3 by 23.63 ± 4.49% (n = 9, p < 0.0001) and 19.94 ± 6.18% (n = 9, p = 0.005), respectively, but had no significant effect on Ca V 3.2 (Fig. 6B). These results suggest that Ca V 3.1 and Ca V 3.3 membrane expression levels were up-regulated by GDF-15. Further, we investigated the role of the TβRII/ERK pathway in the GDF-15-induced up-regulation of Ca V 3.1 and Ca V 3.3 using the corresponding inhibitors. Co-incubation of PFC slices with GDF-15 and LY2109761 (10 μM) reduced the up-regulation of Ca V 3.1 and Ca V 3.3 membrane expression to 3.43% (n = 7, p = 0.785) and 6.16% (n = 7, p = 0.555), respectively, which were significantly different from the results with GDF-15 alone (Fig. 6C). As expected, SB431542 had no influence on the effects of GDF-15 on membrane expression. GDF-15 still up-regulated Ca V 3.1 Brefeldin A is a lactone antibiotic produced by fungi, which can indirectly inhibit protein transport from the endoplasmic reticulum to the Golgi apparatus by preventing formation of coat protein I-mediated transport vesicles 23,24 . We used brefeldin A to determine if the effects of GDF-15 on I T-type VGCC and the surface expression of Ca V 3.1 and Ca V 3.3, as well as the subsequent increases in glutamate release and mEPSC frequency, were caused by Ca V 3.1 and Ca V 3.3 protein trafficking. Incubation with brefeldin A alone for 1 h had no effect on the surface expression of Ca V 3.1 and Ca V 3.3, suggesting no effect on T-type VGCCs already present in the membrane (Fig. 7A). However, co-exposure of brain slices to GDF-15 and brefeldin A (10 μM) inhibited the increase in membrane expression of T-type VGCC protein; membrane expression levels of Ca V 3.1 were increased by −3.65% (n = 8, p = 0.76) and levels of Ca V 3.3 by −7.99%, compared with controls (n = 6, p = 0.521) (Fig. 7A). Similarly, GDF-15 failed to increase I T-type VGCC in the presence of brefeldin A, and the I T-type VGCC was increase by 9.87% compared with brefeldin A alone (n = 26, p = 0.394) (Fig. 7B). HPLC analysis indicated that brefeldin A inhibited the GDF-15-induced increase in glutamate release (from 147.05% with GDF-15 alone to 29.5% with GDF-15 plus brefeldin A, n = 6, p = 0.314) (Fig. 7C). The frequency of mEPSCs in the presence of GDF-15 and brefeldin A was 1.31 ± 0.14 Hz (n = 22), which was similar to that in the presence of brefeldin A alone (1.39 ± 0.11, n = 26, p = 0.623) (Fig. 7D).
Overall, these results indicated that GDF-15 increased glutamate release through promoting T-type VGCC trafficking to the membrane.

Discussion
GDF-15 is known to play pivotal roles in neuroprotection, neural regeneration, and axonal elongation 7,12,25 ; however, little is known about its precise function in neuronal excitability, its mechanism of action, and its downstream effectors. We previously showed that incubation of CGNs with GDF-15 activated TβRII and phosphoinositide 3-kinase/Akt/mammalian target of rapamycin (mTOR) signaling to increase I K amplitude and K V 2.1 expression, with possible developmental significance 13 . In the current study, incubation of cortical neurons with GDF-15 for 60 min up-regulated expression levels of the Ca V 3.1 and Ca V 3.3 subunits of T-type VGCC on the membrane, thereby increasing glutamate release and the frequency of mEPSCs, involving activation of the same receptor and downstream signaling components as those previously reported in CGNs 13 . TGF-β superfamily ligands mediate their effects via the transmembrane TβRI and TβRII receptor heterodimer 14,26 . There is good evidence to indicate that activation of TGF-β receptors not only activates the Smad signaling pathway, but also non-Smad pathways, such as p38, JNK/MAPK, mTOR and Ras-ERK [26][27][28] . We previously found that the Akt/mTOR and MAPK/ERK pathways were activated in CGNs by GDF-15 treatment, though activation of ERK signaling was not required for GDF-15-induced transcriptional regulation of K V 2.1 expression 13 . Use of pharmacological inhibitors demonstrated that activation of TβRII was associated with the effect of GDF-15 on cortical neurons; similar to the results in CGNs, but activation of ERK signaling was also required for GDF-15-induced up-regulation of Ca V 3.1 and Ca V 3.3 expression. This apparent difference is likely because GDF-15 up-regulated K V 2.1 expression in CGNs at the Akt/mTOR-mediated transcriptional level, but induced increases in Ca V 3.1 and Ca V 3.3 expression at MAPK/ERK-associated posttranslational levels, similar to the Hu et al. report, in which MAPK/ERK regulated the I A channel by direct phosphorylation of K V 4.2 subunits 29 . We also noted that there is a significant increase in pERK2 level in LY2109761 treated slice, suggesting that blocking TβRI/TβRII with LY2109761 for 1 h is capable of activating ERK2 phosphorylation via an unknown pathway in mPFC. It is likely because TGF-β receptors mediate both of Smad and non-Smad signal pathways, the crosstalk with other downstream signaling may activate ERK2. However, this effect did not affect LY2109761's role on inhibiting GDF-15-induced the increase of pERK1/2 through blocking TβRI/TβRII.
VGCCs are voltage sensors that convert membrane depolarization into intracellular Ca 2+ signals. In neurons, VGCCs include L-, N-, P/Q-, R-, and T-type Ca 2+ channels 30,31 . T-type VGCCs are transient, low-voltage activated Ca 2+ channels that control Ca 2+ entry during depolarization near resting potential 32 . Increasing evidence suggests that T-type VGCCs are loosely coupled to neurotransmission near the resting potential and sustain neurotransmitter release during mild stimulation 18 . In our study, the T-type VGCC blockers NiCl 2 , mibefradil and TTA-P2 eliminated the increases in mEPSC frequency and glutamate release induced by GDF-15, suggesting the involvement of T-type VGCCs. However, NiCl 2 , mibefradil and TTA-P2 alone did not affect the frequency of mEPSCs in cortical neurons under control conditions. It is possible that a large percentage of the T-type calcium channels are tonically inactivated at normal neural resting membrane potentials [33][34][35][36][37] , and only a small proportion of channels remains tonically activated at membrane potentials within the window current 38 . We therefore hypothesized that mibefradil and NiCl 2 only inhibited the fraction of T-type VGCCs that were increased by GDF-15. This phenomenon is consistent with the situation in which the contribution of T-type channels is indirect, and requires either the activation of coupled presynaptic receptors or depolarization of the membrane potential by blocking the ion channels 39,40 . A previous study found that mEPSCs in striatopallidal medium spiny neurons were mediated by the Ca V 1.3α1 subunit of L-type VGCCs 41 . This difference with the present study may be attributable to different neuron types, different developmental states, and/or the different animals used. In addition, we have noticed that recent study from the juvenile mice calyx of Held synapse and neocortical neurons indicated that spontaneous glutamate release can be triggered indirectly by the Ca 2+ entry through VGCCs and be mediated via a different Ca 2+ -sensing mechanism 42,43 , whether GDF-15-induced the increase of glutamate release is associated with those Ca 2+ -sensing mechanism is worthy of further study.
Three genes, CACNA1G, CACNA1H, and CACNA1I, have been identified as coding for the T-type VGCC subunits Ca V 3.1/α1G, Ca V 3.2/α1H, and Ca V 3.3/α1I, respectively 34 . The biophysical properties, structure-function relationships, and divergent physiological roles of the three Ca V 3 channels of T-type VGCCs have been documented 33,44,45 . Moreover, previous research has focused more on the structure and function of Ca V 3.2 rather than Ca V 3.1, Ca V 3.3 46,47 . However, although all three Ca V 3 subunits were detected in mPFC neurons by immunofluorescence, Ca V 3.2 was the least expressed, consistent with a previous in situ hybridization study 22 . More than this, GDF-15 mainly increased the expression of Ca V 3.1 and Ca V 3.3, rather than Ca V 3.2. T-type VGCCs are known to lack an α-interaction domain and are thus unable to interact with the β-subunit, which usually controls trafficking of the other VGCCs to the plasma membrane 48 . The intracellular loop connecting repeats I and II (I-II loop) of T-type VGCCs is thus an important regulator for trafficking, with distinct effects on the three channel types 44,49,50 . In addition to the lower expression of Ca V 3.2 compared with Ca V 3.1 and Ca V 3.3 in the mPFC, the selective up-regulation of Ca V 3.1 and Ca V 3.3 by GDF-15 may be associated with differences in structural properties and trafficking mechanisms among the different Ca V 3 subunits. Further studies are needed to clarify the precise mechanisms.
As membrane proteins, the expression of functional ion channel subunits can be modulated at multiple levels, including transcription, translation and trafficking. Long-term up-regulation of protein expression is mainly associated with transcription and translation 21,51 , while short-term modulation of ion-channel densities may be the result of rapid mechanisms involving changes in intracellular trafficking of channel proteins 52 . In this study, treatment with GDF-15 for 1 h was enough to enhance the surface expression of Ca V 3.1 and Ca V 3.3, suggesting the involvement of a short-term modulatory mechanism. This speculation that GDF-15 up-regulate Ca V 3.1 or Ca V 3.3 surface expression by activation of ERK-mediated trafficking was supported by the effect of the protein-transport inhibitor, brefeldin A. Although ERK-mediated protein trafficking is known to play a role in the regulation of T-type VGCC expression 24,53,54 , the opposite regulatory effect has also been found in L-type VGCC protein 55 . In addition, a recent study identified the actin-binding protein Kelch-like 1 as a regulator of T-type VGCC protein, responsible for enhanced cell surface expression 56 . However, the mechanisms whereby ERK up-regulates Ca V 3.1 or Ca V 3.3 trafficking, and the role of Kelch-like 1 in the GDF-15-mediated effect on increased cell surface expression of Ca V 3.1 or Ca V 3.3 remain to be elucidated.
Both electrophysiological and behavioral studies have suggested that the mPFC may be involved in recognition memory 57 . mPFC neurons have been shown to carry information concerning the relative familiarity of individual stimuli 58,59 . In our study, short-term application of GDF-15 significantly enhanced neurotransmitter release and mEPSCs in mPFC neurons, indicating a previously unreported role for GDF-15 in rapidly regulating neuronal excitability. It has been noted that GDF-15 mRNA and protein levels were dramatically up-regulated at the sites of cryolesions or ischemic lesions 6,12 , suggesting that GDF-15 may not only promote survival and protect neurons against lesions, but may also affect neuronal excitability or synaptic activity in the lesioned region, thus affecting neural-network excitability, and ultimately processes such as learning and memory. Fuchs et al. found that GDF-15 levels were associated with cognitive performance and age-related cognitive decline in humans, indicating a negative role for GDF-15 in recognition memory 60 . Further animal behavioral tests after overexpression of GDF-15 in different brain areas are needed to determine the physiological and pathophysiological effects of GDF-15 on recognition memory.
In conclusion, the results of this study demonstrated that GDF-15 increased release of the neurotransmitter glutamate in mPFC pyramidal neurons via posttranscriptional regulation of Ca V 3.1 and Ca V 3.3 trafficking. Furthermore, the same signaling pathways and receptors identified in CGNs were activated by GDF-15 in pyramidal neurons. T-type currents occur in neurons throughout the brain, with particularly large currents in the thalamic, septal, and sensory neurons. Their preferential localization in dendrites suggests that T-type channels play an important role in synaptic integration 34 . This study thus provides an important insight into the mechanisms underlying the functions of GDF-15 in the brain.

Materials and Methods
Experimental animals. Female C57BL/6 mice 3-4 weeks old (15-20 g) were purchased from Slac Laboratory Animals (Shanghai, China) and housed under a 12-h light/dark cycle, with food and water available ad libitum. All the experiments were performed in accordance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals. The protocol was approved by the Committee on the Ethics of Animal Experiments of Fudan University (permit number: 20090614-001). Efforts were made to minimize the number of animals used and their suffering.
Slice preparation and whole-cell recording. Mice were anesthetized with sodium pentobarbital and then decapitated. The brain was quickly removed into cold, pre-oxygenated cutting solution containing 220 mM sucrose, 3 mM KCl, 5 mM MgCl 2 , 1 mM CaCl 2 , 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , and 10 mM glucose. The brain was glued to the stage and submerged in the cold cutting solution, and 200-μm slices containing the mPFC were cut coronally from the PFC using a vibratome slicer (DOSAKA, Kyoto, Japan). About three slices per hemisphere were placed into ACSF at 34 °C to recover for at least 1 h, and bubbled with 95% O 2 -5% CO 2 , before recording. ASCF contained 125 mM NaCl, 2.5 mM KCl, 1.5 mM MgSO 4 , 2.5 mM CaCl 2 , 1 mM NaH 2 PO 4 , 26 mM NaHCO 3 , and 10 mM glucose.
Whole-cell recordings were made according to standard procedures, at room temperature 61 . PFC slices were transferred to a recording chamber and continuously perfused with oxygenated ACSF. mEPSCs were recorded from pyramidal cells in layers II/III using an Axon 700B amplifier (Molecular Devices, Union City, California, USA) under visual control, using differential interference contrast and infrared optics via a water-immersion objective (Olympus, Tokyo, Japan) and a CCD camera (Qimaing, Surrey, Canada). The recording pipettes solution contained 125 mM CsCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 10 mM HEPES, 10 mM EGTA, 3 mM Mg-ATP, and 0.3 mM Tris-GTP. Series resistance (Rs) was monitored during recording. Cells in which the Rs varied by >20% and the Rs was >60 MΩ were excluded from subsequent analyses. mEPSCs and I T-type VGCC were collected using pClamp10.2 software (Molecular Devices) and analyzed using the Mini-analysis program (Synpatosoft, Decatur, Georgia, USA) and Clampfit 9.2 (Molecular Devices), respectively.
High-performance liquid chromatography. Brain slices were incubated with GDF-15 or inhibitors in ACSF bubbled with 95% O 2 -5% CO 2 for 1 h. The ACSF was then collected and frozen at −80 °C for later analyses. The slices were homogenized in ice lysis buffer containing protease inhibitor (Sigma, St Louis, Missouri, USA), rotated on ice for 40 min, and centrifuged at 13,800 × g for 20 min at 4 °C. The supernatants were collected to measure the protein concentrations. The samples were subjected to reversed-phase HPLC (ThermoFisher, Waltham, Massachusetts, USA) with fluorometric detection following pre-column derivatization with o-phthalaldehyde to analyze glutamate concentrations, as described previously 62 . Chromatography was performed on a reversed-phase C-18 column using a pH sodium acetate methanol gradient. Methionine sulfone was added to each sample as an internal standard. External standards containing 40, 400, or 4000 pmol/20 ml glutamate were run at the beginning and end of every group. The peak heights of glutamate were initially normalized to the methionine sulfone peak and then quantified according to the linear relationship between peak height and the amounts of the corresponding standards. Glutamate release was normalized by the total protein in each single brain slice, and expressed as ng/mg protein 63 . Immunohistochemistry. Mice were anesthetized with sodium pentobarbital and perfused transcardially with normal saline followed by 4% paraformaldehyde. Brains were post-fixed in the 4% paraformaldehyde overnight at 4 °C. Coronal brain sections (40 μm) were cut using a vibratome slicer (Leica, Wetzlar, Germany) and processed for immunofluorescence.
Sections were transferred into 0.5 ml of blocking solution (5% bovine serum albumin; 0.5% Triton X-100, and 0.05% sodium azide in PBS) in a multi-well plate, placed on a shaker and shaken gently at room temperature for 1.5-2 h. The blocking solution was then replaced with the following antibody solutions for 2 days at 4 °C: mouse anti-Ca V 3.1 (1:50, NeuroMab, Davis, California, USA), mouse anti-Ca V 3.2 (1:50, NeuroMab), and rabbit anti-Ca V 3.3 (1:100, Santa Cruz, Dallas, Texas, USA) in 1% bovine serum albumin, 0.5% Triton X-100, and 0.05% sodium azide in PBS. After 2 days, the antibody solution was removed and the sections were washed with PBST (0.1% Triton X-100 in PBS) three times for 10 min, with a final wash for 4-5 h. The sections were incubated with secondary antibody solution (FITC-labeled goat anti-mouse IgG, Cy3-labeled goat anti-rabbit IgG, 1:500) overnight at 4 °C. The antibody solution was replaced with PBST and the sections were washed three times for 10 min each in PBS. DAPI was added to the slices to stain the nucleus. All sections were covered with coverslips using an anti-fade mounting medium, and then observed under a Leica SP2 confocal laser scanning microscope. Immunoreactivity was examined at optimal resolution. Confocal photomicrographs were further processed to adjust scaling, brightness, and contrast.
Western blotting. Mice were anesthetized with sodium pentobarbital and decapitated. The brain was removed, sectioned, and incubated as described previously for total protein extraction. After incubation, the slices were homogenized in ice lysis buffer containing protease inhibitor (Sigma), rotated on ice for 40 min, and centrifuged at 13,800 × g for 20 min at 4 °C. The supernatants were frozen at −80 °C for later western blotting. For membrane-protein extraction, the slices were lysed using Membrane and Cytosol Protein Extraction Kit (Beyotime, Shanghai, China). Before analysis, the supernatant was diluted in sample buffer containing β-mercaptoethanol. Equal amounts of protein were loaded and separated by electrophoresis in 8% SDS-PAGE gels, and transferred onto PVDF membranes (Merck Millipore/Merck KGaA, Darmstadt, Germany). The membranes with proteins were blocked by 10% defatted milk at room temperature for 1 h and then incubated overnight at 4 °C with rabbit antibodies to phosphorylated ERK (pERK) (1:1000, Cell Signaling Technology, Danvers, Massachusetts, USA), rabbit anti-ERK (1:1000, Cell Signaling Technology), mouse anti-Ca V 3.1 (1:200, NeuroMab), mouse anti-Ca V 3.2 (1:200, NeuroMab), or rabbit anti-Ca V 3.3 (1:500, Santa Cruz) primary antibody. After three washes for 10 min each, the protein blots were incubated with secondary goat anti-rabbit IgG conjugated with horseradish peroxidase (1:1000, Kang Cheng, Shanghai, China) for 2 h at room temperature. Signals were finally visualized using enhanced chemiluminescence (ThermoFisher), and the blots were exposed in a gel-imaging analyzer (Bio-Rad, Hercules, California, USA). pERK1/pERK2 was normalized against total ERK1/ ERK2 and expressed as fold-increase compared with control. Na + /K + -ATPase (1:1000, Cell Signaling Technology) was used as an internal control for membrane proteins, to ensure that the total protein levels were equal. Data analysis. Data are expressed as mean ± SEM. Differences among multiple groups were analyzed using one-way ANOVA, and differences between two groups using Student's t-tests. A p value <0.05 indicated statistical significance.