Reciprocal signals between microglia and neurons regulate α-synuclein secretion by exophagy through a neuronal cJUN-N-terminal kinase-signaling axis

We previously noted that p25α expression in differentiated PC12 nerve cells caused massive and protracted activation of JNK and its downstream target cJUN alongside a greatly increased emission of α-syn to the surroundings. We therefore analyzed JNK activation in more detail. To detect activated JNK, we performed western blotting of phosphorylated (p)-JNK and its downstream target cJUN in whole cell lysates from differentiated PC12 cells expressing various combinations of α-SNC (wt or A30P) and p25α over a 6-day culture period after transgene induction with doxycycline (Fig. 1a). When expressing α-SNC_wt or α-SNC_A30P alone, there was a small increase in levels of p-JNK and p-cJUN compared to control cells transduced with β-synuclein (β-SNC). In contrast, p25α caused a dramatic and persistent activation of JNK and downstream target cJUN regardless of co-expression of α-SNC.

JNK is known to be involved in regulated trafficking of vesicular elements along microtubuli tracks [33]. To investigate if activated JNK plays a role in secretion of α-SNC, we first tested the effect of SP600125, which is a potent and reversible inhibitor of JNK. Differentiated PC12 cells expressing α-SNC_A30P either alone or with p25α were incubated with 10 or 20 μM SP600125 for 48 h (Fig. 1b), and the conditioned medium was analyzed for secreted α-SNC by TCA precipitation and western blotting. Following exposure to 10 or 20 μM SP600125, p-JNK was decreased without affecting cell death (measured by caspase 3 activation, data not shown) in a dose-dependent manner in cells expressing α-SNC_A30P/p25α, while the response was biphasic in PC12 α-SNC_A30P cells. ER stress and induction of the unfolded protein response (UPR) is a common feature of protein malconformation disease. As JNK can be activated following ER stress [47], we therefore also examined the effect of salubrinal, which inhibits dephosphorylation of phosphorylated eukaryotic Initiation Factor 2α (eIF2α), and thereby prolongs the UPR to relieve in part the cytotoxicity of ER stress [48]. As shown in Fig. 1d–f, 10–20 μM salubrinal decreased levels of p-JNK as well as secretion of α-SNC from PC12 α-SNC/p25α expressing cells.

Our observations that salubrinal decreased p-JNK levels and α-SNC release prompted us to investigate whether induction of ER stress is sufficient to activate JNK in p25α-expressing nerve cells. ER stress mediates JNK activation through an inositol requiring ezyme-1α (IRE1α)—TNFα receptor-associated factor 2 (TRAF2)—apotosis signal-regulating kinase-1 (ASK1) signaling axis [49]. Incidentally, ASK1 is also a major activator of JNK signaling in response to oxidative stress [50]. We therefore first examined the effect of expression of dominant-negative ASK1-K709R in differentiated PC12 nerve cells. As shown in Fig. 2a, b, ASK1-K709R was efficiently transduced into PC12 cells; however, p-JNK levels were in fact moderately upregulated in these cells (Fig. 2c, d), excluding the IRE1α-TRAF2-ASK1 axis in the activation of JNK. Secondly, we exposed ATRA-differentiated SH-SY5Y cells expressing α-SNC_A30P to the ER stressors tunicamycin and thapsigargin, which induced the UPR as evidenced by upregulation of C/EBP homologous protein (CHOP) (Fig. 2e). However, neither agent induced a significant change in the levels of p-JNK (Fig. 2e, f), and in fact, in BDNF-differentiated SH-SY5Y cells, these agents moderately decreased JNK activity (see Fig. 9b, e). In contrast, we note that incubation with bafilomycin A, which inhibits fusion of lysosomes with late endosomal or autophagosomal organelles, caused accumulation of the autophagosomal marker LC3B-II (Fig. 2e) and increased p-JNK levels more than twofold in α-SNC_A30P-expressing SH-SY5Y cells (p < 0.05). Bafilomycin A inhibits the vacuolar proton ATPase, and as a consequence, inhibits late endosome or amphisome fusion with lysomes, thereby mimicking the effect of p25α overexpression [10]. Therefore, the effect of bafilomycin A was expectedly reduced in SH-SY5Y cells co-expressing p25α (Fig. 2f). In conclusion, the observed activation of JNK in PC12 and SHSY-5Y nerve cells is not due to ER stress or oxidative stress but rather related to perturbation of membrane trafficking in the endosomal and/or autophagosomal pathways.

To identify the JNK isoform(s) in question, we established PC12 and SH-SY5Y cell lines conditionally expressing α-SNC_wt/p25α or α-SNC_A30P/p25α, respectively, together with stable expression of JNK1, JNK2, or JNK3 shRNA. These cells were then differentiated to mature neurons before lysate and conditioned medium were collected for analysis. We confirmed the knockdown of specific JNK isoforms by western blotting (Fig. 3a, b) and analyzed α-SNC secretion by TCA precipitation of the media and western blotting. We found that knockdown specifically of JNK2 in PC12 cells (Fig. 3a, d) and JNK3 in SH-SY5Y cells (Fig. 3b, e) significantly decreased α-SNC secretion compared to control cells expressing scrambled shRNA. Of note, PC12 cells do not express JNK3 [51]. In differentiated PC12 cells, JNK1 shRNA significantly increased α-SNC secretion, while JNK1 and JNK2 knockdown in the differentiated SH-SY5Y cells also increased secretion. To further substantiate that active JNK in neurons is associated with α-SNC secretion, we utilized dominant-positive JNK1, JNK2, and JNK3 fusion constructs (FLAG-MKK7B2-JNK1/2/3). The FLAG-tagged constructs encode JNK fused to a constitutively active upstream mitogen-activated protein kinase kinase 7 (MKK7), which persistently activates the JNK moiety [46]. The FLAG-MKK7B2-JNK1 (AFP) construct is a JNK kinase dead control. Constructs were transiently transfected into differentiated SH-SY5Y cells expressing α-SNC_A30P (Fig. 3c, f–i) to observe whether constitutive JNK activity could replicate the effect of p25α on α-SNC secretion. The kinase dead AFP-construct afforded no change in α-SNC release relative to control cells, active JNK1 moderately increased secretion, but notably JNK2 and JNK3 isoform transfection many-fold increased the release of α-SNC (p < 0.05) (Fig. 3f). Both transfected JNK (FLAG-MKK7B2-p-JNK, Fig. 3c, h) and endogenous JNK (endo p-JNK, Fig. 3c, g) phosphorylation were increased, and the highest amounts of endogenous p-JNK were found after JNK2 and JNK3 transfections (p < 0.05 compared to control). No significant changes in cell death were detected as assayed by lactate dehydrogenase (LDH) activity in the medium (data not shown). The ratio of endogenous p-JNK relative to FLAG-tagged-p-JNK (Fig. 3i) proved to be highest for the JNK3 fusion construct, suggesting that similar amounts of transfected JNK activity (FLAG-p-JNK) pertaining to different isoforms (JNK1-3) caused the highest amount of endogenous JNK activation (endo p-JNK) for the JNK3 isoform. Hence, JNK3 appears as the most efficient signaling entity for α-SNC secretion of the three investigated.

cJUN is an important transcription factor downstream of JNK activation, and as shown in Fig. 1a, the persistent JNK activation correlated with the activation of cJUN by phosphorylation. We therefore asked whether cJUN-induced transcription is important for the effects of p-JNK on α-SNC secretion. For this purpose, PC12 cells expressing α-SNC/p25α were transduced with a FLAG-tagged and truncated dominant-negative cJUN-∆169 construct, which retains DNA binding, but has lost its transactivation domain. Figure 4a, b shows that cJUN-∆169-FLAG was efficiently transduced into the vast majority of cells and expressed as a protein of the expected size (ca. 20 kDa), which localized to the nucleus. However, we found that dominant-negative cJUN did not significantly alter the level of secreted α-SNC to the medium (Fig. 4c, d), and we conclude that the effect of p-JNK on α-SNC release is not related to cJUN-induced transcriptional effects.

It has recently been shown that oligomeric α-SNC via binding to TLR2 or CR3 can activate microglia to produce inflammatory cytokines (e.g., TNFα and interleukin-1β) and mediate chemotaxis [40, 41]. The phagocyte NADPH oxidase (NOX2) is activated in response to many different stimuli, and the degree of either activation (superoxide production) or priming (hyper-responsiveness to stimuli) is a very sensitive measure of cell activation. When we incubated Ra2 microglia expressing the genetically encoded, fluorescent H₂O₂-sensor HYPER3 [45] with differentiated PC12 cells expressing either α-SNC_wt or α-SNC_wt/p25α in direct co-culture, we found that PC12 nerve cells expressing α-SNC, and even more so α-SNC/p25α, enhanced microglial NADPH oxidase activity compared to microglia monoculture. When microglia were primed with LPS (100 ng/ml), HYPER3 fluorescence was augmented further and again showed a p25α-dependent increase. Analysis of the conditioned media from these co-cultures showed that regardless of p25α expression, secretion of α-SNC was upregulated by LPS stimulation of the microglia (p < 0.05, Fig. 5a, b). The factor(s) emitted from neurons to activate microglia were soluble mediators as the effect could be replicated by passive transfer of PC12 conditioned medium to the Ra2 microglia (Fig. 5c‐e). Resting or LPS-stimulated Ra2 microglia expressing HYPER3 were incubated with conditioned medium from PC12 cells expressing α-SNC with or without p25α for 2 hours before basal or PMA-induced oxidant production was measured by ratiometric HYPER3 fluorescence in a microtiter plate reader. Again, we found that PC12 cells expressing α-SNC increased microglial oxidant production, and this was further increased by p25α co-expression either under basal or PMA-stimulated conditions (Fig. 5c‐e). Microglia cells incubated with LPS were particularly susceptible to activation by conditioned medium from α-SNC-expressing PC12 cells compared to control cells (p < 0.05) (Fig. 5e). Thus, substance(s) released from synucleinopathic differentiated neurons can activate microglia. It has been described that both CD11b [41] and TLR2 [40] function as signaling and/or phagocytic receptors of α-SNC. We therefore conducted co-culture experiments with or without LPS activation of microglia and then performed immunofluorescence to gauge internalization of α-SNC and potential interaction with these surface receptors. As shown in Fig. 5, the increased secretion of α-SNC afforded by LPS activation of microglia correlated with uptake into Ra2 microglia, where colocalization with TLR2 was evident (arrows in Fig. 5f).

As seen in Fig. 5a, b, LPS activation of microglia in co-culture greatly increased the secretion of α-SNC from neurons regardless of p25α expression. We further validated these results with primary rat microglia. Figure 6a, b shows that primary microglia activated by LPS caused a fourfold increased release of α-SNC from PC12 cells expressing α-SNC alone (p < 0.05 compared to control), again without excessive cell death as measured by LDH release (data not shown). The microglia redox state is intimately tied to microglia activation [52]. Therefore, we also analyzed the effect of overexpression of gp91^phox, the rate-limiting subunit of the cyt b₅₅₈ flavocytochrome complex of NADPH oxidase, in Ra2 microglia. As seen in Fig. 6c, gp91^phox overexpression in microglia caused an increased secretion of α-SNC from neurons in direct co-culture both in basal or LPS-primed states of the microglia. The effect of gp91^phox in microglia on neuronal secretion of α-SNC was not due to a direct effect of (short-lived) oxidants on the nerve cells, as the effect could be replicated by passive transfer of microglia conditioned medium to neuron monoculture (Fig. 6d). Secondly, we addressed if microglial signaling to increase neuronal α-SNC secretion was associated with neuronal JNK activation. To assess p-JNK levels accurately, these assays were performed in the absence of serum to avoid JNK activation associated with mitogenic signaling. Differentiated PC12 cells expressing α-SNC in monoculture were exposed to conditioned media from Ra2 microglia monocultures +/−LPS (100 ng/ml) exposure (Fig. 6e, f) with or without the addition of NGF. NGF has been shown to be important for maintaining PC12 cells in a differentiated state and to prevent apoptosis [53]. In agreement with previous experiments (Fig. 6a, b), we found that LPS alone had either no or little effect on neuronal JNK phosphorylation and α-SNC secretion. When differentiated PC12 cells were exposed to conditioned medium from control Ra2 microglia, there was no change in p-JNK levels but a decrease (although not statistically significant) in α-SNC secretion. However, conditioned medium transferred from inflammatory Ra2 microglia activated with LPS caused a statistically significant increase in p-JNK levels (p < 0.05 compared to PC12 cells with unexposed Ra2) associated with an increased α-SNC release, which in serum-free conditions is somewhat lower than in the presence of serum (compare Figs. 5b with 6f). The presence of NGF did not alter the response (Fig. 6f). Lastly, we performed confocal imaging with differentiated PC12 cells expressing α-SNC_wt and Ra2 microglia +/−LPS in co-culture to verify by independent methods that LPS-stimulated Ra2 microglia indeed caused an increase in p-JNK immunoreactivity in PC12 nerve cells (Fig. 6g). LDH release to medium was measured to estimate cell death (data not shown). In all cases, LPS exposure or microglia conditioned medium did not cause a biologically significant increase in cell death (the highest measure was a mean increase of Δ3.3 ± 0.9 % (p < 0.05) for cells exposed overnight to conditioned medium from LPS-primed microglia compared to control cells in monoculture).

The pro-inflammatory cytokine TNFα is an important effector of inflammatory microglia and is associated with disease progression in animal models of PD [39]. Therefore, we wanted to investigate if exogenously added TNFα could influence neuronal p-JNK levels and α-SNC secretion. Differentiated PC12 or SH-SY5Y cells were exposed to recombinant rat or human TNFα, respectively, in a concentration (Fig. 7a–c)- and time (Fig. 7d–f)-dependent manner. In differentiated PC12 and SH-SY5Y cells expressing α-SNC_wt or α-SNC_A30P, respectively, there was a TNFα dose-dependent increase in α-SNC secretion to the medium accompanied by an increase in p-JNK and its downstream target p-cJUN levels after 18 h exposure (Fig. 7a–c). In both cell types, exposure to TNFα for 18 h in the low ng/ml range (2–5 ng/ml) led to a significant increase in α-SNC secretion compared to control (0 ng/ml TNFα) (p < 0.05). Two concentrations of TNFα were chosen for 6 and 12 h exposures of differentiated PC12 cells expressing α-SNC_wt (10 and 25 ng/ml) and differentiated SH-SY5Y cells expressing α-SNC_A30P (2 and 5 ng/ml) (Fig. 7d–f). Already after 6-h exposure to the highest TNFα concentration, we detected a slightly increased α-SNC level in the conditioned media of both PC12 and SH-SY5Y cells. After 12 h of exposure to the highest concentration, there was a clear and statistically significant increase in α-SNC levels in the media of both differentiated PC12 and SH-SY5Y cells (p < 0.05) (Fig. 7e–f). The increased levels of secreted α-SNC were correlated with increased p-JNK levels in the whole cell lysates (Fig. 7d). Additionally, when comparing conditions with low and high TNFα concentrations, it appeared that within each time point was a dose-dependency as observed above. Collectively, the data indicate that direct exposure of neurons to TNFα suffices to increase JNK activation and α-SNC secretion.

To address the possibility of a causal relationship for JNK in the TNFα-mediated increase in α-SNC secretion from neurons, we knocked each of the three JNK isoforms down with shRNAs. Then differentiated PC12 and SH-SY5Y cells expressing JNK shRNAs together with α-SNC_wt or α-SNC_A30P, respectively, were subjected to TNFα exposure (Fig. 8a–d). For each cell line, we chose two TNFα concentrations and since JNK3 is not present in PC12 cells, we used only JNK1 or JNK2 shRNA in PC12 cells. In all cases, p-JNK levels and JNK knockdown were confirmed by p-JNK and JNK immunoblotting of whole cell lysates. Stable expression of JNK2 or JNK3 shRNA decreased α-SNC secretion from PC12 α-SNC_wt and SHSY5Y α-SNC_A30P cells (Fig. 8c, d, control) indicating that JNK is involved in α-synuclein metabolism also in the absence of p25α expression (see Fig. 3). Expectedly (Fig. 7), TNFα mediated an increased α-SNC secretion, which correlated with the activation of JNK (Fig. 8a–d, scrambled shRNA [Scr]). Importantly, the specific knockdown of JNK2 in PC12 cells or JNK3 in SHSY-5Y cells reduced the ability of TNFα to increase α-SNC secretion from the nerve cells indicating that JNK activation is an essential component of inflammatory signaling-induced α-SNC secretion. Cell death was assessed by LDH assay. Exposure to 10 and 25 ng/ml recombinant TNFα in the presence of JNK1 shRNA in PC12 cells (Fig. 8a, c) significantly increased cell death from 3.1 % ± 0.7 to 23.9 % ± 7.2 (p < 0.05) and 28.6 ± 11.7 (p < 0.05 compared to controls), respectively. Exposure of SH-SY5Y cells expressing JNK2 shRNA to 2 and 5 ng/ml recombinant TNFα increased (although not statistically significant) cell death from 8.2 % ± 3.5 to 28.9 % ± 7.2 and 28.6 % ± 11.7, respectively. The large increases in α-SNC levels in the conditioned medium from these conditions (only) are therefore not due to an active release mechanism.

Since exposure to the pro-inflammatory cytokine TNFα increased neuronal secretion of α-SNC, we also examined the effect of several non-inflammatory cytokines on neuronal α-SNC secretion. Differentiated PC12 cells expressing α-SNC_A30P in monoculture were incubated with interleukin (IL)34, transforming growth factor (TGF)β1, or interferon (IFN)β1 at a concentration of 10 ng/ml. TGFβ1 is considered a master negative regulator of inflammation; IFNβ1 is a classical anti-viral and anti-bacterial cytokine, while IL34, which binds to colony-stimulating factor-1 receptors present on PC12 cells, was recently shown to increase microglia-afforded neuroprotection in an animal model of Alzheimer’s disease [54]. In addition, we also added LPS previously shown to have no effect in monoculture (Fig. 6a–d). As seen in Fig. 8e, f, both TGFβ1 and IFNβ1 affected neuronal α-SNC secretion as seen by a statistically significant increase in α-SNC levels in the conditioned media from PC12 monocultures (p < 0.01 and p < 0.001, respectively, compared to control). Neither LPS nor IL34 affected secretion.

As a stress kinase and a mitogen-activated protein kinase, JNK is activated by many different factors and conditions in neurons. Therefore, we examined the universality of the correlation between activation of JNK and secretion of α-SNC into the surroundings. For the purpose, we imposed either ER stress with tunicamycin or thapsigargin exposure or alternatively oxidative stress generated by xanthine/xanthine oxidase production of superoxide, on PC12 and BDNF-differentiated SH-SY5Y cells expressing α-SNC_wt or α-SNC_A30P, respectively. Both stress conditions are highly relevant in the context of neurodegenerative disease and PD. As shown in Fig. 9a–c, ER stress or oxidative stress specifically upregulated CHOP as a marker of ER stress (p < 0.05 tunicamycin in PC12 and thapsigargin in SH-SY5Y compared to Cont.) and Nrf2 as a marker of oxidative stress (p < 0.05 compared to control), respectively, in either cell type. The expression levels of transcription factor ATF3, which is transcriptionally regulated by Nrf2, closely followed the expression levels of Nrf2 in both cell types (data not shown). While ER stress induced with thapsigargin caused JNK activation in PC12 cells (p < 0.05 compared to control), both tunicamycin and thapsigargin moderately downregulated p-JNK in SH-SY5Y cells (p < 0.05 compared to control, Fig. 9a, d). Note that JNK in SH-SY5Y cells is activated differentially following ER-stress induction with thapsigargin depending on either ATRA (Fig. 2e, f) or BDNF differentiation (Fig. 9a, d). In PC12 cells, both tunicamycin (not significant) and thapsigargin (p < 0.05 compared to control) increased α-SNC secretion to the medium, while preferentially thapsigargin caused α-SNC secretion in SH-SY5Y cells (p < 0.05). In neither cell type did oxidative stress under the experimental conditions cause activation of JNK, but nevertheless, α-SNC secretion was increased in both cell types particularly SHSY-5Y cells (p < 0.05, Fig. 9a, e). We therefore conclude that secretion of α-SNC under conditions of cellular stress (different from lysosomal perturbation and inflammatory signaling) does not necessarily correlate with JNK activation. Note that in the above experiment, released α-SNC may not necessarily derive from exophagy but could be due to other mechanisms of unconventional secretion. The essentiality of JNK2 and JNK3 for α-SNC secretion that we have described in the preceding sections therefore seems to be specifically tied to certain types of either intrinsic (autophagosomal and/or lysosomal perturbation) or extrinsic (inflammatory signaling) stress.