The JAK Inhibitor AG490 Specifically Blocks the Induction of NMDAR-LTD
We first investigated the role of JAK in synaptic transmission and synaptic plasticity in the Schaffer collateral-commissural pathway (). For the initial set of experiments we used extracellular recording in acutely prepared rat hippocampal slices and stimulated two independent inputs onto the same population of neurons. We decided to test the effects of the JAK inhibitor AG490 (10 μM), since this inhibitor has been shown to interfere with learning and memory (Chiba et al., 2009b
). We found that AG490 had no effect on baseline transmission (100% ± 1% before and 101% ± 1% during AG490 application, n = 13). Next we tested the effects of AG490 on NMDAR-LTP, since this is the most widely studied cellular correlate of learning and memory (Bliss and Collingridge, 1993
). However, we found no difference between the level of LTP induced in the control input, in which AG490 was applied immediately after the tetanus, or in the input tetanized in the presence of AG490 (A). Thus, the level of LTP obtained 30 min following the tetanus, expressed as a percentage of baseline, was 135% ± 4% and 145% ± 3% (n = 4), respectively. These values were similar to the level of LTP induced in untreated inputs (140% ± 3% of baseline, n = 6; C).
AG490 Blocks the Induction of NMDAR-LTD but Not LTP or Depotentiation
Since more recent evidence has suggested that NMDAR-LTD is also involved in some forms of learning and memory (see Collingridge et al., 2010
) we next tested AG490 on this form of synaptic plasticity. In all experiments, AG490 completely prevented the induction of NMDAR-LTD induced by low-frequency stimulation (LFS; comprising of 900 stimuli delivered at 1 Hz), though usually a short-term depression remained (B). In all cases, the block of NMDAR-LTD was fully reversible since a second, identical period of LFS induced LTD that was similar to that observed under control conditions. Thus, 60 min following the first LFS, delivered in presence of AG490, the responses were 99% ± 4% of baseline and 60 min following the second LFS, delivered after washout of AG490, they were 74% ± 11% of baseline (n = 6). In contrast to the dramatic effect on the induction of NMDAR-LTD, AG490 had no effect on the expression phase of this process. Thus, LFS induced an LTD that was 71% ± 9% and 72% ± 9% of baseline (n = 6), before and following the application of AG490, respectively. Since these experiments were all performed using two inputs, the ability of AG490 to selectively and reversibly block the induction of NMDAR-LTD without affecting baseline transmission or the expression of NMDAR-LTD were all internally controlled.
Next, we explored whether the effects of AG490 were specific for de novo NMDAR-LTD or whether it blocked all forms of LTD. To do this we investigated depotentiation, the reversal of a previously potentiated input. For these experiments we compared, in the two inputs, the level of depotentiation before the application and in the presence of AG490. Under both sets of conditions, LFS reversed LTP to baseline conditions (C). For example, in the test input the synaptic response was 137% ± 3% and 93% ± 3% of baseline (n = 6), before and following the LFS.
In conclusion, these experiments have shown that the JAK inhibitor AG490 has a highly specific effect on the induction of NMDAR-LTD.
Evidence for a Role of Postsynaptic JAK in NMDAR-LTD
To establish the locus of action of AG490 we made whole-cell recordings and added the compound to the filling solution (). In all neurons loaded with AG490 (10 μM) it was not possible to induce NMDAR-LTD using a pairing protocol (300 pulses, 0.66 Hz, at −40 mV; A), whereas in interleaved control experiments NMDAR-LTD was readily induced (G). Thus, the responses were 99% ± 2% (n = 6) and 63% ± 4% (n = 7) of baseline, measured at least 30 min after pairing, respectively. These experiments demonstrate that the likely locus of AG490 inhibition is within the postsynaptic neuron. However they do not establish beyond reasonable doubt that the target is JAK since all kinase inhibitors have off-target effects (Bain et al., 2003
), due largely to the huge diversity of protein kinases expressed in neurons. The best way to establish the target is to apply a panel of different inhibitors, on the realistic assumption that the off-target effects of the structurally distinct compounds will vary (Peineau et al., 2009
). We therefore used three additional JAK inhibitors (CP690550 [1 μM], JAK inhibitor I [0.1 μM], and WP1066 [10 μM]). We also included two src inhibitors (PP2 [10 or 20 μM] or SU6656 [10 μM]) in the study, given that src family PTKs are expressed postsynaptically and regulate neuronal function (Lu et al., 1998; Yu et al., 1997
), including insulin-induced LTD (Ahmadian et al., 2004
). Similar to the effects of AG490, we found that the other three JAK inhibitors all fully blocked the induction of NMDAR-LTD (101% ± 2% of baseline, n = 5, B; 99% ± 2% of baseline, n = 6, C; and 99% ± 2% of baseline, n = 4, D; respectively). In contrast, neither PP2 nor SU6656 affected the induction of NMDAR-LTD (64% ± 3% of baseline, n = 7, E; and 64% ± 3% of baseline, n = 11, F; respectively). Apart from blocking the induction of NMDAR-LTD none of the inhibitors affected baseline transmission or other measured properties. The results are summarized in G and collectively demonstrate that JAK is required for the induction of NMDAR-LTD.
JAK2 Knockdown Blocks the Induction of NMDAR-LTD
The available JAK inhibitors do not effectively distinguish between the JAK isoforms. Of the four JAK isoforms present in the body (JAK1, JAK2, JAK3, and TYK2), JAK2 is the most highly expressed in the brain and has been found in the postsynaptic density (PSD) fraction (De-Fraja et al., 1998; Murata et al., 2000
). Therefore, to investigate the role of JAK2 in NMDAR-LTD directly, we used constructs coding for two different shRNAs against rat JAK2 or a control shRNA, plus GFP as a transfection marker. The JAK2 shRNAs could effectively knockdown JAK2, as assessed biochemically (shRNA-1: to 32% ± 8% and shRNA-2 to 13% ± 5% of control, n = 5; A) and with immunocytochemistry (B). We then transfected hippocampal organotypic slices with these constructs to assess their effects on basal synaptic transmission, by comparing AMPAR and NMDAR-mediated EPSCs between pairs of transfected and neighboring untransfected neurons, 48–72 hr after transfection. There was no difference under any condition, showing that knocking down JAK2 has no effect on basal synaptic transmission (C–3E). In the next set of experiments we investigated the effects of these constructs on NMDAR-LTD. In all cells examined, NMDAR-LTD was absent in neurons transfected with the JAK2 shRNA constructs (shRNA-1: 88% ± 9% of baseline, n = 7, F; shRNA-2: 94% ± 15%, n = 6, G). In contrast, NMDAR-LTD was observed in all neurons transfected with the control shRNA (51% ± 5% of baseline, n = 8; H), and this was similar to that observed in non-transfected cells (Amici et al., 2009
). These experiments further substantiate the pharmacological results identifying a role of JAK in NMDAR-LTD and support the idea that the JAK2 isoform is critically involved in this process.
JAK2 Knockdown Blocks the Induction of NMDAR-LTD
JAK2 Is Present in Dendritic Spines and Is Regulated during NMDAR-LTD
We investigated the distribution of JAK2 in cultured hippocampal neurons using confocal microscopy (A). JAK2 showed a highly punctate distribution that decorated dendrites, labeled with microtubule-associated protein 2 (MAP2, Aa–4Ac″). A high proportion of JAK2 immunostaining was colocalized with PSD-95 (45% ± 3% of PSD-95 positive puncta colocalized with JAK2; 54% ± 3% of JAK2 positive puncta colocalized with PSD-95, Ad–4Ae″). We also confirmed, using differential centrifugation, that JAK2 is expressed in the synaptosomal (LP1) fraction (B).
JAK2 Is Present and Regulated at Synapses
If JAK2 is indeed the isoform involved in NMDAR-LTD then it would be expected that its activity would be regulated during the induction of the process. We therefore measured the level of phosphorylation of Tyr 1007/1008, as an indicator of its activity (Feng et al., 1997
). In the initial experiments we applied NMDA (20 μM, 3 min), a treatment that induces a chemical form of NMDAR-LTD (Lee et al., 1998
). We found that at the three initial time points measured (0, 5, and 30 min after NMDA treatment) the activity of JAK2 in hippocampal slices was significantly increased (145% ± 10%, n = 10; 167% ± 13%, n = 18; 150% ± 18% compared to control, n = 7, respectively; C and 4D). However, the activation was transient since there was no significant difference in the level of phosphorylation measured 60 or 120 min later. The activation of JAK was dependent on the presence of Ca2+
and was specific for NMDARs, since neither an mGluR agonist (DHPG) nor a muscarinic agonist (carbachol) affected JAK2 phosphorylation (C and 4D). Consistent with the lack of effect of DHPG on JAK2 phosphorylation, AG490 had no effect on DHPG-LTD (E), a form of LTD induced by the activation of mGluRs (Palmer et al., 1997
Since there are some differences in the mechanism of LTD induced by bath perfusion of NMDA compared with the LTD induced by synaptic activation of NMDARs (Morishita et al., 2001
), we also measured JAK2 phosphorylation in CA1 dendrites following LFS (F and 4G). Electrical stimulation also resulted in an increase in JAK2 activity (158% ± 16% compared to nonstimulated slices, n = 24; G) and this required the synaptic activation of NMDARs since the increase in phosphorylation was absent in slices treated with AP5 during the LFS (116% ± 14%, n = 10; F and 4G). Treatment with inhibitors for the Ser/Thr protein phosphatases PP1 and PP2B also prevented activation of JAK2 during LFS (okadaic acid [1 μM]: 103% ± 17%, n = 9; cyclosporine A [50–250 μM]: 112% ± 27%, n = 5; F and 4G).
In summary, the finding that JAK2 is enriched at synapses, colocalizes with PSD-95 and is activated during LTD in an NMDAR, Ca2+ and PP1/PP2B dependent manner, suggests that this isoform is involved in NMDAR-LTD.
STAT3 Is Required for the Induction of NMDAR-LTD
The next question we wished to address is what the downstream effector of JAK2 is in NMDAR-LTD. JAK2 is known to signal via the PI3K/Akt pathway and the ras/MAPK pathway (Lanning and Carter-Su, 2006; Zhu et al., 2001
). However, inhibitors of these pathways do not affect NMDAR-LTD, under our experimental conditions (Peineau et al., 2009
). Another possibility is via STATs. The JAK/STAT pathway is a major signaling pathway involved in many nonneuronal processes where JAK activation leads to phosphorylation of STATs, which results in their activation and translocation to the nucleus. We focused on STAT3, since this isoform is present in the hippocampus and PSD (Cattaneo et al., 1999; De-Fraja et al., 1998; Murata et al., 2000
). Therefore, we tested the effects of two compounds that inhibit the activation of STAT3: Stattic (50 μM) and STA-21 (30 μM). We found that both STAT3 inhibitors prevented the induction of NMDAR-LTD (99% ± 2% of baseline, n = 4, A; and 96% ± 4% of baseline, n = 7, B; respectively), with a similar time-course as the JAK inhibitors.
STAT3 Is Required for the Induction of NMDAR-LTD
These data are consistent with a scheme in which, during NMDAR-LTD, activation of JAK2 leads to activation of STAT3. In which case, inhibition of STAT3 would not be expected to affect the activation of JAK2 (Beales and Ogunwobi, 2009; Schust et al., 2006
). To test whether this was indeed the case, we treated cultured hippocampal neurons with Stattic and found that this completely prevented the activation of STAT3 without affecting the activation of JAK2 in response to the stimulation of NMDARs (C). This treatment also reduced basal levels of STAT3 activity suggesting that there is a degree of constitutive activation of STAT3.
To substantiate the involvement of STAT3 in NMDAR-LTD, we used two different shRNAs against STAT3, which efficiently knocked down the target protein in hippocampal cultured neurons as assessed with immunocytochemistry (D). As with JAK2 shRNAs, the knockdown of STAT3 had no effect on basal synaptic transmission, as assessed by comparing AMPAR and NMDAR-mediated EPSCs between pairs of transfected and neighboring untransfected neurons, 48–72 hr after transfection in organotypic slices (E and 5F). However, no NMDAR-LTD could be observed in the cells transfected with the shRNAs (G and 5H), whereas NMDAR-LTD was reliably induced in interleaved experiments in neurons transfected with control shRNA (H). With both shRNAs against STAT3 there was a small decrease in the synaptic response following the LTD stimulus protocol but this was similar for both the test and control inputs, and significantly smaller than for control LTD. When all these data are considered together it strongly suggests that STAT3 is the isoform involved in NMDAR-LTD.
STAT3 Is Activated and Translocated to the Nucleus during NMDAR-LTD
Since, when activated, STAT3 translocates to the nucleus, we wanted to see if this activation and translocation also occurs during NMDAR-LTD. In cultured hippocampal neurons under control conditions, STAT3 immunoreactivity was fairly evenly distributed throughout the neuron, including the nucleus (A). NMDA treatment (20 μM, 10 min) resulted in nuclear translocation and activation of STAT3 (A). Maximal nuclear accumulation was observed immediately following NMDAR stimulation and the effect persisted for between 1 and 2 hr (A and 6B). There was a corresponding activation of nuclear STAT3, as assessed by the phosphorylation of Tyr 705 (P-STAT3), which also lasted for between 1 and 2 hr (A and 6B). Consistent with the activation of STAT3 being mediated by JAK2, treatment of cultures with AG490 prevented both the translocation of STAT3 and activation of nuclear STAT3 (C).
STAT3 Is Activated and Translocated to the Nucleus during NMDAR-LTD
To investigate whether STAT3 is also activated by LFS in hippocampal slices, we analyzed the levels of STAT3 and P-STAT3 in the CA1 region of hippocampal slices by western blotting. For these experiments, we microdissected both stratum radiatum, which is enriched in CA1 dendrites, and stratum pyramidale, which is correspondingly enriched in CA1 cell soma (D). We prepared a nuclear fraction from the microdissected cell soma preparation and examined the expression of P-STAT3 relative to total STAT3. LFS resulted in a pronounced activation of nuclear STAT3 (199% ± 23%, n = 14, D and 6F), which was absent if LFS was delivered in the presence of AP5 (94% ± 8%, n = 10), okadaic acid (87% ± 17%, n = 5) or cyclosporine A (136% ± 46%, n = 5; E and 6F). Interestingly, LFS also resulted in activation of dendritic STAT3 (135% ± 10%, n = 14; D and 6F) and this effect was also dependent on the synaptic activation of NMDARs (110% ± 11% in presence of AP5, n = 10; E and 6F). These results are consistent with the immunocytochemistry (A and 6B) in cultured neurons and extend them by showing the dependence of nuclear STAT3 activation on the PP1/PP2B protein phosphatase cascade. Significantly, these results show that the synaptic activation of NMDARs can lead to the activation of STAT3 in dendrites.
The Translocation of STAT3 to the Nucleus Is Not Required for NMDAR-LTD
Once activated, cytoplasmic STATs are translocated to the nucleus where they bind to DNA specific sequences within the promoter region to control gene expression. There is evidence that rapid transcription may be involved in LTD (Kauderer and Kandel, 2000
). Therefore, to investigate whether the rapid effect of inhibition of STAT3 on NMDAR-LTD was due to interference with gene transcription we performed a variety of different experiments.
We first tested galiellalactone, a STAT3 inhibitor that blocks STAT3 binding to DNA without affecting STAT3 activation. In all neurons loaded with galiellalactone (50 μM) NMDAR-LTD was readily induced (58% ± 8% of baseline, n = 5; A). To further explore whether nuclear signaling is required for NMDAR-LTD, we used a nuclear export inhibitor (leptomycin B, 50 nM) and this also failed to inhibit NMDAR-LTD (57% ± 3% of baseline, n = 6; B). To investigate transcription more generally, we tested the effects of actinomycin D (25 μM). In field recordings we followed NMDAR-LTD for 3 hr after induction and observed no difference between the level of LTD in the control and test inputs (69% ± 3% and 71% ± 2% of baseline, n = 4, respectively; C). We also performed experiments in slices from which the cell body region of the slice had been completely removed. Once again, NMDAR-LTD that lasted at least 3 hr could be readily observed (76% ± 3% of baseline, n = 5, D). These data collectively suggest that NMDAR-LTD can be readily induced and expressed for at least 3 hr, without the need for gene transcription and that the effects of inhibition of STAT3 are independent of an action within the nucleus. As a final test of this, we blocked transcription using actinomycin D (25 μM) in the patch pipette and tested the effects of Stattic under these conditions. In all neurons tested, NMDAR-LTD was readily induced in the presence of actinomycin D (63% ± 3% of baseline, n = 5; E) but was fully blocked by the additional inclusion of Stattic (50 μM) in the patch solution (98% ± 2% of baseline, n = 5; E).
Translocation of STAT3 to the Nucleus Is Not Necessary for NMDAR-LTD
In summary, activation of STAT3, but not its binding to DNA, is required for the induction and early expression of NMDAR-LTD.