U.S.A. 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Protein kinase A (PKA) plays essential roles in diverse cellular functions. However, the spatiotemporal dynamics of endogenous PKA upon activation remain debated. The classical model predicts that PKA catalytic subunits dissociate from regulatory subunits in the presence of cAMP, whereas a second model proposes that catalytic subunits remain associated with regulatory subunits following physiological activation. Here we report that different PKA subtypes, as defined by the regulatory subunit, exhibit distinct subcellular localization at rest in CA1 neurons of cultured hippocampal slices. Nevertheless, when all tested PKA subtypes are activated via the β-adrenergic receptor by norepinephrine, catalytic subunits translocate to dendritic spines but regulatory subunits remain unmoved. These differential spatial dynamics between the subunits indicate that at least a significant fraction of PKA dissociates. Furthermore, PKA-dependent regulation of synaptic plasticity and transmission can be supported only by wildtype, dissociable PKA, but not by nondissociable PKA. These results argue for the classical model in which endogenous PKA regulatory and catalytic subunits dissociate to achieve PKA function. 31 32 33 34 35 36 37 38 39 40 41 42

Significance Protein kinase A (PKA) is the major effector of the second messenger cAMP and plays a critical role in numerous cell biology processes. The function and specificity of PKA rely on spatial compartmentalization. However, the precise subcellular distribution and dynamics of activated PKA remain debated. Textbooks illustrate that upon activation PKA catalytic subunits dissociate from regulatory subunits. However, some recent work has proposed that PKA catalytic subunits remain associated with regulatory subunits following physiological activation. The present study tests these two models in the context of neuronal function. The results support the classical model of PKA activation. A clarified model of PKA action will contribute to the understanding of how the specificity of PKA phosphorylation may be achieved. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

INTRODUCTION Cyclic adenosine monophosphate (cAMP)-dependent kinase, or protein kinase A (PKA), regulates diverse critical functions in nearly all mammalian cells. PKA is a tetrameric protein consisting of two regulatory subunits (PKA-Rs) and two catalytic subunits (PKA-Cs) (1, 2). In the inactive state, each PKA-R binds to and inhibits a PKA-C. Binding of cAMP to PKA-R activates PKA-C. However, there are different proposals on the molecular events that follow activation.

For decades, PKA-C is thought to dissociate from PKA-R upon cAMP binding. Freed PKA-C molecules then move to phosphorylate their substrates. However, two recent studies propose an alternative model, in which physiological concentrations of cAMP can activate PKA-C but do not result in its dissociation from PKA-R (3, 4). Testing these two models will not only elucidate the biophysical mechanism of PKA activation, but also have distinct implications in how PKA may achieve its specificity, which is thought to rely on spatial compartmentalization (5).

We have previously found that the majority of type IIβ PKA, as defined by PKA-R, is anchored to microtubules in the dendritic shaft of hippocampal CA1 pyramidal neurons where PKA-RIIβ is bound to the abundant microtubule associated protein MAP-2 (6). Upon activation of the β-adrenergic receptor with norepinephrine, a fraction of PKA-C dissociated from PKA-RIIβ (7). The freed PKA-C redistributed into dendritic spines, whereas PKA-RIIβ remained anchored at the dendritic shaft (7, 8). These results are consistent with the classical PKA activation model. However, recent studies suggest that PKA-Rs other than PKA-RIIβ may be the more abundant isoforms in CA1 neurons ( 9, 10 ). It 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 remains unknown whether these PKA isoforms dissociate upon physiologically relevant stimulations.

Here, we examined whether PKA-C dissociates from the major PKA-R isoforms, PKA-RIβ and PKA-RIIα, in CA1 neurons. The rescue of function following knockdown of PKA-C was examined using wildtype dissociable PKA versus a non-dissociable PKA variant in which PKA-C is covalently linked to PKA-R. The result supports the classical model of PKA activation via dissociation.

Recent studies have suggested that PKA-RIIα and PKA-RIβ may be the prevalent neurons of organotypic hippocampal slice cultures (Figs. 1A and 1B, upper left panels). PKA-C-EGFP exhibited a resting distribution that was dependent on the co-expressed PKAR: when co-expressed with PKA-RIIα, PKA-C was enriched in dendritic shafts; when coexpressed with PKA-Riβ, PKA-C was more evenly distributed (quantified using the spine enrichment indexes, or SEI, see Methods) (Fig. 1C). This distribution largely mimicked that of the corresponding PKA-R (Figs. 1A-1C).

Notably, upon application of norepinephrine (10 μM), PKA-C of both subtypes translocated to dendritic spines, but the subcellular localization of PKA-Rs remained unchanged (Figs. 1D & 1E). This differential re-distribution between PKA-C and PKA-R was more prominent following activation with forskolin (25 μM) and IBMX (50 μM). These results can only be explained if at least a fraction of PKA-C dissociated from both PKA-RIIα 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 and PKA-RIβ when a physiological stimulant was used. Together with our earlier results regarding PKA-C/PKA-RIIβ, we conclude that PKA-C dissociates from PKA-R with physiologically relevant stimuli.

Next, we asked whether PKA regulation of neuronal function is dependent on the dissociation of PKA-C from PKA-R. A key experiment supporting the non-dissociating PKA activation model was that PKA regulation of cell growth could be sustained by a construct in which PKA-C was fused to PKA-RIIα in one polypeptide chain via a flexible linker (named R-C; Fig. 1F). We therefore asked whether this R-C construct could support PKA regulation of neuronal function. When R-C-EGFP was expressed in CA1 neurons, this construct exhibited a distribution highly similar to that of RIIα at the resting state (Fig. 1G and 1H). The tendency of this construct to translocate to the spine was largely diminished compared to PKA-C-EGFP co-expressed with wildtype, unlinked PKA-RIIα (Fig. 1I).

To determine the function of R-C, a previously established shRNA construct was used to selectively knock down PKA-Cα in CA1 neurons in cultured hippocampal slices (7). Given that PKA activation is required for the late phase of long-term potentiation (L-LTP) ( 11 ), we examined the structural LTP of individual dendritic spines of CA1 neurons elicited by focal two-photon glutamate uncaging (Figs. 2A–2C) (12). The shRNA knockdown of PKA-C resulted in attenuated LTP at around 90 min after induction (Figs. 2A–2C). This attenuation was not observed when a control shRNA against LacZ was expressed (Fig. 2C). The attenuated structural LTP was rescued by co-expression of shRNA-resistant wild-type PKA-C-EGFP together with PKA-RIIα (Figs. 2A–2C). However, the R-C construct in which 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129

PKA-C was also resistant to shRNA knockdown but could not leave PKA-RIIα failed to rescue the phenotype.

PKA activity has also been shown to regulate synaptic transmission. We therefore examined evoked AMPA and NMDA receptor (AMPAR and NMDAR, respectively) currents in paired, transfected and adjacent untransfected CA1 neurons in cultured hippocampal slices. As shown previously (7), neurons expressing the shRNA construct against PKA-C exhibited significantly lower AMPAR currents (Fig. 2D), but not NMDAR currents (Fig. 2E). As a result, the AMPAR/NMDAR current ratio was also reduced (Fig. 2F). The reduced AMPAR currents and reduced AMAPR/NMDAR current ratios were rescued by coexpression of shRNA-resistant, wild-type dissociable PKA-C-EGFP and PKA-RIIα (Figs. 2D– 2F). However, the non-dissociable R-C construct failed to rescue the phenotype. Taken together, the R-C construct did not support normal PKA-dependent synaptic function.

The results indicate that at least a fraction of PKA-C molecules dissociate from all tested PKA-R isoforms when activated by physiological stimuli. Given that PKA activity increases by two orders of magnitude when dissociated from PKA-R ( 13 ), even a small fraction of PKA-C dissociation will result in a marked increase of PKA kinase activity. Notably, our results corroborate the recent results in HEK cells that used biochemical measurements to show that PKA-C and PKA-R indeed dissociate in intact cells ( 14 ). Furthermore, the PKA-C that is covalently linked to PKA-RIIα cannot functionally replace wildtype PKA for normal neuronal transmission or plasticity. Although the R-C construct has been shown to functionally replace endogenous PKA in terms of supporting the growth 130 131 132 133 134 135 136 137 138 139 140 141 142 of a heterologous cell line (4), our results indicate that it cannot support all necessary PKA functions. Overall, we conclude that PKA-C dissociation from PKA-R is essential for PKA regulation of neuronal function. Additionally, PKA specificity is mediated by spatial compartmentalization (5). This is likely mediated by mechanism downstream of PKA dissociation, such as membrane tethering of freed PKA-C or its buffering by extra PKA-Rs ( 7, 14 ).

This study also demonstrates that PKA is essential for long-term structural LTP of individual spines. PKA has been shown to facilitate the induction of LTP (i.e., metaplasticity) (15), and is required for the maintenance of L-LTP ( 11 ), as assayed using electrophysiological recording of postsynaptic currents. However, it has been suggested that the structural and synaptic current changes may not be causal ( 16 ). Our results fill the gap to show that PKA is also required for the late phase of structural LTP. 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164

Materials and Methods

Constructs were made using standard mutagenesis and subcloning methods. All previously unpublished constructs and their sequences will be submitted to Addgene. In the R-C construct, mouse PKA-RIIα and PKA-Cα were fused via a linker with residues WDPGSGSLEAGCKNFFPRSFTSCGSLEGGSAAA that were previously used (4). Organotypic hippocampal slice cultures and transfections Cultured rat hippocampal slices were prepared from P6 – P8 (typically P7) pups, as described previously ( 6, 17 ). Animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, and were approved by the Institutional Animal Care and Use Committee (IACUC) of the Oregon Health & Science University (#IP00002274). cDNA constructs were transfected after 1.5–3 weeks in vitro via the biolistic gene transfer method using the Helios gene gun and 1.6 μm gold beads (Fig. 1) or, where long-term expression (~ 1 week) was required, with singlecell electroporation (Fig. 2) ( 18 ).

Two-photon imaging and two-photon glutamate uncaging A custom built two-photon microscope based on an Olympus BW51WI microscope body was used. Laser beams from two different Ti:Sapphire lasers (Maitai, Newport) were aligned to allow for simultaneous two-photon excitation and photoactivation. Laser intensities were controlled by Pockels cells (Conoptics). Imaging and photoactivation were controlled by ScanImage (Vidrio Tech) ( 19 ). Slices were perfused in gassed artificial cerebral spinal fluid (ACSF) containing 4 mM Ca, 4 mM Mg, and 0.5 µM tetrodotoxin (TTX) 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 during imaging. mEGFP fluorescence (green) was unmixed from that of the cytosolic marker (mCherry or DsRed Express) using a dichroic (Chroma 565DCXR) and band-pass filters (Chroma HQ510/70 for green and Semrock FF01-630/92 for red).

For single-spine structural LTP experiments, 2.25 mM MNI-caged-L-glutamate (Tocris) was added to ACSF containing 4 mM calcium, 0.05 mM magnesium, 1 μM TTX and 4 µM 2-chloroadenosine, as previously described ( 20 ). To trigger structural plasticity, 30 pulses of 4-ms 16-mW (at back focal plane) 720-nm laser light were delivered to the spine head at 0.5 Hz.

Image analysis was performed using custom software written in MATLAB called SI_View (https://github.com/HZhongLab/SI_View) (21). Using the software, regions of interest (ROIs) were manually drawn to isolate spines or their immediately adjacent dendritic shaft. Only the spines well isolated from the dendrite laterally throughout the entire experiments were included. Spine enrichment index was calculated as:

SEI = log2[(Fgreen/Fred)spine/(Fgreen/Fred)shaft] in which F is the average fluorescence intensity in an ROI.

Whole-cell voltage-clamp recordings were performed using a MultiClamp 700B amplifier (Molecular Devices). Electrophysiological signals were filtered at 2 kHz and digitized and acquired at 20 kHz using custom software written in MATLAB. Slices were perfused with artificial cerebrospinal fluid containing 4 mM Ca and 4 mM Mg. The internal solution contained (in mM) 132 Cs-gluconate, 10 HEPES, 10 Na-phosphocreatine, 4 MgCl2, 4 Na2186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207

ATP, 0.4 Na-GTP, 3 Na-ascorbate, 3 QX314, and 0.2 EGTA with an osmolarity of 295 mOsmol/kg. The junction potential was calculated to be -17 mV using a built-in function in the Clampfit software (Molecular Devices). Several less abundant anions (phosphocreatine, ATP, GTP and ascorbate) were omitted in the calculation due to lack of data in the program. The Cl reversal potential was -75 mV.

To reduce recurrent activities, cultured hippocampal slices were cut on both sides of CA1 and 4 µM 2-chloroadenosine (Sigma) was present in all recording experiments. 10 µM GABAzine (SR 95531, Tocris) was also included to suppress GABA currents. For electrical stimulation, a bipolar, θ-glass stimulating electrode (Warner Instruments) was positioned in the stratum radiatum 100–150 μm lateral to the recorded neuron. For all recordings, a transfected neuron and an untransfected neuron located within 50 µm of each other were sequentially recorded without repositioning the stimulation electrode. Measurements were carried out on averaged traces from approximately 20 trials under each condition. For AMPAR currents, the cells were held at -60 mV (before correcting for the junction potential) and the current was measured as the baseline-subtracted peak current within a window of 2–50 ms after electric stimulation. For NMDAR currents, the average currents at 140 to 160 ms after stimulation were used when the cells were held at +55 mV (before correcting for the junction potential) Data analysis, presentation, and statistics Quantification and statistical tests were performed using custom software written in MATLAB. Averaged data are presented as mean ± s.e.m., unless noted otherwise. p values were obtained from one-way ANOVA tests, unless noted otherwise. In all figures, *: p ≤ 0.05 208 209 and is considered statistically significant after Bonferroni correction for multiple tests, **: p ≤ 0.01, and ***: p ≤ 0.001. 210 211 212 213 214 215 216 217 218 219 220 221 222

Acknowledgements We thank all members of the Mao and Zhong laboratories at the Vollum Institute for constructive discussions. We thank Drs John Williams and Michael Muniak for critical comments and edits on the manuscript. This work was supported by two NIH BRAIN Initiative awards to H.Z. (U01NS094247 and R01NS127013).

H.Z. conceived the project. W.X. and H.Z. designed the experiments. W.X. and H.Z. performed the experiments and data analyses. M.Q. produced critical reagents used in the experiments. H.Z. wrote the manuscript.

The authors declare no competing interests. 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242

S. H. Francis, J. D. Corbin, Structure and Function of Cyclic Nucleotide-Dependent Protein Kinases. Annu. Rev. Physiol. (1994)

D. A. Johnson, P. Akamine, E. Radzio-Andzelm, Madhusudan, S. S. Taylor, Dynamics of cAMP-dependent protein kinase. Chem. Rev. (2001)

F. D. Smith, et al., Intrinsic disorder within an AKAP-protein kinase A complex guides local substrate phosphorylation. Elife 2013 (2013).

F. D. Smith, et al., Local protein kinase A action proceeds through intact holoenzymes. Science (80-. ). 356 (2017).

W. Wong, J. D. Scott, AKAP signalling complexes: Focal points in space and time. Nat. Rev. Mol. Cell Biol. (2004) https:/doi.org/10.1038/nrm1527.

H. Zhong, et al., Subcellular Dynamics of Type II PKA in Neurons. Neuron 62, 363–374 (2009). 617–629 (2017).

S. E. Tillo, et al., Liberated PKA Catalytic Subunits Associate with the Membrane via Myristoylation to Preferentially Phosphorylate Membrane Substrates. Cell Rep. 19, W. H. Xiong, M. Qin, H. Zhong, Myristoylation alone is sufficient for PKA catalytic subunits to associate with the plasma membrane to regulate neuronal functions. 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262

M. Weisenhaus, et al., Mutations in AKAP5 disrupt dendritic signaling complexes and lead to electrophysiological and behavioral phenotypes in mice. PLoS One 5 (2010). a identified by mosaic imaging of mouse brain. Elife 6 (2017). hippocampus-based long-term memory. Cell 88, 615–626 (1997). 12.

M. Matsuzaki, N. Honkura, G. C. R. Ellis-Davies, H. Kasai, Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761–766 (2004).

Purification and properties of adenosine 3’,5’-monophosphate-dependent protein kinase from bovine brain. J. Biol. Chem. 244 (1969). protein kinase revealed by subunit quantitation and cross-linking approaches. Proc.

Natl. Acad. Sci. U. S. A. 114 (2017). 15.

M. J. Thomas, T. D. Moody, M. Makhinson, T. J. O’Dell, Activity-dependent βadrenergic modulation of low frequency stimulation induced LTP in the hippocampal CA1 region. Neuron 17, 475–482 (1996). spine enlargement during chemically induced long-term potentiation. J. Neurosci. 26 263 264 265 266 267 268 269 270 271 272 273 274 275

(2006). activation of a single dendritic spine. Science (80-. ). 321, 136–140 (2008). 21. L. Ma, et al., Locomotion activates PKA through dopamine and adenosine in striatal neurons. Nature 611, 762–768 (2022). 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 Fig. 1. PKA-C but not PKA-R or a non-dissociable R-C redistribute to spines upon activation. (A, B) Representative two-photon images of PKA-C-EGFP co-expressed with PKA-RIIα or PKA-RIβ at rest, or in the presence of norepinephrine (NE) or forskolin and IBMX (F + I). mCherry (magenta) was co-expressed to reveal the neuronal morphology. (C– E) Quantification and comparison of the spine enrichment index at the resting state (C) and upon activation (D & E). As in panel C from left to right, n (spines/neurons) = 53/11, 34/7, 33/6, and 36/7. (F) Schematic of wildtype PKA versus R-C. In both cases PKA-C was tagged by mEGFP. (G–I) Representative images (G), quantifications of resting distribution (H), and the distribution change upon stimulation by forskolin and IBMX (I) of R-C compared to PKA-RIIα-EGFP and co-expressed PKA-C-EGFP/PKA-RIIα. n (spines/neurons) = 48/10. Fig. 2. PKA regulation of synaptic plasticity and synaptic transmission cannot be sustained using a non-dissociable PKA variant. (A–C) Representative image (A), time course (B), and the degree of potentiation (C) at the indicated timepoints in panel B of single-spine LTP experiments as triggered by focal glutamate uncaging at the marked spines (gray dot). In panel B, both stimulated spines (solid circles) and non-stimulated control spines (open circles) are shown. As in panel C from left to right, n (spines, each from a different neuron) = 8, 7, 17, 11, 9. (D–F) Representative traces (red) normalized to the paired control (blue) (insets) and scatter plots of paired AMPA (D) and NMDA (E) receptor currents and AMPA/NMDA receptor current ratios (F) from neighboring untransfected CA1 neurons paired with those transfected with shRNA against PKA-C and the indicated shRNA-resistant 297 298 rescue constructs. Statistical p values were obtained using a sign test (MATLAB). From left to right, n (neuron pairs) = 13, 11, 11, and 15. basal

NE

F+I basal

NE

F+I A PKA-C / RIIα / mCherry

RIIα / mCherry C x1 e d n i t n0 e m h c i r-1 n e e n i p-2 s t s e R 2 µm F