A high-performance genetically encoded fluorescent indicator for in vivo cAMP imaging

cAMP is a key second messenger that regulates diverse cellular functions including neural plasticity. However, the spatiotemporal dynamics of intracellular cAMP in intact organisms are largely unknown due to low sensitivity and/or brightness of current genetically encoded fluorescent cAMP indicators. Here, we report the development of the new circularly permuted GFP (cpGFP)-based cAMP indicator G-Flamp1, which exhibits a large fluorescence increase (a maximum ΔF/F0 of 1100% in HEK293T cells), relatively high brightness, appropriate affinity (a Kd of 2.17 µM) and fast response kinetics (an association and dissociation half-time of 0.20 s and 0.087 s, respectively). Furthermore, the crystal structure of the cAMP-bound G-Flamp1 reveals one linker connecting the cAMP-binding domain to cpGFP adopts a distorted β-strand conformation that may serve as a fluorescence modulation switch. We demonstrate that G-Flamp1 enables sensitive monitoring of endogenous cAMP signals in brain regions that are implicated in learning and motor control in living organisms such as fruit flies and mice.


Introduction
and R-FlincA were found to localize mainly in the cytosol ( Fig. 2a and Fig. 1b), with the latter forming puncta 48 hours post transfection ( Supplementary Fig.   231 1d) and thus likely being toxic to mammalian cells 26 . Under one-photon (488 nm) 232 illumination, the basal fluorescence intensities of G-Flamp1, cAMPr and Flamindo2 were 233 49%, 109% and 11% of that of GCaMP6s 17 , respectively ( Fig. 2a-b). At 450 nm, which 234 gives the largest ΔF/F 0 , the brightness is reduced by half and is ~25% of that of 12 We further determined the fluorescence change and sensitivity of G-Flamp1. Forskolin 252 (Fsk), a potent activator of transmembrane AC 30 , was used to induce a high level of 253 cAMP to assess the maximum fluorescence change. Under 450 nm illumination, G-254 Flamp1 expressed in HEK293T cells exhibited a maximum ΔF/F 0 of 1100% in response 255 to 60 µM Fsk, which was 9−47 times larger than those of other cAMP probes ( Fig. 2c- Supplementary Fig. 1b). G-Flamp1 also showed large fluorescence increases with a 257 maximum ΔF/F 0 of 340% and 820% in HeLa and CHO cells, respectively 258 ( Supplementary Fig. 12). To rule out possible unspecific responses, we generated a 259 cAMP-insensitive indicator G-Flamp1-mut by introducing the R307E mutation into 260 mlCNBD of G-Flamp1 (Supplementary Fig. 13) 20 . As expected, G-Flamp1-mut showed 261 no detectable signal change in living cells (Fig. 2c). To demonstrate the sensitivity of G-262 Flamp1, 2.5 nM Iso was exploited to produce a small amount of cAMP in HEK293T 263 cells. G-Flamp1 exhibited an obvious fluorescence increase with a ΔF/F 0 > 100% after 5 264 min stimulation while other sensors showed little signal changes (|ΔF/F 0 | < 10%) in our 265 setup (Fig. 2e). Under two-photon excitation (920 nm), G-Flamp1 exhibited a maximum 266 ΔF/F 0 of 1240%, which is much larger than those of Flamindo2 and cAMPr (-79% and 267 72%, respectively) ( Supplementary Fig. 10b-c). Meanwhile, G-Flamp1 had a 13-and 268 90-fold higher signal-to-noise ratio (SNR) compared with Flamindo2 and cAMPr, 269 respectively ( Supplementary Fig. 10d).  Fig. 2f and Supplementary Fig. 14), indicating the high specificity of G-Flamp1 280 towards cAMP over cGMP. Regarding reversibility, HEK293T cells expressing G-281 Flamp1 exhibited increased fluorescence upon 100 nM Iso treatment and then returned to 282 basal level after addition of 15 µM β-AR anti-agonist propranolol (Prop) (Fig. 2g).  In vivo two-photon imaging of cAMP dynamics in zebrafish 295 To test whether G-Flamp1 can function in intact living organisms, we first utilized 296 optically transparent zebrafish embryos under Fsk stimulation. We injected UAS:G-297 Flamp1(or G-Flamp1-mut)-T2A-NLS-mCherry (nuclear localized mCherry) plasmid into 298 the embryos of EF1α:Gal4 transgenic zebrafish at one-cell stage (Supplementary Fig.   299   16a). The expression of G-Flamp1 or G-Flamp1-mut sensor was confirmed by green 300 fluorescence in cells of the developing central nervous system. Brain ventricular injection 301 of 120 µM Fsk but not PBS elicited a robust fluorescence increase with a ΔF/F 0 of 450% 302 for G-Flamp1, whereas no signal changes were observed for G-Flamp1-mut 303 (Supplementary Fig. 16b-d). These data indicate that G-Flamp1 sensor has high 304 sensitivity for in vivo cAMP detection in zebrafish.

306
In vivo two-photon imaging of cAMP dynamics in Drosophila 307 The importance of cAMP in associative learning, where it serves as a coincidence 308 detector by integrating concurrent signal inputs from both conditioned and unconditioned 309 stimuli, has been well documented across phyla 33, 34 . In Drosophila, cAMP signaling in . When the fly was exposed to either 1 s odor puff or subsequent 0.5 s electrical shock, 315 we observed time-locked fluorescence responses with a ΔF/F 0 of ~100% (Fig. 3d-e). 316 Compared with the MB β' lobe that has similar responses among different compartments, 317 the MB γ lobe exhibited compartmentally heterogeneous responses to specific stimuli, as 318 the largest responses were observed in γ4 to odor and in γ2 to electrical shock. These 319 compartmentalized signals were not due to the unequal expression level or saturation of 320 the sensor, since 100 µM Fsk perfusion elicited a homogeneous ΔF/F 0 of around 250% 321 (Fig. 3f). G-Flamp1 specifically reported cAMP changes since the GFP alone expressed 322 in KCs showed no significant response to 1 s odor, 0.5 s shock or 100 µM Fsk perfusion 323 ( Fig. 3d-f). Moreover, both the rise and decay time (τ on and τ off ) for cAMP changes 324 evoked by odor or shock were similar in different compartments (Fig. 3g-h). 325 Collectively, these results show that G-Flamp1 allows detection of physiologically 326 relevant cAMP dynamics in Drosophila with high fidelity and good spatiotemporal 327 resolution, and sheds lights on the role of compartmentally separated cAMP signaling in 328 the olfactory learning process.

330
In vivo two-photon imaging of cAMP dynamics in mouse cortex 331 To demonstrate the utility of G-Flamp1 sensor to detect physiologically relevant cAMP 332 dynamics in living animals, we performed head-fixed two-photon imaging in the motor 333 cortex (M1) of awake mice during forced locomotion (Fig. 4a), which was reported to be 334 associated with increased neuromodulator and PKA activities 37 . We co-expressed G-335 Flamp1 (or G-Flamp1-mut) and the red calcium sensor jRGECO1a in the neurons of 336 motor cortex and imaged the layer 2/3 region (Fig. 4b). We observed running-induced, 337 cell-specific, cAMP and calcium signals with no correlation (Fig. 4c). Interestingly, 338 neurons in M1 area could be further divided into three groups based on the cAMP 339 dynamics: ~60% neurons with fast increase of cAMP (higher average response during the 340 first 30 s after the onset of forced running) and no significant change of calcium, ~30% 341 neurons with slow increase of cAMP and little change of calcium, and ~6% neurons with 342 decrease of cAMP and increase of calcium (Fig. 4c). As a control, G-Flamp1-mut 343 showed little fluorescence change (Fig. 4d). Distribution analysis and averaged traces of 344 ΔF/F 0 of G-Flamp1 and jRGECO1a further confirmed the heterogeneity of neuronal 345 responses ( Fig. 4e-i). Therefore, dual-color imaging of calcium and cAMP revealed cell-346 specific neuronal activity and neuromodulation of cortical neurons in mice during forced 347 locomotion.

349
In vivo fiber photometry recording of cAMP dynamics in mouse nucleus accumbens 350 To test the ability of G-Flamp1 sensor to report cAMP dynamics in deep brain regions, 351 we measured cAMP levels in the nucleus accumbens (NAc) using fiber photometry in 352 mice performing a classical conditioning task. The NAc was chosen because it is recently 353 reported that PKA, a downstream molecule in the cAMP signaling pathway, plays a 354 critical role in dopamine-guided reinforcement learning behavior 38 . We first injected an 355 adeno-associated virus (AAV) expressing G-Flamp1 into the NAc and measured 356 fluorescence signals using fiber photometry while the mice were trained to perform the 357 conditioning task (Fig. 5a). In the task, the mice were trained to learn the associations 358 between three auditory cues (conditioned stimulus, CS) and respective outcomes 359 (unconditioned stimulus, US) ( Fig. 5b; 8 kHz pure tone → water; white noise → brief air 360 puff to the animal's face; 2 kHz pure tone → nothing). Well-trained mice had a high 361 licking rate selectively to the water-predictive sound, and the G-Flamp1 signal showed a 362 large increase immediately after the onset of the water-predictive sound, while responses 363 to the other two sounds were much smaller ( Fig. 5c-e).

365
Interestingly, the G-Flamp1 signal in the water trials exhibits characteristic dynamics 366 during the learning process: in naïve mice, there was a notable signal increase to water 367 delivery; throughout the training, the magnitude of the water-evoked response decreased, 368 while a response to the reward-predictive sound gradually increased ( Fig. 5f-h). This 369 dynamic change mimics the dopamine signal during classical conditioning 39, 40 , 370 suggesting that the increase in cAMP in the NAc is mainly driven by dopamine release.

371
To confirm this, we thus blocked the dopamine D1 receptor using SCH22390 (i.p.) and 372 observed a significantly reduced cAMP signal ( Fig. 5i-j). Together, these results 373 demonstrate that the G-Flamp1 sensor has a high signal-to-noise ratio and high temporal 374 resolution to report the dynamic changes of cAMP in behaving mice.

377
In this study, we described G-Flamp1, a high-performance GEAI engineered by inserting  Our in vivo two-photon imaging experiments in mouse cortex showed that G-Flamp1 is 386 able to detect bidirectional cAMP changes with single-neuron resolution (Fig. 4). Given 387 that multiple neuromodulators can be released in the motor cortex 37 , different downstream signaling processes are expected to be induced in cortex neurons, which 389 might partially explain the discrepancy between cAMP signal and calcium activity in our 390 results (Fig. 4f). Further studies are needed to dissect out the underlying regulation 391 mechanisms and potential functions. Nevertheless, together with other spectrally 392 compatible sensors, G-Flamp1 will be a useful tool for investigating signal transduction 393 networks in behaving animals.   The association constant (k on ) and dissociation constant (k off ) between G-Flamp1 and 522 cAMP were determined using Chirascan spectrometer equipped with an SX20 Stopped- The association and dissociation half-time t on and t off were calculated as ln2/(k on × 532 [cAMP]) and ln2/k off , respectively. The coding sequence of G-Flamp1 was cloned into pJFRC28 (Addgene plasmid #36431).

662
The vector was injected into embryos and integrated into attP40 via phiC31 by the Core

33.
Kandel, E.R. The molecular biology of memory storage: a dialogue between genes and synapses. 883  Flamp1 and GFP groups, respectively. Two-tailed Student's t-tests were performed in d3.

1061
(e) Similar to d except that 0.5 s electrical shock was applied to the fly. n = 9 and 5 for G-1062 Flamp1 and GFP groups, respectively. Two-tailed Student's t-tests were performed in e3.