Bioorthogonal chemical labeling of endogenous neurotransmitter receptors in living mouse brains

We chose AMPARs in the cerebellum as the initial target receptors to test our strategy for in vivo LD chemistry. AMPARs, which are heteromeric tetramers of GluA1–4 (15, 16), are abundantly expressed in the central nervous system and have a characteristic distribution in the cerebellar molecular layer, which can be used as a good indicator to evaluate the efficiency and specificity of the chemical labeling using imaging techniques. We have previously developed a series of ligand-directed alkoxyacylimidazole (LDAI)-based reagents (10–13), and CAM2-Ax647 was selected for the present study because, among the developed compounds, this reagent had the least background signal derived from the autofluorescence of tissues. In addition, we expected that the hydrophilic and cell-impermeable features of this reagent would suppress nonspecific adsorption to hydrophobic materials in tissues and promote the labeling of only cell-surface (functional) receptors. For the delivery of CAM2-Ax647 to living brains, we used a direct injection protocol that is widely used for virus injections. Initially, we injected the reagent (50 µM, 4.5 µL) into the cerebellum (Cb) region of live C57BL/6N mice (5 wk old) under anesthesia (Fig. 2A). At 20 h after the injection of CAM2-Ax647, the mice appeared to behave normally (Movie S1). In western blot (WB) analysis of the Cb homogenates, a strong band corresponding to the labeled AMPAR (100 kDa, a GluA subunit) was detected using an anti-Ax647 antibody (Fig. 2B). No bands were detected with a control compound lacking the ligand moiety (NLC) or vehicle (DMSO) control, indicating that the labeling was driven by selective ligand–receptor recognition. We also confirmed the high stability of the formed covalent bond (no degradation of this bond was observed after 8 d in Neurobasal medium, SI Appendix, Fig. S2). Confocal laser scanning microscopy (CLSM) imaging of the Cb slices, which were prepared from a mouse injected with CAM2-Ax647, showed obvious and selective fluorescence signals of Ax647 from the molecular layer and almost negligible fluorescence from the granular layer (Fig. 2 C and D), which was consistent with the localization of endogenous AMPARs in the Cb (17). In contrast, only dim fluorescence was observed in the slices prepared from mice brains injected with NLC or DMSO. Whole-cell patch clamp recordings from PCs in the Cb acute slices were performed to examine the influence of the injection of CAM2-Ax647 on the electrophysiological properties of AMPAR-mediated synaptic responses, such as PF- and climbing fibers (CF)-mediated excitatory postsynaptic current (PF- and CF-EPSC, respectively) amplitude and kinetics and the paired-pulse ratio of each EPSC, which reflect a presynaptic neurotransmitter release property (Fig. 2 E–P). Both the CF- and PF-EPSCs were identical between samples with and without CAM2-Ax647 treatment. This result indicated that the detrimental impacts of the chemical labeling on the characteristics of AMPARs were minimal or negligible because the ligand moiety is cleaved upon labeling in the LD chemistry.

We next investigated lateral ventricle (LV) injection with the aim of labeling AMPARs over a wide brain region (Fig. 3A). The mice administered CAM2-Ax647 by LV injection showed no behavioral abnormalities (Movie S2). After labeling, the mouse brains were separated into Cb and non-Cb regions including the cerebral cortex, hippocampus, and striatum and analyzed by WB. A single band corresponding to the labeled AMPARs was clearly observed at approximately 100 kDa (Fig. 3B) from both Cb and non-Cb regions. Immunoprecipitation with anti-Ax647 followed by mass analysis (immunoprecipitation mass spectrometry: IP-MS) predominantly resulted in the detection of the four subunits of the AMPAR (GluA1–4), which had the highest enrichment scores, demonstrating the selective labeling of AMPARs with CAM2-Ax647 over a wide brain region (Fig. 3C and Dataset S1). CLSM imaging of whole sagittal brain slices showed strong fluorescence signals of Ax647 from the hippocampus and the molecular layer of Cb and relatively weak but clear fluorescence from the cortex and striatum, indicating that reagents injected into the LV were well distributed over the whole-brain tissues via the cerebrospinal fluid (CSF) flow (19). In contrast, these fluorescent signals were not observed in slices with the NLC or the non-reactive control reagent (NAIC) (Fig. 3D). These negative control experiments indicated that the fluorescence observed with CAM2-Ax647 injection came from covalently labeled AMPARs.

The AMPAR distribution visualized by Ax647 labeling in the whole brain was in good accordance with that observed with anti-GluA2 antibody staining (Fig. 3D). The high-resolution images of the hippocampal CA1 region showed abundant fluorescent puncta of Ax647 and most of these puncta were colocalized with anti-GluA2 antibody signals (Fig. 3E). These puncta were located in between pre- and post-synapses stained with anti-Bassoon and anti-Homer1a antibodies, respectively (Fig. 3F). These data demonstrated that endogenous AMPARs accumulated in the synaptic cleft could be labeled and visualized with a synthetic Ax647 fluorophore. Collectively, these results indicated that AMPARs were labeled with high selectivity after CAM2-Ax647 administration into the brain.

To obtain structure–activity insights into the in-brain LD receptor labeling, we additionally prepared two compounds, CAM2-Cy5 and CAM2-SulfoCy5, which have no and two negatively charged sulfonate groups in the Cy5 fluorophore, respectively, and another labeling reagent (CAM2-Ax555) with a Cy3-based fluorophore (which has a different emission wavelength) bearing four sulfonate groups. We then investigated the distribution properties of these compounds in the brain compared with CAM2-Ax647. Fig. 3G shows that the unsulfonated CAM2-Cy5 scarcely spreads from the injection site, while the more negatively charged CAM2-SulfoCy5 (divalent anion) and CAM2-Ax647 (tetravalent anion) are less adsorbed at the injection site and are widely diffused throughout the brain. In particular, CAM2-Ax647 showed excellent diffusion, reaching areas of the Cb far from the injection site and had the highest signal-to-noise ratio. Similar high diffusibility was also observed for CAM2-Ax555 (tetravalent anion), which showed almost the same labeling pattern as CAM2-Ax647 (SI Appendix, Fig. S3). These results indicated that the negative (net) charge of these reactive small molecules had a considerable impact on the diffusion in the live brain.

We subsequently conducted CUBIC-based tissue clearing (20) of the whole brain after CAM2-Ax647 LV injection. Light-sheet fluorescence microscopy (LSFM) showed stronger fluorescence signals in the regions of the hippocampus and cerebellar molecular layer, which were in good accordance with the high expression areas of AMPARs (Fig. 3 H and I and Movie S3) (17). In CLSM imaging of this transparent brain sample, many small punctate signals (less than 1 µm in diameter), presumably originating from dendritic spines, were observed (Fig. 3J). We also performed whole-brain AMPAR labeling in the established 5xFAD transgenic mouse model of Alzheimer’s disease that has a different brain shape, compared to the C57BL/6 strain for which a standard brain atlas exists. To more efficiently label the cerebral cortex, the region of interest, we optimized the injection method. By using cortex and LV injection methods, the distribution of the labeled AMPARs was detected over a wide area in normal and 5xFAD mice (SI Appendix, Fig. S4 and Fig. 3K). In 5xFAD mouse, labeled AMPARs were not observed around senile plaques, which are extracellular deposits of highly neurotoxic Aβ proteins, probably because of a loss of excitatory neurons (Fig. 3L and SI Appendix, Fig. S5) (21). 3D analysis using a transparent brain sample showed that the density of AMPAR puncta in the regions containing senile plaques was less than those in other areas, 0.295 ± 0.008 and 0.316 ± 0.005 spots/µm³, respectively (Fig. 3M and Movie S4). These results demonstrated that our chemical labeling method enabled the visualization of the spatial distribution of endogenous AMPARs in the whole brains of genetically intact, transgenic, and disease-model mice without requiring the preparation of numerous tissue slices. Notably, the receptor labeling was completed before fixation/clearing processes in our method, unlike antibody-based staining, which enabled snapshot images of the target endogenous receptors under more physiologically relevant conditions of the brain to be obtained.

Owing to the modular features of LDAI reagents, this chemistry can be readily extended for targeting other receptors by changing the ligand moiety. To demonstrate the robustness of our in-brain LD chemistry, we targeted several receptors with different structures and functions, namely, mGlu1 (22), a class-C G protein-coupled (metabotropic glutamate) receptor; NMDAR (23), an ionotropic glutamate receptor involved in excitatory synaptic transmission together with AMPAR; GABA_AR (24), a major inhibitory neurotransmitter receptor. Leveraging on the vast knowledge obtained from past drug discovery research on these receptors, we employed CNITM (25), L-689,560 (23), and flumazenil (24) as the ligand moieties for labeling reagents targeting mGlu1, NMDAR (NR1 subunit), and GABA_AR, respectively (Fig. 1B and SI Appendix, Fig. S1). Although previously we have used gabazine and a benzodiazepine derivative as the ligand moiety in labeling reagents for GABA_AR in model HEK293T cells (13), flumazenil was employed in the present live brain study because of its minimal toxicity. The corresponding ligand and the tetrasulfonated Ax647 were connected by an acyl imidazole group through a linker with an appropriate length as optimized in in vitro experiments (SI Appendix, Fig. S6). We then injected these LDAI reagents into the LVs of a mouse brain and confirmed that there were no serious effects on the mouse behavior. Note that a previously reported GABA_AR-labeling reagent with a gabazine ligand (orthosteric antagonist) caused markedly intense seizures in mice, indicating the importance of appropriate ligand selection and dosage in in-brain LD chemistry. After mGlu1 labeling with CmGlu1M, the WB of the Cb homogenate exhibited three bands (300, 150, and 80 kDa) corresponding to aggregated, monomeric, and partially truncated mGlu1, respectively (Fig. 4A). These bands were not detected in a mGlu1 knockout mouse (26), clearly indicating the specific labeling of mGlu1 in the brain (Fig. 4A). Similarly, a strong band corresponding to the labeled NMDAR (110 kDa, NR1 subunit) and GABA_A R (45 kDa, γ2 subunit) was predominantly detected with CNR1M and CGABAaRM, respectively, while no obvious bands were detected with NLC. The IP-MS data also indicated the high target selectivity of these reagents as shown in Datasets S2–S4. In the CLSM imaging of the brain sections, fluorescence signals derived from the labeled receptors were observed from the particular brain regions where the corresponding endogenous receptors are reported to be abundantly expressed (Fig. 4B). These fluorescence signals were well merged with those from immunostaining using anti-mGlu1 (27), anti-NR1A (28), and anti-GABA_AR-α1 (29) antibodies (Fig. 4C and SI Appendix, Fig. S7). We also conducted electrophysiology experiments for NMDA and GABA receptors and calcium responses for mGlu1. Any significant difference between data for the administration of labeling reagent and for DMSO administration was not observed. These results indicated that, like AMPA receptors, detrimental impacts of the chemical labeling on the functions of these receptors were minimal or negligible (SI Appendix, Figs. S8 and S9). The labeling was compatible with tissue clearing, which facilitated the 3D mapping of mGlu1, NMDAR, and GABA_AR endogenously expressed in the brain (SI Appendix, Fig. S10).

We then sought to simultaneously label and visualize two different target receptors in an individual mouse. CAM2-Ax555 and CGABAaRM were mixed and injected into the mouse brain. As shown in Fig. 4D, strong fluorescence signals were observed from brain regions identical to the individual (single) labeling of AMPARs and GABA_ARs, indicating negligible interference between these reagents (Fig. 4E and SI Appendix, Figs. S11A–S13). The high-resolution images of Cb regions exhibited a number of punctate signals derived from Ax555 or Ax647. These puncta were never overlapped with each other and were distinguishable in not only 2D but also 3D images (Fig. 4F and SI Appendix, Fig. S11B). We counted these bright spots in the 1,000 µm³ region near cerebellar Purkinje cells, which revealed that the 3D density of AMPAR puncta and GABA_AR puncta are 0.751 ± 0.131 and 0.216 ± 0.069/µm³, respectively (Fig. 4F and Movie S5). Note that the value for the AMPAR density was similar to the density of excitatory synapses in the molecular layer of rat Cb (0.817/µm³) previously determined by a stereological method using electron microscopy (30). In addition, our finding that AMPAR puncta exhibit a higher density than GABA_AR puncta is consistent with previous reports that excitatory synapses are more abundant than inhibitory synapses (31). These results obtained by quantitative imaging analysis clearly highlight the power of our in vivo LD chemistry coupled with tissue-clearing techniques.

To apply our chemical labeling to more intricate biological experiments, determining how long it takes to complete the reaction and how long the labeled receptors can be followed in the living brain is important. We thus quantitatively characterized the reaction kinetics and the degradation lifetimes of the labeled receptors (Fig. 5A). After injection of CAM2-Ax647, followed by various incubation times prior to dissection, the homogenates of mice brains were subjected to SDS-PAGE in-gel fluorescence analysis. As shown in Fig. 5B and SI Appendix, Fig. S14 A–E, the labeling band intensity increased for the initial 12 h and thereafter the band intensity decreased over several days. This biphasic profile can be fitted to a typical stepwise reaction comprising two distinct processes that presumably are the chemical labeling process (the first increasing step) and the following degradation of the labeled AMPARs (the second decay step). Given that the CSF exchanges every 1.8 h in the mouse brain (32), it is conceivable that most of the CAM2-Ax647 was extruded a few hours after injection without interacting with AMPARs, halting further AMPAR labeling. Curve fitting analyses provided two kinetic values, 2.2 ± 0.6 h for the half-life of the first step (T_{1/2, label}) and 97.2 ± 15.8 h for the second step (T_{1/2, degradation}) in the Cb, and T_{1/2, label} and T_{1/2, degradation} values of 3.9 ± 1.9 and 92.6 ± 18.2 h, respectively, in the brain–Cb/olfactory bulb (OB). The labeling rates (T_1/2 = 2 to 4 h) were sufficiently faster than the subsequent degradation rates (approximately 4 d). We observed similar biphasic kinetics for receptor labeling in HEK293T cells with a labeling rate almost comparable to that of in-brain labeling, i.e., 2.2 to 3.9 h in mice brains and 3.7 ± 1.4 h in HEK293T cells (SI Appendix, Fig. S14E). Interestingly, these data imply that the labeling step (i.e., the acyl transfer reaction) is the rate-determining step rather than the in-brain diffusion of the LDAI reagent in the case of LV injection. Almost the same trends were observed for the other receptors (mGlu1, NMDAR, and GABA_AR) in the brain and HEK cells as shown in Fig. 5C and SI Appendix, Fig. S14E. A signal decay curve of the labeled surface receptors gave degradation lifetimes of ca. 40 to 95 h (T_{1/2, degradation}), which is slightly shorter than the literature values determined by mass spectroscopy coupled with stable isotope labeling (33, 34). Given that the LD chemistry relies on receptor-ligand recognition and our labeling reagents are cell impermeable, the obtained values may reflect the degradation-lifetime of the surface-exposed (functional) receptors under live brain conditions.

We finally applied this labeling technique to in vivo pulse–chase analysis. We investigated the dynamics of AMPARs during cerebellar development because synaptogenesis and synaptic pruning in the neonatal cerebellum have dynamically occurred (35–38), but the detailed molecular basis remains yet to be explored; for example, from where, how, and when are the synaptic proteins synthesized, degraded, and transported. Prior to the pulse–chase experiment, we confirmed that the selective labeling of cerebellar AMPARs in postnatal mice of various ages (at days P4, P7, and P18) could be performed by injection of CAM2-Ax647 (Fig. 6 A–F and Movies S6–S8). The snapshot imaging data showed that the Ax647-tagged AMPARs were localized in the areas surrounding PC soma in the P4 mouse, and the expression pattern changed from the soma to the dendrites of PCs from P7 to P18, which was consistent with previous reports using immunostaining (Fig. 6 D–F) (38).

We then proceeded to perform pulse–chase labeling in the neonatal mouse brain. Given the labeling kinetics and the degradation-lifetime of AMPARs as determined above, pulse–chase analysis in the live brain could conceivably be performed over a 3 to 4 d range with a dead time of ca. 12 h after injection (Figs. 5 and 7A and SI Appendix, Fig. S15). Thus, we injected CAM2-Ax647 into P4 mice (pulse) and maintained the mice for 13 to 67 h (chase), followed by imaging of the cerebellar tissues. As shown in Fig. 7B, the labeled AMPARs were located homogeneously around the soma of PCs at the P4 stage (chase at 13 h). Notably, at chase after 42 and 67 h, abundant fluorescence punctate signals were observed in the dendrites of PCs, and the distribution of the fluorescence that remained in the soma after 13 h became heterogenous. A PC will receive inputs from CFs on soma in the early developmental stage and the PC will extend its dendrites toward the molecular layer to form new synapses with PFs in the distal area (PF synapses) during the period from P4 to P7 (Fig. 7A). Given our findings obtained from the pulse–chase analysis and the established model of synaptogenesis in PCs, we suspected that some of the AMPARs expressed on the surface of PCs at P4 gradually translocate to the PF synapses in dendrites during functional neural circuit formation. To investigate this hypothesis, we conducted colocalization analysis by immunostaining with anti-vGluT1 and anti-vGluT2 antibodies, which are conventional PF- and CF-presynapse markers, respectively (Fig. 7C and SI Appendix, Figs. S16–S18). Although vGluT2 is expressed in both CF and PF at the early developmental stage, CF synapses could be identified by their larger puncta size (39). High-resolution images of PCs showed that the labeled AMPAR signals were distributed over the cell surfaces, and some signals were observed from CF synapses at P4, while PF synapses (vGluT1-positive puncta) were barely detected at this time (Fig. 7C and SI Appendix, Fig. S16). At P7 (chase 67 h), numerous puncta from Ax647 signals along the dendrites were observed adjacent to the fluorescent spots derived from vGluT1, indicating that the nascent PF synapses in the dendrites contained AMPARs that were once present in the soma (Fig. 7 C and D and SI Appendix, Figs. S17 and S18). Overall, these results validated our hypothesis and provided unique insights into AMPAR dynamics during synaptogenesis, i.e., some early expressed (old) AMPARs in the developing mouse brain may be transported over a long distance (~35 µm) and become part of new synapses rather than a simple scrap-and-build scenario (Fig. 7E). Although the physiological significance and mechanism of the distal transport of old AMPARs are beyond the scope of this study, the obtained results highlight the power of our in vivo pulse–chase labeling method using LDAI chemistry for the analysis of previously unknown behaviors of endogenous neurotransmitter receptors in their natural contexts.

All synthesis procedures and characterizations are described in SI Appendix.

C57BL6/N mice were purchased from Japan SLC, Inc (Shizuoka, Japan). The animals were housed in a controlled environment (23 °C, 12 h light/dark cycle) and had free access to food and water, according to the regulations of the Guidance for Proper Conduct of Animal Experiments by the Ministry of Education, Culture, Sports, Science, and Technology of Japan. All experimental procedures were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals, and were approved by the Institutional Animal Use Committees of Kyoto University and Keio University.

Experimental details for injection of the labeling reagents, sample preparation, electrophysiology, and fluorescence imaging are described in SI Appendix.

H.N. and I.H. designed research; H.N., S.S., K.S., M.I., T.T., K.O., T.K., S.K., E.A.S., C.S., H.R.U., W.K., I.A., M.Y., and I.H. performed research; and H.N., S.S., K.S., T.T., S.K., W.K., I.A., and I.H. wrote the paper.

The authors declare no competing interest.