The fluorescence intensity of SEP is sensitive to pH (Miesenböck et al., 1998; Sankaranarayanan et al., 2000; Kopec et al., 2006); therefore, we first tested if SEP fluorescence lifetime also changes in response to different extracellular pHs. The GluA2 subunit of the AMPAR was tagged on the N-terminus with SEP, which resulted in the localization of the SEP fluorescent protein on the extracellular side of neurons and direct exposure to the imaging solution. Neurons expressing SEP-GluA2 were imaged in solutions with pH ranging from 6.2 to 7.5, and both fluorescence lifetime and intensity were measured in dendritic spines (Figure 1A). As expected, both the fluorescence intensity and lifetime were low at acidic pH, which results from fluorescence quenching (Lakowicz, 2006) and increased proportionally with increasing pH (Figures 1B, C). However, because of different expression levels of SEP-GluA2 between neurons, the coefficient of variation (CV) was 25–80 times higher in intensity-based measurements compared to lifetime measurements (Figures 1B, C). Next, we tested if SEP-GluA2 fluorescence lifetime is sensitive to AMPAR endocytosis by measuring fluorescence lifetime and intensity before and after insulin application, a manipulation shown to produce AMPAR internalization by inducing phosphorylation of three tyrosine sites in the C-terminal domain of GluA2 (Man et al., 2000; Ahmadian et al., 2004). We found that 0.5 uM insulin for 30 min reduced SEP-GluA2 fluorescence lifetime in dendritic spines (Figures 1D–F), consistent with the published effect of insulin on AMPAR endocytosis (Man et al., 2000; Ahmadian et al., 2004). The 3Y peptide, which includes the three unique GluA2 tyrosine residues, can serve as a substrate for Src kinase and effectively occlude GluA2 phosphorylation and endocytosis (Ahmadian et al., 2004). Accordingly, we did not observe any effect of insulin application in neurons incubated with the 3Y peptide (Figures 1D–F). Fluorescence intensity did not significantly change in the same spines (Figure 1E), demonstrating the power of fluorescence lifetime measurements to detect small changes.

Previous experiments showed that the expression of the amyloid precursor protein (APP) significantly reduced SEP-GluA2 intensity in spines and dendrites in hippocampal neurons (Hsieh et al., 2006). However, co-expression of a cytoplasmic red fluorescent protein was needed to normalize the SEP fluorescence signal and a large number of dendritic spines had to be measured (~1,000 individual data points; Hsieh et al., 2006). To test if our FLIM approach can detect Aβ-induced GluA2 endocytosis more effectively and reliably, we expressed SEP-GluA2 in two different preparations: hippocampal cultures and organotypic slices. We then measured the effect of expressing APP_CT100, the C-terminal fragment of APP that leads to elevated levels of Aβ, and compared it to a control virus, APP_CT84, a fragment of APP that does not change Aβ levels and was shown to have no effects on synaptic transmission (Reinders et al., 2016; Uyaniker et al., 2019; Figure 2A). We observed similar SEP fluorescence lifetimes in these two different preparations and a clear effect of APP_CT100 expression, consistent with GluA2 endocytosis in dendritic spines (Figure 2C). In contrast, we did not observe any reductions in fluorescence intensity in these same neurons (Figure 2B). This indicates that GluA2 is similarly trafficked to dendritic spines in hippocampal cultures and organotypic slices and that elevated Aβ equally induces the endocytosis of GluA2-containing receptors in these two different neuronal preparations. Moreover, this experiment demonstrates that FLIM leads to reproducible and consistent data without needing any normalization or adjustments.

AMPARs are composed of four subunits and are mostly found as pairs of GluA1 and GluA2 (GluA1/2) or GluA2 and GluA3 (GluA2/3) dimers in pyramidal neurons (Wenthold et al., 1996). To test if GluA1/2 or GluA2/3 receptors are differentially affected by elevated levels of Aβ, we expressed SEP-GluA1 and SEP-GluA3 along with either APP_CT100 or the APP_CT84 control (Figure 3A). Similarly, as with SEP-GluA2 (Figure 2), APP_CT100 did not lead to any significant reductions in fluorescence intensity in spines of neurons expressing SEP-GluA1 or SEP-GluA3 as compared with neurons expressing APP_CT84 (Figure 3B). Interestingly, in dendrites, a significant decrease in fluorescence intensity was seen in APP_CT100-expressing neurons for both SEP-GluA1 and SEP-GluA3 (Figure 3B). For GluA1-containing receptors, the effect of elevated Aβ on SEP-GluA1 fluorescence lifetime was very obvious both in spines and dendrites (Figure 3C). However, in SEP-GluA3-expressing neurons, compared to APP_CT84 control neurons, APP_CT100 expression induced only a small decrease in fluorescence lifetime in spines and no significant effect in dendrites (Figure 3C).

To gain more insight into AMPAR trafficking, we looked at the spine intensity to dendrite intensity ratio, which can be used as a proxy for receptor enrichment in spines (Kopec et al., 2006). Elevated Aβ significantly increased the spine/dendrite ratio in GluA3-containing receptors (Figure 3D), indicating that SEP-GluA3 fluorescence intensity in dendrites was significantly more reduced than in spines. In contrast, elevated Aβ did not affect GluA1 spine enrichment (Figure 3D). Comparing the spine/dendrite ratio for GluA1 and GluA3 in control conditions (APP_CT84), we found that GluA3-containing receptors were significantly more enriched in spines than GluA1-containing receptors, consistent with previous literature (Kopec et al., 2006). We also looked at the difference between spine and dendrite lifetime, which gives an indication of the relative proportion of surface receptors in spines vs. dendrites. For SEP-GluA1, this difference was positive in control conditions, suggesting that there are more GluA1-containing receptors at the surface in spines than in dendrites (Figure 3E). APP_CT100 caused a significant reduction in the difference in lifetime for SEP-GluA1, indicating an increase in internalized receptors in spines specifically (Figure 3E). For SEP-GluA3, the difference in lifetime was close to 0 and was not affected by APP_CT100 expression, suggesting that there is a similar surface/internalized ratio in spines and dendrites and that this ratio is not affected by elevated Aβ (Figure 3E).

To study the effect of Aβ on AMPAR trafficking in more detail, we used hippocampal cultures made from APP/PS1 mice and their WT littermates. APP/PS1 mice are double transgenic Alzheimer's disease (AD) model mice expressing human APP and presenilin 1 with familial mutations causing early disease onset (Jankowsky et al., 2004). In these AD model mice, spine density was shown to be significantly reduced in both cortical and hippocampal primary cultures, indicating early synaptic deficits (Priller et al., 2009; Kashyap et al., 2019). Since AMPAR removal most likely mediates synaptic loss in amyloidosis AD models (Hsieh et al., 2006), we assessed AMPAR trafficking in APP/PS1 neurons. To do so, we expressed SEP-GluA1, SEP-GluA2, or SEP-GluA3 in hippocampal cultures from either WT or APP/PS1 mice (Figure 4A). We found that the fluorescence intensity in spines and dendrites of neurons expressing SEP-GluA1, SEP-GluA2, and SEP-GluA3 was significantly reduced in APP/PS1 neurons compared to WT controls, indicating significantly lower amounts of all AMPARs (Figure 4B). Additionally, we found that the fluorescence lifetime in spines and dendrites for APP/PS1 neurons expressing SEP-GluA1 and SEP-GluA2 was significantly decreased (Figure 4C). This is consistent with relatively more GluA1/2 receptors in endocytic vesicles in APP/PS1 neurons compared to WT neurons. Surprisingly, for SEP-GluA3, we saw a significant fluorescence lifetime increase in dendrites and no change in spines (Figure 4C). This fluorescence lifetime increase suggests that there are more surface receptors and fewer internalized GluA3-containing receptors in APP/PS1 dendrites compared to WT neuron dendrites.

Using the spine/dendrite ratio to examine AMPAR enrichment in spines, we found that neurons expressing SEP-GluA2 had a significant decrease in this ratio for APP/PS1 compared to WT, indicating that the enrichment of GluA2-containing receptors was reduced in APP/PS1 neurons, possibly due to more endocytosis in spines (Figure 4D). We also looked at the difference between spine and dendrite lifetime, which is similar to what we found by acutely increasing Aβ with APP_CT100 in Figure 3. For SEP-GluA1, this difference was positive in WT neurons and changed to negative in APP/PS1 neurons. This suggests that there are more surface GluA1-containing receptors in the spines of WT mice compared to dendrites and more surface receptors in dendrites of APP/PS1 neurons (Figure 4E). For neurons expressing SEP-GluA2 or SEP-GluA3, this difference between spine and dendrite lifetime also decreased, suggesting reductions in surface receptors at spines, specifically in APP/PS1 neurons (Figure 4E).

For GluA1- and GluA2-containing AMPARs, we see consistent results using either APP_CT100 to express Aβ (Figures 2, 3) or in APP/PS1 neurons (Figure 4). However, the trafficking of GluA3-containing receptors is not clear and, importantly, seems inconsistent with a previous study indicating the requirement of the GluA3 subunit for Aβ-induced depression and memory deficits in APP/PS1 mice (Reinders et al., 2016). A possible explanation for this discrepancy is that GluA2/3 AMPARs are more susceptible to lysosomal degradation than GluA1/2 AMPARs (Kessels et al., 2009). Therefore, SEP-GluA3 subunits may be degraded too quickly for us to measure their trafficking to intracellular vesicles. In order to test this idea, we used leupeptin, a lysosomal protease inhibitor that prevents degradation of internalized AMPAR receptors up to 60 min after their endocytosis (Ehlers, 2000). APP/PS1 neurons expressing SEP-GluA1 or SEP-GluA3 were treated with leupeptin and compared with control neurons (Figure 4F). In APP/PS1 neurons expressing SEP-GluA1, leupeptin induced a significant increase in both fluorescence intensity and lifetime compared to untreated neurons (Figures 4G, H), which is consistent with increased amounts of surface GluA1-containing receptors both in spines and dendrites. Strikingly, we observed the opposite effect in neurons expressing SEP-GluA3. In APP/PS1 cultures treated with leupeptin, SEP-GluA3 fluorescence intensity in dendrites was decreased, as was SEP-GluA3 fluorescence lifetime in both spines and dendrites (Figures 4G, H). Therefore, by blocking lysosomal degradation, we were able to see results consistent with endocytosis of GluA3-containing AMPARs in APP/PS1 mouse neurons. Moreover, the fact that the same treatment instead increased surface GluA1-containing AMPARs suggests largely different degradation dynamics of GluA1- and GluA3-containing AMPARs, highlighting the opposite trafficking properties of these two types of AMPARs.

To further demonstrate the applicability of our approach to different contexts, we decided to test if the expression of PP1 had any effects on the trafficking of GluA1-containing AMPARs. PP1 is a protein phosphatase that participates in LTD (Mulkey et al., 1993; Morishita et al., 2001; Aow et al., 2015) and was recently found to be activated by APP_CT100 expression (Dore et al., 2021). During LTD, PP1 dephosphorylates GluA1 subunits at Ser845 (Lee et al., 2000), which is thought to induce its endocytosis (Guntupalli et al., 2017). To study how Aβ or PP1 affects synaptic transmission, we sparsely infected organotypic hippocampal slices with Sindbis viruses expressing APP_CT100 or the catalytic subunit of PP1 (PP1_cat, designated as PP1) and 18–24 h later obtained whole-cell recordings simultaneously from pairs of infected and uninfected neurons (Figures 5A, B, as in Dore et al., 2021). As expected (Kessels et al., 2013; Dore et al., 2021), neurons expressing APP_CT100 displayed depressed synaptic transmission compared to uninfected control neurons (Figure 5B). Interestingly, the expression of PP1 by itself or along with Aβ produced a similar effect (Figure 5B), suggesting that these processes may be mediated by common signaling pathways. Additionally, we found that PP1 expression led to a significant reduction in both SEP-GluA1 fluorescence intensity and lifetime, consistent with SEP-GluA1 endocytosis (Figures 5C–E). The ratio of the spine to dendrite intensity increased with PP1 expression, indicating that fluorescence intensity in dendrites was significantly more affected (Figure 5F). Accordingly, we also saw an increase in the difference between the spine and dendrite lifetime (Figure 5F), suggesting that more AMPARs are being removed from the dendrites compared to the spines. In these experiments, the PP1 protein expressed was tagged with mCherry, which could induce FRET between GluA1 and PP1 if these proteins were close enough. However, the SEP tag on GluA1 is extracellular and PP1 is a cytoplasmic protein, so this possibility is very unlikely. Nevertheless, to make sure that the change in SEP fluorescence lifetime is not due to proximity with mCherry, we expressed GFP-tagged GluA1 (at its N-terminus, exactly like the SEP tag) and the same PP1-mCherry (Figure 5G). We found that PP1-mCherry had no effect on GFP-GluA1 fluorescence intensity or lifetime, confirming that these proteins are not close enough to allow FRET between GFP and mCherry. This indicates that the changes we measured using SEP-GluA1 (Figures 5C–E) are due to the endocytosis of GluA1-containing AMPARs.

C57BL/6J mice were used for most experiments (#000664-JAX) and referred to as wild-type (WT) mice. For Figure 4, APP/PS1 Alzheimer's model mice (Jankowsky et al., 2004; #034832-JAX) and their WT littermates were used. The mice were kept in the UCSD School of Medicine facility on a 12-h light–dark cycle, given ad libitum access to food and water, and their genotypes were confirmed using genotyping PCR. All animal procedures were approved by UCSD's IACUC.

Primary hippocampal neuronal cultures were made using P0–P1 mice pups according to previously published protocols (Dore et al., 2021). Briefly, hippocampi were dissected using ice-cold dissection media, cut into small pieces using a scalpel, resuspended in dissociation media [dissection media with 2 mM L-cysteine hydrochloride (Sigma), 10 mg/mL of papain (Sigma), pH 7.4], and incubated for 12 min at 37°C. After cell filtration and precipitation, the neurons were resuspended in plating media [Neurobasal-A (Gibco), 10% FBS, 0.5% Pen/Strep (Gibco), and 0.25% GlutaMAX (Gibco)] at a concentration of 1–2 × 10⁶ cells/mL and plated on 18 mm PDL-coated glass coverslips (Neuvitro). At 7–10 DIV, the neurons were transfected using ~2 μg total DNA and ~4 μL of lipofectamine-2000 (Thermo Fisher) per well. SEP-GluA1 and SEP-GluA2 were gifts from Roberto Malinow (Addgene plasmids # 24000 and # 24002; RRID:Addgene_24000, RRID:Addgene_24002, respectively). SEP-GluA3 was a gift from Helmut Kessels (Renner et al., 2017). For Figures 2, 3, 18–24 h before imaging, the neurons were infected with 1–2 μL of Sindbis virus to express APP_CT84 or APP_CT100 as in Dore et al. (2021). For Figure 5, the PPP1CA sequence from the Addgene plasmid # 155843 (RRID:Addgene_155843, a gift from Eugene Yeo) was tagged with mCherry at its N-terminus and cloned into a PCI vector (for FLIM experiments) or a Sindbis vector (for electrophysiology) using Gibson assembly. GFP-GluA1 was also a gift from Roberto Malinow (Kopec et al., 2006). To account for biological variables, all experiments were conducted in at least three different neuronal culture preparations.

Organotypic hippocampal slices were prepared from P5 to P7 C57BL/6J mouse pups as described (Stoppini et al., 1991) and maintained for 8–12 days before infecting them with Sindbis viruses. The slices were infected in the CA1 region using an injection pipette and a Picospritzer 18–24 h before imaging (Figure 2) or electrophysiological recordings (Figure 5). For Figure 2, Sindbis viruses expressing either APP_CT84 or APP_CT100 were mixed with a SEP-GluA2 Sindbis virus (pSIN REP5-GFP-GluR2 (Q), a gift from Roberto Malinow (Addgene plasmid # 24003; RRID:Addgene_24003). For Figures 5A, B, the slices were infected with Sindbis viruses expressing either: APP_CT100, PP1-mCherry, or APP_CT100 and PP1 by means of a double promoter. Similarly, as with the experiments conducted in primary neurons, experiments in organotypic slices were replicated in three independent slice culture preparations.

FLIM was performed using a SliceScope Two-Photon Microscope (Scientifica, UK) with excitation from a Chameleon Ultra II IR laser tuned to 900 nm. Fluorescence emission was captured using a hybrid PMT detector (HPM-100-10, Becker and Hickl) and a GFP emission filter (ET 515/50, Chroma). The arrival time of each photon is calculated with a TCSPC (time-correlated single photon counting) module (SPM-150, Becker and Hickl). The FLIM acquisition software uses this information to construct a fluorescence decay trace, which indicates the number of photons detected in each time bin and is used to calculate the fluorescence lifetime (see Analysis section below). To minimize phototoxicity and photobleaching, a maximum of 3 mW of laser power was used for excitation, scanning speed was high (pixel dwell time of 3.2 μs) and image acquisition time was 120 s maximum per image. See Dore et al. (2021) for more information on FLIM acquisition and analysis parameters. Primary hippocampal neurons were imaged at 14–21 DIV in a circulating perfusion of an HBSS-based imaging solution comprised of: 0.87x HBSS, 10 mM HEPES, 2 mM Glucose, 1 mM MgCl₂, and 1.2 mM CaCl₂. For all experiments in this study, living neurons were used. For Figures 1A–C, HEPES was replaced with MES [membrane impermeable acid, (Sigma)] to obtain solutions with pH levels varying from 6.2 to 7.4. For experiments shown in Figures 1D–F, after imaging neurons in regular HBSS solution, perfusion was switched to the same solution supplemented with 0.5 μM insulin (Sigma). When indicated, the neurons were preincubated with the Tat-GluA2-3Y peptide (AnaSpec) for 1 h prior to imaging at a final concentration of 2 μM, and this peptide was also added to the imaging solution. Similarly, for experiments shown in Figures 4F–H, 20 μM leupeptin (AG Scientific) was preincubated for 3 h and also present during imaging.

Whole-cell recordings were performed in hippocampal organotypic slices infected with Sindbis viruses, as described above in the Organotypic Slices section. The slices were transferred to a recording chamber with a 1.5–2.0 ml/min flow of oxygenated artificial cerebrospinal fluid (aCSF), containing 119 mM NaCl, 2.5 mM KCl, 26 mM NaHCO₃, 1 mM NaH₂PO₄, 10 mM glucose, 4 mM CaCl₂, 4 mM MgCl₂, 10 μM gabazine, and 4 μM 2-chloroadenosine (pH 7.4). Healthy control and infected pyramidal neurons in CA1 were found using differential interference contrast and fluorescence microscopy. Pipettes with 3–5 MΩ resistance were filled with an internal solution containing (in mM): 115 cesium methanesulfonate, 20 CsCl, 10 HEPES, 2.5 MgCl₂, 4 Na₂ATP, 0.4 Na₃GTP, 10 sodium phosphocreatine, and 0.6 EGTA at a pH of 7.3 and utilized to obtain whole-cell recordings with an Axopatch-1D amplifier (Molecular Devices). A stimulating electrode [contact Pt/Ir cluster bipolar electrode (Frederick Haer)] was placed in Stratum Radiatum ~200–300 μm down the apical dendrite to evoke AMPAR-mediated excitatory post-synaptic currents (EPSCs) under voltage-clamp at a holding potential of −60 mV. Evoked responses of each recorded cell were analyzed using Igor Pro software, and amplitudes were averaged from 30 to 100 sweeps.

Fluorescence lifetime images were generated with SPCImage (Becker and Hickl), which calculates the fluorescence lifetime from the acquired fluorescence decay traces at each pixel. For all experiments, a binning factor between 2 and 6 pixels, a minimum threshold of 10 photons at the peak time bin, a single exponential model, and the same calculated instrumental response function were used. The FLIM images shown in all figures were processed in SPCImage and consisted of the fluorescence lifetime (color-coded value) merged with the intensity information (total number of photons) at each pixel. For further analysis, each FLIM image was exported as a matrix containing lifetimes, photon counts, and goodness-of-fit values (chi-square) and analyzed blindly to condition using a custom MATLAB script; refer to Dore et al. (2015) for details. All data are presented as mean ± standard error of the mean (SEM). Statistics were conducted using Prism 9 software. The unpaired t-test or the Kolmogorov–Smirnov test was used when comparing only two groups. In Figure 5B, a paired t-test was used. For all other experiments comparing multiple groups, a two-way ANOVA followed by Tukey's multiple comparison test was used to determine statistical significance.