Introduction

Structural long-term potentiation (sLTP) is the structural basis of long-term potentiation (LTP) and plays a critical role in learning and memory^1–3. Upon electrical synaptic stimulation or glutamate uncaging, Ca²⁺ influx through NMDA-type glutamate receptors (NMDARs) into dendritic spines triggers diverse downstream signaling cascades, including Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) and extracellular signal-regulated kinase (ERK), which lead to the increase in spine volume and synaptic efficacy^4,5. sLTP has two distinct phases: it starts with a rapid and transient spine enlargement over the first few minutes (transient phase), followed by a sustained enlargement over hours (sustained phase)^3,6. Different pharmacological or genetic manipulations can selectively inhibit these two phases, indicating distinct induction mechanisms⁶. The sustained phase is known to be coupled with LTP while the physiological role of the transient phase is less clear^3,6. Nevertheless, the transient phase is associated with the rapid ultrastructural changes of PSD and surrounding membrane and thus may be important for the dramatic synaptic increase in the initial phase of sLTP⁷. In addition, the spine enlargement during the transient phase induces a presynaptic potentiation through mechanical force by pushing on the presynaptic bouton⁸.

During LTP and sLTP, various internal membranes are exocytosed in dendrites in an activity-dependent manner to add spine membrane area, increase surface glutamate receptors, and release plasticity-related peptides^9–13. Particularly, activity-dependent synaptic delivery of GluA1 subunit of AMPA-type glutamate receptor (AMPAR) is considered one of the major mechanisms to increase postsynaptic glutamate sensitivity during LTP^14–18. Several members of the postsynaptic exocytosis machinery have been identified, including the soluble NSF-attachment protein receptor (SNARE) proteins, complexin, synaptotagmins, and myosin Vb^12,19–24.

Among the molecules regulating intracellular trafficking, Rab GTPases constitute the largest Ras subfamily, with more than 60 members localized to distinct intracellular domains^25,26. As small GTPases, Rab proteins switch between two states, the guanosine-5’-triphosphate (GTP)-bound “active” state and the guanosine diphosphate (GDP)-bound “inactive” state. Specific guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) regulate the conversion between these two states²⁵. Once activated, Rab GTPases recruit diverse downstream effectors to coordinate intracellular transport during vesicle budding, movement, tethering, and fusion²⁷. Several members of Rab GTPases have been shown to coordinate AMPAR trafficking during synaptic plasticity²⁸. Rab8 and Rab11 regulate the exocytosis of AMPARs during LTP^29–31, whereas Rab5 drives the endocytosis of AMPARs during long-term depression (LTD)³². Rab4 is dispensable for LTP but is required for the maintenance of spine morphology²⁹. Rab4 also regulates both the fast and slow endosomal recycling to the plasma membrane^33–35. However, the function of other Rabs in synaptic plasticity is little known. Notably, Rab10 is expressed in dendrites^36–39 and has been implicated in Alzheimer’s disease resilience^40,41. Thus, Rab10 is potentially involved in membrane trafficking during synaptic plasticity.

In the present study, we developed a general design of FRET/FLIM-based sensors for various Rab proteins. Among these sensors, we further focused on the spatiotemporal dynamics of Rab10 and Rab4 activity during sLTP. We found that postsynaptic stimulation leads to the persistent inactivation of Rab10 and transient activation of Rab4 in spines undergoing sLTP. These Rab activity changes are dependent on NMDAR and CaMKII activation. Moreover, knock-down analyses demonstrate that Rab10 serves as a negative regulator of sLTP, whereas Rab4 contributes to the transient phase of sLTP. In addition, postnatal deletion of Rab10 from excitatory neurons enhanced sLTP and electrophysiological LTP at Schaeffer collateral pathway. Furthermore, Rab10 negatively regulates activity-dependent GluA1 trafficking into the stimulated spines during sLTP, while Rab4 positively controls this process. Therefore, our results suggest that Rab10 inhibits AMPAR trafficking during synaptic potentiation and NMDAR-dependent inactivation of Rab10 facilitates LTP induction. On the contrary, Rab4 promotes AMPAR trafficking during sLTP and NMDAR-dependent activation of Rab4 regulates the transient phase of sLTP.

Results

Highly sensitive and selective FRET sensors for Rab proteins

To image Rab signaling activity in single dendritic spines, we developed FRET sensors for Rab4, 5, 7, 8, and 10. The Rab sensors have two components: (1) the Rab protein tagged with a fluorescent protein that serves as a FRET donor (monomeric enhanced green fluorescent protein (mEGFP)-Rab or mTurquoise2-Rab) and (2) a Rab binding domain (RBD) from a specific effector protein that is tagged with two fluorescent proteins as FRET acceptors (mCherry-RBD-mCherry or mVenus-RBD-mVenus). When the Rab protein is activated, RBD and Rab increase their binding and thus increase FRET between the donor and acceptor fluorophores (Fig. 1a). The fraction of Rab bound to the RBD (binding fraction) was calculated by measuring the fluorescence lifetime of the donor^42,43.

Fig. 1

For Rab4, 5, 7, 8 and 10, we used Rabenosyn5 [439-503], EEA1 [36-128], FYCO1 [963-206], Rim2 [27-175] and Rim1 [20-277] as RBDs, respectively^44–47. To test the sensitivity and specificity of these Rab sensors, we took three approaches. First, we transfected wild type (WT)-Rab, dominant-negative (DN)-Rab, and constitutively active (CA)-Rab sensors in HEK 293T cells. As expected, DN-Rab and CA-Rab sensors displayed lower and higher binding fractions than the WT-Rab sensor, respectively, indicating the sensitivity of Rab sensors (Fig. 1b,c). Particularly, Rab4 and Rab10 sensors showed significantly different binding fractions between WT- and DN- or CA-Rabs (Fig. 1b,c). For Rab10 sensor, we used the mTurquoise2-mVenus pair instead of the mEGFP-mCherry pair since it reported higher binding fraction differences between WT- and DN-, as well as WT- and CA-Rab10 sensors in HEK 293T cells (Fig. 1b,c, Extended Data Fig. 1a-d)^48,49. Second, we coexpressed a Rab sensor with the corresponding Rab GAPs or GEFs to test whether the sensor could respond to the known upstream signaling^50–56. Indeed, these GAPs and GEFs respectively decreased and increased the activity of Rab proteins as reported by the sensor (Fig. 1d-i). Compared with Rab5 and Rab8, Rab4, Rab7 and Rab10 sensors displayed higher binding fraction changes in response to GAPs or GEFs (Fig. 1d-i). Third, we measured Rab activity change in response to N-Methyl-D-aspartic acid (NMDA) application in neurons⁶. We ballistically transfected rat organotypic hippocampal slices with each Rab sensor and imaged the proximal apical dendrites of CA1 pyramidal neurons. Bath application of NMDA (15 µM, 2 min) in zero extracellular Mg²⁺ triggered a robust activation of Rab sensors in the dendrites, suggesting that Rab sensors could report neuronal Rab activities (Fig. 1j). Nevertheless, these Rab sensors showed different activation kinetics upon NMDA stimulation (Fig. 1j and Extended Data Fig. 1e). Rab4 sensor was rapidly activated, peaked at 3 min and subsequently decayed (Fig. 1j and Extended Data Fig. 1e). On the contrary, Rab5, 7, 8 and 10 sensors displayed a gradually accumulated activation pattern, which peaked at 6-11 min and decrased afterwards (Fig. 1j and Extended Data Fig. 1e). Notably, Rab10 sensor was transiently inactivated in the first 2 min, possibly reflecting a fast response to NMDAR activation (Fig. 1j and Extended Data Fig. 1e). Overall, these results demonstrate that our Rab sensor strategy is generalizable to many Rab proteins.

Rab10 is persistently inactivated in the stimulated spines during sLTP

We biolistically transfected cultured organotypic hippocampal slices of rats with Rab10 sensor, and imaged the secondary apical dendrites of CA1 pyramidal neurons. mTurquoise2-Rab10, the sensor donor, was continuously distributed in the dendrites and spines, without showing an endosomal punctate pattern (Fig. 2a). To eliminate the effect of protein overexpression on its localization, we probed endogenous Rab10 by CRISPR-Cas9-mediated homology-directed repair (SLENDR) technique in vivo (Extended Data Fig. 1f-j)⁵⁹. Endogenous Rab10 displayed a heterougenous pattern: it is widely distributed along the dendrite and protrudes into the spines (Extended Data Fig. 1i). To elucidate the subcellular compartmantalization of Rab10, we further examined its colocalization with exogenously expressed endosomal markers in vivo (Extended Data Fig. 1j). Rab10 is significanltly overlapped with Rab7-labelled lysosomes, partially colocalized with Rab11-labelled recycling endosomes, but separated from Rab5-labelled early endosomes (Extended Data Fig. 1j).

Fig. 2

To characterize the spatiotemporal dynamics of Rab10 activity in single dendritic spines during sLTP, we combined two-photon glutamate uncaging with 2pFLIM^6,13,57,58, and imaged the secondary apical dendrites of CA1 pyramidal neurons expressing the Rab10 sensor at 25-27°C. Upon focal glutamate uncaging (0.5 Hz, 60 s) in zero extracellular Mg²⁺, a single spine underwent a rapid volume increase within the first few minutes (ΔV_transient = 275.9 ± 25.0%; Fig. 2a,e,f), which decayed to a smaller but sustained volume increase lasting more than 30 min (ΔV_sustained = 79.6 ± 6.7%; Fig. 2a,e,g), consistent with the previous studies^3,6,13,57,58. Concomitant with the spine enlargement, Rab10 showed a rapid and spine-specific decrease of activity in the stimulated spines, which lasted for more than 30 min (Fig. 2a-d). Neither the changes in spine volume, basal binding fraction nor the changes in the binding fraction of Rab10 sensor correlated with the initial spine size (Extended Data Fig. 2-e). To verify the specificity of the sensor response, we replaced the RBD of the Rab10 sensor with the RBD of Rab4 sensor (Rabenosyn5 [439-503], false acceptor). This false acceptor sensor did not show a significant change in FRET (measured as the binding fraction change) in the stimulated spines (Fig. 2c,d,f,g and Extended Data Fig. 3). In addition, the Rab10 sensor displayed a similar inactivation pattern at a near physiological temperature (33-35°C; Extended Data Fig. 4). In summary, our results demonstrate that Rab10 is persistently inactivated in the stimulated spines during sLTP, and this inactivation is compartmentalized in the stimulated spines.

The persistent inactivation of Rab10 during sLTP may be caused by the dilution of active Rab10 proteins due to the rapid spine enlargement. To examine this possibility, we compared the basal binding fraction of Rab10 sensor in spines and dendrites, and analyzed the subset of spines with lower Rab10 basal activity than the dendrite (Extended Data Fig. 5b-d). We found that Rab10 activity still decreased against the gradient at the spine neck upon sLTP induction (Fig. 2 and Extended Data Fig. 5d). In addition, binding fraction of Rab10 sensor is independent of donor intensity (Extended Data Fig. 5a). Thus, instead of passive signal dilution, the persistent inactivation of Rab10 requires active biological processes, presumably by translocating Rab10 positive vesicles out of the spine or directly inactivating Rab10 in the spine. To examine whether Rab10 distribution is changed during sLTP, we further analyzed the intensity of Rab10 donor in the stimulated spines and dendrites. Upon sLTP induction, Rab10 donor intensity significantly increased in the stimulated spines but remained unchaged in the dendrites (Extended Data Fig. 6a). The dramatic increase of Rab10 intensity in the stimulated spines was probably due to passive diffusion by spine enlargement. Therefore, we further analyzed the mean intensities of Rab10 donor in the stimulated spines and dendrites, which were similar prior and post sLTP induction (Extended Data Fig. 6b,c).

To further identify signaling pathways that inactivate Rab10 during sLTP, we applied pharmacological inhibitors targeting potential upstream components^4,57. Inhibition of NMDARs by 2-amino-5-phosphonopentanoic acid (AP5, 50 µM) completely abolished Rab10 inactivation and spine enlargement (Fig. 2c,d,f,g and Extended Data Fig. 7a,b). Application of CN21 (10 µM), a CaMKII inhibitory peptide^60,61, abolished Rab10 inactivation while partially attenuating the volume change (Fig. 2c,d,f,g and Extended Data Fig. 7c,d). In contrast, mitogen-activated protein kinase (MAPK)/ERK kinase (MEK) inhibitor U0126 (20 µM) had no effect on Rab10 inactivation, although it impaired sLTP during the sustained phase (Fig. 2c,d,f,g and Extended Data Fig. 7e,f). These results demonstrate that Rab10 is persistently inactivated in the stimulated spines during sLTP, and this inactivation is dependent on NMDARs and CaMKII but not on the MAPK/ERK signaling pathway.

Rab4 is transiently activated in the stimulated spines during sLTP

Next, we measured the spatiotemporal profile of Rab4 activity in dendrites during sLTP induced in single spines. Consistent with its localization in early and recycling endosomes^33–35, mEGFP-Rab4, the sensor donor, showed a punctate distribution pattern (Fig. 3a). In the basal state, Rab4 sensor activity is not correlated with the donor intensity, and was lower in the spines than dendrites (Extended Data Fig. 5e-g). Glutamate uncaging (0.5 Hz, 60 s) induced a transient and sustained volume increase in the stimulated spines of neurons expressing Rab4 sensor (ΔV_transient = 341.0 ± 29.4% and ΔV_sustained = 77.6 ± 8.0%; Fig. 3a,e-g). In contrast to the persistent inactivation of Rab10, Rab4 activity was transiently elevated in the stimulated spines and decayed, with no activity change in the adjacent spines or dendrites (Fig. 3a-d). This enhanced spine activity exceeds that of dendrite during sLTP, suggesting that the activation requires active processes (Extended Data Fig. 5h). Similarly to Rab10, we examined the intensity and mean intensity of Rab4 donor in the stimulated spines and dendrites. In the transient phase of sLTP, both the intensity and mean intensity of Rab4 donor significantly increased in the stimulated spines, suggesting a recruitment of Rab4 into the stimulated spines (Extended Data Fig. 6d-f). In contrast, there was no donor intensity or mean intensity change in the dendrites (Extended Data Fig. 6d-f). Moreover, the initial spine volume was not correlated either with the volume changes, the basal activity of Rab4 reported or the level of Rab4 activation reported by the sensor (Extended Data Fig. 2f-j). Replacing the RBD of Rab4 sensor with the RBD for Rab10 sensor (Rim1 [20-227], false acceptor) showed no change of the binding fraction in the stimulated spines during sLTP, indicating the specificity of the sensor response (Fig. 3c,d,f,g and Extended Data Fig. 8). In addition, the Rab4 sensor showed a similar transient activation pattern at a near-physiological temperature (33-35°C; Extended Data Fig. 9). Thus, our results demonstrate that Rab4 is transiently activated in the stimulated spines during sLTP, and this activation is compartmentalized in the stimulated spines.

Fig. 3

To identify signaling pathways that activate Rab4 during sLTP, we applied pharmacological inhibitors targeting putative upstream components. Inhibition of NMDARs by AP5 (50 µM) completely abolished Rab4 activation and spine enlargement (Fig. 3c,d,f,g and Extended Data Fig. 10a,b). Application of CN21 (10 µM) decreased Rab4 activity and volume changes both in the transient phase and sustained phase (Fig. 3c,d,f,g and Extended Data Fig. 10c,d). In contrast, MEK inhibitor U0126 (20 µM) had no effect on Rab4 activation, although it impaired sLTP during the sustained phase (Fig. 3c,d,f,g and Extended Data Fig. 10e,f). Altogether, Rab4 activation during sLTP is dependent on NMDARs and CaMKII, but not on the MAPK/ERK signaling pathway.

Disruption of Rab10 signaling enhances structural and electrophysiological LTP, whereas disruption of Rab4 signaling inhibits the transient phase of structural LTP

Given that Rab10 and Rab4 display opposing activity profiles during sLTP, we hypothesized that they would have opposite functions in spine structural plasticity. To test this hypothesis, we knocked down Rab10 or Rab4 by respective shRNA, and examined the effects of these manipulations on sLTP. We biolistically transfected cultured organotypic hippocampal slices of rats with scrambled shRNA control or shRNA against Rab10 or Rab4, together with mEGFP as the volume marker. Compared with scrambled shRNA control, knockdown of Rab10 had no effect on spine size, but increased spine density (Fig. 4a and Extended Data Fig. 11a,b,f,g,h,i). Knockdown of Rab4 had no effect on spine size or density (Fig. 4a and Extended Data Fig. 11a,b,e,g,h,i).

Fig. 4

We further induced sLTP by two-photon glutamate uncaging in single spines of neurons expressing mEGFP^6,62. Under control condition with scrambled shRNA, application of a train of glutamate uncaging pulses (0.5 Hz, 60 s) in zero extracellular Mg²⁺ induced a rapid spine volume increase in the transient phase, which decayed to a sustained enlarged volume for more than 30 min (Fig. 4a,c). However, knockdown of Rab10 by shRNA enhanced spine enlargement both in the transient and sustained phase of sLTP, which was rescued by co-expressing shRNA-resistant Rab10 (for scrambled shRNA, ΔV_transient = 215.5 ± 16.6% and ΔV_sustained = 64.4 ± 9.9%; for Rab10 shRNA, ΔV_transient = 309.0 ± 36.3% and ΔV_sustained = 140.6 ± 16.4%; for Rab10 rescue, ΔV_transient = 213.5 ± 19.7% and ΔV_sustained = 87.4 ± 11.7%; Fig. 4a,c,e and Extended Data Fig. 11a,b,d,f,g). In contrast, knockdown of Rab4 by shRNA significantly impaired the transient phase of sLTP while leaving the sustained phase intact (for scrambled shRNA, ΔV_transient = 291.6 ± 35.6% and ΔV_sustained = 76.1 ± 10.0%; for Rab4a/4b shRNA, ΔV_transient = 119.8 ± 15.4% and ΔV_sustained = 73.8 ± 12.1%; Fig. 4a,c,e and Extended Data Fig. 11a,b,e,g). This phenotype was rescued by coexpression of shRNA-resistant Rab4a (ΔV_transient = 223.1 ± 32.8% and ΔV_sustained = 84.2 ± 12.3%; Fig. 4a,c,e and Extended Data Fig. 11c). Overall, these results suggest that Rab10 negatively regulates both the transient and sustained phases of sLTP, while Rab4 is required for the transient phase of sLTP.

As an alternative strategy to inhibit Rab10 and Rab4 functions, we examined the effects of overexpressing DN-Rab mutants on spine structural plasticity. Consistent with the shRNA results, DN-Rab10 enhanced both the transient and sustained phase of sLTP (for control, ΔV_transient = 209.2 ± 31.1% and ΔV_sustained = 64.8 ± 9.4%; for Rab10 DN, ΔV_transient = 332.0 ± 31.8% and ΔV_sustained = 125.0 ± 17.4%; Fig. 4b,d,f), while DN-Rab4a selectively inhibited the transient phase of sLTP (for control, ΔV_transient = 357.4 ± 52.0% and ΔV_sustained = 83.5 ± 10.8%; for Rab4a DN, ΔV_transient = 198.1 ± 32.1% and ΔV_sustained = 70.8 ± 11.0%; Fig. 4b,d,f). Moreover, we evaluated the effect of overexpressing CA-Rab10 or CA-Rab4 on sLTP. These manipulations in general caused opposite phenotypes to the DN mutants: CA-Rab10 decreased both the transient and sustained phase of sLTP (for control, ΔV_transient = 209.2 ± 31.1% and ΔV_sustained = 64.8 ± 9.4%; for Rab10 CA, ΔV_transient = 97.7± 16.5% and ΔV_sustained = 23.0 ± 7.1%; Fig. 4b,d,f), while CA-Rab4 slightly increased the transient phase of sLTP (but not statistically significant; for control, ΔV_transient = 357.4 ± 52.0% and ΔV_sustained = 83.5 ± 10.8%; for Rab4a CA, ΔV_transient = 463.1 ± 61.1% and ΔV_sustained = 112.7 ± 18.3%; Fig. 4b,d,f).

We further evaluated the effects of Rab10 deletion on structural and electrophysiological LTP using Rab10 conditional knockout mice (Rab10^fl/fl)⁶³. For sLTP measurement, we biolistically transfected cultured organotypic hippocampal slices of Rab10^fl/fl mice with tdTomato-fused Cre recombinase and mEGFP, or tdTomato and mEGFP as a control⁶⁴. Consistent with the Rab10-knockdown results, deletion of the Rab10 gene increased sLTP in the stimulated spines of CA1 pyramidal neurons (Fig. 4g,h,i). Furthermore, we crossed Rab10^fl/fl mice with Camk2a-Cre mice (Rab10^fl/fl:Camk2a-Cre^+/-) to postnatally remove Rab10 from forebrain excitatory neurons⁶⁵. These animals showed enhanced LTP upon theta burst stimulation (TBS) at Schaeffer collateral synapses (Fig. 4k,l), with the basal synaptic transmission unchanged (Fig. 4j). Moreover, we monitored Schaffer collateral synaptic transmission in these mice, and found no difference in the amplitude of AMPAR- and NMDAR-EPSCs, or AMPAR/NMDAR EPSC ratio (Extended Data Fig. 12). These results indicate that Rab10 is a negative regulator for electrophysiological LTP.

Overall, our results demonstrate that Rab10 negatively regulates both the transient and sustained phase of sLTP, while Rab4 positively regulates the transient phase of sLTP. These functions are consistent with the direction and time window of their activity changes during sLTP. Moreover, Rab10 negatively modulates electrophysiological LTP.

Rab10 and Rab4 oppositely regulate activity-dependent SEP-GluA1 exocytosis during sLTP

We further examined whether Rab10 and Rab4 play roles in the exocytosis of GluA1-containing vesicles during sLTP⁶⁶. To visualize newly exocytosed AMPARs from the intracellular compartments during sLTP, we combined fluorescence recovery after photobleaching (FRAP) with two-photon glutamate uncaging and two-photon imaging^66,67. Organotypic hippocampal slices were biolistically transfected with N-terminal superecliptic pHluorin (SEP)-tagged GluA1, mCherry, and scrambled shRNA. CA1 pyramidal neurons expressing mCherry and SEP-GluA1 were imaged under two-photon microscopy. Since SEP-GluA1 is quenched in the acidic environment of endosomes, only the population on the surface emits fluorescence^66–68.

We pre-bleached surface SEP-GluA1 in a whole secondary dendrite with two-photon excitation, and measured the fluorescence recovery due to exocytosis in the spines after the induction of sLTP. Upon glutamate uncaging, the volume of the stimulated spines, measured with mCherry fluorescence, was increased by 352.8 ± 34.3% at 2 min (Fig. 5a,b). Meanwhile, the fluorescence intensity of SEP-GluA1 was rapidly recovered from 15.4 ± 1.6% (0 min) to 108.8 ± 10.8% (2 min) in the stimulated spines (Fig. 5a,b). However, in the non-stimulated adjacent spines and dendrites, SEP-GluA1 recovery was smaller and slower (for adjacent spines, from 17.3 ± 1.3% at 0 min to 28.9 ± 2.7% at 2 min; for dendrites, from 27.8 ± 1.4% at 0 min to 43.9 ± 1.9% at 2 min; Fig. 5a,b). Overexpression of tetanus toxin light chain (TeTxLC), which cleaves vesicle-associated membrane protein (VAMP)⁶⁹, significantly decreased both the spine enlargement and the SEP-GluA1 recovery in the stimulated spines, suggesting that the fluorescence recovery requires exocytosis (Fig. 5c-f and Extended Data Fig. 13a).

Fig. 5

To investigate whether Rab10 and Rab4 are involved in this activity-dependent postsynaptic AMPAR exocytosis, we knocked down endogenous Rab10 or Rab4 by respective shRNA, and monitored SEP-GluA1 exocytosis and sLTP in the stimulated and adjacent spines. No significant difference was seen in the adjacent spines or dendrites in either SEP-GluA1 recovery or mCherry intensity change among all groups (Extended Data Fig. 14a-h). Compared with scrambled shRNA, expression of Rab10 shRNA enhanced the spine enlargement as well as SEP-GluA1 incorporation (Fig. 5a,c-f and Extended Data Fig. 13d; measured at 2 min). These phenotypes were rescued by coexpressing shRNA-resistant Rab10 (Fig. 5a,c-f and Extended Data Fig. 13e; measured at 2 min). Therefore, Rab10 negatively regulates GluA1 exocytosis in the stimulated spines during sLTP. In contrast, the expression of Rab4a and Rab4b shRNAs impaired spine enlargement (Fig. 5a,e,f and Extended Data Fig. 13b; measured at 2 min). It also significantly attenuated SEP-GluA1 recovery in the stimulated spines (Fig. 5a,c,d and Extended Data Fig. 13b; measured at 2 min). These phenotypes were rescued by coexpressing shRNA-resistant Rab4a (Fig. 5a,c-f and Extended Data Fig. 13c; measured at 2 min). These findings suggest that Rab4 is required for GluA1 exocytosis in the stimulated spines during sLTP. Moreover, compensating for the effect of SEP-GluA1’s lateral diffusion by subtracting the change in the spine surface area (ΔVolume ^2/3) (Extended Data Fig. 14i)⁶⁶ did not alter our findings. Therefore, during sLTP Rab10 limits and Rab4 enhances SEP-GluA1 incorporation in the stimulated spines.

Discussion

The development of novel FRET-based biosensors for Rab proteins has revealed how Rab signaling pathways regulate sLTP in single dendritic spines. In brief, NMDAR activation triggers Ca²⁺ influx (∼ms) and CaMKII activation (∼s)^62,70, leading to the persistent inactivation of Rab10 (∼min) and transient activation of Rab4 (∼min) (Fig. 5g). Consistent with the directions and the duration of their activity change, Rab10 negatively regulates both the transient and sustained phase of sLTP and electrophysiological LTP. In contrast, Rab4 positively regulates the transient phase of sLTP. Thus, the temporal dynamics of Rab10 and Rab4 mirror the time courses of their functions in sLTP. Furthermore, Rab10 inhibits activity-dependent AMPAR insertion during sLTP, while Rab4 promotes this process (Fig. 5g). These results suggest that Rab4 and Rab10 play critical roles in two membrane trafficking events – AMPAR insertion and spine enlargement – during sLTP.

Previous studies have suggested that gating of signaling regulates kinase-phosphatase balance during LTP. For example, activation of the cyclic adenosine monophosphate (cAMP) pathway inactivates protein phosphatase, thereby gating CaMKII signaling^71–73. Here we demonstrated that Rab10 is inactivated during sLTP and plays inhibitory roles in AMPAR trafficking and spine enlargement, suggesting that Rab10 acts as a gate for the membrane trafficking events during sLTP. Thus, gating of inhibitory signaling is likely a common mechanism in synaptic plasticity.

On the other hand, Rab4 facilitates membrane expansion and AMPAR insertion, specifically during the transient phase of sLTP. In a previous study, LTP onset appears to be delayed in Rab4 knock-down neurons, implicating that Rab4 is required for the rapid increase of AMPAR current during LTP induction³⁰. Indeed, Rab4 is rapidly recruited into the spines in the transient phase of sLTP (Extended Data Fig. 6d-f). Consistently, recent ultrastructural analyses also demonstrated the increase of PSD complexity and membrane expansion during the transient phase of sLTP⁷. RhoA is another signaling molecule that plays a role specifically in the transient phase, likely by organizing the actin cytoskeleton in spines⁶. These signaling processes to rapidly reorganize spine structure during the transient phases would be critical for shaping the onset of synaptic transmission during LTP and sLTP.

Interestingly, Rab4, 5, 7, 8 and 10 sensors are all activated upon low-dose NMDA application (Fig. 1j and Extended Data Fig. 1e). However, they displayed distinct time courses and durations of activity change (Fig. 1j and Extended Data Fig. 1e). It is known that bath application of low-concentration NMDA induces LTD and dephosphorylation of AMPARs in the hippocampus⁸⁰. During LTD, the rate of AMPAR internalization outweighs the rate of AMPAR exocytosis, resulting in a reduced number of synaptic AMPARs. Since Rab proteins are localized on different endosomes and coordinate individual parts of receptor trafficking, their distinct time courses and activity durations may reflect the involvement of related endosomes in LTD. Particularly, Rab10 sensor displayed bi-directional activity changes in response to sLTP or chemaical LTD induction. To fully understand Rab10’s function in synaptic plasticity, it would be necessary to investigate its involvement in LTD.

Notably, the inactivation of Rab10 is persistent and lasts for over 30 min. Despite different time courses of activity during sLTP, other small GTPases RhoA, Cdc42, Rac1 and Ras also remain activated over 20 min^6,57,58. Moreover, BDNF-TrkB signaling is rapidly activated in the spines and remain elevated for at least 60 min during sLTP¹³. The persistent inactivation pattern of Rab10 is possibly defined by the activity of its associated endosomal organelle in sLTP. Previous studies showed that Rab10 is localized on various membranes, including the early endosome, recycling endosome, endoplasmic reticulum (ER), Golgi and trans-Golgi network (TGN). With SLENDR-mediated knockin techinique, we found that endogenous Rab10 is majorly overlapped with Rab7-labelled lysosome, partically colozalized with Rab11-labelled recycling endosome, but separated from Rab5-labelled early endosome in vivo (Extended Data Fig. 1j). These results implicate that Rab10 may mediate the transport from recycling endosome to lysosome for protein degradation. Therefore, inactivation of Rab10 possibly inhibits AMPAR degradation pathway, resulting in more available AMPARs for synaptic insertion and synaptic potetiation.

Pharmacological analysis of Rab10 and Rab4 signaling pathways indicated that they are downstream of CaMKII, but not ERK. Previously the Ras-ERK pathway has been implicated in regulating AMPAR delivery and sLTP^58,66,74. Our data suggest that Rab10 and Rab4 act as parallel pathways that are independent of Ras-ERK signaling. The steps between CaMKII activation and Rab10 inactivation or Rab4 activation are still unclear. One possibility is through the direct phosphorylation of Rab proteins^75,76, or indirect regulations by Rab GAPs or GEFs. Interestingly, in contrast to the Ras-ERK pathway, which shows extensive spreading into dendrites or nucleus^58,77,78, the activity of Rab10 and Rab4 is restricted to the stimulated spines during sLTP. Thus, Rab10 and Rab4 may regulate local membrane trafficking in spines, whereas ERK may regulate AMPAR exocytosis in dendrites⁶⁶, through different downstream effectors. Notably, a spine-restricted pattern of signaling activity has also been observed for Cdc42 and Cofilin activation during sLTP, which promotes the actin polymerization^6,79. Thus, the combination of local membrane trafficking and actin polymerization in spines appears to be important for spine expansion during sLTP.

Finally, our results highlight the diverse roles of Rab proteins in the orchestrated regulation of membrane trafficking during sLTP. It appears that Rab proteins operate with different directions and time windows. Whereas Rab10 negatively regulates sLTP for a long time (∼30 min), Rab4 positively regulates the initial ∼5 min of sLTP. Coordination of the upregulation and downregulation of the activity of various Rab proteins with unique functional and temporal properties would allow for the flexible and reliable control of membrane trafficking in various forms of spine structural plasticity. Although we measured the activity of only two Rab proteins in sLTP, it is likely that other members among the ∼60 Rab family proteins are critical for different spatiotemporal aspects of spine structural plasticity. Imaging the activity of other Rab proteins using similar Rab sensors will hopefully reveal the coordinated signaling that regulates membrane trafficking during synaptic plasticity.

Author contributions

JW, JN and RY designed experiments. JW performed most experiments. JN contributed to the general design and optimization scheme for Rab sensors. JW developed sensors for Rab4, 7 and 10. JN developed sensors for Rab5 and 8. PP performed electrophysiological recordings. EO performed structural plasticity measurement in Rab10 cKO mice. XL examined the localization of endogenous Rab10 by the SLENDR technique. TEM contributed Rab10 cKO mice. EMS performed western blot experiments. GO, TW and IS contributed technically. JW and RY wrote the manuscript with inputs from all authors.

Acknowledgements

This work was supported by grants from JSPS Overseas Research Fellowship (667 (JN)), NRF (NRFF12-2020-0008 (JN)), NIH (NS068410 (RY), MH080047 (RY), MH095090 (JN and RY)), HHMI (RY) and Max Planck Florida Institute for Neuroscience. We thank Drs. Sridhar Raghavachari, Scott Soderling, Fan Wang, and Anne West for important discussions. We thank Dr. Scott Soderling for SEP-GluA1 plasmid. We thank Dr. Boris Kantor and Marguerita Klein at Duke University Viral Vector Core for lentivirus production, Dr. Long Yan and light microscopy facility at Max Planck Florida Institute for technical supports, David Kloetzer for laboratory management and Dr. Lesley Colgan for critical reading of the manuscript. We thank all other Yasuda Lab members for discussions.

Methods

HEK 293T cells

HEK 293T cell lines were obtained from Fisher Scientific (Cat# NC0260915), and grown in DMEM (Gibco) supplemented with 10% fetal bovine serum (Invitrogen) and 1% penicillin-streptomycin (Invitrogen). The cell cultures were maintained at 37°C in a 5% CO₂ humidified atmosphere.

Rat

CD wild-type rats (CD IGS) purchased from Charles River Laboratories were used for preparation of hippocampal organotypic slice culture and dissociated postnatal cortical neuron culture. Both male and female animals were used and randomly allocated to experimental groups. All the experiments were performed in accordance with guidelines from the US National Institutes of Health, and were approved by Duke University Medical Center and the Institutional Animal Care and Use Committee of Max Planck Florida Institute for Neuroscience.

Mice

Rab10 conditional knockout (Rab10^fl/fl) mouse was a gift from Dr. Timothy E McGraw⁶³. Camk2a-Cre mouse was as previously⁶⁵. Swiss Webster mice were obtained from Charles River for endogenous Rab10 knockin experiments. Both of male and female mice were used. All the experiments were performed in accordance with guidelines from the US National Institutes of Health, and were approved by the Institutional Animal Care and User Committee of Max Planck Florida Institute for Neuroscience.

Organotypic slice culture

Organotypic rat or mouse hippocampal slices were prepared at postnatal day 6 or 7, as previously described⁸¹. Briefly, coronal hippocampal slices were dissected at 400 μm thickness using a McIlwain tissue chopper (Ted Pella, Inc). The slices were cultured on hydrophilic PTFE membranes (Millicell, Millipore), which were inserted in the culture medium containing 8.4 mg/ml MEM (Sigma), 20% horse serum (Gibco), 1 mM L-Glutamine (Sigma), 5.2 mM NaHCO₃, 12.9 mM D-Glucose, 0.075% Ascorbic acid, 30 mM Hepes,1 µg/ml Insulin, 1 mM CaCl₂ and 2 mM MgSO₄. The slice cultures were maintained at 35°C in a 5% CO₂ humidified atmosphere.

Dissociated neuron culture

Dissociated postnatal cortical cultures were prepared as previously published⁸². Briefly, cortices dissected from newborn rats (ramdom male and female) were triturated and plated into 5 cm dishes coated with 50 µg/ml PLL (Sigma) in culture medium consisting of Basal Medium Eagle (BME) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 35 mM glucose (Sigma), 1 mM L-glutamine (Sigma), 100 U/ml penicillin (Sigma), and 0.1 mg/ml streptomycin (Sigma). Neuron cultures were maintained at 35°C in a 5% CO₂ humidified atmosphere.

DNA constructs

To generate pCAGGS-mCherry, pCAGGS-mTurquoise2, pCAGGS-mCherry-mCherry, and pCAGGS-mVenus-mVenus, the respective fluorescence protein sequence was cloned into pCAGGS backbone from Raichu-2517KX gifted from Dr. Michiyuki Matsuda⁸³. Rat full length Rab4a, Rab5a, Rab7a, Rab8a, Rab11a and Rab10 were amplified by PCR from rat brain cDNA library (Dharmacon, Cat# LRN1205) and cloned into pmEGFP-C1⁶, pCAGGS-mEGFP, pCAGGS-mCherry and pCAGGS-mTurquoise2. All FRET donors were tagged at the amino terminus of Rab proteins. The linker between FRET donors (mEGFP or m Turquoise2) and Rab GTPases is SGLRSRG. Rabenosyn5 [439-503], EEA1[36-128], FYCO1[963-206], Rim2 [27-175] and Rim1 [20-227] cDNAs were amplified by PCR from rat brain cDNA library and inserted into pCAGGS-mCherry-mCherry and pCAGGS-mVenus-mVenus constructs. The linkers between mCherry and RBDs is SGLRSRA for the amino terminus and GSG for the carboxy terminus. The linkers between mVenus and Rim1 [20-227] is SGLRSRG for the amino terminus and GSG for the carboxy terminus. Rab dominant negative (DN) and constitutively active (CA) mutants were generated from wild type Rab GTPases by site-directed mutagenesis, and subcloned into pCAGGS-mEGFP, pCAGGS-3Flag and pCAGGS-mCherry constructs. PIK3R1 [114-313], Rin1, RNTre, RabGAP5, GRAB, Rabin8, TBC1D10 was amplified from rat brain cDNA library and cloned into pCAGGS-3HA construct. Full length Dennd4c, Evi5 and Rab3gap1 were amplified by PCR from MGC mouse cDNA (Dharmacon) and subcloned into pCAGGS-3HA construct. psiCHECK-2-Sal4-wt_3’UTR was a gift from Robert Blelloch (Addgene plasmid # 31862). psiCHECK-2-Rab GTPases were generated by inserting Rab GTPases into psiCHECK-2-Sal4-wt_3’UTR by XhoI/NotI.Tetanus toxin light chain⁸⁴ was subcloned into pCAGGS-3Flag construct. SEP-GluA1 was a gift from Dr. Scott Soderling at Duke University²³. mTurquoise2-pBAD and mVenus-pBAD were gifts from Michael Davidson (Addgene plasmid # 54844 and # 54845). The human codon-optimized S. pyogenes Cas9 (SpCas9) and single guide RNA (sgRNA) expression plasmid was a gift from F. Zhang (pX330, Addgene plasmid # 42230)⁸⁵.

RNA interference

For shRNA-mediated knock-down of Rab4 and Rab10, we used SHCLND-NM_009003 plasmid for Rab4a (Sigma-Aldrich, TRCN0000088975), SHCLND-NM_016154 palsmid for Rab4b (Sigma-Aldrich, TRCN0000380038), and TRC-Mm1.0 plasmid for Rab10 (Dharmacon, TRCN0000100838). The respective shRNA sequences (according to manufacture and sequencing confirmation) are CCGGAGATGACTCAAATCATACCATC TCGAGATGGTATGATTTGAGTCATCTTTTTTG for Rab4a, GTACCGGGGTCATCCTC TGTGGCAACAACTCGAGTTGTTGCCACAGAGGATGACCTTTTTTG for Rab4b, and T TGCCTTTCGGTACAACTCTC (mature antisense) for Rab10. For shRNA control, we used scrambled shRNA with the following sequence: CCTAAGGTTAAGTCGCCCTCG CTCGAGCGAGGGCGACTTAACCTTAGG (Addgene plasmid # 1864). To visualize transfected neurons in sLTP experiments, mEGFP was inserted into scrambled shRNA, Rab4a and Rab10 shRNA by KpnI/BamHI, and into Rab4b shRNA by BamHI/BstEII. The mEGFP expression was driven by a separate hPGK promoter (shRNA/mEGFP). For the rescue experiments, silent mutations of three amino acids were induced at the targeted region for Rab4a and Rab10 by site-directed mutagenesis (for shRNA-resistant Rab4a, AAAGATGACTCCAACCACACCATA; for shRNA-resistant Rab10, GAGAGTTGTGCCCAAGGGCAA).

Transfection of FRET sensors in HEK 293T cells and organotypic slice cultures

HEK 293T cells were transfected with Lipofectamine 2000 following the manufacture’s recommendations (Invitrogen), and imaged 24-48 hours after transfection. For Rab GTPase FRET sensors, the ratio of transfected FRET donor and acceptor was 1:3.

After 9-13 days in culture, organotypic hippocampal slices were transfected biolistically with gene gun⁸⁶ using gold beads (Bio-Rad, 1.6 µm) coated with plasmids, and imaged 3-4 days after transfection. For Rab4 FRET sensor, pmEGFP-Rab4a and pCAGGS-mCherry-Rabenosyn5 [439-503]-mCherry (1:1, 20 µg) were expressed for 3 days. For Rab10 FRET sensor, pCAGGS-mTurquoise2-Rab10 and pCAGGS-mVenus-Rim1 [20-227]-mVenus (1:3, 40-60 µg) were expressed for 3-4 days.

Two-photon fluorescence lifetime imaging and two-photon glutamate uncaging

We used a custom-built two-photon fluorescence lifetime imaging microscope (2pFLIM) with two Ti:Sapphire lasers (Chameleon, Coherent) as previously described^42,87. One laser was tuned to 920 nm to excite both donor for lifetime measurement and acceptor for morphology. The second laser was tuned to 720 nm for glutamate uncaging. The imaging power for two lasers was controlled independently by electro-optical modulators (Conoptics). The fluorescence was collected by an objective (60X, 1.0 numerical aperture, Olympus), separated by a dichroic mirror (Chroma, 565 nm for mEGFP/mCherry and 505 nm for mTurquoise2/mVenus), filtered by wavelength filters (Chroma, ET520/60M-2p for mEGFP, ET620/60M-2p for mCherry, ET480/40M-2p for mTurquoise2, ET535/50M-2p for mVenus), and finally detected by two independent photoelectron multiplier tubes (PMTs). We used 1.2-1.5 mW imaging power for mEGFP/mCherry sensor, and 1.6-1.8 mW for mTurquoise2/mVenus sensor.

Two-photon fluorescence lifetime imaging in HEK 293T cells was performed in imaging solution containing 20 mM HEPES (pH 7.3), 130 mM NaCl, 2.5 mM KCl, 2 mM MgCl₂, 2 mM NaHCO₃, 1.25 mM NaH₂PO₄ and 25 mM D-glucose.

Two-photon lifetime imaging and glutamate uncaging in organotypic slices was performed in Mg²⁺-free artificial cerebrospinal fluid (ACSF; 127 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 25 mM D-glucose, aerated with 95% O₂ and 5% CO₂) with 4 mM CaCl₂, 4 mM MNI-caged glutamate (Tocris) and 1 μM tetrodotoxin (TTX, Enzo). Uncaging pulses (0.5 Hz, 60 s, 4-6 ms, 3.5-3.8 mW) were delivered to the back focal aperture of the objective, which was around 0.5 µm from the tip of the spine head. We used a heater controller (Warner Instruments TC-344B) to monitor the temperature at 25-27°C or 33-35°C. For pharmacological experiments, inhibitors or vehicles were applied in ACSF 30 min before experiments. Please note that Fig. 2b,e include control samples from Fig. 2c-g, and Fig. 3b,e include control samples from Fig. 3c-g. Images were analyzed by MATLAB (MathWorks) and ImageJ.

2pFLIM data analysis

As described previously⁵⁸, to measure the donor fluorescence lifetime, we imaged a neuron expressing the donor, summed all pixels in the image, and fitted the fluorescence lifetime curve with a single exponential function convolved with the Gaussian pulse response function:

in which F₀ is the constant, and

in which F₀ is the peak fluorescence before convolution, t₀ is the time offset, τ_Dis the fluorescence lifetime of the free donor, τ_G is the width of the Gaussian pulse response function, and erf is the error function. We measured τ_D as 2.46 ns, 2.60 ns and 4.15 ns for the free mEGFP–Rab4a, mEGFP-Rab10 and mTurquoise2-Rab10 donors, respectively.

To measure the binding fraction of donor bound to its acceptor, we summed all pixels over a whole image and fitted the fluorescence lifetime curve with a double exponential function convolved with the Gaussian pulse response function:

in which P_D and P_,D are the fractions of free donor and donor bound with its acceptor, respectively, and τ_,D is the fluorescence lifetime of donor bound with its acceptor (1.10 ns for mEGFP/mCherry pair and 1.60 ns for mTurquoise2/mVenus pair).

For small regions of interest in an image, such as spines and dendrites, the binding fraction P_,D is calculated as follows:

In which τ_m is the mean fluorescence lifetime, τ_D and τ_,D are fixed values.

Spine volume measurement

To estimate the spine volume in neurons expressing Rab sensors, we measured the integrated fluorescence intensity of mCherry-RBD-mCherry or mVenus-RBD-mCherry in the spine, which is proportional to the spine volume⁸⁸, and normalized it by the fluorescence intensity in the thick apical dendrite from the same neuron. We further multiplied this normalized value by the volume of the point spread function, which gives the spine volume in fL^58,89.

NMDA application

Rat organotypic hippocampal slices (DIV 9-DIV 13) were ballistically transfected with indicated Rab sensors. After 3-4 day expression, CA1 pyramidal neurons were imaged in the basal solution (ACSF with 2 mM CaCl₂, 2 mM MgCl₂ and 1 μM TTX) for 6 min. NMDA (Tocris) was bath-applied in the zero Mg²⁺ solution (ACSF with 4 mM CaCl₂, 15 µM NMDA and 1 μM TTX) for 2 min, and replaced by the washout solution (ACSF with 2 mM CaCl₂, 2 mM MgCl₂, 1 μM TTX and 50 µM AP5) for 32 min.

sLTP measurement

Rat organotypic hippocampal slices were biolistically transfected with indicated constructs at days in vitro 9-13 (DIV 9-DIV 13). The constructs for Rab shRNA knockdown experiments were: scrambled shRNA/mEGFP (Ctrl shRNA); Rab4a shRNA/mEGFP and Rab4b shRNA/mEGFP (Rab4a/4b shRNA); Rab4a shRNA/mEGFP, Rab4b shRNA/mEGFP, and pCAGGS-mCherry-shRNA-resistant Rab4a (Rab4a rescue); Rab10 shRNA/mEGFP (Rab10 shRNA); Rab10 shRNA/mEGFP and pCAGGS-mCherry-shRNA-resistant Rab10 (Rab10 rescue). The constructs for DN- and CA-Rab overexpression experiments were: pCAGGS-mEGFP (Ctrl); pCAGGS-mEGFP and pCAGGS-mCherry-Rab DN (Rab DN); pCAGGS-mEGFP and pCAGGS-mCherry-Rab CA (Rab CA).The constructs were expressed for 4-5 days for Rab shRNA knock-down, and 2-3 days for DN- or CA-Rab mutant overexpression. CA1 pyramidal neurons were imaged in ACSF (aerated with 95% O₂ and 5% CO₂) with 4 mM CaCl₂, 4 mM MNI-caged glutamate and 1 μM TTX. Two-photon glutamate uncaging (0.5 Hz, 60 s, 4-6 ms, 3.5-3.8 mW) was performed at a single spine. All experiments were paired with the same day controls from the same batch of slices. The acquired images were analyzed by MATLAB (MathWorks).

Organotypic hippocampal slices of Rab10^fl/fl mice slices were biolistically transfected with tdTomato-Cre and mEGFP, or tdTomato and mEGFP as a control at DIV 10. At DIV 14, CA1 pyramidal neurons were imaged in ACSF (aerated with 95% O₂ and 5% CO₂) with 4 mM CaCl₂, 4 mM MNI-caged glutamate and 1 μM TTX. Two-photon glutamate uncaging (0.5 Hz, 60 s) was performed at a single spine. All experiments were paired with the same day controls from the same batch of slices. The acquired images were analyzed by MATLAB (MathWorks).

Spine size and spine density measurement

Rat organotypic hippocampal slices were biolistically transfected with indicated constructs at DIV 9. The constructs for Rab shRNA knockdown experiments were: scrambled shRNA/mEGFP (Ctrl shRNA), Rab4a shRNA/mEGFP and Rab4b shRNA/mEGFP (Rab4a/4b shRNA), or Rab10 shRNA/mEGFP (Rab10 shRNA). After 4 days expression, CA1 pyramidal neurons were imaged in ACSF by two photon microscopy at 25-27°C. All experiments were paired with the same day controls from the same batch of slices. The acquired images were analyzed by MATLAB (MathWorks) and ImageJ.

Activity-dependent SEP-GluA1 exocytosis

Rat organotypic hippocampal slices were biolistically transfected with indicated constructs at DIV 9-DIV 13. The DNA constructs for each condition were: pCAGGS-mCherry, SEP-GluA1, and scrambled shRNA (Ctrl shRNA); pCAGGS-mCherry, SEP-GluA1, and pCAGGS-3Flag-TeTxLC (TeTxLC); pCAGGS-mCherry, SEP-GluA1, Rab4a shRNA, and Rab4b shRNA (Rab4a/4b shRNA); pCAGGS-mCherry, SEP-GluA1, and Rab10 shRNA (Rab10 shRNA); pCAGGS-mCherry, SEP-GluA1, Rab4a shRNA, Rab4b shRNA, and pCAGGS-3Flag-shRNA-resistant Rab4a (Rab4a rescue); pCAGGS-mCherry, SEP-GluA1, Rab10 shRNA, and pCAGGS-3Flag-shRNA-resistant Rab10 (Rab10 rescue). After 4 days expression, CA1 pyramidal neurons were imaged in ACSF (aerated with 95% O₂ and 5% CO₂) with 4 mM CaCl₂, 4 mM MNI-caged glutamate and 1 μM TTX at 25-27°C. All experiments were paired with the same day controls from the same batch of slices. After taking five baseline images of a secondary apical dendrite (1 min interval) with 1.2-1.5 mW imaging laser power, we bleached the whole dendrite by increasing the imaging laser power to 4.0-4.5 mW for 2 min. We then performed two-photon glutamate uncaging (0.5 Hz, 60 s, 4-6 ms, 3.5-3.8 mW) at a single spine and collected eight images (1 min interval). The acquired images were analyzed by MATLAB (MathWorks) and ImageJ.

Validation of SLENDR-mediated Rab10 knockin

The SLENDR technique was as previously described⁵⁹. Briefly, Rab10 sgRNA (IDT) was ligated into the sgRNA scaffold of pX330⁹⁰ to generate Rab10 sgRNA-expressing plasmid. Rab10 sgRNA-expressing plasmid and single-stranded oligodeoxynucleotides (ssODNs; IDT) for homology-directed repair (HDR) were transfected into Neuro 2a cells by electroporation. Genomic DNA from the electroporation-transduced Neuro 2a cells was isolated with DNeasy Blood & Tissue Kit (Qiagen) following the manufacturer’s instruction. Genomic PCR was performed using extracted DNA as a template with corresponding primer set. The PCR product was purified by QiaQuick gel extraction kit (Qiagen) and then proceeded to Sanger sequencing to detect 2XHA insertion.

sgRNA target sequences (5’-3’)

Rab10: ACGTCTTCTTCGCCATTGGGAGG

Rab4a: GCGGAGCTGTGGCGGCAGAA

ssODNs sequence (5’-3’, upper case: 2XHA tag sequence) ggagttggttgtagtgagcagttccgatcccttggggctaccggcggcgagcgcccgagccgctcctcccaatgTACC CATACGATGTTCCAGATTACGCTTACCCATACGATGTTCCAGATTACGCTgcgaagaa gacgtacgacctgcttttcaagctgctcctgatcggggactcgggagtgggcaagacctgcgtc

Rab10 primer set, recombination

Rab10-2xHA-Fw: GCTCCTCCCAATGTACCCAT

Rab10-RV: AGAAACCGGATTCTGGAACG

Rab10 primer set, control

Rab10-FW: TTTCAAGCTGCTCCTGATCG

Rab10-RV: AGAAACCGGATTCTGGAACG

In Utero Electroporation and histology for SLENSR-mediated Rab10 knockin

In utero electroporation (IUE) was performed as previously described⁵⁹. Electroporation was performed at E13-E14 in anesthetized mice. For endogenous Rab10 localization, 1-2 μl mixture of plasmids (pX330-Rab10 sgRNA and pPB-CAG-mEGFP, 1μg/μl) and ssODNs (20 μM) was injected into the lateral ventricle of each pup. For colocalization examination of endogenous Rab10 and endosomal makers, 1-2 μl mixture of plasmids (pX330-Rab10 sgRNA and pCAGGS-mEGFP-Rab5a, pCAGGS-mEGFP-Rab7a or pCAGGS-mCherry-Rab11a, 1μg/μl) and ssODNs (20 μM) was injected into the lateral ventricle of each pup.

With ketamine-xylazine anesthesia (100 μg of ketamine -10 μg of xylazine per g of body weight, i.p.), mice were perfused with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, and the brain was fixed for 4–12 h. After wash with PBS, coronal vibratome sections (50 μm in thickness) were prepared (VT1200, Leica). For immunohistochemistry, slices were permeabilized with 0.3-0.4% Triton X-100 in PBS, blocked with 5% normal goat serum and 2% BSA or 5% normal donkey serum in PBS, and incubated overnight with rabbit anti-HA primary antibody (1:1000, Cell Signaling Technology). After 1–3 h incubation with Alexa Fluor-conjugated secondary antibodies (Invitrogen) followed by DAPI staining (0.1 μg/ml, Life technologies), slices were imaged using a confocal laser-scanning microscope (LSM880 with Airyscan, Zeiss). Images were processed and analyzed using the Zen (Zeiss) and the Fiji softwares.

Dual-luciferase reporter assay

HEK 293T cells were plated into 24-well plates and cotransfected with psiCHECK-2-Rab GTPases and the respective shRNA at a 1:3 ratio. For positive control, we used shRNA against hRluc with the following sequence: TCATAGTAGTTGATGAAGGAG (mature antisense). After 48 hours transfection, luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega) following the manufacturer’s protocol. After removing the culture medium, cells were briefly rinsed in pre-warmed 0.1M phosphate-buffered saline (PBS), and lysed in 100µL of 1X passive lysis buffer in the luciferase assay kit. After gently shaking for 15 min at room temperature, samples were prepared in a 96-well plate for luminescence measurement using the GloMax-Multi Detection System (Promega). For data analysis, the hRluc (firefly luciferase) luminescence was normalized by the hluc+ (Renilla Luciferase) luminescence in each well to control for transfection efficiency. All experiments were paired with the same day controls from the same batch of HEK 293T cells.

Lentivirus infection in dissociated neuron cultures

Dissociated postnatal cortical cultures were prepared as previously published⁸². Proliferation of non-neuronal cells was inhibited by adding Cytosine arabinoside (2.5 μM) at DIV 2. At DIV 6 cultures were infected with lentiviral particles containing Rab4a shRNA/mEGFP (Rab4a shRNA); Rab4b shRNA/mEGFP (Rab4b shRNA); Rab4a shRNA/mEGFP and Rab4b shRNA/mEGFP (Rab4a/4b shRNA); Rab10 shRNA/mEGFP (Rab10 shRNA) or scrambled shRNA/mEGFP (Ctrl shRNA). At DIV 17 cells were washed with ice-cold PBS and immediately extracted with ice-cold T-PER protein extraction buffer (Pierce) supplemented with inhibitors for proteases and phosphatases (Roche). The lysates were centrifuged at 15000 g for 15 minutes at 4°C and the supernatants were used for further analysis.

SDS-PAGE and immunoblotting

Samples were prepared for standard SDS-PAGE and separated on 12% acrylamide gel (Mini-PROTEAN TGX precast gels, Bio-Rad), then transferred onto 0.2 µm pore size PVDF membranes (Millipore) using semi-dry immunoblotting (transfer buffer containing 25 mM Tris, 200 mM glycine and 20% methanol). Membranes were blocked with 5% nonfat milk (Great Value) in TBS-T (Tris Buffered Saline with 0.1% Tween-20) for 1 hour at room temperature, then incubated overnight at 4°C with primary antibodies diluted in 5% BSA in TBS-T. We used the following commercially available antibodies: rabbit anti-Rab10 (1:500; Cell Signaling Technology), rabbit anti-Rab4b (1:500; ThermoFisher Scientific), mouse anti-Rab4a (1:500; ThermoFisher Scientific) and mouse anti-β-actin (1:2000; Sigma). Membranes were washed 3 times for 15 minutes in TBS-T, followed by incubation for 2 hours at room temperature with HRP-conjugated goat anti-rabbit or goat anti-mouse secondary antibodies (Bio-Rad), diluted 1:5000 in 5% nonfat milk in TBS-T. Membranes were washed 3 times for 15 minutes in TBS-T, then incubated with Pierce ECL Plus western blotting substrate (for Rab10, Rab4a and Rab4b) or Pierce ECL western blotting substrate (for β-actin) for detection of western blotted proteins. We used the Image Quant LAS4000 Imaging System (GE Healthcare) to visualize protein bands.

Acute slice preparation

Rab 10 Cre – and Cre + littermate mice (P 30-P 50) were anesthetized by isoflurane inhalation and perfused intracardially with a chilled choline chloride solution. The brain was removed and placed in the same choline chloride solution composed of 124 mM Choline Chloride, 2.5 mM KCl, 26 mM NaHCO₃, 3.3 mM MgCl₂, 1.2 mM NaH₂PO₄, 10 mM Glucose, and 0.5 mM CaCl₂, pH 7.4 equilibrated with 95% O₂ and 5% CO₂. Coronal slices (300 µm) containing the hippocampus were cut using a vibratome (Leica) and maintained in a submerged chamber at 32 °C for 1 h and then at room temperature in oxygenated ACSF.

Electrophysiology

Slices were perfused with oxygenated ACSF containing 2 mM CaCl₂, 2 mM MgCl₂ and 100 µM picrotoxin. One glass electrode (resistance ∼4 MΩ) containing the same ACSF solution was placed in the dendritic layer of CA1 area (∼100–200 µm away from the soma) while stimulating Schaffer Collateral fibers with current square pulses (0.1 ms) using a concentric bipolar stimulation electrode (FHC). The initial slope of the fEPSP was monitored with custom software (Matlab). The stimulation strength was set to ∼50% saturation. A 20 min stable baseline was first recorded before induction of LTP. LTP was induced by applying 5 trains of TBS. One TBS train consists 10 bursts of 4 stimulations at 100 Hz. The inter-burst interval is 200 ms and the interval between trains is 2 s. fEPSPs responses were recorded for an hour after the stimulation protocol. All data were analyzed with an in-house program written with Matlab. Data are presented as mean ± SEM.

Whole cell recording

Animals were sedated by isoflurane inhalation, and perfused intracardially with ice-cold choline chloride solution containing 124 mM choline chloride, 2.5 mM KCl, 26 mM NaHCO₃, 3.3 mM MgCl₂, 1.2 mM NaH₂PO₄, 10 mM Glucose and 0.5 mM CaCl₂ (pH 7.4 equilibrated with 95% O₂ and 5% CO₂). Brains were then removed and placed in the same chilled choline chloride solution and coronal acute slices of 300 µm from left and right hemispheres were collected and placed in oxygenated (95% O₂ and 5% CO₂) ACSF (in mM: NaCl 127, KCl 2.5, Glucose 10, NaHCO₃ 25, NaH₂PO₄ 1.25, MgCl₂ 2, CaCl₂ 2 mM) at 32°C for 1 h and maintained at room temperature for the rest of the experiment. Whole cell voltage clamp recordings of hippocampal neurons of Cre+ and Cre-Rab10 animals were made with a Multiclamp 700B. Patch pipettes (3-6 ΩM) were filled with a Cs Methanesulfonate solution (in mM: Cs MeSO₄ 120, NaCl 5, TEA-Cl 10, HEPES 5, QX314-Br 5, EGTA 5, NaATP 4, MgGTP 0.3, pH 7.4). Experiments were performed at room temperature (18-20°C) and slices were perfused with oxygenated ACSF (in mM: NaCl 127, KCl 2.5, Glucose 10, NaHCO₃ 25, NaH₂PO₄ 1.25, MgCl₂ 2, CaCl₂ 2 mM, picrotoxin 100 μM). AMPAR and NMDAR evoked responses were measured clamping the cells at -70 mV and +40 mV, respectively. Postsynaptic currents were evoked by electrical stimulation using a concentric bipolar electrode (THC), with a pulse of 0.1 ms. Input-output curves were first assessed by changing the stimulation intensity from 0 to 200 μA. For the rest of the recordings, the stimulation intensity was set to the amplitude that elicited a 50% of EPSC response. Recordings were filtered at 2 kHz and digitized at 10 kHz. Series and input resistances were monitored throughout the experiment. All data were acquired and analyzed with an in-house program written in Matlab. Data are presented as mean ± SEM.

Quantification and statistical analysis

Results are reported as mean ± SEM. Statistical analysis was performed with GraphPad Prism 7 and 10. Comparisons between two groups were performed using unpaired two-tailed Student’s t-tests (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001). Comparisons for more than two groups were calculated using one-way ANOVA followed by Bonferroni’s multiple comparison tests or two-way ANOVA (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001).

Extended Data

Extended Data Fig. 1

Extended Data Fig. 2

Extended Data Fig. 3

Extended Data Fig. 4

Extended Data Fig. 5

Extended Data Fig. 6

Extended Data Fig. 7

Extended Data Fig. 8

Extended Data Fig. 9

Extended Data Fig. 10

Extended Data Fig. 11

Extended Data Fig. 12

Extended Data Fig. 13

Extended Data Fig. 14