Animal care and experiments were conducted in accordance with the European Communities Council Directive for the Care and the Use of Laboratory Animals (Regulation EU 2019/1010) and in compliance with the Ministère de l’Agriculture et de la Forêt, Service Vétérinaire de la Santé et de la Protection Animale (authorization number 01482.01). All efforts were made to minimize the number of animals used in the course of the study and to ensure their wellbeing. The animals were housed in a temperature-controlled room (21 ± 2°C) with ad libitum access to water and food under a 12 h light/dark cycle (lights on 7:30 a.m. to 7:30 p.m.).

Experiments were performed in two groups of animals aged of 2–3 months. The first group was used for the analysis of the localization of VAChT and VGLUT3 in CINs cell bodies and varicosities by immunohistochemistry on striatal sections. The second group was used for the visualization of SVs by scanning electron microscopy (SEM) and for the localization of VAChT and VGLUT3 by immunohistochemistry on striatal isolated SVs. This latter group was subdivided into two subgroups: wild type mice (genetic background C57BL/6) and VGLUT3 knockout mice (VGLUT3^–/–) (Gras et al., 2008).

Statistical analyses were performed using Graphpad Prism 9 (San Diego, CA 92108, United States). Data were tested for normality using D’Agostino-Pearson omnibus test. Comparisons were made by appropriate non-parametric tests. The Wilcoxon matched-pairs signed rank test compared the mean diameter of immunofluorescent spots for VAChT and VGLUT3 from confocal and STED microscopies. The same test was used to compare the mean diameter of immunofluorescent spots for VAChT and VGLUT3 with and without deconvolution. The Kruskal-Wallis test followed by the Dunn’s post-hoc test compared the number of VAChT, VGLUT3 or VAChT/VGLUT3 immunopositive spots per surface of microscope slide in isolated vesicles. The Kolmogorov-Smirnov test was used to compare the frequency distribution of the nearest neighbor distances (NNDs) for VAChT-VGLUT3 and VGLUT1-VGLUT3 or VGLUT2-VGLUT3-immunopositive spots. The significance level was set at 5%. The size and type of individual samples, n, for given experiments is indicated and specified in the results and extending data sections, in figure legends and in tables. Data were expressed as means ± SEM and p-values < 0.05 were considered as statistically significant. The experiments were repeated at least 3 times. Asterisks indicate p-values as follows: *p < 0.01, **p < 0.0001.

SEM was used to analyze the morphology of SVs and to check the absence of clusterization. After wash in PBS (see above), the vesicle solution (25 μg/300 μl PBS) was dropped off on glass slides (Ibidi 12 well chamber slide; Biovalley, France) where the vesicles settled at their surface for 30 min at 37°C. Then, SVs were fixed in 2% glutaraldehyde in cacodylate buffer 0.1 M. After washes in cacodylate buffer 0.1 M, SVs were post-fixed with 1% osmium tetroxide in 0.1 M cacodylate buffer for 10 min, washed and dehydrated in ascending concentrations of ethanol (50–100%), then in hexamethyldisilazane, before drying by evaporation under primary vacuum. SVs were covered by a thin layer (5 nm) of platinum using a Sputter Coater Leica EM ACE600 (Leica Microsystems, Wetzlar, Germany), in order to enhance both the contrast and the electrical conductivity of the samples. They were observed under a Field Emission scanning electron microscope GeminiSEM 500 (Zeiss microscopy, Jena, Germany) operating in high vacuum, at 3 kV, with a 20 μm aperture, high current mode, and a working distance around 2.5 mm. Secondary electrons were collected with the “inLens” detector. Images were acquired with a 1,024 × 768 definition, with a pixel dwell time of 1.6 μs and a line averaging of 10. The diameter of SVs was measured using the Fiji software.

Confocal and STED microscopy were used to codetect VAChT and VGLUT3 in striatal sections and in isolated SVs. Imaging was performed at the surface of brain sections (first 10 μM) where the tissue permeabilization is the most efficient. Confocal microscopy imaging was performed using a Leica TCS SP5 (Leica Microsystems) at the Imaging facility of the IBPS. For that, images were acquired in the dorsolateral striatum (DLS) using a hybrid detector (HyD) and a x63 oil immersion objective, NA 1.4, with a zoom x5. STED imaging was carried out with a Leica SP8—STED 3X microscope (Leica Microsystems) within NeurImag facility (IPNP). Sections were observed with a x93/1.3 NA glycerol immersion objective with two HyDs within DLS. AlexaFluor 594 and AbberiorStar 635P-coupled secondary antibodies were excited at 590 and 631 nm, respectively, with a pulsed white laser, and fluorescence was depleted using a pulsed depletion laser (775 nm). Typically, images of 1,024 × 1,024 pixels were acquired with a x5 magnification resulting in a pixel size in the range of 20–25 nm for tissue and 12–15 nm for isolated SVs. Using the STED microscope, the same fields may be observed alternatively with the confocal and with the STED modes. STED images were deconvolved with the Huygens 3.7 software (Scientific Volume Imaging, Hilversum, Netherlands), which permits the recovery of objects that are degraded by optical blurring and noise. Deconvolution allows also to improve the signal to noise ratio and the resolution. We used an automatically computing theoretical PSF based on the microscopic parameters including the Microscope Type, the numerical aperture, the lens and medium refractive indexes, the channel label, the excitation and emission wavelengths. We used the Classical Maximum Likelihood Estimation (CMLE) algorithm. Signal-to-noise ratio also called the R-parameter was set to a value of 2.6. The number of iterations was 25. The background intensity was averaged from the pixels with lowest intensity. We checked that deconvolution did not create artifacts and we measured the diameter of the spots obtained under the STED microscope before and after deconvolution (Supplementary Figure 1). For the purpose of minimizing noise, we applied a Gaussian 2D kernel with a Gaussian width, σ, to the raw and deconvolved images. In our case, we have used the Gaussian Blur filtering function in the Fiji compilation of ImageJ 1.47N with a Gaussian width of 1. The full width at half maximum corresponding to the diameter of the fluorescent spots was calculated with the Fiji plugin “plot profile.” Finally, images were analyzed using ImageJ and mounted with Adobe Photoshop, Adobe Indisign and Abobe illustrator.

Therefore, as expected, deconvolution increased the resolution of STED images without introducing major alteration of the fluorescent signal as previously shown (Wegel et al., 2016).

VAChT and VGLUT3 were first visualized with classical confocal microscopy in cell bodies and axonal varicosities of CINs (Figures 1A,B). Low magnification images showed that VAChT and VGLUT3 were codetected in the same perikaryia (Figure 1A) and varicosities (Figure 1B). Surprisingly, at higher confocal magnification, we noticed that VAChT and VGLUT3 immunolabeling mainly did not overlap in axonal varicosities (Figure 1C). To confirm this observation, the same varicosity was visualized with STED super-resolution microscopy (Figure 1C′). STED imaging bore out that VAChT-positive and VGLUT3-positive fluorescent puncta were largely non-overlapping (Figures 1C′,D,D′). Manual quantification confirmed that VAChT and VGLUT3 fluorescent spots were mostly dissociated. Only a very small proportion strictly overlapped and probably represent VAChT and VGLUT3 fluorescent spots whose centers are located ≈ 0–20 nm away from each other (1%, Figures 1E,F). This finding suggests that VAChT and VGLUT3 are frequently distant from each other. Similar results were found in the dorsolateral striatum, dorsomedial striatum and nucleus accumbens (data not shown).

In order to analyze more precisely the distances between VAChT and VGLUT3 in CINs varicosities from the dorsolateral striatum, we performed an automatic batch analysis of the NNDs between the centers of VAChT- and VGLUT3-immonopositive spots (Figure 1G). The frequency histogram revealed a peak with NNDs between VAChT and VGLUT3 centered around 40.0 ± 5 nm (Figure 1G). The size of a SVs is estimated to be ≈50 nm and the size of the complex composed of primary/secondary antibodies and a fluorochrome to ≈45 nm (Figure 1H; Früh et al., 2021). A cut-off of 95 nm was used to estimate the proportion of cholinergic VAChT-immunolabeled SVs containing VGLUT3 (< 95 nm) or not (> 95 nm) (Figure 1H). In striatal sections, 30% of VAChT-immunopositive spots were located between 0 and 95 nm from their closest VGLUT3 spot (Figure 1I). Our finding suggests that about one third of VAChT and VGLUT3 molecules could be located on the same SVs.

In striatal sections, SVs are very tightly packed. Different parameters including the lack of penetration of reagents in tissue and the steric hindrance of the primary/secondary antibody/fluorochrome complex (Figure 1H) may lead to misinterpretation of the precise localization of fluorescent spots. To overcome this difficulty and to better estimate the average NND between VAChT and VGLUT3 in less intricated conditions than in a varicosity, we then used purified SVs preparations from the mouse striatum. Previous studies suggested that purified SVs might be associated in clusters (Huttner et al., 1983). This clusterization could bias co-localization analyses in purified striatal SVs preparation. Thus, we first processed preparation of striatal SVs for SEM to analyze their dispersion. Under a SEM microscope, our preparations contained SVs that were not clustered (Figure 2A). The mean diameter of SVs was 49.7 ± 0.8 nm (Quantification from 484 SVs; Figure 2B). This result shows that our striatal SV preparations are adapted for colocalization analysis.

Then, we used immunofluorescent detection of VAChT and VGLUT3 with a confocal or STED microscope to assess their distribution in purified striatal SVs (Figures 2C,C′,D,E–G). As expected, relatively to confocal microscopy, STED microscopy increased the resolution of VGLUT3 immunofluorescent spots (Figures 2C,C′). Indeed, the diameter of VGLUT3-immunofluorescent puncta was 260 ± 3 nm at the confocal level compared to 60 ± 1 nm with STED microscopy (Figure 2D and Supplementary Table 1; Quantification from 110 SVs; Wilcoxon matched-pairs signed rank test, p < 0.0001). Similarly, the diameter of VAChT immunofluorescent spots significantly decreased from 173 ± 2 nm in confocal microscopy to 67 ± 1 nm with STED microscopy (Supplementary Figure 2 and Supplementary Table 1; Quantification from 110 SVs; Wilcoxon matched-pairs signed rank test, p < 0.0001).

The first step for analyzing the distribution of VAChT and VGLUT3 immunofluorescent spots from STED images included deconvolution. We checked that the deconvolution of STED images did not introduce artifacts (see section “Material and methods”; Supplementary Figure 1 and Supplementary Table 2). As expected, deconvolution increased the resolution of STED images without introducing major alteration of the fluorescent signal as previously reported (Wegel et al., 2016).

The next validation step included control for specificity of immunofluorescent labeling. When SVs were prepared from striatum of VGLUT3^–/– mice lacking VGLUT3 (Figure 2F) or when the anti-VGLUT3 antiserum was omitted (Figure 2G), the number of VGLUT3-positive spots was dramatically reduced (Figure 2H, Kruskal Wallis test, Dunn’s post-hoc test: p < 0.0001; Supplementary Table 3). From the quantification of VGLUT3 spots in VGLUT3^–/– mice, we estimated false positive to represent 5% of the total number of VGLUT3-positive spots. In these three conditions, the number of VAChT-immunopositive spots remained constant (Figure 2H and Supplementary Table 3). The specificity of VAChT labeling was assessed by the disappearance of the signal in VAChT null mice (Janickova et al., 2017). These data confirm the specificity of immunolabeling.

Manual quantification of STED images showed that only 6% of VAChT and VGLUT3 immunopositive spots were strictly superposed (Figure 2I). This observation suggests that, in purified vesicles, a minority of VGLUT3 and VAChT molecules are located at distance smaller than the resolution of STED microscopy (50 nm; Hell and Wichmann, 1994).

To better estimate the distribution of the distance between VAChT and VGLUT3 in striatal purified SVs (Figures 3A–D), we quantified NNDs between the center of the fluorescent spots for VAChT and their closest VGLUT3-fluorescent spot. Interestingly, the frequency histogram showed a first peak in a NND range between 0 and 120 nm with a maximum at 40 ± 5 nm (Figure 3A). In addition to this peak, we also noticed a more spread-out low frequency distribution above 120 nm of NNDs (Figure 3A). Therefore, if VGLUT3 and VAChT are present on the same SV, it can be estimated that the distance between immuno-labeled transporters could be constrained between 0 and 95 nm (Figures 1H, 3A–C, and see above for the choice of 95 nm as a cut-off). Quantification showed that 43% of VAChT-positive spots displayed a NND below 95 nm with their closest VGLUT3-immunopositive spot (Figures 3B,C,E). These data suggest that 43% of VAChT and VGLUT3 molecules could be constrained within a distance compatible with a localization of both transporters on the same SV.

We then wondered whether the large pic (0–120 nm) observed with purified SVs (Figure 3A) could be due or not to clusterization (Huttner et al., 1983). CIN varicosities have a mean diameter of 500 nm and contain and average of 10 VAChT-positive spots and 8 VGLUT3-positive spots (Figures 1C′,D′). If SVs originate from the same varicosity, all VAChT and VGLUT3 spots should present NNDs smaller than 500 nm (clusterization model Supplementary Figure 3A). Alternatively, if SVs in our preparations are independent or dispersed, in agreement with our EM observations (Figure 2A), VAChT-VGLUT3 NNDs are larger than 500 nm (dispersion model Supplementary Figure 3B). In order to validate the clusterized or the dispersed distribution of VAChT- or VGLUT3-immunospots we tested the clusterization model and the dispersion model (Supplementary Figure 3). We calculated, on isolated SVs preparations, NNDs for each VAChT-immunofluorescent spot with its first, second, third and fourth closest VGLUT3-immunofluorescent spots in a radius of 500 nm, using the DiAna plugin (Gilles et al., 2017). The clusterization model predicted that close to 100% of VAChT spots have a NND < 500 nm with the first to fourth closest VGLUT3 spot (Supplementary Figure 3A). With the dispersion model, decreasing proportions for NNDs < 500 nm from the first to fourth closest VGLUT3 spot would be expected (Supplementary Figure 3B). As shown in Supplementary Figure 3C, relatively to VAChT, 64% of the first closest VGLUT3 spots were located in a radius of 500 nm. The proportion of second, third, and fourth closest VGLUT3 spots are only 21, 7 or 3%, respectively, (Quantification from 1,506 pairs of VAChT-VGLUT3 spots, from 3 experiments). These data are thus in favor of the dispersion model. Therefore, in our SVs preparation, a majority of immunofluorescent spots correspond to transporters located in independent varicosities. This result is well in line with EM imaging of our SVs preparations.

In the striatum, VAChT and VGLUT3 are present in the same CINs axonal varicosities. In contrast, VGLUT1 or VGLUT2 are expressed in two independent sets of glutamatergic terminals and cannot be expressed by VGLUT3-positive terminals and thus by VGLUT3-positve SVs (Herzog et al., 2001). We therefore estimated NNDs between VGLUT1 and VGLUT3 and between VGLUT2 and VGLUT3. Relatively to VGLUT3-positive spots, both VGLUT1-positive and VGLUT2-positive fluorescent spots appeared evenly distributed in our striatal SVs preparations (Figures 3A,F,H). We observed that 13 and 5% of VGLUT1- or VGLUT2-fluorescent spots were less than 95 nm away from their closest VGLUT3-fluorescent spot (Figures 3G,I), respectively. Statistical analysis showed that the frequency distribution of the NNDs between VAChT and VGLUT3 immunofluorescent spots was significantly different from the one observed between VGLUT1 and VGLUT3 or between VGLUT2 and VGLUT3 immunofluorescent spots (Figure 3A; Kolmogorov-Smirnov test: p < 0.0001; Supplementary Table 4).

Altogether, these data suggest that a subset of cholinergic SVs co-express both VAChT and VGLUT3 whereas other SVs express only VAChT or only VGLUT3 (Figure 3J). Based on quantification from Figures 2H, 3A,E, we estimate that VAChT and VGLUT3 are both present in 34% of CINs SVs whereas VGLUT3 alone and VAChT alone are found in 26 and 40% of CINs SVs, respectively (Figure 3J; see Supplementary Data Sheet 1 for the method of calculation).

The limited spatial resolution (200 nm) of conventional fluorescent microscopic approaches is not compatible with precise imaging of proteins in organelles as small as SVs (50 nm). This limitation can be overcome by fluorescence super-resolution microscopies like STED that allow the detection of multiple proteins at a resolution compatible with the size of SVs (Hell and Wichmann, 1994; Choquet et al., 2021; Upmanyu et al., 2022). In this study, we used STED super-resolution microscopy to perform a detailed analysis of the relative distribution of VAChT and VGLUT3.

We observed that in striatal sections, VAChT and VGLUT3-immunofluorescent spots observed under STED microscopy, rarely overlap. Using the quantification of NNDs between VAChT and VGLUT3, we observed a broad peak of distribution of the NNDs between VAChT and VGLUT3 centered around 40 nm. Since the broad peak spans from 0 to ≈300 nm (Figure 1G), it is difficult to reliably estimate the size of this subpopulation of VAChT/VGLUT3-immunopositive SVs. This difficulty can be due to the fact that, inside CINs varicosities, SVs are tightly packed in SV clusters that may interfere with the analysis of the distances separating VAChT and VGLUT3. To overcome this difficulty, we used isolated striatal SVs and checked with SEM and STED (clusterization vs. dispersion model analysis), that our SVs were well separated.

Using the NND-based quantification between VAChT and VGLUT3 immunofluorescent spots, we observed that among all VAChT spots, 43% of them are present located at 0–95 nm with a maximal frequency value at 46 nm. This proximity between VAChT and VGLUT3 could be due either to transporter located on the same SV or on independent SVs. To examine this specific point, we applied NND analysis to VGLUT1/VGLUT3 and VGLUT2/VGLUT3 distributions; VGLUT1 and VGLUT2 being expressed in distinct axon terminals and SVs than VAChT and VGLUT3. The frequency distributions of the NNDs between VGLUT1 and VGLUT3 or between VGLUT2 and VGLUT3 is more evenly spread than the one between VAChT and VGLUT3. However, some pairs of spots for vesicular transporters were less than 95 nm away (13% for VGLUT1/VGLUT3 and 5% for VGLUT2/VGLUT3). Therefore, a small part of VAChT-spots and VGLUT3-spots within the 0–95 nm peak (5–13% of SVs) may correspond to transporters located on overlapping or very close SVs rather than on the same SV. Taking into account this “background labeling,” we estimate that ≈34% of CINS SVs express both VAChT and VGLUT3, ≈26% VGLUT3 alone and ≈40% VAChT alone (see Supplementary Data Sheet 1 and Figure 3J). Therefore, in theory, 40% of CINs SVs have the capacity to accumulate and release only ACh, more than a quarter only Glut and one third simultaneously ACh and Glut. This distribution is well in line with findings that cholinergic varicosities from the IPN differentially release ACh or Glut depending on their firing activity (Ren et al., 2011). Together, our findings reveal an unsuspected level of sophistication in the sorting of VAChT and VGLUT3 in CINs varicosities. This is interesting since CINs have different patterns of firing when they signal for reward (Atallah et al., 2014). Based on findings in the IPN, it can be proposed that CINs could preferentially release Glut during tonic activity (low frequency) and ACh and/or Glut during burst activity (high frequency). Importantly, ACh and Glut released by CINs have opposite effects on dopamine effluxes. Given the crucial role of dopamine to shape reward-guided behaviors (Morris et al., 2004; Cragg, 2006; Frahm et al., 2015), an important challenge for the coming years will be to identify when these different modes of cholinergic and glutamatergic transmission are used by CINs.

Previous reports show that the presence of VGLUT3 increases the vesicular accumulation of ACh in CINs through vesicular synergy (Gras et al., 2008; El Mestikawy et al., 2011). The current mechanistic explanation of vesicular synergy is based on an increased acidification of cholinergic SVs due to concomitant VGLUT3-dependent Glut uptake (Gras et al., 2008; El Mestikawy et al., 2011). Therefore, vesicular synergy implies that VAChT and VGLUT3 are expressed on the same population of SVs. Our present data bring further support and morphological basis for the current mechanistic understanding of vesicular synergy. Interestingly, we previously reported that VGLUT3, through vesicular synergy, was able to increase ACh vesicular loading by a factor of two to three-folds (Gras et al., 2008). Our data showing that only a third of SVs co-express VAChT and VGLUT3, suggest that vesicular synergy within this subpopulation of SVs could probably be up to a 10-fold increase. The existence and functions of such a subpopulation of VAChT/VGLUT3-positive SVs putatively containing very high amounts of ACh remains to be established.

Interestingly, a recent study using SVs from the whole brain demonstrated that only 1.5% of SVs express both VAChT and VGLUT3 (Upmanyu et al., 2022). Our data suggest that striatal SVs co-expressing VAChT and VGLUT3 are 20 times more abundant than in total brain extracts. It is well established that most of the brain ACh is found in the striatum. Furthermore, in this structure CINs form a dense network of varicosities and cholinergic fibers represent ≈ 15% of striatal terminal/varicosities. It is therefore not surprising that VAChT/VGLUT3-positive SVs are more abundant in the striatum than in total brain extracts. Interestingly, total brain SVs show a high level of colocalization of VGLUT1 with ZnT3 [the Zn²⁺ transporter (Upmanyu et al., 2022) and ZN²⁺ positive ions facilitate the vesicular accumulation of negatively charged glutamate (Upmanyu et al., 2022). This observation confirms and extends the concept that multiple transporters and accumulation of “counter ions” could be indispensable for vesicular synergy to occur (El Mestikawy et al., 2011). However, it should be mentioned that VGLUT-operated vesicular synergy has been proposed to increase ACh or DA vesicular concentration by augmenting the ΔpH (Gras et al., 2008; Hnasko et al., 2010), whereas ZnT3 and Zn²⁺ vesicular accumulation should accelerate glutamate uptake by increasing ΔΨ.