Synapses pass information from one neuron to another, but their performance is idiosyncratic and unreliable. It was long recognised from electrophysiological work that isolating single synapses reveals variability that is otherwise lost when population responses are measured. In electrophysiology, separating out individual synaptic responses can be achieved either by reducing release probability (del Castillo and Katz, 1954), by blocking all other responses (McAllister and Stevens, 2000), using minimal stimulation (Isaac et al., 1996), or by examining connected pairs of neurons (Vyleta and Jonas, 2014) with time resolution in the ms range. In contrast, optical report of neurotransmission via membrane dyes (Griesinger et al., 2005), calcium dyes (Oertner et al., 2002; Enoki et al., 2009), voltage-sensing dyes (Popovic et al., 2015), or neurotransmitter-binding reporters offers direct spatial resolution of synaptic events, usually with a time resolution about two orders of magnitude slower. Quantitation and analysis of the relevant optical signals remain challenging (Helassa et al., 2018; James et al., 2019; Soares et al., 2019; Tagliatti et al., 2020).

Synapses are diverse in their composition and size (Jontes and Smith, 2000), and it is assumed that these architectural features correspond to functional differences (Cizeron et al., 2020). In principle, accessing individual synaptic inputs to a neuron in space and time should allow correlation of molecular architecture with functional synapse properties and thus reveal synaptic diversity, but functional properties might be less diverse than synapse composition itself (Farsi et al., 2021). Plastic changes in synaptic responses, generated by sensory input (Gambino et al., 2014), are widely taken to represent learning and memory mechanisms, and might in turn increase diversity. However, synaptic contributions to sensory selectivity seem to depend most on the number of synapses that are activated (Scholl et al., 2020) suggesting that synaptic diversity may not serve immediate functional purposes. Instead, because synapses use small numbers of molecules, and tend to operate randomly (Ribrault et al., 2011), diversity may be needed to provide robustness through redundancy. The lack of reliability of certain synapses may be key to their computational power and role in decision making (Evans et al., 2018). These observations strongly motivate us to develop and test optical tools that can report activity with individual synapse resolution, and with a dynamic range sufficient to discern different modes of release, and failures to release.

Early generations of genetically encoded reporters of glutamate release relied on FRET (Okumoto et al., 2005; Namiki et al., 2007). The advent of bright, single-wavelength genetically encoded reporters that directly bind neurotransmitters, such as iGluSnFR, have expanded the scope of optical report of synaptic transmission by speeding up report to the limit of optical microscopy (Marvin et al., 2013). This class of sensor has also been adapted to fold more quickly (using superfolder Green Fluorescent Protein (GFP)), and offer affinity and colour variants (Marvin et al., 2018; Helassa et al., 2018). For the purpose of subcellular targeting, the compact sensor architecture of iGluSnFR is easy to transplant into other host proteins.

Fusion of fluorescent probes to proteins with selective, organellar targeting allows their subcellular expression. For example, targeting of probes to synaptic vesicles (such as Synaptophluorin; Miesenböck et al., 1998) has enabled quantitation of vesicle fusion and recovery with high specificity (Balaji and Ryan, 2007; Chanaday and Kavalali, 2018) albeit with slow off-kinetics (Sankaranarayanan et al., 2000) compared to vesicle fusion. On the other hand, most genetically-encoded sensors of neurotransmitters themselves lack targeting to synaptic sites. Even when signals from the latter are localised to putative dendritic spines, the unknown proximity of the sensor to the release site introduces unwanted variability (Soares et al., 2019), because the glutamate signal diverges rapidly in space and time due to diffusion (Raghavachari and Lisman, 2004). Where spatial confinement of the reporter was achieved by exploiting targeting (Kim et al., 2020) or anatomy (Helassa et al., 2018; James et al., 2019; Duerst et al., 2020; Jensen et al., 2019; Mendonça et al., 2022), the quantitative improvement of the fluorescent report is striking.

Here, we report that fusion of iGluSnFR to the glutamate receptor auxiliary proteins γ-2 or γ-8 produces postsynaptic-targeted sensors with improved spatial resolution. The SnFR-γ2 reporter gives a high-contrast signal through enrichment at glutamatergic synapses, and surprisingly also shows more stable activity. Spontaneous and evoked release can be resolved at up to 5 Hz, allowing dissection of the presynaptic contribution to short-term plasticity and release parameters in cultured neurons at the single synapse level.

Several groups have pursued mutagenesis of the Glt1 glutamate-binding domain to reduce the affinity of iGluSnFR for glutamate (Helassa et al., 2018; Marvin et al., 2018), with aim of improving its spatial and temporal resolution. We examined several mutants using patch-clamp fluorometry (Figure 1—figure supplement 1). A particularly interesting mutant that has not been previously reported is the Y230F substitution that removes a hydroxyl group that coordinates glutamate in the Glt1 domain. This mutation increased ∆F/F, the fractional fluorescence change upon glutamate binding (possibly by reducing fluorescence in the absence of glutamate), and sped the kinetics of the off-response (Figure 1—figure supplement 1). As expected, the Y230F mutation also reduced the apparent affinity for glutamate. To assess the kinetics of this mutant in synapses, we turned to autaptic microisland neuronal cultures. In this preparation, a single patch-clamp electrode can produce escaping action potentials that trigger glutamate release from synaptic terminals and record the resulting postsynaptic currents. However, we were surprised to discover that the Y230F mutant gave almost no fluorescent response in autaptic hippocampal neurons following stimulation (Figure 1—figure supplement 2), whereas in the same cultures, the regular iGluSnFR gave robust fluorescent responses, phase-locked to the postsynaptic currents, as did Synaptophluorin (with about 100× slower off kinetics). Paradoxically, the Y230F was expressed on the neuronal surface and responded robustly to 10 mM glutamate perfusion over the autaptic neuron (Figure 1—figure supplement 2). Combined with the observation that the fluorescence from unmodified iGluSnFR on average occurs over a spatial extent much larger than an individual synapse (~10 µm², see below) this observation led us to hypothesise that the low-affinity sensor was on average not sufficiently close to the synaptic glutamate signal to properly report it. Other low-affinity sensors have been reported (Helassa et al., 2018; Marvin et al., 2018) but as outlined above, their localisation in relation to postsynaptic densities (and therefore the signal they report) is uncertain.

iGluSnFR is membrane targeted through a single pass pDisplay peptide (Marvin et al., 2013), which includes no localisation signal to concentrate sensors at or near synapses, even if its responses themselves provide a spatial readout of glutamate release. To address this aspect, we fused the iGluSnFR extracellular domain to AMPA-type glutamate receptor in order to target it to excitatory synapses. Inline fusion at a permissive site in AMPA receptor subunits GluA1 and GluA2 extracellular region produced a sensor with a very small ∆F/F (Figure 1—figure supplement 3). Instead, we turned to the auxiliary proteins (TARPs) that decorate the periphery of AMPA receptors. To place the iGluSnFR sensor domain on the extracellular side of the membrane we fused a single pass transmembrane helix (from the NETO2 kainate receptor auxiliary protein; Zhang et al., 2009 to avoid any adventitious competition against the AMPA receptor–TARP interaction) to the N-terminus of Stargazin (γ-2) or γ-8 (Figure 1A). These chimeric sensors had similar performance to the original iGluSnFR and associated normally with AMPA receptor complexes in HEK cells (Figure 1), also acting to modulate AMPA receptor gating. The two reporters had similar apparent affinity for glutamate (within a factor of 2) to iGluSnFR, quite distinct from published ‘low-affinity’ variants (Helassa et al., 2018). We named these reporters SnFR-γ2 and SnFR-γ8.

Figure 1 with 3 supplements see all

Download asset Open asset

Bulk rat hippocampal cultures infected with the AAV of iGluSnFR gave wide, diffuse and, in our hands, relatively infrequent fluorescent signals (see Figure 2 and Video 1). We made adeno-associated viruses (AAVs) of both SnFR-γ2 and SnFR-γ8. Hippocampal bulk cultures infected with these viruses revealed regular fluorescence spikes (typically about 1 Hz frequency, see Figure 2 and Video 2). At hand-picked regions of interest (ROIs), SnFR-γ2 and SnFR-γ8 showed a more localised signal (Figure 2).

Figure 2

Download asset Open asset

Video 1

Download asset

Video 2

Download asset

To address synaptic localisation quantitatively, we performed immunocytochemistry on bulk cultured hippocampal neurons (Figure 3). We live-labelled cells for GFP (corresponding to the cpGFP domain of iGluSnFR) to show surface localisation. Following fixing, we stained for PSD95 and MAP2 to mark postsynaptic densities and dendritic shafts, respectively. The signal for SnFR-γ2 and SnFR-γ8 was markedly more punctate than for iGluSnFR, as expected. We detected about 50% greater average intensity in PSD spots for SnFR-γ2 and SnFR-γ8 (p = 0.03 and 0.06, respectively), as compared to iGluSnFR, suggesting that SnFR-γ2 and SnFR-γ8 are weakly enriched at synapses and probably exchange with extrasynaptic sites. The PSD size was slightly reduced for SnFR-γ2 compared to iGluSnFR, but the number of PSD spots per neuron imaged was similar across all conditions.

Figure 3

Download asset Open asset

The demonstration that SnFR-γ2 and SnFR-γ8 are enriched at synapses suggested they might show reduced spatial spread of the fluorescent glutamate report in live cells. To investigate this, we co-infected the cultures with a lentivirus for Homer-tdTomato. The complexity of the SnFR-γ2 signal between intracellular (probably endoplasmic reticulum [ER]) membranes and synaptic sites forbade a simple measure of colocalisation between Homer and SnFR-γ2 in these experiments. Consistent with our results from immunocytochemistry (Figure 3), responsive SnFR-γ2 spots were enriched at Homer positive spots, presumably corresponding to synaptic connections (Figure 4A). Background-subtracted heat maps revealed a sharper signal for SnFR-γ2 compared to iGluSnFR (Figure 4B). Taking a line profile through the centre of the peak response showed that the apparent half-width of the responses during spontaneous neurotransmission was on average less than 500 nm (Figure 4D, E) or about three-fold narrower than for iGluSnFR in our cultures. When averaging across frames, the half-width at some sites was of the order of 300 nm, or in other words, probably limited by diffraction of the microscope (Figure 4D). For this spontaneous neurotransmission, mean amplitudes from both the peak and subsequent two frames from SnFR-γ2 were slightly increased compared to iGluSnFR (Figure 4F, G), possibly because SnFR-γ2 and SnFR-γ8 experienced a higher degree of saturation than their non-targeted version (see Discussion). The normalised peak fluorescence of the SnFR-γ2 response was also greater than that for iGluSnFR.

Figure 4

Download asset Open asset

Infection of autapses with SnFR-γ2 and SnFR-γ8 gave punctate GFP signal in the majority of cases, but for some neurons, the signal remained diffuse. Although not obviously related, we also regularly measured a near absence of evoked neurotransmission (Figure 5B) in about half the autaptic neurons following infection before DIV 3. Robust evoked glutamate release (as reported by fluorescence changes from the sensors) was present, and we failed to detect any change in VGLUT puncta size or number on MAP2-positive neurites in immunostaining experiments (Figure 5—figure supplement 1), when comparing to non-infected neurons or neurons infected with iGluSnFR. These observations suggested a profound postsynaptic deficit due to overexpression of SnFR-γ2 and SnFR-γ8. We reasoned that, given that cultured neurons are still developing in this time window, we could infect the neurons later (after 6 days in vitro, DIV) and recover the evoked currents. Indeed, a majority of autaptic neurons retained an appreciable evoked response following late infection (neurons giving excitatory postsynaptic current [EPSC] >100 pA: 7 of 31, DIV 3 infection and 24 of 45, DIV 6; Figure 5C). The fluorescent signal and presynaptic markers (Figure 5—figure supplement 2) were unaffected by the late infection. Given that infection with iGluSnFR does not affect evoked currents much (Figure 5C), we reasoned that auxiliary proteins in overexpression might be causing the postsynaptic deficit. Indeed, early infection (before DIV 3) of either of two Stg constructs tagged with mScarlet (with or without the fused N-terminal NETO transmembrane section) were highly deleterious to evoked currents in autapses (Figure 5F, G). Miniature currents were also strongly reduced in amplitude and frequency, to the point that we failed to detect minis in a large proportion of cells (Figure 5—figure supplement 2), whereas paired-pulse ratio (20-ms interval) was unaffected. However, a late infection (after DIV 6) had much less impact, suggesting a previously unheralded developmental pathology at the postsynapse from early Stargazin overexpression.

Figure 5 with 2 supplements see all

Download asset Open asset

In autaptic neurons infected with AAVs at DIV 6, the signal from iGluSnFR in response to evoked transmission at 5 Hz was again more diffuse than the response of SnFR-γ2 (Figure 6). We selected small ROIs and adjacent neighbour regions, again by hand. Whilst neighbour and principal fluorescence time series usually matched well for iGluSnFR (Figure 6C), for most ROIs in the SnFR-γ2 cells, the neighbour region was silent (Figure 6F). Correspondingly, the correlation coefficient between fluorescence time series for principal and neighbour ROIs was less for SnFR-γ2 (Figure 6).

Figure 6

Download asset Open asset

Having established these prerequisites, we returned to our original motivation for improving the spatial and temporal response of iGluSnFR, which was to evaluate synaptic transmission at individual connections. We recorded movies of autaptic neurons infected with iGluSnFR (Video 3), SnFR-γ2 (Video 4), or SnFR-γ8. At the same time, we subjected the neurons to trains of 10 stimuli at 5 Hz in 0.5, 2, and 4 mM calcium. Escaping action potentials were generated from short depolarising pulses (see Methods), and we recorded the evoked postsynaptic currents under voltage clamp. From these movies, we examined the simultaneous evoked fluorescent responses from individual ROIs.

Video 3

Download asset

Video 4

Download asset

Rather than choosing ROIs by eye, we employed a custom-written imageJ macro to systematically scan movies (25-Hz frame rate) for ROIs that show similar statistics to the expected responses. Responses at 2 mM calcium were most reliable across the different sensors and we used these to identify ROIs with >20 positive-going fluorescence transients per movie in which we made 50 stimulations. Composite ROIs were constructed from the contiguous regions identified. Strikingly, but perhaps unsurprisingly, the ROIs found in this way (Figure 7) were much smaller on average for SnFR-γ2 (mean areas: 2.5 vs. 9.6 μm²) than for iGluSnFR. On average, neurons infected with iGluSnFR gave about twice as many ROIs as either SnFR-γ2 or SnFR-γ8 (Figure 7C; 207 ROIs for iGluSnFR, 107 for SnFR-γ2, and 97 for SnFR-γ8, each over 6 neurons). For all but one neuron, the number of ROIs from SnFR-γ8 was too few to pursue a further analysis.

Figure 7 with 2 supplements see all

Download asset Open asset

Although iGluSnFR gave more ROIs, when we analysed the peak responses from each ROI, we saw that the responses of iGluSnFR in a subset of ROIs tended to run down (assessed as the magnitude of the last 10% of responses to the first 10%, Figure 7—figure supplement 1). This rundown represented as much as fourfold loss of the peak signal for some ROIs. Indeed at least half of automatically identified ROIs from the iGluSnFR movies showed rundown greater than 50% (Figure 7—figure supplement 1), whereas less than 7% of ROIs selected from SnFR-γ2 cells did. Classifying the ROIs in this way arguably lacks statistical rigour but indicates a large qualitative difference in the data obtained from the two sets of ROIs. There was no relation between the size of the compound ROI and the extent of rundown (Figure 7—figure supplement 1). Recordings were made on the same day with the same illumination conditions. Moreover, the average baseline fluorescence intensity (F₀) was stable during the recordings, and showed little change for either iGluSnFR or SnFR-γ2 in our recording conditions (Figure 7—figure supplement 2). This observation speaks against photobleaching as a source of the aberrant behaviour of iGluSnFR in some ROIs.

To understand if the rundown was a problem of the sensor, or perhaps simply an accurate report of the neurotransmission running down in these cells, we looked at how well the responses of the fluorescence reporters followed the summed postsynaptic electrical response to AP-induced release of glutamate, which we measured simultaneously. The EPSC peak amplitude typically reduced by about 40% over an entire run (see Figure 7—figure supplement 2), for both iGluSnFR and SnFR-γ2. Therefore, the iGluSnFR rundown exaggerated the current rundown and the SnFR-γ2 construct perhaps underestimated it, consistent with reporting different sources of glutamate (e.g., iGluSnFR preferentially reporting spillover). However, we note that changes in the peak response of only 30% are less easy to see in the noisy optical measurement (SNR ~3) than in the electrophysiology of evoked EPSCs (SNR ~1000).

Postsynaptic currents in iGluSnFR and SnFR-γ2 showed similar short-term plasticity (Figure 8A, B). With 2 mM calcium, the peak current amplitude exhibited mild depression during trains of five pulses. The initial amplitude was larger and showed a stronger reduction at 4 mM calcium, consistent with multivesicular release and vesicle depletion, whereas responses were small and did not change during the train for 0.5 mM calcium. We described the degree of depression with the ratio of amplitudes in response to the first and last stimulation in the group (Amplitude₅/Amplitude₁). For fluorescence responses, we obtained peak responses from ROI time series following automatic (but supervised) background subtraction and discarded ROIs with a low peak signal-to-noise ratio (>3, see Methods).

Figure 8

Download asset Open asset

In movies recorded from iGluSnFR expressing cells, there was no relation between the depression of the postsynaptic AMPA receptor response during each train and the average depression of the fluorescence signal over the ROIs (Figure 8C). In contrast, there was an excellent correlation between the SnFR-γ2 depression and the postsynaptic current depression at 2 and 4 mM calcium (Figure 8C) indicating that the synaptic release reported by SnFR-γ2 across tens of sites per neuron accurately represents the average synaptic response to glutamate release. At 0.5 mM calcium, one SnFR-γ2 cell gave an abnormal high potentiation (seven-fold). However, omitting this point did not reveal any correlation in the remaining data, so we kept this point in the dataset. The lack of correlation of signals from iGluSnFR is likely because of rundown and also possibly because indistinct localisation leads it to report stray glutamate as well as synaptic glutamate, particularly in conditions where repeated glutamate release occurs (5 Hz or above, see Discussion). The higher correlation in fluorescence between neighbouring sites that we observed for iGluSnFR (Figure 6) exemplifies this phenomenon.

Taking each train of five responses, we could plot the standard deviation of the five responses against the mean. In this way, we could estimate the coefficient of variation (CV) for a large number of synapses and responses (Figure 8D). Quantal analysis (del Castillo and Katz, 1954) suggests that the square of the CV is a simple function of the number of sites and the probability of release (see Methods). Even though we evoked EPSCs over presumably hundreds of connections, the selected ROIs allowed us to analyse the stoichiometry of release at putative single synapses optically. The number of ROIs is limited by the field of view and the focal depth of the objective lens. Taking the slopes from linear fits to paired SD/mean values, we performed a non-linear fit to the estimated squared CV values across the whole population of synaptic ROIs, allowing for a different release probability at each calcium condition, but common N (corresponding to the number of sites in an ROI). Most generally, N here relates to the number of release sites per ROI (independent of the number of synapses in the ROI), and therefore P_R is the neurotransmitter release probability at a single release site. Even though N is averaged over a large number of ROIs (synapses), we fixed N to be an integer to facilitate the search. For both conditions, N could not be uniquely determined (three data points but four unknowns) and the data were fit equally well with a range of N values. N and P_R are inversely correlated. However, for N < 3 the release probabilities were unfeasibly high (Figure 8E), so we took a conservative approach, selecting the lowest N value that gave P_R less than 50% for 0.5 mM Ca. Values of N above one could indicate either that each ROI contains multiple synapses, or, more likely given previous results on autapses, that the synapses show multivesicular release. As our results on individual ROIs below show, because practically no ROIs show N = 1 behaviour, the latter interpretation is also more likely here.

For CV analysis, independent of the choice of N value, the average release probability over all ROIs, estimated from SnFR-γ2, increased monotonically with Ca²⁺ concentration, as expected from decades of electrophysiological analysis (Figure 8G, 6/6 cells). By contrast, the release probability estimated from iGluSnFR responses only increased monotonically for two out of six cells (Figure 8G), suggesting it is a poor reporter of synaptic activity in this context. These results indicate a clear quantitative benefit of targeted SnFR-γ2 over the non-targeted iGluSnFR. Previous analysis on connected cultured hippocampal neurons using electrophysiology gave similar values (N = 5 and P_R = 0.4 from CV analysis Bekkers and Stevens, 1990) to the ones that we obtained from SnFR-γ2 at 2 mM Ca²⁺. Although the congruence between connections between single neurons and the per-ROI parameters we found here is encouraging, we note that the CV method applied in this way cannot resolve N and therefore only gives an approximate rank order of synaptic release probabilities across the different conditions. We would therefore not recommend to use it.

In order to provide a more detailed analysis of synaptic release parameters, we next asked if SnFR-γ2 could be used to measure the quantal properties of individual synapses, and if it offered any advantages over the non-targeted iGluSnFR. To do this, we performed automated baseline subtraction from each ROI and detected peak responses according to a manually determined pattern from the mean of all responses. For both iGluSnFR and SnFR-γ2, we observed fluorescence time series with clear calcium dependence and quantal behaviour (Figure 9—figure supplement 1). From 50 stimulations at each concentration of calcium (150 events in total), at about half of the ROIs from iGluSnFR we could build histograms and fit globally (see Figure 9—figure supplement 2 for workflow). For SnFR-γ2, 80% of the automatically-found ROIs were usable for quantal analysis, and these ROIs reported a range of release probabilities similar to those determined from the CV analysis (Figure 9). In the majority of ROIs, according to the Kolmogorov–Smirnov test for the goodness of fit, the binomial model gave a better fit than the Poisson model (Bhumbra and Beato, 2013; Malagon et al., 2016). Examples of imperfect fits are shown in Figure 9—figure supplement 3. For SnFR-γ2, the distributions of peak amplitudes gave clearer multiple peaks (Figure 9C, D) than iGluSnFR responses. However, these peaks were washed out above 3 or 4 components, even if up to 8–10 components were needed to cover the entire spread of amplitude data. The ranges of numbers of components and open probabilities determined from global fitting of individual synapses were slightly narrower for SnFR-γ2 than for iGluSnFR (Figure 9F, G).

Figure 9 with 3 supplements see all

Download asset Open asset

In summary, the results from the SnFR-γ2 sensor were more convincingly quantal. Yet global fitting allowed similar, if less consistent, quantal parameters to be determined from iGluSnFR responses. Identification of single quanta at 0.5 mM Ca²⁺ was challenging for both indicators, and this difficulty was more pronounced for SnFR-γ2, possibly because the smaller ROIs we obtained in this condition are inherently more noisy. This drawback meant that our quantal analysis (and particularly the estimation of the quantal size) relied on the equal spacing of the responses to multiple quanta at higher calcium concentrations. In turn, the best fits could only describe the histograms with limited accuracy (Figure 9C, D)—we expect that this could be improved by collecting more responses. The ROIs for iGluSnFR were on average much larger in size and number, and a larger fraction of them needed to be manually discarded because they showed strong rundown. A greater proportion of ROIs that were responsive for SnFR-γ2 could be described by quantal parameters, and very few were affected by rundown in the peak response that was seen in ~50% of the responsive ROIs in iGluSnFR-infected neurons.

The principal advantage of optical reporting of neurotransmission is direct access to individual synaptic responses. Taking optical responses from the field of view as group, SnFR-γ2 (but not iGluSnFR) accurately reported short-term depression, as measured simultaneously with electrophysiology, except in the noisy conditions of 0.5 mM calcium. On average, SnFR-γ2 also gave similar quantal parameters from global estimates of the CV to previous electrophysiological work (N = 5 and P_R = 0.4 from CV analysis on pairs of cultured rat neurons; Bekkers and Stevens, 1990). These observations suggest that SnFR-γ2 is a fair reporter of presynaptic release, and that postsynaptic short-term plasticity (e.g., from desensitisation) is limited enough to be neglected at the frequencies we used (up to 5 Hz) (Malagon et al., 2016). However, both iGluSnFR and SnFR-γ2 failed to report gross variation in synaptic release properties across individual cells. Are synapses then truly variable (Rosenmund et al., 1993)? The deviation of individual mean/SD values from responsive ROIs (stemming from distinct synapses, either neighbouring groups or individual sites) from the fitted line representing the global CV was limited, and the R² value was quite high for all conditions. However, the range of putative maximum release probabilities detected in individual cells was substantial (typically from 40% to 80%, at 4 mM calcium). One caveat is that our approach may be biased towards detecting synapses with higher release rates. Another important consideration is that the number of ROIs we typically observed (tens) is consistently fewer than expected from the number of synapses that would be expected to generate nA-scale evoked synaptic currents (hundreds). The reason for this disparity remains unclear. Correlation of the parameters of synaptic release determined from optical analysis to synaptic architectures, synapses altered by plasticity or experience and synapses whose properties are altered by pathological mutants, should yield information about cell-autonomous synaptic diversity (Matz et al., 2010).

Unlike at the neuromuscular junction (del Castillo and Katz, 1954), peaks in the histograms of evoked EPSC amplitudes are much less prominent at central synapses (Bekkers, 1994). Several mathematical treatments were proposed to deal with the smeared distribution that apparently comes from quantal variability at central synapses, including using Bayesian statistics and the gamma distribution (Bhumbra and Beato, 2013; Soares et al., 2019). Optical report of glutamate concentration signal eliminates postsynaptic receptors as a source of variability. Miniature currents are variable in amplitude, introducing unavoidable variance when measuring evoked synaptic currents or when indirectly reporting activity from NMDA receptor calcium flux (Oertner et al., 2002). Simulations suggest that most variability comes from the variation in glutamate released by each vesicle (Franks et al., 2003), but we saw little evidence for this. Perhaps SnFR-γ2 is saturated by even poorly loaded vesicles, but fitting distributions with release sites (N) up to 8 or more, corresponding to up to 8 quanta in high Ca²⁺, suggests this is not the case. AMPA-type glutamate receptors have a lower affinity for glutamate than SnFR-γ2, and so the invariant quantal size reported by SnFR-γ2 would in any case translate to less robust AMPA receptor activation. Synaptic AMPARs are probably often not saturated due to a steep dependence on geometric factors (Savtchenko and Rusakov, 2014), or because the number of AMPA receptors available is so variable. A similar “lack of dips” in histograms has been observed in distributions from iGluSnFR in postsynaptic sites (Soares et al., 2019; Heck et al., 2019), requiring extensive mathematical modelling to extract quantal parameters. In clear contrast, confining iGluSnFR to the presynaptic terminal (in the case that anatomy allows this) gives quantal amplitude distributions (Duerst et al., 2020; Dürst et al., 2019). In contrast with some previous work, for example using field stimulation (Heck et al., 2019), in our experiments in autapses, we stimulated glutamate release with a single action potential and obtained quantal histograms with separated peaks. In particular, we can confirm that changing the calcium concentration and examining the same terminal appears to be beneficial in determining quantal parameters (Duerst et al., 2020). On the other hand, in native tissue, the environment of synaptic terminals may reduce the ability of SnFR-γ2 to report quantal distributions, and future work should address this point.

Single AP stimulation enabled a clear view of the systematic fast loss of the iGluSnFR signal (Figure 8). Why does iGluSnFR run down and why does SnFR-γ2 instead have a stable response? We have no decisive mechanistic explanation. One possibility is that bleaching reduces ∆F/F because a large proportion of the background fluorescence signal comes from either intracellular or occluded sensor molecules. If bleaching (or another process that eliminates signal, like endocytosis) were more intense when glutamate is bound, and iGluSnFR were to in turn have more average occupancy by glutamate than the confined SnFR-γ2, this could specifically reduce ∆F. In such a scheme, SnFR-γ2 would be protected from glutamate exposure by seeing only fast glutamate transients that saturate the sensor and immediately disperse, as opposed to a larger iGluSnFR capacity that is saturated for longer periods by spillover (Figure 10). The diffusion within synapses of the same single TM protein with and without a PSD-binding ligand (from the Stg C-terminus) provides an excellent comparison for our purposes (Li and Blanpied, 2016). These studies show that even though the diffusion of any protein is impeded within the PSD, the membrane protein without any PSD ligand is not excluded. However, our results are consistent with the idea that the PDGFR motif of the original iGluSnFR is excluded from synapses, as corroborated by expansion microscopy (Aggarwal et al., 2022).

Figure 10

Download asset Open asset

Assuming equal expression, and the concentration of membrane proteins with a PDZ ligand inside PSDs (Li and Blanpied, 2016), the density of SnFR-γ2 must be less outside the synapse (also see Figure 6). On average, a greater capacity outside the synapse would mean that, even without being excluded from synapses, iGluSnFR is biased against measuring synaptic glutamate. The basal response of iGluSnFR might be greater in neurons (even though in HEK cells the response was similar) because it is always moving in relation to release sites, and can respond to distant glutamate better by covering greater areas (even within an ROI).

Neither iGluSnFR nor SnFR-γ2 can identify single release events in our conditions, but for different reasons. The iGluSnFR signal often runs down intensely, meaning that the quantal estimate ∆F changes during a recording and cannot be considered constant across different conditions. SnFR-γ2 cannot reliably report isolated single release events because the amplitude of the signal is too small. However, spontaneous neurotransmission in cultures (in the absence of TTX) could be readily resolved by SnFR-γ2, at spot-like ROIs. These optical responses may correspond to single or multiple vesicles released by action potentials. For short-term plasticity, whereby synapses are reactivated over 50-ms intervals or similar, the improved spatial profiles of SnFR-γ2 are expected to be beneficial (Helassa et al., 2018). In our experiments, we did not focus on ROIs with high-frame rate, but this is an obvious future application.

iGluSnFR has been used in a variety of organisms (James et al., 2019; Marvin et al., 2013; Borghuis et al., 2013), and here we expressed SnFR-γ2 only in cultured rodent neurons. Using mouse auxiliary proteins to anchor the glutamate reporter opposite to release sites may not be universal, and may influence synaptic transmission in other systems in undesirable ways. Stargazin overexpression reduced miniature current amplitude and frequency but did not affect paired-pulse response (Figure 5—figure supplement 2) and glutamate release was intact (from SnFR-γ2 signals). This strongly suggests a postsynaptic deficit in cultured hippocampal cells but in other neuronal types, effects might be different. When we made inadvertent recordings of GABAergic cells, which we subsequently discarded, we saw no response of the iGluSnFR sensors, but nor was there any obvious effect on evoked inhibitory current magnitudes. It is unclear how well this approach can translate across species, even though PSD binding should be conserved. Notably, sensor performance is always worse in the in vivo context, and our sensor, along with a similar recently reported approach (Aggarwal et al., 2022), is probably better suited to quantitative studies of synaptic transmission in vitro, rather than in vivo work, even though its performance should not be worse than the original iGluSnFR construct.

Why does overexpressing various forms of Stargazin in the first few days of culture create a specific postsynaptic deficit? Stargazin has a very long first intron (Letts et al., 1998). This long first intron of Stg is the target of three mutations (including the original stargazer mouse) which reduce or abolish γ-2 mRNA (Letts et al., 2003). Because long introns extend transcription times, they are thought to delay expression. The most curious aspect of our observations was that release was apparently normal, and presynaptic VGLUT1 markers were not affected. It is not routine to perform electrophysiological characterisation of synapses following molecular interventions before imaging, however, the synapses in neurons infected with Stargazin or chimeras on or before DIV 3 were non-functional, despite their otherwise normal appearance. It is likely that Stargazin is not the ideal host for the iGluSnFR sensor domain. Auxiliary proteins that are not complexed to AMPA receptors probably fill slots in the postsynaptic density, reducing postsynaptic currents. The performance of the SnFR-γ8 construct was similar to SnFR-γ2, but neuronal labelling was often less broad, giving fewer ROIs that responded. It is likely that tethered sensors preferentially select stronger synapses. The additional transmembrane helix from NETO2 appears to reduce expression (although this may indeed be desirable) and is quantitatively more damaging to synaptic transmission than Stargazin in a more native form (Figure 5).

Rather than obtaining an optimised sensor, we have instead demonstrated further proof of principle that postsynaptic tethering improves reporter characteristics (Soares et al., 2019), even if it reduces overall signal. Our targeted sensor did not have distinctly better fluorescence performance in terms of signal to noise compared to iGluSnFR. A recent report used the principle of tethering to localise the signal (Aggarwal et al., 2022). One disadvantage of targeting may be that non-synaptic signals of interest could be lost. More critically, the synaptic localisation does not guarantee to measure exclusively synaptic glutamate; waves of elevated glutamate from other sources including astroglia, substantial spillover or even changes in the optical properties of tissue could be mistaken for synaptic glutamate in this context, and this is a limitation of our approach. In organised brain tissue, two-photon excitation will not avoid this difficulty. Spectral variants with distinct (or no) targeting could be combined to get a fuller picture of both synaptic and spillover glutamate (see Hao and Plested, 2022 for discussion). Various approaches are known to augment the signal from iGluSnFR including using the superfolder variant of GFP (Marvin et al., 2018), mutagenesis (Figure 1—figure supplement 1), or a different promoter (we used human Synapsin). Indeed, the superfolder variant of iGluSnFR could be used very successfully for quantal analysis in cell culture, when analysis was confined to axonal projections (i.e., presynaptic sites, Mendonça et al., 2022). More work is needed to understand why targeting reduced the size of ROIs at postsynaptic sites in our hands, whereas well-confined signals are seen in axons. Finally, other tethers, either minimal ones (Li and Blanpied, 2016) or those based on other synaptic proteins, may also give better, more defined, or even orthogonal signals.

Custom software is available at https://github.com/agplested/SAFT (copy archived at swh:1:rev:c3f5615648978f9fdc82237c443ed29eadb2fc15).

1. 1. Hao Y
   2. Toulmé E
   3. König B
   4. Rosenmund C
   5. Plested A

   (2023) Zenodo

   Targeted sensors for glutamatergic neurotransmission.

   https://doi.org/10.5281/zenodo.7512561

1. Preprint

   1. Aggarwal A
   2. Liu R
   3. Chen Y
   4. Ralowicz AJ
   5. Bergerson SJ
   6. Tomaska F
   7. Hanson TL
   8. Hasseman JP
   9. Reep D
   10. Tsegaye G
   11. Yao P
   12. Ji X
   13. Kloos M
   14. Walpita D
   15. Patel R
   16. Mohr MA
   17. Tilberg PW
   18. Mohar B
   19. Looger LL
   20. Marvin JS
   21. Hoppa MB
   22. Konnerth A
   23. Kleinfeld D
   24. Schreiter ER
   25. Podgorski K
   26. The GENIE Project Team

   (2022) Glutamate Indicators with Improved Activation Kinetics and Localization for Imaging Synaptic Transmission

   bioRxiv.

   https://doi.org/10.1101/2022.02.13.480251

   • Google Scholar
2. 1. Balaji J
   2. Ryan TA

   (2007) Single-vesicle imaging reveals that synaptic vesicle exocytosis and endocytosis are coupled by a single stochastic mode

   PNAS 104:20576–20581.

   https://doi.org/10.1073/pnas.0707574105

   • PubMed
   • Google Scholar
3. 1. Bekkers JM
   2. Stevens CF

   (1990) Presynaptic mechanism for long-term potentiation in the hippocampus

   Nature 346:724–729.

   https://doi.org/10.1038/346724a0

   • PubMed
   • Google Scholar
4. 1. Bekkers JM

   (1994) Quantal analysis of synaptic transmission in the central nervous system

   Current Opinion in Neurobiology 4:360–365.

   https://doi.org/10.1016/0959-4388(94)90097-3

   • PubMed
   • Google Scholar
5. 1. Bhumbra GS
   2. Beato M

   (2013) Reliable evaluation of the quantal determinants of synaptic efficacy using Bayesian analysis

   Journal of Neurophysiology 109:603–620.

   https://doi.org/10.1152/jn.00528.2012

   • PubMed
   • Google Scholar
6. 1. Bindels DS
   2. Haarbosch L
   3. van Weeren L
   4. Postma M
   5. Wiese KE
   6. Mastop M
   7. Aumonier S
   8. Gotthard G
   9. Royant A
   10. Hink MA
   11. Gadella TJ

   (2017) MScarlet: a bright monomeric red fluorescent protein for cellular imaging

   Nature Methods 14:53–56.

   https://doi.org/10.1038/nmeth.4074

   • PubMed
   • Google Scholar
7. 1. Borghuis BG
   2. Marvin JS
   3. Looger LL
   4. Demb JB

   (2013) Two-Photon imaging of nonlinear glutamate release dynamics at bipolar cell synapses in the mouse retina

   The Journal of Neuroscience 33:10972–10985.

   https://doi.org/10.1523/JNEUROSCI.1241-13.2013

   • PubMed
   • Google Scholar
8. 1. Chanaday NL
   2. Kavalali ET

   (2018) Optical detection of three modes of endocytosis at hippocampal synapses

   eLife 7:e36097.

   https://doi.org/10.7554/eLife.36097

   • PubMed
   • Google Scholar
9. 1. Choi JH
   2. Yu NK
   3. Baek GC
   4. Bakes J
   5. Seo D
   6. Nam HJ
   7. Baek SH
   8. Lim CS
   9. Lee YS
   10. Kaang BK

   (2014) Optimization of AAV expression cassettes to improve packaging capacity and transgene expression in neurons

   Molecular Brain 7:17.

   https://doi.org/10.1186/1756-6606-7-17

   • PubMed
   • Google Scholar
10. 1. Cizeron M
    2. Qiu Z
    3. Koniaris B
    4. Gokhale R
    5. Komiyama NH
    6. Fransén E
    7. Grant SGN

    (2020) A brainwide atlas of synapses across the mouse life span

    Science 369:270–275.

    https://doi.org/10.1126/science.aba3163

    • PubMed
    • Google Scholar
11. 1. del Castillo J
    2. Katz B

    (1954) Quantal components of the end-plate potential

    The Journal of Physiology 124:560–573.

    https://doi.org/10.1113/jphysiol.1954.sp005129

    • PubMed
    • Google Scholar
12. Preprint

    1. Duerst C
    2. Wiegert JS
    3. Schulze C
    4. Helassa N
    5. Török K
    6. Oertner TG

    (2020) The Vesicular Release Probability Sets the Strength of Individual Schaffer Collateral Synapses

    bioRxiv.

    https://doi.org/10.1101/2020.08.02.232850

    • Google Scholar
13. 1. Dürst CD
    2. Wiegert JS
    3. Helassa N
    4. Kerruth S
    5. Coates C
    6. Schulze C
    7. Geeves MA
    8. Török K
    9. Oertner TG

    (2019) High-speed imaging of glutamate release with genetically encoded sensors

    Nature Protocols 14:1401–1424.

    https://doi.org/10.1038/s41596-019-0143-9

    • PubMed
    • Google Scholar
14. 1. Edelstein AD
    2. Tsuchida MA
    3. Amodaj N
    4. Pinkard H
    5. Vale RD
    6. Stuurman N

    (2014) Advanced methods of microscope control using μmanager software

    Journal of Biological Methods 1:e10.

    https://doi.org/10.14440/jbm.2014.36

    • PubMed
    • Google Scholar
15. Report

    1. Eilers PHC
    2. Boelens HFM

    (2005)

    Baseline correction with asymmetric least squares smoothing

    Leiden University Medical Centre.

    • Google Scholar
16. 1. Enoki R
    2. Hu YL
    3. Hamilton D
    4. Fine A

    (2009) Expression of long-term plasticity at individual synapses in hippocampus is graded, bidirectional, and mainly presynaptic: optical quantal analysis

    Neuron 62:242–253.

    https://doi.org/10.1016/j.neuron.2009.02.026

    • PubMed
    • Google Scholar
17. 1. Evans DA
    2. Stempel AV
    3. Vale R
    4. Ruehle S
    5. Lefler Y
    6. Branco T

    (2018) A synaptic threshold mechanism for computing escape decisions

    Nature 558:590–594.

    https://doi.org/10.1038/s41586-018-0244-6

    • PubMed
    • Google Scholar
18. 1. Faber DS
    2. Korn H

    (1991) Applicability of the coefficient of variation method for analyzing synaptic plasticity

    Biophysical Journal 60:1288–1294.

    https://doi.org/10.1016/S0006-3495(91)82162-2

    • PubMed
    • Google Scholar
19. 1. Farsi Z
    2. Walde M
    3. Klementowicz AE
    4. Paraskevopoulou F
    5. Woehler A

    (2021) Single synapse glutamate imaging reveals multiple levels of release mode regulation in mammalian synapses

    IScience 24:101909.

    https://doi.org/10.1016/j.isci.2020.101909

    • PubMed
    • Google Scholar
20. 1. Franks KM
    2. Stevens CF
    3. Sejnowski TJ

    (2003) Independent sources of quantal variability at single glutamatergic synapses

    The Journal of Neuroscience 23:3186–3195.

    https://doi.org/10.1523/JNEUROSCI.23-08-03186.2003

    • PubMed
    • Google Scholar
21. 1. Gambino F
    2. Pagès S
    3. Kehayas V
    4. Baptista D
    5. Tatti R
    6. Carleton A
    7. Holtmaat A

    (2014) Sensory-Evoked LTP driven by dendritic plateau potentials in vivo

    Nature 515:116–119.

    https://doi.org/10.1038/nature13664

    • PubMed
    • Google Scholar
22. 1. Granseth B
    2. Odermatt B
    3. Royle SJ
    4. Lagnado L

    (2006) Clathrin-Mediated endocytosis is the dominant mechanism of vesicle retrieval at hippocampal synapses

    Neuron 51:773–786.

    https://doi.org/10.1016/j.neuron.2006.08.029

    • PubMed
    • Google Scholar
23. 1. Griesinger CB
    2. Richards CD
    3. Ashmore JF

    (2005) Fast vesicle replenishment allows indefatigable signalling at the first auditory synapse

    Nature 435:212–215.

    https://doi.org/10.1038/nature03567

    • PubMed
    • Google Scholar
24. 1. Hao Y
    2. Plested AJR

    (2022) Seeing glutamate at central synapses

    Journal of Neuroscience Methods 375:109531.

    https://doi.org/10.1016/j.jneumeth.2022.109531

    • PubMed
    • Google Scholar
25. 1. Heck J
    2. Parutto P
    3. Ciuraszkiewicz A
    4. Bikbaev A
    5. Freund R
    6. Mitlöhner J
    7. Andres-Alonso M
    8. Fejtova A
    9. Holcman D
    10. Heine M

    (2019) Transient confinement of Cav2.1 Ca2+-channel splice variants shapes synaptic short-term plasticity

    Neuron 103:66–79.

    https://doi.org/10.1016/j.neuron.2019.04.030

    • PubMed
    • Google Scholar
26. 1. Helassa N
    2. Dürst CD
    3. Coates C
    4. Kerruth S
    5. Arif U
    6. Schulze C
    7. Wiegert JS
    8. Geeves M
    9. Oertner TG
    10. Török K

    (2018) Ultrafast glutamate sensors resolve high-frequency release at schaffer collateral synapses

    PNAS 115:5594–5599.

    https://doi.org/10.1073/pnas.1720648115

    • PubMed
    • Google Scholar
27. 1. Isaac JT
    2. Hjelmstad GO
    3. Nicoll RA
    4. Malenka RC

    (1996) Long-term potentiation at single fiber inputs to hippocampal CA1 pyramidal cells

    PNAS 93:8710–8715.

    https://doi.org/10.1073/pnas.93.16.8710

    • PubMed
    • Google Scholar
28. 1. James B
    2. Darnet L
    3. Moya-Díaz J
    4. Seibel SH
    5. Lagnado L

    (2019) An amplitude code transmits information at a visual synapse

    Nature Neuroscience 22:1140–1147.

    https://doi.org/10.1038/s41593-019-0403-6

    • PubMed
    • Google Scholar
29. 1. Jensen TP
    2. Zheng K
    3. Cole N
    4. Marvin JS
    5. Looger LL
    6. Rusakov DA

    (2019) Multiplex imaging relates quantal glutamate release to presynaptic ca2+ homeostasis at multiple synapses in situ

    Nature Communications 10:1414.

    https://doi.org/10.1038/s41467-019-09216-8

    • PubMed
    • Google Scholar
30. 1. Jontes JD
    2. Smith SJ

    (2000) Filopodia, spines, and the generation of synaptic diversity

    Neuron 27:11–14.

    https://doi.org/10.1016/s0896-6273(00)00003-9

    • PubMed
    • Google Scholar
31. 1. Kim Y
    2. Lee U
    3. Choi C
    4. Chang S

    (2020) Release mode dynamically regulates the RRP refilling mechanism at individual hippocampal synapses

    The Journal of Neuroscience 40:8426–8437.

    https://doi.org/10.1523/JNEUROSCI.3029-19.2020

    • PubMed
    • Google Scholar
32. 1. Letts VA
    2. Felix R
    3. Biddlecome GH
    4. Arikkath J
    5. Mahaffey CL
    6. Valenzuela A
    7. Bartlett FS
    8. Mori Y
    9. Campbell KP
    10. Frankel WN

    (1998) The mouse stargazer gene encodes a neuronal Ca2+-channel gamma subunit

    Nature Genetics 19:340–347.

    https://doi.org/10.1038/1228

    • PubMed
    • Google Scholar
33. 1. Letts VA
    2. Kang MG
    3. Mahaffey CL
    4. Beyer B
    5. Tenbrink H
    6. Campbell KP
    7. Frankel WN

    (2003) Phenotypic heterogeneity in the stargazin allelic series

    Mammalian Genome 14:506–513.

    https://doi.org/10.1007/s00335-003-2268-x

    • PubMed
    • Google Scholar
34. 1. Li TP
    2. Blanpied TA

    (2016) Control of transmembrane protein diffusion within the postsynaptic density assessed by simultaneous single-molecule tracking and localization microscopy

    Frontiers in Synaptic Neuroscience 8:19.

    https://doi.org/10.3389/fnsyn.2016.00019

    • PubMed
    • Google Scholar
35. 1. Lois C
    2. Hong EJ
    3. Pease S
    4. Brown EJ
    5. Baltimore D

    (2002) Germline transmission and tissue-specific expression of transgenes delivered by lentiviral vectors

    Science 295:868–872.

    https://doi.org/10.1126/science.1067081

    • PubMed
    • Google Scholar
36. 1. Malagon G
    2. Miki T
    3. Llano I
    4. Neher E
    5. Marty A

    (2016) Counting vesicular release events reveals binomial release statistics at single glutamatergic synapses

    The Journal of Neuroscience 36:4010–4025.

    https://doi.org/10.1523/JNEUROSCI.4352-15.2016

    • PubMed
    • Google Scholar
37. 1. Marvin JS
    2. Borghuis BG
    3. Tian L
    4. Cichon J
    5. Harnett MT
    6. Akerboom J
    7. Gordus A
    8. Renninger SL
    9. Chen TW
    10. Bargmann CI
    11. Orger MB
    12. Schreiter ER
    13. Demb JB
    14. Gan WB
    15. Hires SA
    16. Looger LL

    (2013) An optimized fluorescent probe for visualizing glutamate neurotransmission

    Nature Methods 10:162–170.

    https://doi.org/10.1038/nmeth.2333

    • PubMed
    • Google Scholar
38. 1. Marvin JS
    2. Scholl B
    3. Wilson DE
    4. Podgorski K
    5. Kazemipour A
    6. Müller JA
    7. Schoch S
    8. Quiroz FJU
    9. Rebola N
    10. Bao H
    11. Little JP
    12. Tkachuk AN
    13. Cai E
    14. Hantman AW
    15. Wang SS-H
    16. DePiero VJ
    17. Borghuis BG
    18. Chapman ER
    19. Dietrich D
    20. DiGregorio DA
    21. Fitzpatrick D
    22. Looger LL

    (2018) Stability, affinity, and chromatic variants of the glutamate sensor iglusnfr

    Nature Methods 15:936–939.

    https://doi.org/10.1038/s41592-018-0171-3

    • PubMed
    • Google Scholar
39. 1. Matz J
    2. Gilyan A
    3. Kolar A
    4. McCarvill T
    5. Krueger SR

    (2010) Rapid structural alterations of the active zone lead to sustained changes in neurotransmitter release

    PNAS 107:8836–8841.

    https://doi.org/10.1073/pnas.0906087107

    • PubMed
    • Google Scholar
40. 1. McAllister AK
    2. Stevens CF

    (2000) Nonsaturation of AMPA and NMDA receptors at hippocampal synapses

    PNAS 97:6173–6178.

    https://doi.org/10.1073/pnas.100126497

    • PubMed
    • Google Scholar
41. 1. Mendonça PRF
    2. Tagliatti E
    3. Langley H
    4. Kotzadimitriou D
    5. Zamora-Chimal CG
    6. Timofeeva Y
    7. Volynski KE

    (2022) Asynchronous glutamate release is enhanced in low release efficacy synapses and dispersed across the active zone

    Nature Communications 13:3497.

    https://doi.org/10.1038/s41467-022-31070-4

    • PubMed
    • Google Scholar
42. 1. Miesenböck G
    2. De Angelis DA
    3. Rothman JE

    (1998) Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins

    Nature 394:192–195.

    https://doi.org/10.1038/28190

    • PubMed
    • Google Scholar
43. 1. Namiki S
    2. Sakamoto H
    3. Iinuma S
    4. Iino M
    5. Hirose K

    (2007) Optical glutamate sensor for spatiotemporal analysis of synaptic transmission

    The European Journal of Neuroscience 25:2249–2259.

    https://doi.org/10.1111/j.1460-9568.2007.05511.x

    • PubMed
    • Google Scholar
44. 1. Oertner TG
    2. Sabatini BL
    3. Nimchinsky EA
    4. Svoboda K

    (2002) Facilitation at single synapses probed with optical quantal analysis

    Nature Neuroscience 5:657–664.

    https://doi.org/10.1038/nn867

    • PubMed
    • Google Scholar
45. 1. Okumoto S
    2. Looger LL
    3. Micheva KD
    4. Reimer RJ
    5. Smith SJ
    6. Frommer WB

    (2005) Detection of glutamate release from neurons by genetically encoded surface-displayed FRET nanosensors

    PNAS 102:8740–8745.

    https://doi.org/10.1073/pnas.0503274102

    • PubMed
    • Google Scholar
46. Software

    1. Plested A

    (2023) SAFT, version swh:1:rev:c3f5615648978f9fdc82237c443ed29eadb2fc15

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:ab8c9acfa92cfe80552c882c443cb16f70e2788e;origin=https://github.com/AGPlested/SAFT;visit=swh:1:snp:34fbc252c5eec382433674f5972bf22c4527e537;anchor=swh:1:rev:c3f5615648978f9fdc82237c443ed29eadb2fc15
47. 1. Popovic MA
    2. Carnevale N
    3. Rozsa B
    4. Zecevic D

    (2015) Electrical behaviour of dendritic spines as revealed by voltage imaging

    Nature Communications 6:8436.

    https://doi.org/10.1038/ncomms9436

    • PubMed
    • Google Scholar
48. 1. Raghavachari S
    2. Lisman JE

    (2004) Properties of quantal transmission at CA1 synapses

    Journal of Neurophysiology 92:2456–2467.

    https://doi.org/10.1152/jn.00258.2004

    • PubMed
    • Google Scholar
49. 1. Ribrault C
    2. Sekimoto K
    3. Triller A

    (2011) From the stochasticity of molecular processes to the variability of synaptic transmission

    Nature Reviews. Neuroscience 12:375–387.

    https://doi.org/10.1038/nrn3025

    • PubMed
    • Google Scholar
50. 1. Rosenmund C
    2. Clements JD
    3. Westbrook GL

    (1993) Nonuniform probability of glutamate release at a hippocampal synapse

    Science 262:754–757.

    https://doi.org/10.1126/science.7901909

    • PubMed
    • Google Scholar
51. 1. Rost BR
    2. Schneider F
    3. Grauel MK
    4. Wozny C
    5. Bentz C
    6. Blessing A
    7. Rosenmund T
    8. Jentsch TJ
    9. Schmitz D
    10. Hegemann P
    11. Rosenmund C

    (2015) Optogenetic acidification of synaptic vesicles and lysosomes

    Nature Neuroscience 18:1845–1852.

    https://doi.org/10.1038/nn.4161

    • PubMed
    • Google Scholar
52. 1. Sankaranarayanan S
    2. De Angelis D
    3. Rothman JE
    4. Ryan TA

    (2000) The use of phluorins for optical measurements of presynaptic activity

    Biophysical Journal 79:2199–2208.

    https://doi.org/10.1016/S0006-3495(00)76468-X

    • PubMed
    • Google Scholar
53. 1. Savtchenko LP
    2. Rusakov DA

    (2014) Moderate AMPA receptor clustering on the nanoscale can efficiently potentiate synaptic current

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 369:20130167.

    https://doi.org/10.1098/rstb.2013.0167

    • PubMed
    • Google Scholar
54. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez JY
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
55. 1. Scholl B
    2. Thomas CI
    3. Ryan MA
    4. Kamasawa N
    5. Fitzpatrick D

    (2020) Cortical response selectivity derives from strength in numbers of synapses

    Nature 590:111–114.

    https://doi.org/10.1038/s41586-020-03044-3

    • PubMed
    • Google Scholar
56. 1. Soares C
    2. Trotter D
    3. Longtin A
    4. Béïque JC
    5. Naud R

    (2019) Parsing out the variability of transmission at central synapses using optical quantal analysis

    Frontiers in Synaptic Neuroscience 11:22.

    https://doi.org/10.3389/fnsyn.2019.00022

    • PubMed
    • Google Scholar
57. 1. Tagliatti E
    2. Bello OD
    3. Mendonça PRF
    4. Kotzadimitriou D
    5. Nicholson E
    6. Coleman J
    7. Timofeeva Y
    8. Rothman JE
    9. Krishnakumar SS
    10. Volynski KE

    (2020) Synaptotagmin 1 oligomers clamp and regulate different modes of neurotransmitter release

    PNAS 117:3819–3827.

    https://doi.org/10.1073/pnas.1920403117

    • PubMed
    • Google Scholar
58. 1. Virtanen P
    2. Gommers R
    3. Oliphant TE
    4. Haberland M
    5. Reddy T
    6. Cournapeau D
    7. Burovski E
    8. Peterson P
    9. Weckesser W
    10. Bright J
    11. van der Walt SJ
    12. Brett M
    13. Wilson J
    14. Millman KJ
    15. Mayorov N
    16. Nelson ARJ
    17. Jones E
    18. Kern R
    19. Larson E
    20. Carey CJ
    21. Polat İ
    22. Feng Y
    23. Moore EW
    24. VanderPlas J
    25. Laxalde D
    26. Perktold J
    27. Cimrman R
    28. Henriksen I
    29. Quintero EA
    30. Harris CR
    31. Archibald AM
    32. Ribeiro AH
    33. Pedregosa F
    34. van Mulbregt P
    35. SciPy 1.0 Contributors

    (2020) Author correction: scipy 1.0: fundamental algorithms for scientific computing in python

    Nature Methods 17:261–272.

    https://doi.org/10.1038/s41592-020-0772-5

    • PubMed
    • Google Scholar
59. 1. Vyleta NP
    2. Jonas P

    (2014) Loose coupling between Ca2+ channels and release sensors at a plastic hippocampal synapse

    Science 343:665–670.

    https://doi.org/10.1126/science.1244811

    • PubMed
    • Google Scholar
60. 1. Zhang W
    2. St-Gelais F
    3. Grabner CP
    4. Trinidad JC
    5. Sumioka A
    6. Morimoto-Tomita M
    7. Kim KS
    8. Straub C
    9. Burlingame AL
    10. Howe JR
    11. Tomita S

    (2009) A transmembrane accessory subunit that modulates kainate-type glutamate receptors

    Neuron 61:385–396.

    https://doi.org/10.1016/j.neuron.2008.12.014

    • PubMed
    • Google Scholar

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Targeted sensors for glutamatergic neurotransmission" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a myself as Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Kevin J Bender (Reviewer #3).

Comments to the Authors:

We are sorry to report that, after a joint discussion with the reviewers, we have decided that your work will not be considered further for publication by eLife. The reviewers understood the potential value of the approach to develop a synaptically targeted version of iGluSnFR, but raised a number of technical and conceptual issues that we think would require more than modest revisions to support the conclusions.

Reviewer #1 (Recommendations for the authors):

Hao et al. targeted the genetically encoded glutamate sensor iGluSnFR to synapses by fusion with auxiliary subunits of the AMPA receptor, y2 (stargazin) and y8. They report decreased affinity and increased response stability compared to non-anchored iGluSnFR. Early infection with SnFR-γ2 (or just stargazin) blocked EPSCs, suggesting that AMPARs were displaced from their postsynaptic sites. This effect was less extreme when cultures were transfected late (DIV6), but currents were still down by 50% compared to iGluSnFR-transfected neurons, while presynaptic function appeared to be normal.

To analyze the imaging data, the authors developed a strategy to use the signal (stimulation-evoked increase in fluorescence) to select their regions of interest. This allowed them to identify sites of glutamate release, but is not an unbiased sampling of the synaptic population. The authors show only one example of colocalization with homer (Figure 3a), leaving some doubt as to what fraction of indicator molecules was successfully targeted. For a well-targeted indicator, it should be possible to use resting fluorescence spots to select ROIs.

During repeated stimulation at 5 Hz in 2 ca²⁺, SnFR-y2 produced stable responses while iGluSnFR responses decreased. EPSCs in SnFR-y2 neurons were smaller (Figure 4C, note split axis) but displayed similar short-term plasticity (Figure 7C). Comparing neurons, SnFR-y2 was highly correlated with short term facilitation / depression while iGluSnFR signals were not. The authors speculate that the poor correlation of iGluSnFR is due to run-down, but this would manifest as exaggerated depression, not spurious facilitation as the data suggest. So the reason for the improved performance of SnFR-y2 is not entirely clear. The authors then use the SnFR signals to analyze single synapses in autaptic culture. They show very nicely that the glutamate output is a function of extracelluar ca²⁺, providing direct proof for multivesicular release at individual synapses. Full optical quantal analysis requires measurable responses to the release of a single quantum, which SnFR-γ2 and SnFR-γ8 do not seem to provide (Note that the spontaneous fluorescence transients recorded without TTX (Figure 2) are potentially multivesicular events).

In summary, the improvement of the new variants compared to iGluSnFR could be due to their decreased affinity for glutamate, resulting in selection of the strongest synapses and very focal signals, but lack of sensitivity to the fusion of a single vesicle. The price to pay for synaptic targeting, strong alteration of postsynaptic receptor composition, seems relatively high and may prevent widespread adoption of the new variants.

Concerns:

1) Synaptic targeting: The appearance of the targeted indicator is punctate, but the ROIs that were selected by their signal often have no higher resting fluorescence than their surroundings (Figure 2) while the brightest spots apparently produce no signal. By co-expressing homer, the authors tried to quantify colocalization, but show only one single SnFR-y2 spot (n=1) that is colocalized with homer. (As a side remark, iGluSnFR SIGNALS should also be colocalized with homer, but the example is not). In the text, they state that the complexity of expression (?) precluded colocalization analysis. Thus, in spite of the author's efforts, evidence for successful synaptic tagging is lacking, and the schema presented in Figure 9 (88% of SnFR-y2 inside PSD) seems optimistic. All differences in y2-sensor responses compared to iGluSnFR (e.g. spatially restricted responses) could be due to its reduced affinity for glutamate and in consequence, selection bias towards the most powerful synapses.

2) Run-down of iGluSnFR vs. stability of SnFR-y8 (Figure 6): As SnFR-y8 has a lower affinity for glutamate than iGluSnFR, the ROI detection algorithm finds fewer active pixels on SnFR-y8 neurons. I would expect these to correspond to the strongest and most reliable sources of glutamate. iGluSnFR with its higher affinity will also pick up synapses with a small pool of release-ready vesicles. Small synapses are likely to display pronounced depression during a train. So, the stability of SnFR-y8 responses might reflect a selection bias for the strongest synapses. Figure 6g does little to rule out this possibility, since the size of the compound ROI does not reflect the strength of the synapse (but may pool the response of several synapses). The interpretation of the authors, that synaptic targeting somehow stabilized the fluorescence of the SnFR, I find much less likely. Another scenario that cannot be ruled out is stronger illumination of the iGluSnFR cells (perhaps due to low expression levels), resulting in increased photobleaching during the stimulation train. To monitor indicator bleaching, the resting fluorescence (F0) during the train should be reported for both indicators. In figure 6e, fluctuations in the time course are highly correlated across ROIs, which points to an artifact (non-stationary stimulation or illumination). To solve the mystery of run-down, it would be very helpful to see the EPSCs generated by this stimulation in iGluSnFR and in SnFR-y8 neurons.

3) Quantal analysis: As the biological meaning of the quantal parameters is best defined in single synapses, I will first comment on the single synapse experiments (Figure 8, Figure S6, Figure S8). The authors analyze single ROIs at three different calcium concentrations, resulting in increasing response amplitudes. This is interpreted as multi-vesicular release from a single synapse. While I follow this general interpretation, it seems the sensitivity of sensor/imaging system is not sufficient to detect the release of individual vesicles. In 0.5 mM ca²⁺, the recorded SnFR-y8 traces show no difference between simulated and non-stimulated epochs, strongly suggesting that the fluctuations in fluorescence intensity are noise. Consequently, the histograms of 0.5 mM ca²⁺ responses show no separation between failures and sucesses, just a single peak. To assume that this peak corresponds to the release of a single vesicle, the quantal response (q), is wrong. The example traces from iGluSnFR at 0.5 mM ca²⁺ look slightly more promising (Figure S6), but again, the histogram does not show a separation between failures and successes, and the compound histogram treats the entire first peak as the quantal response (i.e., assumes there were no failures in 0.5 mM ca²⁺). A method that lacks the sensitivity to detect single vesicle fusion is not useful for quantal analysis: Due to increasing variance and indicator saturation, the separation between multi-vesicular events will be less clear, not better, than the separation between uni-vesicular events and failures.

4) Quantal analysis of autaptic connections (Figure 7): The authors use the CV method developed for paired recordings, which seems appropriate for an autaptic neuron. Compared to the analysis of single synapses, the meaning of N is less well defined here. Early studies assumed that one synapse can only release one vesicle at a time, which makes N the number of connecting synapses and Pr the synaptic release probability. For autaptic cultures, multi-vesicular release is established, so N is the number of "release sites", several of which can be located in a single synapse. Thus, N must be larger than the number of recorded synapses, which in this case is the number of ROIs.

It is very difficult to understand Figure 7. Trains of 5 AP are evoked at 5 Hz ten times in 3 different Ca concentrations and recorded in an unspecified number of ROIs (synapses). In panel c, facilitation/depression during the train is compared for electrical and optical recordings. I assume these points (n=6) correspond to different cells, not different ROIs (please indicate the point corresponding to the cell analyzed in panels a and b). SnFR-y2 responses are better correlated with the electrical responses in 2 and 4 Ca, but SnFR-y2 synapses tend towards depression compared to iGluSnFR synapses. This population difference in EPSC dynamics raises the nasty possibility of presynaptic effects caused by SnFR-y2 expression (but due to low sample size, may just be a fluke).

In Figure 7D, mean and standard deviations are calculated, but the basis is not clear. I am assuming each ROI produces 5 points in panel D, each averaged over the 10 repeats. It would be helpful to explain this more clearly. The CVs (slopes in panel D) are related to N and Pr. As the authors state, they end up with 3 numbers for an equation system with 4 variables. It is not clear to me how fitting can help in this situation (as there should be no error, regardless which number is chosen for N). So perhaps panel F shows the numbers they chose for N, perhaps based on the number of ROIs? (The legend of F is cryptic. Please indicate in E and F which points correspond to the cell shown in D.) This would be based on the assumption of univesicular synapses, though, which they refute in Figure 8. In any case, they point out that this N is similar to the N reported by Bekkers and Stevens, who were interested in the number of release sites connecting two neurons. The numbers (N = 4-7) seem very low for a connection producing 2 nA EPSCs (400 pA per uniquantal synapse?). It could of course be argued that the autaptic connection is made of many more synapses than the few that are sampled optically. The N determined by optical quantal analysis, however, has to be higher than the number of ROIs (active synapses), it cannot be lower. If the N is assumed to be equal to the number of ROIs (uniquantal synapses), it makes no sense to call the process quantal analysis (as in this case, the distribution of N values (Figure 7F) is not the result of any fitting procedure). Without understanding the basis of N, I cannot interpret the meaning of Pr (synaptic? vesicular?).

Suggested improvements:

Provide conclusive evidence for successful synaptic targeting, e.g. by homer/y2 co-transfection of individual neurons and high-resolution imaging. Targeting is expected to be most specific at low expression levels.

I am not sure how to improve the quantal analysis (more excitation light?). A symmetrical failure peak should appear exactly at df/f = 0 which is absent from all histograms. Perhaps the problem is background subtraction (leading to division by zero) or the handling of negative df/f values. If there is no way to separate failures from low probability release events in individual trials, I will remain skeptical about multi-peaked compound histograms apparently separating 4 from 3 simultaneously released vesicles.

Please commit to a clear definition / interpretation of the extracted parameters from the coefficient of variance method, specifically in relation to the number of ROIs. At present, the biological interpretation of N and Pr with respect to autaptic synapses is unclear, at least to me.

Reviewer #2 (Recommendations for the authors):

1. According to the authors, one principal advantage of their approach is that SnFR-y2/8 provides a more 'spatially precise' signal compared with iGluSnFR. Clearly, an optical sensor that is proposed as expressed within a nanoscopic membrane domain will yield a more spatially constrained signal compared with the sensor expressed evenly over cell membranes. However, this simply reflects the sensor distribution properties rather than anything else. In words, SnFR-y2/8 will tend to report the brightest glutamate signal where SnFR-y2/8 is accumulated, rather than where the glutamate concentration is highest (e.g., release site proximity). In contrast, the sensor homogenously distributed in space, such as the original iGluSnFR variants, should provide unbiased readout of glutamate hotspots. It appears therefore that the authors strategy is somewhat self-defeating.

2. The authors use TIRF imaging, and single-line diode lasers as an excitation source. This type of imaging is only suitable for monolayer cultures: whether their sensor will be efficient when imaged in organized brain tissue using two-photon excitation is not clear. The claim on a methodological advance appears therefore premature.

3. The other key statement is that iGluSnFR is prone to photobleaching (rundown) more than is SnFR-y2/8. First, fluorophore's photobleaching properties in 1P as opposed to 2P mode could be very different, which has not been addressed or explored. Second, this observation is surprising because several recent studies have documented a fairly stable iGluSnFR signal over multiple cycles of glutamate release imaged at 'quantal' resolution, both in 1P and 2P excitation regimes, both in cultures and in acute slices (e.g., Tagliatti et al., 2020 PNAS 117: 3819; Jensen et al. 2019 Nat Commun 10: 1414). The authors do not seem familiar with these studies. Dye photobleaching depends on multiple imaging parameters starting with laser power: this has not been investigated consistently in the present work.

4. Stargazin overexpression has been used as a principal tool for the iGluSnFR targeting to synapses, but the authors report that this renders a proportion of synapses nonresponsive to glutamate (Figure 4). That the method interferes with the physiological integrity of synaptic circuits, or at least requires some additional experiment-specific manipulations to minimize it, does not speak in its favor.

5. The signal-to-noise ratio of the SnFR-y2 signal does not appear improved compared with that of iGluSnFR (Figure 8A).

6. The quantal-analysis histograms presented here (Figure 8C-D) appear noisier hence less reliable than similar or related analyses in the aforementioned publications that employed iGluSnFR.

Reviewer #3 (Recommendations for the authors):

Hao and colleagues developed new variants of the glutamate sensor iGluSnFR, termed SnFR-γ2 and SnFR-γ8, that are fusion proteins with postsynaptic density (PSD) proteins Stargazin and γ-8. This chimeric protein thus localizes specifically at the PSD. These new variants are characterized, using heterologous expression systems and hippocampal neuron microisland cultures, allowing one to monitor autapses with simultaneous electrical and optical access. Overall, SnFR-γ2 outperforms traditional iGluSnFR in terms of signal localization; presumed single synapses are observed with limited "spillover" of signal to neighboring regions, and imaged transients are amenable to traditional noise analysis. Some concern regarding competition for PSD membrane is raised due to overexpression of these variants, but it appears that such overexpression artifacts can be avoided by ensuring that SnFR-γ2 is delivered after synapses are largely formed. This could be an important tool for the field, as it improves one's ability to resolve the activity of single glutamatergic synapses with sufficient signal to noise. Work here has shown feasibility in cultures systems. Future work will need to show similar performance levels in acute slice and in vivo, though based on past observations with iGluSnFR, this performance is likely within reach.

The main question I have is one of extensibility: can this approach work in more intact systems, or even in vivo? I recognize that asking such a question involves an entirely new dataset, and am not proposing that the authors engage in such an effort after already doing an excellent job characterizing these GluSnFR variants. Rather, I'd hope that the authors would expand on their discussions, which is largely focused on questions of release dynamics observed with their sensors, to also include potential technical advantages or limitations of these variants in other preparations.

Comments below are aimed at improving interpretation of data reported herein:

1) A control for infection is needed for autapse data. Please make parallel recordings in cells infected with control viruses that lack any glutamate sensor to determine if these currents are in the normal range at these ages.

2) Data in Figure 5C and F should be analyzed quantitatively by calculating correlation coefficients for each pair of data. If there are differences in the relative separation of each region (I understand that these were chosen by hand) then a potential comparison could be made by plotting correlation coefficients vs. centroid distance of paired ROIs.

3) Data in Figure 7C, 0.5 mM. There is an obvious outlier near 7 P5/P1 for the SnFR-gamma2. This is likely due to a very low P1 value for this one cell, indicative of excess failures. I'd be curious to know if any correlation holds if this one datapoint is held out of the dataset. If so, perhaps the authors could explain this observation in the main text.

4) Figure 7, data related to optical quantal analysis. Considering that these are autaptic recordings, with superb electrical access, one should be able to perform traditional electrophysiological quantal analysis and determine whether SnFR is allowing for identification of all synapses. This is a critical analysis that should be made on a cell-by-cell basis, paired with optical analyses; however, if this is not feasible at this time, some information could be gleaned from separate recordings, given that you observed fairly consistent numbers of sites optically (4 to 7). This would address your concern that the detection methods bias towards high Pr synapses. Though, if there are more synapses made a significant distance from the coverslip, then they would never be imaged under TIRF microscopy, obviating this request.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Targeted sensors for glutamatergic neurotransmission" for further consideration by eLife. Your revised article has been evaluated by Gary Westbrook (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

This manuscript is an attempt towards synapse specificity for glutamate probes. The current probe's utility is limited to specific cases, and it has what could be considered advantages and drawbacks, even in that specific case (CMOS imaging of cultured cells). It is commendable how carefully the authors characterized side effects of this sensor on synaptic function. The reviewers agreed that several issues require clarifications in the text.

1. Please explain that any glutamate rises or waves in the tissue, any significant glutamate spillover signaling, or even fluctuations in tissue optical conditions, could be falsely perceived as local synaptic signals by an PSD-constrained sensor rather than by an evenly distributed sensor. This is a key limitation of the current method.

2. The authors seem to insist that the use of iGluSnFR or any sensor labeling is generally disadvantageous. It is advantageous in most cases. In terms of this sensor's fluorescence properties, there is no evidence for improved performance in the manuscript.

3. Please consider the specific points raised by the reviewers and address them in the text to the extent possible.

Reviewer #1 (Recommendations for the authors):

The authors have provided their explanations and rebuttal regarding the previous comments, albeit without new experimental evidence.

1. It appears that the authors did not fully understand the main objection. Or perhaps it was not explained clearly enough. To reiterate: SnFR-y2/8 expressed locally at the synapse may in fact sense and report glutamate that is released elsewhere in the vicinity, thus giving a false impression of its local synaptic release. In other words, SnFR-y2/8 may report the spillover signal as well as the actual synaptic signal.

It is true that, in 2PE mode, in organized brain tissue, iGluSnFR will report optical signal integrated, due to diffraction of light, within the PSF depth of ~800 nm, etc., and thus may include glutamate release events, if any, from nearest synaptic neighbors. However, the same will happen if such neighbors express SnFR-y2/8: their optical signal will still be integrated within the diffraction-limited PSF in 2PE imaging mode.

In the case when no glutamate spillover ever occurs in the preparation of interest, both iGluSnFR and SnFr-y2/8 will report true synaptic signals when such happen. In this case, however, glutamate rises or waves arising from other sources, such as astroglia, will still be reported by SnFr-y2/8 very locally, giving an impression of local synaptic events – whereas iGluSnFR will report the entire glutamate 'landscape'. Thus, SnFr-y2/8 does not seem to have any principal spatial-resolution advantage over iGluSnFR.

2. Authors' attention is drawn to yet another recent publication (Mendonca et al. 2022 Nat Commun 3497) showing an excellent stability, S/N ratio, etc. for highly localized glutamate release events detected with iGluSnFR. That this sensor is performing not as satisfactory in the present authors' hands cannot be a basis for their extrapolated claim.

3. The authors concede that the S/N ratio of their probe readout is probably no better than that of iGluSnFR.

4. The authors have made no attempt to check their probe performance in 2PE mode and/ or in organized brain tissue.

5. The amplitude histograms shown in Figure 9C-D do not match satisfactorily the best-fit quantal analysis curves. While the authors acknowledge the difficulty, the reasons for displaying unsatisfactory quantal analysis data are not clear.

6. The entire concept would have made a much greater impact if the authors aimed to express the sensor at a specific, functionally or genetically distinct, sub-population of synapses.

Reviewer #2 (Recommendations for the authors):

The authors have addressed concerns raised by all three reviewers to the best of their ability. Their commentary that untagged SnFR is likely detecting large hotspots of both synaptic and spillover is taken well. Though the concern regarding how well the tagged variant works remains in question. If one can only resolve a few synapses (less than 10 per cell were analyzed) of what the authors suggest are ~100 potential autapses, are these few ROIs representative of the whole? This may be a good point of discussion.

Given the way in which eLife is modifying its review process and overall publishing criteria, this work should be accepted. It represents a new tool that is well characterized. Advantages and flaws are discussed well.

Reviewer #3 (Recommendations for the authors):

In their revised version, the authors address questions about synaptic localization of their GEGI by demonstrating good overlap with the PSD95 signal (new Figure 3). The explanation of Figure 8 has been much improved, quantal parameters are now well defined. The correlation between electrical and optically measured depression during a train (Figure 8c) is indeed much better for their targeted indicator compared to iGluSnFR, which is a strong argument that synaptic (and not extrasynaptic) glutamate is reliably measured by their sensor. In this context, it is an advantage of the autaptic model that all synapses on a given neuron have the identical history of activity and therefore express similar (but cell-specific) short-term plasticity. I do not like the analysis of rundown in Figure 7F: Splitting a continuous distribution into 2 groups, using an arbitrary threshold, then counting the number of cases (ROIs) on either side of the threshold, is not good statistical practice. The analysis in Figure S5 I like much better, it shows no significant difference between the two indicators. As the authors cannot offer a convincing mechanistic explanation why a difference in rundown would be expected, I suggest downplaying this point (getting rid of Figure 7E-G, sticking with the message of S5).

Apart from this quibble, the science is sound and well presented. The separation of quantal histogram peaks is impressive and certainly aided by localizing the indicator to the places of highest glutamate concentration. Extrasynaptic indicator molecules, exposed to a near-continuous range of glutamate concentrations, would be expected to widen the peaks considerably (as shown in the iGluSnFR example in Figure 9c). For future tool development and targeting efforts, the technical information and precise measurements are very useful, even as a somewhat cautionary tale with regard to potentially severe side effects of tool expression.

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Reviewer #1 (Recommendations for the authors):

Hao et al. targeted the genetically encoded glutamate sensor iGluSnFR to synapses by fusion with auxiliary subunits of the AMPA receptor, y2 (stargazin) and y8. They report decreased affinity and increased response stability compared to non-anchored iGluSnFR. Early infection with SnFR-γ2 (or just stargazin) blocked EPSCs, suggesting that AMPARs were displaced from their postsynaptic sites. This effect was less extreme when cultures were transfected late (DIV6), but currents were still down by 50% compared to iGluSnFR-transfected neurons, while presynaptic function appeared to be normal.

We disagree with this otherwise good summary in only one respect: we do not claim any change in affinity. We apologise that we did not underline that a change of apparent glutamate affinity by a factor of 2 is not meaningful. We now added text near line 153 to emphasize that this is a slight change, compared to other published variants.

To analyze the imaging data, the authors developed a strategy to use the signal (stimulation-evoked increase in fluorescence) to select their regions of interest. This allowed them to identify sites of glutamate release, but is not an unbiased sampling of the synaptic population. The authors show only one example of colocalization with homer (Figure 3a), leaving some doubt as to what fraction of indicator molecules was successfully targeted. For a well-targeted indicator, it should be possible to use resting fluorescence spots to select ROIs.

Thank you. We take this criticism seriously. Originally, we worked hard to get a good measure of colocalization from the sensor itself in live cell imaging but, as the referee identified, this was not possible. We now include new Figure 3, where we measure the localization of SnFR variants by immunofluorescence microscopy on live-labelled, fixed cells. These experiments show the enrichment at synaptic sites for SnFR-g2. However, the enrichment is not absolute, presumably because sensors are not anchored at the PSD but rather exchange during an experiment. The mild enrichment perhaps relates to the somewhat paradoxical observation that we do not have major bleaching despite tethering.

During repeated stimulation at 5 Hz in 2 ca²⁺, SnFR-y2 produced stable responses while iGluSnFR responses decreased. EPSCs in SnFR-y2 neurons were smaller (Figure 4C, note split axis) but displayed similar short-term plasticity (Figure 7C). Comparing neurons, SnFR-y2 was highly correlated with short term facilitation / depression while iGluSnFR signals were not. The authors speculate that the poor correlation of iGluSnFR is due to run-down, but this would manifest as exaggerated depression, not spurious facilitation as the data suggest. So the reason for the improved performance of SnFR-y2 is not entirely clear. The authors then use the SnFR signals to analyze single synapses in autaptic culture. They show very nicely that the glutamate output is a function of extracelluar ca²⁺, providing direct proof for multivesicular release at individual synapses. Full optical quantal analysis requires measurable responses to the release of a single quantum, which SnFR-γ2 and SnFR-γ8 do not seem to provide (Note that the spontaneous fluorescence transients recorded without TTX (Figure 2) are potentially multivesicular events).

During repeated stimulation, some iGluSnFR responses ran down dramatically, and some did not. But the advantage of SnFR-Gamma2 is that almost no responses ran down. So there is no need to select between them.

We do not agree that rundown would manifest as depression (STD). Rundown (the overall loss of signal) is happening not during the train (that would be depression) but instead to all the responses of certain ROIs as the experiment progresses (see now Figure 7). This effect tends to make the later responses closer to the noise and therefore apparently more random than the earlier ones. This effect does not happen with our sensor (Figure 7).

For full optical quantal analysis, in the ideal case, single responses can be well resolved. The reviewer is right, our sensors are not ideal in this regard. However, we feel that idea that this is absolutely required is challenged by our excellent estimation of quantal size from global fitting at different calcium concentrations.

We now added a note to point out that the poor resolution of the single quanta is a drawback (near line 478).

In summary, the improvement of the new variants compared to iGluSnFR could be due to their decreased affinity for glutamate, resulting in selection of the strongest synapses and very focal signals, but lack of sensitivity to the fusion of a single vesicle. The price to pay for synaptic targeting, strong alteration of postsynaptic receptor composition, seems relatively high and may prevent widespread adoption of the new variants.

Briefly, no, the affinity is not changed enough for this to be the case. We measured this in close to native conditions (mammalian cells, not in a cuvette) in Figure 1. We did not particularly expect widespread adoption of these variants (a few labs have requested them from Addgene).

What we show is the principle of a massive quantitative improvement (Figure 6 and 7) from targeting, and this should be taken seriously. The performance of our sensor is somewhat different to the prevailing view that tethering the sensor should lead to bleaching. These are important observations for the field in general.

During our process of revising this manuscript, the team at Janelia have developed their new best in class indicator, including the Stg-C-terminus to localize it. It seems this works very well in their hands too but is not so bright in vivo. They recommend it for in vitro work. This development follows exactly from what we present in this manuscript, and their data complement ours. It seems the principle will have good adoption. We are very glad that open science (we put our MS on Biorxiv and gave the plasmids and vectors away) helping to promote innovation in this way.

Concerns:

1) Synaptic targeting: The appearance of the targeted indicator is punctate, but the ROIs that were selected by their signal often have no higher resting fluorescence than their surroundings (Figure 2) while the brightest spots apparently produce no signal. By co-expressing homer, the authors tried to quantify colocalization, but show only one single SnFR-y2 spot (n=1) that is colocalized with homer. (As a side remark, iGluSnFR SIGNALS should also be colocalized with homer, but the example is not). In the text, they state that the complexity of expression (?) precluded colocalization analysis. Thus, in spite of the author's efforts, evidence for successful synaptic tagging is lacking, and the schema presented in Figure 9 (88% of SnFR-y2 inside PSD) seems optimistic. All differences in y2-sensor responses compared to iGluSnFR (e.g. spatially restricted responses) could be due to its reduced affinity for glutamate and in consequence, selection bias towards the most powerful synapses.

We agree, our original analysis of colocalization with Homer in live cells was not convincing. We think this is because of a large amount of ER labelling by all membrane-bound constructs. We now point this out directly in the text near to line 201. We have addressed this concern with a new Figure 3 where we show an enrichment of surface SnFR-g2 (using a GFP antibody) and SnFR-g8 against PSD95 sites using immunohistochemistry.

To the point about iGluSnFR signal and colocalization: the Homer signal trivially corresponds to a synapse where there was no release in this part of the movie.

We take the point that the sketch (now) Figure 10 might appear that we are claiming something optimistic, but we do not mean imply any ratio of 88%. The ratio of extrasynaptic membrane to synaptic membrane is massive, and we do not depict this to scale, nor do we claim to. We added a note to the legend to make this clear.

We think that a really important result in our paper is that by using a non-targeted high-affinity sensor (iGluSnFR), you may unfortunately sample glutamate that is irrelevant for synaptic responses. This is consistent with the recent preprint from Janelia that provides evidence that iGluSnFR is excluded from synapses, and prefers extrasynaptic glutamate. All the successful work so far has used presynaptic anatomy to restrict the spatial aspect of expression, or some kind of filtering in the analysis to restrict it to “good” spots. We show that many of the responses of iGluSnFR to evoked release are massive in area, making their relation to synaptic glutamate implausible. This fact has been overlooked or ignored in previous work that concentrated on a presynaptic locus.

Finally, there is now evidence (from the Janelia Farm preprint) that the Stg C-terminus (inspired by our work) gives the best synaptic localization of iGluSnFR.

2) Run-down of iGluSnFR vs. stability of SnFR-y8 (Figure 6): As SnFR-y8 has a lower affinity for glutamate than iGluSnFR,

The affinity of the three sensors (iGluSnFR, SnFR-g2, SnFR-g8) is essentially the same.

The ROI detection algorithm finds fewer active pixels on SnFR-y8 neurons. I would expect these to correspond to the strongest and most reliable sources of glutamate. iGluSnFR with its higher affinity will also pick up synapses with a small pool of release-ready vesicles. Small synapses are likely to display pronounced depression during a train. So, the stability of SnFR-y8 responses might reflect a selection bias for the strongest synapses. Figure 6g does little to rule out this possibility, since the size of the compound ROI does not reflect the strength of the synapse (but may pool the response of several synapses). The interpretation of the authors, that synaptic targeting somehow stabilized the fluorescence of the SnFR, I find much less likely. Another scenario that cannot be ruled out is stronger illumination of the iGluSnFR cells (perhaps due to low expression levels), resulting in increased photobleaching during the stimulation train. To monitor indicator bleaching, the resting fluorescence (F0) during the train should be reported for both indicators. In figure 6e, fluctuations in the time course are highly correlated across ROIs, which points to an artifact (non-stationary stimulation or illumination). To solve the mystery of run-down, it would be very helpful to see the EPSCs generated by this stimulation in iGluSnFR and in SnFR-y8 neurons.

We are grateful for this careful critique and the excellent ideas for giving more insight into what is going on. The expression levels are similar between the different constructs, although the baseline fluorescence was generally a bit less for the SnFR-g2 constructs. The illumination was not changed between different experiments, which were always done on the same day. We now note this (near line 296), and we made a new supplementary figure S5 to look show all of these data, to answer this point. We make comparisons between iGluSnFR and SnFR- g2, not SnFR-g8 because few cells in this condition gave enough ROIs for this comparison to make sense.

We now report in new Supplementary Figure S5 the baseline fluorescence (F0) for the cells where we also recorded EPSCs. We show exemplary traces for the cells that we analyse in detail in other sections, and summarise the (lack of) rundown for the F0 signal across all the cells. The baseline fluorescence was less for SnFR-g2. We never mentioned bleaching in our original submission and we now explicitly point out that there is no substantial bleaching (near line 311). We also made the comparison that the referee suggests between EPSC rundown and fluorescence rundown. We describe it near line 320.

The observation of the correlation of peaks across ROIs is a fascinating prospect. It is much clearer for iGluSnFR than for SnFR. The strong correlation that the reviewer sees in now-Figure 7E is related to the phenomenon shown in now-figure 6 – the responses of iGluSnFR at neighbouring sites are highly correlated. Large ROIs are also to some extent contiguous in SnfR (see merged view, panel B) and that therefore this is a spurious correlation. If you don’t manually discard neighbouring ROIs, this is what you get. We added a note to clarify in the figure legend.

The idea that we select for stronger synapses is probably correct, even if we do not agree with the argument based on sensor affinities for glutamate (which are essentially the same). We added a note to mention this at line 617. But weak synapses contribute little to electrophysiological responses, and possibly little to physiological responses in general. iGluSnFR also undoubtedly reports extrasynaptic glutamate too (the responses are too broad for anything else).

Our correlation between depression in electrophysiology and fluorescence for SnFR-gamma2 is exceptionally good. For iGluSnFR it is non-existent. We think this is an argument that targeting selects the most meaningful signals in a physiological sense, and that we have demonstrated this.

3) Quantal analysis: As the biological meaning of the quantal parameters is best defined in single synapses, I will first comment on the single synapse experiments (Figure 8, Figure S6, Figure S8). The authors analyze single ROIs at three different calcium concentrations, resulting in increasing response amplitudes. This is interpreted as multi-vesicular release from a single synapse. While I follow this general interpretation, it seems the sensitivity of sensor/imaging system is not sufficient to detect the release of individual vesicles. In 0.5 mM ca²⁺, the recorded SnFR-y8 traces show no difference between simulated and non-stimulated epochs, strongly suggesting that the fluctuations in fluorescence intensity are noise. Consequently, the histograms of 0.5 mM ca²⁺ responses show no separation between failures and sucesses, just a single peak. To assume that this peak corresponds to the release of a single vesicle, the quantal response (q), is wrong. The example traces from iGluSnFR at 0.5 mM ca²⁺ look slightly more promising (Figure S6), but again, the histogram does not show a separation between failures and successes, and the compound histogram treats the entire first peak as the quantal response (i.e., assumes there were no failures in 0.5 mM ca²⁺). A method that lacks the sensitivity to detect single vesicle fusion is not useful for quantal analysis: Due to increasing variance and indicator saturation, the separation between multi-vesicular events will be less clear, not better, than the separation between uni-vesicular events and failures.

We have been quite open about limits of the quantal analysis in the paper. The major point is that a targeted sensor is better for quantal analysis, because many more ROIs are usable (can be fit) and we very clearly see regularly-spaced dips in, for example, the 2 mM histogram for SnFR-g2, but not for regular SnFR. This difference matches well previous work where spatially confined SnFR could give histograms with dips (Jensen, T. P., et al. 2019. Nat Commun 10, 1414; Duerst, C., et al. 2020. Biorxiv), but unconstrained SnFR could not (Soares, C., et al. 2019. Front Synaptic Neurosci 11, 22).

What we want to show here is that we can estimate N and Pr from a comparatively simple set of measurements, only 150 stimuli, because the SnFR-g2 construct is stable, unlike responses from iGluSnFR. We already mentioned in the discussion that it is hard to resolve the release of individual vesicles with SnFR-g2 at 0.5 mM (the reviewer mentions SnFR-g8, we do not show this, perhaps SnFR-g2 is intended?). We now reiterated this point at the end of the results (line 477), However, global fitting gives us an excellent estimate of quantal size. In this global fit, we found that omitting the failure peak helps to get a more robust fit. We are fitting 20+ ROIs in series, semi-automatically.

We understand the reviewer’s concern, could not the first peak just be noise? Please look at the pooled histogram from SnFR-g2 (now Figure 9D). There are three perfectly spaced peaks (multiples of 0.06), with a dip after each one. In fact, we get an excellent estimate of quantal size (which we use for the rest of the fits) from the spacing of these peaks. This is not consistent with the first peak being noise from overzealous fitting. These are typical data from a single ROI. We added text to describe this aspect better (after line 875 in the results).

Of course, the reviewer is right that detection of multivesicular events would be hurt by saturation and increasing variance, but for our sensor this appears only to happen after the 3^rd peak or so. We used the summed histogram to estimate the peak spacing. Our fit for individual calcium concentrations assumes that there is a failure peak at zero. Our method of peak detection does not give negative numbers, meaning we cannot plot a true “failure” peak (symmetrical around zero) as the reviewer would like. But we estimate failures very well with 0.5 mM glutamate – they are in the first bin of the histogram, they are fitted, and we have a P_r of about 20% across the 6 cells we include for this condition. We did initially try to do peak detection including negative values from failures, but the results were much less reliable, so we settled on this approach to give more reliable output from our detection routine (we give details around line 875). However, we will revisit this point and endeavour to obtain a more convincing “failure peak” in future work.

It seems like new versions of iGluSnFR using the principle that we established here will be brighter, and this will combine favourably to give a really nice quantal response. We will explore modifying our software accordingly but this is beyond the scope of this work.

The difference between stimulated and non-stimulated epochs is not pretty, we admit, but there is a difference. We already showed this in detail in the supplementary material (now Figure S7).

4) Quantal analysis of autaptic connections (Figure 7): The authors use the CV method developed for paired recordings, which seems appropriate for an autaptic neuron. Compared to the analysis of single synapses, the meaning of N is less well defined here. Early studies assumed that one synapse can only release one vesicle at a time, which makes N the number of connecting synapses and Pr the synaptic release probability. For autaptic cultures, multi-vesicular release is established, so N is the number of "release sites", several of which can be located in a single synapse. Thus, N must be larger than the number of recorded synapses, which in this case is the number of ROIs.

We apologise that the answers to these concerns were not made clear enough in the original submission. If we follow the logic outlined above, it is important to note that we are looking at the coefficient of variation at individual ROIs. Therefore, N is either the number of release sites in an ROI, or more generally the number of release sites multiplied by the number of synapses in the ROI.

We compare single ROI data to literature data on single connections from electrophysiology- in our opinion this is quite meaningful.

We now explain this point around line 393, and commit to explicit definitions of N and P_r.

It is very difficult to understand Figure 7. Trains of 5 AP are evoked at 5 Hz ten times in 3 different Ca concentrations and recorded in an unspecified number of ROIs (synapses). In panel c, facilitation/depression during the train is compared for electrical and optical recordings. I assume these points (n=6) correspond to different cells, not different ROIs (please indicate the point corresponding to the cell analyzed in panels a and b). SnFR-y2 responses are better correlated with the electrical responses in 2 and 4 Ca, but SnFR-y2 synapses tend towards depression compared to iGluSnFR synapses. This population difference in EPSC dynamics raises the nasty possibility of presynaptic effects caused by SnFR-y2 expression (but due to low sample size, may just be a fluke).

We apologise that Figure 7 (now Figure 8) was hard to understand and led the referee astray. In panel C, we show the results from 6 cells for each condition. SnFR-g2 shows an indistinguishable range of depression/potentiation in electrophysiology from iGluSnFR (range of data on y-axis for each condition for iGluSnFR and SnFR-g2 respectively: at 2mM ca²⁺ : 0.8-1.1 and 0.7-1.25, and at 4 mM ca²⁺ 0.58-0.8, 0.53-0.78). It’s a small sample size but there is no systematic “nasty” effect of the sensor.

On the other hand, the correlation between the extent of depression in electrophysiology and the automatically-selected ROIs is stunning (but for SnFR-g2 only).

We now indicate the cell analysed in panels a and b with a filled symbol and note this in the legend. We now summarise in each figure panel what the experiment is with an explanatory title.

In Figure 7D, mean and standard deviations are calculated, but the basis is not clear. I am assuming each ROI produces 5 points in panel D, each averaged over the 10 repeats. It would be helpful to explain this more clearly. The CVs (slopes in panel D) are related to N and Pr. As the authors state, they end up with 3 numbers for an equation system with 4 variables. It is not clear to me how fitting can help in this situation (as there should be no error, regardless which number is chosen for N). So perhaps panel F shows the numbers they chose for N, perhaps based on the number of ROIs? (The legend of F is cryptic. Please indicate in E and F which points correspond to the cell shown in D.) This would be based on the assumption of univesicular synapses, though, which they refute in Figure 8. In any case, they point out that this N is similar to the N reported by Bekkers and Stevens, who were interested in the number of release sites connecting two neurons. The numbers (N = 4-7) seem very low for a connection producing 2 nA EPSCs (400 pA per uniquantal synapse?). It could of course be argued that the autaptic connection is made of many more synapses than the few that are sampled optically. The N determined by optical quantal analysis, however, has to be higher than the number of ROIs (active synapses), it cannot be lower. If the N is assumed to be equal to the number of ROIs (uniquantal synapses), it makes no sense to call the process quantal analysis (as in this case, the distribution of N values (Figure 7F) is not the result of any fitting procedure). Without understanding the basis of N, I cannot interpret the meaning of Pr (synaptic? vesicular?).

Apologies, we could have again explained much better here in order not to leave the reader confused. First of all, the autapse has many connections to produce nA synaptic responses, but we are looking at individual ROIs and as explained above, the N value refers to each ROI. We took the mean and SD from individual ROIs, for each pulse in the train (average and SD of the 10 first pulses (one from each train of 5), average and SD of the 10 second pulses, and so on). These averages and SDs were similar across the 5 pulses in the trains so we averaged them and plotted these points. Therefore, each point represents the mean and SD of a single ROI during the 10 x 5 pulse train stimulation. We could have separated out the responses according to their order in the train, but it seemed excessive, particularly given the limitations of this CV method.

Therefore the N, as we explained in the text, is related to the number of sites per ROI. We think this relates well to the original (and common) meaning from Katz: the probability of vesicular release from a synapse. This is a global analysis, we could have calculated for each ROI (synapse) because we have these ratios. But as we noted, the fit is not well defined for N. It means overall that the approach is not particularly satisfactory (and we would not recommend it, we now state this explicitly in the text around line 454).

We are grateful to the referee for asking the question as to how N is constrained. We went back an investigated this point again, which deserved more attention. What is really clear is that N < 3 does not work– the associated Pr values are nonsensically saturated at a high P. We now show this in a new panel (Figure 8E), and took an alternative approach. We now took the lowest value of N which gave Pr < 50% at 0.5 mM Calcium. There might be an interesting connection here to the optical method selecting for stronger synapses – the Pr values from CV analysis are generally too high compared to electrophysiological estimates which have been studied in much more detail.

We now indicate in panels E, F and G and in the legend which points and lines correspond to the relevant cells in D (with solid line and solid symbols). We now try to explain better (around line 400) and added notes to the text to point out why we think that we observe multivesicular release optically.

Suggested improvements:

Provide conclusive evidence for successful synaptic targeting, e.g. by homer/y2 co-transfection of individual neurons and high-resolution imaging. Targeting is expected to be most specific at low expression levels.

We appreciate this suggestion. We now live labelled SnFR and variants with antibodies, and compare with PSD labelling in fixed cells. Labelling SnFR variants allowed us to see the mild enrichment of our targeted sensors at synapses (new Figure 3).

I am not sure how to improve the quantal analysis (more excitation light?). A symmetrical failure peak should appear exactly at df/f = 0 which is absent from all histograms. Perhaps the problem is background subtraction (leading to division by zero) or the handling of negative df/f values. If there is no way to separate failures from low probability release events in individual trials, I will remain skeptical about multi-peaked compound histograms apparently separating 4 from 3 simultaneously released vesicles.

The symmetrical failure peak is not seen due to the way in which we have done the analysis. The referee is right, we do not take negative values. We now discuss this explicitly in the methods (starting at line 878). Failures are simply in the first bin (starting at zero), and are accounted for in the fits. We are also skeptical about the high order events (like 3 from 4) and now note explicitly in the text that the peaks above 3 or 4 components are washed out (around line 450).

Please commit to a clear definition / interpretation of the extracted parameters from the coefficient of variance method, specifically in relation to the number of ROIs. At present, the biological interpretation of N and Pr with respect to autaptic synapses is unclear, at least to me.

As outlined above, we rewrote the text here. We commit to clear definitions of N and Pr (around line 392), and emphasise that this CV approach is not ideal – we think the quantal analysis of histograms for individual ROIs is better.

Reviewer #2 (Recommendations for the authors):

1. According to the authors, one principal advantage of their approach is that SnFR-y2/8 provides a more 'spatially precise' signal compared with iGluSnFR. Clearly, an optical sensor that is proposed as expressed within a nanoscopic membrane domain will yield a more spatially constrained signal compared with the sensor expressed evenly over cell membranes. However, this simply reflects the sensor distribution properties rather than anything else. In words, SnFR-y2/8 will tend to report the brightest glutamate signal where SnFR-y2/8 is accumulated, rather than where the glutamate concentration is highest (e.g., release site proximity). In contrast, the sensor homogenously distributed in space, such as the original iGluSnFR variants, should provide unbiased readout of glutamate hotspots. It appears therefore that the authors strategy is somewhat self-defeating.

This comment to some extent represents the canonical view in the field, which we felt with this work we convincingly rebut. The reviewer is entitled to their opinion, but there are some observations in our work and published work that we believe speak against it. Most importantly, the team at Janelia have adopted exactly our approach (targeting with the Stg-C-terminus) for their new best in class indicator.

A. We show that iGluSnFR ROIs from postsynaptic sites (detected in an unbiased way) are massive. They cannot correspond to synapses. Perhaps this is only in the autapse system but we think it’s why most successful quantitative work to date was done in the Schaffer Collateral terminal. A non-targeted sensor is limited in terms of what you can look at. In contrast, the principle underlying our sensor should allow less bias in the choice of neuron, type of neuron, or its geometry.

B. Synapses can be rather densely packed on the dendrite (>1 per micron). In this case, regular SnFR is not useful, the signals from neighbouring synapses will overlap. Particularly, the dominant signal from SnFR could be spillover, whenever release occurs reasonably frequently. This effect might also cause saturation of the (high-affinity) sensor, compressing responses in an unpredictable way.

C. If the referee’s contention that iGluSnFR should be used as an unbiased reporter of glutamate hotspots were correct, one might expect published, quantitative work expressing iGluSnFR at postsynaptic sites. This is not the case. Rather, almost all quantitative work to date used presynaptic terminals to get spatial restriction to individual sites. In contrast, one paper using CA1 pyramidal cell dendrites needed extensive mathematical modelling to approach quantitative insight (and did not see dips in the quantal histogram, Soares, C., et al. (2019). Front Synaptic Neurosci 11, 22.).

D. The idea that PSD ligands might accumulate at a non-trivial distance from release sites is refuted by published work (for example Li, T. P., and Blanpied, T. A. (2016). Front Synaptic Neurosci 8, 19.) Particularly, the PSD and the release sites are aligned on the nanoscale. We now include new data on colocalization to show the better overlap at PSD sites – the new Janelia preprint shows similar data.

E. We also show that a non-targeted sensor can report spurious signals that should be discarded. Unfortunately, spatial bias is not the only possible source of bias.

Overall, our sensor certainly has some limitations. This being said, we feel we have given enough evidence to indicate that the approach is not self-defeating. In fact, it seems to be taken up as the way forward.

We now mention that a potential disadvantage of our sensor is that non-synaptic signals would be missed (around line 623).

2. The authors use TIRF imaging, and single-line diode lasers as an excitation source. This type of imaging is only suitable for monolayer cultures: whether their sensor will be efficient when imaged in organized brain tissue using two-photon excitation is not clear. The claim on a methodological advance appears therefore premature.

We did not use TIRF imaging. -we never mention it (the acronym is only in the name of the objective). We used diode lasers for some experiments and LED illumination in others. Regular SnFR has been used a lot in 2P imaging, and our sensor arguably has better signal to noise, and very little reason to believe the photophysics should be different. We didn’t change that part of the sensor at all.

What we do is to record an arbitrary number of inputs to a single neuron (an autapse in our case). This is our methodological advance. This experiment was not possible before.

3. The other key statement is that iGluSnFR is prone to photobleaching (rundown) more than is SnFR-y2/8. First, fluorophore's photobleaching properties in 1P as opposed to 2P mode could be very different, which has not been addressed or explored. Second, this observation is surprising because several recent studies have documented a fairly stable iGluSnFR signal over multiple cycles of glutamate release imaged at 'quantal' resolution, both in 1P and 2P excitation regimes, both in cultures and in acute slices (e.g., Tagliatti et al., 2020 PNAS 117: 3819; Jensen et al. 2019 Nat Commun 10: 1414). The authors do not seem familiar with these studies. Dye photobleaching depends on multiple imaging parameters starting with laser power: this has not been investigated consistently in the present work.

Because we do not know exactly what is happening to SnFR, we did not talk about photobleaching. Whatever happens occurs only at a subset of sites (about half of them), suggesting it is not due to illumination per se. Overall, the photophysical performance of our sensor seems quite similar to SnFR. The fluorophore is the same and we would expect 2P performance to be similar.

We now include a supplementary figure (new Figure S5) where we analyse the rundown and compare to the electrophysiological recording. Please see our response to referee 1 for more details. The results are interesting, we are grateful for the push to examine more closely. There is rundown of the synaptic currents, but iGluSnFR substantially overestimates it. Our sensor seems if anything to underestimate the rundown in current. We write about these new analyses near line 320

We know the papers suggested. We cited similar manuscripts; those mentioned by the referee above are now also cited for the same purposes (on lines 48 and 88, respectively). These are very good papers, but in the context of our work, we felt we might have to criticise their limitations, and we wanted to avoid this and be as neutral as possible. We apologise for being timid. These papers select individual examples of optical signals that work well for them. In the Tagliatti paper, they select broadly spaced presynaptic terminals by hand to localize SnFR. The Jensen paper is done on single identified connections on Schaffer Collateral, using presynaptic localisation

In our case, looking at a large number of inputs to a single neuron, this means going through movies, checking and selecting responses from hundreds of active ROIs by hand. This approach is neither scalable, nor reproducible. Our point is that, rather than collecting a mixture of gold and junk and manually selecting the gold, why not just collect good ROIs (our case)? This is essential if you want to examine arbitrary geometries, not just presynaptic terminals (for example, inputs to neurons in the brain). This approach is scalable to the number of synapses on a neuron (100-1000), rather than the number of Shaffer Collateral terminals at a connection (a single terminal).

4. Stargazin overexpression has been used as a principal tool for the iGluSnFR targeting to synapses, but the authors report that this renders a proportion of synapses nonresponsive to glutamate (Figure 4). That the method interferes with the physiological integrity of synaptic circuits, or at least requires some additional experiment-specific manipulations to minimize it, does not speak in its favor.

Point taken, but most experiments require “experiment-specific manipulations” to minimize problems in the data collection. We measured a problem (that few would bother to check) and we showed how we largely ameliorated it (through later infection), and learned something new in the process. Agreed, not ideal for our sensor, but valuable information, we feel.

5. The signal-to-noise ratio of the SnFR-y2 signal does not appear improved compared with that of iGluSnFR (Figure 8A).

In some contexts the SnFR-g2 SNR is higher (now Figure 4), in some not (now Figure 9). But high signal to noise is not everything, if some of the signals turn out to be unrelated to what you are trying to measure (see our now-Figure 8). Our work shows that the signal to noise of the targeted sensor is quite similar to the original SnFR, but the targeted sensor is in other ways a better reporter.

6. The quantal-analysis histograms presented here (Figure 8C-D) appear noisier hence less reliable than similar or related analyses in the aforementioned publications that employed iGluSnFR.

Again, this might be explained by selection and the type of experiment. The “nicer” histograms come from presynaptic expression, perhaps using hundreds of stimuli. We used only 50 stimuli per condition, so the individual histograms are not so beautiful. But overall the data are very rich, because we have 40 or so ROIs per field of view, and collect across three different conditions. The overall lack of bleaching of the SnFR-g2 sensor (now shown in supplementary Figure S5) suggests one could collect more data to have more reliable histograms.

Reviewer #3 (Recommendations for the authors):

Hao and colleagues developed new variants of the glutamate sensor iGluSnFR, termed SnFR-γ2 and SnFR-γ8, that are fusion proteins with postsynaptic density (PSD) proteins Stargazin and γ-8. This chimeric protein thus localizes specifically at the PSD. These new variants are characterized, using heterologous expression systems and hippocampal neuron microisland cultures, allowing one to monitor autapses with simultaneous electrical and optical access. Overall, SnFR-γ2 outperforms traditional iGluSnFR in terms of signal localization; presumed single synapses are observed with limited "spillover" of signal to neighboring regions, and imaged transients are amenable to traditional noise analysis. Some concern regarding competition for PSD membrane is raised due to overexpression of these variants, but it appears that such overexpression artifacts can be avoided by ensuring that SnFR-γ2 is delivered after synapses are largely formed. This could be an important tool for the field, as it improves one's ability to resolve the activity of single glutamatergic synapses with sufficient signal to noise. Work here has shown feasibility in cultures systems. Future work will need to show similar performance levels in acute slice and in vivo, though based on past observations with iGluSnFR, this performance is likely within reach.

We thank the reviewer for their balanced critique.

The main question I have is one of extensibility: can this approach work in more intact systems, or even in vivo? I recognize that asking such a question involves an entirely new dataset, and am not proposing that the authors engage in such an effort after already doing an excellent job characterizing these GluSnFR variants. Rather, I'd hope that the authors would expand on their discussions, which is largely focused on questions of release dynamics observed with their sensors, to also include potential technical advantages or limitations of these variants in other preparations.

Thank you for this kind recognition. We added a line that better performance is needed in vivo, and that our sensors (also with reference to the new preprint from Janelia) are likely better for quantitative work in vitro (around line 599). We do not want to speculate further at this stage.

Comments below are aimed at improving interpretation of data reported herein:

1) A control for infection is needed for autapse data. Please make parallel recordings in cells infected with control viruses that lack any glutamate sensor to determine if these currents are in the normal range at these ages.

Thank you for this suggestion. We added these data to the figure (now Figure 6C) – a separate set of recordings of autaptic neurons infected with a GFP AAV. The currents are a bit bigger in this condition.

2) Data in Figure 5C and F should be analyzed quantitatively by calculating correlation coefficients for each pair of data. If there are differences in the relative separation of each region (I understand that these were chosen by hand) then a potential comparison could be made by plotting correlation coefficients vs. centroid distance of paired ROIs.

We now provide the correlation coefficients for the data we show, which give a very conclusive answer along the lines that the referee imagined. Thanks for this great suggestion.

3) Data in Figure 7C, 0.5 mM. There is an obvious outlier near 7 P5/P1 for the SnFR-gamma2. This is likely due to a very low P1 value for this one cell, indicative of excess failures. I'd be curious to know if any correlation holds if this one datapoint is held out of the dataset. If so, perhaps the authors could explain this observation in the main text.

Thank you for this suggestion. Briefly, we looked to see if removing this data point gives a better correlation, but the correlation is still poor so in the interest of transparency, we left it in. We now explain this in the text (near line 373).

4) Figure 7, data related to optical quantal analysis. Considering that these are autaptic recordings, with superb electrical access, one should be able to perform traditional electrophysiological quantal analysis and determine whether SnFR is allowing for identification of all synapses. This is a critical analysis that should be made on a cell-by-cell basis, paired with optical analyses; however, if this is not feasible at this time, some information could be gleaned from separate recordings, given that you observed fairly consistent numbers of sites optically (4 to 7). This would address your concern that the detection methods bias towards high Pr synapses. Though, if there are more synapses made a significant distance from the coverslip, then they would never be imaged under TIRF microscopy, obviating this request.

Unfortunately not – the autapses have way too many synapses (possibly hundreds) for quantal analysis with electrophysiology. The N we derive is per ROI. Although we did not use TIRF microscopy, the reviewer is still correct, we cannot pick up all synapses because of the 3-D nature of the cell, and lack of focal depth with high NA objective. However, those that we do detect with SnFR-g2 offer a very good measure of the behaviour (depression) of the entire group as ascertained from electrophysiology (whereas iGluSnFR ones do not) – this is shown in the correlation in the now Figure 8C.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

This manuscript is an attempt towards synapse specificity for glutamate probes. The current probe's utility is limited to specific cases, and it has what could be considered advantages and drawbacks, even in that specific case (CMOS imaging of cultured cells). It is commendable how carefully the authors characterized side effects of this sensor on synaptic function. The reviewers agreed that several issues require clarifications in the text.

We agree with these points and we are grateful for the recognition of the care we have taken.

1. Please explain that any glutamate rises or waves in the tissue, any significant glutamate spillover signaling, or even fluctuations in tissue optical conditions, could be falsely perceived as local synaptic signals by an PSD-constrained sensor rather than by an evenly distributed sensor. This is a key limitation of the current method.

This is a good point and we have clarified it in the revised text (around line 662).

2. The authors seem to insist that the use of iGluSnFR or any sensor labeling is generally disadvantageous. It is advantageous in most cases. In terms of this sensor's fluorescence properties, there is no evidence for improved performance in the manuscript.

We have clarified that our sensor does not have improved fluorescence properties (around line 658), and already noted that untargeted sensors certainly still can have excellent performance, particularly if confinement can be achieved (line 87). We direct the readers to our recent review on this topic in JNS Methods for a wider discussion of this theme (near line 667),

3. Please consider the specific points raised by the reviewers and address them in the text to the extent possible.

Our point by point outline of the changes to the text follows below.

Reviewer #1 (Recommendations for the authors):

The authors have provided their explanations and rebuttal regarding the previous comments, albeit without new experimental evidence.

We did actually provide some new experiments (Figure 3), showing the localisation of the sensor.

1. It appears that the authors did not fully understand the main objection. Or perhaps it was not explained clearly enough. To reiterate: SnFR-y2/8 expressed locally at the synapse may in fact sense and report glutamate that is released elsewhere in the vicinity, thus giving a false impression of its local synaptic release. In other words, SnFR-y2/8 may report the spillover signal as well as the actual synaptic signal.

It is true that, in 2PE mode, in organized brain tissue, iGluSnFR will report optical signal integrated, due to diffraction of light, within the PSF depth of ~800 nm, etc., and thus may include glutamate release events, if any, from nearest synaptic neighbors. However, the same will happen if such neighbors express SnFR-y2/8: their optical signal will still be integrated within the diffraction-limited PSF in 2PE imaging mode.

In the case when no glutamate spillover ever occurs in the preparation of interest, both iGluSnFR and SnFr-y2/8 will report true synaptic signals when such happen. In this case, however, glutamate rises or waves arising from other sources, such as astroglia, will still be reported by SnFr-y2/8 very locally, giving an impression of local synaptic events – whereas iGluSnFR will report the entire glutamate 'landscape'. Thus, SnFr-y2/8 does not seem to have any principal spatial-resolution advantage over iGluSnFR.

As suggested by the editor, we have added some clarifying text that our targeted sensor will still report glutamate from *all* sources. The examples given by the reviewer are excellent and we use them (around line 662). We moved the final sentence of the discussion forward to offer the potential solution: to have a multiplexed output (as suggested by reviewer 3 in the first round of reviews).

2. Authors' attention is drawn to yet another recent publication (Mendonca et al. 2022 Nat Commun 3497) showing an excellent stability, S/N ratio, etc. for highly localized glutamate release events detected with iGluSnFR. That this sensor is performing not as satisfactory in the present authors' hands cannot be a basis for their extrapolated claim.

We are aware of this work and are happy to cite it (in the introduction line 88 and around line 673). It’s a beautiful study measuring the outputs of a single neuron. It does not correspond directly with what we have done – we measured at confined, postsynaptic sites, in order to detect the inputs to a single neuron, a complementary approach. We congratulate the authors of the Mendonca paper on their work, it’s impressive. Some of the techniques they use certainly exceed the sophistication of our work.

As the referee probably knows, but for the sake of transparency, Mendonca and colleagues used a newer,brighter variant (SF-iGluSnFR) that was not available when we started our work. In the Mendonca paper, the sensor is not tethered, but importantly, the sensor is confined to the presynaptic neuron. This microanatomical restriction apparently helps with the signal localisation (the same effect was seen at the Schaffer terminals). We already noted that in some other work, physical confinement of the sensor was enough (around line 87). Presynaptic tethering did improve performance in other work that we cited (Kim et al., JNS 2020) although the Mendonca work suggests that this is not really necessary.

We do not understand these discrepancies, and we agree, our work is not definitive in this question. More work is needed. We have added a note in the text to this effect (line 673).

3. The authors concede that the S/N ratio of their probe readout is probably no better than that of iGluSnFR.

This point is now quite clear in the text, we hope (explicitly stated at line 658).

4. The authors have made no attempt to check their probe performance in 2PE mode and/ or in organized brain tissue.

We are sorry, we do not have the resources for that (yet). But we are grateful for the reminder about how important this is. We hope to address this point adequately with future work.

5. The amplitude histograms shown in Figure 9C-D do not match satisfactorily the best-fit quantal analysis curves. While the authors acknowledge the difficulty, the reasons for displaying unsatisfactory quantal analysis data are not clear.

We take the point. As reviewer 3 has commented, the separation of quantal peaks is really what determines the fit – to get a good agreement with the small amplitude components we might need to sample more data. We feel that the more important point is that we could semi-automatically obtain similar information from 40+ such sites in the same neuron at once. There remain only a few examples of such an analysis.

We added a note to the text (around line 505) to mention that the incomplete agreement between the fits and the histogram might be ameliorated by collecting more responses.

6. The entire concept would have made a much greater impact if the authors aimed to express the sensor at a specific, functionally or genetically distinct, sub-population of synapses.

We agree, this would be fantastic. We would like to point out, that the prevailing view (up to this preprint) was that targeting iGluSnFR was a very bad idea. Now we showed, and the Janelia preprint also shows, that targeting can be neutral, or even advantageous in terms of the signal collected. The greater impact that the referee seeks from molecular/cellular targeting? We seek it too and we hope now to aggressively pursue targeting in various contexts (as the reviewer describes). Maybe other investigators will too. We hope so.

Reviewer #2 (Recommendations for the authors):

The authors have addressed concerns raised by all three reviewers to the best of their ability. Their commentary that untagged SnFR is likely detecting large hotspots of both synaptic and spillover is taken well. Though the concern regarding how well the tagged variant works remains in question. If one can only resolve a few synapses (less than 10 per cell were analyzed) of what the authors suggest are ~100 potential autapses, are these few ROIs representative of the whole? This may be a good point of discussion.

Thank you. We added a line of discussion (around line 530) about the disparity between the amplitude of the synaptic current and the number of ROIs that display activity. We know that this question has been bothering a few groups for years. Perhaps with further work we can resolve it.

Given the way in which eLife is modifying its review process and overall publishing criteria, this work should be accepted. It represents a new tool that is well characterized. Advantages and flaws are discussed well.

Thank you.

Reviewer #3 (Recommendations for the authors):

In their revised version, the authors address questions about synaptic localization of their GEGI by demonstrating good overlap with the PSD95 signal (new Figure 3). The explanation of Figure 8 has been much improved, quantal parameters are now well defined. The correlation between electrical and optically measured depression during a train (Figure 8c) is indeed much better for their targeted indicator compared to iGluSnFR, which is a strong argument that synaptic (and not extrasynaptic) glutamate is reliably measured by their sensor. In this context, it is an advantage of the autaptic model that all synapses on a given neuron have the identical history of activity and therefore express similar (but cell-specific) short-term plasticity. I do not like the analysis of rundown in Figure 7F: Splitting a continuous distribution into 2 groups, using an arbitrary threshold, then counting the number of cases (ROIs) on either side of the threshold, is not good statistical practice. The analysis in Figure S5 I like much better, it shows no significant difference between the two indicators. As the authors cannot offer a convincing mechanistic explanation why a difference in rundown would be expected, I suggest downplaying this point (getting rid of Figure 7E-G, sticking with the message of S5).

We are grateful for these supportive comments and for the thoughtful analysis of the data. We agree that the arbitrary split is not ideal, but we did it in order to orient the reader with a concrete example of a difference between broadly usable and unusable data.

Aiming to follow the spirit of the referee’s suggestion, we demoted Figure 7E-G to supplementary material (Figure 7—figure supplement 1) and added a caveat that this approach is illustrative and not statistically rigorous (around line 315). We also provide a stronger statement that we have no mechanistic explanation for the rundown of individual ROIs (around line 571).

Apart from this quibble, the science is sound and well presented. The separation of quantal histogram peaks is impressive and certainly aided by localizing the indicator to the places of highest glutamate concentration. Extrasynaptic indicator molecules, exposed to a near-continuous range of glutamate concentrations, would be expected to widen the peaks considerably (as shown in the iGluSnFR example in Figure 9c). For future tool development and targeting efforts, the technical information and precise measurements are very useful, even as a somewhat cautionary tale with regard to potentially severe side effects of tool expression.

Thank you for these supportive comments.

Author details

1. Yuchen Hao

   1. Institute of Biology, Cellular Biophysics, Humboldt-Universität zu Berlin, Berlin, Germany
   2. Leibniz-Forschungsinstitut für Molekulare Pharmakologie, Berlin, Germany

   Contribution

   Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0042-6576

2. Estelle Toulmé

   Institute for Neurophysiology, Charité - Universitätsmedizin Berlin, Berlin, Germany

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8734-3484

3. Benjamin König

   1. Institute of Biology, Cellular Biophysics, Humboldt-Universität zu Berlin, Berlin, Germany
   2. Leibniz-Forschungsinstitut für Molekulare Pharmakologie, Berlin, Germany

   Contribution

   Software, Formal analysis

   Competing interests

   No competing interests declared

4. Christian Rosenmund

   1. Institute for Neurophysiology, Charité - Universitätsmedizin Berlin, Berlin, Germany
   2. NeuroCure Cluster of Excellence, Berlin, Germany

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Project administration

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-3905-2444

5. Andrew JR Plested

   1. Institute of Biology, Cellular Biophysics, Humboldt-Universität zu Berlin, Berlin, Germany
   2. Leibniz-Forschungsinstitut für Molekulare Pharmakologie, Berlin, Germany
   3. NeuroCure Cluster of Excellence, Berlin, Germany

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   andrew.plested@hu-berlin.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6062-0832

• 2,333

  Page views

• 350

  Downloads

• 1

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.