Critical Analysis of Particle Detection Artifacts in Synaptosome Flow Cytometry

FSC and FM4-64 triggering of crude synaptosome fraction P2 and comparison to bead standards

Conventional flow cytometers generally use FSC measurements for detection of cells, but the sensitivity of each instrument in FSC-triggered detection of submicron particles varies widely depending on the refractive index of the particles, the angle of collection for scattered light, detector sensitivity, and other factors (van der Pol et al., 2013). We used PS and silica beads to determine the relative sensitivity of two cytometers: a BD LSR Fortessa analyzer and a BD Influx sorter. We optimized FSC-threshold parameters on each instrument to maintain a mixture of PS and silica bead standards ranging 585–1300 nm in the dynamic range of detection, while simultaneously avoiding electronic and buffer noise.

As shown in Figure 1A, the lowest FSC signal resolvable above the background noise on the Influx was the 585-nm silica bead. Cross-referencing the FSC signal with the green fluorescence signal allowed us to clearly identify PS beads apart from the non-fluorescent silica beads (Fig. 1A, right). As expected due to their higher refractive index, the FSC signals for PS beads were substantially higher than for similarly sized silica beads. We also identified events that appeared to be simultaneous detection of two 500-nm PS beads, which we confirmed by comparing their FSC and green fluorescence to the “singlet” gate as previously described (Extended Data Fig. 1-2; Stoner et al., 2016). Although the FSC and fluorescent signals appeared additive in the 500-nm PS doublet population (1.7-fold higher FSC, 1.8-fold higher green fluorescence), the FSC signal for these events still fell below the 880-nm silica beads. In contrast to the Influx, the Fortessa displayed poor FSC resolution when running the silica/PS bead mixture (Fig. 1B), which required separately running the individual PS beads [500- and 800-nm PS labeled with (B) in Influx plots]. The lowest FSC signal resolvable above the noise background on the Fortessa included 880-nm silica as well as 500-nm PS singlet and “doublet” bead populations, which could only be separated by SSC or green fluorescence intensity (Fig. 1B). A complete summary of FSC and SSC measurements for gated bead populations can be found in Extended Data Figure 1-2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Determining detection sensitivity using bead standards and FSC versus FM triggering of P2 fraction. A, Mixture of non-fluorescent silica/fluorescent PS beads (Apogee; A) and fluorescent PS beads (Bangs Labs; B) detected in FSC-trigger mode on the Influx. Forward and side scatter (left) FSC (right) and green fluorescence detector (FITC) allows fluorescent PS beads and bead doublets to be clearly distinguished from non-fluorescent silica beads. B, Mixture of non-fluorescent silica/fluorescent PS beads (Apogee; A) detected in FSC-trigger mode on the Fortessa. Note that the Bangs Labs 500- and 800-nm fluorescent PS beads could not be run simultaneously with the Apogee mixture on the Fortessa due to poor FSC resolution. C, FM4-64-stained P2 sample on the Influx. Left, FSC-trigger mode. Right, FM-trigger mode. FSC noise gate was set by running clean PBS. D, FM4-64 stained P2 sample on the Fortessa. Left, FSC-trigger mode. Right, FM-trigger mode. FSC noise gate was set by running clean PBS. See Extended Data Figure 1-2 for median FSC and SSC values from gated bead populations. See also Extended Data Figure 1-1.

Having established FSC-triggering limits and relative bead positions on the instruments, we next sought to compare FSC and fluorescence triggering of P2 crude synaptosome (hereafter P2) samples. Despite contamination by a heterogeneous mixture of free mitochondria, myelin, and membranous debris, P2 samples are enriched with synaptosomes and have been widely employed in flow cytometry studies (Gylys et al., 2000; Fein et al., 2008; Postupna et al., 2014; Prieto et al., 2017). As previously described by Biesemann et al. (2014), FM4-64 is a styryl dye with favorable properties for fluorescence-triggered detection of synaptosomes. FM4-64 is minimally fluorescent in aqueous media but becomes intensely fluorescent on partitioning into membranes (Vida and Emr, 1995), and its far-red emission is efficiently excited by a 488nm laser, which is required for triggering on our cytometers. As shown in Figure 1C,D, FM triggering enabled highly sensitive detection of P2 samples compared to FSC triggering, with more than half of particles in the P2 sample below the FSC noise threshold on both cytometers. These results are consistent with the work of Biesemann (2010; his Results 3.5, Fig. 15, pp 87–90), who found ∼70% of particles in sucrose gradient synaptosome preparations were undetectable by FSC triggering on the BD FACS Aria.

Since previous studies have analyzed putative synaptosomal particles with FSC signals as high as 4.5-µm PS beads (Gylys et al., 2000), we sought to assess the size of biological particles in such an FSC range. In the absence of a refractive index mismatch between PS beads and biological particles, one would expect neuronal nuclei, which sediment in the P1 fraction, to exhibit significantly higher FSC than 2-µm PS beads due to their size. Although we had to lower the FSC voltage to place 0.5-, 0.8-, and 2-µm PS beads within the same dynamic range, 2.0-µm PS beads were clearly resolved from 0.5- to 0.8-µm PS beads by FSC on both the Influx and Fortessa (Extended Data Fig. 1-1A). We identified neuronal nuclei based on Hoechst 33342 intensity and NeuN immunofluorescence (Extended Data Fig. 1-1B) and found that they were adjacent to and partially overlapping with the FSC range defined by 2-µm PS beads on the Influx (Extended Data Fig. 1-1C). These results confirm that FSC-based sizing relative to PS beads underestimates the size of biological particles with lower refractive indices, and suggest that a majority of nuclei would be included in FSC gates containing 4.5-µm PS beads.

Serial dilutions define a range of linear detection for FM4-64 triggering

It has previously been demonstrated that high concentrations of particles below the trigger threshold can be detected when they simultaneously occupy the focal point of illumination (van der Pol et al., 2012; Nolan and Stoner, 2013). This phenomenon of coincident detection (also known as “coincidence” or swarm detection) is thought to be particularly problematic for FSC-triggered detection of submicron particles (Nolan and Stoner, 2013). to define a linear range of particle detection and further compare FSC versus FM triggering, we conducted dilution series on both Influx and Fortessa. Compared to FM triggering, FSC-triggered P2 samples did not maintain a constant FSC versus FM4-64 profile on either the Influx or Fortessa (Fig. 2A,B; Extended Data Fig. 2-1). Specifically, the abundance of events near the FSC threshold with low FM4-64 fluorescence decreased with dilution, and the appearance of “noise” events (although <10 events/s) became more apparent with dilution on the Fortessa. On the Influx, quantification of the event rate across the dilution series revealed a sub-linear profile for both FSC and FM triggering, which was fit with high accuracy using a quadratic model (R² = 0.99 for both series; Fig. 2C, left panel). FM triggering was similar on the Fortessa, but FSC triggering was better fit with a linear model (R² = 0.93 for FSC and R² = 0.99 for FM series; Fig. 2D, left panel). These results confirm that event rate saturates at high particle concentrations (Kormelink et al., 2016) and that FSC triggering underestimates the true event rate across the entire dilution series.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Dilution series define a range of linear particle detection in P2 samples. A, Representative density plots of a P2 dilution series detected on the Influx in FSC-trigger mode. FM4-64 low gate was set arbitrarily to quantify disappearance of events in this region with dilution. B, Representative density plots of same P2 dilution series as in A, but detected on the Influx in FM-trigger mode. C, Measurements from full P2 dilution series on the Influx. Left, Event rate data fit with a quadratic model. Middle, Median FSC of all detected events. Right, Median FM4-64 fluorescence of all detected events. D, Measurements from full P2 dilution series on the Fortessa. Left, FM-trigger series fit with a quadratic model and FSC-trigger series fit with a linear model. Middle, Median FSC of all detected events. Right, Median FM4-64 fluorescence of all detected events. See also Extended Data Figure 2-1.

Given that much of the P2 remains undetected by FSC triggering, we also wondered whether particle concentration might alter the FSC and FM fluorescence measurements. On the Influx, the median FSC remained relatively constant for both FSC and FM triggering (Fig. 2C, middle panel), while the median FM4-64 fluorescence varied dramatically for FSC triggering but not for FM triggering (Fig. 2C, right panel). On the Fortessa, both median FSC and median FM4-64 fluorescence varied dramatically across the dilution series for FSC triggering but not FM triggering (Fig. 2D, middle and right panels). The increase in median FM4-64 fluorescence with dilution seen in FSC-triggered samples seems related to the “FM4-64 low” events described above. Given the heterogeneous population of particles contained in P2 samples, we cannot make strong conclusions about the mechanisms underlying these effects. Nonetheless, it is clear that FM triggering is superior to FSC triggering in terms of the stability and accuracy of event rate, FM4-64 fluorescence, and FSC of P2 particles across the dilution series on both instruments. We conducted all further studies using only FM triggering and maintaining reasonably dilute samples so as to stay in the most linear range of the event rate dilution curves (fewer than 10,000 events/s on the Influx and fewer than ∼3000 events/s on the Fortessa; Fig. 2C,D, left panels). Since sample flow rates and event rates may vary across cytometers, we also measured the absolute particle concentration across a typical dilution series on the Influx using volumetric count beads (Extended Data Fig. 2-1C). We found that the most linear range of the event rate curve was for samples below ∼1000 particles/µL. A previous study on the Influx (Kormelink et al., 2016) showed that the trigger pulse baseline becomes continuously elevated at high event rates where coincidence is prominent, so we used a digital oscilloscope on the Influx to confirm that the trigger pulse remained at baseline under these conditions (data not shown).

Development of a multicolor fluorescence assay for false colocalization

Although FM triggering and dilution should help to reduce coincident particle detection, another reported source of “false double-positive” events in submicron flow cytometry is the physical aggregation of multiple particles (Erdbrügger et al., 2014). Multiple microscopy studies have noted the presence of aggregated synaptosomes (Gray and Whittaker, 1962; Choi et al., 2009; Daniel et al., 2012), but conventional flow cytometry does not provide a means to verify whether events truly represent single particles. We sought to design a fluorescence-based flow cytometry assay to estimate false double-positive events in our P2 samples (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Experimental assay for measuring coincidence and aggregation. A, Schematic representation of the P2 sample labeling workflow. A single sample is split into equal portions for single-color labeling, followed by two washes to remove residual dye. The samples are then mixed and washed a third time before flow cytometry. B, Schematic representation of particle detection and artifacts in microparticle flow cytometry. Particles flow in the sample fluid, which is surrounded by sheath fluid. Note that the width of the sample stream is enlarged for illustrative purposes; we run our cytometers at the minimum sample pressure (and thus sample stream width) to ensure optimal hydrodynamic focusing. When the trigger parameter detector crosses a user-defined threshold, the time duration of this pulse is then integrated on all detectors to generate an event. Single events that are detection artifacts occur when multiple particles simultaneously occupy the focal point of laser illumination (coincidence) or are physically bound (aggregates). C, Representative density plots of a P2 sample from the double-calcein assay on the Influx. Left, Calcein AM-labeled P2 sample shows no cross-emission in the calcein red detector. Mid-left, Calcein red-labeled P2 sample shows no cross-emission in the calcein AM detector. Mid-right, Mixture of single-labeled P2 samples is gated for calcein AM+ and calcein red+ particles, which are then cross-checked for the other fluorophore (diagonal arrows). Right, Calcein AM+ or calcein red+ gated particles are analyzed to quantify double calcein+ events. See also Extended Data Figure 3-1.

The assay workflow is similar to previously described “cell barcoding” flow cytometry methods (Krutzik and Nolan, 2006; Krutzik et al., 2011). A P2 sample is first split and single-labeled with spectrally separated, non-transferable fluorescent dyes. After washing, single-labeled samples are mixed before flow cytometry (Fig. 3A). Single particle events should therefore bear only one fluorescent label, while coincidences and aggregates derived from separate single-labeled samples will be positive for multiple fluorescent labels (Fig. 3B). Although the samples are washed twice to remove any residual dye, it is critical to use dyes that cannot be transferred between particles. Synaptosomes are well-labeled with calcein AM dyes (Gylys et al., 2000; Daniel et al., 2012; Prieto et al., 2017), lipophilic dyes that rapidly cross cell membranes and become trapped following hydrolysis of the AM ester. In addition to calcein AM and calcein red, we also used violet amine-reactive dye (covalent reaction with proteins) and MitoTracker Deep Red FM (thiol-conjugation and retention in mitochondria) to expand the number of dyes in some experiments. As shown in Extended Data Figure 3-1A, samples single-labeled with any of the four dyes do not exhibit any spectral overlap.

A representative set of density plots for a P2 sample single-labeled with calcein AM or calcein red, as well as a mixture of the two, on the Influx is shown in Figure 3C. Note that because even the background fluorescence is linearly related to FSC on the log-log plots, quadrant-based gating of double-positive events underestimates the number of positive events. To avoid this problem, we used polygon gating of each fluorophore individually versus FSC (Fig. 3C). Boolean logic is used to determine the number of double-positive events, which can be visualized by displaying only the positive-gated population of one fluorophore on a plot of FSC versus the other fluorophore (Fig. 3C, right).

Having verified the presence of calcein double-positive events in P2 mixtures (Fig. 3C), we sought experimental confirmation that the dyes did not transfer between particles. Because only the cleaved, polar dye molecules are retained, we reasoned that even if some membrane disruption and dye leak occurred during mixing and centrifugation, the polar molecules would be unable to label surrounding intact particles bearing the other calcein dye. We confirmed this by sonicating a P2 sample truly double-labeled with both calcein dyes and adding an unlabeled P2 sample to the sonicated solution using the exact same experimental workflow. As shown in Extended Data Figure 3-1B, no calcein AM-positive or calcein red-positive events were detected. This demonstrates that even after complete membrane disruption to liberate the entire dye content of one sample, we do not observe measurable transfer to an unlabeled sample after washing and mixing.

Aggregation, but not coincidence, causes false double-positive events in P2 samples

With our multicolor fluorescence assay in hand, we turned toward distinguishing between coincidence and aggregation. Similar to previous studies (Libregts et al., 2018), we used mixtures of green- and red-fluorescent 500-nm PS beads to model coincidence. We repeated the dilution series experiments with mixtures of beads or mixtures of single calcein-labeled P2 samples on both Influx (Fig. 4A,B) and Fortessa (Extended Data Fig. 4-1A,B). For bead mixtures, we observed a clear population of double-positive events (red- and green-bead coincidences) on both instruments. At high concentrations, the Fortessa performed particularly poorly with regard to coincidence, with a wide smear of red-fluorescent bead multiplets and double-positive events (Extended Data Fig. 4-1A). On both instruments, we observed a linear reduction in the number of double-positive bead events with dilution, with virtually none detected at event rates <500 events/s (Fig. 4C,D). In contrast, the percentage of double-positive events from P2 samples remained relatively constant across the entire dilution series. P2 double-positive events persisted even at event rates <100 events/s, which strongly suggests that coincidence is not the source of these false double-positive events (Fig. 4C,D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Dilution series reduces false double-positive bead events but not double-calcein positive P2 events. A, Representative density plots of a dilution series for a mixture of fluorescent 0.5-µm PS beads detected using FSC triggering on the Influx. Individual PS beads and double-positive events are distinguished by their green (FITC detector) or red (PE detector) fluorescence. B, Representative density plots of a dilution series for a mixture of single calcein-labeled P2 samples detected using FM triggering on the Influx. Only events gated positive for calcein red are displayed. C, Measurements from full dilution series of bead mixture or single calcein-labeled P2 mixtures (n = 3) on the Influx. P2 data are displayed as mean ± SEM. Left, Event rate. Right, Double-positive events expressed as a percentage of all fluorescent events. D, Measurements from full dilution series of bead or single calcein-labeled P2 mixtures on the Fortessa. Left, Event rate. Right, Double-positive events expressed as a percentage of all fluorescent events. See also Extended Data Figure 4-1.

We speculated that the double-positive events might come from the physical association of two separate fluorescent particles in the P2 samples. To further probe the nature of these events, we sorted double-positive events from bead or P2 mixtures on the Influx and reanalyzed the sorted material (Fig. 5A,B). Sorting of bead double-positive events failed to increase the frequency of these events in the sorted sample, with reanalysis showing a ∼4-fold decrease (Fig. 5C). Consistent with the dilution experiments, this result suggests that the majority of green- and red-fluorescent beads are not physically associated with each other during detection of double-positive events. In contrast, sorting of P2 double-positive events increased the frequency of these events ∼3-fold in the sorted sample reanalysis (Fig. 5C). We note that a 3-fold increase in fluorescent particles during sort reanalysis is similar to that achieved by (Biesemann et al., 2014) for fluorescent VGLUT1+ synaptosomes, suggesting that P2 double-positive events behave as stable, “single particles” in this experiment. It is also possible that the high pressures encountered during sorting physically disrupt some aggregates, as suggested by Biesemann (2010; his Results 3.2, Fig. 12, pp 79–81), which would further decrease the frequency of double-positive events in reanalysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Effects of fluorescence-activated sorting, centrifugation resuspension, and time lapse on double-positive P2 events. These experiments were conducted using FM triggering on the Influx. A, Representative density plots of pre-sorted samples. Left, Mixture of fluorescent 0.5-µm PS beads on the Influx. The “bead double+” gate was used for sorting. Mid-left and right, Mixture of single calcein-labeled P2 samples. Right, Calcein AM+ gated events from single calcein-labeled P2 samples were gated for calcein red+. This “double cal+” gate was used for sorting. B, Representative density plots of sorted samples re-analyzed on the Influx. Left, Sample sorted for bead double+ gate above. Mid-left and right, Sample sorted for double cal+ gate above. Right, Calcein AM+ gated events from double cal+ sorted sample. C, Summary of pre-sort versus reanalysis of sorted double-positive bead and P2 (n = 3) samples. P2 data are plotted as mean ± SEM. D, Comparison of time lapse data for centrifugation-resuspension (“spin”) versus mixing without centrifugation (“mix”) of single-labeled P2 samples. Left, Time lapse of double-positive events plotted expressed as a percentage of all fluorescent events. Samples include the following mixtures of single-labeled P2 samples: violet amine-reactive dye/MitoTracker Deep Red FM (n = 2), calcein AM/calcein red (n = 3), violet amine-reactive dye/calcein AM (n = 1), violet amine-reactive dye/calcein AM/calcein red (n = 2), all four dyes (n = 2). Right, Comparison of averages for the individual trajectories shown on the left, plotted as mean ± SEM. A Student’s t test was performed comparing mix versus spin at each time point; *p < 0.05, **p < 0.01. See also Extended Data Figure 5-1.

The divergent behavior of bead and P2 samples in dilution and sorting experiments strongly suggests that the majority of P2 double-positive events are aggregates. Since we routinely vortex samples at high speed for 5 s before flow cytometry, this suggests that the double-positive aggregates are relatively stable. We hypothesized that these aggregates form during centrifugation and are not fully disrupted during resuspension and filtration. To test this hypothesis, we conducted the multicolor dye-labeling assay with and without centrifuging the sample mixture and monitored the samples over time on the Influx. As shown in Figure 5D, centrifugation resuspension significantly increased the frequency of double-positive events compared to mixing without centrifugation. Intriguingly, the frequency of double-positive events increased over time for nearly every sample (Fig. 5D, left panel).

Taken together with the dilution and sorting experiments, these results suggest that particle aggregation in P2 samples is an ongoing process that is accelerated by centrifugation. We emphasize that our assay actually underestimates the frequency of true aggregates; green-green or red-red aggregates are not detected as double-positive events, but such aggregates are surely formed during centrifugation and washing of single-labeled samples. The mixing of single-labeled samples without centrifugation shown in Figure 5D, mix samples, actually underestimates the true proportion of aggregates, and we therefore conducted all other experiments in this study with centrifugation of the mixtures.

Aggregation of cells can also affect conventional flow cytometry experiments, and we wondered whether commonly employed “singlet gating” procedures could discriminate aggregates in our P2 samples. We plotted the trigger pulse width (in this case, FM4-64) versus FSC on the Influx, or FSC-H versus FSC-A on the Fortessa, for events gated as single or double calcein-positive (Extended Data Fig. 5-1). Cell doublets generally deviate from linearity on such plots, forming a “cloud” that can be gated apart from the linear singlet profile (Wersto et al., 2001). We found no such deviation from linearity in single-positive gated events, which represent the majority of the sample. Although they bear only one fluorophore, single-positive events are expected to contain at least as many or even more aggregates than double-positive gated events (i.e., green-green, red-red, red-nonfluorescent, green-nonfluorescent, etc.). The lack of deviation from linearity for any events in these pulse profiles suggests that a “doublet profile” does not exist for our P2 samples. Indeed, a recent flow cytometry study using this gating procedure in analysis of FSC-triggered P2 samples found that all their events obeyed this “linear singlet profile” (Prieto et al., 2017). The lack of a distinguishing doublet profile becomes especially apparent when looking only at double-positive events, which are predominantly aggregates and display a similar linear pulse profile to single-positive events (Extended Data Fig. 5-1). Thus, we found no evidence that pulse profile is an effective means of removing false double-positive aggregates from downstream analysis of submicron particles.

Double-positive event frequency increases with FSC and is reduced in nonionic buffer

Although bead standards do not provide accurate size estimation of biological particles, we wondered whether the relative size information afforded by FSC intensity would correlate with the presumably larger size of aggregates. We analyzed five discrete regions ranging from low to high FSC on the Influx, based on beads and the FSC noise threshold (Fig. 6A,B). A representative overlay of the FSC gates onto gated double-positive P2 events is shown in Figure 6C. Strikingly, the top two FSC gates contain ∼50% of all double-positive events (percentages in red) but only ∼19% of all calcein-positive events (percentages in black), while the bottom two FSC gates contain only ∼25% of all double-positive events among ∼65% of all calcein-positive events. Accordingly, the percentage of fluorescent events within each FSC gate that were double-positive steadily increased with FSC (Fig. 6C, percentages in blue). For all fluorescent events within each FSC gate, we quantified the percentage of single-, double-, and triple-positive events on the Influx (Fig. 6D–F). Although the overall frequency of single-positives was ∼90%, this steadily decreased from ∼97% in the FSC Noise region to ∼70% in the region above or equal to 1300-nm silica beads (Fig. 6D). The overall frequency of double-positives was ∼10%, but steadily increased from ∼2% in the FSC Noise region to ∼27% in the region above or equal to 1300-nm silica beads (Fig. 6E). Similarly, the overall frequency of triple-positives was ∼2%, but increased from ∼0% in the FSC noise range to ∼12% in the region above or equal to 1300-nm silica beads (Fig. 6F). These results suggest that although aggregates are present in all FSC regions, their presence is strongly correlated with increasing FSC. Again, we emphasize that our fluorescence assay underestimates aggregation, even in samples where the measured double-positive frequency exceeds 40% in the highest FSC regions (Fig. 6C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Distribution of double-positive P2 events across FSC ranges and in different sample buffers. These experiments were conducted using FM triggering on the Influx. A, Representative density plot of non-fluorescent silica and fluorescent PS beads (Apogee) showing the five FSC-gated regions based on noise threshold and bead positions. B, Same sample and FSC gates as in A but plotted to display green fluorescence of PS beads (FITC detector). C, Representative density plot showing the five FSC-gated regions on a mixture of single calcein-labeled P2 samples. Only gated double-positive events are displayed. The percentage of all fluorescent (i.e., calcein-labeled) events that falls into each FSC gate is displayed in black (all five FSC gates sum to 100%). The percentage of all calcein double-positive events that falls into each FSC gate is displayed in red (all five FSC gates sum to 100%). The number of double-positive events within each FSC gate, expressed as a percentage of all fluorescent events in that FSC gate, is displayed in blue. D–F, Box and whiskers plots of single-positive, double-positive, or triple-positive events, expressed as a percentage of all fluorescent events within the respective FSC gate. Central bar represents the median. Lower and upper edges correspond to 25th and 75th percentiles. Lower and upper whiskers extend to the smallest or largest value no greater than 1.5 times the interquartile range away from the corresponding edge. Data for single-positive and double-positive events are derived from two-color mixtures of P2 samples single-labeled with calcein AM/calcein red (n = 4), violet amine-reactive dye/MitoTracker Deep Red FM (n = 1) or calcein AM (n = 1) or calcein red (n = 1), as well as three-color mixtures single-labeled with violet amine-reactive/calcein AM/calcein red (n = 3). Triple-positive event data are derived only from the three-color mixtures. A Student’s t test was performed comparing each FSC gate to that directly above or below it; *p < 0.05, **p < 0.01. G, Diagram of buffer scheme during sample preparation. PBS and SET samples are labeled, washed, and run on the flow cytometer in their respective buffers. PBS-SET samples undergo calcein labeling and the first two washes in PBS before mixing, a third wash, and flow cytometry in SET buffer. H, Box and whiskers plot of double-positive events, expressed as a percentage of all fluorescent events within the respective FSC gate. Data are derived from two-color mixtures of P2 samples single-labeled with calcein AM/calcein red using the indicated buffer scheme: PBS (n = 10), PBS-SET (n = 8), or SET (n = 10). Two-way ANOVA was performed, revealing significant main effects of FSC gate (Df = 5, F = 20.06, p = 2.7e-15) and buffer (Df = 2, F = 20.14, p = 1.8e-8), as well as a significant interaction (Df = 10, F = 2.78, p = 0.004). See also Extended Data Figure 6-1.

Several groups familiar with microscopic analysis of synaptosomal preparations have noted that ionic buffers (e.g., PBS) cause synaptosomes to aggregate, while nonionic media (e.g., sucrose buffer) favors an even distribution of single particles (Choi et al., 2009; Daniel et al., 2012). We repeated our double calcein assay on the Influx using the SET buffer recommended by Daniel et al. (2012). Because many assays are not possible in nonionic media, we tested labeling in PBS and then switching to SET for flow cytometry (designated PBS-SET) in addition to conducting the entire procedure in SET (Fig. 6G). Compared to PBS, both PBS-SET and SET samples displayed similar FSC versus FM4-64 profiles (Extended Data Fig. 6-1A), indicating that SET did not impair FM-triggered particle detection or FSC measurement. FSC and green-fluorescence measurements of PS beads in PBS and SET were also indistinguishable (Extended Data Fig. 6-1B). However, we noticed that the total number of calcein-positive events was reduced to ∼20% in when calcein labeling was conducted in SET, compared to ∼40% for PBS or PBS-SET (Extended Data Fig. 6-1C). We suspect that the nonionic media and divalent cation sequestration by EDTA somewhat reduce the esterase activity of synaptosomes; nonetheless, the labeling efficiency was sufficient to analyze a large number of labeled particles present in these mixtures. As shown in Figure 6H, the percentage of fluorescent events within each FSC gate that were double-positive was reduced in SET compared to PBS. A stepwise pattern was observed, wherein the double-positive frequency was highest for PBS, lower for PBS-SET, and lowest for SET. Two-way ANOVA revealed highly significant main effects of both FSC gate (Df = 5, F = 20.06, p = 2.7e-15) and buffer (Df = 2, F = 20.14, p = 1.8e-8) as well as a significant interaction (Df = 10, F = 2.78, p = 0.004). The significant interaction makes sense given that buffer has little effect in the lower FSC regions, while the reduction of double-positive events is greater in the higher FSC regions (Fig. 6H). These results provide further evidence that double-positive events are aggregates, and that such aggregates are less abundant in nonionic SET buffer.

False colocalization of antigens in immunostained P2 samples

Given reports claiming high purity of synaptosomes in FSC ranges between 500- to 1500-nm PS beads (Gylys et al., 2004; Postupna et al., 2014; Prieto et al., 2017), we suspected that many of the aggregates in our P2 samples contain synaptosomes. Although the violet amine reactive and calcein dyes efficiently label many particles in P2 samples, they do not provide information about the identity of these particles. We conducted three-color immunostaining of P2 samples on the Influx, targeting presynaptic markers expressed by excitatory, inhibitory, and monoaminergic neurons, respectively: VGLUT1, VGAT, and vesicular monoamine transporter 2 (VMAT2). Although we did not attempt immunostaining in SET buffer, we did mimic the PBS-SET condition described above by resuspending stained and washed samples in SET for flow cytometry. As shown in Figure 7A, immunopositive events for all three transporters were observed across the entire FSC range and reflect their abundance in the brain (VGLUT1 > VGAT > VMAT2). Looking at all immunopositive events across FSC gates (Extended Data Fig. 7-1A), VGAT+ and VMAT2+ events displayed an asymmetric distribution centered below 500-nm PS beads, while VGLUT1+ events were distributed fairly uniformly just below 880-nm silica beads. Although the distribution of all immunopositive events across FSC gates was not affected by buffer (Extended Data Fig. 7-1B), the percentage of immunopositive events within each FSC gate was slightly reduced in PBS-SET samples (Fig. 7B). Two-way ANOVA revealed a significant main effect of buffer for VGAT (Df = 1, F = 4.70, p = 0.034) and VMAT2 (Df = 1, F = 5.21, p = 0.026) but not VGLUT1 (Df = 1, F = 0.36, p = 0.55). A highly significant main effect of FSC gate was observed for all three markers (Df = 5, F = 21.51, p = 2.9e-12 for VGAT; Df = 5, F = 68.9, p < 2e-16 for VGLUT1; Df = 5, F = 12.54, p = 2.4e-8 for VMAT2), where the percentage of immunopositive events within each FSC gate increased with FSC (Fig. 7B). All buffer × FSC gate interactions were non-significant. Consistent with recent studies (Prieto et al., 2017), the percentage of VGLUT1+ events in the highest FSC gate approaches ∼60%.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Fractional abundance of immunostained P2 events in different FSC ranges and in different sample buffers. These experiments were conducted using FM triggering on the Influx. A, Representative density plots of P2 samples immunostained for VGAT, VGLUT1, and VMAT2 and run in PBS. Top, Secondary-only controls in which primary antibodies were omitted. Bottom, Immunostained with indicated primary antibodies. B, Box and whiskers plot of immunostained events as indicated, expressed as a percentage of all events within the respective FSC gate. Central bar represents the median. Lower and upper edges correspond to 25th and 75th percentiles. Lower and upper whiskers extend to the smallest or largest value no greater than 1.5 times the interquartile range away from the corresponding edge. Data are derived from P2 samples triple stained and washed in PBS, and subsequently run in either PBS (n = 6) or SET (n = 6) as sample buffer. Two-way ANOVA was performed for each dataset. Left, For VGAT+ events, FSC gate (Df = 5, F = 21.51, p = 2.9e-12) and buffer (Df = 1, F = 4.70, p = 0.034) were both significant, with no significant interaction (Df = 5, F = 0.40, p = 0.85). Middle, For VGLUT1+ events, only FSC gate (Df = 5, F = 68.9, p < 2e-16) was significant, while buffer (Df = 1, F = 0.36, p = 0.55) and the interaction (Df = 5, F = 0.19, p = 0.97) were not. Right, For VMAT2+ events, FSC gate (Df = 5, F = 12.54, p = 2.4e-8) and buffer (Df = 1, F = 5.21, p = 0.026) were both significant, with no significant interaction (Df = 5, F = 0.27, p = 0.93). C, Box and whiskers plots of single-positive, double-positive, and triple-positive events expressed as a percentage of all fluorescent events within the respective FSC gate. Data are derived from the same immunostained P2 samples as in B. Two-way ANOVA was performed for each dataset. Left, For single-positive events, FSC gate (Df = 5, F = 19.83, p = 1.3e-11) and buffer (Df = 1, F = 6.74, p = 0.012) were both significant, with no significant interaction (Df = 5, F = 0.15, p = 0.98). Middle, For double-positive events, FSC gate (Df = 5, F = 23.25, p = 6.6e-13) and buffer (Df = 1, F = 8.00, p = 0.006) were both significant, with no significant interaction (Df = 5, F = 0.40, p = 0.85). Right, For triple-positive events, only FSC gate (Df = 5, F = 8.37, p = 4.7e-6) was significant, while buffer (Df = 1, F = 1.91, p = 0.17) and the interaction (Df = 5, F = 0.19, p = 0.97) were not. See also Extended Data Figures 7-1, 7-2, 7-3.

Although specific cases of neurotransmitter co-transmission and co-release have been observed (for review, see Hnasko and Edwards, 2012; Vaaga et al., 2014), the expression of VGLUT1, VGAT, and VMAT2 should be mutually exclusive for the majority of presynaptic terminals. For all immunopositive events within each FSC gate, we quantified the percentage of single-, double-, and triple-positive events (Fig. 7C). Similar to double calcein-positive events, we observed a steady decrease in single-positive and a steady increase in double-positive and triple-positive events with increasing FSC. Two-way ANOVA revealed a significant main effect of FSC gate for each event type (Df = 5, F = 19.83, p = 1.3e-11 for single-positive; Df = 5, F = 23.25, p = 6.6e-13 for double-positive; Df = 5, F = 8.37, p = 4.7e-6 for triple-positive). PBS-SET samples had higher single-positive event frequencies and lower double-positive event frequencies compared to PBS (Fig. 7C). The main effect of buffer was significant for single-positive (Df = 1, F = 6.74, p = 0.012) and double-positive (Df = 1, F = 8.00, p = 0.006), but not for triple-positive events (Df = 1, F = 1.91, p = 0.17). None of the buffer × FSC gate interactions were significant. Thus, although PBS-SET samples did display reduced aggregation, the effect was modest, especially for triple-positive events. Similar to the percentage of immunopositive events within each FSC gate, the frequency of false co-localization increases with FSC. To further illustrate this point, we calculated the percentage of immunopositive events for each antigen that were single-, double-, or triple-positive within each FSC gate (Extended Data Fig. 7-2). In the FSC noise region we found that ∼80% of VGAT+, ∼75% of VGLUT1+, and ∼63% of VMAT2+ events were single-positive. Strikingly, in the highest FSC gate, the single-positive percentage was only ∼12% for VGAT+, ∼51% for VGLUT1+, and ∼9% for VMAT2. These results suggest that many of the putative “synaptosomal” events in the upper FSC regions are aggregates containing multiple synaptosomes.

The quantification of double-positive events described above represents the sum of the three possible subtypes (i.e., VGAT+/VGLUT1+, VGLUT1+/VMAT2+, and VMAT2+/VGAT+). We wondered whether the frequency of these double-positive subtypes was related to the overall abundance of their respective antigens (Extended Data Fig. 7-1C). We used the percentage of all events immunopositive for each antigen to construct a simple probability model. In this model, a double-positive event represents the random sampling of two events with replacement. The probability of a particular double-positive subtype is therefore obtained by multiplying the frequencies of its respective event types (Extended Data Fig. 7-1D). Consistent with the expected frequencies of double-positive subtypes, we found that VGLUT1+/VGAT+ events were by far the most abundant (∼75%), followed by VGLUT1+/VMAT2+ (∼20%), and finally VMAT2+/VGAT+ (∼5%; Extended Data Fig. 7-1D). These results are consistent with a simple “collision” model of aggregation in the P2 crude synaptosome preparation. Furthermore, they strongly suggest that many double-positive events observed in these immunostaining experiments are aggregated particles rather than single synaptosomes co-expressing multiple vesicular transporters.

Although scatter measurements of submicron particles may vary dramatically across cytometers, fluorescence measurements can be readily calibrated. We determined the fluorescence intensity of VGAT/anti-mouse Alexa Fluor 488 immunostaining in molecules of equivalent soluble fluorophore (MESF) using calibrated bead standards (Extended Data Fig. 7-1E). Because particle fluorescence is directly related to FSC intensity regardless of the immunostaining procedure, many immunopositive events at a given FSC intensity would exhibit fluorescence intensity less than or equal to immunonegative events at a higher FSC intensity nearby. We therefore analyzed the median fluorescence intensity of gated immunopositive and immunonegative events within each FSC gate (Extended Data Fig. 7-1F). Numerical MESF data are also included in Extended Data Figure 7-3, which should facilitate comparison across laboratories.

Particle recovery following repeated centrifugation

Given the increased frequency of aggregates after centrifugation (Fig. 5D), in higher FSC regions (Fig. 6D–F), and in PBS (Fig. 6H), we wondered whether buffer and centrifugation might interact to bias sample composition toward aggregates following repeated centrifugations. Indeed, Gray and Whittaker (1962) previously reported increased recovery of acetylcholine in the low-speed P1 pellet when using saline instead of sucrose buffer, which was interpreted as aggregation in saline increasing the effective size of synaptosomes (and thus sedimentation at lower speed). To assess whether multiple centrifugation steps might bias particle recovery in PBS versus SET, we conducted two sequential 5-min centrifugations of P2 samples in PBS or SET buffer at varying speeds and used flow cytometry with absolute counting beads to determine the particle concentration of the pellets and supernatants (Fig. 8A). We found that under carefully controlled conditions with the absolute counting beads, PBS reduced particle abundance by ∼3-fold before any centrifugation (Fig. 8B). It is possible that certain particles in the P2 sample are susceptible to destruction in PBS, but several features of the data suggest that aggregation contributes to the reduction in particle counts. First, the median fluorescence intensity of membrane staining by FM4-64 was higher in PBS (Fig. 8B), which was due to increased relative abundance of brighter, larger events with higher FSC (Fig. 8C,D). In line with the findings of Gray and Whittaker (1962), we found that these “FSC high” particles were more efficiently pelleted in PBS at lower relative centrifugal force (2500–5000 × g) compared to SET buffer (Fig. 8D). The relative abundance of these particles was dramatically increased in the two pellets, P2-1 and P2-2, compared to the input sample and two supernatants (Fig. 8E). The relative abundance of the large particles in PBS samples was highest in 2500 × g pellets and lowest in 10,000 × g pellets, while no clear relationship between relative abundance and centrifugation speed was observed in SET buffer (Fig. 8F). A slight increase in these FSC high particles was observed when comparing spin 1 versus spin 2 for both buffers, but this effect was modest compared to the increase in PBS versus SET (Fig. 8F).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Effects of repeated centrifugation in different sample buffer on particle abundance and composition of P2 samples. These experiments were conducted using FM triggering on the Influx. A, Diagram of buffer/centrifugation scheme. A P2 sample resuspended in PBS or SET was aliquoted to measure input (P2-0), followed by two centrifugations for 5 min at the indicated relative centrifugal force (× g). An aliquot of the first pellet (P2-1) is reserved before the second spin. All samples are then analyzed by flow cytometry using absolute counting beads to measure particle number in each fraction. B, Input samples (P2-0) in PBS or SET buffer (n = 4 each). Left, Total particle counts. Right, Median fluorescence intensity of FM4-64 for all events. C, Representative histogram of FM4-64 fluorescence of all events for two replicate input samples (P2-0) in PBS and SET buffer. D, FSC versus FM4-64 density plots for P2-1 (top) or S2-1 (bottom) samples in PBS or SET buffer centrifuged at 2500 × g (left) or 5000 × g (right). FSC high gate corresponds to the top two FSC gates from Figures 6, 7. E, Gated FSC high events, expressed as % of all events, plotted for PBS (left) or SET (right) in each of five fractions moving through the centrifugation protocol at indicated relative centrifugal force (× g). Left to right in each panel, Input (P2-0), first pellet (P2-1) and supernatant (S2-1), second pellet (P2-2) and supernatant (S2-2). F, Gated FSC high events, expressed as % of all events, for P2-1 and P2-2 samples in PBS or SET at indicated relative centrifugal force (× g). See also Extended Data Figure 8-1.

As expected, the overall efficiency of particle recovery in the pellet of each spin increased with centrifugation speed, with the exception of the first spin in PBS (Extended Data Fig. 8-1A–C). Similar to the input samples, it appears that a combination of aggregation and/or destruction contribute to the loss of particle counts in the first PBS spin, although the recovery in the second spin was higher than in SET buffer. Critically, the first 5-min spin at 2500–5000 × g in either buffer leaves over half the P2 particles in the supernatant (Extended Data Fig. 8-1D). Collectively, these data suggest that P2 samples do undergo a certain degree of further fractionation during repeated centrifugation-based wash steps. Although some centrifugation bias is inevitable regardless of buffer and centrifugation speed, our data suggest that lower centrifugation speeds and ionic buffers will bias the overall sample composition toward larger particles that are more likely to be aggregates.