Survey of Red Fluorescence Proteins as Markers for Secretory Granule Exocytosis

Fluorescent proteins (FPs) have proven to be valuable tools for high-resolution imaging studies of vesicle transport processes, including exo- and endocytosis. Since the pH of the vesicle lumen changes between acidic and neutral during these events, pH-sensitive FPs with near neutral pKa, such as pHluorin, are particularly useful. FPs with pKa>6 are readily available in the green spectrum, while red-emitting pH-sensitive FPs are rare and often not well characterized as reporters of exo- or endocytosis. Here we tested a panel of ten orange/red and two green FPs in fusions with neuropeptide Y (NPY) for use as secreted vesicle marker and reporter of dense core granule exocytosis and release. We report relative brightness, bleaching rate, targeting accuracy, sensitivity to vesicle pH, and their performance in detecting exocytosis in live cells. Tandem dimer (td)-mOrange2 was identified as well-targeted, bright, slowly bleaching and pH-sensitive FP that performed similar to EGFP. Single exocytosis events were readily observed, which allowed measurements of fusion pore lifetime and the dynamics of the exocytosis protein syntaxin at the release site during membrane fusion and cargo release.


Introduction
Fluorescent proteins (FP) are extensively used as genetically encoded florescent tags [1,2] to address a multitude of biological questions. Their popularity is primarily due to the ease of labeling proteins of interest in living cells and organisms. Systematic development of novel and improved FPs has led to a palette covering the entire visible spectrum, but every FP comes with its own advantages and disadvantages. Newly developed FPs are usually benchmarked for brightness, photochemistry, photostability, cytotoxicity, pH-and ion sensitivity, as well as subcellular targeting accuracy with a fairly standardized panel of subcellular localization sequences [3,4,5]. However, the latter does not include specialized organelles such as synaptic vesicles or dense core secretory granules, organelles that are involved in regulated exocytosis and release of neurotransmitters and hormones.
Regulated exocytosis involves Ca 2+ -and SNARE-dependent fusion of secretory vesicles with the plasma membrane to empty its cargo [6,7], and FP-tagged markers are increasingly being used to study exocytosis in neuronal synapses [8] and even release of individual vesicles [7,9] by live cell microscopy. Since opening of the initial fusion pore during exocytosis leads to neutralization of the acidic pH of the vesicle interior, pH-sensitive FP's such as EGFP or pHluorin have been particularly useful in this respect. Depending on the pKa of the FP, the fluorescence of the intact acidic vesicle is dim but visible (EGFP, Venus), or absent (pHluorin). PHluorin has a neutral pKa [8,10] and is widely used for studying synapse function [11,12,13,14,15,16,17], exocytosis and fusion pore behavior [18,19,20,21,22], endocytosis [23,24], receptor trafficking [25,26], and endosome function [27]. It is essentially dark in the intact acidic vesicle lumen and reaches >70% of its maximal brightness at neutral pH [8,28], which is useful to reduce background fluorescence and therefore increase the signal to noise ratio of the measurements [8,10]. EGFP and Venus have slightly more acidic pKa's which lead to 2-3 fold fluorescence increases when acidic organelles are neutralized [9,19,29,30,31]. Vesicles labeled with these proteins remain visible even in the intact acidified state. If soluble vesicle cargo is labeled with such a FP, exocytosis leads to a rapid increase (flash) in vesicular fluorescence in the moment of fusion, followed by loss of the fluorescence as the cargo is released [9,30,32,33,34,35]. If instead a vesicle membrane protein is labeled, recycling and endocytosis processes can be studied [30,34,36,37].
For most other applications, it is desirable that FPs are insensitive to pH changes in the physiological range, and the development of novel FPs has consequentially focused on those with pKa <5. This has led to a paucity of pH-sensitive red FPs, as required for the study of exoand endocytosis. Despite this, red FP's such as mCherry [38] have been used for imaging secretory granule release in dual color applications. These essentially pH insensitive FPs (pKa<5) only report cargo release and cannot be used to study pore behavior during endocytic retrieval. Anecdotal reports also indicate that these FPs do not target as precisely as GFP derived markers, and that they may not report exocytosis as reliably as their green counterparts. More recently, several novel red FP's have become available, some of which are pH-sensitive (pKa>6). These include mApple [5], mNectarine [39], mRuby2 [40], mOrange2 [5] and td-mOrange2 [41], td-tomato [38], pHred [42], pHtomato [43], pHuji and TYG-Cherry [28]. Their usefulness for the study of dense core granule exocytosis has not been explored systematically, although some have occasionally been used to image exo-or endocytosis [37,41,44]. Here we systematically survey available red FPs as tags for the study of dense core granule exocytosis.

Image analysis
In Fig 1B-1D, the fluorescence emitted by the cell (F) was calculated as the pixel value in the footprint of the cell, corrected for out-of-cell background (bg). F is then normalized (FNH 4 + / F 0 ) to the first frame of the movie (F 0 ). All other measurements are specific for granules: Granules that were well separated from other granules or the edge of the cell were identified and an  algorithm implemented as MetaMorph journal then read the average pixel fluorescence in 1) a central circle (c) of 3 pxl (0.5 μm) diameter, 2) a surrounding annulus (a) with an outer diameter of 5 pxl (0.8 μm) and 3) an area not containing any cell as background (bg). Since the granule site is far smaller than the resolution of the microscope, the circle will contain all of the fluorescence originating from it. To obtain the specific granule fluorescence ΔF, the annulus value (a) was therefore subtracted from that of the circle (c) (ΔF = c-a). In some cases, ΔF was normalized to the first frame of the movie (ΔF 0 ) or the value obtained with NPY-mEGFP (ΔF mEGFP ). Syntaxin was quantified similarly at granule sites (ΔF syx ) and normalized with the off-granule syntaxin fluorescence (S = a-bg) [47]. S represents the local unbound concentration of syntaxin-EGFP. Positive ΔF/S values indicate binding of syntaxin-EGFP to the granule site, negative values indicate exclusion. Note that ΔF is given as per-pixel average for the entire 0.5 μm -2 circle, and ΔF/S values are therefore seemingly small. Assuming a cluster size of 50 nm, ΔF/S = 0.1 corresponds to at least 20-fold enrichment in the cluster beneath a granule.
Exocytosis events were found by eye as events with an obvious change in the fluorescence from the pre-exocytosis baseline and eventual rapid loss of the signal. This definition applied to both types of event, with or without preceding flash. The times of exocytosis and release were then obtained by non-linear fitting with a discontinuous function. The assumptions were (1) constant fluorescence before fusion, (2) inverted exponential decay just after fusion, and (3) exponential decay during content release (see Fig 6B and 6C). Fitting was carried out in Origin (OriginLab Corp, Northampton, MA, USA).
where t is time; c is average fluorescence in a 0.48 μm wide circle at the granule site; A1, A2 and A3 are the fluorescence values at the plateaus; τ1 and τ2 are the decay constants for the fluorescence increase after fusion and content release; t1 and t2 are the times of fusion and release, respectively, and s1 is the pre-exocytosis slope.
Co-localization was estimated by an observer, as follows. A journal in MetaMorph presented an observer (unaware of the image context) with square cutouts of the green channel (11 μm 2 ) that were centered on the position of each previously identified granule. The user then made a yes/no choice based on whether the center of the nearest perceived cluster was within one pixel of the center of the square, guided by an overlaid circle. Granule or nongranular structure (NGS) density was calculated using a script that used the built-in 'find maxima' function in ImageJ (http://rsbweb.nih.gov/ij) for spot detection.

Statistics
Experiments were repeated with at least 3 independent preparations. Data are presented as mean ± SEM unless otherwise stated. One-way ANOVA was used for testing statistical significance in Fig 1. All the other data was tested for statistical significance using Students t-test for two-tailed, paired or unpaired samples, as appropriate. SigniFIcant difference is indicated by asterisks ( Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001).

Brightness and bleaching of red fluorescent granule markers
We targeted a panel of ten red FPs or EGFP/pHluorin to insulin granules with the goal of comparing their brightness and bleaching rates. Fusions of the FPs with neuropeptide Y (NPY) were transfected in Ins1 cells. In all cases the footprint of the cells showed fluorescent puncta representing individual granules when imaged using TIRF microscopy ( Fig 1A). Neutralization of the granule pH with NH 4 + caused a rapid increase in granule fluorescence in case of EGFP but had no effect with mCherry (Fig 1A and 1C). In presence of ammonia, the granule brightness varied several-fold between red FPs with td-mOrange2 and td-tomato being brightest and mRuby2 and pHtomato being dimmest. Only the tandem dimers resulted in brighter granules (1.5-fold) than mEGFP or pHluorin (Fig 1C). The fluorescence increase observed with NH 4 + was by far the strongest with pHluorin (~9-fold). The increase with mEGFP was >2-fold, td-mOrange2, mNectarine, pHtomato and pHuji about 1.5-fold, and mOrange2, NPY-mApple and TYG-Cherry about 1.3-fold. For mCherry, mRuby2 and tdTomato the changes were negligible, indicating pH-independent fluorescence in the physiological range (Fig 1D). Hence only td-mOrange2, mOrange2, mNectarine, mApple, pHtomato, pHuji and TYG-Cherry are sufficiently pH-sensitive to monitor the intravesicular pH change during granule neutralization. As expected from pHuji's high pKa (7.6), it showed larger changes at extracellular pH 8.2 than at neutral pH (7.4) (Fig 1D). Transfection of granule-targeted pHred did not result in detectable fluorescence (not shown). Fluorescence bleaching rates were measured in the absence of NH 4 + during continuous exposure to excitation light and quantified by fitting with single exponential functions (Fig 2A  and 2B). Bleaching was slowest for mEGFP (τ = 15 s) and pHluorin (τ = 14.7 s), followed by TYG-Cherry, pHuji, mCherry, td-mOrange2 and tdTomato (τ = 6-13.4 s). Worst were mNectarine, mRuby2, mApple and pHtomato (τ = 2.7-5 s). Considering that td-mOrange2 and tdTomato were twice as bright as mEGFP, reducing the excitation power would have resulted in brightness and bleaching behavior similar to mEGFP for both RFPs. Since the bleaching rate in the acidic environment in the granule might be different from that at neutral pH we repeated the measurements in the presence of ammonia (Fig 2A and 2B, NH 4 + ). With the exception of NPY-TYG-Cherry, the bleaching rate was unaffected by the luminal pH.

Most NPY-RFPs are reliable markers of dense core granules
Targeting accuracy of reporter proteins is important for many applications. We used the granuphilin tagged with EGFP (EGFP-granuphilin) as fiduciary granule label to quantify correct targeting of NPY-FPs. Since granuphilin is a soluble protein that selectively binds to granule membranes and is involved in granule tethering and docking [52], it labels the large majority of mature granules in insulin secreting cells. Indeed, the density of EGFP-granuphilin labeled granules was similar to previous estimates of granule density [53]. The density of labeled granules was close to that with granuphilin for most of the FPs tested, with the exception of mNectarine, pHuji, and TYG-Cherry where it was reduced by about a third relative to granuphilin (Fig 3B). About 80% of granuphilin-positive granules were labelled with NPY-mCherry, -td-mOrange2 or-mApple and 50-60% were labelled with NPY-pHuji or-TYG-Cherry (Fig 3A  and 3C). Conversely, about 90% of NPY-mCherry, -td-mOrange2 and-mApple labeled puncta, and about 40-50% of NPY-pHuji or-TYG-cherry labeled puncta were granuphilin-positive (Fig 3A and 3D). RFPs have been reported to be somewhat more cytotoxic than EGFP [54] and expression of some of the NPY-RFPs led to larger non-granule structures (NGS) that are unlikely to be granules (Fig 4A). We quantified mistargeting by co-expression with the endosomal marker EGF-P-Rab7 [55,56] (Figs 3D and 4A-4C). With mCherry, td-mOrange2 and mApple fewer than 10% of all NPY-FP labeled structures co-localized with the endosome marker (Fig 3D). However, about 30% of NPY-pHuji and TYG-Cherry labeled structures co-localized with rab7 ( Fig  3D). The density of NGS varied among the different RFPs used, and was lowest with td-mOr-ange2, slightly higher with mCherry and mApple, and strongly increased with NPY-pHuji and-TYG-Cherry (Fig 4A and 4B). The data suggest that tandem-dimer RFPs may be better tolerated than the monomeric FPs (Fig 4A and 4B). The reason why pHuji and TYG-Cherry target worse is not understood. It should be pointed out that most rab7-positive structures were large and most NPY-labeled NGS co-localized with rab7 (Fig 4A-4C), indicating their endosomal nature. Only about 2% of the small (diffraction limited) NPY-RFP positive structures carried EGFP-rab7. Small NPY-FP labeled structures therefore most likely represent secretory granules.

Performance of NPY-RFPs for detection of single exocytosis events
Next we tested the performance of RFPs in reporting single granule exocytosis events. Exocytosis was evoked by local application of K + , which leads to depolarization and Ca 2+ -influx. After onset of the stimulation, exocytosis events were readily observed as sudden loss of granule fluorescence due to release of the label into the extracellular buffer. Cells expressing NPY-mEGFP and NPY-mCherry served as pH-sensitive and-insensitive references. With mEGFP, but not mCherry, this loss of fluorescence was often preceded by a temporary increase in fluorescence (flash), which we interpret as the opening of the fusion pore and concomitant change in luminal pH [9] (Fig 5A-5C). With NPY-mEGFP both flash and non-flash events were observed, presumably because pore opening and release can occur simultaneously. With both td-mOr-ange2 and mApple both types of events were observed. Flash events were somewhat more frequent with td-mOrange2 than with mEGFP, possibly reflecting the differences in their sensitivity to pH changes. The all-events average ( Fig 5C) shows a notable transient increase with td-mOrange2, similar to what is observed with mEGFP. In contrast, this was hardly noticeable with mApple ( Fig 5A and 5D), likely due to the fact that flash events were rare ( Fig 5D). As expected from their pKa's, pHluorin and pHuji resulted in flash events where the granules were not visible before exocytosis (Fig 5B). However, with pHuji about half of the events lacked the flash (upper example in Fig 5B, pHuji), consistent with the relatively small changes observed with ammonia (Fig 1). The total number of exocytotic events was similar for all FPs tested (Fig 5E and 5F). Large NGS (cf. Fig 4) never underwent exocytosis.

NPY-td-mOrange2 reports exocytosis and release in dual color applications
Red granule markers are compatible with the large number of existing green FP-labeled proteins for dual color experiments. To demonstrate the usefulness of this we co-transfected NPY-td-mOrange2 together with NPY-mTurquoise2 ( Fig 6A); both labels co-localized well. mTur-quoise2 has a pKa of 3.1 and should not respond to physiological pH changes [50]. Indeed, the mTurquoise2 signal did not change during the rising phase of the td-mOrange2 signal, but both proteins were lost simultaneously during K + -evoked exocytosis (Fig 6B and 6C). To formally test what phase of the td-mOrange2 flash coincided with loss of mTurquoise2, a tri-part discontinuous function was fit to the individual traces (Fig 6C, red lines). From this we obtained the times of rising and falling phases (arrows) and calculated delays relative to these reference events. The analysis confirms that NPY-mTurquoise2 is lost simultaneous with the falling phase of the NPYtd-mOrange2 flash (Fig 6E). The results corroborate that the rising phase is due to luminal pH change rather than granule movement in the z-axis, and the falling phase due to content release.
We next analyzed cells co-expressing NPY-td-mOrange2 and syntaxin1a-EGFP. Syntaxin is a plasma membrane SNARE protein that lies at the core of the exocytotic machinery [47,57,58]. It is well-established that syntaxin forms small clusters beneath docked secretory granules, which are 50-100 nm in diameter and contain~70 syntaxin molecules [47,53,59,60]. We confirmed that syntaxin clusters were readily observed and co-localized with docked granules (Fig 6A). During exocytosis of single granules, the associated syntaxin clusters rapidly dissolved (Fig 6A and 6B), as observed earlier [47,53]. To test whether syntaxin loss conincides with pore opening or content release, we used again non-linear fitting. The delay between content release and onset of syntaxin loss was -0.76±0.13 s, compared with only 0.11±0.16 s between pore opening and syntaxin loss (Fig 6D and 6E). Finally, all traces were temporally aligned to either the falling phase of NPY-td-mOrange2 (release) and averaged (Fig 6F). This analysis again shows that the syntaxin signal begins to decay before content release from the granule, while NPY-mTurquoise2 remains trapped until NPY-td-mOrange2 is released. Thus, statistical analysis ( Fig 6D) and alignment of the traces (Fig 6B and 6E) indicate that loss of syntaxin coincides with the rising phase of td-mOrange2, suggesting that syntaxin dispersal is initiated already during pore formation. Experiments like this illustrate the utility pH-sensitive red FP's for imaging the behavior of the exocytosis machinery.

Discussion
In our survey of red FP's utility to tag secretory granule cargo, we identified td-mOrange2 [41] as bright, photostable label that is efficiently targeted as fusion protein with NPY (see Table 1). As marker for secretory granule exocytosis, NPY-td-mOrange2 is comparable to the widely used NPY-mEGFP and other EGFP based fusion proteins and reports both fusion pore opening and content release due to its pH sensitivity. td-mOrange2 facilitates dual labeling experiments, in which exocytosis is studied together with other labeled proteins [30,32,33,34,41,53,61], sensors such as Ca 2+ -sensitive dyes (e.g. fluo-4), or knockdown approaches using fluorescently labeled siRNA, where it is often desirable or more convenient to use the red channel for a well-characterized organelle marker, and the green channel for other purposes. However, as suggested by its pKa, mOrange2 is much less sensitive to physiological pH changes than pHluorin.
In vitro, mOrange2 and tdTomato are among the brightest and slowest bleaching red FP's [5]. Since this may not reflect the behavior in live cells, we compared brightness and photostability for red FP's that were targeted to secretory granules. In intact granules, the td-tomato fusion was brightest, followed by td-mOrange2. However, given the 1.7-fold fluorescence increase upon neutralization suggests that NPY-td-mOrange2 is the brightest available red granule marker. Td-tomato and td-mOrange2 were somewhat more photostable than other tested red FP's, which together with their improved brightness should enable long term imaging of vesicle behavior. Td-mOrange2 was originally created by duplexing mOrange2 [5] to facilitate detection of synaptic exo-and endocytosis [41]. For labeling small synaptic vesicles, the main advantage of the tandem dimer over monomeric mOrange2 is increased brightness, presumably due higher fluorophore density on the vesicles. We confirm here that granules labeled with td-mOrange2 are brighter than those labeled with the mOrange2 monomer, without changes in the release kinetics during exocytosis. Most FPs tested here labeled at least 85% of the granuphilin positive granules, with the notable exceptions of pHuji and TYG-Cherry. Since granuphilin is soluble in the cytosol and bound to granules by weak interactions [62,63,64,65], it presumably labels all granules regardless of their history. NPY fusion constructs therefore lead to rapid, near quantitative labeling, which is important for many applications where knowledge about the absolute number of granules is required. However, mistargeting differed between the FP's and the tandem dimer RFP fusions were generally better in this regard than the monomeric fusions, as judged by the lower density of large, rab7 labeled structures. Targeting accuracy of the tandem dimer to granules was comparable to that of EGFP, as was the fraction and density of labeled granules. We speculate that mistargeting results from protein aggregation at high luminal concentration, which is more prominent for the monomeric FPs than tandem dimers. Alternatively, lysosomal degradation of the fusion proteins might be more efficient for tandem dimers than monomeric versions, which leads to less accumulation within such structures. Steric effects as previously observed with dimers [5] do not appear to be a problem within secretory granules.
Green/Yellow FP's with pKa>6 such as EGFP (pKa = 6) and pHluorin (pKa = 7.6) detect the luminal pH change during exocytosis and are therefore powerful tools to study membrane fusion and fission [9,10,24,30,34,66]. In (neuro)endocrine cells, dense core granules are large and sparse enough to be imaged individually and allow pH-sensitive FP's to be used to detect the opening of the exocytotic fusion pore and to measure its life time [9,19,35,61]. Since EGFP retains some of its fluorescence in intact granules, it is a preferred granule marker as it allows observation and tracking of the intact organelles before exocytosis. In neurons, where synaptic vesicles are usually too small and too densely packed for individual observation, average signals from entire synapses are recorded. For this purpose, background suppression is important to increase the signal-to-noise ratio. PHluorin, which is non-fluorescent at the pH of intact synaptic vesicles, is therefore the protein of choice to detect neuronal exocytosis. However, red FP's were originally developed to lack the often undesirable pH dependence [5,38,67], although some pH dependent variants have appeared (mOrange, mApple) [5]. Only recently, several novel pH-sensitive red FP's have been specifically developed for imaging of endo-and exocytosis, including pHred [42], pHtomato [43] and pHuji [28]. Td-mOrange2 retains some of its fluorescence in intact granules. In contrast, pHuji is nearly non-fluorescent while in the lumen of some acidic vesicles [28], and is therefore expected to be an important complement to td-mOrange2. It is not clear why a significant fraction of the pHuji-labeled granules were bright before exocytosis (Fig 5B), while this was not observed with pHluorin. It is possible that these granules failed to acidify. Of the other pH dependent RFPs, mApple and mNectarine (reported pKa = 6.5 and 6.9, respectively; [5]) responded as expected to neutralizing luminal pH with NH 4 + , but were both dim and bleached rapidly. For reasons not understood, mApple did not detect the opening of the fusion pore during exocytosis. In summary, we identify td-mOrange2 [41] as useful label for dense core granules; it is bright, well-targeted and spectrally separated from many blue and green FPs. It is likely suitable also for the study of many other membrane fusion or fission reactions, such as endocytosis and constitutive exocytosis.
Council, Swedish Diabetes Society, European Foundation for the Study of Diabetes, Diabetes Wellness Network Sweden, Swedish Strategic Research area in diabetes EXODIAB, Barndiabetesfonden, EXODIAB, and the NovoNordisk-, Family Ernfors-, and OE&E Johanssonsfoundations.

Author Contributions
Conceived and designed the experiments: NRG SB. Performed the experiments: NRG KV MH. Analyzed the data: NRG KV MH. Contributed reagents/materials/analysis tools: JS PY. Wrote the paper: NRG SB.