|Home | About | Journals | Submit | Contact Us | Français|
Proteins prone to misfolding form large macroscopic deposits in many neurodegenerative diseases. Yet the in situ aggregation kinetics remains poorly understood because of an inability to demarcate precursor oligomers from monomers. We developed a novel strategy for mapping the localization of soluble oligomers and monomers directly in live cells. Sensors for mutant huntingtin, which forms aggregates in Huntington’s disease, were made by introducing a tetracysteine motif into huntingtin that becomes occluded from binding biarsenical fluorophores in oligomers, but not monomers. Up to 70% of the diffusely distributed huntingtin molecules appeared as submicroscopic oligomers in individual neuroblastoma cells expressing mutant huntingtin. We anticipate the sensors to enable insight into cellular mechanisms mediated by oligomers and monomers, and for the approach to be adaptable more generally in the study of protein-protein self-association.
The misfolding of proteins into macroscopic aggregates such as plaques and inclusion bodies is a hallmark of Alzheimer’s, Creutzfeldt-Jakob, Huntington’s, and other neurodegenerative diseases (Chiti et al., 2006). However, the role of aggregation per se in disease is controversial. Indeed, in many of these diseases, notably Huntington’s, Alzheimer’s and Parkinson’s, small soluble oligomers or other diffuse forms of Htt, which may be precursors to the macroscopic aggregates, seem to correlate best with cellular dysfunction (Arrasate et al., 2004, Haass et al., 2007). Despite evidence that soluble oligomers are more toxic than the macroscopic aggregates, quantitative assessment of the formation of the earliest, submicroscopic oligomers in live cells is limited by the lack of suitable tools. To overcome these limitations, we developed a novel approach to rapidly distinguish monomers from all other soluble aggregates directly in live cells.
Our approach centres on the mutant exon 1 fragment of huntingtin (Httex1), which has been used extensively to model Huntington’s disease in transgenic mice, mammalian cells, Drosophila, yeast, and Caenorhabditis (Brignull et al., 2006, Duennwald et al., 2006, Jackson et al., 1998, Mangiarini et al., 1996). Huntington’s disease is caused by CAG-codon duplications in the Htt gene that translate into an abnormally long polyglutamine (polyQ) sequence near the amino terminus of the protein (MacDonald et al., 1993). The pathogenic polyQ expansions cause Htt to aggregate and redistribute from a diffuse cytosolic pool into inclusion bodies during disease progression. Normally Htt has 7–36 glutamines in the polyQ repeat, while in disease there are 37 or more (MacDonald et al., 1993). The number of glutamines beyond this threshold is highly correlated with earlier onset of disease and the rate of protein aggregation (Chen et al., 2001, Perutz et al., 1994, Scherzinger et al., 1999). PolyQ expansions have been described in at least eight unrelated proteins that also cause disease, suggesting that aggregation of polyQ is concomitant with cellular dysfunction more generally (Zoghbi et al., 2000).
Our approach to distinguish Httex1 monomers from other oligomers relies on genetically tagging the protein with two labels. The first is Cerulean, a brighter, cyan-fluorescent variant of enhanced green fluorescent protein (EGFP) (Rizzo et al., 2004), that detects Httex1 localization. Httex1 fused to EGFP has been used to report on polyQ-dependent inclusion body formation and key cellular toxicities associated with that process (Arrasate et al., 2004, Bennett et al., 2007). The second is a tetracysteine (TC) motif, Cys-Cys-Pro-Gly-Cys-Cys, that binds with high affinity (nanomolar dissociation constant) and specificity to the cell permeable biarsenical dyes FlAsH and ReAsH (Adams et al., 2002). Upon binding to a TC motif, FlAsH (green fluorescence) and ReAsH (red fluorescence) increase in fluorescence up to 50,000-fold; in the unbound state, the dyes are essentially nonfluorescent (Adams et al., 2002).
Developed as small, and hence less structurally disruptive, alternatives to fluorescent proteins, TC tags and biarsenical dyes have also been used to report on protein folding and conformational changes (Coleman et al., 2009, Ignatova et al., 2004, Luedtke et al. 2007, Roberti et al., 2007). We devised a novel adaptation of the TC motif so that the conversion of Httex1 monomer into oligomers rendered the TC tag inaccessible to ReAsH. We extended this approach to other colour combinations by using Httex1 fused to the red fluorescent protein Cherry to track Httex1 localization and the use of FlAsH to selectively label the monomeric forms of Httex1. Hence, two colour fluorescence detection for ReAsH reactivity (red) and Cerulean (cyan) or the alternative pair of FlAsH reactivity (green) and Cherry (red) enabled the quantitative monitoring of protein oligomerization directly in situ. Here we describe the development and validation of the approach and its use to visually distinguish monomers from all other oligomers of Htt within living cells.
A single TC tag was initially inserted into the Httex1-Cerulean sequence at various locations with the goal of identifying sequence positions in which TC was exposed in Httex1-Cerulean monomers and buried in oligomers. Because high-resolution structures of Httex1 are not available, we used a systematic scanning approach guided by known conformational properties of Httex1. Httex1 monomers sample a series of compact and non-compact conformations with little stable secondary structure, but the polyQ region is predicted to form a highly stable, compact β-sheet core in aggregated Htt (Bennett et al., 2002, Crick et al., 2006, Dougan et al., 2009, Perutz et al., 1994). Thus, TC tags inserted near the polyQ region are likely to bind biarsenical dyes in the monomer due to the high solvent accessibility but be inaccessible in oligomers and larger aggregates (Fig. 1A). Hence, in our first generation of variants, the TC tag was placed at the junction of Htt and Cerulean (TC1), within the proline-rich region of Htt (TC2), or at the N-terminus of the polyQ repeat sequence (TC3) (Fig. 1B; Fig. S1A).
To test TC1–3 as aggregation reporters, we expressed and purified the recombinant proteins from Escherichia coli using a pathogenic polyQ length (Httex146Q) that readily forms amyloid-like fibrils in vitro (Scherzinger et al., 1999). To regulate the rate of aggregation in vitro, we fused Httex146Q-Cerulean to the C-terminus of maltose binding protein, which inhibited aggregation in a manner similar to that of Httex1 fused to glutathione S-transferase (Schaffar et al. 2004, Scherzinger et al., 1999) (Fig. 2A). Tobacco etch virus (TEV) protease liberated monomeric Httex146Q-Cerulean, which over time formed aggregates that were insoluble in sodium dodecyl sulfate (SDS) (Fig. 2A).
To validate Cerulean as a robust indicator of protein levels regardless of aggregation status, we assessed its fluorescence in soluble protein (monomer) and aggregated states. We found no difference (Fig. S2A–C), consistent with the ability of EGFP, which is similar to Cerulean in structure and sequence, to report on Httex1 quantity in cells (Arrasate et al., 2004).
Next, we determined whether the soluble protein or aggregates of each variant of Httex146Q-Cerulean could bind to ReAsH in a native slot blot assay. In this assay, recombinant protein in nondenaturing buffers becomes affixed to a nitrocellulose membrane, while unbound ReAsH passes through (Fig. 2B). Httex146Q-Cerulean soluble protein and aggregates without the TC tag bound to the membrane as assessed by Cerulean fluorescence, and ReAsH did not bind appreciably. In contrast, all TC-tagged Httex146Q-Cerulean soluble protein bound to ReAsH (Fig. 2B). TC1- and TC2-Httex146Q-Cerulean also bound ReAsH in the aggregated forms, but TC3-Httex146Q-Cerulean bound poorly, which suggests that TC3 is inaccessible to ReAsH upon aggregation (Fig. 1A).
To verify that TC3 reports on aggregation without perturbing the conformation and aggregation potential, we examined the aggregates by negative-staining electron microscopy. The fibrillar architecture of the TC1 3 variants and the non-TC-tagged forms were similar (Fig S3). In an assay monitoring the quantity of pelleted protein, the kinetics of fibril formation were also similar for TC3-Httex146Q-Cerulean and the non-TC counterpart (Fig. 3A). Thus, the TC tags have little, if any, influence on the core aggregation properties of Httex146Q-Cerulean into amyloid-like fibrils.
We next assessed the ability of TC3 to report on the aggregation kinetics of Httex1-Cerulean, as judged by ReAsH reactivity (Fig. 3B). TC3 ReAsH reactivity decreased as the quantity of pelleted aggregates increased, indicating that TC3 demarcates soluble protein from aggregates in constructs of variable polyQ length (Fig. 3B).
To determine whether the TC3 reporter distinguishes monomers from oligomers in live neuroblastoma (Neuro2a) cells (Fig. 4A), we assessed in situ ReAsH reactivity with TC3 in Httex125Q-Cerulean, which has a subpathogenic polyQ length. Untagged (control) Httex125Q-Cerulean had a diffuse, cytosolic distribution with no ReAsH fluorescence and no inclusion bodies. The TC1 counterpart had an identical phenotype but had substantial ReAsH reactivity. In contrast, TC3-Httex125Q-Cerulean was unexpectedly distributed into small globular structures that bound poorly to ReAsH (Fig. 4A). TC3-Httex146Q-Cerulean was also distributed into small globular structures, whereas the TC1 and the non-tagged counterparts formed inclusion bodies in many cells, consistent with the expected behaviour of the 46Q context (Arrasate et al., 2004).
The first 18 residues of Htt are critical for its cellular localization and targeting to the endoplasmic reticulum and mitochondria (Atwal et al., 2007, Rockabrand et al., 2007, Steffan et al., 2004). Since TC3 lies between residues 17 and 18 of Httex1, our data suggest that TC3 disrupts normal interactions with cellular components and ligands through this motif. Yet while the kinetics studies of recombinant protein (Fig. 3) suggested that this defect is independent of its intrinsic aggregation properties, we cannot exclude the possibility that TC3 changed the aggregation potential of the polyQ sequence in the cellular environment. Indeed, a recent study implicated the first 17 residues as important in modulating aggregation rates (Thakur et al., 2009).
Hence, to further investigate the cause of this defect, we introduced TC3 into polyQ25-EGFP and polyQ46-EGFP, which lack Htt-specific sequences other than the polyQ sequence (Fig. S1B). TC3 did not influence the localization or inclusion body formation of polyQ25-or polyQ46-EGFP, confirming that TC3 does not intrinsically alter aggregation in cells (Fig. 4B). Additionally, ReAsH bound readily to the TC3-polyQ-EGFP constructs, suggesting that TC3 can be adapted to other polyQ-containing proteins.
To prevent the defects TC3 causes in Httex1-Cerulean, we initially tried two approaches. First, since a TC tag reports on oligomerization when juxtaposed to the N-terminus of polyQ and might also do so when juxtaposed to the C-terminus, we shifted the TC tag outside the first 18 residues of Htt (TC4, Fig. 1B). However, in the native slot blot assay, there was no difference in ReAsH reactivity between soluble protein and aggregates of recombinant TC4-Httex146Q-Cerulean (Fig. 2B).
Second, we retained the position of TC3 but inserted a new polyQ6 linker sequence immediately N-terminal to the TC tag to restore the 18-residue homology of Htt (TC5; Fig. 1B). Recombinant TC5-Httex146Q-Cerulean showed a marked decrease in ReAsH reactivity from soluble protein to aggregate in the native slot blot assay, providing confidence that this second approach might work (Fig. 2B). In neuro2a cells, TC5-Httex125Q-Cerulean had a mostly normal cellular distribution; however, approximately 25% cells also displayed a more punctate cellular distribution, suggesting that the linker did not fully rescue the defects of TC3 (Fig. S4A).
Next, we extended the linker sequence before TC3 from 6Q in TC5 to 11Q (TC6), a polyQ repeat length comparable to the shortest natural variants in human Htt (Zuhlke et al., 1993) (Fig. 1B). To our surprise, TC6-Httex125Q-Cerulean, which has an 11Q sequence followed by the TC tag and endogenous 25Q sequence, formed inclusion bodies in many cells (Fig. S4A), perhaps because the new polyQ sequence N-terminal to TC3 increases the aggregation potential of the endogenous polyQ sequence. To determine if the enhanced inclusion body formation was independent of Httex1-specific trafficking defects caused by TC3, we introduced TC5 into polyQ25-EGFP. TC5-polyQ25-EGFP had a phenotype similar to that of TC5 in the Httex1-Cerulean context, confirming that the enhanced aggregation is independent of the localization defects (Fig. 4B).
We also investigated whether shortening the linker would reduce the aggregation potential without causing mislocalization (2 or 5 glutamines; TC7 and TC8, respectively; Fig. 1B). TC7 displayed the same defect as TC3, indicating that at least the first 19 residues of Htt are necessary for normal cellular localization; however, TC8-Httex125Q-Cerulean was mostly normally distributed, but over time did form punctate aggregates, albeit to a lesser extent than TC5 and TC6 (Fig. S4A).
NMR studies have shown that a TC tag forms a β-hairpin centered around the Pro-Gly residues when complexed to ReAsH, suggesting that TC tags stabilize a polyQ hairpin configuration that favors aggregation (Madani et al., 2009). Repeating short polyQ sequences separated by Pro-Gly β-hairpins synergistically promote aggregation to a similar or greater extent than the contiguous polyQ sequence lacking the β-hairpins (Thakur et al., 2002). Conversely, aggregation is inhibited when the Pro-Gly β-hairpins are replaced with a single Pro residue, which by itself destabilizes configurations favorable for aggregation (Thakur et al., 2002). To inhibit this mode of TC-facilitated aggregation, we inserted a polyQ linker necessary for proper Htt trafficking (5Q) flanked at the C-terminus by a single Pro, immediately N-terminal to the TC3 tag (TC9; Fig. 1B). We also examined a construct containing a tandem polyQ5-proline-polyQ4-proline immediately N-terminal to the TC3 tag (TC10; Fig. 1B). In neuro2a cells, both TC9- and TC10-Httex1-Cerulean showed a normal cellular distribution indistinguishable from that of untagged versions for both the 25Q and 46Q contexts, indicating that we had both rescued the localization defect and prevented TC-facilitated aggregation (Fig. S4B).
To validate TC9 and TC10 as aggregation reporters, we examined recombinant TC9- and TC10-Httex1-Cerulean for ReAsH reactivity of soluble protein versus aggregates in the native slot blot assay. TC9 and TC10 bound ReAsH in soluble protein (25Q and 46Q) but bound ReAsH poorly as aggregates (46Q) (Fig. 2C). TC9 and TC10 also tracked Htt aggregation kinetics by ReAsH reactivity, whereas TC1, whose ReAsH reactivity is insensitive to aggregation state, did not (Fig. 5A). TC9 and TC10 did not influence aggregate morphology (Fig. 5A,B). Furthermore, TC1 and TC9 reported on aggregation similarly with the green-fluorescent biarsenical dye FlAsH, indicating that both of the most commonly used biarsenical dyes can be used to demarcate monomers from oligomers (Fig. S5).
To determine precisely which oligomeric forms of TC9- and TC10-Httex146Q-Cerulean bind ReAsH, we subjected recombinant TC1, TC9- and TC10-tagged Httex146Q-Cerulean to gel-filtration chromatography. After a 6-h aggregation, the intermediate time point for aggregation kinetics, approximately 5–10% of Httex146Q-Cerulean consisted of material that pelleted by centrifugation (Fig. 5A). Low-speed pelleting (1500 g, 5 min) was used to remove the largest aggregates that cannot traverse the Superdex 75 column, and the remaining soluble material was assessed for oligomeric size. The predominant species eluted at a time point consistent with monomer (Fig. 5C). However, a small amount of higher-mass material was found in the void volume, which likely comprises non-pelleted oligomers or material which had self-associated during the chromatography fractionation. Fractions corresponding to eluted protein were assessed for ReAsH reactivity and the TC1, TC9-10 monomer fractions all bound ReAsH. In contrast, only TC1-tagged Httex146Q-Cerulean eluting in the void volume bound ReAsH. The absence of TC9 and TC10 binding to the oligomers in the void volume indicates that the smallest detectable aggregates cannot bind ReAsH. This finding supports the conclusion that TC9 and TC10 binds ReAsH only in monomers, whereas TC1 binds ReAsH similarly in monomers, oligomers, and pelleted aggregates (Fig. 5C).
The TC1, TC9, and TC10 constructs were used to examine the extent of aggregation of Httex1-Cerulean in neuro2a cells. We first focused on inclusion bodies formed by Httex146QCerulean (Fig. 6). TC9 and TC10 showed far less ReAsH reactivity than TC1, which is consistent with inclusion bodies consisting primarily of aggregates (Kim et al., 2002). In Httex125Q-Cerulean, TC1, TC9, and TC10 all bound ReAsH similarly as indicated by tight colocalization of Cerulean and ReAsH fluorescence intensities, consistent with all tags being accessible to bind ReAsH and a lack of oligomers in the cytosol (Fig. 7).
We next focused on the cytosolic pool of Httex146Q-Cerulean. TC1 bound ReAsH to about the same extent as it did in Httex125Q-Cerulean (Fig. 7). In contrast, TC9 and TC10 showed markedly less ReAsH reactivity than TC1 in the context of Httex146Q-Cerulean and TC9 and TC10 in the context of Httex125Q-Cerulean (Fig. 7). Thus, the diffuse pool of cytosolic Httex146Q-Cerulean evidently comprises a large, evenly distributed proportion of oligomers that are otherwise not visible by confocal microscopy. Analysis of pixel intensity correlation plots and ReAsH fluorescence standardized to the Cerulean fluorescence intensity indicated that the binding of ReAsH to cytosolic TC9- and TC10-Httex146Q was approximately 50–70% less than the binding of ReAsH to TC9- and TC10-Httex125Q-Cerulean in individual cells (Fig. S7).
We also examined whether FlAsH could be similarly used to report on aggregation in live cells. While FlAsH did bind to TC1 and TC9-Httex1-Cerulean, data was difficult to interpret due to a highly efficient FRET from Cerulean to bound FlAsH. Hence, while FRET may have a useful application in the study of Htt conformations, it hinders the utility of Cerulean as an independent marker of protein localization and concentration. To further explore the use of FlAsH to report on Httex1 aggregation, we replaced the Cerulean moiety with the monomeric red fluorescent protein Cherry. The Cherry/FlAsH labeling combination produced the same result as the Cerulean/ReAsH, validating its use as an alternative spectral toolset for imaging (Fig. S8). Examination of cell-cell differences in ReAsH (or FlAsH) reactivity using the TC9- and TC10-Httex146Q-Cerulean (or Cherry) suggested that different cells have different proportions of cytosolic aggregate/monomer content. In several cells that contained a small inclusion body, overall biarsenical dye reactivity for the cytosolic pool of protein was lower than in neighboring cells (Figs. 7 and S8). Other cells devoid of inclusion bodies also had lower biarsenical dye reactivity than neighboring cells, which raises the possibility that these cells are on track to forming an inclusion (Fig. S8).
The development of TC tags as sensors to monitor protein self-association by tetracysteine burial has provided a novel means to distinguish between monomers and higher-order oligomers directly in live cells. Here we showed that this approach can be used for FlAsH (green) or ReAsH (red) as sensors of monomers and with Cerulean (cyan) or Cherry (red) as markers of protein localization. Indeed, the ability to monitor the proportion of monomeric relative to oligomeric or microscopically aggregated Htt in situ will be a powerful toolset for investigating how oligomers form, how they are trafficked, how they relate to inclusion body formation and cellular dysfunction.
A key application of the huntingtin sensors will be in deciphering the precise molecular steps that lead to aggregation and inclusion formation within living cells. The formation of classic inclusions, often used as an indicator of overall aggregation, is a complex cell-mediated process driven by dynein-mediated retrograde transport of misfolded proteins or microaggregates (Kopito, 1999). The nature and toxicity of the forms of protein that precede inclusion formation are largely unknown. However, different aggregation-prone proteins (eg. SOD1 mutants, PABPN1, and pathogenic Httex1) form inclusions with distinct biophysical properties or cellular localizations, suggesting that multiple cellular mechanisms exist to process misfolded or soluble oligomers of proteins (Berger et al., 2006, Matsumoto et al. 2006). In support of this hypothesis, accessory proteins in yeast, such as Bmh1 and Cdc48, alter the intracellular distribution of aggregates and/or inclusion formation (Wang et al., 2009). Moreover, misfolded proteins have been reported to partition into two distinct quality-control compartments that appear visually as inclusions—a juxtanuclear compartment where soluble misfolded proteins are concentrated and a perivacuolar compartment for terminally aggregated proteins (Kaganovich et al., 2008).
In combination, the TC1, TC9, and TC10 markers suggested that the cytosolic pool of Httex146Q had a high proportion of oligomers in many cells 50–70% in Fig. 7) expressing the pathogenic form of Httex1—even some cells lacking inclusion bodies. Hence, Httex1 aggregation appears to occur evenly first in the cytosol before trafficking to inclusions. This conclusion is similar to that reached for a FRET study of another polyQ-containing protein, truncated atrophin-1, whereby pathogenic polyQ lengths induced more oligomers in the cytosol than the non-pathogenic counterpart (Takahashi et al., 2008). These findings highlight a disconnect between visible inclusions and microscopic aggregation and raises the possibility that in disease cellular activities are compromised by pools of oligomers, supporting the notion that inclusions are a protective response by cells to sequester diffuse oligomers.
However, while our data are consistent with Httex1 forming evenly distributed oligomers in the cytosol, we cannot entirely exclude the possibility that pathogenic Httex1 acquires a novel monomeric conformation non-permeable to ReAsH binding. This hypothesis is supported by the finding that monomers of polyQ-fusions to thioredoxin convert from an α-helical to more compact β-sheet conformer before conversion into amyloid-like fibrils (Nagai et al., 2007). In addition, Httex1 monomers collapse into a more compact conformation immediately after cleavage from glutathione-S-transferase, which like MBP, inhibits Htt aggregation (Schaffar et al., 2004). These and other results suggest that monomeric huntingtin, while lacking regular secondary structure, does sample a range of partially compact conformers (Crick et al., 2006, Dougan et al., 2009).
The work is significant for two reasons. First, our study shows soluble oligomers of mutant Htt, constituting up to 70% of the diffuse huntingtin molecules, which are otherwise not discernable from monomers by conventional confocal microscopy, to be evenly distributed in the cytosol of many cells and are distinct to the visible inclusion bodies. Hence, small soluble aggregates of Htt are likely to confer a profound burden on cellular functioning in a manner uncorrelated to IBs, which have been previously postulated to play a direct role in Htt toxicity (reviewed in [Hatters, 2008]). The second significance is that we describe a new use of TC tags for reporting on protein self-association directly in live cells. The TC tags only bind ReAsH and FlAsH in monomers due to ReAsH occlusion upon oligomerization, which enables monomers to be quantitatively demarcated from oligomers in situ in live cells when used in tandem with a fluorescent protein reporter for localization.. This methodology, which we refined for Htt should be readily adaptable to the other proteins containing polyQ, enabling insight to be gained as to the role of oligomers in the 8 other diseases caused by polyQ expansions. Furthermore, we anticipate the fundamental TC burial methodology to be useful in the study of other proteins that undergo oligomerization as a part of normal cellular biology, or with proteins associated with misfolding- and aggregation-related diseases such as Alzheimer’s, Creutzfeldt-Jakob, and Parkinson’s.
The cloning of the constructs, the expression and purification of recombinant protein, and the preparation of monomeric Httex1-Cerulean or Cerulean-TC are described in the Supplementary Experimental Procedures.
Aliquots of TEV-cleaved fusion proteins (2 µM fusion protein) were either freshly thawed from −80 °C storage (10-min TEV cleaved samples) or previously thawed and incubated at 37 °C for 72 h under mineral oil (72 h samples) prior to analysis. Samples of protein were diluted to a final concentration of 0.8 µM and volume of 100 µL in TNBT buffer (20 mM Tris, 150 mM NaCl, 250 µM BAL (2, 3-dimercapto-1-propanol), and 1 mM tris-carboxyethyl-phosphine (TCEP), pH 7.4) with or without 1 µM ReAsH (Invitrogen), which was added from a freshly prepared 2 µM working stock in TNBT. The protein was incubated for 30 min at room temperature and filtered through a nitrocellulose membrane (0.2 µm pore size) using an SF microfiltration unit (Bio-Rad; filtration takes less than 2 minutes). The sample slots were washed twice by filtration with 100 µl of TNBT buffer and immediately scanned for Cerulean fluorescence (excitation 488 nm, emission 528 nm) and ReAsH fluorescence (excitation 633 nm, emission 670 nm) with a Typhoon scanner (GE Biosciences).
Aliquots (200 µl) of TEV-cleaved fusion protein (2 µM) in TNBT were thawed, incubated under mineral oil at 37 °C for various times, and divided in half. One half was centrifuged at 18,000 g for 30 min at room temperature, and the supernatant (50 µl) was removed and added to a black-bottomed fluorescence microplate containing 100 µL of TNIT (50 mM Tris, 150 mM NaCl, 12.5 mM imidazole, and 1 mM TCEP). A portion of the uncentrifuged half (50 µl) was added to the microplate as a control for total protein. The samples were analyzed for Cerulean fluorescence intensity (excitation 434 nm, emission 475 nm, bandwidth 5 nm) with a fluorescence plate reader (Varioskan, Thermo Electron Corporation).
For ReAsH/FlAsH reactivity, samples were prepared in the same fashion as the noncentrifuged protein except that 150 µL TNIT buffer containing 1.33 µM ReAsH or FlAsH was added to the microplate, to provide a final biarsenical dye concentration of 1 µM when the 50 µL protein was added. After a 20-min incubation at room temperature, ReAsH fluorescence (excitation 593 nm, emission 650 nm, bandwidth 5 nm) or FlAsH fluorescence (excitation 519 nm, emission 540 nm, bandwidth 5 nm) was measured with a fluorescence plate reader (Varioskan).
Protein samples (2.5 µM) were negatively stained with 1% (w/v) uranyl acetate. Images were collected on a TF30 electron microscope (FEI) fitted with a US1000 2k×2k CCD camera (Gatan).
Neuro-2a cells were maintained in OptiMem (Invitrogen) supplemented with 10% fetal calf serum, 1 mM glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin in a humidified incubator with 5% atmospheric CO2.
For imaging, 8 × 104 cells were plated into a tissue-culture-treated eight-well µ-slide (Ibidi). The next day, cells were transfected with the pcDNA3.2 vectors with Lipofectamine 2000 according to the manufacturers instructions (Invitrogen). The following day, the medium was replaced, and the cells were cultured for 24 h, rinsed twice with 300 µl of pre-warmed HBSS buffer (Invitrogen), and labelled for 30 min with pre-warmed HBSS containing 1 µM ReAsH or FlAsH and 10 µM EDT at 37 °C. The cells were washed in pre-warmed HBSS containing 250 µM BAL for 15 min, and the solution was replaced with pre-warmed HBSS.
The cells were imaged live on a TCS SP2 Leica confocal microscope with an HCX PL APO CS 40.0×1.25 or HCX PL APO lbd.BL 63×1.40 objective, 1 Airy pinhole and at a scan rate of 200 Hz with 4 line averages. Cerulean fluorescence was collected at an excitation wavelength of 458 nm and emission wavelengths of 470–520 nm. GFP fluorescence was measured with an excitation laser of 488 nm and emission wavelengths of 500–538 nm. Cherry fluorescence was measured with an excitation laser of 543 nm and emission wavelengths of 575–640 nm. ReAsH fluorescence was collected at an excitation wavelength of 543 nm and emission wavelengths of 580–640 nm. FlAsH fluorescence was collected at an excitation wavelength of 488 nm and emission wavelengths of 500–540 nm. The PMT voltage was kept constant within each experiment and varied between some experiments to optimize signal/noise. Data were obtained in 12-bit format and analysed with ImageJ. Masks for filtering Cerulean fluorescence were created with a Cerulean threshold range from 3 times the background fluorescence intensity up to, and excluding, the maximum (i.e., saturated) pixel values. For Cherry, the mask threshold range was from twice the background fluorescence up to, and excluding, the maximum pixel value. Data were corrected for background fluorescence before the ReAsH/Cerulean ratio images were calculated. For the pixel correlation plots, images were converted to 8-bit after data rescaling (to the same scale for multiple images within each experiment) and analysed with the Image Correlator plugin.
We thank Ken Goldie for performing the electron microscopy experiments. This work was funded by grants to AFH (National Health and Medical Research Council program grant) and DMH (NHMRC project grant 566640). DMH is a CR Roper Fellow. SF was supported by the National Institutes of Neurological Disease and Stroke (R01 2NS039074 and R01 2NS045191), the CHDI Foundation, and the Taube-Koret Center for Huntington’s Disease Research. JM is supported by a predoctoral fellowship from the Hillbloom Foundation as well as the NIH-NIGMS UCSF Medical Scientist Training Program.