We are searching data for your request:
Upon completion, a link will appear to access the found materials.
We want study if 2 proteins A and B are co-located, for that we use 2 FTP(Fluorescent tag proteins) for each protein?and after the expirement these 2 FTP are co-located. Does that mean necessarily protein A and B are co-located?
"Are protein A and B co-located?" The answer is that it depends. Since your test question included the word "necessarily", this means that the answer to the test question is no. All you have to do is think of at least one plausible counter-example.
Here's a couple:
The flourescently tagged constructs were screwed up somehow, so A, B, and each of their respective tags were actually expressed as 4 separate proteins.
(As per BPinto's comment) the tagged constructs may co-localize even if A and B would not normally do so. This can happen if, for example, the tags bind to each other, or if the tags somehow cause both constructs to get transported to a particular organelle.
Fluorescent protein tags
Fluorescent proteins (FPs) have been used as protein tags since the mid-1990s mainly for cell biology and fluorescence microscopy. These tags have not only revolutionized cell biology by enabling the imaging of almost any protein, they are also used in biochemical applications. An important example is the immunoprecipitation and affinity purification of FP-tagged proteins, which was enabled by the development of affinity resins with high yield, purity, and affinity such as ChromoTek’s Nano-Traps ( https://www.chromotek.com/products/detail/product-detail/nano-traps/ ).
In this blog we provide a review of
Types of fluorescent proteins
Most researchers use intrinsically fluorescent proteins GFP, mNeonGreen, TurboGFP, RFP, or mCherry. Alternatively, extrinsically fluorescent or self-labelling proteins have been introduced that require the covalent coupling of a fluorescent molecule to the non-fluorescent protein, e.g. the protein tags SNAP, CLIP, and Halo. These self-labelling fluorescent proteins have certain performance advantages over intrinsically FPs due to their fluorescent dyes’ properties.
Figure 1: Structures of fluorescent proteins.
(A) Intrinsically fluorescent proteins (FPs) such as EGFP, GFP, RFP, mNeonGreen, turboGFP etc. share only a small number of common residues, but fold all into a conserved β-barrel structure. Their fluorescence arises through the backbone cyclization and oxidation of three amino acid residues in the center of this barrel (highlighted on the right), which results in a two-ring chromophore. This chemical process is described as maturation of the chromophore, is inherent to the protein fold, and depends only on environmental variables such as temperature and oxygen concentration, but not on additional enzymes. The colour, photostability, quantum yield, and other spectral properties of intrinsically fluorescent proteins are a result of mutations within the amino acids that make up the chromophore or that are located in the vicinity of the chromophore.
(B) Extrinsically fluorescent proteins such as HaloTag are non-fluorescent in their basal apo-form. Only if a suitable activated fluorophore is added to the HaloTag protein, this fluorophore will be captured and covalently bound by the HaloTag residue D106, turning HaloTag fluorescent. (PDB IDs for structures: EGFP, 2y0g apo-HaloTag, 5uy1 holo-HaloTag, 5uxz.)
Jellyfish Green Fluorescent Protein (GFP) and its derivatives are still the most frequently used fluorescent proteins in biomedical research. Recently, additional green fluorescent proteins have been introduced that are derived from other organisms. These FPs own the same basic fold as GFP but diverge widely on sequence-level. Therefore, they require novel, dedicated research tools such as antibodies.
The original: GFP
Green Fluorescent Protein was first isolated from the jellyfish Aequorea victoria in 1962 by Osamu Shimomura. It has a long Stokes shift green fluorescence (ex 395nm em 509 nm). 30 years later, Douglas Prasher eventually managed to clone the sequence of GFP and Martin Chalfie expressed this sequence in vivo. Later, the Roger Tsien lab developed GFP into a suite of veritable research tools. Shimomura, Chalfie, and Tsien were awarded the Nobel Prize in 2008. See Roger Tsien’s Nobel Prize lecture here: https://www.nobelprize.org/prizes/chemistry/2008/tsien/lecture/ .
Scientists developed a plethora of GFP variants with varying properties. These FPs have different functional and spectral properties. The first significant improvement of GFP was a mutation (S65T) that increased the intensity and stability of the fluorescence signal. The main excitation peak has been shifted to 488 nm (Heim et al., 1995). The common variant EGFP is an engineered version of GFP, which facilitates the practical use of GFP in a variety of different organisms and cells.
Green Fluorescent Protein GFP is also known as avGFP, wtGFP, and gfp10 EGFP as enhanced GFP and GFPmut1.
Reported in 2004, TurboGFP is a fast maturing and bright dimeric green fluorescent protein derived from CopGFP from the copepod Pontellina plumata. Copepod TurboGFP is evolutionarily distant from jellyfish-derived fluorescent proteins such as EGFP and shares only about 20% sequence identity with the commonly used GFP variants. Therefore, most anti-GFP antibodies including the GFP-Nanobody used in GFP-Trap do not bind to TurboGFP.
mNeonGreen is derived from a multimeric yellow fluorescence protein of the lancelet Branchiostoma lanceolatum. Hence, mNeonGreen is evolutionarily distant from jellyfish-derived FPs. mNeonGreen and common GFP derivatives share just about 20% sequence identity. Due to the low sequence similarity, it is expected that affinity tools (i.e. antibodies) for GFP variants will not bind to mNeonGreen. Certainly, this has been shown for ChromoTek’s affinity reagents (anti-GFP Nanobodies/ VHHs and antibodies).
First published in 2013, mNeonGreen is up to three times brighter than GFP. It is an emerging, monomeric versatile green/yellow fluorescent protein for imaging applications including super- resolution microscopy. Furthermore, mNeonGreen acts as acceptor for cyan fluorescent proteins in fluorescence resonance energy transfer (FRET) applications. It seems to be the brightest monomeric green fluorescent protein known so far and has a fast maturation rate.
Protein functional activities correspond with their subcellular expression and molecular complexing interactions. Localization can be effectively demonstrated with fluorescence microscopy based techniques or fractionation procedures. A broad spectrum of fluorescence imaging can be accomplished by using recombinant reporter proteins, (i.e. GFP, SNAP-tag ® ), or fluorescent dyes (i.e. Alexa), or fluorophore labeled molecules (i.e. protein specific antibodies). Cells can be genetically modified to overexpress protein targets, or regulatory components of non-coding sequences, for purposes such as determining fundamental cellular processes, disease mechanisms, gene therapy, and response to therapies. Fusion protein tags can be detected by antibodies or have functional properties to enable localization. Bioluminescent protein and fluorescent protein labeling systems are such genetically engineered optical imaging tools, with greater specificity at lower concentrations than other methods. With this approach, imaging can be in fixed or living cells, tissues or animals. A very attractive feature of the labeling of fusion proteins is that the labeling itself can be restricted to certain locations of a cell (i.e. SNAP-Cell ® or SNAP-Surface ® ). Such discrimination cannot be easily achieved when using bioluminescent proteins. Some fluorescent protein labeling systems enable small nonfluorescent molecules to become fluorescent when bound to a small genetically inserted peptide sequence in the target protein of interest (i.e. FlAsH). Advances in genetically engineered fluorescence systems and microscopy optic machinery, have made imaging a core method for protein localization.
SNAP-tag ® , SNAP-Cell ® and SNAP-Surface ® are registered trademarks of New England Biolabs, Inc.
- Application Notes
- Tools & Resources
- Features of SNAP-tag/CLIP-tag
- Applications of SNAP-tag/CLIP-tag
- Advantages of ACP/MCP-tag
- Protein Labeling with SNAP-tag and CLIP-tag
- Protein Labeling with ACP-tag
- SNAP-tag®, CLIP-tag&trade and ACP/MCP-tag Substrate Selection Chart
- Legal Information
Read about the NEB’s set of protein tools for the specific labeling (SNAP-, CLIP-, ACP- and MCP-tags) of fusion proteins.
- Clone and express once, then use with a variety of substrates
- Non-toxic to living cells
- Wide selection of fluorescent substrates
- Highly specific covalent labeling
- Simultaneous dual labeling
- Simultaneous dual protein labeling inside live cells
- Protein localization and translocation
- Pulse-chase experiments
- Receptor internalization studies
- Selective cell surface labeling
- Protein pull-down assays
- Protein detection in SDS-PAGE
- Flow cytometry
- High throughput binding assays in microtiter plates
- Biosensor interaction experiments
- FRET-based binding assays
- Single molecule labeling
- Super-resolution microscopy
- Small &ndash Expressed tag is only 8 kDa (77aa)
- Versatile &ndash ACP- and MCP-tagged fusions can be co-expressed and sequentially labeled for two color applications on cell surfaces
- Specific &ndash Components remain exclusively extracellular, preventing intracellular labeling
- Precise &ndash Label is covalently bound under biological conditions in a defined position
- Non-toxic &ndash Substrates are non-toxic to living cells
- Selection &ndash Choice of substrates available, including 488, 547, 647 nm and biotin
This product is covered by one or more patents, trademarks and/or copyrights owned or controlled by New England Biolabs, Inc (NEB).
While NEB develops and validates its products for various applications, the use of this product may require the buyer to obtain additional third party intellectual property rights for certain applications.
For more information about commercial rights, please contact NEB's Global Business Development team at [email protected]
This product is intended for research purposes only. This product is not intended to be used for therapeutic or diagnostic purposes in humans or animals.
SNAP-TAG ® Overview Tutorial
View an interactive tutorial explaining the mechanism of our SNAP-tag® technologies and reagents available for researchers wishing to study the function and localization of proteins in live or fixed cells.
Fluorescent Labeling of COS-7 Expressing SNAP-TAG ® Fusion Proteins for Live Cell Imaging
Watch as Chris Provost, of New England Biolabs, performs fluorescent imaging of live COS-7 cells expressing SNAP-tag ® fusion proteins.
Results and discussion
Rationally designed variants of split BFP and CFP
To expand our color palette of split FPs, we split EBFP2 at the same site as GFP1-10/11 (note that EBFP2 is 4-fold brighter and >500-fold more photostable than EBFP 16 ). While the short fragment is identical to the amino acid sequence of GFP11, six substitutions have been introduced into the large fragment through site-directed mutagenesis (N40I/T106K/E112V/K166T/I167V/S206T the numbering of amino acids follows that of EBFP2). These substitutions have been previously shown to enhance complementation efficiency of GFP1-10 variants 1 . To verify in vivo complementation between the two fragments, we used GFP11-tagged β-actin and histone 2B. Co-expressing each one with EBFP21-10 in HeLa cells, we observed blue fluorescence in images of the actin stress fibers and the nucleoplasm (Fig. 1a, b). In some cases (e.g., the actin image), autofluorescence limits the usefulness of this split construct because its overall fluorescent signal is extremely weak. In fact, high autofluorescence background with UV light is often observed in the perinuclear region (Supplementary Fig. 2). To improve its overall brightness, we decided to add six more substitutions to EBFP21-10 (S65T/Q80R/F99S/V128T/M153T/V163A some of these have previously been characterized to promote the stability and folding rate of GFP 1,17 ). This new split FP, termed split Capri for its cyan–blue color, has the same absorption spectrum as split EBFP2 (Supplementary Fig. 3). The emission spectrum, however, is red-shifted from split EBFP2 by 20 nm (λabs/λem = 379/469 nm). Furthermore, its peak extinction coefficient of 37,300 M −1 cm −1 and quantum yield of 0.13, are greater than those of split EBFP2 (Supplementary Table 1). When associated with GFP11-tagged β-actin or H2B, Capri1–10 exhibits very bright fluorescence in HeLa cells (Fig. 1c, d). To assess the improvement in the resulting brightness, we co-expressed GFP11-H2B in HEK 293T cells with either EBFP21-10 or Capri1-10. Quantifying the fluorescence intensity of cells by flow cytometry, we found that Capri1-10/11 had a four-fold brighter fluorescence than EBFP21-10/11 (Supplementary Fig. 4).
a–f Confocal images of HeLa cells. Cells co-expressing EBFP21-10 (a, b), Capri1-10 (c, d), and Cerulean1-10 (e, f) with GFP11-tagged β-actin or histone 2B.
In addition to BFP variants, cyan-colored FPs have been widely studied. When we introduced substitutions into GFP1-10 (Y66W to make CFP1-10), complementation fluorescence was observed for GFP11-β-actin or GFP11-H2B fusions co-expressing with CFP1-10 in HeLa cells (Supplementary Fig. 5). However, it is noticeable in the figure that the overall brightness of CFP1-10 is relatively weak, making it difficult to visualize thin actin filaments (A recent in vitro assessment also reported that split CFP has a low brightness 10 ). Therefore, we sought to produce a cyan-colored FP that has enhanced brightness. A recent improvement of full-length CFP, named Cerulean, increases the brightness by
1.6 times 18 . Because the known substitutions are located only on GFP1-10 (Y66W/S72A/Y145A/H148D for Cerulean), Cerulean1-10 can associate with GFP11. To evaluate the enhancement in its overall brightness for cellular microscopy, we prepared plasmids encoding Cerulean1-10 or CFP1-10. We co-transfected each one of these plasmids with a GFP11-H2B plasmid in HEK 293T cells. Imaging by confocal microscopy, we quantified the signal level of these split FPs. We found that Cerulean1-10 signal was
1.7 times brighter than that of CFP1-10 (p < 0.0001, Student’s t-test Supplementary Fig. 6). We next assessed the performance of Cerulean1-10 when used as a fusion tag. We used GFP11-fused β-actin or H2B and co-expressed each one with Cerulean1-10 in HeLa cells (Fig. 1e, f). Although we observed complementation of Cerulean1-10 in the appropriate locations, some cells exhibited thicker actin bundles, which we have never seen in cells expressing a full-length Cerulean fusion (Fig. 1e and Supplementary Fig. 7). Because this artifact is common for dimeric or tetrameric FPs when they are targeted to two-dimensional structures 19,20 , we suspect that Cerulean1-10/11 is an oligomeric split FP. Nonetheless, an in-depth investigation is required to validate such a property in split Cerulean and under more various experimental conditions.
Engineering of a red-colored split FP variant based on mRuby3
Although developmental efforts are ongoing to improve the brightness of split sfCherry 2,13,14 , having spectrally distinct split red FPs would foster the gross usefulness of FP11-tags. Since split sfCherry2 has a far-red shifted emission peak at 610 nm, we sought to explore the evolution of orange-red FPs such as mKO2, mRuby3, mApple, and mScarlet-I 15,21,22,23 in E. coli. Following the previously established approach 13 , we inserted a 30 amino-acid spacer between the 10th and 11th β-strand of the four FPs. The long spacer insertion greatly diminished colony fluorescence of mKO2, mApple, and mScarlet-I, while colonies expressing spacer-inserted mRuby3 remained fluorescent (Supplementary Fig. 8). To improve the brightness of the spacer-inserted mRuby3, we mutagenized it using error-prone PCR and then transformed into E. coli the three brightest candidates were pooled and subjected to another round. After three rounds, brightness of the best candidate revamped six-fold relative to that of spacer-inserted mRuby3 (Supplementary Fig. 9). We found seven substitutions in mRuby31-10 (M15T/Q27H/T31I/V106I/S113C/R126S/A154V) and termed this variant split mRuby4 (λabs/λem = 557/592 nm see also Supplementary Fig. 3 for its absorbance spectrum). Compared to split mRuby3, we created a particularly bright variant that has a higher extinction coefficient and increased quantum yield (Supplementary Table 1).
To assess whether split mRuby4 could fluoresce in human cells, we over-expressed mRuby411-β-actin with mRuby41-10 in HeLa cells. We observed that complemented split mRuby4 has a bright signal in fluorescent images of actin and various fusion proteins (Fig. 2a–f). To determine the signal level of split mRuby4, we performed a cellular fluorescence measurement by flow cytometry and compared the signal to full-length mRuby3. With expression of spacer-inserted mRuby4 in HEK 239T cells, we found that its signal level became around 69% of full-length mRuby3 (Supplementary Fig. 10). We have also demonstrated that mRuby41-10/11 has sufficient efficiency to detect proteins expressed at endogenous levels. We employed CRISPR/Cas9-mediated HDR and introduced a 200-nucleotide ssDNA donor into the HIST2H2BE locus of HEK 293FT cells expressing mRuby41-10. Subsequently, we found that split mRuby4 complementation had a prominent signal in images of the mRuby411 knock-in (Supplementary Fig. 11).
a–f Cells co-expressing mRuby41-10 with mRuby411-fused cellular proteins. For each, the name of the fusion partner and its normal subcellular location are indicated, respectively β-actin, actin stress fibers (a) Zyxin, focal adhesion (b) histone 2B, nuclei (c), Clathrin light chain, clathrin-coated pits (d) Keratin, intermediate filaments (e) Lamin A/C, nuclear envelops (f). g Normalized fluorescence emission spectra of FP1-10/11 variants in HeLa cells. (h) HEK 293 cells expressing H2B labeled with EBFP21-10/11, Capri1-10/11, Cerulean1-10/11, GFP1-10/11, mRuby41-10/11, or sfCherry21-10/11 were co-cultured in the same plate. Spectrally unmixed images at the different stages of mitosis are represented (see also Supplementary Fig. 14). Scale bars,10 μm.
As shown in Fig. 2g, the emission peaks for split mRuby4 and split sfCherry2 are only 20 nm apart yet still visually distinguishable (Supplementary Fig. 12). To further evaluate how many split FPs could be simultaneously visualized in different cells, we performed spectral imaging of HEK 293 cells expressing H2B fused proteins (Fig. 2g). We co-cultured six types of HEK 293 cells, each of which expressed H2B labeled with either EBFP21-10/11, Capri1-10/11, Cerulean1-10/11, GFP1-10/11, mRuby41-10/11, or sfCherry21-10/11. After we synchronized at the G2/M phase by release from a cyclin-dependent kinase inhibitor, we imaged the cells (Supplementary Fig. 13). Within a couple of hours, 20% of the cell population was in cytokinesis, which is consistent with previous literature 24 . We then captured cells with each split FP fusion at different stages of mitosis (Fig. 2h). Overall, these experiments illustrate six-color spectral imaging of cellular proteins.
Evaluation of split FP1-10/11 systems for multiplexed imaging in single cells
An orthogonal interaction between GFP1-10/11 and sfCherry21-10/11—meaning GFP11 can interact with GFP1-10, but cannot interact with sfCherry21-10—is the basis of simultaneous labeling of two different fusion proteins in single cells. With an array of FP1-10/11 pairs developed, we sought to systematically test their binding specificities by flow cytometry. GFP1-10/11, sfCherry21-10/11, mNeonGreen21-10/11 13 , and mRuby41-10/11 were examined for complementation in HEK 293T cells. Each FP1-10 fragment was co-expressed with any one of four FP11-fused β-actin, and the interactions were tested along the grid diagonal (Fig. 3a). As shown in Fig. 3a, all four FP1-10 fragments reconstituted with their corresponding partners. Interestingly, mRuby41-10 and sfCherry211 formed complementation signal as did mRuby411 and sfCherry21-10. Because the FP1-10/11 fragments encoded by closely related FPs, we expected there to be some crosstalk (Fig. 3b). We used HeLa cells co-expressing GFP1-10/11-β-actin with Zyxin-mRuby41-10/11, or mNeonGreen21-10/11-β-actin with mRuby41-10/11-Clathrin to verify dual-color labeling with mRuby411 in single cells. We found that the two distinct fluorescence channels did not overlap in the cells (Fig. 3c, d).
a Characterizing the binding specificities of GFP1-10/11, sfCherry21-10/11, mNeonGreen21-10/11, and mRuby41-10/11 by flow cytometry (see also Supplementary Fig. 15). Each of the FP11 fragments was tested for complementation to all of the FP1-10 fragments. Complementation is indicated as the percentage of fluorescent cells by a color scale and the number in each block. b, c Dual-color fluorescence images of HeLa cells expressing GFP1-10/11-β-actin and Zyxin-mRuby41-10/11 (b), and mNeonGreen21-10/11-β-actin and mRuby41-10/11-Clathrin (c). d This dendrogram is based on the similarities of the following fluorescence protein sequences: EBFP2, Capri, Cerulean, CFP, GFP, mNeonGreen2, mRuby4, and sfCherry2. Proteins that share sequences are separated by smaller branch lengths. Scale bar, 20% dissimilarity. The dendrogram was constructed using MEGA 7.0 software. Scale bars, 10 μm.
A strategy to create new orthogonal split FPs using circularly permutated FP fragments
In order to provide more variants of split FPs orthogonal to existing FP1-10/11, we took advantage of circularly permutated GFP variants 25 . By linking the N-termini and C-termini and cutting out a single β-strand, any one of the eleven β-strands could be a split GFP-tag. We chose to measure complemented signal of the β-strands 7, 8, 9, and 10. (The β-strands 1–6 were excluded because the complementary fragments of these strands are unlikely to be water-soluble 26 ). To this end, we prepared DNA constructs encoding each of the β-strands fused to β-actin and measured the overall complemented signal of each construct in HEK 293T cells by flow cytometry. We observed fluorescence signal reconstituted from the β-strands 7, 8, and 10 (hereafter named GFP8-6/7, GFP9-7/8, and GFP11-9/10) with their corresponding partners (Fig. 4a). These split GFPs retained 7–57% brightness of GFP1-10/11, albeit leaving room for improvement. To validate protein labeling using the β-strands, we generated constructs encoding various cellular proteins fused with GFP8 and co-expressed each one of them with GFP9-7 in HeLa cells. For three proteins tested, we observed their expected localizations (Fig. 4b–d).
a Fluorescence intensity of HEK 293T cells expressing actin labeled with circularly permutated split GFP variants, measured by flow cytometry (see also Supplementary Fig. 16). n = 698 cells for GFP8-6/7 n = 7792 for GFP9-7/8 n = 274 for GFP11-9/10 n = 11017 for GFP1-10/11. Error bars are SEM. b–d Confocal images of HeLa cells co-expressing GFP9-7 with GFP8 fusions β-actin (b), Clathrin light chain (c), and β-tubulin (d). e The binding specificities of GFP8-6/7, GFP9-7/8, GFP11-9/10, and GFP1-10/11 were characterized by flow cytometry (see also Supplementary Fig. 17). f Dual-color fluorescence images of a U2OS cell expressing two different fusions, GFP11-H2B and GFP8-Lamin A/C. g Four-color images of a U2OS cell co-expressing GFP11-H2B, GFP8-LaminA/C, mNeonGreen211-β-actin, and mRuby411-Zyxin with GFP1-10, GFP9-7, mNeonGreen21-10, and mRuby41-10. Scale bars, 10 μm.
Next, we assessed the binding specificities of GFP8-6/7, GFP9-7/8, GFP11-9/10, and GFP1-10/11. We performed the flow cytometry assay conducted in a grid format as described earlier. Either GFP8-6, GFP9-7, GFP11-9, or GFP1-10 was co-expressed with β-actin fused with the β-strands 7, 8, 9, or 11 in HEK 293T cells. In this experiment, each of the β-strands only binds to its corresponding partner (Fig. 4e). For instance, GFP8 interacts with GFP9-7, but not GFP1-10. This orthogonal interaction was validated by dual-color imaging of U2OS cells, in which GFP11-H2B and GFP8-Lamin A/C were co-expressed with Capri1-10 and GFP9-7. We observed the exclusion of GFP8-Lamin A/C from the nucleoplasm where GFP11-H2B predominately localized (Fig. 4f). Taken together, circularly permutated FP fragments can be used to generate additional orthogonal pairs for multiplexed split FP-labeling.
Multicolor images reveal nuclear localization of Zyxin
Finally, we assessed the potential of split FP systems for multiplexed labeling of proteins in single cells. As a proof-of-principle, we used four orthogonal split FPs that we thoroughly investigated in this report (Capri1-10/11, GFP9-7/8, mNeonGreen21-10/11, and mRuby41-10/11), which are distinguishable from each other by using spectral imaging methodology (Fig. 4g). U2OS cells were transfected to express these split FPs targeted to four distinct proteins (H2B, Lamin A/C, β-actin, and Zyxin), and we observed their correct localization. Strikingly, a few cells displayed some portion of Zyxin proteins localized to the nucleus, although the proteins predominantly localized at focal adhesions in these cells. By visual inspection of a total of 145 cells, we found that 37 of these cells exhibited nuclear localization of Zyxin during interphase (Supplementary Fig. 18). Because Zyxin is a relatively large molecule (>80 kDa) but does not have a nuclear localization signal, Zyxin must enter the nucleus in contact with other components. We observed a similar nuclear localization pattern of Zyxin tagged with a full-length FP tag in U2OS cells (50 out of 174 cells), and found that this observation has also been confirmed in other cell lines 27,28,29 .
For the initial demonstration of multiplexed labeling, split FPs were over-expressed as fusions to target proteins in single cells (Fig. 4g). However, these fusion proteins might be subject to limitations because of the potential for overexpression artifacts (e.g., aberrant organelle and/or cellular morphology). To further verify our observation in the future, this approach will be extended to label endogenous proteins by methods such as CRISPR/Cas9-mediated gene knock-in 12 . Because a split FP tag is
42–63 nucleotides long (which is
10 times smaller than the size of an intact FP), a short donor oligo can be directly synthesized, making this a cloning free approach (see also Supplementary Fig. 11) 2,12 . In addition, a small tag such as GFP11-tag can be introduced into a host cell genome at high homologous recombination efficiencies 11 . Such a simple and efficient approach would facilitate the generation of multiple insertions in single cells. Thus, the sequences for multiple split FP tags could then be inserted either sequentially or simultaneously into targeted loci in individual cells stably expressing the complementary fragments, enabling multiplexed visualization of endogenous proteins.
Overview of EzColocalization workflow
The workflow for EzColocalization is divided into four modules each with its own tab on the GUI. The tabs are: (i) “Inputs” where images, masks or regions of interest (ROI) lists are selected and aligned (ii) “Cell Filters” where cells can be selected based on physical features and signal intensity (iii) “Visualization” where heat maps, scatterplots, and metric matrices (defined below) are created and (iv) “Analysis” where the colocalization metrics and outputs are chosen. Not all modules and not all processes within a module have to be used. Some tabs have a “Preview” button to run a specific module instead of the “Analyze” button which runs all selected processes in all modules.
Image files, which are chosen in the “Inputs” tab (Fig. 1A), must be: (i) monochromatic (i.e. not RGB or CMYK formats) (ii) 8-bit, 16-bit, or 32-bit and (iii) in a format such as TIFF that retains the original pixel intensity values. Large images may be compressed for file transfer using a lossless format such as ZIP or LZW, and then decompressed for analyses. In addition to images, EzColocalization can accept masks and ROI lists for cell identification (see below). If there are multiple images for each channel, the images should be stacked for more efficient analysis in the “Stack” menu (see ImageJ guide for further details 24 ). Images in a stack may be different fields of view or a time series, but must have the same dimensions, magnification and image order for each channel. The input tab also provides options for setting thresholds for signal intensity and aligning misaligned images from different channels (Fig. 1B and Supplementary Information). Recommendations for acquiring suitable images for colocalization analysis are provided in the Supplementary Information. Note: alignment operates on the assumption that an appropriate threshold for signal intensity can be chosen to distinguish pixels inside and outside of cells if thresholding includes areas outside the cell or only a limited area within cells, then the alignment may not function properly. For this reason, all alignments should be checked by visually by examining the ROIs to confirm that appropriate cell areas are selected.
EzColocalization is primarily designed for one “cell identification” channel and two or three “reporter” channel images. However, it can operate with other input combinations (Table S1). The cell identification channel is used to identify individual cells, and consequently to distinguish intracellular and extracellular pixels. The cell identification channel can be any type of image that permits identification of the cell boundaries including: light microscopy images (e.g. phase contrast 25,26 and bright-field), images with a reporter that labels the cell membrane or that is throughout the cytoplasm (e.g. Cy5, Fig. 1B), and images with an extracellular dye that outlines cells. Differential interference contrast (DIC) images create shadows that make it difficult for automated selection of cells using threshold methods 27 therefore for DIC images we recommend that ROIs be created using the “selection tools” in ImageJ to manually outline cell areas, and then adding them to a list by choosing “Add to Manager” (in “Selection” submenu of the “Edit” menu). Once the ROIs for all cells of interest in an image are selected, a binary mask can be created using the “Clear Outside” and “Autothreshold” functions of ImageJ.
The “Cell Filters” tab is used to help select cells in images (Fig. 2A) and distinguish intracellular and extracellular pixels. Cells are identified by: (i) choosing one of the ImageJ threshold algorithms 24 , or manually selecting the thresholds (which is done by selecting “*Manual*” from a drop-down list in the Inputs tab and pressing the “Show threshold(s)” button), to identify regions corresponding to cells in the cell identification channel (Fig. 2B) (ii) using watershed segmentation to separate touching objects in the cell identification channel images (optional) (Fig. 2B) (iii) selecting objects from the cell identification channel images based on physical parameters (Fig. 2C) and signal intensity (Fig. 2D). EzColocalization will attempt to automatically detect whether input images have dark or light background using skewness. Assuming there are more pixels in the background than in the cells, an image with positive skewness indicates a dark background and negative skewness indicates a light background. Users can also manually select whether the input images have dark or light background in the “Parameters…” options of the “Settings” menu. Cells that are only partly within an image, and therefore could provide misleading values, are automatically removed from analyses.
EzColocalization has one optional “Pre-watershed filter” and eight optional post-watershed filters (with the option to select more). Watershed segmentation can aid the separation of dividing and touching cells 28 but it can also divide large objects such as aggregates of extracellular material into smaller fragments that are the same size as cells. To avoid the latter, the Pre-watershed filter can be used to exclude objects with large areas from the analysis. The Preview button in the Cell Filters tab allows users to see which objects on the current image will be selected when the minimum and maximum bounds of all the filters are adjusted. There are two classes of parameters for the post-watershed cell filters (Table S2): (i) physical parameters based on measurements from the cell identification channel and (ii) signal intensity parameters from the reporter channels. Physical parameters apply to all channels whereas signal intensity parameters apply only to the reporter channel for which they are selected (because reporters may have very different levels of signal). In addition to filtering based on predefined options in ImageJ, EzColocalization has filters for the “MeanBgndRatio” or “MedianBgndRatio”, which are calculated by dividing the mean or median signal intensity of pixels inside an object by the respective mean or median signal intensity of extracellular pixels.
The “Visualization” tab displays signals or metrics in cells as: (i) “heat maps” (ii) scatterplots and (iii) “metric matrices” (Fig. 3A).
Heat maps are pseudocolor images that show the relative magnitude of reporter signals (Fig. 3B). They are generated by normalization and rescaling so that the minimum and maximum pixel values are 0 and 255 respectively in each cell, image, or stack. There are eight options for coloring the heat maps, and the intensity values for each color are obtained from the “Show LUT” function (within the “Color” submenu of the “Image” menu in ImageJ). Cell heat maps are suited for determining where each reporter occurs with highest intensity in cells. Image heat maps can show if different cells within a field of view have substantially different intensities, which may indicate biological heterogeneity or unevenness in labeling. Stack heat maps can show if cells in different images have substantially different levels of signal intensity, which may indicate unevenness in labeling or measurements across a slide (e.g. due to photobleaching) or changes in signal over time (if the stack is a time series). Note: heat map appearances are affected by brightness and contrast settings.
Scatterplots show the relationship between the signal intensity for two or three reporter channels for individual cells and images (Fig. 3C). This relationship is important in choosing the appropriate colocalization metric (Supplementary Information). Scatterplots can also reveal heterogeneity in the localization patterns 8 , which may require removal of background pixels or separate analyses for different cell types.
Metric matrices provide an overview of localization patterns by showing the calculated values of a colocalization metric for many threshold combinations. Metric matrices for the threshold overlap score (TOS) have been shown to be useful for the analysis of localization patterns for two reporter channels 8,15 (Fig. 3D). For completeness, EzColocalization has the option to calculate metric matrices for two reporter channels using five other metrics: threshold overlap score with logarithmic scaling 8 , Pearson correlation coefficient (PCC), Manders’ colocalization coefficients (M1 and M2), Spearman’s rank correlation coefficient (SRCC), and intensity correlation quotient (ICQ) 8,15 . Colocalization for three channels can also be measured using ICQ, Manders’ colocalization coefficients and TOS 29 (Supplementary Information).
Thresholds for all metrics are measured as the top percentile (FT) of pixels for signal intensity 8,15 . For example, FT = 0.1 is the 10% of pixels with the highest signal. For the metric matrices, FT is also used to specify the step size for the threshold combinations. That is, FT = 0.1 also selects thresholds for the 10%, 20%, …, and 100% of pixels with the highest signal. If FT does not divide evenly into 100, then the remaining percent is the last step size. For metrics that do not need a threshold (i.e. PCC, SRCC, and ICQ) the values are calculated assuming that only the pixels above the thresholds exist. The metric matrix window has options for the results to be saved as text or image, for changing the FT or type of metric, viewing individual cell metric values as a list, and calculating the mean, median or mode of the metric for each threshold combination. The “Proc” (processed) and “Raw” button determines whether the list of data displayed, copied, or saved with the “List”, “Copy”, or “Save…” buttons respectively is the average value for the sample for each threshold combination (e.g. median value) or all values for each cell in the sample for all threshold combinations.
The “Analysis” tab has three subtabs (“Analysis Metrics”, “Metrics Info” and “Custom”). The Analysis Metrics subtab has six metrics for measuring colocalization for two reporters (Fig. 4A) and three metrics for three reporters (see previous section). Users may choose a threshold or no threshold for PCC, SRCC and ICQ. TOS and Manders’ colocalization coefficients must have a threshold to be calculated. The Metrics Info subtab contains information and resources about the metrics used in the Analysis Metrics subtab (more details in the Supplementary Information). Thresholds can be selected using Costes’ method 30 or manually. In the Custom subtab (see Supplementary Information for additional information), users can write their own code in Java TM to analyze images (note: the example provided is for calculating PCC) (Fig. 4B). The “Compile” button tests the code and creates a temporary file in the Java temporary directory and displays the outcome of the compiling with a “Succeeded” or “Failed” label. If successful, the compiled custom code is read to the memory again and applied to the selected cells.
The output of every analysis is a table that specifies the image and an identifier number for every cell (Fig. 4C), and for each cell, values are provided for: (i) the selected metric (ii) physical parameters and (iii) average signal intensity for each channel (if selected). Note: “NaN” in the output table indicates the failure to calculate a value. Users can also generate summary windows (with the cell number, mean, median and standard deviation for the selected metric) (Fig. 4D), histograms of metric values (Fig. 4E), binary mask images, and a list of ROIs that represent each cell’s position and number on each image in the ROI manager. ROI lists and binary mask images can be saved for re-analysis of the same cells.
Applications of EzColocalization
EzColocalization is designed to be used in a modular manner to facilitate customization of analyses for a wide variety of experiments and researcher needs. This section focuses on demonstrating specific tools in EzColocalization to solve real-world problems in diverse image sets.
In the first application of EzColocalization, images of rat hippocampal neurons from the Cell Image Library (CIL:8773, 8775–8788, which are attributed to Dieter Brandner and Ginger Withers) are used to demonstrate: (i) the use of a reporter channel for cell identification when an experiment does not have separate non-reporter images for cell identification (ii) cell filters for selecting cells and (iii) visualization tools for choosing metrics. The workflow of the analysis is outlined in Fig. 5A. In the first step, two reporter image stacks were created: one stack with images where F-actin is labelled (using a phalloidin peptide conjugated to rhodamine) and the second stack with images where tubulin is labelled (using an antibody conjugated to Alexa 488) (Fig. 5B). The interaction of F-actin and tubulin is important for the growth and migration of neurons 31,32 . We used the F-actin images for cell identification because it is present in all cells and it shows the cell boundaries 8 . Individual cells were selected from the F-actin images by applying a threshold to identify cells 24 and using a cell filter to remove cell debris (note: parameter values in Fig. 5A).
After the cells were selected, the intensity of reporter signals were examined using cellular heat maps and scatterplots. We found the reporters did not colocalize at high signal levels and there was a complex relationship between the signal intensities (Fig. 5C,D). Due to the latter, localization was quantified using Manders’ M1 and M2 and TOS (Supplementary Information). M1 and M2 were evaluated at thresholds selected by Costes’ method for the cell outlined in Fig. 5B, and the values were 0.289 and 0.995 respectively. These values are usually interpreted as indicating that tubulin has high colocalization with F-actin, and F-actin has low colocalization with tubulin. TOS values were evaluated by generating a metric matrix with median TOS values. The matrix showed colocalization, anticolocalization and noncolocalization at different thresholds for the signal intensities of tubulin and F-actin (Fig. 5E). At sites in cells where F-actin and tubulin have the highest intensity signal (top 10% of pixels for each channel), the median TOS value is −0.36 (n = 20). This negative value indicates anticolocalization, which is consistent with the impression obtained from the heat maps and scatterplots, and with other reports 8 .
In the second application, images of Saccharomyces cerevisiae undergoing mitosis were obtained from the Cell Image Library 33 to demonstrate: (i) cell identification via hand-drawn outlines (for experiments where automated methods of cell identification cannot be applied) and (ii) image alignment. The reporter inputs were an image from a wild type strain (“control” CIL: 13871) that has the BFA1 protein that loads TEM1 onto the spindle pole body, and an image from a strain without the BFA1 protein (∆bfa1 deletion mutant CIL: 13870). In these reporter images, cells expressed TEM1 protein fused to GFP and the DNA was labelled with DAPI (4′, 6-diamidino-2-phenylindole). TEM1 localizes to spindle pole bodies during mitosis and is implicated in triggering exit from mitosis 33 . The workflow is shown in Fig. 6A. In this application, ROIs were manually drawn around cells using the “Freehand” selection tool in ImageJ on DIC images. Binary masks, which were used to select cell areas, were created by selecting the ROIs and using the “Clear Outside” and then “Auto Threshold” functions of ImageJ 24 (Fig. 6B). The cell areas were used for cell identification and to correct alignment between the DIC images and the reporter channels using the “Default” threshold algorithm (Fig. 6C). Following this cell identification and image alignment, the images are now ready for visualization and analysis as described in the previous example.
In the third application, images of whole adult Caenorhabditis elegans obtained from the Broad Bioimage Benchmark Collection (BBBC012v1, M14) 34 were used to demonstrate that: (i) EzColocalization can analyze colocalization in whole organisms and (ii) “cell” filters can select individual organisms based on reporter signal intensity. The images in this example are from the same dataset used in our study describing TOS (but they are not the same images) 8 . The workflow is shown in Fig. 7A. Outlines of individual C. elegans were drawn in ImageJ on bright-field images to create ROIs, and the ROIs were added to the ROI manager for “cell” identification. GFP expressed from the clec-60 promoter in the anterior intestine was reporter 1 and mCherry expressed from the myo-2 promoter within the pharynx, which is an organ next to the anterior intestine 35 , was reporter 2. Cell filters for physical parameters were unnecessary because only those objects considered to be suitable C. elegans had outlines drawn around them in the first place. However, cell filters for signal intensity were necessary because some C. elegans had low GFP signal, possibly due to transgene silencing 36,37 (Fig. 7B). Subsequent visualization and analysis can be performed as described in the first application.
In the fourth application, we demonstrate the analysis of colocalization for three reporter channels. The workflow was the same as for two reporter channels except “3 reporter channels” was first selected in the “Settings” main menu (Fig. 8A). Images were obtained from the Broad Bioimage Benchmark Collection (BBBC025, Version 1, Image set: 37983, image: p23_s9) of U2OS bone cancer cells (n = 66) 38 . The three reporter images had DNA, endoplasmic reticulum (ER) and mitochondria respectively stained with Hoechst 33342, concanavalin A/Alexa Fluor488 conjugate, and MitoTracker Deep Red (upper row, Fig. 8B). Cell identification was performed with an image of the plasma membrane labeled with wheat germ agglutinin (WGA)/Alexa Fluor 555 conjugate (upper left, Fig. 8B). Note: the image also had the Golgi apparatus and F-actin network labeled 38 . The plasma membrane was traced using the polygon selection tool in ImageJ to create ROIs for the individual cells, and the ROI manager containing the ROIs was selected for cell identification.
The localization patterns were visualized in the same manner as for two reporters except that: (i) there are three sets of heat maps for the reporters instead of two (lower row, Fig. 8B) and (ii) scatterplots and metric matrices are in three dimensions (Fig. 8C–F). There is the option in the Visualization tab and the Analysis tab (Fig. 8G) to measure colocalization for the three reporters using ICQ, TOS or Manders’ M1, M2 and M3 metrics. Of the three metrics, we found that TOS was the easiest to interpret. TOS has a single value for measuring the colocalization of all three reporter signals, and it clearly showed the reporter signals for the nucleus, mitochondria and ER overlapped at low thresholds (i.e. at high FT values there is colocalization red color in Fig. 8E) and did not overlap at high thresholds (i.e. at low FT values there is anticolocalization blue color in Fig. 8E). These observations are consistent with the nucleus, mitochondria and ER organelles overlapping at their edges (where the signal from their reporters is typically lower) due to known physical interactions, but not at their centers (where the signal from their reporters is typically higher) because they are distinct structures in cells 39,40,41 .
The primary function of slam RNA is encoding Slam protein. In addition to coding information, slam RNA contains 2 more pieces of noncoding information: (1) information for specific subcellular localization of the RNA at the FC, which is at least partially mediated by an interaction with Slam protein, and (2) information for spatial and temporal control of translation, which is high at the FC during the onset of cellularization. By comparing wild-type RNA with a variant RNA, slam[ACU] with the same coding information, the relevance of the noncoding information was uncovered. slam[ACU] RNA is widely distributed in the cell and gives rise to much less Slam protein. Containing coding and noncoding information distinguishes slam RNA from generic mRNAs, which contain only coding information. slam RNA may be related to the growing class of mRNAs with coding and noncoding functions (cncRNA) .
Using the injection assay with synthetic RNA transcribed from truncation constructs, we identified several regions of slam RNA that are sufficient for localization to the FC in wild-type embryos. This includes the 5′ untranslated region and at least 2 large regions within the coding sequence, which we have so far not further defined. Each of these regions is able to localize to the FC on its own in wild-type embryos, which indicates redundancy in the mechanism of RNA localization. Interpretation of these data is difficult, however, given that endogenous slam RNA and protein were present in our assay, which may lead to localization by oligomerization or other indirect binding mechanisms.
slam RNA is subject to spatiotemporal control of translation. Although the RNA is present in high levels during the first half of cellularization, translation is restricted to the initial minutes of cellularization. The almost constant protein levels throughout cellularization are due to the stability of Slam protein, as inhibition of new synthesis by cycloheximide leads only to a weak decrease of GFP-slam fluorescence. In contrast to these constant levels during cellularization is the sharp increase in protein levels during the onset of cellularization.
This initial rise in protein levels is partly due to the transcriptional up-regulation of slam. The transcriptional regulation appears not to be sufficient, as we observed a striking difference between slam[wild-type] RNA with slam[ACU] RNA. Although both RNAs contain the same coding information and give rise to similar amounts of Slam protein in cultured cells, slam[wild-type] RNA is more efficiently translated than slam[ACU] RNA in blastoderm embryos. Based on the correlation of impaired RNA localization and reduced translational efficiency, we favor the model that efficient translation is linked to RNA localization or interaction with Slam protein at the FC. However, the embryo-specific lower efficiency of slam[ACU] translation may alternatively be due to secondary RNA structures or disadvantageous codons, which were introduced in our mutagenesis.
A particular feature of slam is that the encoded protein is required for the elaboration of noncoding features. slam RNA requires Slam protein for accumulation at the FC. Not only do we observe a functional interaction of the RNA and protein but also colocalization and biochemical association, indicating molecular interactions. These molecular interactions between RNA and protein may be direct or indirect. They are likely to be indirect, as Slam protein does not contain a canonical RNA binding domain or does not share any detectable sequence homology to RNA binding proteins. Analysis of transcripts associated with Slam protein provided distinct lists depending on the experimental procedure. Importantly, slam RNA was identified by both approaches, which is consistent with our quantitative polymerase chain reaction (qPCR) analysis for a few selected genes. The list of associated transcripts contained transcripts with high and low abundance, indicating that the procedure was sufficiently sensitive to also detect weakly expressed genes. Screening through the gene functions, we did not detect a specific set of genes, such as genes involved in cellularization or cytoskeletal organization.
The biological function of the intimate relationship of slam mRNA and its protein has been unclear. Given the observed specific spatiotemporal profile of slam translation and Slam dynamics, we propose the following model (Fig 8A). Initially, Slam protein accumulates in low levels at the FC independently of its mRNA. Starting with the onset in zygotic transcription, slam mRNA exits the nucleus as part of a complex that is not competent for efficient translation. At least a fraction of slam RNA molecules accumulates at the basal domain forming the FC prior to the first round of translation. At the basal domain, slam mRNA becomes competent for efficient translation, which leads to an increase in Slam protein at the FC. The increased amounts of Slam protein further promote accumulation of slam mRNA, leading to even more Slam protein. Such a mechanism constitutes a positive feedback loop, which provides an explanation for the observed switch-like profile of Slam protein staining at the FC (Fig 8B). Some minutes later, when full Slam levels have been reached, translation is turned off. Slam protein then functions in spatially restricted activation of Rho signaling, actomyosin organization, Patj recruitment, cortical compartmentalization, and furrow invagination (Fig 8) [10, 17, 25]. As slam is a key regulator of cellularization, the rapid increase of Slam protein may be important for a coordinated and timely onset of its downstream processes.
Where Can I Buy Fluorescent Probes?
Most major chemical suppliers sell specific fluorophores. Organic fluorescent dyes are the most commonly available types but now more unusual types like quantum dots are becoming available. Many suppliers produce protein labeling kits that contain everything needed to complete the process in a short time. Some offer the ability to label up to 10 mg of protein.
Fluorescent probes can be bought from most life science suppliers such as Anaspec, ThermoFisher, and LiCor.
Here are some examples of protein labeling kits found on the market:
AnaTag 5 microscale protein labeling kits. These labeling kits use fluorescein isothiocyanate (FITC) as a fluorophore. The kit is suitable for biological applications and can label 3 x 200 ug of protein.
Alexa Fluor protein labeling kits from ThermoFisher. These kits offer a very straightforward way to covalently label 1 – 10 mg of protein with a fluorescent dye. They offer a range of dyes with a broad range of excitation and emission wavelengths.
LI-COR IR Dye protein labeling kits. These kits are designed for labeling antibodies for use in flow cytometry where fluorophore-conjugated antibodies are required.
The fluorescent protein color palette
Because fluorescence is intrinsically a color-resolved technique, the most important consideration in choosing a FP is its spectral profile, that is, the color of its fluorescence. A broad range of FP variants that span nearly the entire visible spectrum has been developed and optimized (see poster panel ‘Excitation and emission spectral properties of the brightest FPs’). The poster highlights the spectral and imaging properties of a few widely used FPs from across the spectrum. The FPs were purified and their extinction coefficients, quantum yields and spectral properties measured as previously described (Patterson et al., 1997). To compare the brightness of various FPs, each normalized excitation spectrum was multiplied by its peak molar extinction coefficient, and then divided by the peak molar extinction coefficient of EGFP. Similarly, each normalized emission spectrum was multiplied by its molecular brightness (molar extinction coefficient × quantum yield) and then divided by the brightness of EGFP.
The choices for the current best-performing FPs in each color class are based on a number of crucial factors, including maturation efficiency, spectral properties, photostability, monomeric character, brightness, fidelity in fusions and potential efficiency as a Förster resonance energy transfer (FRET) donor or acceptor. There are several key mutations that are repeatedly used in different FPs to enhance function as assayed by these parameters (see ‘Critical mutations’ in the poster). On the basis of overall performance, we consider mTagBFP (Subach et al., 2009) and mTurquoise (Goedhart et al., 2010) the brightest and most photostable blue and cyan FPs, respectively. mEGFP was the first generally reliable FP yet, because of its combination of positive attributes, it remains the gold standard with which to compare the performance of all other FPs. In the yellow and orange spectral regions, mVenus (Nagai et al., 2002) and mKO2 (Sakaue-Sawano et al., 2008) are useful because they mature rapidly and are both bright monomeric variants, even though they lack the level of photostability exhibited by mEGFP. In the orange-red spectral region, mCherry (Shaner et al., 2004) is widely used for many applications, but has been reported to aggregate when expressed within some fusions (Katayama et al., 2008). mApple (Shaner et al., 2008) can be used as an effective substitute for mCherry in most proteins fusions (such as connexins, α-tubulin and focal adhesions). Use of mApple helps to reduce artifacts, but its emission is blue-shifted by
18 nm, which increases its spectral overlap with the yellow and orange FP variants. In the red to far-red region, mKate2 (Shcherbo et al., 2009) is currently the best choice in terms of brightness, photostability and performance in fusion proteins. The important photophysical properties of the selected FPs are summarized in the poster panel ‘Fluorescent protein properties’.
By using multi-color fluorescence microscopy, FPs are often used in combination to examine interactions between their fusion partners (see poster panel ‘Multi-color imaging with fluorescent protein fusions’). In the 2-color image shown here, pig kidney epithelial cells (the LLC-PK1 cell line) express mApple fused to human histone H2B, and mEGFP fused to human α-tubulin. The 3-color image shows HeLa cells that express mVenus fused to the SV40 T-antigen nuclear targeting signal to stain the nucleus, whereas mTurquoise and mCherry are fused to peptides targeted to the Golgi complex and mitochondria, respectively. In the 4-color panel, rabbit kidney (RK-13) cells are shown that express mCherry (fused to pyruvate dehydrogenase), mEGFP (fused to Lifeact), mTurquoise (fused to peroxisomal membrane protein), and mTagBFP (fused to H2B) to visualize the mitochondria, filamentous actin, peroxisomes, and nucleus. Finally, the 5-color assay combines the expression of mTagBFP, mTurquoise, mEGFP, mKO2 and mKate2 to label the nucleus, peroxisomes, endoplasmic reticulum, focal adhesions and mitochondria, respectively.
Choosing compatible fluorescent proteins
To choose a set of fluorescent proteins to be imaged together, you will need to consider the same factors as when choosing an individual fluorescent protein (brightness, photostability, and so on see the previous blog post for more discussion of these factors). In addition, you will also need to choose fluorescent proteins that can be distinguished from one another and that can be imaged with the optics on the microscope(s) you intend to use. An accurate determination of whether two fluorescent proteins can be separated from each other requires knowledge of their excitation and emission spectra, but a good rule of thumb is that both the peak excitation wavelengths and peak emission wavelength of the two proteins should be separated by 50-60 nm. For example, CFP (ex 430 nm / em 474 nm) and YFP (ex 514 nm / em 527 nm) can be imaged together but CFP and GFP (ex 488 nm / em 507 nm) show some crosstalk between the two fluorescent proteins. If you must image fluorescent proteins whose spectra overlap, there are techniques, like spectral unmixing, which can be used to separate the fluorescent proteins, but these are beyond the scope of this post.
The Future of SNAP- and CLIP-Tag protein?
The aforementioned examples from the literature demonstrate the potential of specific chemical labeling of fusion proteins, in particular the SNAP- and CLIP-tags, to address central questions in cell biology and protein science. Innovation in chemistry via the synthesis of new labeling substrates with advantageous properties will open up completely new ways to study protein function: the recently introduced optical switches for increased sensitivity in FRET experiments (18) or the use of environmentally sensitive dyes (19) are good examples. The simplicity of the synthesis of such substrates and the availability of the necessary building blocks from NEB permit even those scientists with little background in chemistry to assemble their own substrates. Whereas advances in the development of new intrinsically fluorescent proteins force their users to go through continuous cycles of subcloning and characterization, users of the SNAP-tag and its relatives can be assured to directly benefit from such future inventions.