Imaging Proteins Sensitive to Direct Fusions Using Transient Peptide–Peptide Interactions

Fluorescence microscopy enables specific visualization of proteins in living cells and has played an important role in our understanding of the protein subcellular location and function. Some proteins, however, show altered localization or function when labeled using direct fusions to fluorescent proteins, making them difficult to study in live cells. Additionally, the resolution of fluorescence microscopy is limited to ∼200 nm, which is 2 orders of magnitude larger than the size of most proteins. To circumvent these challenges, we previously developed LIVE-PAINT, a live-cell super-resolution approach that takes advantage of short interacting peptides to transiently bind a fluorescent protein to the protein-of-interest. Here, we successfully use LIVE-PAINT to image yeast membrane proteins that do not tolerate the direct fusion of a fluorescent protein by using peptide tags as short as 5-residues. We also demonstrate that it is possible to resolve multiple proteins at the nanoscale concurrently using orthogonal peptide interaction pairs.


Figure S1
Representative structures of interacting peptide pairs used in this study.

Figure S2
Can1 function is not impaired by the 101B tag used for LIVE-PAINT.

Figure S3
Gap1 and Bap2 function is not impaired by the 101B tag used for LIVE-PAINT.
Figure S4 101A-mNeonGreen alone does not localize to the plasma membrane.

Figure S5
Full-cell super-resolution images of membrane proteins labelled with the 101A/B peptide pair shown in Figure 3.

Figure S6
Analysis of whole population protein cluster data for each membrane protein imaged in Figure 3.

Figure S7
Quantification of the number of localizations per cluster for each membrane protein imaged in Figure 3.

Figure S8
Histograms of photon counts for SR images of each membrane protein presented in Figure 3.

Figure S9
Localizations are detected throughout three minutes of imaging with LIVE-PAINT.
Figure S10 Representative plots showing localizations classified as membrane, cellular or external for membrane proteins labelled with the 101A/B peptide pair.
Figure S11 KQTSV can be used as a peptide label for LIVE-PAINT imaging with two tandem repeats of PDZ3 fused to mNeonGreen.
Figure S12 Graphs of diffusion lag time as a function of PDZ concentration for the 1xPDZ and 2xPDZ proteins binding to the KQTSV peptide.
Figure S13 Full-cell super-resolution images of membrane proteins labelled with the KQTSV/2xPDZ3 peptide-protein pair shown in Figure 4.
Figure S14 Representative plots showing localizations classified as membrane, cellular or external for membrane proteins labelled with the KQTSV/2xPDZ3 peptideprotein pair.

Table S1
Membrane transporter proteins tagged and imaged in this study.

Table S2
Summary statistics of super-resolution images of membrane associated proteins imaged using LIVE-PAINT in Figure 3.

Table S3
Summary statistics of super-resolution images of membrane associated proteins imaged using LIVE-PAINT in Figure 4.

Table S4
Summary statistics of super-resolution images of Arc35 and Pil1 imaged using two color LIVE-PAINT and shown in Figure 5.

Table S5
Constructs used to measure binding kinetics ex vivo.

Table S6
Primers for tagging membrane proteins at their genomic loci.
Table S7 Primers for generating a yeast strain with two membrane-associated proteins tagged at their genomic loci.
Table S8 List of yeast strains used in this study.

Molecular biology
All cloning was performed in TOP10 E. coli, using standard techniques.All constructs were generated in the pFA6a-His3MX6 and pFA6a-KanMX6 yeast integration vectors.All linker sequences used are GGSGSGLQ.Plasmids were constructed via Gibson assembly using NEBuilder® HiFi DNA Assembly Master Mix (New England Biolabs) by amplifying the plasmid backbone via PCR and inserting other parts of the construct either using gBlocks (Integrated DNA Technologies) or other PCR amplified sequences.

Yeast strain construction
Except where otherwise noted, standard methods for genetically modifying yeast and preparing growth media were used. 1 To label membrane transporter proteins we used the 101A/101B coiled coil interaction pair, 2 which we have previously demonstrated to be compatible with LIVE-PAINT imaging. 3At the gene level, 101B was fused to the C-terminus of the target membrane protein at its endogenous locus.101A was fused to mNG, under control of the galactose-inducible promoter, and integrated into the genome, replacing GAL2.When GAL2 is deleted, the galactose-inducible promoter has a linear response with increasing galactose concentration. 4This same approach was used when tagging Pma1 using the KQTSV/PDZ3 interaction pair.
Yeast strains were produced by amplifying the insert and selection marker from a yeast integration plasmid (pFA6a-His3MX6 or pFA6a-KanMX6) with ~45 bp overhang sequences matching the 45 bp upstream of the target protein's stop codon and ~45 bp downstream of the desired integration site in the genome.The primers used to tag the various proteins studied in this work are listed in Supplementary Table 1.To transform the yeast, a 3 mL culture of yeast was grown overnight in YPD, back-diluted to an OD600 of 0.1 in 3 mL in the morning and grown to OD600 of ~0.6-0.8.Cells were pelleted, washed with 300 µL of 0.1 M LiAc twice, then pelleted and resuspended in 30 µL of 0.1 M LiAc.Then, with the cells on ice, the following were then added to the cells in order: 100 µL 50% w/v PEG 3350, 15 µL 1.0 M LiAc, 6 µL 7 mg/mL ss carrier DNA, 18 µL DMSO, 15 µL PCR product, and 14 µL sterile water.This was mixed and incubated in a GeneAMP TM PCR System 9700 (Thermofisher Scientific) for 30 minutes at 30˚C, followed by 15 minutes at 42˚C.Cells were pelleted, the transformation buffer was then removed by aspiration, and cells were resuspended in 100 µL water pre-warmed to 30˚C.All 100 µL of cells were then spread on an agar plate containing the appropriate selection.For histidine selection, plates were made from synthetic complete media lacking histidine.For G418 selection, cells were first plated on a YPD plate and then replica plated to a YPD plate including 600 µg/mL G418 the next morning.Plates were incubated at 30˚C for 2-3 days.A full list of strains used in this study are listed in Supplementary Table 2. Insertion of the desired construct in the genome was checked via colony PCR.

Preparing cells for microscopy
To prepare cells for microscopy, a single colony was picked from an agar plate into 500 µL synthetic complete media.100 µL of this cell suspension was then pipetted into 400 uL synthetic complete media, to obtain a 1:5 dilution as well.The cells were grown for 16-20 hrs in a shaking incubator at 30 ˚C.Whichever culture had an approximate OD600 of 0.1-0.8 was then used for imaging.
To prepare slides for imaging, 22x40 mm glass coverslips with thickness no. 1 (VWR) were cleaned using a 40-minute exposure to argon plasma in a 2.6 L Zepto plasma laboratory unit (Diener Electronic).Frame-Seal slide chambers (9 × 9 mm2, Biorad, Hercules, CA) were attached to the plasma-cleaned coverslips and the surface treated with 100 μL concanavalin A (Sigma-Aldrich) (2 mg/mL in PBS).The concanavalin A was removed from the surface after 30 s by aspirating with a pipette.100 μL of prepared yeast culture was then pipetted onto the slide and left for approximately 5 minutes to allow cells to attach.The cells were then aspirated from the slide and the surface washed using 100 μL fresh PBS and gently pipetting up and down.100 μL fresh PBS was then added to the slide before imaging.

TIRF microscopy
Microscopy data was primarily collected using a commercial TIRF microscope (Oxford Nanoimager).Images were acquired with an exposure of 50 ms for between 2000 and 4000 frames using the NimOS software.mNeonGreen was activated using the 488 nm laser at 15% power and the angle of illumination (TIR angle) was set at 48.5° for all acquisitions.Pixel length was 117 nm.All images were collected at room temperature.
Two color, Cdc12 and Pil1 microscopy data was collected using on a custom built TIRF microscope.mNG was excited using a 488 nm laser (Cobolt MLD 488-200 Diode Laser System, Cobalt, Sweden) ~40W/cm 2 and mCherry was excited using a 561 nm laser (Cobolt DPL Series 561-100 DPSS Laser System, Cobolt, Sweden) ~25W/cm 2 .Both lasers were aligned and directed parallel to the optical axis at the edge of a 1.49 NA TIRF objective (CFI Apochromat TIRF 60XC Oil, Nikon, Japan), mounted on an inverted Nikon TI2 microscope (Nikon, Japan).To prevent zstage drift during imaging a perfect focus system was utilized.Fluorescence collected by the same objective was separated from the returning TIR beam by a dichroic mirror (Di01-R405/488/561/635 (Semrock, Rochester, NY, USA)), and was passed through appropriate filters (488 nm: BLP01-488R, FF01-520/44 (Semrock, NY, USA), 561 nm: LP02-568-RS, FF01-587/35 (Semrock, NY, USA)).Fluorescence was then passed through a 2.5× beam expander and recorded on an EMCCD camera (Delta Evolve 512, Photometrics) operating in frame transfer mode (EMGain = 11.5 e-/ADU and 250 ADU/photon).Each pixel was 103 nm in length.The microscope was automated by using the open-source microscopy platform Micromanager.For two color samples, images were recorded by collecting 200 frames of 50 ms using the 561 nm laser followed by 200 frames of 50 ms using the 488 nm laser.This was repeated 10 times to result in a total of 2,000 frames for both mCherry and mNeonGreen.All Cdc12 and Pil1 images, were acquired using the 488 nm laser using an exposure of 50 ms for between 2000 and 4000 frames.
Prior to two color imaging, tetraspeck beads immobilized on a glass coverslip were used to check the image registration and alignment between the two channels.All images were collected at room temperature.

Confocal microscopy
For localization of Pma1 throughout entire cells, a Zeiss LSM880 confocal microscope (with alpha Plan-Apochromat 100x/1.46Oil DIC M27 Elyra objective) and Airyscan was used to capture mNG fluorescence data (ex/em 488/522) with z-stack increments of 0.170 micron.Airyscan image processing for generation of super-resolution 3D data was carried out in Zen Blue (Zeiss).

Super-resolution analysis
Super-resolution images were analyzed using Fiji (Java 8 2017 release).Single localizations were processed using the Peak Fit function of the Fiji GDSC SMLM plugin, using a signal strength threshold of 30, a minimum photon threshold of 100, and a precision threshold of 15-30 nm.All localization files are available at: https://doi.org/10.5281/zenodo.8101098.The overall resolution of the images was calculated using Fourier Ring Correlation analysis using the Python code available at: https://doi.org/10.5281/zenodo.7275952.

Quantification of membrane specific localizations
First, cell segmentation on the maximum intensity projections was performed using an adaptive Otsu algorithm on CellProfiler. 5Segmented cell objects were shrunk down by 4 pixels and subtracted from the cell mask to generate membrane masks.The code is available at https://doi.org/10.5281/zenodo.7817411was then used to tag localizations contained within localization files generated through super-resolution analysis described previously as cell, membrane or external based on overlap with the respective masks.Cells with fewer than 50 localizations were excluded from further analysis.

Cluster analysis
DBSCAN 6 complied in Python 3.8 using epsilon = 0.8 pixels and a minimum points threshold of 3 was used to identify clusters of localizations in the localization files generated through superresolution analysis described previously.For cells with more than 10 clusters, several parameters were then calculated and recorded: the number of localizations within each cluster, the ID of the nearest cluster, and the distance to the nearest cluster (determined using unsupervised nearest neighbors learning with the Ball Tree algorithm).The eccentricity of the cluster and the maximum length of the cluster was also calculated by fitting a 2 SD confidence ellipse to the cluster.The eccentricity was determined using the following formula: where a is the semi-major axis length and b is the semi-minor axis length.The maximum length of the clusters was taken to be the major axis length.This information was collated for all images of cells with the same protein target and peptide pair used for labelling.The Python code is available at: https://doi.org/10.5281/zenodo.8359460.

Two color cluster analysis
For two color images, cluster analysis was carried out on localization data for each color as described previously.The distance between each mCherry (Pil1) cluster and the nearest mNeonGreen (Arc35) cluster in the same cell was determined by calculating the Euclidian distance to all the mNG clusters in the cell and recording the shortest distance.This information was collated for all two-color images of Arc35 and Pil1.The Python code is available at: https://doi.org/10.5281/zenodo.8060654.

Localization rate analysis
Localization files generated through super-resolution analysis, described previously were loaded into Python 3.8.The localizations were then plotted to enable identification of localizations from neighboring cells which were then excluded from further analysis.The localizations recorded over intervals of 600 frames (30 s) were totaled for each localization file.The Python code is available at: https://doi.org/10.5281/zenodo.8360920.The total localizations recorded in each 30 s interval were then plotted against total imaging time using GraphPad (Prism).

Determining binding kinetics of peptides ex vivo
In order to quantify the binding of PDZ peptide to its target, Fluorescence Correlation Spectroscopy (FCS) was used together with an ex vivo system: a chemically synthesized and fluorescent-labelled KQTSV peptide and PDZ domain proteins (two constructs: 1xPDZ domainbiotin, 2xPDZ domain-biotin, purified from E. Coli, Table S5).
The diffusion through a confocal volume was observed using a home-built single molecule confocal instrument, previously described by Chappard et al.. 8 The confocal volume was illuminated by a 488 nm laser (LBX-405-100-CSB-OE, Oxxius; 1.2 mW irradiation at the backport), and 5x 30 second traces were recorded for each sample.The bursts of fluorescence as the peptide/peptide-protein passed the confocal volume were recorded on Avalanche Photodiode (APD) detector (PerkinElmer).
A custom-written Python script was used to transform the burst data into autocorrelation curves, and further fit diffusion lag time and diffusion coefficient (https://doi.org/10.5281/zenodo.8388699).Further analysis was performed in GraphPad (Prism).
The red channel diffusion lag time data was averaged for each condition (N=3, or N=2 for 1xPDZ where the residual value for a repeat exceeded the other residuals by an order of magnitude), and the standard deviation between repeat measurements calculated.The resulting plots were fit using a non-linear regression function (binding kinetics, one site specific binding) to the following function: where Bmax is the max diffusion lag time (plateau) and KD is the equilibrium dissociation constant.
Three fits were performed to obtain the average KD, using the mean from the three measurements and the values obtained on either extreme (+/-standard deviation).

Growth rate experiments
For growth studies, 100 μL microplate cultures were incubated in Tecan Sunrise plate readers at 30 °C, with 'normal' shaking level, and A595 was recorded at 15 min intervals, which correlates with cell density.Cells from pre-cultures (5 mL 6.7 g/L Yeast Nitrogen Base Sigma Y0626, 790 mg/L Complete Supplement Mixture Formedium DCS001, 2 % glucose) were washed with sterile water, and inoculated into test media at starting OD600 0.2 (corresponding to A595 0.04 in the plate reader).For canavanine (SLS C9758) toxicity studies, Complete Supplement Mixture -ARG (Formedium DCS0051) was used.The average OD600 measured in blank wells was subtracted from the OD600 recorded for the cell cultures and the blank corrected OD600 was plotted against incubation time using GraphPad (Prism).This permease has a broad specificity towards purines, and also transport cytosine and 5-methylcytosine but neither uracil nor thymine.

Gap1 3,634
General amino-acid permease involved in the uptake of all the naturally occurring L-amino-acids, related compounds such as ornithine and citrulline, some D-amino acids, toxic amino acid analogs such as azetidine-2-carboxylate, and the polyamines putrescine and spermidine.

25,100
Table S6.Primers for tagging membrane proteins at their genomic loci.p6h_int_F and p6h_int_R were used to amplify the plasmid containing pGAL1 101A-mNG, pGAL1 PDZ3-mNG, or pGAL1 2xPDZ3-mNG and HIS3 selection marker and to integrate it into the genome replacing GAL2.PMA1_URA3_R was used with PMA1_F to amplify the plasmid containing the KQTSV peptide and the URA3 selection marker.The ADY2, AGP1, ATO3, and BAP3 primers were used to amplify the plasmid containing the 101B peptide and the G418 selection marker.All other primers were used to amplify the plasmid containing the 101B peptide and the URA3 selection marker.The primer name contains the name of the gene being tagged and the primers come in pairs, with the "_F" and "_R" primers being used as forward and reverse primers.

Name
Table S7.Primers for generating a yeast strain with two membrane-associated proteins tagged at their genomic loci.The primers come in pairs, with the "_F" and "_R" primers being used as forward and reverse primers.Primers ARC35_F/R were used to fuse 101B to Arc35 by amplifying the plasmid containing 101B and the G418 selection marker.PIL1_MULTI_F/R were used to fuse 108B to PIL1 and insert pGAL1 101A-mCherry at the same gentic loci by amplifying the plasmid containing 108B, pGAL1-108A-mCherry and the selection marker LEU2.Table S8.List of yeast strains used in this study.

# Gene
Coiled-coil pair The gene being tagged is listed in the "Gene" column and the full genotype of the strain used for imaging is given in the "Genotype" column.

Figure S1 .
Figure S1.Representative structures of interacting peptide pairs used in this study.(a) Schematic

Figure S2 .
Figure S2.Can1 function is not impaired by the 101B tag used for LIVE-PAINT.(a) OD600 over

Figure S3 .
Figure S3.Gap1 and Bap2 function is not impaired by the 101B tag used for LIVE-PAINT.(a)

Figure S5 .
Figure S5.Full-cell super-resolution images of membrane proteins labelled with the 101A/B

Figure S6 .
Figure S6.Analysis of whole population protein cluster data for each membrane protein imaged

Figure S7 .
Figure S7.Quantification of the number of localizations per cluster for each membrane protein

Figure S8 .Figure 3 .
Figure S8.Histograms of photon counts for SR images of each membrane protein presented in

Figure S9 .
Figure S9.Localizations are detected throughout three minutes of imaging with LIVE-PAINT.

Figure S10 .
Figure S10.Representative plots showing localizations classified as membrane, cellular or

Figure S11 .
Figure S11.KQTSV can be used as a peptide label for LIVE-PAINT with two tandem repeats of

Figure S12 .
Figure S12.Graphs of diffusion lag time as a function of PDZ concentration for the 1xPDZ and

Figure S13 .
Figure S13.Full-cell super-resolution images of membrane proteins labelled with the

Table S1 .
Membrane transporter proteins tagged and imaged in this study.

Table S5 .
Constructs used to measure binding kinetics ex vivo.