Small Peptide-Protein Interaction Pair for Genetically Encoded, Fixation Compatible Peptide-PAINT.

Super-resolution microscopy via PAINT has been widely adopted in life sciences to interrogate the nanoscale architecture of many cellular structures. However, obtaining quantitative information in fixed cellular samples remains challenging because control of labeling stoichiometry is hampered in current approaches due to click-chemistry and additional targeting probes. To overcome these challenges, we have identified a small, PDZ-based, peptide-protein interaction pair that is genetically encodable and compatible with super-resolution imaging upon cellular fixation without additional labeling. Stoichiometric labeling control by genetic incorporation of this probe into the cellular vimentin network and mitochondria resulted in super-resolved 3D reconstructions with high specificity and spatial resolution. Further characterization reveals that this peptide-protein interaction is compatible with quantitative PAINT and that its binding kinetics remains unaltered upon fixation. Finally, by fusion of our probe to nanobodies against conventional expression markers, we show that this approach provides a versatile addition to the super-resolution toolbox.

Super-resolution microscopy has emerged as a revolutionary tool to study complex molecular assemblies with diffraction unlimited resolution.[1,2] One widely used super-resolution technique is single-molecule localization microscopy (SMLM). This technique relies on sequential detection and localization of fluorophores on a target to reconstruct its diffraction unlimited organization with nanometer spatial resolution.[3] So far, SMLM has already provided exciting new insights into the nanoscale architecture of many cellular structures.[4−7] However, obtaining quantitative, spatial information about the exact number of proteins required for molecular processes remains challenging. Recently, point accumulation for imaging in nanoscale topography (PAINT) has been proposed as a powerful strategy for SMLM that closes this gap.[8] One major advantage of this approach is that the probe is continuously recycled from solution, generating unlimited and continuous events without bleaching. As a consequence, quantitative PAINT (qPAINT) can be performed which allows counting of the number of molecules that were labeled.[9,10] In life sciences, PAINT via transient DNA hybridization (DNA-PAINT) has been widely adopted. In this strategy, the structure is targeted with a DNA-docking strand and imaged using a complementary, fluorophore-labeled, imager sequence.[11−13] Thus, far, DNA-PAINT has already been applied to resolve many cellular structures with nanometer resolution.[14−17] However, to achieve quantitative labeling in fixed cellular samples several challenges remain in order to reduce aspecific binding and control fluorophore and probe stoichiometry, while maintaining minimal linkage errors and high labeling efficiency. Controlling these parameters in DNA-PAINT is specifically challenging because labeling uncertainties are introduced by click-chemistry and targeting probes (e.g., nanobodies, antibodies, affimers, Halo/SNAP-tags) that are required to couple the DNA-docking strand to the protein of interest.[18−22] As a result, determining the effective labeling stoichiometry remains demanding.

One promising alternative to DNA hybridization is PAINT via transient protein-based interactions. Previous studies, using highly selective endogenous binding proteins as probes, were able to show robust SMLM reconstructions of a selection of cytoskeletal components in fixed cells.[23−28] Recently, a more general approach to such protein-based PAINT has been implemented by chemical labeling of secondary antibodies with one α-helical peptide of a well-characterized, transiently interacting coiled-coil pair. The other α-helix was conjugated to an organic dye for detection. This implementation, termed peptide-PAINT, provided the first proof-of-principle for PAINT via standardized peptide interaction pairs, similar to the approach in DNA-PAINT.[29] The potential of such standardized peptide-based probes for PAINT is high since their interaction is very specific and can be genetically encoded, allowing for controlled stoichiometric labeling of cellular proteins without additional chemistry. However, due to the presence of lysines within the coiled-coils, genetic expression and preservation of these interactions is likely perturbed upon fixation, limiting these interactions to indirect labeling via antibody coupling. Therefore, novel peptide-based interactions that are fixation compatible are highly interesting to provide small genetically encoded PAINT probes with controlled labeling stoichiometry. We argue that genetic engineering of such probes to a protein of interest within the cellular environment would allow for direct, controlled quantitative measurements and circumvent the need for additional chemistry for labeling.

In this work, we have identified such a small, genetically encodable peptide–protein interaction that is compatible with peptide-PAINT super-resolution imaging upon cellular fixation. This interaction is based on the transient interaction between the well-characterized Erbin PDZ domain and its small four amino acid peptide ligands.[30,31] We show that decoration of nanoparticles or genetic fusion of cellular proteins with this peptide allows for direct (q)PAINT acquisitions with purified mNeonGreen labeled PDZ domains. This interaction is unperturbed upon fixation and results in high quality, background-free reconstructions of vimentin and mitochondria in COS7 cells without any additional chemistry. Additionally, detection of the peptide with an ePDZ-based high affinity clamp, that has been widely used in live cells and reconstituted systems,[32−37] showed that this peptide could also serve as a genetic handle for easy one-step detection of target proteins in STED microscopy. Finally, we extended the application of our probe as a general PAINT strategy against standard expression markers by fusing anti-mCherry nanobodies to this peptide to stain and resolve expressed vimentin-mCherry networks in cells.

Results

One essential condition for PAINT imaging is that the interaction between the peptide–protein pair is transient with equilibrium dissociation constants (KD) in the micromolar range. The interaction of the Erbin PDZ domain with a short four amino acid C-terminal peptide sequence (DTWV-COOH) has previously been well-characterized to be highly specific with low micromolar affinity.[31,37] We speculated that it may therefore be compatible with transient detection in PAINT acquisitions. Furthermore, we hypothesized that as this sequence is only four amino acids and does not contain any lysines it would result in a minimal linkage error and fixation compatibility upon expression in cells. Moreover, previously directed domain-interface evolution studies have generated a nanomolar affinity clamp (KD ∼ <5 nM) to this DTWV sequence,[37,38] consisting of a fusion of an enhanced PDZ domain to domain II of fibronectin type III, termed ePDZ-b1. Therefore, in addition to performing PAINT via labeled PDZ domains, this high affinity ePDZ-b1 clamp can potentially be used as a one-step staining method for widefield or STED imaging (Figure a). One major advantage of such protein-based probes is that they can be expressed directly fused to a fluorescent protein, requiring no further labeling and allowing direct detection of the DTWV ligand. Therefore, to detect the peptDTWV via either PAINT or STED microscopy, mNeonGreen fused PDZ (PDZ-mNG) and ePDZ-b1 (ePDZb1-mNG) were purified from bacteria.

Figure 1

A small PDZ-based peptide–protein interaction pair for super-resolution microscopy. (a) Schematic representation of the experimental setup to image surface immobilized polystyrene nanoparticles coated with the biotinylated-peptDTWV sequence. SMLM and widefield/confocal microscopy can be performed via the transient interaction of PDZ-mNG or the high affinity interaction with ePDZb1-mNG with the peptDTWV respectively. (b) Representative STED and PAINT reconstruction images for the ePDZb1-mNG and PDZ-mNG probe, respectively. Inset shows the diffraction limited confocal (left panel) or TIRF (right panel) image of the nanoparticle highlighted by the white box. (c) Thirty nanometers thick cross section and three-dimensional rendering of a single nanoparticle imaged with astigmatism-based 3D-PAINT with the PDZ-mNG probe. Intensity profile of the indicated line in the cross-section is shown. Details of the experimental setup and conditions for STED and SMLM can be found in Supporting Information (Materials and Methods) and Table S2. Scale bar: 1 μm (b), 100 nm (c).

To test whether the interaction of the peptide with PDZ-mNG or ePDZb1-mNG is indeed compatible with PAINT or STED, respectively, 300 nm streptavidin-coated polystyrene nanoparticles were decorated with the biotinylated DTWV peptide sequence separated by a short linker (biotin-peptDTWV) (Figure a). Subsequently, imaging of the immobilized nanoparticles was performed. PAINT acquisitions with PDZ-mNG in solution displayed transient binding and allowed for diffraction unlimited reconstructions of the nanoparticles (Figure b,c). Furthermore, control PAINT experiments in the absence of biotin-peptDTWV showed that this peptide/protein interaction is highly specific with very few off-target interactions of the PDZ-mNG probe, resulting in high signal-to-noise ratios upon reconstruction (Figure S1). In contrast to transient binding by PDZ-mNG, purified ePDZb1-mNG bound with high affinity to the biotin-peptDTWV decorated nanoparticles and allowed STED microscopy (Figure b). These STED images already showed a markable improvement in resolution compared to confocal imaging (Figure b, Figure S2). Gratifyingly, these results demonstrate that the interaction of PDZ-mNG and ePDZb1-mNG with the small c-terminal peptDTWV ligand are highly suitable for PAINT and STED super-resolution microscopy, respectively.

Next, we set out to further explore the potential of the peptDTWV/PDZ-mNG pair as a genetically encodable single molecule probe that is compatible with qPAINT. First, to characterize the resolving power, peptDTWV distributions along the nanoparticle were reconstructed through astigmatism-based 3D-PAINT in the presence of ∼10 nM PDZ-mNG probe in solution (Table S2). Acquisitions were performed at high laser intensities to rapidly observe single events with high precision. The resulting three-dimensional reconstructions and corresponding cross sections could resolve the biotin-peptDTWV decoration along the surface of the nanoparticles with a localization accuracy of 15 ± 7 nm, far beyond the diffraction limit (Figure c). Further dilution of PDZ-mNG to 0.1 nM then enabled imaging under conditions where single binding events could clearly be distinguished under low, nonbleaching laser intensities. This showed that the number of events remained constant over time, which is an essential requirement for molecular counting in qPAINT (Figure S3). Therefore, in addition to diffraction unlimited mapping at the nanoscale, this interaction may be utilized to quantitatively map the number of peptide functionalized targets.

To test the latter and further characterize the peptDTWV/PDZ-mNG interaction, we set out to count the number of immobilized biotin-peptDTWV molecules at the nanoparticle surface. To achieve this, kinetic information is required to determine the number of molecules according to the formula n = (kON × ci × τd*)−1, formulated in the previously developed DNA-based qPAINT strategy.[9,10] Here the number of peptides (n) is dependent on the association constant (kON), the concentration (ci), and the average dark times (τd*). To extract the value of kON, a well-controlled experiment was set up where individual biotin-peptDTWV molecules were immobilized on passivated, streptavidin-functionalized surfaces. The applied, biotin-peptDTWV concentrations were diluted until single molecules, spaced apart, resulted in a single cluster of detections upon PAINT imaging (Figure a, Figure S4). Consequently, solving the equation for each localization cluster now containing only one molecule (n = 1) leads to kON = (ci × τd*)−1. After immobilization, the qPAINT calibration was performed in the presence of 26 nM PDZ-mNG. Each cluster of localizations, around a single biotin-peptDTWV, was automatically detected so that the corresponding binding kinetics could be assessed. This resulted in observed average dark-times of 102 ± 60 s per cluster, corresponding to kON values of ∼3.6 × 105 M–1 s–1. Furthermore, the average bright times (τb*) were 159 ± 39 ms, which is in good agreement with typical integration times for PAINT imaging[29] (Figure b). Using these parameters, the ability to accurately count molecules was demonstrated by determining the number of peptides from pooled binding events for a known number of observed clusters in the calibration experiment.[10] This showed that there is indeed a linear relationship between the “ground truth” and the qPAINT-determined amount using these parameters as input (Figure S4b). These results therefore show that even though the association constant of the PDZ-mNG/peptDTWV interaction is lower as would be selected in typical peptide- and DNA-PAINT experiments,[9,29] the observed kinetics combined with the low signal-to-noise levels and high specificity of the PDZ-mNG with peptDTWV are well compatible with qPAINT.

Figure 2

The transient interaction between peptDTWV and PDZ-mNG is qPAINT compatible (a) Schematic representation of the experimental setup to generate a qPAINT calibration to extract k. Inset shows localization clusters for single peptDTWV strands after acquisition with PDZ-mNG (26 nM). (b) Normalized frequency distributions of the average dark (τd*) and bright times (τb*) for single peptDTWV interacting with PDZ-mNG (n = 1000, equally divided over four acquisitions). Solid line shows the result of a fit with a Gaussian model. (c) Representative PAINT reconstruction of peptDTWV decorated nanoparticles imaged at low PDZ-mNG concentrations (0.1 nM) so that single binding events could be observed for qPAINT. (d) Normalized frequency distributions for the number of biotin-peptDTWV peptides docked to a single nanoparticle (n = 449). The solid line shows the result of a fit with a Gaussian model to the data to determine the mean and standard deviation (1690 ± 712). Details of the experimental setup and conditions can be found in Supporting Information (Materials and Methods) and Table S2. Scale bar: 500 nm (a), 1 μm (c).

Having determined kON, we next performed qPAINT on the nanoparticles to quantify the number of accessible, peptDTWV-tagged streptavidin complexes on their surface. Imaging was performed under identical illumination conditions to those of the calibration and now in the presence of 0.1 nM PDZ-mNG to detect single binding events (Figure c). Measurements of the average dark times per nanoparticle resulted in an average of 1690 ± 712 sites per nanoparticle which corresponds to 422 streptavidin tetramers (Figure c,d, Figure S4c–e). These values therefore indicate an approximate streptavidin spacing of ∼25 nm on the surface of the 300 nm nanoparticles. Interestingly, these values are in close agreement with previous bulk measurements of the accessible biotin-binding sites on commercial streptavidin-coated magnetic nanoparticles of 500 nm,[39] yet higher than previously reported for DNA based qPAINT experiments.[9] One possible reconciliation for this observation is a differential occupation of tetrameric streptavadin by the peptDTWV or DNA sequences as DNA is generally bigger and highly charged causing steric and electrostatic hindrance.

So far, our results show that the interaction of PDZ-mNG with its peptDTWV ligand is compatible with (q)PAINT to resolve the nanoscale structure of molecular assemblies and provide quantitative information about the number of labeled molecules. Next, we wanted to explore whether this interaction remains functional after genetic expression in complex cellular environments followed by chemical fixation. This direct fusion of the peptDTWV sequence would be highly advantageous over previous strategies, as it directly controls stoichiometry and does not require additional specific targeting agents (e.g., antibodies, affimers, nanodies) or labeling steps. As conventional methods to preserve the nanoscale structure of different cellular components contain either paraformaldehyde (PFA) or a combination of PFA and glutaraldehyde (GA), targets that require fixation via either of these methods were selected. For PFA, the cytoskeletal filament vimentin was chosen[23] and for PFA/GA fixation the mitochondrial outer membrane transporter subunit TOMM20[40] was selected. To perform peptide-PAINT with the PDZ-mNG probe, both targets were cloned as fusion constructs with a C-terminal peptDTWV sequence. Additionally, to identify the transfected cells and have a diffraction limited reference, mCherry was engineered in between the C-terminus of these proteins and the peptDTWV. After 1 day of expression with either construct in COS7 cells, the cells were fixed and permeabilized to allow accessibility of the purified PDZ-mNG probe from solution to bind to the expressed peptDTWV sequence. Subsequently, the probe was washed in at high concentrations and imaged at high laser powers so that single binding events could be detected with high localization accuracy and frequency during 3D-PAINT acquisitions. These acquisitions resulted in high-quality super-resolved reconstructions of both cellular vimentin and TOMM20 under both fixation conditions (Figure ). Further assessment of the image quality showed that the PDZ-mNG/peptDTWV interaction allows for single-molecule localization with high precision (15.4 ± 8.5 nm) to achieve an estimated Fourier ring correlation (FRC) image resolution of 53 ± 4 nm (Figure S5). Additionally, in our approach, very little aspecific interactions of the PDZ-mNG could be observed in the fixed cellular samples which is favorable for optimal PAINT probes (Figure S6). Next, to characterize binding kinetics in fixed cells, additional imaging was performed where PDZ-mNG was diluted significantly and imaged at low laser powers to eliminate bleaching and detect single binding and unbinding events. This showed that the number of localizations over time remained constant and revealed bright times of 169 ms. (Figure S7). These kinetics are highly consistent with the observed kinetics at the nanoparticles (Figure ) suggesting that the interaction between peptDTWV and PDZ-mNG is not affected by fixation. Additionally, to extend the cell biological applications of this small four amino acid sequence for fluorescence detection, we tested whether the ePDZb1-mNG affinity clamp could be used as a one-step staining probe in fixed cells. Again vimentin-mCherry-peptDTWV and TOMM20-mCherry-peptDTWV were expressed and fixed in COS7 cells. Subsequently, they were imaged in the presence of ePDZb1-mNG (Figure S8). Confocal microscopy showed that the ePDZb1-mNG fully decorated the expressed proteins marked by mCherry. Subsequent STED acquisitions then improved the resolution for both Vimentin and TOMM20 labeling (Figure S8). These experiments therefore show that the high affinity interaction of the peptDTWV with ePDZb1 is applicable in a one-step staining strategy with a genetically encoded tag.

Figure 3

PAINT via the peptDTWV/PDZ-mNG interaction is compatible with conventional fixation strategies upon cellular expression. The 3D-PAINT reconstructions of a paraformaldehyde fixed vimentin-mCherry-peptDTWV network (a) and paraformaldehyde/glutaraldehyde fixed mitochondria (b) after expression in COS7 cells. Three-dimensional reconstructions are color-coded for depth (top panels); diffraction limited mCherry preacquisition and corresponding reconstruction (bottom panels) are depicted. For the zoom of mitochondria, a 100 nm slice was selected for the reconstruction to show the mitochondrial lumen (bottom right panel). Details of the experimental setup and conditions can be found in the experimental section and Table S2. Scale bar: 5 μm (top panels); 2 μm (bottom panels); depth in nanometer color coded as indicated in figure.

Finally, we set out to implement PAINT via the peptDTWV sequence in a more general labeling approach. We argued that recombinant fusion of peptDTWV to high affinity probes, such as nanobodies or affimers, would allow for an additional convenient method to perform PAINT on many already available epitopes used in SMLM.[18,41−44] Furthermore, as these probes only recognize one epitope, fusion of the small peptDTWV sequence to their C-terminus theoretically maintains stoichiometric control by targeting only one peptDTWV to the protein of interest. To test this approach, we choose the previously described nanobody against mCherry (VHHmCherry)[45] in combination with the vimentin-mCherry construct expressed in COS7 cells (Figure a). After fixation, the samples were stained with the purified VHHmCherry-peptDTWV and PAINT with the PDZ-mNG probe (1 nM) was performed. Again, this resulted in super-resolved reconstructions of the vimentin network with high spatial resolutions and little aspecific interactions (Figure b, Figure S9).

Figure 4

Targeting conventional epitopes via a nanobody-peptDTWV fusion. (a) Schematic representation of the experimental setup. Vimentin-mCherry is expressed in COS7 cells, fixed and subsequently detected with VHHmCherry-peptDTWV to perform peptide-PAINT. (b) Representative image of a whole COS7 cell after 3D-PAINT acquisition. Two-dimensional visualization of the whole cell (left panel) and the diffraction limited mCherry signal and 3D reconstruction for a selected area (right panels) are depicted. Details of the experimental setup and conditions can be found in Supporting Information (Materials and Methods) and Table S2. Scale bar: 10 μm (full cell); 5 μm (zoom), depth in nanometer color-coded as indicated in figure.

Conclusions and Discussion

In conclusion, we have introduced a novel peptide-PAINT probe for SMLM based on the transient interaction of the PDZ domain with a small four amino acid (DTWV) C-terminal sequence. PAINT imaging of 300 nm nanoparticles with the PDZ domain revealed that this interaction is compatible with imaging at high spatial resolutions, specificity, and signal-to-noise ratios. Further characterization showed that this probe is qPAINT compatible and allowed quantification of the number of functional sites on the nanoparticles. Expression of the peptDTWV sequence fused to both vimentin and TOMM20 in COS7 cells showed that this interaction is fixation compatible while maintaining stoichiometric control via genetic expression. This allowed us to map the three-dimensional nanoscale organization of both structures within the complex cellular environment. Furthermore, binding of the PDZ-mNG with the peptDTWV, after fixation with both PFA or GA, displayed kinetics similar to those at the nanoparticles, indicating that the interaction was not perturbed by conventional fixation methods. Additionally, we could show that the peptDTWV sequence can be used as a minimal on-step staining tag upon detection with its high affinity ePDZb1 binding partner. Finally, fusion of the short peptDTWV sequence to nanobodies extended the application of this interaction as a general PAINT approach targeted against commonly used expression markers and epitopes.

We predict that the PDZ/peptDTWV and other genetically encodable and fixation compatible interactions will have promising implications in the quantitative assessment of the structure and composition of molecular assemblies with nanometer resolution. Specifically, application of these interactions in combination with novel genetic tools enables direct quantitative nanoscopy in cell biology without the need for additional chemical labeling steps. For example, CRISPR/Cas-based engineering of proteins of interest fused to the small peptDTWV will result in complete labeling of all target proteins with 1:1 stoichiometry. Subsequent fixation and qPAINT with the PDZ-mNG will then allow direct molecular counting. Furthermore, our PAINT approach complements previously established SMLM strategies, such as quantitative STORM[46,47] and qPALM,[48,49] that are based on the calibrated blinking kinetics of a finite number of localizations of organic dyes or fluorescent proteins to quantify protein heterogeneity and numbers. Currently, one limitation of the approach described in this work is that its peptDTWV ligand needs to be located at the C-terminus of the target construct. Whereas this could partially be overcome with label controlled nanobody intermediates against other conventional targets as shown for the VHHmCherry, future work should focus to extend the peptide-PAINT based toolbox with other fixation compatible interactions to enable multiplex detection of several proteins of interest simultaneously. Additional interesting candidates might be found in previous live-PAINT experiments where such transient interactions are implemented.[50−53] However, their compatibility with cellular fixation should be assessed. Alternatively, synthetic routes to rationally engineer compatible and highly tunable short peptide interactors would be highly interesting. In conclusion, because of its versatility, we think that our PDZ-based PAINT approach will provide a valuable new addition to the super-resolution toolbox to quantitatively study both reconstituted and cellular assemblies at the molecular level.