A novel "bio-tag" for cryo-EM studies based on the small, electron-dense protein Csp1

Summary

Small proteins can be challenging to study by single-particle cryogenic electron microscopy (cryo-EM) techniques because they have low signal-to-noise ratios, making them difficult to identify and analyze. Here we investigated the use of Csp1, a small (∼50 kDa) tetrameric metal-binding protein, to act as a “bio-tag” to help overcome this problem. We find Csp1 is compact, stable, and exhibits enhanced electron scattering and excellent particle contrast in cryo-EM micrographs. As a result, we could determine the structure of Csp1 to 2.98-Å resolution using standard cryo-EM approaches. We also tested if Csp1 could be used as a tag or fiducial to help determine the structure of a protein bound to it. Specifically, we analyzed an epitope-tagged Csp1 bound to a ∼40 kDa Fab fragment from the antibody 1D4. Data from these complexes yielded medium-resolution structures of the complex (5.70 Å) and the bound 1D4 Fab (5.40 Å). These results suggest that, with further optimization, electron-rich Csp1 is a promising bio-tag for use in cryo-EM studies.

Significance

Small proteins can be difficult to study using single-particle cryo-EM because they lack contrast and recognizable features in images. Here, we tested if Csp1, a small (∼50 kDa) tetrameric metal-binding protein, can be used as an electron-dense “tag” to help overcome these challenges and improve particle alignment. We show Csp1 by itself has excellent contrast and visibility in cryo-EM micrographs, allowing its structure to be determined to 2.98-Å resolution. We also find an epitope-tagged Csp1 could be used to facilitate structural analysis of a bound ∼40 kDa Fab fragment to a medium resolution (5.40 Å). Our results suggest Csp1 is a promising candidate for further development as an electron-dense protein-tag in cryo-EM studies.

Introduction

Cryogenic electron microscopy (cryo-EM) has dramatically advanced the speed and ability to solve protein structures due to new methods and hardware improvements (1). Compared to other high-resolution structural methods, cryo-EM approaches have less strict demands on sample quantity, purity, or buffering conditions. Increasingly, cryo-EM is the method of choice for studying proteins that are flexible, oligomeric, complexed, or embedded in membrane-like environments.

However, some proteins, especially small proteins (less than ∼100 kDa) can be difficult to identify and analyze due to their low particle contrast and signal to noise in cryo-EM images. Technical approaches have been developed to address these problems, such as minimizing ice thickness to increase contrast (2,3,4) and the use of phase plates, graphene oxide grids, or other imaging schemes. Using these approaches, structures for several small proteins like streptavidin (52 kDa) or various globins (64 kDa) have been solved (5,6,7). However, there is still a need for more and better tools for studying small proteins by cryo-EM.

Our goal in the present work was to develop a “bio-tag” to provide an additional tool for addressing the above problems. Surprisingly, although tag-based approaches are widely used in structural biology, there are still relatively few for cryo-EM research. Among these are electron-dense chemical tags (such as quantum dots and gold nanoclusters), which are helpful for identifying particles in images (8), and epitope tags that bind specific antibody Fabs or nanobodies to both aid in particle detection as well as break pseudo symmetry or identify hetero-oligomers (3,9). A parallel approach has involved using symmetric scaffolds, wherein a small protein is bound to a larger one to facilitate imaging (9). Ideally, a protein-based cryo-EM bio-tag would incorporate all of these approaches into one unique, electron-dense, symmetric scaffold.

We started with the idea that small heavy-metal-binding proteins could be optimal contenders for a cryo-EM bio-tag, as their increased electron density should scatter electrons strongly yielding higher image contrast. Attaching such a tag to other proteins should then make particles both easier to identify and to align and classify. The result should facilitate achieving higher-resolution reconstructions.

With these criteria in mind, we identified the family of copper storage proteins (CSPs) as a source of potential candidates for such tags. CSP proteins are small yet bind numerous copper ions within a series of four-helix bundles (10). One CSP family member, Csp1, is an especially promising candidate for use as a cryo-EM tag. Csp1 is a small, stable, tetrameric protein (∼50 kDa) that can bind up to 52 copper ions (13 per protomer), and its D2 symmetry could potentially be exploited to act as an imaging scaffold. The structure of Csp1 (Fig. 1 a) is already known to 1.2-Å resolution (10), thus simplifying protein engineering efforts. Below, we report how these properties facilitate single-particle cryo-EM analysis of Csp1 by itself and when bound to another protein.

Figure 1.

Scheme for using the copper-binding protein Csp1 as a bio-tag in cryo-EM studies. (a) Structure of a Csp1 tetramer (PDB: 5FJE) showing the four identical four-helix subunits in different colors, as well as bound copper (gold balls). (b) 2D cartoon representation of Csp1_1D4-ET, a modified Csp1 that has a 1D4 antibody epitope at the C terminus and a prodomain tag at the N terminus. The 1D4 epitope enables binding of the 1D4 Fab and the prodomain enables purification using the Profinity eXact system, which is cleaved off during the process. (c) Cartoon scheme showing how a Csp1_1D4-ET tetramer (center) can potentially bind up to four 1D4 Fab fragments.

Materials and methods

Materials and reagents

All restriction enzymes, ligase, and DNA polymerase were from New England Biolabs (Ipswich, MA). Profinity eXact purification resin was purchased from Bio-Rad (Hercules, CA). Amicon Ultra Protein concentrator (10-kDa cutoff) and Microcon YM-50 centrifugal filters were from Millipore (Burlington, MA). 1D4 antibody was obtained from the Monoclonal Antibody Core at the Vaccine and Gene Therapy Institute of Oregon Health and Science University (Portland, OR). Immobilized papain resin and L-cysteine for antibody digestion were from Bio-Rad. All other chemicals and reagents were purchased from Sigma-Aldrich (Burlington, MA) and Fisher Scientific (Pittsburgh, PA).

Negative-stain EM (electron microscopy) grids are extra-thick carbon film 400-mesh grids purchased from Electron Microscopy Services (EMS; Hatfield, PA). The cryo-EM grids were Quantifoil R1.2/1.3 400-mesh copper and Quantifoil R2/2 300-mesh copper, both purchased from 2spi. C-clips and rings were purchased from FEI (Hillsboro, OR). Vitrobot filter paper was purchased from EMS.

Buffers

Lysis buffer is 20 mM Tris-acetate, 0.5% (w/v) octylthioglucoside, 0.1% (v/v) Triton X-100, pH 8.5. Wash buffer 1 is 20 mM Tris-acetate, 1 mM DTT, pH 8.5. Wash buffer 2 is 20 mM Tris-acetate, 300 mM (NH₄)₂SO₄, 1 mM DTT, pH 8.5. Elution buffer is 20 mM Tris-acetate, 1 mM DTT, 100 mM NaF, pH 8.5. Size-exclusion chromatography (SEC) buffer is 20 mM HEPES, 200 mM NaCl, pH 7.5. EM buffer is 20 mM HEPES, pH 7.5. Digestion buffer is 20 mM Na₂HPO₄, 10 mM EDTA, 20 mM L-cysteine, pH 7. PBSSC (phosphate buffered saline) is 137 mM NaCl, 2.7 mM KCl, 1.5 mM K₂HPO₄, 1 mM Na₂HPO₄, pH 7.2.

Csp1 cloning

For fast purification of tag-less Csp1 protein, the pro-subtilisin affinity tag method was used. We first designed a gene encoding Csp1 protein without the Tat leader sequences, which was then then synthesized by GenScript (Piscataway, NJ). To create a protein for subsequent Fab-binding studies, a construct called Csp1_1D4-ET was designed in which the epitope for 1D4 antibody (TETSQVAPA) was added to the C- terminus of Csp1. The gene encoding Csp1 was amplified with forward primer containing NcoI restriction site and reverse primer containing HindIII restriction site and sequence encoding 1D4 epitope. The amplified gene fragment was excised with NcoI and HindIII and cloned into a pET28(a) plasmid containing the 77-amino acid long ProR8FKAM subtilisin BPN' prodomain coding sequence.

Csp1_1D4-ET expression and purification

Competent BL21(DE3) cells were transformed with pET28-ProCsp1_1D4-ET and grown in LB (lysogeny broth) medium at 37°C (100 μg/mL kanamycin) overnight and then diluted 10 times in LB medium (100 μg/mL kanamycin) and grown at 37°C for 1 h. Expression of the Csp1_1D4-ET was induced with 1 mM isopropyl β-D-thiogalactopyranoside at 37°C. Cells were harvested after 4 h of induction and stored as cell pellets at −80°C. Cells were lysed in 20 mL of lysis buffer. Small amounts of DNaseI and lysozyme were added to the lysis reaction. The lysis reaction was nutated at room temperature for 30 min and then 1 mM DTT was added. Cell lysate was clarified at 41,000 × g for 30 min at 4°C.

Supernatant was filtered through a 0.45-μm polyvinylidene fluoride (PVDF) syringe filter and nutated with 10 mL of Profinity eXact resin (pre-equilibrated with wash buffer 1) at 4°C for 30 min. The resin was washed with five column volumes (CV) of wash buffer 1, 5 CV of wash buffer 2, and 10 CV of wash buffer 1. The resin was then rinsed with 1 CV of elution buffer and incubated with 1 CV of elution buffer at 4°C for 3 h to overnight. The resins were eluted with elution buffer and all elution fractions containing protein were combined and concentrated with Amicon Ultra-15 centrifugal filter unit (10-kDa cutoff).

Concentrated Csp1_1D4-ET was further purified by SEC. Csp1_1D4-ET (500 μL) was injected onto a Superdex200 10/300 GL gel filtration column in SEC buffer at a flow rate of 0.5 mL/min and 0.5-mL fractions were collected. Absorbance at 280 nm was monitored and purity of elution fractions containing protein were assessed with SDS-PAGE. Concentration of SEC-purified Csp1_1D4-ET was calculated based on the absorbance at 280 nm with extinction coefficient of 4120 M⁻¹cm⁻¹. Purified Csp1_1D4-ET was stored at 4°C for immediate use or snap-frozen with liquid nitrogen for long-term storage.

Csp1_1D4-ET metal loading

Fresh Cu(I) stock solution was made before use by dissolving tetrakis(acetonitrile)-copper(I)hexafluorophosphate in acetonitrile to 100 mM and diluted to between 1 and 10 mM in EM buffer. Before adding metal to Csp1_1D4-ET, the protein was buffer exchanged into EM buffer with Zeba desalting column (7-kDa cutoff). To add Cu(I) to Csp1_1D4-ET, about 14- to 15-fold molar excess of Cu(I) was added to SEC-purified Csp1_1D4-ET proteins, and excess metal ions were removed with Zeba desalting column (7-kDa cutoff).

1D4 Fab preparation and purification

In brief, Fab fragments were prepared by digesting the antibody 1D4 on immobilized papain resin. The Fc regions of anti-1D4 were then removed with protein A resin and 1D4 Fab further purified by SEC.

Specifically, first the 1D4 Fab was generated by papain digestion. Immobilized papain resin (250 μL) was equilibrated with 4 mL of freshly made digestion buffer, and 250 μL of digestion buffer was added to the resin to make a 50% slurry mixture of the resin ready to be used. 1D4 antibody was diluted with digestion buffer to 1 mg/mL, and 500 μL of the diluted antibody was added to the resin slurry. The mixture was agitated to disperse the resin and incubated at 37°C for more than 20 h. The digestion reaction was quenched with addition of 125 μL of 20 mM Tris, pH 7.5, and the supernatant was removed from the papain resin. The collected supernatant was added to protein A resin (100 μL, pre-equilibrated with PBSSC) and nutated at 4°C overnight. Unbound fraction was collected from the protein A resin and the resin was rinsed with 3 CV of PBSSC before eluting with 1 CV of 100 mM glycine, 100 mM NaCl, pH 2. All fractions from protein A resin were examined with SDS-PAGE. 1D4 Fab fragments were further purified with SEC and stored at 4°C for immediate use or snap-frozen with liquid nitrogen and stored at −80°C for long term storage. Concentration of purified Fab was estimated based on the absorbance at 280 nm with extinction coefficient of 1.4 and molecular weight of ∼40 kDa.

1D4 Fab with Csp1_1D4-ET complex preparation

Cu(I) bound Csp1_1D4-ET was incubated with 1D4 Fab fragments at a molar ratio of 2 Csp1 monomer to 1 1D4 Fab in EM buffer at 4°C overnight. The binding reaction was concentrated with Microcon YM-50 at 13,000 × g, 4°C for 5 min and stored at 4°C until further use. Complex formation and protein concentrations were examined using negative-staining EM before cryo-EM.

SEC analysis of 1D4 Fab with Csp1_1D4-ET complex

Apo-state Csp1_1D4-ET (12 μM) was incubated with 1D4 Fab (6 μM) at a molar ratio of two Csp1 monomer to one 1D4 Fab at 4°C for 4 h and assessed with SEC using Superdex 200 increased at flowrate of 0.75 mL/min with 20 mM HEPES, 200 mM NaCl, pH 7.5 as running buffer.

Csp1_1D4-ET negative-stain screening

The quality of purified Csp1_1D4-ET protein molecules was first validated by negative-stain screening. Negative-stain grids were made as previously discussed above and imaged on a FEI Tecnai T12 120 kV microscope. A small dataset of 25 micrographs were acquired and processed in Relion (11). The resultant 2D classification results provided sufficient quality and orientations for further EM studies.

Cu-bound Csp1_1D4-ET cryo-EM studies

Following successful negative-stain data, 2.5 μL of the Csp1_1D4-ET sample was loaded onto glow-discharged (60 s, 15 mA) Quantifoil R1.2/1.3 300-mesh copper grids, blotted for 1.5 s at ∼100% humidity, and flash frozen in liquid ethane using a Vitrobot Mark III (FEI). The grids were then transferred to a 300-keV Titan Krios cryoelectron microscope with a standalone Gatan K3 DED camera and Bioquantum energy filter with a slit width of 20 eV for single-particle cryo-EM data acquisition. A total of 7,020 movies were recorded at 130,000× magnification in super-resolution mode with a pixel size of 0.32415 Å/pixel and a defocus range of −0.8 to −1.7 μm. Each of the 7,020 movies contains 42 frames with a total dose of 42 e⁻/pix² (dose rate of ∼18 e⁻/pix²/s, frame rate of 0.024 s/exposure, for ∼1 s at a spot size of 5).

Raw images were binned two times and patch motion corrected with dose weighting using CryoSPARC’s (12) implementation. The aligned micrographs were passed to CryoSPARC’s patch CTF correction and inspected afterward to select 6,888 micrographs for further processing. Next, 6,383,264 particles were picked using 2D templates and extracted at a binned pixel size of 1.2966 Å/pixel. After iterative 2D and 3D classifications, 125,008 particles were selected as the final particle stacks and re-extracted at 0.6483 Å/pixel for refinement using CryoSPARC’s nonuniform refinement algorithm.

Apo-Csp1_1D4-ET cryo-EM studies

Apo-Csp1_1D4-ET sample (2.5 μL) was loaded onto glow-discharged (60 s, 15 mA) Quantifoil R1.2/1.3 300-mesh copper grids, blotted for 1.5 s at ∼100% humidity, and flash frozen in liquid ethane using a Vitrobot Mark III (FEI). The grids were then transferred to a 300-keV Titan Krios cryoelectron microscope with a standalone Gatan K3 DED camera and Bioquantum energy filter with a slit width of 20 eV for single-particle cryo-EM data acquisition. A total of 6,018 movies were recorded at 130,000× magnification in super-resolution mode with a pixel size of 0.32415 Å/pixel and a defocus range of −0.8 to −1.7 μm. Each of the 6,018 movies contains 42 frames with a total dose of 42 e⁻/pix² (dose rate of ∼18 e⁻/pix²/s, frame rate of 0.024 s/exposure, for ∼1 s at a spot size of 5).

Raw images were binned two times and patch motion corrected with dose weighting using CryoSPARC’s (12) implementation. The aligned micrographs were passed to CryoSPARC’s patch CTF correction and inspected afterward to select 5,342 micrographs for further processing. Next, 8,482,428 particles were picked using 2D templates and extracted at a binned pixel size of 1.21556 Å/pixel. After iterative 2D and 3D classifications, 110,533 particles were selected as the final particle stacks and re-extracted at 0.6483 Å/pixel for refinement using CryoSPARC’s nonuniform refinement algorithm.

Csp1_1D4-ET-1D4 Fab complex negative-stain screening

The quality of purified Csp1_1D4-ET-1D4 Fab protein molecules was first validated by negative-stain screening. Negative-stain grids were made as previously discussed above and imaged on a FEI Tecnai T12 120-kV microscope. A small dataset of 37 micrographs were acquired and processed in Relion. The resultant 2D classification and initial model results provided a comprehensive quality check as well as a low 20-A initial model for use in processing of cryo-EM data.

Csp1_1D4-ET-1D4 Fab complex cryo-EM studies

Following successful negative-stain data, 5 μL of the Csp1_1D4-ET-1D4 Fab bound sample was loaded onto glow-discharged (60 s, 15 mA) Quantifoil R1.2/1.3 400-mesh copper grids, blotted for 4.0 s at ∼100% humidity, and flash frozen in liquid ethane using a Vitrobot Mark III (FEI).

The grids were then transferred to a 300-keV Titan Krios cryoelectron microscope with a Gatan K3 DED camera and Bioquantum energy filter with a slit width of 20 eV for single-particle cryo-EM data acquisition. A total of 1,926 movies were recorded at 130,000× magnification in super-resolution mode with a pixel size of 0.324 Å/pixel and a defocus range of −1.2 to −3 μm. Each of the 1926 movies contains 79 frames with a total dose of 42 e⁻/pix² (dose rate of 21 e⁻/pix²/s, frame rate of 0.025 s/exposure, for 2.0 s at a spot size of 8).

Raw image stacks were patch motion corrected with dose weighting using CryoSPARC’s implementation (12). The aligned micrographs were passed to CryoSPARC’s patch CTF correction and inspected afterward to select 1,055 micrographs for further processing. Next, 1,181,918 particles were picked using 2D templates and extracted at a binned pixel size of 1.597 Å/pixel. Iterative 2D classifications resulted in 965,710 particles, and 3D classifications were performed resulting in a particle stack of 326,835 used for refinement. To resolve the complex of Csp1_1D4-ET-1D4, a particle stack of 58,921 was used, and further local refinement of the 1D4 region using the same particle stack produced the final map of the 1D4 Fab fragment. To resolve the Csp1 in the complex, a particle stack of 47,664 was used, and nonuniform refinement followed by local refinement of the Csp1 region were performed to produce the final map of the Csp1 protein.

Results

Csp1 cloning, expression, purification, and biochemical characterization

To enable Fab binding, we created a construct called Csp1_1D4-ET (Figs. 1, b and c). Csp1_1D4-ET has the epitope for the anti-rhodopsin antibody 1D4 (TETSQVAPA) introduced at the Csp1 C terminus and a Profinity Fusion tag introduced at the Csp1 N terminus to enable purification using the Profinity eXact purification method (13,14). Importantly, the Profinity tag is cleaved off during the purification process. The Csp1_1D4-ET construct was introduced into a pET28 plasmid, expressed in E. coli, and purified using the Profinity eXact column (described in the materials and methods and supporting material).

The Csp1_1D4-ET construct was expressed in E. coli with reasonable yield (∼10–15 mg/L culture). Purification using the Profinity approach was rapid (less than 3 h) and resulted in efficient cleavage and removal of the Profinity tag. SDS-PAGE analysis showed the Csp1_1D4-ET protein thus produced was ∼90% pure (Fig. 2 a, inset; analyzed with ImageJ). Subsequent purification of these samples by SEC showed a relatively uniform peak consistent with a size of ∼50 kDa, appropriate for the expected molecular volume for a Csp1_1D4-ET tetramer (Fig. 2 a).

Figure 2.

Csp1_1D4-ET purification, characterization, and cryo-EM structural analysis to 2.98 Å resolution. (a) Csp1_1D4-ET purified using Profinity eXact system (SDS-PAGE inset), then further purified by size-exclusion chromatography (SEC). The SEC trace shows the Csp1_1D4-ET is highly pure, homogenous, and of the expected molecular weight. Gray arrows indicate elution of molecular weight standards. (b) Representative cryo-EM micrograph of Csp1_1D4-ET and three representative 2D class averages (scale bars, 5 nm). (c) (Left ) Tetrameric cryo-EM map of Csp1_1D4-ET solved to 2.98 Å with each subunit colored according to scheme in Fig. 1. (Right) Individual helices from a subunit docked in the cryo-EM map highlighting side-chain densities (PDB: 9E7J). Further details provided in the materials and methods section and the supporting material.

The ability of the purified Csp1_1D4-ET to bind metal ions was assessed by titrating in Cu(I) and monitoring the spectral changes in UV-visible absorption (supporting material, Fig. S1). These experiments showed the purified Csp1_1D4-ET are properly folded, as they readily bound copper (up to >10 Cu(I) ions per Csp1_1D4-ET monomer), consistent with previous reports for wild-type Csp1 (10). Interestingly, we found the Csp1_1D4-ET could still bind copper even though it was exposed to ambient oxygen during purification. Further details are provided in the supporting material.

Negative-stain and cryo-EM analysis of Csp1_1D4-ET

Negative-stain imaging was next used to assess the initial protein quality and concentration. Purified Csp1_1D4-ET was loaded with Cu(I) using ∼15-fold molar excess of metal ion/protein monomer, and any unbound metal ions were then removed with a desalting column. Grids were prepared with standard protocols using 1% uranyl acetate. A small dataset of 25 micrographs were acquired on a FEI Tecnai T12 120kV microscope showing monodisperse Csp1_1D4-ET particles of around 50 Å, consistent with the expected size of a Csp1 tetramer (supporting material, Fig. S2). Additionally, the micrographs were processed with Relion (11) and low-resolution 2D class averages were determined which showed multiple unique orientations (SM Appendix, Fig. S2 inset).

Based on these promising results, we next studied the Csp1_1D4-ET samples by cryo-EM, collecting 7,020 movies on a 300-kV Titan Krios with a GIF and K3 camera. Fig. 2 b shows a representative cryo-EM image and 2D class averages of Csp1_1D4-ET bound with copper. We note that the Csp1_1D4-ET 2D class averages show many distinct views with high-resolution features and match the predicted compact shape of the protein. The most prominent 2D class appears as a pair of high-density lines (presumably a side view showing two of the copper-loaded helical bundles). There is also another distinct class showing a unique-looking diamond-shape with two parallel lines crossing diagonally, forming a high-density parallelogram image (Fig. 2 b, inset). Another distinct 2D class displays two dense lines on top of two round-shaped density (presumably from the side of the protein showing two helical bundles from the top and two helical bundles from the side; SM Appendix, Fig. S3). Finally, there is also a 2D class with four intense dots with some rounded lines of density around those dots (presumably showing the top view of all four-helix bundles; SM Appendix, Fig. S3).

The data were further processed to obtain a 3D reconstruction of the Csp1_1D4-ET structure. To accomplish this, 6,383,264 particles were extracted from 6,888 inspected micrographs and then processed through iterative 2D and 3D classifications. A final particle subset of 125,008 particles was fed into 3D refinement leading to the aforementioned 3D reconstruction at 2.98 Å resolution.

Once the 3D reconstruction was obtained, modeling was performed using the crystal structure of copper-bound Csp1 (PDB: 5FJE) as the initial model. The initial model was docked into the density map and real space refinement was performed using Phenix (15). The quality of the final Csp1_1D4-ET reconstruction and the strong map-to-model fit with clear density for sidechains are shown in Fig. 2 c. The corresponding coordinates were deposited into the PDB (PDB: 9E7J), and the final maps were deposited into the EMDB (Electron Microscopy Data Bank) (EMDB: EMD-47678).

Preparation of Csp1_1D4-ET-1D4 Fab complex and analysis by negative-stain EM

We next tested whether Csp1_1D4-ET could bind to and facilitate structural analysis of another small protein, in this case, a Fab fragment generated from the antibody 1D4. First, we expressed and purified Csp1_1D4-ET in the absence of copper and then confirmed its ability to bind 1D4 Fab with SEC (SM Appendix, Fig. S4A). Next, to prepare complexes with the 1D4 Fab for EM analysis, the apo-state Csp1_1D4-ET was loaded with Cu(I), incubated overnight with 1D4 Fab fragments at a 2:1 M ratio, and then concentrated before further characterization as outlined below.

These Csp1_1D4-ET-1D4 Fab complexes were analyzed by negative-stain EM (SM Appendix, Fig. S4B). The data are consistent with the complex formation observed in the SEC data, with several class averages consistent with the expected size (∼90 Å) and shape of the particle. Together, these data confirmed that the samples were appropriate for further analysis by single-particle cryo-EM.

Cryo-EM analysis of Csp1_1D4-ET-1D4 Fab complex structure

We next carried out large-scale data acquisition of the Csp1_1D4-ET-1D4 Fab complex on grids prepared using the optimized conditions discussed above. Cryo-EM data were collected for 18 h on frozen-hydrated Csp1_1D4-ET bound to the 1D4 Fab using a 300-kV Titan Krios, resulting in 1,926 total movies collected. A representative example of the raw cryo-EM micrograph images highlights monodisperse particles of the correct shape and size (Fig. S4C). The higher electron density of the Cu-bound Csp1_1D4-ET was used to help to identify individual particles and to identify the Csp1_1D4-ET in the 2D class averages.

The cryo-EM data resulted in three high- to medium-resolution maps (Fig. 3). Each map was the result of iterative 2D and 3D classifications followed by different refinement techniques. The first was a 3D reconstruction of the Csp1_1D4-ET-1D4 Fab complex to 5.70 Å (Fig. 3 a). This map used 58,921 particles with CryoSPARC’s nonuniform refinement. The second was a 3D reconstruction of only the Csp1_1D4-ET at 4.09 Å using local refinement and 47,664 particles (Fig. 3 b). The third and final reconstruction was the 1D4 Fab at 5.40 Å using 58,921 particles (Fig. 3 c). Further details are provided in SM Appendix, Figs. S5–S8.

Figure 3.

Cryo-EM analysis of the Csp1_1D4E-ET-1D4 Fab complex. Top panels (a–c) show ribbon representations of the different models docked into their respective maps. (a) Csp1_1D4-ET (PDB: 9E7J) and a Fab fragment (PDB: 1XGY) docked into the cryo-EM map of Csp1_1D4E-ET-1D4 Fab complex. (b) Csp1_1D4-ET (PDB: 9E7J) docked into the locally refined cryo-EM map for Csp1_1D4E-ET. (c) Fab fragment (PDB: 1XGY) docked into the locally refined cryo-EM map corresponding to the 1D4 Fab. Insets show representative 2D class averages. Bottom panels (d–f) show the local resolution plots for each cryo-EM map corresponding to (a–c), colored according to local resolution. More details are provided in the supporting material.

Once the 3D reconstructions were completed for the Csp1_1D4-ET, the 1D4 Fab, and the complex, models generated from cryo-EM analysis of Csp1 (PDB: 9E7J) and crystal structure of a Fab fragment (PDB: 1XGY) (Fig. 3) were docked into the respective cryo-EM reconstructions, with the latter being used as a representative Fab fragment as there is no published structure of 1D4. Since the resolutions of the final reconstructions were not high enough for accurate model building, only the final maps for the Csp1_1D4-ET-1D4 complex, Csp1_1D4-ET, and the 1D4 Fab in the complex were deposited into the EMDB as ascensions 47673, 47674, and 47675, respectively.

Comparison of cryo-EM data for Csp1_1D4-ET in metal- and nonmetal-bound forms

We also attempted to assess if the effect of metal binding on the contrast of Csp1_1D4-ET in the cryo-EM images could be quantified, but the results were somewhat inconclusive. Two approaches were used as described below.

The first involved comparing the analyses of cryo-EM data obtained from Csp1_1D4-ET in the apo-state and Csp1_1D4-ET incubated with Cu(I). We collected a cryo-EM dataset of apo-state Csp1_1D4-ET using the same imaging conditions as used for the Cu-bound Csp1_1D4-ET data shown in Fig. 2. Comparing the raw images of apo-state and Cu-bound Csp1_1D4-ET at the same defocus level, Cu-bound Csp1_1D4-ET appears to have slightly enhanced contrast (SM Appendix, Fig. S9). The apo-state dataset was processed similarly as the Cu-bound dataset (SM Appendix, Fig. S9), and we were able to obtain a reconstruction of apo-state Csp1_1D4-ET at 3.55 Å (SM Appendix, Fig. S9). Similar numbers of particles were used for reconstruction of apo-state (110,553 particles) and Cu-bound (125,008 particles) Csp1_1D4-ET. This implies that the difference in resolution between the final reconstructions is likely due to the enhanced contrast from the bound copper ions.

The second approach involved preparing Csp1_1D4-ET-1D4 Fab complexes both with and without Cu(I), using the same experimental and imaging conditions as above, and then subjecting them to cryo-EM analysis. This approach enabled us to directly compare Cu-bound and apo-state Csp1_1D4-ET in the complexes, since the intensity of the Fab in the complexes is unchanged, and, thus, the 1D4 Fab acts as an internal control between datasets and overall ice thicknesses.

Specifically, we collected 752 movies of the Cu-bound Csp1_1D4-ET-1D4 Fab complex and 3,431 movies of the apo-state Csp1_1D4-ET-1D4 Fab complex on a 300-kV Titan Krios with a GIF and K3 DED. In total, 177,373 Cu-bound Csp1_1D4-ET-1D4 particles and 414,673 particles of apo-state Csp1_1D4-ET-1D4 particles were picked and analyzed to generate 2D class averages. Importantly, the same number of particles per class and orientation across datasets were used for this analysis to enable direct comparison between datasets. Example raw images for Cu-bound Csp1_1D4-ET-1D4 complexes and for the apo-Csp1_1D4-ET-1D4 Fab complexes as well as representative 2D classes used for the attempted quantitative analysis are shown in the SM Appendix (Fig. S10).

Interestingly, Csp1_1D4-ET can be easily identified in these 2D class averages, as it exhibits the same compact, parallelogram shape it shows in the Csp1-alone data (compare 2D class averages in Fig. 2 with those in Fig. 3). The rough outline of a Fab fragment is also apparent in some class averages. Qualitatively, Cu-bound Csp1_1D4-ET exhibits more contrast than apo Csp1 in the micrographs. Similarly, in some 2D classes, the Cu-bound Csp1_1D4-ET shows a greater difference in its intensity versus the bound Fab, compared to the Fab bound to apo-Csp1_1D4-ET. These intensity differences are difficult to quantify definitively, due to the normalization and scaling of particles involved in cryo-EM processing algorithms (Fig. S10). However, we qualitatively compared the intensity differences as follows. Three similar-looking representative 2D classes were chosen for both the apo unbound and Cu-bound Csp1_1D4-ET-1D4 datasets. We then split each of the 2D classes into subsets containing the same number of particles, in this way generating replicate 2D class averages for each chosen class. The resulting 2D classes were analyzed as described in Fig. S5. This limited analysis suggests that, when compared to the bound 1D4 Fab, the intensity of the Cu-bound Csp1_1D4-ET is greater than for the apo Csp1_1D4-ET (p < 0.05 for all three 2D classes chosen for analysis; see SM Appendix, Fig. S10). As the 1D4 Fab was the same in both datasets, this intensity difference is likely due to the enhanced scattering by the densely packed copper ions in the Cu-bound Csp1_1D4-ET. The implications of this observation are covered in the discussion.

Discussion

Our goal in this work was to test whether the copper-binding protein Csp1 could be used as a GFP equivalent for single-particle cryo-EM protein structural studies. Our data suggest that Csp1 shows promise in this role—it is a small, compact protein that is readily apparent in cryo-EM images. We also find that Csp1 is an easily modified protein scaffold that, with further optimization, could potentially be used to help analyze the structure of a bound protein. Insights gained from these studies are briefly reviewed below, as well as future areas for improvement.

Expression, purification, and structural analysis of Csp1_1D4-ET by EM

We find that Csp1 is relatively easy to express and purify. Somewhat unexpectedly, we also find it is very stable and can even be purified under ambient (aerobic) conditions while still binding close to the maximal number of Cu(I) ions (SM Appendix, Fig. S1). This finding is noteworthy, as it means complicated and cumbersome anaerobic conditions are not required when using Csp1 for cryo-EM studies. It also suggests that Csp1 in the absence of metal ions is rigid enough to keep most of the cysteine residues in the metal-binding region from interacting and forming irreversible disulfide bonds.

Interestingly, the compact structure of Csp1 is readily observed in both negative-stain and cryo-EM images made with standard techniques, and its apparent enhanced contrast helped particle picking (Fig. 2B). Importantly, we did not need to use a phase plate, graphene oxide coating, or any specialized grid-making technologies to visualize Csp1_1D4-ET. The 2D class averages generated from the cryo-EM data are especially intriguing, as many have a unique, compact shape that roughly looks like a parallelogram (see insets in Figs. 2 and 3 c). We do notice a minor preferred orientation of Csp1_1D4-ET in the 2D classes, with about three times more particles of Csp1_1D4-ET appearing oriented in a side view with two helical bundles vs. in a top view with the diamond shape. Additionally, we attribute the low percentage of particles in the final reconstruction to protein that associates with the air-water interface given its low molecular weight. Both of these observations are compensated in the final 3D reconstruction by the initial yield of 927 particles per micrograph, even at high magnification.

We were able to determine the ∼50 kDa structure of Cu-bound Csp1_1D4-ET to 2.98 Å from our dataset of Csp1_1D4-ET construct alone (6,888 selected movies and a final stack of 125,008 particles). In this small protein, high-resolution cryo-EM reconstruction (Fig. 2), the tetrameric nature of Csp1_1D4-ET is clearly evident and we can see interactions between each of the four helical bundles. Additionally, there is density in the center of each four-helix bundle mostly near cysteine residues (SM Appendix, Fig. S11). This density is presumably due to the copper ions but is poorly resolved, most likely due to differing numbers of copper ions in varying positions per Csp1_1D4-ET monomer that are averaged out in the consensus cryo-EM reconstruction. As it was not the focus of this study, we did not try to resolve the location of the individual coppers in the electron-density map or final structure. However, the Csp1_1D4-ET protein is well resolved in our cryo-EM map, with clear sidechains evident. Modeling of our cryo-EM reconstruction yields a Csp1_1D4-ET structure that is very similar to the original Csp1 structure determined by crystallography (PDB: 5FJE), with only a root-mean-square deviation (RMSD) 0.432-Å difference between the two (as estimated in Chimera (16)).

Csp1_1D4ET-1D4 Fab complex formation and identification in EM images

Our tests to see whether Csp1 could help determine the structure of other proteins are promising but still not definitive. Specifically, they show that Csp1 can be modified at its C terminus with an antibody-binding epitope to which Fab fragments can be bound, and the complex can be visualized by EM methods. In principle, this approach could be broadened to other proteins and binding motifs, as discussed later.

We purposely used conditions that favored only one Fab binding per Csp1_1D4-ET tetramer (although up to four Fabs could theoretically bind), because of concerns that flexibility in the 1D4-epitope/linker region could prove destructive to high-resolution refinement with multiple-bound Fab particles. If flexibility in the linker region could be reduced and thus more Fabs bound per Csp1, it is likely even better results could be obtained by exploiting the higher order D2 symmetry of Csp1.

Analysis of 1D4 structure and insights gained about epitope binding

Our original goal was to use the structure of the Csp1_1D4-ET-1D4 Fab complex to gain insights into how 1D4 interacts with its epitope, the 9-amino acid sequence at the extreme C terminus of the visual GPCR (G protein coupled receptor) rhodopsin. The 1D4 epitope is relatively uncommon and unique to visual proteins, and thus it has been used in a number of immunochemical approaches ranging from western blots to imaging and immuno-purification of fusion proteins (17,18).

Thus, we collected and analyzed an extensive cryo-EM dataset for the Cu-bound Csp1_1D4-ET-1D4 complex to determine a 3D structure of the complex. Different refinement tools were used for each of the three maps resolved from this dataset, depending on the signal intensity and flexibility of the protein. To obtain the full map of Csp1_1D4-ET –1D4 complex, nonuniform refinement proved the best compromise between the strong signal of Csp1 and the weaker signal of the 1D4 Fab. The robust signal of Csp1_1D4-ET enabled local refinement of the section that greatly improved resolution, whereas the weaker signal of the 1D4 Fab paired with local refinement minimally improved the map resolution and quality. In the future, newly developed flexible refinement algorithms (19,20) could improve the quality and resolution of the inherent flexibility caused by two linked proteins.

The resulting map of the Csp1_1D4-ET-1D4 Fab complex is shown in Fig. 3. Regrettably, we were only able to obtain medium-resolution structures for the complex (5.7 Å) and for the Fab fragment (5.4 Å). We attribute these lower-resolution reconstructions to the fact that we did not attempt to optimize the rigidity of the introduced epitope sequence and linker. As the structure of the 1D4 antibody has not been published, we could not compare it with our map here. However, the “classical” Fab form is readily apparent in the reconstruction, with varying resolutions apparent in the local resolution as one extends further away from the epitope-binding region (Fig. 3). Because of this, although it is possible to see where the epitope continues from the C- terminus of Csp1_1D4-ET into the base of the 1D4 Fab, at our current resolution, it is not possible to observe the key residues in the 1D4 epitope identified by previous alanine scanning studies of Molday and colleagues (17,21). Together, these results point out the current limitations of using Csp1 for cryo-EM analysis of protein structures.

Cu(I) binding enhances Csp1 analysis in cryo-EM analysis but the extent it does so is difficult to quantify

Importantly, our studies suggest that copper binding does enhance Csp1 visibility in cryo-EM images, although the extent to which it does so is still not defined. For example, Cu-bound Csp1 shows higher apparent intensity in 2D class averages, and its structure can be resolved to higher resolution (compare Figs. S5–S8). Similarly, Cu-bound Csp1_1D4-ET shows an apparent greater difference in intensity compared to the bound 1D4 Fab (an internal control) than does the apo Csp1_1D4-ET in some 2D class averages (SM Appendix, Fig. S10). Together, this apparent increase in intensity of Cu-bound Csp1_1D4-ET and its unique, compact shape help identify Csp1 in complexes with other proteins in class averages, even when both proteins are similarly sized, such as the ∼50 kDa Csp1_1D4-ET vs. the ∼40 kDa 1D4 Fab.

However, determining exactly how much the bound Cu(I) enhances Csp1 in cryo-EM images is challenging. Differences in ice thickness between sample grids can impact the comparison of apparent particle intensities. Moreover, picking algorithms select the best-defined particles, and thus may miss less “intense” apo-Csp1_1D4-ET particles and thus skew an analysis comparing 2D class averages. We also cannot ensure that the apo-Csp1_1D4-ET samples analyzed are completely Cu(I) free, given the high affinity of Csp1 for Cu(I) (10), and the fact that Cu-grids were used in these EM studies.

Conclusions

In summary, our results show Csp1 is an excellent prospect for use as a tag and/or platform for protein-based cryo-EM studies, for several reasons.

First, Csp1 is a very compact, stable protein that is easy to work with and is readily modified and seen on cryo-EM grids. Even without bound metals, its apo form can be identified in cryo-EM images, and the binding of Cu(I) to Csp1 further increases its visibility. Moreover, both the Csp1 N and C termini are exposed in its structure and thus can potentially be modified to introduce tags or even other proteins. Our data show both regions tolerate modification, as we introduced the 1D4 epitope tag on the C terminus and the Profinity tag on the N terminus for purification. Importantly, the termini on each monomer unit of the Csp1 tetramer are on maximally distant corners, leaving potential space for the introduction of other tagged proteins.

Second, Csp1 binds Cu(I) with high affinity (10), suggesting the bound metals have a slow off rate and thus will remain predominantly bound after loading onto the protein (a practical consideration for use as a tag). Moreover, previous studies have shown Csp1 can also bind other metals (22,23), opening the possibility of using other, even more electron-dense metals such as silver or gold.

Finally, as a potential platform for EM studies, Csp1 has a relatively high symmetry operator (D2) for its small size (∼12.6 kDa/monomer, ∼50 kDa per tetramer). This high symmetry potentially allows for a fourfold increase in the number of useful monomeric symmetrically related subunits that could be used in refinement and classification. Since detector sizes are fixed in EM instruments, a small dispersed protein like Csp1 can populate each high-magnification micrograph with a very high number of particles (compared to large proteins). This, in principle, could potentially yield thousands of particles per image at high magnifications, leading to very large particle sets that yield high-resolution structures, although potential caveats would need to be addressed, as noted below.

Areas for improvement and future directions

Our results show Csp1 is a promising candidate for facilitating protein analysis by cryo-EM methods, but further optimization is clearly needed before it can be used as a general-purpose tag. Some key obstacles as well as potential solutions and promising areas of future study are noted below.

Minimizing the flexibility between the epitope/linker on Csp1 and the target molecule must be done to enable high-resolution reconstructions and detailed structural analysis (we think linker flexibility likely contributed to our lower-resolution reconstructions of the Csp1_1D4-ET-1D4 complex). Possible ways to achieve this include better-engineered linkers or epitope sequences, the use of disulfide-constrained Fabs (24), and the use of flexible refinement algorithms during structural analysis of the bound fragment. Successful use of these strategies would also enable Csp1 to be used as a fiducial or platform for analyzing other proteins after determining whether the fused proteins impact copper entrance into Csp1.

Optimized fusion-protein strategies that limit linker flexibility would also enable the D2 symmetry of Csp1 to be exploited, as discussed above. However, for studies using binding epitopes fused to Csp1, the tetrameric aspect of Csp1 raises another problem—the risk of heterogeneity in the binding stoichiometry for the target protein, which would reduce the number of particles suitable for analysis. Symmetry expansion of the cryo-EM data could potentially overcome some of the induced heterogeneity, but working with even smaller particles might complicate analysis. However, in some cases, higher-affinity-binding tags could help overcome problems with binding-induced heterogeneity. Alternatively, monomerized Csp1 could be used, but these would have fewer copper atoms bound (13 maximum), and thus it is not clear if the reduced Cu load would still be sufficient for contrast enhancement (our data demonstrating that copper binding enhances contrast are still qualitative). Nonetheless, monomeric Csp1 warrants further investigation as a cryo-EM tag.

Finally, we think the higher electron density of metal-bound Csp1 could be useful in other structural approaches, such as the growing field of ptychographic hard X-ray tomography (25,26), a novel method that enables imaging with quantitative electron-density contrast.

Acknowledgments

This work was funded by NIH grant R01 EY029343A. A portion of this research was supported by NIH grant U24GM129547 and performed at the PNCC at OHSU and accessed through EMSL (grid.436923.9), a DOE Office of Science User Facility sponsored by the Office of Biological and Environmental Research. A portion of this research was also facilitated by the OHSU Biophysics Shared Resource Core (RRID: SCR_022744), and the OHSU Multiscale Microscopy Core (RRID: SCR_009969). We thank Dr. Eric Gouaux, Dr. Steven E. Mansoor, and Dr. Craig Yoshioka for reading and commenting on this manuscript, as well as Dr. James Chen for help in the early stage of this project.

Author contributions

A.C.O., W.Y., and D.L.F. designed the study, analyzed the results, and wrote the manuscript. A.C.O. and W.Y. conducted the experiments.

Declaration of interests

The authors declare no competing interests.