Quantitative Analysis and Optimization of Site-Specific Protein Bioconjugation in Mammalian Cells

Abstract

Despite a range of covalent protein modifications, few techniques exist for quantification of protein bioconjugation in cells. Here, we describe a novel method for quantifying in cellulo protein bioconjugation through covalent bond formation with HaloTag. This approach utilizes unnatural amino acid (UAA) mutagenesis to selectively install a small and bioorthogonally reactive handle onto the surface of a protein. We utilized the fast kinetics and high selectivity of inverse electron-demand Diels–Alder cycloadditions to evaluate reactions of tetrazine phenylalanine (TetF) with strained trans-cyclooctene-chloroalkane (sTCO-CA) and trans-cyclooctene lysine (TCOK) with tetrazine-chloroalkane (Tet-CA). Following bioconjugation, the chloroalkane ligand is exposed for labeling by the HaloTag enzyme, allowing for straightforward quantification of bioconjugation via simple western blot analysis. We demonstrate the versatility of this tool for quickly and accurately determining the bioconjugation efficiency of different UAA/chloroalkane pairs and for different sites on different proteins of interest, including EGFP and the estrogen-related receptor ERRα.

Graphical Abstract

INTRODUCTION

Covalent protein modification, or protein bioconjugation, is a method used for many applications, including protein immobilization, protein PEGylation, antibody–drug conjugation, and protein labeling with fluorophores or biophysical probes.^1–6 With a wide range of substrates, protein bioconjugation has been used for drug discovery and delivery,^5,7,8 protein imaging,^2,7,9 biochemical and biophysical assay development,³ and the study of protein function in native environments.^3,10,11 Bioconjugation techniques often require significant target manipulation, which limits analysis of reactive partners in live cells.¹² Fast, robust, and specific reactivity is critical for the real-time observation of protein bioconjugation in live cells due to the low concentration of reactants.^1,13 A minimally disruptive method for bioconjugation is the site-specific installation small reactive handles via genetic code expansion with unnatural amino acids (UAAs).^14,15

Genetic code expansion is a well-developed and versatile methodology with a wide variety of applications.¹⁶ It has been used in a range of organisms, including bacterial and mammalian cell culture, as well as worm, fly, and fish models.^17,18 UAA incorporation is enabled by an orthogonal aminoacyl-tRNA synthetase and tRNA_CUA pair, which is evolved from non-endogenous cellular machinery. While there are more than a dozen orthogonal pairs developed for genetic code expansion, orthogonality is dependent on the host organism. Namely, the Escherichia coli tyrosyl- and leucyl-tRNA synthetases have been evolved for use in eukaryotic cells, while Methanosarcina barkeri and Methanosarcina mazei pyrolysyl-tRNA synthetases are orthogonal in both eukaryotic and bacterial cells.^19,20 In mammalian cells, the amber stop codon-modified gene of interest is delivered with the necessary orthogonal machinery most often through transient transfection. Multiple plasmid systems have been developed and optimized for incorporation efficiency through the adjustment of factors such as tRNA copy number and promoters.²¹ Cell lines have been generated for stable expression of orthogonal machinery,^22,23 or a mutant UAG release factor, to suppress the amber stop codon without significantly affecting the opal or ochre stop codons.²⁴ Strategic UAA placement enables a high degree of selectivity with regard to the protein of interest (POI) as well as the specific site of bioconjugation.¹⁰ The native protein conformation and functionally relevant posttranslational modifications are preserved by the endogenous expression of proteins bearing reactive handles of a few hundred Da.²⁵

UAAs have been developed for bioorthogonal labeling reactions that employ azide and alkyne reactive handles for copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC), azide and phosphine handles for Staudinger ligation, azide and strained alkyne handles for strain-promoted azide-alkyne cycloaddition (SPAAC), or tetrazine and dienophile handles for inverse electron-demand Diels–Alder (IEDDA).^26–29 IEDDA reactions have become the preferred methods for labeling biomolecules in complex physiological conditions due to their fast kinetics, high conversion rates, excellent bioorthogonality, and nontoxic reaction components.²⁸ While SPAAC^30,31 and IEDDA cycloadditions^30–36 have been utilized in both E. coli and mammalian cells, IEDDA reactions have the advantage of further enhanced reaction kinetics. Among the IEDDA-compatible UAAs genetically encoded in cells, lysine derivatives with trans-cyclooctene (TCOK^30,34,37 or TCO*K^37,38), bicyclononyne (BCNK),^30,31 norbornene (NorK),^32–34 and strained cyclooctyne (SCOK)^34,35,37 are commonly encoded dienophiles, while tetrazine phenylalanine (TetF) is a suitable tetrazine UAA.^36,39

Qualitative evaluation of protein bioconjugation has traditionally been achieved by utilizing fluorescent labels and monitoring the reaction through imaging.²⁶ Building on this, FRET has been cleverly adapted to determine the intracellular kinetics for common bioconjugation reactions.¹⁵ However, the quantification of cellular biomolecule labeling has been limited to in-gel fluorescence measurement of fluorophore conjugates (e.g., using a TAMRA label) relative to the fluorescence of the total protein (e.g., GFP^UAA).^39,40 These techniques are limited by the permeability and intracellular distribution of the labeling fluorophore, as well as the use of fluorescent proteins as cellular targets.^41–43 The field of protein bioconjugation via UAA mutagenesis would benefit from the development of a quantitative, readily performed method using only basic equipment available in any lab for the analysis of intracellular bioconjugation efficiency in a protein-specific manner. To address this need, we have developed a methodology utilizing HaloTag for the quantification of the intracellular bioconjugation reaction between an IEDDA-compatible UAA and its chloroalkane-derivatized conjugate.

HaloTag has previously been used to quantify peptide cell permeability via the chloroalkane penetration assay, which has since been established as the “gold-standard” of permeability assays.^44–46 This methodology takes advantage of the low reactivity and excellent cell permeability of the chloroalkane functionality, specifically in contrast to fluorescent dyes.^{41–43,47,48} Based on these validated properties, we have designed a chloroalkane-based assay for the quantification of protein bioconjugation reactions using UAAs in live cells. Following bioconjugation to the UAA-containing protein, the chloroalkane handle is exposed through cell lysis to mediate covalent bond formation with recombinant HaloTag enzyme (Figure 1). This protein fusion enables the straightforward and direct quantification of the extent of bioconjugation via gel analysis while avoiding issues of probe permeability, protein stability, and off-target reactivity. This methodology requires no specialized equipment, allows for the quantitative comparison between different conjugation reactions (to different UAAs), and enables the assessment of different UAA incorporation sites in any POI, regardless of its fluorescent properties, directly in a cellular environment.

Figure 1.

Experimental workflow to evaluate intracellular protein bioconjugation through the covalent modification of an installed chloroalkane handle with the HaloTag protein.

RESULTS AND DISCUSSION

To enable the accurate quantification of in cellulo protein bioconjugation, we designed an assay using Western blot analysis of conjugate formation. We hypothesized that bioconjugation efficiency could be determined through site-specific installation of a reactive residue on a POI, followed by intracellular ligand-mediated protein labeling with a chloroalkane (CA) handle. Upon cell lysis, ternary complex formation between the ligand-modified POI and HaloTag will differentiate the pool of labeled protein and enable its quantification using simple and accessible techniques (Figure 1).

We first attempted to establish the utility of this methodology using an mCherry-EGFP-HA fusion protein for efficient protein expression and Western blot detection. We chose to test the bioconjugation of the TetF/sTCO pair due to its efficient encoding, good stability, and high labeling rate.³⁹ The IEDDA-reactive TetF was genetically incorporated at surface residue Y151 (pmCherry-EGFP-HA-Y151TAG). The cellular toxicity of both the TetF amino acid and the chloroalkane-modified reactive dienophile (sTCO-CA) were tested prior to use in assessing protein bioconjugation (Supporting Figure S1). The IC₅₀ values determined for TetF and sTCO-CA incubation in HEK293T cells over the course of 72 h were 753 μM and 6.32 μM, respectively. TetF incorporation experiments were conducted with 250 μM of UAA to achieve good incorporation efficiency while minimizing toxicity, as was observed through the titration of TetF for mCherry-TAG-EGFP expression (Supporting Figure S1). While the toxicity of sTCO-CA is apparent during multiday exposures at low micromolar concentration, the tetrazine/sTCO IEDDA reaction is fast,³⁹ requiring only very short (minute to hour time scale) incubation periods and no toxicity was observed due to the 4 h treatment (Supporting Figure S1). We therefore used sTCO-CA concentrations between 1 and 500 μM.

HEK293T cells expressing mCherry-EGFP-HA-Y151TetF were exposed to sTCO-CA to determine the lowest ligand concentration suitable for bioconjugation (Figure 2A). Upon incubation of the cells with or without sTCO-CA (1–500 μM for 4 h) and two subsequent ligand washout steps, the cells were lysed, exposing the bioconjugated protein and thus the chloroalkane handle. Cell lysates were then incubated with recombinant HaloTag⁴⁹ protein (20 μM) to yield the resulting high molecular weight protein dimer. Unreacted POI and protein conjugate were visualized through anti-HA Western blot, and the extent of bioconjugation was determined through quantification of the POI band intensity relative to the GAPDH loading control (Figure 2B).^50,51 Quantification values were normalized to the vehicle control (absence of sTCO-CA). Near-complete bioconjugation was observed with only 1 μM of sTCO-CA, indicating the high reactivity of the IEDDA pair, and no statistical difference was observed with further increasing ligand concentrations. An sTCO-CA concentration of 10 μM was used to ensure consistent and complete bioconjugation in all subsequent experiments. Furthermore, this concentration of sTCO-CA did not show any cell toxicity after a 4 h incubation. To determine the rate of intracellular bioconjugation between sTCO-CA and mCherry-EGFP-Y151TetF, the POI was expressed in HEK293T cells and incubated with sTCO-CA (10 μM) for up to 2 h. The assay was conducted as previously described, revealing near-complete bioconjugation within 1 h of sTCO-CA exposure (Figure 2C).

Figure 2.

Quantitative evaluation of TCO-chloroalkane bioconjugation to mCherry-EGFP Y151TetF in HEK293T cells. (A) EGFP-TetF bioconjugation to sTCO-CA and chemical structures of TetF, sTCO-CA, and the small molecule conjugate. Western blot and bar graph quantification of conjugate formation following mCherry-EGFP-HA-Y151TetF incubation with (B) increasing concentrations (0–500 μM, 4 h) or (C) increasing incubation times (10 μM, 0–4 h) of sTCO-CA, followed by overnight incubation with HaloTag. **** p < 0.0001, *** p = 0.0005, determined by one-way ANOVA of biological triplicates.

The ligand washout step was observed to be important for maximizing HaloTag labeling efficiency. For cells expressing mCherry-EGFP-Y151TetF, treatment with sTCO-CA (10 μM for 2 h) followed by three consecutive 15-min incubations with fresh media revealed significantly decreased HaloTag labeling efficiency, suggesting incomplete removal of the chloroalkane from the cellular environment. Thus, we adjusted the sTCO-CA washout step to a convenient 2-h incubation of cells in fresh media, which yielded greatly improved HaloTag conjugation after lysis (Supporting Figure S3). Additionally, HaloTag and HaloTag/POI conjugates formed two distinct bands instead of the expected single band. Recently, a disulfide bond between two internal cysteines in HaloTag7 has been identified as the cause.⁵² Recombinant HaloTag protein was expressed and purified regularly (approximately one month shelf life) to reduce the prominence of double band formation (Supporting Figure S4).

We next aimed to demonstrate the adaptability of this methodology for assessment of different UAA/chloroalkane pairs. Using the same pmCherry-EGFP-Y151TAG reporter construct in HEK293T cells, we transiently expressed the protein bearing trans-cyclooctene-lysine (TCOK). TCO was selected for its synthetic accessibility, despite its slightly less favorable stability in the presence of thiols when compared to TCO*A.³⁷ Following incubation for 36 h, we tested subsequent bioconjugation with the corresponding tetrazine-chloroalkane ligand (Tet-CA, Figure 3A). Cells were incubated with or without the Tet-CA ligand (0.1–100 μM) for 8 h to observe the lowest allowable concentration yielding maximal bioconjugation. Following overnight incubation with HaloTag and Western blot analysis of conjugate formation, a limited extent of bioconjugation was observed (Figure 3B). Bioconjugation efficiency was maximized with 1 μM Tet-CA, resulting in nearly 50% reduction in POI band density. We hypothesized that the reduced efficiency of TCOK is due to the instability of the trans isomer of cyclooctene under physiological conditions. Previous studies have identified the half-life (t_1/2) of a PEG-TCO derivative in mouse serum to be a matter of hours,⁵³ and a fluorophore-modified TCO displayed a maximum labeling efficiency of 60% with its tetrazine pair in cells. While silver complexation of TCO (Ag-TCO) ligands has been proven to stabilize the trans conformation in storage,⁵⁴ we found the use of Ag-TCOK and Ag-sTCO-CA failed to increase bioconjugation efficiency, and had the added effect of inducing cytotoxicity at higher concentrations (Supporting Figure S2).

Figure 3.

Quantitative evaluation of Tet-CA bioconjugation to mCherry-EGFP-Y151TCOK. (A) Cartoon depicting TCOK bioconjugation to Tet-CA and chemical structures of TCOK, Tet-CA, and the small molecule conjugate. Western blot and bar graph quantification of conjugate formation following mCherry-EGFP-Y151TCOK incubation with (B) increasing concentrations (0–100 μM, 8 h) or (C) increasing incubation times (10 μM, 0–4 h) of Tet-CA, followed by overnight incubation with HaloTag. * p = 0.0363, *** p < 0.001, **** p < 0.0001, determined by one-way ANOVA of biological triplicates.

We next chose to determine the shortest incubation time necessary for maximum observable protein bioconjugation. Due to the limited bioconjugation efficiency by intracellular TCOK isomerization, we decreased the POI expression time to limit trans-to-cis isomerization of TCOK. HEK293T cells transiently expressed mCherry-EGFP-Y151TCOK for 16 h, upon which time the Tet-CA ligand was applied at the optimal concentration (10 μM) for up to 2 h (Figure 3C). POI band density was decreased to about 15% of the Tet-CA control within 1 h of ligand exposure.

To further demonstrate the adaptability of our method for direct comparison of bioconjugation efficiency at different UAA incorporation sites, we generated two additional TAG mutants of the mCherry-EGFP fusion protein (Figure 4A). The I128 site was chosen due to its protein surface accessibility and lack of impact on EGFP folding and fluorescence. In contrast, the G134 residue is located in an unstructured stretch of amino acids and is therefore afforded greater flexibility compared to the Y151 and I128 mutations. The mCherry-EGFP-HA-I128TetF, -G134TetF, and -Y151TetF constructs were expressed in HEK293T cells and incubated with sTCO-CA using increasing concentrations (0–100 μM) for 1 h. The cell lysates were incubated overnight with HaloTag, followed by Western blot analysis of conjugate formation. Western blot results are shown for the optimal concentration, which was revealed to be 10 μM sTCO-CA for all three constructs (Figure 4B). Full blots are shown in Supporting Figure S5. Near-complete bioconjugation was observed between sTCO-CA and the I128TetF and Y151TetF constructs. The mCherry-EGFP-HA-G134TetF mutant demonstrated around 60% decrease in band density relative to the unconjugated control, suggesting a significant difference in the accessibility of G134TetF. TCO/tetrazine reactivity has previously been observed to be significantly restricted by the local hydrophobic environment. Not only can UAA accessibility be masked by surrounding hydrophobic residues,⁵⁵ but TCO ligand reactivity can be impacted by interactions with proximal protein domains.⁵⁶ It is possible that the enhanced flexibility of the region of TetF incorporation enables interior burying of the hydrophobic UAA, reducing reactivity.

Figure 4.

Quantitative evaluation of TCO-chloroalkane bioconjugation to EGFP protein surface mutations. (A) Crystal structure of EGFP depicting I128, G134, and Y151 residues. PDB: 2Y0G. (B) Western blot and bar graph quantification of conjugate formation following sTCO-CA incubation (10 μM, 1 h) with I128TetF, G134TetF, and Y151TetF mCherry-EGFP-HA mutants and overnight incubation with HaloTag. **** p < 0.0001, determined by two-way ANOVA of biological triplicates.

Given the success of this approach in monitoring conjugation efficiency at various amino acid positions on EGFP, we next chose to demonstrate its general applicability in quantifying bioconjugation to another POI. The EGFP protein structure is comprised of a central α-helix surrounded by a set of 11 β-sheets, lending to a highly hydrogen bonded and inherently stable protein conformation.⁵⁷ For an alternative target, we aimed to use a protein with a more complex tertiary structure and greater conformational flexibility. We therefore chose to target the nuclear estrogen-related receptor alpha, or ERRα, specifically in a region that is predominantly comprised of layered α-helical structures.⁵⁸ We hypothesized that while the protein offers a more challenging target, our optimized method would allow for the quantification of small molecule bioconjugation to multiple unique ERRα mutants. Three sites were chosen for comparison of bioconjugation efficiency, including surface residue V321 and hydrophobic pocket residues F382 and F495 (Figure 5A).^58,59 The F382 residue is located in a relatively open pocket, while F495 is spatially constricted by surrounding hydrophobic residues.

Figure 5.

Quantitative evaluation of TCO-chloroalkane bioconjugation to ERRα mutants. (A) Crystal structure of the ERRα ligand binding domain depicting V299, F382, and F495 residues. PDB: 3K6P. (B) Western blot and bar graph quantification of conjugate formation following ERRα-HA-V321TetF, -F382TetF, and -F495TetF incubation with sTCO-CA (1 h) and overnight incubation with HaloTag. V321TetF was incubated with 10 μM of sTCO-CA; F382TetF and F495TetF were incubated with 100 μM of sTCO-CA. **** p < 0.0001 determined by two-way ANOVA of biological triplicates.

ERRα⁶⁰ was expressed with an HA tag for Western blot analysis of POI and conjugate protein levels. As previously described, HEK293T cells transiently expressing ERRα-HA-V321TetF, -F382TetF, or -F495TetF were incubated with increasing concentrations of sTCO-CA (0–100 μM) for 1 h, followed by cell lysis, HaloTag conjugation, and Western blot analysis. ERRα-HA-F382TetF demonstrated the highest conjugation efficiency in the presence of 10 μM ligand, matching previous EGFP results. However, both V321TetF and F495TetF mutants displayed the greatest conversion with 100 μM sTCO-CA (Figure 5B, full blots in Supporting Figure S6). Interestingly, sTCO-CA incubation with internal residue F382TetF resulted in excellent (~90%) bioconjugation, while the surface residue V321TetF showed moderate (~70%) conversion to the conjugate. ERRα-HA-F495TetF displayed poor bioconjugation efficiency, with only ~40% of the protein being conjugated. Similar to our rationale for the reduced labeling efficiency of mCherry-EGFP-HA-G134TetF, this might be attributed to the localized hydrophobicity of the area surrounding V321 and F495. While all three sites are relatively close to each other, F495 in particular is surrounded by a number of hydrophobic residues and V321 lies on the exterior of this hydrophobic core.^58,61 It is possible that TetF is thus shielded from sTCO-CA docking within the binding pocket, resulting in a lower conjugation efficiency. Overall, these results support the utility of our bioconjugation quantification approach in assessing the site-specific labeling efficiency in completely different proteins.

CONCLUSION

We developed a quantitative, Western blot-based methodology for the assessment of efficient protein bioconjugation in live mammalian cells. This was achieved through the site-specific incorporation of an inverse electron-demand Diels–Alder-reactive UAA at different sites on the surface of different proteins, followed by in-cell reaction with a bioconjugation-compatible, chloroalkane-derivatized ligand. Cell lysis followed by HaloTag protein addition then led to protein dimer formation with the chloroalkane-modified target. The resulting molecular weight shift enabled quantification of protein conjugation by a simple Western blot. We evaluated two bioreactive pairs, TetF/sTCO and TCOK/Tet, using EGFP as an initial proof-of-concept. Near-quantitative bioconjugation was observed for the in-cell reaction between mCherry-EGFP-HA-Y151TetF and sTCO-CA (10 μM) in just one hour. In contrast, Tet-CA conjugation to the TCOK-modified protein plateaued at around 60% conjugation, potentially due to the isomerization of the UAA to the cis isomer under physiological conditions over the course of the 36-h protein expression. Reducing the protein expression time to 16 h significantly improved bioconjugation efficiency, reaching nearly 85% conversion within an hour of Tet-CA application (10 μM). However, incorporation of the more stable tetrazine handle into the protein of interest followed by conjugation with the more reactive sTCO ligand overall yielded improved results.

Upon demonstrating the applicability of this methodology in assessing bioconjugation reactions, we aimed to next determine the potential impact of different sites of TetF incorporation on bioconjugation efficiency. Out of three EGFP mutants, I128TetF, G134TetF, and Y151TetF, only reaction at the I128 and Y151 sites resulted in excellent conversion to the conjugate. Reaction of the ligand with G134TetF yielded only around 60% bioconjugation. While the reasons for this reduced reactivity are unclear, it highlights the necessity of quantitatively assessing multiple different sites independently, which is uniquely enabled by the complete residue specificity provided by unnatural amino acid mutagenesis.

Given the high rigidity of EGFP due to its β-barrel structure, we next chose to evaluate the bioconjugation efficiency at different locations on a protein with a fundamentally different fold, alpha-helices rich ERRα. We selected three sites within relative proximity of each other, but with varying degrees of protein surface accessibility. The first residue, V299, resides at the protein surface, while the second and third, F382 and F495, are located within a hydrophobic site. Surprisingly, F382 displayed the highest reactivity, followed by the protein surface modification V299TetF. Bioconjugation of the F495TetF mutant was limited, failing to exceed more than 60% conversion. These observations again support the need to explore multiple sites, but also highlight the difficulties of predicting site reactivity based on structural data for a given protein.

The spatial precision enabled through genetic insertion of a reactive handle, combined with our straightforward method for bioconjugation quantification, will allow for direct and facile analysis of protein site reactivity and conjugate pair compatibility. Western blot is a common and established biochemical technique, making our approach highly accessible to any lab using a mammalian cell model. Due to the high membrane permeability of the chloroalkane tag, the bioconjugation efficiency will directly reflect the reactivity of the conjugate pair, without interference by other cargo, such as fluorescent dyes. While we deliberately designed the reagents used here to be cell permeable, future probes (e.g., fluorophores) can be readily evaluated for both their cell permeability and their ability to undergo protein bioconjugation in the same experiment. Overall, we envision that this method will be valuable for determining ideal protein conjugation sites and the validation of future protein bioconjugation chemistries. The high selectivity of bioorthogonal ligands combined with the reliable reactivity of chloroalkanes further allows for optimization of this methodology for different and more sensitive protein detection techniques, such as ELISA or immuno-PCR.⁶² Furthermore, the intracellular assembly of covalent protein dimers, and potentially higher order structures, offers future opportunities for biomolecule targeting.^63–65 Given the extensive precedence for genetic code expansion in animals, we believe this technique could be easily expanded into mouse, fly, fish, and worm systems.¹⁷

Supplementary Material

ASSOCIATED CONTENT

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.bioconjchem.2c00451.

Biological protocols, Synthetic protocols, as well as supporting figures and tables, NMR spectra, and references (PDF)