Incorporation of azide groups into an engineered Fn3 scaffold protein for site-specific, bioorthogonal conjugation

Abstract

Protein bioconjugates have a wide range of applications, including for targeted drug delivery and in vivo diagnostics. Expanding the chemistry available to link protein to cargo remains a critical consideration for the development of novel bioconjugates for unmet needs. Currently, most FDA-approved protein bioconjugates are based on the antibody structure, which constrains the available conjugation chemistries and applications. Alternative scaffold proteins have potential as targeting molecules that advance the synthetic possibilities. Here, we report two methods of high-yield incorporation of azide-containing groups into an Fn3 domain protein scaffold to enable click chemistry conjugation. One approach incorporates noncanonical amino acid azidohomoalanine during recombinant protein expression, and the other method develops a linker system using a unique thiol residue in the Fn3 structure. These synthetic approaches enabled orthogonal, dual labeling. The resulting conjugates retain functionality in receptor binding assays, validating these methods for diverse applications in engineering scaffold proteins for bioconjugate applications.

Introduction

Proteins have been engineered to bind a variety of receptors and other targets of interest for broad applications in biotechnology and medicine. In recent years, protein bioconjugates have found success as targeted therapeutics, targeted drug delivery agents, and molecular diagnostics, where a systemically delivered target-binding protein is able to localize a conjugated group, such as a toxin or radioisotope, to the site of disease. Antibody-drug conjugates (ADCs) are one such class of protein-bioconjugates, including 15 current FDA-approved ADCs for indications in oncology, primarily for hematologic malignancies, with many more related molecules in the therapeutic pipeline (Chen et al., 2025). For some applications, protein bioconjugates using polypeptide folds other than antibodies are warranted, enabling application-specific tailoring of circulation half-life and greater modularity in developing chemical conjugation strategies. For instance, ADCs have not fared well for many solid malignancies, partially due to barriers in accessing the entirety of a dense tumor (Cruz and Kayser, 2019). Long circulation times of antibodies are unfavorable for molecular imaging, radiotherapy, and other contexts where rapid clearance from non-target tissues is desirable (Tsai and Wu, 2018; Luo et al., 2022; Ninatti et al., 2025; To et al., 2025). A large number of alternative protein scaffolds have been developed for engineering as target-binding molecules, including affibodies, fibronectin domains, anticalins, nanobodies, cysteine knot proteins, DARPins, Gp2 proteins, and others (Beste et al., 1999; Binz et al., 2003; Hamers-Casterman et al., 1993; Kimura et al., 2009; Koide et al., 1998; Kruziki et al., 2015; Nord et al., 1997; Luo et al., 2022). These alternative scaffolds have the potential to enable greater flexibility compared to antibodies for modifications, such as through bioconjugation, where the complex structure of antibodies and manufacturing requirements can constrain the set of feasible chemical modifications (Gordon et al., 2015).

Historically, protein bioconjugation has largely relied on native functional groups to act as linkers, including primary amines found at the N-terminus and in lysine residues, and cysteine thiol moieties (Backer et al., 2008; Chalker et al., 2009; Li et al., 2018; Pickens et al., 2018; Hennrich and Kopka, 2019; Chavda et al., 2023). The majority of current FDA-approved ADCs utilize thiol-maleimide chemistry for bioconjugation (Wang et al., 2022). However, thiol-maleimide linkages have been demonstrated to be unstable in physiological conditions due to hydrolysis, retro-Michael reactions, and thiol exchanges with endogenous functional groups, making off-target cytotoxic activity a concern (Huang et al., 2019; Wang et al., 2022; Dudchak et al., 2024). While efforts to enhance the stability of thiol-maleimide linkages are ongoing, bioorthogonal conjugation chemistries offer the potential to eliminate these concerns (Bird et al., 2021; Wang et al., 2022). The development of a wide variety of noncanonical amino acids (ncAAs) is enabling the incorporation of novel chemistries into proteins, such as for the development of bioconjugate therapeutics (Rezhdo et al., 2019; Xu et al., 2021; Lino et al., 2024; Pigula and Schultz, 2024). Control over the site of payload incorporation into a protein-drug conjugate is critical when designing an overall biologic so that added functional groups do not impede protein binding function, and ncAAs can enable such site-specific incorporation of chemical reactive groups. Towards industrial scale production of proteins with ncAAs, microbial systems that can incorporate ncAAs during protein synthesis have been reported, including auxotrophic strains that incorporate ncAAs following a media shift (Johnson et al., 2010; Merten et al., 2015, 2019; Majekodunmi et al., 2024). For the production of therapeutic bioconjugates, ncAAs that include an azide group are attractive due to their bioorthogonal, high yield click chemistry reactions with strained alkyne rings, yielding no byproducts.

Azide-alkyne [3+2] cycloadditions, a form of click chemistry (Kolb et al., 2001), are simple, high-yielding, and bioorthogonal reactions that provide an alternative to thiol-maleimide and amineester coupling chemistries in protein bioconjugation (Scinto et al., 2021; Dudchak et al., 2024). These reactions are irreversible in water at neutral pH, occur at biological temperatures, and generate no byproducts, making them a promising candidate for linker chemistry. Azide-alkyne [3+2] cycloadditions can be copper-catalyzed (CuAAC) or strain-promoted (SPAAC) (Agard et al., 2004; Scinto et al., 2021). SPAAC avoids copper toxicity by using strained cycloalkyne rings like dibenzocyclooctyne (DBCO), bis-(1,3-dibutylbarbituric acid)trimethine oxonol (DIBAC), or bicyclonon-4-yne (BCN) to drive the reaction in order to relieve ring strain (Agard et al., 2004; Dudchak et al., 2024). SPAAC has shown comparable kinetics to CuAAC and has been used in protein-drug conjugates (Agard et al., 2004; Baskin et al., 2007; Dennler et al., 2014; Zimmerman et al., 2014; Currier et al., 2016; Donnelly et al., 2018; Pickens et al., 2018; Dudchak et al., 2024). When SPAAC and thiol-maleimide chemistry were compared for generation of an trastuzumab antibody base with a DBCO-monomethyl auristatin E (MMAE) payload, the SPAAC linker demonstrated a better drug-to-antibody ratio and greater stability in vivo, highlighting SPAAC as a promising linker option (Dennler et al., 2014; Dudchak et al., 2024).

The addition of an azide functional group into proteins for SPAAC can be achieved through post-translational modifications or the incorporation of ncAAs, often relying on using cell-free synthesis approaches or solid-phase peptide synthesis (Zimmerman et al., 2014; Ma and Yates, 2018; Pickens et al., 2018; Elledge et al., 2020; Iwamoto et al., 2025). Azide-containing ncAAs have previously been used to enable SPAAC bioconjugation in alternative protein scaffold-based protein-drug conjugates. An integrin-targeting knottin-Fc-MMAF conjugate was synthesized by incorporating L-para-azidomethyl-L-phenylalanine (pAMF) into the knottin-Fc fusion protein in solid-phase peptide synthesis and conjugating with MMAF through SPAAC (Currier et al., 2016). Azidohomoalanine (AHA) is another azide-containing ncAA that can facilitate click chemistry reactions with proteins (Lehner et al., 2017; Ma and Yates, 2018; Jecmen et al., 2023). Given that AHA is a methionine analog, when methionine-auxotrophic E. coli are grown in media lacking methionine but supplemented with AHA, the translational machinery can readily incorporate AHA into newly synthesized proteins where methionine is encoded (Lehner et al., 2017; Ma and Yates, 2018; Jecmen et al., 2023). Since AHA incorporation has largely been developed in E. coli, translationally incorporating AHA has been pursued in scaffold proteins, such as DARPins (Merten et al., 2015). Conference reports have shown the incorporation of AHA into fibronectin proteins as a part of a larger artificial protein hydrogel network (Baccile and Tirrell, 2020). An azide moiety has recently been incorporated into an Fn3 scaffold through post-translational modification of an acetone staple via oxime-ligation, without compromising target binding (Iwamoto et al., 2025).

In this present work, we successfully demonstrated two methods that can be used to introduce an azide group into Fn3 variants, which can be extended to other alternative protein scaffolds that are produced through bacterial expression. We confirmed the functionality of the azide in facilitating site-specific, bioorthogonal click chemistry bioconjugation reactions, and we demonstrated maintenance of the bioconjugates’ binding affinity. We used an Fn3 variant that had previously been engineered to bind αvβ3 integrin receptors with high affinity (Richards et al., 2003). In one approach, we incorporated AHA into Fn3 proteins using a methionine auxotrophic E. coli, replacing a single methionine residue in Fn3 with AHA during protein synthesis. In a second approach, we engineered a unique cysteine residue within the C-terminal region of the Fn3 proteins, which we then post-translationally modified with azido-PEG3-maleimide to introduce the azide moiety. In both methods of azide incorporation, the azide was readily available for a ring-strained click chemistry reaction with a fluorophore with a DBCO reactive group, resulting in specific labeling with high yield. Finally, we demonstrated the ability to label Fn3 proteins simultaneously and orthogonally using one cysteine residue to conjugate a maleimide-PEG group and one AHA residue to conjugate a DBCO-fluorophore, demonstrating how azide incorporation can enable the synthesis of more complex, specifically designed Fn3-drug molecules than would have been possible using only native amino acid chemistries. All bioconjugates maintained binding affinity post-modifications. To our knowledge, incorporation of AHA into purified, engineered Fn3 domain proteins enabling dual labeling in a protein-drug conjugate context has yet to be reported. Similarly, this work is the first to report simple incorporation of an azide group into an engineered scaffold protein via a maleimide-PEG3-azide linker. Collectively, this report establishes methods to incorporate azide reactive groups into engineered Fn3 domain proteins, expanding the available bioconjugate chemistry approaches possible for the synthesis of scaffold protein-small molecule conjugates, towards applications including targeted drug delivery, synthesis of molecular imaging agents, and development of higher order macromolecular structures.

Results

Design of Fn3 variants with a reactive azide group

We designed two parallel approaches to incorporate a reactive azide group into an Fn3 scaffold protein (Figure 1). Fn3 non-antibody scaffold proteins have been primarily engineered to bind novel targets mediated by the BC, DE, and FG loops, which are structurally proximal to the N-terminus of the protein (Figure 1A). Therefore, to reduce the potential for steric hindrance from conjugated molecules interfering with binding to targets, we chose to incorporate site-specific reactive groups at the C-terminus. As a model system to study azide incorporation into Fn3 scaffold proteins, we used an Fn3 variant that had previously been engineered to bind αvβ3 integrin receptors with high affinity, reported as having a dissociation constant in the single-digit nanomolar range, primarily mediated by an RGDWNE sequence in the FG loop (Richards et al., 2003). RGD refers to the tripeptide sequence of arginine, glycine, and aspartic acid that mediates interactions with integrin receptors. We refer to this variant as FnRGD. As a non-binding control, we worked with an Fn3 variant with the native RGD binding motif replaced with an RDG amino acid sequence, starting with a wildtype, lower affinity Fn3 variant; we refer to this variant as FnRDG. All protein sequences of described variants are documented in Supporting Information.

Figure 1. Bioconjugation of a small molecule to Fn3-azide proteins.

(A) Engineered Fn3 proteins have binding loops proximal to the N-terminus; the FG loop with the RGD binding motif is highlighted in red. Azide was incorporated near the C-terminus via either (B) AHA replacement of Met during bacterial protein expression or (C) post-purification modification of a Cys with azido-PEG3-maleimide. (D) A small molecule fluorophore (fluor) was then conjugated using SPAAC with DBCO.

To facilitate click chemistry-mediated bioconjugates, we incorporated the noncanonical amino acid AHA using auxotrophic E. coli that substitute AHA in place of Met during protein translation (Figure 1B). We used standard recombinant DNA methods to introduce a Met codon at the C-terminus of the Fn3 sequences (Figure 1B). We designed the protein sequences considering that post-translational processing in E. coli would remove the Met/AHA coded for by the start codon by having the second codon encode Ala (Wang et al., 2008). This processing ensures that only the AHA in the C-terminal region of the Fn3 would remain when proteins were expressed. Due to the lack of endogenous Met residues in the Fn3 variants, we proposed that addition of a Met codon would enable site-specific incorporation of AHA and subsequent site-specific conjugation using click chemistry (Figure 1D). These variants are referred to as FnRGD-AHA and FnRDG-AHA.

As an alternative approach to enable click chemistry reactions with Fn3 molecules that would rely only on canonical amino acids for comparison to the methods using AHA, we designed FnRGD and FnRDG variants with a C-terminal 15-amino acid cysteine tag (FnRGD-Cys and FnRDG-Cys, Figure 1C) (Backer et al., 2008). Due to the lack of endogenous cysteine residues in the Fn3 scaffold, this cysteine tag allows for site-specific modification of the Fn3-Cys variants via thiol-maleimide chemistry. To introduce an azide group, we reacted the Cys with azido-PEG3-maleimide to enable subsequent click chemistry reactions (Figure 1D).

Mass spectrometry of Fn3 variants confirmed AHA incorporation

Variants FnRGD-AHA and FnRDG-AHA were expressed in methionine auxotrophic B834(DE3) E. coli and induced in media lacking Met and supplemented with AHA (Jecmen et al., 2023). Successful incorporation of AHA into the Fn3 variants was confirmed using mass spectrometry. The percent error between the expected molecular weight and measured monoisotopic mass of FnRGD-AHA was 0.007%, and the percent error between the expected molecular weight and measured monoisotopic mass of FnRDG-AHA was 0.06%, indicating successful incorporation of AHA into the variants (Table 1). For each variant, the measured molecular weight was closer to the expected molecular weight with AHA incorporation, compared to with Met incorporation.

Table 1.

Mass spectrometry of Fn3 variants.

Speciesa	Expected MW, Fn3-Met (Da)b	Expected MW, Fn3- AHA (Da)b	Measured MW (Da)c
FnRGD	11379.65	11374.57	11373.7733
FnRDG	11279.49	11274.41	11267.6657

^aData were retrieved by ESI-TOF mass spectrometry.

^bExpected and

^cmeasured molecular weight (MW) of Fn3 proteins.

Click chemistry reaction of Fn3-AHA with ring-strained fluorophore

To further confirm that AHA was correctly incorporated into the Fn3 variants and that the azide was functional, FnRGD-AHA and FnRDG-AHA were reacted with AZDye-488-DBCO, a fluorophore containing a strained dibenzocyclooctyne group. Following removal of excess dye, equivalent amounts of Fn3 variants, with and without AHA, and before and after dye conjugation, were analyzed by SDS-PAGE. Fluorescent gel imaging revealed that FnRGD-AHA-DBCO-488 and FnRDG-AHA-DBCO-488 were fluorescent, and no other Fn3 variants exhibited fluorescence (Figure 2A, B). The degree of labeling of both labeled proteins was ~1, consistent with site-specific conjugation and high reaction yields. While there is a less-favored possibility in some biological systems of a slower side reaction of free thiols reacting with DBCO, there are no native cysteine residues in Fn3 variants, so we expect all DBCO reactions to be with the intended AHA residue.

Figure 2. Fn3-AHA was expressed and purified with functional azidohomoalanine.

Fn3 variants were successfully expressed and purified with and without azidohomoalanine (AHA) incorporation. Control proteins and variants with AHA reacted with DBCO-488 were analyzed by SDS-PAGE, (A) stained with SimplyBlue SafeStain or (B) detected using fluorescence imaging. Lanes: (M) standard marker; (1) FnRDG (MW ~ 11.1 kDa); (2) FnRGD (MW ~11.2 kDa); (3) FnRGD-AHA (MW ~ 11.4 kDa); (4) FnRDG-AHA-DBCO-488; (5) FnRGD-AHA-DBCO-488; (FM) fluorescent marker. Only Fn3 variants with AHA that reacted with DBCO-488 are fluorescent.

Synthesis of Fn3-Cys-PEG3-azide and click chemistry reaction with ring-strained fluorophore

As a parallel approach to conferring azide functionality to Fn3 variants, we designed a system using only canonical amino acids, utilizing the thiol in cysteine as a reactive handle and through maleimide chemistry conjugating a short linker with a PEG3 group and terminal azide. Fn3 variants containing a C-terminal tag with a unique cysteine residue were cloned, expressed in bacteria, and purified. These variants were named FnRGD-Cys, which binds αvβ3 integrins, and FnRDG-Cys, the negative control. A commercially available maleimide-PEG3-azide linker was conjugated to the cysteine-containing variants using standard thiol-maleimide reaction conditions. To confirm successful incorporation and reactivity of the azide-maleimide linker, FnRGD-Cys-PEG3-azide and FnRDG-Cys-PEG3-azide were treated with the same DBCO-488 fluorophore. Following removal of excess dye via a dye removal column and dialysis, equivalent amounts of post-translationally modified Fn3-Cys-PEG3-azide proteins before and after dye conjugation were analyzed by SDS-PAGE (Figure 3). Fluorescent gel imaging revealed that FnRGD-Cys-PEG3-azide-DBCO-488 and FnRDG-Cys-PEG3-azide-DBCO-488 exhibited fluorescence after dye conjugation but not before, demonstrating successful incorporation of an azide via maleimide-thiol reactions and subsequent successful conjugation to DBCO-488 via DBCO-azide reactions (Figure 3B). Degree of labeling measurements showed ~90% of FnRGD-Cys molecules and ~60% of FnRDG-Cys molecules were labeled with fluorophore, which required both the thiol-maleimide and azide-DBCO reactions to be successful on the same molecule.

Figure 3. Fn3-Cys proteins were expressed, post-translationally modified to include an azide, and were functional in a click chemistry reaction.

Fn3 variants with a C-terminal tag containing a unique cysteine residue (MW: FnRGD-Cys ~ 13.2 kDa) were reacted with maleimide-PEG3-azide (MW ~ 0.4 kDa). The azide group was then conjugated to a DBCO-488 fluorophore (MW ~ 0.8 kDa). Protein gels were (A) stained with SimplyBlue SafeStain and (B) analyzed using fluorescence imaging. Lanes: (M) standard marker; (1) FnRGD-Cys-PEG3-azide; (2) FnRDG-Cys-PEG3-azide-DBCO-488; (3)FnRGD-Cys-PEG3-azide-DBCO-488; (FM) fluorescent marker. Only Fn3 variants that were functionalized with the PEG3-azide linker and reacted with DBCO-488 are fluorescent.

Fn3 variants fluorescently labeled via click chemistry maintain functional binding

Equilibrium cell binding assays using flow cytometry were conducted using αvβ3 integrin-expressing U-87 MG glioblastoma cells to determine the αvβ3 integrin binding affinity (dissociation constant, K_D) of the azide-containing, fluorescently-labeled Fn3 variants (Figure 4). As a control with no azide chemistry, FnRGD-Cys-488 was synthesized by a standard thiol-maleimide reaction and measured to have a K_D = 1.3 nM (range: 1.0-1.4 nM) for U-87 MG cells (Figure 4B). FnRGD variants bind U-87 MG cells, but non-binding control variants (FnRDG) do not (Figure 4 C, D). The K_D of FnRGD-AHA-DBCO-488 and FnRGD-Cys-PEG3-azide-DBCO-488 were measured to be K_D = 1.9 nM (range: 1.7 - 2.3 nM) and K_D = 1.9 nM (range: 1.0 - 5.5 nM), respectively, confirming maintained binding after C-terminal modification and strain-promoted click chemistry conjugation (Figure 4E, F). We see slightly more variability in the binding measurements of FnRGD-Cys-PEG3-azide-DBCO-488, which may be due to the short PEG linker causing possible structural variability with partial steric hindrance in some conformations.

Figure 4. FnRGD-AHA-DBCO-488 and FnRGD-Cys-PEG3-azide-DBCO-488 maintain binding to cells expressing αvβ3 integrins.

(A) U-87 MG glioblastoma cells express high levels of αvβ3 integrins, as detected by an anti-αvβ3 fluorescein conjugated antibody (white - cells only; gray - αvβ3 antibody). (B) FnRGD-Cys-488 protein without azide modification binds U-87 MG cells with high affinity (K_D = 1.3 nM; range: 1.0-1.4 nM). (C) FnRGD-AHA-DBCO-488 binds U-87 MG cells (30 nM, gray), but non-binding control FnRDG-AHA-DBCO-488 does not (30 nM, white). (D) FnRGD-Cys-PEG3-azide-DBCO-488 binds U-87 MG cells (30 nM, gray), but non-binding control FnRDG-Cys-PEG3-azide-DBCO-488 does not (30 nM, white). (E) Equilibrium titration assays demonstrated that FnRGD-AHA-DBCO-488 (K_D = 1.9 nM; range: 1.7 - 2.3 nM) and (F) FnRGD-Cys-PEG3-azide-DBCO-488 (K_D = 1.9 nM; range: 1.0 - 5.5 nM) maintained binding to U-87 MG cells. K_D values were determined as the half-maximal signal of a sigmoidal fit and are reported as geometric mean and range of independent values from triplicate binding data collected on different days to capture experimental variation. Normalized, averaged data is shown for binding curves, with arithmetic means plotted for triplicate experiments. Error bars represent the range of values measured. Concentrations from 0.01 nM - 100 nM of conjugates were used in titration binding assays.

Simultaneous incorporation of Cys and AHA enables orthogonal dual labeling of Fn3 scaffold protein

Orthogonal functionalities for bioconjugation enable simultaneous loading of a receptor-targeting protein with multiple distinct payloads. Given the results that the Fn3 scaffold could be readily labeled via a cysteine residue or an AHA residue, we designed Fn3 constructs incorporating both a single cysteine residue and a single AHA residue, for orthogonal conjugation with two distinct payloads. A single methionine residue had previously been included in the C-terminal Cys-tag of the FnRGD-Cys and FnRDG-Cys variants, with eight amino acids separating Cys and Met. We proposed that if we expressed these variants in the bacterial system that incorporated AHA in place of Met, then we would have Fn3 variants that offered a single thiol and single azide for synthetic opportunities. Using AlphaFold 3, we modeled the structure of the Fn3-Cys variants to observe the expected orientation of the thiol in the cysteine and the proposed azide in place of methionine (Figure 5A). Interpreting the five top-ranked models and the predicted Local Distance Difference Test (pLDDT) scores, the Cys-tag was predicted to adopt a flexible structure with potential alpha helical content (Figure S1), consistent with its original design (Backer et al., 2008). The Cys residue is proximal to the Fn3 immunoglobulin-like fold, and we had already shown that this location was available for bioconjugation. We anticipated that the proposed location for AHA incorporation would be sufficiently distant from the Fn3 fold and the Cys residue, and at a rotation of ~180° from the Cys residue if in an alpha helix, such that it was reasonable to expect that each residue could be used simultaneously for cargo conjugation.

Figure 5. Dual labeling of FnRGD using orthogonal reactions.

(A) An FnRGD protein with a unique Cys and unique AHA was expressed, with the Cys and AHA residues in a C-terminal tag. Structure predicted by AlphaFold 3. The protein was labeled in a simultaneous one-pot reaction with maleimide-PEG and DBCO-488. (B) Conjugation with maleimide-PEG was visible as a shift with a higher apparent molecular weight, as analyzed by SDS-PAGE (left). Conjugation with DBCO-488 was visible when imaging the gel with fluorescence (right). Gel was run under non-reducing conditions. Lanes: (M) standard marker; (1) FnRGD; (2) FnRGD-Cys-AHA unlabeled, with monomer and disulfide dimer populations visible; (3) dual labeled reaction, purified with a ~5 kDa MW cutoff Centrispin column; (FM) fluorescent marker. Fluorescent gel shows all populations that were labeled with DBCO-488, including monomer with no PEG, dimer with no PEG, and desired product dual-labeled with maleimide-PEG and DBCO-488. (C) Dual-labeled reaction was purified using a ~25 kDa MW cutoff Centrispin column and analyzed by SDS-PAGE (left) and with fluorescence (right). Lanes: (M) standard marker; (1) FnRGD; (P) purified dual labeled reaction of FnRGD-Cys-AHA with maleimide-PEG and DBCO-488. Predominant band is the apparent MW of protein with PEG and is fluorescent. (D) Mixture of FnRGD-Cys(PEG)-AHA(488) and FnRGD-Cys(PEG)-AHA(no label) (50 nM total protein, green) binds to U-87 MG glioblastoma cells, compared to cells only (white). Unlabeled FnRGD (200 nM) competes off binding of dual-labeled mixture (50 nM) (gray), demonstrating specific binding of dual-labeled FnRGD-Cys(PEG)-AHA(488).

We expressed FnRGD-Cys and FnRDG-Cys in the methionine auxotrophic bacterial system supplemented with AHA to generate FnRGD-Cys-AHA and FnRDG-Cys-AHA. These proteins were able to be expressed and purified at high yield and purity, routinely resulting in greater than 10 mg purified protein per liter of standard shake flask culture, without any optimization. Acknowledging that TCEP is known to react with azide groups via a Staudinger-type reduction (Faucher and Grand-Maître, 2003; Kantner et al., 2017), we did several preliminary experiments to inform dual labeling of Fn3-Cys-AHA proteins. We examined how the order of conjugation affected conjugation efficiencies with maleimide-AF647 and DBCO-488 fluorophores, analyzing conjugates by SDS-PAGE (Figure S2) and measuring extent of labeling using UV-visible spectroscopy (Figure S3). When FnRGD-Cys-AHA was labeled only with maleimide-AF647 (10 eq), in the presence of TCEP (10 eq) for 4 h at room temperature, and following purification by a dye removal column, 88% of the proteins were labeled by maleimide-AF647 (Figure S3A). When FnRGD-Cys-AHA was labeled only with DBCO-488 (10 eq) for 4 h at room temperature, without any attempt to reduce disulfide bonds, only 33% of molecules were labeled by DBCO-488 (Figure S3B). This reaction yield was substantially lower than the yields seen for Fn3-AHA proteins, which could be attributed to the ability of the FnRGD-Cys-AHA molecules to form disulfide bonded dimers. Such dimerization would sterically hinder access to the AHA side chain, preventing the SPAAC reaction from occurring. To examine a two-step, dual labeling protocol, the proteins from these single-labeled reactions were then reacted with the second fluorophore. When the mixture of FnRGD-Cys(AF647)-AHA (88%) and unlabeled FnRGD-Cys-AHA (12%) was reacted with DBCO-488 (10 eq) for 4 h at room temperature, then purified by dialysis, the degree of labeling with DBCO-488 was 91% (Figure S3C). When the mixture of FnRGD-Cys-AHA(488) (33%) and FnRGD-Cys-AHA (67%) was reacted with TCEP (10 eq) and maleimide-AF647, then purified by dialysis, the degree of labeling by maleimide-AF647 was 55% (Figure S3D).

Based on these preliminary results, we designed a simultaneous one-pot reaction, minimizing the amount of time that the azides were exposed to TCEP, while also preventing disulfide-mediated dimerization and steric hindrance of the azide. A maleimide-PEG (M_n ~ 10 kDa) was conjugated to the Cys residue while DBCO-488 was conjugated to AHA, in reaction conditions containing TCEP (Figure 5A). We anticipated that once an Fn3 molecule was successfully reacted with maleimide-PEG, that molecule would then be unable to form a disulfide-bonded dimer, and therefore the AHA would be unhindered and able to rapidly undergo a click reaction with DBCO-488. With the kinetics of the TCEP reaction with azide being slower than the DBCO reaction with azide, we hypothesized that this one-pot reaction would address the issues observed with stepwise synthesis (Heiss et al., 2021; Schauenburg and Weil, 2025).

The simultaneous one-pot, dual labeling reaction was subsequently initially purified using a gel filtration chromatography spin column with a ~5 kDa MW cutoff column, to retain all proteins and enable quantification of yield of proteins conjugated to maleimide-PEG while removing most unreacted DBCO-488. The maleimide-PEG conjugation resulted in a substantial visual shift in apparent molecular weight of the protein conjugate when analyzed by SDS-PAGE, shifting the band to >30 kDa (Figure 5B, lane 3), compared to monomer (MW ~ 11.2 or 13.2 kDa) and disulfide-mediated dimer (MW ~ 26.4 kDa) bands in unreacted proteins (Figure 5B, lanes 1 and 2). The yield of the maleimide-PEG reaction was 61% in the gel shown, determined by the ratio of the mass of the band corresponding to PEG conjugate compared to the total protein mass in the lane (Figure 5B, lane 3). Fluorescent imaging enabled visualization of those populations that included conjugation to DBCO-488 (Figure 5B, right). In the dual labeled reaction lane, fluorescence is visible in monomer, dimer, and PEG-conjugated monomer bands (Figure 5B, lane 3, right gel). There is a faint band at higher molecular weight between 40 and 63 kDa, visible only in the highly sensitive fluorescent gel image (Figure 5B, lanes 3). We attribute this band to a very small amount of fluorescently labeled species that may have two maleimide-PEG conjugations. Two PEG groups are possible if a second maleimide-PEG conjugates to a protein molecule by a primary amine, which is known to react with maleimides in a reaction that is less favored than reactions via thiols. Incompletely removed DBCO-488 at low molecular weight is also visible.

To remove remaining unreacted DBCO-488, unreacted maleimide-PEG, and protein not conjugated to maleimide-PEG, the reaction was applied to a gel filtration column with a ~25 kDa MW cutoff, and analyzed by SDS-PAGE (Figure 5C, lane P) and UV-visible spectroscopy (Figure S4). Using quantitative gel analysis, comparing the band with apparent MW consistent with the addition of maleimide-PEG compared to total protein in the same lane, we determined that ~92% of the sample was conjugated to maleimide-PEG after this purification step. Quantitative analysis with UV-visible spectroscopy using the extinction coefficients and correction values included in Supporting Information and described in Materials and Methods showed that 60% of the sample purified by the ~25 kDa gel filtration column was conjugated to DBCO-488. The overall yield of the one-pot reaction defined as (mol dual labeled protein) / (mol initial protein) was therefore 30-40%, calculated as the product of the maleimide-PEG degree of labeling and DBCO-488 degree of labeling, consistent with experimental replication.

To assess maintained binding and specificity for αvβ3 integrin receptors, we used flow cytometry with αvβ3 integrin-expressing U-87 MG cells (Figure 5D). The sample in Figure 5C, lane P was a mixture of FnRGD-Cys(PEG)-AHA(488) (~60%) and FnRGD-Cys(PEG)-AHA(no 488) (40%). Since the sample was not fully fluorescently labeled, an equilibrium binding assay with a full titration curve was not appropriate, and non-fluorescent molecules would compete off fluorescent molecules. Instead, to assess if the FnRGD-Cys(PEG)-AHA(488) could still bind receptor, we tested a single concentration of mixture (50 nM, with ~30 nM FnRGD-Cys(PEG)-AHA(488) and ~20 nM FnRGD-Cys(PEG)-AHA(no 488)) in a flow cytometry binding assay with U-87 MG cells (Figure 5D). This sample showed a fluorescent signal significantly shifted from the cells-only control, indicating that the dual-labeled molecules maintain binding, although this experiment does not indicate if the dissociation constant was altered. To confirm that the binding was specific, we then incubated the same amount of mixture of FnRGD-Cys(PEG)-AHA(488) and FnRGD-Cys(PEG)-AHA(no 488) (50 nM total) with the cells, in competition with excess, unlabeled FnRGD protein (200 nM). In the competition assay, the binding fluorescent signal was reduced to near baseline signal of autofluorescence, confirming specificity of the observed binding of dual-labeled FnRGD-Cys(PEG)-AHA(488) (Figure 5D). Non-specific binding would not have been competed off by unlabeled FnRGD protein.

Discussion

Development of alternative protein scaffolds with diverse bioconjugation chemistries can expand the applications of protein bioconjugates. In this work, we report two strategies to incorporate azides into Fn3 scaffolds using either an ncAA or a post-translational modification, enabling site-specific, bioorthogonal bioconjugation, towards synthesis of protein-drug conjugates with Fn3 variants and other scaffold proteins. The methods reported maintain high yields in bacterial expression and rely on commercially available reagents that can be adopted in many settings. The post-translational approach of azide incorporation using a maleimide-PEG-azide linker required no special media and could easily be incorporated into standard microbial expression workflows. Using SPAAC with DBCO eliminates the need for a copper catalyst, avoiding copper purification challenges and biotoxicity, which is advantageous for biological applications. For Fn3 proteins, the binding interface is proximal to the N-terminus, necessitating that conjugated groups be on the opposite pole of the protein, a constraint that is not universal for other scaffold proteins. When desiring two orthogonal chemical handles for conjugation of two distinct groups, we addressed this constraint by incorporating a C-terminal Cys-tag with a Cys residue and AHA residue sufficiently separated to enable efficient labeling of each. This tag with alpha-helical properties and one Cys and one AHA could be included at the terminus of any scaffold protein that can be produced in bacteria for a straightforward method to produce variants for dual labeling. The simultaneous one-pot reaction reported results with high yields that did not require intermediate purification steps. In our example reaction, we did not obtain a homogeneous product, not separating FnRGD-Cys(PEG)-AHA(488) from FnRGD-Cys(PEG)-AHA(no 488).

Optimization in labeling reaction conditions and purification protocols tailored to a particular set of two conjugate groups would be required to obtain a homogeneous sample of molecules for downstream applications. A conjugated PEG group, for instance, can be used for half-life modification and increased therapeutic stability when paired with cytotoxin conjugation. The overall approach allows for the incorporation of multiple distinct therapeutic agents, potentially resulting in enhanced potency through synergistic mechanisms, or the ability to simultaneously include separate diagnostic and therapeutic groups within the same molecule.

Materials and Methods

Plasmid construction

Genes for all Fn3 variants were subcloned into a pET24b protein expression plasmid containing a T7 promoter, a C-terminal hexahistidine tag for metal affinity chromatography, and a kanamycin resistance gene for selection. The FnRGD protein sequence with high affinity for αvβ3 integrin was a variant previously reported as variant 3JCLI4 (Richards et al., 2003). A non-binding FnRDG variant has been previously described (Sirois et al., 2020). To construct methionine variants for substitution with AHA, recombinant DNA methods were used to incorporate an ATG codon for Met immediately following the last amino acid in the Fn3 sequence, upstream of the hexahistidine tag. To construct variants containing a single Cys residue, a Cys-tag was added using recombinant DNA methods at the C-terminus, upstream of the hexahistidine tag. The Cys tag codes for 15 amino acids and takes an α-helical structure: Lys-Glu-Ser-Cys-Ala-Lys-Lys-Phe-Gln-Arg-Gln-His-Met-Asp-Ser (Backer et al., 2008). All plasmid sequences were confirmed by standard Sanger sequencing.

Standard protein expression and purification

Plasmids containing Fn3 variants were transformed into BL21(DE3) protein expression E. coli. Cultures were grown in LB broth with kanamycin and induced (3 hr, 37°C) with 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG). Cells were resuspended in lysis buffer (35 mM Na₂HPO₄×dibasic, 15 mM NaH₂PO₄×monobasic, 500 mM NaCl, 5 mM CHAPS, 25 mM imidazole, 5% glycerol) supplemented with an EDTA-free protease inhibitor (Pierce), and lysed by repeated freezing and thawing. Soluble fractions were isolated by centrifugation. Fn3 variants were purified by cobalt affinity chromatography with HisPur cobalt resin (Thermo Fisher). Protein samples were dialyzed into water, lyophilized, reconstituted with PBS to the desired concentration, and analyzed for purity by SDS-PAGE. For longer term storage, lyophilized samples were stored at −80°C.

Protein expression including AHA

The pET-Fn3 plasmids with a C-terminal Met codon or with a Met codon in the C-terminal Cys-tag were transformed into B834(DE3) E. coli (Millipore Sigma, 69041), a methionine auxotroph, for expression with substitution of AHA in place of Met. Bacteria were first grown in LB broth with kanamycin until the culture reached an optical density at 600 nm (OD₆₀₀) of ~1.0. Bacteria were washed twice with PBS by pelleting and resuspending to remove all traces of methionine. Bacteria were then resuspended in induction media, consisting of M9 minimal media supplemented with kanamycin and amino acids, without methionine and with AHA, as described previously (Jecmen et al., 2023). Protein expression was induced with 1 mM IPTG for 18 h at 20 °C with shaking for aeration. Bacterial pellets were collected by centrifugation and proteins were purified as described above.

Post-purification modification of Fn3-Cys variants with azido-PEG3-maleimide

Azido-PEG3-maleimide was purchased as a kit containing azido-PEG3-amine as an oil and maleimide-NHS ester as a solid (Thermo Scientific, J65984MC). Upon receiving, each component was separately resuspended in anhydrous DMF, aliquoted, and stored at −20 °C until use. Immediately before reacting with Fn3 containing a Cys residue, maleimide-NHS-ester and azido-PEG3-amine were combined and stirred (30 min, 300 rpm shaking, room temperature). The Fn3-Cys was treated with TCEP (10 eq, 15 min, room temperature) to reduce any disulfide bonds. The resulting azido-PEG3-maleimide (15 eq) was then reacted with FnRGD-Cys or FnRDG-Cys (75 μM, 1 mg/mL, 1 eq) in PBS. The thiol-maleimide reaction was allowed to proceed for 2 h at room temperature with gentle mixing. Glutathione (150 eq) was added to consume excess thiol-reactive reagent. Excess azido-PEG3-maleimide was immediately removed using Centrispin separation columns (Centrispin 10, ~5 kDa MW cutoff, Princeton Separations, CS-100).

Click chemistry reaction of Fn3 variants with DBCO-488

AZDye 488 DBCO (Vector Laboratories, CCT-1278) was received as a lyophilized powder, resuspended in anhydrous DMF, aliquoted, and stored at −20 °C until use. Fn3-AHA protein (1 eq, concentration ~100 μM) or Fn3-Cys-PEG3-azide protein (1 eq, concentration ~100 μM) were incubated with AZDye 488 DBCO (10 eq or 15 eq, respectively) in PBS for 4 h at room temperature with gentle mixing by rotation. Excess DBCO-488 was removed following manufacturer protocol using a dye removal column (Thermo Scientific, 22858) and/or a 10 kDa MW cutoff dialysis cassette (Thermo Scientific, 66383).

Dual labeling reaction to conjugate maleimide-PEG and DBCO-488 to Fn3-Cys-AHA proteins

Methoxypolyethylene glycol maleimide (referred to as maleimide-PEG, average M_n~10 kDa, Sigma Aldrich 712469) was received as a lyophilized powder and resuspended in PBS. Fn3-Cys-AHA proteins (1 eq, concentration ~100 μM) were incubated with TCEP (10 eq), maleimide-PEG (10 eq), and AZDye 488 DBCO (10 eq) in PBS at room temperature overnight with gentle rotation. Excess TCEP, maleimide-PEG, and DBCO-488 were removed by gel filtration chromatography using a Centrispin 10 (MW cutoff ~ 5 kDa, Princeton Separations, CS-100) or Centrispin 20 (MW cutoff ~ 25 kDa, Princeton Separations, CS-200) column following the manufacturer's protocol.

SDS-PAGE analysis

For analysis by SDS-PAGE, samples were non-reduced. Each protein sample (approximately 1 μg, determined as described below) in LDS buffer was heated, then loaded onto a NuPAGE 4-12% Bis-Tris protein gel (Fisher Scientific, NP0322) and run in MES buffer (Fisher Scientific, NP0002). For fluorescence analysis, a fluorescent protein standard (Invitrogen, LC5928) was used. For staining by SimplyBlue SafeStain, a prestained protein standard was used (Invitrogen, LC5800). Images were taken on a Biorad ChemiDoc MP gel imaging system, collecting fluorescence data before staining with SafeStain, if applicable.

Determination of protein concentration and degree of labeling

Protein concentration was initially measured using absorbance at 280 nm with a Nanodrop spectrophotometer and an extinction coefficient estimated with the Expasy ProtParam tool (Gasteiger, 2003), with appropriate correction factors used for conjugated groups. Due to the confounding absorbance from conjugated groups making it difficult to ascertain a clear peak at 280 nm for bioconjugated protein samples, these samples were further analyzed using SDS-PAGE and imaged on a Biorad ChemiDoc MP gel imaging system following manufacturer mass analysis protocols and ImageLab Software to refine the protein concentration measurement. To determine degree of labeling for fluorescently labeled samples, the gel-analyzed concentration for protein was used, and the absorbance at 494 nm for the fluorophore was used to determine fluorophore concentration with the AZDye 488 extinction coefficient of 73,000 cm⁻¹M⁻¹. The degree of labeling is reported as the ratio of fluorophore to protein. Extent of labeling by maleimide-PEG in dual labeled conjugates was determined as the ratio of the mass of the higher molecular weight band that is consistent with PEG conjugation to the total mass of protein loaded in the lane, analyzed by SDS-PAGE and using the ImageLab Software mass analysis protocol.

Mass spectrometry

To measure incorporation of AHA, the molecular weight of purified Fn3 molecules was evaluated through electrospray ionization (ESI) mass spectrometry (Thermo Scientific Q Exactive HF-X Hybrid Quadrupole-Orbitrap Mass Spectrometer). Before injection into the mass spectrometer, protein samples in PBS were either desalted using StageTips (Thermo Scientific, SP301) and exchanged into a solution of 50% acetonitrile, 0.1% formic acid, and ultrapure water or dialyzed into water, lyophilized, and resuspended directly in the same solution. Acquisition range was set between 500-2250 m/z, and quantification was performed on the 10+ cluster.

Binding assays with fluorescently labeled proteins and their bioconjugates

U-87 MG glioblastoma cell line was obtained from ATCC (HTB-14). Cells were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37 °C in a humidified chamber containing 5% CO₂. When conducting cell assays, cells were removed from the culture surface using trypsin and resuspended in integrin binding buffer (IBB: 1 mM MgCl₂ hexahydrate, 1 mM MnCl₂ tetrahydrate, 2mM CaCl₂ dihydrate, 1 mg/mL bovine serum albumin, 100 mM NaCl, 20 mM Tris HCL pH 7.5). To measure the presence of αvβ3 integrins, an anti-αvβ3 FITC conjugate antibody was used (1:100 dilution, CD51/CD61 monoclonal antibody clone 23C6, FITC conjugate, eBioscience, #11-0519-42), incubating for 1 h at 4 °C with gentle rotation, then washed with IBB before analysis on a Guava EasyCyte flow cytometer. When comparing binding to U-87 MG cells by RGD binding bioconjugates versus RDG non-binding conjugates, a concentration of 30 nM for each conjugate was used. To measure dissociation constants of binding variants, 400,000 cells were incubated with a range of concentrations (0.01 nM - 100 nM) of binding conjugate for 1 h at 4 °C with gentle rotation, washed, and immediately analyzed by flow cytometry. The mean fluorescence intensity for each sample was measured, and the data were fit to a sigmoidal curve using Kaleidagraph (Synergy Software). Individual dissociation constants (K_D) were determined as the conjugate concentration yielding the half-maximal binding signal for each replicate. K_D values for three replicates from different days were used to determine a geometric mean, and the range of individually fit K_D values are given. In order to plot data from different days on the same plot, data in a single trial were normalized by subtracting the value of the asymptote of the lower bound of the curve fit from each fluorescence value for each concentration, then dividing each value by (maximum bound value - lower bound value). The arithmetic mean of the normalized fluorescence value was then plotted at each concentration tested, with error bars showing the range of individual replicates.

Protein structure modeling with AlphaFold 3

The relative positions and orientation of the cysteine residue and proposed AHA site in FnRGD-Cys-AHA protein was modeled using the AlphaFold 3 server (Abramson et al., 2024). The primary amino acid sequence for FnRGD-Cys was uploaded, with the original methionine residue in the C-terminal cysteine tag rather than AHA, and the default AlphaFold 3 settings were used. The five top-ranked models were selected for further analysis and visualized using ChimeraX. Model quality and structural flexibility was assessed using the pLDDT scores.

Binding and competition binding assay with FnRGD-cys-AHA

To detect binding of FnRGD-Cys(PEG)-AHA(488) dual-labeled sample, containing a heterogeneous mixture of protein conjugated to PEG alone or PEG and 488 fluorophore, 50 nM of mixture sample was incubated with U-87 MG cells, using protocols described above for previous cell binding assays. To confirm that the binding was specific, the same amount of mixture of FnRGD-Cys(PEG)-AHA(488) and FnRGD-Cys(PEG)-AHA(no 488) (50 nM) was incubated with U-87 MG cells in competition with excess, unlabeled FnRGD protein (200 nM), allowing the excess unlabeled protein to compete off dual-labeled protein.

Supplementary Material

Supplementary data are available, and includes information on:

Protein Sequences, Molecular Weights, and Predicted Extinction Coefficients

AlphaFold 3 structure prediction of FnRGD-Cys-AHA

Preliminary studies to develop protocols for dual labeling of FnRGD-Cys-AHA

UV-visible spectroscopy of dual-labeled FnRGD-Cys-AHA

Data availability

The data underlying this article are available in the article and in its online supplementary material.