Supramolecular coiled-coil peptide platform for site-specific antibody drug conjugate engineering

Abstract

Antibody conjugates play a central role across multiple healthcare sectors with a prime example being antibody-drug conjugates (ADCs). Although widely used lysine and hinge cysteine conjugation methods yield products, the lack of site-specificity and spatial control along with the highly heterogeneous composition are significant limitations. Herein, we describe a facile supramolecular assembly method based on heterodimer coiled-coil formation for site-specific antibody conjugation that perseveres antibody antigen binding sites for target engagement. The platform method affords uniform loading of diverse payloads including anti-cancer agents, polymers, enzymes, fluorophores, etc. under mild aqueous conditions. Further, the facile convergent approach capitalizes on the independent strengths and flexibility of protein expression and peptide chemistry culminating in a final self-assembly step. An ADC loaded with monomethyl auristatin E targeting ErbB2/Her2 positive tumors significantly reduces tumor volume in a human ovarian cancer xenograft model outperforming the antibody alone with validated performance against a best-in-class therapeutic.

Introduction

Antibody-drug conjugates (ADCs) are transformative treatments against cancer^1,2,3,4,5. Classical and still widely used bioconjugation methods for ADCs include reacting protein primary amines with an entity containing an activated ester or converting primary amines to thiols, via Traut’s reagent, or using hinge cysteines for subsequent coupling to a maleimide functionalized moiety^6,7,8. Although they are useful reactions, their selectivity and specificity are poor, with ill-defined, highly heterogeneous products being obtained from a single reaction^9,10. Alternative methods use unique reactive sites, such as a C-terminal thioester or N-terminal amine; however, yields remain low^11,12. Newer techniques rely on biorthogonal chemistry, the selective chemical transformation within a complex biological milieu^13,14, and include: (1) incorporation of a non-natural amino acid into the protein, such as azide-modified phenylalanine^15,16,17,18 or glycan remodeling^19,20,21 for a subsequent specific click-chemistry reaction (e.g., azide fragment with an alkyne functionalized molecule); (2) disulfide rebridging^22,23; and (3) enzymatic ligation reactions (e.g., AviTag, SorTag)^24,25,26,27 and split protein tags (e.g., SpyTag, Intein Tag)^28,29,30,31. Each of these methods possesses a number of advantages for the preparation of ADCs, but none have advanced to the preparation of clinically approved products^32,33.

A supramolecular assembly synthesis strategy offers benefits such as ease of use, selectivity, specificity, bio-orthogonality, and complementarity to other methods. Supramolecular assembly, the ability of smaller entities to spontaneously organize into well-defined larger structures through non-covalent interactions, is widely used in materials science, chemistry, and biological chemistry^{34,35,36,37,38}. Alpha-helical coil peptides are protein structural motifs capable of orthogonal and selective self-assembly into coiled-coil structures (Fig. 1A). Although they rely on non-covalent interactions, the presence of multiple contact points within the assembly significantly contributes to stability with dissociation constants in the sub-nanomolar range³⁹. Due to modular design and easily controlled molecular features, by varying the peptide sequences, coiled-coils are used in many biomedical applications^{40,41,42,43,44} including hydrogel fabrication^45,46, nanoparticle modification^47,48, fusion protein preparation^49,50, and cell therapy engineering^51,52.

Fig. 1: Antibody-receiving peptide characterization.

A AlphaFold illustration of a coiled-coil heterodimer formed from the receiving and docking peptides. B Schematic diagram illustrating the drug loading process to an antibody through the formation of a coiled-coil structure between the docking peptide conjugated to the drug and the receiving peptide fused to the heavy chain of an antibody. C SEC-RI chromatograms of native trastuzumab (αHER2) and trastuzumab fused to different receiving peptides (αHER2-P1, αHER2-P2, αHER2-P3, αHER2-P4). D Coomassie-stained SDS-PAGE gel analysis under non-reducing (left) and reducing (right) conditions of commercial trastuzumab (commercial αHER2), in-house produced native trastuzumab (αHER2), and trastuzumab fused to a coil peptide (αHER2-coil peptide). E ELISA assay demonstrating antigen-binding ability of native αHER2 and αHER2 fused to P3 peptide (αHER2-P3). Data are presented as mean ± s.d., n = 3 biological replicates. F Binding affinity of αHER2-P3 to biotinylated P4 peptide measured by biolayer interferometry (BLI). G ELISA measurements of binding of native αHER2 and peptide-modified αHER2-P3 antibodies to human FcRn receptor. Data are presented as mean ± s.d., n = 3 biological replicates. Source data are provided as a Source Data file.

Here, we report a biorthogonal platform based on supramolecular assembly of coiled-coil peptides for preparing antibody conjugates, including antibody-drug conjugates (ADCs). The Fc region of an IgG with two heavy chains is poised for a heterodimer coiled-coil supramolecular interaction: two C-terminal receiving peptides on the antibody assemble with two separately synthesized docking peptides containing the conjugated payload of interest (Fig. 1B). This facile synthetic strategy enables uniform stoichiometric site-specific antibody conjugation while maintaining antibody functionality. It is a convergent bioconjugation method where the final self-assembly step occurs under mild aqueous conditions, facilitating an easy, modular pairing of antibodies to payload(s), simplifying final production and purification. We site-specifically pair antibodies with fluorophores, cytotoxic drugs, peptides, polymers, nucleic acids, and enzymes. The resulting antibody conjugates retain function and bioactivity, and a prepared ADC excels in a murine xenograft human ovarian cancer model.

Results

Antibody-receiving peptide production and characterization

We identified two sets of peptides that form dimeric coiled-coil structures without self-interaction to create the bioconjugate platform⁵³. Each of the four peptide sequences (P1, P2, P3, and P4) comprises four amino acid heptad repeats containing different patterns of glutamic acid or lysine for promoting electrostatic interaction between the coils and leucine or isoleucine for hydrophobic interaction (Supplementary Table 1). We introduced a single unnatural azide-bearing amino acid into the C-terminus of these peptides to enable subsequent conjugation, generating a novel set labeled as - P1N₃, P2N₃, P3N₃, P4N₃. Small peptides bearing unnatural amino acids are readily synthesized via solid-phase peptide synthesis at milligram scale^54,55, unlike large antibodies that usually require elaborate optimization of genetic encoding and protein expression^16,56,57. Individually, the peptides do not form an organized structure but spontaneously form a heterodimeric coiled-coil structure when mixed at a 1:1 molar ratio with their pairing peptide (P1N₃ when mixed with P2N₃, and P3N₃ when mixed with P4N₃). The assembled coiled-coil structure is stable over time and under endosomal-lysosomal pH conditions (Supplementary Fig. 1).

Next, we designed a set of plasmids encoding the trastuzumab heavy chain fused at the C-terminal end with one of the receiving peptide sequences (P1, P2, P3, or P4; Supplementary Fig. 2). We expressed the antibodies in Expi293F cells and collected the protein via Protein A column chromatography. Incorporation of the peptide at the C-terminal end of the antibody does not impair antibody production, as titers are similar between peptide-modified and native antibodies and ranged from ≈40 to 80 mg/L (Supplementary Tables 2 and 3). A fully folded antibody possesses two identical peptide tags linked to the two heavy chains at the Fc region. The far-UV CD spectrum displays a broad minimum at 218 nm for the unmodified αHER2 antibodies and for the peptide-tagged antibodies (αHER2-P1, αHER2-P2, αHER2-P3, and αHER2-P4), characteristic of a significant presence of β-sheets in IgG (Supplementary Fig. 3), indicating that the coil tag does not interfere with the protein folding. The peptide-modified antibodies, except αHER2-P2, elute slightly earlier (αHER2-P1 t = 6.95 min; αHER2-P2 t = 7.15 min; αHER2-P3 t = 7.03 min; αHER2-P4 t = 7.00 min) than the native αHER2 antibodies (t = 7.15 min) during SEC chromatography, consistent with an increase in mass due to peptide insertion (Fig. 1C). The slower elution of αHER2-P2 is likely attributed to the lysine-rich P2 sequence, which also contributes to peak tailing. Antibodies modified with the P3 peptide display a uniform peak and were used for further experiments. SDS-PAGE analysis additionally confirms the modification of the heavy chain with the coil peptide (Fig. 1D, Supplementary Fig. 4A). Non-reduced antibodies with the inserted peptide exhibit a shift toward a higher molecular weight region compared to unmodified antibodies. Upon reduction, the light chain remains intact, while the heavy chain-coil peptide conjugate shifts to a higher molecular weight region, as anticipated based on the conjugate design. Finally, the MALDI-TOF MS spectrum reveals a major peak at 156 kDa for αHER2-P3, corresponding to the expected molecular weight of the full antibody (trastuzumab M_w ≈ 148 kDa) with two receiving peptides (P3 M_w ≈ 4 kDa each) (Supplementary Fig. 4B, C). To confirm the recombinant antibody binds the antigen (ErbB2/Her2 protein), we performed an ELISA assay and compared it to the commercial αHER2 without the peptide tag and αHER2 with a P3 receiving peptide tag. The EC₅₀ values for all tested samples are: 0.110 ± 0.024 nM and 0.133 ± 0.003 nM, respectively, indicating that the peptide tag does not compromise target binding of the antibody (Fig. 1E). We then measured the affinity of coiled-coil formation when one of the peptides is fused to the antibody. We immobilized the biotin-labeled docking peptide (P4N₃-biotin) on a BLI sensor and exposed it to αHER2-P3 receiving peptide antibody at various concentrations. A shift in the interference pattern occurs only between the complementary pair of a docking peptide (P4N₃) and the corresponding P3 receiving peptide on αHER2 antibodies (αHER2-P3). The calculated dissociation constant (Kd) is 6 × 10⁻¹⁰ M, indicating a highly stable structure and similar in magnitude as that reported by Plaper et al.⁵⁸ (Fig. 1F). Repeating the affinity experiment using microscale thermophoresis (MST), an immobilization-free assay format, affords similar Kd values in nanomolar range (Supplementary Fig. 5). The interaction between the docking and receiving peptides is specific as antibodies without a receiving peptide and those with a non- complementary receiving peptide tail, do not elicit a BLI response when mixed with the docking peptide (Supplementary Fig. 6). Next, we evaluated if conjugation at the Fc region interferes with its interaction with the neonatal Fc receptor (FcRn), which regulates IgG recycling. We plated native trastuzumab antibody or αHER2-P3 on an ELISA plate and incubated with biotin-labeled human FcRn receptor, followed by detection with streptavidin-HRP conjugates. The αHER2-P3 antibodies maintain their affinity to FcRn after introducing the coil; however, the binding is slightly lower than for native trastuzumab (Fig. 1G). Overall, these findings indicate that the introduction of a coiled-coil peptide tag preserves the key properties of the mAb and its antigen-binding capabilities.

Conjugation principle

With the introduction of a coil peptide tag preserving the properties of the mAb and its antigen-binding capabilities, we next investigated payload loading onto the antibody, which occurs in two steps: (1) attachment of the payload to the docking peptide via a strain-promoted azide-alkyne cycloaddition reaction; and (2) self-assembly of the payload-conjugated docking peptide with the receiving peptide on the antibody. This convergent approach enables expanded payload chemistry on the docking peptide independent of antibody modification, thereby separating the steps, minimizing the reactions with the antibody, and avoiding denaturation or aggregation of the monoclonal antibody in organic solvents and/or harsh pH conditions. Thus, this facile plug-and-play supramolecular assembly opens opportunities for novel chemistries and payloads to be conjugated to the docking peptide.

Docking peptide conjugation

To demonstrate the utility of the methodology, we prepared docking peptides linked with different entity classes to include a fluorophore (e.g., BP Fluor647), antimitotic agents mertansine (DM1) and monomethyl auristatin E (MMAE), a DNA oligo (e.g., polyA₁₅CCC), a polymer (e.g., PEG), an enzyme (e.g., HRP), a biotin molecule, and a phospholipid (e.g., DSPE). DBCO-modified payloads, except HRP, were mixed is a small excess (1.2 eq) with azide-bearing docking peptide (P4N₃) (Fig. 2A). The coupling reactions occur in minutes as monitored through depletion of DBCO absorbance peak at 310 nm (Fig. 2B). The estimated reactions half-life with P4N₃ is 26 ± 2 min for DBCO BP Fluor647, 12 ± 3 min for DBCO-PEG₄-DM1, 42 ± 10 min for DBCO-PEG₄-Val-Cit-MMAE, 77 ± 13 min for DBCO-N-polyA₁₅CCC oligo, 27 ± 4 min DBCO-mPEG_10k, 16 ± 2 min for DBCO-PEG₄-biotin, and 25 ± 8 min DBCO-PEG₄-DSPE (Fig. 2C–I). Mass spectrometry confirms the sequential conjugation and desired product formation (Supplementary Fig. 7). To enable assembly of the antibody-enzyme conjugate using this technology, we expressed a recombinant enzyme with the docking peptide sequence fused into the C-terminal of the peroxidase C1A. The activity of the engineered HRP-P4 enzyme is > 60 U/mg, as measured via an enzymatic reaction with the colorimetric 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Fig. 2J and Supplementary Fig. 8). The introduction of the peptide to the HRP enzyme doesn’t interfere with its activity and provides a handle for assembly with the receiving peptide.

Fig. 2: Docking peptide conjugation.

A Schematic illustration of azide-modified coil peptide reacting with DBCO-bearing payload B DBCO-PEG₄-biotin absorbance spectra change upon conjugation to P4N₃. At the beginning of the reaction (0 min), the intact DBCO-moiety absorbs in the UV spectrum with a characteristic maxima at 310 nm and stops absorbing when the structure is rearranged as a result of click conjugation. Conjugation kinetics of DBCO-modified payloads to azide-modified docking peptide P4N₃ at room temperature in PBS buffer solution (pH 7.4): C DBCO BP Fluor647, D DM1-PEG₄-DBCO, E MMAE-CitVal-PEG₄-DBCO, F DBCO-mPEG (M_w ~ 10,000), G DBCO-PEG₄-biotin, H DBCO-PEG₄-DSPE, I 5’-DBCO-N-polyA₁₅GGG. Data are presented as mean ± s.d., n = 3 biological replicates. J TMB-based HRP activity assay measuring the activity of recombinant HRP-P4 conjugate compared to commercially available native HRP (comHRP). Data are presented as mean ± s.d., n = 3 biological replicates. Source data are provided as a Source Data file.

Coiled-coil mediated assembly of antibody-payload conjugates and their functions

We then assembled the antibody conjugates by simply mixing the antibody-receiving peptide (αHER2-P3) with the payload-modified docking peptide in PBS for one hour, followed by spin column centrifugation to isolate the desired product (Fig. 3A). To confirm the conjugation and the payload’s activity, we performed functional assays. The antibody conjugate with the BP Fluor647 payload absorbs at 647 nm while the native antibody does not (Supplementary Fig. 9). Based on absorbance measurements at 280 nm and 647 nm, the fluorophore-to-antibody ratio is 2.02, consistent with the designed coiled-coil assembly. Additionally, mixing αHER2-P3 with a solution containing P1, P2, P3, and P4 peptides labeled with different fluorophores (P1-BP Fluor647; P2-TAMRA 555; P3-BP Fluor 594 and P4-BP Fluor488) results in a conjugate with an absorbance in the BP Fluor488 region only, indicating selective assembly of P4-BP Fluor488 and αHER2-P3 (Supplementary Fig. 10). These data are consistent with the results with the coiled-coils themselves⁵³. The antibody-fluorophore conjugate specifically targets ErbB2/HER2-positive cells in an ErbB2/Her2 dependent manner as determined by flow cytometry using this fluorescent tag (Supplementary Fig. 11).

Fig. 3: Coiled-coil mediated assembly of antibody-payload conjugates.

A Schematic illustration of an antibody reacting with payload-modified coil peptide to give antibody-payload conjugates (room temperature (RT); 1 h of incubation). B Direct ELISA assay assessing the binding and enzyme activity of αHER2-HRP conjugate. Data are presented as mean ± s.d., n = 3 biological replicates. C UV-Vis analysis of antibody conjugated to polyA₁₅GGG oligo annealed with a fluorophore-labeled polyT₁₅CCC-TAMRA oligo. D Cytotoxicity of the αHER2 ADC containing MMAE drug against SKBR-3 cells compared to equivalent unloaded αHER2-P3 antibody. Data are presented as mean ± s.d., n = 3 biological replicates. E Schematic illustration of an antibody reacting with two orthogonal payload-modified coil peptides to afford an antibody bearing two different payloads under the same conditions. UV-Vis analyses of antibodies assembled with a single F Fluor488 payload or G Fluor647 payload and H dual-loaded conjugates. Source data are provided as a Source Data file.

To evaluate the αHER2-HRP conjugate activity, we conducted a direct ELISA using ErbB2/Her2 protein immobilized on a plate and TMB as the HRP substrate (Fig. 3B). The αHER2-HRP conjugate catalyzes the oxidation of TMB in a dose-dependent manner, demonstrating that, despite the presence of a large payload, αHER2-HRP retains its binding affinity to ErbB2/Her2 post-conjugation, while also maintaining HRP enzymatic activity. For the antibody-PEG_5k-DSPE lipid, mPEG_10k polymer, and biotin conjugates, we performed indirect ELISA assays using anti-PEG secondary antibodies or streptavidin-HRP for payload detection. All the assays confirm the successful conjugation of antibodies with different payloads (Supplementary Fig. 12). For oligonucleotide-loaded antibodies, we mixed the single-stranded DNA-loaded (polyA₁₅GGG) antibody with a complementary TAMRA fluorophore-labeled strand (polyT₁₅CCC-TAMRA) and incubated the mixture to allow hybridization, followed by purification via spin filtration. The results confirm successful hybridization of the complementary oligos, as indicated by the TAMRA fluorophore absorbance of the final product, demonstrating the capability of formation of a double-stranded oligonucleotide structure post-conjugation to the αHER2 antibody (Fig. 3C and Supplementary Fig. 13). Finally, we assessed the cytotoxicity of the synthesized αHER2-MMAE conjugates using the MTS assay against three cell lines with different ErbB2/Her2 expression levels: 4T1 (Negative ErbB2/Her2−), A549 (Low ErbB2/Her2+), SKOV-3 (Medium-high ErbB2/Her2++) and SKBR-3 (High ErbB2/Her2+++). We treated cells with varying concentrations of ADCs, unloaded αHER2 antibodies, or free MMAE drug for 72 h (Fig. 3D and Supplementary Fig. 14). Treatment with the ADC shows a dose-dependent and ErbB2/Her2 expression-dependent decrease in cell viability for SKBR-3 (ADC EC₅₀ ~ 0.002 μM), SKOV-3 (ADC EC₅₀ ~ 0.09 μM), and A549 cell lines, while showing minimal toxicity in ErbB2/Her2 negative 4T1.

Dual-payload loaded antibody conjugates

An additional beneficial feature of this coiled-coil design is that two or more, similar or different, entities can be attached to the antibody by extending the coiled-coil structure with the same peptide or an orthogonal peptide sequence set. To demonstrate the assembly of dual-payload loaded antibodies, we fused receiving peptide P1, followed by another receiving peptide P3 sequence, into the heavy chain of trastuzumab, resulting in antibodies with two sequential orthogonal conjugation sites (Fig. 3E and Supplementary Fig. 15). As a model system, we chose DBCO-modified BP Fluor488 (Fluor488) and DBCO-modified BP Fluor647 (Fluor647), which we conjugated to azide-modified P2N₃ and P4N₃ peptides, respectively, resulting in P2-Fluor488 and P4-Fluor647 conjugates. Coil peptide P1 forms an orthogonal coiled-coil structure with peptide P2, while peptide P3 forms a dimer with peptide P4. Next, we mixed P2-Fluor488 or P4-Fluor647 with the antibodies separately. After one hour of incubation and spin-filtering purification, we obtained the conjugates bearing either Fluor488 or Fluor647 fluorophores. The absorbance spectra of the conjugation products exhibit maxima at 280 nm, corresponding to protein absorbance, and either at 488 nm or 647 nm, depending on which payload is loaded, indicating that the extended peptide tag structure does not interfere with payload loading (Fig. 3F, G). We then mixed both P2-Fluor488 and P4-Fluor647 with antibodies bearing P1 and P3 peptides. The corresponding peptide pairs self-assemble, leading to the formation of dual-loaded antibody conjugates both with Fluor488 and Fluor647 (Fig. 3H). This approach offers the opportunity to increase the drug-antibody ratio or load two different payloads in a defined stoichiometric ratio and at a specific location on the antibody, enabling the development of new co-delivery strategies.

Stability of antibody-payload conjugates

To evaluate the stability of the proposed conjugation method, we compared the stability of conjugates formed via conventional thiol-maleimide covalent linkage with those generated through coiled-coil-mediated conjugation. In addition to the P3 and P4 pair (4 repeating amino acid heptads, total 28-amino-acid-long coil peptides) described above, we investigated two additional peptide pair designs based on prior work on coiled-coil interactions. Woolfson et al. report that increasing the number of heptads in a coiled-coil structure increases the stability affording nM or lower Kd values³⁹. Also, Xia et al. and Lee et al. show that the presence of complementary reactive groups capable of forming a bond (e.g., a disulfide) between the peptides stabilizes the coiled-coils^50,59. Based on this knowledge, for the first design, we extended the P3 and P4 peptide sequence by one coiled-coil heptad pattern, creating 35-amino-acid-long peptides (P3_5mer and P4_5merN₃, see Supplementary Table 1). We used the same repeating units for the second design, except that we replaced the isoleucine with a cysteine residue in the last heptad unit of each peptide to allow a disulfide bond formation between coils (P3_5mer-cys and P4_5mer-cysN₃, see Supplementary Table 1). A MALDI laser ionization spectrum of the coiled-coil lacking cysteines (P3_5mer:P4_5mer) shows a primary peak at a m/z of ~5000 – the expected molecular weight of a single peptide - consistent with disruption of the coiled-coil non-covalent interactions. In contrast, the spectrum for the cysteine-containing dimer (P3_5mer-cys:P4_5mer-cys) reveals a high-intensity peak at m/z of ~10,000. This peak reduces upon addition of TCEP, consistent with cleavage of the disulfide bond between the peptides, with more resulting single peptides present – analogous to the results obtained with coiled-coil lacking cysteines (Supplementary Fig. 16). We produced αHER2-P3, αHER2-P3_5mer, and αHER2-P3_5mer-cys antibodies (Supplementary Fig. 17) and assembled them with corresponding docking peptides conjugated to DBCO BP Fluor647. We then incubated the antibody-fluorophore conjugates (AFCs) in human serum at 37 °C in 5% CO₂ atmosphere. We collected serum samples over a 28-day period and analyzed them for intact AFC levels using an ELISA assay that detects the payload portion of AFC. The AFC prepared by the standard thiol-maleimide hinge conjugation technique degrades by 25%, while the coiled-coil-based conjugates showed varying stability levels depending on the coil design (Fig. 4A, B). The P3:P4-N₃ coils (short 4-heptad sequence) disassemble over the first few days, with only 13% of the AFC intact by the end of the study. In contrast, the AFCs prepared using P3_5mer:P4_5merN₃, and P3_5mer-cys:P4_5mer-cysN₃ coil pairs are stable with only a 4% loss in structure over 4 weeks, outperforming the conventional conjugation approach (Fig. 4C, D). From a design perspective, we selected the 5-heptad construct with the disulfide bond for the subsequent in vivo experiments because it possesses two stabilizing mechanisms.

Fig. 4: Stability and efficacy of antibody-drug conjugates.

A Stability of conventionally prepared AFCs via hinge cysteine conjugation, B via short 28-amino-acid-long coiled-coil peptides, C via long 35-amino-acid-long coiled-coil peptides, and D via long 35-amino-acid-long coiled-coil peptides with cysteine residue substitution in human plasma incubated at 37 °C. A–D Data are presented as mean ± s.d., n = 3 biological replicates. E Schematic illustration of ADC conjugation strategy and molecular structure of the linker-payload consisting of valine-citrulline (Val-Cit) cleavable portion and self-immolative p-aminobenzyl carbamate (PAB) moiety. F PK of unmodified trastuzumab (αHER2), unloaded ADC (αHER2-P3_5mer-cys), and fully assembled coiled-coil ADC (ADC) in female Balb/c. Blood was collected at the indicated time points, and the total amount of human IgG was quantified by indirect ELISA using a standard curve. Data are presented as mean ± s.d., n = 3 biological replicates for the trastuzumab group, n = 4 for ADC, and n = 2 for the unloaded ADC group. G Ex vivo organ biodistribution of AFC injected i.v. at a 10 mg/kg dose into SKOV-3 tumor-bearing BALB/c nude mice at different time points. Data are presented as mean ± s.d., n = 3 biological replicates. The total radiant efficiency was calculated using IVIS software, keeping the area of the ROI the same across all the samples. H Representative fluorescent picture of ex vivo organs post AFC injection. Color scale: min = 1.05 × 10⁹, max = 1.1 × 10¹⁰ $(\frac{p/\sec \frac{/{{cm}}^{2}}{{sr}}}{\mu W/{{cm}}^{2}})$. I Orthotopic xenograft model of ErbB2/Her2-positive cancer study. Female NSG mice were engrafted with 2 million (2 M) SKOV-3 tumor cells by intraperitoneal injection (IP), were randomized on day 14, and treated with a single intravenous (IV) dose of PBS or trastuzumab antibody (10 mg/kg) or coiled-coil ADC (10 mg/kg) or high dose coiled-coil ADC (2xADC, 20 mg/kg) or disitamab vedotin (10 mg/kg), or trastuzumab vedotin (10 mg/kg), or disitamab antibody (10 mg/kg). All mice were euthanized at week 7 post-treatment, when PBS-treated mice exhibited severe hunching posture and signs of morbidity, to record the endpoint tumor volume and mass. J Average mouse body weight for selected treatment groups over the observation period. Data are presented as mean ± s.d., n =  5 biological replicates. K Exemplary images of harvested tumors at study termination. L Recorded tumor size, and M tumor mass. Data are presented as mean ± s.d., n = 5 biological replicates for Trastuzumab, Trastuzumab vedotin, ADC, 2xADC, and Disitamab treatment groups, n = 10 biological replicates for Disitamab vedotin, and n = 13 for PBS group. Statistical significance determined by the Mann–Whitney U test. Source data are provided as a Source Data file.

Efficacy of antibody-drug conjugates in vivo

Given our interest in ovarian cancer and the dearth of available treatments^60,61,62,63, we prepared MMAE-loaded ADC conjugates to assess the overall therapeutic potential of the coiled-coil conjugation. We conjugated P4_5mer-cysN₃ to DBCO-PEG₄-Val-Cit-PAB-MMAE (Fig. 4E and Supplementary Fig. 18A), where the Val-Cit-PAB portion ensures scar-less drug release⁶⁴. MMAE-peptide loading to the αHER2-P3_5mer-cys antibody results in a SEC trace shift (Supplementary Fig. 16B) with a drug-antibody ratio of two as determined by UV-vis absorbance at 247 nm and 280 nm^65,66 (Supplementary Fig. 16C). An indirect ELISA using anti-MMAE secondary antibodies additionally confirms the modification of αHER2-P3_5mer-cys with MMAE drug compared to unloaded antibodies (Supplementary Fig. 19).

First, we evaluated the pharmacokinetic (PK) profile of the ADC in vivo. We intravenously (i.v.) administered native mAb (trastuzumab), unloaded αHER2-P3cys mAb, and MMAE-loaded ADC at a dose of 10 mg/kg to healthy Balb/c mice. We collected blood samples at predefined time points and measured the concentration of mAb in the blood by an ELISA assay. The half-life for trastuzumab (native αHER2) is 6.5 h (95% CI 5.5–7.3 h), similar to the unloaded ADC (αHER2-P3_5mer-cys) and MMAE-loaded ADC conjugates at 6.8 (95% CI 6.2–7.4 h) and 6.5 (95% CI 5.0–7.7 h) hours, respectively (Fig. 4F). The coiled-coil peptide tag does not negatively affect antibody PK and clearance in the mouse.

To assess tissue biodistribution of the antibody conjugate, we used an orthotopic xenograft model of ErbB2/Her2-positive human ovarian tumors by implanting SKOV-3 cells intraperitoneally into immunodeficient BALB/c nude mice. When tumors are established, we i.v. administered a BP Fluor647-loaded AFC at a dose of 10 mg/kg. At 1 h, 1 day, 3 days, 5 days, and 7 days post-injection, we euthanized the animals and quantified the fluorescent signal in the tumor and major organs by IVIS. (Fig. 4G, H and Supplementary Fig. 20). The AFC accumulates in the tumor throughout the 7-day study period, along with the liver and kidneys at the earlier time points. These results are consistent with the literature^67,68 and demonstrate the ability of the coiled-coil-based antibody conjugate to deliver a payload to ErbB2/Her2-positive tumors.

To identify an effective dose, we performed a dose–response study. We intraperitoneally administered 2 × 10⁶ SKOV-3-Fluc cells to immunodeficient BALB/c nude mice and treated two weeks post-implantation with ADC doses of 1, 3, 5, or 10 mg/kg, or with PBS as a negative control (Supplementary Fig. 21). Treatment with the ADC at 1 mg/kg does not result in IVIS signal reduction. Doses of 3 and 5 mg/kg show a modest reduction in IVIS signal, whereas treatment at 10 mg/kg leads to a reduction in IVIS signal consistent with tumor regression. Based on these results, we selected a dose of 10 mg/kg for the subsequent in vivo efficacy studies. For comparison, the effective dose for disitimab vedotin is between 5 and 10 mg/kg^69,70, and for trastuzumab vedotin is between 3 and 10 mg/kg^71,72 in gastric and ovarian cancer models in mice.

Using the established orthotopic xenograft model (Fig. 4I), we performed an efficacy study and i.v. administered a single dose of 10 mg/kg of trastuzumab (native αHER2 antibodies) or 10 mg/kg of coiled-coil ADC (n = 5 per group). Additionally, we administered PBS as the negative control and two positive controls: (1) trastuzumab vedotin and (2) disitamab vedotin—the latest HER2-targeting ADC to advance to Phase III clinical trials⁷³. Trastuzumab vedotin and disitamab vedotin consist of the trastuzumab and hertuzumab antibody, respectively, and, similar to our ADC, use the cytotoxic payload MMAE linked via a Val-Cit cleavable linker. However, both commercial antibodies employ cysteine-based conjugation, resulting in approximately four MMAE molecules per antibody. To match the drug dose of the positive control groups, we also administered a double dose of coiled-coil ADC (20 mg/kg, n = 5 per group). The antibody vehicle controls for this study are trastuzumab mAb and disitamab mAb. All groups tolerated the treatment with no significant changes in body mass (weight loss of >15%) (Fig. 4J and Supplementary Fig. 22) or observable adverse effects on animal behavior at earlier time points. However, at week 7, the PBS treatment group and one trastuzumab-treated animal exhibited hunched posture and signs of gait rigidity. We euthanized the mice on week 7 post single dose treatment, and tumor size and mass were measured. Compared to the PBS-treated animals, all treatment groups significantly reduce tumor size and mass (Fig. 4K–M and Supplementary Fig. 23). Mice treated with 10 mg/kg coiled-coil ADC exhibit a significant reduction in tumor size, with a 42-fold decrease in tumor mass and 37-fold decrease in tumor volume (m = 13 ± 9 mg, V = 24 ± 18 mm³) compared to the untreated control (m = 542 ± 322 mg, V = 880 ± 479 mm³). In contrast, those receiving the antibody vehicle controls (trastuzumab mAb or disitamab mAb) show only a 10-fold or 5-fold and 5-fold or 6-fold reduction in tumor mass and volume, respectively (m = 52 ± 27 mg, V = 166 ± 81 mm³ for trastuzumab and m = 78 ± 32 mg, V = 145 ± 97 mm³ for disitamab mAb). Treatment with 20 mg/kg ADC, 10 mg/kg trastuzumab vedotin, or 10 mg/kg disitamab vedotin completely eradicates tumor in one mouse from each group, with average tumor masses of 5 ± 3 mg, 6 ± 7 mg, and 2 ± 3 mg, respectively, with a similar trend in the tumor volume: V = 6 ± 5 mm³, V = 9 ± 11 mm³ and V = 4 ± 4 mm³, respectively (Fig. 4K–M and Supplementary Fig. 23). Disitamab vedotin and trastuzumab vedotin are highly efficacious, consistent with its elevated drug-to-antibody ratio for both ADCs and enhanced affinity for the HER2 target of disitamab vedotin⁷⁴. Our comparable results from the coiled-coil ADC treatment group bode well for further use of the coiled-coil conjugation method for antibody conjugation.

Finally, we confirmed the dependency of the therapeutic efficacy of the ADC on ErbB2/Her2 expression levels in vivo. We compared ErbB2/HER2 expression in human lung A549 and SKOV-3 cell lines, as well as in their corresponding tumor tissues. SKOV-3 cells exhibit an 83-fold higher ErbB2/Her2 targeting at the cellular level compared to the isotype control and an 11-fold increase for tumor tissues, whereas for A549 both for the cell line and tumor tissue the ErbB2/Her2 targeting is modest, with 1.6-fold and 2.5-fold increases, respectively (Supplementary Fig. 24). We then established an A549 xenograft model via subcutaneous implantation. At the 10 mg/kg i.v. dose, neither the trastuzumab antibody (mAb), MMAE coiled-coil ADC, nor disitamab vedotin demonstrates any significant tumor growth inhibition, consistent with the low ErbB2/Her2 expression levels observed in A549 cells and tumors (Supplementary Fig. 26).

Discussion

During the writing of this article, but following our issued patent^75,76, two articles were published on using coiled-coil peptides for antibody modification, further supporting this strategy. Nyesiga et. al.⁷⁷ report an approach to antigenic peptide delivery. A docking peptide fused to an OVA peptide self-assembles with an anti-CD40 mAb containing the receiving peptide. The antigenic peptide conjugated mAb binds CD40-expressing cells in vitro and expands OVA peptide-specific CD8+ T cells when administered to human CD40 transgenic mice. The antigenic peptide conjugated mAb reduces tumor volume in a subcutaneous B49-epCam-OvA tumor model; however, administration of the individual anti-CD40 mAb and OVA peptide shows greater efficacy. Baniahmad et. al.⁷⁸ employ this coiled-coil approach to prepare a fluorophore-labeled antibody for tumor targeting. A docking peptide conjugated to a near-infrared dye (CF750), via its thiol residue, joins with the anti-HER2 mAb containing the receiving peptide. Unfortunately, the fluorophore-labeled antibody undergoes partial payload dissociation after in vivo administration. In both reports, the authors note the limited performance and the need for further refinement. We similarly observed this limitation early in the project and, thus, we lengthened the coiled-coil region and introduced a disulfide linkage between the peptides to improve stability. As shown in Fig. 4, the extended coiled-coil sequence itself provides a robust structure, as does the one with the disulfide linkage, as a supplemental stabilization strategy. Incorporation of these two changes in the coiled-coil design affords a functional ADC for in vivo use. However, if incomplete disulfide linkage forms (e.g., unfavorable pH or steric limitations), the cysteines may cause heterogenicity introducing a possible limitation to the ADC. Another potential limitation of a coiled-coil design is that it may induce a host immune response even though short peptides are usually non-immunogenic^74,75. Ljubetič et al. report that a complex coiled-coil protein origami structure, formed in part from modified-like P3 and P4 peptides, induces minimal immune activation in vitro and causes no signs of inflammation or other side effects in vivo⁷⁶. Therefore, we do not expect our coiled-coils to be strongly immunogenic, but this will need to be tested.

In summary, we describe a facile method for site-specific antibody conjugation that enables uniform loading of distinct payloads under mild aqueous conditions. Coiled-coil conjugation perseveres both antibody antigen-binding sites for target engagement and heavy chains constant domains for Fc binding and recycling. The convergent approach bifurcates the conventional antibody method, thereby minimizing the reaction steps and conditions exposed to the antibody while increasing the synthetic flexibility for payload conjugation as it employs peptide-based chemistry. In doing so, this strategy also provides access to readily synthesizing several antibody conjugates or a library of antibody conjugates, whereby one or more antibodies are prepared, followed by mixing and matching with different conjugates. Additionally, the step-wise assembly of the coiled-coils affords relative spatial control over the site-specific location of multiple conjugates on the antibody. The coiled-coil composition controls conjugate stability and structures with stability beyond that of conventional hinge cysteine conjugation, which is known to prematurely release payload. An MMAE-ADC targeting ErbB2/Her2+ tumors, prepared via self-assembly, significantly reduces tumor volume in a human ovarian cancer xenograft model, outperforming the antibody alone, with validated similar performance against best-in-class therapeutics. Heterodimer coiled-coil driven bioconjugation opens new avenues for well-defined stoichiometric site-specific antibody labeling with one or more payload compositions, and as a supramolecular chemistry approach expands existing bio-orthogonal strategies.

Methods

Data representation

Data represented as mean ± s.d. with biological replicates.

Cells

SKOV-3 (cat.# HTB-77), SKBR-3 (cat.# HTB-30), A549 (cat.# CRM-CCL-185), and 4T1 (cat.# CRL-2539) cell lines were obtained from American Type Culture Collection (ATCC, USA). Cell media supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin was used to maintain the cell lines: McCoy’s 5A Medium (ATCC 30-2007) for SKBR-3 and SKOV-3, F-12K Medium (ATCC 30-2004) for A549, and RPMI 1640 (ATCC 30-2001) for 4T1. The cell lines were cultured at 37 °C in a humidified CO₂ (5%) incubator. Expi293F cells (cat.# A14527, Gibco™) were cultured in Expi293™ Expression Medium (Gibco™) at 37 °C under constant shaking (120 r.p.m.) in a humidified CO₂ (8%) incubator.

Plasmids

The pcDNA3.1(-) backbone (Invitrogen, cat.# V79520) was pre-cut with EcoRV and BamHI restriction enzymes (NEB) and purified using the QIAquick PCR Purification Kit (Qiagen). The 5′ ends were dephosphorylated with Quick CIP (NEB) to prevent re-ligation. The gBlocks of interest were synthesized by Integrated DNA Technologies (IDT) and inserted into the prepared pcDNA3.1(-) backbone via Gibson Assembly (NEB). The assembled plasmids were transformed into NEB Stable Escherichia coli and purified using the ZymoPURE II Plasmid Midiprep Kit (Zymo Research). Sequence confirmation was performed by Sanger sequencing using CMV-F and BGH-R primers (Genewiz).

Protein expression and purification

Native αHER2 and receiving-peptide-modified αHER2 antibodies, as well as HRP-P4 conjugates, were produced in the Expi293F expression system (Gibco™) according to the manufacturer’s protocol. Expi293F cells were co-transfected with a mixture of light and heavy chains containing plasmids for mAb production or with a single plasmid containing HRP-5xHis docking peptide fusion sequence. Proteins were harvested on days 6–7 by centrifugation at 3220 × g for 20 min. The supernatant was filtered through a 0.22-µm membrane for subsequent purification and mixed 1:1 with equilibration buffer. mAbs were purified using a gravity-flow column (Bio-Rad) packed with rProtein A Sepharose Fast Flow resin (Cytiva), according to the resin manufacturer's protocol. Phosphate-buffered saline (pH = 7.4) was used as equilibration and wash buffer, 0.2 M glycine-HCl (pH = 2.5) and 1 M TRIS-HCl (pH = 9.0) were used as elution and neutralization buffers, respectively. HRP-P4 was purified using HisPur™ Ni-NTA Resin (Thermo Fisher Scientific) according to the resin manufacturer's protocol. High-salt phosphate buffer (20 mM sodium phosphate, 300 mM sodium chloride) with 10 mM imidazole (pH = 7.4) was used as equilibration buffer, high-salt phosphate buffer with 25 mM imidazole (pH = 7.4) and high-salt phosphate buffer with 250 mM imidazole (pH = 7.4) were used as wash and elution buffers, respectively. Collected proteins were concentrated and buffer-exchanged to PBS with centrifugal filters (10 kDa or 50 kDa molecular weight cut-off, MilliporeSigma). The concentration of all proteins was determined by absorbance at 280 nm using sequence-predicted extinction coefficients (mAb ε₂₈₀ = 216,340 M⁻¹ cm⁻¹; and HRP-P4 ε₂₈₀ = 15,065 M⁻¹ cm⁻¹).

Proteins characterization

Size exclusion chromatography (SEC-RI)

Antibodies were analyzed using an AdvanceBio SEC column (130 Å, 2.7 µm, 4.6 × 150 mm Agilent Technologies). The column was equilibrated and operated with phosphate-buffered saline (PBS, pH 7.4) as the mobile phase at room temperature under isocratic conditions at 0.7 mL/min flow rate, and elution was monitored by an RI detector (Agilent 1260 Infinity II).

SDS-PAGE gel

Produced proteins (αHER2, αHER2-coil peptide, HRP-P4) were separated by SDS polyacrylamide gel electrophoresis (SDS-PAGE). 1 μg of each protein was mixed with gel loading buffer (reducing or non-reducing) and loaded into wells of a 4–15% precast gel (Bio-Rad). The gel was subjected to 120 constant volts for 40–60 min. Following this, the gel was stained with Coomassie Blue (Thermo Fisher Scientific).

Circular dichroism (CD)

Antibodies and peptides were prepared in PBS buffer (pH 7.4) at a concentration of 0.2 g/L. The measurements were performed on a CS/2 Chirascan (Applied Photophysics) CD spectrometer using a 1 mm QS 100 (Hellma) cuvette. PBS solution was used as a blank. To study the stability of the coiled-coil assembly at different pH levels, the peptides were buffer-exchanged using 3 kDa MWCO spin filters (MilliporeSigma) into a PBS solution adjusted with HCl to pH 5.4, 6.4, or 7.4. For the overtime stability study of the P3-P4 coiled-coil assembly, the peptides were incubated at 37 °C between measurements. The proportion of intact coiled-coil assemblies relative to single α-helices was estimated by calculating the ratio of ellipticity at 222 nm to that at 208 nm⁷⁹.

Recombinant antibodies antigen-binding assay

The ability of recombinant antibodies to recognize ErbB2/Her2 protein was evaluated using indirect ELISA. Nunc MaxiSorp 96-well ELISA plates were coated overnight at 4 °C with 1 μg/mL of recombinant human ErbB2/Her2 protein (Cat. #10126-ER, R&D Systems Inc.) in 0.05 M carbonate–bicarbonate buffer (pH 9.6). Plates were then blocked with 5% bovine serum albumin (BSA) in PBS and washed with PBS containing 0.05% Tween-20 (PBS-T) between each step, following a standard ELISA protocol. Serial dilutions of native αHER2 (Human ErbB2/Her2 Antibody, Cat. # MAB9589-100, R&D Systems Inc) or antibodies fused to a receiving peptide (αHER2-P3) were prepared in blocking buffer (starting at 3 μg/mL). Wells without antigen were used as negative controls. Rabbit anti-Human IgG (H+L) Secondary Antibody (1:8000, Invitrogen, Cat. #A18903) was used as the secondary antibody. After washing, 100 μL of TMB substrate (1-Step™ TMB ELISA Substrate Solution, Thermo Scientific™) was added to each well. Following incubation, the reaction was stopped with 100 μL of 2 N sulfuric acid. Absorbance was measured at 450 nm using a microplate reader (SpectraMax iD3, Molecular Devices).

In vitro FcRn recognition assay

The interaction between recombinant antibodies with fused receiving peptides was studied according to the SinoBiological activity protocol for recombinant human FCGRT & B2M heterodimer protein (Cat. # CT071-H27H-B, SinoBiological). Nunc MaxiSorp 96-well ELISA plates were coated overnight at 4 °C with 2 μg/mL of native αHER2 or antibodies fused to receiving peptide αHER2-P3 in 0.05 M carbonate–bicarbonate buffer (pH 9.6). Plates were then blocked with 5% bovine serum albumin (BSA) in PBS and washed with PBS containing 0.05% Tween-20 (PBS-T) between each step. Biotinylated recombinant human FcRn (FCGRT & B2M) heterodimer (SinoBiological, Cat. #CT071-H27H-B) was serially diluted two-fold (starting at 5 µg/mL) in dilution buffer (PBS, pH 6.0, containing 0.2% Tween-20 and 0.1% BSA) and incubated in an ELISA plate for 1 h at 37 °C. A solution of streptavidin-HRP (1:40, R&D Systems Inc. part # 893975) in dilution buffer (0.5% BSA in PBS-T, pH = 6.0) was used for detection. After washing steps, 60 μL of TMB solution (1-Step™ TMB ELISA Substrate Solutions, Thermo Scientific™) was added to the wells. The reaction was stopped with 60 μL of 2 N sulfuric acid. Absorbance was measured at 450 nm using a microplate reader (SpectraMax iD3, Molecular Devices).

Biolayer interferometry (BLI)

P2N₃, P3N₃, and P4N₃ peptides were first biotinylated with DBCO-PEG₄-biotin. Streptavidin SA tips were baselined in BLI buffer (28 mM Tris-HCl, pH 8.0, 5 mM MgCl, 4.25% Glycerol, 25 mM NaCl, 1.67 mg/mL BSA) and dipped into 100 nM docking peptide solution for 60 s. Then the peptide-coated tips were baselined in BLI buffer again and dipped into wells containing αHER2-P3 (12.5 nM, 25 nM, 50 nM, and 150 nM) or native αHER2 (100 nM) or αHER2-P2 (100 nM). After reaching the equilibrium tips were dipped in BLI buffer for complex dissociation. The assay was performed using ForteBio OctetRed96, and data analysis was performed in Octet Analysis Studio 13.0. Tips with no docking peptide loading were used as an association control, and tips with no dipping into antibody solutions were used as a dissociation control. Trace Y-Axes were aligned to the last 5 s of the second baseline step, and data was fit to a Mass Transport model.

Docking peptide synthesis

Docking peptides were synthesized by ABclonal (Woburn, MA) using solid-phase synthesis with >95% purity and verified by HPLC and mass spectrometry by the manufacturer.

Payload loading to the docking peptide

The DBCO-bearing payloads were dissolved according to manufacturer recommendations: BP Fluor647 DBCO in DMF, BroadPharm; DM1-PEG₄-DBCO in DMF, BroadPharm; DBCO-PEG₄-Val-Cit-MMAE in DMF, BroadPharm; DBCO-N-polyA₁₅GGG in water, Integrated DNA Technologies; DBCO-mPEG_10k and DBCO-PEG_2k-OH in water, Biopharma PEG; DBCO-biotin in DMF, BroadPharm; DBCO-PEG₄-DSPE and DBCO-PEG_5k-DSPE in DMF:H₂O:EtOH 1:1:2, BroadPharm. The small excess of DBCO-bearing payload (1.2 eq.) was added dropwise to a solution of P4 docking peptide in PBS (20 ug). The absorbance at 310 nm was measured immediately using a NanoDrop spectrophotometer.

MALDI-TOF analysis of peptides and antibodies

Antibodies, peptides, and their conjugates were analyzed using a Bruker Autoflex Speed MALDI-TOF mass spectrometer with sinapinic acid (SA) or CHCA (α-Cyano-4-hydroxycinnamic acid) as the matrix (10 mg/mL in 50% acetonitrile and 0.1% trifluoroacetic acid in water). Prior to analysis, samples were buffer-exchanged into ultrapure water using centrifugal filter units (50 kDa MWCO for antibodies and 3.5 kDa MWCO for peptides; Millipore). The resulting solutions were mixed 1:1 (v/v) with the SA or CHCA matrix solution immediately before spotting onto the MALDI target plate.

HRP-P4 docking peptide activity

Serial 10-fold dilutions of commercially available HRP (Sigma, CAS #9003-99-0) were prepared in phosphate-buffered saline (in PBS; 1 mg/L - 0.001 μg/L) as standards. A volume of 60 µL from each dilution or recombinant HRP-P4 was transferred into a well of a flat-bottom 96-well microplate. Subsequently, 60 µL of 3,3′,5,5′-tetramethylbenzidine (TMB) solution (1-Step™ TMB ELISA Substrate Solutions, Thermo Scientific™) was added to each well and absorbance at 650 nm was measured immediately using a microplate reader (SpectraMax iD3, Molecular devices) at one-minute intervals. In 15 min the reaction was stopped by the addition of 60 μL of 2 N sulfuric acid, and absorbance at 450 nm was recorded. Recombinant HRP-P4 activity was quantified by comparing the endpoint values of recombinant HRP samples to a standard curve generated using serial dilutions of a commercially available HRP standard.

Antibody-payload assembly

Solution of αHER2-docking peptide (1 eq., in PBS) was mixed with corresponding payload-modified peptide (2.2 eq., in PBS) and incubated for 1 h at RT or overnight at 4 °C to let the coiled-coil peptides self-assemble. The solution was centrifuged using a spin filter to isolate the conjugated antibody-payload.

Conjugates functional assays

HRP conjugates functional assay

Activity of the αHER2-HRP conjugates was assessed by direct ELISA. Nunc MaxiSorp 96-well ELISA plates were coated overnight at 4 °C with 2 μg/mL of recombinant human ErbB2/Her2 protein (Cat. #10126-ER, R&D Systems Inc.) in 0.05 M carbonate–bicarbonate buffer (pH 9.6). Plates were then blocked with 5% bovine serum albumin (BSA) in PBS and washed with PBS containing 0.05% Tween-20 (PBS-T) between each step, following a standard ELISA protocol. The αHER2-HRP conjugates were serially diluted (two-fold) in blocking buffer (2 μg/mL - 1 ng/mL, conjugate concentration determined via nanodrop using ε₂₈₀ = 237,385 M⁻¹ cm⁻¹) and incubated in ELISA plate for 1 h at 37 °C. After washing steps, 60 μL of TMB solution (1-Step™ TMB ELISA Substrate Solutions, Thermo Scientific™) was added to the wells. After 30 min of incubation, the reaction was stopped with 60 μL of 2 N sulfuric acid. Absorbance was measured at 450 nm using a microplate reader (SpectraMax iD3, Molecular Devices).

DNA oligo conjugates functional assay

A solution of antibodies conjugated to 5’-DBCO-N-polyA₁₅GGG-3’ DNA oligo (αHER2-P3-P4-polyA₁₅CCC, 1 eq., 0.17 mM, 27 μL in PBS) was mixed with the complementary 5’-TAMRA-CCCpolyT₁₅-3’ (2.1 eq., 0.37 mM, 13 μL in water). The mixture was incubated at 34.4 °C for 30 min in a thermal cycler (T100 Thermal Cycler, Bio-Rad) to allow oligo hybridization. The product was purified by sequential washing using an Amicon Ultra 30 kDa MWCO spin column until no TAMRA absorbance was detected in the flowthrough (3 times). αHER2-P3 with no P4-polyA₁₅CCC was used as a negative control. The absorbance spectra of the resulting conjugate were measured using a NanoDrop spectrophotometer. DNA oligos were synthesized by Integrated DNA Technologies (IDT).

MMAE conjugates a functional assay

Activity of the αHER2-MMAE conjugates was assessed by MTS cell viability assay. SKBR-3 cells were plated at a density of 5000 cells per well, A549 cells were plated at a density of 2500 cells per well, SKOV-3 cells were plated at a density of 2000 cells per well, and the 4T1 cells were plated at a density of 1000 cells per well the day before treatment. The next day, cell media was removed, and the cells were treated with varying concentrations of mAb (αHER2-P3), ADC (αHER2-MMAE), or MMAE diluted in the cell media. The MMAE was serially diluted starting from 2.5 μM in media (100 μL per well), and the ADC concentration was selected to match the MMAE dose (assuming two MMAE molecules per antibody). mAb concentration matched ADC concentration. After 72 h cell media was removed, and the solution of MTS was added. The signal development (30–60 min) was monitored with a microplate reader (SpectraMax iD3, Molecular Devices) as absorbance at 490 nm. Media only and untreated cells were used as controls.

Conjugates plasma stability

AFCs were prepared using conventional hinge cysteine conjugation or coiled-coil–based strategy for subsequent stability studies. Hinge cysteine conjugation was performed via TCEP reduction (2.75 eq.). The reduction was carried out in 1 M sodium borate buffer containing 25 mM NaCl and 1 mM EDTA for 90 min at 37 °C. Without purification, the antibodies were cooled to 0 °C and reacted with 5 equivalents of maleimide - BP Fluor647 overnight at 4 °C covered with foil wrap. The resulting conjugates were purified using a centrifugal spin filter with a 10 kDa MWCO. Then, the conjugates were reoxidized using 2 eq. of DTNB for 20 min and subsequently re-purified using a 50 kDa MWCO spin filter to obtain the final product. Coiled-coil AFCs were prepared as described previously using short coiled-coil peptide pairs P3-P4N₃ or long coiled-coil peptides P3_5mer-P4_5merN₃ or cysteine-modified P3_5mer-cys-P4_5mer-cys-N₃. The AFCs (5 µg) were dissolved in 0.5 mL of human plasma and incubated at 37 °C under constant shaking conditions. Aliquots (10 µL each) were collected at predetermined time points, flash-frozen in liquid nitrogen, and stored at −80 °C for downstream analysis. The samples were analyzed using an indirect ELISA assay using 2 μg/mL of recombinant human ErbB2/Her2 protein (Cat. #10126-ER, R&D Systems Inc.) as the antigen and anti-AF647 antibodies secondary antibodies (1:10,000, mouse anti-Alexa Fluor647 Antibody Immunology Consultant Laboratory, Cat. #M647-65A-400; HRP labeled with Abcam – HRP Conjugation Kit - Lightning-Link® Cat. #ab102890). Flash-frozen AFCs collected prior to plasma incubation (day 0) were used as an intact AFC standard.

Animal studies

All animal experiments described in this study were performed in accordance with protocol (PROTO201900105) that was reviewed and approved by the Institutional Animal Care and Use Committee of Boston University. All mice were maintained with a 12-h light/dark cycle at 20–26 °C and 30–70% humidity in facilities accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care.

Pharmacokinetic studies and serum preparation

Non-tumor-bearing Balb/c mice (6–7 weeks old) were treated with 10 mg/kg of native mAb (αHER2) or unloaded ADC (αHER3-P3_5mer-cys) or MMAE conjugated ADC (αHER3- P3_5mer-cys -P4_5mer-cys-MMAE) intravenously (100 µL, PBS solution). Blood was collected post-treatment via tail vein bleeding at 30 min, 1 day, and 3 days, and via cheek bleeding at the 7-day time point. Serum was isolated by centrifuging blood in a benchtop centrifuge at 1000 × g for 10 min at room temperature. Serum was then transferred to a new Eppendorf tube and stored at −20 °C for downstream analysis. The data was analyzed using an indirect ELISA assay detecting the total amount of human IgG and a standard curve.

Efficacy study

Female NSG mice (6–8  weeks old) were obtained from the Jackson Laboratory (#005557). The orthotopic xenograft model of SKOV-3 was established by injecting SKOV-3 cells (2 × 10⁶ cells in 0.2 ml PBS) intraperitoneally. Mice were administered a single dose of 100 µl of PBS or trastuzumab or ADCs at a dose of 10 mg/kg or 20 mg/kg in 100 µl PBS solution by i.v. injection in the tail vein. Animals were monitored regularly for changes in body weight and body condition. The study was terminated at week 7 post-treatment when animals began to exhibit severe hunched posture and signs of gait rigidity. Tumor tissues were collected, and their mass and weight were recorded ex vivo. The formula, volume = 1/2 (length × width²), was used to calculate the tumor volumes. Disitamab vedotin (cat.# HY-P9985), disitamab (cat.# HY-P99854), trastuzumab (cat.# HY-P9907), and trastuzumab vedotin (cat.# HY-164992) were purchased from MedChemExpress.

Biodistribution

SKOV-3 tumor-bearing BALB/c nude mice were administered a single intravenous dose of 10 mg/kg of the AFC (100 µL, PBS solution) on day 60 following tumor cell inoculation. Animals were sacrificed at predetermined time points—1 h, 1 day, 3 days, 5 days, and 7 days post-injection (n = 3 per time point). Major organs were harvested, and fluorescence signals were quantified ex vivo using IVIS Spectrum (PerkinElmer) to assess the biodistribution of the AFC. An untreated animal was used as a control. Total flux analysis was performed with Living Image® 4.8.2.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.