De novo design of a two-step approach targeting Claudin-6 for enhanced drug delivery to solid tumors

Abstract

Background

Although antibody-conjugated drugs have achieved success in clinical practice for cancer treatment, challenges remain in developing a highly efficient drug delivery system with specific accumulation in tumors and reduction in side effects. With improved pharmacokinetics, strong covalent bonding and quick binding reactions, a pre-targeting approach via molecular pairs represents an attractive platform for two-step delivery system construction.

Methods

Bioinformatics and immunohistochemistry assays were performed to assess Claudin-6 (CLDN6) as a highly specific tumor target in solid tumors. A phage-displayed library was used to screen and optimize anti-CLDN6 designed ankyrin repeat proteins (DARPins), which were incorporated into a two-step delivery system based on SpyTag/SpyCatcher. Fluorescent staining, flow cytometry and near-infrared imaging were performed to assess the tumor-targeting ability and biodistribution of this delivery system. The cytotoxic drug, Monomethyl auristatin E (MMAE), was conjugated with the delivery system to evaluate its anti-tumor efficacy and safety profile.

Results

Anti-CLDN6 DARPins exhibited specific binding to CLDN6⁺ cancer cells with high affinity instead of negative cells in vitro, ex vivo and in vivo. The DARPins-based two-step delivery system improved background clearance with a high signal-to-noise ratio, enhancing the specific accumulation of payloads in tumors. The cytotoxic drug delivered via the two-step system appeared superior to the one-step approach in IC50, biodistribution, and tumor growth inhibition.

Conclusions

Our study presented the de novo design of a two-step drug delivery system targeting Claudin-6 with enhanced anti-tumor efficacy and improved biosafety. These findings highlighted the potential of this approach to enhance the efficacy of tumor-targeting therapies and reduce adverse effects, paving the way for more effective cancer treatments.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07316-2.

Background

Antibody-conjugated drugs (ADCs) or drug conjugation with tumor-targeting domains have been widely developed and clinically applied in patients with solid tumors in recent years [1, 2]. However, off-target toxicity in normal tissues and insufficient accumulation in tumors can still be observed in clinical practice [3–5]. To improve treatment efficacy with reduced side effects, several key components should be considered, such as tumor target selection, targeting domain identification and delivery system construction [6–8]. An ideal tumor target should be expressed on the cell membrane limited to cancer cells, but absent in normal cells. The targeting domain is supposed to be developed against a specific extracellular epitope with high affinity. Furthermore, the biodistribution of cytotoxic drugs also relies on drug delivery systems, which strongly correlate with drug accumulation in normal organs, resulting in potential off-target effects [9].

Traditional ADCs are usually designed with full-length antibodies to provide cysteines for drug conjugation. However, the massive molecular weight and conformational structure of IgG-format ADCs might limit their application in solid tumors because of poor penetration into the tumor mass. Most of ADCs accumulate at tumor sites via specific tumor targeting, as expected. Moreover, they can also be enriched in the liver and spleen metabolically, thus inducing potential off-target release of cytotoxic drugs during circulation with a prolonged half-life of IgG-format ADCs [10]. To overcome the above challenges, novel targeting domains (such as DARPins, affibodies, AI-designed scaffolds, etc.) are characterized by a specific high affinity for tumor targets and enhanced extravasation with penetration into tumors, which could be engineered for drug conjugation [11–13].

Compared with single-step delivery of traditional ADCs, a pre-targeting approach with two-step delivery has been developed to increase tumor-targeting efficiency. According to previous research [14], the concept of pre-targeting derived from in vivo imaging has been validated as a delivery system with a high signal-to-noise ratio (SNR) and quick clearance of background signals. Usually, a two-step delivery system consists of two components: ① a targeting agent without toxic compounds; and ② an effector agent conjugated with cytotoxic drugs. Since the first agent targeting tumors is not toxic, its circulation would not cause side effects in off-target tissues, facilitating background clearance. The second agent, as an effector, is equipped with a quick-binding domain (high K_on value), which can be ligated to the first agent via molecular pairs (such as Barnase-Barstar or Biotin-Streptavidin). Moreover, the second agent carrying drugs has favorable clearance kinetics [15]. Taking advantage of the improved biodistribution of drugs, reduced toxicity to normal tissues and enhanced penetration into tumors, a two-step delivery system for drug conjugation could effectively kill tumor cells and decrease side effects on normal cells [16].

In this study, we chose Claudin-6, a tetraspanin membrane protein involved in tight junction formation, as a tumor target because of its unique expression patterns. It is widely expressed during the embryonic period but is absent in normal adult tissues, whereas it is upregulated in malignant tissues [17, 18]. To target Claudin-6 expressed on the tumor cell membrane, we screened a phage-displayed library and identified anti-CLDN6 DARPins (C6DP) sequences with high affinity for the Claudin-6 extracellular epitope. As alternatives to antibodies, DARPins derived from natural ankyrin repeat proteins are characterized by small size, low immunogenicity and high affinity [19, 20]. To construct a two-step delivery system, we used SpyTag and SpyCatcher derived from Streptococcus pyogenes with modified kinetics (K_on = 6.2 × 10⁵ M^{− 1}s^{− 1}) in the SpyTag/SpyCatcher003 version as molecular pairs for ligation through isopeptide bonds [21]. For the targeting agent in the first step (DARPins-fused SpyTag, DST), we modified DARPins with albumin-binding domains (ABDs) [22] and SpyTag. For the effector agent in the second step (SpyCatcher with Cysteines, SCC), we fused SpyCatcher with a linker-K5 self-assembling peptide (space linker: APIAQKDELE; K5 peptide: KLVFFAE) and extended sequences (4 × CGG) to provide free cysteines for drug conjugation. The K5 peptide derived from β-amyloid (Aβ) peptides can self-assemble to enhance the endocytosis of drugs in vivo [23]. Both in vitro and in vivo anti-tumor potency assays have shown promising results for enhanced drug delivery to solid tumors, supporting further clinical translation.

Methods

Reagents, cell lines and mice

The biotinylated Claudin-6 antigen peptide and plasmids were synthesized by GenScript (Nanjing, China). The µMACS Streptavidin Kit was purchased from Miltenyi Biotec (Cologne, Germany). MMAE and maleimidocaproyl-valine-citrulline-PABC-MMAE (mc-vc-PABC-MMAE) were purchased from MedChemExpress (New Jersey, USA). Cyanine5 maleimide (Cy5-Mal) and Tris (2-carboxyethyl) phosphine hydrochloride (TCEP) were purchased from Psaitong Biotechnology (Beijing, China).

The human hepatocellular carcinoma cell line HepG-2 (ATCC, Manassas, VA) was cultured in DMEM (Thermo Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA) at 37 °C in an atmosphere containing 5% CO₂. The human gastric cancer cell line (AGS, NUGC-4), human breast cancer cell line MDA-MB-231, human normal colon cell line NCM460, human normal gastric cell line GES-1, human bronchial epithelial cell line HBE and human normal hepatic cell line L02 (ATCC, Manassas, VA) were cultured in RPMI 1640 medium (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum at 37 °C in an atmosphere containing 5% CO₂.

In this study, ethics approval statements for animal work were provided by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Drum Tower Hospital. The procedures for the animal experiments were carried out in accordance with the Guide for Care and Use of Laboratory Animals, 8th Edition (2011). All animal experiments were carried out according to the IACUC guidelines, and all studies followed the protocols approved by the IACUC at Nanjing Drum Tower Hospital. Male and female BALB/c nude mice (4–5 weeks old, half male and half female, 16–18 g) were purchased from Cavens Laboratory Animal Technology Co., Ltd. (Changzhou, China). The mice had free access to sterilized water and food and were maintained under specific pathogen-free conditions with a controlled temperature (~ 25 °C), humidity (50–70%) and circadian rhythm (12-h light/dark cycle). All essential procedures were performed to minimize discomfort and avoid wasting animals. NUGC-4 and AGS animal models were generated via the subcutaneous injection of 8 × 10⁶ cells into the left abdomen of athymic nude mice. The length and width of the tumors were measured every other day, and the tumor volume was calculated via the following formula: volume = length × width × width × 1/2. The mice were subjected to biodistribution or anti-tumor response assessment after the tumor volume reached 150–200 mm³.

Bioinformatics analysis

RNA-Seq data acquisition, analysis and visualization were performed in R software (version 4.5.0, R Core Team, Vienna, Austria, https://www.r-project.org/) [24]. The pan-cancer dataset TCGA TARGET GTEx (PANCAN, N = 19131, G = 60499) was acquired from UCSC (https://xenabrowser.net/) for further analysis [25]. The expression levels of CLDN6 in cancer and paracancerous tissues were extracted from the dataset, and analyzed in 25 different cancer types, including Head and Neck squamous cell carcinoma (HNSC), Ovarian serous cystadenocarcinoma (OV), Sarcoma (SARC), Stomach adenocarcinoma (STAD), Stomach and Esophageal carcinoma (STES), Testicular Germ Cell Tumors (TGCT), Skin Cutaneous Melanoma- Metastatic (SKCM-M), Thymoma (THYM), Cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), and Adrenocortical carcinoma (ACC).

For analysis of immune cell infiltration in various cancers, CLDN6 expression data were mapped onto GeneSymbol and analyzed via TIMER solution (version 2.0) in the R package (Immuno-Oncology Biological Research, IOBR, version 2.0) [26]. This strategy was used to explore the correlation between CLDN6 expression and tumor-infiltrating immune cell subtypes (B cells, CD4 + T cells, CD8 + T cells, neutrophils, macrophages and dendritic cells).

For analysis of stem-like properties, the RNA-based and DNA methylation-based stemness scores derived from the stemness group were evaluated in various cancers [27]. The correlation between the stemness score and expression level of CLDN6 was analyzed in 37 types of cancer. The clinical data and CLDN6 expression data were acquired from TCGA and GEO database. Survival analysis was performed to generate Kaplan-Meier survival curve [28].

For immunogenicity assessment, protein sequences were submitted to the Immune Epitope Database (IEDB, National Institute of Allergy and Infectious Diseases, https://www.iedb.org/) [29]. T-cell epitope prediction was performed on the TepiTool (IEDB Analysis Resource) [30]. B-cell rpitope prediction was performed on the Bepipred (version 3.0, IEDB Analysis Resource) [31].

Isolation of Anti-CLDN6 DARPins

The M13 phage-displayed DARPins library was constructed by Abiocenter (Jiangsu, China) and biopanned with a µMACS Streptavidin Kit following the manufacturer’s protocol (Miltenyi Biotec, Cologne, Germany). Briefly, in the preparation steps, µMACS streptavidin microbeads were mixed with 4 × 10¹¹ pfu phages and collected into a new tube after running through the µColumn to remove non-specific phages. For biopanning, the biotinylated Claudin-6 antigen peptide (IRDFYNPLVAEAQKREL) was incubated with the collected phages (~ 10¹¹ pfu) at room temperature (RT) for 1 h with gentle rotation. Next, the µMACS Streptavidin Microbeads were added to the tube and incubated at RT for 1 h. Magnetic separation was performed with a new µColumn in the µMACS Separator, and the Microbeads were washed with TBS-T (0.5% Tween-20) 5 times to remove unbound phages. To elute specific bound phages, 0.2 M glycine-HCl (pH 2.2) containing 1 mg/mL BSA was added to the µColumn and neutralized with 1 M Tris–HCl (pH 9.1). The eluted phages were amplified for the next round of biopanning. The stringency of biopanning was increased by adjusting the number of washing steps and the concentration of Tween-20 in TBST. During each round, the phages added to the microbeads and eluted from the µColumn were titered on the culture plates. The enrichment rate was calculated by the output/input ratio of phages recovered after each round of biopanning. For DNA Sanger sequencing, random phage clones were selected from the eluted phages to identify anti-CLDN6 DARPins candidates on the basis of enriched DARPins sequences.

Production of DARPins proteins

All the proteins were expressed in E. coli and purified as described previously. In brief, DNA fragments of proteins modified with a N-terminal 6 × His tag were synthesized and cloned into the pET-30a (+) vector. E. coli strain BL21 (DE3) cells (New England Biolabs, Ipswich, MA, USA) were transformed with the respective pET-30a (+) expression plasmid and grown in LB medium until an optical density of 0.6 ~ 0.8 was reached. Protein production was induced by the addition of 0.5 mM Isopropyl β-D-thiogalactoside (IPTG) for 16 h at 15 °C with 220 rpm. After induction, the cell pellet was harvested via centrifugation, resuspended in lysis buffer (50 mM Tris, 150 mM NaCl, 5% glycerol, pH 8.0) supplemented with protease inhibitors (Vazyme, Nanjing, China), and lysed via sonification. After centrifugation, the supernatant was collected for purification via an AKTA HisTrap HP column and an AKTA pure protein purification system (Cytiva, Danaher, Uppsala, Sweden) according to the manufacturer’s instructions. After dialysis against PBS buffer, the purified proteins were evaluated via a BCA assay (Vazyme, Nanjing, China), SDS‒PAGE and SEC‒HPLC (Shimadzu Corporation, Kyoto, Japan). The secondary structure of the protein was determined via Chirascan circular dichroism (CD) spectroscopy (Applied Photophysics, Surrey, UK).

For characterization of SpyTag/SpyCatcher pairs (SpyTag003: RGVPHIVMVDAYKRYK; SpyCatcher003: VTTLSGLSGEQGPSGDMTTEEDSATHIKFSKRDEDGRELAGATMELRDSSGKTISTWISDGHVKDFYLYPGKYTFVETAAPDGYEVATPIEFTVNEDGQVTVDGEATEGDAHT), DARPins-fused SpyTag (DST) and SpyCatcher with extra cysteine (SCC) were produced in E. coli as described above. To form a complex of Tag/Catcher, SCC and DST were incubated at various ratios (SCC: DST = 1:1, 1:2, 2:1 or 10:1) under different conditions (4 °C for 12 h, 37 °C for 4 h, or RT for 0 h). After incubation, the reaction was terminated with loading buffer. SCC, DST and the mixture were evaluated via SDS‒PAGE.

Preparation and characterization of DARPins-conjugated drugs

DST and SCC proteins were produced in E. coli and purified as described previously. The SCC proteins were designed with 4 extra cysteines at the C-terminus for drug conjugation (optimized DAR = 4). Conjugation of the proteins to maleimidocaproyl-valine-citrulline-PABC-MMAE (Vc-MMAE), was carried out via thiol-maleimide chemistry following the manufacturer’s protocol (MedChemExpress, New Jersey, USA). Briefly, reduction of the C-terminal cysteine introduced into the SCC protein and the cysteines forming disulfide bonds between SCC proteins were performed in 0.5 mM TCEP for 1.5 h at RT with gentle agitation. After reduction, the TCEP buffer was removed with a PD-10 desalting column (5000 MWCO, Cytiva, Danaher, Uppsala, Sweden). SCC proteins with free C-terminal cysteines were immediately mixed with Vc-MMAE at a ratio of 1:8 (SCC: Vc-MMAE) for 3 h at RT, after which the buffer was changed to PBS in a PD-10 desalting column (the final product was SCC-MMAE). After the reaction, all the samples were evaluated via SDS‒PAGE, and the particle size was detected via Dynamic Light Scattering (DLS) with a Zetasizer Nano ZS (Malvern Panalytical, Great Malvern, UK) at room temperature.

Additionally, SCC proteins were conjugated to Cy5-Mal (SCC-Cy5) via a thiol-maleimide chemistry reaction following the procedures described above. For the one-step group, DST-fused SCC proteins (DST-SCC) were conjugated to Vc-MMAE (DST-SCC-MMAE) or Cy5-Mal (DST-SCC-Cy5) in a reaction identical to that described above. The schemes were produced by IBS 2.0 [32].

Biophysical characterization of self- assembly

DST and SCC proteins were dissolved in PBS solution at a concentration of 20 or 40 µM. Then, the solution (Group 1: DST 20 µM; Group 2: SCC 20 µM; Group 3: DST 40 µM + SCC 40 µM) was transferred into 96-well black opaque plates (#265301, Thermo Scientific, Waltham, MA, USA). The 8-anilino-1-naphthalenesulfonic acid (ANS, Beyotime, Shanghai, China) was added into the well with a fixed concentration of 20 µM each well. Changes in the fluorescence intensity of each well (assembly reaction for 180 min) were measured by a Varioskan LUX multimode microplate reader (Thermo Fisher Scientific, Waltham, MA, USA) under the excitation wavelength of 375 nm and the emission wavelength of 480 nm.

The secondary structures of DST, SCC and DST + SCC were prepared in deionized water with a concentration of 20 µM. The assembly reaction time was set as 180 min and then measured with Chirascan CD spectroscopy (Applied Photophysics, Surrey, UK) at room temperature, measuring range 200 ~ 260 nm.

Immunohistochemical and immunofluorescence staining

All human tissue samples used in this study were obtained from the Pathology Department of Nanjing Drum Tower Hospital. This study was performed in accordance with the Declaration of Helsinki. All patients included consented to participate in the study and to use their tissue samples in research. Our study protocol was approved by the Ethics Committee of Nanjing Drum Tower Hospital. To detect the expression of Claudin-6 proteins in clinical samples, tumor tissues from gastric cancer (GC) and ovarian cancer (OV) patients, and normal tissues from adults were collected from Nanjing Drum Tower Hospital. Deparaffined formalin-fixed, paraffin-embedded (FFPE) tumor slices were generated for immunohistochemistry (IHC) and Hematoxylin and eosin (H&E) staining. To assess the expression of Claudin-6, all samples were immunohistochemically (IHC) stained with an anti-CLDN6 antibody (Cell Signaling Technology, Cambridge, MA, USA) and visualized with an ABC peroxidase standard staining kit (Thermo Scientific, Waltham, MA, USA). The expression of Claudin-6 was evaluated as different levels (0+, 1+, 2+, 3+) according to the staining area and intensity by pathologists. For the quantitative assessment of Claudin-6 staining, the H-score method was used, considering both the intensity of the staining (0-negative, 1-weak, 2-moderate, 3-strong) and the percentage of positive cells (0-100%). Final H-score was calculated using the following formula: H-score = (%stained cells at 0) × 0 + (% stained cells at 1+) × 1 + (% stained cells at 2+) × 2 + (%stained cells at 3+) × 3. The H-score value ranges from 0 to 300.

For immunofluorescence staining (IF) with anti-CLDN6 DARPins, cryosections were made from frozen tissues, including cancer tissues and normal tissues. The slices were blocked with 1% (w/v) BSA at RT for 1 h. After the washing steps, the slices were incubated with DARPins at 4°C overnight. To detect DARPins binding, DyLight 650-conjugated anti-6 × His tag antibody (Abcam, Cambridge, MA, USA) was added to the slices, which were subsequently incubated at 37°C for 2 h. After mounted with 4’,6-diamidino-2-phenylindole (DAPI, Beyotime, Shanghai, China), all the slices were observed with a Leica TCS SP8 confocal laser scanning microscope (Leica, Wetzlar, Germany). The images of each slice were analyzed via ImageJ software (National Institutes of Health, Maryland, USA).

For the Cy5 delivery assays, the cryosections were blocked and washed as described above. In the first step, the slices were incubated with DARPins-fused SpyTag (Ctrl DP-SpyTag or C6DP-SpyTag) at 4 °C overnight. After washing with PBS, SCC labeled with Cy5-Mal were added to the slices at the second step. To remove unbound Cy5, the slices were washed with PBS and then mounted with DAPI for imaging and analysis as described above.

Cell binding assays, blocking assays and endocytosis assays

For immunofluorescence staining in cell lines, NUGC-4 and AGS cells were seeded into a 12-well plate at a concentration of 1.5 × 10⁴ cells/well. After incubation overnight, the cells were fixed with 4% paraformaldehyde (PFA) for 30 min at RT. After being washed with PBS, the cells were blocked with 1% (w/v) BSA at 37 °C for 1 h. To detect the expression of the Claudin-6 protein, NUGC-4 and AGS cells were incubated with a rabbit anti-CLDN6 primary antibody (CST, Cambridge, MA, USA) overnight at 4°C. After washing with PBS, NUGC-4 and AGS cells were visualized with a FITC-conjugated goat anti-rabbit secondary antibody (Abcam, Cambridge, MA, USA).

For binding assays via DARPins, cells were incubated with anti-CLDN6 DARPins or control DARPins containing unrelated sequences (Ctrl DP) overnight at 4 °C after being fixed and blocked. Then, the cells were visualized with a DyLight 650-conjugated anti-6 × His tag antibody (Abcam, Cambridge, MA, USA).

For blocking assays, fixed AGS cells were blocked with 1% BSA containing an anti-CLDN6 mAb (CST, Cambridge, MA, USA) or with PBS for 1 h prior to incubation with anti-CLDN6 DARPins. Then, the cells were visualized as described above.

For the Cy5 delivery assays, the cells were blocked and washed as described above. In the first step, the cells were incubated with DST containing C6DP or Ctrl DP at 4 °C overnight. After washing with PBS, SCC proteins labeled with Cy5 were added to the cells at the second step for 3 h incubation.

For endocytosis assays, AGS cells were cultured with C6DP at 37 °C in an atmosphere containing 5% CO₂. At 0 h, 2 h, 4 h and 8 h, endocytosed DARPins were visualized with a DyLight 650-conjugated anti-6 × His tag antibody, and lysosomes were visualized with a rabbit anti-LAMP1 antibody and a FITC-conjugated goat anti-rabbit secondary antibody (Abcam, Cambridge, MA, USA).

After the nuclei were labeled with DAPI (Beyotime, Shanghai, China), all the cells were observed with a Leica TCS SP8 confocal laser scanning microscope (Leica, Wetzlar, Germany). The semiquantitative analysis of fluorescence presented as the integrated optical density was performed via ImageJ software (National Institutes of Health, Maryland, USA).

Flow cytometry and affinity measurement

For cell surface staining with a commercial mAb, NUGC-4 and AGS cells were incubated with a rabbit anti-CLDN6 mAb at 4 °C for 30 min and visualized with an Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (Abcam, Cambridge, MA, USA). For DARPins staining, NUGC-4 and AGS cells were incubated with anti-CLDN6 DARPins or control DARPins containing unrelated sequences at 4 °C for 30 min and then visualized with a DyLight 650-conjugated anti-6 × His tag antibody.

For affinity measurement, AGS cells were incubated with anti-CLDN6 DARPins at different concentrations (0 ~ 500 nM) at 4 °C for 30 min. After washing with PBS, the cells were incubated with DyLight 650-conjugated anti-6 × His tag antibody for cell surface detection. The affinity of anti-CLDN6 DARPins was calculated by the EC50 of the mean fluorescence intensity (MFI). The interaction of anti-CLDN6 DARPins with the Claudin-6 protein was predicted by AlphaFold (https://alphafoldserver.com/) and PDBePISA (https://www.ebi.ac.uk/pdbe/pisa/).

For detection of Cy5 delivery to the cell surface, all the cells, including both cancer cell lines and normal cell lines, were incubated with DST containing C6DP or Ctrl DP at 4 °C for 30 min at the first step. After washing with PBS, the cells were incubated with Cy5-labeled SCC proteins in the second step to evaluate the delivery system.

All the FACS data were collected with a Beckman CytoFlex flow cytometer (Beckman Coulter, Brea, CA, USA) and analyzed with FlowJo software (https://www.flowjo.com/).

Biodistribution of DARPins in vivo

To assess tumor-targeting delivery systems in vivo, BALB/c nude mice bearing NUGC-4 or AGS tumors were established as described previously. In the two-step group, DST proteins were intravenously injected and allowed to circulate for 3 h at the first step, and then, SCC proteins labeled with Cy5 were intravenously injected at the second step. In the one-step group, DST proteins were fused with SCC proteins in advance. After labeled with Cy5, DST-SCC-Cy5 was then intravenously injected into tumor-bearing mice. In the free Cy5 group, Cy5 dissolved in PBS was intravenously injected as described above.

After being injected with Cy5 or Cy5-labeled proteins, the tumor-bearing mice were anesthetized with isoflurane and imaged with an IVIS Lumina Series III (PerkinElmer, Waltham, MA, USA) at 0 h, 2 h, 4 h, 8 h, and 24 h. At 24 h, major organs and tumor tissues in AGS mice were collected and imaged. The signal-to-noise ratio of tumor to its surrounding tissues was calculated via the formula: tumor radiance / surrounding background radiance [33, 34].

Tumor sections from the two-step group were co-stained with anti-CD31 antibody (ab9498, Abcam) and anti-CLDN6 antibody (#18932, Cell Signaling Technology) as primary antibody. Then, the co-staining markers were visualized by a Multiplex Fluorescence Staining Kit (Beyotime, Shanghai, China) (CD31, green; Claudin-6, purple; nuclei, blue; SCC-Cy5, red). In step-by-step visualization, DST labeled with FITC (green) and SCC labeled with Cy5 (red) were administrated in the two-step regimen. Tumor sections were observed with a Leica TCS SP8 confocal laser scanning microscope (Leica, Wetzlar, Germany). The quantitative analysis of payloads fluorescence (Cy5) was performed via ImageJ software (National Institutes of Health, Maryland, USA). The perivascular region was the area within 50 μm of CD31-positive blood vessels, while the tumor core was the region ≥ 200 μm from the nearest blood vessel [35, 36].

In pharmacokinetics assays, the blood and tumor samples were collected at 0.012 h, 0.5 h, 1 h, 4 h, 12 h, 24 h, 48 h, 72 h after injection of Cy5 in AGS mice. Plasma and grinded tumor tissues were placed in a 96-well plated and imaged. All the images were captured with Living Image software (PerkinElmer, Waltham, MA, USA).

Cell viability assays

A Cell Counting Kit-8 (CCK8 kit, Vazyme, Nanjing, China) was used to assess 2D cell viability and cytotoxicity induced by DARPins-conjugated drugs. The cells were seeded in 96-well plates and incubated overnight at 37 °C in a 5% CO₂ humidified atmosphere. The next day, the growth medium was replaced with fresh medium containing proteins, MMAE or DARPins-conjugated drugs at different concentrations. For the two-step group (also referred to as DST + SCC-MMAE), the cells were incubated with DST proteins for 3 h at 37 °C prior to SCC-MMAE treatment. After 72 h of incubation, the medium was removed, and the CCK8 reagent was added to each well. After the reaction, the absorbance at 450 nm was measured with a Varioskan LUX multimode microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). The absorbance values were analyzed via the following formula: cell viability = (treatment group - blank)/(control group - blank) × 100%. The IC50 value of each drug was calculated via GraphPad Prism software (GraphPad, San Diego, CA, USA). Additionally, 3D cell models established in ultralow-attachment 96-well plates were utilized for the cell viability assay. Three-dimensional tumor spheroids were incubated with DARPins-conjugated drugs as described above, and the diameter of the spheroids was monitored under a Zeiss microscope (ZEISS, Baden-Württemberg, Germany).

Colony formation assay

Six hundred AGS cells were plated in six-well plates. After overnight incubation, the cells were treated with different concentrations of DST and SCC-MMAE. At the end of 10 days, colonies were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet (Beyotime, Shanghai, China), washed, dried, and imaged.

In vivo antitumor efficacy and biosafety

To assess the therapeutic effects of DARPins-conjugated MMAE, BALB/c nude mice bearing AGS tumors were established as described above and randomized into 5 groups with equal tumor sizes (n = 4). SCC-MMAE, DST, and DST-SCC-MMAE were prepared as described above. DST proteins were generated by fusing SpyTag and ABD to anti-CLDN6 DARPins.

In the one-step group, the mice were treated intravenously with DST-SCC-MMAE (3 mg/kg, thrice per week). In the two-step group, DST protein was administered first and SCC-MMAE was given 3 h later on each dosing day, where the SCC-MMAE was injected at an equal molar mass of 3 mg/kg DST-SCC-MMAE (MMAE-equivalents). The two-step treatment was administered thrice per week. In the DARPins group and MMAE group, the DST protein or free MMAE was intravenously injected respectively (thrice per week). They were injected at an equal molar mass of 3 mg/kg DST-SCC-MMAE on each dosing day. The body weights of the mice were recorded every three days. Tumor volumes were measured via a digital caliper every three days and calculated as described previously.

For biosafety assessment, the mice were sacrificed, and normal organs together with the tumors were harvested for H&E staining. H&E images of normal organs were acquired with a Zeiss microscope (ZEISS, Baden-Württemberg, Germany). For routine hematological and biochemical analysis, peripheral blood was collected and examined in a clinical laboratory.

To assess red blood cell (RBC) hemolysis, RBCs were isolated from fresh whole blood of healthy donors. After centrifugation, the RBCs were resuspended in PBS to obtain a 5% hematocrit. DST and SCC proteins were incubated with RBCs for 12 h at RT. After centrifugation, the RBCs were removed, and the absorbance of the supernatant at 540 nm was measured. The samples of RBCs lysed with RBC lysis buffer (Biosharp, Beijing, China) or PBS were used as controls. The quantity of released hemoglobin was presented in percent of the optical density of lysed samples with lysis buffer (equal to full lysis).

Statistical analysis

Statistical analysis was performed with GraphPad Prism software (GraphPad, San Diego, CA, USA). The results were presented as the means ± SD. The significant difference between the perivascular region fluorescence and tumor core fluorescence was calculated via a paired two-tailed t test. Other significant differences between the two groups were calculated via an unpaired two-tailed t test. Kaplan‒Meier survival curves were analyzed via the log-rank (Mantel‒Cox) test. All the experiments were repeated at least twice. Data were considered statistically significant if the p value was < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001), and ns indicated non-significance.

Results

Identification of CLDN6 as a potential therapeutic target

To evaluate CLDN6 as a therapeutic target, we analyzed CLDN6 expression patterns, stemness scores and immune cell infiltration. RNA-seq of the pan-cancer dataset revealed that CLDN6 expression in cancer tissues was significantly upregulated in 25 kinds of solid tumors compared with normal tissues. In the ovarian cancer, CLDN6 expression was increased by 15-fold compared with that in normal tissues (Fig. 1A). In addition, we summarized cancer stemness scores related to CLDN6 expression. In most cancer types, CLDN6 expression was positively correlated with stemness scores, indicating that CLDN6 likely enhanced the stem-like properties of cancer cells (Fig. 1B). Among various cancers, we observed that the ovarian cancer stemness scores had the strongest correlation with CLDN6 (Fig. 1B). We also performed immune cell infiltration analysis and observed that CLDN6 decreased B-cell and CD4⁺/CD8 + T-cell infiltration in OV, STES, STAD, SARC, SKCM-M, THYM, CESC, HNSC, ACC and TGCT (Fig. 1C). Furthermore, we plotted the infiltration profiles of B cells, CD4⁺ T cells, CD8⁺ T cells, neutrophil cells, macrophages and dendritic cells in these cancers. Most immune cells were negatively correlated with CLDN6 expression in these cancers (Fig. 1D-I). In OV, STAD and STES, we found that the immune cells involved in anti-tumor immunity (B cells, CD8⁺ T cells and dendritic cells) were strongly decreased due to CLDN6 expression (Fig. 1E, G, H).

Fig. 1.

Bioinformatics analysis of CLDN6 in pan-cancer. (A) RNA-seq data of CLDN6. (B) Correlation between CLDN6 expression and cancer stem-like properties. (C) Heatmap analysis of immune cells infiltration corelated with CLDN6. (D-I) Scatter plot of Immune cells infiltration correlated with CLDN6 in (D) HNSC, (E) OV, (F) SARC, (G) STAD, (H) STES, and (I) TGCT. Significant differences between the two groups were calculated via an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P <0.001, ****P < 0.0001 and ns indicated non-significance

Correlation of CLDN6 with clinical prognosis in GC and OV patients

To investigate CLDN6 expression in GC and OV patients, we performed multiple assays to confirm the results of the bioinformatics analysis for CLDN6. IHC assays with commercial anti-CLDN6 mAbs confirmed that the Claudin-6 protein was strongly localized on the cancer cell membrane (Fig. 2A, B) and was absent in normal adult organs, including the ovary and stomach (Fig. 2B). RNA-seq assays also confirmed a significant increase in the abundance of CLDN6 in GC and OV patients (Fig. 2D). To determine its correlation with clinical outcomes, we performed Kaplan‒Meier survival analysis. In the overall survival (OS), progression-free survival (PFS) and post-progression survival (PPS) analysis, higher expression of CLDN6 was associated with poorer prognosis in OV patients (Fig. 2E-H). Similarly, in the OS, first progression (FP), PPS, and recurrence-free survival (RFS) analysis of GC patients, higher expression of CLDN6 also suggested poorer prognosis (Fig. 2I-L).

Fig. 2.

CLDN6 expression in clinical samples and correlation with prognosis of ovarian and gastric cancer patients. (A) IHC staining of Claudin-6 in ovarian cancer and gastric cancer. (B) IHC staining of Claudin-6 in adult normal tissues. (C) H-Score of Claudin-6 staining. (D) Comparison of CLDN6 expressed in GC/OV cancers and relative normal tissues based on RNA-seq data. (E-H) Clinical prognosis of ovarian cancer patients: (E) OS, (F) PFS, (G) PPS, and (H) RFS. (I-L) Clinical prognosis of gastric cancer patients: (I) OS, (J) FP, (K) PPS, and (L) RFS. Significant differences between the two groups were calculated via an unpaired two-tailed t test. Kaplan-Meier survival curves were analyzed by the log-rank (Mantel-Cox) test. *P < 0.05, **P < 0.01

Characterization of anti-CLDN6 DARPins with high affinity

Considering that Claudin-6 was specifically expressed on the membrane of cancer cells and was correlated with tumor biology in multiple assays, we proposed that Claudin-6 was a potentially ideal target for tumor-targeting drug delivery. To target Claudin-6, DARPins with specific binding to Claudin-6 were screened from a phage-displayed library. After four rounds of biopanning against the biotinylated Claudin-6 antigen peptide, the enrichment rate peaked in the third round and decreased next, indicating a saturated state (Fig. S1A). After phage DNA sequencing and amino acids analysis, the top four clones were identified on the basis of triple-code theory (Fig. S1B). The top one candidate accounting for 32.4% of DARPins was selected for further investigation and was identified as C6DP (anti-CLDN6 DARPins).

After production and purification, we acquired high-quality purified C6DP proteins (purity > 95%), as confirmed by SDS‒PAGE and SEC‒HPLC (Fig. 3A, C). We confirmed the secondary structure of C6DP via CD spectroscopy (Fig. 3B). We also confirmed NUGC-4 cells as CLDN6⁻ and AGS cells as CLDN6⁺ via anti-CLDN6 mAbs (p = 0.0006. Fig S2A, B). Using these cell lines, we performed cell binding assays via DARPins. As shown in Fig. 3D, C6DP significantly bound to AGS cells instead of NUGC-4 cells (p = 0.0008), whereas Ctrl DP weakly bound to both (p > 0.05 to C6DP in NUGC-4. p = 0.0003 for C6DP in AGS. Figure 3G). To further assess C6DP binding specificity, we performed cell blocking assays in AGS cells in which mAbs were used as blockades (Fig. 3E). We observed a significant reduction in C6DP binding in the mAb-blocking group compared with the PBS-blocking group (p = 0.006) (Fig. 3H). To evaluate the potential of C6DP for drug delivery, we performed endocytosis assays with C6DP in AGS cells. After incubation at 37 °C, DARPins colocalized with LAMP1-positive lysosomes at 4 h and 8 h (Fig. 3F, I).

Fig. 3.

Identification and characterization of anti-CLDN6 DARPins. (A) SDS-PAGE of C6DP before and after purification. Red arrow indicated the purified proteins. (B) CD spectroscopy of C6DP. (C) SEC-HPLC of C6DP. (D) Representative confocal images of AGS and NUGC-4 cells stained with C6DP (red) in binding assays. Scale bar: 100 μm. (E) Blocking assays in AGS cells blocked by anti-CLDN6 mAb or PBS. The cells were stained with C6DP (red). Scale bar: 200 μm. Nuclei were stained with DAPI (blue). (F) Endocytosis assays of C6DP (red) in AGS cells. LAMP-1 (green) was used as a lysosome marker. Scale bar: 200 μm. (G-I) Fluorescence analysis of DARPins IF in (G)binding, (H) blocking, and (I) endocytosis assays. n = 3. (J) Flow cytometry assays of anti-CLDN6 mAb (Alexa Flour 488) and DARPins (Dylight 650) in AGS and NUGC-4 cells. (K) MFI analysis of AGS and NUGC04 cells stained with C6DP. n = 3. (L) Relative affinity measurement of C6DP based on MFI analysis in AGS cells. (M) Interaction of DARPins with Claudin-6 predicted by AlphaFold. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. **P < 0.01, ***P < 0.001

We also performed flow cytometry in cell lines via mAbs and DARPins. Claudin-6 expression on the cell surface was confirmed by anti-CLDN6 mAbs (Fig. 3J). C6DP, instead of Ctrl DP, significantly interacted with AGS cells without binding to NUGC-4 cells, which was in a DARPin sequence-dependent manner (Fig. 3J, K). Considering that Claudin-6 had a four-transmembrane domain, we measured the affinity between C6DP DARPins and Claudin-6 via flow cytometry in AGS cells. Based on the EC50 value from MFI, we determined the relative affinity to be 7.98 × 10^− 9 M (Fig. 3L). In the binding model predicted by AlphaFold (Fig. 3M), ankyrin repeat domains with designed variable regions in DARPins interacted with extracellular loops in the Claudin-6 protein. Among them, Asp65, Tyr66, Tyr67, Glu69, Asp131 and Lys135 of C6DP were predicted to form hydrogen bonds with Claudin-6. Besides, His73, Leu74, Thr77 and Asp78 of C6DP were predicted to be involved in salt bridge or covalent link with Claudin-6.

Detection of Claudin-6 with DARPins ex vivo

Since Claudin-6 was a potential target in GC and OV, we performed fluorescence staining ex vivo to evaluate DARPins in clinical applications. Firstly, we stained cancer tissues from OV and GC patients via IHC to confirm Claudin-6 expression (Fig. 4A). Then, we stained cancer tissues and adult normal tissues with DARPins. In CLDN6-positive cancer tissues, the fluorescence signal of C6DP was significantly stronger than that in CLDN6-negative cancer tissues (p = 0.0004 in OV. p = 0.0007 in GC. Figure 4A, B). Considering that claudin proteins were widely expressed in the gastrointestinal tract as tight junctions, we performed fluorescence staining via C6DP in these tissues from healthy donors. In the esophagus, stomach and intestinal tracts, we did not detect obvious signals of C6DP (Fig. 4C), indicating that the cross-reaction between C6DP and other claudin proteins was significantly weak (Fig. 4F). We also performed C6DP staining in other normal tissues, including those of the cardiovascular system, respiratory system, digestive system, urinary system, etc. (Fig. 4D, E). Without obvious staining in these normal tissues (Fig. 4G), C6DP could be considered as a promising agent for tumor-targeting drug delivery.

Fig. 4.

Detection of Claudin-6 in clinical samples with DARPins. (A) IF staining with C6DP (red) in ovarian cancer and gastric cancer tissues. Scale bar: 100 μm. (B) Fluorescence analysis of DARPins in (A). n = 4. (C-E) IF staining with C6DP (red) in adult (C) esophagus, stomach, jejunum and colon; (D) heart, lung, artery, bronchus, liver and spleen; (E) pancreas, thyroid, bladder, skeletal muscle and skin. Scale bar: 400 μm. Nuclei were stained with DAPI (blue). (F-G) Fluorescence analysis of DARPins in (F) digestive tract and (G) other adult normal tissues. n = 4. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001

DARPins-fused two-step delivery system targeting Claudin-6 in vitro

Using anti-CLDN6 DARPins, we constructed a two-step delivery system via SpyTag/SpyCatcher to achieve dual-functional proteins (DST and SCC), where we also fused SpyCatcher to a K5 self-assembling peptide (Fig. 5A). We produced and purified proteins with high purity (> 95%) via SDS‒PAGE (Fig. S3A). To confirm the ligation of the two agents, we performed SDS‒PAGE after incubation. The two-step delivery system could work at 4 °C and 37 °C with a ratio of 1:1 or 1:2 (SCC: DST) to form a stable complex (Fig. 5B). Besides, a classic hydrophobic environment-reactive fluorescence probe, ANS, was employed for K5 peptide self-assembly detection. As a result, a strong fluorescence signal was detected in DST + SCC group compared with DST or SCC alone (Fig. S5A), demonstrating the formation of 𝛽-sheet structure among the self-assembly with hydrophobic surfaces exposed. Furthermore, circular dichroism (CD) spectrum was utilized to detect the secondary structure of self-assembled DST + SCC. In Fig. S5B, DST or SCC alone remained a random coil conformation. Nevertheless, DST + SCC exhibited signals of 𝛽-sheet structure.

Fig. 5.

Construction of DST/SCC system and delivery of Cy5 to cells in vitro. (A) Scheme of DST and SCC as dual-functional proteins. The ligation induced by isopeptide bond. (B) Ligation of DST/SCC under 4 ℃ for 12 h, 37 ℃ for 4 h or RT for 0 h with different ratios. (C) IF staining with DST and SCC-Cy5 (red) in NUGC-4 and AGS cells. Unrelated DARPins were applied in DST as a control. (D) Fluorescence analysis of Cy5 in (C). (E) IF staining with C6DP-fused DST and SCC-Cy5 (red) in HepG-2 and MDA-MB-231 cells. (F) Fluorescence analysis of Cy5 in (E). (G) Flow cytometry assays with DST/SCC-Cy5 in cancer cell lines. (H) MFI analysis of (G). (I) Flow cytometry assays with DST/SCC-Cy5 in normal cell lines. (J) MFI analysis of (I). Scale Bar: 200 μm. Nuclei were stained with DAPI (blue). n = 4. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001

To evaluate this delivery system in vitro, we labeled SCC with Cy5-Mal and performed fluorescence staining in NUGC-4 (CLDN6⁻) and AGS (CLDN6⁺) cells (Fig. 5C). The C6DP-fused two-step system significantly enhanced Cy5 delivery to AGS cells instead of NUGC-4 cells (p = 0.0003), whereas the Ctrl DP-fused system failed to deliver Cy5 to any cells (p = 0.0004 to C6DP in AGS), indicating that the tumor-targeting delivery depended on C6DP (Fig. 5D). We observed similar results in CLDN6⁺ HepG-2 cells and CLDN6⁻ MDA-MB-231 cells (p = 0.0004. Figure 5E, F). Consistently, in flow cytometry assays, the C6DP-fused two-step delivery system successfully delivered Cy5 onto the surface of AGS and HepG-2 cells (Fig. 5G, H). To investigate Cy5 delivery to adult normal cell lines, we assessed delivery efficiency in the normal gastrointestinal tract cell lines (NCM460 and GES-1), the normal bronchial epithelial cell line HBE and the normal hepatic cell line L02 (Fig. 5I). Consistent with our findings with anti-CLDN6 DARPins, the C6DP-fused two-step system delivered few Cy5 to normal cells (Fig. 5J).

DST/SCC targeting Claudin-6 ex vivo and in vivo

To further investigate a two-step delivery system (DST/SCC) for clinical application, we chose Cy5 as a drug mimetic and a fluorescent reporter for assessment in GC tissues and adult normal tissues. After fluorescence staining, the C6DP-fused DST/SCC could specifically deliver Cy5 to the CLDN6⁺ GC tissues (p = 0.0004. Figure 6 A). On the contrary, Ctrl DP-fused DST/SCC (Ctrl DP-SpyTag + SCC-Cy5) failed to deliver Cy5 to the CLDN6⁺ GC tissues (Fig. S6A). Moreover, few Cy5 was delivered to either CLDN6⁻ GC tissues or adult normal tissues (Fig. 6B, C). In particular, in the gastrointestinal tract stained with a two-step system, the Cy5 signal was significantly weak in normal gastrointestinal tract tissues (p = 0.0006. Figure 6D).

Fig. 6.

Anti-CLDN6 DST/SCC delivered Cy5 to CLDN6⁺ cancer cells ex vivo and in vivo. (A) IF staining with DST and SCC-Cy5 (red) in gastric cancer tissues. Scale bar: 100 μm. (B-C) IF staining with DST and SCC-Cy5 (red) in adult digestive tract and other normal tissues. Scale bar: 400 μm. Nuclei were stained with DAPI (blue). (D) Fluorescence analysis of Cy5 in different tissues. n = 4. (E) NIR images of Cy5 delivered by different systems in AGS and NUGC-4 tumor models. (F) Biodistribution of Cy5 in heart, kidneys, liver, lungs, spleen and tumor of AGS mice at 24 h post injection. (G) Pharmacokinetics of Cy5 in plasma and tumor. (H) Signal-to-noise ratio (SNR) of Cy5 in AGS mice. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001

Considering that a pre-targeting approach could improve pharmacokinetics in vivo, we performed biodistribution assays in tumor-bearing mice. In the AGS and NUGC-4 tumor models, DST-SCC-Cy5 was administered to the one-step group. In the two-step group, we administered DST without fluorescence dye (after 3 h) followed by i.v. injection of SCC-Cy5. After injection with free or conjugated Cy5, the mice were subjected to live-animal imaging at different time points. We observed that, in the one-step group, DST-SCC-Cy5 was enriched in AGS tumors at 4 h (Fig. 6E), whereas in the two-step group, DST followed by SCC-Cy5 showed efficient accumulation in AGS tumors with a higher signal-to-noise ratio (Fig. 6H). The two-step system maintained high-SNR tumor-targeting delivery for up to 24 h in AGS tumor-bearing mice (Fig. 6F, H). Conversely, in NUGC-4 tumor models, Cy5 was administered i.v. without obvious accumulation in tumors, revealing that the DST/SCC delivery system specifically targeted Claudin-6 (Fig. 6E). Besides, we analyzed the payloads Cy5 distribution in tumor-microenvironment and found that more payloads accumulated in the tumor core instead of the perivascular region (Fig. S7A and C), which indicated improved tumor penetration. We also observed that the two-step delivery system extended the half-life time of cargo in CLDN6⁺ tumors than one-step system (Fig. 6G).

DST/SCC-conjugated MMAE efficiently killed CLDN6⁺ cancer cells in vitro

On the basis of the DST/SCC targeting Claudin-6, we constructed SCC-conjugated MMAE (SCC-MMAE) via thiol-maleimide chemistry between Vc-MMAE and extra 4 × cysteines at the C-terminus of SCC protein (Fig. 7A). To validate the construction of SCC-MMAE and the interaction between DST and SCC-MMAE, we performed SDS‒PAGE under reducing or nonreducing conditions (Fig. 7B). The molecular weight of SCC-MMAE (Lane 3, 7) was obviously increased after conjugation, compared with that of SCC proteins alone (Lane 2, 6). As expected, SCC-MMAE could still ligate with DST forming a complex with increasing molecular weight (Lane 5, 9), suggesting that MMAE conjugation did not disturb the ligation of molecular pairs. In addition, the particle size of SCC-MMAE was increased which was detected by dynamic light scattering (DLS) as an indicator of successful drug conjugation (Fig. 7C).

Fig. 7.

Anti-CLDN6 DST/SCC-MMAE inhibited CLDN6⁺ cancer cell growth in vitro. (A) Construction of SCC-MMAE. (B) SDS-PAGE test of drug conjugation. Lane① Marker; Lane②-⑤ under reducing condition: ②Free SCC; ③SCC-MMAE; ④Free DST; ⑤Complex of DST and SCC-MMAE; Lane⑥-⑨ under non-reducing condition: ⑥Free SCC; ⑦SCC-MMAE; ⑧Free DST; ⑨Complex of DST and SCC-MMAE. (C) Particle size of SCC and SCC-MMAE. (D) CCK-8 killing assays in AGS and NUGC-4 cells. (E) Cell viability assessment of different drugs in AGS cells after 72 h incubation. Two-Steps: DST followed by SCC-MMAE. One-Step: DST-SCC-MMAE. n = 3. (F) Colony formation assays in AGS cells. Unrelated DARPins were used in DST as a control. (G) Colony formation rate analysis. n = 3. (H) Tumor growth inhibition assays in 3D cell spheroids. (I) AGS tumor spheroids inhibition analysis. n = 3. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. *P < 0.05, **P < 0.01, ***P < 0.001

To evaluate the potency of DST/SCC-conjugated MMAE (C6DP-SpyTag + SCC-MMAE) in vitro, we performed cell killing assays with a CCK-8 kit. NUGC-4 and AGS cells were treated with DST alone or DST followed by SCC-MMAE (DST/SCC-MMAE) at the indicated concentrations. The therapeutic effect of the naked DST protein was minor in both cell lines (Fig. 7D). However, in contrast to NUGC-4 cells, AGS cells were specifically sensitive to DST/SCC-MMAE (Fig. 7D). To determine the IC50 values of various drug delivery systems, we assessed cell viability in AGS cells after treatment for 72 h. Compared with MMAE alone (IC50 = 8.2 nM), DST-SCC-MMAE (One-step IC50 = 12.6 nM) similarly reduced cell viability (Fig. 7E). Moreover, DST/SCC-MMAE (Two-step IC50 = 7.5 nM) killed AGS cells efficiently, whereas the killing potency of naked DARPins (DST) was minor and undetectable (Fig. 7E). Meanwhile, an additional control group (Ctrl DP + SCC-MMAE) showed limited anti-tumor activity compared with C6DP-SpyTag + SCC-MMAE treatment (Fig. S6C).

The DST/SCC-MMAE targeting Claudin-6 also demonstrated an inhibitory effect on the colony formation of AGS cells (Fig. 7F, G). To further investigate the killing potency in 3D cell models, we established AGS and NUGC-4 tumor spheroids, and incubated them with different drugs for 72 h (Fig. 7H). Consistent with the results of the CCK-8 assays, DST/SCC-MMAE significantly inhibited AGS tumor spheroids growth compared with DST-SCC-MMAE or MMAE (p = 0.04 to one-step. p = 0.03 to MMAE. Figure 8I), whereas both drugs showed limited potency in NUGC-4 tumor spheroids.

Fig. 8.

Tumor suppression efficacy of anti-CLDN6 DST/SCC-MMAE in AGS tumor-bearing mice. (A) Schematic illustration of anti-tumor treatment in vivo. One-Step: DST-SCC-MMAE, 3 mg/kg, thrice per week; Two-Step: DST protein was administered first and SCC-MMAE was given 3 h later on each dosing day. The two-step treatment was administered thrice per week. (B) Body weight profiles of mice with different treatment. (C) Images of tumors on Day 27 after sacrifice. Scale bar: 1 cm. (D) H&E staining of normal tissues from mice on Day 27 after sacrifice. Scale bar: 200 μm. (E) Tumor growth curves of mice with different treatment. (F) Hemotoxicity assessment of DST and SCC proteins in blood samples from healthy adult donors. (G) Hemolysis curves of DST and SCC compared to lysis buffer (represent 100% lysis). n = 4. All data represent the means ± standard error. Significant differences between the two groups were calculated via an unpaired two-tailed t test. ***P < 0.001 and ns indicated non-significance

Enhanced therapeutic potency of DST/SCC-conjugated MMAE in vivo

We assessed the therapeutic potential of DST/SCC-MMAE in AGS subcutaneous tumor models (Fig. 8A). The body weights of tumor-bearing mice in all the groups increased (Fig. 8B). Therapeutic treatment with DST-SCC-MMAE (one-step) or DST followed by SCC-MMAE (two-step, MMAE-equivalents) inhibited the growth of AGS xenograft tumors, whereas DST (DARPins) without MMAE conjugation had few therapeutic effects (Fig. 8C). The two-step treatment was much more effective at inhibiting AGS tumor growth than one-step and free MMAE (p = 0.0005 to one-step. p = 0.0006 to MMAE. Fig. 8E).

For biosafety assessment, at the end of the animal experiments, we collected blood samples for laboratory tests and harvested tissues from normal organs for pathological evaluation. H&E staining revealed that the important organs in all groups maintained normal histomorphological features (Fig. 8D). Given that K5 peptide in SCC proteins was derived from an amyloid-𝛽 core aggregation motif, we also evaluated potential aggregation-related or deposition toxicities in cerebrum via H&E staining. We didn’t observe any amyloid plaques in hippocampus and cortex (Fig. S8A), which were common deposition sites for amyloid plaques. In the laboratory test, the numbers of white blood cells, red blood cells and platelets were similar among all the groups without significant difference (Fig. S4A-C). For liver and kidney function tests, we examined AST, ALT, Crea and Urea in blood samples, and no significant differences were detected (Fig. S4D-G).

Considering that SpyTag/SpyCatcher was derived from Streptococcus pyogenes, we performed a hemotoxicity study to assess the possible risks (hemolysis and hemagglutination) associated with systemic repeated administration of the DST/SCC. We collected human blood samples from healthy donors and incubated them with DST or SCC proteins at various concentrations. After incubation, all proteins did not cause hemagglutination in a U-bottom 96-well plate without forming a film over the entire surface (Fig. 8F). Compared with lysis buffer, which induced complete lysis, DST or SCC did not cause hemolysis of erythrocytes (Fig. 8G). Besides, we performed immunogenicity assessment of DST and SCC proteins. They were predicted as low-immunogenicity via IEDB database, and only few of residues in DST or SCC were predicted as high immunogenicity scores (score > 0.6, Fig. S9).

Discussion

Our study has identified several key findings related to novel drug delivery systems. First, our data suggest that Claudin-6 is a tumor-specific antigen expressed on the cell membrane that appears suitable for ADC targeting. According to our bioinformatics and IHC analysis, the expression of CLDN6 in adult humans is strictly silenced in healthy tissues but robustly reactivated in various solid tumors. We also note that high expression levels of CLDN6 enhance the stem-like properties of cancer cells, reduce the infiltration of immune cells and lead to poor clinical outcomes in patients with GC and OV. Second, via phage display technology, we have identified a novel DARPin clone with specificity and high affinity (7.98 × 10^− 9 M) that targets Claudin-6, as confirmed by binding and blocking assays. Taking advantage of their small size, high affinity and enhanced tumor penetration, anti-CLDN6 DARPins are suitable for drug delivery as tumor-targeting domains.

Furthermore, we present a two-step delivery system, DST/SCC, via SpyTag/SpyCatcher003 ligation as a pre-targeting approach to enhance the tumor-specific accumulation and anti-tumor efficacy of cytotoxic MMAE conjugates. In addition, a self-assembling K5 peptide is incorporated into the SCC to stimulate endocytosis. In biodistribution assays and tumor killing assays, DST/SCC efficiently delivers conjugated Cy5 or MMAE to CLDN6⁺ tumors, which is characterized by powerful anti-tumor potency, a high SNR and quick clearance from metabolism. The IC50 of DST/SCC-MMAE (two-step) is 7.5 nM, which is superior to that of one-step delivery (12.6 nM) and free MMAE (8.2 nM). In addition, in tumor-bearing mice, the two-step delivery system triggers specific accumulation in AGS tumors before 4 h, sooner than one-step delivery does. And more enrichment of cargo in tumors is observed at 24 h. Therefore, MMAE conjugation delivered by DST/SCC efficiently kills AGS tumors without toxicity to normal tissues and erythrocytes. Our data indicate that DST/SCC-MMAE achieves superior anti-tumor efficacy with improved biodistribution and biosafety.

The basic principle in cancer biomarker selection and targeting domain identification is to distinguish tumor cells from complicated tissue components, including normal cells. As an oncofetal protein, Claudin-6 is highly expressed during embryonic development but is gradually silenced in terminally differentiated cells until it is completely absent in most adult normal cells [17]. This unique expression pattern provides a potential opportunity for tumor-targeting therapy. Our data has confirmed significantly greater expression of CLDN6 in various solid tumors than in normal tissues, which resulted in enhanced cancer stem-like properties, insufficient immune cell infiltration and poor clinical prognosis. Consistently, recent basic research has revealed that CLDN6 interacted with LATS in gastric cancer cells, increasing the expression of YAP1, thus activating its downstream target genes to enhance epithelial mesenchymal transition progression [37]. Owing to its specific expression on the cancer cell membrane and oncogenic features, Claudin-6 is considered as a promising tumor target for anti-tumor therapy, and more than 20 novel drugs targeting Claudin-6 have entered clinical trials (NCT04503278, NCT06223256, NCT06690775), including CAR-T cells, BiTEs, and ADCs. These novel drugs under intensive preclinical or clinical assessment have achieved remarkable anti-tumor efficacy without serious side effects, indicating that Claudin-6 is a promising target in clinical practice [38]. For Claudin-6 epitope selection, we aligned Claudin-6 extracellular sequences with other homologous claudin proteins and selected the unique peptide (IRDFYNPLVAEAQKREL) in extracellular loop 2 with the lowest sequence identity and optimized water solubility. In our study, we have identified anti-CLDN6 DARPins with high affinity after multiple rounds of biopanning and successfully applied them in drug delivery systems, indicating their potential in tumor imaging probes, CAR-T cells and bispecific cell engagers for further clinical investigations.

In our design, the unique kinetics of molecular pairs in the two-step delivery system contributes to the improvement in drug biodistribution. With the development of protein technology, the SpyTag/SpyCatcher molecular pair has been updated to the 003 version and applied in our design. Their enhanced kinetics approach infinite affinity, inducing a quick binding reaction even at a low concentration (10 nM) [21]. In previous studies, they were successfully applied in nanodisc delivery [39], universal CAR-T construction [40] and tumor vaccine modification [41] due to their covalent bonding. Recently, different pre-targeting approaches have been explored for two-step delivery, such as click chemistry and Barnase-Barstar. They all exhibit high efficiency in drug delivery [42, 43]. However, click chemistry, which relies on chemical groups, is usually applied in peptide modification because of its chemical synthesis progress. However, it requires complicated chemical reactions for protein modification and purification (such as DARPins). Moreover, molecules applied in click chemistry are tiny compared with proteins, which could cause steric hindrance, interrupting the binding reaction. For Barnase-Barstar, this molecular pair derived from Bacillus amyloliquefaciens could induce cytosolic inclusions during production, and proteins fused with Barnase/Barstar might require additional steps for denaturation-refolding and purification [44]. Considering the convenience of DARPins production and purification in the E. coli system, either click chemistry or Barnase-Barstar is not suitable for DARPins modification. Instead, SpyTag/SpyCatcher proteins could be fused with DARPins and produced in E. coli with soluble expression in high yield. Above all, on the basis of the molecular pair, a novel targeting agent aimed at other tumor biomarkers could be quickly generated and purified in E. coli, indicating that the two-step delivery system is a universally applicable approach, as well as a matched effector agent conjugated with other toxic drugs.

To enhance the endocytosis of cytotoxic drugs, the K5 peptide applied in this study exhibits several distinctive features. This peptide is characterized by self-assembly forming β-sheets in nano-fibers, which could be triggered by unbalanced hydrophilic–hydrophobic interactions during proteins ligation [45]. As previously reported, nanostructures can induce cell membranes perturbation when they interact with cell membranes [46]; meanwhile, K5-induced nanofibers are capable of forming transient pores on cell membranes [47]. As a result, the influx of chemotherapy drug was increased in tumor cells by this drug delivery strategy. In this study, a key advantage is that the K5 peptide within SCC-MMAE initiates in situ self-assembly upon ligation between DST and SCC-MMAE, thereby forming nanofibers that promote endocytosis and the subsequent release of MMAE. This transformation occurs exclusively in situ, triggered by the covalent ligation of DST and SCC-MMAE, which alters the conformational structure, hydrophilicity, and hydrophobicity of the entire complex [48]. Preclinical studies have demonstrated that the K5 self-assembling peptide effectively enhances tumor-targeted delivery of imaging probes, chemotherapeutic agents, immunomodulators, and other cargoes, validating its suitability for improving drug delivery efficiency [33, 34, 45]. By leveraging both the enhanced internalization mediated by the K5 peptide and the multivalent loading of MMAE (with a drug-to-antibody ratio of 4), the two-step MMAE delivery strategy (IC₅₀ = 7.5 nM) exhibits stronger cytotoxicity compared to free MMAE (IC₅₀ = 8.2 nM). Although the K5 peptide is derived from an amyloid-𝛽 core aggregation motif, it only triggers self-assembly upon DST/SCC interaction in CLDN6⁺ tumor tissues. It won’t induce severe aggregation-related or deposition toxicities in normal tissues, as the K5 peptide within SCC cannot aggregate during in vivo circulation until SCC binding to DST in tumor sites. In future clinical translational research, we will focus on monitoring respiratory symptoms (e.g., pulmonary embolism) and neurologic symptoms (e.g., Alzheimer’s disease).

Our data establish the feasibility and biosafety of the DST/SCC delivery system, a novel approach in which Claudin-6 is targeted via DARPins with high affinity. The cytotoxic drug delivered via the two-step system appeared superior to the one-step approach in IC50, biodistribution, and tumor growth inhibition. The DST/SCC system improves drugs accumulation in tumors with higher SNRs, increasing their anti-tumor potency. Future studies should explore whether higher drug concentrations in plasma and tumors could be achieved by diverse intervals between the targeting step and the effector step administered in vivo. The cargo delivered by this novel system is not limited to only cytotoxic drugs. With the evolution of immunotherapy, engineered immune cells and cell engagers could also be split and delivered in two steps, which might enhance antitumor immunity. In recent studies, two-step delivery methods for switchable BsCAR-T cells (barstar-based) have been developed, in which a targeting agent with Barnase was administered first, followed by BsCAR-T cells. This approach increased anti-tumor immunity of CAR-T cells [44]. In fact, with the development of CAR-like immune cells without genetic engineering [49, 50], cell engagers without tumor-binding domains could be pre-incubated in vitro with normal immune cells and function as effector agents after the administration of targeting agents with tumor-binding domains, in which two agents could be ligated via fused molecular pairs, such as SpyTag/SpyCatcher. Compared with traditional CAR strategies, enhancing the potency of cellular immunotherapies without genetic engineering through this approach is facile and possible, which could be explored in further investigations.

As a proof-of-concept study, the findings of this study have several limitations. In this study, subcutaneous xenograft models were employed to preliminarily validate the in vivo biological effects of the CLDN6-targeted strategy, providing foundational data support for subsequent research. Nevertheless, subcutaneous tumors fail to recapitulate the organ-specific microenvironmental characteristics of the primary tumor site (e.g., stromal cell interactions, extracellular matrix composition), physiologically relevant perfusion states and tissue barrier functions that are critical to drug delivery in solid tumors. These factors not only affect the penetration efficiency and uniform distribution of drugs in tumor tissues but may also lead to discrepancies between the efficacy observed in subcutaneous models and the actual clinical responses of patients, thereby limiting the reliability of translating research conclusions to clinical practice. To systematically address this issue, the gastric orthotopic xenograft models based on AGS/NUGC-4 cells are necessary. Subsequent efficacy studies using this model will focus on comparing the tumor penetration of Claudin-6-targeted drugs and their inhibitory effects on tumor growth and metastasis under orthotopic conditions. Moreover, patient-derived xenograft (PDX) models derived from fresh post-operative gastric cancer specimens (including both CLDN6-positive and CLDN6-negative subtypes) are another way to address this limitation. These models will subsequently be used to assess the individualized efficacy of drugs (e.g., response differences among patients with varying Claudin-6 expression levels) and pharmacokinetic profiles, further solidifying the clinical application basis of the Claudin-6-targeted strategy. Besides, this study lacks a non-targeting two-step regimen for antitumor efficacy comparison. Although the Ctrl DP-SpyTag + SCC-MMAE has shown limited tumor inhibition in cell killing assays, additional control groups are still needed for in vivo assays, which demonstrates that antitumor activity requires DARPins specifically targeting Claudin-6, instead of the two-step schedule or non-specific EPR effects.

SpyTag/Catcher’s sequences were derived from the second immunoglobulin-like collagen adhesin domain (CnaB2) of Streptococcus pyogenes fibronectin-binding protein (FbaB). The bacterial origin and fusion proteins therapeutics may elicit anti-drug antibodies (ADA) upon repeat dosing, as bacterial proteins might contain T-cell and B-cell epitopes recognized by the adaptive immune system [51]. While in silico analyses of the SpyTag/Catcher system suggested low immunogenic potential via the Immune Epitope Database (IEDB), these predictions were limited by their reliance on sequence homology rather than functional immune assays. Repeat-dosing studies in immunocompetent models with ADA measurements and functional neutralization assessments are required in future investigation. Notably, SpyCatcher’s molecular size (~ 15 kD) and exposed reactive lysine residues correlates with immunogenicity in bacterial proteins [52, 53]. However, previous research assessed the SpyCatcher’s immunogenicity in C57BL/6 mice and found that SpyCatcher with truncated sequences, similar to SpyCatcher003 sequence used here, induced less serum antibody response [54]. Recently, the SpyTag/Catcher system was successfully applied in outer membrane vesicles modification for cancer vaccines and in chimeric antigen receptors modification for engineered immune cell therapies, where few of immunogenicity related to SpyTag/Catcher was observed [55, 56]. Should ADA responses be confirmed, we propose multiple mitigation strategies based on T-cell epitope engineering and site-specific PEGylation technologies. Using in silico tools (e.g., NetMHCpan, TCPro), we will identify and modify predicted MHC binding peptides in a truncated SpyTag/Catcher. Besides, conjugating polyethylene glycol (PEG) to surface-exposed residues in predicted immunodominant regions could mask epitopes while preserving SpyTag/Catcher ligation activity [57].

Accurate pharmacokinetic (PK) and intratumoral distribution data are vital for validating targeted drug delivery systems. While Cy5-labeled drugs provided initial spatial insights into our two-step delivery system, they have critical limitations: Cy5’s molecular size may alter the complex’s physicochemical properties (e.g., hydrodynamic radius, clearance); ester/amide-linked Cy5 can dissociate in vivo, causing “signal-payload decoupling”; and tissue autofluorescence/light scattering prevent absolute quantification. Radioisotope labeling (e.g., ⁶⁴Cu and ⁸⁹Zr) with DOTA/DFO stable chelators tracks the entire drug complex, which is suitable for in vivo imaging and biodistribution investigation. In further studies, we will conjugate the DOTA to the amine groups of DST or SCC, and then labeled proteins with ⁸⁹Zr. The biodistribution of protein drugs will be monitored at several time points through a micro-PET/CT. We will also collect plasma and tissue samples to measure radioactivity via a γ-counter and analyze pharmacokinetics quantitatively. Moreover, LC-MS/MS can directly measure the biologically active payload (MMAE), which is critical for analyzing PK and antitumor efficacy. This measurement also addresses the risk of complex dissociation (e.g., SCC-MMAE may release free MMAE in vivo) that radioisotope labeling cannot detect. In further investigation, we will collect plasma and tumor samples at each time point to separate and detect MMAE in a UHPLC system coupled to a triple quadrupole mass spectrometer. The accurate concentration of MMAE will be analyzed according to a MMAE standard curve, which can derive PK parameters for free MMAE (t½, Cmax, AUC) and calculate tumor/plasma concentration ratio (to assess tumor-specific enrichment of active payload).

High-affinity (10^{− 9}~10^{− 10} M, or even higher) targeting domains and enough targets expressed on the cell surface are crucial for tumor-targeting therapy. Cell membrane biomarkers without enough expression due to tumor heterogeneity requires a targeting domain with stronger affinity. In our design, the affinity of the targeting domain relies on monovalent DARPins with an affinity of 7.98 × 10^{− 9} M. One possible way to address this is constructing multivalent targeting domains, increasing the avidity of entire targeting agents, which is suitable for recombinant protein modification. For example, AMG509, a STEAP1-targeted T-cell engager (TCE) for prostate cancer, aims to target STEAP1 on the cell membrane, and its avidity is greatly improved by fusion with two STEAP1-binding Fab domains [58]. We will also explore the possibility of constructing bivalent or trivalent DARPins that target Claudin-6 via the DST/SCC delivery system in the future.

Taken together, our findings demonstrate that Claudin-6 is an ideal biomarker for tumor-targeting therapy, and a de novo design of a two-step delivery system has been constructed on the basis of molecular pairs and novel DARPins targeting Claudin-6. Anti-CLDN6 DARPins specifically bind to CLDN6⁺ cancer cells with high affinity. Through DARPins and SpyTag/SpyCatcher, DST/SCC has achieved improved biodistribution and enhanced drug delivery, resulting in impressive anti-tumor efficacy. Overall, the preclinical findings of this study provide new insights into drug delivery system design and have translational implications for the treatment of solid tumors.

Conclusions

The anti-CLDN6 DARPins and pre-targeting strategies provide an efficient tumor-targeting delivery system with rapid background clearance and enhanced accumulation in tumor tissues. By improving the specificity and efficacy of cytotoxic drugs, this approach could lead to more effective tumor-targeting treatments with reduced side effects.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CLDN6

Claudin-6

DARPins

Designed ankyrin repeat proteins

ADCs

Antibody-conjugated drugs

SNR

Signal-to-noise ratio

ABDs

Albumin-binding domains

DST

DARPins-fused SpyTag

SCC

SpyCatcher with Cysteines

Aβ

β-Amyloid

C6DP

Anti-CLDN6 DARPins

Ctrl DP

Control DARPins containing unrelated sequences

MMAE

Monomethyl auristatin E

Vc-MMAE

Maleimidocaproyl-valine-citrulline-PABC-MMAE

Cy5-Mal

Cyanine5 maleimide

TCEP

Tris(2-carboxyethyl)phosphine hydrochloride

HNSC

Head and Neck squamous cell carcinoma

OV

Ovarian serous cystadenocarcinoma

SARC

Sarcoma

STAD

Stomach adenocarcinoma

STES

Stomach and Esophageal carcinoma

TGCT

Testicular Germ Cell Tumors

SKCM-M

Skin Cutaneous Melanoma- Metastatic

THYM

Thymoma

CESC

Cervical squamous cell carcinoma and endocervical adenocarcinoma

ACC

Adrenocortical carcinoma

OS

Overall survival

PFS

Progression-free survival

PPS

Post-progression survival

RFS

Recurrence-free survival

FP

First progression

Author contributions

Conceptualization: BL, RL, QL. Methodology: JY, XY, XC, YZ. Investigation: JY, LZ. Visualization: JY, LZ, FL, XW, LL, JS, JG. Funding acquisition: BL. Project administration: JY, LZ, XC, YW, TC, JS. Supervision: BL, RL, QL. Writing – original draft: JY, BL. Writing – review & editing: JY, BL, RL, QL. All authors read and approved the final manuscript.

Funding

This work was funded by grants from the National Natural Science Foundation of China grant 82272811 (BL), Jiangsu Provincial Medical Key Discipline grant ZDXK202233 (BL), and Jiangsu Province Key Research and Development Program grant BE2023654 (BL). The funding sources had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.

Data availability

All the data are publicly available and can be requested from the authors.

Declarations

Ethics approval and consent to participate

All human tissue samples applied in this study were obtained from Pathology Department of Nanjing Drum Tower Hospital. This study was performed in accordance with the Declaration of Helsinki. All patients included were consented to participate in the study and to use their tissue samples in research. Our study protocol was approved by the Ethics Committee of Nanjing Drum Tower Hospital. In this study, the ethics approval statements for animal work were provided by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Drum Tower Hospital. The procedures for animal experiments were carried out in accordance with the Guide of Care and Use of Laboratory Animals, 8th Edition (2011). All animal experiments were carried out according to the IACUC guidelines, and all studies followed the approved protocols by the IACUC at Nanjing Drum Tower Hospital.

Consent for publication

All authors contributed significantly to the conception, design, execution, and interpretation of the research. They reviewed and approved the manuscript. All authors agreed to be listed as co-authors.

Competing interests

All authors declare that they have no competing interests.