EpCAM- and EGFR-Specific Antibody Drug Conjugates for Triple-Negative Breast Cancer Treatment

Triple-negative breast cancer (TNBC) is a group of heterogeneous and refractory breast cancers with the absence of estrogen receptor (ER), progesterone receptor (PgR) and epidermal growth factor receptor 2 (HER2). Over the past decade, antibody drug conjugates (ADCs) have ushered in a new era of targeting therapy. Since the epidermal growth factor receptor (EGFR) and epithelial cell adhesion molecule (EpCAM) are over expressed on triple-negative breast cancer, we developed novel ADCs by conjugating benzylguanine (BG)-modified monomethyl auristatin E (MMAE) to EpCAM- and EGFR-specific SNAP-tagged single chain antibody fragments (scFvs). Rapid and efficient conjugation was achieved by SNAP-tag technology. The binding and internalization properties of scFv-SNAP fusion proteins were confirmed by flow cytometry and fluorescence microscopy. The dose-dependent cytotoxicity was evaluated in cell lines expressing different levels of EGFR and EpCAM. Both ADCs showed specific cytotoxicity to EGFR or EpCAM positive cell lines via inducing apoptosis at a nanomolar concentration. Our study demonstrated that EGFR specific scFv-425-SNAP-BG-MMAE and EpCAM-specific scFv-EpCAM-SNAP-BG-MMAE could be promising ADCs for the treatment of TNBC.


Introduction
Female breast cancer is the leading cause of global cancer incidence worldwide in 2020 and ranks fifth in terms of cancer mortality [1]. Based on the expression of estrogen receptor (ER), progesterone receptor (PgR) and epidermal growth factor receptor 2 (HER2, also known as ERBB2), Perou and Sørlie's team first classified breast cancer into five molecular subtypes: luminal subtype A; luminal subtype B; basal-like subtype; HER2 positive subtype; and normal breast-like subtype [2,3]. Triple negative breast cancer (TNBC) is characterized by the clinical absence of ER, PgR and HER2 diagnosed mainly via immunohistochemistry and accounts for 10-20% of all breast cancer [4]. Even though defined by different methods, TNBC is often, but not always, equal to the basal-like subtype, as most of the basal-like subtypes are TNBC and about 80% of TNBC are also basal-like breast cancer [5].
Patients positive for ER and/or PgR benefitted from endocrine therapy such as tamoxifen, which is confirmed to reduce recurrence rates throughout the first 10 years and mortality throughout the first 15 years [6]. Meanwhile, the application of HER2 targeted monoclonal antibodies (mAbs) such as pertuzumab or trastuzumab demonstrated a prolonged median overall survival (OS) of 16 months in patients with HER2-overexpressing breast cancer [7]. TNBC showed more aggressive biology compared with other subtypes with increased likelihood of distant recurrence and death ratio [8]. Due to the absence of hormone receptors and distinctive targetable antigens, TNBC is still struggling with limited therapeutic options including surgery and systematic chemotherapy [9].
One hundred years ago, Paul Ehrlich proposed a "magic bullet" concept aiming to target and eliminate tumor cells while sparing normal cells. Since then, numerous efforts had been made to find a tumor specific agent. From nitrogen mustard, to small molecules targeting mutated proteins, to multi-targeted inhibitors, antibodies now exhibit excellent potency for tumor-selective targeting [10]. Whereas most tumor-selective antibodies have limited therapeutic activity, antibody-drug conjugates (ADCs), which are made up of monoclonal antibodies tethered to cytotoxic agents, provide the possibility to achieve "magic bullet" activity by integrating the specific-binding ability of antibodies and tumorkilling capacity of cytotoxic agents.
The first generated ADCs rely on conjugating the cytotoxic agents to the antibodies using non-specific conjugation methods, which take advantage of naturally occurring amino acids such as lysine and cysteine without modification of antibodies, but result in an undesirable heterogeneous mixture of ADCs with varying drug to antibody ratios (DARs) [11]. In 2008, Junutula and co-workers produced homogeneous ADCs by engineering reactive cysteine residues site-specifically without disruption of interchain disulfide bonds, followed by cytotoxic drug attachment to these unpaired cysteines [12]. Since then, more and more site-specific conjugation methods have been developed, among which enzyme-based conjugation methods have emerged as promising alternative site-specific conjugation methods. SNAP-tag is a modified version of human DNA repair enzyme O6alkylguanine-DNA-alkyltransferase (AGT), specifically reacting with O6-benzylguanine (BG) derivatives by irreversible transfer of an alkyl group to a cysteine residue and forms a stable thioether bond [13,14]. This site-specific conjugation strategy enabled the production of homogeneous ADCs with uniform pharmacokinetic properties.
Targeting therapy by means of ADCs emerged as a promising strategy to treat TNBC. Here, we developed a targeting method based on epidermal growth factor receptor (EGFR) and epithelial cell adhesion molecule (EpCAM), which have been reported to be overexpressed in TNBC patients [15][16][17]. In particular, EGFR associated with poor prognosis is more frequently overexpressed in TNBC than in other subtypes, ranging from 13% to 76% depending on population and methods of the evaluation and antibodies [15,18]. We genetically engineered fusion proteins consisting of SNAP-tag fused to single chain variable fragments (scFvs). The constructs are named scFv-425-SNAP (anti-EGFR) and scFv-EpCAM-SNAP (anti-EpCAM) respectively. Purified scFvs were then conjugated with BG-modified cytotoxic agent monomethyl auristatin E (MMAE) [19,20]. The specific cytotoxicity of both ADCs was confirmed in vitro. Compared to full-length antibodies, the small size (27 KDa) and short half-life (within hours) of scFvs improve the permeability of ADCs and reduce exposure time to normal cells which express low levels of targeting receptors. The scFvs-SNAP fusion proteins either labelled with fluorescence dyes or cytotoxic agents have already been evaluated in different tumors for imaging application or targeting therapy which exhibited impressive results, indicating the suitability of scFv-SNAP fusion proteins to be used as delivery vehicles [19,21,22].
Our study produced two MMAE-based ADCs with controlled stoichiometry targeting EGFR and EpCAM using SNAP-tag technology. The high-efficiency conjugation method, homogeneous products and selective cytotoxicity provide a promising novel targeting strategy for TNBC.

Production of scFv-SNAP Fusion Proteins
The scFv-SNAP fusion proteins (~52 kDa) were expressed by HEK293T cells, and purified by an Ni-NTA superflow cartridge using the C-terminal 6 × His-tag. All fractions during purification were collected. After incubation with SNAP-Surface ® Alexa Fluor ® 488, the presence of fusion protein and the activity of SNAP-tag in collected fractions were confirmed in SDS-PAGE. ScFv-EpCAM-SNAP is shown as an example in Figure S1. ScFv-EpCAM-SNAP was present in supernatant. Proteins were eluted with 40 mM and 250 mM imidazole ( Figure S1a). Considering that the amount of samples loaded in E2 were more than threefold that in E3, most proteins were eluted with 250 mM imidazole and proteins eluted with 40 mM imidazole were negligible. Coomassie brilliant blue stain indicated that only scFv-EpCAM-SNAP was eluted with 250 mM imidazole, and other proteins from the culture medium were removed stepwise ( Figure S1b). ScFv-425-SNAP was produced and confirmed in the same way. This expression system and purification method yielded proteins up to 10 mg per liter of culture supernatant. The conjugation of scFv-SNAP fusion proteins was further explored. ScFv-425-SNAP and scFv-EpCAM-SNAP were conjugated with BG-MMAE (used as therapeutic agents) or SNAP-Surface ® Alexa Fluor ® 647 (used as imaging agents) (Figure 1a). The site-specific conjugation was confirmed by post-incubation with SNAP-Surface ® Alexa Fluor ® 488. While unconjugated proteins retain the full activity to couple with SNAP-Surface ® Alexa Fluor ® 488, the specific coupling sites were blocked by SNAP-Surface ® Alexa Fluor ® 647, or BG-MMAE ( Figure 1b). SNAP-Surface ® Alexa Fluor ® 488 was observed under UV light using ChemiDoc XRS+ System, and the SNAP-Surface ® Alexa Fluor ® 647 signal was observed at 685 nm using an Odyssey DLx Imager (Figure 1c). The presence of all proteins is shown in SDS-PAGE stained with Coomassie brilliant blue ( Figure 1d). Our results demonstrated that BG-modified agents could be conjugated to scFv-SNAP fusion proteins site-specifically and sufficiently within 2 h at room temperature.

Binding and Internalization Properties of scFv-SNAP Fusion Proteins
Targeting properties of Alexa Fluor ® 647-labeled scFv-425-SNAP and scFv-EpCAM-SNAP were validated by flow cytometry and fluorescence microscopy.

Specific Cytotoxicity of MMAE Conjugated scFv-SNAP Fusion Proteins
Since the specific binding and internalization of scFv-SNAP fusion proteins have been confirmed, we expected that these properties were retained after conjugation with BG-MMAE and therefore a cytotoxicity assay was performed to determine the specific cytotoxicity of scFv-   The IC 50 values indicate the concentration that inhibit cell viability by 50% relative to untreated control cells. The data represented three independent experiments carried out in triplicate and were presented as mean ± standard error of the mean (SEM).

Discussion
TNBC is a heterogeneous, aggressive type of breast cancer associated with limited treatment options and comparably poor prognosis. Although immunotherapy in combination with chemotherapy showed promising activity in PD-L1-expressing TNBC, systematic chemotherapy still remains the standard for TNBC treatment [9]. The emergence and development of targeted therapy has dramatically changed the management of antitumor therapy. Monoclonal antibodies and small molecule drugs are the most common types of targeted therapy. Albeit a superior selective binding property, only a few mAbs exhibit moderate antitumor activity by themselves and are often used in combination with other anticancer drugs. Meanwhile, most small molecular weight cytotoxins, which are commonly used in chemotherapy, have potent cytotoxicity, but suffer from rapid plasma clearance and low specificity. ADCs assuredly widened the scope of mAb-based targeting therapy by endowing mAbs with cytotoxicity and retaining high specificity. To date, there are approximately one hundred mAb products and twelve ADCs have been approved by the FDA (U.S. Food and Drug Administration), among which three mAbs and three ADCs are designed to treat breast cancer [29,30]. Of note, only sacituzumab and govitecan are targeting the human trophoblast cell-surface antigen 2 (Trop-2) in metastatic TNBC, while all others are targeting Her2. Since sacituzumab govitecan was the first and sole ADC for TNBC, our study estimated the specific cytotoxicity of two other ADCs for TNBC.
An ideal ADC consists of three core components: a highly specific mAb targeting tumor-associated antigen expressed on the tumor surface with minimal expression on non-malignant cells, a stable and flexible linker that can survive during blood circulation and release cytotoxic agents effectively at target sites, and a potent cytotoxic payload. The majority of ADCs are built on full-length antibodies, especially IgG1, limiting tumor penetration due to their large molecular size of approximately 150 kDa [31,32]. The half-life for IgG1, IgG2 and IgG4 is around 21 days, and the mAbs were found to be retained in circulation instead of targeted tumor sites [32]. The long half-life and poor permeability of these drug conjugated mAbs increase the exposure risk of normal cells to toxic drugs and cause nonspecific cell death during circulation [33]. Compared to full-length antibodies, scFvs consist exclusively of variable fragments of heavy and light chain, resulting in a much smaller molecular weight (27 kDa). The smaller size extends the possibility of antibodies to penetrate tissues, while variable regions are still capable of binding to antigens. Nevertheless, given that antibodies less than 25 kDa can be fast filtered by glomerulus, scFv exhibit blood half-life for only 11 min and whole body half-life for 1.4 h [34,35]. Fusing of the scFv to SNAP-tag improves its half-life by increasing the size of the antibody to around 52 kDa, and ∼50 kDa has been proven to be the optimal size to obtain a maximum tumor-toplasma exposure ratio [36]. Furthermore, ADCs are supposed to be internalized rapidly to avoid off-target effects. The internalization time varies from antibody to antibody. Our flow cytometry and fluorescence microscopy have proven that both conjugated scFv-425-SNAP and scFv-EpCAM-SNAP could specifically bind to EGFR or EpCAM overexpressing cell lines, respectively, and be rapidly internalized within 3 h, indicating the usability of these scFvs as delivery vehicles.
The mechanisms of therapeutic mAbs action are mainly due to signal transduction changes, antibody-dependent cellular cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC) and antibody-dependent cellular phagocytosis (ADCP) [37][38][39]. The immune reactions triggered by therapeutic mAbs are highly associated with Fc domains. However, the immune responses might be superfluous for mAbs in ADCs and even lead to adverse effects in an FcR-dependent manner [40,41]. The efficacy of ADCs is mainly conducted by highly potent cytotoxic drugs such as auristatin, calicheamicin, maytansinoid, camptothecin and pyrrolobenzodiazepine dimer. Auristatins are commonly used payloads, accounting for five out of twelve of the current ADCs approved by the FDA and six out of fourteen approved worldwide [29]. Mechanistically, it destabilizes microtubules and blocks the microtubule assembly by binding tubulin, causing G2/M phase cell cycle arrest in the nanomolar concentration range [42]. Conjugating different payloads such as DNA cleavage agents, microtubule inhibitors and TOPO1 inhibitors might be a potential way to solve the drug resistance.
Linkers that connect the cytotoxin to the mAb maintain the stability of ADCs in systemic circulation and play a key role in pharmacokinetic and therapeutic properties [43,44]. Generally, linkers can be divided into cleavable and non-cleavable linkers. Cleavable linkers increase the possibility of a "bystander" effect, which means the released cytotoxins kill surrounding cells expressing low or no ADC-targeted antigens. Dipeptide-containing linker valine-citrulline (val-cit) used in our experiment is the most popular enzymatic cleavable linker used in current clinical research, and Brentuximab vedotin typically used for lymphoma treatment is a case in point [11,45].
In addition to three core components mentioned above, the conjugation method is also crucial for ADCs. The conventional drug conjugation methods are either using lysine sidechains or cysteine residues on the mAb backbone, yielding over 10 6 different ADC species with different DARs and pharmacological properties [43,46]. Site-specific conjugation methods enable the generation of homogeneous products with defined DAR and unified pharmacokinetics, as exemplified by non-natural amino acids incorporation, engineered cysteine conjugation, and enzymatic conjugation [47]. Nevertheless, some site-specific conjugation methods require complicated procedures such as redox reaction and metal-dependent catalytic reaction, or huge amounts of cytotoxic agents [48][49][50]. These may affect the activity of antibodies and are also cost-prohibitive. The SNAP-tag technology we used here is an enzymatic conjugation method. Apart from site-specific conjugation and product homogeneity, this self-labeling tag achieved rapid and efficient reaction under physiological conditions with a 1:1 stoichiometry [13]. Theoretically, ADCs with more drug molecules are supposed to achieve higher cytotoxicity. However, Hamblett et al. revealed that mAb with reduced drug loading could increase the therapeutic index [51]. The research group compared the antitumor activity of mAb with two, four or eight MMAE molecules (E2, E4, and E8, respectively) in vitro and in vivo. The drug potency was undoubtedly improved with increased drug loading when applied to cell lines. However, a xenograft model revealed unexpected results when E4 showed comparable effectiveness with E8 at same mAb doses, and E2 exhibited the best response after doubling the dose to reach an equal amount of MMAE with E4 and E8 [51]. In this study we conjugated scFv-SNAP fusion proteins with one MMAE molecule and confirmed that conjugation could be fulfilled with a two-fold molar excess of BG derivatives within 2 h at room temperature. The SNAPtag technology along with scalable recombinant proteins production could facilitate an economic production of ADCs. Overall, our research established two homogeneous ADCs based on SNAP-tag technology, targeting the overexpressed antigens EGFR and EpCAM in TNBC. Compared to current available ADCs, highly potent cytotoxic drugs armed to scFv are more likely to be delivered into deep tumor regions due to their smaller size, and their short half-life improves the safety profile in the meantime. Combined with SNAP-tag technology, both of the scFv-SNAP-MMAE exhibited high specificity, rapid internalization and specific cytotoxicity in TNBC cell lines, indicating the promising application of scFv and SNAP-tag in the ADC field. Despite the impressive results, our study still needs to be further confirmed in vivo, and pharmacokinetic properties should be determined. Additionally, given the drug susceptibility as we mentioned above, SNAP-tag could also facilitate the coupling of different or more toxic agents to overcome drug resistance or improve cytotoxicity.

SNAP-Fusion Protein Expression and Purification
The scFv-SNAP fusion proteins were expressed as previously described [13]. Briefly, the SNAP-tag and scFv fragment were inserted into the mammalian expression vector pMS. The plasmids were transfected into HEK293T cells using Roti ® Fect (Carl Roth, Karlsruhe, Germany). Transfected cells were selected by zeocin (InvivoGen, Toulouse, France) (0.1 mg/mL) supplemented in medium. ScFv-425-SNAP and scFv-EpCAM-SNAP were secreted into culture medium, and then isolated from the supernatant with different imidazole concentrations (10, 40 and 250 mM, respectively) using an Ni-NTA Superflow cartridge (Qiagen, Hilden, Germany) on an ÄKTA start system (GE Healthcare Bio-Sciences AB, Uppsala, Sweden,). All eluted fractions were collected during purification. After incubation with SNAP-Surface ® Alexa Fluor ® 488 (New Englands Biolabs, Ipswich, MA, USA) at room temperature for 20 min in the dark, fractions were run in 10% SDS-PAGE to confirm the activity of SNAP-tag and the presence of proteins followed by Coomassie brilliant blue staining.

Fluorescence Microscopy
MDA-MB-468, MDA-MB-231, MDA-MB-453, Hs578T and MCF-7 were seeded in black 96-well plate with a clear bottom (Greiner Bio-One, Frickenhausen, Germany) to a density of 40,000 cells/well and incubated overnight at 37 • C. To validate the internalization, cells were washed with PBS two times, and then incubated with 1 µg of each SNAP-Surface ® Alexa Fluor ® 647 labeled scFv-SNAP fusion protein at 37 • C for 3 h. After two washing steps, cells were incubated with Hoechst 33,342 fluorescent nuclear counterstain (1:500 in PBS) (Thermo Fisher Scientific, Darmstadt, Germany) for 15 min at 37 • C. Cells were washed with PBS twice and incubated in 100 µL of PBS. To determine the binding property, cells were fixed before incubation with antibodies and all steps were carried out at 4 • C. Briefly, cells were fixed by 4% formaldehyde solution for 10 min, followed by two washing steps with PBS. Residual formaldehyde was quenched by incubating with 50 nM ammonium chloride for 5 min. The internalization and binding were visualized with a DMi8 S Live-cell microscope (Leica Microsystems, Wetzlar, Germany) using a 100× oil objective.

Induction of Apoptosis
The induction of apoptosis was determined by Alexa Fluor ® 488 annexin V/dead cell apoptosis kit (Thermo Fisher Scientific, Eugene, OR, USA) using the manufacturer's instructions. Briefly, MDA-MB-468, MDA-MB-231, MDA-MB-453, Hs758T and MCF-7 were seeded in 24-well plates at a density of 50,000 cells/well in triplicates, and then incubated with 640 nM of scFv-425-SNAP, scFv-EpCAM-SNAP, scFv-425-SNAP-MMAE, scFv-EpCAM-SNAP-MMAE or MMAE at 37 • C for 48 h. Cells treated with PBS or camptothecin (Merck KGaA, Darmstadt, Germany) were used as negative or positive control, respectively. Cells were washed by cold PBS after harvesting, and then resuspended in 100 µL of 1 X annexin-binding buffer (10 mM HEPES, 140 nM NaCl, 2.5 mM CaCl2, pH 7.4). After incubation with Alexa Fluor ® 488 annexin V and propidium iodide (PI), early and late apoptotic cells were detected on BD FACSCanto TM II Flow Cytometer(BD Biosciences, San Jose, CA, USA). The experiment was repeated independently three times in triplicate.

Statistical Analysis
All experiments were carried out independently for at least three times in triplicate. The half maximal inhibitory concentration (IC 50 ) was calculated by GraphPad Prism 9.0.0 (GraphPad Software, San Diego, CA, USA) and shown as mean ± SEM. The data of induction of apoptosis was presented as mean ± SEM and significance was calculated by one-way analysis of variance (ANOVA) followed by Tukey's test using GraphPad Prism 9.0.0 (**** p ≤ 0.0001).

Conflicts of Interest:
The authors declare that they have no conflicts of interest.